diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,366462 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 8.0, + "eval_steps": 500, + "global_step": 61072, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 1.0911074740861974e-08, + "loss": 1.0479, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 2.182214948172395e-08, + "loss": 1.1155, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 3.273322422258593e-08, + "loss": 1.0246, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 4.36442989634479e-08, + "loss": 1.0661, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 5.4555373704309885e-08, + "loss": 1.0981, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 6.546644844517186e-08, + "loss": 1.0996, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 7.637752318603383e-08, + "loss": 1.0209, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 8.72885979268958e-08, + "loss": 1.0641, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 9.819967266775778e-08, + "loss": 1.1584, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 1.0911074740861977e-07, + "loss": 1.0417, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 1.2002182214948173e-07, + "loss": 1.0948, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 1.3093289689034372e-07, + "loss": 1.0282, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 1.4184397163120568e-07, + "loss": 0.9452, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 1.5275504637206767e-07, + "loss": 1.0645, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 1.6366612111292966e-07, + "loss": 1.0341, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 1.745771958537916e-07, + "loss": 1.1215, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 1.8548827059465358e-07, + "loss": 1.0733, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 1.9639934533551557e-07, + "loss": 1.0773, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 2.0731042007637755e-07, + "loss": 1.1137, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 2.1822149481723954e-07, + "loss": 1.0629, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 2.2913256955810148e-07, + "loss": 1.0217, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 2.4004364429896346e-07, + "loss": 0.9414, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 2.5095471903982545e-07, + "loss": 0.9763, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 2.6186579378068744e-07, + "loss": 1.0455, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 2.7277686852154937e-07, + "loss": 1.0455, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 2.8368794326241136e-07, + "loss": 1.1128, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 2.9459901800327335e-07, + "loss": 1.1807, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 3.0551009274413534e-07, + "loss": 1.01, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 3.1642116748499727e-07, + "loss": 0.9309, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 3.273322422258593e-07, + "loss": 1.0056, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 3.3824331696672125e-07, + "loss": 0.9746, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 3.491543917075832e-07, + "loss": 0.9866, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 3.600654664484452e-07, + "loss": 1.0662, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 3.7097654118930716e-07, + "loss": 0.9883, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 3.818876159301692e-07, + "loss": 1.0536, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 3.9279869067103113e-07, + "loss": 1.0052, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 4.0370976541189307e-07, + "loss": 0.9908, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 4.146208401527551e-07, + "loss": 0.9659, + "step": 38 + }, + { + "epoch": 0.01, + "learning_rate": 4.2553191489361704e-07, + "loss": 0.8591, + "step": 39 + }, + { + "epoch": 0.01, + "learning_rate": 4.364429896344791e-07, + "loss": 0.8516, + "step": 40 + }, + { + "epoch": 0.01, + "learning_rate": 4.47354064375341e-07, + "loss": 0.8932, + "step": 41 + }, + { + "epoch": 0.01, + "learning_rate": 4.5826513911620295e-07, + "loss": 0.9124, + "step": 42 + }, + { + "epoch": 0.01, + "learning_rate": 4.69176213857065e-07, + "loss": 0.8881, + "step": 43 + }, + { + "epoch": 0.01, + "learning_rate": 4.800872885979269e-07, + "loss": 0.8825, + "step": 44 + }, + { + "epoch": 0.01, + "learning_rate": 4.909983633387889e-07, + "loss": 0.9047, + "step": 45 + }, + { + "epoch": 0.01, + "learning_rate": 5.019094380796509e-07, + "loss": 0.8781, + "step": 46 + }, + { + "epoch": 0.01, + "learning_rate": 5.128205128205128e-07, + "loss": 0.8772, + "step": 47 + }, + { + "epoch": 0.01, + "learning_rate": 5.237315875613749e-07, + "loss": 0.9097, + "step": 48 + }, + { + "epoch": 0.01, + "learning_rate": 5.346426623022368e-07, + "loss": 0.8682, + "step": 49 + }, + { + "epoch": 0.01, + "learning_rate": 5.455537370430987e-07, + "loss": 0.7825, + "step": 50 + }, + { + "epoch": 0.01, + "learning_rate": 5.564648117839608e-07, + "loss": 0.8284, + "step": 51 + }, + { + "epoch": 0.01, + "learning_rate": 5.673758865248227e-07, + "loss": 0.8308, + "step": 52 + }, + { + "epoch": 0.01, + "learning_rate": 5.782869612656848e-07, + "loss": 0.8371, + "step": 53 + }, + { + "epoch": 0.01, + "learning_rate": 5.891980360065467e-07, + "loss": 0.8117, + "step": 54 + }, + { + "epoch": 0.01, + "learning_rate": 6.001091107474086e-07, + "loss": 0.7724, + "step": 55 + }, + { + "epoch": 0.01, + "learning_rate": 6.110201854882707e-07, + "loss": 0.6956, + "step": 56 + }, + { + "epoch": 0.01, + "learning_rate": 6.219312602291326e-07, + "loss": 0.662, + "step": 57 + }, + { + "epoch": 0.01, + "learning_rate": 6.328423349699945e-07, + "loss": 0.7113, + "step": 58 + }, + { + "epoch": 0.01, + "learning_rate": 6.437534097108565e-07, + "loss": 0.6371, + "step": 59 + }, + { + "epoch": 0.01, + "learning_rate": 6.546644844517186e-07, + "loss": 0.633, + "step": 60 + }, + { + "epoch": 0.01, + "learning_rate": 6.655755591925806e-07, + "loss": 0.6516, + "step": 61 + }, + { + "epoch": 0.01, + "learning_rate": 6.764866339334425e-07, + "loss": 0.6407, + "step": 62 + }, + { + "epoch": 0.01, + "learning_rate": 6.873977086743044e-07, + "loss": 0.5885, + "step": 63 + }, + { + "epoch": 0.01, + "learning_rate": 6.983087834151664e-07, + "loss": 0.5897, + "step": 64 + }, + { + "epoch": 0.01, + "learning_rate": 7.092198581560285e-07, + "loss": 0.6153, + "step": 65 + }, + { + "epoch": 0.01, + "learning_rate": 7.201309328968904e-07, + "loss": 0.5937, + "step": 66 + }, + { + "epoch": 0.01, + "learning_rate": 7.310420076377524e-07, + "loss": 0.6546, + "step": 67 + }, + { + "epoch": 0.01, + "learning_rate": 7.419530823786143e-07, + "loss": 0.5816, + "step": 68 + }, + { + "epoch": 0.01, + "learning_rate": 7.528641571194762e-07, + "loss": 0.5968, + "step": 69 + }, + { + "epoch": 0.01, + "learning_rate": 7.637752318603384e-07, + "loss": 0.6042, + "step": 70 + }, + { + "epoch": 0.01, + "learning_rate": 7.746863066012003e-07, + "loss": 0.5238, + "step": 71 + }, + { + "epoch": 0.01, + "learning_rate": 7.855973813420623e-07, + "loss": 0.5912, + "step": 72 + }, + { + "epoch": 0.01, + "learning_rate": 7.965084560829242e-07, + "loss": 0.5562, + "step": 73 + }, + { + "epoch": 0.01, + "learning_rate": 8.074195308237861e-07, + "loss": 0.5425, + "step": 74 + }, + { + "epoch": 0.01, + "learning_rate": 8.183306055646483e-07, + "loss": 0.5306, + "step": 75 + }, + { + "epoch": 0.01, + "learning_rate": 8.292416803055102e-07, + "loss": 0.5506, + "step": 76 + }, + { + "epoch": 0.01, + "learning_rate": 8.401527550463721e-07, + "loss": 0.5229, + "step": 77 + }, + { + "epoch": 0.01, + "learning_rate": 8.510638297872341e-07, + "loss": 0.5135, + "step": 78 + }, + { + "epoch": 0.01, + "learning_rate": 8.61974904528096e-07, + "loss": 0.4877, + "step": 79 + }, + { + "epoch": 0.01, + "learning_rate": 8.728859792689582e-07, + "loss": 0.5038, + "step": 80 + }, + { + "epoch": 0.01, + "learning_rate": 8.837970540098201e-07, + "loss": 0.5003, + "step": 81 + }, + { + "epoch": 0.01, + "learning_rate": 8.94708128750682e-07, + "loss": 0.489, + "step": 82 + }, + { + "epoch": 0.01, + "learning_rate": 9.05619203491544e-07, + "loss": 0.4867, + "step": 83 + }, + { + "epoch": 0.01, + "learning_rate": 9.165302782324059e-07, + "loss": 0.4398, + "step": 84 + }, + { + "epoch": 0.01, + "learning_rate": 9.27441352973268e-07, + "loss": 0.4727, + "step": 85 + }, + { + "epoch": 0.01, + "learning_rate": 9.3835242771413e-07, + "loss": 0.4866, + "step": 86 + }, + { + "epoch": 0.01, + "learning_rate": 9.492635024549919e-07, + "loss": 0.4794, + "step": 87 + }, + { + "epoch": 0.01, + "learning_rate": 9.601745771958539e-07, + "loss": 0.4855, + "step": 88 + }, + { + "epoch": 0.01, + "learning_rate": 9.710856519367158e-07, + "loss": 0.476, + "step": 89 + }, + { + "epoch": 0.01, + "learning_rate": 9.819967266775777e-07, + "loss": 0.4606, + "step": 90 + }, + { + "epoch": 0.01, + "learning_rate": 9.929078014184399e-07, + "loss": 0.4144, + "step": 91 + }, + { + "epoch": 0.01, + "learning_rate": 1.0038188761593018e-06, + "loss": 0.4316, + "step": 92 + }, + { + "epoch": 0.01, + "learning_rate": 1.0147299509001637e-06, + "loss": 0.4885, + "step": 93 + }, + { + "epoch": 0.01, + "learning_rate": 1.0256410256410257e-06, + "loss": 0.4707, + "step": 94 + }, + { + "epoch": 0.01, + "learning_rate": 1.0365521003818876e-06, + "loss": 0.4757, + "step": 95 + }, + { + "epoch": 0.01, + "learning_rate": 1.0474631751227498e-06, + "loss": 0.4324, + "step": 96 + }, + { + "epoch": 0.01, + "learning_rate": 1.0583742498636117e-06, + "loss": 0.414, + "step": 97 + }, + { + "epoch": 0.01, + "learning_rate": 1.0692853246044736e-06, + "loss": 0.4533, + "step": 98 + }, + { + "epoch": 0.01, + "learning_rate": 1.0801963993453356e-06, + "loss": 0.4197, + "step": 99 + }, + { + "epoch": 0.01, + "learning_rate": 1.0911074740861975e-06, + "loss": 0.4438, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 1.1020185488270596e-06, + "loss": 0.4285, + "step": 101 + }, + { + "epoch": 0.01, + "learning_rate": 1.1129296235679216e-06, + "loss": 0.4252, + "step": 102 + }, + { + "epoch": 0.01, + "learning_rate": 1.1238406983087835e-06, + "loss": 0.3962, + "step": 103 + }, + { + "epoch": 0.01, + "learning_rate": 1.1347517730496454e-06, + "loss": 0.4189, + "step": 104 + }, + { + "epoch": 0.01, + "learning_rate": 1.1456628477905074e-06, + "loss": 0.4627, + "step": 105 + }, + { + "epoch": 0.01, + "learning_rate": 1.1565739225313695e-06, + "loss": 0.4347, + "step": 106 + }, + { + "epoch": 0.01, + "learning_rate": 1.1674849972722315e-06, + "loss": 0.4404, + "step": 107 + }, + { + "epoch": 0.01, + "learning_rate": 1.1783960720130934e-06, + "loss": 0.4039, + "step": 108 + }, + { + "epoch": 0.01, + "learning_rate": 1.1893071467539553e-06, + "loss": 0.3945, + "step": 109 + }, + { + "epoch": 0.01, + "learning_rate": 1.2002182214948173e-06, + "loss": 0.4174, + "step": 110 + }, + { + "epoch": 0.01, + "learning_rate": 1.2111292962356794e-06, + "loss": 0.3971, + "step": 111 + }, + { + "epoch": 0.01, + "learning_rate": 1.2220403709765413e-06, + "loss": 0.413, + "step": 112 + }, + { + "epoch": 0.01, + "learning_rate": 1.2329514457174033e-06, + "loss": 0.3852, + "step": 113 + }, + { + "epoch": 0.01, + "learning_rate": 1.2438625204582652e-06, + "loss": 0.4115, + "step": 114 + }, + { + "epoch": 0.02, + "learning_rate": 1.2547735951991271e-06, + "loss": 0.3671, + "step": 115 + }, + { + "epoch": 0.02, + "learning_rate": 1.265684669939989e-06, + "loss": 0.3922, + "step": 116 + }, + { + "epoch": 0.02, + "learning_rate": 1.276595744680851e-06, + "loss": 0.3667, + "step": 117 + }, + { + "epoch": 0.02, + "learning_rate": 1.287506819421713e-06, + "loss": 0.3442, + "step": 118 + }, + { + "epoch": 0.02, + "learning_rate": 1.2984178941625753e-06, + "loss": 0.3943, + "step": 119 + }, + { + "epoch": 0.02, + "learning_rate": 1.3093289689034372e-06, + "loss": 0.3621, + "step": 120 + }, + { + "epoch": 0.02, + "learning_rate": 1.3202400436442992e-06, + "loss": 0.3715, + "step": 121 + }, + { + "epoch": 0.02, + "learning_rate": 1.3311511183851611e-06, + "loss": 0.4172, + "step": 122 + }, + { + "epoch": 0.02, + "learning_rate": 1.342062193126023e-06, + "loss": 0.4011, + "step": 123 + }, + { + "epoch": 0.02, + "learning_rate": 1.352973267866885e-06, + "loss": 0.3565, + "step": 124 + }, + { + "epoch": 0.02, + "learning_rate": 1.363884342607747e-06, + "loss": 0.3704, + "step": 125 + }, + { + "epoch": 0.02, + "learning_rate": 1.3747954173486089e-06, + "loss": 0.3644, + "step": 126 + }, + { + "epoch": 0.02, + "learning_rate": 1.3857064920894708e-06, + "loss": 0.3835, + "step": 127 + }, + { + "epoch": 0.02, + "learning_rate": 1.3966175668303327e-06, + "loss": 0.3824, + "step": 128 + }, + { + "epoch": 0.02, + "learning_rate": 1.407528641571195e-06, + "loss": 0.3717, + "step": 129 + }, + { + "epoch": 0.02, + "learning_rate": 1.418439716312057e-06, + "loss": 0.3495, + "step": 130 + }, + { + "epoch": 0.02, + "learning_rate": 1.429350791052919e-06, + "loss": 0.3703, + "step": 131 + }, + { + "epoch": 0.02, + "learning_rate": 1.4402618657937809e-06, + "loss": 0.3523, + "step": 132 + }, + { + "epoch": 0.02, + "learning_rate": 1.4511729405346428e-06, + "loss": 0.3769, + "step": 133 + }, + { + "epoch": 0.02, + "learning_rate": 1.4620840152755048e-06, + "loss": 0.3672, + "step": 134 + }, + { + "epoch": 0.02, + "learning_rate": 1.4729950900163667e-06, + "loss": 0.3633, + "step": 135 + }, + { + "epoch": 0.02, + "learning_rate": 1.4839061647572286e-06, + "loss": 0.3493, + "step": 136 + }, + { + "epoch": 0.02, + "learning_rate": 1.4948172394980906e-06, + "loss": 0.3405, + "step": 137 + }, + { + "epoch": 0.02, + "learning_rate": 1.5057283142389525e-06, + "loss": 0.343, + "step": 138 + }, + { + "epoch": 0.02, + "learning_rate": 1.5166393889798144e-06, + "loss": 0.3444, + "step": 139 + }, + { + "epoch": 0.02, + "learning_rate": 1.5275504637206768e-06, + "loss": 0.3563, + "step": 140 + }, + { + "epoch": 0.02, + "learning_rate": 1.5384615384615387e-06, + "loss": 0.3353, + "step": 141 + }, + { + "epoch": 0.02, + "learning_rate": 1.5493726132024007e-06, + "loss": 0.3424, + "step": 142 + }, + { + "epoch": 0.02, + "learning_rate": 1.5602836879432626e-06, + "loss": 0.346, + "step": 143 + }, + { + "epoch": 0.02, + "learning_rate": 1.5711947626841245e-06, + "loss": 0.3149, + "step": 144 + }, + { + "epoch": 0.02, + "learning_rate": 1.5821058374249865e-06, + "loss": 0.3406, + "step": 145 + }, + { + "epoch": 0.02, + "learning_rate": 1.5930169121658484e-06, + "loss": 0.3204, + "step": 146 + }, + { + "epoch": 0.02, + "learning_rate": 1.6039279869067103e-06, + "loss": 0.3097, + "step": 147 + }, + { + "epoch": 0.02, + "learning_rate": 1.6148390616475723e-06, + "loss": 0.2999, + "step": 148 + }, + { + "epoch": 0.02, + "learning_rate": 1.6257501363884342e-06, + "loss": 0.3042, + "step": 149 + }, + { + "epoch": 0.02, + "learning_rate": 1.6366612111292966e-06, + "loss": 0.3043, + "step": 150 + }, + { + "epoch": 0.02, + "learning_rate": 1.6475722858701585e-06, + "loss": 0.3042, + "step": 151 + }, + { + "epoch": 0.02, + "learning_rate": 1.6584833606110204e-06, + "loss": 0.3396, + "step": 152 + }, + { + "epoch": 0.02, + "learning_rate": 1.6693944353518824e-06, + "loss": 0.3022, + "step": 153 + }, + { + "epoch": 0.02, + "learning_rate": 1.6803055100927443e-06, + "loss": 0.2957, + "step": 154 + }, + { + "epoch": 0.02, + "learning_rate": 1.6912165848336062e-06, + "loss": 0.317, + "step": 155 + }, + { + "epoch": 0.02, + "learning_rate": 1.7021276595744682e-06, + "loss": 0.3027, + "step": 156 + }, + { + "epoch": 0.02, + "learning_rate": 1.71303873431533e-06, + "loss": 0.316, + "step": 157 + }, + { + "epoch": 0.02, + "learning_rate": 1.723949809056192e-06, + "loss": 0.3131, + "step": 158 + }, + { + "epoch": 0.02, + "learning_rate": 1.734860883797054e-06, + "loss": 0.3155, + "step": 159 + }, + { + "epoch": 0.02, + "learning_rate": 1.7457719585379163e-06, + "loss": 0.2835, + "step": 160 + }, + { + "epoch": 0.02, + "learning_rate": 1.7566830332787783e-06, + "loss": 0.2999, + "step": 161 + }, + { + "epoch": 0.02, + "learning_rate": 1.7675941080196402e-06, + "loss": 0.3049, + "step": 162 + }, + { + "epoch": 0.02, + "learning_rate": 1.7785051827605021e-06, + "loss": 0.2853, + "step": 163 + }, + { + "epoch": 0.02, + "learning_rate": 1.789416257501364e-06, + "loss": 0.3059, + "step": 164 + }, + { + "epoch": 0.02, + "learning_rate": 1.800327332242226e-06, + "loss": 0.3025, + "step": 165 + }, + { + "epoch": 0.02, + "learning_rate": 1.811238406983088e-06, + "loss": 0.2737, + "step": 166 + }, + { + "epoch": 0.02, + "learning_rate": 1.8221494817239499e-06, + "loss": 0.293, + "step": 167 + }, + { + "epoch": 0.02, + "learning_rate": 1.8330605564648118e-06, + "loss": 0.2447, + "step": 168 + }, + { + "epoch": 0.02, + "learning_rate": 1.8439716312056737e-06, + "loss": 0.2741, + "step": 169 + }, + { + "epoch": 0.02, + "learning_rate": 1.854882705946536e-06, + "loss": 0.2687, + "step": 170 + }, + { + "epoch": 0.02, + "learning_rate": 1.865793780687398e-06, + "loss": 0.267, + "step": 171 + }, + { + "epoch": 0.02, + "learning_rate": 1.87670485542826e-06, + "loss": 0.2688, + "step": 172 + }, + { + "epoch": 0.02, + "learning_rate": 1.887615930169122e-06, + "loss": 0.2653, + "step": 173 + }, + { + "epoch": 0.02, + "learning_rate": 1.8985270049099838e-06, + "loss": 0.2825, + "step": 174 + }, + { + "epoch": 0.02, + "learning_rate": 1.9094380796508458e-06, + "loss": 0.2589, + "step": 175 + }, + { + "epoch": 0.02, + "learning_rate": 1.9203491543917077e-06, + "loss": 0.2676, + "step": 176 + }, + { + "epoch": 0.02, + "learning_rate": 1.9312602291325696e-06, + "loss": 0.2862, + "step": 177 + }, + { + "epoch": 0.02, + "learning_rate": 1.9421713038734316e-06, + "loss": 0.2655, + "step": 178 + }, + { + "epoch": 0.02, + "learning_rate": 1.9530823786142935e-06, + "loss": 0.2589, + "step": 179 + }, + { + "epoch": 0.02, + "learning_rate": 1.9639934533551554e-06, + "loss": 0.2462, + "step": 180 + }, + { + "epoch": 0.02, + "learning_rate": 1.974904528096018e-06, + "loss": 0.2473, + "step": 181 + }, + { + "epoch": 0.02, + "learning_rate": 1.9858156028368797e-06, + "loss": 0.2684, + "step": 182 + }, + { + "epoch": 0.02, + "learning_rate": 1.9967266775777417e-06, + "loss": 0.2443, + "step": 183 + }, + { + "epoch": 0.02, + "learning_rate": 2.0076377523186036e-06, + "loss": 0.2429, + "step": 184 + }, + { + "epoch": 0.02, + "learning_rate": 2.0185488270594655e-06, + "loss": 0.2474, + "step": 185 + }, + { + "epoch": 0.02, + "learning_rate": 2.0294599018003275e-06, + "loss": 0.2667, + "step": 186 + }, + { + "epoch": 0.02, + "learning_rate": 2.0403709765411894e-06, + "loss": 0.2676, + "step": 187 + }, + { + "epoch": 0.02, + "learning_rate": 2.0512820512820513e-06, + "loss": 0.2351, + "step": 188 + }, + { + "epoch": 0.02, + "learning_rate": 2.0621931260229133e-06, + "loss": 0.2072, + "step": 189 + }, + { + "epoch": 0.02, + "learning_rate": 2.073104200763775e-06, + "loss": 0.2349, + "step": 190 + }, + { + "epoch": 0.03, + "learning_rate": 2.0840152755046376e-06, + "loss": 0.2489, + "step": 191 + }, + { + "epoch": 0.03, + "learning_rate": 2.0949263502454995e-06, + "loss": 0.2366, + "step": 192 + }, + { + "epoch": 0.03, + "learning_rate": 2.1058374249863614e-06, + "loss": 0.2366, + "step": 193 + }, + { + "epoch": 0.03, + "learning_rate": 2.1167484997272234e-06, + "loss": 0.2211, + "step": 194 + }, + { + "epoch": 0.03, + "learning_rate": 2.1276595744680853e-06, + "loss": 0.2532, + "step": 195 + }, + { + "epoch": 0.03, + "learning_rate": 2.1385706492089472e-06, + "loss": 0.2184, + "step": 196 + }, + { + "epoch": 0.03, + "learning_rate": 2.149481723949809e-06, + "loss": 0.2052, + "step": 197 + }, + { + "epoch": 0.03, + "learning_rate": 2.160392798690671e-06, + "loss": 0.2217, + "step": 198 + }, + { + "epoch": 0.03, + "learning_rate": 2.171303873431533e-06, + "loss": 0.2419, + "step": 199 + }, + { + "epoch": 0.03, + "learning_rate": 2.182214948172395e-06, + "loss": 0.2074, + "step": 200 + }, + { + "epoch": 0.03, + "learning_rate": 2.1931260229132573e-06, + "loss": 0.2346, + "step": 201 + }, + { + "epoch": 0.03, + "learning_rate": 2.2040370976541193e-06, + "loss": 0.2457, + "step": 202 + }, + { + "epoch": 0.03, + "learning_rate": 2.214948172394981e-06, + "loss": 0.2236, + "step": 203 + }, + { + "epoch": 0.03, + "learning_rate": 2.225859247135843e-06, + "loss": 0.2266, + "step": 204 + }, + { + "epoch": 0.03, + "learning_rate": 2.236770321876705e-06, + "loss": 0.2167, + "step": 205 + }, + { + "epoch": 0.03, + "learning_rate": 2.247681396617567e-06, + "loss": 0.2299, + "step": 206 + }, + { + "epoch": 0.03, + "learning_rate": 2.258592471358429e-06, + "loss": 0.2113, + "step": 207 + }, + { + "epoch": 0.03, + "learning_rate": 2.269503546099291e-06, + "loss": 0.2171, + "step": 208 + }, + { + "epoch": 0.03, + "learning_rate": 2.280414620840153e-06, + "loss": 0.2221, + "step": 209 + }, + { + "epoch": 0.03, + "learning_rate": 2.2913256955810148e-06, + "loss": 0.2233, + "step": 210 + }, + { + "epoch": 0.03, + "learning_rate": 2.3022367703218767e-06, + "loss": 0.2287, + "step": 211 + }, + { + "epoch": 0.03, + "learning_rate": 2.313147845062739e-06, + "loss": 0.218, + "step": 212 + }, + { + "epoch": 0.03, + "learning_rate": 2.324058919803601e-06, + "loss": 0.2009, + "step": 213 + }, + { + "epoch": 0.03, + "learning_rate": 2.334969994544463e-06, + "loss": 0.2146, + "step": 214 + }, + { + "epoch": 0.03, + "learning_rate": 2.345881069285325e-06, + "loss": 0.2036, + "step": 215 + }, + { + "epoch": 0.03, + "learning_rate": 2.3567921440261868e-06, + "loss": 0.1997, + "step": 216 + }, + { + "epoch": 0.03, + "learning_rate": 2.3677032187670487e-06, + "loss": 0.2148, + "step": 217 + }, + { + "epoch": 0.03, + "learning_rate": 2.3786142935079107e-06, + "loss": 0.211, + "step": 218 + }, + { + "epoch": 0.03, + "learning_rate": 2.3895253682487726e-06, + "loss": 0.198, + "step": 219 + }, + { + "epoch": 0.03, + "learning_rate": 2.4004364429896345e-06, + "loss": 0.2161, + "step": 220 + }, + { + "epoch": 0.03, + "learning_rate": 2.4113475177304965e-06, + "loss": 0.2234, + "step": 221 + }, + { + "epoch": 0.03, + "learning_rate": 2.422258592471359e-06, + "loss": 0.205, + "step": 222 + }, + { + "epoch": 0.03, + "learning_rate": 2.4331696672122207e-06, + "loss": 0.2199, + "step": 223 + }, + { + "epoch": 0.03, + "learning_rate": 2.4440807419530827e-06, + "loss": 0.2014, + "step": 224 + }, + { + "epoch": 0.03, + "learning_rate": 2.4549918166939446e-06, + "loss": 0.2105, + "step": 225 + }, + { + "epoch": 0.03, + "learning_rate": 2.4659028914348066e-06, + "loss": 0.2037, + "step": 226 + }, + { + "epoch": 0.03, + "learning_rate": 2.4768139661756685e-06, + "loss": 0.1967, + "step": 227 + }, + { + "epoch": 0.03, + "learning_rate": 2.4877250409165304e-06, + "loss": 0.2015, + "step": 228 + }, + { + "epoch": 0.03, + "learning_rate": 2.4986361156573924e-06, + "loss": 0.1942, + "step": 229 + }, + { + "epoch": 0.03, + "learning_rate": 2.5095471903982543e-06, + "loss": 0.1962, + "step": 230 + }, + { + "epoch": 0.03, + "learning_rate": 2.5204582651391162e-06, + "loss": 0.2066, + "step": 231 + }, + { + "epoch": 0.03, + "learning_rate": 2.531369339879978e-06, + "loss": 0.2071, + "step": 232 + }, + { + "epoch": 0.03, + "learning_rate": 2.54228041462084e-06, + "loss": 0.2119, + "step": 233 + }, + { + "epoch": 0.03, + "learning_rate": 2.553191489361702e-06, + "loss": 0.2129, + "step": 234 + }, + { + "epoch": 0.03, + "learning_rate": 2.564102564102564e-06, + "loss": 0.1916, + "step": 235 + }, + { + "epoch": 0.03, + "learning_rate": 2.575013638843426e-06, + "loss": 0.1974, + "step": 236 + }, + { + "epoch": 0.03, + "learning_rate": 2.5859247135842887e-06, + "loss": 0.1951, + "step": 237 + }, + { + "epoch": 0.03, + "learning_rate": 2.5968357883251506e-06, + "loss": 0.2006, + "step": 238 + }, + { + "epoch": 0.03, + "learning_rate": 2.6077468630660126e-06, + "loss": 0.2029, + "step": 239 + }, + { + "epoch": 0.03, + "learning_rate": 2.6186579378068745e-06, + "loss": 0.2038, + "step": 240 + }, + { + "epoch": 0.03, + "learning_rate": 2.6295690125477364e-06, + "loss": 0.2051, + "step": 241 + }, + { + "epoch": 0.03, + "learning_rate": 2.6404800872885984e-06, + "loss": 0.1978, + "step": 242 + }, + { + "epoch": 0.03, + "learning_rate": 2.6513911620294603e-06, + "loss": 0.2085, + "step": 243 + }, + { + "epoch": 0.03, + "learning_rate": 2.6623022367703222e-06, + "loss": 0.168, + "step": 244 + }, + { + "epoch": 0.03, + "learning_rate": 2.673213311511184e-06, + "loss": 0.1921, + "step": 245 + }, + { + "epoch": 0.03, + "learning_rate": 2.684124386252046e-06, + "loss": 0.1961, + "step": 246 + }, + { + "epoch": 0.03, + "learning_rate": 2.695035460992908e-06, + "loss": 0.2067, + "step": 247 + }, + { + "epoch": 0.03, + "learning_rate": 2.70594653573377e-06, + "loss": 0.188, + "step": 248 + }, + { + "epoch": 0.03, + "learning_rate": 2.716857610474632e-06, + "loss": 0.1868, + "step": 249 + }, + { + "epoch": 0.03, + "learning_rate": 2.727768685215494e-06, + "loss": 0.1796, + "step": 250 + }, + { + "epoch": 0.03, + "learning_rate": 2.7386797599563558e-06, + "loss": 0.1824, + "step": 251 + }, + { + "epoch": 0.03, + "learning_rate": 2.7495908346972177e-06, + "loss": 0.192, + "step": 252 + }, + { + "epoch": 0.03, + "learning_rate": 2.7605019094380796e-06, + "loss": 0.1951, + "step": 253 + }, + { + "epoch": 0.03, + "learning_rate": 2.7714129841789416e-06, + "loss": 0.1958, + "step": 254 + }, + { + "epoch": 0.03, + "learning_rate": 2.7823240589198035e-06, + "loss": 0.1926, + "step": 255 + }, + { + "epoch": 0.03, + "learning_rate": 2.7932351336606654e-06, + "loss": 0.1856, + "step": 256 + }, + { + "epoch": 0.03, + "learning_rate": 2.8041462084015274e-06, + "loss": 0.1924, + "step": 257 + }, + { + "epoch": 0.03, + "learning_rate": 2.81505728314239e-06, + "loss": 0.1781, + "step": 258 + }, + { + "epoch": 0.03, + "learning_rate": 2.825968357883252e-06, + "loss": 0.2033, + "step": 259 + }, + { + "epoch": 0.03, + "learning_rate": 2.836879432624114e-06, + "loss": 0.183, + "step": 260 + }, + { + "epoch": 0.03, + "learning_rate": 2.847790507364976e-06, + "loss": 0.196, + "step": 261 + }, + { + "epoch": 0.03, + "learning_rate": 2.858701582105838e-06, + "loss": 0.198, + "step": 262 + }, + { + "epoch": 0.03, + "learning_rate": 2.8696126568467e-06, + "loss": 0.1853, + "step": 263 + }, + { + "epoch": 0.03, + "learning_rate": 2.8805237315875618e-06, + "loss": 0.1695, + "step": 264 + }, + { + "epoch": 0.03, + "learning_rate": 2.8914348063284237e-06, + "loss": 0.1735, + "step": 265 + }, + { + "epoch": 0.03, + "learning_rate": 2.9023458810692856e-06, + "loss": 0.2021, + "step": 266 + }, + { + "epoch": 0.03, + "learning_rate": 2.9132569558101476e-06, + "loss": 0.1834, + "step": 267 + }, + { + "epoch": 0.04, + "learning_rate": 2.9241680305510095e-06, + "loss": 0.1902, + "step": 268 + }, + { + "epoch": 0.04, + "learning_rate": 2.9350791052918714e-06, + "loss": 0.1879, + "step": 269 + }, + { + "epoch": 0.04, + "learning_rate": 2.9459901800327334e-06, + "loss": 0.1978, + "step": 270 + }, + { + "epoch": 0.04, + "learning_rate": 2.9569012547735953e-06, + "loss": 0.1733, + "step": 271 + }, + { + "epoch": 0.04, + "learning_rate": 2.9678123295144572e-06, + "loss": 0.1791, + "step": 272 + }, + { + "epoch": 0.04, + "learning_rate": 2.978723404255319e-06, + "loss": 0.2068, + "step": 273 + }, + { + "epoch": 0.04, + "learning_rate": 2.989634478996181e-06, + "loss": 0.1903, + "step": 274 + }, + { + "epoch": 0.04, + "learning_rate": 3.000545553737043e-06, + "loss": 0.1769, + "step": 275 + }, + { + "epoch": 0.04, + "learning_rate": 3.011456628477905e-06, + "loss": 0.1818, + "step": 276 + }, + { + "epoch": 0.04, + "learning_rate": 3.022367703218767e-06, + "loss": 0.159, + "step": 277 + }, + { + "epoch": 0.04, + "learning_rate": 3.033278777959629e-06, + "loss": 0.1789, + "step": 278 + }, + { + "epoch": 0.04, + "learning_rate": 3.0441898527004916e-06, + "loss": 0.1936, + "step": 279 + }, + { + "epoch": 0.04, + "learning_rate": 3.0551009274413536e-06, + "loss": 0.1915, + "step": 280 + }, + { + "epoch": 0.04, + "learning_rate": 3.0660120021822155e-06, + "loss": 0.1668, + "step": 281 + }, + { + "epoch": 0.04, + "learning_rate": 3.0769230769230774e-06, + "loss": 0.184, + "step": 282 + }, + { + "epoch": 0.04, + "learning_rate": 3.0878341516639394e-06, + "loss": 0.1778, + "step": 283 + }, + { + "epoch": 0.04, + "learning_rate": 3.0987452264048013e-06, + "loss": 0.168, + "step": 284 + }, + { + "epoch": 0.04, + "learning_rate": 3.1096563011456632e-06, + "loss": 0.1753, + "step": 285 + }, + { + "epoch": 0.04, + "learning_rate": 3.120567375886525e-06, + "loss": 0.1765, + "step": 286 + }, + { + "epoch": 0.04, + "learning_rate": 3.131478450627387e-06, + "loss": 0.1749, + "step": 287 + }, + { + "epoch": 0.04, + "learning_rate": 3.142389525368249e-06, + "loss": 0.1676, + "step": 288 + }, + { + "epoch": 0.04, + "learning_rate": 3.153300600109111e-06, + "loss": 0.1632, + "step": 289 + }, + { + "epoch": 0.04, + "learning_rate": 3.164211674849973e-06, + "loss": 0.1636, + "step": 290 + }, + { + "epoch": 0.04, + "learning_rate": 3.175122749590835e-06, + "loss": 0.1797, + "step": 291 + }, + { + "epoch": 0.04, + "learning_rate": 3.1860338243316968e-06, + "loss": 0.1701, + "step": 292 + }, + { + "epoch": 0.04, + "learning_rate": 3.1969448990725587e-06, + "loss": 0.1712, + "step": 293 + }, + { + "epoch": 0.04, + "learning_rate": 3.2078559738134207e-06, + "loss": 0.1712, + "step": 294 + }, + { + "epoch": 0.04, + "learning_rate": 3.2187670485542826e-06, + "loss": 0.1831, + "step": 295 + }, + { + "epoch": 0.04, + "learning_rate": 3.2296781232951445e-06, + "loss": 0.1861, + "step": 296 + }, + { + "epoch": 0.04, + "learning_rate": 3.2405891980360065e-06, + "loss": 0.1879, + "step": 297 + }, + { + "epoch": 0.04, + "learning_rate": 3.2515002727768684e-06, + "loss": 0.1655, + "step": 298 + }, + { + "epoch": 0.04, + "learning_rate": 3.262411347517731e-06, + "loss": 0.1771, + "step": 299 + }, + { + "epoch": 0.04, + "learning_rate": 3.273322422258593e-06, + "loss": 0.1836, + "step": 300 + }, + { + "epoch": 0.04, + "learning_rate": 3.284233496999455e-06, + "loss": 0.1704, + "step": 301 + }, + { + "epoch": 0.04, + "learning_rate": 3.295144571740317e-06, + "loss": 0.1738, + "step": 302 + }, + { + "epoch": 0.04, + "learning_rate": 3.306055646481179e-06, + "loss": 0.1805, + "step": 303 + }, + { + "epoch": 0.04, + "learning_rate": 3.316966721222041e-06, + "loss": 0.1771, + "step": 304 + }, + { + "epoch": 0.04, + "learning_rate": 3.3278777959629028e-06, + "loss": 0.1605, + "step": 305 + }, + { + "epoch": 0.04, + "learning_rate": 3.3387888707037647e-06, + "loss": 0.1803, + "step": 306 + }, + { + "epoch": 0.04, + "learning_rate": 3.3496999454446266e-06, + "loss": 0.1758, + "step": 307 + }, + { + "epoch": 0.04, + "learning_rate": 3.3606110201854886e-06, + "loss": 0.1706, + "step": 308 + }, + { + "epoch": 0.04, + "learning_rate": 3.3715220949263505e-06, + "loss": 0.1771, + "step": 309 + }, + { + "epoch": 0.04, + "learning_rate": 3.3824331696672125e-06, + "loss": 0.1745, + "step": 310 + }, + { + "epoch": 0.04, + "learning_rate": 3.3933442444080744e-06, + "loss": 0.183, + "step": 311 + }, + { + "epoch": 0.04, + "learning_rate": 3.4042553191489363e-06, + "loss": 0.1744, + "step": 312 + }, + { + "epoch": 0.04, + "learning_rate": 3.4151663938897983e-06, + "loss": 0.1907, + "step": 313 + }, + { + "epoch": 0.04, + "learning_rate": 3.42607746863066e-06, + "loss": 0.1689, + "step": 314 + }, + { + "epoch": 0.04, + "learning_rate": 3.436988543371522e-06, + "loss": 0.1714, + "step": 315 + }, + { + "epoch": 0.04, + "learning_rate": 3.447899618112384e-06, + "loss": 0.1586, + "step": 316 + }, + { + "epoch": 0.04, + "learning_rate": 3.458810692853246e-06, + "loss": 0.1806, + "step": 317 + }, + { + "epoch": 0.04, + "learning_rate": 3.469721767594108e-06, + "loss": 0.1722, + "step": 318 + }, + { + "epoch": 0.04, + "learning_rate": 3.48063284233497e-06, + "loss": 0.178, + "step": 319 + }, + { + "epoch": 0.04, + "learning_rate": 3.4915439170758326e-06, + "loss": 0.1707, + "step": 320 + }, + { + "epoch": 0.04, + "learning_rate": 3.5024549918166946e-06, + "loss": 0.1702, + "step": 321 + }, + { + "epoch": 0.04, + "learning_rate": 3.5133660665575565e-06, + "loss": 0.1622, + "step": 322 + }, + { + "epoch": 0.04, + "learning_rate": 3.5242771412984185e-06, + "loss": 0.1664, + "step": 323 + }, + { + "epoch": 0.04, + "learning_rate": 3.5351882160392804e-06, + "loss": 0.1683, + "step": 324 + }, + { + "epoch": 0.04, + "learning_rate": 3.5460992907801423e-06, + "loss": 0.1615, + "step": 325 + }, + { + "epoch": 0.04, + "learning_rate": 3.5570103655210043e-06, + "loss": 0.18, + "step": 326 + }, + { + "epoch": 0.04, + "learning_rate": 3.567921440261866e-06, + "loss": 0.1647, + "step": 327 + }, + { + "epoch": 0.04, + "learning_rate": 3.578832515002728e-06, + "loss": 0.18, + "step": 328 + }, + { + "epoch": 0.04, + "learning_rate": 3.58974358974359e-06, + "loss": 0.1796, + "step": 329 + }, + { + "epoch": 0.04, + "learning_rate": 3.600654664484452e-06, + "loss": 0.1611, + "step": 330 + }, + { + "epoch": 0.04, + "learning_rate": 3.611565739225314e-06, + "loss": 0.1753, + "step": 331 + }, + { + "epoch": 0.04, + "learning_rate": 3.622476813966176e-06, + "loss": 0.1814, + "step": 332 + }, + { + "epoch": 0.04, + "learning_rate": 3.633387888707038e-06, + "loss": 0.1828, + "step": 333 + }, + { + "epoch": 0.04, + "learning_rate": 3.6442989634478997e-06, + "loss": 0.1696, + "step": 334 + }, + { + "epoch": 0.04, + "learning_rate": 3.6552100381887617e-06, + "loss": 0.1595, + "step": 335 + }, + { + "epoch": 0.04, + "learning_rate": 3.6661211129296236e-06, + "loss": 0.1758, + "step": 336 + }, + { + "epoch": 0.04, + "learning_rate": 3.6770321876704855e-06, + "loss": 0.1663, + "step": 337 + }, + { + "epoch": 0.04, + "learning_rate": 3.6879432624113475e-06, + "loss": 0.1769, + "step": 338 + }, + { + "epoch": 0.04, + "learning_rate": 3.6988543371522094e-06, + "loss": 0.1571, + "step": 339 + }, + { + "epoch": 0.04, + "learning_rate": 3.709765411893072e-06, + "loss": 0.1577, + "step": 340 + }, + { + "epoch": 0.04, + "learning_rate": 3.720676486633934e-06, + "loss": 0.1591, + "step": 341 + }, + { + "epoch": 0.04, + "learning_rate": 3.731587561374796e-06, + "loss": 0.1495, + "step": 342 + }, + { + "epoch": 0.04, + "learning_rate": 3.742498636115658e-06, + "loss": 0.1695, + "step": 343 + }, + { + "epoch": 0.05, + "learning_rate": 3.75340971085652e-06, + "loss": 0.1696, + "step": 344 + }, + { + "epoch": 0.05, + "learning_rate": 3.764320785597382e-06, + "loss": 0.171, + "step": 345 + }, + { + "epoch": 0.05, + "learning_rate": 3.775231860338244e-06, + "loss": 0.1764, + "step": 346 + }, + { + "epoch": 0.05, + "learning_rate": 3.7861429350791057e-06, + "loss": 0.1566, + "step": 347 + }, + { + "epoch": 0.05, + "learning_rate": 3.7970540098199677e-06, + "loss": 0.1548, + "step": 348 + }, + { + "epoch": 0.05, + "learning_rate": 3.8079650845608296e-06, + "loss": 0.1654, + "step": 349 + }, + { + "epoch": 0.05, + "learning_rate": 3.8188761593016915e-06, + "loss": 0.1623, + "step": 350 + }, + { + "epoch": 0.05, + "learning_rate": 3.8297872340425535e-06, + "loss": 0.1682, + "step": 351 + }, + { + "epoch": 0.05, + "learning_rate": 3.840698308783415e-06, + "loss": 0.1597, + "step": 352 + }, + { + "epoch": 0.05, + "learning_rate": 3.851609383524277e-06, + "loss": 0.1625, + "step": 353 + }, + { + "epoch": 0.05, + "learning_rate": 3.862520458265139e-06, + "loss": 0.1674, + "step": 354 + }, + { + "epoch": 0.05, + "learning_rate": 3.873431533006001e-06, + "loss": 0.1643, + "step": 355 + }, + { + "epoch": 0.05, + "learning_rate": 3.884342607746863e-06, + "loss": 0.1683, + "step": 356 + }, + { + "epoch": 0.05, + "learning_rate": 3.895253682487725e-06, + "loss": 0.1814, + "step": 357 + }, + { + "epoch": 0.05, + "learning_rate": 3.906164757228587e-06, + "loss": 0.1725, + "step": 358 + }, + { + "epoch": 0.05, + "learning_rate": 3.917075831969449e-06, + "loss": 0.1584, + "step": 359 + }, + { + "epoch": 0.05, + "learning_rate": 3.927986906710311e-06, + "loss": 0.1681, + "step": 360 + }, + { + "epoch": 0.05, + "learning_rate": 3.938897981451174e-06, + "loss": 0.1688, + "step": 361 + }, + { + "epoch": 0.05, + "learning_rate": 3.949809056192036e-06, + "loss": 0.1479, + "step": 362 + }, + { + "epoch": 0.05, + "learning_rate": 3.9607201309328975e-06, + "loss": 0.1625, + "step": 363 + }, + { + "epoch": 0.05, + "learning_rate": 3.9716312056737595e-06, + "loss": 0.1614, + "step": 364 + }, + { + "epoch": 0.05, + "learning_rate": 3.982542280414621e-06, + "loss": 0.1522, + "step": 365 + }, + { + "epoch": 0.05, + "learning_rate": 3.993453355155483e-06, + "loss": 0.1693, + "step": 366 + }, + { + "epoch": 0.05, + "learning_rate": 4.004364429896345e-06, + "loss": 0.1716, + "step": 367 + }, + { + "epoch": 0.05, + "learning_rate": 4.015275504637207e-06, + "loss": 0.1612, + "step": 368 + }, + { + "epoch": 0.05, + "learning_rate": 4.026186579378069e-06, + "loss": 0.1455, + "step": 369 + }, + { + "epoch": 0.05, + "learning_rate": 4.037097654118931e-06, + "loss": 0.1619, + "step": 370 + }, + { + "epoch": 0.05, + "learning_rate": 4.048008728859793e-06, + "loss": 0.1617, + "step": 371 + }, + { + "epoch": 0.05, + "learning_rate": 4.058919803600655e-06, + "loss": 0.1499, + "step": 372 + }, + { + "epoch": 0.05, + "learning_rate": 4.069830878341517e-06, + "loss": 0.16, + "step": 373 + }, + { + "epoch": 0.05, + "learning_rate": 4.080741953082379e-06, + "loss": 0.1637, + "step": 374 + }, + { + "epoch": 0.05, + "learning_rate": 4.091653027823241e-06, + "loss": 0.1583, + "step": 375 + }, + { + "epoch": 0.05, + "learning_rate": 4.102564102564103e-06, + "loss": 0.1551, + "step": 376 + }, + { + "epoch": 0.05, + "learning_rate": 4.113475177304965e-06, + "loss": 0.151, + "step": 377 + }, + { + "epoch": 0.05, + "learning_rate": 4.1243862520458266e-06, + "loss": 0.1671, + "step": 378 + }, + { + "epoch": 0.05, + "learning_rate": 4.1352973267866885e-06, + "loss": 0.1596, + "step": 379 + }, + { + "epoch": 0.05, + "learning_rate": 4.14620840152755e-06, + "loss": 0.1623, + "step": 380 + }, + { + "epoch": 0.05, + "learning_rate": 4.157119476268412e-06, + "loss": 0.1522, + "step": 381 + }, + { + "epoch": 0.05, + "learning_rate": 4.168030551009275e-06, + "loss": 0.1647, + "step": 382 + }, + { + "epoch": 0.05, + "learning_rate": 4.178941625750137e-06, + "loss": 0.1454, + "step": 383 + }, + { + "epoch": 0.05, + "learning_rate": 4.189852700490999e-06, + "loss": 0.1538, + "step": 384 + }, + { + "epoch": 0.05, + "learning_rate": 4.200763775231861e-06, + "loss": 0.1493, + "step": 385 + }, + { + "epoch": 0.05, + "learning_rate": 4.211674849972723e-06, + "loss": 0.1547, + "step": 386 + }, + { + "epoch": 0.05, + "learning_rate": 4.222585924713585e-06, + "loss": 0.1645, + "step": 387 + }, + { + "epoch": 0.05, + "learning_rate": 4.233496999454447e-06, + "loss": 0.1537, + "step": 388 + }, + { + "epoch": 0.05, + "learning_rate": 4.244408074195309e-06, + "loss": 0.163, + "step": 389 + }, + { + "epoch": 0.05, + "learning_rate": 4.255319148936171e-06, + "loss": 0.1511, + "step": 390 + }, + { + "epoch": 0.05, + "learning_rate": 4.2662302236770325e-06, + "loss": 0.1599, + "step": 391 + }, + { + "epoch": 0.05, + "learning_rate": 4.2771412984178945e-06, + "loss": 0.1646, + "step": 392 + }, + { + "epoch": 0.05, + "learning_rate": 4.288052373158756e-06, + "loss": 0.1574, + "step": 393 + }, + { + "epoch": 0.05, + "learning_rate": 4.298963447899618e-06, + "loss": 0.16, + "step": 394 + }, + { + "epoch": 0.05, + "learning_rate": 4.30987452264048e-06, + "loss": 0.1588, + "step": 395 + }, + { + "epoch": 0.05, + "learning_rate": 4.320785597381342e-06, + "loss": 0.1582, + "step": 396 + }, + { + "epoch": 0.05, + "learning_rate": 4.331696672122204e-06, + "loss": 0.1541, + "step": 397 + }, + { + "epoch": 0.05, + "learning_rate": 4.342607746863066e-06, + "loss": 0.1724, + "step": 398 + }, + { + "epoch": 0.05, + "learning_rate": 4.353518821603928e-06, + "loss": 0.149, + "step": 399 + }, + { + "epoch": 0.05, + "learning_rate": 4.36442989634479e-06, + "loss": 0.1734, + "step": 400 + }, + { + "epoch": 0.05, + "learning_rate": 4.375340971085652e-06, + "loss": 0.1516, + "step": 401 + }, + { + "epoch": 0.05, + "learning_rate": 4.386252045826515e-06, + "loss": 0.1567, + "step": 402 + }, + { + "epoch": 0.05, + "learning_rate": 4.397163120567377e-06, + "loss": 0.1575, + "step": 403 + }, + { + "epoch": 0.05, + "learning_rate": 4.4080741953082385e-06, + "loss": 0.1499, + "step": 404 + }, + { + "epoch": 0.05, + "learning_rate": 4.4189852700491005e-06, + "loss": 0.159, + "step": 405 + }, + { + "epoch": 0.05, + "learning_rate": 4.429896344789962e-06, + "loss": 0.1727, + "step": 406 + }, + { + "epoch": 0.05, + "learning_rate": 4.440807419530824e-06, + "loss": 0.1581, + "step": 407 + }, + { + "epoch": 0.05, + "learning_rate": 4.451718494271686e-06, + "loss": 0.147, + "step": 408 + }, + { + "epoch": 0.05, + "learning_rate": 4.462629569012548e-06, + "loss": 0.1563, + "step": 409 + }, + { + "epoch": 0.05, + "learning_rate": 4.47354064375341e-06, + "loss": 0.1422, + "step": 410 + }, + { + "epoch": 0.05, + "learning_rate": 4.484451718494272e-06, + "loss": 0.1569, + "step": 411 + }, + { + "epoch": 0.05, + "learning_rate": 4.495362793235134e-06, + "loss": 0.1591, + "step": 412 + }, + { + "epoch": 0.05, + "learning_rate": 4.506273867975996e-06, + "loss": 0.1489, + "step": 413 + }, + { + "epoch": 0.05, + "learning_rate": 4.517184942716858e-06, + "loss": 0.1446, + "step": 414 + }, + { + "epoch": 0.05, + "learning_rate": 4.52809601745772e-06, + "loss": 0.1375, + "step": 415 + }, + { + "epoch": 0.05, + "learning_rate": 4.539007092198582e-06, + "loss": 0.1704, + "step": 416 + }, + { + "epoch": 0.05, + "learning_rate": 4.549918166939444e-06, + "loss": 0.1701, + "step": 417 + }, + { + "epoch": 0.05, + "learning_rate": 4.560829241680306e-06, + "loss": 0.1589, + "step": 418 + }, + { + "epoch": 0.05, + "learning_rate": 4.5717403164211676e-06, + "loss": 0.1628, + "step": 419 + }, + { + "epoch": 0.06, + "learning_rate": 4.5826513911620295e-06, + "loss": 0.1577, + "step": 420 + }, + { + "epoch": 0.06, + "learning_rate": 4.5935624659028914e-06, + "loss": 0.1626, + "step": 421 + }, + { + "epoch": 0.06, + "learning_rate": 4.604473540643753e-06, + "loss": 0.1467, + "step": 422 + }, + { + "epoch": 0.06, + "learning_rate": 4.615384615384616e-06, + "loss": 0.1582, + "step": 423 + }, + { + "epoch": 0.06, + "learning_rate": 4.626295690125478e-06, + "loss": 0.1656, + "step": 424 + }, + { + "epoch": 0.06, + "learning_rate": 4.63720676486634e-06, + "loss": 0.1547, + "step": 425 + }, + { + "epoch": 0.06, + "learning_rate": 4.648117839607202e-06, + "loss": 0.1643, + "step": 426 + }, + { + "epoch": 0.06, + "learning_rate": 4.659028914348064e-06, + "loss": 0.1426, + "step": 427 + }, + { + "epoch": 0.06, + "learning_rate": 4.669939989088926e-06, + "loss": 0.154, + "step": 428 + }, + { + "epoch": 0.06, + "learning_rate": 4.680851063829788e-06, + "loss": 0.1672, + "step": 429 + }, + { + "epoch": 0.06, + "learning_rate": 4.69176213857065e-06, + "loss": 0.1513, + "step": 430 + }, + { + "epoch": 0.06, + "learning_rate": 4.702673213311512e-06, + "loss": 0.152, + "step": 431 + }, + { + "epoch": 0.06, + "learning_rate": 4.7135842880523736e-06, + "loss": 0.1546, + "step": 432 + }, + { + "epoch": 0.06, + "learning_rate": 4.7244953627932355e-06, + "loss": 0.1526, + "step": 433 + }, + { + "epoch": 0.06, + "learning_rate": 4.7354064375340974e-06, + "loss": 0.1621, + "step": 434 + }, + { + "epoch": 0.06, + "learning_rate": 4.746317512274959e-06, + "loss": 0.1463, + "step": 435 + }, + { + "epoch": 0.06, + "learning_rate": 4.757228587015821e-06, + "loss": 0.1757, + "step": 436 + }, + { + "epoch": 0.06, + "learning_rate": 4.768139661756683e-06, + "loss": 0.154, + "step": 437 + }, + { + "epoch": 0.06, + "learning_rate": 4.779050736497545e-06, + "loss": 0.1557, + "step": 438 + }, + { + "epoch": 0.06, + "learning_rate": 4.789961811238407e-06, + "loss": 0.1624, + "step": 439 + }, + { + "epoch": 0.06, + "learning_rate": 4.800872885979269e-06, + "loss": 0.1655, + "step": 440 + }, + { + "epoch": 0.06, + "learning_rate": 4.811783960720131e-06, + "loss": 0.1459, + "step": 441 + }, + { + "epoch": 0.06, + "learning_rate": 4.822695035460993e-06, + "loss": 0.1521, + "step": 442 + }, + { + "epoch": 0.06, + "learning_rate": 4.833606110201856e-06, + "loss": 0.1506, + "step": 443 + }, + { + "epoch": 0.06, + "learning_rate": 4.844517184942718e-06, + "loss": 0.1491, + "step": 444 + }, + { + "epoch": 0.06, + "learning_rate": 4.8554282596835796e-06, + "loss": 0.1458, + "step": 445 + }, + { + "epoch": 0.06, + "learning_rate": 4.8663393344244415e-06, + "loss": 0.1537, + "step": 446 + }, + { + "epoch": 0.06, + "learning_rate": 4.8772504091653034e-06, + "loss": 0.1462, + "step": 447 + }, + { + "epoch": 0.06, + "learning_rate": 4.888161483906165e-06, + "loss": 0.1392, + "step": 448 + }, + { + "epoch": 0.06, + "learning_rate": 4.899072558647027e-06, + "loss": 0.1414, + "step": 449 + }, + { + "epoch": 0.06, + "learning_rate": 4.909983633387889e-06, + "loss": 0.1524, + "step": 450 + }, + { + "epoch": 0.06, + "learning_rate": 4.920894708128751e-06, + "loss": 0.1673, + "step": 451 + }, + { + "epoch": 0.06, + "learning_rate": 4.931805782869613e-06, + "loss": 0.1492, + "step": 452 + }, + { + "epoch": 0.06, + "learning_rate": 4.942716857610475e-06, + "loss": 0.1478, + "step": 453 + }, + { + "epoch": 0.06, + "learning_rate": 4.953627932351337e-06, + "loss": 0.1454, + "step": 454 + }, + { + "epoch": 0.06, + "learning_rate": 4.964539007092199e-06, + "loss": 0.1505, + "step": 455 + }, + { + "epoch": 0.06, + "learning_rate": 4.975450081833061e-06, + "loss": 0.1556, + "step": 456 + }, + { + "epoch": 0.06, + "learning_rate": 4.986361156573923e-06, + "loss": 0.1353, + "step": 457 + }, + { + "epoch": 0.06, + "learning_rate": 4.997272231314785e-06, + "loss": 0.1477, + "step": 458 + }, + { + "epoch": 0.06, + "learning_rate": 5.0081833060556475e-06, + "loss": 0.1446, + "step": 459 + }, + { + "epoch": 0.06, + "learning_rate": 5.019094380796509e-06, + "loss": 0.1473, + "step": 460 + }, + { + "epoch": 0.06, + "learning_rate": 5.030005455537371e-06, + "loss": 0.1561, + "step": 461 + }, + { + "epoch": 0.06, + "learning_rate": 5.0409165302782325e-06, + "loss": 0.1601, + "step": 462 + }, + { + "epoch": 0.06, + "learning_rate": 5.051827605019095e-06, + "loss": 0.1527, + "step": 463 + }, + { + "epoch": 0.06, + "learning_rate": 5.062738679759956e-06, + "loss": 0.1569, + "step": 464 + }, + { + "epoch": 0.06, + "learning_rate": 5.073649754500819e-06, + "loss": 0.1489, + "step": 465 + }, + { + "epoch": 0.06, + "learning_rate": 5.08456082924168e-06, + "loss": 0.1456, + "step": 466 + }, + { + "epoch": 0.06, + "learning_rate": 5.095471903982543e-06, + "loss": 0.1525, + "step": 467 + }, + { + "epoch": 0.06, + "learning_rate": 5.106382978723404e-06, + "loss": 0.1482, + "step": 468 + }, + { + "epoch": 0.06, + "learning_rate": 5.117294053464267e-06, + "loss": 0.1344, + "step": 469 + }, + { + "epoch": 0.06, + "learning_rate": 5.128205128205128e-06, + "loss": 0.1434, + "step": 470 + }, + { + "epoch": 0.06, + "learning_rate": 5.139116202945991e-06, + "loss": 0.1423, + "step": 471 + }, + { + "epoch": 0.06, + "learning_rate": 5.150027277686852e-06, + "loss": 0.159, + "step": 472 + }, + { + "epoch": 0.06, + "learning_rate": 5.160938352427715e-06, + "loss": 0.1428, + "step": 473 + }, + { + "epoch": 0.06, + "learning_rate": 5.171849427168577e-06, + "loss": 0.1445, + "step": 474 + }, + { + "epoch": 0.06, + "learning_rate": 5.1827605019094385e-06, + "loss": 0.1397, + "step": 475 + }, + { + "epoch": 0.06, + "learning_rate": 5.193671576650301e-06, + "loss": 0.1508, + "step": 476 + }, + { + "epoch": 0.06, + "learning_rate": 5.204582651391162e-06, + "loss": 0.1548, + "step": 477 + }, + { + "epoch": 0.06, + "learning_rate": 5.215493726132025e-06, + "loss": 0.1598, + "step": 478 + }, + { + "epoch": 0.06, + "learning_rate": 5.226404800872886e-06, + "loss": 0.151, + "step": 479 + }, + { + "epoch": 0.06, + "learning_rate": 5.237315875613749e-06, + "loss": 0.1419, + "step": 480 + }, + { + "epoch": 0.06, + "learning_rate": 5.24822695035461e-06, + "loss": 0.1437, + "step": 481 + }, + { + "epoch": 0.06, + "learning_rate": 5.259138025095473e-06, + "loss": 0.1496, + "step": 482 + }, + { + "epoch": 0.06, + "learning_rate": 5.270049099836334e-06, + "loss": 0.1436, + "step": 483 + }, + { + "epoch": 0.06, + "learning_rate": 5.280960174577197e-06, + "loss": 0.1431, + "step": 484 + }, + { + "epoch": 0.06, + "learning_rate": 5.291871249318058e-06, + "loss": 0.1469, + "step": 485 + }, + { + "epoch": 0.06, + "learning_rate": 5.302782324058921e-06, + "loss": 0.1576, + "step": 486 + }, + { + "epoch": 0.06, + "learning_rate": 5.313693398799782e-06, + "loss": 0.1417, + "step": 487 + }, + { + "epoch": 0.06, + "learning_rate": 5.3246044735406444e-06, + "loss": 0.1628, + "step": 488 + }, + { + "epoch": 0.06, + "learning_rate": 5.3355155482815055e-06, + "loss": 0.1541, + "step": 489 + }, + { + "epoch": 0.06, + "learning_rate": 5.346426623022368e-06, + "loss": 0.1505, + "step": 490 + }, + { + "epoch": 0.06, + "learning_rate": 5.357337697763229e-06, + "loss": 0.1621, + "step": 491 + }, + { + "epoch": 0.06, + "learning_rate": 5.368248772504092e-06, + "loss": 0.146, + "step": 492 + }, + { + "epoch": 0.06, + "learning_rate": 5.379159847244953e-06, + "loss": 0.1336, + "step": 493 + }, + { + "epoch": 0.06, + "learning_rate": 5.390070921985816e-06, + "loss": 0.1498, + "step": 494 + }, + { + "epoch": 0.06, + "learning_rate": 5.400981996726679e-06, + "loss": 0.1551, + "step": 495 + }, + { + "epoch": 0.06, + "learning_rate": 5.41189307146754e-06, + "loss": 0.1473, + "step": 496 + }, + { + "epoch": 0.07, + "learning_rate": 5.422804146208403e-06, + "loss": 0.14, + "step": 497 + }, + { + "epoch": 0.07, + "learning_rate": 5.433715220949264e-06, + "loss": 0.1442, + "step": 498 + }, + { + "epoch": 0.07, + "learning_rate": 5.4446262956901266e-06, + "loss": 0.1357, + "step": 499 + }, + { + "epoch": 0.07, + "learning_rate": 5.455537370430988e-06, + "loss": 0.1398, + "step": 500 + }, + { + "epoch": 0.07, + "learning_rate": 5.4664484451718504e-06, + "loss": 0.1485, + "step": 501 + }, + { + "epoch": 0.07, + "learning_rate": 5.4773595199127115e-06, + "loss": 0.1413, + "step": 502 + }, + { + "epoch": 0.07, + "learning_rate": 5.488270594653574e-06, + "loss": 0.1295, + "step": 503 + }, + { + "epoch": 0.07, + "learning_rate": 5.499181669394435e-06, + "loss": 0.1273, + "step": 504 + }, + { + "epoch": 0.07, + "learning_rate": 5.510092744135298e-06, + "loss": 0.148, + "step": 505 + }, + { + "epoch": 0.07, + "learning_rate": 5.521003818876159e-06, + "loss": 0.1486, + "step": 506 + }, + { + "epoch": 0.07, + "learning_rate": 5.531914893617022e-06, + "loss": 0.1628, + "step": 507 + }, + { + "epoch": 0.07, + "learning_rate": 5.542825968357883e-06, + "loss": 0.1518, + "step": 508 + }, + { + "epoch": 0.07, + "learning_rate": 5.553737043098746e-06, + "loss": 0.1416, + "step": 509 + }, + { + "epoch": 0.07, + "learning_rate": 5.564648117839607e-06, + "loss": 0.1548, + "step": 510 + }, + { + "epoch": 0.07, + "learning_rate": 5.57555919258047e-06, + "loss": 0.1278, + "step": 511 + }, + { + "epoch": 0.07, + "learning_rate": 5.586470267321331e-06, + "loss": 0.1399, + "step": 512 + }, + { + "epoch": 0.07, + "learning_rate": 5.597381342062194e-06, + "loss": 0.1309, + "step": 513 + }, + { + "epoch": 0.07, + "learning_rate": 5.608292416803055e-06, + "loss": 0.1438, + "step": 514 + }, + { + "epoch": 0.07, + "learning_rate": 5.6192034915439175e-06, + "loss": 0.1534, + "step": 515 + }, + { + "epoch": 0.07, + "learning_rate": 5.63011456628478e-06, + "loss": 0.1472, + "step": 516 + }, + { + "epoch": 0.07, + "learning_rate": 5.641025641025641e-06, + "loss": 0.1566, + "step": 517 + }, + { + "epoch": 0.07, + "learning_rate": 5.651936715766504e-06, + "loss": 0.1374, + "step": 518 + }, + { + "epoch": 0.07, + "learning_rate": 5.662847790507365e-06, + "loss": 0.1531, + "step": 519 + }, + { + "epoch": 0.07, + "learning_rate": 5.673758865248228e-06, + "loss": 0.1586, + "step": 520 + }, + { + "epoch": 0.07, + "learning_rate": 5.684669939989089e-06, + "loss": 0.1266, + "step": 521 + }, + { + "epoch": 0.07, + "learning_rate": 5.695581014729952e-06, + "loss": 0.1429, + "step": 522 + }, + { + "epoch": 0.07, + "learning_rate": 5.706492089470813e-06, + "loss": 0.1385, + "step": 523 + }, + { + "epoch": 0.07, + "learning_rate": 5.717403164211676e-06, + "loss": 0.1312, + "step": 524 + }, + { + "epoch": 0.07, + "learning_rate": 5.728314238952537e-06, + "loss": 0.1429, + "step": 525 + }, + { + "epoch": 0.07, + "learning_rate": 5.7392253136934e-06, + "loss": 0.1415, + "step": 526 + }, + { + "epoch": 0.07, + "learning_rate": 5.750136388434261e-06, + "loss": 0.1497, + "step": 527 + }, + { + "epoch": 0.07, + "learning_rate": 5.7610474631751235e-06, + "loss": 0.1454, + "step": 528 + }, + { + "epoch": 0.07, + "learning_rate": 5.771958537915985e-06, + "loss": 0.1477, + "step": 529 + }, + { + "epoch": 0.07, + "learning_rate": 5.782869612656847e-06, + "loss": 0.1496, + "step": 530 + }, + { + "epoch": 0.07, + "learning_rate": 5.7937806873977085e-06, + "loss": 0.1417, + "step": 531 + }, + { + "epoch": 0.07, + "learning_rate": 5.804691762138571e-06, + "loss": 0.1305, + "step": 532 + }, + { + "epoch": 0.07, + "learning_rate": 5.815602836879432e-06, + "loss": 0.1403, + "step": 533 + }, + { + "epoch": 0.07, + "learning_rate": 5.826513911620295e-06, + "loss": 0.1554, + "step": 534 + }, + { + "epoch": 0.07, + "learning_rate": 5.837424986361156e-06, + "loss": 0.1243, + "step": 535 + }, + { + "epoch": 0.07, + "learning_rate": 5.848336061102019e-06, + "loss": 0.1432, + "step": 536 + }, + { + "epoch": 0.07, + "learning_rate": 5.859247135842882e-06, + "loss": 0.1422, + "step": 537 + }, + { + "epoch": 0.07, + "learning_rate": 5.870158210583743e-06, + "loss": 0.1345, + "step": 538 + }, + { + "epoch": 0.07, + "learning_rate": 5.881069285324606e-06, + "loss": 0.16, + "step": 539 + }, + { + "epoch": 0.07, + "learning_rate": 5.891980360065467e-06, + "loss": 0.1469, + "step": 540 + }, + { + "epoch": 0.07, + "learning_rate": 5.9028914348063295e-06, + "loss": 0.1382, + "step": 541 + }, + { + "epoch": 0.07, + "learning_rate": 5.913802509547191e-06, + "loss": 0.1369, + "step": 542 + }, + { + "epoch": 0.07, + "learning_rate": 5.924713584288053e-06, + "loss": 0.1383, + "step": 543 + }, + { + "epoch": 0.07, + "learning_rate": 5.9356246590289145e-06, + "loss": 0.1369, + "step": 544 + }, + { + "epoch": 0.07, + "learning_rate": 5.946535733769777e-06, + "loss": 0.1598, + "step": 545 + }, + { + "epoch": 0.07, + "learning_rate": 5.957446808510638e-06, + "loss": 0.1347, + "step": 546 + }, + { + "epoch": 0.07, + "learning_rate": 5.968357883251501e-06, + "loss": 0.1434, + "step": 547 + }, + { + "epoch": 0.07, + "learning_rate": 5.979268957992362e-06, + "loss": 0.1243, + "step": 548 + }, + { + "epoch": 0.07, + "learning_rate": 5.990180032733225e-06, + "loss": 0.1364, + "step": 549 + }, + { + "epoch": 0.07, + "learning_rate": 6.001091107474086e-06, + "loss": 0.1351, + "step": 550 + }, + { + "epoch": 0.07, + "learning_rate": 6.012002182214949e-06, + "loss": 0.1533, + "step": 551 + }, + { + "epoch": 0.07, + "learning_rate": 6.02291325695581e-06, + "loss": 0.13, + "step": 552 + }, + { + "epoch": 0.07, + "learning_rate": 6.033824331696673e-06, + "loss": 0.1418, + "step": 553 + }, + { + "epoch": 0.07, + "learning_rate": 6.044735406437534e-06, + "loss": 0.1534, + "step": 554 + }, + { + "epoch": 0.07, + "learning_rate": 6.055646481178397e-06, + "loss": 0.12, + "step": 555 + }, + { + "epoch": 0.07, + "learning_rate": 6.066557555919258e-06, + "loss": 0.1429, + "step": 556 + }, + { + "epoch": 0.07, + "learning_rate": 6.0774686306601205e-06, + "loss": 0.1345, + "step": 557 + }, + { + "epoch": 0.07, + "learning_rate": 6.088379705400983e-06, + "loss": 0.1467, + "step": 558 + }, + { + "epoch": 0.07, + "learning_rate": 6.099290780141844e-06, + "loss": 0.1484, + "step": 559 + }, + { + "epoch": 0.07, + "learning_rate": 6.110201854882707e-06, + "loss": 0.145, + "step": 560 + }, + { + "epoch": 0.07, + "learning_rate": 6.121112929623568e-06, + "loss": 0.1564, + "step": 561 + }, + { + "epoch": 0.07, + "learning_rate": 6.132024004364431e-06, + "loss": 0.1428, + "step": 562 + }, + { + "epoch": 0.07, + "learning_rate": 6.142935079105292e-06, + "loss": 0.133, + "step": 563 + }, + { + "epoch": 0.07, + "learning_rate": 6.153846153846155e-06, + "loss": 0.124, + "step": 564 + }, + { + "epoch": 0.07, + "learning_rate": 6.164757228587016e-06, + "loss": 0.14, + "step": 565 + }, + { + "epoch": 0.07, + "learning_rate": 6.175668303327879e-06, + "loss": 0.1405, + "step": 566 + }, + { + "epoch": 0.07, + "learning_rate": 6.18657937806874e-06, + "loss": 0.1418, + "step": 567 + }, + { + "epoch": 0.07, + "learning_rate": 6.197490452809603e-06, + "loss": 0.1464, + "step": 568 + }, + { + "epoch": 0.07, + "learning_rate": 6.208401527550464e-06, + "loss": 0.1449, + "step": 569 + }, + { + "epoch": 0.07, + "learning_rate": 6.2193126022913265e-06, + "loss": 0.1332, + "step": 570 + }, + { + "epoch": 0.07, + "learning_rate": 6.2302236770321876e-06, + "loss": 0.1284, + "step": 571 + }, + { + "epoch": 0.07, + "learning_rate": 6.24113475177305e-06, + "loss": 0.1447, + "step": 572 + }, + { + "epoch": 0.08, + "learning_rate": 6.2520458265139114e-06, + "loss": 0.1405, + "step": 573 + }, + { + "epoch": 0.08, + "learning_rate": 6.262956901254774e-06, + "loss": 0.1289, + "step": 574 + }, + { + "epoch": 0.08, + "learning_rate": 6.273867975995635e-06, + "loss": 0.1369, + "step": 575 + }, + { + "epoch": 0.08, + "learning_rate": 6.284779050736498e-06, + "loss": 0.1331, + "step": 576 + }, + { + "epoch": 0.08, + "learning_rate": 6.295690125477361e-06, + "loss": 0.1344, + "step": 577 + }, + { + "epoch": 0.08, + "learning_rate": 6.306601200218222e-06, + "loss": 0.1571, + "step": 578 + }, + { + "epoch": 0.08, + "learning_rate": 6.317512274959085e-06, + "loss": 0.1448, + "step": 579 + }, + { + "epoch": 0.08, + "learning_rate": 6.328423349699946e-06, + "loss": 0.1347, + "step": 580 + }, + { + "epoch": 0.08, + "learning_rate": 6.339334424440809e-06, + "loss": 0.1435, + "step": 581 + }, + { + "epoch": 0.08, + "learning_rate": 6.35024549918167e-06, + "loss": 0.1324, + "step": 582 + }, + { + "epoch": 0.08, + "learning_rate": 6.3611565739225325e-06, + "loss": 0.151, + "step": 583 + }, + { + "epoch": 0.08, + "learning_rate": 6.3720676486633936e-06, + "loss": 0.1358, + "step": 584 + }, + { + "epoch": 0.08, + "learning_rate": 6.382978723404256e-06, + "loss": 0.1306, + "step": 585 + }, + { + "epoch": 0.08, + "learning_rate": 6.3938897981451174e-06, + "loss": 0.137, + "step": 586 + }, + { + "epoch": 0.08, + "learning_rate": 6.40480087288598e-06, + "loss": 0.143, + "step": 587 + }, + { + "epoch": 0.08, + "learning_rate": 6.415711947626841e-06, + "loss": 0.145, + "step": 588 + }, + { + "epoch": 0.08, + "learning_rate": 6.426623022367704e-06, + "loss": 0.126, + "step": 589 + }, + { + "epoch": 0.08, + "learning_rate": 6.437534097108565e-06, + "loss": 0.1407, + "step": 590 + }, + { + "epoch": 0.08, + "learning_rate": 6.448445171849428e-06, + "loss": 0.1443, + "step": 591 + }, + { + "epoch": 0.08, + "learning_rate": 6.459356246590289e-06, + "loss": 0.1286, + "step": 592 + }, + { + "epoch": 0.08, + "learning_rate": 6.470267321331152e-06, + "loss": 0.1322, + "step": 593 + }, + { + "epoch": 0.08, + "learning_rate": 6.481178396072013e-06, + "loss": 0.1452, + "step": 594 + }, + { + "epoch": 0.08, + "learning_rate": 6.492089470812876e-06, + "loss": 0.1428, + "step": 595 + }, + { + "epoch": 0.08, + "learning_rate": 6.503000545553737e-06, + "loss": 0.1222, + "step": 596 + }, + { + "epoch": 0.08, + "learning_rate": 6.5139116202945996e-06, + "loss": 0.1334, + "step": 597 + }, + { + "epoch": 0.08, + "learning_rate": 6.524822695035462e-06, + "loss": 0.1251, + "step": 598 + }, + { + "epoch": 0.08, + "learning_rate": 6.5357337697763234e-06, + "loss": 0.1287, + "step": 599 + }, + { + "epoch": 0.08, + "learning_rate": 6.546644844517186e-06, + "loss": 0.1285, + "step": 600 + }, + { + "epoch": 0.08, + "learning_rate": 6.557555919258047e-06, + "loss": 0.1414, + "step": 601 + }, + { + "epoch": 0.08, + "learning_rate": 6.56846699399891e-06, + "loss": 0.138, + "step": 602 + }, + { + "epoch": 0.08, + "learning_rate": 6.579378068739771e-06, + "loss": 0.1325, + "step": 603 + }, + { + "epoch": 0.08, + "learning_rate": 6.590289143480634e-06, + "loss": 0.127, + "step": 604 + }, + { + "epoch": 0.08, + "learning_rate": 6.601200218221495e-06, + "loss": 0.1366, + "step": 605 + }, + { + "epoch": 0.08, + "learning_rate": 6.612111292962358e-06, + "loss": 0.1379, + "step": 606 + }, + { + "epoch": 0.08, + "learning_rate": 6.623022367703219e-06, + "loss": 0.1452, + "step": 607 + }, + { + "epoch": 0.08, + "learning_rate": 6.633933442444082e-06, + "loss": 0.1461, + "step": 608 + }, + { + "epoch": 0.08, + "learning_rate": 6.644844517184943e-06, + "loss": 0.1298, + "step": 609 + }, + { + "epoch": 0.08, + "learning_rate": 6.6557555919258056e-06, + "loss": 0.1308, + "step": 610 + }, + { + "epoch": 0.08, + "learning_rate": 6.666666666666667e-06, + "loss": 0.135, + "step": 611 + }, + { + "epoch": 0.08, + "learning_rate": 6.6775777414075294e-06, + "loss": 0.1388, + "step": 612 + }, + { + "epoch": 0.08, + "learning_rate": 6.6884888161483905e-06, + "loss": 0.1464, + "step": 613 + }, + { + "epoch": 0.08, + "learning_rate": 6.699399890889253e-06, + "loss": 0.1397, + "step": 614 + }, + { + "epoch": 0.08, + "learning_rate": 6.710310965630114e-06, + "loss": 0.1495, + "step": 615 + }, + { + "epoch": 0.08, + "learning_rate": 6.721222040370977e-06, + "loss": 0.1372, + "step": 616 + }, + { + "epoch": 0.08, + "learning_rate": 6.732133115111838e-06, + "loss": 0.1357, + "step": 617 + }, + { + "epoch": 0.08, + "learning_rate": 6.743044189852701e-06, + "loss": 0.1338, + "step": 618 + }, + { + "epoch": 0.08, + "learning_rate": 6.753955264593564e-06, + "loss": 0.146, + "step": 619 + }, + { + "epoch": 0.08, + "learning_rate": 6.764866339334425e-06, + "loss": 0.1362, + "step": 620 + }, + { + "epoch": 0.08, + "learning_rate": 6.775777414075288e-06, + "loss": 0.1235, + "step": 621 + }, + { + "epoch": 0.08, + "learning_rate": 6.786688488816149e-06, + "loss": 0.1506, + "step": 622 + }, + { + "epoch": 0.08, + "learning_rate": 6.7975995635570116e-06, + "loss": 0.1467, + "step": 623 + }, + { + "epoch": 0.08, + "learning_rate": 6.808510638297873e-06, + "loss": 0.1417, + "step": 624 + }, + { + "epoch": 0.08, + "learning_rate": 6.8194217130387354e-06, + "loss": 0.1376, + "step": 625 + }, + { + "epoch": 0.08, + "learning_rate": 6.8303327877795965e-06, + "loss": 0.1572, + "step": 626 + }, + { + "epoch": 0.08, + "learning_rate": 6.841243862520459e-06, + "loss": 0.148, + "step": 627 + }, + { + "epoch": 0.08, + "learning_rate": 6.85215493726132e-06, + "loss": 0.1426, + "step": 628 + }, + { + "epoch": 0.08, + "learning_rate": 6.863066012002183e-06, + "loss": 0.1355, + "step": 629 + }, + { + "epoch": 0.08, + "learning_rate": 6.873977086743044e-06, + "loss": 0.134, + "step": 630 + }, + { + "epoch": 0.08, + "learning_rate": 6.884888161483907e-06, + "loss": 0.1348, + "step": 631 + }, + { + "epoch": 0.08, + "learning_rate": 6.895799236224768e-06, + "loss": 0.1213, + "step": 632 + }, + { + "epoch": 0.08, + "learning_rate": 6.906710310965631e-06, + "loss": 0.1335, + "step": 633 + }, + { + "epoch": 0.08, + "learning_rate": 6.917621385706492e-06, + "loss": 0.1403, + "step": 634 + }, + { + "epoch": 0.08, + "learning_rate": 6.928532460447355e-06, + "loss": 0.1327, + "step": 635 + }, + { + "epoch": 0.08, + "learning_rate": 6.939443535188216e-06, + "loss": 0.1334, + "step": 636 + }, + { + "epoch": 0.08, + "learning_rate": 6.950354609929079e-06, + "loss": 0.1456, + "step": 637 + }, + { + "epoch": 0.08, + "learning_rate": 6.96126568466994e-06, + "loss": 0.1086, + "step": 638 + }, + { + "epoch": 0.08, + "learning_rate": 6.9721767594108025e-06, + "loss": 0.1337, + "step": 639 + }, + { + "epoch": 0.08, + "learning_rate": 6.983087834151665e-06, + "loss": 0.1463, + "step": 640 + }, + { + "epoch": 0.08, + "learning_rate": 6.993998908892526e-06, + "loss": 0.1479, + "step": 641 + }, + { + "epoch": 0.08, + "learning_rate": 7.004909983633389e-06, + "loss": 0.1445, + "step": 642 + }, + { + "epoch": 0.08, + "learning_rate": 7.01582105837425e-06, + "loss": 0.1413, + "step": 643 + }, + { + "epoch": 0.08, + "learning_rate": 7.026732133115113e-06, + "loss": 0.1377, + "step": 644 + }, + { + "epoch": 0.08, + "learning_rate": 7.037643207855974e-06, + "loss": 0.1359, + "step": 645 + }, + { + "epoch": 0.08, + "learning_rate": 7.048554282596837e-06, + "loss": 0.1389, + "step": 646 + }, + { + "epoch": 0.08, + "learning_rate": 7.059465357337698e-06, + "loss": 0.1362, + "step": 647 + }, + { + "epoch": 0.08, + "learning_rate": 7.070376432078561e-06, + "loss": 0.1505, + "step": 648 + }, + { + "epoch": 0.09, + "learning_rate": 7.081287506819422e-06, + "loss": 0.1319, + "step": 649 + }, + { + "epoch": 0.09, + "learning_rate": 7.092198581560285e-06, + "loss": 0.125, + "step": 650 + }, + { + "epoch": 0.09, + "learning_rate": 7.103109656301146e-06, + "loss": 0.1332, + "step": 651 + }, + { + "epoch": 0.09, + "learning_rate": 7.1140207310420085e-06, + "loss": 0.122, + "step": 652 + }, + { + "epoch": 0.09, + "learning_rate": 7.12493180578287e-06, + "loss": 0.1302, + "step": 653 + }, + { + "epoch": 0.09, + "learning_rate": 7.135842880523732e-06, + "loss": 0.1391, + "step": 654 + }, + { + "epoch": 0.09, + "learning_rate": 7.1467539552645935e-06, + "loss": 0.1438, + "step": 655 + }, + { + "epoch": 0.09, + "learning_rate": 7.157665030005456e-06, + "loss": 0.1419, + "step": 656 + }, + { + "epoch": 0.09, + "learning_rate": 7.168576104746317e-06, + "loss": 0.1238, + "step": 657 + }, + { + "epoch": 0.09, + "learning_rate": 7.17948717948718e-06, + "loss": 0.1394, + "step": 658 + }, + { + "epoch": 0.09, + "learning_rate": 7.190398254228041e-06, + "loss": 0.1458, + "step": 659 + }, + { + "epoch": 0.09, + "learning_rate": 7.201309328968904e-06, + "loss": 0.1254, + "step": 660 + }, + { + "epoch": 0.09, + "learning_rate": 7.212220403709767e-06, + "loss": 0.1311, + "step": 661 + }, + { + "epoch": 0.09, + "learning_rate": 7.223131478450628e-06, + "loss": 0.1335, + "step": 662 + }, + { + "epoch": 0.09, + "learning_rate": 7.234042553191491e-06, + "loss": 0.1426, + "step": 663 + }, + { + "epoch": 0.09, + "learning_rate": 7.244953627932352e-06, + "loss": 0.1362, + "step": 664 + }, + { + "epoch": 0.09, + "learning_rate": 7.2558647026732145e-06, + "loss": 0.1219, + "step": 665 + }, + { + "epoch": 0.09, + "learning_rate": 7.266775777414076e-06, + "loss": 0.1313, + "step": 666 + }, + { + "epoch": 0.09, + "learning_rate": 7.277686852154938e-06, + "loss": 0.1272, + "step": 667 + }, + { + "epoch": 0.09, + "learning_rate": 7.2885979268957995e-06, + "loss": 0.1211, + "step": 668 + }, + { + "epoch": 0.09, + "learning_rate": 7.299509001636662e-06, + "loss": 0.1375, + "step": 669 + }, + { + "epoch": 0.09, + "learning_rate": 7.310420076377523e-06, + "loss": 0.1455, + "step": 670 + }, + { + "epoch": 0.09, + "learning_rate": 7.321331151118386e-06, + "loss": 0.1382, + "step": 671 + }, + { + "epoch": 0.09, + "learning_rate": 7.332242225859247e-06, + "loss": 0.1285, + "step": 672 + }, + { + "epoch": 0.09, + "learning_rate": 7.34315330060011e-06, + "loss": 0.138, + "step": 673 + }, + { + "epoch": 0.09, + "learning_rate": 7.354064375340971e-06, + "loss": 0.1263, + "step": 674 + }, + { + "epoch": 0.09, + "learning_rate": 7.364975450081834e-06, + "loss": 0.1257, + "step": 675 + }, + { + "epoch": 0.09, + "learning_rate": 7.375886524822695e-06, + "loss": 0.1215, + "step": 676 + }, + { + "epoch": 0.09, + "learning_rate": 7.386797599563558e-06, + "loss": 0.1359, + "step": 677 + }, + { + "epoch": 0.09, + "learning_rate": 7.397708674304419e-06, + "loss": 0.1356, + "step": 678 + }, + { + "epoch": 0.09, + "learning_rate": 7.408619749045282e-06, + "loss": 0.1374, + "step": 679 + }, + { + "epoch": 0.09, + "learning_rate": 7.419530823786144e-06, + "loss": 0.1327, + "step": 680 + }, + { + "epoch": 0.09, + "learning_rate": 7.4304418985270055e-06, + "loss": 0.1322, + "step": 681 + }, + { + "epoch": 0.09, + "learning_rate": 7.441352973267868e-06, + "loss": 0.1411, + "step": 682 + }, + { + "epoch": 0.09, + "learning_rate": 7.452264048008729e-06, + "loss": 0.1385, + "step": 683 + }, + { + "epoch": 0.09, + "learning_rate": 7.463175122749592e-06, + "loss": 0.1265, + "step": 684 + }, + { + "epoch": 0.09, + "learning_rate": 7.474086197490453e-06, + "loss": 0.1231, + "step": 685 + }, + { + "epoch": 0.09, + "learning_rate": 7.484997272231316e-06, + "loss": 0.1294, + "step": 686 + }, + { + "epoch": 0.09, + "learning_rate": 7.495908346972177e-06, + "loss": 0.1314, + "step": 687 + }, + { + "epoch": 0.09, + "learning_rate": 7.50681942171304e-06, + "loss": 0.1193, + "step": 688 + }, + { + "epoch": 0.09, + "learning_rate": 7.517730496453901e-06, + "loss": 0.1341, + "step": 689 + }, + { + "epoch": 0.09, + "learning_rate": 7.528641571194764e-06, + "loss": 0.1319, + "step": 690 + }, + { + "epoch": 0.09, + "learning_rate": 7.539552645935625e-06, + "loss": 0.1427, + "step": 691 + }, + { + "epoch": 0.09, + "learning_rate": 7.550463720676488e-06, + "loss": 0.1263, + "step": 692 + }, + { + "epoch": 0.09, + "learning_rate": 7.561374795417349e-06, + "loss": 0.1292, + "step": 693 + }, + { + "epoch": 0.09, + "learning_rate": 7.5722858701582115e-06, + "loss": 0.1259, + "step": 694 + }, + { + "epoch": 0.09, + "learning_rate": 7.5831969448990725e-06, + "loss": 0.1345, + "step": 695 + }, + { + "epoch": 0.09, + "learning_rate": 7.594108019639935e-06, + "loss": 0.1352, + "step": 696 + }, + { + "epoch": 0.09, + "learning_rate": 7.605019094380796e-06, + "loss": 0.1297, + "step": 697 + }, + { + "epoch": 0.09, + "learning_rate": 7.615930169121659e-06, + "loss": 0.1453, + "step": 698 + }, + { + "epoch": 0.09, + "learning_rate": 7.62684124386252e-06, + "loss": 0.1413, + "step": 699 + }, + { + "epoch": 0.09, + "learning_rate": 7.637752318603383e-06, + "loss": 0.1277, + "step": 700 + }, + { + "epoch": 0.09, + "learning_rate": 7.648663393344246e-06, + "loss": 0.1281, + "step": 701 + }, + { + "epoch": 0.09, + "learning_rate": 7.659574468085107e-06, + "loss": 0.1269, + "step": 702 + }, + { + "epoch": 0.09, + "learning_rate": 7.67048554282597e-06, + "loss": 0.1423, + "step": 703 + }, + { + "epoch": 0.09, + "learning_rate": 7.68139661756683e-06, + "loss": 0.1301, + "step": 704 + }, + { + "epoch": 0.09, + "learning_rate": 7.692307692307694e-06, + "loss": 0.1462, + "step": 705 + }, + { + "epoch": 0.09, + "learning_rate": 7.703218767048555e-06, + "loss": 0.1528, + "step": 706 + }, + { + "epoch": 0.09, + "learning_rate": 7.714129841789417e-06, + "loss": 0.1342, + "step": 707 + }, + { + "epoch": 0.09, + "learning_rate": 7.725040916530279e-06, + "loss": 0.1371, + "step": 708 + }, + { + "epoch": 0.09, + "learning_rate": 7.735951991271141e-06, + "loss": 0.1401, + "step": 709 + }, + { + "epoch": 0.09, + "learning_rate": 7.746863066012002e-06, + "loss": 0.1189, + "step": 710 + }, + { + "epoch": 0.09, + "learning_rate": 7.757774140752865e-06, + "loss": 0.1405, + "step": 711 + }, + { + "epoch": 0.09, + "learning_rate": 7.768685215493726e-06, + "loss": 0.1353, + "step": 712 + }, + { + "epoch": 0.09, + "learning_rate": 7.779596290234589e-06, + "loss": 0.1481, + "step": 713 + }, + { + "epoch": 0.09, + "learning_rate": 7.79050736497545e-06, + "loss": 0.1396, + "step": 714 + }, + { + "epoch": 0.09, + "learning_rate": 7.801418439716313e-06, + "loss": 0.1325, + "step": 715 + }, + { + "epoch": 0.09, + "learning_rate": 7.812329514457174e-06, + "loss": 0.1206, + "step": 716 + }, + { + "epoch": 0.09, + "learning_rate": 7.823240589198037e-06, + "loss": 0.1414, + "step": 717 + }, + { + "epoch": 0.09, + "learning_rate": 7.834151663938898e-06, + "loss": 0.1288, + "step": 718 + }, + { + "epoch": 0.09, + "learning_rate": 7.84506273867976e-06, + "loss": 0.1285, + "step": 719 + }, + { + "epoch": 0.09, + "learning_rate": 7.855973813420622e-06, + "loss": 0.1345, + "step": 720 + }, + { + "epoch": 0.09, + "learning_rate": 7.866884888161485e-06, + "loss": 0.1313, + "step": 721 + }, + { + "epoch": 0.09, + "learning_rate": 7.877795962902347e-06, + "loss": 0.1375, + "step": 722 + }, + { + "epoch": 0.09, + "learning_rate": 7.888707037643208e-06, + "loss": 0.1366, + "step": 723 + }, + { + "epoch": 0.09, + "learning_rate": 7.899618112384071e-06, + "loss": 0.1212, + "step": 724 + }, + { + "epoch": 0.09, + "learning_rate": 7.910529187124932e-06, + "loss": 0.1175, + "step": 725 + }, + { + "epoch": 0.1, + "learning_rate": 7.921440261865795e-06, + "loss": 0.1357, + "step": 726 + }, + { + "epoch": 0.1, + "learning_rate": 7.932351336606656e-06, + "loss": 0.1304, + "step": 727 + }, + { + "epoch": 0.1, + "learning_rate": 7.943262411347519e-06, + "loss": 0.1301, + "step": 728 + }, + { + "epoch": 0.1, + "learning_rate": 7.95417348608838e-06, + "loss": 0.1313, + "step": 729 + }, + { + "epoch": 0.1, + "learning_rate": 7.965084560829243e-06, + "loss": 0.1202, + "step": 730 + }, + { + "epoch": 0.1, + "learning_rate": 7.975995635570104e-06, + "loss": 0.1458, + "step": 731 + }, + { + "epoch": 0.1, + "learning_rate": 7.986906710310967e-06, + "loss": 0.1516, + "step": 732 + }, + { + "epoch": 0.1, + "learning_rate": 7.997817785051828e-06, + "loss": 0.1285, + "step": 733 + }, + { + "epoch": 0.1, + "learning_rate": 8.00872885979269e-06, + "loss": 0.1332, + "step": 734 + }, + { + "epoch": 0.1, + "learning_rate": 8.019639934533552e-06, + "loss": 0.1316, + "step": 735 + }, + { + "epoch": 0.1, + "learning_rate": 8.030551009274414e-06, + "loss": 0.1268, + "step": 736 + }, + { + "epoch": 0.1, + "learning_rate": 8.041462084015275e-06, + "loss": 0.1366, + "step": 737 + }, + { + "epoch": 0.1, + "learning_rate": 8.052373158756138e-06, + "loss": 0.1237, + "step": 738 + }, + { + "epoch": 0.1, + "learning_rate": 8.063284233497e-06, + "loss": 0.135, + "step": 739 + }, + { + "epoch": 0.1, + "learning_rate": 8.074195308237862e-06, + "loss": 0.1365, + "step": 740 + }, + { + "epoch": 0.1, + "learning_rate": 8.085106382978723e-06, + "loss": 0.1182, + "step": 741 + }, + { + "epoch": 0.1, + "learning_rate": 8.096017457719586e-06, + "loss": 0.1148, + "step": 742 + }, + { + "epoch": 0.1, + "learning_rate": 8.106928532460449e-06, + "loss": 0.1422, + "step": 743 + }, + { + "epoch": 0.1, + "learning_rate": 8.11783960720131e-06, + "loss": 0.123, + "step": 744 + }, + { + "epoch": 0.1, + "learning_rate": 8.128750681942173e-06, + "loss": 0.1391, + "step": 745 + }, + { + "epoch": 0.1, + "learning_rate": 8.139661756683034e-06, + "loss": 0.1288, + "step": 746 + }, + { + "epoch": 0.1, + "learning_rate": 8.150572831423897e-06, + "loss": 0.1349, + "step": 747 + }, + { + "epoch": 0.1, + "learning_rate": 8.161483906164758e-06, + "loss": 0.1333, + "step": 748 + }, + { + "epoch": 0.1, + "learning_rate": 8.17239498090562e-06, + "loss": 0.1379, + "step": 749 + }, + { + "epoch": 0.1, + "learning_rate": 8.183306055646481e-06, + "loss": 0.1341, + "step": 750 + }, + { + "epoch": 0.1, + "learning_rate": 8.194217130387344e-06, + "loss": 0.1479, + "step": 751 + }, + { + "epoch": 0.1, + "learning_rate": 8.205128205128205e-06, + "loss": 0.1381, + "step": 752 + }, + { + "epoch": 0.1, + "learning_rate": 8.216039279869068e-06, + "loss": 0.1272, + "step": 753 + }, + { + "epoch": 0.1, + "learning_rate": 8.22695035460993e-06, + "loss": 0.1285, + "step": 754 + }, + { + "epoch": 0.1, + "learning_rate": 8.237861429350792e-06, + "loss": 0.1313, + "step": 755 + }, + { + "epoch": 0.1, + "learning_rate": 8.248772504091653e-06, + "loss": 0.1297, + "step": 756 + }, + { + "epoch": 0.1, + "learning_rate": 8.259683578832516e-06, + "loss": 0.1321, + "step": 757 + }, + { + "epoch": 0.1, + "learning_rate": 8.270594653573377e-06, + "loss": 0.1468, + "step": 758 + }, + { + "epoch": 0.1, + "learning_rate": 8.28150572831424e-06, + "loss": 0.1375, + "step": 759 + }, + { + "epoch": 0.1, + "learning_rate": 8.2924168030551e-06, + "loss": 0.1332, + "step": 760 + }, + { + "epoch": 0.1, + "learning_rate": 8.303327877795964e-06, + "loss": 0.1371, + "step": 761 + }, + { + "epoch": 0.1, + "learning_rate": 8.314238952536825e-06, + "loss": 0.1213, + "step": 762 + }, + { + "epoch": 0.1, + "learning_rate": 8.325150027277687e-06, + "loss": 0.1281, + "step": 763 + }, + { + "epoch": 0.1, + "learning_rate": 8.33606110201855e-06, + "loss": 0.1379, + "step": 764 + }, + { + "epoch": 0.1, + "learning_rate": 8.346972176759411e-06, + "loss": 0.1453, + "step": 765 + }, + { + "epoch": 0.1, + "learning_rate": 8.357883251500274e-06, + "loss": 0.1337, + "step": 766 + }, + { + "epoch": 0.1, + "learning_rate": 8.368794326241135e-06, + "loss": 0.1387, + "step": 767 + }, + { + "epoch": 0.1, + "learning_rate": 8.379705400981998e-06, + "loss": 0.1484, + "step": 768 + }, + { + "epoch": 0.1, + "learning_rate": 8.390616475722859e-06, + "loss": 0.1235, + "step": 769 + }, + { + "epoch": 0.1, + "learning_rate": 8.401527550463722e-06, + "loss": 0.1267, + "step": 770 + }, + { + "epoch": 0.1, + "learning_rate": 8.412438625204583e-06, + "loss": 0.1352, + "step": 771 + }, + { + "epoch": 0.1, + "learning_rate": 8.423349699945446e-06, + "loss": 0.1236, + "step": 772 + }, + { + "epoch": 0.1, + "learning_rate": 8.434260774686307e-06, + "loss": 0.1309, + "step": 773 + }, + { + "epoch": 0.1, + "learning_rate": 8.44517184942717e-06, + "loss": 0.1377, + "step": 774 + }, + { + "epoch": 0.1, + "learning_rate": 8.45608292416803e-06, + "loss": 0.1254, + "step": 775 + }, + { + "epoch": 0.1, + "learning_rate": 8.466993998908893e-06, + "loss": 0.1319, + "step": 776 + }, + { + "epoch": 0.1, + "learning_rate": 8.477905073649755e-06, + "loss": 0.1223, + "step": 777 + }, + { + "epoch": 0.1, + "learning_rate": 8.488816148390617e-06, + "loss": 0.1206, + "step": 778 + }, + { + "epoch": 0.1, + "learning_rate": 8.499727223131478e-06, + "loss": 0.1388, + "step": 779 + }, + { + "epoch": 0.1, + "learning_rate": 8.510638297872341e-06, + "loss": 0.1204, + "step": 780 + }, + { + "epoch": 0.1, + "learning_rate": 8.521549372613202e-06, + "loss": 0.129, + "step": 781 + }, + { + "epoch": 0.1, + "learning_rate": 8.532460447354065e-06, + "loss": 0.1143, + "step": 782 + }, + { + "epoch": 0.1, + "learning_rate": 8.543371522094928e-06, + "loss": 0.1259, + "step": 783 + }, + { + "epoch": 0.1, + "learning_rate": 8.554282596835789e-06, + "loss": 0.146, + "step": 784 + }, + { + "epoch": 0.1, + "learning_rate": 8.565193671576652e-06, + "loss": 0.1293, + "step": 785 + }, + { + "epoch": 0.1, + "learning_rate": 8.576104746317513e-06, + "loss": 0.1452, + "step": 786 + }, + { + "epoch": 0.1, + "learning_rate": 8.587015821058376e-06, + "loss": 0.1259, + "step": 787 + }, + { + "epoch": 0.1, + "learning_rate": 8.597926895799237e-06, + "loss": 0.1335, + "step": 788 + }, + { + "epoch": 0.1, + "learning_rate": 8.6088379705401e-06, + "loss": 0.1386, + "step": 789 + }, + { + "epoch": 0.1, + "learning_rate": 8.61974904528096e-06, + "loss": 0.1335, + "step": 790 + }, + { + "epoch": 0.1, + "learning_rate": 8.630660120021823e-06, + "loss": 0.1262, + "step": 791 + }, + { + "epoch": 0.1, + "learning_rate": 8.641571194762684e-06, + "loss": 0.1275, + "step": 792 + }, + { + "epoch": 0.1, + "learning_rate": 8.652482269503547e-06, + "loss": 0.1415, + "step": 793 + }, + { + "epoch": 0.1, + "learning_rate": 8.663393344244408e-06, + "loss": 0.1157, + "step": 794 + }, + { + "epoch": 0.1, + "learning_rate": 8.674304418985271e-06, + "loss": 0.1334, + "step": 795 + }, + { + "epoch": 0.1, + "learning_rate": 8.685215493726132e-06, + "loss": 0.1258, + "step": 796 + }, + { + "epoch": 0.1, + "learning_rate": 8.696126568466995e-06, + "loss": 0.1246, + "step": 797 + }, + { + "epoch": 0.1, + "learning_rate": 8.707037643207856e-06, + "loss": 0.1346, + "step": 798 + }, + { + "epoch": 0.1, + "learning_rate": 8.717948717948719e-06, + "loss": 0.1278, + "step": 799 + }, + { + "epoch": 0.1, + "learning_rate": 8.72885979268958e-06, + "loss": 0.1255, + "step": 800 + }, + { + "epoch": 0.1, + "learning_rate": 8.739770867430443e-06, + "loss": 0.121, + "step": 801 + }, + { + "epoch": 0.11, + "learning_rate": 8.750681942171304e-06, + "loss": 0.1242, + "step": 802 + }, + { + "epoch": 0.11, + "learning_rate": 8.761593016912167e-06, + "loss": 0.1427, + "step": 803 + }, + { + "epoch": 0.11, + "learning_rate": 8.77250409165303e-06, + "loss": 0.1356, + "step": 804 + }, + { + "epoch": 0.11, + "learning_rate": 8.78341516639389e-06, + "loss": 0.1245, + "step": 805 + }, + { + "epoch": 0.11, + "learning_rate": 8.794326241134753e-06, + "loss": 0.1196, + "step": 806 + }, + { + "epoch": 0.11, + "learning_rate": 8.805237315875614e-06, + "loss": 0.1307, + "step": 807 + }, + { + "epoch": 0.11, + "learning_rate": 8.816148390616477e-06, + "loss": 0.1446, + "step": 808 + }, + { + "epoch": 0.11, + "learning_rate": 8.827059465357338e-06, + "loss": 0.1326, + "step": 809 + }, + { + "epoch": 0.11, + "learning_rate": 8.837970540098201e-06, + "loss": 0.1277, + "step": 810 + }, + { + "epoch": 0.11, + "learning_rate": 8.848881614839062e-06, + "loss": 0.1282, + "step": 811 + }, + { + "epoch": 0.11, + "learning_rate": 8.859792689579925e-06, + "loss": 0.1405, + "step": 812 + }, + { + "epoch": 0.11, + "learning_rate": 8.870703764320786e-06, + "loss": 0.1235, + "step": 813 + }, + { + "epoch": 0.11, + "learning_rate": 8.881614839061649e-06, + "loss": 0.1495, + "step": 814 + }, + { + "epoch": 0.11, + "learning_rate": 8.89252591380251e-06, + "loss": 0.1453, + "step": 815 + }, + { + "epoch": 0.11, + "learning_rate": 8.903436988543373e-06, + "loss": 0.1262, + "step": 816 + }, + { + "epoch": 0.11, + "learning_rate": 8.914348063284234e-06, + "loss": 0.1275, + "step": 817 + }, + { + "epoch": 0.11, + "learning_rate": 8.925259138025096e-06, + "loss": 0.1398, + "step": 818 + }, + { + "epoch": 0.11, + "learning_rate": 8.936170212765958e-06, + "loss": 0.1287, + "step": 819 + }, + { + "epoch": 0.11, + "learning_rate": 8.94708128750682e-06, + "loss": 0.1288, + "step": 820 + }, + { + "epoch": 0.11, + "learning_rate": 8.957992362247681e-06, + "loss": 0.1271, + "step": 821 + }, + { + "epoch": 0.11, + "learning_rate": 8.968903436988544e-06, + "loss": 0.1142, + "step": 822 + }, + { + "epoch": 0.11, + "learning_rate": 8.979814511729405e-06, + "loss": 0.1203, + "step": 823 + }, + { + "epoch": 0.11, + "learning_rate": 8.990725586470268e-06, + "loss": 0.1397, + "step": 824 + }, + { + "epoch": 0.11, + "learning_rate": 9.00163666121113e-06, + "loss": 0.1458, + "step": 825 + }, + { + "epoch": 0.11, + "learning_rate": 9.012547735951992e-06, + "loss": 0.1292, + "step": 826 + }, + { + "epoch": 0.11, + "learning_rate": 9.023458810692855e-06, + "loss": 0.1204, + "step": 827 + }, + { + "epoch": 0.11, + "learning_rate": 9.034369885433716e-06, + "loss": 0.1335, + "step": 828 + }, + { + "epoch": 0.11, + "learning_rate": 9.045280960174579e-06, + "loss": 0.1208, + "step": 829 + }, + { + "epoch": 0.11, + "learning_rate": 9.05619203491544e-06, + "loss": 0.1385, + "step": 830 + }, + { + "epoch": 0.11, + "learning_rate": 9.067103109656302e-06, + "loss": 0.1142, + "step": 831 + }, + { + "epoch": 0.11, + "learning_rate": 9.078014184397164e-06, + "loss": 0.132, + "step": 832 + }, + { + "epoch": 0.11, + "learning_rate": 9.088925259138026e-06, + "loss": 0.1283, + "step": 833 + }, + { + "epoch": 0.11, + "learning_rate": 9.099836333878887e-06, + "loss": 0.123, + "step": 834 + }, + { + "epoch": 0.11, + "learning_rate": 9.11074740861975e-06, + "loss": 0.1465, + "step": 835 + }, + { + "epoch": 0.11, + "learning_rate": 9.121658483360611e-06, + "loss": 0.1194, + "step": 836 + }, + { + "epoch": 0.11, + "learning_rate": 9.132569558101474e-06, + "loss": 0.133, + "step": 837 + }, + { + "epoch": 0.11, + "learning_rate": 9.143480632842335e-06, + "loss": 0.1296, + "step": 838 + }, + { + "epoch": 0.11, + "learning_rate": 9.154391707583198e-06, + "loss": 0.1363, + "step": 839 + }, + { + "epoch": 0.11, + "learning_rate": 9.165302782324059e-06, + "loss": 0.1211, + "step": 840 + }, + { + "epoch": 0.11, + "learning_rate": 9.176213857064922e-06, + "loss": 0.132, + "step": 841 + }, + { + "epoch": 0.11, + "learning_rate": 9.187124931805783e-06, + "loss": 0.1281, + "step": 842 + }, + { + "epoch": 0.11, + "learning_rate": 9.198036006546646e-06, + "loss": 0.1401, + "step": 843 + }, + { + "epoch": 0.11, + "learning_rate": 9.208947081287507e-06, + "loss": 0.1266, + "step": 844 + }, + { + "epoch": 0.11, + "learning_rate": 9.21985815602837e-06, + "loss": 0.1323, + "step": 845 + }, + { + "epoch": 0.11, + "learning_rate": 9.230769230769232e-06, + "loss": 0.1321, + "step": 846 + }, + { + "epoch": 0.11, + "learning_rate": 9.241680305510093e-06, + "loss": 0.131, + "step": 847 + }, + { + "epoch": 0.11, + "learning_rate": 9.252591380250956e-06, + "loss": 0.1227, + "step": 848 + }, + { + "epoch": 0.11, + "learning_rate": 9.263502454991817e-06, + "loss": 0.1208, + "step": 849 + }, + { + "epoch": 0.11, + "learning_rate": 9.27441352973268e-06, + "loss": 0.1303, + "step": 850 + }, + { + "epoch": 0.11, + "learning_rate": 9.285324604473541e-06, + "loss": 0.1381, + "step": 851 + }, + { + "epoch": 0.11, + "learning_rate": 9.296235679214404e-06, + "loss": 0.1254, + "step": 852 + }, + { + "epoch": 0.11, + "learning_rate": 9.307146753955265e-06, + "loss": 0.136, + "step": 853 + }, + { + "epoch": 0.11, + "learning_rate": 9.318057828696128e-06, + "loss": 0.1196, + "step": 854 + }, + { + "epoch": 0.11, + "learning_rate": 9.328968903436989e-06, + "loss": 0.1295, + "step": 855 + }, + { + "epoch": 0.11, + "learning_rate": 9.339879978177852e-06, + "loss": 0.1163, + "step": 856 + }, + { + "epoch": 0.11, + "learning_rate": 9.350791052918713e-06, + "loss": 0.1288, + "step": 857 + }, + { + "epoch": 0.11, + "learning_rate": 9.361702127659576e-06, + "loss": 0.1405, + "step": 858 + }, + { + "epoch": 0.11, + "learning_rate": 9.372613202400437e-06, + "loss": 0.1162, + "step": 859 + }, + { + "epoch": 0.11, + "learning_rate": 9.3835242771413e-06, + "loss": 0.13, + "step": 860 + }, + { + "epoch": 0.11, + "learning_rate": 9.39443535188216e-06, + "loss": 0.1338, + "step": 861 + }, + { + "epoch": 0.11, + "learning_rate": 9.405346426623023e-06, + "loss": 0.1329, + "step": 862 + }, + { + "epoch": 0.11, + "learning_rate": 9.416257501363884e-06, + "loss": 0.1134, + "step": 863 + }, + { + "epoch": 0.11, + "learning_rate": 9.427168576104747e-06, + "loss": 0.1336, + "step": 864 + }, + { + "epoch": 0.11, + "learning_rate": 9.438079650845608e-06, + "loss": 0.1311, + "step": 865 + }, + { + "epoch": 0.11, + "learning_rate": 9.448990725586471e-06, + "loss": 0.1409, + "step": 866 + }, + { + "epoch": 0.11, + "learning_rate": 9.459901800327334e-06, + "loss": 0.1357, + "step": 867 + }, + { + "epoch": 0.11, + "learning_rate": 9.470812875068195e-06, + "loss": 0.1285, + "step": 868 + }, + { + "epoch": 0.11, + "learning_rate": 9.481723949809058e-06, + "loss": 0.1443, + "step": 869 + }, + { + "epoch": 0.11, + "learning_rate": 9.492635024549919e-06, + "loss": 0.1255, + "step": 870 + }, + { + "epoch": 0.11, + "learning_rate": 9.503546099290782e-06, + "loss": 0.1316, + "step": 871 + }, + { + "epoch": 0.11, + "learning_rate": 9.514457174031643e-06, + "loss": 0.1225, + "step": 872 + }, + { + "epoch": 0.11, + "learning_rate": 9.525368248772505e-06, + "loss": 0.143, + "step": 873 + }, + { + "epoch": 0.11, + "learning_rate": 9.536279323513366e-06, + "loss": 0.139, + "step": 874 + }, + { + "epoch": 0.11, + "learning_rate": 9.54719039825423e-06, + "loss": 0.1378, + "step": 875 + }, + { + "epoch": 0.11, + "learning_rate": 9.55810147299509e-06, + "loss": 0.1166, + "step": 876 + }, + { + "epoch": 0.11, + "learning_rate": 9.569012547735953e-06, + "loss": 0.1309, + "step": 877 + }, + { + "epoch": 0.12, + "learning_rate": 9.579923622476814e-06, + "loss": 0.133, + "step": 878 + }, + { + "epoch": 0.12, + "learning_rate": 9.590834697217677e-06, + "loss": 0.1296, + "step": 879 + }, + { + "epoch": 0.12, + "learning_rate": 9.601745771958538e-06, + "loss": 0.1267, + "step": 880 + }, + { + "epoch": 0.12, + "learning_rate": 9.612656846699401e-06, + "loss": 0.1232, + "step": 881 + }, + { + "epoch": 0.12, + "learning_rate": 9.623567921440262e-06, + "loss": 0.1306, + "step": 882 + }, + { + "epoch": 0.12, + "learning_rate": 9.634478996181125e-06, + "loss": 0.1164, + "step": 883 + }, + { + "epoch": 0.12, + "learning_rate": 9.645390070921986e-06, + "loss": 0.1023, + "step": 884 + }, + { + "epoch": 0.12, + "learning_rate": 9.656301145662849e-06, + "loss": 0.1146, + "step": 885 + }, + { + "epoch": 0.12, + "learning_rate": 9.667212220403711e-06, + "loss": 0.1362, + "step": 886 + }, + { + "epoch": 0.12, + "learning_rate": 9.678123295144572e-06, + "loss": 0.1257, + "step": 887 + }, + { + "epoch": 0.12, + "learning_rate": 9.689034369885435e-06, + "loss": 0.1152, + "step": 888 + }, + { + "epoch": 0.12, + "learning_rate": 9.699945444626296e-06, + "loss": 0.131, + "step": 889 + }, + { + "epoch": 0.12, + "learning_rate": 9.710856519367159e-06, + "loss": 0.1176, + "step": 890 + }, + { + "epoch": 0.12, + "learning_rate": 9.72176759410802e-06, + "loss": 0.1333, + "step": 891 + }, + { + "epoch": 0.12, + "learning_rate": 9.732678668848883e-06, + "loss": 0.1187, + "step": 892 + }, + { + "epoch": 0.12, + "learning_rate": 9.743589743589744e-06, + "loss": 0.1191, + "step": 893 + }, + { + "epoch": 0.12, + "learning_rate": 9.754500818330607e-06, + "loss": 0.1322, + "step": 894 + }, + { + "epoch": 0.12, + "learning_rate": 9.765411893071468e-06, + "loss": 0.1308, + "step": 895 + }, + { + "epoch": 0.12, + "learning_rate": 9.77632296781233e-06, + "loss": 0.1136, + "step": 896 + }, + { + "epoch": 0.12, + "learning_rate": 9.787234042553192e-06, + "loss": 0.1293, + "step": 897 + }, + { + "epoch": 0.12, + "learning_rate": 9.798145117294055e-06, + "loss": 0.1375, + "step": 898 + }, + { + "epoch": 0.12, + "learning_rate": 9.809056192034916e-06, + "loss": 0.1338, + "step": 899 + }, + { + "epoch": 0.12, + "learning_rate": 9.819967266775778e-06, + "loss": 0.1316, + "step": 900 + }, + { + "epoch": 0.12, + "learning_rate": 9.83087834151664e-06, + "loss": 0.1277, + "step": 901 + }, + { + "epoch": 0.12, + "learning_rate": 9.841789416257502e-06, + "loss": 0.1237, + "step": 902 + }, + { + "epoch": 0.12, + "learning_rate": 9.852700490998363e-06, + "loss": 0.1188, + "step": 903 + }, + { + "epoch": 0.12, + "learning_rate": 9.863611565739226e-06, + "loss": 0.1324, + "step": 904 + }, + { + "epoch": 0.12, + "learning_rate": 9.874522640480087e-06, + "loss": 0.1167, + "step": 905 + }, + { + "epoch": 0.12, + "learning_rate": 9.88543371522095e-06, + "loss": 0.1252, + "step": 906 + }, + { + "epoch": 0.12, + "learning_rate": 9.896344789961813e-06, + "loss": 0.1263, + "step": 907 + }, + { + "epoch": 0.12, + "learning_rate": 9.907255864702674e-06, + "loss": 0.1272, + "step": 908 + }, + { + "epoch": 0.12, + "learning_rate": 9.918166939443537e-06, + "loss": 0.1324, + "step": 909 + }, + { + "epoch": 0.12, + "learning_rate": 9.929078014184398e-06, + "loss": 0.1237, + "step": 910 + }, + { + "epoch": 0.12, + "learning_rate": 9.93998908892526e-06, + "loss": 0.1224, + "step": 911 + }, + { + "epoch": 0.12, + "learning_rate": 9.950900163666122e-06, + "loss": 0.1134, + "step": 912 + }, + { + "epoch": 0.12, + "learning_rate": 9.961811238406984e-06, + "loss": 0.1321, + "step": 913 + }, + { + "epoch": 0.12, + "learning_rate": 9.972722313147846e-06, + "loss": 0.1211, + "step": 914 + }, + { + "epoch": 0.12, + "learning_rate": 9.983633387888708e-06, + "loss": 0.1118, + "step": 915 + }, + { + "epoch": 0.12, + "learning_rate": 9.99454446262957e-06, + "loss": 0.1248, + "step": 916 + }, + { + "epoch": 0.12, + "learning_rate": 1.0005455537370432e-05, + "loss": 0.1221, + "step": 917 + }, + { + "epoch": 0.12, + "learning_rate": 1.0016366612111295e-05, + "loss": 0.1291, + "step": 918 + }, + { + "epoch": 0.12, + "learning_rate": 1.0027277686852154e-05, + "loss": 0.1179, + "step": 919 + }, + { + "epoch": 0.12, + "learning_rate": 1.0038188761593017e-05, + "loss": 0.1282, + "step": 920 + }, + { + "epoch": 0.12, + "learning_rate": 1.004909983633388e-05, + "loss": 0.1241, + "step": 921 + }, + { + "epoch": 0.12, + "learning_rate": 1.0060010911074743e-05, + "loss": 0.1246, + "step": 922 + }, + { + "epoch": 0.12, + "learning_rate": 1.0070921985815602e-05, + "loss": 0.138, + "step": 923 + }, + { + "epoch": 0.12, + "learning_rate": 1.0081833060556465e-05, + "loss": 0.125, + "step": 924 + }, + { + "epoch": 0.12, + "learning_rate": 1.0092744135297328e-05, + "loss": 0.1311, + "step": 925 + }, + { + "epoch": 0.12, + "learning_rate": 1.010365521003819e-05, + "loss": 0.1296, + "step": 926 + }, + { + "epoch": 0.12, + "learning_rate": 1.0114566284779053e-05, + "loss": 0.1188, + "step": 927 + }, + { + "epoch": 0.12, + "learning_rate": 1.0125477359519913e-05, + "loss": 0.1319, + "step": 928 + }, + { + "epoch": 0.12, + "learning_rate": 1.0136388434260775e-05, + "loss": 0.135, + "step": 929 + }, + { + "epoch": 0.12, + "learning_rate": 1.0147299509001638e-05, + "loss": 0.1122, + "step": 930 + }, + { + "epoch": 0.12, + "learning_rate": 1.0158210583742501e-05, + "loss": 0.133, + "step": 931 + }, + { + "epoch": 0.12, + "learning_rate": 1.016912165848336e-05, + "loss": 0.1251, + "step": 932 + }, + { + "epoch": 0.12, + "learning_rate": 1.0180032733224223e-05, + "loss": 0.1171, + "step": 933 + }, + { + "epoch": 0.12, + "learning_rate": 1.0190943807965086e-05, + "loss": 0.1166, + "step": 934 + }, + { + "epoch": 0.12, + "learning_rate": 1.0201854882705949e-05, + "loss": 0.1206, + "step": 935 + }, + { + "epoch": 0.12, + "learning_rate": 1.0212765957446808e-05, + "loss": 0.1252, + "step": 936 + }, + { + "epoch": 0.12, + "learning_rate": 1.0223677032187671e-05, + "loss": 0.128, + "step": 937 + }, + { + "epoch": 0.12, + "learning_rate": 1.0234588106928534e-05, + "loss": 0.1238, + "step": 938 + }, + { + "epoch": 0.12, + "learning_rate": 1.0245499181669396e-05, + "loss": 0.1336, + "step": 939 + }, + { + "epoch": 0.12, + "learning_rate": 1.0256410256410256e-05, + "loss": 0.1288, + "step": 940 + }, + { + "epoch": 0.12, + "learning_rate": 1.0267321331151119e-05, + "loss": 0.1217, + "step": 941 + }, + { + "epoch": 0.12, + "learning_rate": 1.0278232405891981e-05, + "loss": 0.1296, + "step": 942 + }, + { + "epoch": 0.12, + "learning_rate": 1.0289143480632844e-05, + "loss": 0.1333, + "step": 943 + }, + { + "epoch": 0.12, + "learning_rate": 1.0300054555373704e-05, + "loss": 0.1236, + "step": 944 + }, + { + "epoch": 0.12, + "learning_rate": 1.0310965630114566e-05, + "loss": 0.1318, + "step": 945 + }, + { + "epoch": 0.12, + "learning_rate": 1.032187670485543e-05, + "loss": 0.1368, + "step": 946 + }, + { + "epoch": 0.12, + "learning_rate": 1.0332787779596292e-05, + "loss": 0.1289, + "step": 947 + }, + { + "epoch": 0.12, + "learning_rate": 1.0343698854337155e-05, + "loss": 0.1275, + "step": 948 + }, + { + "epoch": 0.12, + "learning_rate": 1.0354609929078014e-05, + "loss": 0.1312, + "step": 949 + }, + { + "epoch": 0.12, + "learning_rate": 1.0365521003818877e-05, + "loss": 0.1206, + "step": 950 + }, + { + "epoch": 0.12, + "learning_rate": 1.037643207855974e-05, + "loss": 0.1371, + "step": 951 + }, + { + "epoch": 0.12, + "learning_rate": 1.0387343153300602e-05, + "loss": 0.1267, + "step": 952 + }, + { + "epoch": 0.12, + "learning_rate": 1.0398254228041462e-05, + "loss": 0.1279, + "step": 953 + }, + { + "epoch": 0.12, + "learning_rate": 1.0409165302782325e-05, + "loss": 0.1353, + "step": 954 + }, + { + "epoch": 0.13, + "learning_rate": 1.0420076377523187e-05, + "loss": 0.125, + "step": 955 + }, + { + "epoch": 0.13, + "learning_rate": 1.043098745226405e-05, + "loss": 0.1245, + "step": 956 + }, + { + "epoch": 0.13, + "learning_rate": 1.044189852700491e-05, + "loss": 0.1264, + "step": 957 + }, + { + "epoch": 0.13, + "learning_rate": 1.0452809601745772e-05, + "loss": 0.1256, + "step": 958 + }, + { + "epoch": 0.13, + "learning_rate": 1.0463720676486635e-05, + "loss": 0.1168, + "step": 959 + }, + { + "epoch": 0.13, + "learning_rate": 1.0474631751227498e-05, + "loss": 0.1261, + "step": 960 + }, + { + "epoch": 0.13, + "learning_rate": 1.0485542825968357e-05, + "loss": 0.1358, + "step": 961 + }, + { + "epoch": 0.13, + "learning_rate": 1.049645390070922e-05, + "loss": 0.1299, + "step": 962 + }, + { + "epoch": 0.13, + "learning_rate": 1.0507364975450083e-05, + "loss": 0.1285, + "step": 963 + }, + { + "epoch": 0.13, + "learning_rate": 1.0518276050190946e-05, + "loss": 0.1055, + "step": 964 + }, + { + "epoch": 0.13, + "learning_rate": 1.0529187124931805e-05, + "loss": 0.1247, + "step": 965 + }, + { + "epoch": 0.13, + "learning_rate": 1.0540098199672668e-05, + "loss": 0.1212, + "step": 966 + }, + { + "epoch": 0.13, + "learning_rate": 1.055100927441353e-05, + "loss": 0.1317, + "step": 967 + }, + { + "epoch": 0.13, + "learning_rate": 1.0561920349154393e-05, + "loss": 0.1284, + "step": 968 + }, + { + "epoch": 0.13, + "learning_rate": 1.0572831423895256e-05, + "loss": 0.1165, + "step": 969 + }, + { + "epoch": 0.13, + "learning_rate": 1.0583742498636116e-05, + "loss": 0.1272, + "step": 970 + }, + { + "epoch": 0.13, + "learning_rate": 1.0594653573376978e-05, + "loss": 0.1228, + "step": 971 + }, + { + "epoch": 0.13, + "learning_rate": 1.0605564648117841e-05, + "loss": 0.1249, + "step": 972 + }, + { + "epoch": 0.13, + "learning_rate": 1.0616475722858704e-05, + "loss": 0.1135, + "step": 973 + }, + { + "epoch": 0.13, + "learning_rate": 1.0627386797599563e-05, + "loss": 0.1158, + "step": 974 + }, + { + "epoch": 0.13, + "learning_rate": 1.0638297872340426e-05, + "loss": 0.1351, + "step": 975 + }, + { + "epoch": 0.13, + "learning_rate": 1.0649208947081289e-05, + "loss": 0.1201, + "step": 976 + }, + { + "epoch": 0.13, + "learning_rate": 1.0660120021822152e-05, + "loss": 0.1259, + "step": 977 + }, + { + "epoch": 0.13, + "learning_rate": 1.0671031096563011e-05, + "loss": 0.1222, + "step": 978 + }, + { + "epoch": 0.13, + "learning_rate": 1.0681942171303874e-05, + "loss": 0.1308, + "step": 979 + }, + { + "epoch": 0.13, + "learning_rate": 1.0692853246044737e-05, + "loss": 0.133, + "step": 980 + }, + { + "epoch": 0.13, + "learning_rate": 1.07037643207856e-05, + "loss": 0.1183, + "step": 981 + }, + { + "epoch": 0.13, + "learning_rate": 1.0714675395526459e-05, + "loss": 0.1098, + "step": 982 + }, + { + "epoch": 0.13, + "learning_rate": 1.0725586470267322e-05, + "loss": 0.1309, + "step": 983 + }, + { + "epoch": 0.13, + "learning_rate": 1.0736497545008184e-05, + "loss": 0.1118, + "step": 984 + }, + { + "epoch": 0.13, + "learning_rate": 1.0747408619749047e-05, + "loss": 0.1263, + "step": 985 + }, + { + "epoch": 0.13, + "learning_rate": 1.0758319694489907e-05, + "loss": 0.1223, + "step": 986 + }, + { + "epoch": 0.13, + "learning_rate": 1.076923076923077e-05, + "loss": 0.1157, + "step": 987 + }, + { + "epoch": 0.13, + "learning_rate": 1.0780141843971632e-05, + "loss": 0.1289, + "step": 988 + }, + { + "epoch": 0.13, + "learning_rate": 1.0791052918712495e-05, + "loss": 0.1208, + "step": 989 + }, + { + "epoch": 0.13, + "learning_rate": 1.0801963993453358e-05, + "loss": 0.1218, + "step": 990 + }, + { + "epoch": 0.13, + "learning_rate": 1.0812875068194217e-05, + "loss": 0.1271, + "step": 991 + }, + { + "epoch": 0.13, + "learning_rate": 1.082378614293508e-05, + "loss": 0.1215, + "step": 992 + }, + { + "epoch": 0.13, + "learning_rate": 1.0834697217675943e-05, + "loss": 0.1224, + "step": 993 + }, + { + "epoch": 0.13, + "learning_rate": 1.0845608292416805e-05, + "loss": 0.1367, + "step": 994 + }, + { + "epoch": 0.13, + "learning_rate": 1.0856519367157665e-05, + "loss": 0.1192, + "step": 995 + }, + { + "epoch": 0.13, + "learning_rate": 1.0867430441898528e-05, + "loss": 0.1266, + "step": 996 + }, + { + "epoch": 0.13, + "learning_rate": 1.087834151663939e-05, + "loss": 0.1266, + "step": 997 + }, + { + "epoch": 0.13, + "learning_rate": 1.0889252591380253e-05, + "loss": 0.1263, + "step": 998 + }, + { + "epoch": 0.13, + "learning_rate": 1.0900163666121113e-05, + "loss": 0.1001, + "step": 999 + }, + { + "epoch": 0.13, + "learning_rate": 1.0911074740861975e-05, + "loss": 0.1175, + "step": 1000 + }, + { + "epoch": 0.13, + "learning_rate": 1.0921985815602838e-05, + "loss": 0.1263, + "step": 1001 + }, + { + "epoch": 0.13, + "learning_rate": 1.0932896890343701e-05, + "loss": 0.1117, + "step": 1002 + }, + { + "epoch": 0.13, + "learning_rate": 1.094380796508456e-05, + "loss": 0.1282, + "step": 1003 + }, + { + "epoch": 0.13, + "learning_rate": 1.0954719039825423e-05, + "loss": 0.1246, + "step": 1004 + }, + { + "epoch": 0.13, + "learning_rate": 1.0965630114566286e-05, + "loss": 0.1102, + "step": 1005 + }, + { + "epoch": 0.13, + "learning_rate": 1.0976541189307149e-05, + "loss": 0.1379, + "step": 1006 + }, + { + "epoch": 0.13, + "learning_rate": 1.0987452264048008e-05, + "loss": 0.124, + "step": 1007 + }, + { + "epoch": 0.13, + "learning_rate": 1.099836333878887e-05, + "loss": 0.1225, + "step": 1008 + }, + { + "epoch": 0.13, + "learning_rate": 1.1009274413529734e-05, + "loss": 0.1284, + "step": 1009 + }, + { + "epoch": 0.13, + "learning_rate": 1.1020185488270596e-05, + "loss": 0.1151, + "step": 1010 + }, + { + "epoch": 0.13, + "learning_rate": 1.1031096563011459e-05, + "loss": 0.1291, + "step": 1011 + }, + { + "epoch": 0.13, + "learning_rate": 1.1042007637752319e-05, + "loss": 0.1244, + "step": 1012 + }, + { + "epoch": 0.13, + "learning_rate": 1.1052918712493181e-05, + "loss": 0.1249, + "step": 1013 + }, + { + "epoch": 0.13, + "learning_rate": 1.1063829787234044e-05, + "loss": 0.1238, + "step": 1014 + }, + { + "epoch": 0.13, + "learning_rate": 1.1074740861974907e-05, + "loss": 0.1266, + "step": 1015 + }, + { + "epoch": 0.13, + "learning_rate": 1.1085651936715766e-05, + "loss": 0.1188, + "step": 1016 + }, + { + "epoch": 0.13, + "learning_rate": 1.1096563011456629e-05, + "loss": 0.127, + "step": 1017 + }, + { + "epoch": 0.13, + "learning_rate": 1.1107474086197492e-05, + "loss": 0.1276, + "step": 1018 + }, + { + "epoch": 0.13, + "learning_rate": 1.1118385160938355e-05, + "loss": 0.1224, + "step": 1019 + }, + { + "epoch": 0.13, + "learning_rate": 1.1129296235679214e-05, + "loss": 0.113, + "step": 1020 + }, + { + "epoch": 0.13, + "learning_rate": 1.1140207310420077e-05, + "loss": 0.1265, + "step": 1021 + }, + { + "epoch": 0.13, + "learning_rate": 1.115111838516094e-05, + "loss": 0.1182, + "step": 1022 + }, + { + "epoch": 0.13, + "learning_rate": 1.1162029459901802e-05, + "loss": 0.1121, + "step": 1023 + }, + { + "epoch": 0.13, + "learning_rate": 1.1172940534642662e-05, + "loss": 0.1302, + "step": 1024 + }, + { + "epoch": 0.13, + "learning_rate": 1.1183851609383525e-05, + "loss": 0.1319, + "step": 1025 + }, + { + "epoch": 0.13, + "learning_rate": 1.1194762684124387e-05, + "loss": 0.1146, + "step": 1026 + }, + { + "epoch": 0.13, + "learning_rate": 1.120567375886525e-05, + "loss": 0.1228, + "step": 1027 + }, + { + "epoch": 0.13, + "learning_rate": 1.121658483360611e-05, + "loss": 0.1152, + "step": 1028 + }, + { + "epoch": 0.13, + "learning_rate": 1.1227495908346972e-05, + "loss": 0.1226, + "step": 1029 + }, + { + "epoch": 0.13, + "learning_rate": 1.1238406983087835e-05, + "loss": 0.1266, + "step": 1030 + }, + { + "epoch": 0.14, + "learning_rate": 1.1249318057828698e-05, + "loss": 0.128, + "step": 1031 + }, + { + "epoch": 0.14, + "learning_rate": 1.126022913256956e-05, + "loss": 0.1208, + "step": 1032 + }, + { + "epoch": 0.14, + "learning_rate": 1.127114020731042e-05, + "loss": 0.1177, + "step": 1033 + }, + { + "epoch": 0.14, + "learning_rate": 1.1282051282051283e-05, + "loss": 0.1282, + "step": 1034 + }, + { + "epoch": 0.14, + "learning_rate": 1.1292962356792146e-05, + "loss": 0.1145, + "step": 1035 + }, + { + "epoch": 0.14, + "learning_rate": 1.1303873431533008e-05, + "loss": 0.13, + "step": 1036 + }, + { + "epoch": 0.14, + "learning_rate": 1.1314784506273868e-05, + "loss": 0.1184, + "step": 1037 + }, + { + "epoch": 0.14, + "learning_rate": 1.132569558101473e-05, + "loss": 0.1171, + "step": 1038 + }, + { + "epoch": 0.14, + "learning_rate": 1.1336606655755593e-05, + "loss": 0.1153, + "step": 1039 + }, + { + "epoch": 0.14, + "learning_rate": 1.1347517730496456e-05, + "loss": 0.1017, + "step": 1040 + }, + { + "epoch": 0.14, + "learning_rate": 1.1358428805237315e-05, + "loss": 0.123, + "step": 1041 + }, + { + "epoch": 0.14, + "learning_rate": 1.1369339879978178e-05, + "loss": 0.1201, + "step": 1042 + }, + { + "epoch": 0.14, + "learning_rate": 1.1380250954719041e-05, + "loss": 0.1205, + "step": 1043 + }, + { + "epoch": 0.14, + "learning_rate": 1.1391162029459904e-05, + "loss": 0.1139, + "step": 1044 + }, + { + "epoch": 0.14, + "learning_rate": 1.1402073104200763e-05, + "loss": 0.1028, + "step": 1045 + }, + { + "epoch": 0.14, + "learning_rate": 1.1412984178941626e-05, + "loss": 0.1281, + "step": 1046 + }, + { + "epoch": 0.14, + "learning_rate": 1.1423895253682489e-05, + "loss": 0.1187, + "step": 1047 + }, + { + "epoch": 0.14, + "learning_rate": 1.1434806328423352e-05, + "loss": 0.1188, + "step": 1048 + }, + { + "epoch": 0.14, + "learning_rate": 1.1445717403164211e-05, + "loss": 0.128, + "step": 1049 + }, + { + "epoch": 0.14, + "learning_rate": 1.1456628477905074e-05, + "loss": 0.1281, + "step": 1050 + }, + { + "epoch": 0.14, + "learning_rate": 1.1467539552645937e-05, + "loss": 0.115, + "step": 1051 + }, + { + "epoch": 0.14, + "learning_rate": 1.14784506273868e-05, + "loss": 0.1233, + "step": 1052 + }, + { + "epoch": 0.14, + "learning_rate": 1.1489361702127662e-05, + "loss": 0.1304, + "step": 1053 + }, + { + "epoch": 0.14, + "learning_rate": 1.1500272776868521e-05, + "loss": 0.118, + "step": 1054 + }, + { + "epoch": 0.14, + "learning_rate": 1.1511183851609384e-05, + "loss": 0.1241, + "step": 1055 + }, + { + "epoch": 0.14, + "learning_rate": 1.1522094926350247e-05, + "loss": 0.1232, + "step": 1056 + }, + { + "epoch": 0.14, + "learning_rate": 1.153300600109111e-05, + "loss": 0.1229, + "step": 1057 + }, + { + "epoch": 0.14, + "learning_rate": 1.154391707583197e-05, + "loss": 0.1184, + "step": 1058 + }, + { + "epoch": 0.14, + "learning_rate": 1.1554828150572832e-05, + "loss": 0.139, + "step": 1059 + }, + { + "epoch": 0.14, + "learning_rate": 1.1565739225313695e-05, + "loss": 0.0985, + "step": 1060 + }, + { + "epoch": 0.14, + "learning_rate": 1.1576650300054558e-05, + "loss": 0.1281, + "step": 1061 + }, + { + "epoch": 0.14, + "learning_rate": 1.1587561374795417e-05, + "loss": 0.1195, + "step": 1062 + }, + { + "epoch": 0.14, + "learning_rate": 1.159847244953628e-05, + "loss": 0.1275, + "step": 1063 + }, + { + "epoch": 0.14, + "learning_rate": 1.1609383524277143e-05, + "loss": 0.1128, + "step": 1064 + }, + { + "epoch": 0.14, + "learning_rate": 1.1620294599018005e-05, + "loss": 0.1251, + "step": 1065 + }, + { + "epoch": 0.14, + "learning_rate": 1.1631205673758865e-05, + "loss": 0.1224, + "step": 1066 + }, + { + "epoch": 0.14, + "learning_rate": 1.1642116748499727e-05, + "loss": 0.1341, + "step": 1067 + }, + { + "epoch": 0.14, + "learning_rate": 1.165302782324059e-05, + "loss": 0.11, + "step": 1068 + }, + { + "epoch": 0.14, + "learning_rate": 1.1663938897981453e-05, + "loss": 0.1376, + "step": 1069 + }, + { + "epoch": 0.14, + "learning_rate": 1.1674849972722312e-05, + "loss": 0.1158, + "step": 1070 + }, + { + "epoch": 0.14, + "learning_rate": 1.1685761047463175e-05, + "loss": 0.1243, + "step": 1071 + }, + { + "epoch": 0.14, + "learning_rate": 1.1696672122204038e-05, + "loss": 0.1089, + "step": 1072 + }, + { + "epoch": 0.14, + "learning_rate": 1.17075831969449e-05, + "loss": 0.1261, + "step": 1073 + }, + { + "epoch": 0.14, + "learning_rate": 1.1718494271685764e-05, + "loss": 0.1134, + "step": 1074 + }, + { + "epoch": 0.14, + "learning_rate": 1.1729405346426623e-05, + "loss": 0.1284, + "step": 1075 + }, + { + "epoch": 0.14, + "learning_rate": 1.1740316421167486e-05, + "loss": 0.1212, + "step": 1076 + }, + { + "epoch": 0.14, + "learning_rate": 1.1751227495908349e-05, + "loss": 0.1062, + "step": 1077 + }, + { + "epoch": 0.14, + "learning_rate": 1.1762138570649211e-05, + "loss": 0.1206, + "step": 1078 + }, + { + "epoch": 0.14, + "learning_rate": 1.177304964539007e-05, + "loss": 0.1222, + "step": 1079 + }, + { + "epoch": 0.14, + "learning_rate": 1.1783960720130933e-05, + "loss": 0.1113, + "step": 1080 + }, + { + "epoch": 0.14, + "learning_rate": 1.1794871794871796e-05, + "loss": 0.1225, + "step": 1081 + }, + { + "epoch": 0.14, + "learning_rate": 1.1805782869612659e-05, + "loss": 0.1103, + "step": 1082 + }, + { + "epoch": 0.14, + "learning_rate": 1.1816693944353518e-05, + "loss": 0.1289, + "step": 1083 + }, + { + "epoch": 0.14, + "learning_rate": 1.1827605019094381e-05, + "loss": 0.1292, + "step": 1084 + }, + { + "epoch": 0.14, + "learning_rate": 1.1838516093835244e-05, + "loss": 0.1266, + "step": 1085 + }, + { + "epoch": 0.14, + "learning_rate": 1.1849427168576107e-05, + "loss": 0.1273, + "step": 1086 + }, + { + "epoch": 0.14, + "learning_rate": 1.1860338243316966e-05, + "loss": 0.1168, + "step": 1087 + }, + { + "epoch": 0.14, + "learning_rate": 1.1871249318057829e-05, + "loss": 0.1186, + "step": 1088 + }, + { + "epoch": 0.14, + "learning_rate": 1.1882160392798692e-05, + "loss": 0.1244, + "step": 1089 + }, + { + "epoch": 0.14, + "learning_rate": 1.1893071467539555e-05, + "loss": 0.1211, + "step": 1090 + }, + { + "epoch": 0.14, + "learning_rate": 1.1903982542280414e-05, + "loss": 0.1243, + "step": 1091 + }, + { + "epoch": 0.14, + "learning_rate": 1.1914893617021277e-05, + "loss": 0.1156, + "step": 1092 + }, + { + "epoch": 0.14, + "learning_rate": 1.192580469176214e-05, + "loss": 0.1199, + "step": 1093 + }, + { + "epoch": 0.14, + "learning_rate": 1.1936715766503002e-05, + "loss": 0.1344, + "step": 1094 + }, + { + "epoch": 0.14, + "learning_rate": 1.1947626841243865e-05, + "loss": 0.1301, + "step": 1095 + }, + { + "epoch": 0.14, + "learning_rate": 1.1958537915984724e-05, + "loss": 0.1306, + "step": 1096 + }, + { + "epoch": 0.14, + "learning_rate": 1.1969448990725587e-05, + "loss": 0.1101, + "step": 1097 + }, + { + "epoch": 0.14, + "learning_rate": 1.198036006546645e-05, + "loss": 0.1124, + "step": 1098 + }, + { + "epoch": 0.14, + "learning_rate": 1.1991271140207313e-05, + "loss": 0.1173, + "step": 1099 + }, + { + "epoch": 0.14, + "learning_rate": 1.2002182214948172e-05, + "loss": 0.1164, + "step": 1100 + }, + { + "epoch": 0.14, + "learning_rate": 1.2013093289689035e-05, + "loss": 0.1221, + "step": 1101 + }, + { + "epoch": 0.14, + "learning_rate": 1.2024004364429898e-05, + "loss": 0.1091, + "step": 1102 + }, + { + "epoch": 0.14, + "learning_rate": 1.203491543917076e-05, + "loss": 0.122, + "step": 1103 + }, + { + "epoch": 0.14, + "learning_rate": 1.204582651391162e-05, + "loss": 0.1253, + "step": 1104 + }, + { + "epoch": 0.14, + "learning_rate": 1.2056737588652483e-05, + "loss": 0.1135, + "step": 1105 + }, + { + "epoch": 0.14, + "learning_rate": 1.2067648663393345e-05, + "loss": 0.1245, + "step": 1106 + }, + { + "epoch": 0.15, + "learning_rate": 1.2078559738134208e-05, + "loss": 0.1266, + "step": 1107 + }, + { + "epoch": 0.15, + "learning_rate": 1.2089470812875068e-05, + "loss": 0.1158, + "step": 1108 + }, + { + "epoch": 0.15, + "learning_rate": 1.210038188761593e-05, + "loss": 0.1283, + "step": 1109 + }, + { + "epoch": 0.15, + "learning_rate": 1.2111292962356793e-05, + "loss": 0.1194, + "step": 1110 + }, + { + "epoch": 0.15, + "learning_rate": 1.2122204037097656e-05, + "loss": 0.1149, + "step": 1111 + }, + { + "epoch": 0.15, + "learning_rate": 1.2133115111838515e-05, + "loss": 0.1212, + "step": 1112 + }, + { + "epoch": 0.15, + "learning_rate": 1.2144026186579378e-05, + "loss": 0.1255, + "step": 1113 + }, + { + "epoch": 0.15, + "learning_rate": 1.2154937261320241e-05, + "loss": 0.1346, + "step": 1114 + }, + { + "epoch": 0.15, + "learning_rate": 1.2165848336061104e-05, + "loss": 0.1113, + "step": 1115 + }, + { + "epoch": 0.15, + "learning_rate": 1.2176759410801967e-05, + "loss": 0.1245, + "step": 1116 + }, + { + "epoch": 0.15, + "learning_rate": 1.2187670485542826e-05, + "loss": 0.1173, + "step": 1117 + }, + { + "epoch": 0.15, + "learning_rate": 1.2198581560283689e-05, + "loss": 0.1285, + "step": 1118 + }, + { + "epoch": 0.15, + "learning_rate": 1.2209492635024551e-05, + "loss": 0.1288, + "step": 1119 + }, + { + "epoch": 0.15, + "learning_rate": 1.2220403709765414e-05, + "loss": 0.1268, + "step": 1120 + }, + { + "epoch": 0.15, + "learning_rate": 1.2231314784506274e-05, + "loss": 0.1225, + "step": 1121 + }, + { + "epoch": 0.15, + "learning_rate": 1.2242225859247136e-05, + "loss": 0.1134, + "step": 1122 + }, + { + "epoch": 0.15, + "learning_rate": 1.2253136933988e-05, + "loss": 0.1217, + "step": 1123 + }, + { + "epoch": 0.15, + "learning_rate": 1.2264048008728862e-05, + "loss": 0.1232, + "step": 1124 + }, + { + "epoch": 0.15, + "learning_rate": 1.2274959083469721e-05, + "loss": 0.1119, + "step": 1125 + }, + { + "epoch": 0.15, + "learning_rate": 1.2285870158210584e-05, + "loss": 0.1141, + "step": 1126 + }, + { + "epoch": 0.15, + "learning_rate": 1.2296781232951447e-05, + "loss": 0.1201, + "step": 1127 + }, + { + "epoch": 0.15, + "learning_rate": 1.230769230769231e-05, + "loss": 0.1185, + "step": 1128 + }, + { + "epoch": 0.15, + "learning_rate": 1.2318603382433169e-05, + "loss": 0.1126, + "step": 1129 + }, + { + "epoch": 0.15, + "learning_rate": 1.2329514457174032e-05, + "loss": 0.1308, + "step": 1130 + }, + { + "epoch": 0.15, + "learning_rate": 1.2340425531914895e-05, + "loss": 0.1291, + "step": 1131 + }, + { + "epoch": 0.15, + "learning_rate": 1.2351336606655757e-05, + "loss": 0.1168, + "step": 1132 + }, + { + "epoch": 0.15, + "learning_rate": 1.236224768139662e-05, + "loss": 0.1159, + "step": 1133 + }, + { + "epoch": 0.15, + "learning_rate": 1.237315875613748e-05, + "loss": 0.1282, + "step": 1134 + }, + { + "epoch": 0.15, + "learning_rate": 1.2384069830878342e-05, + "loss": 0.1162, + "step": 1135 + }, + { + "epoch": 0.15, + "learning_rate": 1.2394980905619205e-05, + "loss": 0.1212, + "step": 1136 + }, + { + "epoch": 0.15, + "learning_rate": 1.2405891980360068e-05, + "loss": 0.1278, + "step": 1137 + }, + { + "epoch": 0.15, + "learning_rate": 1.2416803055100927e-05, + "loss": 0.1108, + "step": 1138 + }, + { + "epoch": 0.15, + "learning_rate": 1.242771412984179e-05, + "loss": 0.1148, + "step": 1139 + }, + { + "epoch": 0.15, + "learning_rate": 1.2438625204582653e-05, + "loss": 0.1221, + "step": 1140 + }, + { + "epoch": 0.15, + "learning_rate": 1.2449536279323516e-05, + "loss": 0.1128, + "step": 1141 + }, + { + "epoch": 0.15, + "learning_rate": 1.2460447354064375e-05, + "loss": 0.1048, + "step": 1142 + }, + { + "epoch": 0.15, + "learning_rate": 1.2471358428805238e-05, + "loss": 0.1214, + "step": 1143 + }, + { + "epoch": 0.15, + "learning_rate": 1.24822695035461e-05, + "loss": 0.1103, + "step": 1144 + }, + { + "epoch": 0.15, + "learning_rate": 1.2493180578286963e-05, + "loss": 0.1241, + "step": 1145 + }, + { + "epoch": 0.15, + "learning_rate": 1.2504091653027823e-05, + "loss": 0.1203, + "step": 1146 + }, + { + "epoch": 0.15, + "learning_rate": 1.2515002727768686e-05, + "loss": 0.1316, + "step": 1147 + }, + { + "epoch": 0.15, + "learning_rate": 1.2525913802509548e-05, + "loss": 0.1278, + "step": 1148 + }, + { + "epoch": 0.15, + "learning_rate": 1.2536824877250411e-05, + "loss": 0.1321, + "step": 1149 + }, + { + "epoch": 0.15, + "learning_rate": 1.254773595199127e-05, + "loss": 0.1212, + "step": 1150 + }, + { + "epoch": 0.15, + "learning_rate": 1.2558647026732133e-05, + "loss": 0.1313, + "step": 1151 + }, + { + "epoch": 0.15, + "learning_rate": 1.2569558101472996e-05, + "loss": 0.1163, + "step": 1152 + }, + { + "epoch": 0.15, + "learning_rate": 1.2580469176213859e-05, + "loss": 0.1167, + "step": 1153 + }, + { + "epoch": 0.15, + "learning_rate": 1.2591380250954722e-05, + "loss": 0.1225, + "step": 1154 + }, + { + "epoch": 0.15, + "learning_rate": 1.2602291325695581e-05, + "loss": 0.1296, + "step": 1155 + }, + { + "epoch": 0.15, + "learning_rate": 1.2613202400436444e-05, + "loss": 0.1198, + "step": 1156 + }, + { + "epoch": 0.15, + "learning_rate": 1.2624113475177307e-05, + "loss": 0.1198, + "step": 1157 + }, + { + "epoch": 0.15, + "learning_rate": 1.263502454991817e-05, + "loss": 0.108, + "step": 1158 + }, + { + "epoch": 0.15, + "learning_rate": 1.2645935624659029e-05, + "loss": 0.1231, + "step": 1159 + }, + { + "epoch": 0.15, + "learning_rate": 1.2656846699399892e-05, + "loss": 0.1174, + "step": 1160 + }, + { + "epoch": 0.15, + "learning_rate": 1.2667757774140754e-05, + "loss": 0.1186, + "step": 1161 + }, + { + "epoch": 0.15, + "learning_rate": 1.2678668848881617e-05, + "loss": 0.1231, + "step": 1162 + }, + { + "epoch": 0.15, + "learning_rate": 1.2689579923622477e-05, + "loss": 0.1115, + "step": 1163 + }, + { + "epoch": 0.15, + "learning_rate": 1.270049099836334e-05, + "loss": 0.123, + "step": 1164 + }, + { + "epoch": 0.15, + "learning_rate": 1.2711402073104202e-05, + "loss": 0.1076, + "step": 1165 + }, + { + "epoch": 0.15, + "learning_rate": 1.2722313147845065e-05, + "loss": 0.1128, + "step": 1166 + }, + { + "epoch": 0.15, + "learning_rate": 1.2733224222585924e-05, + "loss": 0.1167, + "step": 1167 + }, + { + "epoch": 0.15, + "learning_rate": 1.2744135297326787e-05, + "loss": 0.1213, + "step": 1168 + }, + { + "epoch": 0.15, + "learning_rate": 1.275504637206765e-05, + "loss": 0.1303, + "step": 1169 + }, + { + "epoch": 0.15, + "learning_rate": 1.2765957446808513e-05, + "loss": 0.1277, + "step": 1170 + }, + { + "epoch": 0.15, + "learning_rate": 1.2776868521549372e-05, + "loss": 0.1253, + "step": 1171 + }, + { + "epoch": 0.15, + "learning_rate": 1.2787779596290235e-05, + "loss": 0.1074, + "step": 1172 + }, + { + "epoch": 0.15, + "learning_rate": 1.2798690671031098e-05, + "loss": 0.103, + "step": 1173 + }, + { + "epoch": 0.15, + "learning_rate": 1.280960174577196e-05, + "loss": 0.1112, + "step": 1174 + }, + { + "epoch": 0.15, + "learning_rate": 1.2820512820512823e-05, + "loss": 0.1175, + "step": 1175 + }, + { + "epoch": 0.15, + "learning_rate": 1.2831423895253683e-05, + "loss": 0.1229, + "step": 1176 + }, + { + "epoch": 0.15, + "learning_rate": 1.2842334969994545e-05, + "loss": 0.1258, + "step": 1177 + }, + { + "epoch": 0.15, + "learning_rate": 1.2853246044735408e-05, + "loss": 0.134, + "step": 1178 + }, + { + "epoch": 0.15, + "learning_rate": 1.2864157119476271e-05, + "loss": 0.1101, + "step": 1179 + }, + { + "epoch": 0.15, + "learning_rate": 1.287506819421713e-05, + "loss": 0.1204, + "step": 1180 + }, + { + "epoch": 0.15, + "learning_rate": 1.2885979268957993e-05, + "loss": 0.117, + "step": 1181 + }, + { + "epoch": 0.15, + "learning_rate": 1.2896890343698856e-05, + "loss": 0.1223, + "step": 1182 + }, + { + "epoch": 0.15, + "learning_rate": 1.2907801418439719e-05, + "loss": 0.1124, + "step": 1183 + }, + { + "epoch": 0.16, + "learning_rate": 1.2918712493180578e-05, + "loss": 0.1314, + "step": 1184 + }, + { + "epoch": 0.16, + "learning_rate": 1.2929623567921441e-05, + "loss": 0.1234, + "step": 1185 + }, + { + "epoch": 0.16, + "learning_rate": 1.2940534642662304e-05, + "loss": 0.1184, + "step": 1186 + }, + { + "epoch": 0.16, + "learning_rate": 1.2951445717403166e-05, + "loss": 0.1156, + "step": 1187 + }, + { + "epoch": 0.16, + "learning_rate": 1.2962356792144026e-05, + "loss": 0.1256, + "step": 1188 + }, + { + "epoch": 0.16, + "learning_rate": 1.2973267866884889e-05, + "loss": 0.1179, + "step": 1189 + }, + { + "epoch": 0.16, + "learning_rate": 1.2984178941625751e-05, + "loss": 0.131, + "step": 1190 + }, + { + "epoch": 0.16, + "learning_rate": 1.2995090016366614e-05, + "loss": 0.1271, + "step": 1191 + }, + { + "epoch": 0.16, + "learning_rate": 1.3006001091107474e-05, + "loss": 0.1171, + "step": 1192 + }, + { + "epoch": 0.16, + "learning_rate": 1.3016912165848336e-05, + "loss": 0.1184, + "step": 1193 + }, + { + "epoch": 0.16, + "learning_rate": 1.3027823240589199e-05, + "loss": 0.1085, + "step": 1194 + }, + { + "epoch": 0.16, + "learning_rate": 1.3038734315330062e-05, + "loss": 0.1151, + "step": 1195 + }, + { + "epoch": 0.16, + "learning_rate": 1.3049645390070925e-05, + "loss": 0.108, + "step": 1196 + }, + { + "epoch": 0.16, + "learning_rate": 1.3060556464811784e-05, + "loss": 0.1182, + "step": 1197 + }, + { + "epoch": 0.16, + "learning_rate": 1.3071467539552647e-05, + "loss": 0.1083, + "step": 1198 + }, + { + "epoch": 0.16, + "learning_rate": 1.308237861429351e-05, + "loss": 0.12, + "step": 1199 + }, + { + "epoch": 0.16, + "learning_rate": 1.3093289689034372e-05, + "loss": 0.1133, + "step": 1200 + }, + { + "epoch": 0.16, + "learning_rate": 1.3104200763775232e-05, + "loss": 0.1197, + "step": 1201 + }, + { + "epoch": 0.16, + "learning_rate": 1.3115111838516095e-05, + "loss": 0.1086, + "step": 1202 + }, + { + "epoch": 0.16, + "learning_rate": 1.3126022913256957e-05, + "loss": 0.1175, + "step": 1203 + }, + { + "epoch": 0.16, + "learning_rate": 1.313693398799782e-05, + "loss": 0.1041, + "step": 1204 + }, + { + "epoch": 0.16, + "learning_rate": 1.314784506273868e-05, + "loss": 0.1199, + "step": 1205 + }, + { + "epoch": 0.16, + "learning_rate": 1.3158756137479542e-05, + "loss": 0.1256, + "step": 1206 + }, + { + "epoch": 0.16, + "learning_rate": 1.3169667212220405e-05, + "loss": 0.1161, + "step": 1207 + }, + { + "epoch": 0.16, + "learning_rate": 1.3180578286961268e-05, + "loss": 0.1146, + "step": 1208 + }, + { + "epoch": 0.16, + "learning_rate": 1.3191489361702127e-05, + "loss": 0.1251, + "step": 1209 + }, + { + "epoch": 0.16, + "learning_rate": 1.320240043644299e-05, + "loss": 0.1271, + "step": 1210 + }, + { + "epoch": 0.16, + "learning_rate": 1.3213311511183853e-05, + "loss": 0.1128, + "step": 1211 + }, + { + "epoch": 0.16, + "learning_rate": 1.3224222585924716e-05, + "loss": 0.1142, + "step": 1212 + }, + { + "epoch": 0.16, + "learning_rate": 1.3235133660665575e-05, + "loss": 0.1258, + "step": 1213 + }, + { + "epoch": 0.16, + "learning_rate": 1.3246044735406438e-05, + "loss": 0.1299, + "step": 1214 + }, + { + "epoch": 0.16, + "learning_rate": 1.32569558101473e-05, + "loss": 0.1299, + "step": 1215 + }, + { + "epoch": 0.16, + "learning_rate": 1.3267866884888163e-05, + "loss": 0.1245, + "step": 1216 + }, + { + "epoch": 0.16, + "learning_rate": 1.3278777959629026e-05, + "loss": 0.1108, + "step": 1217 + }, + { + "epoch": 0.16, + "learning_rate": 1.3289689034369886e-05, + "loss": 0.1226, + "step": 1218 + }, + { + "epoch": 0.16, + "learning_rate": 1.3300600109110748e-05, + "loss": 0.1251, + "step": 1219 + }, + { + "epoch": 0.16, + "learning_rate": 1.3311511183851611e-05, + "loss": 0.1328, + "step": 1220 + }, + { + "epoch": 0.16, + "learning_rate": 1.3322422258592474e-05, + "loss": 0.1288, + "step": 1221 + }, + { + "epoch": 0.16, + "learning_rate": 1.3333333333333333e-05, + "loss": 0.1105, + "step": 1222 + }, + { + "epoch": 0.16, + "learning_rate": 1.3344244408074196e-05, + "loss": 0.1328, + "step": 1223 + }, + { + "epoch": 0.16, + "learning_rate": 1.3355155482815059e-05, + "loss": 0.1194, + "step": 1224 + }, + { + "epoch": 0.16, + "learning_rate": 1.3366066557555922e-05, + "loss": 0.114, + "step": 1225 + }, + { + "epoch": 0.16, + "learning_rate": 1.3376977632296781e-05, + "loss": 0.1138, + "step": 1226 + }, + { + "epoch": 0.16, + "learning_rate": 1.3387888707037644e-05, + "loss": 0.1225, + "step": 1227 + }, + { + "epoch": 0.16, + "learning_rate": 1.3398799781778507e-05, + "loss": 0.1201, + "step": 1228 + }, + { + "epoch": 0.16, + "learning_rate": 1.340971085651937e-05, + "loss": 0.1097, + "step": 1229 + }, + { + "epoch": 0.16, + "learning_rate": 1.3420621931260229e-05, + "loss": 0.1134, + "step": 1230 + }, + { + "epoch": 0.16, + "learning_rate": 1.3431533006001092e-05, + "loss": 0.1156, + "step": 1231 + }, + { + "epoch": 0.16, + "learning_rate": 1.3442444080741954e-05, + "loss": 0.1142, + "step": 1232 + }, + { + "epoch": 0.16, + "learning_rate": 1.3453355155482817e-05, + "loss": 0.1129, + "step": 1233 + }, + { + "epoch": 0.16, + "learning_rate": 1.3464266230223677e-05, + "loss": 0.1111, + "step": 1234 + }, + { + "epoch": 0.16, + "learning_rate": 1.347517730496454e-05, + "loss": 0.1179, + "step": 1235 + }, + { + "epoch": 0.16, + "learning_rate": 1.3486088379705402e-05, + "loss": 0.1096, + "step": 1236 + }, + { + "epoch": 0.16, + "learning_rate": 1.3496999454446265e-05, + "loss": 0.1337, + "step": 1237 + }, + { + "epoch": 0.16, + "learning_rate": 1.3507910529187128e-05, + "loss": 0.1291, + "step": 1238 + }, + { + "epoch": 0.16, + "learning_rate": 1.3518821603927987e-05, + "loss": 0.1159, + "step": 1239 + }, + { + "epoch": 0.16, + "learning_rate": 1.352973267866885e-05, + "loss": 0.1092, + "step": 1240 + }, + { + "epoch": 0.16, + "learning_rate": 1.3540643753409713e-05, + "loss": 0.1329, + "step": 1241 + }, + { + "epoch": 0.16, + "learning_rate": 1.3551554828150575e-05, + "loss": 0.1113, + "step": 1242 + }, + { + "epoch": 0.16, + "learning_rate": 1.3562465902891435e-05, + "loss": 0.1189, + "step": 1243 + }, + { + "epoch": 0.16, + "learning_rate": 1.3573376977632298e-05, + "loss": 0.1325, + "step": 1244 + }, + { + "epoch": 0.16, + "learning_rate": 1.358428805237316e-05, + "loss": 0.1348, + "step": 1245 + }, + { + "epoch": 0.16, + "learning_rate": 1.3595199127114023e-05, + "loss": 0.1342, + "step": 1246 + }, + { + "epoch": 0.16, + "learning_rate": 1.3606110201854883e-05, + "loss": 0.1158, + "step": 1247 + }, + { + "epoch": 0.16, + "learning_rate": 1.3617021276595745e-05, + "loss": 0.1252, + "step": 1248 + }, + { + "epoch": 0.16, + "learning_rate": 1.3627932351336608e-05, + "loss": 0.104, + "step": 1249 + }, + { + "epoch": 0.16, + "learning_rate": 1.3638843426077471e-05, + "loss": 0.1291, + "step": 1250 + }, + { + "epoch": 0.16, + "learning_rate": 1.364975450081833e-05, + "loss": 0.1304, + "step": 1251 + }, + { + "epoch": 0.16, + "learning_rate": 1.3660665575559193e-05, + "loss": 0.1221, + "step": 1252 + }, + { + "epoch": 0.16, + "learning_rate": 1.3671576650300056e-05, + "loss": 0.1182, + "step": 1253 + }, + { + "epoch": 0.16, + "learning_rate": 1.3682487725040919e-05, + "loss": 0.126, + "step": 1254 + }, + { + "epoch": 0.16, + "learning_rate": 1.3693398799781778e-05, + "loss": 0.1167, + "step": 1255 + }, + { + "epoch": 0.16, + "learning_rate": 1.370430987452264e-05, + "loss": 0.1276, + "step": 1256 + }, + { + "epoch": 0.16, + "learning_rate": 1.3715220949263504e-05, + "loss": 0.1151, + "step": 1257 + }, + { + "epoch": 0.16, + "learning_rate": 1.3726132024004366e-05, + "loss": 0.1217, + "step": 1258 + }, + { + "epoch": 0.16, + "learning_rate": 1.3737043098745229e-05, + "loss": 0.1222, + "step": 1259 + }, + { + "epoch": 0.17, + "learning_rate": 1.3747954173486089e-05, + "loss": 0.1157, + "step": 1260 + }, + { + "epoch": 0.17, + "learning_rate": 1.3758865248226951e-05, + "loss": 0.1295, + "step": 1261 + }, + { + "epoch": 0.17, + "learning_rate": 1.3769776322967814e-05, + "loss": 0.1244, + "step": 1262 + }, + { + "epoch": 0.17, + "learning_rate": 1.3780687397708677e-05, + "loss": 0.1295, + "step": 1263 + }, + { + "epoch": 0.17, + "learning_rate": 1.3791598472449536e-05, + "loss": 0.1203, + "step": 1264 + }, + { + "epoch": 0.17, + "learning_rate": 1.3802509547190399e-05, + "loss": 0.1114, + "step": 1265 + }, + { + "epoch": 0.17, + "learning_rate": 1.3813420621931262e-05, + "loss": 0.1316, + "step": 1266 + }, + { + "epoch": 0.17, + "learning_rate": 1.3824331696672125e-05, + "loss": 0.1035, + "step": 1267 + }, + { + "epoch": 0.17, + "learning_rate": 1.3835242771412984e-05, + "loss": 0.104, + "step": 1268 + }, + { + "epoch": 0.17, + "learning_rate": 1.3846153846153847e-05, + "loss": 0.1234, + "step": 1269 + }, + { + "epoch": 0.17, + "learning_rate": 1.385706492089471e-05, + "loss": 0.1139, + "step": 1270 + }, + { + "epoch": 0.17, + "learning_rate": 1.3867975995635572e-05, + "loss": 0.1126, + "step": 1271 + }, + { + "epoch": 0.17, + "learning_rate": 1.3878887070376432e-05, + "loss": 0.1175, + "step": 1272 + }, + { + "epoch": 0.17, + "learning_rate": 1.3889798145117295e-05, + "loss": 0.1117, + "step": 1273 + }, + { + "epoch": 0.17, + "learning_rate": 1.3900709219858157e-05, + "loss": 0.1152, + "step": 1274 + }, + { + "epoch": 0.17, + "learning_rate": 1.391162029459902e-05, + "loss": 0.1195, + "step": 1275 + }, + { + "epoch": 0.17, + "learning_rate": 1.392253136933988e-05, + "loss": 0.1359, + "step": 1276 + }, + { + "epoch": 0.17, + "learning_rate": 1.3933442444080742e-05, + "loss": 0.1204, + "step": 1277 + }, + { + "epoch": 0.17, + "learning_rate": 1.3944353518821605e-05, + "loss": 0.1168, + "step": 1278 + }, + { + "epoch": 0.17, + "learning_rate": 1.3955264593562468e-05, + "loss": 0.1277, + "step": 1279 + }, + { + "epoch": 0.17, + "learning_rate": 1.396617566830333e-05, + "loss": 0.1163, + "step": 1280 + }, + { + "epoch": 0.17, + "learning_rate": 1.397708674304419e-05, + "loss": 0.1181, + "step": 1281 + }, + { + "epoch": 0.17, + "learning_rate": 1.3987997817785053e-05, + "loss": 0.1135, + "step": 1282 + }, + { + "epoch": 0.17, + "learning_rate": 1.3998908892525916e-05, + "loss": 0.1255, + "step": 1283 + }, + { + "epoch": 0.17, + "learning_rate": 1.4009819967266778e-05, + "loss": 0.1228, + "step": 1284 + }, + { + "epoch": 0.17, + "learning_rate": 1.4020731042007638e-05, + "loss": 0.1209, + "step": 1285 + }, + { + "epoch": 0.17, + "learning_rate": 1.40316421167485e-05, + "loss": 0.1202, + "step": 1286 + }, + { + "epoch": 0.17, + "learning_rate": 1.4042553191489363e-05, + "loss": 0.1174, + "step": 1287 + }, + { + "epoch": 0.17, + "learning_rate": 1.4053464266230226e-05, + "loss": 0.1091, + "step": 1288 + }, + { + "epoch": 0.17, + "learning_rate": 1.4064375340971085e-05, + "loss": 0.1252, + "step": 1289 + }, + { + "epoch": 0.17, + "learning_rate": 1.4075286415711948e-05, + "loss": 0.1116, + "step": 1290 + }, + { + "epoch": 0.17, + "learning_rate": 1.4086197490452811e-05, + "loss": 0.1225, + "step": 1291 + }, + { + "epoch": 0.17, + "learning_rate": 1.4097108565193674e-05, + "loss": 0.1038, + "step": 1292 + }, + { + "epoch": 0.17, + "learning_rate": 1.4108019639934533e-05, + "loss": 0.1342, + "step": 1293 + }, + { + "epoch": 0.17, + "learning_rate": 1.4118930714675396e-05, + "loss": 0.1109, + "step": 1294 + }, + { + "epoch": 0.17, + "learning_rate": 1.4129841789416259e-05, + "loss": 0.1009, + "step": 1295 + }, + { + "epoch": 0.17, + "learning_rate": 1.4140752864157122e-05, + "loss": 0.1226, + "step": 1296 + }, + { + "epoch": 0.17, + "learning_rate": 1.4151663938897981e-05, + "loss": 0.1189, + "step": 1297 + }, + { + "epoch": 0.17, + "learning_rate": 1.4162575013638844e-05, + "loss": 0.1185, + "step": 1298 + }, + { + "epoch": 0.17, + "learning_rate": 1.4173486088379706e-05, + "loss": 0.1332, + "step": 1299 + }, + { + "epoch": 0.17, + "learning_rate": 1.418439716312057e-05, + "loss": 0.1249, + "step": 1300 + }, + { + "epoch": 0.17, + "learning_rate": 1.4195308237861432e-05, + "loss": 0.1208, + "step": 1301 + }, + { + "epoch": 0.17, + "learning_rate": 1.4206219312602291e-05, + "loss": 0.1244, + "step": 1302 + }, + { + "epoch": 0.17, + "learning_rate": 1.4217130387343154e-05, + "loss": 0.1212, + "step": 1303 + }, + { + "epoch": 0.17, + "learning_rate": 1.4228041462084017e-05, + "loss": 0.1196, + "step": 1304 + }, + { + "epoch": 0.17, + "learning_rate": 1.423895253682488e-05, + "loss": 0.0914, + "step": 1305 + }, + { + "epoch": 0.17, + "learning_rate": 1.424986361156574e-05, + "loss": 0.0999, + "step": 1306 + }, + { + "epoch": 0.17, + "learning_rate": 1.4260774686306602e-05, + "loss": 0.141, + "step": 1307 + }, + { + "epoch": 0.17, + "learning_rate": 1.4271685761047465e-05, + "loss": 0.1207, + "step": 1308 + }, + { + "epoch": 0.17, + "learning_rate": 1.4282596835788328e-05, + "loss": 0.1144, + "step": 1309 + }, + { + "epoch": 0.17, + "learning_rate": 1.4293507910529187e-05, + "loss": 0.1102, + "step": 1310 + }, + { + "epoch": 0.17, + "learning_rate": 1.430441898527005e-05, + "loss": 0.1147, + "step": 1311 + }, + { + "epoch": 0.17, + "learning_rate": 1.4315330060010912e-05, + "loss": 0.1135, + "step": 1312 + }, + { + "epoch": 0.17, + "learning_rate": 1.4326241134751775e-05, + "loss": 0.1204, + "step": 1313 + }, + { + "epoch": 0.17, + "learning_rate": 1.4337152209492635e-05, + "loss": 0.1147, + "step": 1314 + }, + { + "epoch": 0.17, + "learning_rate": 1.4348063284233497e-05, + "loss": 0.1174, + "step": 1315 + }, + { + "epoch": 0.17, + "learning_rate": 1.435897435897436e-05, + "loss": 0.1146, + "step": 1316 + }, + { + "epoch": 0.17, + "learning_rate": 1.4369885433715223e-05, + "loss": 0.1215, + "step": 1317 + }, + { + "epoch": 0.17, + "learning_rate": 1.4380796508456082e-05, + "loss": 0.1204, + "step": 1318 + }, + { + "epoch": 0.17, + "learning_rate": 1.4391707583196945e-05, + "loss": 0.1153, + "step": 1319 + }, + { + "epoch": 0.17, + "learning_rate": 1.4402618657937808e-05, + "loss": 0.1396, + "step": 1320 + }, + { + "epoch": 0.17, + "learning_rate": 1.441352973267867e-05, + "loss": 0.1182, + "step": 1321 + }, + { + "epoch": 0.17, + "learning_rate": 1.4424440807419534e-05, + "loss": 0.1065, + "step": 1322 + }, + { + "epoch": 0.17, + "learning_rate": 1.4435351882160393e-05, + "loss": 0.1089, + "step": 1323 + }, + { + "epoch": 0.17, + "learning_rate": 1.4446262956901256e-05, + "loss": 0.1144, + "step": 1324 + }, + { + "epoch": 0.17, + "learning_rate": 1.4457174031642118e-05, + "loss": 0.1053, + "step": 1325 + }, + { + "epoch": 0.17, + "learning_rate": 1.4468085106382981e-05, + "loss": 0.1142, + "step": 1326 + }, + { + "epoch": 0.17, + "learning_rate": 1.447899618112384e-05, + "loss": 0.1191, + "step": 1327 + }, + { + "epoch": 0.17, + "learning_rate": 1.4489907255864703e-05, + "loss": 0.1217, + "step": 1328 + }, + { + "epoch": 0.17, + "learning_rate": 1.4500818330605566e-05, + "loss": 0.1084, + "step": 1329 + }, + { + "epoch": 0.17, + "learning_rate": 1.4511729405346429e-05, + "loss": 0.1198, + "step": 1330 + }, + { + "epoch": 0.17, + "learning_rate": 1.4522640480087288e-05, + "loss": 0.1139, + "step": 1331 + }, + { + "epoch": 0.17, + "learning_rate": 1.4533551554828151e-05, + "loss": 0.1165, + "step": 1332 + }, + { + "epoch": 0.17, + "learning_rate": 1.4544462629569014e-05, + "loss": 0.1153, + "step": 1333 + }, + { + "epoch": 0.17, + "learning_rate": 1.4555373704309877e-05, + "loss": 0.1235, + "step": 1334 + }, + { + "epoch": 0.17, + "learning_rate": 1.4566284779050736e-05, + "loss": 0.1086, + "step": 1335 + }, + { + "epoch": 0.18, + "learning_rate": 1.4577195853791599e-05, + "loss": 0.1273, + "step": 1336 + }, + { + "epoch": 0.18, + "learning_rate": 1.4588106928532462e-05, + "loss": 0.1158, + "step": 1337 + }, + { + "epoch": 0.18, + "learning_rate": 1.4599018003273324e-05, + "loss": 0.1226, + "step": 1338 + }, + { + "epoch": 0.18, + "learning_rate": 1.4609929078014187e-05, + "loss": 0.1192, + "step": 1339 + }, + { + "epoch": 0.18, + "learning_rate": 1.4620840152755047e-05, + "loss": 0.1177, + "step": 1340 + }, + { + "epoch": 0.18, + "learning_rate": 1.463175122749591e-05, + "loss": 0.1309, + "step": 1341 + }, + { + "epoch": 0.18, + "learning_rate": 1.4642662302236772e-05, + "loss": 0.1106, + "step": 1342 + }, + { + "epoch": 0.18, + "learning_rate": 1.4653573376977635e-05, + "loss": 0.1164, + "step": 1343 + }, + { + "epoch": 0.18, + "learning_rate": 1.4664484451718494e-05, + "loss": 0.1005, + "step": 1344 + }, + { + "epoch": 0.18, + "learning_rate": 1.4675395526459357e-05, + "loss": 0.1119, + "step": 1345 + }, + { + "epoch": 0.18, + "learning_rate": 1.468630660120022e-05, + "loss": 0.1047, + "step": 1346 + }, + { + "epoch": 0.18, + "learning_rate": 1.4697217675941083e-05, + "loss": 0.114, + "step": 1347 + }, + { + "epoch": 0.18, + "learning_rate": 1.4708128750681942e-05, + "loss": 0.1323, + "step": 1348 + }, + { + "epoch": 0.18, + "learning_rate": 1.4719039825422805e-05, + "loss": 0.1176, + "step": 1349 + }, + { + "epoch": 0.18, + "learning_rate": 1.4729950900163668e-05, + "loss": 0.1255, + "step": 1350 + }, + { + "epoch": 0.18, + "learning_rate": 1.474086197490453e-05, + "loss": 0.1183, + "step": 1351 + }, + { + "epoch": 0.18, + "learning_rate": 1.475177304964539e-05, + "loss": 0.1116, + "step": 1352 + }, + { + "epoch": 0.18, + "learning_rate": 1.4762684124386253e-05, + "loss": 0.0991, + "step": 1353 + }, + { + "epoch": 0.18, + "learning_rate": 1.4773595199127115e-05, + "loss": 0.1242, + "step": 1354 + }, + { + "epoch": 0.18, + "learning_rate": 1.4784506273867978e-05, + "loss": 0.1189, + "step": 1355 + }, + { + "epoch": 0.18, + "learning_rate": 1.4795417348608838e-05, + "loss": 0.0933, + "step": 1356 + }, + { + "epoch": 0.18, + "learning_rate": 1.48063284233497e-05, + "loss": 0.119, + "step": 1357 + }, + { + "epoch": 0.18, + "learning_rate": 1.4817239498090563e-05, + "loss": 0.119, + "step": 1358 + }, + { + "epoch": 0.18, + "learning_rate": 1.4828150572831426e-05, + "loss": 0.1234, + "step": 1359 + }, + { + "epoch": 0.18, + "learning_rate": 1.4839061647572289e-05, + "loss": 0.1112, + "step": 1360 + }, + { + "epoch": 0.18, + "learning_rate": 1.4849972722313148e-05, + "loss": 0.1162, + "step": 1361 + }, + { + "epoch": 0.18, + "learning_rate": 1.4860883797054011e-05, + "loss": 0.1163, + "step": 1362 + }, + { + "epoch": 0.18, + "learning_rate": 1.4871794871794874e-05, + "loss": 0.1329, + "step": 1363 + }, + { + "epoch": 0.18, + "learning_rate": 1.4882705946535736e-05, + "loss": 0.1172, + "step": 1364 + }, + { + "epoch": 0.18, + "learning_rate": 1.4893617021276596e-05, + "loss": 0.1157, + "step": 1365 + }, + { + "epoch": 0.18, + "learning_rate": 1.4904528096017459e-05, + "loss": 0.1256, + "step": 1366 + }, + { + "epoch": 0.18, + "learning_rate": 1.4915439170758321e-05, + "loss": 0.1086, + "step": 1367 + }, + { + "epoch": 0.18, + "learning_rate": 1.4926350245499184e-05, + "loss": 0.117, + "step": 1368 + }, + { + "epoch": 0.18, + "learning_rate": 1.4937261320240044e-05, + "loss": 0.1189, + "step": 1369 + }, + { + "epoch": 0.18, + "learning_rate": 1.4948172394980906e-05, + "loss": 0.1092, + "step": 1370 + }, + { + "epoch": 0.18, + "learning_rate": 1.495908346972177e-05, + "loss": 0.1241, + "step": 1371 + }, + { + "epoch": 0.18, + "learning_rate": 1.4969994544462632e-05, + "loss": 0.1354, + "step": 1372 + }, + { + "epoch": 0.18, + "learning_rate": 1.4980905619203491e-05, + "loss": 0.12, + "step": 1373 + }, + { + "epoch": 0.18, + "learning_rate": 1.4991816693944354e-05, + "loss": 0.1252, + "step": 1374 + }, + { + "epoch": 0.18, + "learning_rate": 1.5002727768685217e-05, + "loss": 0.1188, + "step": 1375 + }, + { + "epoch": 0.18, + "learning_rate": 1.501363884342608e-05, + "loss": 0.1228, + "step": 1376 + }, + { + "epoch": 0.18, + "learning_rate": 1.5024549918166939e-05, + "loss": 0.115, + "step": 1377 + }, + { + "epoch": 0.18, + "learning_rate": 1.5035460992907802e-05, + "loss": 0.117, + "step": 1378 + }, + { + "epoch": 0.18, + "learning_rate": 1.5046372067648665e-05, + "loss": 0.1233, + "step": 1379 + }, + { + "epoch": 0.18, + "learning_rate": 1.5057283142389527e-05, + "loss": 0.1278, + "step": 1380 + }, + { + "epoch": 0.18, + "learning_rate": 1.506819421713039e-05, + "loss": 0.1191, + "step": 1381 + }, + { + "epoch": 0.18, + "learning_rate": 1.507910529187125e-05, + "loss": 0.1121, + "step": 1382 + }, + { + "epoch": 0.18, + "learning_rate": 1.5090016366612112e-05, + "loss": 0.1227, + "step": 1383 + }, + { + "epoch": 0.18, + "learning_rate": 1.5100927441352975e-05, + "loss": 0.1089, + "step": 1384 + }, + { + "epoch": 0.18, + "learning_rate": 1.5111838516093838e-05, + "loss": 0.1128, + "step": 1385 + }, + { + "epoch": 0.18, + "learning_rate": 1.5122749590834697e-05, + "loss": 0.1246, + "step": 1386 + }, + { + "epoch": 0.18, + "learning_rate": 1.513366066557556e-05, + "loss": 0.1198, + "step": 1387 + }, + { + "epoch": 0.18, + "learning_rate": 1.5144571740316423e-05, + "loss": 0.1175, + "step": 1388 + }, + { + "epoch": 0.18, + "learning_rate": 1.5155482815057286e-05, + "loss": 0.1222, + "step": 1389 + }, + { + "epoch": 0.18, + "learning_rate": 1.5166393889798145e-05, + "loss": 0.1238, + "step": 1390 + }, + { + "epoch": 0.18, + "learning_rate": 1.5177304964539008e-05, + "loss": 0.117, + "step": 1391 + }, + { + "epoch": 0.18, + "learning_rate": 1.518821603927987e-05, + "loss": 0.1423, + "step": 1392 + }, + { + "epoch": 0.18, + "learning_rate": 1.5199127114020733e-05, + "loss": 0.1169, + "step": 1393 + }, + { + "epoch": 0.18, + "learning_rate": 1.5210038188761593e-05, + "loss": 0.1264, + "step": 1394 + }, + { + "epoch": 0.18, + "learning_rate": 1.5220949263502456e-05, + "loss": 0.1191, + "step": 1395 + }, + { + "epoch": 0.18, + "learning_rate": 1.5231860338243318e-05, + "loss": 0.1119, + "step": 1396 + }, + { + "epoch": 0.18, + "learning_rate": 1.5242771412984181e-05, + "loss": 0.1095, + "step": 1397 + }, + { + "epoch": 0.18, + "learning_rate": 1.525368248772504e-05, + "loss": 0.1186, + "step": 1398 + }, + { + "epoch": 0.18, + "learning_rate": 1.5264593562465903e-05, + "loss": 0.1084, + "step": 1399 + }, + { + "epoch": 0.18, + "learning_rate": 1.5275504637206766e-05, + "loss": 0.1131, + "step": 1400 + }, + { + "epoch": 0.18, + "learning_rate": 1.528641571194763e-05, + "loss": 0.1114, + "step": 1401 + }, + { + "epoch": 0.18, + "learning_rate": 1.5297326786688492e-05, + "loss": 0.1168, + "step": 1402 + }, + { + "epoch": 0.18, + "learning_rate": 1.530823786142935e-05, + "loss": 0.1165, + "step": 1403 + }, + { + "epoch": 0.18, + "learning_rate": 1.5319148936170214e-05, + "loss": 0.1251, + "step": 1404 + }, + { + "epoch": 0.18, + "learning_rate": 1.5330060010911077e-05, + "loss": 0.1338, + "step": 1405 + }, + { + "epoch": 0.18, + "learning_rate": 1.534097108565194e-05, + "loss": 0.1153, + "step": 1406 + }, + { + "epoch": 0.18, + "learning_rate": 1.53518821603928e-05, + "loss": 0.1067, + "step": 1407 + }, + { + "epoch": 0.18, + "learning_rate": 1.536279323513366e-05, + "loss": 0.1181, + "step": 1408 + }, + { + "epoch": 0.18, + "learning_rate": 1.5373704309874524e-05, + "loss": 0.119, + "step": 1409 + }, + { + "epoch": 0.18, + "learning_rate": 1.5384615384615387e-05, + "loss": 0.1118, + "step": 1410 + }, + { + "epoch": 0.18, + "learning_rate": 1.5395526459356247e-05, + "loss": 0.1184, + "step": 1411 + }, + { + "epoch": 0.18, + "learning_rate": 1.540643753409711e-05, + "loss": 0.1137, + "step": 1412 + }, + { + "epoch": 0.19, + "learning_rate": 1.5417348608837972e-05, + "loss": 0.1165, + "step": 1413 + }, + { + "epoch": 0.19, + "learning_rate": 1.5428259683578835e-05, + "loss": 0.1277, + "step": 1414 + }, + { + "epoch": 0.19, + "learning_rate": 1.5439170758319694e-05, + "loss": 0.1254, + "step": 1415 + }, + { + "epoch": 0.19, + "learning_rate": 1.5450081833060557e-05, + "loss": 0.1177, + "step": 1416 + }, + { + "epoch": 0.19, + "learning_rate": 1.546099290780142e-05, + "loss": 0.1235, + "step": 1417 + }, + { + "epoch": 0.19, + "learning_rate": 1.5471903982542283e-05, + "loss": 0.1214, + "step": 1418 + }, + { + "epoch": 0.19, + "learning_rate": 1.5482815057283142e-05, + "loss": 0.124, + "step": 1419 + }, + { + "epoch": 0.19, + "learning_rate": 1.5493726132024005e-05, + "loss": 0.0992, + "step": 1420 + }, + { + "epoch": 0.19, + "learning_rate": 1.5504637206764868e-05, + "loss": 0.1227, + "step": 1421 + }, + { + "epoch": 0.19, + "learning_rate": 1.551554828150573e-05, + "loss": 0.1114, + "step": 1422 + }, + { + "epoch": 0.19, + "learning_rate": 1.5526459356246593e-05, + "loss": 0.111, + "step": 1423 + }, + { + "epoch": 0.19, + "learning_rate": 1.5537370430987453e-05, + "loss": 0.1194, + "step": 1424 + }, + { + "epoch": 0.19, + "learning_rate": 1.5548281505728315e-05, + "loss": 0.1167, + "step": 1425 + }, + { + "epoch": 0.19, + "learning_rate": 1.5559192580469178e-05, + "loss": 0.1161, + "step": 1426 + }, + { + "epoch": 0.19, + "learning_rate": 1.557010365521004e-05, + "loss": 0.1251, + "step": 1427 + }, + { + "epoch": 0.19, + "learning_rate": 1.55810147299509e-05, + "loss": 0.1156, + "step": 1428 + }, + { + "epoch": 0.19, + "learning_rate": 1.5591925804691763e-05, + "loss": 0.1173, + "step": 1429 + }, + { + "epoch": 0.19, + "learning_rate": 1.5602836879432626e-05, + "loss": 0.1195, + "step": 1430 + }, + { + "epoch": 0.19, + "learning_rate": 1.561374795417349e-05, + "loss": 0.1173, + "step": 1431 + }, + { + "epoch": 0.19, + "learning_rate": 1.5624659028914348e-05, + "loss": 0.1169, + "step": 1432 + }, + { + "epoch": 0.19, + "learning_rate": 1.563557010365521e-05, + "loss": 0.1133, + "step": 1433 + }, + { + "epoch": 0.19, + "learning_rate": 1.5646481178396074e-05, + "loss": 0.1199, + "step": 1434 + }, + { + "epoch": 0.19, + "learning_rate": 1.5657392253136936e-05, + "loss": 0.1241, + "step": 1435 + }, + { + "epoch": 0.19, + "learning_rate": 1.5668303327877796e-05, + "loss": 0.1201, + "step": 1436 + }, + { + "epoch": 0.19, + "learning_rate": 1.567921440261866e-05, + "loss": 0.112, + "step": 1437 + }, + { + "epoch": 0.19, + "learning_rate": 1.569012547735952e-05, + "loss": 0.1091, + "step": 1438 + }, + { + "epoch": 0.19, + "learning_rate": 1.5701036552100384e-05, + "loss": 0.1113, + "step": 1439 + }, + { + "epoch": 0.19, + "learning_rate": 1.5711947626841244e-05, + "loss": 0.1059, + "step": 1440 + }, + { + "epoch": 0.19, + "learning_rate": 1.5722858701582106e-05, + "loss": 0.1248, + "step": 1441 + }, + { + "epoch": 0.19, + "learning_rate": 1.573376977632297e-05, + "loss": 0.1125, + "step": 1442 + }, + { + "epoch": 0.19, + "learning_rate": 1.5744680851063832e-05, + "loss": 0.1357, + "step": 1443 + }, + { + "epoch": 0.19, + "learning_rate": 1.5755591925804695e-05, + "loss": 0.1072, + "step": 1444 + }, + { + "epoch": 0.19, + "learning_rate": 1.5766503000545554e-05, + "loss": 0.1364, + "step": 1445 + }, + { + "epoch": 0.19, + "learning_rate": 1.5777414075286417e-05, + "loss": 0.1252, + "step": 1446 + }, + { + "epoch": 0.19, + "learning_rate": 1.578832515002728e-05, + "loss": 0.1085, + "step": 1447 + }, + { + "epoch": 0.19, + "learning_rate": 1.5799236224768142e-05, + "loss": 0.1149, + "step": 1448 + }, + { + "epoch": 0.19, + "learning_rate": 1.5810147299509002e-05, + "loss": 0.1222, + "step": 1449 + }, + { + "epoch": 0.19, + "learning_rate": 1.5821058374249865e-05, + "loss": 0.102, + "step": 1450 + }, + { + "epoch": 0.19, + "learning_rate": 1.5831969448990727e-05, + "loss": 0.1152, + "step": 1451 + }, + { + "epoch": 0.19, + "learning_rate": 1.584288052373159e-05, + "loss": 0.111, + "step": 1452 + }, + { + "epoch": 0.19, + "learning_rate": 1.585379159847245e-05, + "loss": 0.1224, + "step": 1453 + }, + { + "epoch": 0.19, + "learning_rate": 1.5864702673213312e-05, + "loss": 0.1152, + "step": 1454 + }, + { + "epoch": 0.19, + "learning_rate": 1.5875613747954175e-05, + "loss": 0.1153, + "step": 1455 + }, + { + "epoch": 0.19, + "learning_rate": 1.5886524822695038e-05, + "loss": 0.1242, + "step": 1456 + }, + { + "epoch": 0.19, + "learning_rate": 1.5897435897435897e-05, + "loss": 0.1181, + "step": 1457 + }, + { + "epoch": 0.19, + "learning_rate": 1.590834697217676e-05, + "loss": 0.108, + "step": 1458 + }, + { + "epoch": 0.19, + "learning_rate": 1.5919258046917623e-05, + "loss": 0.1164, + "step": 1459 + }, + { + "epoch": 0.19, + "learning_rate": 1.5930169121658486e-05, + "loss": 0.1121, + "step": 1460 + }, + { + "epoch": 0.19, + "learning_rate": 1.5941080196399345e-05, + "loss": 0.129, + "step": 1461 + }, + { + "epoch": 0.19, + "learning_rate": 1.5951991271140208e-05, + "loss": 0.1108, + "step": 1462 + }, + { + "epoch": 0.19, + "learning_rate": 1.596290234588107e-05, + "loss": 0.1248, + "step": 1463 + }, + { + "epoch": 0.19, + "learning_rate": 1.5973813420621933e-05, + "loss": 0.1098, + "step": 1464 + }, + { + "epoch": 0.19, + "learning_rate": 1.5984724495362796e-05, + "loss": 0.1149, + "step": 1465 + }, + { + "epoch": 0.19, + "learning_rate": 1.5995635570103656e-05, + "loss": 0.12, + "step": 1466 + }, + { + "epoch": 0.19, + "learning_rate": 1.6006546644844518e-05, + "loss": 0.1093, + "step": 1467 + }, + { + "epoch": 0.19, + "learning_rate": 1.601745771958538e-05, + "loss": 0.1128, + "step": 1468 + }, + { + "epoch": 0.19, + "learning_rate": 1.6028368794326244e-05, + "loss": 0.1066, + "step": 1469 + }, + { + "epoch": 0.19, + "learning_rate": 1.6039279869067103e-05, + "loss": 0.1061, + "step": 1470 + }, + { + "epoch": 0.19, + "learning_rate": 1.6050190943807966e-05, + "loss": 0.1022, + "step": 1471 + }, + { + "epoch": 0.19, + "learning_rate": 1.606110201854883e-05, + "loss": 0.1203, + "step": 1472 + }, + { + "epoch": 0.19, + "learning_rate": 1.607201309328969e-05, + "loss": 0.1139, + "step": 1473 + }, + { + "epoch": 0.19, + "learning_rate": 1.608292416803055e-05, + "loss": 0.1285, + "step": 1474 + }, + { + "epoch": 0.19, + "learning_rate": 1.6093835242771414e-05, + "loss": 0.1037, + "step": 1475 + }, + { + "epoch": 0.19, + "learning_rate": 1.6104746317512277e-05, + "loss": 0.1172, + "step": 1476 + }, + { + "epoch": 0.19, + "learning_rate": 1.611565739225314e-05, + "loss": 0.1051, + "step": 1477 + }, + { + "epoch": 0.19, + "learning_rate": 1.6126568466994e-05, + "loss": 0.1095, + "step": 1478 + }, + { + "epoch": 0.19, + "learning_rate": 1.613747954173486e-05, + "loss": 0.1065, + "step": 1479 + }, + { + "epoch": 0.19, + "learning_rate": 1.6148390616475724e-05, + "loss": 0.1127, + "step": 1480 + }, + { + "epoch": 0.19, + "learning_rate": 1.6159301691216587e-05, + "loss": 0.1135, + "step": 1481 + }, + { + "epoch": 0.19, + "learning_rate": 1.6170212765957446e-05, + "loss": 0.1061, + "step": 1482 + }, + { + "epoch": 0.19, + "learning_rate": 1.618112384069831e-05, + "loss": 0.1077, + "step": 1483 + }, + { + "epoch": 0.19, + "learning_rate": 1.6192034915439172e-05, + "loss": 0.1147, + "step": 1484 + }, + { + "epoch": 0.19, + "learning_rate": 1.6202945990180035e-05, + "loss": 0.1085, + "step": 1485 + }, + { + "epoch": 0.19, + "learning_rate": 1.6213857064920898e-05, + "loss": 0.1229, + "step": 1486 + }, + { + "epoch": 0.19, + "learning_rate": 1.6224768139661757e-05, + "loss": 0.112, + "step": 1487 + }, + { + "epoch": 0.19, + "learning_rate": 1.623567921440262e-05, + "loss": 0.1188, + "step": 1488 + }, + { + "epoch": 0.2, + "learning_rate": 1.6246590289143483e-05, + "loss": 0.1188, + "step": 1489 + }, + { + "epoch": 0.2, + "learning_rate": 1.6257501363884345e-05, + "loss": 0.1229, + "step": 1490 + }, + { + "epoch": 0.2, + "learning_rate": 1.6268412438625205e-05, + "loss": 0.1108, + "step": 1491 + }, + { + "epoch": 0.2, + "learning_rate": 1.6279323513366068e-05, + "loss": 0.119, + "step": 1492 + }, + { + "epoch": 0.2, + "learning_rate": 1.629023458810693e-05, + "loss": 0.104, + "step": 1493 + }, + { + "epoch": 0.2, + "learning_rate": 1.6301145662847793e-05, + "loss": 0.1342, + "step": 1494 + }, + { + "epoch": 0.2, + "learning_rate": 1.6312056737588652e-05, + "loss": 0.1201, + "step": 1495 + }, + { + "epoch": 0.2, + "learning_rate": 1.6322967812329515e-05, + "loss": 0.1077, + "step": 1496 + }, + { + "epoch": 0.2, + "learning_rate": 1.6333878887070378e-05, + "loss": 0.1119, + "step": 1497 + }, + { + "epoch": 0.2, + "learning_rate": 1.634478996181124e-05, + "loss": 0.1255, + "step": 1498 + }, + { + "epoch": 0.2, + "learning_rate": 1.63557010365521e-05, + "loss": 0.1191, + "step": 1499 + }, + { + "epoch": 0.2, + "learning_rate": 1.6366612111292963e-05, + "loss": 0.1218, + "step": 1500 + }, + { + "epoch": 0.2, + "learning_rate": 1.6377523186033826e-05, + "loss": 0.11, + "step": 1501 + }, + { + "epoch": 0.2, + "learning_rate": 1.638843426077469e-05, + "loss": 0.1126, + "step": 1502 + }, + { + "epoch": 0.2, + "learning_rate": 1.6399345335515548e-05, + "loss": 0.1169, + "step": 1503 + }, + { + "epoch": 0.2, + "learning_rate": 1.641025641025641e-05, + "loss": 0.1067, + "step": 1504 + }, + { + "epoch": 0.2, + "learning_rate": 1.6421167484997274e-05, + "loss": 0.1126, + "step": 1505 + }, + { + "epoch": 0.2, + "learning_rate": 1.6432078559738136e-05, + "loss": 0.0973, + "step": 1506 + }, + { + "epoch": 0.2, + "learning_rate": 1.6442989634479e-05, + "loss": 0.1081, + "step": 1507 + }, + { + "epoch": 0.2, + "learning_rate": 1.645390070921986e-05, + "loss": 0.1256, + "step": 1508 + }, + { + "epoch": 0.2, + "learning_rate": 1.646481178396072e-05, + "loss": 0.1136, + "step": 1509 + }, + { + "epoch": 0.2, + "learning_rate": 1.6475722858701584e-05, + "loss": 0.1053, + "step": 1510 + }, + { + "epoch": 0.2, + "learning_rate": 1.6486633933442447e-05, + "loss": 0.113, + "step": 1511 + }, + { + "epoch": 0.2, + "learning_rate": 1.6497545008183306e-05, + "loss": 0.1165, + "step": 1512 + }, + { + "epoch": 0.2, + "learning_rate": 1.650845608292417e-05, + "loss": 0.1159, + "step": 1513 + }, + { + "epoch": 0.2, + "learning_rate": 1.6519367157665032e-05, + "loss": 0.105, + "step": 1514 + }, + { + "epoch": 0.2, + "learning_rate": 1.6530278232405895e-05, + "loss": 0.1216, + "step": 1515 + }, + { + "epoch": 0.2, + "learning_rate": 1.6541189307146754e-05, + "loss": 0.1009, + "step": 1516 + }, + { + "epoch": 0.2, + "learning_rate": 1.6552100381887617e-05, + "loss": 0.1085, + "step": 1517 + }, + { + "epoch": 0.2, + "learning_rate": 1.656301145662848e-05, + "loss": 0.1187, + "step": 1518 + }, + { + "epoch": 0.2, + "learning_rate": 1.6573922531369342e-05, + "loss": 0.1006, + "step": 1519 + }, + { + "epoch": 0.2, + "learning_rate": 1.65848336061102e-05, + "loss": 0.1103, + "step": 1520 + }, + { + "epoch": 0.2, + "learning_rate": 1.6595744680851064e-05, + "loss": 0.1218, + "step": 1521 + }, + { + "epoch": 0.2, + "learning_rate": 1.6606655755591927e-05, + "loss": 0.1223, + "step": 1522 + }, + { + "epoch": 0.2, + "learning_rate": 1.661756683033279e-05, + "loss": 0.113, + "step": 1523 + }, + { + "epoch": 0.2, + "learning_rate": 1.662847790507365e-05, + "loss": 0.1191, + "step": 1524 + }, + { + "epoch": 0.2, + "learning_rate": 1.6639388979814512e-05, + "loss": 0.117, + "step": 1525 + }, + { + "epoch": 0.2, + "learning_rate": 1.6650300054555375e-05, + "loss": 0.1238, + "step": 1526 + }, + { + "epoch": 0.2, + "learning_rate": 1.6661211129296238e-05, + "loss": 0.1266, + "step": 1527 + }, + { + "epoch": 0.2, + "learning_rate": 1.66721222040371e-05, + "loss": 0.1176, + "step": 1528 + }, + { + "epoch": 0.2, + "learning_rate": 1.668303327877796e-05, + "loss": 0.1122, + "step": 1529 + }, + { + "epoch": 0.2, + "learning_rate": 1.6693944353518823e-05, + "loss": 0.1231, + "step": 1530 + }, + { + "epoch": 0.2, + "learning_rate": 1.6704855428259686e-05, + "loss": 0.1267, + "step": 1531 + }, + { + "epoch": 0.2, + "learning_rate": 1.6715766503000548e-05, + "loss": 0.1075, + "step": 1532 + }, + { + "epoch": 0.2, + "learning_rate": 1.6726677577741408e-05, + "loss": 0.1246, + "step": 1533 + }, + { + "epoch": 0.2, + "learning_rate": 1.673758865248227e-05, + "loss": 0.0989, + "step": 1534 + }, + { + "epoch": 0.2, + "learning_rate": 1.6748499727223133e-05, + "loss": 0.1098, + "step": 1535 + }, + { + "epoch": 0.2, + "learning_rate": 1.6759410801963996e-05, + "loss": 0.1291, + "step": 1536 + }, + { + "epoch": 0.2, + "learning_rate": 1.6770321876704855e-05, + "loss": 0.1187, + "step": 1537 + }, + { + "epoch": 0.2, + "learning_rate": 1.6781232951445718e-05, + "loss": 0.1128, + "step": 1538 + }, + { + "epoch": 0.2, + "learning_rate": 1.679214402618658e-05, + "loss": 0.0924, + "step": 1539 + }, + { + "epoch": 0.2, + "learning_rate": 1.6803055100927444e-05, + "loss": 0.1122, + "step": 1540 + }, + { + "epoch": 0.2, + "learning_rate": 1.6813966175668303e-05, + "loss": 0.1157, + "step": 1541 + }, + { + "epoch": 0.2, + "learning_rate": 1.6824877250409166e-05, + "loss": 0.1091, + "step": 1542 + }, + { + "epoch": 0.2, + "learning_rate": 1.683578832515003e-05, + "loss": 0.1072, + "step": 1543 + }, + { + "epoch": 0.2, + "learning_rate": 1.684669939989089e-05, + "loss": 0.1355, + "step": 1544 + }, + { + "epoch": 0.2, + "learning_rate": 1.6857610474631754e-05, + "loss": 0.1077, + "step": 1545 + }, + { + "epoch": 0.2, + "learning_rate": 1.6868521549372614e-05, + "loss": 0.1252, + "step": 1546 + }, + { + "epoch": 0.2, + "learning_rate": 1.6879432624113476e-05, + "loss": 0.1118, + "step": 1547 + }, + { + "epoch": 0.2, + "learning_rate": 1.689034369885434e-05, + "loss": 0.1307, + "step": 1548 + }, + { + "epoch": 0.2, + "learning_rate": 1.6901254773595202e-05, + "loss": 0.1051, + "step": 1549 + }, + { + "epoch": 0.2, + "learning_rate": 1.691216584833606e-05, + "loss": 0.1246, + "step": 1550 + }, + { + "epoch": 0.2, + "learning_rate": 1.6923076923076924e-05, + "loss": 0.1246, + "step": 1551 + }, + { + "epoch": 0.2, + "learning_rate": 1.6933987997817787e-05, + "loss": 0.1201, + "step": 1552 + }, + { + "epoch": 0.2, + "learning_rate": 1.694489907255865e-05, + "loss": 0.1215, + "step": 1553 + }, + { + "epoch": 0.2, + "learning_rate": 1.695581014729951e-05, + "loss": 0.1157, + "step": 1554 + }, + { + "epoch": 0.2, + "learning_rate": 1.6966721222040372e-05, + "loss": 0.1163, + "step": 1555 + }, + { + "epoch": 0.2, + "learning_rate": 1.6977632296781235e-05, + "loss": 0.117, + "step": 1556 + }, + { + "epoch": 0.2, + "learning_rate": 1.6988543371522098e-05, + "loss": 0.1054, + "step": 1557 + }, + { + "epoch": 0.2, + "learning_rate": 1.6999454446262957e-05, + "loss": 0.1143, + "step": 1558 + }, + { + "epoch": 0.2, + "learning_rate": 1.701036552100382e-05, + "loss": 0.1054, + "step": 1559 + }, + { + "epoch": 0.2, + "learning_rate": 1.7021276595744682e-05, + "loss": 0.1049, + "step": 1560 + }, + { + "epoch": 0.2, + "learning_rate": 1.7032187670485545e-05, + "loss": 0.1156, + "step": 1561 + }, + { + "epoch": 0.2, + "learning_rate": 1.7043098745226405e-05, + "loss": 0.1114, + "step": 1562 + }, + { + "epoch": 0.2, + "learning_rate": 1.7054009819967267e-05, + "loss": 0.1186, + "step": 1563 + }, + { + "epoch": 0.2, + "learning_rate": 1.706492089470813e-05, + "loss": 0.1165, + "step": 1564 + }, + { + "epoch": 0.21, + "learning_rate": 1.7075831969448993e-05, + "loss": 0.1286, + "step": 1565 + }, + { + "epoch": 0.21, + "learning_rate": 1.7086743044189856e-05, + "loss": 0.1189, + "step": 1566 + }, + { + "epoch": 0.21, + "learning_rate": 1.7097654118930715e-05, + "loss": 0.113, + "step": 1567 + }, + { + "epoch": 0.21, + "learning_rate": 1.7108565193671578e-05, + "loss": 0.1279, + "step": 1568 + }, + { + "epoch": 0.21, + "learning_rate": 1.711947626841244e-05, + "loss": 0.1176, + "step": 1569 + }, + { + "epoch": 0.21, + "learning_rate": 1.7130387343153304e-05, + "loss": 0.1038, + "step": 1570 + }, + { + "epoch": 0.21, + "learning_rate": 1.7141298417894163e-05, + "loss": 0.1038, + "step": 1571 + }, + { + "epoch": 0.21, + "learning_rate": 1.7152209492635026e-05, + "loss": 0.1154, + "step": 1572 + }, + { + "epoch": 0.21, + "learning_rate": 1.716312056737589e-05, + "loss": 0.1174, + "step": 1573 + }, + { + "epoch": 0.21, + "learning_rate": 1.717403164211675e-05, + "loss": 0.1187, + "step": 1574 + }, + { + "epoch": 0.21, + "learning_rate": 1.718494271685761e-05, + "loss": 0.1261, + "step": 1575 + }, + { + "epoch": 0.21, + "learning_rate": 1.7195853791598473e-05, + "loss": 0.1174, + "step": 1576 + }, + { + "epoch": 0.21, + "learning_rate": 1.7206764866339336e-05, + "loss": 0.1263, + "step": 1577 + }, + { + "epoch": 0.21, + "learning_rate": 1.72176759410802e-05, + "loss": 0.1144, + "step": 1578 + }, + { + "epoch": 0.21, + "learning_rate": 1.722858701582106e-05, + "loss": 0.118, + "step": 1579 + }, + { + "epoch": 0.21, + "learning_rate": 1.723949809056192e-05, + "loss": 0.1119, + "step": 1580 + }, + { + "epoch": 0.21, + "learning_rate": 1.7250409165302784e-05, + "loss": 0.1149, + "step": 1581 + }, + { + "epoch": 0.21, + "learning_rate": 1.7261320240043647e-05, + "loss": 0.1075, + "step": 1582 + }, + { + "epoch": 0.21, + "learning_rate": 1.7272231314784506e-05, + "loss": 0.1182, + "step": 1583 + }, + { + "epoch": 0.21, + "learning_rate": 1.728314238952537e-05, + "loss": 0.1134, + "step": 1584 + }, + { + "epoch": 0.21, + "learning_rate": 1.729405346426623e-05, + "loss": 0.1043, + "step": 1585 + }, + { + "epoch": 0.21, + "learning_rate": 1.7304964539007094e-05, + "loss": 0.1073, + "step": 1586 + }, + { + "epoch": 0.21, + "learning_rate": 1.7315875613747957e-05, + "loss": 0.1217, + "step": 1587 + }, + { + "epoch": 0.21, + "learning_rate": 1.7326786688488817e-05, + "loss": 0.114, + "step": 1588 + }, + { + "epoch": 0.21, + "learning_rate": 1.733769776322968e-05, + "loss": 0.1332, + "step": 1589 + }, + { + "epoch": 0.21, + "learning_rate": 1.7348608837970542e-05, + "loss": 0.1092, + "step": 1590 + }, + { + "epoch": 0.21, + "learning_rate": 1.7359519912711405e-05, + "loss": 0.1212, + "step": 1591 + }, + { + "epoch": 0.21, + "learning_rate": 1.7370430987452264e-05, + "loss": 0.1141, + "step": 1592 + }, + { + "epoch": 0.21, + "learning_rate": 1.7381342062193127e-05, + "loss": 0.1053, + "step": 1593 + }, + { + "epoch": 0.21, + "learning_rate": 1.739225313693399e-05, + "loss": 0.1243, + "step": 1594 + }, + { + "epoch": 0.21, + "learning_rate": 1.7403164211674853e-05, + "loss": 0.1024, + "step": 1595 + }, + { + "epoch": 0.21, + "learning_rate": 1.7414075286415712e-05, + "loss": 0.1109, + "step": 1596 + }, + { + "epoch": 0.21, + "learning_rate": 1.7424986361156575e-05, + "loss": 0.1139, + "step": 1597 + }, + { + "epoch": 0.21, + "learning_rate": 1.7435897435897438e-05, + "loss": 0.1167, + "step": 1598 + }, + { + "epoch": 0.21, + "learning_rate": 1.74468085106383e-05, + "loss": 0.113, + "step": 1599 + }, + { + "epoch": 0.21, + "learning_rate": 1.745771958537916e-05, + "loss": 0.1244, + "step": 1600 + }, + { + "epoch": 0.21, + "learning_rate": 1.7468630660120023e-05, + "loss": 0.1206, + "step": 1601 + }, + { + "epoch": 0.21, + "learning_rate": 1.7479541734860885e-05, + "loss": 0.1124, + "step": 1602 + }, + { + "epoch": 0.21, + "learning_rate": 1.7490452809601748e-05, + "loss": 0.1154, + "step": 1603 + }, + { + "epoch": 0.21, + "learning_rate": 1.7501363884342608e-05, + "loss": 0.1276, + "step": 1604 + }, + { + "epoch": 0.21, + "learning_rate": 1.751227495908347e-05, + "loss": 0.1225, + "step": 1605 + }, + { + "epoch": 0.21, + "learning_rate": 1.7523186033824333e-05, + "loss": 0.1092, + "step": 1606 + }, + { + "epoch": 0.21, + "learning_rate": 1.7534097108565196e-05, + "loss": 0.1066, + "step": 1607 + }, + { + "epoch": 0.21, + "learning_rate": 1.754500818330606e-05, + "loss": 0.1353, + "step": 1608 + }, + { + "epoch": 0.21, + "learning_rate": 1.7555919258046918e-05, + "loss": 0.1375, + "step": 1609 + }, + { + "epoch": 0.21, + "learning_rate": 1.756683033278778e-05, + "loss": 0.118, + "step": 1610 + }, + { + "epoch": 0.21, + "learning_rate": 1.7577741407528644e-05, + "loss": 0.116, + "step": 1611 + }, + { + "epoch": 0.21, + "learning_rate": 1.7588652482269506e-05, + "loss": 0.1257, + "step": 1612 + }, + { + "epoch": 0.21, + "learning_rate": 1.7599563557010366e-05, + "loss": 0.1145, + "step": 1613 + }, + { + "epoch": 0.21, + "learning_rate": 1.761047463175123e-05, + "loss": 0.1067, + "step": 1614 + }, + { + "epoch": 0.21, + "learning_rate": 1.762138570649209e-05, + "loss": 0.1188, + "step": 1615 + }, + { + "epoch": 0.21, + "learning_rate": 1.7632296781232954e-05, + "loss": 0.1181, + "step": 1616 + }, + { + "epoch": 0.21, + "learning_rate": 1.7643207855973814e-05, + "loss": 0.1192, + "step": 1617 + }, + { + "epoch": 0.21, + "learning_rate": 1.7654118930714676e-05, + "loss": 0.115, + "step": 1618 + }, + { + "epoch": 0.21, + "learning_rate": 1.766503000545554e-05, + "loss": 0.0923, + "step": 1619 + }, + { + "epoch": 0.21, + "learning_rate": 1.7675941080196402e-05, + "loss": 0.1181, + "step": 1620 + }, + { + "epoch": 0.21, + "learning_rate": 1.768685215493726e-05, + "loss": 0.1267, + "step": 1621 + }, + { + "epoch": 0.21, + "learning_rate": 1.7697763229678124e-05, + "loss": 0.1202, + "step": 1622 + }, + { + "epoch": 0.21, + "learning_rate": 1.7708674304418987e-05, + "loss": 0.1098, + "step": 1623 + }, + { + "epoch": 0.21, + "learning_rate": 1.771958537915985e-05, + "loss": 0.121, + "step": 1624 + }, + { + "epoch": 0.21, + "learning_rate": 1.773049645390071e-05, + "loss": 0.1187, + "step": 1625 + }, + { + "epoch": 0.21, + "learning_rate": 1.7741407528641572e-05, + "loss": 0.1085, + "step": 1626 + }, + { + "epoch": 0.21, + "learning_rate": 1.7752318603382435e-05, + "loss": 0.1085, + "step": 1627 + }, + { + "epoch": 0.21, + "learning_rate": 1.7763229678123297e-05, + "loss": 0.1126, + "step": 1628 + }, + { + "epoch": 0.21, + "learning_rate": 1.777414075286416e-05, + "loss": 0.1062, + "step": 1629 + }, + { + "epoch": 0.21, + "learning_rate": 1.778505182760502e-05, + "loss": 0.1097, + "step": 1630 + }, + { + "epoch": 0.21, + "learning_rate": 1.7795962902345882e-05, + "loss": 0.112, + "step": 1631 + }, + { + "epoch": 0.21, + "learning_rate": 1.7806873977086745e-05, + "loss": 0.1117, + "step": 1632 + }, + { + "epoch": 0.21, + "learning_rate": 1.7817785051827608e-05, + "loss": 0.1058, + "step": 1633 + }, + { + "epoch": 0.21, + "learning_rate": 1.7828696126568467e-05, + "loss": 0.1173, + "step": 1634 + }, + { + "epoch": 0.21, + "learning_rate": 1.783960720130933e-05, + "loss": 0.1173, + "step": 1635 + }, + { + "epoch": 0.21, + "learning_rate": 1.7850518276050193e-05, + "loss": 0.0897, + "step": 1636 + }, + { + "epoch": 0.21, + "learning_rate": 1.7861429350791056e-05, + "loss": 0.1159, + "step": 1637 + }, + { + "epoch": 0.21, + "learning_rate": 1.7872340425531915e-05, + "loss": 0.1297, + "step": 1638 + }, + { + "epoch": 0.21, + "learning_rate": 1.7883251500272778e-05, + "loss": 0.1181, + "step": 1639 + }, + { + "epoch": 0.21, + "learning_rate": 1.789416257501364e-05, + "loss": 0.113, + "step": 1640 + }, + { + "epoch": 0.21, + "learning_rate": 1.7905073649754503e-05, + "loss": 0.1161, + "step": 1641 + }, + { + "epoch": 0.22, + "learning_rate": 1.7915984724495363e-05, + "loss": 0.1009, + "step": 1642 + }, + { + "epoch": 0.22, + "learning_rate": 1.7926895799236226e-05, + "loss": 0.1145, + "step": 1643 + }, + { + "epoch": 0.22, + "learning_rate": 1.793780687397709e-05, + "loss": 0.1251, + "step": 1644 + }, + { + "epoch": 0.22, + "learning_rate": 1.794871794871795e-05, + "loss": 0.114, + "step": 1645 + }, + { + "epoch": 0.22, + "learning_rate": 1.795962902345881e-05, + "loss": 0.1214, + "step": 1646 + }, + { + "epoch": 0.22, + "learning_rate": 1.7970540098199673e-05, + "loss": 0.1101, + "step": 1647 + }, + { + "epoch": 0.22, + "learning_rate": 1.7981451172940536e-05, + "loss": 0.1044, + "step": 1648 + }, + { + "epoch": 0.22, + "learning_rate": 1.79923622476814e-05, + "loss": 0.1035, + "step": 1649 + }, + { + "epoch": 0.22, + "learning_rate": 1.800327332242226e-05, + "loss": 0.1057, + "step": 1650 + }, + { + "epoch": 0.22, + "learning_rate": 1.801418439716312e-05, + "loss": 0.1068, + "step": 1651 + }, + { + "epoch": 0.22, + "learning_rate": 1.8025095471903984e-05, + "loss": 0.0951, + "step": 1652 + }, + { + "epoch": 0.22, + "learning_rate": 1.8036006546644847e-05, + "loss": 0.1101, + "step": 1653 + }, + { + "epoch": 0.22, + "learning_rate": 1.804691762138571e-05, + "loss": 0.1118, + "step": 1654 + }, + { + "epoch": 0.22, + "learning_rate": 1.805782869612657e-05, + "loss": 0.1077, + "step": 1655 + }, + { + "epoch": 0.22, + "learning_rate": 1.806873977086743e-05, + "loss": 0.1099, + "step": 1656 + }, + { + "epoch": 0.22, + "learning_rate": 1.8079650845608294e-05, + "loss": 0.1136, + "step": 1657 + }, + { + "epoch": 0.22, + "learning_rate": 1.8090561920349157e-05, + "loss": 0.1195, + "step": 1658 + }, + { + "epoch": 0.22, + "learning_rate": 1.8101472995090017e-05, + "loss": 0.1092, + "step": 1659 + }, + { + "epoch": 0.22, + "learning_rate": 1.811238406983088e-05, + "loss": 0.1125, + "step": 1660 + }, + { + "epoch": 0.22, + "learning_rate": 1.8123295144571742e-05, + "loss": 0.1032, + "step": 1661 + }, + { + "epoch": 0.22, + "learning_rate": 1.8134206219312605e-05, + "loss": 0.0985, + "step": 1662 + }, + { + "epoch": 0.22, + "learning_rate": 1.8145117294053464e-05, + "loss": 0.1129, + "step": 1663 + }, + { + "epoch": 0.22, + "learning_rate": 1.8156028368794327e-05, + "loss": 0.1267, + "step": 1664 + }, + { + "epoch": 0.22, + "learning_rate": 1.816693944353519e-05, + "loss": 0.1189, + "step": 1665 + }, + { + "epoch": 0.22, + "learning_rate": 1.8177850518276053e-05, + "loss": 0.1102, + "step": 1666 + }, + { + "epoch": 0.22, + "learning_rate": 1.8188761593016912e-05, + "loss": 0.1083, + "step": 1667 + }, + { + "epoch": 0.22, + "learning_rate": 1.8199672667757775e-05, + "loss": 0.1244, + "step": 1668 + }, + { + "epoch": 0.22, + "learning_rate": 1.8210583742498638e-05, + "loss": 0.0995, + "step": 1669 + }, + { + "epoch": 0.22, + "learning_rate": 1.82214948172395e-05, + "loss": 0.095, + "step": 1670 + }, + { + "epoch": 0.22, + "learning_rate": 1.8232405891980363e-05, + "loss": 0.1158, + "step": 1671 + }, + { + "epoch": 0.22, + "learning_rate": 1.8243316966721223e-05, + "loss": 0.1232, + "step": 1672 + }, + { + "epoch": 0.22, + "learning_rate": 1.8254228041462085e-05, + "loss": 0.1037, + "step": 1673 + }, + { + "epoch": 0.22, + "learning_rate": 1.8265139116202948e-05, + "loss": 0.1076, + "step": 1674 + }, + { + "epoch": 0.22, + "learning_rate": 1.827605019094381e-05, + "loss": 0.122, + "step": 1675 + }, + { + "epoch": 0.22, + "learning_rate": 1.828696126568467e-05, + "loss": 0.1224, + "step": 1676 + }, + { + "epoch": 0.22, + "learning_rate": 1.8297872340425533e-05, + "loss": 0.1034, + "step": 1677 + }, + { + "epoch": 0.22, + "learning_rate": 1.8308783415166396e-05, + "loss": 0.1199, + "step": 1678 + }, + { + "epoch": 0.22, + "learning_rate": 1.831969448990726e-05, + "loss": 0.1032, + "step": 1679 + }, + { + "epoch": 0.22, + "learning_rate": 1.8330605564648118e-05, + "loss": 0.096, + "step": 1680 + }, + { + "epoch": 0.22, + "learning_rate": 1.834151663938898e-05, + "loss": 0.0953, + "step": 1681 + }, + { + "epoch": 0.22, + "learning_rate": 1.8352427714129844e-05, + "loss": 0.1131, + "step": 1682 + }, + { + "epoch": 0.22, + "learning_rate": 1.8363338788870706e-05, + "loss": 0.1144, + "step": 1683 + }, + { + "epoch": 0.22, + "learning_rate": 1.8374249863611566e-05, + "loss": 0.1186, + "step": 1684 + }, + { + "epoch": 0.22, + "learning_rate": 1.838516093835243e-05, + "loss": 0.1185, + "step": 1685 + }, + { + "epoch": 0.22, + "learning_rate": 1.839607201309329e-05, + "loss": 0.1087, + "step": 1686 + }, + { + "epoch": 0.22, + "learning_rate": 1.8406983087834154e-05, + "loss": 0.1061, + "step": 1687 + }, + { + "epoch": 0.22, + "learning_rate": 1.8417894162575013e-05, + "loss": 0.1245, + "step": 1688 + }, + { + "epoch": 0.22, + "learning_rate": 1.8428805237315876e-05, + "loss": 0.1226, + "step": 1689 + }, + { + "epoch": 0.22, + "learning_rate": 1.843971631205674e-05, + "loss": 0.1225, + "step": 1690 + }, + { + "epoch": 0.22, + "learning_rate": 1.8450627386797602e-05, + "loss": 0.1107, + "step": 1691 + }, + { + "epoch": 0.22, + "learning_rate": 1.8461538461538465e-05, + "loss": 0.1227, + "step": 1692 + }, + { + "epoch": 0.22, + "learning_rate": 1.8472449536279324e-05, + "loss": 0.1237, + "step": 1693 + }, + { + "epoch": 0.22, + "learning_rate": 1.8483360611020187e-05, + "loss": 0.1205, + "step": 1694 + }, + { + "epoch": 0.22, + "learning_rate": 1.849427168576105e-05, + "loss": 0.1044, + "step": 1695 + }, + { + "epoch": 0.22, + "learning_rate": 1.8505182760501912e-05, + "loss": 0.1295, + "step": 1696 + }, + { + "epoch": 0.22, + "learning_rate": 1.8516093835242772e-05, + "loss": 0.126, + "step": 1697 + }, + { + "epoch": 0.22, + "learning_rate": 1.8527004909983635e-05, + "loss": 0.1339, + "step": 1698 + }, + { + "epoch": 0.22, + "learning_rate": 1.8537915984724497e-05, + "loss": 0.0998, + "step": 1699 + }, + { + "epoch": 0.22, + "learning_rate": 1.854882705946536e-05, + "loss": 0.1126, + "step": 1700 + }, + { + "epoch": 0.22, + "learning_rate": 1.855973813420622e-05, + "loss": 0.1203, + "step": 1701 + }, + { + "epoch": 0.22, + "learning_rate": 1.8570649208947082e-05, + "loss": 0.1164, + "step": 1702 + }, + { + "epoch": 0.22, + "learning_rate": 1.8581560283687945e-05, + "loss": 0.1141, + "step": 1703 + }, + { + "epoch": 0.22, + "learning_rate": 1.8592471358428808e-05, + "loss": 0.1069, + "step": 1704 + }, + { + "epoch": 0.22, + "learning_rate": 1.8603382433169667e-05, + "loss": 0.1057, + "step": 1705 + }, + { + "epoch": 0.22, + "learning_rate": 1.861429350791053e-05, + "loss": 0.1117, + "step": 1706 + }, + { + "epoch": 0.22, + "learning_rate": 1.8625204582651393e-05, + "loss": 0.0967, + "step": 1707 + }, + { + "epoch": 0.22, + "learning_rate": 1.8636115657392256e-05, + "loss": 0.1233, + "step": 1708 + }, + { + "epoch": 0.22, + "learning_rate": 1.8647026732133115e-05, + "loss": 0.1126, + "step": 1709 + }, + { + "epoch": 0.22, + "learning_rate": 1.8657937806873978e-05, + "loss": 0.1201, + "step": 1710 + }, + { + "epoch": 0.22, + "learning_rate": 1.866884888161484e-05, + "loss": 0.1034, + "step": 1711 + }, + { + "epoch": 0.22, + "learning_rate": 1.8679759956355703e-05, + "loss": 0.11, + "step": 1712 + }, + { + "epoch": 0.22, + "learning_rate": 1.8690671031096566e-05, + "loss": 0.1228, + "step": 1713 + }, + { + "epoch": 0.22, + "learning_rate": 1.8701582105837425e-05, + "loss": 0.1147, + "step": 1714 + }, + { + "epoch": 0.22, + "learning_rate": 1.8712493180578288e-05, + "loss": 0.1123, + "step": 1715 + }, + { + "epoch": 0.22, + "learning_rate": 1.872340425531915e-05, + "loss": 0.1189, + "step": 1716 + }, + { + "epoch": 0.22, + "learning_rate": 1.8734315330060014e-05, + "loss": 0.1246, + "step": 1717 + }, + { + "epoch": 0.23, + "learning_rate": 1.8745226404800873e-05, + "loss": 0.1136, + "step": 1718 + }, + { + "epoch": 0.23, + "learning_rate": 1.8756137479541736e-05, + "loss": 0.1039, + "step": 1719 + }, + { + "epoch": 0.23, + "learning_rate": 1.87670485542826e-05, + "loss": 0.1132, + "step": 1720 + }, + { + "epoch": 0.23, + "learning_rate": 1.877795962902346e-05, + "loss": 0.1201, + "step": 1721 + }, + { + "epoch": 0.23, + "learning_rate": 1.878887070376432e-05, + "loss": 0.105, + "step": 1722 + }, + { + "epoch": 0.23, + "learning_rate": 1.8799781778505184e-05, + "loss": 0.1097, + "step": 1723 + }, + { + "epoch": 0.23, + "learning_rate": 1.8810692853246047e-05, + "loss": 0.1172, + "step": 1724 + }, + { + "epoch": 0.23, + "learning_rate": 1.882160392798691e-05, + "loss": 0.1059, + "step": 1725 + }, + { + "epoch": 0.23, + "learning_rate": 1.883251500272777e-05, + "loss": 0.1103, + "step": 1726 + }, + { + "epoch": 0.23, + "learning_rate": 1.884342607746863e-05, + "loss": 0.1111, + "step": 1727 + }, + { + "epoch": 0.23, + "learning_rate": 1.8854337152209494e-05, + "loss": 0.1169, + "step": 1728 + }, + { + "epoch": 0.23, + "learning_rate": 1.8865248226950357e-05, + "loss": 0.1138, + "step": 1729 + }, + { + "epoch": 0.23, + "learning_rate": 1.8876159301691216e-05, + "loss": 0.1248, + "step": 1730 + }, + { + "epoch": 0.23, + "learning_rate": 1.888707037643208e-05, + "loss": 0.1223, + "step": 1731 + }, + { + "epoch": 0.23, + "learning_rate": 1.8897981451172942e-05, + "loss": 0.0946, + "step": 1732 + }, + { + "epoch": 0.23, + "learning_rate": 1.8908892525913805e-05, + "loss": 0.1162, + "step": 1733 + }, + { + "epoch": 0.23, + "learning_rate": 1.8919803600654668e-05, + "loss": 0.1078, + "step": 1734 + }, + { + "epoch": 0.23, + "learning_rate": 1.8930714675395527e-05, + "loss": 0.1149, + "step": 1735 + }, + { + "epoch": 0.23, + "learning_rate": 1.894162575013639e-05, + "loss": 0.1241, + "step": 1736 + }, + { + "epoch": 0.23, + "learning_rate": 1.8952536824877253e-05, + "loss": 0.1218, + "step": 1737 + }, + { + "epoch": 0.23, + "learning_rate": 1.8963447899618115e-05, + "loss": 0.1168, + "step": 1738 + }, + { + "epoch": 0.23, + "learning_rate": 1.8974358974358975e-05, + "loss": 0.1148, + "step": 1739 + }, + { + "epoch": 0.23, + "learning_rate": 1.8985270049099837e-05, + "loss": 0.1138, + "step": 1740 + }, + { + "epoch": 0.23, + "learning_rate": 1.89961811238407e-05, + "loss": 0.1108, + "step": 1741 + }, + { + "epoch": 0.23, + "learning_rate": 1.9007092198581563e-05, + "loss": 0.1186, + "step": 1742 + }, + { + "epoch": 0.23, + "learning_rate": 1.9018003273322422e-05, + "loss": 0.1074, + "step": 1743 + }, + { + "epoch": 0.23, + "learning_rate": 1.9028914348063285e-05, + "loss": 0.119, + "step": 1744 + }, + { + "epoch": 0.23, + "learning_rate": 1.9039825422804148e-05, + "loss": 0.1313, + "step": 1745 + }, + { + "epoch": 0.23, + "learning_rate": 1.905073649754501e-05, + "loss": 0.1128, + "step": 1746 + }, + { + "epoch": 0.23, + "learning_rate": 1.906164757228587e-05, + "loss": 0.1196, + "step": 1747 + }, + { + "epoch": 0.23, + "learning_rate": 1.9072558647026733e-05, + "loss": 0.0939, + "step": 1748 + }, + { + "epoch": 0.23, + "learning_rate": 1.9083469721767596e-05, + "loss": 0.1054, + "step": 1749 + }, + { + "epoch": 0.23, + "learning_rate": 1.909438079650846e-05, + "loss": 0.0966, + "step": 1750 + }, + { + "epoch": 0.23, + "learning_rate": 1.910529187124932e-05, + "loss": 0.1207, + "step": 1751 + }, + { + "epoch": 0.23, + "learning_rate": 1.911620294599018e-05, + "loss": 0.1043, + "step": 1752 + }, + { + "epoch": 0.23, + "learning_rate": 1.9127114020731043e-05, + "loss": 0.1168, + "step": 1753 + }, + { + "epoch": 0.23, + "learning_rate": 1.9138025095471906e-05, + "loss": 0.1149, + "step": 1754 + }, + { + "epoch": 0.23, + "learning_rate": 1.914893617021277e-05, + "loss": 0.123, + "step": 1755 + }, + { + "epoch": 0.23, + "learning_rate": 1.915984724495363e-05, + "loss": 0.1101, + "step": 1756 + }, + { + "epoch": 0.23, + "learning_rate": 1.917075831969449e-05, + "loss": 0.1067, + "step": 1757 + }, + { + "epoch": 0.23, + "learning_rate": 1.9181669394435354e-05, + "loss": 0.1075, + "step": 1758 + }, + { + "epoch": 0.23, + "learning_rate": 1.9192580469176217e-05, + "loss": 0.119, + "step": 1759 + }, + { + "epoch": 0.23, + "learning_rate": 1.9203491543917076e-05, + "loss": 0.1054, + "step": 1760 + }, + { + "epoch": 0.23, + "learning_rate": 1.921440261865794e-05, + "loss": 0.1208, + "step": 1761 + }, + { + "epoch": 0.23, + "learning_rate": 1.9225313693398802e-05, + "loss": 0.1155, + "step": 1762 + }, + { + "epoch": 0.23, + "learning_rate": 1.9236224768139665e-05, + "loss": 0.1087, + "step": 1763 + }, + { + "epoch": 0.23, + "learning_rate": 1.9247135842880524e-05, + "loss": 0.1028, + "step": 1764 + }, + { + "epoch": 0.23, + "learning_rate": 1.9258046917621387e-05, + "loss": 0.1155, + "step": 1765 + }, + { + "epoch": 0.23, + "learning_rate": 1.926895799236225e-05, + "loss": 0.1122, + "step": 1766 + }, + { + "epoch": 0.23, + "learning_rate": 1.9279869067103112e-05, + "loss": 0.1163, + "step": 1767 + }, + { + "epoch": 0.23, + "learning_rate": 1.929078014184397e-05, + "loss": 0.1016, + "step": 1768 + }, + { + "epoch": 0.23, + "learning_rate": 1.9301691216584834e-05, + "loss": 0.1096, + "step": 1769 + }, + { + "epoch": 0.23, + "learning_rate": 1.9312602291325697e-05, + "loss": 0.1048, + "step": 1770 + }, + { + "epoch": 0.23, + "learning_rate": 1.932351336606656e-05, + "loss": 0.1141, + "step": 1771 + }, + { + "epoch": 0.23, + "learning_rate": 1.9334424440807423e-05, + "loss": 0.1148, + "step": 1772 + }, + { + "epoch": 0.23, + "learning_rate": 1.9345335515548282e-05, + "loss": 0.1302, + "step": 1773 + }, + { + "epoch": 0.23, + "learning_rate": 1.9356246590289145e-05, + "loss": 0.111, + "step": 1774 + }, + { + "epoch": 0.23, + "learning_rate": 1.9367157665030008e-05, + "loss": 0.1118, + "step": 1775 + }, + { + "epoch": 0.23, + "learning_rate": 1.937806873977087e-05, + "loss": 0.1112, + "step": 1776 + }, + { + "epoch": 0.23, + "learning_rate": 1.938897981451173e-05, + "loss": 0.116, + "step": 1777 + }, + { + "epoch": 0.23, + "learning_rate": 1.9399890889252593e-05, + "loss": 0.1057, + "step": 1778 + }, + { + "epoch": 0.23, + "learning_rate": 1.9410801963993455e-05, + "loss": 0.1263, + "step": 1779 + }, + { + "epoch": 0.23, + "learning_rate": 1.9421713038734318e-05, + "loss": 0.1, + "step": 1780 + }, + { + "epoch": 0.23, + "learning_rate": 1.9432624113475178e-05, + "loss": 0.1067, + "step": 1781 + }, + { + "epoch": 0.23, + "learning_rate": 1.944353518821604e-05, + "loss": 0.1042, + "step": 1782 + }, + { + "epoch": 0.23, + "learning_rate": 1.9454446262956903e-05, + "loss": 0.1198, + "step": 1783 + }, + { + "epoch": 0.23, + "learning_rate": 1.9465357337697766e-05, + "loss": 0.1065, + "step": 1784 + }, + { + "epoch": 0.23, + "learning_rate": 1.9476268412438625e-05, + "loss": 0.1189, + "step": 1785 + }, + { + "epoch": 0.23, + "learning_rate": 1.9487179487179488e-05, + "loss": 0.1107, + "step": 1786 + }, + { + "epoch": 0.23, + "learning_rate": 1.949809056192035e-05, + "loss": 0.1028, + "step": 1787 + }, + { + "epoch": 0.23, + "learning_rate": 1.9509001636661214e-05, + "loss": 0.1023, + "step": 1788 + }, + { + "epoch": 0.23, + "learning_rate": 1.9519912711402073e-05, + "loss": 0.1066, + "step": 1789 + }, + { + "epoch": 0.23, + "learning_rate": 1.9530823786142936e-05, + "loss": 0.1277, + "step": 1790 + }, + { + "epoch": 0.23, + "learning_rate": 1.95417348608838e-05, + "loss": 0.1177, + "step": 1791 + }, + { + "epoch": 0.23, + "learning_rate": 1.955264593562466e-05, + "loss": 0.1083, + "step": 1792 + }, + { + "epoch": 0.23, + "learning_rate": 1.9563557010365524e-05, + "loss": 0.1092, + "step": 1793 + }, + { + "epoch": 0.24, + "learning_rate": 1.9574468085106384e-05, + "loss": 0.1114, + "step": 1794 + }, + { + "epoch": 0.24, + "learning_rate": 1.9585379159847246e-05, + "loss": 0.1236, + "step": 1795 + }, + { + "epoch": 0.24, + "learning_rate": 1.959629023458811e-05, + "loss": 0.1192, + "step": 1796 + }, + { + "epoch": 0.24, + "learning_rate": 1.9607201309328972e-05, + "loss": 0.0983, + "step": 1797 + }, + { + "epoch": 0.24, + "learning_rate": 1.961811238406983e-05, + "loss": 0.1101, + "step": 1798 + }, + { + "epoch": 0.24, + "learning_rate": 1.9629023458810694e-05, + "loss": 0.1174, + "step": 1799 + }, + { + "epoch": 0.24, + "learning_rate": 1.9639934533551557e-05, + "loss": 0.1119, + "step": 1800 + }, + { + "epoch": 0.24, + "learning_rate": 1.965084560829242e-05, + "loss": 0.1063, + "step": 1801 + }, + { + "epoch": 0.24, + "learning_rate": 1.966175668303328e-05, + "loss": 0.1047, + "step": 1802 + }, + { + "epoch": 0.24, + "learning_rate": 1.9672667757774142e-05, + "loss": 0.1213, + "step": 1803 + }, + { + "epoch": 0.24, + "learning_rate": 1.9683578832515005e-05, + "loss": 0.1142, + "step": 1804 + }, + { + "epoch": 0.24, + "learning_rate": 1.9694489907255867e-05, + "loss": 0.1189, + "step": 1805 + }, + { + "epoch": 0.24, + "learning_rate": 1.9705400981996727e-05, + "loss": 0.1174, + "step": 1806 + }, + { + "epoch": 0.24, + "learning_rate": 1.971631205673759e-05, + "loss": 0.1094, + "step": 1807 + }, + { + "epoch": 0.24, + "learning_rate": 1.9727223131478452e-05, + "loss": 0.1215, + "step": 1808 + }, + { + "epoch": 0.24, + "learning_rate": 1.9738134206219315e-05, + "loss": 0.1165, + "step": 1809 + }, + { + "epoch": 0.24, + "learning_rate": 1.9749045280960175e-05, + "loss": 0.1239, + "step": 1810 + }, + { + "epoch": 0.24, + "learning_rate": 1.9759956355701037e-05, + "loss": 0.1007, + "step": 1811 + }, + { + "epoch": 0.24, + "learning_rate": 1.97708674304419e-05, + "loss": 0.1088, + "step": 1812 + }, + { + "epoch": 0.24, + "learning_rate": 1.9781778505182763e-05, + "loss": 0.1025, + "step": 1813 + }, + { + "epoch": 0.24, + "learning_rate": 1.9792689579923626e-05, + "loss": 0.1116, + "step": 1814 + }, + { + "epoch": 0.24, + "learning_rate": 1.9803600654664485e-05, + "loss": 0.1212, + "step": 1815 + }, + { + "epoch": 0.24, + "learning_rate": 1.9814511729405348e-05, + "loss": 0.0998, + "step": 1816 + }, + { + "epoch": 0.24, + "learning_rate": 1.982542280414621e-05, + "loss": 0.103, + "step": 1817 + }, + { + "epoch": 0.24, + "learning_rate": 1.9836333878887073e-05, + "loss": 0.0896, + "step": 1818 + }, + { + "epoch": 0.24, + "learning_rate": 1.9847244953627933e-05, + "loss": 0.1211, + "step": 1819 + }, + { + "epoch": 0.24, + "learning_rate": 1.9858156028368796e-05, + "loss": 0.1241, + "step": 1820 + }, + { + "epoch": 0.24, + "learning_rate": 1.986906710310966e-05, + "loss": 0.1175, + "step": 1821 + }, + { + "epoch": 0.24, + "learning_rate": 1.987997817785052e-05, + "loss": 0.1203, + "step": 1822 + }, + { + "epoch": 0.24, + "learning_rate": 1.989088925259138e-05, + "loss": 0.1072, + "step": 1823 + }, + { + "epoch": 0.24, + "learning_rate": 1.9901800327332243e-05, + "loss": 0.1126, + "step": 1824 + }, + { + "epoch": 0.24, + "learning_rate": 1.9912711402073106e-05, + "loss": 0.1189, + "step": 1825 + }, + { + "epoch": 0.24, + "learning_rate": 1.992362247681397e-05, + "loss": 0.11, + "step": 1826 + }, + { + "epoch": 0.24, + "learning_rate": 1.993453355155483e-05, + "loss": 0.1091, + "step": 1827 + }, + { + "epoch": 0.24, + "learning_rate": 1.994544462629569e-05, + "loss": 0.1026, + "step": 1828 + }, + { + "epoch": 0.24, + "learning_rate": 1.9956355701036554e-05, + "loss": 0.109, + "step": 1829 + }, + { + "epoch": 0.24, + "learning_rate": 1.9967266775777417e-05, + "loss": 0.1122, + "step": 1830 + }, + { + "epoch": 0.24, + "learning_rate": 1.9978177850518276e-05, + "loss": 0.1017, + "step": 1831 + }, + { + "epoch": 0.24, + "learning_rate": 1.998908892525914e-05, + "loss": 0.1037, + "step": 1832 + }, + { + "epoch": 0.24, + "learning_rate": 2e-05, + "loss": 0.1038, + "step": 1833 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999999985937767e-05, + "loss": 0.1124, + "step": 1834 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999999943751068e-05, + "loss": 0.1143, + "step": 1835 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999999873439896e-05, + "loss": 0.1139, + "step": 1836 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999999775004263e-05, + "loss": 0.0926, + "step": 1837 + }, + { + "epoch": 0.24, + "learning_rate": 1.999999964844416e-05, + "loss": 0.1065, + "step": 1838 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999999493759593e-05, + "loss": 0.1153, + "step": 1839 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999999310950558e-05, + "loss": 0.1232, + "step": 1840 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999999100017056e-05, + "loss": 0.1052, + "step": 1841 + }, + { + "epoch": 0.24, + "learning_rate": 1.999999886095909e-05, + "loss": 0.1118, + "step": 1842 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999998593776658e-05, + "loss": 0.1258, + "step": 1843 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999998298469768e-05, + "loss": 0.117, + "step": 1844 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999997975038413e-05, + "loss": 0.1088, + "step": 1845 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999997623482596e-05, + "loss": 0.1089, + "step": 1846 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999997243802317e-05, + "loss": 0.102, + "step": 1847 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999996835997576e-05, + "loss": 0.1129, + "step": 1848 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999996400068383e-05, + "loss": 0.115, + "step": 1849 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999995936014725e-05, + "loss": 0.115, + "step": 1850 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999995443836616e-05, + "loss": 0.105, + "step": 1851 + }, + { + "epoch": 0.24, + "learning_rate": 1.999999492353405e-05, + "loss": 0.1043, + "step": 1852 + }, + { + "epoch": 0.24, + "learning_rate": 1.999999437510703e-05, + "loss": 0.0938, + "step": 1853 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999993798555564e-05, + "loss": 0.1228, + "step": 1854 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999993193879644e-05, + "loss": 0.1101, + "step": 1855 + }, + { + "epoch": 0.24, + "learning_rate": 1.999999256107928e-05, + "loss": 0.0932, + "step": 1856 + }, + { + "epoch": 0.24, + "learning_rate": 1.999999190015446e-05, + "loss": 0.1078, + "step": 1857 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999991211105202e-05, + "loss": 0.1195, + "step": 1858 + }, + { + "epoch": 0.24, + "learning_rate": 1.99999904939315e-05, + "loss": 0.1188, + "step": 1859 + }, + { + "epoch": 0.24, + "learning_rate": 1.999998974863336e-05, + "loss": 0.1094, + "step": 1860 + }, + { + "epoch": 0.24, + "learning_rate": 1.999998897521078e-05, + "loss": 0.1127, + "step": 1861 + }, + { + "epoch": 0.24, + "learning_rate": 1.999998817366376e-05, + "loss": 0.099, + "step": 1862 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999987343992307e-05, + "loss": 0.0953, + "step": 1863 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999986486196424e-05, + "loss": 0.1251, + "step": 1864 + }, + { + "epoch": 0.24, + "learning_rate": 1.999998560027611e-05, + "loss": 0.1117, + "step": 1865 + }, + { + "epoch": 0.24, + "learning_rate": 1.999998468623137e-05, + "loss": 0.1063, + "step": 1866 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999983744062206e-05, + "loss": 0.1086, + "step": 1867 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999982773768618e-05, + "loss": 0.1115, + "step": 1868 + }, + { + "epoch": 0.24, + "learning_rate": 1.9999981775350612e-05, + "loss": 0.1161, + "step": 1869 + }, + { + "epoch": 0.24, + "learning_rate": 1.999998074880819e-05, + "loss": 0.1123, + "step": 1870 + }, + { + "epoch": 0.25, + "learning_rate": 1.999997969414136e-05, + "loss": 0.1189, + "step": 1871 + }, + { + "epoch": 0.25, + "learning_rate": 1.999997861135011e-05, + "loss": 0.1124, + "step": 1872 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999977500434457e-05, + "loss": 0.1104, + "step": 1873 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999976361394402e-05, + "loss": 0.1238, + "step": 1874 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999975194229944e-05, + "loss": 0.1107, + "step": 1875 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999973998941087e-05, + "loss": 0.1113, + "step": 1876 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999972775527837e-05, + "loss": 0.1234, + "step": 1877 + }, + { + "epoch": 0.25, + "learning_rate": 1.99999715239902e-05, + "loss": 0.1181, + "step": 1878 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999970244328168e-05, + "loss": 0.102, + "step": 1879 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999968936541755e-05, + "loss": 0.1162, + "step": 1880 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999967600630966e-05, + "loss": 0.1159, + "step": 1881 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999966236595796e-05, + "loss": 0.1063, + "step": 1882 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999964844436256e-05, + "loss": 0.1147, + "step": 1883 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999963424152345e-05, + "loss": 0.1168, + "step": 1884 + }, + { + "epoch": 0.25, + "learning_rate": 1.999996197574407e-05, + "loss": 0.1075, + "step": 1885 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999960499211436e-05, + "loss": 0.1136, + "step": 1886 + }, + { + "epoch": 0.25, + "learning_rate": 1.999995899455445e-05, + "loss": 0.1078, + "step": 1887 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999957461773104e-05, + "loss": 0.1163, + "step": 1888 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999955900867412e-05, + "loss": 0.1317, + "step": 1889 + }, + { + "epoch": 0.25, + "learning_rate": 1.999995431183738e-05, + "loss": 0.1101, + "step": 1890 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999952694683006e-05, + "loss": 0.1118, + "step": 1891 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999951049404295e-05, + "loss": 0.1078, + "step": 1892 + }, + { + "epoch": 0.25, + "learning_rate": 1.999994937600126e-05, + "loss": 0.114, + "step": 1893 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999947674473898e-05, + "loss": 0.1056, + "step": 1894 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999945944822214e-05, + "loss": 0.098, + "step": 1895 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999944187046215e-05, + "loss": 0.111, + "step": 1896 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999942401145906e-05, + "loss": 0.1083, + "step": 1897 + }, + { + "epoch": 0.25, + "learning_rate": 1.999994058712129e-05, + "loss": 0.1202, + "step": 1898 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999938744972377e-05, + "loss": 0.1103, + "step": 1899 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999936874699168e-05, + "loss": 0.1018, + "step": 1900 + }, + { + "epoch": 0.25, + "learning_rate": 1.999993497630167e-05, + "loss": 0.1183, + "step": 1901 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999933049779884e-05, + "loss": 0.0892, + "step": 1902 + }, + { + "epoch": 0.25, + "learning_rate": 1.999993109513382e-05, + "loss": 0.1216, + "step": 1903 + }, + { + "epoch": 0.25, + "learning_rate": 1.999992911236348e-05, + "loss": 0.0949, + "step": 1904 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999927101468874e-05, + "loss": 0.1265, + "step": 1905 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999925062450008e-05, + "loss": 0.1174, + "step": 1906 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999922995306883e-05, + "loss": 0.1187, + "step": 1907 + }, + { + "epoch": 0.25, + "learning_rate": 1.999992090003951e-05, + "loss": 0.1062, + "step": 1908 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999918776647888e-05, + "loss": 0.1007, + "step": 1909 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999916625132028e-05, + "loss": 0.1103, + "step": 1910 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999914445491933e-05, + "loss": 0.1052, + "step": 1911 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999912237727614e-05, + "loss": 0.1064, + "step": 1912 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999910001839077e-05, + "loss": 0.1073, + "step": 1913 + }, + { + "epoch": 0.25, + "learning_rate": 1.999990773782632e-05, + "loss": 0.1059, + "step": 1914 + }, + { + "epoch": 0.25, + "learning_rate": 1.999990544568936e-05, + "loss": 0.1142, + "step": 1915 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999903125428195e-05, + "loss": 0.1244, + "step": 1916 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999900777042836e-05, + "loss": 0.1174, + "step": 1917 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999898400533288e-05, + "loss": 0.1265, + "step": 1918 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999895995899558e-05, + "loss": 0.1152, + "step": 1919 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999893563141658e-05, + "loss": 0.1165, + "step": 1920 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999891102259584e-05, + "loss": 0.1021, + "step": 1921 + }, + { + "epoch": 0.25, + "learning_rate": 1.999988861325335e-05, + "loss": 0.0919, + "step": 1922 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999886096122962e-05, + "loss": 0.1088, + "step": 1923 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999883550868428e-05, + "loss": 0.1093, + "step": 1924 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999880977489753e-05, + "loss": 0.1161, + "step": 1925 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999878375986946e-05, + "loss": 0.1157, + "step": 1926 + }, + { + "epoch": 0.25, + "learning_rate": 1.999987574636001e-05, + "loss": 0.106, + "step": 1927 + }, + { + "epoch": 0.25, + "learning_rate": 1.999987308860896e-05, + "loss": 0.1088, + "step": 1928 + }, + { + "epoch": 0.25, + "learning_rate": 1.99998704027338e-05, + "loss": 0.1063, + "step": 1929 + }, + { + "epoch": 0.25, + "learning_rate": 1.999986768873454e-05, + "loss": 0.1102, + "step": 1930 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999864946611177e-05, + "loss": 0.1082, + "step": 1931 + }, + { + "epoch": 0.25, + "learning_rate": 1.999986217636373e-05, + "loss": 0.1091, + "step": 1932 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999859377992203e-05, + "loss": 0.1249, + "step": 1933 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999856551496604e-05, + "loss": 0.1029, + "step": 1934 + }, + { + "epoch": 0.25, + "learning_rate": 1.999985369687694e-05, + "loss": 0.1143, + "step": 1935 + }, + { + "epoch": 0.25, + "learning_rate": 1.999985081413322e-05, + "loss": 0.1103, + "step": 1936 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999847903265452e-05, + "loss": 0.1179, + "step": 1937 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999844964273645e-05, + "loss": 0.0951, + "step": 1938 + }, + { + "epoch": 0.25, + "learning_rate": 1.999984199715781e-05, + "loss": 0.1184, + "step": 1939 + }, + { + "epoch": 0.25, + "learning_rate": 1.9999839001917948e-05, + "loss": 0.1062, + "step": 1940 + }, + { + "epoch": 0.25, + "learning_rate": 1.999983597855407e-05, + "loss": 0.1204, + "step": 1941 + }, + { + "epoch": 0.25, + "learning_rate": 1.999983292706619e-05, + "loss": 0.0962, + "step": 1942 + }, + { + "epoch": 0.25, + "learning_rate": 1.999982984745431e-05, + "loss": 0.1158, + "step": 1943 + }, + { + "epoch": 0.25, + "learning_rate": 1.999982673971844e-05, + "loss": 0.1338, + "step": 1944 + }, + { + "epoch": 0.25, + "learning_rate": 1.999982360385859e-05, + "loss": 0.1098, + "step": 1945 + }, + { + "epoch": 0.25, + "learning_rate": 1.999982043987477e-05, + "loss": 0.1045, + "step": 1946 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999817247766987e-05, + "loss": 0.1089, + "step": 1947 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999814027535255e-05, + "loss": 0.098, + "step": 1948 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999810779179572e-05, + "loss": 0.1114, + "step": 1949 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999807502699957e-05, + "loss": 0.1174, + "step": 1950 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999804198096414e-05, + "loss": 0.1176, + "step": 1951 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999800865368955e-05, + "loss": 0.0947, + "step": 1952 + }, + { + "epoch": 0.26, + "learning_rate": 1.999979750451759e-05, + "loss": 0.1153, + "step": 1953 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999794115542326e-05, + "loss": 0.1164, + "step": 1954 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999790698443172e-05, + "loss": 0.1178, + "step": 1955 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999787253220143e-05, + "loss": 0.1202, + "step": 1956 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999783779873244e-05, + "loss": 0.1084, + "step": 1957 + }, + { + "epoch": 0.26, + "learning_rate": 1.999978027840248e-05, + "loss": 0.1167, + "step": 1958 + }, + { + "epoch": 0.26, + "learning_rate": 1.999977674880787e-05, + "loss": 0.0964, + "step": 1959 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999773191089424e-05, + "loss": 0.1106, + "step": 1960 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999769605247145e-05, + "loss": 0.1118, + "step": 1961 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999765991281048e-05, + "loss": 0.1023, + "step": 1962 + }, + { + "epoch": 0.26, + "learning_rate": 1.999976234919114e-05, + "loss": 0.1157, + "step": 1963 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999758678977433e-05, + "loss": 0.1012, + "step": 1964 + }, + { + "epoch": 0.26, + "learning_rate": 1.999975498063994e-05, + "loss": 0.0993, + "step": 1965 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999751254178664e-05, + "loss": 0.1224, + "step": 1966 + }, + { + "epoch": 0.26, + "learning_rate": 1.999974749959362e-05, + "loss": 0.109, + "step": 1967 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999743716884822e-05, + "loss": 0.1136, + "step": 1968 + }, + { + "epoch": 0.26, + "learning_rate": 1.999973990605228e-05, + "loss": 0.1079, + "step": 1969 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999736067096e-05, + "loss": 0.1042, + "step": 1970 + }, + { + "epoch": 0.26, + "learning_rate": 1.999973220001599e-05, + "loss": 0.1256, + "step": 1971 + }, + { + "epoch": 0.26, + "learning_rate": 1.999972830481227e-05, + "loss": 0.1113, + "step": 1972 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999724381484845e-05, + "loss": 0.1269, + "step": 1973 + }, + { + "epoch": 0.26, + "learning_rate": 1.999972043003373e-05, + "loss": 0.1051, + "step": 1974 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999716450458933e-05, + "loss": 0.1031, + "step": 1975 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999712442760463e-05, + "loss": 0.1154, + "step": 1976 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999708406938338e-05, + "loss": 0.1143, + "step": 1977 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999704342992567e-05, + "loss": 0.1224, + "step": 1978 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999700250923154e-05, + "loss": 0.1046, + "step": 1979 + }, + { + "epoch": 0.26, + "learning_rate": 1.999969613073012e-05, + "loss": 0.1096, + "step": 1980 + }, + { + "epoch": 0.26, + "learning_rate": 1.999969198241347e-05, + "loss": 0.1163, + "step": 1981 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999687805973224e-05, + "loss": 0.1069, + "step": 1982 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999683601409386e-05, + "loss": 0.1072, + "step": 1983 + }, + { + "epoch": 0.26, + "learning_rate": 1.999967936872197e-05, + "loss": 0.1082, + "step": 1984 + }, + { + "epoch": 0.26, + "learning_rate": 1.999967510791099e-05, + "loss": 0.108, + "step": 1985 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999670818976457e-05, + "loss": 0.1058, + "step": 1986 + }, + { + "epoch": 0.26, + "learning_rate": 1.999966650191838e-05, + "loss": 0.1191, + "step": 1987 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999662156736776e-05, + "loss": 0.1123, + "step": 1988 + }, + { + "epoch": 0.26, + "learning_rate": 1.999965778343165e-05, + "loss": 0.1149, + "step": 1989 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999653382003022e-05, + "loss": 0.1214, + "step": 1990 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999648952450902e-05, + "loss": 0.1066, + "step": 1991 + }, + { + "epoch": 0.26, + "learning_rate": 1.99996444947753e-05, + "loss": 0.0903, + "step": 1992 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999640008976232e-05, + "loss": 0.1049, + "step": 1993 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999635495053708e-05, + "loss": 0.1017, + "step": 1994 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999630953007743e-05, + "loss": 0.1046, + "step": 1995 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999626382838345e-05, + "loss": 0.1144, + "step": 1996 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999621784545533e-05, + "loss": 0.1045, + "step": 1997 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999617158129315e-05, + "loss": 0.1007, + "step": 1998 + }, + { + "epoch": 0.26, + "learning_rate": 1.999961250358971e-05, + "loss": 0.1109, + "step": 1999 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999607820926728e-05, + "loss": 0.1058, + "step": 2000 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999603110140377e-05, + "loss": 0.118, + "step": 2001 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999598371230677e-05, + "loss": 0.117, + "step": 2002 + }, + { + "epoch": 0.26, + "learning_rate": 1.999959360419764e-05, + "loss": 0.1164, + "step": 2003 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999588809041278e-05, + "loss": 0.1081, + "step": 2004 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999583985761603e-05, + "loss": 0.1069, + "step": 2005 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999579134358633e-05, + "loss": 0.1166, + "step": 2006 + }, + { + "epoch": 0.26, + "learning_rate": 1.999957425483238e-05, + "loss": 0.103, + "step": 2007 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999569347182853e-05, + "loss": 0.1093, + "step": 2008 + }, + { + "epoch": 0.26, + "learning_rate": 1.999956441141007e-05, + "loss": 0.1092, + "step": 2009 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999559447514047e-05, + "loss": 0.1119, + "step": 2010 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999554455494795e-05, + "loss": 0.1183, + "step": 2011 + }, + { + "epoch": 0.26, + "learning_rate": 1.999954943535233e-05, + "loss": 0.1162, + "step": 2012 + }, + { + "epoch": 0.26, + "learning_rate": 1.999954438708666e-05, + "loss": 0.1066, + "step": 2013 + }, + { + "epoch": 0.26, + "learning_rate": 1.999953931069781e-05, + "loss": 0.0995, + "step": 2014 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999534206185783e-05, + "loss": 0.1179, + "step": 2015 + }, + { + "epoch": 0.26, + "learning_rate": 1.99995290735506e-05, + "loss": 0.1046, + "step": 2016 + }, + { + "epoch": 0.26, + "learning_rate": 1.999952391279228e-05, + "loss": 0.0987, + "step": 2017 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999518723910823e-05, + "loss": 0.1031, + "step": 2018 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999513506906257e-05, + "loss": 0.1102, + "step": 2019 + }, + { + "epoch": 0.26, + "learning_rate": 1.999950826177859e-05, + "loss": 0.1108, + "step": 2020 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999502988527842e-05, + "loss": 0.1039, + "step": 2021 + }, + { + "epoch": 0.26, + "learning_rate": 1.999949768715402e-05, + "loss": 0.1079, + "step": 2022 + }, + { + "epoch": 0.26, + "learning_rate": 1.9999492357657143e-05, + "loss": 0.1101, + "step": 2023 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999487000037227e-05, + "loss": 0.1082, + "step": 2024 + }, + { + "epoch": 0.27, + "learning_rate": 1.999948161429429e-05, + "loss": 0.1187, + "step": 2025 + }, + { + "epoch": 0.27, + "learning_rate": 1.999947620042834e-05, + "loss": 0.1095, + "step": 2026 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999470758439395e-05, + "loss": 0.1094, + "step": 2027 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999465288327472e-05, + "loss": 0.1133, + "step": 2028 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999459790092586e-05, + "loss": 0.1129, + "step": 2029 + }, + { + "epoch": 0.27, + "learning_rate": 1.999945426373475e-05, + "loss": 0.1112, + "step": 2030 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999448709253984e-05, + "loss": 0.1095, + "step": 2031 + }, + { + "epoch": 0.27, + "learning_rate": 1.99994431266503e-05, + "loss": 0.1119, + "step": 2032 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999437515923714e-05, + "loss": 0.116, + "step": 2033 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999431877074243e-05, + "loss": 0.1003, + "step": 2034 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999426210101903e-05, + "loss": 0.1125, + "step": 2035 + }, + { + "epoch": 0.27, + "learning_rate": 1.999942051500671e-05, + "loss": 0.1172, + "step": 2036 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999414791788678e-05, + "loss": 0.1015, + "step": 2037 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999409040447824e-05, + "loss": 0.1128, + "step": 2038 + }, + { + "epoch": 0.27, + "learning_rate": 1.999940326098416e-05, + "loss": 0.0914, + "step": 2039 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999397453397714e-05, + "loss": 0.109, + "step": 2040 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999391617688496e-05, + "loss": 0.1198, + "step": 2041 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999385753856514e-05, + "loss": 0.1185, + "step": 2042 + }, + { + "epoch": 0.27, + "learning_rate": 1.99993798619018e-05, + "loss": 0.114, + "step": 2043 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999373941824357e-05, + "loss": 0.109, + "step": 2044 + }, + { + "epoch": 0.27, + "learning_rate": 1.999936799362421e-05, + "loss": 0.1093, + "step": 2045 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999362017301373e-05, + "loss": 0.1123, + "step": 2046 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999356012855863e-05, + "loss": 0.1097, + "step": 2047 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999349980287697e-05, + "loss": 0.1009, + "step": 2048 + }, + { + "epoch": 0.27, + "learning_rate": 1.999934391959689e-05, + "loss": 0.1028, + "step": 2049 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999337830783462e-05, + "loss": 0.1262, + "step": 2050 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999331713847426e-05, + "loss": 0.1214, + "step": 2051 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999325568788802e-05, + "loss": 0.1087, + "step": 2052 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999319395607614e-05, + "loss": 0.112, + "step": 2053 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999313194303865e-05, + "loss": 0.1073, + "step": 2054 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999306964877585e-05, + "loss": 0.0932, + "step": 2055 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999300707328785e-05, + "loss": 0.1067, + "step": 2056 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999294421657484e-05, + "loss": 0.0964, + "step": 2057 + }, + { + "epoch": 0.27, + "learning_rate": 1.99992881078637e-05, + "loss": 0.1092, + "step": 2058 + }, + { + "epoch": 0.27, + "learning_rate": 1.999928176594745e-05, + "loss": 0.1039, + "step": 2059 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999275395908754e-05, + "loss": 0.1122, + "step": 2060 + }, + { + "epoch": 0.27, + "learning_rate": 1.999926899774763e-05, + "loss": 0.1106, + "step": 2061 + }, + { + "epoch": 0.27, + "learning_rate": 1.999926257146409e-05, + "loss": 0.1102, + "step": 2062 + }, + { + "epoch": 0.27, + "learning_rate": 1.999925611705816e-05, + "loss": 0.098, + "step": 2063 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999249634529854e-05, + "loss": 0.1147, + "step": 2064 + }, + { + "epoch": 0.27, + "learning_rate": 1.999924312387919e-05, + "loss": 0.1107, + "step": 2065 + }, + { + "epoch": 0.27, + "learning_rate": 1.999923658510619e-05, + "loss": 0.1126, + "step": 2066 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999230018210865e-05, + "loss": 0.1052, + "step": 2067 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999223423193242e-05, + "loss": 0.0981, + "step": 2068 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999216800053332e-05, + "loss": 0.1026, + "step": 2069 + }, + { + "epoch": 0.27, + "learning_rate": 1.999921014879116e-05, + "loss": 0.1188, + "step": 2070 + }, + { + "epoch": 0.27, + "learning_rate": 1.999920346940674e-05, + "loss": 0.1074, + "step": 2071 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999196761900094e-05, + "loss": 0.0985, + "step": 2072 + }, + { + "epoch": 0.27, + "learning_rate": 1.999919002627124e-05, + "loss": 0.1101, + "step": 2073 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999183262520196e-05, + "loss": 0.1154, + "step": 2074 + }, + { + "epoch": 0.27, + "learning_rate": 1.999917647064698e-05, + "loss": 0.1147, + "step": 2075 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999169650651616e-05, + "loss": 0.1046, + "step": 2076 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999162802534118e-05, + "loss": 0.1028, + "step": 2077 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999155926294506e-05, + "loss": 0.0964, + "step": 2078 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999149021932804e-05, + "loss": 0.1167, + "step": 2079 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999142089449025e-05, + "loss": 0.0986, + "step": 2080 + }, + { + "epoch": 0.27, + "learning_rate": 1.999913512884319e-05, + "loss": 0.1031, + "step": 2081 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999128140115322e-05, + "loss": 0.1131, + "step": 2082 + }, + { + "epoch": 0.27, + "learning_rate": 1.999912112326544e-05, + "loss": 0.1086, + "step": 2083 + }, + { + "epoch": 0.27, + "learning_rate": 1.999911407829356e-05, + "loss": 0.102, + "step": 2084 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999107005199704e-05, + "loss": 0.1144, + "step": 2085 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999099903983896e-05, + "loss": 0.0972, + "step": 2086 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999092774646146e-05, + "loss": 0.098, + "step": 2087 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999085617186483e-05, + "loss": 0.1068, + "step": 2088 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999078431604926e-05, + "loss": 0.105, + "step": 2089 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999071217901493e-05, + "loss": 0.1161, + "step": 2090 + }, + { + "epoch": 0.27, + "learning_rate": 1.99990639760762e-05, + "loss": 0.1246, + "step": 2091 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999056706129077e-05, + "loss": 0.1141, + "step": 2092 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999049408060135e-05, + "loss": 0.1119, + "step": 2093 + }, + { + "epoch": 0.27, + "learning_rate": 1.99990420818694e-05, + "loss": 0.109, + "step": 2094 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999034727556896e-05, + "loss": 0.0978, + "step": 2095 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999027345122636e-05, + "loss": 0.1071, + "step": 2096 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999019934566645e-05, + "loss": 0.1087, + "step": 2097 + }, + { + "epoch": 0.27, + "learning_rate": 1.9999012495888943e-05, + "loss": 0.1077, + "step": 2098 + }, + { + "epoch": 0.27, + "learning_rate": 1.999900502908955e-05, + "loss": 0.1003, + "step": 2099 + }, + { + "epoch": 0.28, + "learning_rate": 1.999899753416849e-05, + "loss": 0.0954, + "step": 2100 + }, + { + "epoch": 0.28, + "learning_rate": 1.999899001112578e-05, + "loss": 0.116, + "step": 2101 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998982459961442e-05, + "loss": 0.1057, + "step": 2102 + }, + { + "epoch": 0.28, + "learning_rate": 1.99989748806755e-05, + "loss": 0.1173, + "step": 2103 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998967273267973e-05, + "loss": 0.1203, + "step": 2104 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998959637738882e-05, + "loss": 0.1091, + "step": 2105 + }, + { + "epoch": 0.28, + "learning_rate": 1.999895197408825e-05, + "loss": 0.1076, + "step": 2106 + }, + { + "epoch": 0.28, + "learning_rate": 1.99989442823161e-05, + "loss": 0.1053, + "step": 2107 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998936562422455e-05, + "loss": 0.1185, + "step": 2108 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998928814407324e-05, + "loss": 0.1055, + "step": 2109 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998921038270744e-05, + "loss": 0.1129, + "step": 2110 + }, + { + "epoch": 0.28, + "learning_rate": 1.999891323401273e-05, + "loss": 0.0989, + "step": 2111 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998905401633306e-05, + "loss": 0.1034, + "step": 2112 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998897541132495e-05, + "loss": 0.1119, + "step": 2113 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998889652510312e-05, + "loss": 0.1041, + "step": 2114 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998881735766788e-05, + "loss": 0.1028, + "step": 2115 + }, + { + "epoch": 0.28, + "learning_rate": 1.999887379090194e-05, + "loss": 0.1129, + "step": 2116 + }, + { + "epoch": 0.28, + "learning_rate": 1.999886581791579e-05, + "loss": 0.1171, + "step": 2117 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998857816808367e-05, + "loss": 0.108, + "step": 2118 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998849787579684e-05, + "loss": 0.0932, + "step": 2119 + }, + { + "epoch": 0.28, + "learning_rate": 1.999884173022977e-05, + "loss": 0.1085, + "step": 2120 + }, + { + "epoch": 0.28, + "learning_rate": 1.999883364475865e-05, + "loss": 0.1111, + "step": 2121 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998825531166335e-05, + "loss": 0.1124, + "step": 2122 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998817389452863e-05, + "loss": 0.1083, + "step": 2123 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998809219618245e-05, + "loss": 0.1107, + "step": 2124 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998801021662514e-05, + "loss": 0.0998, + "step": 2125 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998792795585684e-05, + "loss": 0.1109, + "step": 2126 + }, + { + "epoch": 0.28, + "learning_rate": 1.999878454138778e-05, + "loss": 0.1167, + "step": 2127 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998776259068832e-05, + "loss": 0.1136, + "step": 2128 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998767948628853e-05, + "loss": 0.1013, + "step": 2129 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998759610067874e-05, + "loss": 0.1139, + "step": 2130 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998751243385913e-05, + "loss": 0.0983, + "step": 2131 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998742848583e-05, + "loss": 0.0902, + "step": 2132 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998734425659157e-05, + "loss": 0.1139, + "step": 2133 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998725974614402e-05, + "loss": 0.1193, + "step": 2134 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998717495448763e-05, + "loss": 0.1, + "step": 2135 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998708988162264e-05, + "loss": 0.0995, + "step": 2136 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998700452754928e-05, + "loss": 0.0835, + "step": 2137 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998691889226783e-05, + "loss": 0.1108, + "step": 2138 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998683297577845e-05, + "loss": 0.107, + "step": 2139 + }, + { + "epoch": 0.28, + "learning_rate": 1.999867467780815e-05, + "loss": 0.1088, + "step": 2140 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998666029917708e-05, + "loss": 0.1132, + "step": 2141 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998657353906548e-05, + "loss": 0.1055, + "step": 2142 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998648649774705e-05, + "loss": 0.1073, + "step": 2143 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998639917522187e-05, + "loss": 0.1056, + "step": 2144 + }, + { + "epoch": 0.28, + "learning_rate": 1.999863115714903e-05, + "loss": 0.108, + "step": 2145 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998622368655256e-05, + "loss": 0.1051, + "step": 2146 + }, + { + "epoch": 0.28, + "learning_rate": 1.999861355204089e-05, + "loss": 0.1236, + "step": 2147 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998604707305955e-05, + "loss": 0.1025, + "step": 2148 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998595834450474e-05, + "loss": 0.1055, + "step": 2149 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998586933474476e-05, + "loss": 0.1159, + "step": 2150 + }, + { + "epoch": 0.28, + "learning_rate": 1.999857800437799e-05, + "loss": 0.111, + "step": 2151 + }, + { + "epoch": 0.28, + "learning_rate": 1.999856904716103e-05, + "loss": 0.1061, + "step": 2152 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998560061823626e-05, + "loss": 0.1018, + "step": 2153 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998551048365806e-05, + "loss": 0.1045, + "step": 2154 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998542006787597e-05, + "loss": 0.1069, + "step": 2155 + }, + { + "epoch": 0.28, + "learning_rate": 1.999853293708902e-05, + "loss": 0.1048, + "step": 2156 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998523839270095e-05, + "loss": 0.1095, + "step": 2157 + }, + { + "epoch": 0.28, + "learning_rate": 1.999851471333086e-05, + "loss": 0.0996, + "step": 2158 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998505559271334e-05, + "loss": 0.1081, + "step": 2159 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998496377091544e-05, + "loss": 0.1111, + "step": 2160 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998487166791517e-05, + "loss": 0.1111, + "step": 2161 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998477928371274e-05, + "loss": 0.1079, + "step": 2162 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998468661830844e-05, + "loss": 0.1012, + "step": 2163 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998459367170256e-05, + "loss": 0.1108, + "step": 2164 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998450044389532e-05, + "loss": 0.1157, + "step": 2165 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998440693488697e-05, + "loss": 0.1089, + "step": 2166 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998431314467788e-05, + "loss": 0.1145, + "step": 2167 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998421907326815e-05, + "loss": 0.1228, + "step": 2168 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998412472065818e-05, + "loss": 0.0944, + "step": 2169 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998403008684818e-05, + "loss": 0.1291, + "step": 2170 + }, + { + "epoch": 0.28, + "learning_rate": 1.999839351718384e-05, + "loss": 0.0996, + "step": 2171 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998383997562914e-05, + "loss": 0.124, + "step": 2172 + }, + { + "epoch": 0.28, + "learning_rate": 1.999837444982206e-05, + "loss": 0.0987, + "step": 2173 + }, + { + "epoch": 0.28, + "learning_rate": 1.9998364873961317e-05, + "loss": 0.0996, + "step": 2174 + }, + { + "epoch": 0.28, + "learning_rate": 1.99983552699807e-05, + "loss": 0.103, + "step": 2175 + }, + { + "epoch": 0.29, + "learning_rate": 1.999834563788025e-05, + "loss": 0.0981, + "step": 2176 + }, + { + "epoch": 0.29, + "learning_rate": 1.999833597765998e-05, + "loss": 0.1134, + "step": 2177 + }, + { + "epoch": 0.29, + "learning_rate": 1.999832628931992e-05, + "loss": 0.0955, + "step": 2178 + }, + { + "epoch": 0.29, + "learning_rate": 1.99983165728601e-05, + "loss": 0.1142, + "step": 2179 + }, + { + "epoch": 0.29, + "learning_rate": 1.999830682828055e-05, + "loss": 0.1049, + "step": 2180 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998297055581295e-05, + "loss": 0.1066, + "step": 2181 + }, + { + "epoch": 0.29, + "learning_rate": 1.999828725476236e-05, + "loss": 0.1221, + "step": 2182 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998277425823776e-05, + "loss": 0.109, + "step": 2183 + }, + { + "epoch": 0.29, + "learning_rate": 1.999826756876557e-05, + "loss": 0.1072, + "step": 2184 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998257683587766e-05, + "loss": 0.1012, + "step": 2185 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998247770290396e-05, + "loss": 0.1125, + "step": 2186 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998237828873485e-05, + "loss": 0.1076, + "step": 2187 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998227859337065e-05, + "loss": 0.1154, + "step": 2188 + }, + { + "epoch": 0.29, + "learning_rate": 1.999821786168116e-05, + "loss": 0.1057, + "step": 2189 + }, + { + "epoch": 0.29, + "learning_rate": 1.99982078359058e-05, + "loss": 0.1074, + "step": 2190 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998197782011015e-05, + "loss": 0.1051, + "step": 2191 + }, + { + "epoch": 0.29, + "learning_rate": 1.999818769999683e-05, + "loss": 0.1057, + "step": 2192 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998177589863273e-05, + "loss": 0.1078, + "step": 2193 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998167451610375e-05, + "loss": 0.1036, + "step": 2194 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998157285238162e-05, + "loss": 0.1129, + "step": 2195 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998147090746664e-05, + "loss": 0.1069, + "step": 2196 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998136868135912e-05, + "loss": 0.1145, + "step": 2197 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998126617405932e-05, + "loss": 0.1003, + "step": 2198 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998116338556753e-05, + "loss": 0.1104, + "step": 2199 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998106031588404e-05, + "loss": 0.1015, + "step": 2200 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998095696500915e-05, + "loss": 0.1018, + "step": 2201 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998085333294312e-05, + "loss": 0.1117, + "step": 2202 + }, + { + "epoch": 0.29, + "learning_rate": 1.999807494196863e-05, + "loss": 0.102, + "step": 2203 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998064522523894e-05, + "loss": 0.0979, + "step": 2204 + }, + { + "epoch": 0.29, + "learning_rate": 1.999805407496013e-05, + "loss": 0.1093, + "step": 2205 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998043599277377e-05, + "loss": 0.1073, + "step": 2206 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998033095475654e-05, + "loss": 0.107, + "step": 2207 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998022563555e-05, + "loss": 0.1186, + "step": 2208 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998012003515434e-05, + "loss": 0.0967, + "step": 2209 + }, + { + "epoch": 0.29, + "learning_rate": 1.9998001415356995e-05, + "loss": 0.106, + "step": 2210 + }, + { + "epoch": 0.29, + "learning_rate": 1.999799079907971e-05, + "loss": 0.0978, + "step": 2211 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997980154683608e-05, + "loss": 0.1029, + "step": 2212 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997969482168717e-05, + "loss": 0.0954, + "step": 2213 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997958781535073e-05, + "loss": 0.1154, + "step": 2214 + }, + { + "epoch": 0.29, + "learning_rate": 1.99979480527827e-05, + "loss": 0.1064, + "step": 2215 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997937295911633e-05, + "loss": 0.0967, + "step": 2216 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997926510921893e-05, + "loss": 0.1103, + "step": 2217 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997915697813526e-05, + "loss": 0.1087, + "step": 2218 + }, + { + "epoch": 0.29, + "learning_rate": 1.999790485658655e-05, + "loss": 0.1124, + "step": 2219 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997893987240994e-05, + "loss": 0.1191, + "step": 2220 + }, + { + "epoch": 0.29, + "learning_rate": 1.99978830897769e-05, + "loss": 0.097, + "step": 2221 + }, + { + "epoch": 0.29, + "learning_rate": 1.999787216419429e-05, + "loss": 0.0983, + "step": 2222 + }, + { + "epoch": 0.29, + "learning_rate": 1.99978612104932e-05, + "loss": 0.1044, + "step": 2223 + }, + { + "epoch": 0.29, + "learning_rate": 1.999785022867365e-05, + "loss": 0.1162, + "step": 2224 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997839218735688e-05, + "loss": 0.1088, + "step": 2225 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997828180679328e-05, + "loss": 0.1025, + "step": 2226 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997817114504614e-05, + "loss": 0.1198, + "step": 2227 + }, + { + "epoch": 0.29, + "learning_rate": 1.999780602021157e-05, + "loss": 0.1086, + "step": 2228 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997794897800227e-05, + "loss": 0.1029, + "step": 2229 + }, + { + "epoch": 0.29, + "learning_rate": 1.999778374727062e-05, + "loss": 0.1092, + "step": 2230 + }, + { + "epoch": 0.29, + "learning_rate": 1.999777256862278e-05, + "loss": 0.1194, + "step": 2231 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997761361856734e-05, + "loss": 0.1029, + "step": 2232 + }, + { + "epoch": 0.29, + "learning_rate": 1.999775012697252e-05, + "loss": 0.1117, + "step": 2233 + }, + { + "epoch": 0.29, + "learning_rate": 1.999773886397016e-05, + "loss": 0.1072, + "step": 2234 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997727572849697e-05, + "loss": 0.1077, + "step": 2235 + }, + { + "epoch": 0.29, + "learning_rate": 1.999771625361116e-05, + "loss": 0.1068, + "step": 2236 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997704906254575e-05, + "loss": 0.1079, + "step": 2237 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997693530779977e-05, + "loss": 0.1133, + "step": 2238 + }, + { + "epoch": 0.29, + "learning_rate": 1.99976821271874e-05, + "loss": 0.1107, + "step": 2239 + }, + { + "epoch": 0.29, + "learning_rate": 1.999767069547687e-05, + "loss": 0.1106, + "step": 2240 + }, + { + "epoch": 0.29, + "learning_rate": 1.999765923564843e-05, + "loss": 0.1216, + "step": 2241 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997647747702103e-05, + "loss": 0.0957, + "step": 2242 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997636231637922e-05, + "loss": 0.0955, + "step": 2243 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997624687455923e-05, + "loss": 0.099, + "step": 2244 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997613115156137e-05, + "loss": 0.1162, + "step": 2245 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997601514738597e-05, + "loss": 0.1216, + "step": 2246 + }, + { + "epoch": 0.29, + "learning_rate": 1.999758988620334e-05, + "loss": 0.1033, + "step": 2247 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997578229550384e-05, + "loss": 0.1117, + "step": 2248 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997566544779777e-05, + "loss": 0.0979, + "step": 2249 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997554831891546e-05, + "loss": 0.109, + "step": 2250 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997543090885724e-05, + "loss": 0.1112, + "step": 2251 + }, + { + "epoch": 0.29, + "learning_rate": 1.9997531321762346e-05, + "loss": 0.1131, + "step": 2252 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997519524521444e-05, + "loss": 0.1145, + "step": 2253 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997507699163047e-05, + "loss": 0.0987, + "step": 2254 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997495845687194e-05, + "loss": 0.1028, + "step": 2255 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997483964093917e-05, + "loss": 0.1108, + "step": 2256 + }, + { + "epoch": 0.3, + "learning_rate": 1.999747205438325e-05, + "loss": 0.1, + "step": 2257 + }, + { + "epoch": 0.3, + "learning_rate": 1.999746011655522e-05, + "loss": 0.1059, + "step": 2258 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997448150609872e-05, + "loss": 0.112, + "step": 2259 + }, + { + "epoch": 0.3, + "learning_rate": 1.999743615654723e-05, + "loss": 0.1139, + "step": 2260 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997424134367332e-05, + "loss": 0.1017, + "step": 2261 + }, + { + "epoch": 0.3, + "learning_rate": 1.999741208407021e-05, + "loss": 0.1, + "step": 2262 + }, + { + "epoch": 0.3, + "learning_rate": 1.99974000056559e-05, + "loss": 0.1089, + "step": 2263 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997387899124434e-05, + "loss": 0.103, + "step": 2264 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997375764475852e-05, + "loss": 0.088, + "step": 2265 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997363601710174e-05, + "loss": 0.0999, + "step": 2266 + }, + { + "epoch": 0.3, + "learning_rate": 1.999735141082745e-05, + "loss": 0.103, + "step": 2267 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997339191827707e-05, + "loss": 0.1117, + "step": 2268 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997326944710978e-05, + "loss": 0.097, + "step": 2269 + }, + { + "epoch": 0.3, + "learning_rate": 1.99973146694773e-05, + "loss": 0.1179, + "step": 2270 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997302366126706e-05, + "loss": 0.1069, + "step": 2271 + }, + { + "epoch": 0.3, + "learning_rate": 1.999729003465923e-05, + "loss": 0.1081, + "step": 2272 + }, + { + "epoch": 0.3, + "learning_rate": 1.999727767507491e-05, + "loss": 0.1062, + "step": 2273 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997265287373783e-05, + "loss": 0.1005, + "step": 2274 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997252871555874e-05, + "loss": 0.0985, + "step": 2275 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997240427621224e-05, + "loss": 0.1107, + "step": 2276 + }, + { + "epoch": 0.3, + "learning_rate": 1.999722795556987e-05, + "loss": 0.0996, + "step": 2277 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997215455401846e-05, + "loss": 0.0885, + "step": 2278 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997202927117186e-05, + "loss": 0.1078, + "step": 2279 + }, + { + "epoch": 0.3, + "learning_rate": 1.999719037071592e-05, + "loss": 0.1133, + "step": 2280 + }, + { + "epoch": 0.3, + "learning_rate": 1.999717778619809e-05, + "loss": 0.0977, + "step": 2281 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997165173563732e-05, + "loss": 0.1043, + "step": 2282 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997152532812883e-05, + "loss": 0.119, + "step": 2283 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997139863945568e-05, + "loss": 0.0996, + "step": 2284 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997127166961835e-05, + "loss": 0.0978, + "step": 2285 + }, + { + "epoch": 0.3, + "learning_rate": 1.999711444186171e-05, + "loss": 0.1047, + "step": 2286 + }, + { + "epoch": 0.3, + "learning_rate": 1.999710168864523e-05, + "loss": 0.1014, + "step": 2287 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997088907312443e-05, + "loss": 0.1183, + "step": 2288 + }, + { + "epoch": 0.3, + "learning_rate": 1.999707609786337e-05, + "loss": 0.1137, + "step": 2289 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997063260298054e-05, + "loss": 0.1015, + "step": 2290 + }, + { + "epoch": 0.3, + "learning_rate": 1.999705039461653e-05, + "loss": 0.1199, + "step": 2291 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997037500818838e-05, + "loss": 0.1125, + "step": 2292 + }, + { + "epoch": 0.3, + "learning_rate": 1.9997024578905004e-05, + "loss": 0.1155, + "step": 2293 + }, + { + "epoch": 0.3, + "learning_rate": 1.999701162887507e-05, + "loss": 0.0943, + "step": 2294 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996998650729077e-05, + "loss": 0.1056, + "step": 2295 + }, + { + "epoch": 0.3, + "learning_rate": 1.999698564446706e-05, + "loss": 0.1091, + "step": 2296 + }, + { + "epoch": 0.3, + "learning_rate": 1.999697261008905e-05, + "loss": 0.1071, + "step": 2297 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996959547595087e-05, + "loss": 0.1064, + "step": 2298 + }, + { + "epoch": 0.3, + "learning_rate": 1.999694645698521e-05, + "loss": 0.1131, + "step": 2299 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996933338259448e-05, + "loss": 0.0981, + "step": 2300 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996920191417848e-05, + "loss": 0.0955, + "step": 2301 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996907016460443e-05, + "loss": 0.109, + "step": 2302 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996893813387266e-05, + "loss": 0.0936, + "step": 2303 + }, + { + "epoch": 0.3, + "learning_rate": 1.999688058219836e-05, + "loss": 0.1076, + "step": 2304 + }, + { + "epoch": 0.3, + "learning_rate": 1.999686732289376e-05, + "loss": 0.0954, + "step": 2305 + }, + { + "epoch": 0.3, + "learning_rate": 1.99968540354735e-05, + "loss": 0.1068, + "step": 2306 + }, + { + "epoch": 0.3, + "learning_rate": 1.999684071993762e-05, + "loss": 0.1167, + "step": 2307 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996827376286162e-05, + "loss": 0.1049, + "step": 2308 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996814004519158e-05, + "loss": 0.1022, + "step": 2309 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996800604636646e-05, + "loss": 0.1067, + "step": 2310 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996787176638664e-05, + "loss": 0.1086, + "step": 2311 + }, + { + "epoch": 0.3, + "learning_rate": 1.999677372052525e-05, + "loss": 0.1059, + "step": 2312 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996760236296448e-05, + "loss": 0.0987, + "step": 2313 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996746723952284e-05, + "loss": 0.1143, + "step": 2314 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996733183492803e-05, + "loss": 0.1017, + "step": 2315 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996719614918042e-05, + "loss": 0.099, + "step": 2316 + }, + { + "epoch": 0.3, + "learning_rate": 1.999670601822804e-05, + "loss": 0.0983, + "step": 2317 + }, + { + "epoch": 0.3, + "learning_rate": 1.999669239342284e-05, + "loss": 0.1102, + "step": 2318 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996678740502465e-05, + "loss": 0.1024, + "step": 2319 + }, + { + "epoch": 0.3, + "learning_rate": 1.999666505946697e-05, + "loss": 0.1109, + "step": 2320 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996651350316388e-05, + "loss": 0.1135, + "step": 2321 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996637613050753e-05, + "loss": 0.1107, + "step": 2322 + }, + { + "epoch": 0.3, + "learning_rate": 1.999662384767011e-05, + "loss": 0.0999, + "step": 2323 + }, + { + "epoch": 0.3, + "learning_rate": 1.999661005417449e-05, + "loss": 0.1065, + "step": 2324 + }, + { + "epoch": 0.3, + "learning_rate": 1.999659623256394e-05, + "loss": 0.0973, + "step": 2325 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996582382838497e-05, + "loss": 0.1048, + "step": 2326 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996568504998194e-05, + "loss": 0.1014, + "step": 2327 + }, + { + "epoch": 0.3, + "learning_rate": 1.9996554599043078e-05, + "loss": 0.0993, + "step": 2328 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996540664973187e-05, + "loss": 0.11, + "step": 2329 + }, + { + "epoch": 0.31, + "learning_rate": 1.999652670278855e-05, + "loss": 0.1162, + "step": 2330 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996512712489218e-05, + "loss": 0.0998, + "step": 2331 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996498694075228e-05, + "loss": 0.114, + "step": 2332 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996484647546616e-05, + "loss": 0.1057, + "step": 2333 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996470572903424e-05, + "loss": 0.1, + "step": 2334 + }, + { + "epoch": 0.31, + "learning_rate": 1.999645647014569e-05, + "loss": 0.1152, + "step": 2335 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996442339273455e-05, + "loss": 0.1143, + "step": 2336 + }, + { + "epoch": 0.31, + "learning_rate": 1.999642818028676e-05, + "loss": 0.0901, + "step": 2337 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996413993185636e-05, + "loss": 0.1109, + "step": 2338 + }, + { + "epoch": 0.31, + "learning_rate": 1.999639977797014e-05, + "loss": 0.1026, + "step": 2339 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996385534640294e-05, + "loss": 0.1109, + "step": 2340 + }, + { + "epoch": 0.31, + "learning_rate": 1.999637126319615e-05, + "loss": 0.121, + "step": 2341 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996356963637742e-05, + "loss": 0.1102, + "step": 2342 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996342635965112e-05, + "loss": 0.0927, + "step": 2343 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996328280178306e-05, + "loss": 0.1189, + "step": 2344 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996313896277355e-05, + "loss": 0.1143, + "step": 2345 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996299484262303e-05, + "loss": 0.1026, + "step": 2346 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996285044133194e-05, + "loss": 0.1051, + "step": 2347 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996270575890062e-05, + "loss": 0.1159, + "step": 2348 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996256079532955e-05, + "loss": 0.1102, + "step": 2349 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996241555061906e-05, + "loss": 0.1084, + "step": 2350 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996227002476962e-05, + "loss": 0.1131, + "step": 2351 + }, + { + "epoch": 0.31, + "learning_rate": 1.999621242177816e-05, + "loss": 0.1041, + "step": 2352 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996197812965548e-05, + "loss": 0.0909, + "step": 2353 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996183176039156e-05, + "loss": 0.1076, + "step": 2354 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996168510999035e-05, + "loss": 0.1082, + "step": 2355 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996153817845223e-05, + "loss": 0.1064, + "step": 2356 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996139096577758e-05, + "loss": 0.1099, + "step": 2357 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996124347196683e-05, + "loss": 0.1067, + "step": 2358 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996109569702046e-05, + "loss": 0.1067, + "step": 2359 + }, + { + "epoch": 0.31, + "learning_rate": 1.999609476409388e-05, + "loss": 0.1024, + "step": 2360 + }, + { + "epoch": 0.31, + "learning_rate": 1.999607993037223e-05, + "loss": 0.1097, + "step": 2361 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996065068537136e-05, + "loss": 0.1101, + "step": 2362 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996050178588643e-05, + "loss": 0.109, + "step": 2363 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996035260526788e-05, + "loss": 0.104, + "step": 2364 + }, + { + "epoch": 0.31, + "learning_rate": 1.999602031435162e-05, + "loss": 0.104, + "step": 2365 + }, + { + "epoch": 0.31, + "learning_rate": 1.9996005340063176e-05, + "loss": 0.0971, + "step": 2366 + }, + { + "epoch": 0.31, + "learning_rate": 1.99959903376615e-05, + "loss": 0.1076, + "step": 2367 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995975307146633e-05, + "loss": 0.1111, + "step": 2368 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995960248518617e-05, + "loss": 0.1023, + "step": 2369 + }, + { + "epoch": 0.31, + "learning_rate": 1.99959451617775e-05, + "loss": 0.1056, + "step": 2370 + }, + { + "epoch": 0.31, + "learning_rate": 1.999593004692331e-05, + "loss": 0.0991, + "step": 2371 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995914903956107e-05, + "loss": 0.0961, + "step": 2372 + }, + { + "epoch": 0.31, + "learning_rate": 1.999589973287592e-05, + "loss": 0.0972, + "step": 2373 + }, + { + "epoch": 0.31, + "learning_rate": 1.99958845336828e-05, + "loss": 0.1079, + "step": 2374 + }, + { + "epoch": 0.31, + "learning_rate": 1.999586930637679e-05, + "loss": 0.1049, + "step": 2375 + }, + { + "epoch": 0.31, + "learning_rate": 1.999585405095793e-05, + "loss": 0.1139, + "step": 2376 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995838767426256e-05, + "loss": 0.1175, + "step": 2377 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995823455781824e-05, + "loss": 0.1075, + "step": 2378 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995808116024667e-05, + "loss": 0.1097, + "step": 2379 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995792748154835e-05, + "loss": 0.0966, + "step": 2380 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995777352172365e-05, + "loss": 0.1098, + "step": 2381 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995761928077306e-05, + "loss": 0.115, + "step": 2382 + }, + { + "epoch": 0.31, + "learning_rate": 1.99957464758697e-05, + "loss": 0.0995, + "step": 2383 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995730995549584e-05, + "loss": 0.1008, + "step": 2384 + }, + { + "epoch": 0.31, + "learning_rate": 1.999571548711701e-05, + "loss": 0.1137, + "step": 2385 + }, + { + "epoch": 0.31, + "learning_rate": 1.999569995057202e-05, + "loss": 0.1057, + "step": 2386 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995684385914657e-05, + "loss": 0.1025, + "step": 2387 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995668793144963e-05, + "loss": 0.0978, + "step": 2388 + }, + { + "epoch": 0.31, + "learning_rate": 1.999565317226298e-05, + "loss": 0.1025, + "step": 2389 + }, + { + "epoch": 0.31, + "learning_rate": 1.999563752326876e-05, + "loss": 0.1057, + "step": 2390 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995621846162336e-05, + "loss": 0.1103, + "step": 2391 + }, + { + "epoch": 0.31, + "learning_rate": 1.999560614094376e-05, + "loss": 0.1051, + "step": 2392 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995590407613076e-05, + "loss": 0.1043, + "step": 2393 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995574646170323e-05, + "loss": 0.1092, + "step": 2394 + }, + { + "epoch": 0.31, + "learning_rate": 1.999555885661555e-05, + "loss": 0.1062, + "step": 2395 + }, + { + "epoch": 0.31, + "learning_rate": 1.99955430389488e-05, + "loss": 0.0893, + "step": 2396 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995527193170118e-05, + "loss": 0.1057, + "step": 2397 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995511319279547e-05, + "loss": 0.1187, + "step": 2398 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995495417277136e-05, + "loss": 0.0952, + "step": 2399 + }, + { + "epoch": 0.31, + "learning_rate": 1.999547948716292e-05, + "loss": 0.1106, + "step": 2400 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995463528936958e-05, + "loss": 0.1042, + "step": 2401 + }, + { + "epoch": 0.31, + "learning_rate": 1.999544754259928e-05, + "loss": 0.11, + "step": 2402 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995431528149942e-05, + "loss": 0.0998, + "step": 2403 + }, + { + "epoch": 0.31, + "learning_rate": 1.9995415485588985e-05, + "loss": 0.1051, + "step": 2404 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995399414916456e-05, + "loss": 0.0944, + "step": 2405 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995383316132396e-05, + "loss": 0.0984, + "step": 2406 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995367189236852e-05, + "loss": 0.0959, + "step": 2407 + }, + { + "epoch": 0.32, + "learning_rate": 1.999535103422987e-05, + "loss": 0.109, + "step": 2408 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995334851111497e-05, + "loss": 0.1037, + "step": 2409 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995318639881774e-05, + "loss": 0.1094, + "step": 2410 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995302400540756e-05, + "loss": 0.1, + "step": 2411 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995286133088476e-05, + "loss": 0.1104, + "step": 2412 + }, + { + "epoch": 0.32, + "learning_rate": 1.999526983752499e-05, + "loss": 0.1067, + "step": 2413 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995253513850338e-05, + "loss": 0.1124, + "step": 2414 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995237162064568e-05, + "loss": 0.1194, + "step": 2415 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995220782167726e-05, + "loss": 0.0981, + "step": 2416 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995204374159856e-05, + "loss": 0.1129, + "step": 2417 + }, + { + "epoch": 0.32, + "learning_rate": 1.999518793804101e-05, + "loss": 0.0986, + "step": 2418 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995171473811225e-05, + "loss": 0.0965, + "step": 2419 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995154981470556e-05, + "loss": 0.1055, + "step": 2420 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995138461019045e-05, + "loss": 0.1064, + "step": 2421 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995121912456737e-05, + "loss": 0.0877, + "step": 2422 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995105335783686e-05, + "loss": 0.0943, + "step": 2423 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995088730999928e-05, + "loss": 0.1003, + "step": 2424 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995072098105522e-05, + "loss": 0.1108, + "step": 2425 + }, + { + "epoch": 0.32, + "learning_rate": 1.99950554371005e-05, + "loss": 0.096, + "step": 2426 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995038747984924e-05, + "loss": 0.1009, + "step": 2427 + }, + { + "epoch": 0.32, + "learning_rate": 1.999502203075883e-05, + "loss": 0.0982, + "step": 2428 + }, + { + "epoch": 0.32, + "learning_rate": 1.9995005285422268e-05, + "loss": 0.105, + "step": 2429 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994988511975287e-05, + "loss": 0.1177, + "step": 2430 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994971710417935e-05, + "loss": 0.1162, + "step": 2431 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994954880750254e-05, + "loss": 0.1082, + "step": 2432 + }, + { + "epoch": 0.32, + "learning_rate": 1.99949380229723e-05, + "loss": 0.1051, + "step": 2433 + }, + { + "epoch": 0.32, + "learning_rate": 1.999492113708411e-05, + "loss": 0.1051, + "step": 2434 + }, + { + "epoch": 0.32, + "learning_rate": 1.999490422308574e-05, + "loss": 0.0917, + "step": 2435 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994887280977232e-05, + "loss": 0.1047, + "step": 2436 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994870310758633e-05, + "loss": 0.1056, + "step": 2437 + }, + { + "epoch": 0.32, + "learning_rate": 1.999485331243e-05, + "loss": 0.1024, + "step": 2438 + }, + { + "epoch": 0.32, + "learning_rate": 1.999483628599137e-05, + "loss": 0.1034, + "step": 2439 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994819231442794e-05, + "loss": 0.101, + "step": 2440 + }, + { + "epoch": 0.32, + "learning_rate": 1.999480214878432e-05, + "loss": 0.0887, + "step": 2441 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994785038016004e-05, + "loss": 0.1071, + "step": 2442 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994767899137884e-05, + "loss": 0.1166, + "step": 2443 + }, + { + "epoch": 0.32, + "learning_rate": 1.999475073215001e-05, + "loss": 0.0956, + "step": 2444 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994733537052435e-05, + "loss": 0.0989, + "step": 2445 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994716313845203e-05, + "loss": 0.1074, + "step": 2446 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994699062528363e-05, + "loss": 0.1013, + "step": 2447 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994681783101965e-05, + "loss": 0.1152, + "step": 2448 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994664475566054e-05, + "loss": 0.1009, + "step": 2449 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994647139920683e-05, + "loss": 0.0925, + "step": 2450 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994629776165898e-05, + "loss": 0.104, + "step": 2451 + }, + { + "epoch": 0.32, + "learning_rate": 1.999461238430175e-05, + "loss": 0.1141, + "step": 2452 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994594964328288e-05, + "loss": 0.1027, + "step": 2453 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994577516245562e-05, + "loss": 0.0986, + "step": 2454 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994560040053614e-05, + "loss": 0.1027, + "step": 2455 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994542535752498e-05, + "loss": 0.1028, + "step": 2456 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994525003342268e-05, + "loss": 0.0955, + "step": 2457 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994507442822966e-05, + "loss": 0.0955, + "step": 2458 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994489854194644e-05, + "loss": 0.1144, + "step": 2459 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994472237457352e-05, + "loss": 0.1018, + "step": 2460 + }, + { + "epoch": 0.32, + "learning_rate": 1.999445459261114e-05, + "loss": 0.0957, + "step": 2461 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994436919656056e-05, + "loss": 0.104, + "step": 2462 + }, + { + "epoch": 0.32, + "learning_rate": 1.999441921859215e-05, + "loss": 0.1125, + "step": 2463 + }, + { + "epoch": 0.32, + "learning_rate": 1.999440148941947e-05, + "loss": 0.097, + "step": 2464 + }, + { + "epoch": 0.32, + "learning_rate": 1.999438373213807e-05, + "loss": 0.1077, + "step": 2465 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994365946748e-05, + "loss": 0.101, + "step": 2466 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994348133249305e-05, + "loss": 0.1024, + "step": 2467 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994330291642037e-05, + "loss": 0.097, + "step": 2468 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994312421926253e-05, + "loss": 0.1079, + "step": 2469 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994294524101995e-05, + "loss": 0.1036, + "step": 2470 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994276598169314e-05, + "loss": 0.1205, + "step": 2471 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994258644128262e-05, + "loss": 0.1087, + "step": 2472 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994240661978893e-05, + "loss": 0.106, + "step": 2473 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994222651721254e-05, + "loss": 0.1118, + "step": 2474 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994204613355392e-05, + "loss": 0.1093, + "step": 2475 + }, + { + "epoch": 0.32, + "learning_rate": 1.999418654688137e-05, + "loss": 0.1288, + "step": 2476 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994168452299217e-05, + "loss": 0.1057, + "step": 2477 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994150329609006e-05, + "loss": 0.1086, + "step": 2478 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994132178810783e-05, + "loss": 0.1032, + "step": 2479 + }, + { + "epoch": 0.32, + "learning_rate": 1.999411399990459e-05, + "loss": 0.1048, + "step": 2480 + }, + { + "epoch": 0.32, + "learning_rate": 1.9994095792890485e-05, + "loss": 0.1032, + "step": 2481 + }, + { + "epoch": 0.33, + "learning_rate": 1.9994077557768515e-05, + "loss": 0.1065, + "step": 2482 + }, + { + "epoch": 0.33, + "learning_rate": 1.999405929453874e-05, + "loss": 0.0998, + "step": 2483 + }, + { + "epoch": 0.33, + "learning_rate": 1.9994041003201203e-05, + "loss": 0.1037, + "step": 2484 + }, + { + "epoch": 0.33, + "learning_rate": 1.9994022683755955e-05, + "loss": 0.1136, + "step": 2485 + }, + { + "epoch": 0.33, + "learning_rate": 1.9994004336203053e-05, + "loss": 0.1046, + "step": 2486 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993985960542543e-05, + "loss": 0.1021, + "step": 2487 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993967556774484e-05, + "loss": 0.1073, + "step": 2488 + }, + { + "epoch": 0.33, + "learning_rate": 1.999394912489892e-05, + "loss": 0.0983, + "step": 2489 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993930664915907e-05, + "loss": 0.0965, + "step": 2490 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993912176825498e-05, + "loss": 0.1146, + "step": 2491 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993893660627744e-05, + "loss": 0.0996, + "step": 2492 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993875116322693e-05, + "loss": 0.1028, + "step": 2493 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993856543910404e-05, + "loss": 0.0853, + "step": 2494 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993837943390923e-05, + "loss": 0.1099, + "step": 2495 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993819314764304e-05, + "loss": 0.1044, + "step": 2496 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993800658030605e-05, + "loss": 0.0916, + "step": 2497 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993781973189868e-05, + "loss": 0.1093, + "step": 2498 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993763260242156e-05, + "loss": 0.0931, + "step": 2499 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993744519187515e-05, + "loss": 0.1028, + "step": 2500 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993725750026e-05, + "loss": 0.1006, + "step": 2501 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993706952757663e-05, + "loss": 0.0973, + "step": 2502 + }, + { + "epoch": 0.33, + "learning_rate": 1.999368812738256e-05, + "loss": 0.1078, + "step": 2503 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993669273900738e-05, + "loss": 0.1173, + "step": 2504 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993650392312255e-05, + "loss": 0.1116, + "step": 2505 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993631482617163e-05, + "loss": 0.0954, + "step": 2506 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993612544815512e-05, + "loss": 0.1076, + "step": 2507 + }, + { + "epoch": 0.33, + "learning_rate": 1.999359357890736e-05, + "loss": 0.0996, + "step": 2508 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993574584892757e-05, + "loss": 0.1081, + "step": 2509 + }, + { + "epoch": 0.33, + "learning_rate": 1.999355556277176e-05, + "loss": 0.1038, + "step": 2510 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993536512544416e-05, + "loss": 0.11, + "step": 2511 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993517434210787e-05, + "loss": 0.1099, + "step": 2512 + }, + { + "epoch": 0.33, + "learning_rate": 1.999349832777092e-05, + "loss": 0.1033, + "step": 2513 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993479193224875e-05, + "loss": 0.0905, + "step": 2514 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993460030572697e-05, + "loss": 0.1038, + "step": 2515 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993440839814448e-05, + "loss": 0.1033, + "step": 2516 + }, + { + "epoch": 0.33, + "learning_rate": 1.999342162095018e-05, + "loss": 0.099, + "step": 2517 + }, + { + "epoch": 0.33, + "learning_rate": 1.999340237397994e-05, + "loss": 0.0916, + "step": 2518 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993383098903793e-05, + "loss": 0.0954, + "step": 2519 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993363795721788e-05, + "loss": 0.0968, + "step": 2520 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993344464433976e-05, + "loss": 0.1041, + "step": 2521 + }, + { + "epoch": 0.33, + "learning_rate": 1.999332510504042e-05, + "loss": 0.1165, + "step": 2522 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993305717541163e-05, + "loss": 0.1068, + "step": 2523 + }, + { + "epoch": 0.33, + "learning_rate": 1.999328630193627e-05, + "loss": 0.1101, + "step": 2524 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993266858225794e-05, + "loss": 0.1042, + "step": 2525 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993247386409782e-05, + "loss": 0.1107, + "step": 2526 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993227886488295e-05, + "loss": 0.1016, + "step": 2527 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993208358461388e-05, + "loss": 0.1108, + "step": 2528 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993188802329114e-05, + "loss": 0.1044, + "step": 2529 + }, + { + "epoch": 0.33, + "learning_rate": 1.999316921809153e-05, + "loss": 0.0958, + "step": 2530 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993149605748688e-05, + "loss": 0.1028, + "step": 2531 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993129965300647e-05, + "loss": 0.1093, + "step": 2532 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993110296747457e-05, + "loss": 0.0974, + "step": 2533 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993090600089178e-05, + "loss": 0.0998, + "step": 2534 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993070875325865e-05, + "loss": 0.0945, + "step": 2535 + }, + { + "epoch": 0.33, + "learning_rate": 1.999305112245757e-05, + "loss": 0.0892, + "step": 2536 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993031341484353e-05, + "loss": 0.1064, + "step": 2537 + }, + { + "epoch": 0.33, + "learning_rate": 1.9993011532406266e-05, + "loss": 0.093, + "step": 2538 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992991695223368e-05, + "loss": 0.0989, + "step": 2539 + }, + { + "epoch": 0.33, + "learning_rate": 1.999297182993571e-05, + "loss": 0.0987, + "step": 2540 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992951936543353e-05, + "loss": 0.1003, + "step": 2541 + }, + { + "epoch": 0.33, + "learning_rate": 1.999293201504635e-05, + "loss": 0.0976, + "step": 2542 + }, + { + "epoch": 0.33, + "learning_rate": 1.999291206544476e-05, + "loss": 0.1141, + "step": 2543 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992892087738635e-05, + "loss": 0.1025, + "step": 2544 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992872081928035e-05, + "loss": 0.0983, + "step": 2545 + }, + { + "epoch": 0.33, + "learning_rate": 1.999285204801301e-05, + "loss": 0.0957, + "step": 2546 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992831985993626e-05, + "loss": 0.0975, + "step": 2547 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992811895869932e-05, + "loss": 0.1165, + "step": 2548 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992791777641987e-05, + "loss": 0.0982, + "step": 2549 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992771631309848e-05, + "loss": 0.1088, + "step": 2550 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992751456873568e-05, + "loss": 0.1035, + "step": 2551 + }, + { + "epoch": 0.33, + "learning_rate": 1.999273125433321e-05, + "loss": 0.1158, + "step": 2552 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992711023688824e-05, + "loss": 0.1088, + "step": 2553 + }, + { + "epoch": 0.33, + "learning_rate": 1.999269076494047e-05, + "loss": 0.1013, + "step": 2554 + }, + { + "epoch": 0.33, + "learning_rate": 1.999267047808821e-05, + "loss": 0.1069, + "step": 2555 + }, + { + "epoch": 0.33, + "learning_rate": 1.9992650163132094e-05, + "loss": 0.1125, + "step": 2556 + }, + { + "epoch": 0.33, + "learning_rate": 1.999262982007218e-05, + "loss": 0.0933, + "step": 2557 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992609448908533e-05, + "loss": 0.1064, + "step": 2558 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992589049641198e-05, + "loss": 0.1077, + "step": 2559 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992568622270242e-05, + "loss": 0.1002, + "step": 2560 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992548166795715e-05, + "loss": 0.1018, + "step": 2561 + }, + { + "epoch": 0.34, + "learning_rate": 1.999252768321768e-05, + "loss": 0.1014, + "step": 2562 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992507171536194e-05, + "loss": 0.1022, + "step": 2563 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992486631751315e-05, + "loss": 0.0958, + "step": 2564 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992466063863095e-05, + "loss": 0.0922, + "step": 2565 + }, + { + "epoch": 0.34, + "learning_rate": 1.99924454678716e-05, + "loss": 0.0941, + "step": 2566 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992424843776882e-05, + "loss": 0.1035, + "step": 2567 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992404191579004e-05, + "loss": 0.1011, + "step": 2568 + }, + { + "epoch": 0.34, + "learning_rate": 1.999238351127802e-05, + "loss": 0.1068, + "step": 2569 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992362802873987e-05, + "loss": 0.1088, + "step": 2570 + }, + { + "epoch": 0.34, + "learning_rate": 1.999234206636697e-05, + "loss": 0.1086, + "step": 2571 + }, + { + "epoch": 0.34, + "learning_rate": 1.999232130175702e-05, + "loss": 0.0975, + "step": 2572 + }, + { + "epoch": 0.34, + "learning_rate": 1.99923005090442e-05, + "loss": 0.1075, + "step": 2573 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992279688228565e-05, + "loss": 0.1007, + "step": 2574 + }, + { + "epoch": 0.34, + "learning_rate": 1.999225883931018e-05, + "loss": 0.1001, + "step": 2575 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992237962289092e-05, + "loss": 0.0887, + "step": 2576 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992217057165374e-05, + "loss": 0.1183, + "step": 2577 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992196123939074e-05, + "loss": 0.0999, + "step": 2578 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992175162610254e-05, + "loss": 0.0953, + "step": 2579 + }, + { + "epoch": 0.34, + "learning_rate": 1.999215417317898e-05, + "loss": 0.1106, + "step": 2580 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992133155645294e-05, + "loss": 0.1039, + "step": 2581 + }, + { + "epoch": 0.34, + "learning_rate": 1.999211211000927e-05, + "loss": 0.0867, + "step": 2582 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992091036270964e-05, + "loss": 0.0973, + "step": 2583 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992069934430437e-05, + "loss": 0.1102, + "step": 2584 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992048804487738e-05, + "loss": 0.1017, + "step": 2585 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992027646442936e-05, + "loss": 0.1014, + "step": 2586 + }, + { + "epoch": 0.34, + "learning_rate": 1.9992006460296092e-05, + "loss": 0.1025, + "step": 2587 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991985246047262e-05, + "loss": 0.0959, + "step": 2588 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991964003696505e-05, + "loss": 0.0947, + "step": 2589 + }, + { + "epoch": 0.34, + "learning_rate": 1.999194273324388e-05, + "loss": 0.1023, + "step": 2590 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991921434689446e-05, + "loss": 0.1026, + "step": 2591 + }, + { + "epoch": 0.34, + "learning_rate": 1.999190010803327e-05, + "loss": 0.1141, + "step": 2592 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991878753275403e-05, + "loss": 0.1069, + "step": 2593 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991857370415907e-05, + "loss": 0.1007, + "step": 2594 + }, + { + "epoch": 0.34, + "learning_rate": 1.999183595945485e-05, + "loss": 0.1133, + "step": 2595 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991814520392285e-05, + "loss": 0.1066, + "step": 2596 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991793053228274e-05, + "loss": 0.1031, + "step": 2597 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991771557962874e-05, + "loss": 0.091, + "step": 2598 + }, + { + "epoch": 0.34, + "learning_rate": 1.999175003459615e-05, + "loss": 0.0927, + "step": 2599 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991728483128162e-05, + "loss": 0.1029, + "step": 2600 + }, + { + "epoch": 0.34, + "learning_rate": 1.999170690355897e-05, + "loss": 0.0998, + "step": 2601 + }, + { + "epoch": 0.34, + "learning_rate": 1.999168529588863e-05, + "loss": 0.1049, + "step": 2602 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991663660117214e-05, + "loss": 0.1119, + "step": 2603 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991641996244776e-05, + "loss": 0.1143, + "step": 2604 + }, + { + "epoch": 0.34, + "learning_rate": 1.999162030427137e-05, + "loss": 0.0931, + "step": 2605 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991598584197073e-05, + "loss": 0.1011, + "step": 2606 + }, + { + "epoch": 0.34, + "learning_rate": 1.999157683602193e-05, + "loss": 0.0992, + "step": 2607 + }, + { + "epoch": 0.34, + "learning_rate": 1.999155505974601e-05, + "loss": 0.0968, + "step": 2608 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991533255369375e-05, + "loss": 0.1055, + "step": 2609 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991511422892085e-05, + "loss": 0.1028, + "step": 2610 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991489562314203e-05, + "loss": 0.1031, + "step": 2611 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991467673635785e-05, + "loss": 0.0957, + "step": 2612 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991445756856898e-05, + "loss": 0.1048, + "step": 2613 + }, + { + "epoch": 0.34, + "learning_rate": 1.99914238119776e-05, + "loss": 0.0988, + "step": 2614 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991401838997958e-05, + "loss": 0.1051, + "step": 2615 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991379837918028e-05, + "loss": 0.1195, + "step": 2616 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991357808737875e-05, + "loss": 0.0942, + "step": 2617 + }, + { + "epoch": 0.34, + "learning_rate": 1.999133575145756e-05, + "loss": 0.1145, + "step": 2618 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991313666077147e-05, + "loss": 0.1059, + "step": 2619 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991291552596693e-05, + "loss": 0.1085, + "step": 2620 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991269411016264e-05, + "loss": 0.1199, + "step": 2621 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991247241335924e-05, + "loss": 0.1, + "step": 2622 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991225043555733e-05, + "loss": 0.0933, + "step": 2623 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991202817675752e-05, + "loss": 0.0963, + "step": 2624 + }, + { + "epoch": 0.34, + "learning_rate": 1.999118056369605e-05, + "loss": 0.1062, + "step": 2625 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991158281616677e-05, + "loss": 0.0954, + "step": 2626 + }, + { + "epoch": 0.34, + "learning_rate": 1.999113597143771e-05, + "loss": 0.1005, + "step": 2627 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991113633159203e-05, + "loss": 0.0901, + "step": 2628 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991091266781217e-05, + "loss": 0.0929, + "step": 2629 + }, + { + "epoch": 0.34, + "learning_rate": 1.9991068872303824e-05, + "loss": 0.1123, + "step": 2630 + }, + { + "epoch": 0.34, + "learning_rate": 1.999104644972708e-05, + "loss": 0.1074, + "step": 2631 + }, + { + "epoch": 0.34, + "learning_rate": 1.999102399905105e-05, + "loss": 0.1008, + "step": 2632 + }, + { + "epoch": 0.34, + "learning_rate": 1.99910015202758e-05, + "loss": 0.0958, + "step": 2633 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990979013401387e-05, + "loss": 0.1059, + "step": 2634 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990956478427877e-05, + "loss": 0.105, + "step": 2635 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990933915355332e-05, + "loss": 0.0911, + "step": 2636 + }, + { + "epoch": 0.35, + "learning_rate": 1.999091132418382e-05, + "loss": 0.1007, + "step": 2637 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990888704913405e-05, + "loss": 0.1035, + "step": 2638 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990866057544144e-05, + "loss": 0.102, + "step": 2639 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990843382076104e-05, + "loss": 0.0965, + "step": 2640 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990820678509352e-05, + "loss": 0.0854, + "step": 2641 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990797946843946e-05, + "loss": 0.1043, + "step": 2642 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990775187079956e-05, + "loss": 0.1034, + "step": 2643 + }, + { + "epoch": 0.35, + "learning_rate": 1.999075239921744e-05, + "loss": 0.0976, + "step": 2644 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990729583256466e-05, + "loss": 0.103, + "step": 2645 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990706739197096e-05, + "loss": 0.1014, + "step": 2646 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990683867039396e-05, + "loss": 0.0996, + "step": 2647 + }, + { + "epoch": 0.35, + "learning_rate": 1.999066096678343e-05, + "loss": 0.113, + "step": 2648 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990638038429263e-05, + "loss": 0.099, + "step": 2649 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990615081976956e-05, + "loss": 0.0981, + "step": 2650 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990592097426576e-05, + "loss": 0.1032, + "step": 2651 + }, + { + "epoch": 0.35, + "learning_rate": 1.999056908477819e-05, + "loss": 0.1115, + "step": 2652 + }, + { + "epoch": 0.35, + "learning_rate": 1.999054604403186e-05, + "loss": 0.1053, + "step": 2653 + }, + { + "epoch": 0.35, + "learning_rate": 1.999052297518765e-05, + "loss": 0.0987, + "step": 2654 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990499878245625e-05, + "loss": 0.1031, + "step": 2655 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990476753205853e-05, + "loss": 0.0969, + "step": 2656 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990453600068402e-05, + "loss": 0.0879, + "step": 2657 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990430418833323e-05, + "loss": 0.0972, + "step": 2658 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990407209500698e-05, + "loss": 0.1048, + "step": 2659 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990383972070583e-05, + "loss": 0.1, + "step": 2660 + }, + { + "epoch": 0.35, + "learning_rate": 1.999036070654304e-05, + "loss": 0.0959, + "step": 2661 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990337412918145e-05, + "loss": 0.0948, + "step": 2662 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990314091195957e-05, + "loss": 0.0988, + "step": 2663 + }, + { + "epoch": 0.35, + "learning_rate": 1.999029074137654e-05, + "loss": 0.1114, + "step": 2664 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990267363459964e-05, + "loss": 0.11, + "step": 2665 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990243957446295e-05, + "loss": 0.1066, + "step": 2666 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990220523335595e-05, + "loss": 0.1236, + "step": 2667 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990197061127934e-05, + "loss": 0.113, + "step": 2668 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990173570823374e-05, + "loss": 0.1039, + "step": 2669 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990150052421985e-05, + "loss": 0.0996, + "step": 2670 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990126505923826e-05, + "loss": 0.1068, + "step": 2671 + }, + { + "epoch": 0.35, + "learning_rate": 1.999010293132897e-05, + "loss": 0.0981, + "step": 2672 + }, + { + "epoch": 0.35, + "learning_rate": 1.9990079328637484e-05, + "loss": 0.1004, + "step": 2673 + }, + { + "epoch": 0.35, + "learning_rate": 1.999005569784943e-05, + "loss": 0.0972, + "step": 2674 + }, + { + "epoch": 0.35, + "learning_rate": 1.999003203896488e-05, + "loss": 0.1009, + "step": 2675 + }, + { + "epoch": 0.35, + "learning_rate": 1.999000835198389e-05, + "loss": 0.1084, + "step": 2676 + }, + { + "epoch": 0.35, + "learning_rate": 1.998998463690654e-05, + "loss": 0.0885, + "step": 2677 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989960893732886e-05, + "loss": 0.0944, + "step": 2678 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989937122462998e-05, + "loss": 0.0903, + "step": 2679 + }, + { + "epoch": 0.35, + "learning_rate": 1.998991332309695e-05, + "loss": 0.0991, + "step": 2680 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989889495634796e-05, + "loss": 0.0955, + "step": 2681 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989865640076612e-05, + "loss": 0.0963, + "step": 2682 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989841756422463e-05, + "loss": 0.0963, + "step": 2683 + }, + { + "epoch": 0.35, + "learning_rate": 1.998981784467242e-05, + "loss": 0.095, + "step": 2684 + }, + { + "epoch": 0.35, + "learning_rate": 1.998979390482654e-05, + "loss": 0.1032, + "step": 2685 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989769936884903e-05, + "loss": 0.1088, + "step": 2686 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989745940847564e-05, + "loss": 0.108, + "step": 2687 + }, + { + "epoch": 0.35, + "learning_rate": 1.99897219167146e-05, + "loss": 0.1033, + "step": 2688 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989697864486076e-05, + "loss": 0.1057, + "step": 2689 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989673784162054e-05, + "loss": 0.1072, + "step": 2690 + }, + { + "epoch": 0.35, + "learning_rate": 1.998964967574261e-05, + "loss": 0.0949, + "step": 2691 + }, + { + "epoch": 0.35, + "learning_rate": 1.998962553922781e-05, + "loss": 0.102, + "step": 2692 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989601374617714e-05, + "loss": 0.1045, + "step": 2693 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989577181912404e-05, + "loss": 0.1122, + "step": 2694 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989552961111932e-05, + "loss": 0.1004, + "step": 2695 + }, + { + "epoch": 0.35, + "learning_rate": 1.998952871221638e-05, + "loss": 0.093, + "step": 2696 + }, + { + "epoch": 0.35, + "learning_rate": 1.998950443522581e-05, + "loss": 0.0987, + "step": 2697 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989480130140288e-05, + "loss": 0.1018, + "step": 2698 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989455796959886e-05, + "loss": 0.1009, + "step": 2699 + }, + { + "epoch": 0.35, + "learning_rate": 1.998943143568467e-05, + "loss": 0.1133, + "step": 2700 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989407046314716e-05, + "loss": 0.1102, + "step": 2701 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989382628850084e-05, + "loss": 0.1045, + "step": 2702 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989358183290842e-05, + "loss": 0.1112, + "step": 2703 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989333709637067e-05, + "loss": 0.0918, + "step": 2704 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989309207888817e-05, + "loss": 0.1124, + "step": 2705 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989284678046172e-05, + "loss": 0.1117, + "step": 2706 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989260120109196e-05, + "loss": 0.0964, + "step": 2707 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989235534077954e-05, + "loss": 0.0911, + "step": 2708 + }, + { + "epoch": 0.35, + "learning_rate": 1.998921091995252e-05, + "loss": 0.1018, + "step": 2709 + }, + { + "epoch": 0.35, + "learning_rate": 1.9989186277732962e-05, + "loss": 0.1007, + "step": 2710 + }, + { + "epoch": 0.36, + "learning_rate": 1.9989161607419353e-05, + "loss": 0.0987, + "step": 2711 + }, + { + "epoch": 0.36, + "learning_rate": 1.9989136909011754e-05, + "loss": 0.1095, + "step": 2712 + }, + { + "epoch": 0.36, + "learning_rate": 1.998911218251024e-05, + "loss": 0.0949, + "step": 2713 + }, + { + "epoch": 0.36, + "learning_rate": 1.9989087427914885e-05, + "loss": 0.1072, + "step": 2714 + }, + { + "epoch": 0.36, + "learning_rate": 1.998906264522575e-05, + "loss": 0.1031, + "step": 2715 + }, + { + "epoch": 0.36, + "learning_rate": 1.998903783444291e-05, + "loss": 0.1039, + "step": 2716 + }, + { + "epoch": 0.36, + "learning_rate": 1.998901299556643e-05, + "loss": 0.1012, + "step": 2717 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988988128596385e-05, + "loss": 0.1074, + "step": 2718 + }, + { + "epoch": 0.36, + "learning_rate": 1.998896323353284e-05, + "loss": 0.0897, + "step": 2719 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988938310375872e-05, + "loss": 0.0924, + "step": 2720 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988913359125547e-05, + "loss": 0.0967, + "step": 2721 + }, + { + "epoch": 0.36, + "learning_rate": 1.998888837978193e-05, + "loss": 0.0965, + "step": 2722 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988863372345102e-05, + "loss": 0.0898, + "step": 2723 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988838336815128e-05, + "loss": 0.1006, + "step": 2724 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988813273192072e-05, + "loss": 0.0811, + "step": 2725 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988788181476014e-05, + "loss": 0.0989, + "step": 2726 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988763061667024e-05, + "loss": 0.1001, + "step": 2727 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988737913765166e-05, + "loss": 0.1026, + "step": 2728 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988712737770515e-05, + "loss": 0.1063, + "step": 2729 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988687533683142e-05, + "loss": 0.1003, + "step": 2730 + }, + { + "epoch": 0.36, + "learning_rate": 1.998866230150312e-05, + "loss": 0.1006, + "step": 2731 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988637041230513e-05, + "loss": 0.1031, + "step": 2732 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988611752865398e-05, + "loss": 0.1037, + "step": 2733 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988586436407843e-05, + "loss": 0.106, + "step": 2734 + }, + { + "epoch": 0.36, + "learning_rate": 1.998856109185792e-05, + "loss": 0.0939, + "step": 2735 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988535719215702e-05, + "loss": 0.0995, + "step": 2736 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988510318481262e-05, + "loss": 0.0928, + "step": 2737 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988484889654666e-05, + "loss": 0.1089, + "step": 2738 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988459432735988e-05, + "loss": 0.0927, + "step": 2739 + }, + { + "epoch": 0.36, + "learning_rate": 1.99884339477253e-05, + "loss": 0.112, + "step": 2740 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988408434622673e-05, + "loss": 0.0869, + "step": 2741 + }, + { + "epoch": 0.36, + "learning_rate": 1.998838289342818e-05, + "loss": 0.1113, + "step": 2742 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988357324141893e-05, + "loss": 0.1084, + "step": 2743 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988331726763882e-05, + "loss": 0.1189, + "step": 2744 + }, + { + "epoch": 0.36, + "learning_rate": 1.998830610129422e-05, + "loss": 0.1047, + "step": 2745 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988280447732978e-05, + "loss": 0.1089, + "step": 2746 + }, + { + "epoch": 0.36, + "learning_rate": 1.998825476608023e-05, + "loss": 0.0962, + "step": 2747 + }, + { + "epoch": 0.36, + "learning_rate": 1.998822905633605e-05, + "loss": 0.1253, + "step": 2748 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988203318500503e-05, + "loss": 0.1045, + "step": 2749 + }, + { + "epoch": 0.36, + "learning_rate": 1.998817755257367e-05, + "loss": 0.0963, + "step": 2750 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988151758555615e-05, + "loss": 0.1053, + "step": 2751 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988125936446417e-05, + "loss": 0.1091, + "step": 2752 + }, + { + "epoch": 0.36, + "learning_rate": 1.998810008624615e-05, + "loss": 0.0883, + "step": 2753 + }, + { + "epoch": 0.36, + "learning_rate": 1.998807420795488e-05, + "loss": 0.1098, + "step": 2754 + }, + { + "epoch": 0.36, + "learning_rate": 1.998804830157268e-05, + "loss": 0.1057, + "step": 2755 + }, + { + "epoch": 0.36, + "learning_rate": 1.9988022367099633e-05, + "loss": 0.1004, + "step": 2756 + }, + { + "epoch": 0.36, + "learning_rate": 1.99879964045358e-05, + "loss": 0.107, + "step": 2757 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987970413881264e-05, + "loss": 0.1142, + "step": 2758 + }, + { + "epoch": 0.36, + "learning_rate": 1.998794439513609e-05, + "loss": 0.0947, + "step": 2759 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987918348300356e-05, + "loss": 0.0979, + "step": 2760 + }, + { + "epoch": 0.36, + "learning_rate": 1.998789227337413e-05, + "loss": 0.0994, + "step": 2761 + }, + { + "epoch": 0.36, + "learning_rate": 1.998786617035749e-05, + "loss": 0.1015, + "step": 2762 + }, + { + "epoch": 0.36, + "learning_rate": 1.998784003925051e-05, + "loss": 0.1027, + "step": 2763 + }, + { + "epoch": 0.36, + "learning_rate": 1.998781388005326e-05, + "loss": 0.0989, + "step": 2764 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987787692765818e-05, + "loss": 0.1028, + "step": 2765 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987761477388252e-05, + "loss": 0.0947, + "step": 2766 + }, + { + "epoch": 0.36, + "learning_rate": 1.998773523392064e-05, + "loss": 0.1092, + "step": 2767 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987708962363056e-05, + "loss": 0.0953, + "step": 2768 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987682662715572e-05, + "loss": 0.1069, + "step": 2769 + }, + { + "epoch": 0.36, + "learning_rate": 1.998765633497826e-05, + "loss": 0.1113, + "step": 2770 + }, + { + "epoch": 0.36, + "learning_rate": 1.99876299791512e-05, + "loss": 0.0891, + "step": 2771 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987603595234456e-05, + "loss": 0.091, + "step": 2772 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987577183228116e-05, + "loss": 0.1001, + "step": 2773 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987550743132242e-05, + "loss": 0.1031, + "step": 2774 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987524274946915e-05, + "loss": 0.102, + "step": 2775 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987497778672208e-05, + "loss": 0.1029, + "step": 2776 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987471254308194e-05, + "loss": 0.098, + "step": 2777 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987444701854954e-05, + "loss": 0.1, + "step": 2778 + }, + { + "epoch": 0.36, + "learning_rate": 1.998741812131255e-05, + "loss": 0.1062, + "step": 2779 + }, + { + "epoch": 0.36, + "learning_rate": 1.998739151268107e-05, + "loss": 0.0976, + "step": 2780 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987364875960585e-05, + "loss": 0.1008, + "step": 2781 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987338211151163e-05, + "loss": 0.1066, + "step": 2782 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987311518252888e-05, + "loss": 0.0985, + "step": 2783 + }, + { + "epoch": 0.36, + "learning_rate": 1.998728479726583e-05, + "loss": 0.092, + "step": 2784 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987258048190065e-05, + "loss": 0.101, + "step": 2785 + }, + { + "epoch": 0.36, + "learning_rate": 1.9987231271025666e-05, + "loss": 0.1108, + "step": 2786 + }, + { + "epoch": 0.37, + "learning_rate": 1.9987204465772713e-05, + "loss": 0.1026, + "step": 2787 + }, + { + "epoch": 0.37, + "learning_rate": 1.998717763243128e-05, + "loss": 0.0928, + "step": 2788 + }, + { + "epoch": 0.37, + "learning_rate": 1.998715077100144e-05, + "loss": 0.0927, + "step": 2789 + }, + { + "epoch": 0.37, + "learning_rate": 1.998712388148327e-05, + "loss": 0.118, + "step": 2790 + }, + { + "epoch": 0.37, + "learning_rate": 1.9987096963876852e-05, + "loss": 0.0996, + "step": 2791 + }, + { + "epoch": 0.37, + "learning_rate": 1.998707001818225e-05, + "loss": 0.09, + "step": 2792 + }, + { + "epoch": 0.37, + "learning_rate": 1.9987043044399547e-05, + "loss": 0.1043, + "step": 2793 + }, + { + "epoch": 0.37, + "learning_rate": 1.9987016042528817e-05, + "loss": 0.0965, + "step": 2794 + }, + { + "epoch": 0.37, + "learning_rate": 1.998698901257014e-05, + "loss": 0.1012, + "step": 2795 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986961954523577e-05, + "loss": 0.0913, + "step": 2796 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986934868389226e-05, + "loss": 0.1089, + "step": 2797 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986907754167148e-05, + "loss": 0.1122, + "step": 2798 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986880611857423e-05, + "loss": 0.1057, + "step": 2799 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986853441460133e-05, + "loss": 0.0949, + "step": 2800 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986826242975346e-05, + "loss": 0.0978, + "step": 2801 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986799016403142e-05, + "loss": 0.1043, + "step": 2802 + }, + { + "epoch": 0.37, + "learning_rate": 1.99867717617436e-05, + "loss": 0.1027, + "step": 2803 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986744478996793e-05, + "loss": 0.1011, + "step": 2804 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986717168162796e-05, + "loss": 0.0955, + "step": 2805 + }, + { + "epoch": 0.37, + "learning_rate": 1.998668982924169e-05, + "loss": 0.094, + "step": 2806 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986662462233552e-05, + "loss": 0.0885, + "step": 2807 + }, + { + "epoch": 0.37, + "learning_rate": 1.998663506713846e-05, + "loss": 0.1001, + "step": 2808 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986607643956485e-05, + "loss": 0.1006, + "step": 2809 + }, + { + "epoch": 0.37, + "learning_rate": 1.998658019268771e-05, + "loss": 0.1034, + "step": 2810 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986552713332206e-05, + "loss": 0.1028, + "step": 2811 + }, + { + "epoch": 0.37, + "learning_rate": 1.998652520589006e-05, + "loss": 0.0941, + "step": 2812 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986497670361344e-05, + "loss": 0.0986, + "step": 2813 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986470106746128e-05, + "loss": 0.0982, + "step": 2814 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986442515044503e-05, + "loss": 0.1016, + "step": 2815 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986414895256536e-05, + "loss": 0.1147, + "step": 2816 + }, + { + "epoch": 0.37, + "learning_rate": 1.998638724738231e-05, + "loss": 0.098, + "step": 2817 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986359571421904e-05, + "loss": 0.0848, + "step": 2818 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986331867375392e-05, + "loss": 0.099, + "step": 2819 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986304135242857e-05, + "loss": 0.0991, + "step": 2820 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986276375024366e-05, + "loss": 0.1119, + "step": 2821 + }, + { + "epoch": 0.37, + "learning_rate": 1.998624858672001e-05, + "loss": 0.1037, + "step": 2822 + }, + { + "epoch": 0.37, + "learning_rate": 1.998622077032986e-05, + "loss": 0.1015, + "step": 2823 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986192925853998e-05, + "loss": 0.1132, + "step": 2824 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986165053292497e-05, + "loss": 0.1142, + "step": 2825 + }, + { + "epoch": 0.37, + "learning_rate": 1.998613715264544e-05, + "loss": 0.0914, + "step": 2826 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986109223912905e-05, + "loss": 0.1082, + "step": 2827 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986081267094967e-05, + "loss": 0.0991, + "step": 2828 + }, + { + "epoch": 0.37, + "learning_rate": 1.998605328219171e-05, + "loss": 0.1004, + "step": 2829 + }, + { + "epoch": 0.37, + "learning_rate": 1.9986025269203205e-05, + "loss": 0.0942, + "step": 2830 + }, + { + "epoch": 0.37, + "learning_rate": 1.998599722812954e-05, + "loss": 0.103, + "step": 2831 + }, + { + "epoch": 0.37, + "learning_rate": 1.998596915897079e-05, + "loss": 0.0993, + "step": 2832 + }, + { + "epoch": 0.37, + "learning_rate": 1.998594106172703e-05, + "loss": 0.1026, + "step": 2833 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985912936398346e-05, + "loss": 0.0935, + "step": 2834 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985884782984816e-05, + "loss": 0.0912, + "step": 2835 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985856601486513e-05, + "loss": 0.1077, + "step": 2836 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985828391903518e-05, + "loss": 0.1003, + "step": 2837 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985800154235915e-05, + "loss": 0.1046, + "step": 2838 + }, + { + "epoch": 0.37, + "learning_rate": 1.998577188848378e-05, + "loss": 0.0989, + "step": 2839 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985743594647197e-05, + "loss": 0.099, + "step": 2840 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985715272726237e-05, + "loss": 0.0999, + "step": 2841 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985686922720985e-05, + "loss": 0.1054, + "step": 2842 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985658544631525e-05, + "loss": 0.0901, + "step": 2843 + }, + { + "epoch": 0.37, + "learning_rate": 1.998563013845793e-05, + "loss": 0.1003, + "step": 2844 + }, + { + "epoch": 0.37, + "learning_rate": 1.998560170420028e-05, + "loss": 0.0978, + "step": 2845 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985573241858657e-05, + "loss": 0.1004, + "step": 2846 + }, + { + "epoch": 0.37, + "learning_rate": 1.998554475143314e-05, + "loss": 0.0961, + "step": 2847 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985516232923816e-05, + "loss": 0.0953, + "step": 2848 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985487686330754e-05, + "loss": 0.0954, + "step": 2849 + }, + { + "epoch": 0.37, + "learning_rate": 1.998545911165404e-05, + "loss": 0.0849, + "step": 2850 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985430508893758e-05, + "loss": 0.0983, + "step": 2851 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985401878049982e-05, + "loss": 0.103, + "step": 2852 + }, + { + "epoch": 0.37, + "learning_rate": 1.998537321912279e-05, + "loss": 0.0928, + "step": 2853 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985344532112278e-05, + "loss": 0.0952, + "step": 2854 + }, + { + "epoch": 0.37, + "learning_rate": 1.998531581701851e-05, + "loss": 0.1134, + "step": 2855 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985287073841572e-05, + "loss": 0.0954, + "step": 2856 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985258302581546e-05, + "loss": 0.1007, + "step": 2857 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985229503238516e-05, + "loss": 0.0958, + "step": 2858 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985200675812555e-05, + "loss": 0.1004, + "step": 2859 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985171820303752e-05, + "loss": 0.1156, + "step": 2860 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985142936712186e-05, + "loss": 0.0976, + "step": 2861 + }, + { + "epoch": 0.37, + "learning_rate": 1.9985114025037936e-05, + "loss": 0.1095, + "step": 2862 + }, + { + "epoch": 0.38, + "learning_rate": 1.9985085085281085e-05, + "loss": 0.1003, + "step": 2863 + }, + { + "epoch": 0.38, + "learning_rate": 1.9985056117441714e-05, + "loss": 0.0919, + "step": 2864 + }, + { + "epoch": 0.38, + "learning_rate": 1.9985027121519903e-05, + "loss": 0.0973, + "step": 2865 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984998097515732e-05, + "loss": 0.0981, + "step": 2866 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984969045429292e-05, + "loss": 0.0946, + "step": 2867 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984939965260656e-05, + "loss": 0.1031, + "step": 2868 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984910857009908e-05, + "loss": 0.1014, + "step": 2869 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984881720677126e-05, + "loss": 0.0948, + "step": 2870 + }, + { + "epoch": 0.38, + "learning_rate": 1.99848525562624e-05, + "loss": 0.098, + "step": 2871 + }, + { + "epoch": 0.38, + "learning_rate": 1.998482336376581e-05, + "loss": 0.1095, + "step": 2872 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984794143187432e-05, + "loss": 0.0998, + "step": 2873 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984764894527356e-05, + "loss": 0.0918, + "step": 2874 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984735617785656e-05, + "loss": 0.0966, + "step": 2875 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984706312962423e-05, + "loss": 0.1066, + "step": 2876 + }, + { + "epoch": 0.38, + "learning_rate": 1.998467698005773e-05, + "loss": 0.0972, + "step": 2877 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984647619071672e-05, + "loss": 0.1003, + "step": 2878 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984618230004317e-05, + "loss": 0.1074, + "step": 2879 + }, + { + "epoch": 0.38, + "learning_rate": 1.998458881285576e-05, + "loss": 0.1037, + "step": 2880 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984559367626075e-05, + "loss": 0.1094, + "step": 2881 + }, + { + "epoch": 0.38, + "learning_rate": 1.998452989431535e-05, + "loss": 0.1019, + "step": 2882 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984500392923668e-05, + "loss": 0.0848, + "step": 2883 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984470863451108e-05, + "loss": 0.1018, + "step": 2884 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984441305897757e-05, + "loss": 0.0938, + "step": 2885 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984411720263694e-05, + "loss": 0.0777, + "step": 2886 + }, + { + "epoch": 0.38, + "learning_rate": 1.998438210654901e-05, + "loss": 0.1064, + "step": 2887 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984352464753778e-05, + "loss": 0.1015, + "step": 2888 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984322794878087e-05, + "loss": 0.1027, + "step": 2889 + }, + { + "epoch": 0.38, + "learning_rate": 1.998429309692202e-05, + "loss": 0.1076, + "step": 2890 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984263370885665e-05, + "loss": 0.1028, + "step": 2891 + }, + { + "epoch": 0.38, + "learning_rate": 1.998423361676909e-05, + "loss": 0.0971, + "step": 2892 + }, + { + "epoch": 0.38, + "learning_rate": 1.99842038345724e-05, + "loss": 0.0981, + "step": 2893 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984174024295666e-05, + "loss": 0.0955, + "step": 2894 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984144185938973e-05, + "loss": 0.0955, + "step": 2895 + }, + { + "epoch": 0.38, + "learning_rate": 1.998411431950241e-05, + "loss": 0.0967, + "step": 2896 + }, + { + "epoch": 0.38, + "learning_rate": 1.998408442498605e-05, + "loss": 0.1054, + "step": 2897 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984054502389988e-05, + "loss": 0.0983, + "step": 2898 + }, + { + "epoch": 0.38, + "learning_rate": 1.9984024551714307e-05, + "loss": 0.0871, + "step": 2899 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983994572959083e-05, + "loss": 0.1114, + "step": 2900 + }, + { + "epoch": 0.38, + "learning_rate": 1.998396456612441e-05, + "loss": 0.101, + "step": 2901 + }, + { + "epoch": 0.38, + "learning_rate": 1.998393453121037e-05, + "loss": 0.0988, + "step": 2902 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983904468217043e-05, + "loss": 0.1057, + "step": 2903 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983874377144516e-05, + "loss": 0.1077, + "step": 2904 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983844257992873e-05, + "loss": 0.1035, + "step": 2905 + }, + { + "epoch": 0.38, + "learning_rate": 1.99838141107622e-05, + "loss": 0.1078, + "step": 2906 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983783935452586e-05, + "loss": 0.0975, + "step": 2907 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983753732064106e-05, + "loss": 0.0968, + "step": 2908 + }, + { + "epoch": 0.38, + "learning_rate": 1.998372350059685e-05, + "loss": 0.1001, + "step": 2909 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983693241050907e-05, + "loss": 0.0952, + "step": 2910 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983662953426356e-05, + "loss": 0.0962, + "step": 2911 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983632637723287e-05, + "loss": 0.107, + "step": 2912 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983602293941783e-05, + "loss": 0.1047, + "step": 2913 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983571922081927e-05, + "loss": 0.1042, + "step": 2914 + }, + { + "epoch": 0.38, + "learning_rate": 1.998354152214381e-05, + "loss": 0.1123, + "step": 2915 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983511094127508e-05, + "loss": 0.103, + "step": 2916 + }, + { + "epoch": 0.38, + "learning_rate": 1.998348063803312e-05, + "loss": 0.1019, + "step": 2917 + }, + { + "epoch": 0.38, + "learning_rate": 1.998345015386072e-05, + "loss": 0.1027, + "step": 2918 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983419641610396e-05, + "loss": 0.0961, + "step": 2919 + }, + { + "epoch": 0.38, + "learning_rate": 1.998338910128224e-05, + "loss": 0.0954, + "step": 2920 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983358532876333e-05, + "loss": 0.1106, + "step": 2921 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983327936392758e-05, + "loss": 0.1019, + "step": 2922 + }, + { + "epoch": 0.38, + "learning_rate": 1.998329731183161e-05, + "loss": 0.1021, + "step": 2923 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983266659192965e-05, + "loss": 0.0941, + "step": 2924 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983235978476917e-05, + "loss": 0.1056, + "step": 2925 + }, + { + "epoch": 0.38, + "learning_rate": 1.998320526968355e-05, + "loss": 0.0929, + "step": 2926 + }, + { + "epoch": 0.38, + "learning_rate": 1.998317453281295e-05, + "loss": 0.1043, + "step": 2927 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983143767865202e-05, + "loss": 0.1082, + "step": 2928 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983112974840395e-05, + "loss": 0.0986, + "step": 2929 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983082153738614e-05, + "loss": 0.0927, + "step": 2930 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983051304559944e-05, + "loss": 0.1022, + "step": 2931 + }, + { + "epoch": 0.38, + "learning_rate": 1.9983020427304476e-05, + "loss": 0.104, + "step": 2932 + }, + { + "epoch": 0.38, + "learning_rate": 1.9982989521972295e-05, + "loss": 0.107, + "step": 2933 + }, + { + "epoch": 0.38, + "learning_rate": 1.9982958588563486e-05, + "loss": 0.099, + "step": 2934 + }, + { + "epoch": 0.38, + "learning_rate": 1.998292762707814e-05, + "loss": 0.115, + "step": 2935 + }, + { + "epoch": 0.38, + "learning_rate": 1.998289663751634e-05, + "loss": 0.1007, + "step": 2936 + }, + { + "epoch": 0.38, + "learning_rate": 1.9982865619878175e-05, + "loss": 0.1189, + "step": 2937 + }, + { + "epoch": 0.38, + "learning_rate": 1.998283457416373e-05, + "loss": 0.0904, + "step": 2938 + }, + { + "epoch": 0.38, + "learning_rate": 1.99828035003731e-05, + "loss": 0.1124, + "step": 2939 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982772398506363e-05, + "loss": 0.0977, + "step": 2940 + }, + { + "epoch": 0.39, + "learning_rate": 1.998274126856361e-05, + "loss": 0.0926, + "step": 2941 + }, + { + "epoch": 0.39, + "learning_rate": 1.998271011054493e-05, + "loss": 0.1146, + "step": 2942 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982678924450408e-05, + "loss": 0.0828, + "step": 2943 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982647710280135e-05, + "loss": 0.1109, + "step": 2944 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982616468034192e-05, + "loss": 0.098, + "step": 2945 + }, + { + "epoch": 0.39, + "learning_rate": 1.998258519771268e-05, + "loss": 0.0994, + "step": 2946 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982553899315672e-05, + "loss": 0.0969, + "step": 2947 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982522572843263e-05, + "loss": 0.1084, + "step": 2948 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982491218295545e-05, + "loss": 0.0978, + "step": 2949 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982459835672597e-05, + "loss": 0.1074, + "step": 2950 + }, + { + "epoch": 0.39, + "learning_rate": 1.998242842497452e-05, + "loss": 0.0862, + "step": 2951 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982396986201388e-05, + "loss": 0.1022, + "step": 2952 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982365519353296e-05, + "loss": 0.0875, + "step": 2953 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982334024430337e-05, + "loss": 0.1199, + "step": 2954 + }, + { + "epoch": 0.39, + "learning_rate": 1.998230250143259e-05, + "loss": 0.1, + "step": 2955 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982270950360153e-05, + "loss": 0.098, + "step": 2956 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982239371213107e-05, + "loss": 0.0977, + "step": 2957 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982207763991547e-05, + "loss": 0.0941, + "step": 2958 + }, + { + "epoch": 0.39, + "learning_rate": 1.998217612869556e-05, + "loss": 0.0983, + "step": 2959 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982144465325234e-05, + "loss": 0.1049, + "step": 2960 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982112773880657e-05, + "loss": 0.1127, + "step": 2961 + }, + { + "epoch": 0.39, + "learning_rate": 1.9982081054361918e-05, + "loss": 0.0999, + "step": 2962 + }, + { + "epoch": 0.39, + "learning_rate": 1.998204930676911e-05, + "loss": 0.1044, + "step": 2963 + }, + { + "epoch": 0.39, + "learning_rate": 1.998201753110232e-05, + "loss": 0.1034, + "step": 2964 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981985727361634e-05, + "loss": 0.0868, + "step": 2965 + }, + { + "epoch": 0.39, + "learning_rate": 1.998195389554715e-05, + "loss": 0.0943, + "step": 2966 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981922035658945e-05, + "loss": 0.1025, + "step": 2967 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981890147697123e-05, + "loss": 0.0976, + "step": 2968 + }, + { + "epoch": 0.39, + "learning_rate": 1.998185823166176e-05, + "loss": 0.1036, + "step": 2969 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981826287552958e-05, + "loss": 0.1146, + "step": 2970 + }, + { + "epoch": 0.39, + "learning_rate": 1.99817943153708e-05, + "loss": 0.1161, + "step": 2971 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981762315115377e-05, + "loss": 0.0935, + "step": 2972 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981730286786776e-05, + "loss": 0.0957, + "step": 2973 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981698230385094e-05, + "loss": 0.1034, + "step": 2974 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981666145910413e-05, + "loss": 0.1021, + "step": 2975 + }, + { + "epoch": 0.39, + "learning_rate": 1.998163403336283e-05, + "loss": 0.0983, + "step": 2976 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981601892742434e-05, + "loss": 0.1078, + "step": 2977 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981569724049314e-05, + "loss": 0.0991, + "step": 2978 + }, + { + "epoch": 0.39, + "learning_rate": 1.998153752728356e-05, + "loss": 0.107, + "step": 2979 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981505302445262e-05, + "loss": 0.105, + "step": 2980 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981473049534515e-05, + "loss": 0.0967, + "step": 2981 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981440768551402e-05, + "loss": 0.0887, + "step": 2982 + }, + { + "epoch": 0.39, + "learning_rate": 1.998140845949602e-05, + "loss": 0.1082, + "step": 2983 + }, + { + "epoch": 0.39, + "learning_rate": 1.998137612236846e-05, + "loss": 0.0972, + "step": 2984 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981343757168812e-05, + "loss": 0.1044, + "step": 2985 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981311363897163e-05, + "loss": 0.1058, + "step": 2986 + }, + { + "epoch": 0.39, + "learning_rate": 1.998127894255361e-05, + "loss": 0.0975, + "step": 2987 + }, + { + "epoch": 0.39, + "learning_rate": 1.998124649313824e-05, + "loss": 0.101, + "step": 2988 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981214015651145e-05, + "loss": 0.1112, + "step": 2989 + }, + { + "epoch": 0.39, + "learning_rate": 1.998118151009242e-05, + "loss": 0.1001, + "step": 2990 + }, + { + "epoch": 0.39, + "learning_rate": 1.998114897646215e-05, + "loss": 0.1029, + "step": 2991 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981116414760434e-05, + "loss": 0.0839, + "step": 2992 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981083824987357e-05, + "loss": 0.0963, + "step": 2993 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981051207143012e-05, + "loss": 0.108, + "step": 2994 + }, + { + "epoch": 0.39, + "learning_rate": 1.9981018561227493e-05, + "loss": 0.1102, + "step": 2995 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980985887240893e-05, + "loss": 0.1005, + "step": 2996 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980953185183302e-05, + "loss": 0.1045, + "step": 2997 + }, + { + "epoch": 0.39, + "learning_rate": 1.998092045505481e-05, + "loss": 0.0952, + "step": 2998 + }, + { + "epoch": 0.39, + "learning_rate": 1.998088769685551e-05, + "loss": 0.1137, + "step": 2999 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980854910585495e-05, + "loss": 0.0961, + "step": 3000 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980822096244858e-05, + "loss": 0.0985, + "step": 3001 + }, + { + "epoch": 0.39, + "learning_rate": 1.998078925383369e-05, + "loss": 0.0933, + "step": 3002 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980756383352085e-05, + "loss": 0.108, + "step": 3003 + }, + { + "epoch": 0.39, + "learning_rate": 1.998072348480013e-05, + "loss": 0.1031, + "step": 3004 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980690558177927e-05, + "loss": 0.1103, + "step": 3005 + }, + { + "epoch": 0.39, + "learning_rate": 1.998065760348556e-05, + "loss": 0.1197, + "step": 3006 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980624620723127e-05, + "loss": 0.0941, + "step": 3007 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980591609890717e-05, + "loss": 0.1152, + "step": 3008 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980558570988425e-05, + "loss": 0.1061, + "step": 3009 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980525504016345e-05, + "loss": 0.0863, + "step": 3010 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980492408974566e-05, + "loss": 0.1004, + "step": 3011 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980459285863185e-05, + "loss": 0.1028, + "step": 3012 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980426134682295e-05, + "loss": 0.0939, + "step": 3013 + }, + { + "epoch": 0.39, + "learning_rate": 1.9980392955431983e-05, + "loss": 0.0907, + "step": 3014 + }, + { + "epoch": 0.39, + "learning_rate": 1.998035974811235e-05, + "loss": 0.0897, + "step": 3015 + }, + { + "epoch": 0.4, + "learning_rate": 1.998032651272349e-05, + "loss": 0.0876, + "step": 3016 + }, + { + "epoch": 0.4, + "learning_rate": 1.9980293249265493e-05, + "loss": 0.1026, + "step": 3017 + }, + { + "epoch": 0.4, + "learning_rate": 1.998025995773845e-05, + "loss": 0.1021, + "step": 3018 + }, + { + "epoch": 0.4, + "learning_rate": 1.9980226638142456e-05, + "loss": 0.1115, + "step": 3019 + }, + { + "epoch": 0.4, + "learning_rate": 1.998019329047761e-05, + "loss": 0.0996, + "step": 3020 + }, + { + "epoch": 0.4, + "learning_rate": 1.9980159914744e-05, + "loss": 0.1024, + "step": 3021 + }, + { + "epoch": 0.4, + "learning_rate": 1.998012651094172e-05, + "loss": 0.1018, + "step": 3022 + }, + { + "epoch": 0.4, + "learning_rate": 1.9980093079070867e-05, + "loss": 0.0968, + "step": 3023 + }, + { + "epoch": 0.4, + "learning_rate": 1.9980059619131533e-05, + "loss": 0.09, + "step": 3024 + }, + { + "epoch": 0.4, + "learning_rate": 1.9980026131123817e-05, + "loss": 0.0968, + "step": 3025 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979992615047804e-05, + "loss": 0.1133, + "step": 3026 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979959070903593e-05, + "loss": 0.1059, + "step": 3027 + }, + { + "epoch": 0.4, + "learning_rate": 1.997992549869128e-05, + "loss": 0.093, + "step": 3028 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979891898410957e-05, + "loss": 0.0831, + "step": 3029 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979858270062723e-05, + "loss": 0.0848, + "step": 3030 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979824613646666e-05, + "loss": 0.1192, + "step": 3031 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979790929162885e-05, + "loss": 0.1094, + "step": 3032 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979757216611472e-05, + "loss": 0.1061, + "step": 3033 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979723475992524e-05, + "loss": 0.0946, + "step": 3034 + }, + { + "epoch": 0.4, + "learning_rate": 1.997968970730614e-05, + "loss": 0.0895, + "step": 3035 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979655910552403e-05, + "loss": 0.1109, + "step": 3036 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979622085731418e-05, + "loss": 0.1, + "step": 3037 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979588232843277e-05, + "loss": 0.1021, + "step": 3038 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979554351888076e-05, + "loss": 0.0986, + "step": 3039 + }, + { + "epoch": 0.4, + "learning_rate": 1.997952044286591e-05, + "loss": 0.1048, + "step": 3040 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979486505776874e-05, + "loss": 0.102, + "step": 3041 + }, + { + "epoch": 0.4, + "learning_rate": 1.997945254062106e-05, + "loss": 0.0913, + "step": 3042 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979418547398576e-05, + "loss": 0.1149, + "step": 3043 + }, + { + "epoch": 0.4, + "learning_rate": 1.99793845261095e-05, + "loss": 0.0994, + "step": 3044 + }, + { + "epoch": 0.4, + "learning_rate": 1.997935047675394e-05, + "loss": 0.1055, + "step": 3045 + }, + { + "epoch": 0.4, + "learning_rate": 1.997931639933199e-05, + "loss": 0.0876, + "step": 3046 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979282293843743e-05, + "loss": 0.0934, + "step": 3047 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979248160289294e-05, + "loss": 0.1026, + "step": 3048 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979213998668744e-05, + "loss": 0.1014, + "step": 3049 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979179808982184e-05, + "loss": 0.1035, + "step": 3050 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979145591229708e-05, + "loss": 0.1043, + "step": 3051 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979111345411423e-05, + "loss": 0.1001, + "step": 3052 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979077071527416e-05, + "loss": 0.1124, + "step": 3053 + }, + { + "epoch": 0.4, + "learning_rate": 1.997904276957779e-05, + "loss": 0.1039, + "step": 3054 + }, + { + "epoch": 0.4, + "learning_rate": 1.9979008439562633e-05, + "loss": 0.1125, + "step": 3055 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978974081482047e-05, + "loss": 0.0951, + "step": 3056 + }, + { + "epoch": 0.4, + "learning_rate": 1.997893969533613e-05, + "loss": 0.1036, + "step": 3057 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978905281124973e-05, + "loss": 0.097, + "step": 3058 + }, + { + "epoch": 0.4, + "learning_rate": 1.997887083884868e-05, + "loss": 0.1082, + "step": 3059 + }, + { + "epoch": 0.4, + "learning_rate": 1.997883636850734e-05, + "loss": 0.0883, + "step": 3060 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978801870101058e-05, + "loss": 0.1007, + "step": 3061 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978767343629926e-05, + "loss": 0.0991, + "step": 3062 + }, + { + "epoch": 0.4, + "learning_rate": 1.997873278909404e-05, + "loss": 0.1034, + "step": 3063 + }, + { + "epoch": 0.4, + "learning_rate": 1.99786982064935e-05, + "loss": 0.0968, + "step": 3064 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978663595828408e-05, + "loss": 0.0906, + "step": 3065 + }, + { + "epoch": 0.4, + "learning_rate": 1.997862895709885e-05, + "loss": 0.0986, + "step": 3066 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978594290304933e-05, + "loss": 0.0916, + "step": 3067 + }, + { + "epoch": 0.4, + "learning_rate": 1.997855959544675e-05, + "loss": 0.0917, + "step": 3068 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978524872524398e-05, + "loss": 0.1106, + "step": 3069 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978490121537977e-05, + "loss": 0.0927, + "step": 3070 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978455342487583e-05, + "loss": 0.1068, + "step": 3071 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978420535373317e-05, + "loss": 0.0989, + "step": 3072 + }, + { + "epoch": 0.4, + "learning_rate": 1.997838570019527e-05, + "loss": 0.1, + "step": 3073 + }, + { + "epoch": 0.4, + "learning_rate": 1.997835083695355e-05, + "loss": 0.1023, + "step": 3074 + }, + { + "epoch": 0.4, + "learning_rate": 1.997831594564825e-05, + "loss": 0.1104, + "step": 3075 + }, + { + "epoch": 0.4, + "learning_rate": 1.997828102627946e-05, + "loss": 0.0959, + "step": 3076 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978246078847296e-05, + "loss": 0.103, + "step": 3077 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978211103351838e-05, + "loss": 0.0899, + "step": 3078 + }, + { + "epoch": 0.4, + "learning_rate": 1.99781760997932e-05, + "loss": 0.0915, + "step": 3079 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978141068171468e-05, + "loss": 0.0994, + "step": 3080 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978106008486746e-05, + "loss": 0.0984, + "step": 3081 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978070920739134e-05, + "loss": 0.0963, + "step": 3082 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978035804928728e-05, + "loss": 0.1064, + "step": 3083 + }, + { + "epoch": 0.4, + "learning_rate": 1.9978000661055626e-05, + "loss": 0.1152, + "step": 3084 + }, + { + "epoch": 0.4, + "learning_rate": 1.9977965489119933e-05, + "loss": 0.0956, + "step": 3085 + }, + { + "epoch": 0.4, + "learning_rate": 1.997793028912174e-05, + "loss": 0.0932, + "step": 3086 + }, + { + "epoch": 0.4, + "learning_rate": 1.997789506106115e-05, + "loss": 0.1034, + "step": 3087 + }, + { + "epoch": 0.4, + "learning_rate": 1.997785980493826e-05, + "loss": 0.108, + "step": 3088 + }, + { + "epoch": 0.4, + "learning_rate": 1.9977824520753172e-05, + "loss": 0.1027, + "step": 3089 + }, + { + "epoch": 0.4, + "learning_rate": 1.9977789208505986e-05, + "loss": 0.0952, + "step": 3090 + }, + { + "epoch": 0.4, + "learning_rate": 1.99777538681968e-05, + "loss": 0.0934, + "step": 3091 + }, + { + "epoch": 0.41, + "learning_rate": 1.997771849982571e-05, + "loss": 0.095, + "step": 3092 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977683103392816e-05, + "loss": 0.0968, + "step": 3093 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977647678898224e-05, + "loss": 0.1068, + "step": 3094 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977612226342026e-05, + "loss": 0.1017, + "step": 3095 + }, + { + "epoch": 0.41, + "learning_rate": 1.997757674572432e-05, + "loss": 0.0946, + "step": 3096 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977541237045218e-05, + "loss": 0.1016, + "step": 3097 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977505700304815e-05, + "loss": 0.0944, + "step": 3098 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977470135503202e-05, + "loss": 0.1074, + "step": 3099 + }, + { + "epoch": 0.41, + "learning_rate": 1.997743454264049e-05, + "loss": 0.1077, + "step": 3100 + }, + { + "epoch": 0.41, + "learning_rate": 1.997739892171677e-05, + "loss": 0.0981, + "step": 3101 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977363272732153e-05, + "loss": 0.1116, + "step": 3102 + }, + { + "epoch": 0.41, + "learning_rate": 1.997732759568673e-05, + "loss": 0.102, + "step": 3103 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977291890580603e-05, + "loss": 0.0982, + "step": 3104 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977256157413876e-05, + "loss": 0.1084, + "step": 3105 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977220396186644e-05, + "loss": 0.0992, + "step": 3106 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977184606899017e-05, + "loss": 0.0974, + "step": 3107 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977148789551086e-05, + "loss": 0.1155, + "step": 3108 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977112944142956e-05, + "loss": 0.0898, + "step": 3109 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977077070674728e-05, + "loss": 0.1013, + "step": 3110 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977041169146497e-05, + "loss": 0.095, + "step": 3111 + }, + { + "epoch": 0.41, + "learning_rate": 1.9977005239558376e-05, + "loss": 0.0946, + "step": 3112 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976969281910455e-05, + "loss": 0.0903, + "step": 3113 + }, + { + "epoch": 0.41, + "learning_rate": 1.997693329620284e-05, + "loss": 0.0952, + "step": 3114 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976897282435632e-05, + "loss": 0.0984, + "step": 3115 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976861240608933e-05, + "loss": 0.0894, + "step": 3116 + }, + { + "epoch": 0.41, + "learning_rate": 1.997682517072284e-05, + "loss": 0.0894, + "step": 3117 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976789072777455e-05, + "loss": 0.0966, + "step": 3118 + }, + { + "epoch": 0.41, + "learning_rate": 1.997675294677289e-05, + "loss": 0.1067, + "step": 3119 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976716792709233e-05, + "loss": 0.0978, + "step": 3120 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976680610586593e-05, + "loss": 0.1062, + "step": 3121 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976644400405066e-05, + "loss": 0.0889, + "step": 3122 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976608162164766e-05, + "loss": 0.0878, + "step": 3123 + }, + { + "epoch": 0.41, + "learning_rate": 1.997657189586578e-05, + "loss": 0.112, + "step": 3124 + }, + { + "epoch": 0.41, + "learning_rate": 1.997653560150822e-05, + "loss": 0.0992, + "step": 3125 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976499279092184e-05, + "loss": 0.0892, + "step": 3126 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976462928617775e-05, + "loss": 0.1007, + "step": 3127 + }, + { + "epoch": 0.41, + "learning_rate": 1.99764265500851e-05, + "loss": 0.0974, + "step": 3128 + }, + { + "epoch": 0.41, + "learning_rate": 1.997639014349425e-05, + "loss": 0.0924, + "step": 3129 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976353708845335e-05, + "loss": 0.1019, + "step": 3130 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976317246138457e-05, + "loss": 0.0933, + "step": 3131 + }, + { + "epoch": 0.41, + "learning_rate": 1.997628075537372e-05, + "loss": 0.0909, + "step": 3132 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976244236551224e-05, + "loss": 0.0929, + "step": 3133 + }, + { + "epoch": 0.41, + "learning_rate": 1.997620768967107e-05, + "loss": 0.1009, + "step": 3134 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976171114733365e-05, + "loss": 0.097, + "step": 3135 + }, + { + "epoch": 0.41, + "learning_rate": 1.997613451173821e-05, + "loss": 0.0945, + "step": 3136 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976097880685707e-05, + "loss": 0.097, + "step": 3137 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976061221575962e-05, + "loss": 0.0927, + "step": 3138 + }, + { + "epoch": 0.41, + "learning_rate": 1.9976024534409074e-05, + "loss": 0.1109, + "step": 3139 + }, + { + "epoch": 0.41, + "learning_rate": 1.997598781918515e-05, + "loss": 0.1071, + "step": 3140 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975951075904292e-05, + "loss": 0.0959, + "step": 3141 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975914304566598e-05, + "loss": 0.0999, + "step": 3142 + }, + { + "epoch": 0.41, + "learning_rate": 1.997587750517218e-05, + "loss": 0.0942, + "step": 3143 + }, + { + "epoch": 0.41, + "learning_rate": 1.997584067772114e-05, + "loss": 0.0999, + "step": 3144 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975803822213575e-05, + "loss": 0.1065, + "step": 3145 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975766938649596e-05, + "loss": 0.1043, + "step": 3146 + }, + { + "epoch": 0.41, + "learning_rate": 1.99757300270293e-05, + "loss": 0.0997, + "step": 3147 + }, + { + "epoch": 0.41, + "learning_rate": 1.99756930873528e-05, + "loss": 0.1041, + "step": 3148 + }, + { + "epoch": 0.41, + "learning_rate": 1.997565611962019e-05, + "loss": 0.103, + "step": 3149 + }, + { + "epoch": 0.41, + "learning_rate": 1.997561912383158e-05, + "loss": 0.1085, + "step": 3150 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975582099987072e-05, + "loss": 0.0936, + "step": 3151 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975545048086774e-05, + "loss": 0.1117, + "step": 3152 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975507968130782e-05, + "loss": 0.0988, + "step": 3153 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975470860119207e-05, + "loss": 0.0914, + "step": 3154 + }, + { + "epoch": 0.41, + "learning_rate": 1.997543372405215e-05, + "loss": 0.094, + "step": 3155 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975396559929718e-05, + "loss": 0.1008, + "step": 3156 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975359367752015e-05, + "loss": 0.1097, + "step": 3157 + }, + { + "epoch": 0.41, + "learning_rate": 1.997532214751914e-05, + "loss": 0.0924, + "step": 3158 + }, + { + "epoch": 0.41, + "learning_rate": 1.997528489923121e-05, + "loss": 0.1123, + "step": 3159 + }, + { + "epoch": 0.41, + "learning_rate": 1.997524762288832e-05, + "loss": 0.0991, + "step": 3160 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975210318490577e-05, + "loss": 0.1153, + "step": 3161 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975172986038086e-05, + "loss": 0.0922, + "step": 3162 + }, + { + "epoch": 0.41, + "learning_rate": 1.9975135625530952e-05, + "loss": 0.1029, + "step": 3163 + }, + { + "epoch": 0.41, + "learning_rate": 1.997509823696928e-05, + "loss": 0.1024, + "step": 3164 + }, + { + "epoch": 0.41, + "learning_rate": 1.997506082035318e-05, + "loss": 0.1005, + "step": 3165 + }, + { + "epoch": 0.41, + "learning_rate": 1.997502337568275e-05, + "loss": 0.1021, + "step": 3166 + }, + { + "epoch": 0.41, + "learning_rate": 1.9974985902958094e-05, + "loss": 0.1096, + "step": 3167 + }, + { + "epoch": 0.41, + "learning_rate": 1.9974948402179323e-05, + "loss": 0.0973, + "step": 3168 + }, + { + "epoch": 0.42, + "learning_rate": 1.997491087334654e-05, + "loss": 0.0833, + "step": 3169 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974873316459854e-05, + "loss": 0.1092, + "step": 3170 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974835731519367e-05, + "loss": 0.0948, + "step": 3171 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974798118525187e-05, + "loss": 0.0962, + "step": 3172 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974760477477417e-05, + "loss": 0.0984, + "step": 3173 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974722808376163e-05, + "loss": 0.0923, + "step": 3174 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974685111221534e-05, + "loss": 0.1072, + "step": 3175 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974647386013635e-05, + "loss": 0.1124, + "step": 3176 + }, + { + "epoch": 0.42, + "learning_rate": 1.997460963275257e-05, + "loss": 0.0913, + "step": 3177 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974571851438448e-05, + "loss": 0.0992, + "step": 3178 + }, + { + "epoch": 0.42, + "learning_rate": 1.997453404207137e-05, + "loss": 0.1061, + "step": 3179 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974496204651453e-05, + "loss": 0.1047, + "step": 3180 + }, + { + "epoch": 0.42, + "learning_rate": 1.997445833917879e-05, + "loss": 0.0975, + "step": 3181 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974420445653498e-05, + "loss": 0.1053, + "step": 3182 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974382524075677e-05, + "loss": 0.0979, + "step": 3183 + }, + { + "epoch": 0.42, + "learning_rate": 1.997434457444544e-05, + "loss": 0.0965, + "step": 3184 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974306596762886e-05, + "loss": 0.1086, + "step": 3185 + }, + { + "epoch": 0.42, + "learning_rate": 1.997426859102813e-05, + "loss": 0.1039, + "step": 3186 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974230557241273e-05, + "loss": 0.1004, + "step": 3187 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974192495402422e-05, + "loss": 0.0912, + "step": 3188 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974154405511687e-05, + "loss": 0.095, + "step": 3189 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974116287569172e-05, + "loss": 0.1123, + "step": 3190 + }, + { + "epoch": 0.42, + "learning_rate": 1.997407814157499e-05, + "loss": 0.1038, + "step": 3191 + }, + { + "epoch": 0.42, + "learning_rate": 1.997403996752924e-05, + "loss": 0.0975, + "step": 3192 + }, + { + "epoch": 0.42, + "learning_rate": 1.9974001765432038e-05, + "loss": 0.0854, + "step": 3193 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973963535283485e-05, + "loss": 0.0906, + "step": 3194 + }, + { + "epoch": 0.42, + "learning_rate": 1.997392527708369e-05, + "loss": 0.0997, + "step": 3195 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973886990832762e-05, + "loss": 0.1067, + "step": 3196 + }, + { + "epoch": 0.42, + "learning_rate": 1.997384867653081e-05, + "loss": 0.0962, + "step": 3197 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973810334177936e-05, + "loss": 0.1014, + "step": 3198 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973771963774253e-05, + "loss": 0.1026, + "step": 3199 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973733565319867e-05, + "loss": 0.1083, + "step": 3200 + }, + { + "epoch": 0.42, + "learning_rate": 1.997369513881489e-05, + "loss": 0.0998, + "step": 3201 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973656684259423e-05, + "loss": 0.0994, + "step": 3202 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973618201653578e-05, + "loss": 0.1153, + "step": 3203 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973579690997465e-05, + "loss": 0.1017, + "step": 3204 + }, + { + "epoch": 0.42, + "learning_rate": 1.997354115229119e-05, + "loss": 0.0992, + "step": 3205 + }, + { + "epoch": 0.42, + "learning_rate": 1.997350258553486e-05, + "loss": 0.1077, + "step": 3206 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973463990728587e-05, + "loss": 0.1049, + "step": 3207 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973425367872473e-05, + "loss": 0.0989, + "step": 3208 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973386716966636e-05, + "loss": 0.1131, + "step": 3209 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973348038011178e-05, + "loss": 0.0935, + "step": 3210 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973309331006214e-05, + "loss": 0.0909, + "step": 3211 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973270595951845e-05, + "loss": 0.0924, + "step": 3212 + }, + { + "epoch": 0.42, + "learning_rate": 1.997323183284818e-05, + "loss": 0.1009, + "step": 3213 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973193041695337e-05, + "loss": 0.093, + "step": 3214 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973154222493415e-05, + "loss": 0.0962, + "step": 3215 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973115375242532e-05, + "loss": 0.1007, + "step": 3216 + }, + { + "epoch": 0.42, + "learning_rate": 1.997307649994279e-05, + "loss": 0.0927, + "step": 3217 + }, + { + "epoch": 0.42, + "learning_rate": 1.9973037596594305e-05, + "loss": 0.1004, + "step": 3218 + }, + { + "epoch": 0.42, + "learning_rate": 1.997299866519718e-05, + "loss": 0.1182, + "step": 3219 + }, + { + "epoch": 0.42, + "learning_rate": 1.997295970575153e-05, + "loss": 0.0826, + "step": 3220 + }, + { + "epoch": 0.42, + "learning_rate": 1.997292071825746e-05, + "loss": 0.0912, + "step": 3221 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972881702715082e-05, + "loss": 0.0839, + "step": 3222 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972842659124507e-05, + "loss": 0.0905, + "step": 3223 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972803587485838e-05, + "loss": 0.0865, + "step": 3224 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972764487799194e-05, + "loss": 0.0957, + "step": 3225 + }, + { + "epoch": 0.42, + "learning_rate": 1.997272536006468e-05, + "loss": 0.1161, + "step": 3226 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972686204282406e-05, + "loss": 0.0962, + "step": 3227 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972647020452486e-05, + "loss": 0.1013, + "step": 3228 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972607808575027e-05, + "loss": 0.1036, + "step": 3229 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972568568650136e-05, + "loss": 0.0933, + "step": 3230 + }, + { + "epoch": 0.42, + "learning_rate": 1.997252930067793e-05, + "loss": 0.1073, + "step": 3231 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972490004658517e-05, + "loss": 0.0852, + "step": 3232 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972450680592005e-05, + "loss": 0.1018, + "step": 3233 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972411328478505e-05, + "loss": 0.0999, + "step": 3234 + }, + { + "epoch": 0.42, + "learning_rate": 1.997237194831813e-05, + "loss": 0.0964, + "step": 3235 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972332540110995e-05, + "loss": 0.12, + "step": 3236 + }, + { + "epoch": 0.42, + "learning_rate": 1.99722931038572e-05, + "loss": 0.0839, + "step": 3237 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972253639556865e-05, + "loss": 0.0952, + "step": 3238 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972214147210094e-05, + "loss": 0.0983, + "step": 3239 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972174626817006e-05, + "loss": 0.0957, + "step": 3240 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972135078377707e-05, + "loss": 0.0954, + "step": 3241 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972095501892305e-05, + "loss": 0.0973, + "step": 3242 + }, + { + "epoch": 0.42, + "learning_rate": 1.997205589736092e-05, + "loss": 0.0979, + "step": 3243 + }, + { + "epoch": 0.42, + "learning_rate": 1.9972016264783656e-05, + "loss": 0.091, + "step": 3244 + }, + { + "epoch": 0.43, + "learning_rate": 1.997197660416063e-05, + "loss": 0.087, + "step": 3245 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971936915491952e-05, + "loss": 0.0924, + "step": 3246 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971897198777727e-05, + "loss": 0.0964, + "step": 3247 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971857454018076e-05, + "loss": 0.0941, + "step": 3248 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971817681213106e-05, + "loss": 0.0885, + "step": 3249 + }, + { + "epoch": 0.43, + "learning_rate": 1.997177788036293e-05, + "loss": 0.0931, + "step": 3250 + }, + { + "epoch": 0.43, + "learning_rate": 1.997173805146766e-05, + "loss": 0.1189, + "step": 3251 + }, + { + "epoch": 0.43, + "learning_rate": 1.997169819452741e-05, + "loss": 0.0989, + "step": 3252 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971658309542284e-05, + "loss": 0.0976, + "step": 3253 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971618396512406e-05, + "loss": 0.0881, + "step": 3254 + }, + { + "epoch": 0.43, + "learning_rate": 1.997157845543788e-05, + "loss": 0.1081, + "step": 3255 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971538486318817e-05, + "loss": 0.1108, + "step": 3256 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971498489155338e-05, + "loss": 0.1013, + "step": 3257 + }, + { + "epoch": 0.43, + "learning_rate": 1.997145846394755e-05, + "loss": 0.0891, + "step": 3258 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971418410695565e-05, + "loss": 0.1009, + "step": 3259 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971378329399495e-05, + "loss": 0.1107, + "step": 3260 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971338220059458e-05, + "loss": 0.1025, + "step": 3261 + }, + { + "epoch": 0.43, + "learning_rate": 1.997129808267556e-05, + "loss": 0.0945, + "step": 3262 + }, + { + "epoch": 0.43, + "learning_rate": 1.997125791724792e-05, + "loss": 0.1003, + "step": 3263 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971217723776646e-05, + "loss": 0.1048, + "step": 3264 + }, + { + "epoch": 0.43, + "learning_rate": 1.997117750226185e-05, + "loss": 0.1067, + "step": 3265 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971137252703654e-05, + "loss": 0.0827, + "step": 3266 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971096975102166e-05, + "loss": 0.1081, + "step": 3267 + }, + { + "epoch": 0.43, + "learning_rate": 1.9971056669457493e-05, + "loss": 0.1048, + "step": 3268 + }, + { + "epoch": 0.43, + "learning_rate": 1.997101633576976e-05, + "loss": 0.1049, + "step": 3269 + }, + { + "epoch": 0.43, + "learning_rate": 1.997097597403907e-05, + "loss": 0.0956, + "step": 3270 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970935584265544e-05, + "loss": 0.0929, + "step": 3271 + }, + { + "epoch": 0.43, + "learning_rate": 1.997089516644929e-05, + "loss": 0.0916, + "step": 3272 + }, + { + "epoch": 0.43, + "learning_rate": 1.997085472059043e-05, + "loss": 0.1058, + "step": 3273 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970814246689066e-05, + "loss": 0.1023, + "step": 3274 + }, + { + "epoch": 0.43, + "learning_rate": 1.997077374474532e-05, + "loss": 0.0992, + "step": 3275 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970733214759304e-05, + "loss": 0.0857, + "step": 3276 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970692656731132e-05, + "loss": 0.101, + "step": 3277 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970652070660915e-05, + "loss": 0.0943, + "step": 3278 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970611456548775e-05, + "loss": 0.0998, + "step": 3279 + }, + { + "epoch": 0.43, + "learning_rate": 1.997057081439482e-05, + "loss": 0.0984, + "step": 3280 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970530144199163e-05, + "loss": 0.0991, + "step": 3281 + }, + { + "epoch": 0.43, + "learning_rate": 1.997048944596192e-05, + "loss": 0.0918, + "step": 3282 + }, + { + "epoch": 0.43, + "learning_rate": 1.997044871968321e-05, + "loss": 0.0998, + "step": 3283 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970407965363142e-05, + "loss": 0.098, + "step": 3284 + }, + { + "epoch": 0.43, + "learning_rate": 1.997036718300183e-05, + "loss": 0.0939, + "step": 3285 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970326372599397e-05, + "loss": 0.0946, + "step": 3286 + }, + { + "epoch": 0.43, + "learning_rate": 1.997028553415595e-05, + "loss": 0.0955, + "step": 3287 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970244667671604e-05, + "loss": 0.1023, + "step": 3288 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970203773146476e-05, + "loss": 0.0778, + "step": 3289 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970162850580684e-05, + "loss": 0.1005, + "step": 3290 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970121899974337e-05, + "loss": 0.1137, + "step": 3291 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970080921327555e-05, + "loss": 0.1108, + "step": 3292 + }, + { + "epoch": 0.43, + "learning_rate": 1.9970039914640448e-05, + "loss": 0.0973, + "step": 3293 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969998879913138e-05, + "loss": 0.0938, + "step": 3294 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969957817145738e-05, + "loss": 0.0824, + "step": 3295 + }, + { + "epoch": 0.43, + "learning_rate": 1.996991672633836e-05, + "loss": 0.0987, + "step": 3296 + }, + { + "epoch": 0.43, + "learning_rate": 1.996987560749112e-05, + "loss": 0.1008, + "step": 3297 + }, + { + "epoch": 0.43, + "learning_rate": 1.996983446060414e-05, + "loss": 0.113, + "step": 3298 + }, + { + "epoch": 0.43, + "learning_rate": 1.996979328567753e-05, + "loss": 0.0959, + "step": 3299 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969752082711407e-05, + "loss": 0.1002, + "step": 3300 + }, + { + "epoch": 0.43, + "learning_rate": 1.996971085170589e-05, + "loss": 0.0864, + "step": 3301 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969669592661087e-05, + "loss": 0.0859, + "step": 3302 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969628305577123e-05, + "loss": 0.0994, + "step": 3303 + }, + { + "epoch": 0.43, + "learning_rate": 1.996958699045411e-05, + "loss": 0.1108, + "step": 3304 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969545647292162e-05, + "loss": 0.1103, + "step": 3305 + }, + { + "epoch": 0.43, + "learning_rate": 1.99695042760914e-05, + "loss": 0.0948, + "step": 3306 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969462876851934e-05, + "loss": 0.1048, + "step": 3307 + }, + { + "epoch": 0.43, + "learning_rate": 1.996942144957389e-05, + "loss": 0.1033, + "step": 3308 + }, + { + "epoch": 0.43, + "learning_rate": 1.996937999425738e-05, + "loss": 0.0996, + "step": 3309 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969338510902518e-05, + "loss": 0.0998, + "step": 3310 + }, + { + "epoch": 0.43, + "learning_rate": 1.996929699950942e-05, + "loss": 0.1062, + "step": 3311 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969255460078206e-05, + "loss": 0.0942, + "step": 3312 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969213892608992e-05, + "loss": 0.1021, + "step": 3313 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969172297101894e-05, + "loss": 0.1052, + "step": 3314 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969130673557032e-05, + "loss": 0.093, + "step": 3315 + }, + { + "epoch": 0.43, + "learning_rate": 1.996908902197452e-05, + "loss": 0.103, + "step": 3316 + }, + { + "epoch": 0.43, + "learning_rate": 1.996904734235447e-05, + "loss": 0.1054, + "step": 3317 + }, + { + "epoch": 0.43, + "learning_rate": 1.9969005634697015e-05, + "loss": 0.0966, + "step": 3318 + }, + { + "epoch": 0.43, + "learning_rate": 1.9968963899002258e-05, + "loss": 0.1002, + "step": 3319 + }, + { + "epoch": 0.43, + "learning_rate": 1.996892213527032e-05, + "loss": 0.1146, + "step": 3320 + }, + { + "epoch": 0.44, + "learning_rate": 1.996888034350132e-05, + "loss": 0.0965, + "step": 3321 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968838523695377e-05, + "loss": 0.0966, + "step": 3322 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968796675852603e-05, + "loss": 0.0855, + "step": 3323 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968754799973123e-05, + "loss": 0.0777, + "step": 3324 + }, + { + "epoch": 0.44, + "learning_rate": 1.996871289605705e-05, + "loss": 0.1084, + "step": 3325 + }, + { + "epoch": 0.44, + "learning_rate": 1.99686709641045e-05, + "loss": 0.0949, + "step": 3326 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968629004115597e-05, + "loss": 0.0901, + "step": 3327 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968587016090456e-05, + "loss": 0.0934, + "step": 3328 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968545000029192e-05, + "loss": 0.093, + "step": 3329 + }, + { + "epoch": 0.44, + "learning_rate": 1.996850295593193e-05, + "loss": 0.0903, + "step": 3330 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968460883798783e-05, + "loss": 0.102, + "step": 3331 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968418783629867e-05, + "loss": 0.0975, + "step": 3332 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968376655425307e-05, + "loss": 0.1083, + "step": 3333 + }, + { + "epoch": 0.44, + "learning_rate": 1.996833449918522e-05, + "loss": 0.1097, + "step": 3334 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968292314909717e-05, + "loss": 0.1098, + "step": 3335 + }, + { + "epoch": 0.44, + "learning_rate": 1.996825010259893e-05, + "loss": 0.0972, + "step": 3336 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968207862252966e-05, + "loss": 0.1043, + "step": 3337 + }, + { + "epoch": 0.44, + "learning_rate": 1.996816559387195e-05, + "loss": 0.0927, + "step": 3338 + }, + { + "epoch": 0.44, + "learning_rate": 1.9968123297455998e-05, + "loss": 0.091, + "step": 3339 + }, + { + "epoch": 0.44, + "learning_rate": 1.996808097300523e-05, + "loss": 0.1008, + "step": 3340 + }, + { + "epoch": 0.44, + "learning_rate": 1.996803862051977e-05, + "loss": 0.0958, + "step": 3341 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967996239999726e-05, + "loss": 0.0988, + "step": 3342 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967953831445224e-05, + "loss": 0.0954, + "step": 3343 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967911394856385e-05, + "loss": 0.1022, + "step": 3344 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967868930233326e-05, + "loss": 0.0935, + "step": 3345 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967826437576163e-05, + "loss": 0.0971, + "step": 3346 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967783916885023e-05, + "loss": 0.0902, + "step": 3347 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967741368160022e-05, + "loss": 0.0991, + "step": 3348 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967698791401277e-05, + "loss": 0.0966, + "step": 3349 + }, + { + "epoch": 0.44, + "learning_rate": 1.996765618660891e-05, + "loss": 0.0975, + "step": 3350 + }, + { + "epoch": 0.44, + "learning_rate": 1.996761355378304e-05, + "loss": 0.0957, + "step": 3351 + }, + { + "epoch": 0.44, + "learning_rate": 1.996757089292379e-05, + "loss": 0.1142, + "step": 3352 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967528204031275e-05, + "loss": 0.0955, + "step": 3353 + }, + { + "epoch": 0.44, + "learning_rate": 1.996748548710562e-05, + "loss": 0.0923, + "step": 3354 + }, + { + "epoch": 0.44, + "learning_rate": 1.996744274214694e-05, + "loss": 0.1102, + "step": 3355 + }, + { + "epoch": 0.44, + "learning_rate": 1.996739996915536e-05, + "loss": 0.093, + "step": 3356 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967357168131e-05, + "loss": 0.0937, + "step": 3357 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967314339073976e-05, + "loss": 0.0941, + "step": 3358 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967271481984412e-05, + "loss": 0.1188, + "step": 3359 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967228596862428e-05, + "loss": 0.1079, + "step": 3360 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967185683708146e-05, + "loss": 0.0942, + "step": 3361 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967142742521684e-05, + "loss": 0.0946, + "step": 3362 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967099773303162e-05, + "loss": 0.1008, + "step": 3363 + }, + { + "epoch": 0.44, + "learning_rate": 1.9967056776052705e-05, + "loss": 0.0952, + "step": 3364 + }, + { + "epoch": 0.44, + "learning_rate": 1.996701375077043e-05, + "loss": 0.0824, + "step": 3365 + }, + { + "epoch": 0.44, + "learning_rate": 1.996697069745646e-05, + "loss": 0.1026, + "step": 3366 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966927616110916e-05, + "loss": 0.0916, + "step": 3367 + }, + { + "epoch": 0.44, + "learning_rate": 1.996688450673392e-05, + "loss": 0.0945, + "step": 3368 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966841369325588e-05, + "loss": 0.1091, + "step": 3369 + }, + { + "epoch": 0.44, + "learning_rate": 1.996679820388605e-05, + "loss": 0.1095, + "step": 3370 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966755010415422e-05, + "loss": 0.1072, + "step": 3371 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966711788913825e-05, + "loss": 0.0995, + "step": 3372 + }, + { + "epoch": 0.44, + "learning_rate": 1.996666853938138e-05, + "loss": 0.0907, + "step": 3373 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966625261818214e-05, + "loss": 0.1007, + "step": 3374 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966581956224447e-05, + "loss": 0.0977, + "step": 3375 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966538622600194e-05, + "loss": 0.0908, + "step": 3376 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966495260945585e-05, + "loss": 0.1066, + "step": 3377 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966451871260738e-05, + "loss": 0.0991, + "step": 3378 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966408453545775e-05, + "loss": 0.1055, + "step": 3379 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966365007800818e-05, + "loss": 0.0917, + "step": 3380 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966321534025992e-05, + "loss": 0.1034, + "step": 3381 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966278032221417e-05, + "loss": 0.1068, + "step": 3382 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966234502387216e-05, + "loss": 0.09, + "step": 3383 + }, + { + "epoch": 0.44, + "learning_rate": 1.996619094452351e-05, + "loss": 0.0989, + "step": 3384 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966147358630423e-05, + "loss": 0.088, + "step": 3385 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966103744708077e-05, + "loss": 0.0965, + "step": 3386 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966060102756598e-05, + "loss": 0.0938, + "step": 3387 + }, + { + "epoch": 0.44, + "learning_rate": 1.9966016432776104e-05, + "loss": 0.1041, + "step": 3388 + }, + { + "epoch": 0.44, + "learning_rate": 1.996597273476672e-05, + "loss": 0.0912, + "step": 3389 + }, + { + "epoch": 0.44, + "learning_rate": 1.9965929008728567e-05, + "loss": 0.0856, + "step": 3390 + }, + { + "epoch": 0.44, + "learning_rate": 1.996588525466177e-05, + "loss": 0.0928, + "step": 3391 + }, + { + "epoch": 0.44, + "learning_rate": 1.996584147256645e-05, + "loss": 0.1009, + "step": 3392 + }, + { + "epoch": 0.44, + "learning_rate": 1.9965797662442733e-05, + "loss": 0.1014, + "step": 3393 + }, + { + "epoch": 0.44, + "learning_rate": 1.9965753824290742e-05, + "loss": 0.1042, + "step": 3394 + }, + { + "epoch": 0.44, + "learning_rate": 1.9965709958110597e-05, + "loss": 0.1084, + "step": 3395 + }, + { + "epoch": 0.44, + "learning_rate": 1.9965666063902423e-05, + "loss": 0.0912, + "step": 3396 + }, + { + "epoch": 0.44, + "learning_rate": 1.9965622141666346e-05, + "loss": 0.0978, + "step": 3397 + }, + { + "epoch": 0.45, + "learning_rate": 1.9965578191402487e-05, + "loss": 0.0908, + "step": 3398 + }, + { + "epoch": 0.45, + "learning_rate": 1.996553421311097e-05, + "loss": 0.1118, + "step": 3399 + }, + { + "epoch": 0.45, + "learning_rate": 1.9965490206791916e-05, + "loss": 0.1005, + "step": 3400 + }, + { + "epoch": 0.45, + "learning_rate": 1.9965446172445455e-05, + "loss": 0.0964, + "step": 3401 + }, + { + "epoch": 0.45, + "learning_rate": 1.9965402110071707e-05, + "loss": 0.1019, + "step": 3402 + }, + { + "epoch": 0.45, + "learning_rate": 1.9965358019670794e-05, + "loss": 0.107, + "step": 3403 + }, + { + "epoch": 0.45, + "learning_rate": 1.9965313901242848e-05, + "loss": 0.0945, + "step": 3404 + }, + { + "epoch": 0.45, + "learning_rate": 1.996526975478798e-05, + "loss": 0.1035, + "step": 3405 + }, + { + "epoch": 0.45, + "learning_rate": 1.9965225580306326e-05, + "loss": 0.0932, + "step": 3406 + }, + { + "epoch": 0.45, + "learning_rate": 1.9965181377798e-05, + "loss": 0.0818, + "step": 3407 + }, + { + "epoch": 0.45, + "learning_rate": 1.996513714726314e-05, + "loss": 0.0897, + "step": 3408 + }, + { + "epoch": 0.45, + "learning_rate": 1.9965092888701857e-05, + "loss": 0.0879, + "step": 3409 + }, + { + "epoch": 0.45, + "learning_rate": 1.9965048602114283e-05, + "loss": 0.0918, + "step": 3410 + }, + { + "epoch": 0.45, + "learning_rate": 1.996500428750054e-05, + "loss": 0.0945, + "step": 3411 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964959944860754e-05, + "loss": 0.1026, + "step": 3412 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964915574195046e-05, + "loss": 0.088, + "step": 3413 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964871175503546e-05, + "loss": 0.0936, + "step": 3414 + }, + { + "epoch": 0.45, + "learning_rate": 1.996482674878638e-05, + "loss": 0.1061, + "step": 3415 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964782294043662e-05, + "loss": 0.1017, + "step": 3416 + }, + { + "epoch": 0.45, + "learning_rate": 1.996473781127553e-05, + "loss": 0.0801, + "step": 3417 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964693300482102e-05, + "loss": 0.1036, + "step": 3418 + }, + { + "epoch": 0.45, + "learning_rate": 1.99646487616635e-05, + "loss": 0.0916, + "step": 3419 + }, + { + "epoch": 0.45, + "learning_rate": 1.996460419481986e-05, + "loss": 0.0946, + "step": 3420 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964559599951302e-05, + "loss": 0.0967, + "step": 3421 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964514977057946e-05, + "loss": 0.0898, + "step": 3422 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964470326139926e-05, + "loss": 0.0961, + "step": 3423 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964425647197363e-05, + "loss": 0.0889, + "step": 3424 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964380940230383e-05, + "loss": 0.0956, + "step": 3425 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964336205239112e-05, + "loss": 0.1003, + "step": 3426 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964291442223674e-05, + "loss": 0.0898, + "step": 3427 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964246651184202e-05, + "loss": 0.0856, + "step": 3428 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964201832120814e-05, + "loss": 0.0864, + "step": 3429 + }, + { + "epoch": 0.45, + "learning_rate": 1.996415698503364e-05, + "loss": 0.0969, + "step": 3430 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964112109922803e-05, + "loss": 0.0984, + "step": 3431 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964067206788434e-05, + "loss": 0.0897, + "step": 3432 + }, + { + "epoch": 0.45, + "learning_rate": 1.9964022275630654e-05, + "loss": 0.0912, + "step": 3433 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963977316449596e-05, + "loss": 0.0844, + "step": 3434 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963932329245377e-05, + "loss": 0.0944, + "step": 3435 + }, + { + "epoch": 0.45, + "learning_rate": 1.996388731401813e-05, + "loss": 0.0999, + "step": 3436 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963842270767985e-05, + "loss": 0.0961, + "step": 3437 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963797199495062e-05, + "loss": 0.1001, + "step": 3438 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963752100199487e-05, + "loss": 0.0929, + "step": 3439 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963706972881392e-05, + "loss": 0.0921, + "step": 3440 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963661817540903e-05, + "loss": 0.0972, + "step": 3441 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963616634178145e-05, + "loss": 0.1, + "step": 3442 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963571422793243e-05, + "loss": 0.0971, + "step": 3443 + }, + { + "epoch": 0.45, + "learning_rate": 1.996352618338633e-05, + "loss": 0.0988, + "step": 3444 + }, + { + "epoch": 0.45, + "learning_rate": 1.996348091595753e-05, + "loss": 0.0969, + "step": 3445 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963435620506967e-05, + "loss": 0.1061, + "step": 3446 + }, + { + "epoch": 0.45, + "learning_rate": 1.996339029703478e-05, + "loss": 0.1088, + "step": 3447 + }, + { + "epoch": 0.45, + "learning_rate": 1.996334494554108e-05, + "loss": 0.0946, + "step": 3448 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963299566026006e-05, + "loss": 0.0971, + "step": 3449 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963254158489683e-05, + "loss": 0.0943, + "step": 3450 + }, + { + "epoch": 0.45, + "learning_rate": 1.996320872293224e-05, + "loss": 0.0985, + "step": 3451 + }, + { + "epoch": 0.45, + "learning_rate": 1.99631632593538e-05, + "loss": 0.1021, + "step": 3452 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963117767754494e-05, + "loss": 0.1011, + "step": 3453 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963072248134452e-05, + "loss": 0.091, + "step": 3454 + }, + { + "epoch": 0.45, + "learning_rate": 1.9963026700493797e-05, + "loss": 0.0951, + "step": 3455 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962981124832663e-05, + "loss": 0.1034, + "step": 3456 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962935521151174e-05, + "loss": 0.1073, + "step": 3457 + }, + { + "epoch": 0.45, + "learning_rate": 1.996288988944946e-05, + "loss": 0.096, + "step": 3458 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962844229727646e-05, + "loss": 0.0876, + "step": 3459 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962798541985865e-05, + "loss": 0.084, + "step": 3460 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962752826224245e-05, + "loss": 0.0894, + "step": 3461 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962707082442913e-05, + "loss": 0.0912, + "step": 3462 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962661310641996e-05, + "loss": 0.0926, + "step": 3463 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962615510821625e-05, + "loss": 0.0881, + "step": 3464 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962569682981926e-05, + "loss": 0.0932, + "step": 3465 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962523827123035e-05, + "loss": 0.0961, + "step": 3466 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962477943245076e-05, + "loss": 0.1049, + "step": 3467 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962432031348176e-05, + "loss": 0.1023, + "step": 3468 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962386091432466e-05, + "loss": 0.0961, + "step": 3469 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962340123498078e-05, + "loss": 0.1068, + "step": 3470 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962294127545138e-05, + "loss": 0.102, + "step": 3471 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962248103573777e-05, + "loss": 0.0927, + "step": 3472 + }, + { + "epoch": 0.45, + "learning_rate": 1.9962202051584122e-05, + "loss": 0.1147, + "step": 3473 + }, + { + "epoch": 0.46, + "learning_rate": 1.9962155971576305e-05, + "loss": 0.1005, + "step": 3474 + }, + { + "epoch": 0.46, + "learning_rate": 1.9962109863550453e-05, + "loss": 0.1144, + "step": 3475 + }, + { + "epoch": 0.46, + "learning_rate": 1.99620637275067e-05, + "loss": 0.0832, + "step": 3476 + }, + { + "epoch": 0.46, + "learning_rate": 1.9962017563445174e-05, + "loss": 0.0912, + "step": 3477 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961971371366004e-05, + "loss": 0.0972, + "step": 3478 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961925151269318e-05, + "loss": 0.0937, + "step": 3479 + }, + { + "epoch": 0.46, + "learning_rate": 1.996187890315525e-05, + "loss": 0.1161, + "step": 3480 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961832627023925e-05, + "loss": 0.0955, + "step": 3481 + }, + { + "epoch": 0.46, + "learning_rate": 1.996178632287548e-05, + "loss": 0.1011, + "step": 3482 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961739990710036e-05, + "loss": 0.1041, + "step": 3483 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961693630527734e-05, + "loss": 0.091, + "step": 3484 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961647242328697e-05, + "loss": 0.0886, + "step": 3485 + }, + { + "epoch": 0.46, + "learning_rate": 1.996160082611306e-05, + "loss": 0.0975, + "step": 3486 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961554381880948e-05, + "loss": 0.0974, + "step": 3487 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961507909632495e-05, + "loss": 0.0931, + "step": 3488 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961461409367834e-05, + "loss": 0.0916, + "step": 3489 + }, + { + "epoch": 0.46, + "learning_rate": 1.996141488108709e-05, + "loss": 0.1009, + "step": 3490 + }, + { + "epoch": 0.46, + "learning_rate": 1.99613683247904e-05, + "loss": 0.0923, + "step": 3491 + }, + { + "epoch": 0.46, + "learning_rate": 1.996132174047789e-05, + "loss": 0.0954, + "step": 3492 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961275128149698e-05, + "loss": 0.0902, + "step": 3493 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961228487805943e-05, + "loss": 0.1022, + "step": 3494 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961181819446767e-05, + "loss": 0.0864, + "step": 3495 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961135123072298e-05, + "loss": 0.0909, + "step": 3496 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961088398682665e-05, + "loss": 0.0954, + "step": 3497 + }, + { + "epoch": 0.46, + "learning_rate": 1.9961041646278003e-05, + "loss": 0.1069, + "step": 3498 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960994865858442e-05, + "loss": 0.0983, + "step": 3499 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960948057424113e-05, + "loss": 0.1013, + "step": 3500 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960901220975148e-05, + "loss": 0.089, + "step": 3501 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960854356511676e-05, + "loss": 0.0946, + "step": 3502 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960807464033837e-05, + "loss": 0.1124, + "step": 3503 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960760543541755e-05, + "loss": 0.0812, + "step": 3504 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960713595035562e-05, + "loss": 0.0978, + "step": 3505 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960666618515395e-05, + "loss": 0.0979, + "step": 3506 + }, + { + "epoch": 0.46, + "learning_rate": 1.996061961398138e-05, + "loss": 0.0908, + "step": 3507 + }, + { + "epoch": 0.46, + "learning_rate": 1.996057258143366e-05, + "loss": 0.0977, + "step": 3508 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960525520872355e-05, + "loss": 0.0869, + "step": 3509 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960478432297603e-05, + "loss": 0.1035, + "step": 3510 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960431315709536e-05, + "loss": 0.1024, + "step": 3511 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960384171108288e-05, + "loss": 0.0992, + "step": 3512 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960336998493986e-05, + "loss": 0.1047, + "step": 3513 + }, + { + "epoch": 0.46, + "learning_rate": 1.996028979786677e-05, + "loss": 0.1157, + "step": 3514 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960242569226768e-05, + "loss": 0.1037, + "step": 3515 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960195312574116e-05, + "loss": 0.0941, + "step": 3516 + }, + { + "epoch": 0.46, + "learning_rate": 1.996014802790894e-05, + "loss": 0.0983, + "step": 3517 + }, + { + "epoch": 0.46, + "learning_rate": 1.9960100715231384e-05, + "loss": 0.0968, + "step": 3518 + }, + { + "epoch": 0.46, + "learning_rate": 1.996005337454157e-05, + "loss": 0.0946, + "step": 3519 + }, + { + "epoch": 0.46, + "learning_rate": 1.996000600583964e-05, + "loss": 0.0926, + "step": 3520 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959958609125723e-05, + "loss": 0.108, + "step": 3521 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959911184399955e-05, + "loss": 0.0978, + "step": 3522 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959863731662463e-05, + "loss": 0.097, + "step": 3523 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959816250913387e-05, + "loss": 0.095, + "step": 3524 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959768742152857e-05, + "loss": 0.1076, + "step": 3525 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959721205381006e-05, + "loss": 0.1036, + "step": 3526 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959673640597975e-05, + "loss": 0.0929, + "step": 3527 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959626047803886e-05, + "loss": 0.0881, + "step": 3528 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959578426998885e-05, + "loss": 0.0934, + "step": 3529 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959530778183097e-05, + "loss": 0.0811, + "step": 3530 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959483101356658e-05, + "loss": 0.0913, + "step": 3531 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959435396519704e-05, + "loss": 0.1003, + "step": 3532 + }, + { + "epoch": 0.46, + "learning_rate": 1.995938766367237e-05, + "loss": 0.089, + "step": 3533 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959339902814787e-05, + "loss": 0.1106, + "step": 3534 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959292113947092e-05, + "loss": 0.0846, + "step": 3535 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959244297069417e-05, + "loss": 0.0951, + "step": 3536 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959196452181897e-05, + "loss": 0.1036, + "step": 3537 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959148579284672e-05, + "loss": 0.101, + "step": 3538 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959100678377866e-05, + "loss": 0.0929, + "step": 3539 + }, + { + "epoch": 0.46, + "learning_rate": 1.995905274946162e-05, + "loss": 0.1071, + "step": 3540 + }, + { + "epoch": 0.46, + "learning_rate": 1.9959004792536075e-05, + "loss": 0.1084, + "step": 3541 + }, + { + "epoch": 0.46, + "learning_rate": 1.9958956807601354e-05, + "loss": 0.0828, + "step": 3542 + }, + { + "epoch": 0.46, + "learning_rate": 1.9958908794657598e-05, + "loss": 0.1015, + "step": 3543 + }, + { + "epoch": 0.46, + "learning_rate": 1.995886075370494e-05, + "loss": 0.0915, + "step": 3544 + }, + { + "epoch": 0.46, + "learning_rate": 1.995881268474352e-05, + "loss": 0.0977, + "step": 3545 + }, + { + "epoch": 0.46, + "learning_rate": 1.9958764587773466e-05, + "loss": 0.0905, + "step": 3546 + }, + { + "epoch": 0.46, + "learning_rate": 1.995871646279492e-05, + "loss": 0.0987, + "step": 3547 + }, + { + "epoch": 0.46, + "learning_rate": 1.9958668309808015e-05, + "loss": 0.0897, + "step": 3548 + }, + { + "epoch": 0.46, + "learning_rate": 1.995862012881288e-05, + "loss": 0.1024, + "step": 3549 + }, + { + "epoch": 0.47, + "learning_rate": 1.9958571919809662e-05, + "loss": 0.0856, + "step": 3550 + }, + { + "epoch": 0.47, + "learning_rate": 1.995852368279849e-05, + "loss": 0.0884, + "step": 3551 + }, + { + "epoch": 0.47, + "learning_rate": 1.9958475417779496e-05, + "loss": 0.1024, + "step": 3552 + }, + { + "epoch": 0.47, + "learning_rate": 1.9958427124752828e-05, + "loss": 0.0886, + "step": 3553 + }, + { + "epoch": 0.47, + "learning_rate": 1.9958378803718608e-05, + "loss": 0.1072, + "step": 3554 + }, + { + "epoch": 0.47, + "learning_rate": 1.995833045467698e-05, + "loss": 0.0962, + "step": 3555 + }, + { + "epoch": 0.47, + "learning_rate": 1.9958282077628075e-05, + "loss": 0.0958, + "step": 3556 + }, + { + "epoch": 0.47, + "learning_rate": 1.9958233672572036e-05, + "loss": 0.0921, + "step": 3557 + }, + { + "epoch": 0.47, + "learning_rate": 1.995818523950899e-05, + "loss": 0.1087, + "step": 3558 + }, + { + "epoch": 0.47, + "learning_rate": 1.9958136778439084e-05, + "loss": 0.1002, + "step": 3559 + }, + { + "epoch": 0.47, + "learning_rate": 1.995808828936245e-05, + "loss": 0.1007, + "step": 3560 + }, + { + "epoch": 0.47, + "learning_rate": 1.9958039772279218e-05, + "loss": 0.0971, + "step": 3561 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957991227189527e-05, + "loss": 0.0977, + "step": 3562 + }, + { + "epoch": 0.47, + "learning_rate": 1.995794265409352e-05, + "loss": 0.0912, + "step": 3563 + }, + { + "epoch": 0.47, + "learning_rate": 1.995789405299133e-05, + "loss": 0.0949, + "step": 3564 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957845423883097e-05, + "loss": 0.1061, + "step": 3565 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957796766768945e-05, + "loss": 0.0995, + "step": 3566 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957748081649028e-05, + "loss": 0.1031, + "step": 3567 + }, + { + "epoch": 0.47, + "learning_rate": 1.995769936852347e-05, + "loss": 0.1069, + "step": 3568 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957650627392415e-05, + "loss": 0.0905, + "step": 3569 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957601858255998e-05, + "loss": 0.0902, + "step": 3570 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957553061114354e-05, + "loss": 0.1017, + "step": 3571 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957504235967625e-05, + "loss": 0.1086, + "step": 3572 + }, + { + "epoch": 0.47, + "learning_rate": 1.995745538281594e-05, + "loss": 0.1069, + "step": 3573 + }, + { + "epoch": 0.47, + "learning_rate": 1.995740650165945e-05, + "loss": 0.0977, + "step": 3574 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957357592498278e-05, + "loss": 0.0972, + "step": 3575 + }, + { + "epoch": 0.47, + "learning_rate": 1.995730865533257e-05, + "loss": 0.0969, + "step": 3576 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957259690162463e-05, + "loss": 0.0901, + "step": 3577 + }, + { + "epoch": 0.47, + "learning_rate": 1.995721069698809e-05, + "loss": 0.0852, + "step": 3578 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957161675809597e-05, + "loss": 0.1031, + "step": 3579 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957112626627112e-05, + "loss": 0.0913, + "step": 3580 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957063549440784e-05, + "loss": 0.0855, + "step": 3581 + }, + { + "epoch": 0.47, + "learning_rate": 1.9957014444250743e-05, + "loss": 0.0916, + "step": 3582 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956965311057125e-05, + "loss": 0.1042, + "step": 3583 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956916149860073e-05, + "loss": 0.1026, + "step": 3584 + }, + { + "epoch": 0.47, + "learning_rate": 1.995686696065973e-05, + "loss": 0.0953, + "step": 3585 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956817743456225e-05, + "loss": 0.101, + "step": 3586 + }, + { + "epoch": 0.47, + "learning_rate": 1.99567684982497e-05, + "loss": 0.0952, + "step": 3587 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956719225040296e-05, + "loss": 0.0975, + "step": 3588 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956669923828146e-05, + "loss": 0.1067, + "step": 3589 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956620594613395e-05, + "loss": 0.1076, + "step": 3590 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956571237396178e-05, + "loss": 0.1034, + "step": 3591 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956521852176635e-05, + "loss": 0.1063, + "step": 3592 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956472438954907e-05, + "loss": 0.1065, + "step": 3593 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956422997731128e-05, + "loss": 0.0958, + "step": 3594 + }, + { + "epoch": 0.47, + "learning_rate": 1.995637352850544e-05, + "loss": 0.0913, + "step": 3595 + }, + { + "epoch": 0.47, + "learning_rate": 1.995632403127798e-05, + "loss": 0.1034, + "step": 3596 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956274506048887e-05, + "loss": 0.1006, + "step": 3597 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956224952818305e-05, + "loss": 0.0932, + "step": 3598 + }, + { + "epoch": 0.47, + "learning_rate": 1.995617537158637e-05, + "loss": 0.1042, + "step": 3599 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956125762353222e-05, + "loss": 0.0965, + "step": 3600 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956076125119002e-05, + "loss": 0.105, + "step": 3601 + }, + { + "epoch": 0.47, + "learning_rate": 1.9956026459883845e-05, + "loss": 0.0804, + "step": 3602 + }, + { + "epoch": 0.47, + "learning_rate": 1.99559767666479e-05, + "loss": 0.0978, + "step": 3603 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955927045411293e-05, + "loss": 0.0993, + "step": 3604 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955877296174174e-05, + "loss": 0.0957, + "step": 3605 + }, + { + "epoch": 0.47, + "learning_rate": 1.995582751893668e-05, + "loss": 0.0903, + "step": 3606 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955777713698953e-05, + "loss": 0.0975, + "step": 3607 + }, + { + "epoch": 0.47, + "learning_rate": 1.995572788046113e-05, + "loss": 0.1009, + "step": 3608 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955678019223354e-05, + "loss": 0.1058, + "step": 3609 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955628129985763e-05, + "loss": 0.0981, + "step": 3610 + }, + { + "epoch": 0.47, + "learning_rate": 1.99555782127485e-05, + "loss": 0.095, + "step": 3611 + }, + { + "epoch": 0.47, + "learning_rate": 1.99555282675117e-05, + "loss": 0.0979, + "step": 3612 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955478294275508e-05, + "loss": 0.0876, + "step": 3613 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955428293040066e-05, + "loss": 0.0934, + "step": 3614 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955378263805506e-05, + "loss": 0.1045, + "step": 3615 + }, + { + "epoch": 0.47, + "learning_rate": 1.995532820657198e-05, + "loss": 0.0866, + "step": 3616 + }, + { + "epoch": 0.47, + "learning_rate": 1.995527812133962e-05, + "loss": 0.1001, + "step": 3617 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955228008108577e-05, + "loss": 0.086, + "step": 3618 + }, + { + "epoch": 0.47, + "learning_rate": 1.995517786687898e-05, + "loss": 0.0921, + "step": 3619 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955127697650974e-05, + "loss": 0.099, + "step": 3620 + }, + { + "epoch": 0.47, + "learning_rate": 1.9955077500424706e-05, + "loss": 0.1081, + "step": 3621 + }, + { + "epoch": 0.47, + "learning_rate": 1.995502727520031e-05, + "loss": 0.0865, + "step": 3622 + }, + { + "epoch": 0.47, + "learning_rate": 1.9954977021977928e-05, + "loss": 0.1026, + "step": 3623 + }, + { + "epoch": 0.47, + "learning_rate": 1.9954926740757704e-05, + "loss": 0.1044, + "step": 3624 + }, + { + "epoch": 0.47, + "learning_rate": 1.9954876431539782e-05, + "loss": 0.0874, + "step": 3625 + }, + { + "epoch": 0.47, + "learning_rate": 1.99548260943243e-05, + "loss": 0.0999, + "step": 3626 + }, + { + "epoch": 0.48, + "learning_rate": 1.9954775729111398e-05, + "loss": 0.0888, + "step": 3627 + }, + { + "epoch": 0.48, + "learning_rate": 1.995472533590122e-05, + "loss": 0.0932, + "step": 3628 + }, + { + "epoch": 0.48, + "learning_rate": 1.9954674914693905e-05, + "loss": 0.1008, + "step": 3629 + }, + { + "epoch": 0.48, + "learning_rate": 1.99546244654896e-05, + "loss": 0.0954, + "step": 3630 + }, + { + "epoch": 0.48, + "learning_rate": 1.995457398828844e-05, + "loss": 0.0952, + "step": 3631 + }, + { + "epoch": 0.48, + "learning_rate": 1.9954523483090574e-05, + "loss": 0.0861, + "step": 3632 + }, + { + "epoch": 0.48, + "learning_rate": 1.995447294989614e-05, + "loss": 0.0987, + "step": 3633 + }, + { + "epoch": 0.48, + "learning_rate": 1.995442238870528e-05, + "loss": 0.1017, + "step": 3634 + }, + { + "epoch": 0.48, + "learning_rate": 1.995437179951814e-05, + "loss": 0.0968, + "step": 3635 + }, + { + "epoch": 0.48, + "learning_rate": 1.995432118233486e-05, + "loss": 0.0962, + "step": 3636 + }, + { + "epoch": 0.48, + "learning_rate": 1.9954270537155577e-05, + "loss": 0.0942, + "step": 3637 + }, + { + "epoch": 0.48, + "learning_rate": 1.995421986398044e-05, + "loss": 0.0852, + "step": 3638 + }, + { + "epoch": 0.48, + "learning_rate": 1.995416916280959e-05, + "loss": 0.1052, + "step": 3639 + }, + { + "epoch": 0.48, + "learning_rate": 1.995411843364317e-05, + "loss": 0.1079, + "step": 3640 + }, + { + "epoch": 0.48, + "learning_rate": 1.995406767648132e-05, + "loss": 0.1096, + "step": 3641 + }, + { + "epoch": 0.48, + "learning_rate": 1.9954016891324188e-05, + "loss": 0.1005, + "step": 3642 + }, + { + "epoch": 0.48, + "learning_rate": 1.995396607817191e-05, + "loss": 0.0919, + "step": 3643 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953915237024633e-05, + "loss": 0.0994, + "step": 3644 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953864367882506e-05, + "loss": 0.0931, + "step": 3645 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953813470745656e-05, + "loss": 0.0953, + "step": 3646 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953762545614245e-05, + "loss": 0.0899, + "step": 3647 + }, + { + "epoch": 0.48, + "learning_rate": 1.99537115924884e-05, + "loss": 0.0929, + "step": 3648 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953660611368272e-05, + "loss": 0.1026, + "step": 3649 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953609602254005e-05, + "loss": 0.0969, + "step": 3650 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953558565145742e-05, + "loss": 0.1109, + "step": 3651 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953507500043624e-05, + "loss": 0.0909, + "step": 3652 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953456406947795e-05, + "loss": 0.091, + "step": 3653 + }, + { + "epoch": 0.48, + "learning_rate": 1.99534052858584e-05, + "loss": 0.0888, + "step": 3654 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953354136775586e-05, + "loss": 0.0931, + "step": 3655 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953302959699493e-05, + "loss": 0.0918, + "step": 3656 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953251754630262e-05, + "loss": 0.0964, + "step": 3657 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953200521568043e-05, + "loss": 0.0969, + "step": 3658 + }, + { + "epoch": 0.48, + "learning_rate": 1.995314926051298e-05, + "loss": 0.1035, + "step": 3659 + }, + { + "epoch": 0.48, + "learning_rate": 1.995309797146521e-05, + "loss": 0.1015, + "step": 3660 + }, + { + "epoch": 0.48, + "learning_rate": 1.9953046654424885e-05, + "loss": 0.0863, + "step": 3661 + }, + { + "epoch": 0.48, + "learning_rate": 1.9952995309392143e-05, + "loss": 0.0929, + "step": 3662 + }, + { + "epoch": 0.48, + "learning_rate": 1.9952943936367135e-05, + "loss": 0.0924, + "step": 3663 + }, + { + "epoch": 0.48, + "learning_rate": 1.995289253535e-05, + "loss": 0.0935, + "step": 3664 + }, + { + "epoch": 0.48, + "learning_rate": 1.995284110634089e-05, + "loss": 0.0866, + "step": 3665 + }, + { + "epoch": 0.48, + "learning_rate": 1.9952789649339936e-05, + "loss": 0.0888, + "step": 3666 + }, + { + "epoch": 0.48, + "learning_rate": 1.99527381643473e-05, + "loss": 0.106, + "step": 3667 + }, + { + "epoch": 0.48, + "learning_rate": 1.995268665136311e-05, + "loss": 0.091, + "step": 3668 + }, + { + "epoch": 0.48, + "learning_rate": 1.9952635110387524e-05, + "loss": 0.0997, + "step": 3669 + }, + { + "epoch": 0.48, + "learning_rate": 1.995258354142068e-05, + "loss": 0.0868, + "step": 3670 + }, + { + "epoch": 0.48, + "learning_rate": 1.9952531944462726e-05, + "loss": 0.1006, + "step": 3671 + }, + { + "epoch": 0.48, + "learning_rate": 1.9952480319513802e-05, + "loss": 0.0943, + "step": 3672 + }, + { + "epoch": 0.48, + "learning_rate": 1.995242866657406e-05, + "loss": 0.0942, + "step": 3673 + }, + { + "epoch": 0.48, + "learning_rate": 1.995237698564364e-05, + "loss": 0.0961, + "step": 3674 + }, + { + "epoch": 0.48, + "learning_rate": 1.9952325276722696e-05, + "loss": 0.0892, + "step": 3675 + }, + { + "epoch": 0.48, + "learning_rate": 1.995227353981136e-05, + "loss": 0.0941, + "step": 3676 + }, + { + "epoch": 0.48, + "learning_rate": 1.995222177490979e-05, + "loss": 0.0985, + "step": 3677 + }, + { + "epoch": 0.48, + "learning_rate": 1.9952169982018124e-05, + "loss": 0.0996, + "step": 3678 + }, + { + "epoch": 0.48, + "learning_rate": 1.995211816113651e-05, + "loss": 0.097, + "step": 3679 + }, + { + "epoch": 0.48, + "learning_rate": 1.995206631226509e-05, + "loss": 0.1012, + "step": 3680 + }, + { + "epoch": 0.48, + "learning_rate": 1.995201443540402e-05, + "loss": 0.0931, + "step": 3681 + }, + { + "epoch": 0.48, + "learning_rate": 1.9951962530553436e-05, + "loss": 0.0868, + "step": 3682 + }, + { + "epoch": 0.48, + "learning_rate": 1.9951910597713487e-05, + "loss": 0.0991, + "step": 3683 + }, + { + "epoch": 0.48, + "learning_rate": 1.9951858636884323e-05, + "loss": 0.0846, + "step": 3684 + }, + { + "epoch": 0.48, + "learning_rate": 1.995180664806608e-05, + "loss": 0.1177, + "step": 3685 + }, + { + "epoch": 0.48, + "learning_rate": 1.9951754631258916e-05, + "loss": 0.1032, + "step": 3686 + }, + { + "epoch": 0.48, + "learning_rate": 1.995170258646297e-05, + "loss": 0.0952, + "step": 3687 + }, + { + "epoch": 0.48, + "learning_rate": 1.995165051367839e-05, + "loss": 0.089, + "step": 3688 + }, + { + "epoch": 0.48, + "learning_rate": 1.9951598412905328e-05, + "loss": 0.1042, + "step": 3689 + }, + { + "epoch": 0.48, + "learning_rate": 1.995154628414392e-05, + "loss": 0.0905, + "step": 3690 + }, + { + "epoch": 0.48, + "learning_rate": 1.9951494127394318e-05, + "loss": 0.1052, + "step": 3691 + }, + { + "epoch": 0.48, + "learning_rate": 1.995144194265667e-05, + "loss": 0.1015, + "step": 3692 + }, + { + "epoch": 0.48, + "learning_rate": 1.9951389729931123e-05, + "loss": 0.1026, + "step": 3693 + }, + { + "epoch": 0.48, + "learning_rate": 1.995133748921782e-05, + "loss": 0.1092, + "step": 3694 + }, + { + "epoch": 0.48, + "learning_rate": 1.9951285220516912e-05, + "loss": 0.105, + "step": 3695 + }, + { + "epoch": 0.48, + "learning_rate": 1.995123292382854e-05, + "loss": 0.0913, + "step": 3696 + }, + { + "epoch": 0.48, + "learning_rate": 1.995118059915286e-05, + "loss": 0.1003, + "step": 3697 + }, + { + "epoch": 0.48, + "learning_rate": 1.9951128246490016e-05, + "loss": 0.0984, + "step": 3698 + }, + { + "epoch": 0.48, + "learning_rate": 1.995107586584015e-05, + "loss": 0.1046, + "step": 3699 + }, + { + "epoch": 0.48, + "learning_rate": 1.995102345720341e-05, + "loss": 0.0949, + "step": 3700 + }, + { + "epoch": 0.48, + "learning_rate": 1.9950971020579953e-05, + "loss": 0.1, + "step": 3701 + }, + { + "epoch": 0.48, + "learning_rate": 1.9950918555969917e-05, + "loss": 0.1003, + "step": 3702 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950866063373453e-05, + "loss": 0.0857, + "step": 3703 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950813542790708e-05, + "loss": 0.1024, + "step": 3704 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950760994221828e-05, + "loss": 0.0893, + "step": 3705 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950708417666966e-05, + "loss": 0.096, + "step": 3706 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950655813126265e-05, + "loss": 0.1004, + "step": 3707 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950603180599876e-05, + "loss": 0.1097, + "step": 3708 + }, + { + "epoch": 0.49, + "learning_rate": 1.995055052008794e-05, + "loss": 0.0889, + "step": 3709 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950497831590615e-05, + "loss": 0.1019, + "step": 3710 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950445115108045e-05, + "loss": 0.0828, + "step": 3711 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950392370640377e-05, + "loss": 0.0979, + "step": 3712 + }, + { + "epoch": 0.49, + "learning_rate": 1.995033959818776e-05, + "loss": 0.1055, + "step": 3713 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950286797750344e-05, + "loss": 0.0919, + "step": 3714 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950233969328273e-05, + "loss": 0.0974, + "step": 3715 + }, + { + "epoch": 0.49, + "learning_rate": 1.99501811129217e-05, + "loss": 0.1136, + "step": 3716 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950128228530777e-05, + "loss": 0.0942, + "step": 3717 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950075316155643e-05, + "loss": 0.0988, + "step": 3718 + }, + { + "epoch": 0.49, + "learning_rate": 1.9950022375796457e-05, + "loss": 0.095, + "step": 3719 + }, + { + "epoch": 0.49, + "learning_rate": 1.994996940745336e-05, + "loss": 0.1013, + "step": 3720 + }, + { + "epoch": 0.49, + "learning_rate": 1.9949916411126503e-05, + "loss": 0.0952, + "step": 3721 + }, + { + "epoch": 0.49, + "learning_rate": 1.9949863386816036e-05, + "loss": 0.0879, + "step": 3722 + }, + { + "epoch": 0.49, + "learning_rate": 1.994981033452211e-05, + "loss": 0.1007, + "step": 3723 + }, + { + "epoch": 0.49, + "learning_rate": 1.994975725424487e-05, + "loss": 0.1007, + "step": 3724 + }, + { + "epoch": 0.49, + "learning_rate": 1.994970414598447e-05, + "loss": 0.0998, + "step": 3725 + }, + { + "epoch": 0.49, + "learning_rate": 1.9949651009741054e-05, + "loss": 0.0953, + "step": 3726 + }, + { + "epoch": 0.49, + "learning_rate": 1.9949597845514778e-05, + "loss": 0.095, + "step": 3727 + }, + { + "epoch": 0.49, + "learning_rate": 1.9949544653305785e-05, + "loss": 0.1093, + "step": 3728 + }, + { + "epoch": 0.49, + "learning_rate": 1.994949143311423e-05, + "loss": 0.0937, + "step": 3729 + }, + { + "epoch": 0.49, + "learning_rate": 1.994943818494026e-05, + "loss": 0.111, + "step": 3730 + }, + { + "epoch": 0.49, + "learning_rate": 1.9949384908784023e-05, + "loss": 0.0957, + "step": 3731 + }, + { + "epoch": 0.49, + "learning_rate": 1.994933160464567e-05, + "loss": 0.1061, + "step": 3732 + }, + { + "epoch": 0.49, + "learning_rate": 1.994927827252535e-05, + "loss": 0.107, + "step": 3733 + }, + { + "epoch": 0.49, + "learning_rate": 1.994922491242322e-05, + "loss": 0.0909, + "step": 3734 + }, + { + "epoch": 0.49, + "learning_rate": 1.9949171524339425e-05, + "loss": 0.1056, + "step": 3735 + }, + { + "epoch": 0.49, + "learning_rate": 1.9949118108274112e-05, + "loss": 0.106, + "step": 3736 + }, + { + "epoch": 0.49, + "learning_rate": 1.994906466422743e-05, + "loss": 0.0847, + "step": 3737 + }, + { + "epoch": 0.49, + "learning_rate": 1.994901119219954e-05, + "loss": 0.0968, + "step": 3738 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948957692190585e-05, + "loss": 0.0861, + "step": 3739 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948904164200717e-05, + "loss": 0.1188, + "step": 3740 + }, + { + "epoch": 0.49, + "learning_rate": 1.994885060823008e-05, + "loss": 0.0986, + "step": 3741 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948797024278837e-05, + "loss": 0.0974, + "step": 3742 + }, + { + "epoch": 0.49, + "learning_rate": 1.994874341234713e-05, + "loss": 0.0839, + "step": 3743 + }, + { + "epoch": 0.49, + "learning_rate": 1.994868977243511e-05, + "loss": 0.098, + "step": 3744 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948636104542933e-05, + "loss": 0.0969, + "step": 3745 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948582408670746e-05, + "loss": 0.0904, + "step": 3746 + }, + { + "epoch": 0.49, + "learning_rate": 1.99485286848187e-05, + "loss": 0.0966, + "step": 3747 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948474932986946e-05, + "loss": 0.0855, + "step": 3748 + }, + { + "epoch": 0.49, + "learning_rate": 1.994842115317564e-05, + "loss": 0.0956, + "step": 3749 + }, + { + "epoch": 0.49, + "learning_rate": 1.994836734538492e-05, + "loss": 0.0978, + "step": 3750 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948313509614956e-05, + "loss": 0.0986, + "step": 3751 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948259645865886e-05, + "loss": 0.1052, + "step": 3752 + }, + { + "epoch": 0.49, + "learning_rate": 1.994820575413787e-05, + "loss": 0.0949, + "step": 3753 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948151834431048e-05, + "loss": 0.1049, + "step": 3754 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948097886745584e-05, + "loss": 0.0909, + "step": 3755 + }, + { + "epoch": 0.49, + "learning_rate": 1.9948043911081622e-05, + "loss": 0.0928, + "step": 3756 + }, + { + "epoch": 0.49, + "learning_rate": 1.9947989907439318e-05, + "loss": 0.0879, + "step": 3757 + }, + { + "epoch": 0.49, + "learning_rate": 1.994793587581882e-05, + "loss": 0.0959, + "step": 3758 + }, + { + "epoch": 0.49, + "learning_rate": 1.9947881816220285e-05, + "loss": 0.1007, + "step": 3759 + }, + { + "epoch": 0.49, + "learning_rate": 1.994782772864386e-05, + "loss": 0.1163, + "step": 3760 + }, + { + "epoch": 0.49, + "learning_rate": 1.99477736130897e-05, + "loss": 0.0924, + "step": 3761 + }, + { + "epoch": 0.49, + "learning_rate": 1.9947719469557953e-05, + "loss": 0.1045, + "step": 3762 + }, + { + "epoch": 0.49, + "learning_rate": 1.994766529804878e-05, + "loss": 0.093, + "step": 3763 + }, + { + "epoch": 0.49, + "learning_rate": 1.9947611098562324e-05, + "loss": 0.0913, + "step": 3764 + }, + { + "epoch": 0.49, + "learning_rate": 1.9947556871098744e-05, + "loss": 0.1071, + "step": 3765 + }, + { + "epoch": 0.49, + "learning_rate": 1.994750261565819e-05, + "loss": 0.0943, + "step": 3766 + }, + { + "epoch": 0.49, + "learning_rate": 1.9947448332240813e-05, + "loss": 0.0952, + "step": 3767 + }, + { + "epoch": 0.49, + "learning_rate": 1.994739402084677e-05, + "loss": 0.1013, + "step": 3768 + }, + { + "epoch": 0.49, + "learning_rate": 1.994733968147621e-05, + "loss": 0.0856, + "step": 3769 + }, + { + "epoch": 0.49, + "learning_rate": 1.9947285314129284e-05, + "loss": 0.1101, + "step": 3770 + }, + { + "epoch": 0.49, + "learning_rate": 1.994723091880615e-05, + "loss": 0.0896, + "step": 3771 + }, + { + "epoch": 0.49, + "learning_rate": 1.994717649550696e-05, + "loss": 0.0962, + "step": 3772 + }, + { + "epoch": 0.49, + "learning_rate": 1.9947122044231865e-05, + "loss": 0.0846, + "step": 3773 + }, + { + "epoch": 0.49, + "learning_rate": 1.994706756498102e-05, + "loss": 0.0938, + "step": 3774 + }, + { + "epoch": 0.49, + "learning_rate": 1.9947013057754575e-05, + "loss": 0.1035, + "step": 3775 + }, + { + "epoch": 0.49, + "learning_rate": 1.9946958522552683e-05, + "loss": 0.1003, + "step": 3776 + }, + { + "epoch": 0.49, + "learning_rate": 1.9946903959375505e-05, + "loss": 0.0983, + "step": 3777 + }, + { + "epoch": 0.49, + "learning_rate": 1.994684936822319e-05, + "loss": 0.0898, + "step": 3778 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946794749095888e-05, + "loss": 0.0946, + "step": 3779 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946740101993757e-05, + "loss": 0.0947, + "step": 3780 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946685426916944e-05, + "loss": 0.0904, + "step": 3781 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946630723865614e-05, + "loss": 0.1061, + "step": 3782 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946575992839913e-05, + "loss": 0.0807, + "step": 3783 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946521233839996e-05, + "loss": 0.0917, + "step": 3784 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946466446866017e-05, + "loss": 0.098, + "step": 3785 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946411631918132e-05, + "loss": 0.0954, + "step": 3786 + }, + { + "epoch": 0.5, + "learning_rate": 1.994635678899649e-05, + "loss": 0.0918, + "step": 3787 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946301918101256e-05, + "loss": 0.0982, + "step": 3788 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946247019232575e-05, + "loss": 0.0938, + "step": 3789 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946192092390607e-05, + "loss": 0.0892, + "step": 3790 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946137137575498e-05, + "loss": 0.0939, + "step": 3791 + }, + { + "epoch": 0.5, + "learning_rate": 1.994608215478741e-05, + "loss": 0.0958, + "step": 3792 + }, + { + "epoch": 0.5, + "learning_rate": 1.9946027144026496e-05, + "loss": 0.0976, + "step": 3793 + }, + { + "epoch": 0.5, + "learning_rate": 1.994597210529291e-05, + "loss": 0.108, + "step": 3794 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945917038586808e-05, + "loss": 0.0959, + "step": 3795 + }, + { + "epoch": 0.5, + "learning_rate": 1.994586194390834e-05, + "loss": 0.097, + "step": 3796 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945806821257668e-05, + "loss": 0.1011, + "step": 3797 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945751670634945e-05, + "loss": 0.0999, + "step": 3798 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945696492040325e-05, + "loss": 0.0858, + "step": 3799 + }, + { + "epoch": 0.5, + "learning_rate": 1.994564128547396e-05, + "loss": 0.098, + "step": 3800 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945586050936014e-05, + "loss": 0.0955, + "step": 3801 + }, + { + "epoch": 0.5, + "learning_rate": 1.994553078842663e-05, + "loss": 0.0909, + "step": 3802 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945475497945973e-05, + "loss": 0.0878, + "step": 3803 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945420179494198e-05, + "loss": 0.0816, + "step": 3804 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945364833071458e-05, + "loss": 0.1061, + "step": 3805 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945309458677907e-05, + "loss": 0.098, + "step": 3806 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945254056313703e-05, + "loss": 0.0933, + "step": 3807 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945198625979003e-05, + "loss": 0.096, + "step": 3808 + }, + { + "epoch": 0.5, + "learning_rate": 1.994514316767396e-05, + "loss": 0.0884, + "step": 3809 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945087681398735e-05, + "loss": 0.0985, + "step": 3810 + }, + { + "epoch": 0.5, + "learning_rate": 1.9945032167153476e-05, + "loss": 0.1019, + "step": 3811 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944976624938344e-05, + "loss": 0.0891, + "step": 3812 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944921054753497e-05, + "loss": 0.0967, + "step": 3813 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944865456599087e-05, + "loss": 0.0966, + "step": 3814 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944809830475273e-05, + "loss": 0.1036, + "step": 3815 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944754176382214e-05, + "loss": 0.0993, + "step": 3816 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944698494320063e-05, + "loss": 0.0972, + "step": 3817 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944642784288972e-05, + "loss": 0.0961, + "step": 3818 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944587046289107e-05, + "loss": 0.0925, + "step": 3819 + }, + { + "epoch": 0.5, + "learning_rate": 1.994453128032062e-05, + "loss": 0.0835, + "step": 3820 + }, + { + "epoch": 0.5, + "learning_rate": 1.994447548638367e-05, + "loss": 0.1034, + "step": 3821 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944419664478408e-05, + "loss": 0.1078, + "step": 3822 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944363814604996e-05, + "loss": 0.0969, + "step": 3823 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944307936763593e-05, + "loss": 0.0944, + "step": 3824 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944252030954355e-05, + "loss": 0.0957, + "step": 3825 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944196097177434e-05, + "loss": 0.0982, + "step": 3826 + }, + { + "epoch": 0.5, + "learning_rate": 1.994414013543299e-05, + "loss": 0.0964, + "step": 3827 + }, + { + "epoch": 0.5, + "learning_rate": 1.9944084145721184e-05, + "loss": 0.0985, + "step": 3828 + }, + { + "epoch": 0.5, + "learning_rate": 1.994402812804217e-05, + "loss": 0.1002, + "step": 3829 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943972082396107e-05, + "loss": 0.0965, + "step": 3830 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943916008783153e-05, + "loss": 0.0976, + "step": 3831 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943859907203463e-05, + "loss": 0.1024, + "step": 3832 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943803777657197e-05, + "loss": 0.0925, + "step": 3833 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943747620144512e-05, + "loss": 0.0918, + "step": 3834 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943691434665567e-05, + "loss": 0.1037, + "step": 3835 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943635221220522e-05, + "loss": 0.0867, + "step": 3836 + }, + { + "epoch": 0.5, + "learning_rate": 1.994357897980953e-05, + "loss": 0.091, + "step": 3837 + }, + { + "epoch": 0.5, + "learning_rate": 1.994352271043275e-05, + "loss": 0.0948, + "step": 3838 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943466413090342e-05, + "loss": 0.1034, + "step": 3839 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943410087782466e-05, + "loss": 0.1029, + "step": 3840 + }, + { + "epoch": 0.5, + "learning_rate": 1.994335373450928e-05, + "loss": 0.0896, + "step": 3841 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943297353270938e-05, + "loss": 0.0946, + "step": 3842 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943240944067607e-05, + "loss": 0.08, + "step": 3843 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943184506899432e-05, + "loss": 0.0949, + "step": 3844 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943128041766587e-05, + "loss": 0.0994, + "step": 3845 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943071548669223e-05, + "loss": 0.0862, + "step": 3846 + }, + { + "epoch": 0.5, + "learning_rate": 1.9943015027607494e-05, + "loss": 0.0936, + "step": 3847 + }, + { + "epoch": 0.5, + "learning_rate": 1.994295847858157e-05, + "loss": 0.1016, + "step": 3848 + }, + { + "epoch": 0.5, + "learning_rate": 1.9942901901591604e-05, + "loss": 0.0994, + "step": 3849 + }, + { + "epoch": 0.5, + "learning_rate": 1.9942845296637756e-05, + "loss": 0.09, + "step": 3850 + }, + { + "epoch": 0.5, + "learning_rate": 1.9942788663720186e-05, + "loss": 0.1011, + "step": 3851 + }, + { + "epoch": 0.5, + "learning_rate": 1.994273200283905e-05, + "loss": 0.0874, + "step": 3852 + }, + { + "epoch": 0.5, + "learning_rate": 1.994267531399451e-05, + "loss": 0.0869, + "step": 3853 + }, + { + "epoch": 0.5, + "learning_rate": 1.9942618597186726e-05, + "loss": 0.1, + "step": 3854 + }, + { + "epoch": 0.5, + "learning_rate": 1.994256185241586e-05, + "loss": 0.1012, + "step": 3855 + }, + { + "epoch": 0.51, + "learning_rate": 1.994250507968206e-05, + "loss": 0.101, + "step": 3856 + }, + { + "epoch": 0.51, + "learning_rate": 1.99424482789855e-05, + "loss": 0.0841, + "step": 3857 + }, + { + "epoch": 0.51, + "learning_rate": 1.9942391450326333e-05, + "loss": 0.0899, + "step": 3858 + }, + { + "epoch": 0.51, + "learning_rate": 1.994233459370472e-05, + "loss": 0.1028, + "step": 3859 + }, + { + "epoch": 0.51, + "learning_rate": 1.994227770912082e-05, + "loss": 0.0878, + "step": 3860 + }, + { + "epoch": 0.51, + "learning_rate": 1.9942220796574793e-05, + "loss": 0.0949, + "step": 3861 + }, + { + "epoch": 0.51, + "learning_rate": 1.99421638560668e-05, + "loss": 0.1063, + "step": 3862 + }, + { + "epoch": 0.51, + "learning_rate": 1.9942106887597e-05, + "loss": 0.0937, + "step": 3863 + }, + { + "epoch": 0.51, + "learning_rate": 1.994204989116555e-05, + "loss": 0.0973, + "step": 3864 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941992866772618e-05, + "loss": 0.083, + "step": 3865 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941935814418362e-05, + "loss": 0.109, + "step": 3866 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941878734102936e-05, + "loss": 0.0966, + "step": 3867 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941821625826512e-05, + "loss": 0.0766, + "step": 3868 + }, + { + "epoch": 0.51, + "learning_rate": 1.994176448958924e-05, + "loss": 0.0868, + "step": 3869 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941707325391286e-05, + "loss": 0.0972, + "step": 3870 + }, + { + "epoch": 0.51, + "learning_rate": 1.994165013323281e-05, + "loss": 0.104, + "step": 3871 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941592913113972e-05, + "loss": 0.0916, + "step": 3872 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941535665034933e-05, + "loss": 0.0967, + "step": 3873 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941478388995854e-05, + "loss": 0.0971, + "step": 3874 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941421084996896e-05, + "loss": 0.0926, + "step": 3875 + }, + { + "epoch": 0.51, + "learning_rate": 1.994136375303822e-05, + "loss": 0.0937, + "step": 3876 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941306393119993e-05, + "loss": 0.0951, + "step": 3877 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941249005242364e-05, + "loss": 0.1024, + "step": 3878 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941191589405505e-05, + "loss": 0.0935, + "step": 3879 + }, + { + "epoch": 0.51, + "learning_rate": 1.994113414560958e-05, + "loss": 0.0954, + "step": 3880 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941076673854732e-05, + "loss": 0.0981, + "step": 3881 + }, + { + "epoch": 0.51, + "learning_rate": 1.9941019174141144e-05, + "loss": 0.0924, + "step": 3882 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940961646468967e-05, + "loss": 0.0937, + "step": 3883 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940904090838366e-05, + "loss": 0.0901, + "step": 3884 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940846507249502e-05, + "loss": 0.0906, + "step": 3885 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940788895702533e-05, + "loss": 0.1042, + "step": 3886 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940731256197626e-05, + "loss": 0.0913, + "step": 3887 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940673588734942e-05, + "loss": 0.0985, + "step": 3888 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940615893314646e-05, + "loss": 0.098, + "step": 3889 + }, + { + "epoch": 0.51, + "learning_rate": 1.994055816993689e-05, + "loss": 0.0929, + "step": 3890 + }, + { + "epoch": 0.51, + "learning_rate": 1.994050041860185e-05, + "loss": 0.0964, + "step": 3891 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940442639309682e-05, + "loss": 0.0795, + "step": 3892 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940384832060545e-05, + "loss": 0.1004, + "step": 3893 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940326996854608e-05, + "loss": 0.0982, + "step": 3894 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940269133692026e-05, + "loss": 0.0944, + "step": 3895 + }, + { + "epoch": 0.51, + "learning_rate": 1.994021124257297e-05, + "loss": 0.1046, + "step": 3896 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940153323497597e-05, + "loss": 0.0937, + "step": 3897 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940095376466076e-05, + "loss": 0.0965, + "step": 3898 + }, + { + "epoch": 0.51, + "learning_rate": 1.9940037401478564e-05, + "loss": 0.0979, + "step": 3899 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939979398535224e-05, + "loss": 0.0942, + "step": 3900 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939921367636222e-05, + "loss": 0.0971, + "step": 3901 + }, + { + "epoch": 0.51, + "learning_rate": 1.993986330878172e-05, + "loss": 0.0985, + "step": 3902 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939805221971888e-05, + "loss": 0.0887, + "step": 3903 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939747107206877e-05, + "loss": 0.1034, + "step": 3904 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939688964486858e-05, + "loss": 0.0885, + "step": 3905 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939630793811994e-05, + "loss": 0.0915, + "step": 3906 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939572595182447e-05, + "loss": 0.0933, + "step": 3907 + }, + { + "epoch": 0.51, + "learning_rate": 1.993951436859838e-05, + "loss": 0.1117, + "step": 3908 + }, + { + "epoch": 0.51, + "learning_rate": 1.993945611405996e-05, + "loss": 0.0958, + "step": 3909 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939397831567348e-05, + "loss": 0.0884, + "step": 3910 + }, + { + "epoch": 0.51, + "learning_rate": 1.993933952112071e-05, + "loss": 0.0979, + "step": 3911 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939281182720208e-05, + "loss": 0.0956, + "step": 3912 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939222816366008e-05, + "loss": 0.0935, + "step": 3913 + }, + { + "epoch": 0.51, + "learning_rate": 1.993916442205827e-05, + "loss": 0.1019, + "step": 3914 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939105999797166e-05, + "loss": 0.0939, + "step": 3915 + }, + { + "epoch": 0.51, + "learning_rate": 1.9939047549582856e-05, + "loss": 0.0971, + "step": 3916 + }, + { + "epoch": 0.51, + "learning_rate": 1.99389890714155e-05, + "loss": 0.0954, + "step": 3917 + }, + { + "epoch": 0.51, + "learning_rate": 1.9938930565295266e-05, + "loss": 0.0921, + "step": 3918 + }, + { + "epoch": 0.51, + "learning_rate": 1.9938872031222322e-05, + "loss": 0.0902, + "step": 3919 + }, + { + "epoch": 0.51, + "learning_rate": 1.993881346919683e-05, + "loss": 0.0894, + "step": 3920 + }, + { + "epoch": 0.51, + "learning_rate": 1.993875487921895e-05, + "loss": 0.0865, + "step": 3921 + }, + { + "epoch": 0.51, + "learning_rate": 1.9938696261288856e-05, + "loss": 0.101, + "step": 3922 + }, + { + "epoch": 0.51, + "learning_rate": 1.9938637615406706e-05, + "loss": 0.0972, + "step": 3923 + }, + { + "epoch": 0.51, + "learning_rate": 1.993857894157267e-05, + "loss": 0.0937, + "step": 3924 + }, + { + "epoch": 0.51, + "learning_rate": 1.9938520239786906e-05, + "loss": 0.0914, + "step": 3925 + }, + { + "epoch": 0.51, + "learning_rate": 1.9938461510049585e-05, + "loss": 0.0928, + "step": 3926 + }, + { + "epoch": 0.51, + "learning_rate": 1.993840275236087e-05, + "loss": 0.0911, + "step": 3927 + }, + { + "epoch": 0.51, + "learning_rate": 1.9938343966720925e-05, + "loss": 0.1004, + "step": 3928 + }, + { + "epoch": 0.51, + "learning_rate": 1.9938285153129917e-05, + "loss": 0.0912, + "step": 3929 + }, + { + "epoch": 0.51, + "learning_rate": 1.993822631158801e-05, + "loss": 0.0964, + "step": 3930 + }, + { + "epoch": 0.51, + "learning_rate": 1.9938167442095374e-05, + "loss": 0.1086, + "step": 3931 + }, + { + "epoch": 0.52, + "learning_rate": 1.9938108544652173e-05, + "loss": 0.0919, + "step": 3932 + }, + { + "epoch": 0.52, + "learning_rate": 1.993804961925857e-05, + "loss": 0.0978, + "step": 3933 + }, + { + "epoch": 0.52, + "learning_rate": 1.9937990665914725e-05, + "loss": 0.0993, + "step": 3934 + }, + { + "epoch": 0.52, + "learning_rate": 1.9937931684620816e-05, + "loss": 0.1029, + "step": 3935 + }, + { + "epoch": 0.52, + "learning_rate": 1.9937872675377002e-05, + "loss": 0.0973, + "step": 3936 + }, + { + "epoch": 0.52, + "learning_rate": 1.9937813638183452e-05, + "loss": 0.0972, + "step": 3937 + }, + { + "epoch": 0.52, + "learning_rate": 1.993775457304033e-05, + "loss": 0.0944, + "step": 3938 + }, + { + "epoch": 0.52, + "learning_rate": 1.9937695479947804e-05, + "loss": 0.091, + "step": 3939 + }, + { + "epoch": 0.52, + "learning_rate": 1.9937636358906032e-05, + "loss": 0.0957, + "step": 3940 + }, + { + "epoch": 0.52, + "learning_rate": 1.9937577209915196e-05, + "loss": 0.0954, + "step": 3941 + }, + { + "epoch": 0.52, + "learning_rate": 1.9937518032975446e-05, + "loss": 0.0867, + "step": 3942 + }, + { + "epoch": 0.52, + "learning_rate": 1.993745882808696e-05, + "loss": 0.0983, + "step": 3943 + }, + { + "epoch": 0.52, + "learning_rate": 1.99373995952499e-05, + "loss": 0.0976, + "step": 3944 + }, + { + "epoch": 0.52, + "learning_rate": 1.993734033446443e-05, + "loss": 0.0914, + "step": 3945 + }, + { + "epoch": 0.52, + "learning_rate": 1.993728104573072e-05, + "loss": 0.1041, + "step": 3946 + }, + { + "epoch": 0.52, + "learning_rate": 1.993722172904894e-05, + "loss": 0.1054, + "step": 3947 + }, + { + "epoch": 0.52, + "learning_rate": 1.993716238441925e-05, + "loss": 0.0882, + "step": 3948 + }, + { + "epoch": 0.52, + "learning_rate": 1.993710301184182e-05, + "loss": 0.0947, + "step": 3949 + }, + { + "epoch": 0.52, + "learning_rate": 1.9937043611316817e-05, + "loss": 0.0758, + "step": 3950 + }, + { + "epoch": 0.52, + "learning_rate": 1.9936984182844407e-05, + "loss": 0.1126, + "step": 3951 + }, + { + "epoch": 0.52, + "learning_rate": 1.9936924726424764e-05, + "loss": 0.1022, + "step": 3952 + }, + { + "epoch": 0.52, + "learning_rate": 1.993686524205804e-05, + "loss": 0.1007, + "step": 3953 + }, + { + "epoch": 0.52, + "learning_rate": 1.993680572974442e-05, + "loss": 0.089, + "step": 3954 + }, + { + "epoch": 0.52, + "learning_rate": 1.993674618948406e-05, + "loss": 0.1004, + "step": 3955 + }, + { + "epoch": 0.52, + "learning_rate": 1.993668662127713e-05, + "loss": 0.0965, + "step": 3956 + }, + { + "epoch": 0.52, + "learning_rate": 1.99366270251238e-05, + "loss": 0.0957, + "step": 3957 + }, + { + "epoch": 0.52, + "learning_rate": 1.9936567401024233e-05, + "loss": 0.0924, + "step": 3958 + }, + { + "epoch": 0.52, + "learning_rate": 1.99365077489786e-05, + "loss": 0.0971, + "step": 3959 + }, + { + "epoch": 0.52, + "learning_rate": 1.993644806898707e-05, + "loss": 0.0917, + "step": 3960 + }, + { + "epoch": 0.52, + "learning_rate": 1.9936388361049805e-05, + "loss": 0.1001, + "step": 3961 + }, + { + "epoch": 0.52, + "learning_rate": 1.993632862516698e-05, + "loss": 0.0844, + "step": 3962 + }, + { + "epoch": 0.52, + "learning_rate": 1.9936268861338758e-05, + "loss": 0.1015, + "step": 3963 + }, + { + "epoch": 0.52, + "learning_rate": 1.993620906956531e-05, + "loss": 0.0929, + "step": 3964 + }, + { + "epoch": 0.52, + "learning_rate": 1.9936149249846802e-05, + "loss": 0.0901, + "step": 3965 + }, + { + "epoch": 0.52, + "learning_rate": 1.9936089402183407e-05, + "loss": 0.1065, + "step": 3966 + }, + { + "epoch": 0.52, + "learning_rate": 1.9936029526575287e-05, + "loss": 0.0966, + "step": 3967 + }, + { + "epoch": 0.52, + "learning_rate": 1.993596962302261e-05, + "loss": 0.1099, + "step": 3968 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935909691525553e-05, + "loss": 0.0937, + "step": 3969 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935849732084275e-05, + "loss": 0.0946, + "step": 3970 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935789744698953e-05, + "loss": 0.0987, + "step": 3971 + }, + { + "epoch": 0.52, + "learning_rate": 1.993572972936975e-05, + "loss": 0.1011, + "step": 3972 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935669686096834e-05, + "loss": 0.1057, + "step": 3973 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935609614880377e-05, + "loss": 0.0943, + "step": 3974 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935549515720546e-05, + "loss": 0.0831, + "step": 3975 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935489388617516e-05, + "loss": 0.0912, + "step": 3976 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935429233571448e-05, + "loss": 0.0919, + "step": 3977 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935369050582515e-05, + "loss": 0.0964, + "step": 3978 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935308839650883e-05, + "loss": 0.0879, + "step": 3979 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935248600776725e-05, + "loss": 0.1109, + "step": 3980 + }, + { + "epoch": 0.52, + "learning_rate": 1.993518833396021e-05, + "loss": 0.1023, + "step": 3981 + }, + { + "epoch": 0.52, + "learning_rate": 1.993512803920151e-05, + "loss": 0.1036, + "step": 3982 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935067716500787e-05, + "loss": 0.1083, + "step": 3983 + }, + { + "epoch": 0.52, + "learning_rate": 1.9935007365858216e-05, + "loss": 0.0961, + "step": 3984 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934946987273967e-05, + "loss": 0.093, + "step": 3985 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934886580748207e-05, + "loss": 0.1069, + "step": 3986 + }, + { + "epoch": 0.52, + "learning_rate": 1.993482614628111e-05, + "loss": 0.0923, + "step": 3987 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934765683872845e-05, + "loss": 0.0989, + "step": 3988 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934705193523577e-05, + "loss": 0.1115, + "step": 3989 + }, + { + "epoch": 0.52, + "learning_rate": 1.993464467523348e-05, + "loss": 0.0899, + "step": 3990 + }, + { + "epoch": 0.52, + "learning_rate": 1.993458412900272e-05, + "loss": 0.0966, + "step": 3991 + }, + { + "epoch": 0.52, + "learning_rate": 1.993452355483148e-05, + "loss": 0.1055, + "step": 3992 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934462952719914e-05, + "loss": 0.0912, + "step": 3993 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934402322668208e-05, + "loss": 0.0966, + "step": 3994 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934341664676517e-05, + "loss": 0.1109, + "step": 3995 + }, + { + "epoch": 0.52, + "learning_rate": 1.993428097874502e-05, + "loss": 0.0914, + "step": 3996 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934220264873892e-05, + "loss": 0.0962, + "step": 3997 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934159523063293e-05, + "loss": 0.1064, + "step": 3998 + }, + { + "epoch": 0.52, + "learning_rate": 1.99340987533134e-05, + "loss": 0.0947, + "step": 3999 + }, + { + "epoch": 0.52, + "learning_rate": 1.9934037955624384e-05, + "loss": 0.0944, + "step": 4000 + }, + { + "epoch": 0.52, + "learning_rate": 1.993397712999642e-05, + "loss": 0.0951, + "step": 4001 + }, + { + "epoch": 0.52, + "learning_rate": 1.993391627642967e-05, + "loss": 0.0993, + "step": 4002 + }, + { + "epoch": 0.52, + "learning_rate": 1.993385539492431e-05, + "loss": 0.1011, + "step": 4003 + }, + { + "epoch": 0.52, + "learning_rate": 1.9933794485480507e-05, + "loss": 0.0909, + "step": 4004 + }, + { + "epoch": 0.52, + "learning_rate": 1.9933733548098444e-05, + "loss": 0.096, + "step": 4005 + }, + { + "epoch": 0.52, + "learning_rate": 1.9933672582778277e-05, + "loss": 0.0945, + "step": 4006 + }, + { + "epoch": 0.52, + "learning_rate": 1.993361158952019e-05, + "loss": 0.0909, + "step": 4007 + }, + { + "epoch": 0.53, + "learning_rate": 1.9933550568324347e-05, + "loss": 0.1005, + "step": 4008 + }, + { + "epoch": 0.53, + "learning_rate": 1.9933489519190923e-05, + "loss": 0.0831, + "step": 4009 + }, + { + "epoch": 0.53, + "learning_rate": 1.9933428442120086e-05, + "loss": 0.0886, + "step": 4010 + }, + { + "epoch": 0.53, + "learning_rate": 1.9933367337112016e-05, + "loss": 0.0961, + "step": 4011 + }, + { + "epoch": 0.53, + "learning_rate": 1.9933306204166874e-05, + "loss": 0.0984, + "step": 4012 + }, + { + "epoch": 0.53, + "learning_rate": 1.9933245043284838e-05, + "loss": 0.0988, + "step": 4013 + }, + { + "epoch": 0.53, + "learning_rate": 1.993318385446608e-05, + "loss": 0.1014, + "step": 4014 + }, + { + "epoch": 0.53, + "learning_rate": 1.993312263771077e-05, + "loss": 0.0861, + "step": 4015 + }, + { + "epoch": 0.53, + "learning_rate": 1.9933061393019087e-05, + "loss": 0.0977, + "step": 4016 + }, + { + "epoch": 0.53, + "learning_rate": 1.9933000120391192e-05, + "loss": 0.0843, + "step": 4017 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932938819827266e-05, + "loss": 0.0938, + "step": 4018 + }, + { + "epoch": 0.53, + "learning_rate": 1.993287749132748e-05, + "loss": 0.0871, + "step": 4019 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932816134892e-05, + "loss": 0.0923, + "step": 4020 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932754750521008e-05, + "loss": 0.0871, + "step": 4021 + }, + { + "epoch": 0.53, + "learning_rate": 1.993269333821467e-05, + "loss": 0.0925, + "step": 4022 + }, + { + "epoch": 0.53, + "learning_rate": 1.993263189797316e-05, + "loss": 0.1078, + "step": 4023 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932570429796655e-05, + "loss": 0.093, + "step": 4024 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932508933685324e-05, + "loss": 0.0824, + "step": 4025 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932447409639337e-05, + "loss": 0.0876, + "step": 4026 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932385857658872e-05, + "loss": 0.0964, + "step": 4027 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932324277744104e-05, + "loss": 0.0898, + "step": 4028 + }, + { + "epoch": 0.53, + "learning_rate": 1.99322626698952e-05, + "loss": 0.1016, + "step": 4029 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932201034112336e-05, + "loss": 0.1043, + "step": 4030 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932139370395685e-05, + "loss": 0.0876, + "step": 4031 + }, + { + "epoch": 0.53, + "learning_rate": 1.9932077678745426e-05, + "loss": 0.0995, + "step": 4032 + }, + { + "epoch": 0.53, + "learning_rate": 1.993201595916172e-05, + "loss": 0.1084, + "step": 4033 + }, + { + "epoch": 0.53, + "learning_rate": 1.9931954211644754e-05, + "loss": 0.1039, + "step": 4034 + }, + { + "epoch": 0.53, + "learning_rate": 1.9931892436194693e-05, + "loss": 0.0845, + "step": 4035 + }, + { + "epoch": 0.53, + "learning_rate": 1.993183063281171e-05, + "loss": 0.0957, + "step": 4036 + }, + { + "epoch": 0.53, + "learning_rate": 1.993176880149599e-05, + "loss": 0.0933, + "step": 4037 + }, + { + "epoch": 0.53, + "learning_rate": 1.9931706942247695e-05, + "loss": 0.0886, + "step": 4038 + }, + { + "epoch": 0.53, + "learning_rate": 1.9931645055067005e-05, + "loss": 0.0912, + "step": 4039 + }, + { + "epoch": 0.53, + "learning_rate": 1.993158313995409e-05, + "loss": 0.0954, + "step": 4040 + }, + { + "epoch": 0.53, + "learning_rate": 1.9931521196909127e-05, + "loss": 0.0792, + "step": 4041 + }, + { + "epoch": 0.53, + "learning_rate": 1.9931459225932288e-05, + "loss": 0.0968, + "step": 4042 + }, + { + "epoch": 0.53, + "learning_rate": 1.993139722702375e-05, + "loss": 0.0856, + "step": 4043 + }, + { + "epoch": 0.53, + "learning_rate": 1.9931335200183687e-05, + "loss": 0.0995, + "step": 4044 + }, + { + "epoch": 0.53, + "learning_rate": 1.9931273145412273e-05, + "loss": 0.0915, + "step": 4045 + }, + { + "epoch": 0.53, + "learning_rate": 1.993121106270968e-05, + "loss": 0.0955, + "step": 4046 + }, + { + "epoch": 0.53, + "learning_rate": 1.9931148952076086e-05, + "loss": 0.0793, + "step": 4047 + }, + { + "epoch": 0.53, + "learning_rate": 1.9931086813511664e-05, + "loss": 0.0995, + "step": 4048 + }, + { + "epoch": 0.53, + "learning_rate": 1.993102464701659e-05, + "loss": 0.0956, + "step": 4049 + }, + { + "epoch": 0.53, + "learning_rate": 1.9930962452591038e-05, + "loss": 0.0919, + "step": 4050 + }, + { + "epoch": 0.53, + "learning_rate": 1.9930900230235182e-05, + "loss": 0.0807, + "step": 4051 + }, + { + "epoch": 0.53, + "learning_rate": 1.99308379799492e-05, + "loss": 0.1062, + "step": 4052 + }, + { + "epoch": 0.53, + "learning_rate": 1.993077570173326e-05, + "loss": 0.0774, + "step": 4053 + }, + { + "epoch": 0.53, + "learning_rate": 1.9930713395587545e-05, + "loss": 0.101, + "step": 4054 + }, + { + "epoch": 0.53, + "learning_rate": 1.993065106151223e-05, + "loss": 0.0857, + "step": 4055 + }, + { + "epoch": 0.53, + "learning_rate": 1.9930588699507485e-05, + "loss": 0.0846, + "step": 4056 + }, + { + "epoch": 0.53, + "learning_rate": 1.993052630957349e-05, + "loss": 0.107, + "step": 4057 + }, + { + "epoch": 0.53, + "learning_rate": 1.993046389171042e-05, + "loss": 0.0907, + "step": 4058 + }, + { + "epoch": 0.53, + "learning_rate": 1.9930401445918443e-05, + "loss": 0.1074, + "step": 4059 + }, + { + "epoch": 0.53, + "learning_rate": 1.9930338972197744e-05, + "loss": 0.0929, + "step": 4060 + }, + { + "epoch": 0.53, + "learning_rate": 1.99302764705485e-05, + "loss": 0.0953, + "step": 4061 + }, + { + "epoch": 0.53, + "learning_rate": 1.9930213940970876e-05, + "loss": 0.105, + "step": 4062 + }, + { + "epoch": 0.53, + "learning_rate": 1.9930151383465056e-05, + "loss": 0.0944, + "step": 4063 + }, + { + "epoch": 0.53, + "learning_rate": 1.993008879803121e-05, + "loss": 0.0952, + "step": 4064 + }, + { + "epoch": 0.53, + "learning_rate": 1.9930026184669525e-05, + "loss": 0.0968, + "step": 4065 + }, + { + "epoch": 0.53, + "learning_rate": 1.992996354338017e-05, + "loss": 0.1034, + "step": 4066 + }, + { + "epoch": 0.53, + "learning_rate": 1.9929900874163318e-05, + "loss": 0.0848, + "step": 4067 + }, + { + "epoch": 0.53, + "learning_rate": 1.992983817701915e-05, + "loss": 0.0916, + "step": 4068 + }, + { + "epoch": 0.53, + "learning_rate": 1.9929775451947838e-05, + "loss": 0.0866, + "step": 4069 + }, + { + "epoch": 0.53, + "learning_rate": 1.9929712698949565e-05, + "loss": 0.0935, + "step": 4070 + }, + { + "epoch": 0.53, + "learning_rate": 1.9929649918024504e-05, + "loss": 0.0954, + "step": 4071 + }, + { + "epoch": 0.53, + "learning_rate": 1.9929587109172827e-05, + "loss": 0.1012, + "step": 4072 + }, + { + "epoch": 0.53, + "learning_rate": 1.992952427239472e-05, + "loss": 0.0884, + "step": 4073 + }, + { + "epoch": 0.53, + "learning_rate": 1.992946140769035e-05, + "loss": 0.1095, + "step": 4074 + }, + { + "epoch": 0.53, + "learning_rate": 1.9929398515059902e-05, + "loss": 0.0996, + "step": 4075 + }, + { + "epoch": 0.53, + "learning_rate": 1.992933559450355e-05, + "loss": 0.0916, + "step": 4076 + }, + { + "epoch": 0.53, + "learning_rate": 1.992927264602147e-05, + "loss": 0.106, + "step": 4077 + }, + { + "epoch": 0.53, + "learning_rate": 1.992920966961384e-05, + "loss": 0.1029, + "step": 4078 + }, + { + "epoch": 0.53, + "learning_rate": 1.9929146665280833e-05, + "loss": 0.1105, + "step": 4079 + }, + { + "epoch": 0.53, + "learning_rate": 1.992908363302263e-05, + "loss": 0.0979, + "step": 4080 + }, + { + "epoch": 0.53, + "learning_rate": 1.992902057283941e-05, + "loss": 0.0912, + "step": 4081 + }, + { + "epoch": 0.53, + "learning_rate": 1.992895748473135e-05, + "loss": 0.0833, + "step": 4082 + }, + { + "epoch": 0.53, + "learning_rate": 1.992889436869862e-05, + "loss": 0.0963, + "step": 4083 + }, + { + "epoch": 0.53, + "learning_rate": 1.992883122474141e-05, + "loss": 0.0998, + "step": 4084 + }, + { + "epoch": 0.54, + "learning_rate": 1.9928768052859888e-05, + "loss": 0.114, + "step": 4085 + }, + { + "epoch": 0.54, + "learning_rate": 1.9928704853054236e-05, + "loss": 0.0826, + "step": 4086 + }, + { + "epoch": 0.54, + "learning_rate": 1.9928641625324628e-05, + "loss": 0.0968, + "step": 4087 + }, + { + "epoch": 0.54, + "learning_rate": 1.9928578369671244e-05, + "loss": 0.0946, + "step": 4088 + }, + { + "epoch": 0.54, + "learning_rate": 1.9928515086094267e-05, + "loss": 0.0956, + "step": 4089 + }, + { + "epoch": 0.54, + "learning_rate": 1.9928451774593863e-05, + "loss": 0.0921, + "step": 4090 + }, + { + "epoch": 0.54, + "learning_rate": 1.992838843517022e-05, + "loss": 0.1034, + "step": 4091 + }, + { + "epoch": 0.54, + "learning_rate": 1.9928325067823516e-05, + "loss": 0.0861, + "step": 4092 + }, + { + "epoch": 0.54, + "learning_rate": 1.9928261672553925e-05, + "loss": 0.0994, + "step": 4093 + }, + { + "epoch": 0.54, + "learning_rate": 1.9928198249361625e-05, + "loss": 0.0882, + "step": 4094 + }, + { + "epoch": 0.54, + "learning_rate": 1.99281347982468e-05, + "loss": 0.1024, + "step": 4095 + }, + { + "epoch": 0.54, + "learning_rate": 1.992807131920962e-05, + "loss": 0.0844, + "step": 4096 + }, + { + "epoch": 0.54, + "learning_rate": 1.992800781225027e-05, + "loss": 0.0932, + "step": 4097 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927944277368928e-05, + "loss": 0.094, + "step": 4098 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927880714565768e-05, + "loss": 0.0991, + "step": 4099 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927817123840976e-05, + "loss": 0.0953, + "step": 4100 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927753505194723e-05, + "loss": 0.0981, + "step": 4101 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927689858627194e-05, + "loss": 0.0892, + "step": 4102 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927626184138566e-05, + "loss": 0.0856, + "step": 4103 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927562481729018e-05, + "loss": 0.0963, + "step": 4104 + }, + { + "epoch": 0.54, + "learning_rate": 1.992749875139873e-05, + "loss": 0.0888, + "step": 4105 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927434993147884e-05, + "loss": 0.088, + "step": 4106 + }, + { + "epoch": 0.54, + "learning_rate": 1.992737120697665e-05, + "loss": 0.0894, + "step": 4107 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927307392885215e-05, + "loss": 0.0975, + "step": 4108 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927243550873754e-05, + "loss": 0.0887, + "step": 4109 + }, + { + "epoch": 0.54, + "learning_rate": 1.992717968094245e-05, + "loss": 0.0922, + "step": 4110 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927115783091486e-05, + "loss": 0.0802, + "step": 4111 + }, + { + "epoch": 0.54, + "learning_rate": 1.9927051857321037e-05, + "loss": 0.0866, + "step": 4112 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926987903631277e-05, + "loss": 0.0895, + "step": 4113 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926923922022397e-05, + "loss": 0.1025, + "step": 4114 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926859912494572e-05, + "loss": 0.1005, + "step": 4115 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926795875047983e-05, + "loss": 0.1025, + "step": 4116 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926731809682807e-05, + "loss": 0.0952, + "step": 4117 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926667716399226e-05, + "loss": 0.0858, + "step": 4118 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926603595197426e-05, + "loss": 0.0954, + "step": 4119 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926539446077575e-05, + "loss": 0.1067, + "step": 4120 + }, + { + "epoch": 0.54, + "learning_rate": 1.992647526903986e-05, + "loss": 0.0944, + "step": 4121 + }, + { + "epoch": 0.54, + "learning_rate": 1.992641106408447e-05, + "loss": 0.1012, + "step": 4122 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926346831211573e-05, + "loss": 0.096, + "step": 4123 + }, + { + "epoch": 0.54, + "learning_rate": 1.992628257042135e-05, + "loss": 0.088, + "step": 4124 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926218281713995e-05, + "loss": 0.0894, + "step": 4125 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926153965089673e-05, + "loss": 0.0921, + "step": 4126 + }, + { + "epoch": 0.54, + "learning_rate": 1.992608962054857e-05, + "loss": 0.0945, + "step": 4127 + }, + { + "epoch": 0.54, + "learning_rate": 1.9926025248090872e-05, + "loss": 0.0961, + "step": 4128 + }, + { + "epoch": 0.54, + "learning_rate": 1.992596084771676e-05, + "loss": 0.0795, + "step": 4129 + }, + { + "epoch": 0.54, + "learning_rate": 1.9925896419426406e-05, + "loss": 0.083, + "step": 4130 + }, + { + "epoch": 0.54, + "learning_rate": 1.992583196322e-05, + "loss": 0.0827, + "step": 4131 + }, + { + "epoch": 0.54, + "learning_rate": 1.9925767479097716e-05, + "loss": 0.09, + "step": 4132 + }, + { + "epoch": 0.54, + "learning_rate": 1.9925702967059742e-05, + "loss": 0.1007, + "step": 4133 + }, + { + "epoch": 0.54, + "learning_rate": 1.992563842710626e-05, + "loss": 0.0965, + "step": 4134 + }, + { + "epoch": 0.54, + "learning_rate": 1.9925573859237442e-05, + "loss": 0.0899, + "step": 4135 + }, + { + "epoch": 0.54, + "learning_rate": 1.9925509263453483e-05, + "loss": 0.0968, + "step": 4136 + }, + { + "epoch": 0.54, + "learning_rate": 1.9925444639754554e-05, + "loss": 0.0835, + "step": 4137 + }, + { + "epoch": 0.54, + "learning_rate": 1.992537998814084e-05, + "loss": 0.103, + "step": 4138 + }, + { + "epoch": 0.54, + "learning_rate": 1.9925315308612526e-05, + "loss": 0.0866, + "step": 4139 + }, + { + "epoch": 0.54, + "learning_rate": 1.992525060116979e-05, + "loss": 0.0961, + "step": 4140 + }, + { + "epoch": 0.54, + "learning_rate": 1.9925185865812815e-05, + "loss": 0.0906, + "step": 4141 + }, + { + "epoch": 0.54, + "learning_rate": 1.9925121102541786e-05, + "loss": 0.0967, + "step": 4142 + }, + { + "epoch": 0.54, + "learning_rate": 1.992505631135688e-05, + "loss": 0.099, + "step": 4143 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924991492258278e-05, + "loss": 0.0928, + "step": 4144 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924926645246173e-05, + "loss": 0.1001, + "step": 4145 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924861770320737e-05, + "loss": 0.0957, + "step": 4146 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924796867482154e-05, + "loss": 0.0996, + "step": 4147 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924731936730606e-05, + "loss": 0.0919, + "step": 4148 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924666978066284e-05, + "loss": 0.0898, + "step": 4149 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924601991489366e-05, + "loss": 0.0931, + "step": 4150 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924536977000026e-05, + "loss": 0.0971, + "step": 4151 + }, + { + "epoch": 0.54, + "learning_rate": 1.992447193459846e-05, + "loss": 0.0933, + "step": 4152 + }, + { + "epoch": 0.54, + "learning_rate": 1.992440686428484e-05, + "loss": 0.1018, + "step": 4153 + }, + { + "epoch": 0.54, + "learning_rate": 1.992434176605936e-05, + "loss": 0.0956, + "step": 4154 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924276639922197e-05, + "loss": 0.0832, + "step": 4155 + }, + { + "epoch": 0.54, + "learning_rate": 1.992421148587353e-05, + "loss": 0.0891, + "step": 4156 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924146303913548e-05, + "loss": 0.1042, + "step": 4157 + }, + { + "epoch": 0.54, + "learning_rate": 1.992408109404244e-05, + "loss": 0.0885, + "step": 4158 + }, + { + "epoch": 0.54, + "learning_rate": 1.9924015856260374e-05, + "loss": 0.1062, + "step": 4159 + }, + { + "epoch": 0.54, + "learning_rate": 1.9923950590567543e-05, + "loss": 0.0919, + "step": 4160 + }, + { + "epoch": 0.55, + "learning_rate": 1.992388529696413e-05, + "loss": 0.087, + "step": 4161 + }, + { + "epoch": 0.55, + "learning_rate": 1.992381997545032e-05, + "loss": 0.0885, + "step": 4162 + }, + { + "epoch": 0.55, + "learning_rate": 1.992375462602629e-05, + "loss": 0.0914, + "step": 4163 + }, + { + "epoch": 0.55, + "learning_rate": 1.9923689248692235e-05, + "loss": 0.094, + "step": 4164 + }, + { + "epoch": 0.55, + "learning_rate": 1.992362384344833e-05, + "loss": 0.1032, + "step": 4165 + }, + { + "epoch": 0.55, + "learning_rate": 1.9923558410294762e-05, + "loss": 0.1004, + "step": 4166 + }, + { + "epoch": 0.55, + "learning_rate": 1.9923492949231712e-05, + "loss": 0.0857, + "step": 4167 + }, + { + "epoch": 0.55, + "learning_rate": 1.992342746025937e-05, + "loss": 0.0938, + "step": 4168 + }, + { + "epoch": 0.55, + "learning_rate": 1.9923361943377913e-05, + "loss": 0.0872, + "step": 4169 + }, + { + "epoch": 0.55, + "learning_rate": 1.9923296398587534e-05, + "loss": 0.081, + "step": 4170 + }, + { + "epoch": 0.55, + "learning_rate": 1.9923230825888408e-05, + "loss": 0.1037, + "step": 4171 + }, + { + "epoch": 0.55, + "learning_rate": 1.992316522528072e-05, + "loss": 0.0965, + "step": 4172 + }, + { + "epoch": 0.55, + "learning_rate": 1.992309959676467e-05, + "loss": 0.1001, + "step": 4173 + }, + { + "epoch": 0.55, + "learning_rate": 1.9923033940340422e-05, + "loss": 0.1024, + "step": 4174 + }, + { + "epoch": 0.55, + "learning_rate": 1.992296825600817e-05, + "loss": 0.095, + "step": 4175 + }, + { + "epoch": 0.55, + "learning_rate": 1.9922902543768102e-05, + "loss": 0.0922, + "step": 4176 + }, + { + "epoch": 0.55, + "learning_rate": 1.9922836803620395e-05, + "loss": 0.0937, + "step": 4177 + }, + { + "epoch": 0.55, + "learning_rate": 1.992277103556524e-05, + "loss": 0.0997, + "step": 4178 + }, + { + "epoch": 0.55, + "learning_rate": 1.992270523960282e-05, + "loss": 0.0873, + "step": 4179 + }, + { + "epoch": 0.55, + "learning_rate": 1.992263941573332e-05, + "loss": 0.09, + "step": 4180 + }, + { + "epoch": 0.55, + "learning_rate": 1.9922573563956922e-05, + "loss": 0.0957, + "step": 4181 + }, + { + "epoch": 0.55, + "learning_rate": 1.992250768427382e-05, + "loss": 0.0883, + "step": 4182 + }, + { + "epoch": 0.55, + "learning_rate": 1.9922441776684187e-05, + "loss": 0.0914, + "step": 4183 + }, + { + "epoch": 0.55, + "learning_rate": 1.992237584118822e-05, + "loss": 0.0849, + "step": 4184 + }, + { + "epoch": 0.55, + "learning_rate": 1.9922309877786097e-05, + "loss": 0.094, + "step": 4185 + }, + { + "epoch": 0.55, + "learning_rate": 1.9922243886478006e-05, + "loss": 0.0907, + "step": 4186 + }, + { + "epoch": 0.55, + "learning_rate": 1.9922177867264134e-05, + "loss": 0.093, + "step": 4187 + }, + { + "epoch": 0.55, + "learning_rate": 1.9922111820144663e-05, + "loss": 0.0954, + "step": 4188 + }, + { + "epoch": 0.55, + "learning_rate": 1.9922045745119783e-05, + "loss": 0.0994, + "step": 4189 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921979642189677e-05, + "loss": 0.0931, + "step": 4190 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921913511354528e-05, + "loss": 0.1036, + "step": 4191 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921847352614532e-05, + "loss": 0.0855, + "step": 4192 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921781165969866e-05, + "loss": 0.0951, + "step": 4193 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921714951420716e-05, + "loss": 0.0926, + "step": 4194 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921648708967275e-05, + "loss": 0.099, + "step": 4195 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921582438609723e-05, + "loss": 0.0975, + "step": 4196 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921516140348246e-05, + "loss": 0.0962, + "step": 4197 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921449814183037e-05, + "loss": 0.0902, + "step": 4198 + }, + { + "epoch": 0.55, + "learning_rate": 1.992138346011428e-05, + "loss": 0.1013, + "step": 4199 + }, + { + "epoch": 0.55, + "learning_rate": 1.992131707814216e-05, + "loss": 0.0982, + "step": 4200 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921250668266862e-05, + "loss": 0.1005, + "step": 4201 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921184230488576e-05, + "loss": 0.0897, + "step": 4202 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921117764807486e-05, + "loss": 0.0913, + "step": 4203 + }, + { + "epoch": 0.55, + "learning_rate": 1.9921051271223782e-05, + "loss": 0.0975, + "step": 4204 + }, + { + "epoch": 0.55, + "learning_rate": 1.992098474973765e-05, + "loss": 0.0944, + "step": 4205 + }, + { + "epoch": 0.55, + "learning_rate": 1.9920918200349276e-05, + "loss": 0.0922, + "step": 4206 + }, + { + "epoch": 0.55, + "learning_rate": 1.992085162305885e-05, + "loss": 0.086, + "step": 4207 + }, + { + "epoch": 0.55, + "learning_rate": 1.9920785017866554e-05, + "loss": 0.0827, + "step": 4208 + }, + { + "epoch": 0.55, + "learning_rate": 1.992071838477258e-05, + "loss": 0.1024, + "step": 4209 + }, + { + "epoch": 0.55, + "learning_rate": 1.9920651723777117e-05, + "loss": 0.0999, + "step": 4210 + }, + { + "epoch": 0.55, + "learning_rate": 1.9920585034880345e-05, + "loss": 0.1016, + "step": 4211 + }, + { + "epoch": 0.55, + "learning_rate": 1.9920518318082457e-05, + "loss": 0.0973, + "step": 4212 + }, + { + "epoch": 0.55, + "learning_rate": 1.992045157338364e-05, + "loss": 0.0982, + "step": 4213 + }, + { + "epoch": 0.55, + "learning_rate": 1.9920384800784077e-05, + "loss": 0.094, + "step": 4214 + }, + { + "epoch": 0.55, + "learning_rate": 1.9920318000283965e-05, + "loss": 0.0983, + "step": 4215 + }, + { + "epoch": 0.55, + "learning_rate": 1.9920251171883486e-05, + "loss": 0.0849, + "step": 4216 + }, + { + "epoch": 0.55, + "learning_rate": 1.992018431558283e-05, + "loss": 0.099, + "step": 4217 + }, + { + "epoch": 0.55, + "learning_rate": 1.992011743138218e-05, + "loss": 0.0829, + "step": 4218 + }, + { + "epoch": 0.55, + "learning_rate": 1.9920050519281733e-05, + "loss": 0.099, + "step": 4219 + }, + { + "epoch": 0.55, + "learning_rate": 1.9919983579281668e-05, + "loss": 0.0959, + "step": 4220 + }, + { + "epoch": 0.55, + "learning_rate": 1.991991661138218e-05, + "loss": 0.0858, + "step": 4221 + }, + { + "epoch": 0.55, + "learning_rate": 1.9919849615583454e-05, + "loss": 0.1035, + "step": 4222 + }, + { + "epoch": 0.55, + "learning_rate": 1.9919782591885677e-05, + "loss": 0.0991, + "step": 4223 + }, + { + "epoch": 0.55, + "learning_rate": 1.991971554028904e-05, + "loss": 0.0878, + "step": 4224 + }, + { + "epoch": 0.55, + "learning_rate": 1.9919648460793736e-05, + "loss": 0.0899, + "step": 4225 + }, + { + "epoch": 0.55, + "learning_rate": 1.9919581353399946e-05, + "loss": 0.0944, + "step": 4226 + }, + { + "epoch": 0.55, + "learning_rate": 1.9919514218107863e-05, + "loss": 0.0851, + "step": 4227 + }, + { + "epoch": 0.55, + "learning_rate": 1.9919447054917672e-05, + "loss": 0.1033, + "step": 4228 + }, + { + "epoch": 0.55, + "learning_rate": 1.9919379863829563e-05, + "loss": 0.1066, + "step": 4229 + }, + { + "epoch": 0.55, + "learning_rate": 1.991931264484373e-05, + "loss": 0.0919, + "step": 4230 + }, + { + "epoch": 0.55, + "learning_rate": 1.991924539796036e-05, + "loss": 0.1071, + "step": 4231 + }, + { + "epoch": 0.55, + "learning_rate": 1.9919178123179638e-05, + "loss": 0.0998, + "step": 4232 + }, + { + "epoch": 0.55, + "learning_rate": 1.9919110820501756e-05, + "loss": 0.0941, + "step": 4233 + }, + { + "epoch": 0.55, + "learning_rate": 1.99190434899269e-05, + "loss": 0.0941, + "step": 4234 + }, + { + "epoch": 0.55, + "learning_rate": 1.991897613145527e-05, + "loss": 0.0961, + "step": 4235 + }, + { + "epoch": 0.55, + "learning_rate": 1.991890874508704e-05, + "loss": 0.0908, + "step": 4236 + }, + { + "epoch": 0.56, + "learning_rate": 1.991884133082241e-05, + "loss": 0.0836, + "step": 4237 + }, + { + "epoch": 0.56, + "learning_rate": 1.9918773888661567e-05, + "loss": 0.0804, + "step": 4238 + }, + { + "epoch": 0.56, + "learning_rate": 1.99187064186047e-05, + "loss": 0.0962, + "step": 4239 + }, + { + "epoch": 0.56, + "learning_rate": 1.9918638920652e-05, + "loss": 0.0962, + "step": 4240 + }, + { + "epoch": 0.56, + "learning_rate": 1.991857139480366e-05, + "loss": 0.0968, + "step": 4241 + }, + { + "epoch": 0.56, + "learning_rate": 1.991850384105986e-05, + "loss": 0.0883, + "step": 4242 + }, + { + "epoch": 0.56, + "learning_rate": 1.99184362594208e-05, + "loss": 0.0881, + "step": 4243 + }, + { + "epoch": 0.56, + "learning_rate": 1.9918368649886664e-05, + "loss": 0.1032, + "step": 4244 + }, + { + "epoch": 0.56, + "learning_rate": 1.9918301012457647e-05, + "loss": 0.0997, + "step": 4245 + }, + { + "epoch": 0.56, + "learning_rate": 1.991823334713393e-05, + "loss": 0.1026, + "step": 4246 + }, + { + "epoch": 0.56, + "learning_rate": 1.991816565391572e-05, + "loss": 0.0973, + "step": 4247 + }, + { + "epoch": 0.56, + "learning_rate": 1.9918097932803192e-05, + "loss": 0.0867, + "step": 4248 + }, + { + "epoch": 0.56, + "learning_rate": 1.991803018379654e-05, + "loss": 0.0957, + "step": 4249 + }, + { + "epoch": 0.56, + "learning_rate": 1.991796240689596e-05, + "loss": 0.1003, + "step": 4250 + }, + { + "epoch": 0.56, + "learning_rate": 1.991789460210164e-05, + "loss": 0.0877, + "step": 4251 + }, + { + "epoch": 0.56, + "learning_rate": 1.9917826769413764e-05, + "loss": 0.0849, + "step": 4252 + }, + { + "epoch": 0.56, + "learning_rate": 1.991775890883253e-05, + "loss": 0.0975, + "step": 4253 + }, + { + "epoch": 0.56, + "learning_rate": 1.991769102035813e-05, + "loss": 0.1016, + "step": 4254 + }, + { + "epoch": 0.56, + "learning_rate": 1.991762310399075e-05, + "loss": 0.0833, + "step": 4255 + }, + { + "epoch": 0.56, + "learning_rate": 1.991755515973058e-05, + "loss": 0.0876, + "step": 4256 + }, + { + "epoch": 0.56, + "learning_rate": 1.9917487187577823e-05, + "loss": 0.0923, + "step": 4257 + }, + { + "epoch": 0.56, + "learning_rate": 1.9917419187532653e-05, + "loss": 0.0935, + "step": 4258 + }, + { + "epoch": 0.56, + "learning_rate": 1.9917351159595274e-05, + "loss": 0.096, + "step": 4259 + }, + { + "epoch": 0.56, + "learning_rate": 1.9917283103765873e-05, + "loss": 0.104, + "step": 4260 + }, + { + "epoch": 0.56, + "learning_rate": 1.991721502004464e-05, + "loss": 0.0934, + "step": 4261 + }, + { + "epoch": 0.56, + "learning_rate": 1.9917146908431765e-05, + "loss": 0.0883, + "step": 4262 + }, + { + "epoch": 0.56, + "learning_rate": 1.9917078768927446e-05, + "loss": 0.0973, + "step": 4263 + }, + { + "epoch": 0.56, + "learning_rate": 1.9917010601531876e-05, + "loss": 0.0966, + "step": 4264 + }, + { + "epoch": 0.56, + "learning_rate": 1.9916942406245235e-05, + "loss": 0.0801, + "step": 4265 + }, + { + "epoch": 0.56, + "learning_rate": 1.991687418306772e-05, + "loss": 0.1006, + "step": 4266 + }, + { + "epoch": 0.56, + "learning_rate": 1.991680593199953e-05, + "loss": 0.0985, + "step": 4267 + }, + { + "epoch": 0.56, + "learning_rate": 1.9916737653040847e-05, + "loss": 0.0923, + "step": 4268 + }, + { + "epoch": 0.56, + "learning_rate": 1.991666934619187e-05, + "loss": 0.0914, + "step": 4269 + }, + { + "epoch": 0.56, + "learning_rate": 1.991660101145279e-05, + "loss": 0.0968, + "step": 4270 + }, + { + "epoch": 0.56, + "learning_rate": 1.9916532648823794e-05, + "loss": 0.0995, + "step": 4271 + }, + { + "epoch": 0.56, + "learning_rate": 1.9916464258305082e-05, + "loss": 0.0889, + "step": 4272 + }, + { + "epoch": 0.56, + "learning_rate": 1.9916395839896838e-05, + "loss": 0.0906, + "step": 4273 + }, + { + "epoch": 0.56, + "learning_rate": 1.991632739359926e-05, + "loss": 0.096, + "step": 4274 + }, + { + "epoch": 0.56, + "learning_rate": 1.9916258919412542e-05, + "loss": 0.1007, + "step": 4275 + }, + { + "epoch": 0.56, + "learning_rate": 1.991619041733687e-05, + "loss": 0.0836, + "step": 4276 + }, + { + "epoch": 0.56, + "learning_rate": 1.9916121887372442e-05, + "loss": 0.0921, + "step": 4277 + }, + { + "epoch": 0.56, + "learning_rate": 1.991605332951945e-05, + "loss": 0.0969, + "step": 4278 + }, + { + "epoch": 0.56, + "learning_rate": 1.9915984743778086e-05, + "loss": 0.0976, + "step": 4279 + }, + { + "epoch": 0.56, + "learning_rate": 1.9915916130148544e-05, + "loss": 0.096, + "step": 4280 + }, + { + "epoch": 0.56, + "learning_rate": 1.9915847488631015e-05, + "loss": 0.0937, + "step": 4281 + }, + { + "epoch": 0.56, + "learning_rate": 1.9915778819225692e-05, + "loss": 0.0924, + "step": 4282 + }, + { + "epoch": 0.56, + "learning_rate": 1.991571012193277e-05, + "loss": 0.1001, + "step": 4283 + }, + { + "epoch": 0.56, + "learning_rate": 1.9915641396752443e-05, + "loss": 0.0967, + "step": 4284 + }, + { + "epoch": 0.56, + "learning_rate": 1.99155726436849e-05, + "loss": 0.0871, + "step": 4285 + }, + { + "epoch": 0.56, + "learning_rate": 1.991550386273034e-05, + "loss": 0.0921, + "step": 4286 + }, + { + "epoch": 0.56, + "learning_rate": 1.991543505388895e-05, + "loss": 0.1013, + "step": 4287 + }, + { + "epoch": 0.56, + "learning_rate": 1.9915366217160927e-05, + "loss": 0.0968, + "step": 4288 + }, + { + "epoch": 0.56, + "learning_rate": 1.9915297352546467e-05, + "loss": 0.0777, + "step": 4289 + }, + { + "epoch": 0.56, + "learning_rate": 1.9915228460045757e-05, + "loss": 0.0918, + "step": 4290 + }, + { + "epoch": 0.56, + "learning_rate": 1.9915159539659e-05, + "loss": 0.0888, + "step": 4291 + }, + { + "epoch": 0.56, + "learning_rate": 1.9915090591386384e-05, + "loss": 0.084, + "step": 4292 + }, + { + "epoch": 0.56, + "learning_rate": 1.99150216152281e-05, + "loss": 0.0956, + "step": 4293 + }, + { + "epoch": 0.56, + "learning_rate": 1.9914952611184352e-05, + "loss": 0.0914, + "step": 4294 + }, + { + "epoch": 0.56, + "learning_rate": 1.991488357925532e-05, + "loss": 0.0898, + "step": 4295 + }, + { + "epoch": 0.56, + "learning_rate": 1.9914814519441213e-05, + "loss": 0.0944, + "step": 4296 + }, + { + "epoch": 0.56, + "learning_rate": 1.991474543174221e-05, + "loss": 0.0962, + "step": 4297 + }, + { + "epoch": 0.56, + "learning_rate": 1.9914676316158522e-05, + "loss": 0.0961, + "step": 4298 + }, + { + "epoch": 0.56, + "learning_rate": 1.9914607172690333e-05, + "loss": 0.1013, + "step": 4299 + }, + { + "epoch": 0.56, + "learning_rate": 1.9914538001337835e-05, + "loss": 0.0889, + "step": 4300 + }, + { + "epoch": 0.56, + "learning_rate": 1.991446880210123e-05, + "loss": 0.0992, + "step": 4301 + }, + { + "epoch": 0.56, + "learning_rate": 1.991439957498071e-05, + "loss": 0.1087, + "step": 4302 + }, + { + "epoch": 0.56, + "learning_rate": 1.9914330319976466e-05, + "loss": 0.0827, + "step": 4303 + }, + { + "epoch": 0.56, + "learning_rate": 1.9914261037088696e-05, + "loss": 0.108, + "step": 4304 + }, + { + "epoch": 0.56, + "learning_rate": 1.99141917263176e-05, + "loss": 0.0913, + "step": 4305 + }, + { + "epoch": 0.56, + "learning_rate": 1.9914122387663364e-05, + "loss": 0.0877, + "step": 4306 + }, + { + "epoch": 0.56, + "learning_rate": 1.991405302112619e-05, + "loss": 0.0938, + "step": 4307 + }, + { + "epoch": 0.56, + "learning_rate": 1.9913983626706264e-05, + "loss": 0.0798, + "step": 4308 + }, + { + "epoch": 0.56, + "learning_rate": 1.9913914204403793e-05, + "loss": 0.091, + "step": 4309 + }, + { + "epoch": 0.56, + "learning_rate": 1.9913844754218962e-05, + "loss": 0.0854, + "step": 4310 + }, + { + "epoch": 0.56, + "learning_rate": 1.9913775276151972e-05, + "loss": 0.091, + "step": 4311 + }, + { + "epoch": 0.56, + "learning_rate": 1.991370577020302e-05, + "loss": 0.0886, + "step": 4312 + }, + { + "epoch": 0.56, + "learning_rate": 1.9913636236372293e-05, + "loss": 0.1056, + "step": 4313 + }, + { + "epoch": 0.57, + "learning_rate": 1.9913566674659993e-05, + "loss": 0.0849, + "step": 4314 + }, + { + "epoch": 0.57, + "learning_rate": 1.9913497085066316e-05, + "loss": 0.0967, + "step": 4315 + }, + { + "epoch": 0.57, + "learning_rate": 1.991342746759146e-05, + "loss": 0.0878, + "step": 4316 + }, + { + "epoch": 0.57, + "learning_rate": 1.991335782223561e-05, + "loss": 0.0779, + "step": 4317 + }, + { + "epoch": 0.57, + "learning_rate": 1.9913288148998972e-05, + "loss": 0.092, + "step": 4318 + }, + { + "epoch": 0.57, + "learning_rate": 1.9913218447881743e-05, + "loss": 0.1015, + "step": 4319 + }, + { + "epoch": 0.57, + "learning_rate": 1.991314871888411e-05, + "loss": 0.0953, + "step": 4320 + }, + { + "epoch": 0.57, + "learning_rate": 1.9913078962006276e-05, + "loss": 0.0925, + "step": 4321 + }, + { + "epoch": 0.57, + "learning_rate": 1.9913009177248436e-05, + "loss": 0.0905, + "step": 4322 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912939364610784e-05, + "loss": 0.0977, + "step": 4323 + }, + { + "epoch": 0.57, + "learning_rate": 1.991286952409352e-05, + "loss": 0.0859, + "step": 4324 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912799655696835e-05, + "loss": 0.1081, + "step": 4325 + }, + { + "epoch": 0.57, + "learning_rate": 1.991272975942093e-05, + "loss": 0.0948, + "step": 4326 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912659835266e-05, + "loss": 0.0821, + "step": 4327 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912589883232245e-05, + "loss": 0.0937, + "step": 4328 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912519903319857e-05, + "loss": 0.0889, + "step": 4329 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912449895529036e-05, + "loss": 0.099, + "step": 4330 + }, + { + "epoch": 0.57, + "learning_rate": 1.991237985985998e-05, + "loss": 0.0969, + "step": 4331 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912309796312876e-05, + "loss": 0.092, + "step": 4332 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912239704887933e-05, + "loss": 0.09, + "step": 4333 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912169585585343e-05, + "loss": 0.0966, + "step": 4334 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912099438405303e-05, + "loss": 0.0989, + "step": 4335 + }, + { + "epoch": 0.57, + "learning_rate": 1.9912029263348013e-05, + "loss": 0.0949, + "step": 4336 + }, + { + "epoch": 0.57, + "learning_rate": 1.991195906041367e-05, + "loss": 0.1008, + "step": 4337 + }, + { + "epoch": 0.57, + "learning_rate": 1.9911888829602465e-05, + "loss": 0.0896, + "step": 4338 + }, + { + "epoch": 0.57, + "learning_rate": 1.9911818570914607e-05, + "loss": 0.0965, + "step": 4339 + }, + { + "epoch": 0.57, + "learning_rate": 1.991174828435028e-05, + "loss": 0.0948, + "step": 4340 + }, + { + "epoch": 0.57, + "learning_rate": 1.9911677969909694e-05, + "loss": 0.1004, + "step": 4341 + }, + { + "epoch": 0.57, + "learning_rate": 1.991160762759304e-05, + "loss": 0.091, + "step": 4342 + }, + { + "epoch": 0.57, + "learning_rate": 1.9911537257400514e-05, + "loss": 0.0878, + "step": 4343 + }, + { + "epoch": 0.57, + "learning_rate": 1.991146685933232e-05, + "loss": 0.0821, + "step": 4344 + }, + { + "epoch": 0.57, + "learning_rate": 1.9911396433388656e-05, + "loss": 0.1014, + "step": 4345 + }, + { + "epoch": 0.57, + "learning_rate": 1.9911325979569716e-05, + "loss": 0.0989, + "step": 4346 + }, + { + "epoch": 0.57, + "learning_rate": 1.9911255497875697e-05, + "loss": 0.0937, + "step": 4347 + }, + { + "epoch": 0.57, + "learning_rate": 1.99111849883068e-05, + "loss": 0.0912, + "step": 4348 + }, + { + "epoch": 0.57, + "learning_rate": 1.991111445086323e-05, + "loss": 0.0813, + "step": 4349 + }, + { + "epoch": 0.57, + "learning_rate": 1.9911043885545168e-05, + "loss": 0.1044, + "step": 4350 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910973292352828e-05, + "loss": 0.0859, + "step": 4351 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910902671286402e-05, + "loss": 0.1025, + "step": 4352 + }, + { + "epoch": 0.57, + "learning_rate": 1.991083202234609e-05, + "loss": 0.0978, + "step": 4353 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910761345532094e-05, + "loss": 0.0867, + "step": 4354 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910690640844604e-05, + "loss": 0.1016, + "step": 4355 + }, + { + "epoch": 0.57, + "learning_rate": 1.991061990828383e-05, + "loss": 0.0951, + "step": 4356 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910549147849966e-05, + "loss": 0.0847, + "step": 4357 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910478359543204e-05, + "loss": 0.0952, + "step": 4358 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910407543363752e-05, + "loss": 0.0885, + "step": 4359 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910336699311807e-05, + "loss": 0.0995, + "step": 4360 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910265827387565e-05, + "loss": 0.0858, + "step": 4361 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910194927591237e-05, + "loss": 0.0864, + "step": 4362 + }, + { + "epoch": 0.57, + "learning_rate": 1.9910123999923005e-05, + "loss": 0.0965, + "step": 4363 + }, + { + "epoch": 0.57, + "learning_rate": 1.991005304438308e-05, + "loss": 0.1035, + "step": 4364 + }, + { + "epoch": 0.57, + "learning_rate": 1.9909982060971656e-05, + "loss": 0.0991, + "step": 4365 + }, + { + "epoch": 0.57, + "learning_rate": 1.990991104968894e-05, + "loss": 0.0899, + "step": 4366 + }, + { + "epoch": 0.57, + "learning_rate": 1.990984001053512e-05, + "loss": 0.0948, + "step": 4367 + }, + { + "epoch": 0.57, + "learning_rate": 1.990976894351041e-05, + "loss": 0.099, + "step": 4368 + }, + { + "epoch": 0.57, + "learning_rate": 1.9909697848614997e-05, + "loss": 0.1023, + "step": 4369 + }, + { + "epoch": 0.57, + "learning_rate": 1.990962672584909e-05, + "loss": 0.1145, + "step": 4370 + }, + { + "epoch": 0.57, + "learning_rate": 1.9909555575212885e-05, + "loss": 0.0871, + "step": 4371 + }, + { + "epoch": 0.57, + "learning_rate": 1.9909484396706582e-05, + "loss": 0.0879, + "step": 4372 + }, + { + "epoch": 0.57, + "learning_rate": 1.9909413190330382e-05, + "loss": 0.1055, + "step": 4373 + }, + { + "epoch": 0.57, + "learning_rate": 1.9909341956084483e-05, + "loss": 0.091, + "step": 4374 + }, + { + "epoch": 0.57, + "learning_rate": 1.990927069396909e-05, + "loss": 0.0879, + "step": 4375 + }, + { + "epoch": 0.57, + "learning_rate": 1.9909199403984402e-05, + "loss": 0.0901, + "step": 4376 + }, + { + "epoch": 0.57, + "learning_rate": 1.9909128086130615e-05, + "loss": 0.1096, + "step": 4377 + }, + { + "epoch": 0.57, + "learning_rate": 1.9909056740407937e-05, + "loss": 0.0881, + "step": 4378 + }, + { + "epoch": 0.57, + "learning_rate": 1.9908985366816563e-05, + "loss": 0.0849, + "step": 4379 + }, + { + "epoch": 0.57, + "learning_rate": 1.9908913965356695e-05, + "loss": 0.0985, + "step": 4380 + }, + { + "epoch": 0.57, + "learning_rate": 1.990884253602853e-05, + "loss": 0.088, + "step": 4381 + }, + { + "epoch": 0.57, + "learning_rate": 1.9908771078832278e-05, + "loss": 0.0885, + "step": 4382 + }, + { + "epoch": 0.57, + "learning_rate": 1.9908699593768133e-05, + "loss": 0.0982, + "step": 4383 + }, + { + "epoch": 0.57, + "learning_rate": 1.99086280808363e-05, + "loss": 0.0897, + "step": 4384 + }, + { + "epoch": 0.57, + "learning_rate": 1.9908556540036976e-05, + "loss": 0.0976, + "step": 4385 + }, + { + "epoch": 0.57, + "learning_rate": 1.9908484971370368e-05, + "loss": 0.1069, + "step": 4386 + }, + { + "epoch": 0.57, + "learning_rate": 1.990841337483667e-05, + "loss": 0.0901, + "step": 4387 + }, + { + "epoch": 0.57, + "learning_rate": 1.9908341750436086e-05, + "loss": 0.0917, + "step": 4388 + }, + { + "epoch": 0.57, + "learning_rate": 1.9908270098168823e-05, + "loss": 0.0902, + "step": 4389 + }, + { + "epoch": 0.58, + "learning_rate": 1.9908198418035078e-05, + "loss": 0.0819, + "step": 4390 + }, + { + "epoch": 0.58, + "learning_rate": 1.9908126710035046e-05, + "loss": 0.0946, + "step": 4391 + }, + { + "epoch": 0.58, + "learning_rate": 1.990805497416894e-05, + "loss": 0.0921, + "step": 4392 + }, + { + "epoch": 0.58, + "learning_rate": 1.9907983210436957e-05, + "loss": 0.1003, + "step": 4393 + }, + { + "epoch": 0.58, + "learning_rate": 1.99079114188393e-05, + "loss": 0.0777, + "step": 4394 + }, + { + "epoch": 0.58, + "learning_rate": 1.9907839599376166e-05, + "loss": 0.1127, + "step": 4395 + }, + { + "epoch": 0.58, + "learning_rate": 1.9907767752047763e-05, + "loss": 0.0841, + "step": 4396 + }, + { + "epoch": 0.58, + "learning_rate": 1.9907695876854293e-05, + "loss": 0.0798, + "step": 4397 + }, + { + "epoch": 0.58, + "learning_rate": 1.9907623973795953e-05, + "loss": 0.1043, + "step": 4398 + }, + { + "epoch": 0.58, + "learning_rate": 1.990755204287295e-05, + "loss": 0.097, + "step": 4399 + }, + { + "epoch": 0.58, + "learning_rate": 1.9907480084085483e-05, + "loss": 0.0997, + "step": 4400 + }, + { + "epoch": 0.58, + "learning_rate": 1.9907408097433753e-05, + "loss": 0.0942, + "step": 4401 + }, + { + "epoch": 0.58, + "learning_rate": 1.990733608291797e-05, + "loss": 0.0991, + "step": 4402 + }, + { + "epoch": 0.58, + "learning_rate": 1.990726404053833e-05, + "loss": 0.0998, + "step": 4403 + }, + { + "epoch": 0.58, + "learning_rate": 1.990719197029504e-05, + "loss": 0.0845, + "step": 4404 + }, + { + "epoch": 0.58, + "learning_rate": 1.9907119872188298e-05, + "loss": 0.1034, + "step": 4405 + }, + { + "epoch": 0.58, + "learning_rate": 1.990704774621831e-05, + "loss": 0.0854, + "step": 4406 + }, + { + "epoch": 0.58, + "learning_rate": 1.9906975592385274e-05, + "loss": 0.1029, + "step": 4407 + }, + { + "epoch": 0.58, + "learning_rate": 1.99069034106894e-05, + "loss": 0.0971, + "step": 4408 + }, + { + "epoch": 0.58, + "learning_rate": 1.9906831201130886e-05, + "loss": 0.1, + "step": 4409 + }, + { + "epoch": 0.58, + "learning_rate": 1.990675896370994e-05, + "loss": 0.0908, + "step": 4410 + }, + { + "epoch": 0.58, + "learning_rate": 1.990668669842676e-05, + "loss": 0.0943, + "step": 4411 + }, + { + "epoch": 0.58, + "learning_rate": 1.990661440528155e-05, + "loss": 0.0832, + "step": 4412 + }, + { + "epoch": 0.58, + "learning_rate": 1.9906542084274515e-05, + "loss": 0.0868, + "step": 4413 + }, + { + "epoch": 0.58, + "learning_rate": 1.990646973540586e-05, + "loss": 0.1026, + "step": 4414 + }, + { + "epoch": 0.58, + "learning_rate": 1.9906397358675785e-05, + "loss": 0.0974, + "step": 4415 + }, + { + "epoch": 0.58, + "learning_rate": 1.9906324954084497e-05, + "loss": 0.0748, + "step": 4416 + }, + { + "epoch": 0.58, + "learning_rate": 1.9906252521632194e-05, + "loss": 0.0922, + "step": 4417 + }, + { + "epoch": 0.58, + "learning_rate": 1.9906180061319085e-05, + "loss": 0.0726, + "step": 4418 + }, + { + "epoch": 0.58, + "learning_rate": 1.9906107573145374e-05, + "loss": 0.0929, + "step": 4419 + }, + { + "epoch": 0.58, + "learning_rate": 1.990603505711126e-05, + "loss": 0.0895, + "step": 4420 + }, + { + "epoch": 0.58, + "learning_rate": 1.990596251321695e-05, + "loss": 0.0961, + "step": 4421 + }, + { + "epoch": 0.58, + "learning_rate": 1.990588994146265e-05, + "loss": 0.0898, + "step": 4422 + }, + { + "epoch": 0.58, + "learning_rate": 1.990581734184856e-05, + "loss": 0.102, + "step": 4423 + }, + { + "epoch": 0.58, + "learning_rate": 1.9905744714374888e-05, + "loss": 0.0886, + "step": 4424 + }, + { + "epoch": 0.58, + "learning_rate": 1.9905672059041837e-05, + "loss": 0.0794, + "step": 4425 + }, + { + "epoch": 0.58, + "learning_rate": 1.990559937584961e-05, + "loss": 0.0966, + "step": 4426 + }, + { + "epoch": 0.58, + "learning_rate": 1.990552666479841e-05, + "loss": 0.0889, + "step": 4427 + }, + { + "epoch": 0.58, + "learning_rate": 1.9905453925888448e-05, + "loss": 0.0923, + "step": 4428 + }, + { + "epoch": 0.58, + "learning_rate": 1.9905381159119922e-05, + "loss": 0.0877, + "step": 4429 + }, + { + "epoch": 0.58, + "learning_rate": 1.9905308364493035e-05, + "loss": 0.098, + "step": 4430 + }, + { + "epoch": 0.58, + "learning_rate": 1.9905235542008e-05, + "loss": 0.0925, + "step": 4431 + }, + { + "epoch": 0.58, + "learning_rate": 1.9905162691665017e-05, + "loss": 0.0979, + "step": 4432 + }, + { + "epoch": 0.58, + "learning_rate": 1.9905089813464293e-05, + "loss": 0.1091, + "step": 4433 + }, + { + "epoch": 0.58, + "learning_rate": 1.990501690740603e-05, + "loss": 0.0876, + "step": 4434 + }, + { + "epoch": 0.58, + "learning_rate": 1.9904943973490437e-05, + "loss": 0.0923, + "step": 4435 + }, + { + "epoch": 0.58, + "learning_rate": 1.9904871011717712e-05, + "loss": 0.1079, + "step": 4436 + }, + { + "epoch": 0.58, + "learning_rate": 1.990479802208807e-05, + "loss": 0.1068, + "step": 4437 + }, + { + "epoch": 0.58, + "learning_rate": 1.9904725004601704e-05, + "loss": 0.0907, + "step": 4438 + }, + { + "epoch": 0.58, + "learning_rate": 1.990465195925883e-05, + "loss": 0.0892, + "step": 4439 + }, + { + "epoch": 0.58, + "learning_rate": 1.990457888605965e-05, + "loss": 0.085, + "step": 4440 + }, + { + "epoch": 0.58, + "learning_rate": 1.990450578500437e-05, + "loss": 0.0921, + "step": 4441 + }, + { + "epoch": 0.58, + "learning_rate": 1.9904432656093194e-05, + "loss": 0.1007, + "step": 4442 + }, + { + "epoch": 0.58, + "learning_rate": 1.990435949932633e-05, + "loss": 0.0905, + "step": 4443 + }, + { + "epoch": 0.58, + "learning_rate": 1.990428631470398e-05, + "loss": 0.1039, + "step": 4444 + }, + { + "epoch": 0.58, + "learning_rate": 1.9904213102226354e-05, + "loss": 0.0833, + "step": 4445 + }, + { + "epoch": 0.58, + "learning_rate": 1.9904139861893657e-05, + "loss": 0.0863, + "step": 4446 + }, + { + "epoch": 0.58, + "learning_rate": 1.990406659370609e-05, + "loss": 0.1027, + "step": 4447 + }, + { + "epoch": 0.58, + "learning_rate": 1.9903993297663866e-05, + "loss": 0.0949, + "step": 4448 + }, + { + "epoch": 0.58, + "learning_rate": 1.990391997376719e-05, + "loss": 0.091, + "step": 4449 + }, + { + "epoch": 0.58, + "learning_rate": 1.9903846622016263e-05, + "loss": 0.1024, + "step": 4450 + }, + { + "epoch": 0.58, + "learning_rate": 1.99037732424113e-05, + "loss": 0.104, + "step": 4451 + }, + { + "epoch": 0.58, + "learning_rate": 1.9903699834952493e-05, + "loss": 0.0909, + "step": 4452 + }, + { + "epoch": 0.58, + "learning_rate": 1.9903626399640066e-05, + "loss": 0.0873, + "step": 4453 + }, + { + "epoch": 0.58, + "learning_rate": 1.9903552936474215e-05, + "loss": 0.0977, + "step": 4454 + }, + { + "epoch": 0.58, + "learning_rate": 1.9903479445455145e-05, + "loss": 0.0949, + "step": 4455 + }, + { + "epoch": 0.58, + "learning_rate": 1.9903405926583067e-05, + "loss": 0.088, + "step": 4456 + }, + { + "epoch": 0.58, + "learning_rate": 1.9903332379858194e-05, + "loss": 0.0889, + "step": 4457 + }, + { + "epoch": 0.58, + "learning_rate": 1.990325880528072e-05, + "loss": 0.0992, + "step": 4458 + }, + { + "epoch": 0.58, + "learning_rate": 1.990318520285086e-05, + "loss": 0.0946, + "step": 4459 + }, + { + "epoch": 0.58, + "learning_rate": 1.990311157256882e-05, + "loss": 0.0943, + "step": 4460 + }, + { + "epoch": 0.58, + "learning_rate": 1.99030379144348e-05, + "loss": 0.095, + "step": 4461 + }, + { + "epoch": 0.58, + "learning_rate": 1.990296422844902e-05, + "loss": 0.0966, + "step": 4462 + }, + { + "epoch": 0.58, + "learning_rate": 1.990289051461168e-05, + "loss": 0.0846, + "step": 4463 + }, + { + "epoch": 0.58, + "learning_rate": 1.9902816772922984e-05, + "loss": 0.0873, + "step": 4464 + }, + { + "epoch": 0.58, + "learning_rate": 1.9902743003383147e-05, + "loss": 0.0972, + "step": 4465 + }, + { + "epoch": 0.59, + "learning_rate": 1.9902669205992372e-05, + "loss": 0.079, + "step": 4466 + }, + { + "epoch": 0.59, + "learning_rate": 1.9902595380750865e-05, + "loss": 0.106, + "step": 4467 + }, + { + "epoch": 0.59, + "learning_rate": 1.9902521527658834e-05, + "loss": 0.0934, + "step": 4468 + }, + { + "epoch": 0.59, + "learning_rate": 1.9902447646716495e-05, + "loss": 0.1059, + "step": 4469 + }, + { + "epoch": 0.59, + "learning_rate": 1.9902373737924045e-05, + "loss": 0.105, + "step": 4470 + }, + { + "epoch": 0.59, + "learning_rate": 1.9902299801281697e-05, + "loss": 0.0921, + "step": 4471 + }, + { + "epoch": 0.59, + "learning_rate": 1.9902225836789658e-05, + "loss": 0.0917, + "step": 4472 + }, + { + "epoch": 0.59, + "learning_rate": 1.9902151844448138e-05, + "loss": 0.0903, + "step": 4473 + }, + { + "epoch": 0.59, + "learning_rate": 1.990207782425734e-05, + "loss": 0.0962, + "step": 4474 + }, + { + "epoch": 0.59, + "learning_rate": 1.9902003776217478e-05, + "loss": 0.092, + "step": 4475 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901929700328754e-05, + "loss": 0.0942, + "step": 4476 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901855596591387e-05, + "loss": 0.1078, + "step": 4477 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901781465005575e-05, + "loss": 0.0914, + "step": 4478 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901707305571525e-05, + "loss": 0.09, + "step": 4479 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901633118289455e-05, + "loss": 0.0855, + "step": 4480 + }, + { + "epoch": 0.59, + "learning_rate": 1.990155890315957e-05, + "loss": 0.0815, + "step": 4481 + }, + { + "epoch": 0.59, + "learning_rate": 1.990148466018208e-05, + "loss": 0.0786, + "step": 4482 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901410389357186e-05, + "loss": 0.1024, + "step": 4483 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901336090685103e-05, + "loss": 0.0993, + "step": 4484 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901261764166042e-05, + "loss": 0.0912, + "step": 4485 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901187409800208e-05, + "loss": 0.0839, + "step": 4486 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901113027587816e-05, + "loss": 0.0916, + "step": 4487 + }, + { + "epoch": 0.59, + "learning_rate": 1.9901038617529064e-05, + "loss": 0.1016, + "step": 4488 + }, + { + "epoch": 0.59, + "learning_rate": 1.9900964179624175e-05, + "loss": 0.0924, + "step": 4489 + }, + { + "epoch": 0.59, + "learning_rate": 1.990088971387335e-05, + "loss": 0.099, + "step": 4490 + }, + { + "epoch": 0.59, + "learning_rate": 1.9900815220276795e-05, + "loss": 0.0907, + "step": 4491 + }, + { + "epoch": 0.59, + "learning_rate": 1.9900740698834728e-05, + "loss": 0.095, + "step": 4492 + }, + { + "epoch": 0.59, + "learning_rate": 1.9900666149547354e-05, + "loss": 0.0904, + "step": 4493 + }, + { + "epoch": 0.59, + "learning_rate": 1.9900591572414884e-05, + "loss": 0.0882, + "step": 4494 + }, + { + "epoch": 0.59, + "learning_rate": 1.990051696743753e-05, + "loss": 0.0962, + "step": 4495 + }, + { + "epoch": 0.59, + "learning_rate": 1.9900442334615495e-05, + "loss": 0.0911, + "step": 4496 + }, + { + "epoch": 0.59, + "learning_rate": 1.9900367673948998e-05, + "loss": 0.0885, + "step": 4497 + }, + { + "epoch": 0.59, + "learning_rate": 1.9900292985438237e-05, + "loss": 0.1036, + "step": 4498 + }, + { + "epoch": 0.59, + "learning_rate": 1.9900218269083437e-05, + "loss": 0.0957, + "step": 4499 + }, + { + "epoch": 0.59, + "learning_rate": 1.9900143524884793e-05, + "loss": 0.096, + "step": 4500 + }, + { + "epoch": 0.59, + "learning_rate": 1.990006875284253e-05, + "loss": 0.0965, + "step": 4501 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899993952956847e-05, + "loss": 0.095, + "step": 4502 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899919125227957e-05, + "loss": 0.0987, + "step": 4503 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899844269656072e-05, + "loss": 0.1001, + "step": 4504 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899769386241407e-05, + "loss": 0.1047, + "step": 4505 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899694474984163e-05, + "loss": 0.098, + "step": 4506 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899619535884556e-05, + "loss": 0.0874, + "step": 4507 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899544568942796e-05, + "loss": 0.0933, + "step": 4508 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899469574159095e-05, + "loss": 0.1072, + "step": 4509 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899394551533664e-05, + "loss": 0.0896, + "step": 4510 + }, + { + "epoch": 0.59, + "learning_rate": 1.989931950106671e-05, + "loss": 0.1001, + "step": 4511 + }, + { + "epoch": 0.59, + "learning_rate": 1.989924442275845e-05, + "loss": 0.0915, + "step": 4512 + }, + { + "epoch": 0.59, + "learning_rate": 1.989916931660909e-05, + "loss": 0.1042, + "step": 4513 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899094182618847e-05, + "loss": 0.0984, + "step": 4514 + }, + { + "epoch": 0.59, + "learning_rate": 1.9899019020787926e-05, + "loss": 0.0908, + "step": 4515 + }, + { + "epoch": 0.59, + "learning_rate": 1.9898943831116542e-05, + "loss": 0.1031, + "step": 4516 + }, + { + "epoch": 0.59, + "learning_rate": 1.9898868613604903e-05, + "loss": 0.0836, + "step": 4517 + }, + { + "epoch": 0.59, + "learning_rate": 1.9898793368253226e-05, + "loss": 0.0868, + "step": 4518 + }, + { + "epoch": 0.59, + "learning_rate": 1.989871809506172e-05, + "loss": 0.0949, + "step": 4519 + }, + { + "epoch": 0.59, + "learning_rate": 1.9898642794030597e-05, + "loss": 0.0929, + "step": 4520 + }, + { + "epoch": 0.59, + "learning_rate": 1.9898567465160065e-05, + "loss": 0.0758, + "step": 4521 + }, + { + "epoch": 0.59, + "learning_rate": 1.9898492108450342e-05, + "loss": 0.0942, + "step": 4522 + }, + { + "epoch": 0.59, + "learning_rate": 1.9898416723901637e-05, + "loss": 0.0876, + "step": 4523 + }, + { + "epoch": 0.59, + "learning_rate": 1.9898341311514162e-05, + "loss": 0.0932, + "step": 4524 + }, + { + "epoch": 0.59, + "learning_rate": 1.989826587128813e-05, + "loss": 0.0958, + "step": 4525 + }, + { + "epoch": 0.59, + "learning_rate": 1.9898190403223748e-05, + "loss": 0.0974, + "step": 4526 + }, + { + "epoch": 0.59, + "learning_rate": 1.9898114907321236e-05, + "loss": 0.0914, + "step": 4527 + }, + { + "epoch": 0.59, + "learning_rate": 1.98980393835808e-05, + "loss": 0.103, + "step": 4528 + }, + { + "epoch": 0.59, + "learning_rate": 1.989796383200266e-05, + "loss": 0.0882, + "step": 4529 + }, + { + "epoch": 0.59, + "learning_rate": 1.9897888252587022e-05, + "loss": 0.0899, + "step": 4530 + }, + { + "epoch": 0.59, + "learning_rate": 1.9897812645334098e-05, + "loss": 0.0887, + "step": 4531 + }, + { + "epoch": 0.59, + "learning_rate": 1.9897737010244107e-05, + "loss": 0.0987, + "step": 4532 + }, + { + "epoch": 0.59, + "learning_rate": 1.9897661347317255e-05, + "loss": 0.1017, + "step": 4533 + }, + { + "epoch": 0.59, + "learning_rate": 1.9897585656553757e-05, + "loss": 0.0969, + "step": 4534 + }, + { + "epoch": 0.59, + "learning_rate": 1.9897509937953828e-05, + "loss": 0.1026, + "step": 4535 + }, + { + "epoch": 0.59, + "learning_rate": 1.989743419151768e-05, + "loss": 0.083, + "step": 4536 + }, + { + "epoch": 0.59, + "learning_rate": 1.989735841724552e-05, + "loss": 0.0901, + "step": 4537 + }, + { + "epoch": 0.59, + "learning_rate": 1.9897282615137574e-05, + "loss": 0.0829, + "step": 4538 + }, + { + "epoch": 0.59, + "learning_rate": 1.9897206785194046e-05, + "loss": 0.0852, + "step": 4539 + }, + { + "epoch": 0.59, + "learning_rate": 1.9897130927415146e-05, + "loss": 0.094, + "step": 4540 + }, + { + "epoch": 0.59, + "learning_rate": 1.9897055041801096e-05, + "loss": 0.0819, + "step": 4541 + }, + { + "epoch": 0.59, + "learning_rate": 1.9896979128352107e-05, + "loss": 0.095, + "step": 4542 + }, + { + "epoch": 0.6, + "learning_rate": 1.989690318706839e-05, + "loss": 0.0823, + "step": 4543 + }, + { + "epoch": 0.6, + "learning_rate": 1.9896827217950162e-05, + "loss": 0.0855, + "step": 4544 + }, + { + "epoch": 0.6, + "learning_rate": 1.989675122099763e-05, + "loss": 0.0871, + "step": 4545 + }, + { + "epoch": 0.6, + "learning_rate": 1.9896675196211015e-05, + "loss": 0.099, + "step": 4546 + }, + { + "epoch": 0.6, + "learning_rate": 1.989659914359053e-05, + "loss": 0.0941, + "step": 4547 + }, + { + "epoch": 0.6, + "learning_rate": 1.989652306313638e-05, + "loss": 0.1079, + "step": 4548 + }, + { + "epoch": 0.6, + "learning_rate": 1.989644695484879e-05, + "loss": 0.0928, + "step": 4549 + }, + { + "epoch": 0.6, + "learning_rate": 1.9896370818727972e-05, + "loss": 0.0987, + "step": 4550 + }, + { + "epoch": 0.6, + "learning_rate": 1.9896294654774135e-05, + "loss": 0.0994, + "step": 4551 + }, + { + "epoch": 0.6, + "learning_rate": 1.98962184629875e-05, + "loss": 0.0803, + "step": 4552 + }, + { + "epoch": 0.6, + "learning_rate": 1.9896142243368276e-05, + "loss": 0.1085, + "step": 4553 + }, + { + "epoch": 0.6, + "learning_rate": 1.989606599591668e-05, + "loss": 0.0882, + "step": 4554 + }, + { + "epoch": 0.6, + "learning_rate": 1.989598972063292e-05, + "loss": 0.0983, + "step": 4555 + }, + { + "epoch": 0.6, + "learning_rate": 1.989591341751722e-05, + "loss": 0.093, + "step": 4556 + }, + { + "epoch": 0.6, + "learning_rate": 1.989583708656979e-05, + "loss": 0.0947, + "step": 4557 + }, + { + "epoch": 0.6, + "learning_rate": 1.9895760727790845e-05, + "loss": 0.1048, + "step": 4558 + }, + { + "epoch": 0.6, + "learning_rate": 1.98956843411806e-05, + "loss": 0.0934, + "step": 4559 + }, + { + "epoch": 0.6, + "learning_rate": 1.989560792673927e-05, + "loss": 0.087, + "step": 4560 + }, + { + "epoch": 0.6, + "learning_rate": 1.989553148446707e-05, + "loss": 0.0864, + "step": 4561 + }, + { + "epoch": 0.6, + "learning_rate": 1.9895455014364215e-05, + "loss": 0.1003, + "step": 4562 + }, + { + "epoch": 0.6, + "learning_rate": 1.9895378516430915e-05, + "loss": 0.0875, + "step": 4563 + }, + { + "epoch": 0.6, + "learning_rate": 1.9895301990667396e-05, + "loss": 0.1006, + "step": 4564 + }, + { + "epoch": 0.6, + "learning_rate": 1.9895225437073863e-05, + "loss": 0.0976, + "step": 4565 + }, + { + "epoch": 0.6, + "learning_rate": 1.989514885565054e-05, + "loss": 0.095, + "step": 4566 + }, + { + "epoch": 0.6, + "learning_rate": 1.9895072246397634e-05, + "loss": 0.0919, + "step": 4567 + }, + { + "epoch": 0.6, + "learning_rate": 1.9894995609315365e-05, + "loss": 0.0866, + "step": 4568 + }, + { + "epoch": 0.6, + "learning_rate": 1.989491894440395e-05, + "loss": 0.0888, + "step": 4569 + }, + { + "epoch": 0.6, + "learning_rate": 1.98948422516636e-05, + "loss": 0.1002, + "step": 4570 + }, + { + "epoch": 0.6, + "learning_rate": 1.9894765531094532e-05, + "loss": 0.0889, + "step": 4571 + }, + { + "epoch": 0.6, + "learning_rate": 1.9894688782696965e-05, + "loss": 0.1033, + "step": 4572 + }, + { + "epoch": 0.6, + "learning_rate": 1.9894612006471112e-05, + "loss": 0.0951, + "step": 4573 + }, + { + "epoch": 0.6, + "learning_rate": 1.9894535202417188e-05, + "loss": 0.0876, + "step": 4574 + }, + { + "epoch": 0.6, + "learning_rate": 1.9894458370535416e-05, + "loss": 0.0847, + "step": 4575 + }, + { + "epoch": 0.6, + "learning_rate": 1.9894381510826002e-05, + "loss": 0.0965, + "step": 4576 + }, + { + "epoch": 0.6, + "learning_rate": 1.989430462328917e-05, + "loss": 0.0931, + "step": 4577 + }, + { + "epoch": 0.6, + "learning_rate": 1.989422770792513e-05, + "loss": 0.0919, + "step": 4578 + }, + { + "epoch": 0.6, + "learning_rate": 1.98941507647341e-05, + "loss": 0.0914, + "step": 4579 + }, + { + "epoch": 0.6, + "learning_rate": 1.9894073793716302e-05, + "loss": 0.0877, + "step": 4580 + }, + { + "epoch": 0.6, + "learning_rate": 1.989399679487195e-05, + "loss": 0.0887, + "step": 4581 + }, + { + "epoch": 0.6, + "learning_rate": 1.9893919768201252e-05, + "loss": 0.0817, + "step": 4582 + }, + { + "epoch": 0.6, + "learning_rate": 1.989384271370444e-05, + "loss": 0.0967, + "step": 4583 + }, + { + "epoch": 0.6, + "learning_rate": 1.9893765631381713e-05, + "loss": 0.0902, + "step": 4584 + }, + { + "epoch": 0.6, + "learning_rate": 1.9893688521233304e-05, + "loss": 0.0951, + "step": 4585 + }, + { + "epoch": 0.6, + "learning_rate": 1.9893611383259422e-05, + "loss": 0.0875, + "step": 4586 + }, + { + "epoch": 0.6, + "learning_rate": 1.989353421746028e-05, + "loss": 0.1081, + "step": 4587 + }, + { + "epoch": 0.6, + "learning_rate": 1.9893457023836104e-05, + "loss": 0.078, + "step": 4588 + }, + { + "epoch": 0.6, + "learning_rate": 1.989337980238711e-05, + "loss": 0.0958, + "step": 4589 + }, + { + "epoch": 0.6, + "learning_rate": 1.9893302553113508e-05, + "loss": 0.1056, + "step": 4590 + }, + { + "epoch": 0.6, + "learning_rate": 1.9893225276015518e-05, + "loss": 0.0901, + "step": 4591 + }, + { + "epoch": 0.6, + "learning_rate": 1.989314797109336e-05, + "loss": 0.0954, + "step": 4592 + }, + { + "epoch": 0.6, + "learning_rate": 1.989307063834725e-05, + "loss": 0.1052, + "step": 4593 + }, + { + "epoch": 0.6, + "learning_rate": 1.989299327777741e-05, + "loss": 0.0821, + "step": 4594 + }, + { + "epoch": 0.6, + "learning_rate": 1.9892915889384047e-05, + "loss": 0.0914, + "step": 4595 + }, + { + "epoch": 0.6, + "learning_rate": 1.9892838473167388e-05, + "loss": 0.0846, + "step": 4596 + }, + { + "epoch": 0.6, + "learning_rate": 1.9892761029127647e-05, + "loss": 0.1028, + "step": 4597 + }, + { + "epoch": 0.6, + "learning_rate": 1.9892683557265043e-05, + "loss": 0.0958, + "step": 4598 + }, + { + "epoch": 0.6, + "learning_rate": 1.9892606057579792e-05, + "loss": 0.1007, + "step": 4599 + }, + { + "epoch": 0.6, + "learning_rate": 1.9892528530072117e-05, + "loss": 0.0963, + "step": 4600 + }, + { + "epoch": 0.6, + "learning_rate": 1.989245097474223e-05, + "loss": 0.0962, + "step": 4601 + }, + { + "epoch": 0.6, + "learning_rate": 1.989237339159035e-05, + "loss": 0.0869, + "step": 4602 + }, + { + "epoch": 0.6, + "learning_rate": 1.9892295780616696e-05, + "loss": 0.0908, + "step": 4603 + }, + { + "epoch": 0.6, + "learning_rate": 1.989221814182149e-05, + "loss": 0.0866, + "step": 4604 + }, + { + "epoch": 0.6, + "learning_rate": 1.9892140475204943e-05, + "loss": 0.0802, + "step": 4605 + }, + { + "epoch": 0.6, + "learning_rate": 1.989206278076728e-05, + "loss": 0.0837, + "step": 4606 + }, + { + "epoch": 0.6, + "learning_rate": 1.989198505850872e-05, + "loss": 0.0786, + "step": 4607 + }, + { + "epoch": 0.6, + "learning_rate": 1.9891907308429478e-05, + "loss": 0.093, + "step": 4608 + }, + { + "epoch": 0.6, + "learning_rate": 1.989182953052977e-05, + "loss": 0.0866, + "step": 4609 + }, + { + "epoch": 0.6, + "learning_rate": 1.989175172480982e-05, + "loss": 0.1045, + "step": 4610 + }, + { + "epoch": 0.6, + "learning_rate": 1.9891673891269847e-05, + "loss": 0.1081, + "step": 4611 + }, + { + "epoch": 0.6, + "learning_rate": 1.9891596029910065e-05, + "loss": 0.0959, + "step": 4612 + }, + { + "epoch": 0.6, + "learning_rate": 1.9891518140730696e-05, + "loss": 0.0868, + "step": 4613 + }, + { + "epoch": 0.6, + "learning_rate": 1.9891440223731962e-05, + "loss": 0.0761, + "step": 4614 + }, + { + "epoch": 0.6, + "learning_rate": 1.989136227891408e-05, + "loss": 0.095, + "step": 4615 + }, + { + "epoch": 0.6, + "learning_rate": 1.9891284306277264e-05, + "loss": 0.0913, + "step": 4616 + }, + { + "epoch": 0.6, + "learning_rate": 1.989120630582174e-05, + "loss": 0.0905, + "step": 4617 + }, + { + "epoch": 0.6, + "learning_rate": 1.9891128277547723e-05, + "loss": 0.0874, + "step": 4618 + }, + { + "epoch": 0.61, + "learning_rate": 1.9891050221455436e-05, + "loss": 0.0969, + "step": 4619 + }, + { + "epoch": 0.61, + "learning_rate": 1.9890972137545096e-05, + "loss": 0.078, + "step": 4620 + }, + { + "epoch": 0.61, + "learning_rate": 1.9890894025816924e-05, + "loss": 0.0946, + "step": 4621 + }, + { + "epoch": 0.61, + "learning_rate": 1.989081588627114e-05, + "loss": 0.101, + "step": 4622 + }, + { + "epoch": 0.61, + "learning_rate": 1.9890737718907962e-05, + "loss": 0.0984, + "step": 4623 + }, + { + "epoch": 0.61, + "learning_rate": 1.989065952372761e-05, + "loss": 0.0846, + "step": 4624 + }, + { + "epoch": 0.61, + "learning_rate": 1.9890581300730307e-05, + "loss": 0.0991, + "step": 4625 + }, + { + "epoch": 0.61, + "learning_rate": 1.989050304991627e-05, + "loss": 0.0951, + "step": 4626 + }, + { + "epoch": 0.61, + "learning_rate": 1.9890424771285722e-05, + "loss": 0.1039, + "step": 4627 + }, + { + "epoch": 0.61, + "learning_rate": 1.9890346464838875e-05, + "loss": 0.0836, + "step": 4628 + }, + { + "epoch": 0.61, + "learning_rate": 1.989026813057596e-05, + "loss": 0.0802, + "step": 4629 + }, + { + "epoch": 0.61, + "learning_rate": 1.9890189768497192e-05, + "loss": 0.0965, + "step": 4630 + }, + { + "epoch": 0.61, + "learning_rate": 1.9890111378602793e-05, + "loss": 0.0878, + "step": 4631 + }, + { + "epoch": 0.61, + "learning_rate": 1.989003296089298e-05, + "loss": 0.0814, + "step": 4632 + }, + { + "epoch": 0.61, + "learning_rate": 1.988995451536798e-05, + "loss": 0.0822, + "step": 4633 + }, + { + "epoch": 0.61, + "learning_rate": 1.9889876042028008e-05, + "loss": 0.1051, + "step": 4634 + }, + { + "epoch": 0.61, + "learning_rate": 1.9889797540873284e-05, + "loss": 0.085, + "step": 4635 + }, + { + "epoch": 0.61, + "learning_rate": 1.988971901190403e-05, + "loss": 0.0749, + "step": 4636 + }, + { + "epoch": 0.61, + "learning_rate": 1.988964045512047e-05, + "loss": 0.0913, + "step": 4637 + }, + { + "epoch": 0.61, + "learning_rate": 1.988956187052282e-05, + "loss": 0.1082, + "step": 4638 + }, + { + "epoch": 0.61, + "learning_rate": 1.9889483258111312e-05, + "loss": 0.1008, + "step": 4639 + }, + { + "epoch": 0.61, + "learning_rate": 1.988940461788615e-05, + "loss": 0.0743, + "step": 4640 + }, + { + "epoch": 0.61, + "learning_rate": 1.988932594984757e-05, + "loss": 0.0835, + "step": 4641 + }, + { + "epoch": 0.61, + "learning_rate": 1.9889247253995783e-05, + "loss": 0.0816, + "step": 4642 + }, + { + "epoch": 0.61, + "learning_rate": 1.9889168530331016e-05, + "loss": 0.0973, + "step": 4643 + }, + { + "epoch": 0.61, + "learning_rate": 1.9889089778853492e-05, + "loss": 0.0929, + "step": 4644 + }, + { + "epoch": 0.61, + "learning_rate": 1.9889010999563428e-05, + "loss": 0.0998, + "step": 4645 + }, + { + "epoch": 0.61, + "learning_rate": 1.9888932192461047e-05, + "loss": 0.0908, + "step": 4646 + }, + { + "epoch": 0.61, + "learning_rate": 1.9888853357546573e-05, + "loss": 0.0896, + "step": 4647 + }, + { + "epoch": 0.61, + "learning_rate": 1.9888774494820223e-05, + "loss": 0.1031, + "step": 4648 + }, + { + "epoch": 0.61, + "learning_rate": 1.988869560428222e-05, + "loss": 0.085, + "step": 4649 + }, + { + "epoch": 0.61, + "learning_rate": 1.988861668593279e-05, + "loss": 0.0936, + "step": 4650 + }, + { + "epoch": 0.61, + "learning_rate": 1.988853773977215e-05, + "loss": 0.0813, + "step": 4651 + }, + { + "epoch": 0.61, + "learning_rate": 1.9888458765800525e-05, + "loss": 0.0836, + "step": 4652 + }, + { + "epoch": 0.61, + "learning_rate": 1.9888379764018138e-05, + "loss": 0.0925, + "step": 4653 + }, + { + "epoch": 0.61, + "learning_rate": 1.9888300734425206e-05, + "loss": 0.0887, + "step": 4654 + }, + { + "epoch": 0.61, + "learning_rate": 1.988822167702196e-05, + "loss": 0.0905, + "step": 4655 + }, + { + "epoch": 0.61, + "learning_rate": 1.9888142591808613e-05, + "loss": 0.0924, + "step": 4656 + }, + { + "epoch": 0.61, + "learning_rate": 1.9888063478785393e-05, + "loss": 0.1042, + "step": 4657 + }, + { + "epoch": 0.61, + "learning_rate": 1.988798433795252e-05, + "loss": 0.0809, + "step": 4658 + }, + { + "epoch": 0.61, + "learning_rate": 1.988790516931022e-05, + "loss": 0.0958, + "step": 4659 + }, + { + "epoch": 0.61, + "learning_rate": 1.988782597285871e-05, + "loss": 0.1008, + "step": 4660 + }, + { + "epoch": 0.61, + "learning_rate": 1.9887746748598217e-05, + "loss": 0.0869, + "step": 4661 + }, + { + "epoch": 0.61, + "learning_rate": 1.9887667496528964e-05, + "loss": 0.0946, + "step": 4662 + }, + { + "epoch": 0.61, + "learning_rate": 1.9887588216651173e-05, + "loss": 0.1032, + "step": 4663 + }, + { + "epoch": 0.61, + "learning_rate": 1.9887508908965066e-05, + "loss": 0.0982, + "step": 4664 + }, + { + "epoch": 0.61, + "learning_rate": 1.9887429573470865e-05, + "loss": 0.093, + "step": 4665 + }, + { + "epoch": 0.61, + "learning_rate": 1.9887350210168795e-05, + "loss": 0.0923, + "step": 4666 + }, + { + "epoch": 0.61, + "learning_rate": 1.988727081905908e-05, + "loss": 0.1062, + "step": 4667 + }, + { + "epoch": 0.61, + "learning_rate": 1.9887191400141942e-05, + "loss": 0.088, + "step": 4668 + }, + { + "epoch": 0.61, + "learning_rate": 1.9887111953417607e-05, + "loss": 0.1072, + "step": 4669 + }, + { + "epoch": 0.61, + "learning_rate": 1.9887032478886294e-05, + "loss": 0.0835, + "step": 4670 + }, + { + "epoch": 0.61, + "learning_rate": 1.9886952976548232e-05, + "loss": 0.0803, + "step": 4671 + }, + { + "epoch": 0.61, + "learning_rate": 1.9886873446403636e-05, + "loss": 0.0839, + "step": 4672 + }, + { + "epoch": 0.61, + "learning_rate": 1.9886793888452738e-05, + "loss": 0.0849, + "step": 4673 + }, + { + "epoch": 0.61, + "learning_rate": 1.9886714302695757e-05, + "loss": 0.1042, + "step": 4674 + }, + { + "epoch": 0.61, + "learning_rate": 1.988663468913292e-05, + "loss": 0.087, + "step": 4675 + }, + { + "epoch": 0.61, + "learning_rate": 1.988655504776445e-05, + "loss": 0.0926, + "step": 4676 + }, + { + "epoch": 0.61, + "learning_rate": 1.988647537859057e-05, + "loss": 0.0957, + "step": 4677 + }, + { + "epoch": 0.61, + "learning_rate": 1.9886395681611505e-05, + "loss": 0.0863, + "step": 4678 + }, + { + "epoch": 0.61, + "learning_rate": 1.9886315956827476e-05, + "loss": 0.1015, + "step": 4679 + }, + { + "epoch": 0.61, + "learning_rate": 1.988623620423871e-05, + "loss": 0.095, + "step": 4680 + }, + { + "epoch": 0.61, + "learning_rate": 1.9886156423845437e-05, + "loss": 0.0912, + "step": 4681 + }, + { + "epoch": 0.61, + "learning_rate": 1.988607661564787e-05, + "loss": 0.0871, + "step": 4682 + }, + { + "epoch": 0.61, + "learning_rate": 1.988599677964624e-05, + "loss": 0.1042, + "step": 4683 + }, + { + "epoch": 0.61, + "learning_rate": 1.9885916915840767e-05, + "loss": 0.1036, + "step": 4684 + }, + { + "epoch": 0.61, + "learning_rate": 1.9885837024231685e-05, + "loss": 0.0985, + "step": 4685 + }, + { + "epoch": 0.61, + "learning_rate": 1.988575710481921e-05, + "loss": 0.0889, + "step": 4686 + }, + { + "epoch": 0.61, + "learning_rate": 1.988567715760357e-05, + "loss": 0.0858, + "step": 4687 + }, + { + "epoch": 0.61, + "learning_rate": 1.988559718258499e-05, + "loss": 0.0851, + "step": 4688 + }, + { + "epoch": 0.61, + "learning_rate": 1.9885517179763693e-05, + "loss": 0.1025, + "step": 4689 + }, + { + "epoch": 0.61, + "learning_rate": 1.988543714913991e-05, + "loss": 0.0886, + "step": 4690 + }, + { + "epoch": 0.61, + "learning_rate": 1.9885357090713854e-05, + "loss": 0.0913, + "step": 4691 + }, + { + "epoch": 0.61, + "learning_rate": 1.9885277004485762e-05, + "loss": 0.0969, + "step": 4692 + }, + { + "epoch": 0.61, + "learning_rate": 1.9885196890455857e-05, + "loss": 0.0974, + "step": 4693 + }, + { + "epoch": 0.61, + "learning_rate": 1.988511674862436e-05, + "loss": 0.0921, + "step": 4694 + }, + { + "epoch": 0.62, + "learning_rate": 1.9885036578991497e-05, + "loss": 0.0895, + "step": 4695 + }, + { + "epoch": 0.62, + "learning_rate": 1.9884956381557497e-05, + "loss": 0.0908, + "step": 4696 + }, + { + "epoch": 0.62, + "learning_rate": 1.988487615632258e-05, + "loss": 0.1043, + "step": 4697 + }, + { + "epoch": 0.62, + "learning_rate": 1.988479590328698e-05, + "loss": 0.0912, + "step": 4698 + }, + { + "epoch": 0.62, + "learning_rate": 1.9884715622450913e-05, + "loss": 0.0872, + "step": 4699 + }, + { + "epoch": 0.62, + "learning_rate": 1.9884635313814613e-05, + "loss": 0.102, + "step": 4700 + }, + { + "epoch": 0.62, + "learning_rate": 1.9884554977378304e-05, + "loss": 0.0985, + "step": 4701 + }, + { + "epoch": 0.62, + "learning_rate": 1.9884474613142205e-05, + "loss": 0.0853, + "step": 4702 + }, + { + "epoch": 0.62, + "learning_rate": 1.9884394221106552e-05, + "loss": 0.1041, + "step": 4703 + }, + { + "epoch": 0.62, + "learning_rate": 1.9884313801271566e-05, + "loss": 0.1059, + "step": 4704 + }, + { + "epoch": 0.62, + "learning_rate": 1.9884233353637474e-05, + "loss": 0.0719, + "step": 4705 + }, + { + "epoch": 0.62, + "learning_rate": 1.9884152878204502e-05, + "loss": 0.1017, + "step": 4706 + }, + { + "epoch": 0.62, + "learning_rate": 1.9884072374972875e-05, + "loss": 0.097, + "step": 4707 + }, + { + "epoch": 0.62, + "learning_rate": 1.9883991843942822e-05, + "loss": 0.1035, + "step": 4708 + }, + { + "epoch": 0.62, + "learning_rate": 1.9883911285114568e-05, + "loss": 0.094, + "step": 4709 + }, + { + "epoch": 0.62, + "learning_rate": 1.988383069848834e-05, + "loss": 0.0927, + "step": 4710 + }, + { + "epoch": 0.62, + "learning_rate": 1.9883750084064365e-05, + "loss": 0.0914, + "step": 4711 + }, + { + "epoch": 0.62, + "learning_rate": 1.9883669441842868e-05, + "loss": 0.105, + "step": 4712 + }, + { + "epoch": 0.62, + "learning_rate": 1.988358877182408e-05, + "loss": 0.1, + "step": 4713 + }, + { + "epoch": 0.62, + "learning_rate": 1.988350807400822e-05, + "loss": 0.0943, + "step": 4714 + }, + { + "epoch": 0.62, + "learning_rate": 1.9883427348395524e-05, + "loss": 0.0986, + "step": 4715 + }, + { + "epoch": 0.62, + "learning_rate": 1.9883346594986214e-05, + "loss": 0.0896, + "step": 4716 + }, + { + "epoch": 0.62, + "learning_rate": 1.9883265813780515e-05, + "loss": 0.0831, + "step": 4717 + }, + { + "epoch": 0.62, + "learning_rate": 1.988318500477866e-05, + "loss": 0.0907, + "step": 4718 + }, + { + "epoch": 0.62, + "learning_rate": 1.9883104167980874e-05, + "loss": 0.0987, + "step": 4719 + }, + { + "epoch": 0.62, + "learning_rate": 1.9883023303387383e-05, + "loss": 0.0847, + "step": 4720 + }, + { + "epoch": 0.62, + "learning_rate": 1.9882942410998415e-05, + "loss": 0.0835, + "step": 4721 + }, + { + "epoch": 0.62, + "learning_rate": 1.98828614908142e-05, + "loss": 0.1072, + "step": 4722 + }, + { + "epoch": 0.62, + "learning_rate": 1.988278054283496e-05, + "loss": 0.0785, + "step": 4723 + }, + { + "epoch": 0.62, + "learning_rate": 1.9882699567060927e-05, + "loss": 0.0961, + "step": 4724 + }, + { + "epoch": 0.62, + "learning_rate": 1.9882618563492325e-05, + "loss": 0.0933, + "step": 4725 + }, + { + "epoch": 0.62, + "learning_rate": 1.988253753212939e-05, + "loss": 0.0925, + "step": 4726 + }, + { + "epoch": 0.62, + "learning_rate": 1.9882456472972343e-05, + "loss": 0.0899, + "step": 4727 + }, + { + "epoch": 0.62, + "learning_rate": 1.988237538602141e-05, + "loss": 0.1005, + "step": 4728 + }, + { + "epoch": 0.62, + "learning_rate": 1.9882294271276824e-05, + "loss": 0.0788, + "step": 4729 + }, + { + "epoch": 0.62, + "learning_rate": 1.9882213128738813e-05, + "loss": 0.1024, + "step": 4730 + }, + { + "epoch": 0.62, + "learning_rate": 1.9882131958407604e-05, + "loss": 0.0842, + "step": 4731 + }, + { + "epoch": 0.62, + "learning_rate": 1.9882050760283428e-05, + "loss": 0.1008, + "step": 4732 + }, + { + "epoch": 0.62, + "learning_rate": 1.9881969534366505e-05, + "loss": 0.0869, + "step": 4733 + }, + { + "epoch": 0.62, + "learning_rate": 1.988188828065707e-05, + "loss": 0.0848, + "step": 4734 + }, + { + "epoch": 0.62, + "learning_rate": 1.988180699915535e-05, + "loss": 0.087, + "step": 4735 + }, + { + "epoch": 0.62, + "learning_rate": 1.9881725689861577e-05, + "loss": 0.105, + "step": 4736 + }, + { + "epoch": 0.62, + "learning_rate": 1.9881644352775975e-05, + "loss": 0.0989, + "step": 4737 + }, + { + "epoch": 0.62, + "learning_rate": 1.9881562987898774e-05, + "loss": 0.0914, + "step": 4738 + }, + { + "epoch": 0.62, + "learning_rate": 1.9881481595230205e-05, + "loss": 0.0855, + "step": 4739 + }, + { + "epoch": 0.62, + "learning_rate": 1.9881400174770492e-05, + "loss": 0.0871, + "step": 4740 + }, + { + "epoch": 0.62, + "learning_rate": 1.9881318726519872e-05, + "loss": 0.101, + "step": 4741 + }, + { + "epoch": 0.62, + "learning_rate": 1.9881237250478566e-05, + "loss": 0.1052, + "step": 4742 + }, + { + "epoch": 0.62, + "learning_rate": 1.9881155746646806e-05, + "loss": 0.0824, + "step": 4743 + }, + { + "epoch": 0.62, + "learning_rate": 1.9881074215024823e-05, + "loss": 0.087, + "step": 4744 + }, + { + "epoch": 0.62, + "learning_rate": 1.988099265561285e-05, + "loss": 0.1, + "step": 4745 + }, + { + "epoch": 0.62, + "learning_rate": 1.9880911068411102e-05, + "loss": 0.0783, + "step": 4746 + }, + { + "epoch": 0.62, + "learning_rate": 1.9880829453419824e-05, + "loss": 0.1036, + "step": 4747 + }, + { + "epoch": 0.62, + "learning_rate": 1.9880747810639238e-05, + "loss": 0.1001, + "step": 4748 + }, + { + "epoch": 0.62, + "learning_rate": 1.9880666140069574e-05, + "loss": 0.0902, + "step": 4749 + }, + { + "epoch": 0.62, + "learning_rate": 1.988058444171106e-05, + "loss": 0.0948, + "step": 4750 + }, + { + "epoch": 0.62, + "learning_rate": 1.9880502715563933e-05, + "loss": 0.0946, + "step": 4751 + }, + { + "epoch": 0.62, + "learning_rate": 1.9880420961628416e-05, + "loss": 0.09, + "step": 4752 + }, + { + "epoch": 0.62, + "learning_rate": 1.9880339179904743e-05, + "loss": 0.0835, + "step": 4753 + }, + { + "epoch": 0.62, + "learning_rate": 1.988025737039314e-05, + "loss": 0.0905, + "step": 4754 + }, + { + "epoch": 0.62, + "learning_rate": 1.988017553309384e-05, + "loss": 0.1028, + "step": 4755 + }, + { + "epoch": 0.62, + "learning_rate": 1.9880093668007076e-05, + "loss": 0.1022, + "step": 4756 + }, + { + "epoch": 0.62, + "learning_rate": 1.988001177513307e-05, + "loss": 0.1006, + "step": 4757 + }, + { + "epoch": 0.62, + "learning_rate": 1.987992985447206e-05, + "loss": 0.0995, + "step": 4758 + }, + { + "epoch": 0.62, + "learning_rate": 1.987984790602427e-05, + "loss": 0.0939, + "step": 4759 + }, + { + "epoch": 0.62, + "learning_rate": 1.9879765929789936e-05, + "loss": 0.0984, + "step": 4760 + }, + { + "epoch": 0.62, + "learning_rate": 1.9879683925769287e-05, + "loss": 0.0958, + "step": 4761 + }, + { + "epoch": 0.62, + "learning_rate": 1.987960189396255e-05, + "loss": 0.0902, + "step": 4762 + }, + { + "epoch": 0.62, + "learning_rate": 1.987951983436996e-05, + "loss": 0.0989, + "step": 4763 + }, + { + "epoch": 0.62, + "learning_rate": 1.9879437746991752e-05, + "loss": 0.0936, + "step": 4764 + }, + { + "epoch": 0.62, + "learning_rate": 1.9879355631828145e-05, + "loss": 0.097, + "step": 4765 + }, + { + "epoch": 0.62, + "learning_rate": 1.987927348887938e-05, + "loss": 0.0828, + "step": 4766 + }, + { + "epoch": 0.62, + "learning_rate": 1.9879191318145683e-05, + "loss": 0.0917, + "step": 4767 + }, + { + "epoch": 0.62, + "learning_rate": 1.9879109119627285e-05, + "loss": 0.0919, + "step": 4768 + }, + { + "epoch": 0.62, + "learning_rate": 1.987902689332442e-05, + "loss": 0.0898, + "step": 4769 + }, + { + "epoch": 0.62, + "learning_rate": 1.9878944639237318e-05, + "loss": 0.0937, + "step": 4770 + }, + { + "epoch": 0.62, + "learning_rate": 1.9878862357366212e-05, + "loss": 0.0908, + "step": 4771 + }, + { + "epoch": 0.63, + "learning_rate": 1.9878780047711332e-05, + "loss": 0.0961, + "step": 4772 + }, + { + "epoch": 0.63, + "learning_rate": 1.9878697710272907e-05, + "loss": 0.097, + "step": 4773 + }, + { + "epoch": 0.63, + "learning_rate": 1.987861534505117e-05, + "loss": 0.0855, + "step": 4774 + }, + { + "epoch": 0.63, + "learning_rate": 1.9878532952046354e-05, + "loss": 0.098, + "step": 4775 + }, + { + "epoch": 0.63, + "learning_rate": 1.987845053125869e-05, + "loss": 0.0962, + "step": 4776 + }, + { + "epoch": 0.63, + "learning_rate": 1.9878368082688412e-05, + "loss": 0.0859, + "step": 4777 + }, + { + "epoch": 0.63, + "learning_rate": 1.9878285606335746e-05, + "loss": 0.0906, + "step": 4778 + }, + { + "epoch": 0.63, + "learning_rate": 1.9878203102200936e-05, + "loss": 0.0873, + "step": 4779 + }, + { + "epoch": 0.63, + "learning_rate": 1.98781205702842e-05, + "loss": 0.0942, + "step": 4780 + }, + { + "epoch": 0.63, + "learning_rate": 1.9878038010585773e-05, + "loss": 0.0863, + "step": 4781 + }, + { + "epoch": 0.63, + "learning_rate": 1.987795542310589e-05, + "loss": 0.0863, + "step": 4782 + }, + { + "epoch": 0.63, + "learning_rate": 1.987787280784479e-05, + "loss": 0.0853, + "step": 4783 + }, + { + "epoch": 0.63, + "learning_rate": 1.9877790164802695e-05, + "loss": 0.1029, + "step": 4784 + }, + { + "epoch": 0.63, + "learning_rate": 1.987770749397984e-05, + "loss": 0.0934, + "step": 4785 + }, + { + "epoch": 0.63, + "learning_rate": 1.9877624795376463e-05, + "loss": 0.0871, + "step": 4786 + }, + { + "epoch": 0.63, + "learning_rate": 1.987754206899279e-05, + "loss": 0.0837, + "step": 4787 + }, + { + "epoch": 0.63, + "learning_rate": 1.9877459314829057e-05, + "loss": 0.0861, + "step": 4788 + }, + { + "epoch": 0.63, + "learning_rate": 1.987737653288549e-05, + "loss": 0.0847, + "step": 4789 + }, + { + "epoch": 0.63, + "learning_rate": 1.9877293723162335e-05, + "loss": 0.1034, + "step": 4790 + }, + { + "epoch": 0.63, + "learning_rate": 1.9877210885659812e-05, + "loss": 0.0896, + "step": 4791 + }, + { + "epoch": 0.63, + "learning_rate": 1.987712802037816e-05, + "loss": 0.1028, + "step": 4792 + }, + { + "epoch": 0.63, + "learning_rate": 1.9877045127317617e-05, + "loss": 0.0984, + "step": 4793 + }, + { + "epoch": 0.63, + "learning_rate": 1.98769622064784e-05, + "loss": 0.0992, + "step": 4794 + }, + { + "epoch": 0.63, + "learning_rate": 1.987687925786076e-05, + "loss": 0.0869, + "step": 4795 + }, + { + "epoch": 0.63, + "learning_rate": 1.9876796281464927e-05, + "loss": 0.0913, + "step": 4796 + }, + { + "epoch": 0.63, + "learning_rate": 1.987671327729112e-05, + "loss": 0.0855, + "step": 4797 + }, + { + "epoch": 0.63, + "learning_rate": 1.987663024533959e-05, + "loss": 0.0984, + "step": 4798 + }, + { + "epoch": 0.63, + "learning_rate": 1.9876547185610564e-05, + "loss": 0.0808, + "step": 4799 + }, + { + "epoch": 0.63, + "learning_rate": 1.9876464098104274e-05, + "loss": 0.0996, + "step": 4800 + }, + { + "epoch": 0.63, + "learning_rate": 1.9876380982820952e-05, + "loss": 0.0923, + "step": 4801 + }, + { + "epoch": 0.63, + "learning_rate": 1.9876297839760837e-05, + "loss": 0.0851, + "step": 4802 + }, + { + "epoch": 0.63, + "learning_rate": 1.9876214668924158e-05, + "loss": 0.0919, + "step": 4803 + }, + { + "epoch": 0.63, + "learning_rate": 1.9876131470311154e-05, + "loss": 0.1017, + "step": 4804 + }, + { + "epoch": 0.63, + "learning_rate": 1.9876048243922055e-05, + "loss": 0.1, + "step": 4805 + }, + { + "epoch": 0.63, + "learning_rate": 1.9875964989757097e-05, + "loss": 0.089, + "step": 4806 + }, + { + "epoch": 0.63, + "learning_rate": 1.9875881707816508e-05, + "loss": 0.0809, + "step": 4807 + }, + { + "epoch": 0.63, + "learning_rate": 1.9875798398100533e-05, + "loss": 0.0844, + "step": 4808 + }, + { + "epoch": 0.63, + "learning_rate": 1.98757150606094e-05, + "loss": 0.0937, + "step": 4809 + }, + { + "epoch": 0.63, + "learning_rate": 1.9875631695343344e-05, + "loss": 0.095, + "step": 4810 + }, + { + "epoch": 0.63, + "learning_rate": 1.98755483023026e-05, + "loss": 0.0835, + "step": 4811 + }, + { + "epoch": 0.63, + "learning_rate": 1.9875464881487403e-05, + "loss": 0.0859, + "step": 4812 + }, + { + "epoch": 0.63, + "learning_rate": 1.9875381432897988e-05, + "loss": 0.0955, + "step": 4813 + }, + { + "epoch": 0.63, + "learning_rate": 1.9875297956534586e-05, + "loss": 0.0882, + "step": 4814 + }, + { + "epoch": 0.63, + "learning_rate": 1.9875214452397435e-05, + "loss": 0.0873, + "step": 4815 + }, + { + "epoch": 0.63, + "learning_rate": 1.9875130920486768e-05, + "loss": 0.086, + "step": 4816 + }, + { + "epoch": 0.63, + "learning_rate": 1.9875047360802822e-05, + "loss": 0.0958, + "step": 4817 + }, + { + "epoch": 0.63, + "learning_rate": 1.9874963773345832e-05, + "loss": 0.0835, + "step": 4818 + }, + { + "epoch": 0.63, + "learning_rate": 1.9874880158116034e-05, + "loss": 0.1041, + "step": 4819 + }, + { + "epoch": 0.63, + "learning_rate": 1.9874796515113658e-05, + "loss": 0.0922, + "step": 4820 + }, + { + "epoch": 0.63, + "learning_rate": 1.9874712844338943e-05, + "loss": 0.0929, + "step": 4821 + }, + { + "epoch": 0.63, + "learning_rate": 1.9874629145792124e-05, + "loss": 0.0945, + "step": 4822 + }, + { + "epoch": 0.63, + "learning_rate": 1.987454541947344e-05, + "loss": 0.09, + "step": 4823 + }, + { + "epoch": 0.63, + "learning_rate": 1.987446166538312e-05, + "loss": 0.0979, + "step": 4824 + }, + { + "epoch": 0.63, + "learning_rate": 1.9874377883521403e-05, + "loss": 0.0856, + "step": 4825 + }, + { + "epoch": 0.63, + "learning_rate": 1.987429407388852e-05, + "loss": 0.0901, + "step": 4826 + }, + { + "epoch": 0.63, + "learning_rate": 1.9874210236484715e-05, + "loss": 0.0924, + "step": 4827 + }, + { + "epoch": 0.63, + "learning_rate": 1.9874126371310222e-05, + "loss": 0.0849, + "step": 4828 + }, + { + "epoch": 0.63, + "learning_rate": 1.9874042478365274e-05, + "loss": 0.0913, + "step": 4829 + }, + { + "epoch": 0.63, + "learning_rate": 1.9873958557650102e-05, + "loss": 0.0987, + "step": 4830 + }, + { + "epoch": 0.63, + "learning_rate": 1.987387460916495e-05, + "loss": 0.098, + "step": 4831 + }, + { + "epoch": 0.63, + "learning_rate": 1.9873790632910054e-05, + "loss": 0.0963, + "step": 4832 + }, + { + "epoch": 0.63, + "learning_rate": 1.987370662888564e-05, + "loss": 0.0835, + "step": 4833 + }, + { + "epoch": 0.63, + "learning_rate": 1.9873622597091962e-05, + "loss": 0.0993, + "step": 4834 + }, + { + "epoch": 0.63, + "learning_rate": 1.987353853752924e-05, + "loss": 0.088, + "step": 4835 + }, + { + "epoch": 0.63, + "learning_rate": 1.9873454450197716e-05, + "loss": 0.0921, + "step": 4836 + }, + { + "epoch": 0.63, + "learning_rate": 1.987337033509763e-05, + "loss": 0.0843, + "step": 4837 + }, + { + "epoch": 0.63, + "learning_rate": 1.9873286192229218e-05, + "loss": 0.0785, + "step": 4838 + }, + { + "epoch": 0.63, + "learning_rate": 1.9873202021592706e-05, + "loss": 0.0886, + "step": 4839 + }, + { + "epoch": 0.63, + "learning_rate": 1.9873117823188347e-05, + "loss": 0.0884, + "step": 4840 + }, + { + "epoch": 0.63, + "learning_rate": 1.9873033597016368e-05, + "loss": 0.0948, + "step": 4841 + }, + { + "epoch": 0.63, + "learning_rate": 1.9872949343077005e-05, + "loss": 0.0974, + "step": 4842 + }, + { + "epoch": 0.63, + "learning_rate": 1.9872865061370503e-05, + "loss": 0.0944, + "step": 4843 + }, + { + "epoch": 0.63, + "learning_rate": 1.987278075189709e-05, + "loss": 0.09, + "step": 4844 + }, + { + "epoch": 0.63, + "learning_rate": 1.9872696414657006e-05, + "loss": 0.0821, + "step": 4845 + }, + { + "epoch": 0.63, + "learning_rate": 1.9872612049650493e-05, + "loss": 0.1023, + "step": 4846 + }, + { + "epoch": 0.63, + "learning_rate": 1.987252765687778e-05, + "loss": 0.1046, + "step": 4847 + }, + { + "epoch": 0.64, + "learning_rate": 1.9872443236339113e-05, + "loss": 0.0802, + "step": 4848 + }, + { + "epoch": 0.64, + "learning_rate": 1.987235878803472e-05, + "loss": 0.0808, + "step": 4849 + }, + { + "epoch": 0.64, + "learning_rate": 1.987227431196485e-05, + "loss": 0.0886, + "step": 4850 + }, + { + "epoch": 0.64, + "learning_rate": 1.9872189808129726e-05, + "loss": 0.0982, + "step": 4851 + }, + { + "epoch": 0.64, + "learning_rate": 1.9872105276529602e-05, + "loss": 0.0873, + "step": 4852 + }, + { + "epoch": 0.64, + "learning_rate": 1.9872020717164704e-05, + "loss": 0.1045, + "step": 4853 + }, + { + "epoch": 0.64, + "learning_rate": 1.9871936130035272e-05, + "loss": 0.0995, + "step": 4854 + }, + { + "epoch": 0.64, + "learning_rate": 1.987185151514155e-05, + "loss": 0.0768, + "step": 4855 + }, + { + "epoch": 0.64, + "learning_rate": 1.9871766872483765e-05, + "loss": 0.0827, + "step": 4856 + }, + { + "epoch": 0.64, + "learning_rate": 1.9871682202062165e-05, + "loss": 0.0899, + "step": 4857 + }, + { + "epoch": 0.64, + "learning_rate": 1.9871597503876986e-05, + "loss": 0.104, + "step": 4858 + }, + { + "epoch": 0.64, + "learning_rate": 1.9871512777928463e-05, + "loss": 0.0777, + "step": 4859 + }, + { + "epoch": 0.64, + "learning_rate": 1.9871428024216835e-05, + "loss": 0.0809, + "step": 4860 + }, + { + "epoch": 0.64, + "learning_rate": 1.987134324274234e-05, + "loss": 0.0753, + "step": 4861 + }, + { + "epoch": 0.64, + "learning_rate": 1.9871258433505224e-05, + "loss": 0.0883, + "step": 4862 + }, + { + "epoch": 0.64, + "learning_rate": 1.9871173596505714e-05, + "loss": 0.0751, + "step": 4863 + }, + { + "epoch": 0.64, + "learning_rate": 1.9871088731744056e-05, + "loss": 0.0842, + "step": 4864 + }, + { + "epoch": 0.64, + "learning_rate": 1.9871003839220483e-05, + "loss": 0.0967, + "step": 4865 + }, + { + "epoch": 0.64, + "learning_rate": 1.9870918918935242e-05, + "loss": 0.0833, + "step": 4866 + }, + { + "epoch": 0.64, + "learning_rate": 1.9870833970888568e-05, + "loss": 0.0852, + "step": 4867 + }, + { + "epoch": 0.64, + "learning_rate": 1.9870748995080697e-05, + "loss": 0.0962, + "step": 4868 + }, + { + "epoch": 0.64, + "learning_rate": 1.9870663991511867e-05, + "loss": 0.0894, + "step": 4869 + }, + { + "epoch": 0.64, + "learning_rate": 1.9870578960182325e-05, + "loss": 0.0886, + "step": 4870 + }, + { + "epoch": 0.64, + "learning_rate": 1.9870493901092304e-05, + "loss": 0.1034, + "step": 4871 + }, + { + "epoch": 0.64, + "learning_rate": 1.9870408814242043e-05, + "loss": 0.0787, + "step": 4872 + }, + { + "epoch": 0.64, + "learning_rate": 1.987032369963178e-05, + "loss": 0.087, + "step": 4873 + }, + { + "epoch": 0.64, + "learning_rate": 1.9870238557261762e-05, + "loss": 0.089, + "step": 4874 + }, + { + "epoch": 0.64, + "learning_rate": 1.9870153387132224e-05, + "loss": 0.0996, + "step": 4875 + }, + { + "epoch": 0.64, + "learning_rate": 1.9870068189243404e-05, + "loss": 0.1025, + "step": 4876 + }, + { + "epoch": 0.64, + "learning_rate": 1.9869982963595542e-05, + "loss": 0.0818, + "step": 4877 + }, + { + "epoch": 0.64, + "learning_rate": 1.986989771018888e-05, + "loss": 0.1027, + "step": 4878 + }, + { + "epoch": 0.64, + "learning_rate": 1.9869812429023655e-05, + "loss": 0.0764, + "step": 4879 + }, + { + "epoch": 0.64, + "learning_rate": 1.9869727120100108e-05, + "loss": 0.0875, + "step": 4880 + }, + { + "epoch": 0.64, + "learning_rate": 1.986964178341848e-05, + "loss": 0.0915, + "step": 4881 + }, + { + "epoch": 0.64, + "learning_rate": 1.986955641897901e-05, + "loss": 0.0863, + "step": 4882 + }, + { + "epoch": 0.64, + "learning_rate": 1.9869471026781936e-05, + "loss": 0.0871, + "step": 4883 + }, + { + "epoch": 0.64, + "learning_rate": 1.9869385606827504e-05, + "loss": 0.0935, + "step": 4884 + }, + { + "epoch": 0.64, + "learning_rate": 1.986930015911595e-05, + "loss": 0.0946, + "step": 4885 + }, + { + "epoch": 0.64, + "learning_rate": 1.986921468364751e-05, + "loss": 0.0786, + "step": 4886 + }, + { + "epoch": 0.64, + "learning_rate": 1.9869129180422434e-05, + "loss": 0.1012, + "step": 4887 + }, + { + "epoch": 0.64, + "learning_rate": 1.9869043649440957e-05, + "loss": 0.0981, + "step": 4888 + }, + { + "epoch": 0.64, + "learning_rate": 1.9868958090703318e-05, + "loss": 0.0932, + "step": 4889 + }, + { + "epoch": 0.64, + "learning_rate": 1.9868872504209764e-05, + "loss": 0.0935, + "step": 4890 + }, + { + "epoch": 0.64, + "learning_rate": 1.9868786889960526e-05, + "loss": 0.1059, + "step": 4891 + }, + { + "epoch": 0.64, + "learning_rate": 1.9868701247955856e-05, + "loss": 0.0989, + "step": 4892 + }, + { + "epoch": 0.64, + "learning_rate": 1.9868615578195985e-05, + "loss": 0.0875, + "step": 4893 + }, + { + "epoch": 0.64, + "learning_rate": 1.9868529880681162e-05, + "loss": 0.0816, + "step": 4894 + }, + { + "epoch": 0.64, + "learning_rate": 1.986844415541162e-05, + "loss": 0.1069, + "step": 4895 + }, + { + "epoch": 0.64, + "learning_rate": 1.9868358402387612e-05, + "loss": 0.0927, + "step": 4896 + }, + { + "epoch": 0.64, + "learning_rate": 1.9868272621609366e-05, + "loss": 0.0927, + "step": 4897 + }, + { + "epoch": 0.64, + "learning_rate": 1.9868186813077127e-05, + "loss": 0.0848, + "step": 4898 + }, + { + "epoch": 0.64, + "learning_rate": 1.986810097679114e-05, + "loss": 0.0899, + "step": 4899 + }, + { + "epoch": 0.64, + "learning_rate": 1.9868015112751643e-05, + "loss": 0.0982, + "step": 4900 + }, + { + "epoch": 0.64, + "learning_rate": 1.9867929220958884e-05, + "loss": 0.0903, + "step": 4901 + }, + { + "epoch": 0.64, + "learning_rate": 1.9867843301413097e-05, + "loss": 0.088, + "step": 4902 + }, + { + "epoch": 0.64, + "learning_rate": 1.9867757354114526e-05, + "loss": 0.1076, + "step": 4903 + }, + { + "epoch": 0.64, + "learning_rate": 1.9867671379063412e-05, + "loss": 0.0944, + "step": 4904 + }, + { + "epoch": 0.64, + "learning_rate": 1.9867585376259996e-05, + "loss": 0.0891, + "step": 4905 + }, + { + "epoch": 0.64, + "learning_rate": 1.9867499345704523e-05, + "loss": 0.0895, + "step": 4906 + }, + { + "epoch": 0.64, + "learning_rate": 1.9867413287397233e-05, + "loss": 0.0948, + "step": 4907 + }, + { + "epoch": 0.64, + "learning_rate": 1.9867327201338372e-05, + "loss": 0.0867, + "step": 4908 + }, + { + "epoch": 0.64, + "learning_rate": 1.9867241087528177e-05, + "loss": 0.0999, + "step": 4909 + }, + { + "epoch": 0.64, + "learning_rate": 1.9867154945966893e-05, + "loss": 0.0992, + "step": 4910 + }, + { + "epoch": 0.64, + "learning_rate": 1.986706877665476e-05, + "loss": 0.0919, + "step": 4911 + }, + { + "epoch": 0.64, + "learning_rate": 1.986698257959202e-05, + "loss": 0.0931, + "step": 4912 + }, + { + "epoch": 0.64, + "learning_rate": 1.986689635477892e-05, + "loss": 0.0965, + "step": 4913 + }, + { + "epoch": 0.64, + "learning_rate": 1.9866810102215696e-05, + "loss": 0.0997, + "step": 4914 + }, + { + "epoch": 0.64, + "learning_rate": 1.9866723821902595e-05, + "loss": 0.0859, + "step": 4915 + }, + { + "epoch": 0.64, + "learning_rate": 1.986663751383986e-05, + "loss": 0.0795, + "step": 4916 + }, + { + "epoch": 0.64, + "learning_rate": 1.9866551178027735e-05, + "loss": 0.0894, + "step": 4917 + }, + { + "epoch": 0.64, + "learning_rate": 1.9866464814466453e-05, + "loss": 0.0978, + "step": 4918 + }, + { + "epoch": 0.64, + "learning_rate": 1.986637842315627e-05, + "loss": 0.0893, + "step": 4919 + }, + { + "epoch": 0.64, + "learning_rate": 1.986629200409742e-05, + "loss": 0.0924, + "step": 4920 + }, + { + "epoch": 0.64, + "learning_rate": 1.9866205557290152e-05, + "loss": 0.1057, + "step": 4921 + }, + { + "epoch": 0.64, + "learning_rate": 1.9866119082734703e-05, + "loss": 0.075, + "step": 4922 + }, + { + "epoch": 0.64, + "learning_rate": 1.9866032580431323e-05, + "loss": 0.0897, + "step": 4923 + }, + { + "epoch": 0.65, + "learning_rate": 1.986594605038025e-05, + "loss": 0.0819, + "step": 4924 + }, + { + "epoch": 0.65, + "learning_rate": 1.986585949258173e-05, + "loss": 0.0866, + "step": 4925 + }, + { + "epoch": 0.65, + "learning_rate": 1.9865772907036e-05, + "loss": 0.0977, + "step": 4926 + }, + { + "epoch": 0.65, + "learning_rate": 1.9865686293743315e-05, + "loss": 0.0849, + "step": 4927 + }, + { + "epoch": 0.65, + "learning_rate": 1.9865599652703913e-05, + "loss": 0.1031, + "step": 4928 + }, + { + "epoch": 0.65, + "learning_rate": 1.9865512983918036e-05, + "loss": 0.0996, + "step": 4929 + }, + { + "epoch": 0.65, + "learning_rate": 1.9865426287385928e-05, + "loss": 0.0914, + "step": 4930 + }, + { + "epoch": 0.65, + "learning_rate": 1.9865339563107836e-05, + "loss": 0.0983, + "step": 4931 + }, + { + "epoch": 0.65, + "learning_rate": 1.9865252811084e-05, + "loss": 0.0987, + "step": 4932 + }, + { + "epoch": 0.65, + "learning_rate": 1.9865166031314667e-05, + "loss": 0.0902, + "step": 4933 + }, + { + "epoch": 0.65, + "learning_rate": 1.986507922380008e-05, + "loss": 0.0907, + "step": 4934 + }, + { + "epoch": 0.65, + "learning_rate": 1.986499238854048e-05, + "loss": 0.0953, + "step": 4935 + }, + { + "epoch": 0.65, + "learning_rate": 1.9864905525536117e-05, + "loss": 0.0858, + "step": 4936 + }, + { + "epoch": 0.65, + "learning_rate": 1.9864818634787232e-05, + "loss": 0.0896, + "step": 4937 + }, + { + "epoch": 0.65, + "learning_rate": 1.986473171629407e-05, + "loss": 0.0925, + "step": 4938 + }, + { + "epoch": 0.65, + "learning_rate": 1.9864644770056876e-05, + "loss": 0.0905, + "step": 4939 + }, + { + "epoch": 0.65, + "learning_rate": 1.9864557796075894e-05, + "loss": 0.0846, + "step": 4940 + }, + { + "epoch": 0.65, + "learning_rate": 1.9864470794351368e-05, + "loss": 0.0901, + "step": 4941 + }, + { + "epoch": 0.65, + "learning_rate": 1.9864383764883544e-05, + "loss": 0.0913, + "step": 4942 + }, + { + "epoch": 0.65, + "learning_rate": 1.9864296707672664e-05, + "loss": 0.0902, + "step": 4943 + }, + { + "epoch": 0.65, + "learning_rate": 1.986420962271898e-05, + "loss": 0.0913, + "step": 4944 + }, + { + "epoch": 0.65, + "learning_rate": 1.9864122510022724e-05, + "loss": 0.0876, + "step": 4945 + }, + { + "epoch": 0.65, + "learning_rate": 1.9864035369584155e-05, + "loss": 0.0893, + "step": 4946 + }, + { + "epoch": 0.65, + "learning_rate": 1.9863948201403506e-05, + "loss": 0.0814, + "step": 4947 + }, + { + "epoch": 0.65, + "learning_rate": 1.986386100548103e-05, + "loss": 0.0925, + "step": 4948 + }, + { + "epoch": 0.65, + "learning_rate": 1.9863773781816974e-05, + "loss": 0.0965, + "step": 4949 + }, + { + "epoch": 0.65, + "learning_rate": 1.9863686530411573e-05, + "loss": 0.0981, + "step": 4950 + }, + { + "epoch": 0.65, + "learning_rate": 1.9863599251265088e-05, + "loss": 0.0853, + "step": 4951 + }, + { + "epoch": 0.65, + "learning_rate": 1.986351194437775e-05, + "loss": 0.0947, + "step": 4952 + }, + { + "epoch": 0.65, + "learning_rate": 1.9863424609749805e-05, + "loss": 0.0866, + "step": 4953 + }, + { + "epoch": 0.65, + "learning_rate": 1.986333724738151e-05, + "loss": 0.096, + "step": 4954 + }, + { + "epoch": 0.65, + "learning_rate": 1.9863249857273105e-05, + "loss": 0.095, + "step": 4955 + }, + { + "epoch": 0.65, + "learning_rate": 1.9863162439424832e-05, + "loss": 0.0993, + "step": 4956 + }, + { + "epoch": 0.65, + "learning_rate": 1.9863074993836942e-05, + "loss": 0.0962, + "step": 4957 + }, + { + "epoch": 0.65, + "learning_rate": 1.9862987520509675e-05, + "loss": 0.0966, + "step": 4958 + }, + { + "epoch": 0.65, + "learning_rate": 1.9862900019443285e-05, + "loss": 0.0867, + "step": 4959 + }, + { + "epoch": 0.65, + "learning_rate": 1.9862812490638012e-05, + "loss": 0.0924, + "step": 4960 + }, + { + "epoch": 0.65, + "learning_rate": 1.9862724934094105e-05, + "loss": 0.0961, + "step": 4961 + }, + { + "epoch": 0.65, + "learning_rate": 1.986263734981181e-05, + "loss": 0.0879, + "step": 4962 + }, + { + "epoch": 0.65, + "learning_rate": 1.986254973779137e-05, + "loss": 0.0888, + "step": 4963 + }, + { + "epoch": 0.65, + "learning_rate": 1.9862462098033036e-05, + "loss": 0.1007, + "step": 4964 + }, + { + "epoch": 0.65, + "learning_rate": 1.986237443053705e-05, + "loss": 0.0991, + "step": 4965 + }, + { + "epoch": 0.65, + "learning_rate": 1.9862286735303663e-05, + "loss": 0.0959, + "step": 4966 + }, + { + "epoch": 0.65, + "learning_rate": 1.986219901233312e-05, + "loss": 0.0899, + "step": 4967 + }, + { + "epoch": 0.65, + "learning_rate": 1.9862111261625666e-05, + "loss": 0.086, + "step": 4968 + }, + { + "epoch": 0.65, + "learning_rate": 1.9862023483181553e-05, + "loss": 0.1063, + "step": 4969 + }, + { + "epoch": 0.65, + "learning_rate": 1.9861935677001023e-05, + "loss": 0.0927, + "step": 4970 + }, + { + "epoch": 0.65, + "learning_rate": 1.9861847843084322e-05, + "loss": 0.0877, + "step": 4971 + }, + { + "epoch": 0.65, + "learning_rate": 1.98617599814317e-05, + "loss": 0.085, + "step": 4972 + }, + { + "epoch": 0.65, + "learning_rate": 1.98616720920434e-05, + "loss": 0.0895, + "step": 4973 + }, + { + "epoch": 0.65, + "learning_rate": 1.986158417491968e-05, + "loss": 0.0958, + "step": 4974 + }, + { + "epoch": 0.65, + "learning_rate": 1.9861496230060774e-05, + "loss": 0.0992, + "step": 4975 + }, + { + "epoch": 0.65, + "learning_rate": 1.9861408257466938e-05, + "loss": 0.0901, + "step": 4976 + }, + { + "epoch": 0.65, + "learning_rate": 1.9861320257138414e-05, + "loss": 0.0871, + "step": 4977 + }, + { + "epoch": 0.65, + "learning_rate": 1.9861232229075452e-05, + "loss": 0.0877, + "step": 4978 + }, + { + "epoch": 0.65, + "learning_rate": 1.9861144173278298e-05, + "loss": 0.0945, + "step": 4979 + }, + { + "epoch": 0.65, + "learning_rate": 1.9861056089747205e-05, + "loss": 0.0932, + "step": 4980 + }, + { + "epoch": 0.65, + "learning_rate": 1.9860967978482416e-05, + "loss": 0.0936, + "step": 4981 + }, + { + "epoch": 0.65, + "learning_rate": 1.986087983948418e-05, + "loss": 0.0857, + "step": 4982 + }, + { + "epoch": 0.65, + "learning_rate": 1.986079167275274e-05, + "loss": 0.106, + "step": 4983 + }, + { + "epoch": 0.65, + "learning_rate": 1.9860703478288354e-05, + "loss": 0.0853, + "step": 4984 + }, + { + "epoch": 0.65, + "learning_rate": 1.9860615256091262e-05, + "loss": 0.0866, + "step": 4985 + }, + { + "epoch": 0.65, + "learning_rate": 1.9860527006161715e-05, + "loss": 0.084, + "step": 4986 + }, + { + "epoch": 0.65, + "learning_rate": 1.986043872849996e-05, + "loss": 0.0806, + "step": 4987 + }, + { + "epoch": 0.65, + "learning_rate": 1.986035042310625e-05, + "loss": 0.0952, + "step": 4988 + }, + { + "epoch": 0.65, + "learning_rate": 1.9860262089980825e-05, + "loss": 0.0863, + "step": 4989 + }, + { + "epoch": 0.65, + "learning_rate": 1.986017372912394e-05, + "loss": 0.0878, + "step": 4990 + }, + { + "epoch": 0.65, + "learning_rate": 1.9860085340535843e-05, + "loss": 0.081, + "step": 4991 + }, + { + "epoch": 0.65, + "learning_rate": 1.985999692421678e-05, + "loss": 0.0949, + "step": 4992 + }, + { + "epoch": 0.65, + "learning_rate": 1.9859908480166997e-05, + "loss": 0.0819, + "step": 4993 + }, + { + "epoch": 0.65, + "learning_rate": 1.985982000838675e-05, + "loss": 0.087, + "step": 4994 + }, + { + "epoch": 0.65, + "learning_rate": 1.9859731508876287e-05, + "loss": 0.0992, + "step": 4995 + }, + { + "epoch": 0.65, + "learning_rate": 1.985964298163585e-05, + "loss": 0.0936, + "step": 4996 + }, + { + "epoch": 0.65, + "learning_rate": 1.985955442666569e-05, + "loss": 0.0867, + "step": 4997 + }, + { + "epoch": 0.65, + "learning_rate": 1.9859465843966064e-05, + "loss": 0.0894, + "step": 4998 + }, + { + "epoch": 0.65, + "learning_rate": 1.9859377233537214e-05, + "loss": 0.0971, + "step": 4999 + }, + { + "epoch": 0.65, + "learning_rate": 1.9859288595379386e-05, + "loss": 0.0883, + "step": 5000 + }, + { + "epoch": 0.66, + "learning_rate": 1.985919992949284e-05, + "loss": 0.088, + "step": 5001 + }, + { + "epoch": 0.66, + "learning_rate": 1.9859111235877816e-05, + "loss": 0.0927, + "step": 5002 + }, + { + "epoch": 0.66, + "learning_rate": 1.9859022514534566e-05, + "loss": 0.0913, + "step": 5003 + }, + { + "epoch": 0.66, + "learning_rate": 1.9858933765463343e-05, + "loss": 0.0839, + "step": 5004 + }, + { + "epoch": 0.66, + "learning_rate": 1.9858844988664392e-05, + "loss": 0.0863, + "step": 5005 + }, + { + "epoch": 0.66, + "learning_rate": 1.9858756184137963e-05, + "loss": 0.088, + "step": 5006 + }, + { + "epoch": 0.66, + "learning_rate": 1.985866735188431e-05, + "loss": 0.094, + "step": 5007 + }, + { + "epoch": 0.66, + "learning_rate": 1.985857849190368e-05, + "loss": 0.0925, + "step": 5008 + }, + { + "epoch": 0.66, + "learning_rate": 1.9858489604196322e-05, + "loss": 0.0801, + "step": 5009 + }, + { + "epoch": 0.66, + "learning_rate": 1.9858400688762486e-05, + "loss": 0.084, + "step": 5010 + }, + { + "epoch": 0.66, + "learning_rate": 1.9858311745602425e-05, + "loss": 0.086, + "step": 5011 + }, + { + "epoch": 0.66, + "learning_rate": 1.9858222774716384e-05, + "loss": 0.1004, + "step": 5012 + }, + { + "epoch": 0.66, + "learning_rate": 1.985813377610462e-05, + "loss": 0.0996, + "step": 5013 + }, + { + "epoch": 0.66, + "learning_rate": 1.985804474976738e-05, + "loss": 0.0897, + "step": 5014 + }, + { + "epoch": 0.66, + "learning_rate": 1.9857955695704913e-05, + "loss": 0.0971, + "step": 5015 + }, + { + "epoch": 0.66, + "learning_rate": 1.9857866613917468e-05, + "loss": 0.0964, + "step": 5016 + }, + { + "epoch": 0.66, + "learning_rate": 1.98577775044053e-05, + "loss": 0.0919, + "step": 5017 + }, + { + "epoch": 0.66, + "learning_rate": 1.985768836716866e-05, + "loss": 0.0846, + "step": 5018 + }, + { + "epoch": 0.66, + "learning_rate": 1.985759920220779e-05, + "loss": 0.0957, + "step": 5019 + }, + { + "epoch": 0.66, + "learning_rate": 1.9857510009522952e-05, + "loss": 0.0954, + "step": 5020 + }, + { + "epoch": 0.66, + "learning_rate": 1.9857420789114395e-05, + "loss": 0.0863, + "step": 5021 + }, + { + "epoch": 0.66, + "learning_rate": 1.9857331540982364e-05, + "loss": 0.0839, + "step": 5022 + }, + { + "epoch": 0.66, + "learning_rate": 1.985724226512711e-05, + "loss": 0.1028, + "step": 5023 + }, + { + "epoch": 0.66, + "learning_rate": 1.985715296154889e-05, + "loss": 0.0722, + "step": 5024 + }, + { + "epoch": 0.66, + "learning_rate": 1.985706363024795e-05, + "loss": 0.0974, + "step": 5025 + }, + { + "epoch": 0.66, + "learning_rate": 1.9856974271224546e-05, + "loss": 0.0982, + "step": 5026 + }, + { + "epoch": 0.66, + "learning_rate": 1.9856884884478928e-05, + "loss": 0.0879, + "step": 5027 + }, + { + "epoch": 0.66, + "learning_rate": 1.9856795470011344e-05, + "loss": 0.0713, + "step": 5028 + }, + { + "epoch": 0.66, + "learning_rate": 1.9856706027822046e-05, + "loss": 0.0981, + "step": 5029 + }, + { + "epoch": 0.66, + "learning_rate": 1.985661655791129e-05, + "loss": 0.0853, + "step": 5030 + }, + { + "epoch": 0.66, + "learning_rate": 1.9856527060279323e-05, + "loss": 0.0845, + "step": 5031 + }, + { + "epoch": 0.66, + "learning_rate": 1.9856437534926402e-05, + "loss": 0.0854, + "step": 5032 + }, + { + "epoch": 0.66, + "learning_rate": 1.985634798185277e-05, + "loss": 0.0959, + "step": 5033 + }, + { + "epoch": 0.66, + "learning_rate": 1.985625840105869e-05, + "loss": 0.1002, + "step": 5034 + }, + { + "epoch": 0.66, + "learning_rate": 1.9856168792544402e-05, + "loss": 0.0939, + "step": 5035 + }, + { + "epoch": 0.66, + "learning_rate": 1.985607915631017e-05, + "loss": 0.0827, + "step": 5036 + }, + { + "epoch": 0.66, + "learning_rate": 1.9855989492356235e-05, + "loss": 0.0927, + "step": 5037 + }, + { + "epoch": 0.66, + "learning_rate": 1.985589980068286e-05, + "loss": 0.0871, + "step": 5038 + }, + { + "epoch": 0.66, + "learning_rate": 1.9855810081290285e-05, + "loss": 0.0982, + "step": 5039 + }, + { + "epoch": 0.66, + "learning_rate": 1.9855720334178775e-05, + "loss": 0.0863, + "step": 5040 + }, + { + "epoch": 0.66, + "learning_rate": 1.9855630559348573e-05, + "loss": 0.092, + "step": 5041 + }, + { + "epoch": 0.66, + "learning_rate": 1.9855540756799936e-05, + "loss": 0.0866, + "step": 5042 + }, + { + "epoch": 0.66, + "learning_rate": 1.9855450926533115e-05, + "loss": 0.085, + "step": 5043 + }, + { + "epoch": 0.66, + "learning_rate": 1.9855361068548364e-05, + "loss": 0.087, + "step": 5044 + }, + { + "epoch": 0.66, + "learning_rate": 1.9855271182845933e-05, + "loss": 0.0893, + "step": 5045 + }, + { + "epoch": 0.66, + "learning_rate": 1.9855181269426077e-05, + "loss": 0.0897, + "step": 5046 + }, + { + "epoch": 0.66, + "learning_rate": 1.985509132828905e-05, + "loss": 0.0867, + "step": 5047 + }, + { + "epoch": 0.66, + "learning_rate": 1.9855001359435102e-05, + "loss": 0.0988, + "step": 5048 + }, + { + "epoch": 0.66, + "learning_rate": 1.985491136286449e-05, + "loss": 0.0862, + "step": 5049 + }, + { + "epoch": 0.66, + "learning_rate": 1.9854821338577462e-05, + "loss": 0.0941, + "step": 5050 + }, + { + "epoch": 0.66, + "learning_rate": 1.9854731286574275e-05, + "loss": 0.0933, + "step": 5051 + }, + { + "epoch": 0.66, + "learning_rate": 1.985464120685518e-05, + "loss": 0.0913, + "step": 5052 + }, + { + "epoch": 0.66, + "learning_rate": 1.9854551099420432e-05, + "loss": 0.0844, + "step": 5053 + }, + { + "epoch": 0.66, + "learning_rate": 1.9854460964270284e-05, + "loss": 0.0851, + "step": 5054 + }, + { + "epoch": 0.66, + "learning_rate": 1.985437080140499e-05, + "loss": 0.0942, + "step": 5055 + }, + { + "epoch": 0.66, + "learning_rate": 1.9854280610824802e-05, + "loss": 0.0955, + "step": 5056 + }, + { + "epoch": 0.66, + "learning_rate": 1.9854190392529972e-05, + "loss": 0.0913, + "step": 5057 + }, + { + "epoch": 0.66, + "learning_rate": 1.985410014652076e-05, + "loss": 0.0823, + "step": 5058 + }, + { + "epoch": 0.66, + "learning_rate": 1.9854009872797414e-05, + "loss": 0.0858, + "step": 5059 + }, + { + "epoch": 0.66, + "learning_rate": 1.985391957136019e-05, + "loss": 0.0936, + "step": 5060 + }, + { + "epoch": 0.66, + "learning_rate": 1.9853829242209345e-05, + "loss": 0.1009, + "step": 5061 + }, + { + "epoch": 0.66, + "learning_rate": 1.985373888534513e-05, + "loss": 0.095, + "step": 5062 + }, + { + "epoch": 0.66, + "learning_rate": 1.9853648500767793e-05, + "loss": 0.1022, + "step": 5063 + }, + { + "epoch": 0.66, + "learning_rate": 1.98535580884776e-05, + "loss": 0.0866, + "step": 5064 + }, + { + "epoch": 0.66, + "learning_rate": 1.98534676484748e-05, + "loss": 0.0903, + "step": 5065 + }, + { + "epoch": 0.66, + "learning_rate": 1.9853377180759642e-05, + "loss": 0.0812, + "step": 5066 + }, + { + "epoch": 0.66, + "learning_rate": 1.9853286685332393e-05, + "loss": 0.1021, + "step": 5067 + }, + { + "epoch": 0.66, + "learning_rate": 1.985319616219329e-05, + "loss": 0.0923, + "step": 5068 + }, + { + "epoch": 0.66, + "learning_rate": 1.9853105611342606e-05, + "loss": 0.0929, + "step": 5069 + }, + { + "epoch": 0.66, + "learning_rate": 1.9853015032780584e-05, + "loss": 0.1021, + "step": 5070 + }, + { + "epoch": 0.66, + "learning_rate": 1.985292442650748e-05, + "loss": 0.1052, + "step": 5071 + }, + { + "epoch": 0.66, + "learning_rate": 1.9852833792523552e-05, + "loss": 0.0848, + "step": 5072 + }, + { + "epoch": 0.66, + "learning_rate": 1.9852743130829055e-05, + "loss": 0.081, + "step": 5073 + }, + { + "epoch": 0.66, + "learning_rate": 1.9852652441424243e-05, + "loss": 0.0985, + "step": 5074 + }, + { + "epoch": 0.66, + "learning_rate": 1.985256172430937e-05, + "loss": 0.0963, + "step": 5075 + }, + { + "epoch": 0.66, + "learning_rate": 1.9852470979484693e-05, + "loss": 0.0794, + "step": 5076 + }, + { + "epoch": 0.67, + "learning_rate": 1.9852380206950463e-05, + "loss": 0.0778, + "step": 5077 + }, + { + "epoch": 0.67, + "learning_rate": 1.985228940670694e-05, + "loss": 0.0933, + "step": 5078 + }, + { + "epoch": 0.67, + "learning_rate": 1.9852198578754378e-05, + "loss": 0.0936, + "step": 5079 + }, + { + "epoch": 0.67, + "learning_rate": 1.985210772309303e-05, + "loss": 0.09, + "step": 5080 + }, + { + "epoch": 0.67, + "learning_rate": 1.9852016839723157e-05, + "loss": 0.1009, + "step": 5081 + }, + { + "epoch": 0.67, + "learning_rate": 1.985192592864501e-05, + "loss": 0.0991, + "step": 5082 + }, + { + "epoch": 0.67, + "learning_rate": 1.9851834989858845e-05, + "loss": 0.0991, + "step": 5083 + }, + { + "epoch": 0.67, + "learning_rate": 1.985174402336492e-05, + "loss": 0.0967, + "step": 5084 + }, + { + "epoch": 0.67, + "learning_rate": 1.985165302916349e-05, + "loss": 0.08, + "step": 5085 + }, + { + "epoch": 0.67, + "learning_rate": 1.9851562007254808e-05, + "loss": 0.0948, + "step": 5086 + }, + { + "epoch": 0.67, + "learning_rate": 1.9851470957639135e-05, + "loss": 0.1045, + "step": 5087 + }, + { + "epoch": 0.67, + "learning_rate": 1.985137988031672e-05, + "loss": 0.0951, + "step": 5088 + }, + { + "epoch": 0.67, + "learning_rate": 1.9851288775287828e-05, + "loss": 0.0972, + "step": 5089 + }, + { + "epoch": 0.67, + "learning_rate": 1.9851197642552715e-05, + "loss": 0.0968, + "step": 5090 + }, + { + "epoch": 0.67, + "learning_rate": 1.9851106482111625e-05, + "loss": 0.0923, + "step": 5091 + }, + { + "epoch": 0.67, + "learning_rate": 1.9851015293964825e-05, + "loss": 0.0923, + "step": 5092 + }, + { + "epoch": 0.67, + "learning_rate": 1.985092407811257e-05, + "loss": 0.0987, + "step": 5093 + }, + { + "epoch": 0.67, + "learning_rate": 1.9850832834555116e-05, + "loss": 0.1013, + "step": 5094 + }, + { + "epoch": 0.67, + "learning_rate": 1.985074156329272e-05, + "loss": 0.0827, + "step": 5095 + }, + { + "epoch": 0.67, + "learning_rate": 1.9850650264325636e-05, + "loss": 0.0871, + "step": 5096 + }, + { + "epoch": 0.67, + "learning_rate": 1.9850558937654127e-05, + "loss": 0.0879, + "step": 5097 + }, + { + "epoch": 0.67, + "learning_rate": 1.985046758327844e-05, + "loss": 0.1019, + "step": 5098 + }, + { + "epoch": 0.67, + "learning_rate": 1.9850376201198836e-05, + "loss": 0.0961, + "step": 5099 + }, + { + "epoch": 0.67, + "learning_rate": 1.985028479141558e-05, + "loss": 0.1022, + "step": 5100 + }, + { + "epoch": 0.67, + "learning_rate": 1.9850193353928918e-05, + "loss": 0.0871, + "step": 5101 + }, + { + "epoch": 0.67, + "learning_rate": 1.9850101888739116e-05, + "loss": 0.0942, + "step": 5102 + }, + { + "epoch": 0.67, + "learning_rate": 1.9850010395846423e-05, + "loss": 0.0923, + "step": 5103 + }, + { + "epoch": 0.67, + "learning_rate": 1.9849918875251105e-05, + "loss": 0.0903, + "step": 5104 + }, + { + "epoch": 0.67, + "learning_rate": 1.984982732695341e-05, + "loss": 0.0917, + "step": 5105 + }, + { + "epoch": 0.67, + "learning_rate": 1.98497357509536e-05, + "loss": 0.092, + "step": 5106 + }, + { + "epoch": 0.67, + "learning_rate": 1.9849644147251935e-05, + "loss": 0.1048, + "step": 5107 + }, + { + "epoch": 0.67, + "learning_rate": 1.984955251584867e-05, + "loss": 0.0815, + "step": 5108 + }, + { + "epoch": 0.67, + "learning_rate": 1.9849460856744062e-05, + "loss": 0.0875, + "step": 5109 + }, + { + "epoch": 0.67, + "learning_rate": 1.9849369169938372e-05, + "loss": 0.0965, + "step": 5110 + }, + { + "epoch": 0.67, + "learning_rate": 1.9849277455431855e-05, + "loss": 0.0798, + "step": 5111 + }, + { + "epoch": 0.67, + "learning_rate": 1.984918571322477e-05, + "loss": 0.0905, + "step": 5112 + }, + { + "epoch": 0.67, + "learning_rate": 1.984909394331737e-05, + "loss": 0.0871, + "step": 5113 + }, + { + "epoch": 0.67, + "learning_rate": 1.9849002145709923e-05, + "loss": 0.1079, + "step": 5114 + }, + { + "epoch": 0.67, + "learning_rate": 1.984891032040268e-05, + "loss": 0.0989, + "step": 5115 + }, + { + "epoch": 0.67, + "learning_rate": 1.9848818467395902e-05, + "loss": 0.1046, + "step": 5116 + }, + { + "epoch": 0.67, + "learning_rate": 1.9848726586689846e-05, + "loss": 0.1013, + "step": 5117 + }, + { + "epoch": 0.67, + "learning_rate": 1.984863467828477e-05, + "loss": 0.0909, + "step": 5118 + }, + { + "epoch": 0.67, + "learning_rate": 1.9848542742180937e-05, + "loss": 0.0918, + "step": 5119 + }, + { + "epoch": 0.67, + "learning_rate": 1.9848450778378597e-05, + "loss": 0.0916, + "step": 5120 + }, + { + "epoch": 0.67, + "learning_rate": 1.9848358786878015e-05, + "loss": 0.0904, + "step": 5121 + }, + { + "epoch": 0.67, + "learning_rate": 1.9848266767679453e-05, + "loss": 0.0917, + "step": 5122 + }, + { + "epoch": 0.67, + "learning_rate": 1.984817472078316e-05, + "loss": 0.0993, + "step": 5123 + }, + { + "epoch": 0.67, + "learning_rate": 1.9848082646189402e-05, + "loss": 0.0991, + "step": 5124 + }, + { + "epoch": 0.67, + "learning_rate": 1.9847990543898436e-05, + "loss": 0.0907, + "step": 5125 + }, + { + "epoch": 0.67, + "learning_rate": 1.984789841391052e-05, + "loss": 0.099, + "step": 5126 + }, + { + "epoch": 0.67, + "learning_rate": 1.9847806256225918e-05, + "loss": 0.0975, + "step": 5127 + }, + { + "epoch": 0.67, + "learning_rate": 1.984771407084488e-05, + "loss": 0.0898, + "step": 5128 + }, + { + "epoch": 0.67, + "learning_rate": 1.9847621857767674e-05, + "loss": 0.0911, + "step": 5129 + }, + { + "epoch": 0.67, + "learning_rate": 1.9847529616994556e-05, + "loss": 0.0878, + "step": 5130 + }, + { + "epoch": 0.67, + "learning_rate": 1.9847437348525784e-05, + "loss": 0.087, + "step": 5131 + }, + { + "epoch": 0.67, + "learning_rate": 1.984734505236162e-05, + "loss": 0.0832, + "step": 5132 + }, + { + "epoch": 0.67, + "learning_rate": 1.984725272850232e-05, + "loss": 0.095, + "step": 5133 + }, + { + "epoch": 0.67, + "learning_rate": 1.9847160376948146e-05, + "loss": 0.0801, + "step": 5134 + }, + { + "epoch": 0.67, + "learning_rate": 1.984706799769936e-05, + "loss": 0.0972, + "step": 5135 + }, + { + "epoch": 0.67, + "learning_rate": 1.984697559075622e-05, + "loss": 0.0837, + "step": 5136 + }, + { + "epoch": 0.67, + "learning_rate": 1.9846883156118982e-05, + "loss": 0.0855, + "step": 5137 + }, + { + "epoch": 0.67, + "learning_rate": 1.9846790693787913e-05, + "loss": 0.1094, + "step": 5138 + }, + { + "epoch": 0.67, + "learning_rate": 1.984669820376327e-05, + "loss": 0.0863, + "step": 5139 + }, + { + "epoch": 0.67, + "learning_rate": 1.984660568604531e-05, + "loss": 0.0789, + "step": 5140 + }, + { + "epoch": 0.67, + "learning_rate": 1.98465131406343e-05, + "loss": 0.0792, + "step": 5141 + }, + { + "epoch": 0.67, + "learning_rate": 1.984642056753049e-05, + "loss": 0.0899, + "step": 5142 + }, + { + "epoch": 0.67, + "learning_rate": 1.9846327966734153e-05, + "loss": 0.0921, + "step": 5143 + }, + { + "epoch": 0.67, + "learning_rate": 1.9846235338245537e-05, + "loss": 0.0884, + "step": 5144 + }, + { + "epoch": 0.67, + "learning_rate": 1.9846142682064913e-05, + "loss": 0.0863, + "step": 5145 + }, + { + "epoch": 0.67, + "learning_rate": 1.9846049998192536e-05, + "loss": 0.0924, + "step": 5146 + }, + { + "epoch": 0.67, + "learning_rate": 1.9845957286628667e-05, + "loss": 0.0953, + "step": 5147 + }, + { + "epoch": 0.67, + "learning_rate": 1.9845864547373566e-05, + "loss": 0.0873, + "step": 5148 + }, + { + "epoch": 0.67, + "learning_rate": 1.9845771780427497e-05, + "loss": 0.0845, + "step": 5149 + }, + { + "epoch": 0.67, + "learning_rate": 1.984567898579072e-05, + "loss": 0.0704, + "step": 5150 + }, + { + "epoch": 0.67, + "learning_rate": 1.9845586163463495e-05, + "loss": 0.0859, + "step": 5151 + }, + { + "epoch": 0.67, + "learning_rate": 1.9845493313446083e-05, + "loss": 0.0974, + "step": 5152 + }, + { + "epoch": 0.68, + "learning_rate": 1.9845400435738744e-05, + "loss": 0.0961, + "step": 5153 + }, + { + "epoch": 0.68, + "learning_rate": 1.984530753034174e-05, + "loss": 0.1035, + "step": 5154 + }, + { + "epoch": 0.68, + "learning_rate": 1.9845214597255334e-05, + "loss": 0.0826, + "step": 5155 + }, + { + "epoch": 0.68, + "learning_rate": 1.9845121636479786e-05, + "loss": 0.0875, + "step": 5156 + }, + { + "epoch": 0.68, + "learning_rate": 1.9845028648015357e-05, + "loss": 0.0849, + "step": 5157 + }, + { + "epoch": 0.68, + "learning_rate": 1.9844935631862312e-05, + "loss": 0.0888, + "step": 5158 + }, + { + "epoch": 0.68, + "learning_rate": 1.9844842588020906e-05, + "loss": 0.0823, + "step": 5159 + }, + { + "epoch": 0.68, + "learning_rate": 1.9844749516491408e-05, + "loss": 0.0969, + "step": 5160 + }, + { + "epoch": 0.68, + "learning_rate": 1.9844656417274074e-05, + "loss": 0.0951, + "step": 5161 + }, + { + "epoch": 0.68, + "learning_rate": 1.984456329036917e-05, + "loss": 0.0893, + "step": 5162 + }, + { + "epoch": 0.68, + "learning_rate": 1.9844470135776953e-05, + "loss": 0.0911, + "step": 5163 + }, + { + "epoch": 0.68, + "learning_rate": 1.984437695349769e-05, + "loss": 0.0931, + "step": 5164 + }, + { + "epoch": 0.68, + "learning_rate": 1.9844283743531646e-05, + "loss": 0.0799, + "step": 5165 + }, + { + "epoch": 0.68, + "learning_rate": 1.9844190505879072e-05, + "loss": 0.0943, + "step": 5166 + }, + { + "epoch": 0.68, + "learning_rate": 1.9844097240540237e-05, + "loss": 0.0829, + "step": 5167 + }, + { + "epoch": 0.68, + "learning_rate": 1.9844003947515402e-05, + "loss": 0.0928, + "step": 5168 + }, + { + "epoch": 0.68, + "learning_rate": 1.984391062680483e-05, + "loss": 0.0923, + "step": 5169 + }, + { + "epoch": 0.68, + "learning_rate": 1.9843817278408787e-05, + "loss": 0.0839, + "step": 5170 + }, + { + "epoch": 0.68, + "learning_rate": 1.984372390232753e-05, + "loss": 0.0847, + "step": 5171 + }, + { + "epoch": 0.68, + "learning_rate": 1.9843630498561322e-05, + "loss": 0.0982, + "step": 5172 + }, + { + "epoch": 0.68, + "learning_rate": 1.9843537067110432e-05, + "loss": 0.0879, + "step": 5173 + }, + { + "epoch": 0.68, + "learning_rate": 1.9843443607975114e-05, + "loss": 0.079, + "step": 5174 + }, + { + "epoch": 0.68, + "learning_rate": 1.984335012115564e-05, + "loss": 0.0919, + "step": 5175 + }, + { + "epoch": 0.68, + "learning_rate": 1.9843256606652264e-05, + "loss": 0.0947, + "step": 5176 + }, + { + "epoch": 0.68, + "learning_rate": 1.984316306446525e-05, + "loss": 0.0775, + "step": 5177 + }, + { + "epoch": 0.68, + "learning_rate": 1.984306949459487e-05, + "loss": 0.0865, + "step": 5178 + }, + { + "epoch": 0.68, + "learning_rate": 1.9842975897041377e-05, + "loss": 0.0897, + "step": 5179 + }, + { + "epoch": 0.68, + "learning_rate": 1.9842882271805043e-05, + "loss": 0.0907, + "step": 5180 + }, + { + "epoch": 0.68, + "learning_rate": 1.9842788618886124e-05, + "loss": 0.0906, + "step": 5181 + }, + { + "epoch": 0.68, + "learning_rate": 1.9842694938284883e-05, + "loss": 0.087, + "step": 5182 + }, + { + "epoch": 0.68, + "learning_rate": 1.984260123000159e-05, + "loss": 0.0902, + "step": 5183 + }, + { + "epoch": 0.68, + "learning_rate": 1.9842507494036508e-05, + "loss": 0.0871, + "step": 5184 + }, + { + "epoch": 0.68, + "learning_rate": 1.9842413730389894e-05, + "loss": 0.0914, + "step": 5185 + }, + { + "epoch": 0.68, + "learning_rate": 1.9842319939062017e-05, + "loss": 0.0911, + "step": 5186 + }, + { + "epoch": 0.68, + "learning_rate": 1.9842226120053136e-05, + "loss": 0.091, + "step": 5187 + }, + { + "epoch": 0.68, + "learning_rate": 1.984213227336352e-05, + "loss": 0.0953, + "step": 5188 + }, + { + "epoch": 0.68, + "learning_rate": 1.9842038398993435e-05, + "loss": 0.0946, + "step": 5189 + }, + { + "epoch": 0.68, + "learning_rate": 1.984194449694314e-05, + "loss": 0.0967, + "step": 5190 + }, + { + "epoch": 0.68, + "learning_rate": 1.9841850567212896e-05, + "loss": 0.0997, + "step": 5191 + }, + { + "epoch": 0.68, + "learning_rate": 1.9841756609802973e-05, + "loss": 0.0882, + "step": 5192 + }, + { + "epoch": 0.68, + "learning_rate": 1.9841662624713635e-05, + "loss": 0.0941, + "step": 5193 + }, + { + "epoch": 0.68, + "learning_rate": 1.9841568611945142e-05, + "loss": 0.0865, + "step": 5194 + }, + { + "epoch": 0.68, + "learning_rate": 1.9841474571497766e-05, + "loss": 0.0985, + "step": 5195 + }, + { + "epoch": 0.68, + "learning_rate": 1.9841380503371767e-05, + "loss": 0.0912, + "step": 5196 + }, + { + "epoch": 0.68, + "learning_rate": 1.9841286407567403e-05, + "loss": 0.0855, + "step": 5197 + }, + { + "epoch": 0.68, + "learning_rate": 1.984119228408495e-05, + "loss": 0.0837, + "step": 5198 + }, + { + "epoch": 0.68, + "learning_rate": 1.9841098132924667e-05, + "loss": 0.0928, + "step": 5199 + }, + { + "epoch": 0.68, + "learning_rate": 1.984100395408682e-05, + "loss": 0.0864, + "step": 5200 + }, + { + "epoch": 0.68, + "learning_rate": 1.9840909747571672e-05, + "loss": 0.0956, + "step": 5201 + }, + { + "epoch": 0.68, + "learning_rate": 1.984081551337949e-05, + "loss": 0.0839, + "step": 5202 + }, + { + "epoch": 0.68, + "learning_rate": 1.984072125151054e-05, + "loss": 0.0955, + "step": 5203 + }, + { + "epoch": 0.68, + "learning_rate": 1.9840626961965085e-05, + "loss": 0.0855, + "step": 5204 + }, + { + "epoch": 0.68, + "learning_rate": 1.984053264474339e-05, + "loss": 0.1011, + "step": 5205 + }, + { + "epoch": 0.68, + "learning_rate": 1.9840438299845717e-05, + "loss": 0.0902, + "step": 5206 + }, + { + "epoch": 0.68, + "learning_rate": 1.9840343927272342e-05, + "loss": 0.1059, + "step": 5207 + }, + { + "epoch": 0.68, + "learning_rate": 1.984024952702352e-05, + "loss": 0.084, + "step": 5208 + }, + { + "epoch": 0.68, + "learning_rate": 1.984015509909952e-05, + "loss": 0.0853, + "step": 5209 + }, + { + "epoch": 0.68, + "learning_rate": 1.984006064350061e-05, + "loss": 0.0923, + "step": 5210 + }, + { + "epoch": 0.68, + "learning_rate": 1.9839966160227053e-05, + "loss": 0.0854, + "step": 5211 + }, + { + "epoch": 0.68, + "learning_rate": 1.9839871649279113e-05, + "loss": 0.0927, + "step": 5212 + }, + { + "epoch": 0.68, + "learning_rate": 1.9839777110657058e-05, + "loss": 0.0832, + "step": 5213 + }, + { + "epoch": 0.68, + "learning_rate": 1.983968254436116e-05, + "loss": 0.0967, + "step": 5214 + }, + { + "epoch": 0.68, + "learning_rate": 1.983958795039167e-05, + "loss": 0.0815, + "step": 5215 + }, + { + "epoch": 0.68, + "learning_rate": 1.983949332874887e-05, + "loss": 0.0927, + "step": 5216 + }, + { + "epoch": 0.68, + "learning_rate": 1.9839398679433014e-05, + "loss": 0.0907, + "step": 5217 + }, + { + "epoch": 0.68, + "learning_rate": 1.9839304002444377e-05, + "loss": 0.0838, + "step": 5218 + }, + { + "epoch": 0.68, + "learning_rate": 1.983920929778322e-05, + "loss": 0.096, + "step": 5219 + }, + { + "epoch": 0.68, + "learning_rate": 1.983911456544981e-05, + "loss": 0.1003, + "step": 5220 + }, + { + "epoch": 0.68, + "learning_rate": 1.9839019805444412e-05, + "loss": 0.0939, + "step": 5221 + }, + { + "epoch": 0.68, + "learning_rate": 1.98389250177673e-05, + "loss": 0.0941, + "step": 5222 + }, + { + "epoch": 0.68, + "learning_rate": 1.9838830202418732e-05, + "loss": 0.0874, + "step": 5223 + }, + { + "epoch": 0.68, + "learning_rate": 1.983873535939898e-05, + "loss": 0.0843, + "step": 5224 + }, + { + "epoch": 0.68, + "learning_rate": 1.9838640488708306e-05, + "loss": 0.0874, + "step": 5225 + }, + { + "epoch": 0.68, + "learning_rate": 1.983854559034698e-05, + "loss": 0.0872, + "step": 5226 + }, + { + "epoch": 0.68, + "learning_rate": 1.9838450664315272e-05, + "loss": 0.0993, + "step": 5227 + }, + { + "epoch": 0.68, + "learning_rate": 1.9838355710613443e-05, + "loss": 0.0899, + "step": 5228 + }, + { + "epoch": 0.68, + "learning_rate": 1.9838260729241764e-05, + "loss": 0.1041, + "step": 5229 + }, + { + "epoch": 0.69, + "learning_rate": 1.9838165720200497e-05, + "loss": 0.0939, + "step": 5230 + }, + { + "epoch": 0.69, + "learning_rate": 1.9838070683489917e-05, + "loss": 0.0971, + "step": 5231 + }, + { + "epoch": 0.69, + "learning_rate": 1.9837975619110286e-05, + "loss": 0.0837, + "step": 5232 + }, + { + "epoch": 0.69, + "learning_rate": 1.983788052706187e-05, + "loss": 0.0965, + "step": 5233 + }, + { + "epoch": 0.69, + "learning_rate": 1.983778540734494e-05, + "loss": 0.0873, + "step": 5234 + }, + { + "epoch": 0.69, + "learning_rate": 1.9837690259959762e-05, + "loss": 0.0962, + "step": 5235 + }, + { + "epoch": 0.69, + "learning_rate": 1.9837595084906607e-05, + "loss": 0.083, + "step": 5236 + }, + { + "epoch": 0.69, + "learning_rate": 1.9837499882185738e-05, + "loss": 0.1015, + "step": 5237 + }, + { + "epoch": 0.69, + "learning_rate": 1.9837404651797425e-05, + "loss": 0.0962, + "step": 5238 + }, + { + "epoch": 0.69, + "learning_rate": 1.9837309393741933e-05, + "loss": 0.0911, + "step": 5239 + }, + { + "epoch": 0.69, + "learning_rate": 1.9837214108019535e-05, + "loss": 0.0892, + "step": 5240 + }, + { + "epoch": 0.69, + "learning_rate": 1.9837118794630494e-05, + "loss": 0.081, + "step": 5241 + }, + { + "epoch": 0.69, + "learning_rate": 1.983702345357508e-05, + "loss": 0.0895, + "step": 5242 + }, + { + "epoch": 0.69, + "learning_rate": 1.9836928084853565e-05, + "loss": 0.091, + "step": 5243 + }, + { + "epoch": 0.69, + "learning_rate": 1.983683268846621e-05, + "loss": 0.0919, + "step": 5244 + }, + { + "epoch": 0.69, + "learning_rate": 1.983673726441329e-05, + "loss": 0.0949, + "step": 5245 + }, + { + "epoch": 0.69, + "learning_rate": 1.9836641812695066e-05, + "loss": 0.0889, + "step": 5246 + }, + { + "epoch": 0.69, + "learning_rate": 1.9836546333311813e-05, + "loss": 0.0958, + "step": 5247 + }, + { + "epoch": 0.69, + "learning_rate": 1.9836450826263796e-05, + "loss": 0.0947, + "step": 5248 + }, + { + "epoch": 0.69, + "learning_rate": 1.9836355291551287e-05, + "loss": 0.0791, + "step": 5249 + }, + { + "epoch": 0.69, + "learning_rate": 1.9836259729174553e-05, + "loss": 0.0949, + "step": 5250 + }, + { + "epoch": 0.69, + "learning_rate": 1.983616413913386e-05, + "loss": 0.0833, + "step": 5251 + }, + { + "epoch": 0.69, + "learning_rate": 1.983606852142948e-05, + "loss": 0.0877, + "step": 5252 + }, + { + "epoch": 0.69, + "learning_rate": 1.9835972876061682e-05, + "loss": 0.0974, + "step": 5253 + }, + { + "epoch": 0.69, + "learning_rate": 1.9835877203030733e-05, + "loss": 0.0976, + "step": 5254 + }, + { + "epoch": 0.69, + "learning_rate": 1.9835781502336903e-05, + "loss": 0.0932, + "step": 5255 + }, + { + "epoch": 0.69, + "learning_rate": 1.9835685773980465e-05, + "loss": 0.0986, + "step": 5256 + }, + { + "epoch": 0.69, + "learning_rate": 1.983559001796168e-05, + "loss": 0.0975, + "step": 5257 + }, + { + "epoch": 0.69, + "learning_rate": 1.983549423428082e-05, + "loss": 0.0901, + "step": 5258 + }, + { + "epoch": 0.69, + "learning_rate": 1.983539842293816e-05, + "loss": 0.102, + "step": 5259 + }, + { + "epoch": 0.69, + "learning_rate": 1.9835302583933965e-05, + "loss": 0.1001, + "step": 5260 + }, + { + "epoch": 0.69, + "learning_rate": 1.983520671726851e-05, + "loss": 0.0931, + "step": 5261 + }, + { + "epoch": 0.69, + "learning_rate": 1.9835110822942053e-05, + "loss": 0.0905, + "step": 5262 + }, + { + "epoch": 0.69, + "learning_rate": 1.9835014900954876e-05, + "loss": 0.098, + "step": 5263 + }, + { + "epoch": 0.69, + "learning_rate": 1.9834918951307238e-05, + "loss": 0.0941, + "step": 5264 + }, + { + "epoch": 0.69, + "learning_rate": 1.9834822973999417e-05, + "loss": 0.0989, + "step": 5265 + }, + { + "epoch": 0.69, + "learning_rate": 1.9834726969031678e-05, + "loss": 0.0815, + "step": 5266 + }, + { + "epoch": 0.69, + "learning_rate": 1.983463093640429e-05, + "loss": 0.1008, + "step": 5267 + }, + { + "epoch": 0.69, + "learning_rate": 1.983453487611754e-05, + "loss": 0.0859, + "step": 5268 + }, + { + "epoch": 0.69, + "learning_rate": 1.983443878817167e-05, + "loss": 0.0891, + "step": 5269 + }, + { + "epoch": 0.69, + "learning_rate": 1.983434267256697e-05, + "loss": 0.0853, + "step": 5270 + }, + { + "epoch": 0.69, + "learning_rate": 1.9834246529303706e-05, + "loss": 0.0949, + "step": 5271 + }, + { + "epoch": 0.69, + "learning_rate": 1.9834150358382148e-05, + "loss": 0.1095, + "step": 5272 + }, + { + "epoch": 0.69, + "learning_rate": 1.9834054159802562e-05, + "loss": 0.0927, + "step": 5273 + }, + { + "epoch": 0.69, + "learning_rate": 1.9833957933565226e-05, + "loss": 0.0922, + "step": 5274 + }, + { + "epoch": 0.69, + "learning_rate": 1.9833861679670404e-05, + "loss": 0.0832, + "step": 5275 + }, + { + "epoch": 0.69, + "learning_rate": 1.983376539811837e-05, + "loss": 0.0873, + "step": 5276 + }, + { + "epoch": 0.69, + "learning_rate": 1.9833669088909397e-05, + "loss": 0.0974, + "step": 5277 + }, + { + "epoch": 0.69, + "learning_rate": 1.983357275204375e-05, + "loss": 0.0916, + "step": 5278 + }, + { + "epoch": 0.69, + "learning_rate": 1.9833476387521707e-05, + "loss": 0.0953, + "step": 5279 + }, + { + "epoch": 0.69, + "learning_rate": 1.9833379995343532e-05, + "loss": 0.0882, + "step": 5280 + }, + { + "epoch": 0.69, + "learning_rate": 1.98332835755095e-05, + "loss": 0.085, + "step": 5281 + }, + { + "epoch": 0.69, + "learning_rate": 1.9833187128019885e-05, + "loss": 0.0907, + "step": 5282 + }, + { + "epoch": 0.69, + "learning_rate": 1.983309065287495e-05, + "loss": 0.088, + "step": 5283 + }, + { + "epoch": 0.69, + "learning_rate": 1.9832994150074975e-05, + "loss": 0.0791, + "step": 5284 + }, + { + "epoch": 0.69, + "learning_rate": 1.9832897619620227e-05, + "loss": 0.0795, + "step": 5285 + }, + { + "epoch": 0.69, + "learning_rate": 1.9832801061510977e-05, + "loss": 0.0937, + "step": 5286 + }, + { + "epoch": 0.69, + "learning_rate": 1.9832704475747498e-05, + "loss": 0.0849, + "step": 5287 + }, + { + "epoch": 0.69, + "learning_rate": 1.983260786233006e-05, + "loss": 0.0913, + "step": 5288 + }, + { + "epoch": 0.69, + "learning_rate": 1.9832511221258938e-05, + "loss": 0.0964, + "step": 5289 + }, + { + "epoch": 0.69, + "learning_rate": 1.98324145525344e-05, + "loss": 0.0663, + "step": 5290 + }, + { + "epoch": 0.69, + "learning_rate": 1.983231785615672e-05, + "loss": 0.0931, + "step": 5291 + }, + { + "epoch": 0.69, + "learning_rate": 1.9832221132126173e-05, + "loss": 0.0851, + "step": 5292 + }, + { + "epoch": 0.69, + "learning_rate": 1.9832124380443024e-05, + "loss": 0.0867, + "step": 5293 + }, + { + "epoch": 0.69, + "learning_rate": 1.983202760110755e-05, + "loss": 0.1114, + "step": 5294 + }, + { + "epoch": 0.69, + "learning_rate": 1.983193079412002e-05, + "loss": 0.093, + "step": 5295 + }, + { + "epoch": 0.69, + "learning_rate": 1.983183395948071e-05, + "loss": 0.101, + "step": 5296 + }, + { + "epoch": 0.69, + "learning_rate": 1.983173709718989e-05, + "loss": 0.0881, + "step": 5297 + }, + { + "epoch": 0.69, + "learning_rate": 1.9831640207247835e-05, + "loss": 0.0847, + "step": 5298 + }, + { + "epoch": 0.69, + "learning_rate": 1.9831543289654814e-05, + "loss": 0.0905, + "step": 5299 + }, + { + "epoch": 0.69, + "learning_rate": 1.98314463444111e-05, + "loss": 0.0853, + "step": 5300 + }, + { + "epoch": 0.69, + "learning_rate": 1.9831349371516967e-05, + "loss": 0.075, + "step": 5301 + }, + { + "epoch": 0.69, + "learning_rate": 1.983125237097269e-05, + "loss": 0.0863, + "step": 5302 + }, + { + "epoch": 0.69, + "learning_rate": 1.9831155342778538e-05, + "loss": 0.0908, + "step": 5303 + }, + { + "epoch": 0.69, + "learning_rate": 1.9831058286934785e-05, + "loss": 0.0979, + "step": 5304 + }, + { + "epoch": 0.69, + "learning_rate": 1.9830961203441702e-05, + "loss": 0.0919, + "step": 5305 + }, + { + "epoch": 0.7, + "learning_rate": 1.9830864092299566e-05, + "loss": 0.0941, + "step": 5306 + }, + { + "epoch": 0.7, + "learning_rate": 1.9830766953508648e-05, + "loss": 0.0823, + "step": 5307 + }, + { + "epoch": 0.7, + "learning_rate": 1.983066978706922e-05, + "loss": 0.0776, + "step": 5308 + }, + { + "epoch": 0.7, + "learning_rate": 1.9830572592981562e-05, + "loss": 0.0938, + "step": 5309 + }, + { + "epoch": 0.7, + "learning_rate": 1.9830475371245936e-05, + "loss": 0.097, + "step": 5310 + }, + { + "epoch": 0.7, + "learning_rate": 1.9830378121862622e-05, + "loss": 0.1111, + "step": 5311 + }, + { + "epoch": 0.7, + "learning_rate": 1.9830280844831896e-05, + "loss": 0.0829, + "step": 5312 + }, + { + "epoch": 0.7, + "learning_rate": 1.9830183540154027e-05, + "loss": 0.0999, + "step": 5313 + }, + { + "epoch": 0.7, + "learning_rate": 1.983008620782929e-05, + "loss": 0.096, + "step": 5314 + }, + { + "epoch": 0.7, + "learning_rate": 1.9829988847857963e-05, + "loss": 0.0864, + "step": 5315 + }, + { + "epoch": 0.7, + "learning_rate": 1.9829891460240312e-05, + "loss": 0.0898, + "step": 5316 + }, + { + "epoch": 0.7, + "learning_rate": 1.9829794044976617e-05, + "loss": 0.0905, + "step": 5317 + }, + { + "epoch": 0.7, + "learning_rate": 1.9829696602067146e-05, + "loss": 0.0843, + "step": 5318 + }, + { + "epoch": 0.7, + "learning_rate": 1.982959913151218e-05, + "loss": 0.0906, + "step": 5319 + }, + { + "epoch": 0.7, + "learning_rate": 1.9829501633311986e-05, + "loss": 0.0927, + "step": 5320 + }, + { + "epoch": 0.7, + "learning_rate": 1.9829404107466846e-05, + "loss": 0.0965, + "step": 5321 + }, + { + "epoch": 0.7, + "learning_rate": 1.982930655397703e-05, + "loss": 0.0838, + "step": 5322 + }, + { + "epoch": 0.7, + "learning_rate": 1.9829208972842814e-05, + "loss": 0.0837, + "step": 5323 + }, + { + "epoch": 0.7, + "learning_rate": 1.9829111364064467e-05, + "loss": 0.1013, + "step": 5324 + }, + { + "epoch": 0.7, + "learning_rate": 1.982901372764227e-05, + "loss": 0.0825, + "step": 5325 + }, + { + "epoch": 0.7, + "learning_rate": 1.9828916063576497e-05, + "loss": 0.0982, + "step": 5326 + }, + { + "epoch": 0.7, + "learning_rate": 1.9828818371867418e-05, + "loss": 0.0875, + "step": 5327 + }, + { + "epoch": 0.7, + "learning_rate": 1.982872065251531e-05, + "loss": 0.1017, + "step": 5328 + }, + { + "epoch": 0.7, + "learning_rate": 1.9828622905520452e-05, + "loss": 0.0947, + "step": 5329 + }, + { + "epoch": 0.7, + "learning_rate": 1.9828525130883114e-05, + "loss": 0.0966, + "step": 5330 + }, + { + "epoch": 0.7, + "learning_rate": 1.982842732860357e-05, + "loss": 0.095, + "step": 5331 + }, + { + "epoch": 0.7, + "learning_rate": 1.9828329498682103e-05, + "loss": 0.0838, + "step": 5332 + }, + { + "epoch": 0.7, + "learning_rate": 1.982823164111898e-05, + "loss": 0.0808, + "step": 5333 + }, + { + "epoch": 0.7, + "learning_rate": 1.9828133755914477e-05, + "loss": 0.1003, + "step": 5334 + }, + { + "epoch": 0.7, + "learning_rate": 1.9828035843068877e-05, + "loss": 0.0929, + "step": 5335 + }, + { + "epoch": 0.7, + "learning_rate": 1.9827937902582444e-05, + "loss": 0.0977, + "step": 5336 + }, + { + "epoch": 0.7, + "learning_rate": 1.982783993445546e-05, + "loss": 0.0977, + "step": 5337 + }, + { + "epoch": 0.7, + "learning_rate": 1.98277419386882e-05, + "loss": 0.0904, + "step": 5338 + }, + { + "epoch": 0.7, + "learning_rate": 1.9827643915280937e-05, + "loss": 0.0817, + "step": 5339 + }, + { + "epoch": 0.7, + "learning_rate": 1.9827545864233953e-05, + "loss": 0.0892, + "step": 5340 + }, + { + "epoch": 0.7, + "learning_rate": 1.9827447785547518e-05, + "loss": 0.0812, + "step": 5341 + }, + { + "epoch": 0.7, + "learning_rate": 1.9827349679221907e-05, + "loss": 0.0977, + "step": 5342 + }, + { + "epoch": 0.7, + "learning_rate": 1.9827251545257404e-05, + "loss": 0.0933, + "step": 5343 + }, + { + "epoch": 0.7, + "learning_rate": 1.9827153383654273e-05, + "loss": 0.0876, + "step": 5344 + }, + { + "epoch": 0.7, + "learning_rate": 1.98270551944128e-05, + "loss": 0.0938, + "step": 5345 + }, + { + "epoch": 0.7, + "learning_rate": 1.9826956977533255e-05, + "loss": 0.0973, + "step": 5346 + }, + { + "epoch": 0.7, + "learning_rate": 1.9826858733015923e-05, + "loss": 0.0745, + "step": 5347 + }, + { + "epoch": 0.7, + "learning_rate": 1.9826760460861067e-05, + "loss": 0.0997, + "step": 5348 + }, + { + "epoch": 0.7, + "learning_rate": 1.9826662161068974e-05, + "loss": 0.0788, + "step": 5349 + }, + { + "epoch": 0.7, + "learning_rate": 1.9826563833639913e-05, + "loss": 0.0849, + "step": 5350 + }, + { + "epoch": 0.7, + "learning_rate": 1.982646547857417e-05, + "loss": 0.1002, + "step": 5351 + }, + { + "epoch": 0.7, + "learning_rate": 1.9826367095872014e-05, + "loss": 0.089, + "step": 5352 + }, + { + "epoch": 0.7, + "learning_rate": 1.982626868553372e-05, + "loss": 0.1007, + "step": 5353 + }, + { + "epoch": 0.7, + "learning_rate": 1.9826170247559576e-05, + "loss": 0.0913, + "step": 5354 + }, + { + "epoch": 0.7, + "learning_rate": 1.9826071781949846e-05, + "loss": 0.0937, + "step": 5355 + }, + { + "epoch": 0.7, + "learning_rate": 1.9825973288704813e-05, + "loss": 0.0869, + "step": 5356 + }, + { + "epoch": 0.7, + "learning_rate": 1.9825874767824757e-05, + "loss": 0.0902, + "step": 5357 + }, + { + "epoch": 0.7, + "learning_rate": 1.982577621930995e-05, + "loss": 0.1053, + "step": 5358 + }, + { + "epoch": 0.7, + "learning_rate": 1.9825677643160668e-05, + "loss": 0.0812, + "step": 5359 + }, + { + "epoch": 0.7, + "learning_rate": 1.9825579039377194e-05, + "loss": 0.092, + "step": 5360 + }, + { + "epoch": 0.7, + "learning_rate": 1.98254804079598e-05, + "loss": 0.0879, + "step": 5361 + }, + { + "epoch": 0.7, + "learning_rate": 1.9825381748908764e-05, + "loss": 0.0952, + "step": 5362 + }, + { + "epoch": 0.7, + "learning_rate": 1.9825283062224367e-05, + "loss": 0.0831, + "step": 5363 + }, + { + "epoch": 0.7, + "learning_rate": 1.9825184347906888e-05, + "loss": 0.092, + "step": 5364 + }, + { + "epoch": 0.7, + "learning_rate": 1.9825085605956596e-05, + "loss": 0.0852, + "step": 5365 + }, + { + "epoch": 0.7, + "learning_rate": 1.9824986836373775e-05, + "loss": 0.0881, + "step": 5366 + }, + { + "epoch": 0.7, + "learning_rate": 1.9824888039158704e-05, + "loss": 0.0877, + "step": 5367 + }, + { + "epoch": 0.7, + "learning_rate": 1.982478921431166e-05, + "loss": 0.082, + "step": 5368 + }, + { + "epoch": 0.7, + "learning_rate": 1.9824690361832918e-05, + "loss": 0.0885, + "step": 5369 + }, + { + "epoch": 0.7, + "learning_rate": 1.9824591481722754e-05, + "loss": 0.0916, + "step": 5370 + }, + { + "epoch": 0.7, + "learning_rate": 1.9824492573981453e-05, + "loss": 0.0823, + "step": 5371 + }, + { + "epoch": 0.7, + "learning_rate": 1.982439363860929e-05, + "loss": 0.0978, + "step": 5372 + }, + { + "epoch": 0.7, + "learning_rate": 1.9824294675606543e-05, + "loss": 0.0877, + "step": 5373 + }, + { + "epoch": 0.7, + "learning_rate": 1.9824195684973486e-05, + "loss": 0.0908, + "step": 5374 + }, + { + "epoch": 0.7, + "learning_rate": 1.9824096666710405e-05, + "loss": 0.1034, + "step": 5375 + }, + { + "epoch": 0.7, + "learning_rate": 1.9823997620817576e-05, + "loss": 0.0945, + "step": 5376 + }, + { + "epoch": 0.7, + "learning_rate": 1.9823898547295278e-05, + "loss": 0.0799, + "step": 5377 + }, + { + "epoch": 0.7, + "learning_rate": 1.9823799446143785e-05, + "loss": 0.1035, + "step": 5378 + }, + { + "epoch": 0.7, + "learning_rate": 1.9823700317363382e-05, + "loss": 0.0837, + "step": 5379 + }, + { + "epoch": 0.7, + "learning_rate": 1.9823601160954343e-05, + "loss": 0.0879, + "step": 5380 + }, + { + "epoch": 0.7, + "learning_rate": 1.982350197691695e-05, + "loss": 0.0972, + "step": 5381 + }, + { + "epoch": 0.71, + "learning_rate": 1.9823402765251482e-05, + "loss": 0.0992, + "step": 5382 + }, + { + "epoch": 0.71, + "learning_rate": 1.9823303525958216e-05, + "loss": 0.0899, + "step": 5383 + }, + { + "epoch": 0.71, + "learning_rate": 1.9823204259037432e-05, + "loss": 0.0902, + "step": 5384 + }, + { + "epoch": 0.71, + "learning_rate": 1.9823104964489408e-05, + "loss": 0.0915, + "step": 5385 + }, + { + "epoch": 0.71, + "learning_rate": 1.9823005642314425e-05, + "loss": 0.0862, + "step": 5386 + }, + { + "epoch": 0.71, + "learning_rate": 1.982290629251276e-05, + "loss": 0.0857, + "step": 5387 + }, + { + "epoch": 0.71, + "learning_rate": 1.9822806915084698e-05, + "loss": 0.0956, + "step": 5388 + }, + { + "epoch": 0.71, + "learning_rate": 1.9822707510030513e-05, + "loss": 0.0801, + "step": 5389 + }, + { + "epoch": 0.71, + "learning_rate": 1.9822608077350484e-05, + "loss": 0.0831, + "step": 5390 + }, + { + "epoch": 0.71, + "learning_rate": 1.98225086170449e-05, + "loss": 0.1038, + "step": 5391 + }, + { + "epoch": 0.71, + "learning_rate": 1.9822409129114025e-05, + "loss": 0.095, + "step": 5392 + }, + { + "epoch": 0.71, + "learning_rate": 1.982230961355815e-05, + "loss": 0.0895, + "step": 5393 + }, + { + "epoch": 0.71, + "learning_rate": 1.9822210070377554e-05, + "loss": 0.0938, + "step": 5394 + }, + { + "epoch": 0.71, + "learning_rate": 1.9822110499572513e-05, + "loss": 0.0896, + "step": 5395 + }, + { + "epoch": 0.71, + "learning_rate": 1.982201090114331e-05, + "loss": 0.081, + "step": 5396 + }, + { + "epoch": 0.71, + "learning_rate": 1.9821911275090227e-05, + "loss": 0.0901, + "step": 5397 + }, + { + "epoch": 0.71, + "learning_rate": 1.982181162141354e-05, + "loss": 0.086, + "step": 5398 + }, + { + "epoch": 0.71, + "learning_rate": 1.982171194011353e-05, + "loss": 0.0971, + "step": 5399 + }, + { + "epoch": 0.71, + "learning_rate": 1.982161223119048e-05, + "loss": 0.084, + "step": 5400 + }, + { + "epoch": 0.71, + "learning_rate": 1.9821512494644665e-05, + "loss": 0.0904, + "step": 5401 + }, + { + "epoch": 0.71, + "learning_rate": 1.982141273047637e-05, + "loss": 0.0979, + "step": 5402 + }, + { + "epoch": 0.71, + "learning_rate": 1.982131293868588e-05, + "loss": 0.0833, + "step": 5403 + }, + { + "epoch": 0.71, + "learning_rate": 1.9821213119273465e-05, + "loss": 0.0965, + "step": 5404 + }, + { + "epoch": 0.71, + "learning_rate": 1.9821113272239413e-05, + "loss": 0.0866, + "step": 5405 + }, + { + "epoch": 0.71, + "learning_rate": 1.9821013397584002e-05, + "loss": 0.0934, + "step": 5406 + }, + { + "epoch": 0.71, + "learning_rate": 1.982091349530751e-05, + "loss": 0.088, + "step": 5407 + }, + { + "epoch": 0.71, + "learning_rate": 1.982081356541023e-05, + "loss": 0.0831, + "step": 5408 + }, + { + "epoch": 0.71, + "learning_rate": 1.982071360789243e-05, + "loss": 0.0879, + "step": 5409 + }, + { + "epoch": 0.71, + "learning_rate": 1.9820613622754398e-05, + "loss": 0.0954, + "step": 5410 + }, + { + "epoch": 0.71, + "learning_rate": 1.982051360999641e-05, + "loss": 0.079, + "step": 5411 + }, + { + "epoch": 0.71, + "learning_rate": 1.9820413569618754e-05, + "loss": 0.0999, + "step": 5412 + }, + { + "epoch": 0.71, + "learning_rate": 1.9820313501621706e-05, + "loss": 0.095, + "step": 5413 + }, + { + "epoch": 0.71, + "learning_rate": 1.982021340600555e-05, + "loss": 0.0866, + "step": 5414 + }, + { + "epoch": 0.71, + "learning_rate": 1.9820113282770568e-05, + "loss": 0.0902, + "step": 5415 + }, + { + "epoch": 0.71, + "learning_rate": 1.982001313191704e-05, + "loss": 0.0828, + "step": 5416 + }, + { + "epoch": 0.71, + "learning_rate": 1.9819912953445245e-05, + "loss": 0.0922, + "step": 5417 + }, + { + "epoch": 0.71, + "learning_rate": 1.9819812747355468e-05, + "loss": 0.08, + "step": 5418 + }, + { + "epoch": 0.71, + "learning_rate": 1.981971251364799e-05, + "loss": 0.0855, + "step": 5419 + }, + { + "epoch": 0.71, + "learning_rate": 1.98196122523231e-05, + "loss": 0.0876, + "step": 5420 + }, + { + "epoch": 0.71, + "learning_rate": 1.981951196338107e-05, + "loss": 0.0866, + "step": 5421 + }, + { + "epoch": 0.71, + "learning_rate": 1.981941164682218e-05, + "loss": 0.0858, + "step": 5422 + }, + { + "epoch": 0.71, + "learning_rate": 1.9819311302646724e-05, + "loss": 0.0868, + "step": 5423 + }, + { + "epoch": 0.71, + "learning_rate": 1.9819210930854974e-05, + "loss": 0.0885, + "step": 5424 + }, + { + "epoch": 0.71, + "learning_rate": 1.981911053144722e-05, + "loss": 0.0926, + "step": 5425 + }, + { + "epoch": 0.71, + "learning_rate": 1.9819010104423738e-05, + "loss": 0.0883, + "step": 5426 + }, + { + "epoch": 0.71, + "learning_rate": 1.9818909649784814e-05, + "loss": 0.0897, + "step": 5427 + }, + { + "epoch": 0.71, + "learning_rate": 1.9818809167530732e-05, + "loss": 0.0943, + "step": 5428 + }, + { + "epoch": 0.71, + "learning_rate": 1.9818708657661767e-05, + "loss": 0.0951, + "step": 5429 + }, + { + "epoch": 0.71, + "learning_rate": 1.981860812017821e-05, + "loss": 0.0888, + "step": 5430 + }, + { + "epoch": 0.71, + "learning_rate": 1.9818507555080337e-05, + "loss": 0.088, + "step": 5431 + }, + { + "epoch": 0.71, + "learning_rate": 1.981840696236844e-05, + "loss": 0.0919, + "step": 5432 + }, + { + "epoch": 0.71, + "learning_rate": 1.981830634204279e-05, + "loss": 0.0899, + "step": 5433 + }, + { + "epoch": 0.71, + "learning_rate": 1.981820569410368e-05, + "loss": 0.0999, + "step": 5434 + }, + { + "epoch": 0.71, + "learning_rate": 1.981810501855139e-05, + "loss": 0.0759, + "step": 5435 + }, + { + "epoch": 0.71, + "learning_rate": 1.9818004315386202e-05, + "loss": 0.0855, + "step": 5436 + }, + { + "epoch": 0.71, + "learning_rate": 1.9817903584608398e-05, + "loss": 0.0951, + "step": 5437 + }, + { + "epoch": 0.71, + "learning_rate": 1.9817802826218265e-05, + "loss": 0.0997, + "step": 5438 + }, + { + "epoch": 0.71, + "learning_rate": 1.981770204021608e-05, + "loss": 0.084, + "step": 5439 + }, + { + "epoch": 0.71, + "learning_rate": 1.9817601226602138e-05, + "loss": 0.1014, + "step": 5440 + }, + { + "epoch": 0.71, + "learning_rate": 1.981750038537671e-05, + "loss": 0.0818, + "step": 5441 + }, + { + "epoch": 0.71, + "learning_rate": 1.9817399516540087e-05, + "loss": 0.088, + "step": 5442 + }, + { + "epoch": 0.71, + "learning_rate": 1.9817298620092548e-05, + "loss": 0.094, + "step": 5443 + }, + { + "epoch": 0.71, + "learning_rate": 1.9817197696034385e-05, + "loss": 0.0943, + "step": 5444 + }, + { + "epoch": 0.71, + "learning_rate": 1.9817096744365872e-05, + "loss": 0.0865, + "step": 5445 + }, + { + "epoch": 0.71, + "learning_rate": 1.9816995765087297e-05, + "loss": 0.096, + "step": 5446 + }, + { + "epoch": 0.71, + "learning_rate": 1.9816894758198944e-05, + "loss": 0.0841, + "step": 5447 + }, + { + "epoch": 0.71, + "learning_rate": 1.9816793723701102e-05, + "loss": 0.1023, + "step": 5448 + }, + { + "epoch": 0.71, + "learning_rate": 1.9816692661594044e-05, + "loss": 0.095, + "step": 5449 + }, + { + "epoch": 0.71, + "learning_rate": 1.9816591571878067e-05, + "loss": 0.1076, + "step": 5450 + }, + { + "epoch": 0.71, + "learning_rate": 1.9816490454553446e-05, + "loss": 0.0848, + "step": 5451 + }, + { + "epoch": 0.71, + "learning_rate": 1.9816389309620465e-05, + "loss": 0.097, + "step": 5452 + }, + { + "epoch": 0.71, + "learning_rate": 1.9816288137079417e-05, + "loss": 0.0717, + "step": 5453 + }, + { + "epoch": 0.71, + "learning_rate": 1.981618693693058e-05, + "loss": 0.0942, + "step": 5454 + }, + { + "epoch": 0.71, + "learning_rate": 1.9816085709174236e-05, + "loss": 0.08, + "step": 5455 + }, + { + "epoch": 0.71, + "learning_rate": 1.981598445381068e-05, + "loss": 0.0856, + "step": 5456 + }, + { + "epoch": 0.71, + "learning_rate": 1.9815883170840187e-05, + "loss": 0.087, + "step": 5457 + }, + { + "epoch": 0.71, + "learning_rate": 1.9815781860263045e-05, + "loss": 0.0874, + "step": 5458 + }, + { + "epoch": 0.72, + "learning_rate": 1.9815680522079537e-05, + "loss": 0.0801, + "step": 5459 + }, + { + "epoch": 0.72, + "learning_rate": 1.9815579156289955e-05, + "loss": 0.0911, + "step": 5460 + }, + { + "epoch": 0.72, + "learning_rate": 1.9815477762894577e-05, + "loss": 0.1004, + "step": 5461 + }, + { + "epoch": 0.72, + "learning_rate": 1.9815376341893693e-05, + "loss": 0.084, + "step": 5462 + }, + { + "epoch": 0.72, + "learning_rate": 1.9815274893287582e-05, + "loss": 0.0925, + "step": 5463 + }, + { + "epoch": 0.72, + "learning_rate": 1.9815173417076535e-05, + "loss": 0.0989, + "step": 5464 + }, + { + "epoch": 0.72, + "learning_rate": 1.9815071913260837e-05, + "loss": 0.0792, + "step": 5465 + }, + { + "epoch": 0.72, + "learning_rate": 1.981497038184077e-05, + "loss": 0.1008, + "step": 5466 + }, + { + "epoch": 0.72, + "learning_rate": 1.981486882281662e-05, + "loss": 0.1055, + "step": 5467 + }, + { + "epoch": 0.72, + "learning_rate": 1.9814767236188678e-05, + "loss": 0.091, + "step": 5468 + }, + { + "epoch": 0.72, + "learning_rate": 1.9814665621957224e-05, + "loss": 0.0882, + "step": 5469 + }, + { + "epoch": 0.72, + "learning_rate": 1.9814563980122542e-05, + "loss": 0.0861, + "step": 5470 + }, + { + "epoch": 0.72, + "learning_rate": 1.9814462310684927e-05, + "loss": 0.091, + "step": 5471 + }, + { + "epoch": 0.72, + "learning_rate": 1.9814360613644655e-05, + "loss": 0.0889, + "step": 5472 + }, + { + "epoch": 0.72, + "learning_rate": 1.981425888900202e-05, + "loss": 0.0916, + "step": 5473 + }, + { + "epoch": 0.72, + "learning_rate": 1.9814157136757305e-05, + "loss": 0.0974, + "step": 5474 + }, + { + "epoch": 0.72, + "learning_rate": 1.981405535691079e-05, + "loss": 0.0977, + "step": 5475 + }, + { + "epoch": 0.72, + "learning_rate": 1.981395354946277e-05, + "loss": 0.084, + "step": 5476 + }, + { + "epoch": 0.72, + "learning_rate": 1.981385171441353e-05, + "loss": 0.088, + "step": 5477 + }, + { + "epoch": 0.72, + "learning_rate": 1.9813749851763356e-05, + "loss": 0.0828, + "step": 5478 + }, + { + "epoch": 0.72, + "learning_rate": 1.9813647961512528e-05, + "loss": 0.085, + "step": 5479 + }, + { + "epoch": 0.72, + "learning_rate": 1.981354604366134e-05, + "loss": 0.0895, + "step": 5480 + }, + { + "epoch": 0.72, + "learning_rate": 1.9813444098210078e-05, + "loss": 0.0946, + "step": 5481 + }, + { + "epoch": 0.72, + "learning_rate": 1.9813342125159025e-05, + "loss": 0.0887, + "step": 5482 + }, + { + "epoch": 0.72, + "learning_rate": 1.981324012450847e-05, + "loss": 0.0846, + "step": 5483 + }, + { + "epoch": 0.72, + "learning_rate": 1.98131380962587e-05, + "loss": 0.0995, + "step": 5484 + }, + { + "epoch": 0.72, + "learning_rate": 1.9813036040410004e-05, + "loss": 0.0921, + "step": 5485 + }, + { + "epoch": 0.72, + "learning_rate": 1.9812933956962667e-05, + "loss": 0.083, + "step": 5486 + }, + { + "epoch": 0.72, + "learning_rate": 1.981283184591697e-05, + "loss": 0.0945, + "step": 5487 + }, + { + "epoch": 0.72, + "learning_rate": 1.981272970727321e-05, + "loss": 0.091, + "step": 5488 + }, + { + "epoch": 0.72, + "learning_rate": 1.981262754103167e-05, + "loss": 0.1092, + "step": 5489 + }, + { + "epoch": 0.72, + "learning_rate": 1.9812525347192638e-05, + "loss": 0.0754, + "step": 5490 + }, + { + "epoch": 0.72, + "learning_rate": 1.9812423125756403e-05, + "loss": 0.0822, + "step": 5491 + }, + { + "epoch": 0.72, + "learning_rate": 1.9812320876723246e-05, + "loss": 0.0951, + "step": 5492 + }, + { + "epoch": 0.72, + "learning_rate": 1.981221860009346e-05, + "loss": 0.079, + "step": 5493 + }, + { + "epoch": 0.72, + "learning_rate": 1.981211629586733e-05, + "loss": 0.0965, + "step": 5494 + }, + { + "epoch": 0.72, + "learning_rate": 1.981201396404515e-05, + "loss": 0.099, + "step": 5495 + }, + { + "epoch": 0.72, + "learning_rate": 1.9811911604627204e-05, + "loss": 0.0775, + "step": 5496 + }, + { + "epoch": 0.72, + "learning_rate": 1.9811809217613776e-05, + "loss": 0.0901, + "step": 5497 + }, + { + "epoch": 0.72, + "learning_rate": 1.9811706803005156e-05, + "loss": 0.0877, + "step": 5498 + }, + { + "epoch": 0.72, + "learning_rate": 1.981160436080164e-05, + "loss": 0.0795, + "step": 5499 + }, + { + "epoch": 0.72, + "learning_rate": 1.98115018910035e-05, + "loss": 0.0888, + "step": 5500 + }, + { + "epoch": 0.72, + "learning_rate": 1.9811399393611036e-05, + "loss": 0.0956, + "step": 5501 + }, + { + "epoch": 0.72, + "learning_rate": 1.9811296868624536e-05, + "loss": 0.0853, + "step": 5502 + }, + { + "epoch": 0.72, + "learning_rate": 1.9811194316044284e-05, + "loss": 0.0785, + "step": 5503 + }, + { + "epoch": 0.72, + "learning_rate": 1.9811091735870572e-05, + "loss": 0.095, + "step": 5504 + }, + { + "epoch": 0.72, + "learning_rate": 1.9810989128103688e-05, + "loss": 0.0937, + "step": 5505 + }, + { + "epoch": 0.72, + "learning_rate": 1.981088649274392e-05, + "loss": 0.0998, + "step": 5506 + }, + { + "epoch": 0.72, + "learning_rate": 1.9810783829791555e-05, + "loss": 0.0905, + "step": 5507 + }, + { + "epoch": 0.72, + "learning_rate": 1.981068113924688e-05, + "loss": 0.093, + "step": 5508 + }, + { + "epoch": 0.72, + "learning_rate": 1.9810578421110188e-05, + "loss": 0.0918, + "step": 5509 + }, + { + "epoch": 0.72, + "learning_rate": 1.981047567538177e-05, + "loss": 0.098, + "step": 5510 + }, + { + "epoch": 0.72, + "learning_rate": 1.9810372902061906e-05, + "loss": 0.0932, + "step": 5511 + }, + { + "epoch": 0.72, + "learning_rate": 1.9810270101150896e-05, + "loss": 0.1027, + "step": 5512 + }, + { + "epoch": 0.72, + "learning_rate": 1.981016727264902e-05, + "loss": 0.0972, + "step": 5513 + }, + { + "epoch": 0.72, + "learning_rate": 1.9810064416556573e-05, + "loss": 0.0859, + "step": 5514 + }, + { + "epoch": 0.72, + "learning_rate": 1.9809961532873843e-05, + "loss": 0.0913, + "step": 5515 + }, + { + "epoch": 0.72, + "learning_rate": 1.980985862160112e-05, + "loss": 0.0964, + "step": 5516 + }, + { + "epoch": 0.72, + "learning_rate": 1.9809755682738688e-05, + "loss": 0.0839, + "step": 5517 + }, + { + "epoch": 0.72, + "learning_rate": 1.980965271628684e-05, + "loss": 0.0886, + "step": 5518 + }, + { + "epoch": 0.72, + "learning_rate": 1.980954972224587e-05, + "loss": 0.0987, + "step": 5519 + }, + { + "epoch": 0.72, + "learning_rate": 1.980944670061606e-05, + "loss": 0.0895, + "step": 5520 + }, + { + "epoch": 0.72, + "learning_rate": 1.9809343651397706e-05, + "loss": 0.0879, + "step": 5521 + }, + { + "epoch": 0.72, + "learning_rate": 1.9809240574591094e-05, + "loss": 0.0861, + "step": 5522 + }, + { + "epoch": 0.72, + "learning_rate": 1.9809137470196518e-05, + "loss": 0.0995, + "step": 5523 + }, + { + "epoch": 0.72, + "learning_rate": 1.9809034338214265e-05, + "loss": 0.1018, + "step": 5524 + }, + { + "epoch": 0.72, + "learning_rate": 1.9808931178644623e-05, + "loss": 0.0834, + "step": 5525 + }, + { + "epoch": 0.72, + "learning_rate": 1.9808827991487884e-05, + "loss": 0.0755, + "step": 5526 + }, + { + "epoch": 0.72, + "learning_rate": 1.980872477674434e-05, + "loss": 0.0965, + "step": 5527 + }, + { + "epoch": 0.72, + "learning_rate": 1.980862153441428e-05, + "loss": 0.0869, + "step": 5528 + }, + { + "epoch": 0.72, + "learning_rate": 1.9808518264497992e-05, + "loss": 0.1114, + "step": 5529 + }, + { + "epoch": 0.72, + "learning_rate": 1.980841496699577e-05, + "loss": 0.0758, + "step": 5530 + }, + { + "epoch": 0.72, + "learning_rate": 1.9808311641907908e-05, + "loss": 0.0937, + "step": 5531 + }, + { + "epoch": 0.72, + "learning_rate": 1.9808208289234683e-05, + "loss": 0.0741, + "step": 5532 + }, + { + "epoch": 0.72, + "learning_rate": 1.98081049089764e-05, + "loss": 0.0909, + "step": 5533 + }, + { + "epoch": 0.72, + "learning_rate": 1.9808001501133345e-05, + "loss": 0.0767, + "step": 5534 + }, + { + "epoch": 0.73, + "learning_rate": 1.9807898065705808e-05, + "loss": 0.0866, + "step": 5535 + }, + { + "epoch": 0.73, + "learning_rate": 1.9807794602694077e-05, + "loss": 0.0927, + "step": 5536 + }, + { + "epoch": 0.73, + "learning_rate": 1.9807691112098446e-05, + "loss": 0.0893, + "step": 5537 + }, + { + "epoch": 0.73, + "learning_rate": 1.980758759391921e-05, + "loss": 0.0807, + "step": 5538 + }, + { + "epoch": 0.73, + "learning_rate": 1.9807484048156655e-05, + "loss": 0.0857, + "step": 5539 + }, + { + "epoch": 0.73, + "learning_rate": 1.980738047481107e-05, + "loss": 0.0736, + "step": 5540 + }, + { + "epoch": 0.73, + "learning_rate": 1.9807276873882754e-05, + "loss": 0.0841, + "step": 5541 + }, + { + "epoch": 0.73, + "learning_rate": 1.980717324537199e-05, + "loss": 0.0944, + "step": 5542 + }, + { + "epoch": 0.73, + "learning_rate": 1.9807069589279074e-05, + "loss": 0.094, + "step": 5543 + }, + { + "epoch": 0.73, + "learning_rate": 1.98069659056043e-05, + "loss": 0.0873, + "step": 5544 + }, + { + "epoch": 0.73, + "learning_rate": 1.9806862194347953e-05, + "loss": 0.0919, + "step": 5545 + }, + { + "epoch": 0.73, + "learning_rate": 1.980675845551033e-05, + "loss": 0.1036, + "step": 5546 + }, + { + "epoch": 0.73, + "learning_rate": 1.9806654689091723e-05, + "loss": 0.0871, + "step": 5547 + }, + { + "epoch": 0.73, + "learning_rate": 1.980655089509242e-05, + "loss": 0.0874, + "step": 5548 + }, + { + "epoch": 0.73, + "learning_rate": 1.9806447073512715e-05, + "loss": 0.087, + "step": 5549 + }, + { + "epoch": 0.73, + "learning_rate": 1.98063432243529e-05, + "loss": 0.0863, + "step": 5550 + }, + { + "epoch": 0.73, + "learning_rate": 1.980623934761327e-05, + "loss": 0.0893, + "step": 5551 + }, + { + "epoch": 0.73, + "learning_rate": 1.980613544329411e-05, + "loss": 0.0891, + "step": 5552 + }, + { + "epoch": 0.73, + "learning_rate": 1.9806031511395716e-05, + "loss": 0.09, + "step": 5553 + }, + { + "epoch": 0.73, + "learning_rate": 1.980592755191838e-05, + "loss": 0.0995, + "step": 5554 + }, + { + "epoch": 0.73, + "learning_rate": 1.98058235648624e-05, + "loss": 0.0849, + "step": 5555 + }, + { + "epoch": 0.73, + "learning_rate": 1.9805719550228057e-05, + "loss": 0.0917, + "step": 5556 + }, + { + "epoch": 0.73, + "learning_rate": 1.9805615508015654e-05, + "loss": 0.0838, + "step": 5557 + }, + { + "epoch": 0.73, + "learning_rate": 1.980551143822548e-05, + "loss": 0.094, + "step": 5558 + }, + { + "epoch": 0.73, + "learning_rate": 1.9805407340857826e-05, + "loss": 0.0801, + "step": 5559 + }, + { + "epoch": 0.73, + "learning_rate": 1.9805303215912984e-05, + "loss": 0.0894, + "step": 5560 + }, + { + "epoch": 0.73, + "learning_rate": 1.9805199063391253e-05, + "loss": 0.0869, + "step": 5561 + }, + { + "epoch": 0.73, + "learning_rate": 1.9805094883292918e-05, + "loss": 0.0835, + "step": 5562 + }, + { + "epoch": 0.73, + "learning_rate": 1.980499067561828e-05, + "loss": 0.0849, + "step": 5563 + }, + { + "epoch": 0.73, + "learning_rate": 1.9804886440367624e-05, + "loss": 0.0801, + "step": 5564 + }, + { + "epoch": 0.73, + "learning_rate": 1.9804782177541248e-05, + "loss": 0.0822, + "step": 5565 + }, + { + "epoch": 0.73, + "learning_rate": 1.9804677887139446e-05, + "loss": 0.091, + "step": 5566 + }, + { + "epoch": 0.73, + "learning_rate": 1.9804573569162508e-05, + "loss": 0.0938, + "step": 5567 + }, + { + "epoch": 0.73, + "learning_rate": 1.9804469223610727e-05, + "loss": 0.1051, + "step": 5568 + }, + { + "epoch": 0.73, + "learning_rate": 1.98043648504844e-05, + "loss": 0.0863, + "step": 5569 + }, + { + "epoch": 0.73, + "learning_rate": 1.9804260449783823e-05, + "loss": 0.0712, + "step": 5570 + }, + { + "epoch": 0.73, + "learning_rate": 1.980415602150928e-05, + "loss": 0.0964, + "step": 5571 + }, + { + "epoch": 0.73, + "learning_rate": 1.9804051565661072e-05, + "loss": 0.092, + "step": 5572 + }, + { + "epoch": 0.73, + "learning_rate": 1.9803947082239493e-05, + "loss": 0.0828, + "step": 5573 + }, + { + "epoch": 0.73, + "learning_rate": 1.980384257124483e-05, + "loss": 0.1143, + "step": 5574 + }, + { + "epoch": 0.73, + "learning_rate": 1.9803738032677388e-05, + "loss": 0.0997, + "step": 5575 + }, + { + "epoch": 0.73, + "learning_rate": 1.980363346653745e-05, + "loss": 0.1053, + "step": 5576 + }, + { + "epoch": 0.73, + "learning_rate": 1.980352887282532e-05, + "loss": 0.0851, + "step": 5577 + }, + { + "epoch": 0.73, + "learning_rate": 1.9803424251541284e-05, + "loss": 0.0789, + "step": 5578 + }, + { + "epoch": 0.73, + "learning_rate": 1.980331960268564e-05, + "loss": 0.101, + "step": 5579 + }, + { + "epoch": 0.73, + "learning_rate": 1.980321492625868e-05, + "loss": 0.0899, + "step": 5580 + }, + { + "epoch": 0.73, + "learning_rate": 1.98031102222607e-05, + "loss": 0.0986, + "step": 5581 + }, + { + "epoch": 0.73, + "learning_rate": 1.9803005490691997e-05, + "loss": 0.087, + "step": 5582 + }, + { + "epoch": 0.73, + "learning_rate": 1.9802900731552863e-05, + "loss": 0.0928, + "step": 5583 + }, + { + "epoch": 0.73, + "learning_rate": 1.980279594484359e-05, + "loss": 0.0839, + "step": 5584 + }, + { + "epoch": 0.73, + "learning_rate": 1.980269113056448e-05, + "loss": 0.0862, + "step": 5585 + }, + { + "epoch": 0.73, + "learning_rate": 1.9802586288715816e-05, + "loss": 0.0922, + "step": 5586 + }, + { + "epoch": 0.73, + "learning_rate": 1.9802481419297908e-05, + "loss": 0.0933, + "step": 5587 + }, + { + "epoch": 0.73, + "learning_rate": 1.980237652231104e-05, + "loss": 0.0867, + "step": 5588 + }, + { + "epoch": 0.73, + "learning_rate": 1.9802271597755507e-05, + "loss": 0.0973, + "step": 5589 + }, + { + "epoch": 0.73, + "learning_rate": 1.980216664563161e-05, + "loss": 0.0916, + "step": 5590 + }, + { + "epoch": 0.73, + "learning_rate": 1.980206166593964e-05, + "loss": 0.0943, + "step": 5591 + }, + { + "epoch": 0.73, + "learning_rate": 1.9801956658679895e-05, + "loss": 0.0886, + "step": 5592 + }, + { + "epoch": 0.73, + "learning_rate": 1.9801851623852668e-05, + "loss": 0.0931, + "step": 5593 + }, + { + "epoch": 0.73, + "learning_rate": 1.980174656145826e-05, + "loss": 0.0894, + "step": 5594 + }, + { + "epoch": 0.73, + "learning_rate": 1.9801641471496952e-05, + "loss": 0.0977, + "step": 5595 + }, + { + "epoch": 0.73, + "learning_rate": 1.9801536353969057e-05, + "loss": 0.0875, + "step": 5596 + }, + { + "epoch": 0.73, + "learning_rate": 1.980143120887486e-05, + "loss": 0.0873, + "step": 5597 + }, + { + "epoch": 0.73, + "learning_rate": 1.980132603621466e-05, + "loss": 0.0946, + "step": 5598 + }, + { + "epoch": 0.73, + "learning_rate": 1.9801220835988753e-05, + "loss": 0.088, + "step": 5599 + }, + { + "epoch": 0.73, + "learning_rate": 1.9801115608197436e-05, + "loss": 0.0918, + "step": 5600 + }, + { + "epoch": 0.73, + "learning_rate": 1.9801010352841e-05, + "loss": 0.0953, + "step": 5601 + }, + { + "epoch": 0.73, + "learning_rate": 1.9800905069919743e-05, + "loss": 0.0983, + "step": 5602 + }, + { + "epoch": 0.73, + "learning_rate": 1.9800799759433968e-05, + "loss": 0.0804, + "step": 5603 + }, + { + "epoch": 0.73, + "learning_rate": 1.9800694421383962e-05, + "loss": 0.0767, + "step": 5604 + }, + { + "epoch": 0.73, + "learning_rate": 1.9800589055770026e-05, + "loss": 0.1072, + "step": 5605 + }, + { + "epoch": 0.73, + "learning_rate": 1.9800483662592453e-05, + "loss": 0.0839, + "step": 5606 + }, + { + "epoch": 0.73, + "learning_rate": 1.9800378241851546e-05, + "loss": 0.0813, + "step": 5607 + }, + { + "epoch": 0.73, + "learning_rate": 1.980027279354759e-05, + "loss": 0.0989, + "step": 5608 + }, + { + "epoch": 0.73, + "learning_rate": 1.98001673176809e-05, + "loss": 0.0858, + "step": 5609 + }, + { + "epoch": 0.73, + "learning_rate": 1.9800061814251753e-05, + "loss": 0.0958, + "step": 5610 + }, + { + "epoch": 0.74, + "learning_rate": 1.9799956283260452e-05, + "loss": 0.093, + "step": 5611 + }, + { + "epoch": 0.74, + "learning_rate": 1.97998507247073e-05, + "loss": 0.0837, + "step": 5612 + }, + { + "epoch": 0.74, + "learning_rate": 1.979974513859259e-05, + "loss": 0.0832, + "step": 5613 + }, + { + "epoch": 0.74, + "learning_rate": 1.979963952491662e-05, + "loss": 0.0879, + "step": 5614 + }, + { + "epoch": 0.74, + "learning_rate": 1.979953388367968e-05, + "loss": 0.0832, + "step": 5615 + }, + { + "epoch": 0.74, + "learning_rate": 1.9799428214882076e-05, + "loss": 0.1008, + "step": 5616 + }, + { + "epoch": 0.74, + "learning_rate": 1.9799322518524106e-05, + "loss": 0.093, + "step": 5617 + }, + { + "epoch": 0.74, + "learning_rate": 1.979921679460606e-05, + "loss": 0.0932, + "step": 5618 + }, + { + "epoch": 0.74, + "learning_rate": 1.9799111043128236e-05, + "loss": 0.0891, + "step": 5619 + }, + { + "epoch": 0.74, + "learning_rate": 1.9799005264090936e-05, + "loss": 0.0897, + "step": 5620 + }, + { + "epoch": 0.74, + "learning_rate": 1.9798899457494457e-05, + "loss": 0.0859, + "step": 5621 + }, + { + "epoch": 0.74, + "learning_rate": 1.9798793623339093e-05, + "loss": 0.0893, + "step": 5622 + }, + { + "epoch": 0.74, + "learning_rate": 1.9798687761625147e-05, + "loss": 0.079, + "step": 5623 + }, + { + "epoch": 0.74, + "learning_rate": 1.979858187235291e-05, + "loss": 0.0922, + "step": 5624 + }, + { + "epoch": 0.74, + "learning_rate": 1.9798475955522684e-05, + "loss": 0.0902, + "step": 5625 + }, + { + "epoch": 0.74, + "learning_rate": 1.9798370011134764e-05, + "loss": 0.0921, + "step": 5626 + }, + { + "epoch": 0.74, + "learning_rate": 1.9798264039189455e-05, + "loss": 0.083, + "step": 5627 + }, + { + "epoch": 0.74, + "learning_rate": 1.9798158039687046e-05, + "loss": 0.0909, + "step": 5628 + }, + { + "epoch": 0.74, + "learning_rate": 1.9798052012627843e-05, + "loss": 0.0964, + "step": 5629 + }, + { + "epoch": 0.74, + "learning_rate": 1.9797945958012138e-05, + "loss": 0.0943, + "step": 5630 + }, + { + "epoch": 0.74, + "learning_rate": 1.979783987584023e-05, + "loss": 0.0898, + "step": 5631 + }, + { + "epoch": 0.74, + "learning_rate": 1.9797733766112423e-05, + "loss": 0.0941, + "step": 5632 + }, + { + "epoch": 0.74, + "learning_rate": 1.979762762882901e-05, + "loss": 0.0909, + "step": 5633 + }, + { + "epoch": 0.74, + "learning_rate": 1.9797521463990292e-05, + "loss": 0.089, + "step": 5634 + }, + { + "epoch": 0.74, + "learning_rate": 1.9797415271596567e-05, + "loss": 0.0925, + "step": 5635 + }, + { + "epoch": 0.74, + "learning_rate": 1.979730905164813e-05, + "loss": 0.0995, + "step": 5636 + }, + { + "epoch": 0.74, + "learning_rate": 1.9797202804145287e-05, + "loss": 0.0916, + "step": 5637 + }, + { + "epoch": 0.74, + "learning_rate": 1.979709652908833e-05, + "loss": 0.0956, + "step": 5638 + }, + { + "epoch": 0.74, + "learning_rate": 1.9796990226477558e-05, + "loss": 0.0903, + "step": 5639 + }, + { + "epoch": 0.74, + "learning_rate": 1.9796883896313277e-05, + "loss": 0.1027, + "step": 5640 + }, + { + "epoch": 0.74, + "learning_rate": 1.9796777538595783e-05, + "loss": 0.09, + "step": 5641 + }, + { + "epoch": 0.74, + "learning_rate": 1.979667115332537e-05, + "loss": 0.0887, + "step": 5642 + }, + { + "epoch": 0.74, + "learning_rate": 1.9796564740502343e-05, + "loss": 0.0936, + "step": 5643 + }, + { + "epoch": 0.74, + "learning_rate": 1.9796458300127003e-05, + "loss": 0.0836, + "step": 5644 + }, + { + "epoch": 0.74, + "learning_rate": 1.979635183219964e-05, + "loss": 0.0897, + "step": 5645 + }, + { + "epoch": 0.74, + "learning_rate": 1.9796245336720562e-05, + "loss": 0.0802, + "step": 5646 + }, + { + "epoch": 0.74, + "learning_rate": 1.9796138813690064e-05, + "loss": 0.086, + "step": 5647 + }, + { + "epoch": 0.74, + "learning_rate": 1.979603226310845e-05, + "loss": 0.0828, + "step": 5648 + }, + { + "epoch": 0.74, + "learning_rate": 1.9795925684976015e-05, + "loss": 0.0877, + "step": 5649 + }, + { + "epoch": 0.74, + "learning_rate": 1.9795819079293058e-05, + "loss": 0.0885, + "step": 5650 + }, + { + "epoch": 0.74, + "learning_rate": 1.9795712446059883e-05, + "loss": 0.086, + "step": 5651 + }, + { + "epoch": 0.74, + "learning_rate": 1.9795605785276793e-05, + "loss": 0.0908, + "step": 5652 + }, + { + "epoch": 0.74, + "learning_rate": 1.979549909694408e-05, + "loss": 0.0752, + "step": 5653 + }, + { + "epoch": 0.74, + "learning_rate": 1.9795392381062046e-05, + "loss": 0.0961, + "step": 5654 + }, + { + "epoch": 0.74, + "learning_rate": 1.9795285637630996e-05, + "loss": 0.0942, + "step": 5655 + }, + { + "epoch": 0.74, + "learning_rate": 1.9795178866651223e-05, + "loss": 0.0997, + "step": 5656 + }, + { + "epoch": 0.74, + "learning_rate": 1.9795072068123034e-05, + "loss": 0.0863, + "step": 5657 + }, + { + "epoch": 0.74, + "learning_rate": 1.9794965242046726e-05, + "loss": 0.0876, + "step": 5658 + }, + { + "epoch": 0.74, + "learning_rate": 1.97948583884226e-05, + "loss": 0.0929, + "step": 5659 + }, + { + "epoch": 0.74, + "learning_rate": 1.9794751507250954e-05, + "loss": 0.0947, + "step": 5660 + }, + { + "epoch": 0.74, + "learning_rate": 1.979464459853209e-05, + "loss": 0.1012, + "step": 5661 + }, + { + "epoch": 0.74, + "learning_rate": 1.9794537662266316e-05, + "loss": 0.1021, + "step": 5662 + }, + { + "epoch": 0.74, + "learning_rate": 1.9794430698453922e-05, + "loss": 0.0922, + "step": 5663 + }, + { + "epoch": 0.74, + "learning_rate": 1.9794323707095214e-05, + "loss": 0.1021, + "step": 5664 + }, + { + "epoch": 0.74, + "learning_rate": 1.979421668819049e-05, + "loss": 0.0926, + "step": 5665 + }, + { + "epoch": 0.74, + "learning_rate": 1.9794109641740057e-05, + "loss": 0.109, + "step": 5666 + }, + { + "epoch": 0.74, + "learning_rate": 1.9794002567744212e-05, + "loss": 0.0967, + "step": 5667 + }, + { + "epoch": 0.74, + "learning_rate": 1.979389546620325e-05, + "loss": 0.0905, + "step": 5668 + }, + { + "epoch": 0.74, + "learning_rate": 1.9793788337117484e-05, + "loss": 0.0878, + "step": 5669 + }, + { + "epoch": 0.74, + "learning_rate": 1.979368118048721e-05, + "loss": 0.081, + "step": 5670 + }, + { + "epoch": 0.74, + "learning_rate": 1.979357399631273e-05, + "loss": 0.0919, + "step": 5671 + }, + { + "epoch": 0.74, + "learning_rate": 1.979346678459434e-05, + "loss": 0.0924, + "step": 5672 + }, + { + "epoch": 0.74, + "learning_rate": 1.979335954533235e-05, + "loss": 0.0866, + "step": 5673 + }, + { + "epoch": 0.74, + "learning_rate": 1.979325227852706e-05, + "loss": 0.0839, + "step": 5674 + }, + { + "epoch": 0.74, + "learning_rate": 1.9793144984178765e-05, + "loss": 0.0715, + "step": 5675 + }, + { + "epoch": 0.74, + "learning_rate": 1.9793037662287773e-05, + "loss": 0.0895, + "step": 5676 + }, + { + "epoch": 0.74, + "learning_rate": 1.9792930312854384e-05, + "loss": 0.0841, + "step": 5677 + }, + { + "epoch": 0.74, + "learning_rate": 1.9792822935878902e-05, + "loss": 0.0918, + "step": 5678 + }, + { + "epoch": 0.74, + "learning_rate": 1.9792715531361622e-05, + "loss": 0.1005, + "step": 5679 + }, + { + "epoch": 0.74, + "learning_rate": 1.9792608099302856e-05, + "loss": 0.0942, + "step": 5680 + }, + { + "epoch": 0.74, + "learning_rate": 1.9792500639702898e-05, + "loss": 0.0874, + "step": 5681 + }, + { + "epoch": 0.74, + "learning_rate": 1.9792393152562056e-05, + "loss": 0.0865, + "step": 5682 + }, + { + "epoch": 0.74, + "learning_rate": 1.9792285637880627e-05, + "loss": 0.0955, + "step": 5683 + }, + { + "epoch": 0.74, + "learning_rate": 1.9792178095658917e-05, + "loss": 0.0945, + "step": 5684 + }, + { + "epoch": 0.74, + "learning_rate": 1.979207052589723e-05, + "loss": 0.0952, + "step": 5685 + }, + { + "epoch": 0.74, + "learning_rate": 1.979196292859586e-05, + "loss": 0.0942, + "step": 5686 + }, + { + "epoch": 0.74, + "learning_rate": 1.9791855303755122e-05, + "loss": 0.0926, + "step": 5687 + }, + { + "epoch": 0.75, + "learning_rate": 1.979174765137531e-05, + "loss": 0.0817, + "step": 5688 + }, + { + "epoch": 0.75, + "learning_rate": 1.979163997145673e-05, + "loss": 0.0824, + "step": 5689 + }, + { + "epoch": 0.75, + "learning_rate": 1.9791532263999684e-05, + "loss": 0.0842, + "step": 5690 + }, + { + "epoch": 0.75, + "learning_rate": 1.9791424529004473e-05, + "loss": 0.0848, + "step": 5691 + }, + { + "epoch": 0.75, + "learning_rate": 1.9791316766471405e-05, + "loss": 0.0969, + "step": 5692 + }, + { + "epoch": 0.75, + "learning_rate": 1.9791208976400777e-05, + "loss": 0.0886, + "step": 5693 + }, + { + "epoch": 0.75, + "learning_rate": 1.9791101158792894e-05, + "loss": 0.0863, + "step": 5694 + }, + { + "epoch": 0.75, + "learning_rate": 1.9790993313648065e-05, + "loss": 0.0843, + "step": 5695 + }, + { + "epoch": 0.75, + "learning_rate": 1.9790885440966585e-05, + "loss": 0.0872, + "step": 5696 + }, + { + "epoch": 0.75, + "learning_rate": 1.9790777540748765e-05, + "loss": 0.0826, + "step": 5697 + }, + { + "epoch": 0.75, + "learning_rate": 1.9790669612994903e-05, + "loss": 0.0869, + "step": 5698 + }, + { + "epoch": 0.75, + "learning_rate": 1.97905616577053e-05, + "loss": 0.096, + "step": 5699 + }, + { + "epoch": 0.75, + "learning_rate": 1.979045367488027e-05, + "loss": 0.0856, + "step": 5700 + }, + { + "epoch": 0.75, + "learning_rate": 1.9790345664520108e-05, + "loss": 0.0957, + "step": 5701 + }, + { + "epoch": 0.75, + "learning_rate": 1.979023762662512e-05, + "loss": 0.0904, + "step": 5702 + }, + { + "epoch": 0.75, + "learning_rate": 1.9790129561195606e-05, + "loss": 0.083, + "step": 5703 + }, + { + "epoch": 0.75, + "learning_rate": 1.9790021468231878e-05, + "loss": 0.0855, + "step": 5704 + }, + { + "epoch": 0.75, + "learning_rate": 1.9789913347734236e-05, + "loss": 0.0859, + "step": 5705 + }, + { + "epoch": 0.75, + "learning_rate": 1.9789805199702985e-05, + "loss": 0.0886, + "step": 5706 + }, + { + "epoch": 0.75, + "learning_rate": 1.9789697024138426e-05, + "loss": 0.0906, + "step": 5707 + }, + { + "epoch": 0.75, + "learning_rate": 1.9789588821040865e-05, + "loss": 0.0814, + "step": 5708 + }, + { + "epoch": 0.75, + "learning_rate": 1.978948059041061e-05, + "loss": 0.0878, + "step": 5709 + }, + { + "epoch": 0.75, + "learning_rate": 1.978937233224796e-05, + "loss": 0.089, + "step": 5710 + }, + { + "epoch": 0.75, + "learning_rate": 1.9789264046553224e-05, + "loss": 0.095, + "step": 5711 + }, + { + "epoch": 0.75, + "learning_rate": 1.97891557333267e-05, + "loss": 0.0709, + "step": 5712 + }, + { + "epoch": 0.75, + "learning_rate": 1.9789047392568697e-05, + "loss": 0.0913, + "step": 5713 + }, + { + "epoch": 0.75, + "learning_rate": 1.9788939024279525e-05, + "loss": 0.0858, + "step": 5714 + }, + { + "epoch": 0.75, + "learning_rate": 1.9788830628459478e-05, + "loss": 0.0937, + "step": 5715 + }, + { + "epoch": 0.75, + "learning_rate": 1.9788722205108865e-05, + "loss": 0.0834, + "step": 5716 + }, + { + "epoch": 0.75, + "learning_rate": 1.9788613754227997e-05, + "loss": 0.0824, + "step": 5717 + }, + { + "epoch": 0.75, + "learning_rate": 1.9788505275817173e-05, + "loss": 0.0871, + "step": 5718 + }, + { + "epoch": 0.75, + "learning_rate": 1.9788396769876697e-05, + "loss": 0.0946, + "step": 5719 + }, + { + "epoch": 0.75, + "learning_rate": 1.9788288236406877e-05, + "loss": 0.0918, + "step": 5720 + }, + { + "epoch": 0.75, + "learning_rate": 1.9788179675408016e-05, + "loss": 0.0784, + "step": 5721 + }, + { + "epoch": 0.75, + "learning_rate": 1.9788071086880425e-05, + "loss": 0.0861, + "step": 5722 + }, + { + "epoch": 0.75, + "learning_rate": 1.9787962470824406e-05, + "loss": 0.0897, + "step": 5723 + }, + { + "epoch": 0.75, + "learning_rate": 1.9787853827240256e-05, + "loss": 0.1115, + "step": 5724 + }, + { + "epoch": 0.75, + "learning_rate": 1.9787745156128294e-05, + "loss": 0.0968, + "step": 5725 + }, + { + "epoch": 0.75, + "learning_rate": 1.9787636457488822e-05, + "loss": 0.0817, + "step": 5726 + }, + { + "epoch": 0.75, + "learning_rate": 1.978752773132214e-05, + "loss": 0.0966, + "step": 5727 + }, + { + "epoch": 0.75, + "learning_rate": 1.9787418977628557e-05, + "loss": 0.083, + "step": 5728 + }, + { + "epoch": 0.75, + "learning_rate": 1.9787310196408384e-05, + "loss": 0.0957, + "step": 5729 + }, + { + "epoch": 0.75, + "learning_rate": 1.9787201387661917e-05, + "loss": 0.0958, + "step": 5730 + }, + { + "epoch": 0.75, + "learning_rate": 1.978709255138947e-05, + "loss": 0.087, + "step": 5731 + }, + { + "epoch": 0.75, + "learning_rate": 1.9786983687591347e-05, + "loss": 0.0924, + "step": 5732 + }, + { + "epoch": 0.75, + "learning_rate": 1.978687479626785e-05, + "loss": 0.0882, + "step": 5733 + }, + { + "epoch": 0.75, + "learning_rate": 1.978676587741929e-05, + "loss": 0.0829, + "step": 5734 + }, + { + "epoch": 0.75, + "learning_rate": 1.978665693104598e-05, + "loss": 0.1009, + "step": 5735 + }, + { + "epoch": 0.75, + "learning_rate": 1.978654795714821e-05, + "loss": 0.0952, + "step": 5736 + }, + { + "epoch": 0.75, + "learning_rate": 1.9786438955726293e-05, + "loss": 0.0865, + "step": 5737 + }, + { + "epoch": 0.75, + "learning_rate": 1.9786329926780544e-05, + "loss": 0.0903, + "step": 5738 + }, + { + "epoch": 0.75, + "learning_rate": 1.978622087031126e-05, + "loss": 0.0971, + "step": 5739 + }, + { + "epoch": 0.75, + "learning_rate": 1.978611178631875e-05, + "loss": 0.0938, + "step": 5740 + }, + { + "epoch": 0.75, + "learning_rate": 1.9786002674803324e-05, + "loss": 0.0956, + "step": 5741 + }, + { + "epoch": 0.75, + "learning_rate": 1.9785893535765287e-05, + "loss": 0.0904, + "step": 5742 + }, + { + "epoch": 0.75, + "learning_rate": 1.9785784369204947e-05, + "loss": 0.1113, + "step": 5743 + }, + { + "epoch": 0.75, + "learning_rate": 1.9785675175122606e-05, + "loss": 0.0873, + "step": 5744 + }, + { + "epoch": 0.75, + "learning_rate": 1.9785565953518575e-05, + "loss": 0.0828, + "step": 5745 + }, + { + "epoch": 0.75, + "learning_rate": 1.9785456704393163e-05, + "loss": 0.0855, + "step": 5746 + }, + { + "epoch": 0.75, + "learning_rate": 1.9785347427746672e-05, + "loss": 0.0971, + "step": 5747 + }, + { + "epoch": 0.75, + "learning_rate": 1.9785238123579416e-05, + "loss": 0.0886, + "step": 5748 + }, + { + "epoch": 0.75, + "learning_rate": 1.9785128791891697e-05, + "loss": 0.0888, + "step": 5749 + }, + { + "epoch": 0.75, + "learning_rate": 1.9785019432683823e-05, + "loss": 0.0908, + "step": 5750 + }, + { + "epoch": 0.75, + "learning_rate": 1.9784910045956106e-05, + "loss": 0.09, + "step": 5751 + }, + { + "epoch": 0.75, + "learning_rate": 1.9784800631708848e-05, + "loss": 0.1005, + "step": 5752 + }, + { + "epoch": 0.75, + "learning_rate": 1.978469118994236e-05, + "loss": 0.1003, + "step": 5753 + }, + { + "epoch": 0.75, + "learning_rate": 1.978458172065695e-05, + "loss": 0.0998, + "step": 5754 + }, + { + "epoch": 0.75, + "learning_rate": 1.9784472223852926e-05, + "loss": 0.0995, + "step": 5755 + }, + { + "epoch": 0.75, + "learning_rate": 1.9784362699530592e-05, + "loss": 0.0984, + "step": 5756 + }, + { + "epoch": 0.75, + "learning_rate": 1.978425314769026e-05, + "loss": 0.0917, + "step": 5757 + }, + { + "epoch": 0.75, + "learning_rate": 1.978414356833224e-05, + "loss": 0.0984, + "step": 5758 + }, + { + "epoch": 0.75, + "learning_rate": 1.978403396145683e-05, + "loss": 0.101, + "step": 5759 + }, + { + "epoch": 0.75, + "learning_rate": 1.9783924327064352e-05, + "loss": 0.0863, + "step": 5760 + }, + { + "epoch": 0.75, + "learning_rate": 1.978381466515511e-05, + "loss": 0.0891, + "step": 5761 + }, + { + "epoch": 0.75, + "learning_rate": 1.9783704975729405e-05, + "loss": 0.0838, + "step": 5762 + }, + { + "epoch": 0.75, + "learning_rate": 1.978359525878755e-05, + "loss": 0.0794, + "step": 5763 + }, + { + "epoch": 0.76, + "learning_rate": 1.9783485514329858e-05, + "loss": 0.084, + "step": 5764 + }, + { + "epoch": 0.76, + "learning_rate": 1.9783375742356632e-05, + "loss": 0.0939, + "step": 5765 + }, + { + "epoch": 0.76, + "learning_rate": 1.978326594286818e-05, + "loss": 0.0914, + "step": 5766 + }, + { + "epoch": 0.76, + "learning_rate": 1.978315611586482e-05, + "loss": 0.0859, + "step": 5767 + }, + { + "epoch": 0.76, + "learning_rate": 1.978304626134685e-05, + "loss": 0.0799, + "step": 5768 + }, + { + "epoch": 0.76, + "learning_rate": 1.9782936379314586e-05, + "loss": 0.0939, + "step": 5769 + }, + { + "epoch": 0.76, + "learning_rate": 1.9782826469768332e-05, + "loss": 0.0863, + "step": 5770 + }, + { + "epoch": 0.76, + "learning_rate": 1.9782716532708402e-05, + "loss": 0.0943, + "step": 5771 + }, + { + "epoch": 0.76, + "learning_rate": 1.97826065681351e-05, + "loss": 0.0869, + "step": 5772 + }, + { + "epoch": 0.76, + "learning_rate": 1.978249657604874e-05, + "loss": 0.09, + "step": 5773 + }, + { + "epoch": 0.76, + "learning_rate": 1.9782386556449625e-05, + "loss": 0.0817, + "step": 5774 + }, + { + "epoch": 0.76, + "learning_rate": 1.9782276509338075e-05, + "loss": 0.0928, + "step": 5775 + }, + { + "epoch": 0.76, + "learning_rate": 1.978216643471439e-05, + "loss": 0.0842, + "step": 5776 + }, + { + "epoch": 0.76, + "learning_rate": 1.9782056332578882e-05, + "loss": 0.089, + "step": 5777 + }, + { + "epoch": 0.76, + "learning_rate": 1.9781946202931862e-05, + "loss": 0.0903, + "step": 5778 + }, + { + "epoch": 0.76, + "learning_rate": 1.9781836045773643e-05, + "loss": 0.0946, + "step": 5779 + }, + { + "epoch": 0.76, + "learning_rate": 1.9781725861104525e-05, + "loss": 0.0865, + "step": 5780 + }, + { + "epoch": 0.76, + "learning_rate": 1.9781615648924827e-05, + "loss": 0.0951, + "step": 5781 + }, + { + "epoch": 0.76, + "learning_rate": 1.9781505409234855e-05, + "loss": 0.0804, + "step": 5782 + }, + { + "epoch": 0.76, + "learning_rate": 1.9781395142034923e-05, + "loss": 0.0941, + "step": 5783 + }, + { + "epoch": 0.76, + "learning_rate": 1.9781284847325336e-05, + "loss": 0.0958, + "step": 5784 + }, + { + "epoch": 0.76, + "learning_rate": 1.9781174525106405e-05, + "loss": 0.1012, + "step": 5785 + }, + { + "epoch": 0.76, + "learning_rate": 1.9781064175378444e-05, + "loss": 0.0852, + "step": 5786 + }, + { + "epoch": 0.76, + "learning_rate": 1.978095379814176e-05, + "loss": 0.085, + "step": 5787 + }, + { + "epoch": 0.76, + "learning_rate": 1.9780843393396663e-05, + "loss": 0.0967, + "step": 5788 + }, + { + "epoch": 0.76, + "learning_rate": 1.9780732961143463e-05, + "loss": 0.0897, + "step": 5789 + }, + { + "epoch": 0.76, + "learning_rate": 1.9780622501382476e-05, + "loss": 0.1078, + "step": 5790 + }, + { + "epoch": 0.76, + "learning_rate": 1.978051201411401e-05, + "loss": 0.0967, + "step": 5791 + }, + { + "epoch": 0.76, + "learning_rate": 1.978040149933837e-05, + "loss": 0.0922, + "step": 5792 + }, + { + "epoch": 0.76, + "learning_rate": 1.9780290957055873e-05, + "loss": 0.0947, + "step": 5793 + }, + { + "epoch": 0.76, + "learning_rate": 1.9780180387266834e-05, + "loss": 0.0888, + "step": 5794 + }, + { + "epoch": 0.76, + "learning_rate": 1.9780069789971552e-05, + "loss": 0.0889, + "step": 5795 + }, + { + "epoch": 0.76, + "learning_rate": 1.977995916517035e-05, + "loss": 0.0823, + "step": 5796 + }, + { + "epoch": 0.76, + "learning_rate": 1.977984851286353e-05, + "loss": 0.0931, + "step": 5797 + }, + { + "epoch": 0.76, + "learning_rate": 1.9779737833051406e-05, + "loss": 0.0884, + "step": 5798 + }, + { + "epoch": 0.76, + "learning_rate": 1.9779627125734292e-05, + "loss": 0.0893, + "step": 5799 + }, + { + "epoch": 0.76, + "learning_rate": 1.9779516390912498e-05, + "loss": 0.0855, + "step": 5800 + }, + { + "epoch": 0.76, + "learning_rate": 1.9779405628586334e-05, + "loss": 0.0807, + "step": 5801 + }, + { + "epoch": 0.76, + "learning_rate": 1.9779294838756113e-05, + "loss": 0.1034, + "step": 5802 + }, + { + "epoch": 0.76, + "learning_rate": 1.9779184021422146e-05, + "loss": 0.0841, + "step": 5803 + }, + { + "epoch": 0.76, + "learning_rate": 1.9779073176584744e-05, + "loss": 0.0908, + "step": 5804 + }, + { + "epoch": 0.76, + "learning_rate": 1.9778962304244217e-05, + "loss": 0.0985, + "step": 5805 + }, + { + "epoch": 0.76, + "learning_rate": 1.9778851404400882e-05, + "loss": 0.0853, + "step": 5806 + }, + { + "epoch": 0.76, + "learning_rate": 1.9778740477055048e-05, + "loss": 0.0812, + "step": 5807 + }, + { + "epoch": 0.76, + "learning_rate": 1.977862952220703e-05, + "loss": 0.0872, + "step": 5808 + }, + { + "epoch": 0.76, + "learning_rate": 1.9778518539857133e-05, + "loss": 0.0849, + "step": 5809 + }, + { + "epoch": 0.76, + "learning_rate": 1.9778407530005675e-05, + "loss": 0.0922, + "step": 5810 + }, + { + "epoch": 0.76, + "learning_rate": 1.9778296492652965e-05, + "loss": 0.0781, + "step": 5811 + }, + { + "epoch": 0.76, + "learning_rate": 1.9778185427799318e-05, + "loss": 0.0736, + "step": 5812 + }, + { + "epoch": 0.76, + "learning_rate": 1.977807433544505e-05, + "loss": 0.0889, + "step": 5813 + }, + { + "epoch": 0.76, + "learning_rate": 1.9777963215590463e-05, + "loss": 0.093, + "step": 5814 + }, + { + "epoch": 0.76, + "learning_rate": 1.9777852068235878e-05, + "loss": 0.0875, + "step": 5815 + }, + { + "epoch": 0.76, + "learning_rate": 1.9777740893381602e-05, + "loss": 0.0878, + "step": 5816 + }, + { + "epoch": 0.76, + "learning_rate": 1.9777629691027948e-05, + "loss": 0.0848, + "step": 5817 + }, + { + "epoch": 0.76, + "learning_rate": 1.9777518461175237e-05, + "loss": 0.0915, + "step": 5818 + }, + { + "epoch": 0.76, + "learning_rate": 1.977740720382377e-05, + "loss": 0.0748, + "step": 5819 + }, + { + "epoch": 0.76, + "learning_rate": 1.9777295918973872e-05, + "loss": 0.0824, + "step": 5820 + }, + { + "epoch": 0.76, + "learning_rate": 1.977718460662585e-05, + "loss": 0.093, + "step": 5821 + }, + { + "epoch": 0.76, + "learning_rate": 1.9777073266780013e-05, + "loss": 0.0948, + "step": 5822 + }, + { + "epoch": 0.76, + "learning_rate": 1.977696189943668e-05, + "loss": 0.087, + "step": 5823 + }, + { + "epoch": 0.76, + "learning_rate": 1.977685050459616e-05, + "loss": 0.0817, + "step": 5824 + }, + { + "epoch": 0.76, + "learning_rate": 1.977673908225877e-05, + "loss": 0.0722, + "step": 5825 + }, + { + "epoch": 0.76, + "learning_rate": 1.9776627632424824e-05, + "loss": 0.0864, + "step": 5826 + }, + { + "epoch": 0.76, + "learning_rate": 1.9776516155094633e-05, + "loss": 0.0805, + "step": 5827 + }, + { + "epoch": 0.76, + "learning_rate": 1.9776404650268512e-05, + "loss": 0.0814, + "step": 5828 + }, + { + "epoch": 0.76, + "learning_rate": 1.9776293117946766e-05, + "loss": 0.0893, + "step": 5829 + }, + { + "epoch": 0.76, + "learning_rate": 1.9776181558129725e-05, + "loss": 0.082, + "step": 5830 + }, + { + "epoch": 0.76, + "learning_rate": 1.9776069970817692e-05, + "loss": 0.0834, + "step": 5831 + }, + { + "epoch": 0.76, + "learning_rate": 1.9775958356010977e-05, + "loss": 0.0871, + "step": 5832 + }, + { + "epoch": 0.76, + "learning_rate": 1.9775846713709908e-05, + "loss": 0.0941, + "step": 5833 + }, + { + "epoch": 0.76, + "learning_rate": 1.9775735043914786e-05, + "loss": 0.0974, + "step": 5834 + }, + { + "epoch": 0.76, + "learning_rate": 1.977562334662593e-05, + "loss": 0.0908, + "step": 5835 + }, + { + "epoch": 0.76, + "learning_rate": 1.9775511621843655e-05, + "loss": 0.0881, + "step": 5836 + }, + { + "epoch": 0.76, + "learning_rate": 1.9775399869568273e-05, + "loss": 0.085, + "step": 5837 + }, + { + "epoch": 0.76, + "learning_rate": 1.97752880898001e-05, + "loss": 0.0774, + "step": 5838 + }, + { + "epoch": 0.76, + "learning_rate": 1.977517628253945e-05, + "loss": 0.0959, + "step": 5839 + }, + { + "epoch": 0.76, + "learning_rate": 1.977506444778664e-05, + "loss": 0.0927, + "step": 5840 + }, + { + "epoch": 0.77, + "learning_rate": 1.9774952585541976e-05, + "loss": 0.0925, + "step": 5841 + }, + { + "epoch": 0.77, + "learning_rate": 1.977484069580578e-05, + "loss": 0.0883, + "step": 5842 + }, + { + "epoch": 0.77, + "learning_rate": 1.977472877857837e-05, + "loss": 0.0843, + "step": 5843 + }, + { + "epoch": 0.77, + "learning_rate": 1.9774616833860053e-05, + "loss": 0.0956, + "step": 5844 + }, + { + "epoch": 0.77, + "learning_rate": 1.9774504861651143e-05, + "loss": 0.0955, + "step": 5845 + }, + { + "epoch": 0.77, + "learning_rate": 1.9774392861951962e-05, + "loss": 0.0745, + "step": 5846 + }, + { + "epoch": 0.77, + "learning_rate": 1.977428083476282e-05, + "loss": 0.0786, + "step": 5847 + }, + { + "epoch": 0.77, + "learning_rate": 1.9774168780084036e-05, + "loss": 0.0916, + "step": 5848 + }, + { + "epoch": 0.77, + "learning_rate": 1.977405669791592e-05, + "loss": 0.0812, + "step": 5849 + }, + { + "epoch": 0.77, + "learning_rate": 1.9773944588258792e-05, + "loss": 0.0713, + "step": 5850 + }, + { + "epoch": 0.77, + "learning_rate": 1.9773832451112967e-05, + "loss": 0.0832, + "step": 5851 + }, + { + "epoch": 0.77, + "learning_rate": 1.977372028647876e-05, + "loss": 0.1009, + "step": 5852 + }, + { + "epoch": 0.77, + "learning_rate": 1.9773608094356477e-05, + "loss": 0.0943, + "step": 5853 + }, + { + "epoch": 0.77, + "learning_rate": 1.977349587474645e-05, + "loss": 0.0974, + "step": 5854 + }, + { + "epoch": 0.77, + "learning_rate": 1.977338362764898e-05, + "loss": 0.0987, + "step": 5855 + }, + { + "epoch": 0.77, + "learning_rate": 1.9773271353064393e-05, + "loss": 0.0869, + "step": 5856 + }, + { + "epoch": 0.77, + "learning_rate": 1.9773159050993e-05, + "loss": 0.0937, + "step": 5857 + }, + { + "epoch": 0.77, + "learning_rate": 1.977304672143512e-05, + "loss": 0.0923, + "step": 5858 + }, + { + "epoch": 0.77, + "learning_rate": 1.977293436439106e-05, + "loss": 0.0898, + "step": 5859 + }, + { + "epoch": 0.77, + "learning_rate": 1.9772821979861146e-05, + "loss": 0.0911, + "step": 5860 + }, + { + "epoch": 0.77, + "learning_rate": 1.9772709567845693e-05, + "loss": 0.1043, + "step": 5861 + }, + { + "epoch": 0.77, + "learning_rate": 1.9772597128345013e-05, + "loss": 0.0854, + "step": 5862 + }, + { + "epoch": 0.77, + "learning_rate": 1.9772484661359422e-05, + "loss": 0.0907, + "step": 5863 + }, + { + "epoch": 0.77, + "learning_rate": 1.977237216688924e-05, + "loss": 0.0912, + "step": 5864 + }, + { + "epoch": 0.77, + "learning_rate": 1.9772259644934784e-05, + "loss": 0.0828, + "step": 5865 + }, + { + "epoch": 0.77, + "learning_rate": 1.9772147095496367e-05, + "loss": 0.1122, + "step": 5866 + }, + { + "epoch": 0.77, + "learning_rate": 1.9772034518574305e-05, + "loss": 0.0854, + "step": 5867 + }, + { + "epoch": 0.77, + "learning_rate": 1.9771921914168915e-05, + "loss": 0.094, + "step": 5868 + }, + { + "epoch": 0.77, + "learning_rate": 1.9771809282280517e-05, + "loss": 0.1008, + "step": 5869 + }, + { + "epoch": 0.77, + "learning_rate": 1.9771696622909427e-05, + "loss": 0.0796, + "step": 5870 + }, + { + "epoch": 0.77, + "learning_rate": 1.977158393605596e-05, + "loss": 0.0809, + "step": 5871 + }, + { + "epoch": 0.77, + "learning_rate": 1.9771471221720435e-05, + "loss": 0.0852, + "step": 5872 + }, + { + "epoch": 0.77, + "learning_rate": 1.9771358479903162e-05, + "loss": 0.0901, + "step": 5873 + }, + { + "epoch": 0.77, + "learning_rate": 1.977124571060447e-05, + "loss": 0.0746, + "step": 5874 + }, + { + "epoch": 0.77, + "learning_rate": 1.9771132913824665e-05, + "loss": 0.087, + "step": 5875 + }, + { + "epoch": 0.77, + "learning_rate": 1.977102008956407e-05, + "loss": 0.0829, + "step": 5876 + }, + { + "epoch": 0.77, + "learning_rate": 1.9770907237823002e-05, + "loss": 0.0739, + "step": 5877 + }, + { + "epoch": 0.77, + "learning_rate": 1.9770794358601777e-05, + "loss": 0.0822, + "step": 5878 + }, + { + "epoch": 0.77, + "learning_rate": 1.977068145190072e-05, + "loss": 0.0832, + "step": 5879 + }, + { + "epoch": 0.77, + "learning_rate": 1.9770568517720133e-05, + "loss": 0.0881, + "step": 5880 + }, + { + "epoch": 0.77, + "learning_rate": 1.9770455556060345e-05, + "loss": 0.0906, + "step": 5881 + }, + { + "epoch": 0.77, + "learning_rate": 1.977034256692167e-05, + "loss": 0.0921, + "step": 5882 + }, + { + "epoch": 0.77, + "learning_rate": 1.9770229550304426e-05, + "loss": 0.1045, + "step": 5883 + }, + { + "epoch": 0.77, + "learning_rate": 1.9770116506208935e-05, + "loss": 0.0813, + "step": 5884 + }, + { + "epoch": 0.77, + "learning_rate": 1.9770003434635513e-05, + "loss": 0.0933, + "step": 5885 + }, + { + "epoch": 0.77, + "learning_rate": 1.9769890335584473e-05, + "loss": 0.0823, + "step": 5886 + }, + { + "epoch": 0.77, + "learning_rate": 1.976977720905614e-05, + "loss": 0.0798, + "step": 5887 + }, + { + "epoch": 0.77, + "learning_rate": 1.9769664055050826e-05, + "loss": 0.0849, + "step": 5888 + }, + { + "epoch": 0.77, + "learning_rate": 1.976955087356885e-05, + "loss": 0.087, + "step": 5889 + }, + { + "epoch": 0.77, + "learning_rate": 1.9769437664610537e-05, + "loss": 0.0809, + "step": 5890 + }, + { + "epoch": 0.77, + "learning_rate": 1.9769324428176198e-05, + "loss": 0.0922, + "step": 5891 + }, + { + "epoch": 0.77, + "learning_rate": 1.9769211164266157e-05, + "loss": 0.0974, + "step": 5892 + }, + { + "epoch": 0.77, + "learning_rate": 1.9769097872880725e-05, + "loss": 0.0869, + "step": 5893 + }, + { + "epoch": 0.77, + "learning_rate": 1.9768984554020232e-05, + "loss": 0.0934, + "step": 5894 + }, + { + "epoch": 0.77, + "learning_rate": 1.9768871207684985e-05, + "loss": 0.0862, + "step": 5895 + }, + { + "epoch": 0.77, + "learning_rate": 1.976875783387531e-05, + "loss": 0.0855, + "step": 5896 + }, + { + "epoch": 0.77, + "learning_rate": 1.9768644432591523e-05, + "loss": 0.0946, + "step": 5897 + }, + { + "epoch": 0.77, + "learning_rate": 1.9768531003833943e-05, + "loss": 0.0936, + "step": 5898 + }, + { + "epoch": 0.77, + "learning_rate": 1.9768417547602892e-05, + "loss": 0.0891, + "step": 5899 + }, + { + "epoch": 0.77, + "learning_rate": 1.9768304063898688e-05, + "loss": 0.0833, + "step": 5900 + }, + { + "epoch": 0.77, + "learning_rate": 1.9768190552721647e-05, + "loss": 0.0781, + "step": 5901 + }, + { + "epoch": 0.77, + "learning_rate": 1.976807701407209e-05, + "loss": 0.1041, + "step": 5902 + }, + { + "epoch": 0.77, + "learning_rate": 1.9767963447950335e-05, + "loss": 0.0874, + "step": 5903 + }, + { + "epoch": 0.77, + "learning_rate": 1.9767849854356704e-05, + "loss": 0.0939, + "step": 5904 + }, + { + "epoch": 0.77, + "learning_rate": 1.9767736233291518e-05, + "loss": 0.091, + "step": 5905 + }, + { + "epoch": 0.77, + "learning_rate": 1.976762258475509e-05, + "loss": 0.0951, + "step": 5906 + }, + { + "epoch": 0.77, + "learning_rate": 1.9767508908747746e-05, + "loss": 0.095, + "step": 5907 + }, + { + "epoch": 0.77, + "learning_rate": 1.9767395205269805e-05, + "loss": 0.0949, + "step": 5908 + }, + { + "epoch": 0.77, + "learning_rate": 1.9767281474321582e-05, + "loss": 0.0934, + "step": 5909 + }, + { + "epoch": 0.77, + "learning_rate": 1.9767167715903403e-05, + "loss": 0.089, + "step": 5910 + }, + { + "epoch": 0.77, + "learning_rate": 1.9767053930015583e-05, + "loss": 0.0962, + "step": 5911 + }, + { + "epoch": 0.77, + "learning_rate": 1.976694011665844e-05, + "loss": 0.0866, + "step": 5912 + }, + { + "epoch": 0.77, + "learning_rate": 1.9766826275832308e-05, + "loss": 0.096, + "step": 5913 + }, + { + "epoch": 0.77, + "learning_rate": 1.976671240753749e-05, + "loss": 0.0898, + "step": 5914 + }, + { + "epoch": 0.77, + "learning_rate": 1.9766598511774314e-05, + "loss": 0.0873, + "step": 5915 + }, + { + "epoch": 0.77, + "learning_rate": 1.97664845885431e-05, + "loss": 0.0908, + "step": 5916 + }, + { + "epoch": 0.78, + "learning_rate": 1.9766370637844167e-05, + "loss": 0.0925, + "step": 5917 + }, + { + "epoch": 0.78, + "learning_rate": 1.9766256659677842e-05, + "loss": 0.0902, + "step": 5918 + }, + { + "epoch": 0.78, + "learning_rate": 1.9766142654044437e-05, + "loss": 0.0825, + "step": 5919 + }, + { + "epoch": 0.78, + "learning_rate": 1.976602862094427e-05, + "loss": 0.0825, + "step": 5920 + }, + { + "epoch": 0.78, + "learning_rate": 1.9765914560377677e-05, + "loss": 0.0796, + "step": 5921 + }, + { + "epoch": 0.78, + "learning_rate": 1.9765800472344968e-05, + "loss": 0.0836, + "step": 5922 + }, + { + "epoch": 0.78, + "learning_rate": 1.976568635684646e-05, + "loss": 0.0814, + "step": 5923 + }, + { + "epoch": 0.78, + "learning_rate": 1.9765572213882483e-05, + "loss": 0.0987, + "step": 5924 + }, + { + "epoch": 0.78, + "learning_rate": 1.976545804345335e-05, + "loss": 0.0837, + "step": 5925 + }, + { + "epoch": 0.78, + "learning_rate": 1.9765343845559394e-05, + "loss": 0.0915, + "step": 5926 + }, + { + "epoch": 0.78, + "learning_rate": 1.9765229620200923e-05, + "loss": 0.0848, + "step": 5927 + }, + { + "epoch": 0.78, + "learning_rate": 1.9765115367378263e-05, + "loss": 0.0794, + "step": 5928 + }, + { + "epoch": 0.78, + "learning_rate": 1.976500108709174e-05, + "loss": 0.087, + "step": 5929 + }, + { + "epoch": 0.78, + "learning_rate": 1.976488677934167e-05, + "loss": 0.0858, + "step": 5930 + }, + { + "epoch": 0.78, + "learning_rate": 1.9764772444128374e-05, + "loss": 0.0992, + "step": 5931 + }, + { + "epoch": 0.78, + "learning_rate": 1.9764658081452178e-05, + "loss": 0.0915, + "step": 5932 + }, + { + "epoch": 0.78, + "learning_rate": 1.97645436913134e-05, + "loss": 0.0855, + "step": 5933 + }, + { + "epoch": 0.78, + "learning_rate": 1.976442927371236e-05, + "loss": 0.0878, + "step": 5934 + }, + { + "epoch": 0.78, + "learning_rate": 1.9764314828649386e-05, + "loss": 0.0812, + "step": 5935 + }, + { + "epoch": 0.78, + "learning_rate": 1.9764200356124794e-05, + "loss": 0.0781, + "step": 5936 + }, + { + "epoch": 0.78, + "learning_rate": 1.976408585613891e-05, + "loss": 0.0967, + "step": 5937 + }, + { + "epoch": 0.78, + "learning_rate": 1.9763971328692057e-05, + "loss": 0.0832, + "step": 5938 + }, + { + "epoch": 0.78, + "learning_rate": 1.9763856773784552e-05, + "loss": 0.0913, + "step": 5939 + }, + { + "epoch": 0.78, + "learning_rate": 1.976374219141672e-05, + "loss": 0.0879, + "step": 5940 + }, + { + "epoch": 0.78, + "learning_rate": 1.9763627581588884e-05, + "loss": 0.0919, + "step": 5941 + }, + { + "epoch": 0.78, + "learning_rate": 1.9763512944301363e-05, + "loss": 0.0848, + "step": 5942 + }, + { + "epoch": 0.78, + "learning_rate": 1.9763398279554482e-05, + "loss": 0.0864, + "step": 5943 + }, + { + "epoch": 0.78, + "learning_rate": 1.9763283587348568e-05, + "loss": 0.0882, + "step": 5944 + }, + { + "epoch": 0.78, + "learning_rate": 1.9763168867683932e-05, + "loss": 0.0857, + "step": 5945 + }, + { + "epoch": 0.78, + "learning_rate": 1.9763054120560908e-05, + "loss": 0.0851, + "step": 5946 + }, + { + "epoch": 0.78, + "learning_rate": 1.9762939345979813e-05, + "loss": 0.083, + "step": 5947 + }, + { + "epoch": 0.78, + "learning_rate": 1.976282454394097e-05, + "loss": 0.0871, + "step": 5948 + }, + { + "epoch": 0.78, + "learning_rate": 1.9762709714444706e-05, + "loss": 0.0904, + "step": 5949 + }, + { + "epoch": 0.78, + "learning_rate": 1.9762594857491337e-05, + "loss": 0.0945, + "step": 5950 + }, + { + "epoch": 0.78, + "learning_rate": 1.9762479973081193e-05, + "loss": 0.0881, + "step": 5951 + }, + { + "epoch": 0.78, + "learning_rate": 1.9762365061214593e-05, + "loss": 0.0988, + "step": 5952 + }, + { + "epoch": 0.78, + "learning_rate": 1.976225012189186e-05, + "loss": 0.0907, + "step": 5953 + }, + { + "epoch": 0.78, + "learning_rate": 1.976213515511332e-05, + "loss": 0.1064, + "step": 5954 + }, + { + "epoch": 0.78, + "learning_rate": 1.9762020160879295e-05, + "loss": 0.0873, + "step": 5955 + }, + { + "epoch": 0.78, + "learning_rate": 1.9761905139190104e-05, + "loss": 0.0844, + "step": 5956 + }, + { + "epoch": 0.78, + "learning_rate": 1.976179009004608e-05, + "loss": 0.0865, + "step": 5957 + }, + { + "epoch": 0.78, + "learning_rate": 1.9761675013447538e-05, + "loss": 0.088, + "step": 5958 + }, + { + "epoch": 0.78, + "learning_rate": 1.9761559909394808e-05, + "loss": 0.0915, + "step": 5959 + }, + { + "epoch": 0.78, + "learning_rate": 1.976144477788821e-05, + "loss": 0.0799, + "step": 5960 + }, + { + "epoch": 0.78, + "learning_rate": 1.976132961892806e-05, + "loss": 0.0765, + "step": 5961 + }, + { + "epoch": 0.78, + "learning_rate": 1.97612144325147e-05, + "loss": 0.0771, + "step": 5962 + }, + { + "epoch": 0.78, + "learning_rate": 1.9761099218648443e-05, + "loss": 0.0862, + "step": 5963 + }, + { + "epoch": 0.78, + "learning_rate": 1.976098397732961e-05, + "loss": 0.0853, + "step": 5964 + }, + { + "epoch": 0.78, + "learning_rate": 1.9760868708558532e-05, + "loss": 0.0929, + "step": 5965 + }, + { + "epoch": 0.78, + "learning_rate": 1.976075341233553e-05, + "loss": 0.1005, + "step": 5966 + }, + { + "epoch": 0.78, + "learning_rate": 1.976063808866093e-05, + "loss": 0.0908, + "step": 5967 + }, + { + "epoch": 0.78, + "learning_rate": 1.9760522737535055e-05, + "loss": 0.0904, + "step": 5968 + }, + { + "epoch": 0.78, + "learning_rate": 1.976040735895823e-05, + "loss": 0.085, + "step": 5969 + }, + { + "epoch": 0.78, + "learning_rate": 1.9760291952930778e-05, + "loss": 0.0931, + "step": 5970 + }, + { + "epoch": 0.78, + "learning_rate": 1.9760176519453026e-05, + "loss": 0.0864, + "step": 5971 + }, + { + "epoch": 0.78, + "learning_rate": 1.9760061058525295e-05, + "loss": 0.0941, + "step": 5972 + }, + { + "epoch": 0.78, + "learning_rate": 1.9759945570147913e-05, + "loss": 0.0797, + "step": 5973 + }, + { + "epoch": 0.78, + "learning_rate": 1.9759830054321204e-05, + "loss": 0.0975, + "step": 5974 + }, + { + "epoch": 0.78, + "learning_rate": 1.9759714511045492e-05, + "loss": 0.0833, + "step": 5975 + }, + { + "epoch": 0.78, + "learning_rate": 1.97595989403211e-05, + "loss": 0.1085, + "step": 5976 + }, + { + "epoch": 0.78, + "learning_rate": 1.9759483342148364e-05, + "loss": 0.078, + "step": 5977 + }, + { + "epoch": 0.78, + "learning_rate": 1.975936771652759e-05, + "loss": 0.0904, + "step": 5978 + }, + { + "epoch": 0.78, + "learning_rate": 1.9759252063459123e-05, + "loss": 0.0986, + "step": 5979 + }, + { + "epoch": 0.78, + "learning_rate": 1.9759136382943277e-05, + "loss": 0.0969, + "step": 5980 + }, + { + "epoch": 0.78, + "learning_rate": 1.975902067498038e-05, + "loss": 0.0914, + "step": 5981 + }, + { + "epoch": 0.78, + "learning_rate": 1.9758904939570757e-05, + "loss": 0.0978, + "step": 5982 + }, + { + "epoch": 0.78, + "learning_rate": 1.975878917671473e-05, + "loss": 0.0841, + "step": 5983 + }, + { + "epoch": 0.78, + "learning_rate": 1.975867338641263e-05, + "loss": 0.0962, + "step": 5984 + }, + { + "epoch": 0.78, + "learning_rate": 1.975855756866478e-05, + "loss": 0.0916, + "step": 5985 + }, + { + "epoch": 0.78, + "learning_rate": 1.9758441723471507e-05, + "loss": 0.0958, + "step": 5986 + }, + { + "epoch": 0.78, + "learning_rate": 1.975832585083314e-05, + "loss": 0.0949, + "step": 5987 + }, + { + "epoch": 0.78, + "learning_rate": 1.9758209950749995e-05, + "loss": 0.0847, + "step": 5988 + }, + { + "epoch": 0.78, + "learning_rate": 1.9758094023222408e-05, + "loss": 0.0777, + "step": 5989 + }, + { + "epoch": 0.78, + "learning_rate": 1.97579780682507e-05, + "loss": 0.0883, + "step": 5990 + }, + { + "epoch": 0.78, + "learning_rate": 1.97578620858352e-05, + "loss": 0.1095, + "step": 5991 + }, + { + "epoch": 0.78, + "learning_rate": 1.9757746075976234e-05, + "loss": 0.0743, + "step": 5992 + }, + { + "epoch": 0.79, + "learning_rate": 1.9757630038674122e-05, + "loss": 0.0897, + "step": 5993 + }, + { + "epoch": 0.79, + "learning_rate": 1.97575139739292e-05, + "loss": 0.0906, + "step": 5994 + }, + { + "epoch": 0.79, + "learning_rate": 1.9757397881741786e-05, + "loss": 0.0914, + "step": 5995 + }, + { + "epoch": 0.79, + "learning_rate": 1.9757281762112212e-05, + "loss": 0.0913, + "step": 5996 + }, + { + "epoch": 0.79, + "learning_rate": 1.97571656150408e-05, + "loss": 0.0764, + "step": 5997 + }, + { + "epoch": 0.79, + "learning_rate": 1.975704944052788e-05, + "loss": 0.0894, + "step": 5998 + }, + { + "epoch": 0.79, + "learning_rate": 1.975693323857378e-05, + "loss": 0.0897, + "step": 5999 + }, + { + "epoch": 0.79, + "learning_rate": 1.9756817009178825e-05, + "loss": 0.082, + "step": 6000 + }, + { + "epoch": 0.79, + "learning_rate": 1.975670075234334e-05, + "loss": 0.0822, + "step": 6001 + }, + { + "epoch": 0.79, + "learning_rate": 1.9756584468067655e-05, + "loss": 0.0758, + "step": 6002 + }, + { + "epoch": 0.79, + "learning_rate": 1.9756468156352096e-05, + "loss": 0.0769, + "step": 6003 + }, + { + "epoch": 0.79, + "learning_rate": 1.9756351817196986e-05, + "loss": 0.0846, + "step": 6004 + }, + { + "epoch": 0.79, + "learning_rate": 1.975623545060266e-05, + "loss": 0.0891, + "step": 6005 + }, + { + "epoch": 0.79, + "learning_rate": 1.9756119056569442e-05, + "loss": 0.0809, + "step": 6006 + }, + { + "epoch": 0.79, + "learning_rate": 1.9756002635097654e-05, + "loss": 0.0952, + "step": 6007 + }, + { + "epoch": 0.79, + "learning_rate": 1.975588618618763e-05, + "loss": 0.0919, + "step": 6008 + }, + { + "epoch": 0.79, + "learning_rate": 1.9755769709839702e-05, + "loss": 0.0893, + "step": 6009 + }, + { + "epoch": 0.79, + "learning_rate": 1.9755653206054184e-05, + "loss": 0.0863, + "step": 6010 + }, + { + "epoch": 0.79, + "learning_rate": 1.9755536674831413e-05, + "loss": 0.0938, + "step": 6011 + }, + { + "epoch": 0.79, + "learning_rate": 1.9755420116171713e-05, + "loss": 0.0918, + "step": 6012 + }, + { + "epoch": 0.79, + "learning_rate": 1.9755303530075416e-05, + "loss": 0.0871, + "step": 6013 + }, + { + "epoch": 0.79, + "learning_rate": 1.9755186916542848e-05, + "loss": 0.0843, + "step": 6014 + }, + { + "epoch": 0.79, + "learning_rate": 1.9755070275574334e-05, + "loss": 0.0915, + "step": 6015 + }, + { + "epoch": 0.79, + "learning_rate": 1.9754953607170202e-05, + "loss": 0.0951, + "step": 6016 + }, + { + "epoch": 0.79, + "learning_rate": 1.975483691133079e-05, + "loss": 0.0885, + "step": 6017 + }, + { + "epoch": 0.79, + "learning_rate": 1.9754720188056412e-05, + "loss": 0.0939, + "step": 6018 + }, + { + "epoch": 0.79, + "learning_rate": 1.9754603437347404e-05, + "loss": 0.0887, + "step": 6019 + }, + { + "epoch": 0.79, + "learning_rate": 1.9754486659204096e-05, + "loss": 0.0883, + "step": 6020 + }, + { + "epoch": 0.79, + "learning_rate": 1.975436985362681e-05, + "loss": 0.0798, + "step": 6021 + }, + { + "epoch": 0.79, + "learning_rate": 1.975425302061588e-05, + "loss": 0.0841, + "step": 6022 + }, + { + "epoch": 0.79, + "learning_rate": 1.9754136160171633e-05, + "loss": 0.0774, + "step": 6023 + }, + { + "epoch": 0.79, + "learning_rate": 1.9754019272294397e-05, + "loss": 0.0872, + "step": 6024 + }, + { + "epoch": 0.79, + "learning_rate": 1.97539023569845e-05, + "loss": 0.0943, + "step": 6025 + }, + { + "epoch": 0.79, + "learning_rate": 1.975378541424228e-05, + "loss": 0.0794, + "step": 6026 + }, + { + "epoch": 0.79, + "learning_rate": 1.9753668444068047e-05, + "loss": 0.1049, + "step": 6027 + }, + { + "epoch": 0.79, + "learning_rate": 1.9753551446462148e-05, + "loss": 0.09, + "step": 6028 + }, + { + "epoch": 0.79, + "learning_rate": 1.9753434421424905e-05, + "loss": 0.0803, + "step": 6029 + }, + { + "epoch": 0.79, + "learning_rate": 1.9753317368956645e-05, + "loss": 0.0967, + "step": 6030 + }, + { + "epoch": 0.79, + "learning_rate": 1.9753200289057697e-05, + "loss": 0.0944, + "step": 6031 + }, + { + "epoch": 0.79, + "learning_rate": 1.9753083181728394e-05, + "loss": 0.0857, + "step": 6032 + }, + { + "epoch": 0.79, + "learning_rate": 1.9752966046969067e-05, + "loss": 0.0925, + "step": 6033 + }, + { + "epoch": 0.79, + "learning_rate": 1.975284888478004e-05, + "loss": 0.0948, + "step": 6034 + }, + { + "epoch": 0.79, + "learning_rate": 1.9752731695161644e-05, + "loss": 0.0921, + "step": 6035 + }, + { + "epoch": 0.79, + "learning_rate": 1.9752614478114208e-05, + "loss": 0.0904, + "step": 6036 + }, + { + "epoch": 0.79, + "learning_rate": 1.9752497233638066e-05, + "loss": 0.0802, + "step": 6037 + }, + { + "epoch": 0.79, + "learning_rate": 1.9752379961733545e-05, + "loss": 0.1008, + "step": 6038 + }, + { + "epoch": 0.79, + "learning_rate": 1.9752262662400974e-05, + "loss": 0.0855, + "step": 6039 + }, + { + "epoch": 0.79, + "learning_rate": 1.9752145335640685e-05, + "loss": 0.0858, + "step": 6040 + }, + { + "epoch": 0.79, + "learning_rate": 1.9752027981453004e-05, + "loss": 0.0838, + "step": 6041 + }, + { + "epoch": 0.79, + "learning_rate": 1.9751910599838264e-05, + "loss": 0.0961, + "step": 6042 + }, + { + "epoch": 0.79, + "learning_rate": 1.9751793190796794e-05, + "loss": 0.0971, + "step": 6043 + }, + { + "epoch": 0.79, + "learning_rate": 1.975167575432893e-05, + "loss": 0.1037, + "step": 6044 + }, + { + "epoch": 0.79, + "learning_rate": 1.9751558290434995e-05, + "loss": 0.0801, + "step": 6045 + }, + { + "epoch": 0.79, + "learning_rate": 1.975144079911532e-05, + "loss": 0.0887, + "step": 6046 + }, + { + "epoch": 0.79, + "learning_rate": 1.9751323280370235e-05, + "loss": 0.0856, + "step": 6047 + }, + { + "epoch": 0.79, + "learning_rate": 1.9751205734200077e-05, + "loss": 0.086, + "step": 6048 + }, + { + "epoch": 0.79, + "learning_rate": 1.9751088160605167e-05, + "loss": 0.0871, + "step": 6049 + }, + { + "epoch": 0.79, + "learning_rate": 1.9750970559585844e-05, + "loss": 0.0801, + "step": 6050 + }, + { + "epoch": 0.79, + "learning_rate": 1.9750852931142438e-05, + "loss": 0.1044, + "step": 6051 + }, + { + "epoch": 0.79, + "learning_rate": 1.9750735275275272e-05, + "loss": 0.0854, + "step": 6052 + }, + { + "epoch": 0.79, + "learning_rate": 1.975061759198469e-05, + "loss": 0.0902, + "step": 6053 + }, + { + "epoch": 0.79, + "learning_rate": 1.975049988127101e-05, + "loss": 0.0871, + "step": 6054 + }, + { + "epoch": 0.79, + "learning_rate": 1.9750382143134566e-05, + "loss": 0.0857, + "step": 6055 + }, + { + "epoch": 0.79, + "learning_rate": 1.9750264377575695e-05, + "loss": 0.0876, + "step": 6056 + }, + { + "epoch": 0.79, + "learning_rate": 1.9750146584594722e-05, + "loss": 0.0758, + "step": 6057 + }, + { + "epoch": 0.79, + "learning_rate": 1.9750028764191985e-05, + "loss": 0.0978, + "step": 6058 + }, + { + "epoch": 0.79, + "learning_rate": 1.9749910916367808e-05, + "loss": 0.1025, + "step": 6059 + }, + { + "epoch": 0.79, + "learning_rate": 1.9749793041122527e-05, + "loss": 0.0923, + "step": 6060 + }, + { + "epoch": 0.79, + "learning_rate": 1.9749675138456473e-05, + "loss": 0.0893, + "step": 6061 + }, + { + "epoch": 0.79, + "learning_rate": 1.9749557208369977e-05, + "loss": 0.0875, + "step": 6062 + }, + { + "epoch": 0.79, + "learning_rate": 1.974943925086337e-05, + "loss": 0.0946, + "step": 6063 + }, + { + "epoch": 0.79, + "learning_rate": 1.9749321265936983e-05, + "loss": 0.0908, + "step": 6064 + }, + { + "epoch": 0.79, + "learning_rate": 1.974920325359115e-05, + "loss": 0.0971, + "step": 6065 + }, + { + "epoch": 0.79, + "learning_rate": 1.9749085213826203e-05, + "loss": 0.0779, + "step": 6066 + }, + { + "epoch": 0.79, + "learning_rate": 1.9748967146642474e-05, + "loss": 0.0882, + "step": 6067 + }, + { + "epoch": 0.79, + "learning_rate": 1.9748849052040292e-05, + "loss": 0.0948, + "step": 6068 + }, + { + "epoch": 0.79, + "learning_rate": 1.974873093001999e-05, + "loss": 0.0785, + "step": 6069 + }, + { + "epoch": 0.8, + "learning_rate": 1.97486127805819e-05, + "loss": 0.0811, + "step": 6070 + }, + { + "epoch": 0.8, + "learning_rate": 1.974849460372636e-05, + "loss": 0.0833, + "step": 6071 + }, + { + "epoch": 0.8, + "learning_rate": 1.9748376399453697e-05, + "loss": 0.0881, + "step": 6072 + }, + { + "epoch": 0.8, + "learning_rate": 1.9748258167764244e-05, + "loss": 0.0844, + "step": 6073 + }, + { + "epoch": 0.8, + "learning_rate": 1.9748139908658336e-05, + "loss": 0.0876, + "step": 6074 + }, + { + "epoch": 0.8, + "learning_rate": 1.9748021622136303e-05, + "loss": 0.0811, + "step": 6075 + }, + { + "epoch": 0.8, + "learning_rate": 1.9747903308198473e-05, + "loss": 0.0796, + "step": 6076 + }, + { + "epoch": 0.8, + "learning_rate": 1.974778496684519e-05, + "loss": 0.0902, + "step": 6077 + }, + { + "epoch": 0.8, + "learning_rate": 1.974766659807678e-05, + "loss": 0.0916, + "step": 6078 + }, + { + "epoch": 0.8, + "learning_rate": 1.9747548201893574e-05, + "loss": 0.0897, + "step": 6079 + }, + { + "epoch": 0.8, + "learning_rate": 1.974742977829591e-05, + "loss": 0.082, + "step": 6080 + }, + { + "epoch": 0.8, + "learning_rate": 1.974731132728412e-05, + "loss": 0.0902, + "step": 6081 + }, + { + "epoch": 0.8, + "learning_rate": 1.974719284885853e-05, + "loss": 0.1012, + "step": 6082 + }, + { + "epoch": 0.8, + "learning_rate": 1.974707434301949e-05, + "loss": 0.097, + "step": 6083 + }, + { + "epoch": 0.8, + "learning_rate": 1.9746955809767312e-05, + "loss": 0.0864, + "step": 6084 + }, + { + "epoch": 0.8, + "learning_rate": 1.9746837249102344e-05, + "loss": 0.0937, + "step": 6085 + }, + { + "epoch": 0.8, + "learning_rate": 1.9746718661024916e-05, + "loss": 0.0867, + "step": 6086 + }, + { + "epoch": 0.8, + "learning_rate": 1.974660004553536e-05, + "loss": 0.0878, + "step": 6087 + }, + { + "epoch": 0.8, + "learning_rate": 1.9746481402634007e-05, + "loss": 0.088, + "step": 6088 + }, + { + "epoch": 0.8, + "learning_rate": 1.97463627323212e-05, + "loss": 0.0869, + "step": 6089 + }, + { + "epoch": 0.8, + "learning_rate": 1.9746244034597265e-05, + "loss": 0.0925, + "step": 6090 + }, + { + "epoch": 0.8, + "learning_rate": 1.9746125309462538e-05, + "loss": 0.09, + "step": 6091 + }, + { + "epoch": 0.8, + "learning_rate": 1.974600655691735e-05, + "loss": 0.0863, + "step": 6092 + }, + { + "epoch": 0.8, + "learning_rate": 1.974588777696204e-05, + "loss": 0.0833, + "step": 6093 + }, + { + "epoch": 0.8, + "learning_rate": 1.9745768969596937e-05, + "loss": 0.0785, + "step": 6094 + }, + { + "epoch": 0.8, + "learning_rate": 1.974565013482238e-05, + "loss": 0.0999, + "step": 6095 + }, + { + "epoch": 0.8, + "learning_rate": 1.97455312726387e-05, + "loss": 0.0852, + "step": 6096 + }, + { + "epoch": 0.8, + "learning_rate": 1.974541238304623e-05, + "loss": 0.0812, + "step": 6097 + }, + { + "epoch": 0.8, + "learning_rate": 1.9745293466045313e-05, + "loss": 0.0972, + "step": 6098 + }, + { + "epoch": 0.8, + "learning_rate": 1.9745174521636273e-05, + "loss": 0.0916, + "step": 6099 + }, + { + "epoch": 0.8, + "learning_rate": 1.974505554981945e-05, + "loss": 0.0853, + "step": 6100 + }, + { + "epoch": 0.8, + "learning_rate": 1.9744936550595176e-05, + "loss": 0.0899, + "step": 6101 + }, + { + "epoch": 0.8, + "learning_rate": 1.9744817523963788e-05, + "loss": 0.0922, + "step": 6102 + }, + { + "epoch": 0.8, + "learning_rate": 1.9744698469925618e-05, + "loss": 0.0747, + "step": 6103 + }, + { + "epoch": 0.8, + "learning_rate": 1.9744579388481005e-05, + "loss": 0.0728, + "step": 6104 + }, + { + "epoch": 0.8, + "learning_rate": 1.9744460279630276e-05, + "loss": 0.0893, + "step": 6105 + }, + { + "epoch": 0.8, + "learning_rate": 1.9744341143373776e-05, + "loss": 0.0807, + "step": 6106 + }, + { + "epoch": 0.8, + "learning_rate": 1.9744221979711836e-05, + "loss": 0.0946, + "step": 6107 + }, + { + "epoch": 0.8, + "learning_rate": 1.974410278864479e-05, + "loss": 0.0788, + "step": 6108 + }, + { + "epoch": 0.8, + "learning_rate": 1.974398357017297e-05, + "loss": 0.0915, + "step": 6109 + }, + { + "epoch": 0.8, + "learning_rate": 1.974386432429672e-05, + "loss": 0.0925, + "step": 6110 + }, + { + "epoch": 0.8, + "learning_rate": 1.9743745051016366e-05, + "loss": 0.0816, + "step": 6111 + }, + { + "epoch": 0.8, + "learning_rate": 1.9743625750332254e-05, + "loss": 0.0893, + "step": 6112 + }, + { + "epoch": 0.8, + "learning_rate": 1.974350642224471e-05, + "loss": 0.0903, + "step": 6113 + }, + { + "epoch": 0.8, + "learning_rate": 1.9743387066754073e-05, + "loss": 0.0953, + "step": 6114 + }, + { + "epoch": 0.8, + "learning_rate": 1.9743267683860677e-05, + "loss": 0.0892, + "step": 6115 + }, + { + "epoch": 0.8, + "learning_rate": 1.9743148273564864e-05, + "loss": 0.089, + "step": 6116 + }, + { + "epoch": 0.8, + "learning_rate": 1.9743028835866963e-05, + "loss": 0.0841, + "step": 6117 + }, + { + "epoch": 0.8, + "learning_rate": 1.974290937076731e-05, + "loss": 0.0859, + "step": 6118 + }, + { + "epoch": 0.8, + "learning_rate": 1.9742789878266244e-05, + "loss": 0.0914, + "step": 6119 + }, + { + "epoch": 0.8, + "learning_rate": 1.9742670358364103e-05, + "loss": 0.0811, + "step": 6120 + }, + { + "epoch": 0.8, + "learning_rate": 1.974255081106122e-05, + "loss": 0.0785, + "step": 6121 + }, + { + "epoch": 0.8, + "learning_rate": 1.9742431236357928e-05, + "loss": 0.0839, + "step": 6122 + }, + { + "epoch": 0.8, + "learning_rate": 1.974231163425457e-05, + "loss": 0.0861, + "step": 6123 + }, + { + "epoch": 0.8, + "learning_rate": 1.974219200475148e-05, + "loss": 0.0853, + "step": 6124 + }, + { + "epoch": 0.8, + "learning_rate": 1.974207234784899e-05, + "loss": 0.0944, + "step": 6125 + }, + { + "epoch": 0.8, + "learning_rate": 1.9741952663547444e-05, + "loss": 0.0931, + "step": 6126 + }, + { + "epoch": 0.8, + "learning_rate": 1.9741832951847175e-05, + "loss": 0.0988, + "step": 6127 + }, + { + "epoch": 0.8, + "learning_rate": 1.9741713212748517e-05, + "loss": 0.0951, + "step": 6128 + }, + { + "epoch": 0.8, + "learning_rate": 1.9741593446251812e-05, + "loss": 0.0836, + "step": 6129 + }, + { + "epoch": 0.8, + "learning_rate": 1.9741473652357393e-05, + "loss": 0.0792, + "step": 6130 + }, + { + "epoch": 0.8, + "learning_rate": 1.97413538310656e-05, + "loss": 0.0897, + "step": 6131 + }, + { + "epoch": 0.8, + "learning_rate": 1.9741233982376764e-05, + "loss": 0.0737, + "step": 6132 + }, + { + "epoch": 0.8, + "learning_rate": 1.974111410629123e-05, + "loss": 0.0962, + "step": 6133 + }, + { + "epoch": 0.8, + "learning_rate": 1.9740994202809327e-05, + "loss": 0.1002, + "step": 6134 + }, + { + "epoch": 0.8, + "learning_rate": 1.97408742719314e-05, + "loss": 0.0935, + "step": 6135 + }, + { + "epoch": 0.8, + "learning_rate": 1.974075431365778e-05, + "loss": 0.0839, + "step": 6136 + }, + { + "epoch": 0.8, + "learning_rate": 1.974063432798881e-05, + "loss": 0.0704, + "step": 6137 + }, + { + "epoch": 0.8, + "learning_rate": 1.9740514314924827e-05, + "loss": 0.0751, + "step": 6138 + }, + { + "epoch": 0.8, + "learning_rate": 1.974039427446616e-05, + "loss": 0.0825, + "step": 6139 + }, + { + "epoch": 0.8, + "learning_rate": 1.9740274206613156e-05, + "loss": 0.0987, + "step": 6140 + }, + { + "epoch": 0.8, + "learning_rate": 1.9740154111366152e-05, + "loss": 0.0857, + "step": 6141 + }, + { + "epoch": 0.8, + "learning_rate": 1.974003398872548e-05, + "loss": 0.0772, + "step": 6142 + }, + { + "epoch": 0.8, + "learning_rate": 1.973991383869148e-05, + "loss": 0.0886, + "step": 6143 + }, + { + "epoch": 0.8, + "learning_rate": 1.9739793661264495e-05, + "loss": 0.0868, + "step": 6144 + }, + { + "epoch": 0.8, + "learning_rate": 1.9739673456444855e-05, + "loss": 0.0811, + "step": 6145 + }, + { + "epoch": 0.81, + "learning_rate": 1.9739553224232906e-05, + "loss": 0.0943, + "step": 6146 + }, + { + "epoch": 0.81, + "learning_rate": 1.9739432964628978e-05, + "loss": 0.0787, + "step": 6147 + }, + { + "epoch": 0.81, + "learning_rate": 1.9739312677633417e-05, + "loss": 0.0793, + "step": 6148 + }, + { + "epoch": 0.81, + "learning_rate": 1.9739192363246556e-05, + "loss": 0.0837, + "step": 6149 + }, + { + "epoch": 0.81, + "learning_rate": 1.9739072021468737e-05, + "loss": 0.0782, + "step": 6150 + }, + { + "epoch": 0.81, + "learning_rate": 1.9738951652300294e-05, + "loss": 0.0909, + "step": 6151 + }, + { + "epoch": 0.81, + "learning_rate": 1.9738831255741567e-05, + "loss": 0.0811, + "step": 6152 + }, + { + "epoch": 0.81, + "learning_rate": 1.97387108317929e-05, + "loss": 0.104, + "step": 6153 + }, + { + "epoch": 0.81, + "learning_rate": 1.9738590380454627e-05, + "loss": 0.0879, + "step": 6154 + }, + { + "epoch": 0.81, + "learning_rate": 1.9738469901727084e-05, + "loss": 0.097, + "step": 6155 + }, + { + "epoch": 0.81, + "learning_rate": 1.9738349395610615e-05, + "loss": 0.0786, + "step": 6156 + }, + { + "epoch": 0.81, + "learning_rate": 1.9738228862105556e-05, + "loss": 0.0943, + "step": 6157 + }, + { + "epoch": 0.81, + "learning_rate": 1.9738108301212246e-05, + "loss": 0.0864, + "step": 6158 + }, + { + "epoch": 0.81, + "learning_rate": 1.9737987712931027e-05, + "loss": 0.0705, + "step": 6159 + }, + { + "epoch": 0.81, + "learning_rate": 1.9737867097262235e-05, + "loss": 0.08, + "step": 6160 + }, + { + "epoch": 0.81, + "learning_rate": 1.9737746454206214e-05, + "loss": 0.0723, + "step": 6161 + }, + { + "epoch": 0.81, + "learning_rate": 1.9737625783763293e-05, + "loss": 0.0937, + "step": 6162 + }, + { + "epoch": 0.81, + "learning_rate": 1.9737505085933823e-05, + "loss": 0.0895, + "step": 6163 + }, + { + "epoch": 0.81, + "learning_rate": 1.9737384360718133e-05, + "loss": 0.0912, + "step": 6164 + }, + { + "epoch": 0.81, + "learning_rate": 1.9737263608116575e-05, + "loss": 0.0982, + "step": 6165 + }, + { + "epoch": 0.81, + "learning_rate": 1.973714282812948e-05, + "loss": 0.0958, + "step": 6166 + }, + { + "epoch": 0.81, + "learning_rate": 1.9737022020757187e-05, + "loss": 0.0895, + "step": 6167 + }, + { + "epoch": 0.81, + "learning_rate": 1.9736901186000036e-05, + "loss": 0.0916, + "step": 6168 + }, + { + "epoch": 0.81, + "learning_rate": 1.9736780323858374e-05, + "loss": 0.0809, + "step": 6169 + }, + { + "epoch": 0.81, + "learning_rate": 1.9736659434332534e-05, + "loss": 0.0972, + "step": 6170 + }, + { + "epoch": 0.81, + "learning_rate": 1.973653851742286e-05, + "loss": 0.0999, + "step": 6171 + }, + { + "epoch": 0.81, + "learning_rate": 1.9736417573129688e-05, + "loss": 0.0819, + "step": 6172 + }, + { + "epoch": 0.81, + "learning_rate": 1.9736296601453358e-05, + "loss": 0.0967, + "step": 6173 + }, + { + "epoch": 0.81, + "learning_rate": 1.9736175602394214e-05, + "loss": 0.0865, + "step": 6174 + }, + { + "epoch": 0.81, + "learning_rate": 1.9736054575952592e-05, + "loss": 0.0864, + "step": 6175 + }, + { + "epoch": 0.81, + "learning_rate": 1.973593352212884e-05, + "loss": 0.0863, + "step": 6176 + }, + { + "epoch": 0.81, + "learning_rate": 1.973581244092329e-05, + "loss": 0.0796, + "step": 6177 + }, + { + "epoch": 0.81, + "learning_rate": 1.9735691332336287e-05, + "loss": 0.0784, + "step": 6178 + }, + { + "epoch": 0.81, + "learning_rate": 1.9735570196368172e-05, + "loss": 0.0806, + "step": 6179 + }, + { + "epoch": 0.81, + "learning_rate": 1.973544903301928e-05, + "loss": 0.0926, + "step": 6180 + }, + { + "epoch": 0.81, + "learning_rate": 1.973532784228996e-05, + "loss": 0.0787, + "step": 6181 + }, + { + "epoch": 0.81, + "learning_rate": 1.973520662418055e-05, + "loss": 0.0814, + "step": 6182 + }, + { + "epoch": 0.81, + "learning_rate": 1.9735085378691384e-05, + "loss": 0.0829, + "step": 6183 + }, + { + "epoch": 0.81, + "learning_rate": 1.9734964105822816e-05, + "loss": 0.0858, + "step": 6184 + }, + { + "epoch": 0.81, + "learning_rate": 1.9734842805575175e-05, + "loss": 0.0973, + "step": 6185 + }, + { + "epoch": 0.81, + "learning_rate": 1.973472147794881e-05, + "loss": 0.0913, + "step": 6186 + }, + { + "epoch": 0.81, + "learning_rate": 1.9734600122944053e-05, + "loss": 0.0909, + "step": 6187 + }, + { + "epoch": 0.81, + "learning_rate": 1.9734478740561257e-05, + "loss": 0.0902, + "step": 6188 + }, + { + "epoch": 0.81, + "learning_rate": 1.9734357330800756e-05, + "loss": 0.0726, + "step": 6189 + }, + { + "epoch": 0.81, + "learning_rate": 1.9734235893662898e-05, + "loss": 0.0953, + "step": 6190 + }, + { + "epoch": 0.81, + "learning_rate": 1.9734114429148016e-05, + "loss": 0.0837, + "step": 6191 + }, + { + "epoch": 0.81, + "learning_rate": 1.9733992937256453e-05, + "loss": 0.0969, + "step": 6192 + }, + { + "epoch": 0.81, + "learning_rate": 1.973387141798856e-05, + "loss": 0.0885, + "step": 6193 + }, + { + "epoch": 0.81, + "learning_rate": 1.9733749871344664e-05, + "loss": 0.0876, + "step": 6194 + }, + { + "epoch": 0.81, + "learning_rate": 1.973362829732512e-05, + "loss": 0.092, + "step": 6195 + }, + { + "epoch": 0.81, + "learning_rate": 1.9733506695930265e-05, + "loss": 0.0825, + "step": 6196 + }, + { + "epoch": 0.81, + "learning_rate": 1.973338506716044e-05, + "loss": 0.0817, + "step": 6197 + }, + { + "epoch": 0.81, + "learning_rate": 1.9733263411015987e-05, + "loss": 0.0852, + "step": 6198 + }, + { + "epoch": 0.81, + "learning_rate": 1.9733141727497247e-05, + "loss": 0.0843, + "step": 6199 + }, + { + "epoch": 0.81, + "learning_rate": 1.9733020016604567e-05, + "loss": 0.0903, + "step": 6200 + }, + { + "epoch": 0.81, + "learning_rate": 1.9732898278338286e-05, + "loss": 0.0869, + "step": 6201 + }, + { + "epoch": 0.81, + "learning_rate": 1.973277651269875e-05, + "loss": 0.0945, + "step": 6202 + }, + { + "epoch": 0.81, + "learning_rate": 1.9732654719686293e-05, + "loss": 0.0966, + "step": 6203 + }, + { + "epoch": 0.81, + "learning_rate": 1.9732532899301265e-05, + "loss": 0.0883, + "step": 6204 + }, + { + "epoch": 0.81, + "learning_rate": 1.9732411051544006e-05, + "loss": 0.078, + "step": 6205 + }, + { + "epoch": 0.81, + "learning_rate": 1.973228917641486e-05, + "loss": 0.1024, + "step": 6206 + }, + { + "epoch": 0.81, + "learning_rate": 1.973216727391417e-05, + "loss": 0.0835, + "step": 6207 + }, + { + "epoch": 0.81, + "learning_rate": 1.9732045344042276e-05, + "loss": 0.0817, + "step": 6208 + }, + { + "epoch": 0.81, + "learning_rate": 1.9731923386799523e-05, + "loss": 0.0847, + "step": 6209 + }, + { + "epoch": 0.81, + "learning_rate": 1.9731801402186255e-05, + "loss": 0.0847, + "step": 6210 + }, + { + "epoch": 0.81, + "learning_rate": 1.9731679390202812e-05, + "loss": 0.088, + "step": 6211 + }, + { + "epoch": 0.81, + "learning_rate": 1.973155735084954e-05, + "loss": 0.0857, + "step": 6212 + }, + { + "epoch": 0.81, + "learning_rate": 1.973143528412678e-05, + "loss": 0.0877, + "step": 6213 + }, + { + "epoch": 0.81, + "learning_rate": 1.973131319003488e-05, + "loss": 0.0944, + "step": 6214 + }, + { + "epoch": 0.81, + "learning_rate": 1.9731191068574175e-05, + "loss": 0.0942, + "step": 6215 + }, + { + "epoch": 0.81, + "learning_rate": 1.9731068919745014e-05, + "loss": 0.0823, + "step": 6216 + }, + { + "epoch": 0.81, + "learning_rate": 1.9730946743547744e-05, + "loss": 0.0861, + "step": 6217 + }, + { + "epoch": 0.81, + "learning_rate": 1.9730824539982704e-05, + "loss": 0.0813, + "step": 6218 + }, + { + "epoch": 0.81, + "learning_rate": 1.9730702309050236e-05, + "loss": 0.0918, + "step": 6219 + }, + { + "epoch": 0.81, + "learning_rate": 1.9730580050750686e-05, + "loss": 0.1074, + "step": 6220 + }, + { + "epoch": 0.81, + "learning_rate": 1.9730457765084396e-05, + "loss": 0.1072, + "step": 6221 + }, + { + "epoch": 0.82, + "learning_rate": 1.9730335452051715e-05, + "loss": 0.0902, + "step": 6222 + }, + { + "epoch": 0.82, + "learning_rate": 1.973021311165298e-05, + "loss": 0.0867, + "step": 6223 + }, + { + "epoch": 0.82, + "learning_rate": 1.973009074388854e-05, + "loss": 0.0903, + "step": 6224 + }, + { + "epoch": 0.82, + "learning_rate": 1.972996834875874e-05, + "loss": 0.0862, + "step": 6225 + }, + { + "epoch": 0.82, + "learning_rate": 1.9729845926263923e-05, + "loss": 0.0767, + "step": 6226 + }, + { + "epoch": 0.82, + "learning_rate": 1.9729723476404433e-05, + "loss": 0.0823, + "step": 6227 + }, + { + "epoch": 0.82, + "learning_rate": 1.972960099918061e-05, + "loss": 0.0938, + "step": 6228 + }, + { + "epoch": 0.82, + "learning_rate": 1.9729478494592805e-05, + "loss": 0.1019, + "step": 6229 + }, + { + "epoch": 0.82, + "learning_rate": 1.9729355962641355e-05, + "loss": 0.0847, + "step": 6230 + }, + { + "epoch": 0.82, + "learning_rate": 1.9729233403326614e-05, + "loss": 0.0889, + "step": 6231 + }, + { + "epoch": 0.82, + "learning_rate": 1.972911081664892e-05, + "loss": 0.0866, + "step": 6232 + }, + { + "epoch": 0.82, + "learning_rate": 1.972898820260862e-05, + "loss": 0.097, + "step": 6233 + }, + { + "epoch": 0.82, + "learning_rate": 1.9728865561206065e-05, + "loss": 0.087, + "step": 6234 + }, + { + "epoch": 0.82, + "learning_rate": 1.972874289244159e-05, + "loss": 0.0737, + "step": 6235 + }, + { + "epoch": 0.82, + "learning_rate": 1.972862019631554e-05, + "loss": 0.0883, + "step": 6236 + }, + { + "epoch": 0.82, + "learning_rate": 1.9728497472828266e-05, + "loss": 0.0792, + "step": 6237 + }, + { + "epoch": 0.82, + "learning_rate": 1.972837472198011e-05, + "loss": 0.0809, + "step": 6238 + }, + { + "epoch": 0.82, + "learning_rate": 1.9728251943771423e-05, + "loss": 0.0944, + "step": 6239 + }, + { + "epoch": 0.82, + "learning_rate": 1.972812913820254e-05, + "loss": 0.0875, + "step": 6240 + }, + { + "epoch": 0.82, + "learning_rate": 1.9728006305273816e-05, + "loss": 0.0868, + "step": 6241 + }, + { + "epoch": 0.82, + "learning_rate": 1.972788344498559e-05, + "loss": 0.0797, + "step": 6242 + }, + { + "epoch": 0.82, + "learning_rate": 1.9727760557338208e-05, + "loss": 0.0999, + "step": 6243 + }, + { + "epoch": 0.82, + "learning_rate": 1.972763764233202e-05, + "loss": 0.1013, + "step": 6244 + }, + { + "epoch": 0.82, + "learning_rate": 1.972751469996737e-05, + "loss": 0.0879, + "step": 6245 + }, + { + "epoch": 0.82, + "learning_rate": 1.97273917302446e-05, + "loss": 0.0865, + "step": 6246 + }, + { + "epoch": 0.82, + "learning_rate": 1.972726873316406e-05, + "loss": 0.0911, + "step": 6247 + }, + { + "epoch": 0.82, + "learning_rate": 1.9727145708726097e-05, + "loss": 0.088, + "step": 6248 + }, + { + "epoch": 0.82, + "learning_rate": 1.9727022656931052e-05, + "loss": 0.0865, + "step": 6249 + }, + { + "epoch": 0.82, + "learning_rate": 1.9726899577779272e-05, + "loss": 0.0845, + "step": 6250 + }, + { + "epoch": 0.82, + "learning_rate": 1.972677647127111e-05, + "loss": 0.0872, + "step": 6251 + }, + { + "epoch": 0.82, + "learning_rate": 1.9726653337406904e-05, + "loss": 0.0954, + "step": 6252 + }, + { + "epoch": 0.82, + "learning_rate": 1.9726530176187e-05, + "loss": 0.1013, + "step": 6253 + }, + { + "epoch": 0.82, + "learning_rate": 1.9726406987611754e-05, + "loss": 0.099, + "step": 6254 + }, + { + "epoch": 0.82, + "learning_rate": 1.9726283771681504e-05, + "loss": 0.086, + "step": 6255 + }, + { + "epoch": 0.82, + "learning_rate": 1.97261605283966e-05, + "loss": 0.0829, + "step": 6256 + }, + { + "epoch": 0.82, + "learning_rate": 1.9726037257757386e-05, + "loss": 0.0803, + "step": 6257 + }, + { + "epoch": 0.82, + "learning_rate": 1.9725913959764212e-05, + "loss": 0.0901, + "step": 6258 + }, + { + "epoch": 0.82, + "learning_rate": 1.972579063441742e-05, + "loss": 0.08, + "step": 6259 + }, + { + "epoch": 0.82, + "learning_rate": 1.972566728171736e-05, + "loss": 0.0992, + "step": 6260 + }, + { + "epoch": 0.82, + "learning_rate": 1.972554390166438e-05, + "loss": 0.092, + "step": 6261 + }, + { + "epoch": 0.82, + "learning_rate": 1.972542049425883e-05, + "loss": 0.0864, + "step": 6262 + }, + { + "epoch": 0.82, + "learning_rate": 1.9725297059501047e-05, + "loss": 0.0892, + "step": 6263 + }, + { + "epoch": 0.82, + "learning_rate": 1.9725173597391384e-05, + "loss": 0.1059, + "step": 6264 + }, + { + "epoch": 0.82, + "learning_rate": 1.972505010793019e-05, + "loss": 0.089, + "step": 6265 + }, + { + "epoch": 0.82, + "learning_rate": 1.972492659111781e-05, + "loss": 0.09, + "step": 6266 + }, + { + "epoch": 0.82, + "learning_rate": 1.9724803046954593e-05, + "loss": 0.1003, + "step": 6267 + }, + { + "epoch": 0.82, + "learning_rate": 1.972467947544088e-05, + "loss": 0.0941, + "step": 6268 + }, + { + "epoch": 0.82, + "learning_rate": 1.972455587657703e-05, + "loss": 0.0894, + "step": 6269 + }, + { + "epoch": 0.82, + "learning_rate": 1.9724432250363384e-05, + "loss": 0.0849, + "step": 6270 + }, + { + "epoch": 0.82, + "learning_rate": 1.9724308596800287e-05, + "loss": 0.0785, + "step": 6271 + }, + { + "epoch": 0.82, + "learning_rate": 1.9724184915888092e-05, + "loss": 0.0875, + "step": 6272 + }, + { + "epoch": 0.82, + "learning_rate": 1.9724061207627145e-05, + "loss": 0.0782, + "step": 6273 + }, + { + "epoch": 0.82, + "learning_rate": 1.9723937472017794e-05, + "loss": 0.0978, + "step": 6274 + }, + { + "epoch": 0.82, + "learning_rate": 1.9723813709060385e-05, + "loss": 0.0835, + "step": 6275 + }, + { + "epoch": 0.82, + "learning_rate": 1.9723689918755268e-05, + "loss": 0.0865, + "step": 6276 + }, + { + "epoch": 0.82, + "learning_rate": 1.972356610110279e-05, + "loss": 0.0923, + "step": 6277 + }, + { + "epoch": 0.82, + "learning_rate": 1.97234422561033e-05, + "loss": 0.0988, + "step": 6278 + }, + { + "epoch": 0.82, + "learning_rate": 1.972331838375715e-05, + "loss": 0.0929, + "step": 6279 + }, + { + "epoch": 0.82, + "learning_rate": 1.9723194484064684e-05, + "loss": 0.0843, + "step": 6280 + }, + { + "epoch": 0.82, + "learning_rate": 1.972307055702625e-05, + "loss": 0.0837, + "step": 6281 + }, + { + "epoch": 0.82, + "learning_rate": 1.9722946602642197e-05, + "loss": 0.0976, + "step": 6282 + }, + { + "epoch": 0.82, + "learning_rate": 1.9722822620912874e-05, + "loss": 0.0885, + "step": 6283 + }, + { + "epoch": 0.82, + "learning_rate": 1.9722698611838635e-05, + "loss": 0.0877, + "step": 6284 + }, + { + "epoch": 0.82, + "learning_rate": 1.972257457541982e-05, + "loss": 0.0886, + "step": 6285 + }, + { + "epoch": 0.82, + "learning_rate": 1.972245051165678e-05, + "loss": 0.0898, + "step": 6286 + }, + { + "epoch": 0.82, + "learning_rate": 1.972232642054987e-05, + "loss": 0.0853, + "step": 6287 + }, + { + "epoch": 0.82, + "learning_rate": 1.972220230209943e-05, + "loss": 0.0853, + "step": 6288 + }, + { + "epoch": 0.82, + "learning_rate": 1.972207815630582e-05, + "loss": 0.0855, + "step": 6289 + }, + { + "epoch": 0.82, + "learning_rate": 1.9721953983169377e-05, + "loss": 0.077, + "step": 6290 + }, + { + "epoch": 0.82, + "learning_rate": 1.9721829782690462e-05, + "loss": 0.0807, + "step": 6291 + }, + { + "epoch": 0.82, + "learning_rate": 1.972170555486941e-05, + "loss": 0.0921, + "step": 6292 + }, + { + "epoch": 0.82, + "learning_rate": 1.972158129970659e-05, + "loss": 0.0884, + "step": 6293 + }, + { + "epoch": 0.82, + "learning_rate": 1.972145701720233e-05, + "loss": 0.0866, + "step": 6294 + }, + { + "epoch": 0.82, + "learning_rate": 1.9721332707356992e-05, + "loss": 0.0841, + "step": 6295 + }, + { + "epoch": 0.82, + "learning_rate": 1.9721208370170927e-05, + "loss": 0.0912, + "step": 6296 + }, + { + "epoch": 0.82, + "learning_rate": 1.9721084005644478e-05, + "loss": 0.0893, + "step": 6297 + }, + { + "epoch": 0.82, + "learning_rate": 1.9720959613778e-05, + "loss": 0.0826, + "step": 6298 + }, + { + "epoch": 0.83, + "learning_rate": 1.9720835194571842e-05, + "loss": 0.0997, + "step": 6299 + }, + { + "epoch": 0.83, + "learning_rate": 1.972071074802635e-05, + "loss": 0.0859, + "step": 6300 + }, + { + "epoch": 0.83, + "learning_rate": 1.9720586274141875e-05, + "loss": 0.09, + "step": 6301 + }, + { + "epoch": 0.83, + "learning_rate": 1.9720461772918773e-05, + "loss": 0.0868, + "step": 6302 + }, + { + "epoch": 0.83, + "learning_rate": 1.9720337244357387e-05, + "loss": 0.089, + "step": 6303 + }, + { + "epoch": 0.83, + "learning_rate": 1.9720212688458074e-05, + "loss": 0.084, + "step": 6304 + }, + { + "epoch": 0.83, + "learning_rate": 1.9720088105221175e-05, + "loss": 0.1029, + "step": 6305 + }, + { + "epoch": 0.83, + "learning_rate": 1.971996349464705e-05, + "loss": 0.0935, + "step": 6306 + }, + { + "epoch": 0.83, + "learning_rate": 1.9719838856736043e-05, + "loss": 0.0888, + "step": 6307 + }, + { + "epoch": 0.83, + "learning_rate": 1.9719714191488507e-05, + "loss": 0.0746, + "step": 6308 + }, + { + "epoch": 0.83, + "learning_rate": 1.9719589498904795e-05, + "loss": 0.0879, + "step": 6309 + }, + { + "epoch": 0.83, + "learning_rate": 1.9719464778985252e-05, + "loss": 0.0893, + "step": 6310 + }, + { + "epoch": 0.83, + "learning_rate": 1.9719340031730232e-05, + "loss": 0.0933, + "step": 6311 + }, + { + "epoch": 0.83, + "learning_rate": 1.971921525714009e-05, + "loss": 0.0842, + "step": 6312 + }, + { + "epoch": 0.83, + "learning_rate": 1.971909045521517e-05, + "loss": 0.0923, + "step": 6313 + }, + { + "epoch": 0.83, + "learning_rate": 1.9718965625955823e-05, + "loss": 0.0903, + "step": 6314 + }, + { + "epoch": 0.83, + "learning_rate": 1.9718840769362406e-05, + "loss": 0.0868, + "step": 6315 + }, + { + "epoch": 0.83, + "learning_rate": 1.9718715885435265e-05, + "loss": 0.0808, + "step": 6316 + }, + { + "epoch": 0.83, + "learning_rate": 1.9718590974174755e-05, + "loss": 0.087, + "step": 6317 + }, + { + "epoch": 0.83, + "learning_rate": 1.9718466035581226e-05, + "loss": 0.0892, + "step": 6318 + }, + { + "epoch": 0.83, + "learning_rate": 1.9718341069655025e-05, + "loss": 0.0917, + "step": 6319 + }, + { + "epoch": 0.83, + "learning_rate": 1.971821607639651e-05, + "loss": 0.102, + "step": 6320 + }, + { + "epoch": 0.83, + "learning_rate": 1.971809105580603e-05, + "loss": 0.0885, + "step": 6321 + }, + { + "epoch": 0.83, + "learning_rate": 1.9717966007883932e-05, + "loss": 0.0926, + "step": 6322 + }, + { + "epoch": 0.83, + "learning_rate": 1.971784093263058e-05, + "loss": 0.0892, + "step": 6323 + }, + { + "epoch": 0.83, + "learning_rate": 1.9717715830046315e-05, + "loss": 0.0895, + "step": 6324 + }, + { + "epoch": 0.83, + "learning_rate": 1.971759070013149e-05, + "loss": 0.0956, + "step": 6325 + }, + { + "epoch": 0.83, + "learning_rate": 1.971746554288646e-05, + "loss": 0.085, + "step": 6326 + }, + { + "epoch": 0.83, + "learning_rate": 1.971734035831157e-05, + "loss": 0.0697, + "step": 6327 + }, + { + "epoch": 0.83, + "learning_rate": 1.971721514640718e-05, + "loss": 0.0866, + "step": 6328 + }, + { + "epoch": 0.83, + "learning_rate": 1.9717089907173646e-05, + "loss": 0.0929, + "step": 6329 + }, + { + "epoch": 0.83, + "learning_rate": 1.9716964640611312e-05, + "loss": 0.0833, + "step": 6330 + }, + { + "epoch": 0.83, + "learning_rate": 1.971683934672053e-05, + "loss": 0.0973, + "step": 6331 + }, + { + "epoch": 0.83, + "learning_rate": 1.9716714025501657e-05, + "loss": 0.0834, + "step": 6332 + }, + { + "epoch": 0.83, + "learning_rate": 1.971658867695504e-05, + "loss": 0.0864, + "step": 6333 + }, + { + "epoch": 0.83, + "learning_rate": 1.9716463301081038e-05, + "loss": 0.0813, + "step": 6334 + }, + { + "epoch": 0.83, + "learning_rate": 1.9716337897879997e-05, + "loss": 0.0935, + "step": 6335 + }, + { + "epoch": 0.83, + "learning_rate": 1.9716212467352277e-05, + "loss": 0.0867, + "step": 6336 + }, + { + "epoch": 0.83, + "learning_rate": 1.9716087009498225e-05, + "loss": 0.0797, + "step": 6337 + }, + { + "epoch": 0.83, + "learning_rate": 1.9715961524318194e-05, + "loss": 0.0935, + "step": 6338 + }, + { + "epoch": 0.83, + "learning_rate": 1.9715836011812542e-05, + "loss": 0.0926, + "step": 6339 + }, + { + "epoch": 0.83, + "learning_rate": 1.9715710471981614e-05, + "loss": 0.089, + "step": 6340 + }, + { + "epoch": 0.83, + "learning_rate": 1.9715584904825772e-05, + "loss": 0.0896, + "step": 6341 + }, + { + "epoch": 0.83, + "learning_rate": 1.9715459310345363e-05, + "loss": 0.0846, + "step": 6342 + }, + { + "epoch": 0.83, + "learning_rate": 1.9715333688540744e-05, + "loss": 0.0957, + "step": 6343 + }, + { + "epoch": 0.83, + "learning_rate": 1.9715208039412266e-05, + "loss": 0.0812, + "step": 6344 + }, + { + "epoch": 0.83, + "learning_rate": 1.971508236296028e-05, + "loss": 0.0825, + "step": 6345 + }, + { + "epoch": 0.83, + "learning_rate": 1.9714956659185146e-05, + "loss": 0.0793, + "step": 6346 + }, + { + "epoch": 0.83, + "learning_rate": 1.971483092808721e-05, + "loss": 0.0922, + "step": 6347 + }, + { + "epoch": 0.83, + "learning_rate": 1.971470516966683e-05, + "loss": 0.0771, + "step": 6348 + }, + { + "epoch": 0.83, + "learning_rate": 1.9714579383924363e-05, + "loss": 0.0866, + "step": 6349 + }, + { + "epoch": 0.83, + "learning_rate": 1.9714453570860152e-05, + "loss": 0.0917, + "step": 6350 + }, + { + "epoch": 0.83, + "learning_rate": 1.9714327730474565e-05, + "loss": 0.0853, + "step": 6351 + }, + { + "epoch": 0.83, + "learning_rate": 1.9714201862767944e-05, + "loss": 0.0914, + "step": 6352 + }, + { + "epoch": 0.83, + "learning_rate": 1.9714075967740648e-05, + "loss": 0.0749, + "step": 6353 + }, + { + "epoch": 0.83, + "learning_rate": 1.9713950045393032e-05, + "loss": 0.0958, + "step": 6354 + }, + { + "epoch": 0.83, + "learning_rate": 1.9713824095725446e-05, + "loss": 0.0934, + "step": 6355 + }, + { + "epoch": 0.83, + "learning_rate": 1.9713698118738253e-05, + "loss": 0.0905, + "step": 6356 + }, + { + "epoch": 0.83, + "learning_rate": 1.9713572114431794e-05, + "loss": 0.0852, + "step": 6357 + }, + { + "epoch": 0.83, + "learning_rate": 1.9713446082806435e-05, + "loss": 0.0801, + "step": 6358 + }, + { + "epoch": 0.83, + "learning_rate": 1.9713320023862525e-05, + "loss": 0.0951, + "step": 6359 + }, + { + "epoch": 0.83, + "learning_rate": 1.971319393760042e-05, + "loss": 0.0875, + "step": 6360 + }, + { + "epoch": 0.83, + "learning_rate": 1.9713067824020474e-05, + "loss": 0.0816, + "step": 6361 + }, + { + "epoch": 0.83, + "learning_rate": 1.9712941683123044e-05, + "loss": 0.0891, + "step": 6362 + }, + { + "epoch": 0.83, + "learning_rate": 1.9712815514908482e-05, + "loss": 0.0849, + "step": 6363 + }, + { + "epoch": 0.83, + "learning_rate": 1.9712689319377142e-05, + "loss": 0.084, + "step": 6364 + }, + { + "epoch": 0.83, + "learning_rate": 1.971256309652938e-05, + "loss": 0.0798, + "step": 6365 + }, + { + "epoch": 0.83, + "learning_rate": 1.9712436846365556e-05, + "loss": 0.086, + "step": 6366 + }, + { + "epoch": 0.83, + "learning_rate": 1.9712310568886015e-05, + "loss": 0.0964, + "step": 6367 + }, + { + "epoch": 0.83, + "learning_rate": 1.971218426409112e-05, + "loss": 0.0911, + "step": 6368 + }, + { + "epoch": 0.83, + "learning_rate": 1.9712057931981224e-05, + "loss": 0.0822, + "step": 6369 + }, + { + "epoch": 0.83, + "learning_rate": 1.9711931572556683e-05, + "loss": 0.0858, + "step": 6370 + }, + { + "epoch": 0.83, + "learning_rate": 1.971180518581785e-05, + "loss": 0.0883, + "step": 6371 + }, + { + "epoch": 0.83, + "learning_rate": 1.9711678771765084e-05, + "loss": 0.1002, + "step": 6372 + }, + { + "epoch": 0.83, + "learning_rate": 1.9711552330398738e-05, + "loss": 0.0893, + "step": 6373 + }, + { + "epoch": 0.83, + "learning_rate": 1.9711425861719168e-05, + "loss": 0.0931, + "step": 6374 + }, + { + "epoch": 0.84, + "learning_rate": 1.971129936572673e-05, + "loss": 0.0707, + "step": 6375 + }, + { + "epoch": 0.84, + "learning_rate": 1.9711172842421776e-05, + "loss": 0.0851, + "step": 6376 + }, + { + "epoch": 0.84, + "learning_rate": 1.971104629180467e-05, + "loss": 0.0802, + "step": 6377 + }, + { + "epoch": 0.84, + "learning_rate": 1.9710919713875763e-05, + "loss": 0.0803, + "step": 6378 + }, + { + "epoch": 0.84, + "learning_rate": 1.971079310863541e-05, + "loss": 0.0812, + "step": 6379 + }, + { + "epoch": 0.84, + "learning_rate": 1.971066647608397e-05, + "loss": 0.0941, + "step": 6380 + }, + { + "epoch": 0.84, + "learning_rate": 1.9710539816221796e-05, + "loss": 0.0941, + "step": 6381 + }, + { + "epoch": 0.84, + "learning_rate": 1.971041312904925e-05, + "loss": 0.0759, + "step": 6382 + }, + { + "epoch": 0.84, + "learning_rate": 1.9710286414566678e-05, + "loss": 0.0952, + "step": 6383 + }, + { + "epoch": 0.84, + "learning_rate": 1.971015967277445e-05, + "loss": 0.0781, + "step": 6384 + }, + { + "epoch": 0.84, + "learning_rate": 1.971003290367291e-05, + "loss": 0.0915, + "step": 6385 + }, + { + "epoch": 0.84, + "learning_rate": 1.9709906107262417e-05, + "loss": 0.0919, + "step": 6386 + }, + { + "epoch": 0.84, + "learning_rate": 1.9709779283543335e-05, + "loss": 0.0988, + "step": 6387 + }, + { + "epoch": 0.84, + "learning_rate": 1.9709652432516013e-05, + "loss": 0.0896, + "step": 6388 + }, + { + "epoch": 0.84, + "learning_rate": 1.9709525554180816e-05, + "loss": 0.0924, + "step": 6389 + }, + { + "epoch": 0.84, + "learning_rate": 1.9709398648538087e-05, + "loss": 0.0944, + "step": 6390 + }, + { + "epoch": 0.84, + "learning_rate": 1.9709271715588198e-05, + "loss": 0.0867, + "step": 6391 + }, + { + "epoch": 0.84, + "learning_rate": 1.9709144755331495e-05, + "loss": 0.0917, + "step": 6392 + }, + { + "epoch": 0.84, + "learning_rate": 1.9709017767768343e-05, + "loss": 0.0796, + "step": 6393 + }, + { + "epoch": 0.84, + "learning_rate": 1.9708890752899096e-05, + "loss": 0.0764, + "step": 6394 + }, + { + "epoch": 0.84, + "learning_rate": 1.9708763710724108e-05, + "loss": 0.0875, + "step": 6395 + }, + { + "epoch": 0.84, + "learning_rate": 1.970863664124374e-05, + "loss": 0.0857, + "step": 6396 + }, + { + "epoch": 0.84, + "learning_rate": 1.9708509544458346e-05, + "loss": 0.0834, + "step": 6397 + }, + { + "epoch": 0.84, + "learning_rate": 1.970838242036829e-05, + "loss": 0.0866, + "step": 6398 + }, + { + "epoch": 0.84, + "learning_rate": 1.9708255268973922e-05, + "loss": 0.0914, + "step": 6399 + }, + { + "epoch": 0.84, + "learning_rate": 1.9708128090275608e-05, + "loss": 0.0859, + "step": 6400 + }, + { + "epoch": 0.84, + "learning_rate": 1.9708000884273698e-05, + "loss": 0.0898, + "step": 6401 + }, + { + "epoch": 0.84, + "learning_rate": 1.970787365096855e-05, + "loss": 0.0875, + "step": 6402 + }, + { + "epoch": 0.84, + "learning_rate": 1.9707746390360528e-05, + "loss": 0.0876, + "step": 6403 + }, + { + "epoch": 0.84, + "learning_rate": 1.9707619102449983e-05, + "loss": 0.0816, + "step": 6404 + }, + { + "epoch": 0.84, + "learning_rate": 1.9707491787237276e-05, + "loss": 0.091, + "step": 6405 + }, + { + "epoch": 0.84, + "learning_rate": 1.970736444472277e-05, + "loss": 0.0808, + "step": 6406 + }, + { + "epoch": 0.84, + "learning_rate": 1.9707237074906814e-05, + "loss": 0.0963, + "step": 6407 + }, + { + "epoch": 0.84, + "learning_rate": 1.9707109677789772e-05, + "loss": 0.085, + "step": 6408 + }, + { + "epoch": 0.84, + "learning_rate": 1.9706982253372004e-05, + "loss": 0.0927, + "step": 6409 + }, + { + "epoch": 0.84, + "learning_rate": 1.970685480165386e-05, + "loss": 0.0817, + "step": 6410 + }, + { + "epoch": 0.84, + "learning_rate": 1.9706727322635706e-05, + "loss": 0.0905, + "step": 6411 + }, + { + "epoch": 0.84, + "learning_rate": 1.97065998163179e-05, + "loss": 0.0891, + "step": 6412 + }, + { + "epoch": 0.84, + "learning_rate": 1.9706472282700795e-05, + "loss": 0.0843, + "step": 6413 + }, + { + "epoch": 0.84, + "learning_rate": 1.970634472178476e-05, + "loss": 0.0847, + "step": 6414 + }, + { + "epoch": 0.84, + "learning_rate": 1.970621713357014e-05, + "loss": 0.0853, + "step": 6415 + }, + { + "epoch": 0.84, + "learning_rate": 1.9706089518057308e-05, + "loss": 0.0884, + "step": 6416 + }, + { + "epoch": 0.84, + "learning_rate": 1.9705961875246613e-05, + "loss": 0.0864, + "step": 6417 + }, + { + "epoch": 0.84, + "learning_rate": 1.9705834205138413e-05, + "loss": 0.0892, + "step": 6418 + }, + { + "epoch": 0.84, + "learning_rate": 1.9705706507733077e-05, + "loss": 0.0886, + "step": 6419 + }, + { + "epoch": 0.84, + "learning_rate": 1.9705578783030958e-05, + "loss": 0.0897, + "step": 6420 + }, + { + "epoch": 0.84, + "learning_rate": 1.9705451031032413e-05, + "loss": 0.0806, + "step": 6421 + }, + { + "epoch": 0.84, + "learning_rate": 1.9705323251737804e-05, + "loss": 0.084, + "step": 6422 + }, + { + "epoch": 0.84, + "learning_rate": 1.970519544514749e-05, + "loss": 0.0899, + "step": 6423 + }, + { + "epoch": 0.84, + "learning_rate": 1.9705067611261833e-05, + "loss": 0.0848, + "step": 6424 + }, + { + "epoch": 0.84, + "learning_rate": 1.9704939750081185e-05, + "loss": 0.0911, + "step": 6425 + }, + { + "epoch": 0.84, + "learning_rate": 1.970481186160592e-05, + "loss": 0.0832, + "step": 6426 + }, + { + "epoch": 0.84, + "learning_rate": 1.970468394583638e-05, + "loss": 0.0928, + "step": 6427 + }, + { + "epoch": 0.84, + "learning_rate": 1.9704556002772936e-05, + "loss": 0.09, + "step": 6428 + }, + { + "epoch": 0.84, + "learning_rate": 1.9704428032415945e-05, + "loss": 0.0887, + "step": 6429 + }, + { + "epoch": 0.84, + "learning_rate": 1.970430003476577e-05, + "loss": 0.08, + "step": 6430 + }, + { + "epoch": 0.84, + "learning_rate": 1.9704172009822764e-05, + "loss": 0.0913, + "step": 6431 + }, + { + "epoch": 0.84, + "learning_rate": 1.9704043957587292e-05, + "loss": 0.096, + "step": 6432 + }, + { + "epoch": 0.84, + "learning_rate": 1.9703915878059714e-05, + "loss": 0.0765, + "step": 6433 + }, + { + "epoch": 0.84, + "learning_rate": 1.970378777124039e-05, + "loss": 0.0831, + "step": 6434 + }, + { + "epoch": 0.84, + "learning_rate": 1.9703659637129676e-05, + "loss": 0.0961, + "step": 6435 + }, + { + "epoch": 0.84, + "learning_rate": 1.970353147572794e-05, + "loss": 0.0854, + "step": 6436 + }, + { + "epoch": 0.84, + "learning_rate": 1.9703403287035535e-05, + "loss": 0.0937, + "step": 6437 + }, + { + "epoch": 0.84, + "learning_rate": 1.9703275071052827e-05, + "loss": 0.0837, + "step": 6438 + }, + { + "epoch": 0.84, + "learning_rate": 1.9703146827780177e-05, + "loss": 0.0758, + "step": 6439 + }, + { + "epoch": 0.84, + "learning_rate": 1.9703018557217938e-05, + "loss": 0.078, + "step": 6440 + }, + { + "epoch": 0.84, + "learning_rate": 1.970289025936648e-05, + "loss": 0.0873, + "step": 6441 + }, + { + "epoch": 0.84, + "learning_rate": 1.9702761934226157e-05, + "loss": 0.0848, + "step": 6442 + }, + { + "epoch": 0.84, + "learning_rate": 1.9702633581797333e-05, + "loss": 0.0866, + "step": 6443 + }, + { + "epoch": 0.84, + "learning_rate": 1.970250520208037e-05, + "loss": 0.0863, + "step": 6444 + }, + { + "epoch": 0.84, + "learning_rate": 1.970237679507563e-05, + "loss": 0.0938, + "step": 6445 + }, + { + "epoch": 0.84, + "learning_rate": 1.970224836078347e-05, + "loss": 0.0786, + "step": 6446 + }, + { + "epoch": 0.84, + "learning_rate": 1.9702119899204252e-05, + "loss": 0.0924, + "step": 6447 + }, + { + "epoch": 0.84, + "learning_rate": 1.970199141033834e-05, + "loss": 0.0759, + "step": 6448 + }, + { + "epoch": 0.84, + "learning_rate": 1.9701862894186094e-05, + "loss": 0.0928, + "step": 6449 + }, + { + "epoch": 0.84, + "learning_rate": 1.9701734350747874e-05, + "loss": 0.0949, + "step": 6450 + }, + { + "epoch": 0.85, + "learning_rate": 1.9701605780024044e-05, + "loss": 0.0769, + "step": 6451 + }, + { + "epoch": 0.85, + "learning_rate": 1.970147718201496e-05, + "loss": 0.0846, + "step": 6452 + }, + { + "epoch": 0.85, + "learning_rate": 1.9701348556720995e-05, + "loss": 0.0908, + "step": 6453 + }, + { + "epoch": 0.85, + "learning_rate": 1.97012199041425e-05, + "loss": 0.0912, + "step": 6454 + }, + { + "epoch": 0.85, + "learning_rate": 1.970109122427984e-05, + "loss": 0.083, + "step": 6455 + }, + { + "epoch": 0.85, + "learning_rate": 1.9700962517133377e-05, + "loss": 0.0803, + "step": 6456 + }, + { + "epoch": 0.85, + "learning_rate": 1.9700833782703478e-05, + "loss": 0.0953, + "step": 6457 + }, + { + "epoch": 0.85, + "learning_rate": 1.9700705020990496e-05, + "loss": 0.0876, + "step": 6458 + }, + { + "epoch": 0.85, + "learning_rate": 1.97005762319948e-05, + "loss": 0.0863, + "step": 6459 + }, + { + "epoch": 0.85, + "learning_rate": 1.970044741571675e-05, + "loss": 0.0869, + "step": 6460 + }, + { + "epoch": 0.85, + "learning_rate": 1.9700318572156706e-05, + "loss": 0.0915, + "step": 6461 + }, + { + "epoch": 0.85, + "learning_rate": 1.9700189701315035e-05, + "loss": 0.0914, + "step": 6462 + }, + { + "epoch": 0.85, + "learning_rate": 1.97000608031921e-05, + "loss": 0.088, + "step": 6463 + }, + { + "epoch": 0.85, + "learning_rate": 1.9699931877788252e-05, + "loss": 0.0847, + "step": 6464 + }, + { + "epoch": 0.85, + "learning_rate": 1.969980292510387e-05, + "loss": 0.1081, + "step": 6465 + }, + { + "epoch": 0.85, + "learning_rate": 1.9699673945139305e-05, + "loss": 0.0845, + "step": 6466 + }, + { + "epoch": 0.85, + "learning_rate": 1.9699544937894924e-05, + "loss": 0.0873, + "step": 6467 + }, + { + "epoch": 0.85, + "learning_rate": 1.969941590337109e-05, + "loss": 0.0948, + "step": 6468 + }, + { + "epoch": 0.85, + "learning_rate": 1.9699286841568166e-05, + "loss": 0.0806, + "step": 6469 + }, + { + "epoch": 0.85, + "learning_rate": 1.969915775248651e-05, + "loss": 0.0864, + "step": 6470 + }, + { + "epoch": 0.85, + "learning_rate": 1.9699028636126492e-05, + "loss": 0.0791, + "step": 6471 + }, + { + "epoch": 0.85, + "learning_rate": 1.9698899492488475e-05, + "loss": 0.0977, + "step": 6472 + }, + { + "epoch": 0.85, + "learning_rate": 1.969877032157282e-05, + "loss": 0.0941, + "step": 6473 + }, + { + "epoch": 0.85, + "learning_rate": 1.9698641123379887e-05, + "loss": 0.0938, + "step": 6474 + }, + { + "epoch": 0.85, + "learning_rate": 1.9698511897910044e-05, + "loss": 0.0828, + "step": 6475 + }, + { + "epoch": 0.85, + "learning_rate": 1.9698382645163652e-05, + "loss": 0.0873, + "step": 6476 + }, + { + "epoch": 0.85, + "learning_rate": 1.9698253365141073e-05, + "loss": 0.0925, + "step": 6477 + }, + { + "epoch": 0.85, + "learning_rate": 1.9698124057842673e-05, + "loss": 0.0909, + "step": 6478 + }, + { + "epoch": 0.85, + "learning_rate": 1.969799472326882e-05, + "loss": 0.0898, + "step": 6479 + }, + { + "epoch": 0.85, + "learning_rate": 1.969786536141987e-05, + "loss": 0.0819, + "step": 6480 + }, + { + "epoch": 0.85, + "learning_rate": 1.9697735972296193e-05, + "loss": 0.0809, + "step": 6481 + }, + { + "epoch": 0.85, + "learning_rate": 1.9697606555898144e-05, + "loss": 0.0884, + "step": 6482 + }, + { + "epoch": 0.85, + "learning_rate": 1.96974771122261e-05, + "loss": 0.0856, + "step": 6483 + }, + { + "epoch": 0.85, + "learning_rate": 1.9697347641280412e-05, + "loss": 0.0894, + "step": 6484 + }, + { + "epoch": 0.85, + "learning_rate": 1.9697218143061457e-05, + "loss": 0.0774, + "step": 6485 + }, + { + "epoch": 0.85, + "learning_rate": 1.9697088617569586e-05, + "loss": 0.0967, + "step": 6486 + }, + { + "epoch": 0.85, + "learning_rate": 1.9696959064805172e-05, + "loss": 0.0813, + "step": 6487 + }, + { + "epoch": 0.85, + "learning_rate": 1.9696829484768577e-05, + "loss": 0.0988, + "step": 6488 + }, + { + "epoch": 0.85, + "learning_rate": 1.9696699877460168e-05, + "loss": 0.0793, + "step": 6489 + }, + { + "epoch": 0.85, + "learning_rate": 1.9696570242880303e-05, + "loss": 0.0747, + "step": 6490 + }, + { + "epoch": 0.85, + "learning_rate": 1.969644058102935e-05, + "loss": 0.1017, + "step": 6491 + }, + { + "epoch": 0.85, + "learning_rate": 1.969631089190768e-05, + "loss": 0.0944, + "step": 6492 + }, + { + "epoch": 0.85, + "learning_rate": 1.9696181175515646e-05, + "loss": 0.0908, + "step": 6493 + }, + { + "epoch": 0.85, + "learning_rate": 1.9696051431853624e-05, + "loss": 0.0802, + "step": 6494 + }, + { + "epoch": 0.85, + "learning_rate": 1.969592166092197e-05, + "loss": 0.0879, + "step": 6495 + }, + { + "epoch": 0.85, + "learning_rate": 1.9695791862721055e-05, + "loss": 0.0797, + "step": 6496 + }, + { + "epoch": 0.85, + "learning_rate": 1.969566203725124e-05, + "loss": 0.0855, + "step": 6497 + }, + { + "epoch": 0.85, + "learning_rate": 1.9695532184512892e-05, + "loss": 0.0885, + "step": 6498 + }, + { + "epoch": 0.85, + "learning_rate": 1.9695402304506376e-05, + "loss": 0.0905, + "step": 6499 + }, + { + "epoch": 0.85, + "learning_rate": 1.9695272397232056e-05, + "loss": 0.094, + "step": 6500 + }, + { + "epoch": 0.85, + "learning_rate": 1.96951424626903e-05, + "loss": 0.081, + "step": 6501 + }, + { + "epoch": 0.85, + "learning_rate": 1.9695012500881475e-05, + "loss": 0.0877, + "step": 6502 + }, + { + "epoch": 0.85, + "learning_rate": 1.969488251180594e-05, + "loss": 0.1003, + "step": 6503 + }, + { + "epoch": 0.85, + "learning_rate": 1.9694752495464064e-05, + "loss": 0.0822, + "step": 6504 + }, + { + "epoch": 0.85, + "learning_rate": 1.9694622451856214e-05, + "loss": 0.0829, + "step": 6505 + }, + { + "epoch": 0.85, + "learning_rate": 1.9694492380982756e-05, + "loss": 0.0891, + "step": 6506 + }, + { + "epoch": 0.85, + "learning_rate": 1.9694362282844052e-05, + "loss": 0.0922, + "step": 6507 + }, + { + "epoch": 0.85, + "learning_rate": 1.9694232157440473e-05, + "loss": 0.0881, + "step": 6508 + }, + { + "epoch": 0.85, + "learning_rate": 1.969410200477238e-05, + "loss": 0.0914, + "step": 6509 + }, + { + "epoch": 0.85, + "learning_rate": 1.969397182484014e-05, + "loss": 0.0816, + "step": 6510 + }, + { + "epoch": 0.85, + "learning_rate": 1.9693841617644126e-05, + "loss": 0.0767, + "step": 6511 + }, + { + "epoch": 0.85, + "learning_rate": 1.9693711383184696e-05, + "loss": 0.0927, + "step": 6512 + }, + { + "epoch": 0.85, + "learning_rate": 1.9693581121462216e-05, + "loss": 0.0702, + "step": 6513 + }, + { + "epoch": 0.85, + "learning_rate": 1.969345083247706e-05, + "loss": 0.0849, + "step": 6514 + }, + { + "epoch": 0.85, + "learning_rate": 1.9693320516229584e-05, + "loss": 0.0824, + "step": 6515 + }, + { + "epoch": 0.85, + "learning_rate": 1.9693190172720164e-05, + "loss": 0.0806, + "step": 6516 + }, + { + "epoch": 0.85, + "learning_rate": 1.9693059801949162e-05, + "loss": 0.0798, + "step": 6517 + }, + { + "epoch": 0.85, + "learning_rate": 1.9692929403916946e-05, + "loss": 0.0759, + "step": 6518 + }, + { + "epoch": 0.85, + "learning_rate": 1.9692798978623883e-05, + "loss": 0.1014, + "step": 6519 + }, + { + "epoch": 0.85, + "learning_rate": 1.969266852607034e-05, + "loss": 0.0872, + "step": 6520 + }, + { + "epoch": 0.85, + "learning_rate": 1.9692538046256678e-05, + "loss": 0.0874, + "step": 6521 + }, + { + "epoch": 0.85, + "learning_rate": 1.9692407539183275e-05, + "loss": 0.0954, + "step": 6522 + }, + { + "epoch": 0.85, + "learning_rate": 1.969227700485049e-05, + "loss": 0.0907, + "step": 6523 + }, + { + "epoch": 0.85, + "learning_rate": 1.9692146443258687e-05, + "loss": 0.0881, + "step": 6524 + }, + { + "epoch": 0.85, + "learning_rate": 1.9692015854408244e-05, + "loss": 0.0911, + "step": 6525 + }, + { + "epoch": 0.85, + "learning_rate": 1.969188523829952e-05, + "loss": 0.0897, + "step": 6526 + }, + { + "epoch": 0.85, + "learning_rate": 1.9691754594932887e-05, + "loss": 0.0873, + "step": 6527 + }, + { + "epoch": 0.86, + "learning_rate": 1.969162392430871e-05, + "loss": 0.0843, + "step": 6528 + }, + { + "epoch": 0.86, + "learning_rate": 1.9691493226427354e-05, + "loss": 0.0832, + "step": 6529 + }, + { + "epoch": 0.86, + "learning_rate": 1.969136250128919e-05, + "loss": 0.0964, + "step": 6530 + }, + { + "epoch": 0.86, + "learning_rate": 1.969123174889459e-05, + "loss": 0.0777, + "step": 6531 + }, + { + "epoch": 0.86, + "learning_rate": 1.969110096924391e-05, + "loss": 0.0945, + "step": 6532 + }, + { + "epoch": 0.86, + "learning_rate": 1.969097016233753e-05, + "loss": 0.0723, + "step": 6533 + }, + { + "epoch": 0.86, + "learning_rate": 1.969083932817581e-05, + "loss": 0.0795, + "step": 6534 + }, + { + "epoch": 0.86, + "learning_rate": 1.9690708466759118e-05, + "loss": 0.0909, + "step": 6535 + }, + { + "epoch": 0.86, + "learning_rate": 1.9690577578087828e-05, + "loss": 0.0843, + "step": 6536 + }, + { + "epoch": 0.86, + "learning_rate": 1.9690446662162304e-05, + "loss": 0.0821, + "step": 6537 + }, + { + "epoch": 0.86, + "learning_rate": 1.9690315718982912e-05, + "loss": 0.0854, + "step": 6538 + }, + { + "epoch": 0.86, + "learning_rate": 1.969018474855003e-05, + "loss": 0.0784, + "step": 6539 + }, + { + "epoch": 0.86, + "learning_rate": 1.9690053750864012e-05, + "loss": 0.0911, + "step": 6540 + }, + { + "epoch": 0.86, + "learning_rate": 1.968992272592524e-05, + "loss": 0.0883, + "step": 6541 + }, + { + "epoch": 0.86, + "learning_rate": 1.968979167373407e-05, + "loss": 0.0886, + "step": 6542 + }, + { + "epoch": 0.86, + "learning_rate": 1.9689660594290882e-05, + "loss": 0.0799, + "step": 6543 + }, + { + "epoch": 0.86, + "learning_rate": 1.968952948759604e-05, + "loss": 0.0837, + "step": 6544 + }, + { + "epoch": 0.86, + "learning_rate": 1.968939835364991e-05, + "loss": 0.09, + "step": 6545 + }, + { + "epoch": 0.86, + "learning_rate": 1.968926719245286e-05, + "loss": 0.0787, + "step": 6546 + }, + { + "epoch": 0.86, + "learning_rate": 1.9689136004005266e-05, + "loss": 0.0884, + "step": 6547 + }, + { + "epoch": 0.86, + "learning_rate": 1.968900478830749e-05, + "loss": 0.0862, + "step": 6548 + }, + { + "epoch": 0.86, + "learning_rate": 1.9688873545359907e-05, + "loss": 0.0854, + "step": 6549 + }, + { + "epoch": 0.86, + "learning_rate": 1.9688742275162883e-05, + "loss": 0.0852, + "step": 6550 + }, + { + "epoch": 0.86, + "learning_rate": 1.9688610977716786e-05, + "loss": 0.108, + "step": 6551 + }, + { + "epoch": 0.86, + "learning_rate": 1.9688479653021988e-05, + "loss": 0.0867, + "step": 6552 + }, + { + "epoch": 0.86, + "learning_rate": 1.9688348301078858e-05, + "loss": 0.0949, + "step": 6553 + }, + { + "epoch": 0.86, + "learning_rate": 1.968821692188776e-05, + "loss": 0.0843, + "step": 6554 + }, + { + "epoch": 0.86, + "learning_rate": 1.968808551544907e-05, + "loss": 0.1027, + "step": 6555 + }, + { + "epoch": 0.86, + "learning_rate": 1.9687954081763158e-05, + "loss": 0.0817, + "step": 6556 + }, + { + "epoch": 0.86, + "learning_rate": 1.9687822620830388e-05, + "loss": 0.0837, + "step": 6557 + }, + { + "epoch": 0.86, + "learning_rate": 1.9687691132651133e-05, + "loss": 0.0843, + "step": 6558 + }, + { + "epoch": 0.86, + "learning_rate": 1.9687559617225762e-05, + "loss": 0.0838, + "step": 6559 + }, + { + "epoch": 0.86, + "learning_rate": 1.9687428074554648e-05, + "loss": 0.0783, + "step": 6560 + }, + { + "epoch": 0.86, + "learning_rate": 1.9687296504638157e-05, + "loss": 0.091, + "step": 6561 + }, + { + "epoch": 0.86, + "learning_rate": 1.9687164907476658e-05, + "loss": 0.0821, + "step": 6562 + }, + { + "epoch": 0.86, + "learning_rate": 1.968703328307053e-05, + "loss": 0.0842, + "step": 6563 + }, + { + "epoch": 0.86, + "learning_rate": 1.968690163142013e-05, + "loss": 0.0769, + "step": 6564 + }, + { + "epoch": 0.86, + "learning_rate": 1.968676995252584e-05, + "loss": 0.0867, + "step": 6565 + }, + { + "epoch": 0.86, + "learning_rate": 1.968663824638802e-05, + "loss": 0.0855, + "step": 6566 + }, + { + "epoch": 0.86, + "learning_rate": 1.968650651300705e-05, + "loss": 0.1146, + "step": 6567 + }, + { + "epoch": 0.86, + "learning_rate": 1.9686374752383294e-05, + "loss": 0.0862, + "step": 6568 + }, + { + "epoch": 0.86, + "learning_rate": 1.9686242964517128e-05, + "loss": 0.0824, + "step": 6569 + }, + { + "epoch": 0.86, + "learning_rate": 1.968611114940892e-05, + "loss": 0.0904, + "step": 6570 + }, + { + "epoch": 0.86, + "learning_rate": 1.9685979307059035e-05, + "loss": 0.089, + "step": 6571 + }, + { + "epoch": 0.86, + "learning_rate": 1.968584743746785e-05, + "loss": 0.0968, + "step": 6572 + }, + { + "epoch": 0.86, + "learning_rate": 1.968571554063574e-05, + "loss": 0.0837, + "step": 6573 + }, + { + "epoch": 0.86, + "learning_rate": 1.9685583616563065e-05, + "loss": 0.096, + "step": 6574 + }, + { + "epoch": 0.86, + "learning_rate": 1.9685451665250206e-05, + "loss": 0.0915, + "step": 6575 + }, + { + "epoch": 0.86, + "learning_rate": 1.968531968669753e-05, + "loss": 0.0759, + "step": 6576 + }, + { + "epoch": 0.86, + "learning_rate": 1.9685187680905406e-05, + "loss": 0.0835, + "step": 6577 + }, + { + "epoch": 0.86, + "learning_rate": 1.9685055647874206e-05, + "loss": 0.0886, + "step": 6578 + }, + { + "epoch": 0.86, + "learning_rate": 1.9684923587604304e-05, + "loss": 0.0918, + "step": 6579 + }, + { + "epoch": 0.86, + "learning_rate": 1.9684791500096072e-05, + "loss": 0.0924, + "step": 6580 + }, + { + "epoch": 0.86, + "learning_rate": 1.9684659385349876e-05, + "loss": 0.093, + "step": 6581 + }, + { + "epoch": 0.86, + "learning_rate": 1.9684527243366096e-05, + "loss": 0.0857, + "step": 6582 + }, + { + "epoch": 0.86, + "learning_rate": 1.9684395074145097e-05, + "loss": 0.0978, + "step": 6583 + }, + { + "epoch": 0.86, + "learning_rate": 1.9684262877687253e-05, + "loss": 0.0791, + "step": 6584 + }, + { + "epoch": 0.86, + "learning_rate": 1.968413065399293e-05, + "loss": 0.0842, + "step": 6585 + }, + { + "epoch": 0.86, + "learning_rate": 1.9683998403062513e-05, + "loss": 0.0908, + "step": 6586 + }, + { + "epoch": 0.86, + "learning_rate": 1.9683866124896363e-05, + "loss": 0.1021, + "step": 6587 + }, + { + "epoch": 0.86, + "learning_rate": 1.9683733819494855e-05, + "loss": 0.0925, + "step": 6588 + }, + { + "epoch": 0.86, + "learning_rate": 1.9683601486858365e-05, + "loss": 0.0914, + "step": 6589 + }, + { + "epoch": 0.86, + "learning_rate": 1.9683469126987255e-05, + "loss": 0.0891, + "step": 6590 + }, + { + "epoch": 0.86, + "learning_rate": 1.9683336739881907e-05, + "loss": 0.0806, + "step": 6591 + }, + { + "epoch": 0.86, + "learning_rate": 1.968320432554269e-05, + "loss": 0.0839, + "step": 6592 + }, + { + "epoch": 0.86, + "learning_rate": 1.9683071883969976e-05, + "loss": 0.081, + "step": 6593 + }, + { + "epoch": 0.86, + "learning_rate": 1.968293941516414e-05, + "loss": 0.0849, + "step": 6594 + }, + { + "epoch": 0.86, + "learning_rate": 1.968280691912555e-05, + "loss": 0.0837, + "step": 6595 + }, + { + "epoch": 0.86, + "learning_rate": 1.968267439585458e-05, + "loss": 0.088, + "step": 6596 + }, + { + "epoch": 0.86, + "learning_rate": 1.9682541845351606e-05, + "loss": 0.0935, + "step": 6597 + }, + { + "epoch": 0.86, + "learning_rate": 1.9682409267617e-05, + "loss": 0.0803, + "step": 6598 + }, + { + "epoch": 0.86, + "learning_rate": 1.968227666265113e-05, + "loss": 0.0829, + "step": 6599 + }, + { + "epoch": 0.86, + "learning_rate": 1.9682144030454378e-05, + "loss": 0.0863, + "step": 6600 + }, + { + "epoch": 0.86, + "learning_rate": 1.9682011371027108e-05, + "loss": 0.0865, + "step": 6601 + }, + { + "epoch": 0.86, + "learning_rate": 1.9681878684369694e-05, + "loss": 0.1094, + "step": 6602 + }, + { + "epoch": 0.86, + "learning_rate": 1.9681745970482515e-05, + "loss": 0.0827, + "step": 6603 + }, + { + "epoch": 0.87, + "learning_rate": 1.968161322936594e-05, + "loss": 0.1012, + "step": 6604 + }, + { + "epoch": 0.87, + "learning_rate": 1.9681480461020345e-05, + "loss": 0.0892, + "step": 6605 + }, + { + "epoch": 0.87, + "learning_rate": 1.96813476654461e-05, + "loss": 0.0907, + "step": 6606 + }, + { + "epoch": 0.87, + "learning_rate": 1.968121484264358e-05, + "loss": 0.0933, + "step": 6607 + }, + { + "epoch": 0.87, + "learning_rate": 1.968108199261316e-05, + "loss": 0.0766, + "step": 6608 + }, + { + "epoch": 0.87, + "learning_rate": 1.9680949115355215e-05, + "loss": 0.0901, + "step": 6609 + }, + { + "epoch": 0.87, + "learning_rate": 1.9680816210870113e-05, + "loss": 0.077, + "step": 6610 + }, + { + "epoch": 0.87, + "learning_rate": 1.9680683279158234e-05, + "loss": 0.0942, + "step": 6611 + }, + { + "epoch": 0.87, + "learning_rate": 1.968055032021994e-05, + "loss": 0.0942, + "step": 6612 + }, + { + "epoch": 0.87, + "learning_rate": 1.968041733405562e-05, + "loss": 0.0831, + "step": 6613 + }, + { + "epoch": 0.87, + "learning_rate": 1.9680284320665645e-05, + "loss": 0.093, + "step": 6614 + }, + { + "epoch": 0.87, + "learning_rate": 1.968015128005038e-05, + "loss": 0.0904, + "step": 6615 + }, + { + "epoch": 0.87, + "learning_rate": 1.968001821221021e-05, + "loss": 0.0953, + "step": 6616 + }, + { + "epoch": 0.87, + "learning_rate": 1.96798851171455e-05, + "loss": 0.0865, + "step": 6617 + }, + { + "epoch": 0.87, + "learning_rate": 1.9679751994856633e-05, + "loss": 0.085, + "step": 6618 + }, + { + "epoch": 0.87, + "learning_rate": 1.9679618845343977e-05, + "loss": 0.0864, + "step": 6619 + }, + { + "epoch": 0.87, + "learning_rate": 1.9679485668607906e-05, + "loss": 0.0946, + "step": 6620 + }, + { + "epoch": 0.87, + "learning_rate": 1.96793524646488e-05, + "loss": 0.0814, + "step": 6621 + }, + { + "epoch": 0.87, + "learning_rate": 1.967921923346703e-05, + "loss": 0.1004, + "step": 6622 + }, + { + "epoch": 0.87, + "learning_rate": 1.967908597506297e-05, + "loss": 0.0893, + "step": 6623 + }, + { + "epoch": 0.87, + "learning_rate": 1.9678952689436997e-05, + "loss": 0.086, + "step": 6624 + }, + { + "epoch": 0.87, + "learning_rate": 1.9678819376589484e-05, + "loss": 0.0922, + "step": 6625 + }, + { + "epoch": 0.87, + "learning_rate": 1.9678686036520808e-05, + "loss": 0.0916, + "step": 6626 + }, + { + "epoch": 0.87, + "learning_rate": 1.9678552669231343e-05, + "loss": 0.0832, + "step": 6627 + }, + { + "epoch": 0.87, + "learning_rate": 1.9678419274721462e-05, + "loss": 0.0799, + "step": 6628 + }, + { + "epoch": 0.87, + "learning_rate": 1.9678285852991546e-05, + "loss": 0.0823, + "step": 6629 + }, + { + "epoch": 0.87, + "learning_rate": 1.9678152404041966e-05, + "loss": 0.0832, + "step": 6630 + }, + { + "epoch": 0.87, + "learning_rate": 1.9678018927873094e-05, + "loss": 0.099, + "step": 6631 + }, + { + "epoch": 0.87, + "learning_rate": 1.9677885424485312e-05, + "loss": 0.0832, + "step": 6632 + }, + { + "epoch": 0.87, + "learning_rate": 1.9677751893878995e-05, + "loss": 0.1002, + "step": 6633 + }, + { + "epoch": 0.87, + "learning_rate": 1.967761833605451e-05, + "loss": 0.0812, + "step": 6634 + }, + { + "epoch": 0.87, + "learning_rate": 1.9677484751012244e-05, + "loss": 0.0859, + "step": 6635 + }, + { + "epoch": 0.87, + "learning_rate": 1.9677351138752564e-05, + "loss": 0.0872, + "step": 6636 + }, + { + "epoch": 0.87, + "learning_rate": 1.967721749927585e-05, + "loss": 0.0949, + "step": 6637 + }, + { + "epoch": 0.87, + "learning_rate": 1.967708383258248e-05, + "loss": 0.0887, + "step": 6638 + }, + { + "epoch": 0.87, + "learning_rate": 1.9676950138672827e-05, + "loss": 0.107, + "step": 6639 + }, + { + "epoch": 0.87, + "learning_rate": 1.967681641754726e-05, + "loss": 0.0773, + "step": 6640 + }, + { + "epoch": 0.87, + "learning_rate": 1.967668266920617e-05, + "loss": 0.0875, + "step": 6641 + }, + { + "epoch": 0.87, + "learning_rate": 1.967654889364992e-05, + "loss": 0.0872, + "step": 6642 + }, + { + "epoch": 0.87, + "learning_rate": 1.9676415090878898e-05, + "loss": 0.0852, + "step": 6643 + }, + { + "epoch": 0.87, + "learning_rate": 1.9676281260893467e-05, + "loss": 0.0825, + "step": 6644 + }, + { + "epoch": 0.87, + "learning_rate": 1.9676147403694014e-05, + "loss": 0.0848, + "step": 6645 + }, + { + "epoch": 0.87, + "learning_rate": 1.967601351928091e-05, + "loss": 0.0879, + "step": 6646 + }, + { + "epoch": 0.87, + "learning_rate": 1.9675879607654536e-05, + "loss": 0.0959, + "step": 6647 + }, + { + "epoch": 0.87, + "learning_rate": 1.9675745668815262e-05, + "loss": 0.0856, + "step": 6648 + }, + { + "epoch": 0.87, + "learning_rate": 1.9675611702763472e-05, + "loss": 0.0796, + "step": 6649 + }, + { + "epoch": 0.87, + "learning_rate": 1.967547770949954e-05, + "loss": 0.0832, + "step": 6650 + }, + { + "epoch": 0.87, + "learning_rate": 1.967534368902384e-05, + "loss": 0.0774, + "step": 6651 + }, + { + "epoch": 0.87, + "learning_rate": 1.9675209641336748e-05, + "loss": 0.0887, + "step": 6652 + }, + { + "epoch": 0.87, + "learning_rate": 1.967507556643865e-05, + "loss": 0.0909, + "step": 6653 + }, + { + "epoch": 0.87, + "learning_rate": 1.967494146432991e-05, + "loss": 0.0783, + "step": 6654 + }, + { + "epoch": 0.87, + "learning_rate": 1.9674807335010922e-05, + "loss": 0.0928, + "step": 6655 + }, + { + "epoch": 0.87, + "learning_rate": 1.9674673178482045e-05, + "loss": 0.0834, + "step": 6656 + }, + { + "epoch": 0.87, + "learning_rate": 1.9674538994743672e-05, + "loss": 0.0738, + "step": 6657 + }, + { + "epoch": 0.87, + "learning_rate": 1.9674404783796168e-05, + "loss": 0.0817, + "step": 6658 + }, + { + "epoch": 0.87, + "learning_rate": 1.9674270545639916e-05, + "loss": 0.0925, + "step": 6659 + }, + { + "epoch": 0.87, + "learning_rate": 1.9674136280275295e-05, + "loss": 0.0882, + "step": 6660 + }, + { + "epoch": 0.87, + "learning_rate": 1.967400198770268e-05, + "loss": 0.0817, + "step": 6661 + }, + { + "epoch": 0.87, + "learning_rate": 1.967386766792245e-05, + "loss": 0.0904, + "step": 6662 + }, + { + "epoch": 0.87, + "learning_rate": 1.9673733320934985e-05, + "loss": 0.0835, + "step": 6663 + }, + { + "epoch": 0.87, + "learning_rate": 1.9673598946740657e-05, + "loss": 0.0819, + "step": 6664 + }, + { + "epoch": 0.87, + "learning_rate": 1.9673464545339848e-05, + "loss": 0.0935, + "step": 6665 + }, + { + "epoch": 0.87, + "learning_rate": 1.9673330116732933e-05, + "loss": 0.0832, + "step": 6666 + }, + { + "epoch": 0.87, + "learning_rate": 1.9673195660920296e-05, + "loss": 0.0857, + "step": 6667 + }, + { + "epoch": 0.87, + "learning_rate": 1.967306117790231e-05, + "loss": 0.0857, + "step": 6668 + }, + { + "epoch": 0.87, + "learning_rate": 1.9672926667679352e-05, + "loss": 0.0874, + "step": 6669 + }, + { + "epoch": 0.87, + "learning_rate": 1.9672792130251804e-05, + "loss": 0.0812, + "step": 6670 + }, + { + "epoch": 0.87, + "learning_rate": 1.9672657565620045e-05, + "loss": 0.0755, + "step": 6671 + }, + { + "epoch": 0.87, + "learning_rate": 1.967252297378445e-05, + "loss": 0.0955, + "step": 6672 + }, + { + "epoch": 0.87, + "learning_rate": 1.96723883547454e-05, + "loss": 0.0853, + "step": 6673 + }, + { + "epoch": 0.87, + "learning_rate": 1.9672253708503276e-05, + "loss": 0.0767, + "step": 6674 + }, + { + "epoch": 0.87, + "learning_rate": 1.967211903505845e-05, + "loss": 0.0916, + "step": 6675 + }, + { + "epoch": 0.87, + "learning_rate": 1.96719843344113e-05, + "loss": 0.0921, + "step": 6676 + }, + { + "epoch": 0.87, + "learning_rate": 1.9671849606562216e-05, + "loss": 0.0893, + "step": 6677 + }, + { + "epoch": 0.87, + "learning_rate": 1.9671714851511568e-05, + "loss": 0.0868, + "step": 6678 + }, + { + "epoch": 0.87, + "learning_rate": 1.9671580069259738e-05, + "loss": 0.0743, + "step": 6679 + }, + { + "epoch": 0.88, + "learning_rate": 1.96714452598071e-05, + "loss": 0.0857, + "step": 6680 + }, + { + "epoch": 0.88, + "learning_rate": 1.967131042315404e-05, + "loss": 0.0811, + "step": 6681 + }, + { + "epoch": 0.88, + "learning_rate": 1.9671175559300938e-05, + "loss": 0.0772, + "step": 6682 + }, + { + "epoch": 0.88, + "learning_rate": 1.9671040668248166e-05, + "loss": 0.0909, + "step": 6683 + }, + { + "epoch": 0.88, + "learning_rate": 1.9670905749996106e-05, + "loss": 0.0935, + "step": 6684 + }, + { + "epoch": 0.88, + "learning_rate": 1.967077080454514e-05, + "loss": 0.0922, + "step": 6685 + }, + { + "epoch": 0.88, + "learning_rate": 1.967063583189565e-05, + "loss": 0.0835, + "step": 6686 + }, + { + "epoch": 0.88, + "learning_rate": 1.9670500832048003e-05, + "loss": 0.0923, + "step": 6687 + }, + { + "epoch": 0.88, + "learning_rate": 1.9670365805002595e-05, + "loss": 0.0902, + "step": 6688 + }, + { + "epoch": 0.88, + "learning_rate": 1.9670230750759795e-05, + "loss": 0.0798, + "step": 6689 + }, + { + "epoch": 0.88, + "learning_rate": 1.9670095669319986e-05, + "loss": 0.0892, + "step": 6690 + }, + { + "epoch": 0.88, + "learning_rate": 1.9669960560683546e-05, + "loss": 0.0925, + "step": 6691 + }, + { + "epoch": 0.88, + "learning_rate": 1.966982542485086e-05, + "loss": 0.0857, + "step": 6692 + }, + { + "epoch": 0.88, + "learning_rate": 1.9669690261822304e-05, + "loss": 0.1004, + "step": 6693 + }, + { + "epoch": 0.88, + "learning_rate": 1.966955507159826e-05, + "loss": 0.0904, + "step": 6694 + }, + { + "epoch": 0.88, + "learning_rate": 1.9669419854179107e-05, + "loss": 0.0923, + "step": 6695 + }, + { + "epoch": 0.88, + "learning_rate": 1.9669284609565225e-05, + "loss": 0.0947, + "step": 6696 + }, + { + "epoch": 0.88, + "learning_rate": 1.9669149337756995e-05, + "loss": 0.0974, + "step": 6697 + }, + { + "epoch": 0.88, + "learning_rate": 1.96690140387548e-05, + "loss": 0.087, + "step": 6698 + }, + { + "epoch": 0.88, + "learning_rate": 1.9668878712559013e-05, + "loss": 0.0864, + "step": 6699 + }, + { + "epoch": 0.88, + "learning_rate": 1.966874335917002e-05, + "loss": 0.0983, + "step": 6700 + }, + { + "epoch": 0.88, + "learning_rate": 1.9668607978588205e-05, + "loss": 0.0928, + "step": 6701 + }, + { + "epoch": 0.88, + "learning_rate": 1.9668472570813945e-05, + "loss": 0.0899, + "step": 6702 + }, + { + "epoch": 0.88, + "learning_rate": 1.9668337135847616e-05, + "loss": 0.0908, + "step": 6703 + }, + { + "epoch": 0.88, + "learning_rate": 1.9668201673689608e-05, + "loss": 0.0829, + "step": 6704 + }, + { + "epoch": 0.88, + "learning_rate": 1.9668066184340295e-05, + "loss": 0.0948, + "step": 6705 + }, + { + "epoch": 0.88, + "learning_rate": 1.966793066780006e-05, + "loss": 0.082, + "step": 6706 + }, + { + "epoch": 0.88, + "learning_rate": 1.9667795124069288e-05, + "loss": 0.0778, + "step": 6707 + }, + { + "epoch": 0.88, + "learning_rate": 1.9667659553148357e-05, + "loss": 0.0866, + "step": 6708 + }, + { + "epoch": 0.88, + "learning_rate": 1.9667523955037645e-05, + "loss": 0.092, + "step": 6709 + }, + { + "epoch": 0.88, + "learning_rate": 1.966738832973754e-05, + "loss": 0.0908, + "step": 6710 + }, + { + "epoch": 0.88, + "learning_rate": 1.9667252677248418e-05, + "loss": 0.0986, + "step": 6711 + }, + { + "epoch": 0.88, + "learning_rate": 1.9667116997570665e-05, + "loss": 0.0858, + "step": 6712 + }, + { + "epoch": 0.88, + "learning_rate": 1.966698129070466e-05, + "loss": 0.084, + "step": 6713 + }, + { + "epoch": 0.88, + "learning_rate": 1.9666845556650784e-05, + "loss": 0.0759, + "step": 6714 + }, + { + "epoch": 0.88, + "learning_rate": 1.966670979540942e-05, + "loss": 0.0867, + "step": 6715 + }, + { + "epoch": 0.88, + "learning_rate": 1.966657400698095e-05, + "loss": 0.0946, + "step": 6716 + }, + { + "epoch": 0.88, + "learning_rate": 1.9666438191365752e-05, + "loss": 0.0814, + "step": 6717 + }, + { + "epoch": 0.88, + "learning_rate": 1.9666302348564216e-05, + "loss": 0.089, + "step": 6718 + }, + { + "epoch": 0.88, + "learning_rate": 1.9666166478576714e-05, + "loss": 0.0943, + "step": 6719 + }, + { + "epoch": 0.88, + "learning_rate": 1.966603058140364e-05, + "loss": 0.0933, + "step": 6720 + }, + { + "epoch": 0.88, + "learning_rate": 1.9665894657045362e-05, + "loss": 0.086, + "step": 6721 + }, + { + "epoch": 0.88, + "learning_rate": 1.9665758705502276e-05, + "loss": 0.0936, + "step": 6722 + }, + { + "epoch": 0.88, + "learning_rate": 1.9665622726774757e-05, + "loss": 0.0832, + "step": 6723 + }, + { + "epoch": 0.88, + "learning_rate": 1.9665486720863187e-05, + "loss": 0.0964, + "step": 6724 + }, + { + "epoch": 0.88, + "learning_rate": 1.9665350687767954e-05, + "loss": 0.0774, + "step": 6725 + }, + { + "epoch": 0.88, + "learning_rate": 1.9665214627489435e-05, + "loss": 0.0775, + "step": 6726 + }, + { + "epoch": 0.88, + "learning_rate": 1.9665078540028013e-05, + "loss": 0.0838, + "step": 6727 + }, + { + "epoch": 0.88, + "learning_rate": 1.9664942425384073e-05, + "loss": 0.0848, + "step": 6728 + }, + { + "epoch": 0.88, + "learning_rate": 1.9664806283558e-05, + "loss": 0.0781, + "step": 6729 + }, + { + "epoch": 0.88, + "learning_rate": 1.966467011455017e-05, + "loss": 0.0936, + "step": 6730 + }, + { + "epoch": 0.88, + "learning_rate": 1.966453391836097e-05, + "loss": 0.0879, + "step": 6731 + }, + { + "epoch": 0.88, + "learning_rate": 1.9664397694990787e-05, + "loss": 0.0777, + "step": 6732 + }, + { + "epoch": 0.88, + "learning_rate": 1.9664261444439997e-05, + "loss": 0.0891, + "step": 6733 + }, + { + "epoch": 0.88, + "learning_rate": 1.9664125166708988e-05, + "loss": 0.0815, + "step": 6734 + }, + { + "epoch": 0.88, + "learning_rate": 1.966398886179814e-05, + "loss": 0.0799, + "step": 6735 + }, + { + "epoch": 0.88, + "learning_rate": 1.9663852529707837e-05, + "loss": 0.0838, + "step": 6736 + }, + { + "epoch": 0.88, + "learning_rate": 1.966371617043847e-05, + "loss": 0.1, + "step": 6737 + }, + { + "epoch": 0.88, + "learning_rate": 1.966357978399041e-05, + "loss": 0.0818, + "step": 6738 + }, + { + "epoch": 0.88, + "learning_rate": 1.9663443370364045e-05, + "loss": 0.0833, + "step": 6739 + }, + { + "epoch": 0.88, + "learning_rate": 1.9663306929559765e-05, + "loss": 0.0946, + "step": 6740 + }, + { + "epoch": 0.88, + "learning_rate": 1.9663170461577946e-05, + "loss": 0.0886, + "step": 6741 + }, + { + "epoch": 0.88, + "learning_rate": 1.9663033966418973e-05, + "loss": 0.0977, + "step": 6742 + }, + { + "epoch": 0.88, + "learning_rate": 1.9662897444083236e-05, + "loss": 0.0825, + "step": 6743 + }, + { + "epoch": 0.88, + "learning_rate": 1.966276089457111e-05, + "loss": 0.0777, + "step": 6744 + }, + { + "epoch": 0.88, + "learning_rate": 1.9662624317882984e-05, + "loss": 0.0886, + "step": 6745 + }, + { + "epoch": 0.88, + "learning_rate": 1.9662487714019248e-05, + "loss": 0.0909, + "step": 6746 + }, + { + "epoch": 0.88, + "learning_rate": 1.9662351082980273e-05, + "loss": 0.0869, + "step": 6747 + }, + { + "epoch": 0.88, + "learning_rate": 1.966221442476645e-05, + "loss": 0.0846, + "step": 6748 + }, + { + "epoch": 0.88, + "learning_rate": 1.966207773937817e-05, + "loss": 0.0822, + "step": 6749 + }, + { + "epoch": 0.88, + "learning_rate": 1.9661941026815804e-05, + "loss": 0.0947, + "step": 6750 + }, + { + "epoch": 0.88, + "learning_rate": 1.966180428707975e-05, + "loss": 0.0889, + "step": 6751 + }, + { + "epoch": 0.88, + "learning_rate": 1.966166752017038e-05, + "loss": 0.0828, + "step": 6752 + }, + { + "epoch": 0.88, + "learning_rate": 1.9661530726088088e-05, + "loss": 0.0872, + "step": 6753 + }, + { + "epoch": 0.88, + "learning_rate": 1.9661393904833255e-05, + "loss": 0.0857, + "step": 6754 + }, + { + "epoch": 0.88, + "learning_rate": 1.9661257056406265e-05, + "loss": 0.0872, + "step": 6755 + }, + { + "epoch": 0.88, + "learning_rate": 1.9661120180807502e-05, + "loss": 0.0882, + "step": 6756 + }, + { + "epoch": 0.89, + "learning_rate": 1.9660983278037356e-05, + "loss": 0.0951, + "step": 6757 + }, + { + "epoch": 0.89, + "learning_rate": 1.966084634809621e-05, + "loss": 0.1022, + "step": 6758 + }, + { + "epoch": 0.89, + "learning_rate": 1.9660709390984446e-05, + "loss": 0.0935, + "step": 6759 + }, + { + "epoch": 0.89, + "learning_rate": 1.9660572406702452e-05, + "loss": 0.0955, + "step": 6760 + }, + { + "epoch": 0.89, + "learning_rate": 1.9660435395250613e-05, + "loss": 0.0804, + "step": 6761 + }, + { + "epoch": 0.89, + "learning_rate": 1.9660298356629314e-05, + "loss": 0.0888, + "step": 6762 + }, + { + "epoch": 0.89, + "learning_rate": 1.9660161290838936e-05, + "loss": 0.0834, + "step": 6763 + }, + { + "epoch": 0.89, + "learning_rate": 1.9660024197879873e-05, + "loss": 0.0854, + "step": 6764 + }, + { + "epoch": 0.89, + "learning_rate": 1.9659887077752507e-05, + "loss": 0.0884, + "step": 6765 + }, + { + "epoch": 0.89, + "learning_rate": 1.965974993045722e-05, + "loss": 0.0868, + "step": 6766 + }, + { + "epoch": 0.89, + "learning_rate": 1.9659612755994405e-05, + "loss": 0.0771, + "step": 6767 + }, + { + "epoch": 0.89, + "learning_rate": 1.965947555436444e-05, + "loss": 0.0782, + "step": 6768 + }, + { + "epoch": 0.89, + "learning_rate": 1.9659338325567715e-05, + "loss": 0.0783, + "step": 6769 + }, + { + "epoch": 0.89, + "learning_rate": 1.9659201069604616e-05, + "loss": 0.0849, + "step": 6770 + }, + { + "epoch": 0.89, + "learning_rate": 1.9659063786475528e-05, + "loss": 0.081, + "step": 6771 + }, + { + "epoch": 0.89, + "learning_rate": 1.965892647618084e-05, + "loss": 0.0807, + "step": 6772 + }, + { + "epoch": 0.89, + "learning_rate": 1.965878913872093e-05, + "loss": 0.0937, + "step": 6773 + }, + { + "epoch": 0.89, + "learning_rate": 1.9658651774096195e-05, + "loss": 0.0981, + "step": 6774 + }, + { + "epoch": 0.89, + "learning_rate": 1.9658514382307013e-05, + "loss": 0.0895, + "step": 6775 + }, + { + "epoch": 0.89, + "learning_rate": 1.9658376963353775e-05, + "loss": 0.099, + "step": 6776 + }, + { + "epoch": 0.89, + "learning_rate": 1.9658239517236865e-05, + "loss": 0.0881, + "step": 6777 + }, + { + "epoch": 0.89, + "learning_rate": 1.9658102043956673e-05, + "loss": 0.0782, + "step": 6778 + }, + { + "epoch": 0.89, + "learning_rate": 1.965796454351358e-05, + "loss": 0.0859, + "step": 6779 + }, + { + "epoch": 0.89, + "learning_rate": 1.965782701590798e-05, + "loss": 0.0895, + "step": 6780 + }, + { + "epoch": 0.89, + "learning_rate": 1.9657689461140254e-05, + "loss": 0.0805, + "step": 6781 + }, + { + "epoch": 0.89, + "learning_rate": 1.965755187921079e-05, + "loss": 0.0902, + "step": 6782 + }, + { + "epoch": 0.89, + "learning_rate": 1.965741427011998e-05, + "loss": 0.0816, + "step": 6783 + }, + { + "epoch": 0.89, + "learning_rate": 1.9657276633868202e-05, + "loss": 0.0878, + "step": 6784 + }, + { + "epoch": 0.89, + "learning_rate": 1.965713897045585e-05, + "loss": 0.0935, + "step": 6785 + }, + { + "epoch": 0.89, + "learning_rate": 1.9657001279883302e-05, + "loss": 0.0746, + "step": 6786 + }, + { + "epoch": 0.89, + "learning_rate": 1.965686356215096e-05, + "loss": 0.0899, + "step": 6787 + }, + { + "epoch": 0.89, + "learning_rate": 1.96567258172592e-05, + "loss": 0.0966, + "step": 6788 + }, + { + "epoch": 0.89, + "learning_rate": 1.9656588045208416e-05, + "loss": 0.0914, + "step": 6789 + }, + { + "epoch": 0.89, + "learning_rate": 1.9656450245998992e-05, + "loss": 0.0845, + "step": 6790 + }, + { + "epoch": 0.89, + "learning_rate": 1.9656312419631313e-05, + "loss": 0.0881, + "step": 6791 + }, + { + "epoch": 0.89, + "learning_rate": 1.9656174566105772e-05, + "loss": 0.0889, + "step": 6792 + }, + { + "epoch": 0.89, + "learning_rate": 1.9656036685422752e-05, + "loss": 0.0807, + "step": 6793 + }, + { + "epoch": 0.89, + "learning_rate": 1.9655898777582646e-05, + "loss": 0.0722, + "step": 6794 + }, + { + "epoch": 0.89, + "learning_rate": 1.9655760842585838e-05, + "loss": 0.088, + "step": 6795 + }, + { + "epoch": 0.89, + "learning_rate": 1.9655622880432716e-05, + "loss": 0.0886, + "step": 6796 + }, + { + "epoch": 0.89, + "learning_rate": 1.9655484891123666e-05, + "loss": 0.0874, + "step": 6797 + }, + { + "epoch": 0.89, + "learning_rate": 1.9655346874659083e-05, + "loss": 0.0814, + "step": 6798 + }, + { + "epoch": 0.89, + "learning_rate": 1.965520883103935e-05, + "loss": 0.0866, + "step": 6799 + }, + { + "epoch": 0.89, + "learning_rate": 1.9655070760264856e-05, + "loss": 0.079, + "step": 6800 + }, + { + "epoch": 0.89, + "learning_rate": 1.9654932662335988e-05, + "loss": 0.0951, + "step": 6801 + }, + { + "epoch": 0.89, + "learning_rate": 1.965479453725314e-05, + "loss": 0.0922, + "step": 6802 + }, + { + "epoch": 0.89, + "learning_rate": 1.9654656385016695e-05, + "loss": 0.0814, + "step": 6803 + }, + { + "epoch": 0.89, + "learning_rate": 1.965451820562704e-05, + "loss": 0.078, + "step": 6804 + }, + { + "epoch": 0.89, + "learning_rate": 1.9654379999084572e-05, + "loss": 0.0885, + "step": 6805 + }, + { + "epoch": 0.89, + "learning_rate": 1.965424176538967e-05, + "loss": 0.0842, + "step": 6806 + }, + { + "epoch": 0.89, + "learning_rate": 1.9654103504542727e-05, + "loss": 0.0863, + "step": 6807 + }, + { + "epoch": 0.89, + "learning_rate": 1.9653965216544135e-05, + "loss": 0.0863, + "step": 6808 + }, + { + "epoch": 0.89, + "learning_rate": 1.965382690139428e-05, + "loss": 0.0823, + "step": 6809 + }, + { + "epoch": 0.89, + "learning_rate": 1.9653688559093548e-05, + "loss": 0.0762, + "step": 6810 + }, + { + "epoch": 0.89, + "learning_rate": 1.9653550189642333e-05, + "loss": 0.0913, + "step": 6811 + }, + { + "epoch": 0.89, + "learning_rate": 1.9653411793041025e-05, + "loss": 0.0769, + "step": 6812 + }, + { + "epoch": 0.89, + "learning_rate": 1.9653273369290004e-05, + "loss": 0.0786, + "step": 6813 + }, + { + "epoch": 0.89, + "learning_rate": 1.965313491838967e-05, + "loss": 0.0826, + "step": 6814 + }, + { + "epoch": 0.89, + "learning_rate": 1.9652996440340407e-05, + "loss": 0.0902, + "step": 6815 + }, + { + "epoch": 0.89, + "learning_rate": 1.9652857935142604e-05, + "loss": 0.0911, + "step": 6816 + }, + { + "epoch": 0.89, + "learning_rate": 1.9652719402796653e-05, + "loss": 0.1, + "step": 6817 + }, + { + "epoch": 0.89, + "learning_rate": 1.9652580843302944e-05, + "loss": 0.0811, + "step": 6818 + }, + { + "epoch": 0.89, + "learning_rate": 1.9652442256661866e-05, + "loss": 0.0904, + "step": 6819 + }, + { + "epoch": 0.89, + "learning_rate": 1.9652303642873805e-05, + "loss": 0.0898, + "step": 6820 + }, + { + "epoch": 0.89, + "learning_rate": 1.9652165001939155e-05, + "loss": 0.0823, + "step": 6821 + }, + { + "epoch": 0.89, + "learning_rate": 1.9652026333858306e-05, + "loss": 0.0857, + "step": 6822 + }, + { + "epoch": 0.89, + "learning_rate": 1.965188763863165e-05, + "loss": 0.0899, + "step": 6823 + }, + { + "epoch": 0.89, + "learning_rate": 1.965174891625957e-05, + "loss": 0.0935, + "step": 6824 + }, + { + "epoch": 0.89, + "learning_rate": 1.965161016674246e-05, + "loss": 0.0863, + "step": 6825 + }, + { + "epoch": 0.89, + "learning_rate": 1.9651471390080712e-05, + "loss": 0.0873, + "step": 6826 + }, + { + "epoch": 0.89, + "learning_rate": 1.9651332586274715e-05, + "loss": 0.0816, + "step": 6827 + }, + { + "epoch": 0.89, + "learning_rate": 1.9651193755324856e-05, + "loss": 0.0994, + "step": 6828 + }, + { + "epoch": 0.89, + "learning_rate": 1.965105489723153e-05, + "loss": 0.0843, + "step": 6829 + }, + { + "epoch": 0.89, + "learning_rate": 1.965091601199513e-05, + "loss": 0.0914, + "step": 6830 + }, + { + "epoch": 0.89, + "learning_rate": 1.9650777099616037e-05, + "loss": 0.0868, + "step": 6831 + }, + { + "epoch": 0.89, + "learning_rate": 1.9650638160094653e-05, + "loss": 0.0846, + "step": 6832 + }, + { + "epoch": 0.9, + "learning_rate": 1.965049919343136e-05, + "loss": 0.094, + "step": 6833 + }, + { + "epoch": 0.9, + "learning_rate": 1.965036019962655e-05, + "loss": 0.0828, + "step": 6834 + }, + { + "epoch": 0.9, + "learning_rate": 1.965022117868062e-05, + "loss": 0.0826, + "step": 6835 + }, + { + "epoch": 0.9, + "learning_rate": 1.965008213059395e-05, + "loss": 0.0916, + "step": 6836 + }, + { + "epoch": 0.9, + "learning_rate": 1.9649943055366946e-05, + "loss": 0.0931, + "step": 6837 + }, + { + "epoch": 0.9, + "learning_rate": 1.9649803952999985e-05, + "loss": 0.078, + "step": 6838 + }, + { + "epoch": 0.9, + "learning_rate": 1.964966482349347e-05, + "loss": 0.08, + "step": 6839 + }, + { + "epoch": 0.9, + "learning_rate": 1.9649525666847783e-05, + "loss": 0.0839, + "step": 6840 + }, + { + "epoch": 0.9, + "learning_rate": 1.9649386483063317e-05, + "loss": 0.0848, + "step": 6841 + }, + { + "epoch": 0.9, + "learning_rate": 1.9649247272140468e-05, + "loss": 0.0869, + "step": 6842 + }, + { + "epoch": 0.9, + "learning_rate": 1.9649108034079625e-05, + "loss": 0.0899, + "step": 6843 + }, + { + "epoch": 0.9, + "learning_rate": 1.9648968768881178e-05, + "loss": 0.0869, + "step": 6844 + }, + { + "epoch": 0.9, + "learning_rate": 1.964882947654552e-05, + "loss": 0.0841, + "step": 6845 + }, + { + "epoch": 0.9, + "learning_rate": 1.9648690157073047e-05, + "loss": 0.0877, + "step": 6846 + }, + { + "epoch": 0.9, + "learning_rate": 1.9648550810464143e-05, + "loss": 0.0796, + "step": 6847 + }, + { + "epoch": 0.9, + "learning_rate": 1.9648411436719203e-05, + "loss": 0.0881, + "step": 6848 + }, + { + "epoch": 0.9, + "learning_rate": 1.964827203583862e-05, + "loss": 0.0861, + "step": 6849 + }, + { + "epoch": 0.9, + "learning_rate": 1.964813260782279e-05, + "loss": 0.0783, + "step": 6850 + }, + { + "epoch": 0.9, + "learning_rate": 1.9647993152672095e-05, + "loss": 0.0846, + "step": 6851 + }, + { + "epoch": 0.9, + "learning_rate": 1.9647853670386937e-05, + "loss": 0.0881, + "step": 6852 + }, + { + "epoch": 0.9, + "learning_rate": 1.9647714160967705e-05, + "loss": 0.0825, + "step": 6853 + }, + { + "epoch": 0.9, + "learning_rate": 1.9647574624414786e-05, + "loss": 0.0821, + "step": 6854 + }, + { + "epoch": 0.9, + "learning_rate": 1.9647435060728583e-05, + "loss": 0.0912, + "step": 6855 + }, + { + "epoch": 0.9, + "learning_rate": 1.9647295469909478e-05, + "loss": 0.086, + "step": 6856 + }, + { + "epoch": 0.9, + "learning_rate": 1.9647155851957872e-05, + "loss": 0.0812, + "step": 6857 + }, + { + "epoch": 0.9, + "learning_rate": 1.9647016206874147e-05, + "loss": 0.0953, + "step": 6858 + }, + { + "epoch": 0.9, + "learning_rate": 1.964687653465871e-05, + "loss": 0.0783, + "step": 6859 + }, + { + "epoch": 0.9, + "learning_rate": 1.9646736835311944e-05, + "loss": 0.0877, + "step": 6860 + }, + { + "epoch": 0.9, + "learning_rate": 1.9646597108834245e-05, + "loss": 0.0813, + "step": 6861 + }, + { + "epoch": 0.9, + "learning_rate": 1.9646457355226002e-05, + "loss": 0.0806, + "step": 6862 + }, + { + "epoch": 0.9, + "learning_rate": 1.9646317574487613e-05, + "loss": 0.0856, + "step": 6863 + }, + { + "epoch": 0.9, + "learning_rate": 1.9646177766619473e-05, + "loss": 0.0813, + "step": 6864 + }, + { + "epoch": 0.9, + "learning_rate": 1.964603793162197e-05, + "loss": 0.0878, + "step": 6865 + }, + { + "epoch": 0.9, + "learning_rate": 1.9645898069495498e-05, + "loss": 0.0908, + "step": 6866 + }, + { + "epoch": 0.9, + "learning_rate": 1.9645758180240452e-05, + "loss": 0.0856, + "step": 6867 + }, + { + "epoch": 0.9, + "learning_rate": 1.9645618263857224e-05, + "loss": 0.0767, + "step": 6868 + }, + { + "epoch": 0.9, + "learning_rate": 1.9645478320346204e-05, + "loss": 0.0789, + "step": 6869 + }, + { + "epoch": 0.9, + "learning_rate": 1.9645338349707795e-05, + "loss": 0.0882, + "step": 6870 + }, + { + "epoch": 0.9, + "learning_rate": 1.9645198351942387e-05, + "loss": 0.0802, + "step": 6871 + }, + { + "epoch": 0.9, + "learning_rate": 1.964505832705037e-05, + "loss": 0.0983, + "step": 6872 + }, + { + "epoch": 0.9, + "learning_rate": 1.9644918275032138e-05, + "loss": 0.0846, + "step": 6873 + }, + { + "epoch": 0.9, + "learning_rate": 1.964477819588809e-05, + "loss": 0.085, + "step": 6874 + }, + { + "epoch": 0.9, + "learning_rate": 1.9644638089618616e-05, + "loss": 0.0836, + "step": 6875 + }, + { + "epoch": 0.9, + "learning_rate": 1.9644497956224113e-05, + "loss": 0.0946, + "step": 6876 + }, + { + "epoch": 0.9, + "learning_rate": 1.964435779570497e-05, + "loss": 0.091, + "step": 6877 + }, + { + "epoch": 0.9, + "learning_rate": 1.9644217608061585e-05, + "loss": 0.097, + "step": 6878 + }, + { + "epoch": 0.9, + "learning_rate": 1.9644077393294353e-05, + "loss": 0.0886, + "step": 6879 + }, + { + "epoch": 0.9, + "learning_rate": 1.9643937151403668e-05, + "loss": 0.0913, + "step": 6880 + }, + { + "epoch": 0.9, + "learning_rate": 1.964379688238992e-05, + "loss": 0.0769, + "step": 6881 + }, + { + "epoch": 0.9, + "learning_rate": 1.9643656586253508e-05, + "loss": 0.0896, + "step": 6882 + }, + { + "epoch": 0.9, + "learning_rate": 1.9643516262994826e-05, + "loss": 0.094, + "step": 6883 + }, + { + "epoch": 0.9, + "learning_rate": 1.9643375912614267e-05, + "loss": 0.0852, + "step": 6884 + }, + { + "epoch": 0.9, + "learning_rate": 1.964323553511223e-05, + "loss": 0.0919, + "step": 6885 + }, + { + "epoch": 0.9, + "learning_rate": 1.9643095130489103e-05, + "loss": 0.0774, + "step": 6886 + }, + { + "epoch": 0.9, + "learning_rate": 1.9642954698745287e-05, + "loss": 0.0834, + "step": 6887 + }, + { + "epoch": 0.9, + "learning_rate": 1.9642814239881173e-05, + "loss": 0.091, + "step": 6888 + }, + { + "epoch": 0.9, + "learning_rate": 1.9642673753897155e-05, + "loss": 0.0896, + "step": 6889 + }, + { + "epoch": 0.9, + "learning_rate": 1.9642533240793632e-05, + "loss": 0.0858, + "step": 6890 + }, + { + "epoch": 0.9, + "learning_rate": 1.9642392700571e-05, + "loss": 0.0829, + "step": 6891 + }, + { + "epoch": 0.9, + "learning_rate": 1.964225213322965e-05, + "loss": 0.0875, + "step": 6892 + }, + { + "epoch": 0.9, + "learning_rate": 1.964211153876998e-05, + "loss": 0.0886, + "step": 6893 + }, + { + "epoch": 0.9, + "learning_rate": 1.9641970917192385e-05, + "loss": 0.0854, + "step": 6894 + }, + { + "epoch": 0.9, + "learning_rate": 1.964183026849726e-05, + "loss": 0.0974, + "step": 6895 + }, + { + "epoch": 0.9, + "learning_rate": 1.9641689592685e-05, + "loss": 0.0896, + "step": 6896 + }, + { + "epoch": 0.9, + "learning_rate": 1.9641548889756005e-05, + "loss": 0.0789, + "step": 6897 + }, + { + "epoch": 0.9, + "learning_rate": 1.9641408159710664e-05, + "loss": 0.0856, + "step": 6898 + }, + { + "epoch": 0.9, + "learning_rate": 1.9641267402549377e-05, + "loss": 0.0951, + "step": 6899 + }, + { + "epoch": 0.9, + "learning_rate": 1.964112661827254e-05, + "loss": 0.0874, + "step": 6900 + }, + { + "epoch": 0.9, + "learning_rate": 1.9640985806880546e-05, + "loss": 0.0838, + "step": 6901 + }, + { + "epoch": 0.9, + "learning_rate": 1.9640844968373792e-05, + "loss": 0.0812, + "step": 6902 + }, + { + "epoch": 0.9, + "learning_rate": 1.9640704102752677e-05, + "loss": 0.0801, + "step": 6903 + }, + { + "epoch": 0.9, + "learning_rate": 1.9640563210017593e-05, + "loss": 0.0767, + "step": 6904 + }, + { + "epoch": 0.9, + "learning_rate": 1.964042229016894e-05, + "loss": 0.0748, + "step": 6905 + }, + { + "epoch": 0.9, + "learning_rate": 1.9640281343207112e-05, + "loss": 0.0812, + "step": 6906 + }, + { + "epoch": 0.9, + "learning_rate": 1.9640140369132504e-05, + "loss": 0.0907, + "step": 6907 + }, + { + "epoch": 0.9, + "learning_rate": 1.963999936794552e-05, + "loss": 0.0888, + "step": 6908 + }, + { + "epoch": 0.91, + "learning_rate": 1.9639858339646547e-05, + "loss": 0.0891, + "step": 6909 + }, + { + "epoch": 0.91, + "learning_rate": 1.9639717284235987e-05, + "loss": 0.0865, + "step": 6910 + }, + { + "epoch": 0.91, + "learning_rate": 1.9639576201714234e-05, + "loss": 0.093, + "step": 6911 + }, + { + "epoch": 0.91, + "learning_rate": 1.963943509208169e-05, + "loss": 0.0948, + "step": 6912 + }, + { + "epoch": 0.91, + "learning_rate": 1.9639293955338745e-05, + "loss": 0.1002, + "step": 6913 + }, + { + "epoch": 0.91, + "learning_rate": 1.96391527914858e-05, + "loss": 0.0787, + "step": 6914 + }, + { + "epoch": 0.91, + "learning_rate": 1.963901160052325e-05, + "loss": 0.0945, + "step": 6915 + }, + { + "epoch": 0.91, + "learning_rate": 1.9638870382451494e-05, + "loss": 0.0805, + "step": 6916 + }, + { + "epoch": 0.91, + "learning_rate": 1.963872913727093e-05, + "loss": 0.0825, + "step": 6917 + }, + { + "epoch": 0.91, + "learning_rate": 1.963858786498195e-05, + "loss": 0.0763, + "step": 6918 + }, + { + "epoch": 0.91, + "learning_rate": 1.9638446565584957e-05, + "loss": 0.0775, + "step": 6919 + }, + { + "epoch": 0.91, + "learning_rate": 1.9638305239080347e-05, + "loss": 0.0829, + "step": 6920 + }, + { + "epoch": 0.91, + "learning_rate": 1.9638163885468514e-05, + "loss": 0.0946, + "step": 6921 + }, + { + "epoch": 0.91, + "learning_rate": 1.963802250474986e-05, + "loss": 0.0911, + "step": 6922 + }, + { + "epoch": 0.91, + "learning_rate": 1.9637881096924786e-05, + "loss": 0.0838, + "step": 6923 + }, + { + "epoch": 0.91, + "learning_rate": 1.9637739661993677e-05, + "loss": 0.0919, + "step": 6924 + }, + { + "epoch": 0.91, + "learning_rate": 1.963759819995694e-05, + "loss": 0.098, + "step": 6925 + }, + { + "epoch": 0.91, + "learning_rate": 1.9637456710814975e-05, + "loss": 0.096, + "step": 6926 + }, + { + "epoch": 0.91, + "learning_rate": 1.9637315194568174e-05, + "loss": 0.0914, + "step": 6927 + }, + { + "epoch": 0.91, + "learning_rate": 1.9637173651216936e-05, + "loss": 0.0851, + "step": 6928 + }, + { + "epoch": 0.91, + "learning_rate": 1.9637032080761664e-05, + "loss": 0.0835, + "step": 6929 + }, + { + "epoch": 0.91, + "learning_rate": 1.9636890483202746e-05, + "loss": 0.0766, + "step": 6930 + }, + { + "epoch": 0.91, + "learning_rate": 1.9636748858540594e-05, + "loss": 0.0885, + "step": 6931 + }, + { + "epoch": 0.91, + "learning_rate": 1.9636607206775596e-05, + "loss": 0.0824, + "step": 6932 + }, + { + "epoch": 0.91, + "learning_rate": 1.963646552790815e-05, + "loss": 0.0789, + "step": 6933 + }, + { + "epoch": 0.91, + "learning_rate": 1.963632382193866e-05, + "loss": 0.0798, + "step": 6934 + }, + { + "epoch": 0.91, + "learning_rate": 1.9636182088867526e-05, + "loss": 0.0976, + "step": 6935 + }, + { + "epoch": 0.91, + "learning_rate": 1.963604032869514e-05, + "loss": 0.0881, + "step": 6936 + }, + { + "epoch": 0.91, + "learning_rate": 1.9635898541421902e-05, + "loss": 0.0824, + "step": 6937 + }, + { + "epoch": 0.91, + "learning_rate": 1.9635756727048215e-05, + "loss": 0.0898, + "step": 6938 + }, + { + "epoch": 0.91, + "learning_rate": 1.9635614885574478e-05, + "loss": 0.0774, + "step": 6939 + }, + { + "epoch": 0.91, + "learning_rate": 1.9635473017001084e-05, + "loss": 0.0859, + "step": 6940 + }, + { + "epoch": 0.91, + "learning_rate": 1.9635331121328437e-05, + "loss": 0.0828, + "step": 6941 + }, + { + "epoch": 0.91, + "learning_rate": 1.963518919855693e-05, + "loss": 0.0909, + "step": 6942 + }, + { + "epoch": 0.91, + "learning_rate": 1.963504724868697e-05, + "loss": 0.1007, + "step": 6943 + }, + { + "epoch": 0.91, + "learning_rate": 1.963490527171895e-05, + "loss": 0.0754, + "step": 6944 + }, + { + "epoch": 0.91, + "learning_rate": 1.9634763267653273e-05, + "loss": 0.0822, + "step": 6945 + }, + { + "epoch": 0.91, + "learning_rate": 1.9634621236490343e-05, + "loss": 0.0924, + "step": 6946 + }, + { + "epoch": 0.91, + "learning_rate": 1.9634479178230544e-05, + "loss": 0.0733, + "step": 6947 + }, + { + "epoch": 0.91, + "learning_rate": 1.9634337092874294e-05, + "loss": 0.0803, + "step": 6948 + }, + { + "epoch": 0.91, + "learning_rate": 1.9634194980421977e-05, + "loss": 0.0833, + "step": 6949 + }, + { + "epoch": 0.91, + "learning_rate": 1.9634052840874006e-05, + "loss": 0.0753, + "step": 6950 + }, + { + "epoch": 0.91, + "learning_rate": 1.963391067423077e-05, + "loss": 0.0708, + "step": 6951 + }, + { + "epoch": 0.91, + "learning_rate": 1.9633768480492678e-05, + "loss": 0.083, + "step": 6952 + }, + { + "epoch": 0.91, + "learning_rate": 1.963362625966012e-05, + "loss": 0.0732, + "step": 6953 + }, + { + "epoch": 0.91, + "learning_rate": 1.9633484011733504e-05, + "loss": 0.0864, + "step": 6954 + }, + { + "epoch": 0.91, + "learning_rate": 1.9633341736713227e-05, + "loss": 0.0969, + "step": 6955 + }, + { + "epoch": 0.91, + "learning_rate": 1.9633199434599687e-05, + "loss": 0.0858, + "step": 6956 + }, + { + "epoch": 0.91, + "learning_rate": 1.963305710539329e-05, + "loss": 0.0905, + "step": 6957 + }, + { + "epoch": 0.91, + "learning_rate": 1.9632914749094432e-05, + "loss": 0.0898, + "step": 6958 + }, + { + "epoch": 0.91, + "learning_rate": 1.9632772365703513e-05, + "loss": 0.0896, + "step": 6959 + }, + { + "epoch": 0.91, + "learning_rate": 1.9632629955220937e-05, + "loss": 0.0909, + "step": 6960 + }, + { + "epoch": 0.91, + "learning_rate": 1.96324875176471e-05, + "loss": 0.0805, + "step": 6961 + }, + { + "epoch": 0.91, + "learning_rate": 1.9632345052982407e-05, + "loss": 0.0976, + "step": 6962 + }, + { + "epoch": 0.91, + "learning_rate": 1.9632202561227253e-05, + "loss": 0.0853, + "step": 6963 + }, + { + "epoch": 0.91, + "learning_rate": 1.9632060042382044e-05, + "loss": 0.0887, + "step": 6964 + }, + { + "epoch": 0.91, + "learning_rate": 1.963191749644718e-05, + "loss": 0.0811, + "step": 6965 + }, + { + "epoch": 0.91, + "learning_rate": 1.963177492342306e-05, + "loss": 0.0841, + "step": 6966 + }, + { + "epoch": 0.91, + "learning_rate": 1.9631632323310085e-05, + "loss": 0.0891, + "step": 6967 + }, + { + "epoch": 0.91, + "learning_rate": 1.9631489696108657e-05, + "loss": 0.0869, + "step": 6968 + }, + { + "epoch": 0.91, + "learning_rate": 1.963134704181918e-05, + "loss": 0.0861, + "step": 6969 + }, + { + "epoch": 0.91, + "learning_rate": 1.963120436044205e-05, + "loss": 0.0973, + "step": 6970 + }, + { + "epoch": 0.91, + "learning_rate": 1.963106165197767e-05, + "loss": 0.0843, + "step": 6971 + }, + { + "epoch": 0.91, + "learning_rate": 1.9630918916426445e-05, + "loss": 0.0803, + "step": 6972 + }, + { + "epoch": 0.91, + "learning_rate": 1.9630776153788772e-05, + "loss": 0.0801, + "step": 6973 + }, + { + "epoch": 0.91, + "learning_rate": 1.9630633364065052e-05, + "loss": 0.0848, + "step": 6974 + }, + { + "epoch": 0.91, + "learning_rate": 1.963049054725569e-05, + "loss": 0.0778, + "step": 6975 + }, + { + "epoch": 0.91, + "learning_rate": 1.9630347703361088e-05, + "loss": 0.0844, + "step": 6976 + }, + { + "epoch": 0.91, + "learning_rate": 1.9630204832381644e-05, + "loss": 0.0892, + "step": 6977 + }, + { + "epoch": 0.91, + "learning_rate": 1.963006193431776e-05, + "loss": 0.0891, + "step": 6978 + }, + { + "epoch": 0.91, + "learning_rate": 1.9629919009169845e-05, + "loss": 0.0848, + "step": 6979 + }, + { + "epoch": 0.91, + "learning_rate": 1.962977605693829e-05, + "loss": 0.0863, + "step": 6980 + }, + { + "epoch": 0.91, + "learning_rate": 1.9629633077623508e-05, + "loss": 0.0786, + "step": 6981 + }, + { + "epoch": 0.91, + "learning_rate": 1.962949007122589e-05, + "loss": 0.0717, + "step": 6982 + }, + { + "epoch": 0.91, + "learning_rate": 1.962934703774585e-05, + "loss": 0.093, + "step": 6983 + }, + { + "epoch": 0.91, + "learning_rate": 1.9629203977183778e-05, + "loss": 0.0865, + "step": 6984 + }, + { + "epoch": 0.91, + "learning_rate": 1.9629060889540085e-05, + "loss": 0.0911, + "step": 6985 + }, + { + "epoch": 0.92, + "learning_rate": 1.962891777481517e-05, + "loss": 0.0869, + "step": 6986 + }, + { + "epoch": 0.92, + "learning_rate": 1.962877463300944e-05, + "loss": 0.0888, + "step": 6987 + }, + { + "epoch": 0.92, + "learning_rate": 1.9628631464123293e-05, + "loss": 0.0843, + "step": 6988 + }, + { + "epoch": 0.92, + "learning_rate": 1.962848826815713e-05, + "loss": 0.0883, + "step": 6989 + }, + { + "epoch": 0.92, + "learning_rate": 1.9628345045111358e-05, + "loss": 0.0739, + "step": 6990 + }, + { + "epoch": 0.92, + "learning_rate": 1.962820179498638e-05, + "loss": 0.0897, + "step": 6991 + }, + { + "epoch": 0.92, + "learning_rate": 1.9628058517782595e-05, + "loss": 0.0848, + "step": 6992 + }, + { + "epoch": 0.92, + "learning_rate": 1.9627915213500407e-05, + "loss": 0.0885, + "step": 6993 + }, + { + "epoch": 0.92, + "learning_rate": 1.9627771882140223e-05, + "loss": 0.0826, + "step": 6994 + }, + { + "epoch": 0.92, + "learning_rate": 1.9627628523702445e-05, + "loss": 0.0861, + "step": 6995 + }, + { + "epoch": 0.92, + "learning_rate": 1.9627485138187473e-05, + "loss": 0.0878, + "step": 6996 + }, + { + "epoch": 0.92, + "learning_rate": 1.9627341725595707e-05, + "loss": 0.0887, + "step": 6997 + }, + { + "epoch": 0.92, + "learning_rate": 1.962719828592756e-05, + "loss": 0.097, + "step": 6998 + }, + { + "epoch": 0.92, + "learning_rate": 1.962705481918343e-05, + "loss": 0.0906, + "step": 6999 + }, + { + "epoch": 0.92, + "learning_rate": 1.962691132536372e-05, + "loss": 0.086, + "step": 7000 + }, + { + "epoch": 0.92, + "learning_rate": 1.9626767804468833e-05, + "loss": 0.0864, + "step": 7001 + }, + { + "epoch": 0.92, + "learning_rate": 1.9626624256499176e-05, + "loss": 0.0782, + "step": 7002 + }, + { + "epoch": 0.92, + "learning_rate": 1.9626480681455155e-05, + "loss": 0.0923, + "step": 7003 + }, + { + "epoch": 0.92, + "learning_rate": 1.9626337079337165e-05, + "loss": 0.0855, + "step": 7004 + }, + { + "epoch": 0.92, + "learning_rate": 1.9626193450145613e-05, + "loss": 0.0931, + "step": 7005 + }, + { + "epoch": 0.92, + "learning_rate": 1.9626049793880907e-05, + "loss": 0.0854, + "step": 7006 + }, + { + "epoch": 0.92, + "learning_rate": 1.962590611054345e-05, + "loss": 0.1111, + "step": 7007 + }, + { + "epoch": 0.92, + "learning_rate": 1.962576240013364e-05, + "loss": 0.0931, + "step": 7008 + }, + { + "epoch": 0.92, + "learning_rate": 1.962561866265189e-05, + "loss": 0.0936, + "step": 7009 + }, + { + "epoch": 0.92, + "learning_rate": 1.96254748980986e-05, + "loss": 0.0798, + "step": 7010 + }, + { + "epoch": 0.92, + "learning_rate": 1.9625331106474174e-05, + "loss": 0.0816, + "step": 7011 + }, + { + "epoch": 0.92, + "learning_rate": 1.9625187287779014e-05, + "loss": 0.0777, + "step": 7012 + }, + { + "epoch": 0.92, + "learning_rate": 1.962504344201353e-05, + "loss": 0.0783, + "step": 7013 + }, + { + "epoch": 0.92, + "learning_rate": 1.9624899569178124e-05, + "loss": 0.0865, + "step": 7014 + }, + { + "epoch": 0.92, + "learning_rate": 1.96247556692732e-05, + "loss": 0.0874, + "step": 7015 + }, + { + "epoch": 0.92, + "learning_rate": 1.962461174229916e-05, + "loss": 0.0862, + "step": 7016 + }, + { + "epoch": 0.92, + "learning_rate": 1.9624467788256418e-05, + "loss": 0.0892, + "step": 7017 + }, + { + "epoch": 0.92, + "learning_rate": 1.962432380714537e-05, + "loss": 0.0937, + "step": 7018 + }, + { + "epoch": 0.92, + "learning_rate": 1.9624179798966426e-05, + "loss": 0.0729, + "step": 7019 + }, + { + "epoch": 0.92, + "learning_rate": 1.9624035763719986e-05, + "loss": 0.0831, + "step": 7020 + }, + { + "epoch": 0.92, + "learning_rate": 1.9623891701406458e-05, + "loss": 0.0905, + "step": 7021 + }, + { + "epoch": 0.92, + "learning_rate": 1.9623747612026246e-05, + "loss": 0.0775, + "step": 7022 + }, + { + "epoch": 0.92, + "learning_rate": 1.9623603495579758e-05, + "loss": 0.0925, + "step": 7023 + }, + { + "epoch": 0.92, + "learning_rate": 1.96234593520674e-05, + "loss": 0.0927, + "step": 7024 + }, + { + "epoch": 0.92, + "learning_rate": 1.9623315181489575e-05, + "loss": 0.0872, + "step": 7025 + }, + { + "epoch": 0.92, + "learning_rate": 1.9623170983846685e-05, + "loss": 0.089, + "step": 7026 + }, + { + "epoch": 0.92, + "learning_rate": 1.962302675913914e-05, + "loss": 0.0929, + "step": 7027 + }, + { + "epoch": 0.92, + "learning_rate": 1.9622882507367347e-05, + "loss": 0.089, + "step": 7028 + }, + { + "epoch": 0.92, + "learning_rate": 1.9622738228531707e-05, + "loss": 0.0722, + "step": 7029 + }, + { + "epoch": 0.92, + "learning_rate": 1.962259392263263e-05, + "loss": 0.0791, + "step": 7030 + }, + { + "epoch": 0.92, + "learning_rate": 1.9622449589670515e-05, + "loss": 0.0845, + "step": 7031 + }, + { + "epoch": 0.92, + "learning_rate": 1.962230522964578e-05, + "loss": 0.0962, + "step": 7032 + }, + { + "epoch": 0.92, + "learning_rate": 1.9622160842558818e-05, + "loss": 0.0975, + "step": 7033 + }, + { + "epoch": 0.92, + "learning_rate": 1.9622016428410044e-05, + "loss": 0.087, + "step": 7034 + }, + { + "epoch": 0.92, + "learning_rate": 1.962187198719986e-05, + "loss": 0.0881, + "step": 7035 + }, + { + "epoch": 0.92, + "learning_rate": 1.9621727518928677e-05, + "loss": 0.0988, + "step": 7036 + }, + { + "epoch": 0.92, + "learning_rate": 1.9621583023596897e-05, + "loss": 0.0874, + "step": 7037 + }, + { + "epoch": 0.92, + "learning_rate": 1.962143850120492e-05, + "loss": 0.0872, + "step": 7038 + }, + { + "epoch": 0.92, + "learning_rate": 1.9621293951753168e-05, + "loss": 0.0872, + "step": 7039 + }, + { + "epoch": 0.92, + "learning_rate": 1.9621149375242035e-05, + "loss": 0.0806, + "step": 7040 + }, + { + "epoch": 0.92, + "learning_rate": 1.9621004771671933e-05, + "loss": 0.0824, + "step": 7041 + }, + { + "epoch": 0.92, + "learning_rate": 1.9620860141043265e-05, + "loss": 0.0841, + "step": 7042 + }, + { + "epoch": 0.92, + "learning_rate": 1.9620715483356445e-05, + "loss": 0.078, + "step": 7043 + }, + { + "epoch": 0.92, + "learning_rate": 1.962057079861187e-05, + "loss": 0.0837, + "step": 7044 + }, + { + "epoch": 0.92, + "learning_rate": 1.9620426086809954e-05, + "loss": 0.0854, + "step": 7045 + }, + { + "epoch": 0.92, + "learning_rate": 1.9620281347951103e-05, + "loss": 0.0899, + "step": 7046 + }, + { + "epoch": 0.92, + "learning_rate": 1.9620136582035722e-05, + "loss": 0.0888, + "step": 7047 + }, + { + "epoch": 0.92, + "learning_rate": 1.961999178906422e-05, + "loss": 0.0799, + "step": 7048 + }, + { + "epoch": 0.92, + "learning_rate": 1.9619846969037005e-05, + "loss": 0.0879, + "step": 7049 + }, + { + "epoch": 0.92, + "learning_rate": 1.961970212195448e-05, + "loss": 0.0987, + "step": 7050 + }, + { + "epoch": 0.92, + "learning_rate": 1.9619557247817053e-05, + "loss": 0.0867, + "step": 7051 + }, + { + "epoch": 0.92, + "learning_rate": 1.9619412346625137e-05, + "loss": 0.106, + "step": 7052 + }, + { + "epoch": 0.92, + "learning_rate": 1.9619267418379134e-05, + "loss": 0.0951, + "step": 7053 + }, + { + "epoch": 0.92, + "learning_rate": 1.9619122463079455e-05, + "loss": 0.0765, + "step": 7054 + }, + { + "epoch": 0.92, + "learning_rate": 1.9618977480726505e-05, + "loss": 0.0971, + "step": 7055 + }, + { + "epoch": 0.92, + "learning_rate": 1.9618832471320696e-05, + "loss": 0.0858, + "step": 7056 + }, + { + "epoch": 0.92, + "learning_rate": 1.9618687434862433e-05, + "loss": 0.0919, + "step": 7057 + }, + { + "epoch": 0.92, + "learning_rate": 1.961854237135212e-05, + "loss": 0.0955, + "step": 7058 + }, + { + "epoch": 0.92, + "learning_rate": 1.9618397280790172e-05, + "loss": 0.0869, + "step": 7059 + }, + { + "epoch": 0.92, + "learning_rate": 1.961825216317699e-05, + "loss": 0.0817, + "step": 7060 + }, + { + "epoch": 0.92, + "learning_rate": 1.961810701851299e-05, + "loss": 0.0951, + "step": 7061 + }, + { + "epoch": 0.93, + "learning_rate": 1.9617961846798574e-05, + "loss": 0.1001, + "step": 7062 + }, + { + "epoch": 0.93, + "learning_rate": 1.9617816648034155e-05, + "loss": 0.0819, + "step": 7063 + }, + { + "epoch": 0.93, + "learning_rate": 1.9617671422220135e-05, + "loss": 0.0815, + "step": 7064 + }, + { + "epoch": 0.93, + "learning_rate": 1.9617526169356927e-05, + "loss": 0.0819, + "step": 7065 + }, + { + "epoch": 0.93, + "learning_rate": 1.9617380889444942e-05, + "loss": 0.0785, + "step": 7066 + }, + { + "epoch": 0.93, + "learning_rate": 1.9617235582484583e-05, + "loss": 0.0894, + "step": 7067 + }, + { + "epoch": 0.93, + "learning_rate": 1.9617090248476263e-05, + "loss": 0.0909, + "step": 7068 + }, + { + "epoch": 0.93, + "learning_rate": 1.9616944887420385e-05, + "loss": 0.0748, + "step": 7069 + }, + { + "epoch": 0.93, + "learning_rate": 1.9616799499317366e-05, + "loss": 0.0814, + "step": 7070 + }, + { + "epoch": 0.93, + "learning_rate": 1.9616654084167606e-05, + "loss": 0.0869, + "step": 7071 + }, + { + "epoch": 0.93, + "learning_rate": 1.9616508641971522e-05, + "loss": 0.0818, + "step": 7072 + }, + { + "epoch": 0.93, + "learning_rate": 1.961636317272952e-05, + "loss": 0.0793, + "step": 7073 + }, + { + "epoch": 0.93, + "learning_rate": 1.961621767644201e-05, + "loss": 0.0921, + "step": 7074 + }, + { + "epoch": 0.93, + "learning_rate": 1.9616072153109395e-05, + "loss": 0.0916, + "step": 7075 + }, + { + "epoch": 0.93, + "learning_rate": 1.961592660273209e-05, + "loss": 0.101, + "step": 7076 + }, + { + "epoch": 0.93, + "learning_rate": 1.9615781025310504e-05, + "loss": 0.075, + "step": 7077 + }, + { + "epoch": 0.93, + "learning_rate": 1.961563542084505e-05, + "loss": 0.0884, + "step": 7078 + }, + { + "epoch": 0.93, + "learning_rate": 1.9615489789336128e-05, + "loss": 0.0857, + "step": 7079 + }, + { + "epoch": 0.93, + "learning_rate": 1.9615344130784156e-05, + "loss": 0.0953, + "step": 7080 + }, + { + "epoch": 0.93, + "learning_rate": 1.961519844518954e-05, + "loss": 0.0767, + "step": 7081 + }, + { + "epoch": 0.93, + "learning_rate": 1.961505273255269e-05, + "loss": 0.094, + "step": 7082 + }, + { + "epoch": 0.93, + "learning_rate": 1.9614906992874017e-05, + "loss": 0.0792, + "step": 7083 + }, + { + "epoch": 0.93, + "learning_rate": 1.9614761226153928e-05, + "loss": 0.0842, + "step": 7084 + }, + { + "epoch": 0.93, + "learning_rate": 1.9614615432392838e-05, + "loss": 0.0911, + "step": 7085 + }, + { + "epoch": 0.93, + "learning_rate": 1.9614469611591153e-05, + "loss": 0.0748, + "step": 7086 + }, + { + "epoch": 0.93, + "learning_rate": 1.9614323763749286e-05, + "loss": 0.0976, + "step": 7087 + }, + { + "epoch": 0.93, + "learning_rate": 1.9614177888867643e-05, + "loss": 0.0747, + "step": 7088 + }, + { + "epoch": 0.93, + "learning_rate": 1.961403198694664e-05, + "loss": 0.0866, + "step": 7089 + }, + { + "epoch": 0.93, + "learning_rate": 1.9613886057986677e-05, + "loss": 0.0845, + "step": 7090 + }, + { + "epoch": 0.93, + "learning_rate": 1.9613740101988175e-05, + "loss": 0.0862, + "step": 7091 + }, + { + "epoch": 0.93, + "learning_rate": 1.9613594118951542e-05, + "loss": 0.0997, + "step": 7092 + }, + { + "epoch": 0.93, + "learning_rate": 1.961344810887719e-05, + "loss": 0.0885, + "step": 7093 + }, + { + "epoch": 0.93, + "learning_rate": 1.9613302071765523e-05, + "loss": 0.0985, + "step": 7094 + }, + { + "epoch": 0.93, + "learning_rate": 1.9613156007616957e-05, + "loss": 0.0927, + "step": 7095 + }, + { + "epoch": 0.93, + "learning_rate": 1.96130099164319e-05, + "loss": 0.0826, + "step": 7096 + }, + { + "epoch": 0.93, + "learning_rate": 1.961286379821077e-05, + "loss": 0.0985, + "step": 7097 + }, + { + "epoch": 0.93, + "learning_rate": 1.9612717652953964e-05, + "loss": 0.087, + "step": 7098 + }, + { + "epoch": 0.93, + "learning_rate": 1.9612571480661907e-05, + "loss": 0.0917, + "step": 7099 + }, + { + "epoch": 0.93, + "learning_rate": 1.9612425281335004e-05, + "loss": 0.089, + "step": 7100 + }, + { + "epoch": 0.93, + "learning_rate": 1.9612279054973666e-05, + "loss": 0.0932, + "step": 7101 + }, + { + "epoch": 0.93, + "learning_rate": 1.961213280157831e-05, + "loss": 0.0857, + "step": 7102 + }, + { + "epoch": 0.93, + "learning_rate": 1.9611986521149336e-05, + "loss": 0.0934, + "step": 7103 + }, + { + "epoch": 0.93, + "learning_rate": 1.9611840213687165e-05, + "loss": 0.0968, + "step": 7104 + }, + { + "epoch": 0.93, + "learning_rate": 1.96116938791922e-05, + "loss": 0.0804, + "step": 7105 + }, + { + "epoch": 0.93, + "learning_rate": 1.9611547517664865e-05, + "loss": 0.085, + "step": 7106 + }, + { + "epoch": 0.93, + "learning_rate": 1.961140112910556e-05, + "loss": 0.0951, + "step": 7107 + }, + { + "epoch": 0.93, + "learning_rate": 1.9611254713514704e-05, + "loss": 0.0922, + "step": 7108 + }, + { + "epoch": 0.93, + "learning_rate": 1.96111082708927e-05, + "loss": 0.0966, + "step": 7109 + }, + { + "epoch": 0.93, + "learning_rate": 1.9610961801239974e-05, + "loss": 0.0837, + "step": 7110 + }, + { + "epoch": 0.93, + "learning_rate": 1.9610815304556923e-05, + "loss": 0.0854, + "step": 7111 + }, + { + "epoch": 0.93, + "learning_rate": 1.961066878084397e-05, + "loss": 0.0917, + "step": 7112 + }, + { + "epoch": 0.93, + "learning_rate": 1.9610522230101523e-05, + "loss": 0.0804, + "step": 7113 + }, + { + "epoch": 0.93, + "learning_rate": 1.9610375652329993e-05, + "loss": 0.0812, + "step": 7114 + }, + { + "epoch": 0.93, + "learning_rate": 1.961022904752979e-05, + "loss": 0.0929, + "step": 7115 + }, + { + "epoch": 0.93, + "learning_rate": 1.9610082415701336e-05, + "loss": 0.0752, + "step": 7116 + }, + { + "epoch": 0.93, + "learning_rate": 1.9609935756845032e-05, + "loss": 0.0878, + "step": 7117 + }, + { + "epoch": 0.93, + "learning_rate": 1.96097890709613e-05, + "loss": 0.0798, + "step": 7118 + }, + { + "epoch": 0.93, + "learning_rate": 1.960964235805054e-05, + "loss": 0.0838, + "step": 7119 + }, + { + "epoch": 0.93, + "learning_rate": 1.9609495618113178e-05, + "loss": 0.0866, + "step": 7120 + }, + { + "epoch": 0.93, + "learning_rate": 1.9609348851149622e-05, + "loss": 0.0932, + "step": 7121 + }, + { + "epoch": 0.93, + "learning_rate": 1.9609202057160284e-05, + "loss": 0.0818, + "step": 7122 + }, + { + "epoch": 0.93, + "learning_rate": 1.9609055236145574e-05, + "loss": 0.0841, + "step": 7123 + }, + { + "epoch": 0.93, + "learning_rate": 1.960890838810591e-05, + "loss": 0.095, + "step": 7124 + }, + { + "epoch": 0.93, + "learning_rate": 1.9608761513041704e-05, + "loss": 0.0978, + "step": 7125 + }, + { + "epoch": 0.93, + "learning_rate": 1.9608614610953368e-05, + "loss": 0.0856, + "step": 7126 + }, + { + "epoch": 0.93, + "learning_rate": 1.9608467681841312e-05, + "loss": 0.0814, + "step": 7127 + }, + { + "epoch": 0.93, + "learning_rate": 1.9608320725705955e-05, + "loss": 0.0949, + "step": 7128 + }, + { + "epoch": 0.93, + "learning_rate": 1.9608173742547705e-05, + "loss": 0.0783, + "step": 7129 + }, + { + "epoch": 0.93, + "learning_rate": 1.9608026732366977e-05, + "loss": 0.08, + "step": 7130 + }, + { + "epoch": 0.93, + "learning_rate": 1.960787969516419e-05, + "loss": 0.0864, + "step": 7131 + }, + { + "epoch": 0.93, + "learning_rate": 1.9607732630939755e-05, + "loss": 0.0837, + "step": 7132 + }, + { + "epoch": 0.93, + "learning_rate": 1.9607585539694077e-05, + "loss": 0.0791, + "step": 7133 + }, + { + "epoch": 0.93, + "learning_rate": 1.9607438421427577e-05, + "loss": 0.0857, + "step": 7134 + }, + { + "epoch": 0.93, + "learning_rate": 1.9607291276140672e-05, + "loss": 0.0871, + "step": 7135 + }, + { + "epoch": 0.93, + "learning_rate": 1.9607144103833772e-05, + "loss": 0.0818, + "step": 7136 + }, + { + "epoch": 0.93, + "learning_rate": 1.9606996904507287e-05, + "loss": 0.0874, + "step": 7137 + }, + { + "epoch": 0.94, + "learning_rate": 1.9606849678161636e-05, + "loss": 0.0799, + "step": 7138 + }, + { + "epoch": 0.94, + "learning_rate": 1.9606702424797233e-05, + "loss": 0.0948, + "step": 7139 + }, + { + "epoch": 0.94, + "learning_rate": 1.960655514441449e-05, + "loss": 0.0946, + "step": 7140 + }, + { + "epoch": 0.94, + "learning_rate": 1.9606407837013823e-05, + "loss": 0.0856, + "step": 7141 + }, + { + "epoch": 0.94, + "learning_rate": 1.9606260502595645e-05, + "loss": 0.0855, + "step": 7142 + }, + { + "epoch": 0.94, + "learning_rate": 1.960611314116037e-05, + "loss": 0.0999, + "step": 7143 + }, + { + "epoch": 0.94, + "learning_rate": 1.9605965752708415e-05, + "loss": 0.0844, + "step": 7144 + }, + { + "epoch": 0.94, + "learning_rate": 1.9605818337240192e-05, + "loss": 0.0781, + "step": 7145 + }, + { + "epoch": 0.94, + "learning_rate": 1.9605670894756117e-05, + "loss": 0.0913, + "step": 7146 + }, + { + "epoch": 0.94, + "learning_rate": 1.96055234252566e-05, + "loss": 0.0894, + "step": 7147 + }, + { + "epoch": 0.94, + "learning_rate": 1.9605375928742067e-05, + "loss": 0.0852, + "step": 7148 + }, + { + "epoch": 0.94, + "learning_rate": 1.9605228405212925e-05, + "loss": 0.0724, + "step": 7149 + }, + { + "epoch": 0.94, + "learning_rate": 1.9605080854669585e-05, + "loss": 0.0809, + "step": 7150 + }, + { + "epoch": 0.94, + "learning_rate": 1.960493327711247e-05, + "loss": 0.0869, + "step": 7151 + }, + { + "epoch": 0.94, + "learning_rate": 1.960478567254199e-05, + "loss": 0.0886, + "step": 7152 + }, + { + "epoch": 0.94, + "learning_rate": 1.9604638040958562e-05, + "loss": 0.0876, + "step": 7153 + }, + { + "epoch": 0.94, + "learning_rate": 1.9604490382362603e-05, + "loss": 0.0778, + "step": 7154 + }, + { + "epoch": 0.94, + "learning_rate": 1.9604342696754523e-05, + "loss": 0.0859, + "step": 7155 + }, + { + "epoch": 0.94, + "learning_rate": 1.960419498413474e-05, + "loss": 0.0877, + "step": 7156 + }, + { + "epoch": 0.94, + "learning_rate": 1.9604047244503676e-05, + "loss": 0.079, + "step": 7157 + }, + { + "epoch": 0.94, + "learning_rate": 1.9603899477861733e-05, + "loss": 0.0873, + "step": 7158 + }, + { + "epoch": 0.94, + "learning_rate": 1.960375168420934e-05, + "loss": 0.0798, + "step": 7159 + }, + { + "epoch": 0.94, + "learning_rate": 1.9603603863546903e-05, + "loss": 0.0911, + "step": 7160 + }, + { + "epoch": 0.94, + "learning_rate": 1.9603456015874843e-05, + "loss": 0.0972, + "step": 7161 + }, + { + "epoch": 0.94, + "learning_rate": 1.9603308141193576e-05, + "loss": 0.0788, + "step": 7162 + }, + { + "epoch": 0.94, + "learning_rate": 1.9603160239503512e-05, + "loss": 0.1002, + "step": 7163 + }, + { + "epoch": 0.94, + "learning_rate": 1.9603012310805077e-05, + "loss": 0.0835, + "step": 7164 + }, + { + "epoch": 0.94, + "learning_rate": 1.9602864355098678e-05, + "loss": 0.0847, + "step": 7165 + }, + { + "epoch": 0.94, + "learning_rate": 1.9602716372384735e-05, + "loss": 0.092, + "step": 7166 + }, + { + "epoch": 0.94, + "learning_rate": 1.9602568362663662e-05, + "loss": 0.0942, + "step": 7167 + }, + { + "epoch": 0.94, + "learning_rate": 1.960242032593588e-05, + "loss": 0.0812, + "step": 7168 + }, + { + "epoch": 0.94, + "learning_rate": 1.96022722622018e-05, + "loss": 0.1084, + "step": 7169 + }, + { + "epoch": 0.94, + "learning_rate": 1.9602124171461843e-05, + "loss": 0.0815, + "step": 7170 + }, + { + "epoch": 0.94, + "learning_rate": 1.9601976053716418e-05, + "loss": 0.0881, + "step": 7171 + }, + { + "epoch": 0.94, + "learning_rate": 1.960182790896595e-05, + "loss": 0.0738, + "step": 7172 + }, + { + "epoch": 0.94, + "learning_rate": 1.9601679737210855e-05, + "loss": 0.0858, + "step": 7173 + }, + { + "epoch": 0.94, + "learning_rate": 1.960153153845154e-05, + "loss": 0.0848, + "step": 7174 + }, + { + "epoch": 0.94, + "learning_rate": 1.9601383312688435e-05, + "loss": 0.0867, + "step": 7175 + }, + { + "epoch": 0.94, + "learning_rate": 1.9601235059921947e-05, + "loss": 0.0988, + "step": 7176 + }, + { + "epoch": 0.94, + "learning_rate": 1.96010867801525e-05, + "loss": 0.0824, + "step": 7177 + }, + { + "epoch": 0.94, + "learning_rate": 1.9600938473380505e-05, + "loss": 0.0742, + "step": 7178 + }, + { + "epoch": 0.94, + "learning_rate": 1.9600790139606384e-05, + "loss": 0.0638, + "step": 7179 + }, + { + "epoch": 0.94, + "learning_rate": 1.960064177883055e-05, + "loss": 0.0936, + "step": 7180 + }, + { + "epoch": 0.94, + "learning_rate": 1.9600493391053425e-05, + "loss": 0.0809, + "step": 7181 + }, + { + "epoch": 0.94, + "learning_rate": 1.9600344976275423e-05, + "loss": 0.0793, + "step": 7182 + }, + { + "epoch": 0.94, + "learning_rate": 1.9600196534496958e-05, + "loss": 0.0907, + "step": 7183 + }, + { + "epoch": 0.94, + "learning_rate": 1.9600048065718454e-05, + "loss": 0.0843, + "step": 7184 + }, + { + "epoch": 0.94, + "learning_rate": 1.9599899569940327e-05, + "loss": 0.0837, + "step": 7185 + }, + { + "epoch": 0.94, + "learning_rate": 1.9599751047162995e-05, + "loss": 0.0857, + "step": 7186 + }, + { + "epoch": 0.94, + "learning_rate": 1.959960249738687e-05, + "loss": 0.0862, + "step": 7187 + }, + { + "epoch": 0.94, + "learning_rate": 1.9599453920612378e-05, + "loss": 0.0804, + "step": 7188 + }, + { + "epoch": 0.94, + "learning_rate": 1.9599305316839932e-05, + "loss": 0.0985, + "step": 7189 + }, + { + "epoch": 0.94, + "learning_rate": 1.959915668606995e-05, + "loss": 0.0902, + "step": 7190 + }, + { + "epoch": 0.94, + "learning_rate": 1.9599008028302852e-05, + "loss": 0.085, + "step": 7191 + }, + { + "epoch": 0.94, + "learning_rate": 1.9598859343539054e-05, + "loss": 0.0814, + "step": 7192 + }, + { + "epoch": 0.94, + "learning_rate": 1.9598710631778978e-05, + "loss": 0.0874, + "step": 7193 + }, + { + "epoch": 0.94, + "learning_rate": 1.9598561893023037e-05, + "loss": 0.0858, + "step": 7194 + }, + { + "epoch": 0.94, + "learning_rate": 1.959841312727165e-05, + "loss": 0.0855, + "step": 7195 + }, + { + "epoch": 0.94, + "learning_rate": 1.9598264334525244e-05, + "loss": 0.0744, + "step": 7196 + }, + { + "epoch": 0.94, + "learning_rate": 1.9598115514784225e-05, + "loss": 0.0992, + "step": 7197 + }, + { + "epoch": 0.94, + "learning_rate": 1.959796666804902e-05, + "loss": 0.0855, + "step": 7198 + }, + { + "epoch": 0.94, + "learning_rate": 1.959781779432004e-05, + "loss": 0.0897, + "step": 7199 + }, + { + "epoch": 0.94, + "learning_rate": 1.9597668893597716e-05, + "loss": 0.0835, + "step": 7200 + }, + { + "epoch": 0.94, + "learning_rate": 1.9597519965882453e-05, + "loss": 0.0798, + "step": 7201 + }, + { + "epoch": 0.94, + "learning_rate": 1.959737101117468e-05, + "loss": 0.084, + "step": 7202 + }, + { + "epoch": 0.94, + "learning_rate": 1.9597222029474808e-05, + "loss": 0.0837, + "step": 7203 + }, + { + "epoch": 0.94, + "learning_rate": 1.9597073020783263e-05, + "loss": 0.0905, + "step": 7204 + }, + { + "epoch": 0.94, + "learning_rate": 1.959692398510046e-05, + "loss": 0.0922, + "step": 7205 + }, + { + "epoch": 0.94, + "learning_rate": 1.9596774922426822e-05, + "loss": 0.0897, + "step": 7206 + }, + { + "epoch": 0.94, + "learning_rate": 1.9596625832762766e-05, + "loss": 0.0834, + "step": 7207 + }, + { + "epoch": 0.94, + "learning_rate": 1.9596476716108707e-05, + "loss": 0.0881, + "step": 7208 + }, + { + "epoch": 0.94, + "learning_rate": 1.959632757246507e-05, + "loss": 0.0753, + "step": 7209 + }, + { + "epoch": 0.94, + "learning_rate": 1.9596178401832275e-05, + "loss": 0.0844, + "step": 7210 + }, + { + "epoch": 0.94, + "learning_rate": 1.9596029204210736e-05, + "loss": 0.0813, + "step": 7211 + }, + { + "epoch": 0.94, + "learning_rate": 1.9595879979600877e-05, + "loss": 0.0902, + "step": 7212 + }, + { + "epoch": 0.94, + "learning_rate": 1.9595730728003115e-05, + "loss": 0.0961, + "step": 7213 + }, + { + "epoch": 0.94, + "learning_rate": 1.9595581449417874e-05, + "loss": 0.0857, + "step": 7214 + }, + { + "epoch": 0.95, + "learning_rate": 1.959543214384557e-05, + "loss": 0.0914, + "step": 7215 + }, + { + "epoch": 0.95, + "learning_rate": 1.9595282811286625e-05, + "loss": 0.075, + "step": 7216 + }, + { + "epoch": 0.95, + "learning_rate": 1.9595133451741456e-05, + "loss": 0.0762, + "step": 7217 + }, + { + "epoch": 0.95, + "learning_rate": 1.9594984065210486e-05, + "loss": 0.0929, + "step": 7218 + }, + { + "epoch": 0.95, + "learning_rate": 1.9594834651694133e-05, + "loss": 0.0738, + "step": 7219 + }, + { + "epoch": 0.95, + "learning_rate": 1.959468521119282e-05, + "loss": 0.0821, + "step": 7220 + }, + { + "epoch": 0.95, + "learning_rate": 1.959453574370697e-05, + "loss": 0.0948, + "step": 7221 + }, + { + "epoch": 0.95, + "learning_rate": 1.9594386249236993e-05, + "loss": 0.082, + "step": 7222 + }, + { + "epoch": 0.95, + "learning_rate": 1.9594236727783317e-05, + "loss": 0.0918, + "step": 7223 + }, + { + "epoch": 0.95, + "learning_rate": 1.9594087179346362e-05, + "loss": 0.0924, + "step": 7224 + }, + { + "epoch": 0.95, + "learning_rate": 1.9593937603926546e-05, + "loss": 0.0777, + "step": 7225 + }, + { + "epoch": 0.95, + "learning_rate": 1.9593788001524296e-05, + "loss": 0.082, + "step": 7226 + }, + { + "epoch": 0.95, + "learning_rate": 1.9593638372140024e-05, + "loss": 0.0898, + "step": 7227 + }, + { + "epoch": 0.95, + "learning_rate": 1.9593488715774155e-05, + "loss": 0.0848, + "step": 7228 + }, + { + "epoch": 0.95, + "learning_rate": 1.9593339032427112e-05, + "loss": 0.0913, + "step": 7229 + }, + { + "epoch": 0.95, + "learning_rate": 1.9593189322099312e-05, + "loss": 0.0942, + "step": 7230 + }, + { + "epoch": 0.95, + "learning_rate": 1.9593039584791178e-05, + "loss": 0.0885, + "step": 7231 + }, + { + "epoch": 0.95, + "learning_rate": 1.9592889820503133e-05, + "loss": 0.081, + "step": 7232 + }, + { + "epoch": 0.95, + "learning_rate": 1.9592740029235595e-05, + "loss": 0.0805, + "step": 7233 + }, + { + "epoch": 0.95, + "learning_rate": 1.9592590210988987e-05, + "loss": 0.0795, + "step": 7234 + }, + { + "epoch": 0.95, + "learning_rate": 1.9592440365763726e-05, + "loss": 0.0959, + "step": 7235 + }, + { + "epoch": 0.95, + "learning_rate": 1.9592290493560245e-05, + "loss": 0.098, + "step": 7236 + }, + { + "epoch": 0.95, + "learning_rate": 1.959214059437895e-05, + "loss": 0.0747, + "step": 7237 + }, + { + "epoch": 0.95, + "learning_rate": 1.9591990668220275e-05, + "loss": 0.0902, + "step": 7238 + }, + { + "epoch": 0.95, + "learning_rate": 1.959184071508464e-05, + "loss": 0.0942, + "step": 7239 + }, + { + "epoch": 0.95, + "learning_rate": 1.9591690734972454e-05, + "loss": 0.084, + "step": 7240 + }, + { + "epoch": 0.95, + "learning_rate": 1.9591540727884156e-05, + "loss": 0.09, + "step": 7241 + }, + { + "epoch": 0.95, + "learning_rate": 1.9591390693820157e-05, + "loss": 0.0824, + "step": 7242 + }, + { + "epoch": 0.95, + "learning_rate": 1.9591240632780885e-05, + "loss": 0.0835, + "step": 7243 + }, + { + "epoch": 0.95, + "learning_rate": 1.959109054476676e-05, + "loss": 0.0846, + "step": 7244 + }, + { + "epoch": 0.95, + "learning_rate": 1.9590940429778204e-05, + "loss": 0.074, + "step": 7245 + }, + { + "epoch": 0.95, + "learning_rate": 1.9590790287815634e-05, + "loss": 0.0812, + "step": 7246 + }, + { + "epoch": 0.95, + "learning_rate": 1.9590640118879483e-05, + "loss": 0.0959, + "step": 7247 + }, + { + "epoch": 0.95, + "learning_rate": 1.9590489922970164e-05, + "loss": 0.0908, + "step": 7248 + }, + { + "epoch": 0.95, + "learning_rate": 1.95903397000881e-05, + "loss": 0.0718, + "step": 7249 + }, + { + "epoch": 0.95, + "learning_rate": 1.959018945023372e-05, + "loss": 0.0883, + "step": 7250 + }, + { + "epoch": 0.95, + "learning_rate": 1.9590039173407445e-05, + "loss": 0.0849, + "step": 7251 + }, + { + "epoch": 0.95, + "learning_rate": 1.9589888869609693e-05, + "loss": 0.0808, + "step": 7252 + }, + { + "epoch": 0.95, + "learning_rate": 1.9589738538840887e-05, + "loss": 0.0943, + "step": 7253 + }, + { + "epoch": 0.95, + "learning_rate": 1.9589588181101455e-05, + "loss": 0.0843, + "step": 7254 + }, + { + "epoch": 0.95, + "learning_rate": 1.9589437796391816e-05, + "loss": 0.0812, + "step": 7255 + }, + { + "epoch": 0.95, + "learning_rate": 1.9589287384712397e-05, + "loss": 0.0814, + "step": 7256 + }, + { + "epoch": 0.95, + "learning_rate": 1.9589136946063615e-05, + "loss": 0.0861, + "step": 7257 + }, + { + "epoch": 0.95, + "learning_rate": 1.9588986480445893e-05, + "loss": 0.0803, + "step": 7258 + }, + { + "epoch": 0.95, + "learning_rate": 1.958883598785966e-05, + "loss": 0.0854, + "step": 7259 + }, + { + "epoch": 0.95, + "learning_rate": 1.9588685468305336e-05, + "loss": 0.0821, + "step": 7260 + }, + { + "epoch": 0.95, + "learning_rate": 1.9588534921783348e-05, + "loss": 0.0958, + "step": 7261 + }, + { + "epoch": 0.95, + "learning_rate": 1.9588384348294113e-05, + "loss": 0.0828, + "step": 7262 + }, + { + "epoch": 0.95, + "learning_rate": 1.9588233747838057e-05, + "loss": 0.0747, + "step": 7263 + }, + { + "epoch": 0.95, + "learning_rate": 1.9588083120415606e-05, + "loss": 0.0896, + "step": 7264 + }, + { + "epoch": 0.95, + "learning_rate": 1.958793246602718e-05, + "loss": 0.0833, + "step": 7265 + }, + { + "epoch": 0.95, + "learning_rate": 1.9587781784673206e-05, + "loss": 0.0805, + "step": 7266 + }, + { + "epoch": 0.95, + "learning_rate": 1.9587631076354108e-05, + "loss": 0.0874, + "step": 7267 + }, + { + "epoch": 0.95, + "learning_rate": 1.9587480341070305e-05, + "loss": 0.0865, + "step": 7268 + }, + { + "epoch": 0.95, + "learning_rate": 1.9587329578822225e-05, + "loss": 0.0877, + "step": 7269 + }, + { + "epoch": 0.95, + "learning_rate": 1.958717878961029e-05, + "loss": 0.0852, + "step": 7270 + }, + { + "epoch": 0.95, + "learning_rate": 1.958702797343493e-05, + "loss": 0.0862, + "step": 7271 + }, + { + "epoch": 0.95, + "learning_rate": 1.958687713029656e-05, + "loss": 0.0796, + "step": 7272 + }, + { + "epoch": 0.95, + "learning_rate": 1.9586726260195607e-05, + "loss": 0.0772, + "step": 7273 + }, + { + "epoch": 0.95, + "learning_rate": 1.95865753631325e-05, + "loss": 0.0848, + "step": 7274 + }, + { + "epoch": 0.95, + "learning_rate": 1.958642443910766e-05, + "loss": 0.0846, + "step": 7275 + }, + { + "epoch": 0.95, + "learning_rate": 1.958627348812151e-05, + "loss": 0.0867, + "step": 7276 + }, + { + "epoch": 0.95, + "learning_rate": 1.958612251017448e-05, + "loss": 0.0867, + "step": 7277 + }, + { + "epoch": 0.95, + "learning_rate": 1.9585971505266987e-05, + "loss": 0.0769, + "step": 7278 + }, + { + "epoch": 0.95, + "learning_rate": 1.9585820473399463e-05, + "loss": 0.0868, + "step": 7279 + }, + { + "epoch": 0.95, + "learning_rate": 1.9585669414572327e-05, + "loss": 0.0816, + "step": 7280 + }, + { + "epoch": 0.95, + "learning_rate": 1.958551832878601e-05, + "loss": 0.0872, + "step": 7281 + }, + { + "epoch": 0.95, + "learning_rate": 1.9585367216040927e-05, + "loss": 0.0832, + "step": 7282 + }, + { + "epoch": 0.95, + "learning_rate": 1.9585216076337513e-05, + "loss": 0.0854, + "step": 7283 + }, + { + "epoch": 0.95, + "learning_rate": 1.958506490967619e-05, + "loss": 0.0797, + "step": 7284 + }, + { + "epoch": 0.95, + "learning_rate": 1.958491371605738e-05, + "loss": 0.0928, + "step": 7285 + }, + { + "epoch": 0.95, + "learning_rate": 1.958476249548151e-05, + "loss": 0.0808, + "step": 7286 + }, + { + "epoch": 0.95, + "learning_rate": 1.958461124794901e-05, + "loss": 0.0805, + "step": 7287 + }, + { + "epoch": 0.95, + "learning_rate": 1.9584459973460295e-05, + "loss": 0.0798, + "step": 7288 + }, + { + "epoch": 0.95, + "learning_rate": 1.95843086720158e-05, + "loss": 0.0859, + "step": 7289 + }, + { + "epoch": 0.95, + "learning_rate": 1.9584157343615948e-05, + "loss": 0.0883, + "step": 7290 + }, + { + "epoch": 0.96, + "learning_rate": 1.9584005988261164e-05, + "loss": 0.0843, + "step": 7291 + }, + { + "epoch": 0.96, + "learning_rate": 1.9583854605951875e-05, + "loss": 0.0846, + "step": 7292 + }, + { + "epoch": 0.96, + "learning_rate": 1.9583703196688503e-05, + "loss": 0.0825, + "step": 7293 + }, + { + "epoch": 0.96, + "learning_rate": 1.9583551760471475e-05, + "loss": 0.0976, + "step": 7294 + }, + { + "epoch": 0.96, + "learning_rate": 1.9583400297301222e-05, + "loss": 0.0875, + "step": 7295 + }, + { + "epoch": 0.96, + "learning_rate": 1.958324880717816e-05, + "loss": 0.0894, + "step": 7296 + }, + { + "epoch": 0.96, + "learning_rate": 1.9583097290102727e-05, + "loss": 0.095, + "step": 7297 + }, + { + "epoch": 0.96, + "learning_rate": 1.9582945746075343e-05, + "loss": 0.0822, + "step": 7298 + }, + { + "epoch": 0.96, + "learning_rate": 1.958279417509643e-05, + "loss": 0.0885, + "step": 7299 + }, + { + "epoch": 0.96, + "learning_rate": 1.958264257716642e-05, + "loss": 0.0963, + "step": 7300 + }, + { + "epoch": 0.96, + "learning_rate": 1.958249095228574e-05, + "loss": 0.089, + "step": 7301 + }, + { + "epoch": 0.96, + "learning_rate": 1.958233930045481e-05, + "loss": 0.0875, + "step": 7302 + }, + { + "epoch": 0.96, + "learning_rate": 1.9582187621674066e-05, + "loss": 0.0896, + "step": 7303 + }, + { + "epoch": 0.96, + "learning_rate": 1.9582035915943928e-05, + "loss": 0.0819, + "step": 7304 + }, + { + "epoch": 0.96, + "learning_rate": 1.9581884183264827e-05, + "loss": 0.0811, + "step": 7305 + }, + { + "epoch": 0.96, + "learning_rate": 1.958173242363718e-05, + "loss": 0.1, + "step": 7306 + }, + { + "epoch": 0.96, + "learning_rate": 1.958158063706143e-05, + "loss": 0.0917, + "step": 7307 + }, + { + "epoch": 0.96, + "learning_rate": 1.9581428823537987e-05, + "loss": 0.0825, + "step": 7308 + }, + { + "epoch": 0.96, + "learning_rate": 1.958127698306729e-05, + "loss": 0.0847, + "step": 7309 + }, + { + "epoch": 0.96, + "learning_rate": 1.9581125115649758e-05, + "loss": 0.0909, + "step": 7310 + }, + { + "epoch": 0.96, + "learning_rate": 1.9580973221285827e-05, + "loss": 0.0894, + "step": 7311 + }, + { + "epoch": 0.96, + "learning_rate": 1.9580821299975913e-05, + "loss": 0.0886, + "step": 7312 + }, + { + "epoch": 0.96, + "learning_rate": 1.9580669351720452e-05, + "loss": 0.085, + "step": 7313 + }, + { + "epoch": 0.96, + "learning_rate": 1.958051737651987e-05, + "loss": 0.0805, + "step": 7314 + }, + { + "epoch": 0.96, + "learning_rate": 1.9580365374374594e-05, + "loss": 0.0922, + "step": 7315 + }, + { + "epoch": 0.96, + "learning_rate": 1.9580213345285048e-05, + "loss": 0.0851, + "step": 7316 + }, + { + "epoch": 0.96, + "learning_rate": 1.958006128925166e-05, + "loss": 0.0984, + "step": 7317 + }, + { + "epoch": 0.96, + "learning_rate": 1.9579909206274864e-05, + "loss": 0.0912, + "step": 7318 + }, + { + "epoch": 0.96, + "learning_rate": 1.9579757096355084e-05, + "loss": 0.0893, + "step": 7319 + }, + { + "epoch": 0.96, + "learning_rate": 1.9579604959492745e-05, + "loss": 0.0939, + "step": 7320 + }, + { + "epoch": 0.96, + "learning_rate": 1.9579452795688276e-05, + "loss": 0.0912, + "step": 7321 + }, + { + "epoch": 0.96, + "learning_rate": 1.957930060494211e-05, + "loss": 0.0751, + "step": 7322 + }, + { + "epoch": 0.96, + "learning_rate": 1.9579148387254666e-05, + "loss": 0.096, + "step": 7323 + }, + { + "epoch": 0.96, + "learning_rate": 1.957899614262638e-05, + "loss": 0.0859, + "step": 7324 + }, + { + "epoch": 0.96, + "learning_rate": 1.957884387105768e-05, + "loss": 0.0797, + "step": 7325 + }, + { + "epoch": 0.96, + "learning_rate": 1.9578691572548988e-05, + "loss": 0.0894, + "step": 7326 + }, + { + "epoch": 0.96, + "learning_rate": 1.957853924710074e-05, + "loss": 0.0913, + "step": 7327 + }, + { + "epoch": 0.96, + "learning_rate": 1.9578386894713354e-05, + "loss": 0.0943, + "step": 7328 + }, + { + "epoch": 0.96, + "learning_rate": 1.957823451538727e-05, + "loss": 0.0794, + "step": 7329 + }, + { + "epoch": 0.96, + "learning_rate": 1.957808210912291e-05, + "loss": 0.0863, + "step": 7330 + }, + { + "epoch": 0.96, + "learning_rate": 1.9577929675920705e-05, + "loss": 0.0793, + "step": 7331 + }, + { + "epoch": 0.96, + "learning_rate": 1.9577777215781082e-05, + "loss": 0.0863, + "step": 7332 + }, + { + "epoch": 0.96, + "learning_rate": 1.9577624728704473e-05, + "loss": 0.0957, + "step": 7333 + }, + { + "epoch": 0.96, + "learning_rate": 1.9577472214691302e-05, + "loss": 0.0936, + "step": 7334 + }, + { + "epoch": 0.96, + "learning_rate": 1.9577319673742e-05, + "loss": 0.0814, + "step": 7335 + }, + { + "epoch": 0.96, + "learning_rate": 1.9577167105856997e-05, + "loss": 0.0782, + "step": 7336 + }, + { + "epoch": 0.96, + "learning_rate": 1.9577014511036723e-05, + "loss": 0.0843, + "step": 7337 + }, + { + "epoch": 0.96, + "learning_rate": 1.9576861889281604e-05, + "loss": 0.0951, + "step": 7338 + }, + { + "epoch": 0.96, + "learning_rate": 1.957670924059207e-05, + "loss": 0.0901, + "step": 7339 + }, + { + "epoch": 0.96, + "learning_rate": 1.9576556564968555e-05, + "loss": 0.0911, + "step": 7340 + }, + { + "epoch": 0.96, + "learning_rate": 1.957640386241148e-05, + "loss": 0.0937, + "step": 7341 + }, + { + "epoch": 0.96, + "learning_rate": 1.9576251132921286e-05, + "loss": 0.0879, + "step": 7342 + }, + { + "epoch": 0.96, + "learning_rate": 1.9576098376498388e-05, + "loss": 0.0806, + "step": 7343 + }, + { + "epoch": 0.96, + "learning_rate": 1.9575945593143227e-05, + "loss": 0.0931, + "step": 7344 + }, + { + "epoch": 0.96, + "learning_rate": 1.9575792782856227e-05, + "loss": 0.081, + "step": 7345 + }, + { + "epoch": 0.96, + "learning_rate": 1.957563994563782e-05, + "loss": 0.0742, + "step": 7346 + }, + { + "epoch": 0.96, + "learning_rate": 1.957548708148844e-05, + "loss": 0.0904, + "step": 7347 + }, + { + "epoch": 0.96, + "learning_rate": 1.957533419040851e-05, + "loss": 0.0771, + "step": 7348 + }, + { + "epoch": 0.96, + "learning_rate": 1.9575181272398457e-05, + "loss": 0.0814, + "step": 7349 + }, + { + "epoch": 0.96, + "learning_rate": 1.9575028327458724e-05, + "loss": 0.0885, + "step": 7350 + }, + { + "epoch": 0.96, + "learning_rate": 1.957487535558973e-05, + "loss": 0.0905, + "step": 7351 + }, + { + "epoch": 0.96, + "learning_rate": 1.957472235679191e-05, + "loss": 0.0908, + "step": 7352 + }, + { + "epoch": 0.96, + "learning_rate": 1.9574569331065693e-05, + "loss": 0.0935, + "step": 7353 + }, + { + "epoch": 0.96, + "learning_rate": 1.957441627841151e-05, + "loss": 0.0881, + "step": 7354 + }, + { + "epoch": 0.96, + "learning_rate": 1.9574263198829792e-05, + "loss": 0.0838, + "step": 7355 + }, + { + "epoch": 0.96, + "learning_rate": 1.9574110092320967e-05, + "loss": 0.0896, + "step": 7356 + }, + { + "epoch": 0.96, + "learning_rate": 1.957395695888547e-05, + "loss": 0.0817, + "step": 7357 + }, + { + "epoch": 0.96, + "learning_rate": 1.9573803798523727e-05, + "loss": 0.0884, + "step": 7358 + }, + { + "epoch": 0.96, + "learning_rate": 1.9573650611236168e-05, + "loss": 0.088, + "step": 7359 + }, + { + "epoch": 0.96, + "learning_rate": 1.9573497397023228e-05, + "loss": 0.0879, + "step": 7360 + }, + { + "epoch": 0.96, + "learning_rate": 1.957334415588534e-05, + "loss": 0.0771, + "step": 7361 + }, + { + "epoch": 0.96, + "learning_rate": 1.9573190887822927e-05, + "loss": 0.0854, + "step": 7362 + }, + { + "epoch": 0.96, + "learning_rate": 1.9573037592836427e-05, + "loss": 0.0773, + "step": 7363 + }, + { + "epoch": 0.96, + "learning_rate": 1.9572884270926263e-05, + "loss": 0.0823, + "step": 7364 + }, + { + "epoch": 0.96, + "learning_rate": 1.957273092209288e-05, + "loss": 0.098, + "step": 7365 + }, + { + "epoch": 0.96, + "learning_rate": 1.9572577546336698e-05, + "loss": 0.0782, + "step": 7366 + }, + { + "epoch": 0.97, + "learning_rate": 1.957242414365815e-05, + "loss": 0.0931, + "step": 7367 + }, + { + "epoch": 0.97, + "learning_rate": 1.957227071405767e-05, + "loss": 0.0864, + "step": 7368 + }, + { + "epoch": 0.97, + "learning_rate": 1.9572117257535687e-05, + "loss": 0.0816, + "step": 7369 + }, + { + "epoch": 0.97, + "learning_rate": 1.9571963774092633e-05, + "loss": 0.0799, + "step": 7370 + }, + { + "epoch": 0.97, + "learning_rate": 1.9571810263728945e-05, + "loss": 0.0933, + "step": 7371 + }, + { + "epoch": 0.97, + "learning_rate": 1.9571656726445048e-05, + "loss": 0.0835, + "step": 7372 + }, + { + "epoch": 0.97, + "learning_rate": 1.9571503162241377e-05, + "loss": 0.0803, + "step": 7373 + }, + { + "epoch": 0.97, + "learning_rate": 1.9571349571118362e-05, + "loss": 0.0913, + "step": 7374 + }, + { + "epoch": 0.97, + "learning_rate": 1.9571195953076434e-05, + "loss": 0.0839, + "step": 7375 + }, + { + "epoch": 0.97, + "learning_rate": 1.9571042308116032e-05, + "loss": 0.1043, + "step": 7376 + }, + { + "epoch": 0.97, + "learning_rate": 1.957088863623758e-05, + "loss": 0.0798, + "step": 7377 + }, + { + "epoch": 0.97, + "learning_rate": 1.9570734937441512e-05, + "loss": 0.1043, + "step": 7378 + }, + { + "epoch": 0.97, + "learning_rate": 1.9570581211728265e-05, + "loss": 0.0749, + "step": 7379 + }, + { + "epoch": 0.97, + "learning_rate": 1.9570427459098266e-05, + "loss": 0.0807, + "step": 7380 + }, + { + "epoch": 0.97, + "learning_rate": 1.957027367955195e-05, + "loss": 0.1017, + "step": 7381 + }, + { + "epoch": 0.97, + "learning_rate": 1.9570119873089747e-05, + "loss": 0.0848, + "step": 7382 + }, + { + "epoch": 0.97, + "learning_rate": 1.9569966039712097e-05, + "loss": 0.1005, + "step": 7383 + }, + { + "epoch": 0.97, + "learning_rate": 1.9569812179419423e-05, + "loss": 0.0834, + "step": 7384 + }, + { + "epoch": 0.97, + "learning_rate": 1.9569658292212162e-05, + "loss": 0.0876, + "step": 7385 + }, + { + "epoch": 0.97, + "learning_rate": 1.9569504378090748e-05, + "loss": 0.0658, + "step": 7386 + }, + { + "epoch": 0.97, + "learning_rate": 1.9569350437055614e-05, + "loss": 0.0859, + "step": 7387 + }, + { + "epoch": 0.97, + "learning_rate": 1.9569196469107186e-05, + "loss": 0.0862, + "step": 7388 + }, + { + "epoch": 0.97, + "learning_rate": 1.9569042474245904e-05, + "loss": 0.0859, + "step": 7389 + }, + { + "epoch": 0.97, + "learning_rate": 1.9568888452472202e-05, + "loss": 0.0782, + "step": 7390 + }, + { + "epoch": 0.97, + "learning_rate": 1.9568734403786512e-05, + "loss": 0.0954, + "step": 7391 + }, + { + "epoch": 0.97, + "learning_rate": 1.9568580328189264e-05, + "loss": 0.0891, + "step": 7392 + }, + { + "epoch": 0.97, + "learning_rate": 1.9568426225680892e-05, + "loss": 0.0838, + "step": 7393 + }, + { + "epoch": 0.97, + "learning_rate": 1.9568272096261832e-05, + "loss": 0.0911, + "step": 7394 + }, + { + "epoch": 0.97, + "learning_rate": 1.956811793993252e-05, + "loss": 0.0902, + "step": 7395 + }, + { + "epoch": 0.97, + "learning_rate": 1.956796375669338e-05, + "loss": 0.0893, + "step": 7396 + }, + { + "epoch": 0.97, + "learning_rate": 1.9567809546544852e-05, + "loss": 0.0773, + "step": 7397 + }, + { + "epoch": 0.97, + "learning_rate": 1.9567655309487373e-05, + "loss": 0.0904, + "step": 7398 + }, + { + "epoch": 0.97, + "learning_rate": 1.956750104552137e-05, + "loss": 0.0843, + "step": 7399 + }, + { + "epoch": 0.97, + "learning_rate": 1.956734675464728e-05, + "loss": 0.0801, + "step": 7400 + }, + { + "epoch": 0.97, + "learning_rate": 1.956719243686554e-05, + "loss": 0.1029, + "step": 7401 + }, + { + "epoch": 0.97, + "learning_rate": 1.9567038092176576e-05, + "loss": 0.0874, + "step": 7402 + }, + { + "epoch": 0.97, + "learning_rate": 1.956688372058083e-05, + "loss": 0.0834, + "step": 7403 + }, + { + "epoch": 0.97, + "learning_rate": 1.956672932207873e-05, + "loss": 0.0812, + "step": 7404 + }, + { + "epoch": 0.97, + "learning_rate": 1.9566574896670717e-05, + "loss": 0.084, + "step": 7405 + }, + { + "epoch": 0.97, + "learning_rate": 1.956642044435722e-05, + "loss": 0.0763, + "step": 7406 + }, + { + "epoch": 0.97, + "learning_rate": 1.9566265965138675e-05, + "loss": 0.0752, + "step": 7407 + }, + { + "epoch": 0.97, + "learning_rate": 1.956611145901552e-05, + "loss": 0.0851, + "step": 7408 + }, + { + "epoch": 0.97, + "learning_rate": 1.956595692598818e-05, + "loss": 0.0855, + "step": 7409 + }, + { + "epoch": 0.97, + "learning_rate": 1.9565802366057096e-05, + "loss": 0.0861, + "step": 7410 + }, + { + "epoch": 0.97, + "learning_rate": 1.9565647779222705e-05, + "loss": 0.0921, + "step": 7411 + }, + { + "epoch": 0.97, + "learning_rate": 1.956549316548544e-05, + "loss": 0.0781, + "step": 7412 + }, + { + "epoch": 0.97, + "learning_rate": 1.9565338524845732e-05, + "loss": 0.087, + "step": 7413 + }, + { + "epoch": 0.97, + "learning_rate": 1.956518385730402e-05, + "loss": 0.0863, + "step": 7414 + }, + { + "epoch": 0.97, + "learning_rate": 1.9565029162860735e-05, + "loss": 0.086, + "step": 7415 + }, + { + "epoch": 0.97, + "learning_rate": 1.956487444151632e-05, + "loss": 0.0986, + "step": 7416 + }, + { + "epoch": 0.97, + "learning_rate": 1.95647196932712e-05, + "loss": 0.0913, + "step": 7417 + }, + { + "epoch": 0.97, + "learning_rate": 1.956456491812582e-05, + "loss": 0.0734, + "step": 7418 + }, + { + "epoch": 0.97, + "learning_rate": 1.9564410116080606e-05, + "loss": 0.0963, + "step": 7419 + }, + { + "epoch": 0.97, + "learning_rate": 1.9564255287136e-05, + "loss": 0.0855, + "step": 7420 + }, + { + "epoch": 0.97, + "learning_rate": 1.9564100431292434e-05, + "loss": 0.0881, + "step": 7421 + }, + { + "epoch": 0.97, + "learning_rate": 1.9563945548550348e-05, + "loss": 0.0891, + "step": 7422 + }, + { + "epoch": 0.97, + "learning_rate": 1.9563790638910172e-05, + "loss": 0.0776, + "step": 7423 + }, + { + "epoch": 0.97, + "learning_rate": 1.9563635702372347e-05, + "loss": 0.0974, + "step": 7424 + }, + { + "epoch": 0.97, + "learning_rate": 1.95634807389373e-05, + "loss": 0.079, + "step": 7425 + }, + { + "epoch": 0.97, + "learning_rate": 1.956332574860548e-05, + "loss": 0.0827, + "step": 7426 + }, + { + "epoch": 0.97, + "learning_rate": 1.956317073137731e-05, + "loss": 0.0778, + "step": 7427 + }, + { + "epoch": 0.97, + "learning_rate": 1.9563015687253233e-05, + "loss": 0.0863, + "step": 7428 + }, + { + "epoch": 0.97, + "learning_rate": 1.9562860616233683e-05, + "loss": 0.0964, + "step": 7429 + }, + { + "epoch": 0.97, + "learning_rate": 1.95627055183191e-05, + "loss": 0.0945, + "step": 7430 + }, + { + "epoch": 0.97, + "learning_rate": 1.9562550393509916e-05, + "loss": 0.079, + "step": 7431 + }, + { + "epoch": 0.97, + "learning_rate": 1.9562395241806565e-05, + "loss": 0.0855, + "step": 7432 + }, + { + "epoch": 0.97, + "learning_rate": 1.956224006320949e-05, + "loss": 0.0841, + "step": 7433 + }, + { + "epoch": 0.97, + "learning_rate": 1.9562084857719125e-05, + "loss": 0.0877, + "step": 7434 + }, + { + "epoch": 0.97, + "learning_rate": 1.95619296253359e-05, + "loss": 0.076, + "step": 7435 + }, + { + "epoch": 0.97, + "learning_rate": 1.956177436606026e-05, + "loss": 0.0782, + "step": 7436 + }, + { + "epoch": 0.97, + "learning_rate": 1.956161907989264e-05, + "loss": 0.0782, + "step": 7437 + }, + { + "epoch": 0.97, + "learning_rate": 1.9561463766833474e-05, + "loss": 0.0841, + "step": 7438 + }, + { + "epoch": 0.97, + "learning_rate": 1.95613084268832e-05, + "loss": 0.0857, + "step": 7439 + }, + { + "epoch": 0.97, + "learning_rate": 1.956115306004226e-05, + "loss": 0.0802, + "step": 7440 + }, + { + "epoch": 0.97, + "learning_rate": 1.956099766631108e-05, + "loss": 0.0885, + "step": 7441 + }, + { + "epoch": 0.97, + "learning_rate": 1.9560842245690105e-05, + "loss": 0.0818, + "step": 7442 + }, + { + "epoch": 0.97, + "learning_rate": 1.9560686798179772e-05, + "loss": 0.0908, + "step": 7443 + }, + { + "epoch": 0.98, + "learning_rate": 1.9560531323780518e-05, + "loss": 0.0789, + "step": 7444 + }, + { + "epoch": 0.98, + "learning_rate": 1.9560375822492775e-05, + "loss": 0.082, + "step": 7445 + }, + { + "epoch": 0.98, + "learning_rate": 1.9560220294316984e-05, + "loss": 0.0835, + "step": 7446 + }, + { + "epoch": 0.98, + "learning_rate": 1.9560064739253586e-05, + "loss": 0.0907, + "step": 7447 + }, + { + "epoch": 0.98, + "learning_rate": 1.955990915730301e-05, + "loss": 0.1008, + "step": 7448 + }, + { + "epoch": 0.98, + "learning_rate": 1.9559753548465705e-05, + "loss": 0.0836, + "step": 7449 + }, + { + "epoch": 0.98, + "learning_rate": 1.9559597912742097e-05, + "loss": 0.0818, + "step": 7450 + }, + { + "epoch": 0.98, + "learning_rate": 1.955944225013263e-05, + "loss": 0.0931, + "step": 7451 + }, + { + "epoch": 0.98, + "learning_rate": 1.955928656063774e-05, + "loss": 0.0777, + "step": 7452 + }, + { + "epoch": 0.98, + "learning_rate": 1.955913084425787e-05, + "loss": 0.0867, + "step": 7453 + }, + { + "epoch": 0.98, + "learning_rate": 1.9558975100993448e-05, + "loss": 0.0874, + "step": 7454 + }, + { + "epoch": 0.98, + "learning_rate": 1.955881933084492e-05, + "loss": 0.0854, + "step": 7455 + }, + { + "epoch": 0.98, + "learning_rate": 1.9558663533812725e-05, + "loss": 0.0927, + "step": 7456 + }, + { + "epoch": 0.98, + "learning_rate": 1.9558507709897295e-05, + "loss": 0.0929, + "step": 7457 + }, + { + "epoch": 0.98, + "learning_rate": 1.9558351859099072e-05, + "loss": 0.0773, + "step": 7458 + }, + { + "epoch": 0.98, + "learning_rate": 1.955819598141849e-05, + "loss": 0.0957, + "step": 7459 + }, + { + "epoch": 0.98, + "learning_rate": 1.9558040076855996e-05, + "loss": 0.0869, + "step": 7460 + }, + { + "epoch": 0.98, + "learning_rate": 1.9557884145412017e-05, + "loss": 0.0807, + "step": 7461 + }, + { + "epoch": 0.98, + "learning_rate": 1.9557728187087e-05, + "loss": 0.0863, + "step": 7462 + }, + { + "epoch": 0.98, + "learning_rate": 1.9557572201881386e-05, + "loss": 0.0889, + "step": 7463 + }, + { + "epoch": 0.98, + "learning_rate": 1.9557416189795605e-05, + "loss": 0.0829, + "step": 7464 + }, + { + "epoch": 0.98, + "learning_rate": 1.95572601508301e-05, + "loss": 0.089, + "step": 7465 + }, + { + "epoch": 0.98, + "learning_rate": 1.955710408498531e-05, + "loss": 0.092, + "step": 7466 + }, + { + "epoch": 0.98, + "learning_rate": 1.9556947992261673e-05, + "loss": 0.0852, + "step": 7467 + }, + { + "epoch": 0.98, + "learning_rate": 1.9556791872659632e-05, + "loss": 0.0837, + "step": 7468 + }, + { + "epoch": 0.98, + "learning_rate": 1.955663572617962e-05, + "loss": 0.0928, + "step": 7469 + }, + { + "epoch": 0.98, + "learning_rate": 1.955647955282208e-05, + "loss": 0.0904, + "step": 7470 + }, + { + "epoch": 0.98, + "learning_rate": 1.955632335258745e-05, + "loss": 0.089, + "step": 7471 + }, + { + "epoch": 0.98, + "learning_rate": 1.955616712547617e-05, + "loss": 0.0879, + "step": 7472 + }, + { + "epoch": 0.98, + "learning_rate": 1.9556010871488677e-05, + "loss": 0.0801, + "step": 7473 + }, + { + "epoch": 0.98, + "learning_rate": 1.9555854590625413e-05, + "loss": 0.082, + "step": 7474 + }, + { + "epoch": 0.98, + "learning_rate": 1.955569828288682e-05, + "loss": 0.0928, + "step": 7475 + }, + { + "epoch": 0.98, + "learning_rate": 1.955554194827333e-05, + "loss": 0.0796, + "step": 7476 + }, + { + "epoch": 0.98, + "learning_rate": 1.955538558678539e-05, + "loss": 0.0775, + "step": 7477 + }, + { + "epoch": 0.98, + "learning_rate": 1.9555229198423436e-05, + "loss": 0.0856, + "step": 7478 + }, + { + "epoch": 0.98, + "learning_rate": 1.955507278318791e-05, + "loss": 0.0763, + "step": 7479 + }, + { + "epoch": 0.98, + "learning_rate": 1.9554916341079248e-05, + "loss": 0.0935, + "step": 7480 + }, + { + "epoch": 0.98, + "learning_rate": 1.955475987209789e-05, + "loss": 0.0911, + "step": 7481 + }, + { + "epoch": 0.98, + "learning_rate": 1.9554603376244285e-05, + "loss": 0.0929, + "step": 7482 + }, + { + "epoch": 0.98, + "learning_rate": 1.9554446853518864e-05, + "loss": 0.0777, + "step": 7483 + }, + { + "epoch": 0.98, + "learning_rate": 1.955429030392207e-05, + "loss": 0.0787, + "step": 7484 + }, + { + "epoch": 0.98, + "learning_rate": 1.955413372745434e-05, + "loss": 0.0793, + "step": 7485 + }, + { + "epoch": 0.98, + "learning_rate": 1.9553977124116127e-05, + "loss": 0.0899, + "step": 7486 + }, + { + "epoch": 0.98, + "learning_rate": 1.9553820493907856e-05, + "loss": 0.0911, + "step": 7487 + }, + { + "epoch": 0.98, + "learning_rate": 1.9553663836829972e-05, + "loss": 0.0869, + "step": 7488 + }, + { + "epoch": 0.98, + "learning_rate": 1.955350715288292e-05, + "loss": 0.0893, + "step": 7489 + }, + { + "epoch": 0.98, + "learning_rate": 1.9553350442067137e-05, + "loss": 0.0958, + "step": 7490 + }, + { + "epoch": 0.98, + "learning_rate": 1.9553193704383067e-05, + "loss": 0.0769, + "step": 7491 + }, + { + "epoch": 0.98, + "learning_rate": 1.9553036939831145e-05, + "loss": 0.0836, + "step": 7492 + }, + { + "epoch": 0.98, + "learning_rate": 1.9552880148411816e-05, + "loss": 0.0791, + "step": 7493 + }, + { + "epoch": 0.98, + "learning_rate": 1.955272333012552e-05, + "loss": 0.0983, + "step": 7494 + }, + { + "epoch": 0.98, + "learning_rate": 1.95525664849727e-05, + "loss": 0.0872, + "step": 7495 + }, + { + "epoch": 0.98, + "learning_rate": 1.9552409612953794e-05, + "loss": 0.0835, + "step": 7496 + }, + { + "epoch": 0.98, + "learning_rate": 1.9552252714069247e-05, + "loss": 0.0811, + "step": 7497 + }, + { + "epoch": 0.98, + "learning_rate": 1.9552095788319495e-05, + "loss": 0.0869, + "step": 7498 + }, + { + "epoch": 0.98, + "learning_rate": 1.955193883570498e-05, + "loss": 0.0929, + "step": 7499 + }, + { + "epoch": 0.98, + "learning_rate": 1.9551781856226154e-05, + "loss": 0.0779, + "step": 7500 + }, + { + "epoch": 0.98, + "learning_rate": 1.9551624849883443e-05, + "loss": 0.0761, + "step": 7501 + }, + { + "epoch": 0.98, + "learning_rate": 1.95514678166773e-05, + "loss": 0.0852, + "step": 7502 + }, + { + "epoch": 0.98, + "learning_rate": 1.955131075660816e-05, + "loss": 0.095, + "step": 7503 + }, + { + "epoch": 0.98, + "learning_rate": 1.9551153669676467e-05, + "loss": 0.0782, + "step": 7504 + }, + { + "epoch": 0.98, + "learning_rate": 1.9550996555882663e-05, + "loss": 0.0838, + "step": 7505 + }, + { + "epoch": 0.98, + "learning_rate": 1.955083941522719e-05, + "loss": 0.0853, + "step": 7506 + }, + { + "epoch": 0.98, + "learning_rate": 1.9550682247710492e-05, + "loss": 0.0807, + "step": 7507 + }, + { + "epoch": 0.98, + "learning_rate": 1.9550525053333006e-05, + "loss": 0.094, + "step": 7508 + }, + { + "epoch": 0.98, + "learning_rate": 1.9550367832095177e-05, + "loss": 0.08, + "step": 7509 + }, + { + "epoch": 0.98, + "learning_rate": 1.9550210583997452e-05, + "loss": 0.0915, + "step": 7510 + }, + { + "epoch": 0.98, + "learning_rate": 1.955005330904026e-05, + "loss": 0.0863, + "step": 7511 + }, + { + "epoch": 0.98, + "learning_rate": 1.954989600722406e-05, + "loss": 0.0829, + "step": 7512 + }, + { + "epoch": 0.98, + "learning_rate": 1.954973867854928e-05, + "loss": 0.0797, + "step": 7513 + }, + { + "epoch": 0.98, + "learning_rate": 1.9549581323016368e-05, + "loss": 0.0862, + "step": 7514 + }, + { + "epoch": 0.98, + "learning_rate": 1.9549423940625774e-05, + "loss": 0.0796, + "step": 7515 + }, + { + "epoch": 0.98, + "learning_rate": 1.954926653137793e-05, + "loss": 0.0832, + "step": 7516 + }, + { + "epoch": 0.98, + "learning_rate": 1.9549109095273277e-05, + "loss": 0.0834, + "step": 7517 + }, + { + "epoch": 0.98, + "learning_rate": 1.9548951632312268e-05, + "loss": 0.0804, + "step": 7518 + }, + { + "epoch": 0.98, + "learning_rate": 1.9548794142495345e-05, + "loss": 0.0918, + "step": 7519 + }, + { + "epoch": 0.99, + "learning_rate": 1.954863662582294e-05, + "loss": 0.0986, + "step": 7520 + }, + { + "epoch": 0.99, + "learning_rate": 1.9548479082295505e-05, + "loss": 0.0825, + "step": 7521 + }, + { + "epoch": 0.99, + "learning_rate": 1.954832151191348e-05, + "loss": 0.0822, + "step": 7522 + }, + { + "epoch": 0.99, + "learning_rate": 1.9548163914677313e-05, + "loss": 0.0744, + "step": 7523 + }, + { + "epoch": 0.99, + "learning_rate": 1.9548006290587446e-05, + "loss": 0.0761, + "step": 7524 + }, + { + "epoch": 0.99, + "learning_rate": 1.9547848639644314e-05, + "loss": 0.09, + "step": 7525 + }, + { + "epoch": 0.99, + "learning_rate": 1.9547690961848363e-05, + "loss": 0.0853, + "step": 7526 + }, + { + "epoch": 0.99, + "learning_rate": 1.9547533257200046e-05, + "loss": 0.0835, + "step": 7527 + }, + { + "epoch": 0.99, + "learning_rate": 1.9547375525699798e-05, + "loss": 0.0878, + "step": 7528 + }, + { + "epoch": 0.99, + "learning_rate": 1.9547217767348067e-05, + "loss": 0.0848, + "step": 7529 + }, + { + "epoch": 0.99, + "learning_rate": 1.9547059982145293e-05, + "loss": 0.0837, + "step": 7530 + }, + { + "epoch": 0.99, + "learning_rate": 1.954690217009192e-05, + "loss": 0.0918, + "step": 7531 + }, + { + "epoch": 0.99, + "learning_rate": 1.9546744331188396e-05, + "loss": 0.0931, + "step": 7532 + }, + { + "epoch": 0.99, + "learning_rate": 1.9546586465435157e-05, + "loss": 0.079, + "step": 7533 + }, + { + "epoch": 0.99, + "learning_rate": 1.9546428572832654e-05, + "loss": 0.0882, + "step": 7534 + }, + { + "epoch": 0.99, + "learning_rate": 1.954627065338133e-05, + "loss": 0.0925, + "step": 7535 + }, + { + "epoch": 0.99, + "learning_rate": 1.954611270708163e-05, + "loss": 0.0802, + "step": 7536 + }, + { + "epoch": 0.99, + "learning_rate": 1.954595473393399e-05, + "loss": 0.0837, + "step": 7537 + }, + { + "epoch": 0.99, + "learning_rate": 1.9545796733938867e-05, + "loss": 0.0913, + "step": 7538 + }, + { + "epoch": 0.99, + "learning_rate": 1.9545638707096696e-05, + "loss": 0.0902, + "step": 7539 + }, + { + "epoch": 0.99, + "learning_rate": 1.9545480653407927e-05, + "loss": 0.0758, + "step": 7540 + }, + { + "epoch": 0.99, + "learning_rate": 1.9545322572873e-05, + "loss": 0.0782, + "step": 7541 + }, + { + "epoch": 0.99, + "learning_rate": 1.9545164465492362e-05, + "loss": 0.0796, + "step": 7542 + }, + { + "epoch": 0.99, + "learning_rate": 1.9545006331266457e-05, + "loss": 0.0809, + "step": 7543 + }, + { + "epoch": 0.99, + "learning_rate": 1.954484817019573e-05, + "loss": 0.0892, + "step": 7544 + }, + { + "epoch": 0.99, + "learning_rate": 1.9544689982280625e-05, + "loss": 0.085, + "step": 7545 + }, + { + "epoch": 0.99, + "learning_rate": 1.954453176752159e-05, + "loss": 0.082, + "step": 7546 + }, + { + "epoch": 0.99, + "learning_rate": 1.9544373525919065e-05, + "loss": 0.0826, + "step": 7547 + }, + { + "epoch": 0.99, + "learning_rate": 1.95442152574735e-05, + "loss": 0.0899, + "step": 7548 + }, + { + "epoch": 0.99, + "learning_rate": 1.9544056962185336e-05, + "loss": 0.0865, + "step": 7549 + }, + { + "epoch": 0.99, + "learning_rate": 1.9543898640055024e-05, + "loss": 0.0953, + "step": 7550 + }, + { + "epoch": 0.99, + "learning_rate": 1.9543740291083e-05, + "loss": 0.0939, + "step": 7551 + }, + { + "epoch": 0.99, + "learning_rate": 1.9543581915269717e-05, + "loss": 0.0869, + "step": 7552 + }, + { + "epoch": 0.99, + "learning_rate": 1.9543423512615618e-05, + "loss": 0.0746, + "step": 7553 + }, + { + "epoch": 0.99, + "learning_rate": 1.9543265083121146e-05, + "loss": 0.082, + "step": 7554 + }, + { + "epoch": 0.99, + "learning_rate": 1.9543106626786755e-05, + "loss": 0.0978, + "step": 7555 + }, + { + "epoch": 0.99, + "learning_rate": 1.9542948143612885e-05, + "loss": 0.0813, + "step": 7556 + }, + { + "epoch": 0.99, + "learning_rate": 1.9542789633599976e-05, + "loss": 0.0835, + "step": 7557 + }, + { + "epoch": 0.99, + "learning_rate": 1.9542631096748482e-05, + "loss": 0.0859, + "step": 7558 + }, + { + "epoch": 0.99, + "learning_rate": 1.9542472533058845e-05, + "loss": 0.0707, + "step": 7559 + }, + { + "epoch": 0.99, + "learning_rate": 1.954231394253151e-05, + "loss": 0.0794, + "step": 7560 + }, + { + "epoch": 0.99, + "learning_rate": 1.954215532516693e-05, + "loss": 0.0799, + "step": 7561 + }, + { + "epoch": 0.99, + "learning_rate": 1.9541996680965544e-05, + "loss": 0.0832, + "step": 7562 + }, + { + "epoch": 0.99, + "learning_rate": 1.9541838009927803e-05, + "loss": 0.0953, + "step": 7563 + }, + { + "epoch": 0.99, + "learning_rate": 1.9541679312054147e-05, + "loss": 0.0886, + "step": 7564 + }, + { + "epoch": 0.99, + "learning_rate": 1.954152058734503e-05, + "loss": 0.0849, + "step": 7565 + }, + { + "epoch": 0.99, + "learning_rate": 1.954136183580089e-05, + "loss": 0.082, + "step": 7566 + }, + { + "epoch": 0.99, + "learning_rate": 1.954120305742218e-05, + "loss": 0.0833, + "step": 7567 + }, + { + "epoch": 0.99, + "learning_rate": 1.9541044252209344e-05, + "loss": 0.0852, + "step": 7568 + }, + { + "epoch": 0.99, + "learning_rate": 1.9540885420162833e-05, + "loss": 0.0899, + "step": 7569 + }, + { + "epoch": 0.99, + "learning_rate": 1.9540726561283084e-05, + "loss": 0.0953, + "step": 7570 + }, + { + "epoch": 0.99, + "learning_rate": 1.954056767557055e-05, + "loss": 0.092, + "step": 7571 + }, + { + "epoch": 0.99, + "learning_rate": 1.954040876302568e-05, + "loss": 0.0885, + "step": 7572 + }, + { + "epoch": 0.99, + "learning_rate": 1.954024982364892e-05, + "loss": 0.0939, + "step": 7573 + }, + { + "epoch": 0.99, + "learning_rate": 1.954009085744071e-05, + "loss": 0.0831, + "step": 7574 + }, + { + "epoch": 0.99, + "learning_rate": 1.9539931864401507e-05, + "loss": 0.0799, + "step": 7575 + }, + { + "epoch": 0.99, + "learning_rate": 1.9539772844531756e-05, + "loss": 0.0842, + "step": 7576 + }, + { + "epoch": 0.99, + "learning_rate": 1.9539613797831894e-05, + "loss": 0.0831, + "step": 7577 + }, + { + "epoch": 0.99, + "learning_rate": 1.9539454724302382e-05, + "loss": 0.0796, + "step": 7578 + }, + { + "epoch": 0.99, + "learning_rate": 1.9539295623943663e-05, + "loss": 0.0879, + "step": 7579 + }, + { + "epoch": 0.99, + "learning_rate": 1.953913649675618e-05, + "loss": 0.0801, + "step": 7580 + }, + { + "epoch": 0.99, + "learning_rate": 1.9538977342740384e-05, + "loss": 0.0872, + "step": 7581 + }, + { + "epoch": 0.99, + "learning_rate": 1.9538818161896723e-05, + "loss": 0.0879, + "step": 7582 + }, + { + "epoch": 0.99, + "learning_rate": 1.9538658954225643e-05, + "loss": 0.0875, + "step": 7583 + }, + { + "epoch": 0.99, + "learning_rate": 1.9538499719727592e-05, + "loss": 0.0727, + "step": 7584 + }, + { + "epoch": 0.99, + "learning_rate": 1.953834045840302e-05, + "loss": 0.092, + "step": 7585 + }, + { + "epoch": 0.99, + "learning_rate": 1.9538181170252372e-05, + "loss": 0.0879, + "step": 7586 + }, + { + "epoch": 0.99, + "learning_rate": 1.9538021855276098e-05, + "loss": 0.0837, + "step": 7587 + }, + { + "epoch": 0.99, + "learning_rate": 1.9537862513474643e-05, + "loss": 0.0873, + "step": 7588 + }, + { + "epoch": 0.99, + "learning_rate": 1.9537703144848464e-05, + "loss": 0.0889, + "step": 7589 + }, + { + "epoch": 0.99, + "learning_rate": 1.9537543749398e-05, + "loss": 0.0793, + "step": 7590 + }, + { + "epoch": 0.99, + "learning_rate": 1.9537384327123697e-05, + "loss": 0.089, + "step": 7591 + }, + { + "epoch": 0.99, + "learning_rate": 1.9537224878026014e-05, + "loss": 0.0784, + "step": 7592 + }, + { + "epoch": 0.99, + "learning_rate": 1.9537065402105392e-05, + "loss": 0.0691, + "step": 7593 + }, + { + "epoch": 0.99, + "learning_rate": 1.9536905899362283e-05, + "loss": 0.0824, + "step": 7594 + }, + { + "epoch": 0.99, + "learning_rate": 1.953674636979713e-05, + "loss": 0.0812, + "step": 7595 + }, + { + "epoch": 1.0, + "learning_rate": 1.9536586813410388e-05, + "loss": 0.0913, + "step": 7596 + }, + { + "epoch": 1.0, + "learning_rate": 1.9536427230202507e-05, + "loss": 0.091, + "step": 7597 + }, + { + "epoch": 1.0, + "learning_rate": 1.9536267620173927e-05, + "loss": 0.0838, + "step": 7598 + }, + { + "epoch": 1.0, + "learning_rate": 1.9536107983325106e-05, + "loss": 0.0912, + "step": 7599 + }, + { + "epoch": 1.0, + "learning_rate": 1.9535948319656487e-05, + "loss": 0.0853, + "step": 7600 + }, + { + "epoch": 1.0, + "learning_rate": 1.9535788629168522e-05, + "loss": 0.0781, + "step": 7601 + }, + { + "epoch": 1.0, + "learning_rate": 1.953562891186166e-05, + "loss": 0.0711, + "step": 7602 + }, + { + "epoch": 1.0, + "learning_rate": 1.9535469167736347e-05, + "loss": 0.0783, + "step": 7603 + }, + { + "epoch": 1.0, + "learning_rate": 1.9535309396793035e-05, + "loss": 0.0892, + "step": 7604 + }, + { + "epoch": 1.0, + "learning_rate": 1.9535149599032176e-05, + "loss": 0.0788, + "step": 7605 + }, + { + "epoch": 1.0, + "learning_rate": 1.9534989774454213e-05, + "loss": 0.0886, + "step": 7606 + }, + { + "epoch": 1.0, + "learning_rate": 1.9534829923059603e-05, + "loss": 0.0778, + "step": 7607 + }, + { + "epoch": 1.0, + "learning_rate": 1.9534670044848788e-05, + "loss": 0.0833, + "step": 7608 + }, + { + "epoch": 1.0, + "learning_rate": 1.9534510139822224e-05, + "loss": 0.086, + "step": 7609 + }, + { + "epoch": 1.0, + "learning_rate": 1.953435020798036e-05, + "loss": 0.0842, + "step": 7610 + }, + { + "epoch": 1.0, + "learning_rate": 1.9534190249323636e-05, + "loss": 0.0742, + "step": 7611 + }, + { + "epoch": 1.0, + "learning_rate": 1.9534030263852516e-05, + "loss": 0.0801, + "step": 7612 + }, + { + "epoch": 1.0, + "learning_rate": 1.953387025156744e-05, + "loss": 0.087, + "step": 7613 + }, + { + "epoch": 1.0, + "learning_rate": 1.9533710212468863e-05, + "loss": 0.0751, + "step": 7614 + }, + { + "epoch": 1.0, + "learning_rate": 1.9533550146557237e-05, + "loss": 0.0763, + "step": 7615 + }, + { + "epoch": 1.0, + "learning_rate": 1.9533390053833004e-05, + "loss": 0.083, + "step": 7616 + }, + { + "epoch": 1.0, + "learning_rate": 1.9533229934296623e-05, + "loss": 0.0745, + "step": 7617 + }, + { + "epoch": 1.0, + "learning_rate": 1.9533069787948536e-05, + "loss": 0.085, + "step": 7618 + }, + { + "epoch": 1.0, + "learning_rate": 1.95329096147892e-05, + "loss": 0.0919, + "step": 7619 + }, + { + "epoch": 1.0, + "learning_rate": 1.953274941481907e-05, + "loss": 0.0906, + "step": 7620 + }, + { + "epoch": 1.0, + "learning_rate": 1.953258918803858e-05, + "loss": 0.0792, + "step": 7621 + }, + { + "epoch": 1.0, + "learning_rate": 1.9532428934448197e-05, + "loss": 0.0915, + "step": 7622 + }, + { + "epoch": 1.0, + "learning_rate": 1.9532268654048357e-05, + "loss": 0.0788, + "step": 7623 + }, + { + "epoch": 1.0, + "learning_rate": 1.953210834683953e-05, + "loss": 0.0887, + "step": 7624 + }, + { + "epoch": 1.0, + "learning_rate": 1.9531948012822147e-05, + "loss": 0.0991, + "step": 7625 + }, + { + "epoch": 1.0, + "learning_rate": 1.953178765199667e-05, + "loss": 0.0866, + "step": 7626 + }, + { + "epoch": 1.0, + "learning_rate": 1.9531627264363552e-05, + "loss": 0.0865, + "step": 7627 + }, + { + "epoch": 1.0, + "learning_rate": 1.953146684992324e-05, + "loss": 0.0913, + "step": 7628 + }, + { + "epoch": 1.0, + "learning_rate": 1.9531306408676182e-05, + "loss": 0.072, + "step": 7629 + }, + { + "epoch": 1.0, + "learning_rate": 1.953114594062283e-05, + "loss": 0.0819, + "step": 7630 + }, + { + "epoch": 1.0, + "learning_rate": 1.953098544576364e-05, + "loss": 0.0836, + "step": 7631 + }, + { + "epoch": 1.0, + "learning_rate": 1.9530824924099063e-05, + "loss": 0.077, + "step": 7632 + }, + { + "epoch": 1.0, + "learning_rate": 1.9530664375629545e-05, + "loss": 0.0837, + "step": 7633 + }, + { + "epoch": 1.0, + "learning_rate": 1.953050380035554e-05, + "loss": 0.0759, + "step": 7634 + }, + { + "epoch": 1.0, + "learning_rate": 1.9530343198277503e-05, + "loss": 0.0865, + "step": 7635 + }, + { + "epoch": 1.0, + "learning_rate": 1.9530182569395885e-05, + "loss": 0.0802, + "step": 7636 + }, + { + "epoch": 1.0, + "learning_rate": 1.9530021913711135e-05, + "loss": 0.0769, + "step": 7637 + }, + { + "epoch": 1.0, + "learning_rate": 1.9529861231223702e-05, + "loss": 0.0827, + "step": 7638 + }, + { + "epoch": 1.0, + "learning_rate": 1.952970052193405e-05, + "loss": 0.1, + "step": 7639 + }, + { + "epoch": 1.0, + "learning_rate": 1.9529539785842616e-05, + "loss": 0.0818, + "step": 7640 + }, + { + "epoch": 1.0, + "learning_rate": 1.952937902294986e-05, + "loss": 0.082, + "step": 7641 + }, + { + "epoch": 1.0, + "learning_rate": 1.9529218233256233e-05, + "loss": 0.0789, + "step": 7642 + }, + { + "epoch": 1.0, + "learning_rate": 1.952905741676219e-05, + "loss": 0.0836, + "step": 7643 + }, + { + "epoch": 1.0, + "learning_rate": 1.9528896573468178e-05, + "loss": 0.0859, + "step": 7644 + }, + { + "epoch": 1.0, + "learning_rate": 1.9528735703374652e-05, + "loss": 0.0859, + "step": 7645 + }, + { + "epoch": 1.0, + "learning_rate": 1.9528574806482065e-05, + "loss": 0.0771, + "step": 7646 + }, + { + "epoch": 1.0, + "learning_rate": 1.9528413882790866e-05, + "loss": 0.085, + "step": 7647 + }, + { + "epoch": 1.0, + "learning_rate": 1.9528252932301514e-05, + "loss": 0.0828, + "step": 7648 + }, + { + "epoch": 1.0, + "learning_rate": 1.9528091955014458e-05, + "loss": 0.0738, + "step": 7649 + }, + { + "epoch": 1.0, + "learning_rate": 1.9527930950930146e-05, + "loss": 0.0846, + "step": 7650 + }, + { + "epoch": 1.0, + "learning_rate": 1.952776992004904e-05, + "loss": 0.0839, + "step": 7651 + }, + { + "epoch": 1.0, + "learning_rate": 1.9527608862371587e-05, + "loss": 0.079, + "step": 7652 + }, + { + "epoch": 1.0, + "learning_rate": 1.9527447777898244e-05, + "loss": 0.0871, + "step": 7653 + }, + { + "epoch": 1.0, + "learning_rate": 1.9527286666629458e-05, + "loss": 0.0806, + "step": 7654 + }, + { + "epoch": 1.0, + "learning_rate": 1.9527125528565686e-05, + "loss": 0.0848, + "step": 7655 + }, + { + "epoch": 1.0, + "learning_rate": 1.9526964363707385e-05, + "loss": 0.0899, + "step": 7656 + }, + { + "epoch": 1.0, + "learning_rate": 1.9526803172054996e-05, + "loss": 0.0849, + "step": 7657 + }, + { + "epoch": 1.0, + "learning_rate": 1.9526641953608987e-05, + "loss": 0.0866, + "step": 7658 + }, + { + "epoch": 1.0, + "learning_rate": 1.9526480708369804e-05, + "loss": 0.0781, + "step": 7659 + }, + { + "epoch": 1.0, + "learning_rate": 1.95263194363379e-05, + "loss": 0.0869, + "step": 7660 + }, + { + "epoch": 1.0, + "learning_rate": 1.952615813751373e-05, + "loss": 0.0846, + "step": 7661 + }, + { + "epoch": 1.0, + "learning_rate": 1.9525996811897748e-05, + "loss": 0.0819, + "step": 7662 + }, + { + "epoch": 1.0, + "learning_rate": 1.9525835459490408e-05, + "loss": 0.0817, + "step": 7663 + }, + { + "epoch": 1.0, + "learning_rate": 1.952567408029216e-05, + "loss": 0.0766, + "step": 7664 + }, + { + "epoch": 1.0, + "learning_rate": 1.9525512674303465e-05, + "loss": 0.0779, + "step": 7665 + }, + { + "epoch": 1.0, + "learning_rate": 1.952535124152477e-05, + "loss": 0.0855, + "step": 7666 + }, + { + "epoch": 1.0, + "learning_rate": 1.952518978195653e-05, + "loss": 0.0882, + "step": 7667 + }, + { + "epoch": 1.0, + "learning_rate": 1.9525028295599208e-05, + "loss": 0.0806, + "step": 7668 + }, + { + "epoch": 1.0, + "learning_rate": 1.9524866782453245e-05, + "loss": 0.0779, + "step": 7669 + }, + { + "epoch": 1.0, + "learning_rate": 1.9524705242519106e-05, + "loss": 0.0914, + "step": 7670 + }, + { + "epoch": 1.0, + "learning_rate": 1.9524543675797238e-05, + "loss": 0.0835, + "step": 7671 + }, + { + "epoch": 1.0, + "learning_rate": 1.95243820822881e-05, + "loss": 0.0856, + "step": 7672 + }, + { + "epoch": 1.01, + "learning_rate": 1.952422046199214e-05, + "loss": 0.0927, + "step": 7673 + }, + { + "epoch": 1.01, + "learning_rate": 1.952405881490982e-05, + "loss": 0.0714, + "step": 7674 + }, + { + "epoch": 1.01, + "learning_rate": 1.9523897141041594e-05, + "loss": 0.0783, + "step": 7675 + }, + { + "epoch": 1.01, + "learning_rate": 1.9523735440387906e-05, + "loss": 0.0856, + "step": 7676 + }, + { + "epoch": 1.01, + "learning_rate": 1.9523573712949228e-05, + "loss": 0.0811, + "step": 7677 + }, + { + "epoch": 1.01, + "learning_rate": 1.9523411958726005e-05, + "loss": 0.0861, + "step": 7678 + }, + { + "epoch": 1.01, + "learning_rate": 1.952325017771869e-05, + "loss": 0.0796, + "step": 7679 + }, + { + "epoch": 1.01, + "learning_rate": 1.952308836992774e-05, + "loss": 0.0965, + "step": 7680 + }, + { + "epoch": 1.01, + "learning_rate": 1.9522926535353615e-05, + "loss": 0.0754, + "step": 7681 + }, + { + "epoch": 1.01, + "learning_rate": 1.9522764673996768e-05, + "loss": 0.0832, + "step": 7682 + }, + { + "epoch": 1.01, + "learning_rate": 1.9522602785857647e-05, + "loss": 0.083, + "step": 7683 + }, + { + "epoch": 1.01, + "learning_rate": 1.9522440870936716e-05, + "loss": 0.0748, + "step": 7684 + }, + { + "epoch": 1.01, + "learning_rate": 1.9522278929234426e-05, + "loss": 0.0918, + "step": 7685 + }, + { + "epoch": 1.01, + "learning_rate": 1.9522116960751235e-05, + "loss": 0.0883, + "step": 7686 + }, + { + "epoch": 1.01, + "learning_rate": 1.9521954965487596e-05, + "loss": 0.0886, + "step": 7687 + }, + { + "epoch": 1.01, + "learning_rate": 1.9521792943443967e-05, + "loss": 0.0861, + "step": 7688 + }, + { + "epoch": 1.01, + "learning_rate": 1.95216308946208e-05, + "loss": 0.0853, + "step": 7689 + }, + { + "epoch": 1.01, + "learning_rate": 1.9521468819018554e-05, + "loss": 0.0799, + "step": 7690 + }, + { + "epoch": 1.01, + "learning_rate": 1.9521306716637682e-05, + "loss": 0.0897, + "step": 7691 + }, + { + "epoch": 1.01, + "learning_rate": 1.9521144587478646e-05, + "loss": 0.0734, + "step": 7692 + }, + { + "epoch": 1.01, + "learning_rate": 1.9520982431541896e-05, + "loss": 0.0898, + "step": 7693 + }, + { + "epoch": 1.01, + "learning_rate": 1.952082024882789e-05, + "loss": 0.0811, + "step": 7694 + }, + { + "epoch": 1.01, + "learning_rate": 1.9520658039337084e-05, + "loss": 0.0726, + "step": 7695 + }, + { + "epoch": 1.01, + "learning_rate": 1.9520495803069935e-05, + "loss": 0.0811, + "step": 7696 + }, + { + "epoch": 1.01, + "learning_rate": 1.9520333540026896e-05, + "loss": 0.0919, + "step": 7697 + }, + { + "epoch": 1.01, + "learning_rate": 1.9520171250208427e-05, + "loss": 0.0853, + "step": 7698 + }, + { + "epoch": 1.01, + "learning_rate": 1.9520008933614986e-05, + "loss": 0.0922, + "step": 7699 + }, + { + "epoch": 1.01, + "learning_rate": 1.9519846590247023e-05, + "loss": 0.0832, + "step": 7700 + }, + { + "epoch": 1.01, + "learning_rate": 1.9519684220105003e-05, + "loss": 0.0771, + "step": 7701 + }, + { + "epoch": 1.01, + "learning_rate": 1.9519521823189376e-05, + "loss": 0.085, + "step": 7702 + }, + { + "epoch": 1.01, + "learning_rate": 1.95193593995006e-05, + "loss": 0.0778, + "step": 7703 + }, + { + "epoch": 1.01, + "learning_rate": 1.9519196949039133e-05, + "loss": 0.0874, + "step": 7704 + }, + { + "epoch": 1.01, + "learning_rate": 1.9519034471805433e-05, + "loss": 0.0792, + "step": 7705 + }, + { + "epoch": 1.01, + "learning_rate": 1.951887196779995e-05, + "loss": 0.0883, + "step": 7706 + }, + { + "epoch": 1.01, + "learning_rate": 1.9518709437023154e-05, + "loss": 0.078, + "step": 7707 + }, + { + "epoch": 1.01, + "learning_rate": 1.9518546879475487e-05, + "loss": 0.0738, + "step": 7708 + }, + { + "epoch": 1.01, + "learning_rate": 1.951838429515742e-05, + "loss": 0.0816, + "step": 7709 + }, + { + "epoch": 1.01, + "learning_rate": 1.9518221684069402e-05, + "loss": 0.0872, + "step": 7710 + }, + { + "epoch": 1.01, + "learning_rate": 1.951805904621189e-05, + "loss": 0.08, + "step": 7711 + }, + { + "epoch": 1.01, + "learning_rate": 1.9517896381585348e-05, + "loss": 0.0797, + "step": 7712 + }, + { + "epoch": 1.01, + "learning_rate": 1.951773369019023e-05, + "loss": 0.0876, + "step": 7713 + }, + { + "epoch": 1.01, + "learning_rate": 1.951757097202699e-05, + "loss": 0.0726, + "step": 7714 + }, + { + "epoch": 1.01, + "learning_rate": 1.9517408227096085e-05, + "loss": 0.0802, + "step": 7715 + }, + { + "epoch": 1.01, + "learning_rate": 1.951724545539798e-05, + "loss": 0.0995, + "step": 7716 + }, + { + "epoch": 1.01, + "learning_rate": 1.951708265693313e-05, + "loss": 0.0801, + "step": 7717 + }, + { + "epoch": 1.01, + "learning_rate": 1.951691983170199e-05, + "loss": 0.0989, + "step": 7718 + }, + { + "epoch": 1.01, + "learning_rate": 1.951675697970502e-05, + "loss": 0.0777, + "step": 7719 + }, + { + "epoch": 1.01, + "learning_rate": 1.951659410094268e-05, + "loss": 0.079, + "step": 7720 + }, + { + "epoch": 1.01, + "learning_rate": 1.9516431195415424e-05, + "loss": 0.0773, + "step": 7721 + }, + { + "epoch": 1.01, + "learning_rate": 1.951626826312371e-05, + "loss": 0.0767, + "step": 7722 + }, + { + "epoch": 1.01, + "learning_rate": 1.9516105304068004e-05, + "loss": 0.0903, + "step": 7723 + }, + { + "epoch": 1.01, + "learning_rate": 1.9515942318248756e-05, + "loss": 0.0792, + "step": 7724 + }, + { + "epoch": 1.01, + "learning_rate": 1.9515779305666425e-05, + "loss": 0.0818, + "step": 7725 + }, + { + "epoch": 1.01, + "learning_rate": 1.951561626632147e-05, + "loss": 0.0797, + "step": 7726 + }, + { + "epoch": 1.01, + "learning_rate": 1.9515453200214355e-05, + "loss": 0.0761, + "step": 7727 + }, + { + "epoch": 1.01, + "learning_rate": 1.951529010734553e-05, + "loss": 0.0867, + "step": 7728 + }, + { + "epoch": 1.01, + "learning_rate": 1.9515126987715464e-05, + "loss": 0.0873, + "step": 7729 + }, + { + "epoch": 1.01, + "learning_rate": 1.9514963841324607e-05, + "loss": 0.0777, + "step": 7730 + }, + { + "epoch": 1.01, + "learning_rate": 1.951480066817342e-05, + "loss": 0.0887, + "step": 7731 + }, + { + "epoch": 1.01, + "learning_rate": 1.9514637468262364e-05, + "loss": 0.0905, + "step": 7732 + }, + { + "epoch": 1.01, + "learning_rate": 1.9514474241591903e-05, + "loss": 0.0808, + "step": 7733 + }, + { + "epoch": 1.01, + "learning_rate": 1.951431098816248e-05, + "loss": 0.0691, + "step": 7734 + }, + { + "epoch": 1.01, + "learning_rate": 1.9514147707974568e-05, + "loss": 0.0764, + "step": 7735 + }, + { + "epoch": 1.01, + "learning_rate": 1.9513984401028622e-05, + "loss": 0.0828, + "step": 7736 + }, + { + "epoch": 1.01, + "learning_rate": 1.9513821067325103e-05, + "loss": 0.0877, + "step": 7737 + }, + { + "epoch": 1.01, + "learning_rate": 1.9513657706864466e-05, + "loss": 0.0734, + "step": 7738 + }, + { + "epoch": 1.01, + "learning_rate": 1.9513494319647174e-05, + "loss": 0.0917, + "step": 7739 + }, + { + "epoch": 1.01, + "learning_rate": 1.951333090567369e-05, + "loss": 0.0869, + "step": 7740 + }, + { + "epoch": 1.01, + "learning_rate": 1.9513167464944463e-05, + "loss": 0.0831, + "step": 7741 + }, + { + "epoch": 1.01, + "learning_rate": 1.9513003997459964e-05, + "loss": 0.0983, + "step": 7742 + }, + { + "epoch": 1.01, + "learning_rate": 1.9512840503220644e-05, + "loss": 0.0876, + "step": 7743 + }, + { + "epoch": 1.01, + "learning_rate": 1.951267698222697e-05, + "loss": 0.0871, + "step": 7744 + }, + { + "epoch": 1.01, + "learning_rate": 1.9512513434479396e-05, + "loss": 0.097, + "step": 7745 + }, + { + "epoch": 1.01, + "learning_rate": 1.9512349859978386e-05, + "loss": 0.0829, + "step": 7746 + }, + { + "epoch": 1.01, + "learning_rate": 1.95121862587244e-05, + "loss": 0.0966, + "step": 7747 + }, + { + "epoch": 1.01, + "learning_rate": 1.9512022630717896e-05, + "loss": 0.0784, + "step": 7748 + }, + { + "epoch": 1.02, + "learning_rate": 1.951185897595933e-05, + "loss": 0.087, + "step": 7749 + }, + { + "epoch": 1.02, + "learning_rate": 1.9511695294449172e-05, + "loss": 0.0791, + "step": 7750 + }, + { + "epoch": 1.02, + "learning_rate": 1.9511531586187876e-05, + "loss": 0.0856, + "step": 7751 + }, + { + "epoch": 1.02, + "learning_rate": 1.9511367851175908e-05, + "loss": 0.0843, + "step": 7752 + }, + { + "epoch": 1.02, + "learning_rate": 1.951120408941372e-05, + "loss": 0.0796, + "step": 7753 + }, + { + "epoch": 1.02, + "learning_rate": 1.951104030090178e-05, + "loss": 0.0873, + "step": 7754 + }, + { + "epoch": 1.02, + "learning_rate": 1.951087648564054e-05, + "loss": 0.0814, + "step": 7755 + }, + { + "epoch": 1.02, + "learning_rate": 1.9510712643630473e-05, + "loss": 0.0887, + "step": 7756 + }, + { + "epoch": 1.02, + "learning_rate": 1.951054877487203e-05, + "loss": 0.0791, + "step": 7757 + }, + { + "epoch": 1.02, + "learning_rate": 1.9510384879365674e-05, + "loss": 0.0867, + "step": 7758 + }, + { + "epoch": 1.02, + "learning_rate": 1.9510220957111868e-05, + "loss": 0.083, + "step": 7759 + }, + { + "epoch": 1.02, + "learning_rate": 1.9510057008111073e-05, + "loss": 0.0855, + "step": 7760 + }, + { + "epoch": 1.02, + "learning_rate": 1.950989303236375e-05, + "loss": 0.078, + "step": 7761 + }, + { + "epoch": 1.02, + "learning_rate": 1.9509729029870356e-05, + "loss": 0.0914, + "step": 7762 + }, + { + "epoch": 1.02, + "learning_rate": 1.9509565000631356e-05, + "loss": 0.0756, + "step": 7763 + }, + { + "epoch": 1.02, + "learning_rate": 1.9509400944647213e-05, + "loss": 0.0843, + "step": 7764 + }, + { + "epoch": 1.02, + "learning_rate": 1.9509236861918387e-05, + "loss": 0.0829, + "step": 7765 + }, + { + "epoch": 1.02, + "learning_rate": 1.9509072752445336e-05, + "loss": 0.0807, + "step": 7766 + }, + { + "epoch": 1.02, + "learning_rate": 1.9508908616228525e-05, + "loss": 0.0825, + "step": 7767 + }, + { + "epoch": 1.02, + "learning_rate": 1.9508744453268417e-05, + "loss": 0.0888, + "step": 7768 + }, + { + "epoch": 1.02, + "learning_rate": 1.9508580263565466e-05, + "loss": 0.0756, + "step": 7769 + }, + { + "epoch": 1.02, + "learning_rate": 1.9508416047120143e-05, + "loss": 0.0798, + "step": 7770 + }, + { + "epoch": 1.02, + "learning_rate": 1.9508251803932906e-05, + "loss": 0.0864, + "step": 7771 + }, + { + "epoch": 1.02, + "learning_rate": 1.9508087534004222e-05, + "loss": 0.0779, + "step": 7772 + }, + { + "epoch": 1.02, + "learning_rate": 1.9507923237334542e-05, + "loss": 0.0908, + "step": 7773 + }, + { + "epoch": 1.02, + "learning_rate": 1.9507758913924337e-05, + "loss": 0.089, + "step": 7774 + }, + { + "epoch": 1.02, + "learning_rate": 1.9507594563774064e-05, + "loss": 0.0802, + "step": 7775 + }, + { + "epoch": 1.02, + "learning_rate": 1.9507430186884187e-05, + "loss": 0.0782, + "step": 7776 + }, + { + "epoch": 1.02, + "learning_rate": 1.9507265783255174e-05, + "loss": 0.0725, + "step": 7777 + }, + { + "epoch": 1.02, + "learning_rate": 1.950710135288748e-05, + "loss": 0.091, + "step": 7778 + }, + { + "epoch": 1.02, + "learning_rate": 1.9506936895781567e-05, + "loss": 0.0652, + "step": 7779 + }, + { + "epoch": 1.02, + "learning_rate": 1.9506772411937904e-05, + "loss": 0.0919, + "step": 7780 + }, + { + "epoch": 1.02, + "learning_rate": 1.9506607901356944e-05, + "loss": 0.0802, + "step": 7781 + }, + { + "epoch": 1.02, + "learning_rate": 1.950644336403916e-05, + "loss": 0.0861, + "step": 7782 + }, + { + "epoch": 1.02, + "learning_rate": 1.950627879998501e-05, + "loss": 0.0705, + "step": 7783 + }, + { + "epoch": 1.02, + "learning_rate": 1.9506114209194957e-05, + "loss": 0.0841, + "step": 7784 + }, + { + "epoch": 1.02, + "learning_rate": 1.9505949591669464e-05, + "loss": 0.0859, + "step": 7785 + }, + { + "epoch": 1.02, + "learning_rate": 1.9505784947408993e-05, + "loss": 0.0815, + "step": 7786 + }, + { + "epoch": 1.02, + "learning_rate": 1.950562027641401e-05, + "loss": 0.0886, + "step": 7787 + }, + { + "epoch": 1.02, + "learning_rate": 1.9505455578684973e-05, + "loss": 0.0874, + "step": 7788 + }, + { + "epoch": 1.02, + "learning_rate": 1.9505290854222353e-05, + "loss": 0.0805, + "step": 7789 + }, + { + "epoch": 1.02, + "learning_rate": 1.9505126103026605e-05, + "loss": 0.084, + "step": 7790 + }, + { + "epoch": 1.02, + "learning_rate": 1.9504961325098195e-05, + "loss": 0.0788, + "step": 7791 + }, + { + "epoch": 1.02, + "learning_rate": 1.950479652043759e-05, + "loss": 0.0775, + "step": 7792 + }, + { + "epoch": 1.02, + "learning_rate": 1.9504631689045254e-05, + "loss": 0.0835, + "step": 7793 + }, + { + "epoch": 1.02, + "learning_rate": 1.950446683092164e-05, + "loss": 0.0805, + "step": 7794 + }, + { + "epoch": 1.02, + "learning_rate": 1.9504301946067225e-05, + "loss": 0.088, + "step": 7795 + }, + { + "epoch": 1.02, + "learning_rate": 1.9504137034482467e-05, + "loss": 0.0851, + "step": 7796 + }, + { + "epoch": 1.02, + "learning_rate": 1.9503972096167826e-05, + "loss": 0.0771, + "step": 7797 + }, + { + "epoch": 1.02, + "learning_rate": 1.950380713112377e-05, + "loss": 0.0834, + "step": 7798 + }, + { + "epoch": 1.02, + "learning_rate": 1.9503642139350766e-05, + "loss": 0.0828, + "step": 7799 + }, + { + "epoch": 1.02, + "learning_rate": 1.950347712084927e-05, + "loss": 0.0831, + "step": 7800 + }, + { + "epoch": 1.02, + "learning_rate": 1.9503312075619757e-05, + "loss": 0.0609, + "step": 7801 + }, + { + "epoch": 1.02, + "learning_rate": 1.950314700366268e-05, + "loss": 0.0814, + "step": 7802 + }, + { + "epoch": 1.02, + "learning_rate": 1.950298190497851e-05, + "loss": 0.0757, + "step": 7803 + }, + { + "epoch": 1.02, + "learning_rate": 1.9502816779567708e-05, + "loss": 0.0956, + "step": 7804 + }, + { + "epoch": 1.02, + "learning_rate": 1.950265162743074e-05, + "loss": 0.0897, + "step": 7805 + }, + { + "epoch": 1.02, + "learning_rate": 1.950248644856807e-05, + "loss": 0.0798, + "step": 7806 + }, + { + "epoch": 1.02, + "learning_rate": 1.9502321242980165e-05, + "loss": 0.0784, + "step": 7807 + }, + { + "epoch": 1.02, + "learning_rate": 1.9502156010667486e-05, + "loss": 0.0922, + "step": 7808 + }, + { + "epoch": 1.02, + "learning_rate": 1.9501990751630498e-05, + "loss": 0.0693, + "step": 7809 + }, + { + "epoch": 1.02, + "learning_rate": 1.9501825465869668e-05, + "loss": 0.0816, + "step": 7810 + }, + { + "epoch": 1.02, + "learning_rate": 1.9501660153385463e-05, + "loss": 0.0812, + "step": 7811 + }, + { + "epoch": 1.02, + "learning_rate": 1.950149481417834e-05, + "loss": 0.08, + "step": 7812 + }, + { + "epoch": 1.02, + "learning_rate": 1.950132944824877e-05, + "loss": 0.0876, + "step": 7813 + }, + { + "epoch": 1.02, + "learning_rate": 1.950116405559722e-05, + "loss": 0.0855, + "step": 7814 + }, + { + "epoch": 1.02, + "learning_rate": 1.950099863622415e-05, + "loss": 0.0685, + "step": 7815 + }, + { + "epoch": 1.02, + "learning_rate": 1.9500833190130025e-05, + "loss": 0.0762, + "step": 7816 + }, + { + "epoch": 1.02, + "learning_rate": 1.9500667717315317e-05, + "loss": 0.0808, + "step": 7817 + }, + { + "epoch": 1.02, + "learning_rate": 1.9500502217780485e-05, + "loss": 0.08, + "step": 7818 + }, + { + "epoch": 1.02, + "learning_rate": 1.9500336691525998e-05, + "loss": 0.0786, + "step": 7819 + }, + { + "epoch": 1.02, + "learning_rate": 1.9500171138552317e-05, + "loss": 0.0798, + "step": 7820 + }, + { + "epoch": 1.02, + "learning_rate": 1.9500005558859914e-05, + "loss": 0.0887, + "step": 7821 + }, + { + "epoch": 1.02, + "learning_rate": 1.949983995244925e-05, + "loss": 0.0809, + "step": 7822 + }, + { + "epoch": 1.02, + "learning_rate": 1.9499674319320787e-05, + "loss": 0.0937, + "step": 7823 + }, + { + "epoch": 1.02, + "learning_rate": 1.9499508659475e-05, + "loss": 0.0744, + "step": 7824 + }, + { + "epoch": 1.03, + "learning_rate": 1.9499342972912353e-05, + "loss": 0.076, + "step": 7825 + }, + { + "epoch": 1.03, + "learning_rate": 1.9499177259633306e-05, + "loss": 0.0783, + "step": 7826 + }, + { + "epoch": 1.03, + "learning_rate": 1.9499011519638332e-05, + "loss": 0.0823, + "step": 7827 + }, + { + "epoch": 1.03, + "learning_rate": 1.9498845752927892e-05, + "loss": 0.0797, + "step": 7828 + }, + { + "epoch": 1.03, + "learning_rate": 1.9498679959502452e-05, + "loss": 0.0867, + "step": 7829 + }, + { + "epoch": 1.03, + "learning_rate": 1.9498514139362485e-05, + "loss": 0.0791, + "step": 7830 + }, + { + "epoch": 1.03, + "learning_rate": 1.949834829250845e-05, + "loss": 0.0781, + "step": 7831 + }, + { + "epoch": 1.03, + "learning_rate": 1.949818241894082e-05, + "loss": 0.0828, + "step": 7832 + }, + { + "epoch": 1.03, + "learning_rate": 1.949801651866005e-05, + "loss": 0.0822, + "step": 7833 + }, + { + "epoch": 1.03, + "learning_rate": 1.9497850591666624e-05, + "loss": 0.0893, + "step": 7834 + }, + { + "epoch": 1.03, + "learning_rate": 1.949768463796099e-05, + "loss": 0.0798, + "step": 7835 + }, + { + "epoch": 1.03, + "learning_rate": 1.949751865754363e-05, + "loss": 0.0796, + "step": 7836 + }, + { + "epoch": 1.03, + "learning_rate": 1.9497352650415004e-05, + "loss": 0.0866, + "step": 7837 + }, + { + "epoch": 1.03, + "learning_rate": 1.9497186616575576e-05, + "loss": 0.0841, + "step": 7838 + }, + { + "epoch": 1.03, + "learning_rate": 1.9497020556025818e-05, + "loss": 0.0752, + "step": 7839 + }, + { + "epoch": 1.03, + "learning_rate": 1.9496854468766194e-05, + "loss": 0.0848, + "step": 7840 + }, + { + "epoch": 1.03, + "learning_rate": 1.9496688354797178e-05, + "loss": 0.0769, + "step": 7841 + }, + { + "epoch": 1.03, + "learning_rate": 1.9496522214119228e-05, + "loss": 0.0868, + "step": 7842 + }, + { + "epoch": 1.03, + "learning_rate": 1.9496356046732815e-05, + "loss": 0.0807, + "step": 7843 + }, + { + "epoch": 1.03, + "learning_rate": 1.9496189852638406e-05, + "loss": 0.0828, + "step": 7844 + }, + { + "epoch": 1.03, + "learning_rate": 1.949602363183647e-05, + "loss": 0.0894, + "step": 7845 + }, + { + "epoch": 1.03, + "learning_rate": 1.949585738432747e-05, + "loss": 0.0909, + "step": 7846 + }, + { + "epoch": 1.03, + "learning_rate": 1.949569111011188e-05, + "loss": 0.1053, + "step": 7847 + }, + { + "epoch": 1.03, + "learning_rate": 1.9495524809190164e-05, + "loss": 0.0874, + "step": 7848 + }, + { + "epoch": 1.03, + "learning_rate": 1.9495358481562786e-05, + "loss": 0.0815, + "step": 7849 + }, + { + "epoch": 1.03, + "learning_rate": 1.9495192127230224e-05, + "loss": 0.0708, + "step": 7850 + }, + { + "epoch": 1.03, + "learning_rate": 1.9495025746192937e-05, + "loss": 0.0938, + "step": 7851 + }, + { + "epoch": 1.03, + "learning_rate": 1.9494859338451397e-05, + "loss": 0.0833, + "step": 7852 + }, + { + "epoch": 1.03, + "learning_rate": 1.949469290400607e-05, + "loss": 0.0823, + "step": 7853 + }, + { + "epoch": 1.03, + "learning_rate": 1.9494526442857423e-05, + "loss": 0.0732, + "step": 7854 + }, + { + "epoch": 1.03, + "learning_rate": 1.949435995500593e-05, + "loss": 0.0773, + "step": 7855 + }, + { + "epoch": 1.03, + "learning_rate": 1.9494193440452052e-05, + "loss": 0.0845, + "step": 7856 + }, + { + "epoch": 1.03, + "learning_rate": 1.949402689919626e-05, + "loss": 0.0741, + "step": 7857 + }, + { + "epoch": 1.03, + "learning_rate": 1.9493860331239024e-05, + "loss": 0.0855, + "step": 7858 + }, + { + "epoch": 1.03, + "learning_rate": 1.9493693736580813e-05, + "loss": 0.0725, + "step": 7859 + }, + { + "epoch": 1.03, + "learning_rate": 1.949352711522209e-05, + "loss": 0.0784, + "step": 7860 + }, + { + "epoch": 1.03, + "learning_rate": 1.9493360467163335e-05, + "loss": 0.0943, + "step": 7861 + }, + { + "epoch": 1.03, + "learning_rate": 1.9493193792405003e-05, + "loss": 0.0789, + "step": 7862 + }, + { + "epoch": 1.03, + "learning_rate": 1.9493027090947573e-05, + "loss": 0.0882, + "step": 7863 + }, + { + "epoch": 1.03, + "learning_rate": 1.9492860362791504e-05, + "loss": 0.0922, + "step": 7864 + }, + { + "epoch": 1.03, + "learning_rate": 1.9492693607937276e-05, + "loss": 0.0856, + "step": 7865 + }, + { + "epoch": 1.03, + "learning_rate": 1.9492526826385348e-05, + "loss": 0.0958, + "step": 7866 + }, + { + "epoch": 1.03, + "learning_rate": 1.94923600181362e-05, + "loss": 0.0887, + "step": 7867 + }, + { + "epoch": 1.03, + "learning_rate": 1.949219318319029e-05, + "loss": 0.0897, + "step": 7868 + }, + { + "epoch": 1.03, + "learning_rate": 1.9492026321548093e-05, + "loss": 0.0787, + "step": 7869 + }, + { + "epoch": 1.03, + "learning_rate": 1.9491859433210078e-05, + "loss": 0.0861, + "step": 7870 + }, + { + "epoch": 1.03, + "learning_rate": 1.9491692518176715e-05, + "loss": 0.0871, + "step": 7871 + }, + { + "epoch": 1.03, + "learning_rate": 1.9491525576448472e-05, + "loss": 0.0701, + "step": 7872 + }, + { + "epoch": 1.03, + "learning_rate": 1.9491358608025816e-05, + "loss": 0.0847, + "step": 7873 + }, + { + "epoch": 1.03, + "learning_rate": 1.949119161290922e-05, + "loss": 0.0891, + "step": 7874 + }, + { + "epoch": 1.03, + "learning_rate": 1.9491024591099153e-05, + "loss": 0.0878, + "step": 7875 + }, + { + "epoch": 1.03, + "learning_rate": 1.9490857542596088e-05, + "loss": 0.0824, + "step": 7876 + }, + { + "epoch": 1.03, + "learning_rate": 1.949069046740049e-05, + "loss": 0.0873, + "step": 7877 + }, + { + "epoch": 1.03, + "learning_rate": 1.949052336551283e-05, + "loss": 0.089, + "step": 7878 + }, + { + "epoch": 1.03, + "learning_rate": 1.949035623693358e-05, + "loss": 0.0811, + "step": 7879 + }, + { + "epoch": 1.03, + "learning_rate": 1.9490189081663207e-05, + "loss": 0.0932, + "step": 7880 + }, + { + "epoch": 1.03, + "learning_rate": 1.9490021899702183e-05, + "loss": 0.0821, + "step": 7881 + }, + { + "epoch": 1.03, + "learning_rate": 1.9489854691050978e-05, + "loss": 0.0865, + "step": 7882 + }, + { + "epoch": 1.03, + "learning_rate": 1.9489687455710063e-05, + "loss": 0.0779, + "step": 7883 + }, + { + "epoch": 1.03, + "learning_rate": 1.9489520193679903e-05, + "loss": 0.0876, + "step": 7884 + }, + { + "epoch": 1.03, + "learning_rate": 1.9489352904960977e-05, + "loss": 0.0853, + "step": 7885 + }, + { + "epoch": 1.03, + "learning_rate": 1.948918558955375e-05, + "loss": 0.086, + "step": 7886 + }, + { + "epoch": 1.03, + "learning_rate": 1.9489018247458692e-05, + "loss": 0.0867, + "step": 7887 + }, + { + "epoch": 1.03, + "learning_rate": 1.9488850878676278e-05, + "loss": 0.0861, + "step": 7888 + }, + { + "epoch": 1.03, + "learning_rate": 1.9488683483206976e-05, + "loss": 0.0958, + "step": 7889 + }, + { + "epoch": 1.03, + "learning_rate": 1.9488516061051257e-05, + "loss": 0.0991, + "step": 7890 + }, + { + "epoch": 1.03, + "learning_rate": 1.9488348612209592e-05, + "loss": 0.0944, + "step": 7891 + }, + { + "epoch": 1.03, + "learning_rate": 1.948818113668245e-05, + "loss": 0.0816, + "step": 7892 + }, + { + "epoch": 1.03, + "learning_rate": 1.9488013634470303e-05, + "loss": 0.0817, + "step": 7893 + }, + { + "epoch": 1.03, + "learning_rate": 1.948784610557363e-05, + "loss": 0.0885, + "step": 7894 + }, + { + "epoch": 1.03, + "learning_rate": 1.948767854999289e-05, + "loss": 0.0869, + "step": 7895 + }, + { + "epoch": 1.03, + "learning_rate": 1.948751096772856e-05, + "loss": 0.0786, + "step": 7896 + }, + { + "epoch": 1.03, + "learning_rate": 1.9487343358781106e-05, + "loss": 0.083, + "step": 7897 + }, + { + "epoch": 1.03, + "learning_rate": 1.948717572315101e-05, + "loss": 0.0863, + "step": 7898 + }, + { + "epoch": 1.03, + "learning_rate": 1.9487008060838735e-05, + "loss": 0.0768, + "step": 7899 + }, + { + "epoch": 1.03, + "learning_rate": 1.9486840371844757e-05, + "loss": 0.0816, + "step": 7900 + }, + { + "epoch": 1.03, + "learning_rate": 1.9486672656169544e-05, + "loss": 0.0866, + "step": 7901 + }, + { + "epoch": 1.04, + "learning_rate": 1.948650491381357e-05, + "loss": 0.0948, + "step": 7902 + }, + { + "epoch": 1.04, + "learning_rate": 1.9486337144777305e-05, + "loss": 0.0848, + "step": 7903 + }, + { + "epoch": 1.04, + "learning_rate": 1.9486169349061222e-05, + "loss": 0.0772, + "step": 7904 + }, + { + "epoch": 1.04, + "learning_rate": 1.9486001526665793e-05, + "loss": 0.0819, + "step": 7905 + }, + { + "epoch": 1.04, + "learning_rate": 1.9485833677591493e-05, + "loss": 0.0781, + "step": 7906 + }, + { + "epoch": 1.04, + "learning_rate": 1.9485665801838786e-05, + "loss": 0.0879, + "step": 7907 + }, + { + "epoch": 1.04, + "learning_rate": 1.9485497899408152e-05, + "loss": 0.0923, + "step": 7908 + }, + { + "epoch": 1.04, + "learning_rate": 1.948532997030006e-05, + "loss": 0.0785, + "step": 7909 + }, + { + "epoch": 1.04, + "learning_rate": 1.948516201451498e-05, + "loss": 0.083, + "step": 7910 + }, + { + "epoch": 1.04, + "learning_rate": 1.948499403205339e-05, + "loss": 0.0823, + "step": 7911 + }, + { + "epoch": 1.04, + "learning_rate": 1.9484826022915763e-05, + "loss": 0.0827, + "step": 7912 + }, + { + "epoch": 1.04, + "learning_rate": 1.948465798710256e-05, + "loss": 0.085, + "step": 7913 + }, + { + "epoch": 1.04, + "learning_rate": 1.9484489924614266e-05, + "loss": 0.0827, + "step": 7914 + }, + { + "epoch": 1.04, + "learning_rate": 1.9484321835451347e-05, + "loss": 0.0818, + "step": 7915 + }, + { + "epoch": 1.04, + "learning_rate": 1.948415371961428e-05, + "loss": 0.0919, + "step": 7916 + }, + { + "epoch": 1.04, + "learning_rate": 1.9483985577103538e-05, + "loss": 0.0851, + "step": 7917 + }, + { + "epoch": 1.04, + "learning_rate": 1.9483817407919586e-05, + "loss": 0.0897, + "step": 7918 + }, + { + "epoch": 1.04, + "learning_rate": 1.9483649212062906e-05, + "loss": 0.0872, + "step": 7919 + }, + { + "epoch": 1.04, + "learning_rate": 1.9483480989533965e-05, + "loss": 0.091, + "step": 7920 + }, + { + "epoch": 1.04, + "learning_rate": 1.9483312740333243e-05, + "loss": 0.076, + "step": 7921 + }, + { + "epoch": 1.04, + "learning_rate": 1.9483144464461206e-05, + "loss": 0.0722, + "step": 7922 + }, + { + "epoch": 1.04, + "learning_rate": 1.948297616191833e-05, + "loss": 0.0775, + "step": 7923 + }, + { + "epoch": 1.04, + "learning_rate": 1.948280783270509e-05, + "loss": 0.0834, + "step": 7924 + }, + { + "epoch": 1.04, + "learning_rate": 1.9482639476821953e-05, + "loss": 0.081, + "step": 7925 + }, + { + "epoch": 1.04, + "learning_rate": 1.9482471094269403e-05, + "loss": 0.0862, + "step": 7926 + }, + { + "epoch": 1.04, + "learning_rate": 1.9482302685047908e-05, + "loss": 0.0861, + "step": 7927 + }, + { + "epoch": 1.04, + "learning_rate": 1.948213424915794e-05, + "loss": 0.0681, + "step": 7928 + }, + { + "epoch": 1.04, + "learning_rate": 1.9481965786599974e-05, + "loss": 0.0774, + "step": 7929 + }, + { + "epoch": 1.04, + "learning_rate": 1.9481797297374486e-05, + "loss": 0.0884, + "step": 7930 + }, + { + "epoch": 1.04, + "learning_rate": 1.9481628781481948e-05, + "loss": 0.0813, + "step": 7931 + }, + { + "epoch": 1.04, + "learning_rate": 1.948146023892283e-05, + "loss": 0.0834, + "step": 7932 + }, + { + "epoch": 1.04, + "learning_rate": 1.9481291669697616e-05, + "loss": 0.0744, + "step": 7933 + }, + { + "epoch": 1.04, + "learning_rate": 1.9481123073806767e-05, + "loss": 0.0776, + "step": 7934 + }, + { + "epoch": 1.04, + "learning_rate": 1.9480954451250768e-05, + "loss": 0.0862, + "step": 7935 + }, + { + "epoch": 1.04, + "learning_rate": 1.948078580203009e-05, + "loss": 0.0836, + "step": 7936 + }, + { + "epoch": 1.04, + "learning_rate": 1.948061712614521e-05, + "loss": 0.0892, + "step": 7937 + }, + { + "epoch": 1.04, + "learning_rate": 1.9480448423596595e-05, + "loss": 0.0823, + "step": 7938 + }, + { + "epoch": 1.04, + "learning_rate": 1.948027969438472e-05, + "loss": 0.0871, + "step": 7939 + }, + { + "epoch": 1.04, + "learning_rate": 1.948011093851007e-05, + "loss": 0.0901, + "step": 7940 + }, + { + "epoch": 1.04, + "learning_rate": 1.947994215597311e-05, + "loss": 0.0795, + "step": 7941 + }, + { + "epoch": 1.04, + "learning_rate": 1.947977334677432e-05, + "loss": 0.0907, + "step": 7942 + }, + { + "epoch": 1.04, + "learning_rate": 1.947960451091417e-05, + "loss": 0.0788, + "step": 7943 + }, + { + "epoch": 1.04, + "learning_rate": 1.9479435648393138e-05, + "loss": 0.0828, + "step": 7944 + }, + { + "epoch": 1.04, + "learning_rate": 1.9479266759211696e-05, + "loss": 0.0784, + "step": 7945 + }, + { + "epoch": 1.04, + "learning_rate": 1.947909784337032e-05, + "loss": 0.0813, + "step": 7946 + }, + { + "epoch": 1.04, + "learning_rate": 1.947892890086949e-05, + "loss": 0.0944, + "step": 7947 + }, + { + "epoch": 1.04, + "learning_rate": 1.9478759931709677e-05, + "loss": 0.0826, + "step": 7948 + }, + { + "epoch": 1.04, + "learning_rate": 1.9478590935891355e-05, + "loss": 0.083, + "step": 7949 + }, + { + "epoch": 1.04, + "learning_rate": 1.9478421913415e-05, + "loss": 0.0852, + "step": 7950 + }, + { + "epoch": 1.04, + "learning_rate": 1.9478252864281087e-05, + "loss": 0.0859, + "step": 7951 + }, + { + "epoch": 1.04, + "learning_rate": 1.94780837884901e-05, + "loss": 0.0818, + "step": 7952 + }, + { + "epoch": 1.04, + "learning_rate": 1.9477914686042498e-05, + "loss": 0.0807, + "step": 7953 + }, + { + "epoch": 1.04, + "learning_rate": 1.947774555693877e-05, + "loss": 0.0841, + "step": 7954 + }, + { + "epoch": 1.04, + "learning_rate": 1.9477576401179387e-05, + "loss": 0.0788, + "step": 7955 + }, + { + "epoch": 1.04, + "learning_rate": 1.9477407218764822e-05, + "loss": 0.0843, + "step": 7956 + }, + { + "epoch": 1.04, + "learning_rate": 1.947723800969556e-05, + "loss": 0.0846, + "step": 7957 + }, + { + "epoch": 1.04, + "learning_rate": 1.947706877397207e-05, + "loss": 0.0753, + "step": 7958 + }, + { + "epoch": 1.04, + "learning_rate": 1.9476899511594826e-05, + "loss": 0.087, + "step": 7959 + }, + { + "epoch": 1.04, + "learning_rate": 1.9476730222564305e-05, + "loss": 0.0916, + "step": 7960 + }, + { + "epoch": 1.04, + "learning_rate": 1.9476560906880988e-05, + "loss": 0.0809, + "step": 7961 + }, + { + "epoch": 1.04, + "learning_rate": 1.9476391564545346e-05, + "loss": 0.0818, + "step": 7962 + }, + { + "epoch": 1.04, + "learning_rate": 1.9476222195557856e-05, + "loss": 0.076, + "step": 7963 + }, + { + "epoch": 1.04, + "learning_rate": 1.9476052799919e-05, + "loss": 0.0847, + "step": 7964 + }, + { + "epoch": 1.04, + "learning_rate": 1.947588337762925e-05, + "loss": 0.0818, + "step": 7965 + }, + { + "epoch": 1.04, + "learning_rate": 1.947571392868908e-05, + "loss": 0.0839, + "step": 7966 + }, + { + "epoch": 1.04, + "learning_rate": 1.9475544453098968e-05, + "loss": 0.0823, + "step": 7967 + }, + { + "epoch": 1.04, + "learning_rate": 1.9475374950859395e-05, + "loss": 0.082, + "step": 7968 + }, + { + "epoch": 1.04, + "learning_rate": 1.947520542197083e-05, + "loss": 0.0827, + "step": 7969 + }, + { + "epoch": 1.04, + "learning_rate": 1.9475035866433764e-05, + "loss": 0.0673, + "step": 7970 + }, + { + "epoch": 1.04, + "learning_rate": 1.9474866284248657e-05, + "loss": 0.0745, + "step": 7971 + }, + { + "epoch": 1.04, + "learning_rate": 1.9474696675415992e-05, + "loss": 0.0885, + "step": 7972 + }, + { + "epoch": 1.04, + "learning_rate": 1.947452703993625e-05, + "loss": 0.0755, + "step": 7973 + }, + { + "epoch": 1.04, + "learning_rate": 1.9474357377809905e-05, + "loss": 0.088, + "step": 7974 + }, + { + "epoch": 1.04, + "learning_rate": 1.9474187689037435e-05, + "loss": 0.0952, + "step": 7975 + }, + { + "epoch": 1.04, + "learning_rate": 1.9474017973619314e-05, + "loss": 0.077, + "step": 7976 + }, + { + "epoch": 1.04, + "learning_rate": 1.9473848231556024e-05, + "loss": 0.0793, + "step": 7977 + }, + { + "epoch": 1.05, + "learning_rate": 1.947367846284804e-05, + "loss": 0.0778, + "step": 7978 + }, + { + "epoch": 1.05, + "learning_rate": 1.9473508667495838e-05, + "loss": 0.0825, + "step": 7979 + }, + { + "epoch": 1.05, + "learning_rate": 1.94733388454999e-05, + "loss": 0.0805, + "step": 7980 + }, + { + "epoch": 1.05, + "learning_rate": 1.94731689968607e-05, + "loss": 0.0902, + "step": 7981 + }, + { + "epoch": 1.05, + "learning_rate": 1.9472999121578717e-05, + "loss": 0.087, + "step": 7982 + }, + { + "epoch": 1.05, + "learning_rate": 1.947282921965443e-05, + "loss": 0.0846, + "step": 7983 + }, + { + "epoch": 1.05, + "learning_rate": 1.9472659291088312e-05, + "loss": 0.0926, + "step": 7984 + }, + { + "epoch": 1.05, + "learning_rate": 1.947248933588085e-05, + "loss": 0.0726, + "step": 7985 + }, + { + "epoch": 1.05, + "learning_rate": 1.947231935403251e-05, + "loss": 0.0823, + "step": 7986 + }, + { + "epoch": 1.05, + "learning_rate": 1.9472149345543777e-05, + "loss": 0.0885, + "step": 7987 + }, + { + "epoch": 1.05, + "learning_rate": 1.947197931041513e-05, + "loss": 0.0804, + "step": 7988 + }, + { + "epoch": 1.05, + "learning_rate": 1.9471809248647044e-05, + "loss": 0.0895, + "step": 7989 + }, + { + "epoch": 1.05, + "learning_rate": 1.947163916024e-05, + "loss": 0.0864, + "step": 7990 + }, + { + "epoch": 1.05, + "learning_rate": 1.9471469045194476e-05, + "loss": 0.0873, + "step": 7991 + }, + { + "epoch": 1.05, + "learning_rate": 1.9471298903510948e-05, + "loss": 0.0938, + "step": 7992 + }, + { + "epoch": 1.05, + "learning_rate": 1.9471128735189898e-05, + "loss": 0.0824, + "step": 7993 + }, + { + "epoch": 1.05, + "learning_rate": 1.9470958540231802e-05, + "loss": 0.0837, + "step": 7994 + }, + { + "epoch": 1.05, + "learning_rate": 1.947078831863714e-05, + "loss": 0.0746, + "step": 7995 + }, + { + "epoch": 1.05, + "learning_rate": 1.947061807040639e-05, + "loss": 0.0716, + "step": 7996 + }, + { + "epoch": 1.05, + "learning_rate": 1.947044779554003e-05, + "loss": 0.0807, + "step": 7997 + }, + { + "epoch": 1.05, + "learning_rate": 1.9470277494038542e-05, + "loss": 0.0791, + "step": 7998 + }, + { + "epoch": 1.05, + "learning_rate": 1.94701071659024e-05, + "loss": 0.0843, + "step": 7999 + }, + { + "epoch": 1.05, + "learning_rate": 1.9469936811132086e-05, + "loss": 0.0884, + "step": 8000 + }, + { + "epoch": 1.05, + "learning_rate": 1.9469766429728083e-05, + "loss": 0.0827, + "step": 8001 + }, + { + "epoch": 1.05, + "learning_rate": 1.9469596021690864e-05, + "loss": 0.0748, + "step": 8002 + }, + { + "epoch": 1.05, + "learning_rate": 1.946942558702091e-05, + "loss": 0.0724, + "step": 8003 + }, + { + "epoch": 1.05, + "learning_rate": 1.94692551257187e-05, + "loss": 0.0851, + "step": 8004 + }, + { + "epoch": 1.05, + "learning_rate": 1.9469084637784716e-05, + "loss": 0.0929, + "step": 8005 + }, + { + "epoch": 1.05, + "learning_rate": 1.9468914123219432e-05, + "loss": 0.0922, + "step": 8006 + }, + { + "epoch": 1.05, + "learning_rate": 1.9468743582023334e-05, + "loss": 0.0853, + "step": 8007 + }, + { + "epoch": 1.05, + "learning_rate": 1.94685730141969e-05, + "loss": 0.0808, + "step": 8008 + }, + { + "epoch": 1.05, + "learning_rate": 1.9468402419740608e-05, + "loss": 0.0779, + "step": 8009 + }, + { + "epoch": 1.05, + "learning_rate": 1.946823179865494e-05, + "loss": 0.0812, + "step": 8010 + }, + { + "epoch": 1.05, + "learning_rate": 1.946806115094037e-05, + "loss": 0.0857, + "step": 8011 + }, + { + "epoch": 1.05, + "learning_rate": 1.9467890476597385e-05, + "loss": 0.0798, + "step": 8012 + }, + { + "epoch": 1.05, + "learning_rate": 1.9467719775626462e-05, + "loss": 0.081, + "step": 8013 + }, + { + "epoch": 1.05, + "learning_rate": 1.9467549048028083e-05, + "loss": 0.0897, + "step": 8014 + }, + { + "epoch": 1.05, + "learning_rate": 1.9467378293802725e-05, + "loss": 0.0875, + "step": 8015 + }, + { + "epoch": 1.05, + "learning_rate": 1.946720751295087e-05, + "loss": 0.0907, + "step": 8016 + }, + { + "epoch": 1.05, + "learning_rate": 1.9467036705472993e-05, + "loss": 0.0819, + "step": 8017 + }, + { + "epoch": 1.05, + "learning_rate": 1.9466865871369585e-05, + "loss": 0.0779, + "step": 8018 + }, + { + "epoch": 1.05, + "learning_rate": 1.946669501064112e-05, + "loss": 0.0809, + "step": 8019 + }, + { + "epoch": 1.05, + "learning_rate": 1.946652412328808e-05, + "loss": 0.0741, + "step": 8020 + }, + { + "epoch": 1.05, + "learning_rate": 1.9466353209310944e-05, + "loss": 0.083, + "step": 8021 + }, + { + "epoch": 1.05, + "learning_rate": 1.9466182268710195e-05, + "loss": 0.0906, + "step": 8022 + }, + { + "epoch": 1.05, + "learning_rate": 1.946601130148631e-05, + "loss": 0.0796, + "step": 8023 + }, + { + "epoch": 1.05, + "learning_rate": 1.9465840307639773e-05, + "loss": 0.0955, + "step": 8024 + }, + { + "epoch": 1.05, + "learning_rate": 1.9465669287171065e-05, + "loss": 0.0874, + "step": 8025 + }, + { + "epoch": 1.05, + "learning_rate": 1.9465498240080664e-05, + "loss": 0.0779, + "step": 8026 + }, + { + "epoch": 1.05, + "learning_rate": 1.9465327166369057e-05, + "loss": 0.0877, + "step": 8027 + }, + { + "epoch": 1.05, + "learning_rate": 1.9465156066036722e-05, + "loss": 0.0747, + "step": 8028 + }, + { + "epoch": 1.05, + "learning_rate": 1.9464984939084135e-05, + "loss": 0.0869, + "step": 8029 + }, + { + "epoch": 1.05, + "learning_rate": 1.9464813785511785e-05, + "loss": 0.0716, + "step": 8030 + }, + { + "epoch": 1.05, + "learning_rate": 1.946464260532015e-05, + "loss": 0.0732, + "step": 8031 + }, + { + "epoch": 1.05, + "learning_rate": 1.946447139850971e-05, + "loss": 0.0836, + "step": 8032 + }, + { + "epoch": 1.05, + "learning_rate": 1.946430016508095e-05, + "loss": 0.0754, + "step": 8033 + }, + { + "epoch": 1.05, + "learning_rate": 1.9464128905034347e-05, + "loss": 0.072, + "step": 8034 + }, + { + "epoch": 1.05, + "learning_rate": 1.9463957618370387e-05, + "loss": 0.0777, + "step": 8035 + }, + { + "epoch": 1.05, + "learning_rate": 1.946378630508955e-05, + "loss": 0.0863, + "step": 8036 + }, + { + "epoch": 1.05, + "learning_rate": 1.946361496519232e-05, + "loss": 0.0849, + "step": 8037 + }, + { + "epoch": 1.05, + "learning_rate": 1.9463443598679175e-05, + "loss": 0.0883, + "step": 8038 + }, + { + "epoch": 1.05, + "learning_rate": 1.9463272205550604e-05, + "loss": 0.0928, + "step": 8039 + }, + { + "epoch": 1.05, + "learning_rate": 1.9463100785807078e-05, + "loss": 0.0799, + "step": 8040 + }, + { + "epoch": 1.05, + "learning_rate": 1.9462929339449086e-05, + "loss": 0.0818, + "step": 8041 + }, + { + "epoch": 1.05, + "learning_rate": 1.946275786647711e-05, + "loss": 0.0884, + "step": 8042 + }, + { + "epoch": 1.05, + "learning_rate": 1.946258636689163e-05, + "loss": 0.0706, + "step": 8043 + }, + { + "epoch": 1.05, + "learning_rate": 1.946241484069313e-05, + "loss": 0.0745, + "step": 8044 + }, + { + "epoch": 1.05, + "learning_rate": 1.9462243287882093e-05, + "loss": 0.0804, + "step": 8045 + }, + { + "epoch": 1.05, + "learning_rate": 1.9462071708459e-05, + "loss": 0.0802, + "step": 8046 + }, + { + "epoch": 1.05, + "learning_rate": 1.9461900102424336e-05, + "loss": 0.0766, + "step": 8047 + }, + { + "epoch": 1.05, + "learning_rate": 1.946172846977858e-05, + "loss": 0.0819, + "step": 8048 + }, + { + "epoch": 1.05, + "learning_rate": 1.946155681052222e-05, + "loss": 0.0808, + "step": 8049 + }, + { + "epoch": 1.05, + "learning_rate": 1.946138512465573e-05, + "loss": 0.0954, + "step": 8050 + }, + { + "epoch": 1.05, + "learning_rate": 1.94612134121796e-05, + "loss": 0.0919, + "step": 8051 + }, + { + "epoch": 1.05, + "learning_rate": 1.9461041673094312e-05, + "loss": 0.1018, + "step": 8052 + }, + { + "epoch": 1.05, + "learning_rate": 1.946086990740035e-05, + "loss": 0.081, + "step": 8053 + }, + { + "epoch": 1.06, + "learning_rate": 1.946069811509819e-05, + "loss": 0.0841, + "step": 8054 + }, + { + "epoch": 1.06, + "learning_rate": 1.946052629618832e-05, + "loss": 0.091, + "step": 8055 + }, + { + "epoch": 1.06, + "learning_rate": 1.946035445067123e-05, + "loss": 0.0879, + "step": 8056 + }, + { + "epoch": 1.06, + "learning_rate": 1.9460182578547394e-05, + "loss": 0.0849, + "step": 8057 + }, + { + "epoch": 1.06, + "learning_rate": 1.94600106798173e-05, + "loss": 0.0782, + "step": 8058 + }, + { + "epoch": 1.06, + "learning_rate": 1.9459838754481425e-05, + "loss": 0.0827, + "step": 8059 + }, + { + "epoch": 1.06, + "learning_rate": 1.945966680254026e-05, + "loss": 0.0711, + "step": 8060 + }, + { + "epoch": 1.06, + "learning_rate": 1.9459494823994287e-05, + "loss": 0.0723, + "step": 8061 + }, + { + "epoch": 1.06, + "learning_rate": 1.9459322818843988e-05, + "loss": 0.0872, + "step": 8062 + }, + { + "epoch": 1.06, + "learning_rate": 1.9459150787089847e-05, + "loss": 0.0813, + "step": 8063 + }, + { + "epoch": 1.06, + "learning_rate": 1.9458978728732346e-05, + "loss": 0.0794, + "step": 8064 + }, + { + "epoch": 1.06, + "learning_rate": 1.9458806643771976e-05, + "loss": 0.0775, + "step": 8065 + }, + { + "epoch": 1.06, + "learning_rate": 1.9458634532209212e-05, + "loss": 0.0799, + "step": 8066 + }, + { + "epoch": 1.06, + "learning_rate": 1.9458462394044544e-05, + "loss": 0.0891, + "step": 8067 + }, + { + "epoch": 1.06, + "learning_rate": 1.9458290229278455e-05, + "loss": 0.0808, + "step": 8068 + }, + { + "epoch": 1.06, + "learning_rate": 1.9458118037911428e-05, + "loss": 0.0755, + "step": 8069 + }, + { + "epoch": 1.06, + "learning_rate": 1.9457945819943944e-05, + "loss": 0.0852, + "step": 8070 + }, + { + "epoch": 1.06, + "learning_rate": 1.9457773575376496e-05, + "loss": 0.086, + "step": 8071 + }, + { + "epoch": 1.06, + "learning_rate": 1.945760130420956e-05, + "loss": 0.0868, + "step": 8072 + }, + { + "epoch": 1.06, + "learning_rate": 1.9457429006443624e-05, + "loss": 0.08, + "step": 8073 + }, + { + "epoch": 1.06, + "learning_rate": 1.9457256682079173e-05, + "loss": 0.0741, + "step": 8074 + }, + { + "epoch": 1.06, + "learning_rate": 1.9457084331116692e-05, + "loss": 0.082, + "step": 8075 + }, + { + "epoch": 1.06, + "learning_rate": 1.9456911953556665e-05, + "loss": 0.0857, + "step": 8076 + }, + { + "epoch": 1.06, + "learning_rate": 1.9456739549399576e-05, + "loss": 0.0863, + "step": 8077 + }, + { + "epoch": 1.06, + "learning_rate": 1.945656711864591e-05, + "loss": 0.0799, + "step": 8078 + }, + { + "epoch": 1.06, + "learning_rate": 1.9456394661296155e-05, + "loss": 0.0807, + "step": 8079 + }, + { + "epoch": 1.06, + "learning_rate": 1.945622217735079e-05, + "loss": 0.0913, + "step": 8080 + }, + { + "epoch": 1.06, + "learning_rate": 1.9456049666810308e-05, + "loss": 0.0777, + "step": 8081 + }, + { + "epoch": 1.06, + "learning_rate": 1.9455877129675186e-05, + "loss": 0.075, + "step": 8082 + }, + { + "epoch": 1.06, + "learning_rate": 1.9455704565945916e-05, + "loss": 0.0918, + "step": 8083 + }, + { + "epoch": 1.06, + "learning_rate": 1.9455531975622983e-05, + "loss": 0.0857, + "step": 8084 + }, + { + "epoch": 1.06, + "learning_rate": 1.9455359358706864e-05, + "loss": 0.0771, + "step": 8085 + }, + { + "epoch": 1.06, + "learning_rate": 1.945518671519805e-05, + "loss": 0.0933, + "step": 8086 + }, + { + "epoch": 1.06, + "learning_rate": 1.945501404509703e-05, + "loss": 0.0841, + "step": 8087 + }, + { + "epoch": 1.06, + "learning_rate": 1.9454841348404286e-05, + "loss": 0.0753, + "step": 8088 + }, + { + "epoch": 1.06, + "learning_rate": 1.9454668625120304e-05, + "loss": 0.0838, + "step": 8089 + }, + { + "epoch": 1.06, + "learning_rate": 1.9454495875245573e-05, + "loss": 0.0825, + "step": 8090 + }, + { + "epoch": 1.06, + "learning_rate": 1.9454323098780573e-05, + "loss": 0.0915, + "step": 8091 + }, + { + "epoch": 1.06, + "learning_rate": 1.9454150295725793e-05, + "loss": 0.0795, + "step": 8092 + }, + { + "epoch": 1.06, + "learning_rate": 1.9453977466081716e-05, + "loss": 0.0835, + "step": 8093 + }, + { + "epoch": 1.06, + "learning_rate": 1.9453804609848836e-05, + "loss": 0.0773, + "step": 8094 + }, + { + "epoch": 1.06, + "learning_rate": 1.9453631727027628e-05, + "loss": 0.074, + "step": 8095 + }, + { + "epoch": 1.06, + "learning_rate": 1.9453458817618587e-05, + "loss": 0.0793, + "step": 8096 + }, + { + "epoch": 1.06, + "learning_rate": 1.94532858816222e-05, + "loss": 0.0725, + "step": 8097 + }, + { + "epoch": 1.06, + "learning_rate": 1.945311291903895e-05, + "loss": 0.0771, + "step": 8098 + }, + { + "epoch": 1.06, + "learning_rate": 1.9452939929869315e-05, + "loss": 0.0785, + "step": 8099 + }, + { + "epoch": 1.06, + "learning_rate": 1.9452766914113798e-05, + "loss": 0.0769, + "step": 8100 + }, + { + "epoch": 1.06, + "learning_rate": 1.9452593871772874e-05, + "loss": 0.0885, + "step": 8101 + }, + { + "epoch": 1.06, + "learning_rate": 1.9452420802847035e-05, + "loss": 0.0938, + "step": 8102 + }, + { + "epoch": 1.06, + "learning_rate": 1.9452247707336762e-05, + "loss": 0.0857, + "step": 8103 + }, + { + "epoch": 1.06, + "learning_rate": 1.945207458524255e-05, + "loss": 0.0774, + "step": 8104 + }, + { + "epoch": 1.06, + "learning_rate": 1.945190143656488e-05, + "loss": 0.0916, + "step": 8105 + }, + { + "epoch": 1.06, + "learning_rate": 1.945172826130424e-05, + "loss": 0.0843, + "step": 8106 + }, + { + "epoch": 1.06, + "learning_rate": 1.9451555059461115e-05, + "loss": 0.0751, + "step": 8107 + }, + { + "epoch": 1.06, + "learning_rate": 1.9451381831036e-05, + "loss": 0.0856, + "step": 8108 + }, + { + "epoch": 1.06, + "learning_rate": 1.9451208576029373e-05, + "loss": 0.0808, + "step": 8109 + }, + { + "epoch": 1.06, + "learning_rate": 1.9451035294441727e-05, + "loss": 0.0811, + "step": 8110 + }, + { + "epoch": 1.06, + "learning_rate": 1.945086198627355e-05, + "loss": 0.0858, + "step": 8111 + }, + { + "epoch": 1.06, + "learning_rate": 1.9450688651525323e-05, + "loss": 0.0841, + "step": 8112 + }, + { + "epoch": 1.06, + "learning_rate": 1.9450515290197537e-05, + "loss": 0.0842, + "step": 8113 + }, + { + "epoch": 1.06, + "learning_rate": 1.9450341902290684e-05, + "loss": 0.0986, + "step": 8114 + }, + { + "epoch": 1.06, + "learning_rate": 1.9450168487805246e-05, + "loss": 0.0926, + "step": 8115 + }, + { + "epoch": 1.06, + "learning_rate": 1.944999504674171e-05, + "loss": 0.0783, + "step": 8116 + }, + { + "epoch": 1.06, + "learning_rate": 1.944982157910057e-05, + "loss": 0.0665, + "step": 8117 + }, + { + "epoch": 1.06, + "learning_rate": 1.9449648084882306e-05, + "loss": 0.0845, + "step": 8118 + }, + { + "epoch": 1.06, + "learning_rate": 1.944947456408741e-05, + "loss": 0.0807, + "step": 8119 + }, + { + "epoch": 1.06, + "learning_rate": 1.9449301016716374e-05, + "loss": 0.0901, + "step": 8120 + }, + { + "epoch": 1.06, + "learning_rate": 1.944912744276968e-05, + "loss": 0.0846, + "step": 8121 + }, + { + "epoch": 1.06, + "learning_rate": 1.944895384224782e-05, + "loss": 0.0775, + "step": 8122 + }, + { + "epoch": 1.06, + "learning_rate": 1.9448780215151276e-05, + "loss": 0.0863, + "step": 8123 + }, + { + "epoch": 1.06, + "learning_rate": 1.9448606561480545e-05, + "loss": 0.0845, + "step": 8124 + }, + { + "epoch": 1.06, + "learning_rate": 1.944843288123611e-05, + "loss": 0.0734, + "step": 8125 + }, + { + "epoch": 1.06, + "learning_rate": 1.944825917441846e-05, + "loss": 0.0824, + "step": 8126 + }, + { + "epoch": 1.06, + "learning_rate": 1.9448085441028084e-05, + "loss": 0.0877, + "step": 8127 + }, + { + "epoch": 1.06, + "learning_rate": 1.944791168106547e-05, + "loss": 0.0783, + "step": 8128 + }, + { + "epoch": 1.06, + "learning_rate": 1.944773789453111e-05, + "loss": 0.0751, + "step": 8129 + }, + { + "epoch": 1.06, + "learning_rate": 1.9447564081425492e-05, + "loss": 0.0763, + "step": 8130 + }, + { + "epoch": 1.07, + "learning_rate": 1.94473902417491e-05, + "loss": 0.0717, + "step": 8131 + }, + { + "epoch": 1.07, + "learning_rate": 1.9447216375502424e-05, + "loss": 0.0836, + "step": 8132 + }, + { + "epoch": 1.07, + "learning_rate": 1.9447042482685957e-05, + "loss": 0.0902, + "step": 8133 + }, + { + "epoch": 1.07, + "learning_rate": 1.944686856330019e-05, + "loss": 0.0867, + "step": 8134 + }, + { + "epoch": 1.07, + "learning_rate": 1.9446694617345603e-05, + "loss": 0.0874, + "step": 8135 + }, + { + "epoch": 1.07, + "learning_rate": 1.9446520644822692e-05, + "loss": 0.0822, + "step": 8136 + }, + { + "epoch": 1.07, + "learning_rate": 1.9446346645731945e-05, + "loss": 0.0869, + "step": 8137 + }, + { + "epoch": 1.07, + "learning_rate": 1.9446172620073854e-05, + "loss": 0.0769, + "step": 8138 + }, + { + "epoch": 1.07, + "learning_rate": 1.94459985678489e-05, + "loss": 0.0832, + "step": 8139 + }, + { + "epoch": 1.07, + "learning_rate": 1.944582448905758e-05, + "loss": 0.0776, + "step": 8140 + }, + { + "epoch": 1.07, + "learning_rate": 1.9445650383700385e-05, + "loss": 0.0791, + "step": 8141 + }, + { + "epoch": 1.07, + "learning_rate": 1.94454762517778e-05, + "loss": 0.0847, + "step": 8142 + }, + { + "epoch": 1.07, + "learning_rate": 1.9445302093290315e-05, + "loss": 0.0783, + "step": 8143 + }, + { + "epoch": 1.07, + "learning_rate": 1.9445127908238418e-05, + "loss": 0.0833, + "step": 8144 + }, + { + "epoch": 1.07, + "learning_rate": 1.9444953696622606e-05, + "loss": 0.087, + "step": 8145 + }, + { + "epoch": 1.07, + "learning_rate": 1.9444779458443362e-05, + "loss": 0.0963, + "step": 8146 + }, + { + "epoch": 1.07, + "learning_rate": 1.944460519370118e-05, + "loss": 0.0789, + "step": 8147 + }, + { + "epoch": 1.07, + "learning_rate": 1.944443090239655e-05, + "loss": 0.0869, + "step": 8148 + }, + { + "epoch": 1.07, + "learning_rate": 1.944425658452996e-05, + "loss": 0.0751, + "step": 8149 + }, + { + "epoch": 1.07, + "learning_rate": 1.94440822401019e-05, + "loss": 0.0845, + "step": 8150 + }, + { + "epoch": 1.07, + "learning_rate": 1.9443907869112865e-05, + "loss": 0.091, + "step": 8151 + }, + { + "epoch": 1.07, + "learning_rate": 1.944373347156334e-05, + "loss": 0.0877, + "step": 8152 + }, + { + "epoch": 1.07, + "learning_rate": 1.9443559047453814e-05, + "loss": 0.0819, + "step": 8153 + }, + { + "epoch": 1.07, + "learning_rate": 1.944338459678479e-05, + "loss": 0.0889, + "step": 8154 + }, + { + "epoch": 1.07, + "learning_rate": 1.944321011955674e-05, + "loss": 0.0821, + "step": 8155 + }, + { + "epoch": 1.07, + "learning_rate": 1.9443035615770167e-05, + "loss": 0.0824, + "step": 8156 + }, + { + "epoch": 1.07, + "learning_rate": 1.9442861085425563e-05, + "loss": 0.0982, + "step": 8157 + }, + { + "epoch": 1.07, + "learning_rate": 1.944268652852341e-05, + "loss": 0.0784, + "step": 8158 + }, + { + "epoch": 1.07, + "learning_rate": 1.944251194506421e-05, + "loss": 0.0781, + "step": 8159 + }, + { + "epoch": 1.07, + "learning_rate": 1.9442337335048442e-05, + "loss": 0.0905, + "step": 8160 + }, + { + "epoch": 1.07, + "learning_rate": 1.9442162698476607e-05, + "loss": 0.0811, + "step": 8161 + }, + { + "epoch": 1.07, + "learning_rate": 1.944198803534919e-05, + "loss": 0.096, + "step": 8162 + }, + { + "epoch": 1.07, + "learning_rate": 1.9441813345666688e-05, + "loss": 0.083, + "step": 8163 + }, + { + "epoch": 1.07, + "learning_rate": 1.9441638629429584e-05, + "loss": 0.0802, + "step": 8164 + }, + { + "epoch": 1.07, + "learning_rate": 1.9441463886638377e-05, + "loss": 0.0893, + "step": 8165 + }, + { + "epoch": 1.07, + "learning_rate": 1.9441289117293554e-05, + "loss": 0.0786, + "step": 8166 + }, + { + "epoch": 1.07, + "learning_rate": 1.944111432139561e-05, + "loss": 0.0737, + "step": 8167 + }, + { + "epoch": 1.07, + "learning_rate": 1.9440939498945034e-05, + "loss": 0.072, + "step": 8168 + }, + { + "epoch": 1.07, + "learning_rate": 1.9440764649942316e-05, + "loss": 0.0893, + "step": 8169 + }, + { + "epoch": 1.07, + "learning_rate": 1.9440589774387954e-05, + "loss": 0.0865, + "step": 8170 + }, + { + "epoch": 1.07, + "learning_rate": 1.9440414872282436e-05, + "loss": 0.0903, + "step": 8171 + }, + { + "epoch": 1.07, + "learning_rate": 1.944023994362625e-05, + "loss": 0.0879, + "step": 8172 + }, + { + "epoch": 1.07, + "learning_rate": 1.9440064988419895e-05, + "loss": 0.0762, + "step": 8173 + }, + { + "epoch": 1.07, + "learning_rate": 1.9439890006663858e-05, + "loss": 0.0745, + "step": 8174 + }, + { + "epoch": 1.07, + "learning_rate": 1.9439714998358637e-05, + "loss": 0.0877, + "step": 8175 + }, + { + "epoch": 1.07, + "learning_rate": 1.9439539963504717e-05, + "loss": 0.0959, + "step": 8176 + }, + { + "epoch": 1.07, + "learning_rate": 1.9439364902102595e-05, + "loss": 0.0853, + "step": 8177 + }, + { + "epoch": 1.07, + "learning_rate": 1.943918981415276e-05, + "loss": 0.0907, + "step": 8178 + }, + { + "epoch": 1.07, + "learning_rate": 1.9439014699655706e-05, + "loss": 0.0788, + "step": 8179 + }, + { + "epoch": 1.07, + "learning_rate": 1.9438839558611928e-05, + "loss": 0.0918, + "step": 8180 + }, + { + "epoch": 1.07, + "learning_rate": 1.9438664391021914e-05, + "loss": 0.0786, + "step": 8181 + }, + { + "epoch": 1.07, + "learning_rate": 1.943848919688616e-05, + "loss": 0.0796, + "step": 8182 + }, + { + "epoch": 1.07, + "learning_rate": 1.9438313976205162e-05, + "loss": 0.0855, + "step": 8183 + }, + { + "epoch": 1.07, + "learning_rate": 1.9438138728979404e-05, + "loss": 0.0879, + "step": 8184 + }, + { + "epoch": 1.07, + "learning_rate": 1.943796345520938e-05, + "loss": 0.0723, + "step": 8185 + }, + { + "epoch": 1.07, + "learning_rate": 1.9437788154895593e-05, + "loss": 0.0864, + "step": 8186 + }, + { + "epoch": 1.07, + "learning_rate": 1.943761282803853e-05, + "loss": 0.0666, + "step": 8187 + }, + { + "epoch": 1.07, + "learning_rate": 1.943743747463868e-05, + "loss": 0.0881, + "step": 8188 + }, + { + "epoch": 1.07, + "learning_rate": 1.943726209469654e-05, + "loss": 0.0872, + "step": 8189 + }, + { + "epoch": 1.07, + "learning_rate": 1.94370866882126e-05, + "loss": 0.0842, + "step": 8190 + }, + { + "epoch": 1.07, + "learning_rate": 1.943691125518736e-05, + "loss": 0.0759, + "step": 8191 + }, + { + "epoch": 1.07, + "learning_rate": 1.943673579562131e-05, + "loss": 0.0742, + "step": 8192 + }, + { + "epoch": 1.07, + "learning_rate": 1.9436560309514943e-05, + "loss": 0.0821, + "step": 8193 + }, + { + "epoch": 1.07, + "learning_rate": 1.943638479686875e-05, + "loss": 0.0974, + "step": 8194 + }, + { + "epoch": 1.07, + "learning_rate": 1.943620925768323e-05, + "loss": 0.088, + "step": 8195 + }, + { + "epoch": 1.07, + "learning_rate": 1.9436033691958872e-05, + "loss": 0.0874, + "step": 8196 + }, + { + "epoch": 1.07, + "learning_rate": 1.9435858099696174e-05, + "loss": 0.0816, + "step": 8197 + }, + { + "epoch": 1.07, + "learning_rate": 1.943568248089563e-05, + "loss": 0.0722, + "step": 8198 + }, + { + "epoch": 1.07, + "learning_rate": 1.9435506835557725e-05, + "loss": 0.0814, + "step": 8199 + }, + { + "epoch": 1.07, + "learning_rate": 1.9435331163682962e-05, + "loss": 0.0944, + "step": 8200 + }, + { + "epoch": 1.07, + "learning_rate": 1.9435155465271834e-05, + "loss": 0.092, + "step": 8201 + }, + { + "epoch": 1.07, + "learning_rate": 1.943497974032483e-05, + "loss": 0.0713, + "step": 8202 + }, + { + "epoch": 1.07, + "learning_rate": 1.943480398884245e-05, + "loss": 0.0878, + "step": 8203 + }, + { + "epoch": 1.07, + "learning_rate": 1.9434628210825188e-05, + "loss": 0.0934, + "step": 8204 + }, + { + "epoch": 1.07, + "learning_rate": 1.9434452406273536e-05, + "loss": 0.0786, + "step": 8205 + }, + { + "epoch": 1.07, + "learning_rate": 1.9434276575187988e-05, + "loss": 0.0765, + "step": 8206 + }, + { + "epoch": 1.08, + "learning_rate": 1.943410071756904e-05, + "loss": 0.0667, + "step": 8207 + }, + { + "epoch": 1.08, + "learning_rate": 1.9433924833417187e-05, + "loss": 0.0829, + "step": 8208 + }, + { + "epoch": 1.08, + "learning_rate": 1.9433748922732923e-05, + "loss": 0.0759, + "step": 8209 + }, + { + "epoch": 1.08, + "learning_rate": 1.943357298551674e-05, + "loss": 0.0813, + "step": 8210 + }, + { + "epoch": 1.08, + "learning_rate": 1.9433397021769138e-05, + "loss": 0.0857, + "step": 8211 + }, + { + "epoch": 1.08, + "learning_rate": 1.9433221031490608e-05, + "loss": 0.0848, + "step": 8212 + }, + { + "epoch": 1.08, + "learning_rate": 1.9433045014681644e-05, + "loss": 0.0875, + "step": 8213 + }, + { + "epoch": 1.08, + "learning_rate": 1.9432868971342748e-05, + "loss": 0.0753, + "step": 8214 + }, + { + "epoch": 1.08, + "learning_rate": 1.9432692901474407e-05, + "loss": 0.088, + "step": 8215 + }, + { + "epoch": 1.08, + "learning_rate": 1.9432516805077116e-05, + "loss": 0.0816, + "step": 8216 + }, + { + "epoch": 1.08, + "learning_rate": 1.943234068215138e-05, + "loss": 0.0891, + "step": 8217 + }, + { + "epoch": 1.08, + "learning_rate": 1.9432164532697687e-05, + "loss": 0.0784, + "step": 8218 + }, + { + "epoch": 1.08, + "learning_rate": 1.943198835671653e-05, + "loss": 0.0816, + "step": 8219 + }, + { + "epoch": 1.08, + "learning_rate": 1.9431812154208415e-05, + "loss": 0.0866, + "step": 8220 + }, + { + "epoch": 1.08, + "learning_rate": 1.9431635925173825e-05, + "loss": 0.0827, + "step": 8221 + }, + { + "epoch": 1.08, + "learning_rate": 1.9431459669613263e-05, + "loss": 0.086, + "step": 8222 + }, + { + "epoch": 1.08, + "learning_rate": 1.943128338752722e-05, + "loss": 0.0907, + "step": 8223 + }, + { + "epoch": 1.08, + "learning_rate": 1.94311070789162e-05, + "loss": 0.0774, + "step": 8224 + }, + { + "epoch": 1.08, + "learning_rate": 1.943093074378069e-05, + "loss": 0.0799, + "step": 8225 + }, + { + "epoch": 1.08, + "learning_rate": 1.9430754382121192e-05, + "loss": 0.075, + "step": 8226 + }, + { + "epoch": 1.08, + "learning_rate": 1.9430577993938198e-05, + "loss": 0.0668, + "step": 8227 + }, + { + "epoch": 1.08, + "learning_rate": 1.9430401579232204e-05, + "loss": 0.0802, + "step": 8228 + }, + { + "epoch": 1.08, + "learning_rate": 1.9430225138003708e-05, + "loss": 0.0978, + "step": 8229 + }, + { + "epoch": 1.08, + "learning_rate": 1.9430048670253207e-05, + "loss": 0.0898, + "step": 8230 + }, + { + "epoch": 1.08, + "learning_rate": 1.9429872175981197e-05, + "loss": 0.0789, + "step": 8231 + }, + { + "epoch": 1.08, + "learning_rate": 1.9429695655188175e-05, + "loss": 0.0941, + "step": 8232 + }, + { + "epoch": 1.08, + "learning_rate": 1.9429519107874633e-05, + "loss": 0.0857, + "step": 8233 + }, + { + "epoch": 1.08, + "learning_rate": 1.9429342534041072e-05, + "loss": 0.0749, + "step": 8234 + }, + { + "epoch": 1.08, + "learning_rate": 1.9429165933687987e-05, + "loss": 0.0753, + "step": 8235 + }, + { + "epoch": 1.08, + "learning_rate": 1.9428989306815875e-05, + "loss": 0.0962, + "step": 8236 + }, + { + "epoch": 1.08, + "learning_rate": 1.9428812653425233e-05, + "loss": 0.0878, + "step": 8237 + }, + { + "epoch": 1.08, + "learning_rate": 1.942863597351656e-05, + "loss": 0.0872, + "step": 8238 + }, + { + "epoch": 1.08, + "learning_rate": 1.9428459267090346e-05, + "loss": 0.0897, + "step": 8239 + }, + { + "epoch": 1.08, + "learning_rate": 1.9428282534147094e-05, + "loss": 0.0917, + "step": 8240 + }, + { + "epoch": 1.08, + "learning_rate": 1.94281057746873e-05, + "loss": 0.0768, + "step": 8241 + }, + { + "epoch": 1.08, + "learning_rate": 1.942792898871146e-05, + "loss": 0.0818, + "step": 8242 + }, + { + "epoch": 1.08, + "learning_rate": 1.942775217622007e-05, + "loss": 0.0824, + "step": 8243 + }, + { + "epoch": 1.08, + "learning_rate": 1.942757533721363e-05, + "loss": 0.0794, + "step": 8244 + }, + { + "epoch": 1.08, + "learning_rate": 1.942739847169264e-05, + "loss": 0.0931, + "step": 8245 + }, + { + "epoch": 1.08, + "learning_rate": 1.9427221579657592e-05, + "loss": 0.0851, + "step": 8246 + }, + { + "epoch": 1.08, + "learning_rate": 1.9427044661108984e-05, + "loss": 0.0835, + "step": 8247 + }, + { + "epoch": 1.08, + "learning_rate": 1.9426867716047317e-05, + "loss": 0.0794, + "step": 8248 + }, + { + "epoch": 1.08, + "learning_rate": 1.9426690744473084e-05, + "loss": 0.081, + "step": 8249 + }, + { + "epoch": 1.08, + "learning_rate": 1.9426513746386785e-05, + "loss": 0.0867, + "step": 8250 + }, + { + "epoch": 1.08, + "learning_rate": 1.942633672178892e-05, + "loss": 0.0868, + "step": 8251 + }, + { + "epoch": 1.08, + "learning_rate": 1.9426159670679985e-05, + "loss": 0.0798, + "step": 8252 + }, + { + "epoch": 1.08, + "learning_rate": 1.9425982593060476e-05, + "loss": 0.0844, + "step": 8253 + }, + { + "epoch": 1.08, + "learning_rate": 1.9425805488930897e-05, + "loss": 0.082, + "step": 8254 + }, + { + "epoch": 1.08, + "learning_rate": 1.9425628358291737e-05, + "loss": 0.0903, + "step": 8255 + }, + { + "epoch": 1.08, + "learning_rate": 1.9425451201143503e-05, + "loss": 0.0738, + "step": 8256 + }, + { + "epoch": 1.08, + "learning_rate": 1.9425274017486688e-05, + "loss": 0.0831, + "step": 8257 + }, + { + "epoch": 1.08, + "learning_rate": 1.942509680732179e-05, + "loss": 0.0826, + "step": 8258 + }, + { + "epoch": 1.08, + "learning_rate": 1.942491957064931e-05, + "loss": 0.0877, + "step": 8259 + }, + { + "epoch": 1.08, + "learning_rate": 1.9424742307469747e-05, + "loss": 0.0865, + "step": 8260 + }, + { + "epoch": 1.08, + "learning_rate": 1.94245650177836e-05, + "loss": 0.0806, + "step": 8261 + }, + { + "epoch": 1.08, + "learning_rate": 1.9424387701591364e-05, + "loss": 0.0818, + "step": 8262 + }, + { + "epoch": 1.08, + "learning_rate": 1.942421035889354e-05, + "loss": 0.0793, + "step": 8263 + }, + { + "epoch": 1.08, + "learning_rate": 1.9424032989690622e-05, + "loss": 0.0795, + "step": 8264 + }, + { + "epoch": 1.08, + "learning_rate": 1.9423855593983117e-05, + "loss": 0.08, + "step": 8265 + }, + { + "epoch": 1.08, + "learning_rate": 1.9423678171771517e-05, + "loss": 0.0783, + "step": 8266 + }, + { + "epoch": 1.08, + "learning_rate": 1.9423500723056327e-05, + "loss": 0.0871, + "step": 8267 + }, + { + "epoch": 1.08, + "learning_rate": 1.9423323247838043e-05, + "loss": 0.0874, + "step": 8268 + }, + { + "epoch": 1.08, + "learning_rate": 1.942314574611716e-05, + "loss": 0.0837, + "step": 8269 + }, + { + "epoch": 1.08, + "learning_rate": 1.9422968217894187e-05, + "loss": 0.0933, + "step": 8270 + }, + { + "epoch": 1.08, + "learning_rate": 1.9422790663169615e-05, + "loss": 0.0814, + "step": 8271 + }, + { + "epoch": 1.08, + "learning_rate": 1.9422613081943948e-05, + "loss": 0.0885, + "step": 8272 + }, + { + "epoch": 1.08, + "learning_rate": 1.9422435474217682e-05, + "loss": 0.0799, + "step": 8273 + }, + { + "epoch": 1.08, + "learning_rate": 1.9422257839991316e-05, + "loss": 0.0755, + "step": 8274 + }, + { + "epoch": 1.08, + "learning_rate": 1.9422080179265355e-05, + "loss": 0.0809, + "step": 8275 + }, + { + "epoch": 1.08, + "learning_rate": 1.9421902492040294e-05, + "loss": 0.0774, + "step": 8276 + }, + { + "epoch": 1.08, + "learning_rate": 1.9421724778316637e-05, + "loss": 0.0953, + "step": 8277 + }, + { + "epoch": 1.08, + "learning_rate": 1.9421547038094875e-05, + "loss": 0.0938, + "step": 8278 + }, + { + "epoch": 1.08, + "learning_rate": 1.9421369271375518e-05, + "loss": 0.0804, + "step": 8279 + }, + { + "epoch": 1.08, + "learning_rate": 1.942119147815906e-05, + "loss": 0.0898, + "step": 8280 + }, + { + "epoch": 1.08, + "learning_rate": 1.9421013658446003e-05, + "loss": 0.0837, + "step": 8281 + }, + { + "epoch": 1.08, + "learning_rate": 1.9420835812236842e-05, + "loss": 0.0817, + "step": 8282 + }, + { + "epoch": 1.09, + "learning_rate": 1.9420657939532087e-05, + "loss": 0.0852, + "step": 8283 + }, + { + "epoch": 1.09, + "learning_rate": 1.9420480040332236e-05, + "loss": 0.0777, + "step": 8284 + }, + { + "epoch": 1.09, + "learning_rate": 1.942030211463778e-05, + "loss": 0.0791, + "step": 8285 + }, + { + "epoch": 1.09, + "learning_rate": 1.942012416244923e-05, + "loss": 0.0779, + "step": 8286 + }, + { + "epoch": 1.09, + "learning_rate": 1.9419946183767084e-05, + "loss": 0.073, + "step": 8287 + }, + { + "epoch": 1.09, + "learning_rate": 1.9419768178591834e-05, + "loss": 0.0836, + "step": 8288 + }, + { + "epoch": 1.09, + "learning_rate": 1.9419590146923996e-05, + "loss": 0.0771, + "step": 8289 + }, + { + "epoch": 1.09, + "learning_rate": 1.9419412088764057e-05, + "loss": 0.0841, + "step": 8290 + }, + { + "epoch": 1.09, + "learning_rate": 1.9419234004112523e-05, + "loss": 0.0929, + "step": 8291 + }, + { + "epoch": 1.09, + "learning_rate": 1.9419055892969897e-05, + "loss": 0.0746, + "step": 8292 + }, + { + "epoch": 1.09, + "learning_rate": 1.941887775533668e-05, + "loss": 0.0808, + "step": 8293 + }, + { + "epoch": 1.09, + "learning_rate": 1.941869959121337e-05, + "loss": 0.0856, + "step": 8294 + }, + { + "epoch": 1.09, + "learning_rate": 1.9418521400600463e-05, + "loss": 0.0902, + "step": 8295 + }, + { + "epoch": 1.09, + "learning_rate": 1.941834318349847e-05, + "loss": 0.0871, + "step": 8296 + }, + { + "epoch": 1.09, + "learning_rate": 1.9418164939907892e-05, + "loss": 0.0778, + "step": 8297 + }, + { + "epoch": 1.09, + "learning_rate": 1.9417986669829224e-05, + "loss": 0.084, + "step": 8298 + }, + { + "epoch": 1.09, + "learning_rate": 1.941780837326297e-05, + "loss": 0.0822, + "step": 8299 + }, + { + "epoch": 1.09, + "learning_rate": 1.9417630050209627e-05, + "loss": 0.0758, + "step": 8300 + }, + { + "epoch": 1.09, + "learning_rate": 1.9417451700669707e-05, + "loss": 0.0796, + "step": 8301 + }, + { + "epoch": 1.09, + "learning_rate": 1.9417273324643704e-05, + "loss": 0.0964, + "step": 8302 + }, + { + "epoch": 1.09, + "learning_rate": 1.941709492213212e-05, + "loss": 0.0862, + "step": 8303 + }, + { + "epoch": 1.09, + "learning_rate": 1.9416916493135457e-05, + "loss": 0.092, + "step": 8304 + }, + { + "epoch": 1.09, + "learning_rate": 1.941673803765422e-05, + "loss": 0.0969, + "step": 8305 + }, + { + "epoch": 1.09, + "learning_rate": 1.9416559555688912e-05, + "loss": 0.086, + "step": 8306 + }, + { + "epoch": 1.09, + "learning_rate": 1.9416381047240028e-05, + "loss": 0.0849, + "step": 8307 + }, + { + "epoch": 1.09, + "learning_rate": 1.9416202512308072e-05, + "loss": 0.0717, + "step": 8308 + }, + { + "epoch": 1.09, + "learning_rate": 1.941602395089355e-05, + "loss": 0.0824, + "step": 8309 + }, + { + "epoch": 1.09, + "learning_rate": 1.941584536299696e-05, + "loss": 0.0739, + "step": 8310 + }, + { + "epoch": 1.09, + "learning_rate": 1.941566674861881e-05, + "loss": 0.0886, + "step": 8311 + }, + { + "epoch": 1.09, + "learning_rate": 1.9415488107759597e-05, + "loss": 0.0814, + "step": 8312 + }, + { + "epoch": 1.09, + "learning_rate": 1.9415309440419824e-05, + "loss": 0.0921, + "step": 8313 + }, + { + "epoch": 1.09, + "learning_rate": 1.94151307466e-05, + "loss": 0.0909, + "step": 8314 + }, + { + "epoch": 1.09, + "learning_rate": 1.9414952026300615e-05, + "loss": 0.078, + "step": 8315 + }, + { + "epoch": 1.09, + "learning_rate": 1.9414773279522182e-05, + "loss": 0.077, + "step": 8316 + }, + { + "epoch": 1.09, + "learning_rate": 1.9414594506265198e-05, + "loss": 0.0777, + "step": 8317 + }, + { + "epoch": 1.09, + "learning_rate": 1.941441570653017e-05, + "loss": 0.0757, + "step": 8318 + }, + { + "epoch": 1.09, + "learning_rate": 1.94142368803176e-05, + "loss": 0.0823, + "step": 8319 + }, + { + "epoch": 1.09, + "learning_rate": 1.941405802762799e-05, + "loss": 0.0726, + "step": 8320 + }, + { + "epoch": 1.09, + "learning_rate": 1.9413879148461846e-05, + "loss": 0.0859, + "step": 8321 + }, + { + "epoch": 1.09, + "learning_rate": 1.9413700242819664e-05, + "loss": 0.0924, + "step": 8322 + }, + { + "epoch": 1.09, + "learning_rate": 1.941352131070195e-05, + "loss": 0.08, + "step": 8323 + }, + { + "epoch": 1.09, + "learning_rate": 1.9413342352109212e-05, + "loss": 0.0863, + "step": 8324 + }, + { + "epoch": 1.09, + "learning_rate": 1.941316336704195e-05, + "loss": 0.0869, + "step": 8325 + }, + { + "epoch": 1.09, + "learning_rate": 1.9412984355500665e-05, + "loss": 0.0846, + "step": 8326 + }, + { + "epoch": 1.09, + "learning_rate": 1.9412805317485867e-05, + "loss": 0.0833, + "step": 8327 + }, + { + "epoch": 1.09, + "learning_rate": 1.941262625299805e-05, + "loss": 0.0866, + "step": 8328 + }, + { + "epoch": 1.09, + "learning_rate": 1.9412447162037727e-05, + "loss": 0.087, + "step": 8329 + }, + { + "epoch": 1.09, + "learning_rate": 1.9412268044605395e-05, + "loss": 0.0691, + "step": 8330 + }, + { + "epoch": 1.09, + "learning_rate": 1.941208890070156e-05, + "loss": 0.0627, + "step": 8331 + }, + { + "epoch": 1.09, + "learning_rate": 1.941190973032673e-05, + "loss": 0.0846, + "step": 8332 + }, + { + "epoch": 1.09, + "learning_rate": 1.94117305334814e-05, + "loss": 0.0805, + "step": 8333 + }, + { + "epoch": 1.09, + "learning_rate": 1.9411551310166082e-05, + "loss": 0.0788, + "step": 8334 + }, + { + "epoch": 1.09, + "learning_rate": 1.9411372060381274e-05, + "loss": 0.0759, + "step": 8335 + }, + { + "epoch": 1.09, + "learning_rate": 1.941119278412749e-05, + "loss": 0.0804, + "step": 8336 + }, + { + "epoch": 1.09, + "learning_rate": 1.9411013481405223e-05, + "loss": 0.0765, + "step": 8337 + }, + { + "epoch": 1.09, + "learning_rate": 1.9410834152214982e-05, + "loss": 0.0722, + "step": 8338 + }, + { + "epoch": 1.09, + "learning_rate": 1.941065479655727e-05, + "loss": 0.0784, + "step": 8339 + }, + { + "epoch": 1.09, + "learning_rate": 1.9410475414432596e-05, + "loss": 0.0892, + "step": 8340 + }, + { + "epoch": 1.09, + "learning_rate": 1.9410296005841456e-05, + "loss": 0.0777, + "step": 8341 + }, + { + "epoch": 1.09, + "learning_rate": 1.9410116570784362e-05, + "loss": 0.0841, + "step": 8342 + }, + { + "epoch": 1.09, + "learning_rate": 1.940993710926182e-05, + "loss": 0.0912, + "step": 8343 + }, + { + "epoch": 1.09, + "learning_rate": 1.9409757621274324e-05, + "loss": 0.073, + "step": 8344 + }, + { + "epoch": 1.09, + "learning_rate": 1.9409578106822392e-05, + "loss": 0.0855, + "step": 8345 + }, + { + "epoch": 1.09, + "learning_rate": 1.9409398565906522e-05, + "loss": 0.0778, + "step": 8346 + }, + { + "epoch": 1.09, + "learning_rate": 1.9409218998527215e-05, + "loss": 0.0879, + "step": 8347 + }, + { + "epoch": 1.09, + "learning_rate": 1.9409039404684983e-05, + "loss": 0.0883, + "step": 8348 + }, + { + "epoch": 1.09, + "learning_rate": 1.940885978438033e-05, + "loss": 0.0629, + "step": 8349 + }, + { + "epoch": 1.09, + "learning_rate": 1.9408680137613758e-05, + "loss": 0.0808, + "step": 8350 + }, + { + "epoch": 1.09, + "learning_rate": 1.940850046438578e-05, + "loss": 0.0853, + "step": 8351 + }, + { + "epoch": 1.09, + "learning_rate": 1.9408320764696886e-05, + "loss": 0.0817, + "step": 8352 + }, + { + "epoch": 1.09, + "learning_rate": 1.9408141038547596e-05, + "loss": 0.0855, + "step": 8353 + }, + { + "epoch": 1.09, + "learning_rate": 1.940796128593841e-05, + "loss": 0.0828, + "step": 8354 + }, + { + "epoch": 1.09, + "learning_rate": 1.9407781506869834e-05, + "loss": 0.0876, + "step": 8355 + }, + { + "epoch": 1.09, + "learning_rate": 1.9407601701342372e-05, + "loss": 0.08, + "step": 8356 + }, + { + "epoch": 1.09, + "learning_rate": 1.9407421869356535e-05, + "loss": 0.065, + "step": 8357 + }, + { + "epoch": 1.09, + "learning_rate": 1.940724201091282e-05, + "loss": 0.0918, + "step": 8358 + }, + { + "epoch": 1.09, + "learning_rate": 1.940706212601174e-05, + "loss": 0.0759, + "step": 8359 + }, + { + "epoch": 1.1, + "learning_rate": 1.94068822146538e-05, + "loss": 0.0817, + "step": 8360 + }, + { + "epoch": 1.1, + "learning_rate": 1.9406702276839506e-05, + "loss": 0.0829, + "step": 8361 + }, + { + "epoch": 1.1, + "learning_rate": 1.940652231256936e-05, + "loss": 0.0948, + "step": 8362 + }, + { + "epoch": 1.1, + "learning_rate": 1.9406342321843872e-05, + "loss": 0.0813, + "step": 8363 + }, + { + "epoch": 1.1, + "learning_rate": 1.9406162304663545e-05, + "loss": 0.0764, + "step": 8364 + }, + { + "epoch": 1.1, + "learning_rate": 1.940598226102889e-05, + "loss": 0.086, + "step": 8365 + }, + { + "epoch": 1.1, + "learning_rate": 1.940580219094041e-05, + "loss": 0.0876, + "step": 8366 + }, + { + "epoch": 1.1, + "learning_rate": 1.940562209439861e-05, + "loss": 0.0901, + "step": 8367 + }, + { + "epoch": 1.1, + "learning_rate": 1.9405441971404002e-05, + "loss": 0.0799, + "step": 8368 + }, + { + "epoch": 1.1, + "learning_rate": 1.9405261821957088e-05, + "loss": 0.0889, + "step": 8369 + }, + { + "epoch": 1.1, + "learning_rate": 1.9405081646058375e-05, + "loss": 0.0869, + "step": 8370 + }, + { + "epoch": 1.1, + "learning_rate": 1.9404901443708374e-05, + "loss": 0.0733, + "step": 8371 + }, + { + "epoch": 1.1, + "learning_rate": 1.9404721214907585e-05, + "loss": 0.0908, + "step": 8372 + }, + { + "epoch": 1.1, + "learning_rate": 1.9404540959656518e-05, + "loss": 0.081, + "step": 8373 + }, + { + "epoch": 1.1, + "learning_rate": 1.9404360677955685e-05, + "loss": 0.0829, + "step": 8374 + }, + { + "epoch": 1.1, + "learning_rate": 1.9404180369805584e-05, + "loss": 0.0872, + "step": 8375 + }, + { + "epoch": 1.1, + "learning_rate": 1.9404000035206727e-05, + "loss": 0.0735, + "step": 8376 + }, + { + "epoch": 1.1, + "learning_rate": 1.9403819674159622e-05, + "loss": 0.0806, + "step": 8377 + }, + { + "epoch": 1.1, + "learning_rate": 1.9403639286664774e-05, + "loss": 0.088, + "step": 8378 + }, + { + "epoch": 1.1, + "learning_rate": 1.9403458872722694e-05, + "loss": 0.0826, + "step": 8379 + }, + { + "epoch": 1.1, + "learning_rate": 1.9403278432333884e-05, + "loss": 0.0811, + "step": 8380 + }, + { + "epoch": 1.1, + "learning_rate": 1.9403097965498852e-05, + "loss": 0.0868, + "step": 8381 + }, + { + "epoch": 1.1, + "learning_rate": 1.9402917472218114e-05, + "loss": 0.076, + "step": 8382 + }, + { + "epoch": 1.1, + "learning_rate": 1.9402736952492166e-05, + "loss": 0.0731, + "step": 8383 + }, + { + "epoch": 1.1, + "learning_rate": 1.9402556406321524e-05, + "loss": 0.0847, + "step": 8384 + }, + { + "epoch": 1.1, + "learning_rate": 1.940237583370669e-05, + "loss": 0.0734, + "step": 8385 + }, + { + "epoch": 1.1, + "learning_rate": 1.940219523464818e-05, + "loss": 0.0848, + "step": 8386 + }, + { + "epoch": 1.1, + "learning_rate": 1.940201460914649e-05, + "loss": 0.0748, + "step": 8387 + }, + { + "epoch": 1.1, + "learning_rate": 1.940183395720214e-05, + "loss": 0.0886, + "step": 8388 + }, + { + "epoch": 1.1, + "learning_rate": 1.940165327881563e-05, + "loss": 0.079, + "step": 8389 + }, + { + "epoch": 1.1, + "learning_rate": 1.940147257398747e-05, + "loss": 0.0887, + "step": 8390 + }, + { + "epoch": 1.1, + "learning_rate": 1.9401291842718167e-05, + "loss": 0.0857, + "step": 8391 + }, + { + "epoch": 1.1, + "learning_rate": 1.9401111085008236e-05, + "loss": 0.0811, + "step": 8392 + }, + { + "epoch": 1.1, + "learning_rate": 1.940093030085818e-05, + "loss": 0.0918, + "step": 8393 + }, + { + "epoch": 1.1, + "learning_rate": 1.9400749490268507e-05, + "loss": 0.078, + "step": 8394 + }, + { + "epoch": 1.1, + "learning_rate": 1.9400568653239728e-05, + "loss": 0.0876, + "step": 8395 + }, + { + "epoch": 1.1, + "learning_rate": 1.9400387789772347e-05, + "loss": 0.0776, + "step": 8396 + }, + { + "epoch": 1.1, + "learning_rate": 1.940020689986688e-05, + "loss": 0.0797, + "step": 8397 + }, + { + "epoch": 1.1, + "learning_rate": 1.9400025983523832e-05, + "loss": 0.0796, + "step": 8398 + }, + { + "epoch": 1.1, + "learning_rate": 1.9399845040743708e-05, + "loss": 0.0775, + "step": 8399 + }, + { + "epoch": 1.1, + "learning_rate": 1.939966407152702e-05, + "loss": 0.0771, + "step": 8400 + }, + { + "epoch": 1.1, + "learning_rate": 1.939948307587428e-05, + "loss": 0.0872, + "step": 8401 + }, + { + "epoch": 1.1, + "learning_rate": 1.9399302053785997e-05, + "loss": 0.0842, + "step": 8402 + }, + { + "epoch": 1.1, + "learning_rate": 1.9399121005262673e-05, + "loss": 0.0727, + "step": 8403 + }, + { + "epoch": 1.1, + "learning_rate": 1.9398939930304822e-05, + "loss": 0.075, + "step": 8404 + }, + { + "epoch": 1.1, + "learning_rate": 1.939875882891295e-05, + "loss": 0.0755, + "step": 8405 + }, + { + "epoch": 1.1, + "learning_rate": 1.9398577701087576e-05, + "loss": 0.0809, + "step": 8406 + }, + { + "epoch": 1.1, + "learning_rate": 1.93983965468292e-05, + "loss": 0.0863, + "step": 8407 + }, + { + "epoch": 1.1, + "learning_rate": 1.9398215366138334e-05, + "loss": 0.083, + "step": 8408 + }, + { + "epoch": 1.1, + "learning_rate": 1.939803415901549e-05, + "loss": 0.0808, + "step": 8409 + }, + { + "epoch": 1.1, + "learning_rate": 1.939785292546117e-05, + "loss": 0.0824, + "step": 8410 + }, + { + "epoch": 1.1, + "learning_rate": 1.9397671665475893e-05, + "loss": 0.0948, + "step": 8411 + }, + { + "epoch": 1.1, + "learning_rate": 1.9397490379060164e-05, + "loss": 0.0962, + "step": 8412 + }, + { + "epoch": 1.1, + "learning_rate": 1.9397309066214493e-05, + "loss": 0.0817, + "step": 8413 + }, + { + "epoch": 1.1, + "learning_rate": 1.9397127726939394e-05, + "loss": 0.0956, + "step": 8414 + }, + { + "epoch": 1.1, + "learning_rate": 1.939694636123537e-05, + "loss": 0.0768, + "step": 8415 + }, + { + "epoch": 1.1, + "learning_rate": 1.9396764969102936e-05, + "loss": 0.0755, + "step": 8416 + }, + { + "epoch": 1.1, + "learning_rate": 1.93965835505426e-05, + "loss": 0.0654, + "step": 8417 + }, + { + "epoch": 1.1, + "learning_rate": 1.9396402105554876e-05, + "loss": 0.0856, + "step": 8418 + }, + { + "epoch": 1.1, + "learning_rate": 1.939622063414027e-05, + "loss": 0.1001, + "step": 8419 + }, + { + "epoch": 1.1, + "learning_rate": 1.9396039136299292e-05, + "loss": 0.0875, + "step": 8420 + }, + { + "epoch": 1.1, + "learning_rate": 1.9395857612032456e-05, + "loss": 0.0856, + "step": 8421 + }, + { + "epoch": 1.1, + "learning_rate": 1.939567606134027e-05, + "loss": 0.069, + "step": 8422 + }, + { + "epoch": 1.1, + "learning_rate": 1.9395494484223248e-05, + "loss": 0.0907, + "step": 8423 + }, + { + "epoch": 1.1, + "learning_rate": 1.9395312880681897e-05, + "loss": 0.082, + "step": 8424 + }, + { + "epoch": 1.1, + "learning_rate": 1.9395131250716726e-05, + "loss": 0.0796, + "step": 8425 + }, + { + "epoch": 1.1, + "learning_rate": 1.939494959432825e-05, + "loss": 0.0734, + "step": 8426 + }, + { + "epoch": 1.1, + "learning_rate": 1.9394767911516982e-05, + "loss": 0.0818, + "step": 8427 + }, + { + "epoch": 1.1, + "learning_rate": 1.9394586202283422e-05, + "loss": 0.0821, + "step": 8428 + }, + { + "epoch": 1.1, + "learning_rate": 1.9394404466628095e-05, + "loss": 0.0907, + "step": 8429 + }, + { + "epoch": 1.1, + "learning_rate": 1.9394222704551504e-05, + "loss": 0.0948, + "step": 8430 + }, + { + "epoch": 1.1, + "learning_rate": 1.9394040916054163e-05, + "loss": 0.0824, + "step": 8431 + }, + { + "epoch": 1.1, + "learning_rate": 1.939385910113658e-05, + "loss": 0.076, + "step": 8432 + }, + { + "epoch": 1.1, + "learning_rate": 1.939367725979927e-05, + "loss": 0.0776, + "step": 8433 + }, + { + "epoch": 1.1, + "learning_rate": 1.939349539204274e-05, + "loss": 0.0861, + "step": 8434 + }, + { + "epoch": 1.1, + "learning_rate": 1.939331349786751e-05, + "loss": 0.0827, + "step": 8435 + }, + { + "epoch": 1.11, + "learning_rate": 1.9393131577274086e-05, + "loss": 0.0692, + "step": 8436 + }, + { + "epoch": 1.11, + "learning_rate": 1.9392949630262975e-05, + "loss": 0.0802, + "step": 8437 + }, + { + "epoch": 1.11, + "learning_rate": 1.9392767656834696e-05, + "loss": 0.0861, + "step": 8438 + }, + { + "epoch": 1.11, + "learning_rate": 1.9392585656989756e-05, + "loss": 0.0719, + "step": 8439 + }, + { + "epoch": 1.11, + "learning_rate": 1.939240363072867e-05, + "loss": 0.0857, + "step": 8440 + }, + { + "epoch": 1.11, + "learning_rate": 1.9392221578051947e-05, + "loss": 0.0714, + "step": 8441 + }, + { + "epoch": 1.11, + "learning_rate": 1.9392039498960106e-05, + "loss": 0.0783, + "step": 8442 + }, + { + "epoch": 1.11, + "learning_rate": 1.9391857393453654e-05, + "loss": 0.0846, + "step": 8443 + }, + { + "epoch": 1.11, + "learning_rate": 1.93916752615331e-05, + "loss": 0.0686, + "step": 8444 + }, + { + "epoch": 1.11, + "learning_rate": 1.9391493103198957e-05, + "loss": 0.0736, + "step": 8445 + }, + { + "epoch": 1.11, + "learning_rate": 1.9391310918451746e-05, + "loss": 0.0845, + "step": 8446 + }, + { + "epoch": 1.11, + "learning_rate": 1.9391128707291967e-05, + "loss": 0.0865, + "step": 8447 + }, + { + "epoch": 1.11, + "learning_rate": 1.9390946469720142e-05, + "loss": 0.071, + "step": 8448 + }, + { + "epoch": 1.11, + "learning_rate": 1.9390764205736777e-05, + "loss": 0.0812, + "step": 8449 + }, + { + "epoch": 1.11, + "learning_rate": 1.9390581915342395e-05, + "loss": 0.0793, + "step": 8450 + }, + { + "epoch": 1.11, + "learning_rate": 1.9390399598537496e-05, + "loss": 0.0763, + "step": 8451 + }, + { + "epoch": 1.11, + "learning_rate": 1.9390217255322597e-05, + "loss": 0.0707, + "step": 8452 + }, + { + "epoch": 1.11, + "learning_rate": 1.9390034885698218e-05, + "loss": 0.0734, + "step": 8453 + }, + { + "epoch": 1.11, + "learning_rate": 1.938985248966486e-05, + "loss": 0.0931, + "step": 8454 + }, + { + "epoch": 1.11, + "learning_rate": 1.938967006722304e-05, + "loss": 0.0841, + "step": 8455 + }, + { + "epoch": 1.11, + "learning_rate": 1.938948761837328e-05, + "loss": 0.0834, + "step": 8456 + }, + { + "epoch": 1.11, + "learning_rate": 1.9389305143116084e-05, + "loss": 0.087, + "step": 8457 + }, + { + "epoch": 1.11, + "learning_rate": 1.9389122641451966e-05, + "loss": 0.0879, + "step": 8458 + }, + { + "epoch": 1.11, + "learning_rate": 1.938894011338144e-05, + "loss": 0.0872, + "step": 8459 + }, + { + "epoch": 1.11, + "learning_rate": 1.9388757558905023e-05, + "loss": 0.0778, + "step": 8460 + }, + { + "epoch": 1.11, + "learning_rate": 1.9388574978023223e-05, + "loss": 0.0789, + "step": 8461 + }, + { + "epoch": 1.11, + "learning_rate": 1.9388392370736557e-05, + "loss": 0.0678, + "step": 8462 + }, + { + "epoch": 1.11, + "learning_rate": 1.9388209737045532e-05, + "loss": 0.0905, + "step": 8463 + }, + { + "epoch": 1.11, + "learning_rate": 1.9388027076950675e-05, + "loss": 0.0694, + "step": 8464 + }, + { + "epoch": 1.11, + "learning_rate": 1.938784439045249e-05, + "loss": 0.0944, + "step": 8465 + }, + { + "epoch": 1.11, + "learning_rate": 1.9387661677551488e-05, + "loss": 0.0875, + "step": 8466 + }, + { + "epoch": 1.11, + "learning_rate": 1.938747893824819e-05, + "loss": 0.0898, + "step": 8467 + }, + { + "epoch": 1.11, + "learning_rate": 1.9387296172543107e-05, + "loss": 0.0954, + "step": 8468 + }, + { + "epoch": 1.11, + "learning_rate": 1.9387113380436756e-05, + "loss": 0.0815, + "step": 8469 + }, + { + "epoch": 1.11, + "learning_rate": 1.9386930561929644e-05, + "loss": 0.0817, + "step": 8470 + }, + { + "epoch": 1.11, + "learning_rate": 1.9386747717022294e-05, + "loss": 0.0894, + "step": 8471 + }, + { + "epoch": 1.11, + "learning_rate": 1.9386564845715215e-05, + "loss": 0.0749, + "step": 8472 + }, + { + "epoch": 1.11, + "learning_rate": 1.938638194800892e-05, + "loss": 0.0828, + "step": 8473 + }, + { + "epoch": 1.11, + "learning_rate": 1.938619902390393e-05, + "loss": 0.0733, + "step": 8474 + }, + { + "epoch": 1.11, + "learning_rate": 1.9386016073400752e-05, + "loss": 0.0763, + "step": 8475 + }, + { + "epoch": 1.11, + "learning_rate": 1.9385833096499902e-05, + "loss": 0.081, + "step": 8476 + }, + { + "epoch": 1.11, + "learning_rate": 1.9385650093201898e-05, + "loss": 0.0819, + "step": 8477 + }, + { + "epoch": 1.11, + "learning_rate": 1.9385467063507253e-05, + "loss": 0.0912, + "step": 8478 + }, + { + "epoch": 1.11, + "learning_rate": 1.9385284007416483e-05, + "loss": 0.0752, + "step": 8479 + }, + { + "epoch": 1.11, + "learning_rate": 1.9385100924930097e-05, + "loss": 0.0788, + "step": 8480 + }, + { + "epoch": 1.11, + "learning_rate": 1.9384917816048622e-05, + "loss": 0.0601, + "step": 8481 + }, + { + "epoch": 1.11, + "learning_rate": 1.9384734680772557e-05, + "loss": 0.0771, + "step": 8482 + }, + { + "epoch": 1.11, + "learning_rate": 1.938455151910243e-05, + "loss": 0.082, + "step": 8483 + }, + { + "epoch": 1.11, + "learning_rate": 1.938436833103875e-05, + "loss": 0.0754, + "step": 8484 + }, + { + "epoch": 1.11, + "learning_rate": 1.938418511658204e-05, + "loss": 0.0784, + "step": 8485 + }, + { + "epoch": 1.11, + "learning_rate": 1.9384001875732804e-05, + "loss": 0.0778, + "step": 8486 + }, + { + "epoch": 1.11, + "learning_rate": 1.938381860849156e-05, + "loss": 0.0779, + "step": 8487 + }, + { + "epoch": 1.11, + "learning_rate": 1.9383635314858827e-05, + "loss": 0.0801, + "step": 8488 + }, + { + "epoch": 1.11, + "learning_rate": 1.9383451994835122e-05, + "loss": 0.0795, + "step": 8489 + }, + { + "epoch": 1.11, + "learning_rate": 1.938326864842096e-05, + "loss": 0.0819, + "step": 8490 + }, + { + "epoch": 1.11, + "learning_rate": 1.938308527561685e-05, + "loss": 0.0834, + "step": 8491 + }, + { + "epoch": 1.11, + "learning_rate": 1.9382901876423313e-05, + "loss": 0.0656, + "step": 8492 + }, + { + "epoch": 1.11, + "learning_rate": 1.9382718450840865e-05, + "loss": 0.0803, + "step": 8493 + }, + { + "epoch": 1.11, + "learning_rate": 1.9382534998870023e-05, + "loss": 0.0782, + "step": 8494 + }, + { + "epoch": 1.11, + "learning_rate": 1.93823515205113e-05, + "loss": 0.0854, + "step": 8495 + }, + { + "epoch": 1.11, + "learning_rate": 1.938216801576521e-05, + "loss": 0.0849, + "step": 8496 + }, + { + "epoch": 1.11, + "learning_rate": 1.9381984484632274e-05, + "loss": 0.0782, + "step": 8497 + }, + { + "epoch": 1.11, + "learning_rate": 1.9381800927113005e-05, + "loss": 0.0779, + "step": 8498 + }, + { + "epoch": 1.11, + "learning_rate": 1.9381617343207922e-05, + "loss": 0.0831, + "step": 8499 + }, + { + "epoch": 1.11, + "learning_rate": 1.938143373291754e-05, + "loss": 0.081, + "step": 8500 + }, + { + "epoch": 1.11, + "learning_rate": 1.9381250096242374e-05, + "loss": 0.0823, + "step": 8501 + }, + { + "epoch": 1.11, + "learning_rate": 1.9381066433182943e-05, + "loss": 0.0897, + "step": 8502 + }, + { + "epoch": 1.11, + "learning_rate": 1.9380882743739763e-05, + "loss": 0.0899, + "step": 8503 + }, + { + "epoch": 1.11, + "learning_rate": 1.9380699027913347e-05, + "loss": 0.0837, + "step": 8504 + }, + { + "epoch": 1.11, + "learning_rate": 1.9380515285704216e-05, + "loss": 0.0764, + "step": 8505 + }, + { + "epoch": 1.11, + "learning_rate": 1.9380331517112885e-05, + "loss": 0.0829, + "step": 8506 + }, + { + "epoch": 1.11, + "learning_rate": 1.938014772213987e-05, + "loss": 0.0724, + "step": 8507 + }, + { + "epoch": 1.11, + "learning_rate": 1.937996390078569e-05, + "loss": 0.085, + "step": 8508 + }, + { + "epoch": 1.11, + "learning_rate": 1.937978005305086e-05, + "loss": 0.0802, + "step": 8509 + }, + { + "epoch": 1.11, + "learning_rate": 1.9379596178935902e-05, + "loss": 0.09, + "step": 8510 + }, + { + "epoch": 1.11, + "learning_rate": 1.9379412278441328e-05, + "loss": 0.0916, + "step": 8511 + }, + { + "epoch": 1.12, + "learning_rate": 1.9379228351567658e-05, + "loss": 0.0829, + "step": 8512 + }, + { + "epoch": 1.12, + "learning_rate": 1.9379044398315403e-05, + "loss": 0.0949, + "step": 8513 + }, + { + "epoch": 1.12, + "learning_rate": 1.9378860418685085e-05, + "loss": 0.0823, + "step": 8514 + }, + { + "epoch": 1.12, + "learning_rate": 1.9378676412677226e-05, + "loss": 0.0936, + "step": 8515 + }, + { + "epoch": 1.12, + "learning_rate": 1.9378492380292334e-05, + "loss": 0.0871, + "step": 8516 + }, + { + "epoch": 1.12, + "learning_rate": 1.9378308321530938e-05, + "loss": 0.0837, + "step": 8517 + }, + { + "epoch": 1.12, + "learning_rate": 1.9378124236393542e-05, + "loss": 0.0796, + "step": 8518 + }, + { + "epoch": 1.12, + "learning_rate": 1.9377940124880675e-05, + "loss": 0.0913, + "step": 8519 + }, + { + "epoch": 1.12, + "learning_rate": 1.9377755986992853e-05, + "loss": 0.0942, + "step": 8520 + }, + { + "epoch": 1.12, + "learning_rate": 1.937757182273059e-05, + "loss": 0.0827, + "step": 8521 + }, + { + "epoch": 1.12, + "learning_rate": 1.9377387632094403e-05, + "loss": 0.0724, + "step": 8522 + }, + { + "epoch": 1.12, + "learning_rate": 1.9377203415084813e-05, + "loss": 0.092, + "step": 8523 + }, + { + "epoch": 1.12, + "learning_rate": 1.937701917170234e-05, + "loss": 0.097, + "step": 8524 + }, + { + "epoch": 1.12, + "learning_rate": 1.93768349019475e-05, + "loss": 0.0699, + "step": 8525 + }, + { + "epoch": 1.12, + "learning_rate": 1.9376650605820807e-05, + "loss": 0.0847, + "step": 8526 + }, + { + "epoch": 1.12, + "learning_rate": 1.9376466283322787e-05, + "loss": 0.0755, + "step": 8527 + }, + { + "epoch": 1.12, + "learning_rate": 1.9376281934453955e-05, + "loss": 0.0711, + "step": 8528 + }, + { + "epoch": 1.12, + "learning_rate": 1.937609755921483e-05, + "loss": 0.0832, + "step": 8529 + }, + { + "epoch": 1.12, + "learning_rate": 1.937591315760593e-05, + "loss": 0.0897, + "step": 8530 + }, + { + "epoch": 1.12, + "learning_rate": 1.9375728729627772e-05, + "loss": 0.0811, + "step": 8531 + }, + { + "epoch": 1.12, + "learning_rate": 1.9375544275280878e-05, + "loss": 0.0793, + "step": 8532 + }, + { + "epoch": 1.12, + "learning_rate": 1.9375359794565763e-05, + "loss": 0.0825, + "step": 8533 + }, + { + "epoch": 1.12, + "learning_rate": 1.9375175287482947e-05, + "loss": 0.0789, + "step": 8534 + }, + { + "epoch": 1.12, + "learning_rate": 1.937499075403295e-05, + "loss": 0.0759, + "step": 8535 + }, + { + "epoch": 1.12, + "learning_rate": 1.9374806194216296e-05, + "loss": 0.0706, + "step": 8536 + }, + { + "epoch": 1.12, + "learning_rate": 1.9374621608033494e-05, + "loss": 0.0831, + "step": 8537 + }, + { + "epoch": 1.12, + "learning_rate": 1.9374436995485067e-05, + "loss": 0.0813, + "step": 8538 + }, + { + "epoch": 1.12, + "learning_rate": 1.937425235657154e-05, + "loss": 0.0637, + "step": 8539 + }, + { + "epoch": 1.12, + "learning_rate": 1.9374067691293422e-05, + "loss": 0.0795, + "step": 8540 + }, + { + "epoch": 1.12, + "learning_rate": 1.937388299965124e-05, + "loss": 0.0766, + "step": 8541 + }, + { + "epoch": 1.12, + "learning_rate": 1.937369828164551e-05, + "loss": 0.0765, + "step": 8542 + }, + { + "epoch": 1.12, + "learning_rate": 1.9373513537276758e-05, + "loss": 0.0743, + "step": 8543 + }, + { + "epoch": 1.12, + "learning_rate": 1.9373328766545494e-05, + "loss": 0.0823, + "step": 8544 + }, + { + "epoch": 1.12, + "learning_rate": 1.937314396945224e-05, + "loss": 0.0804, + "step": 8545 + }, + { + "epoch": 1.12, + "learning_rate": 1.9372959145997523e-05, + "loss": 0.0928, + "step": 8546 + }, + { + "epoch": 1.12, + "learning_rate": 1.9372774296181856e-05, + "loss": 0.0853, + "step": 8547 + }, + { + "epoch": 1.12, + "learning_rate": 1.9372589420005757e-05, + "loss": 0.0667, + "step": 8548 + }, + { + "epoch": 1.12, + "learning_rate": 1.9372404517469753e-05, + "loss": 0.074, + "step": 8549 + }, + { + "epoch": 1.12, + "learning_rate": 1.9372219588574358e-05, + "loss": 0.0804, + "step": 8550 + }, + { + "epoch": 1.12, + "learning_rate": 1.9372034633320097e-05, + "loss": 0.079, + "step": 8551 + }, + { + "epoch": 1.12, + "learning_rate": 1.9371849651707486e-05, + "loss": 0.0797, + "step": 8552 + }, + { + "epoch": 1.12, + "learning_rate": 1.937166464373705e-05, + "loss": 0.0843, + "step": 8553 + }, + { + "epoch": 1.12, + "learning_rate": 1.93714796094093e-05, + "loss": 0.0896, + "step": 8554 + }, + { + "epoch": 1.12, + "learning_rate": 1.9371294548724768e-05, + "loss": 0.0728, + "step": 8555 + }, + { + "epoch": 1.12, + "learning_rate": 1.9371109461683967e-05, + "loss": 0.0835, + "step": 8556 + }, + { + "epoch": 1.12, + "learning_rate": 1.937092434828742e-05, + "loss": 0.076, + "step": 8557 + }, + { + "epoch": 1.12, + "learning_rate": 1.9370739208535647e-05, + "loss": 0.0806, + "step": 8558 + }, + { + "epoch": 1.12, + "learning_rate": 1.937055404242917e-05, + "loss": 0.0837, + "step": 8559 + }, + { + "epoch": 1.12, + "learning_rate": 1.937036884996851e-05, + "loss": 0.0922, + "step": 8560 + }, + { + "epoch": 1.12, + "learning_rate": 1.9370183631154183e-05, + "loss": 0.0698, + "step": 8561 + }, + { + "epoch": 1.12, + "learning_rate": 1.9369998385986714e-05, + "loss": 0.0764, + "step": 8562 + }, + { + "epoch": 1.12, + "learning_rate": 1.9369813114466627e-05, + "loss": 0.0823, + "step": 8563 + }, + { + "epoch": 1.12, + "learning_rate": 1.9369627816594438e-05, + "loss": 0.0895, + "step": 8564 + }, + { + "epoch": 1.12, + "learning_rate": 1.9369442492370667e-05, + "loss": 0.077, + "step": 8565 + }, + { + "epoch": 1.12, + "learning_rate": 1.936925714179584e-05, + "loss": 0.0815, + "step": 8566 + }, + { + "epoch": 1.12, + "learning_rate": 1.9369071764870475e-05, + "loss": 0.0799, + "step": 8567 + }, + { + "epoch": 1.12, + "learning_rate": 1.9368886361595098e-05, + "loss": 0.0811, + "step": 8568 + }, + { + "epoch": 1.12, + "learning_rate": 1.9368700931970223e-05, + "loss": 0.0714, + "step": 8569 + }, + { + "epoch": 1.12, + "learning_rate": 1.9368515475996376e-05, + "loss": 0.0822, + "step": 8570 + }, + { + "epoch": 1.12, + "learning_rate": 1.936832999367408e-05, + "loss": 0.0796, + "step": 8571 + }, + { + "epoch": 1.12, + "learning_rate": 1.9368144485003854e-05, + "loss": 0.081, + "step": 8572 + }, + { + "epoch": 1.12, + "learning_rate": 1.9367958949986224e-05, + "loss": 0.0934, + "step": 8573 + }, + { + "epoch": 1.12, + "learning_rate": 1.9367773388621705e-05, + "loss": 0.0953, + "step": 8574 + }, + { + "epoch": 1.12, + "learning_rate": 1.936758780091082e-05, + "loss": 0.0837, + "step": 8575 + }, + { + "epoch": 1.12, + "learning_rate": 1.93674021868541e-05, + "loss": 0.0895, + "step": 8576 + }, + { + "epoch": 1.12, + "learning_rate": 1.936721654645205e-05, + "loss": 0.085, + "step": 8577 + }, + { + "epoch": 1.12, + "learning_rate": 1.9367030879705212e-05, + "loss": 0.0882, + "step": 8578 + }, + { + "epoch": 1.12, + "learning_rate": 1.9366845186614093e-05, + "loss": 0.0749, + "step": 8579 + }, + { + "epoch": 1.12, + "learning_rate": 1.9366659467179223e-05, + "loss": 0.079, + "step": 8580 + }, + { + "epoch": 1.12, + "learning_rate": 1.936647372140112e-05, + "loss": 0.098, + "step": 8581 + }, + { + "epoch": 1.12, + "learning_rate": 1.936628794928031e-05, + "loss": 0.0794, + "step": 8582 + }, + { + "epoch": 1.12, + "learning_rate": 1.9366102150817313e-05, + "loss": 0.0903, + "step": 8583 + }, + { + "epoch": 1.12, + "learning_rate": 1.9365916326012655e-05, + "loss": 0.0897, + "step": 8584 + }, + { + "epoch": 1.12, + "learning_rate": 1.9365730474866855e-05, + "loss": 0.0855, + "step": 8585 + }, + { + "epoch": 1.12, + "learning_rate": 1.9365544597380435e-05, + "loss": 0.0835, + "step": 8586 + }, + { + "epoch": 1.12, + "learning_rate": 1.936535869355392e-05, + "loss": 0.0851, + "step": 8587 + }, + { + "epoch": 1.12, + "learning_rate": 1.9365172763387833e-05, + "loss": 0.0751, + "step": 8588 + }, + { + "epoch": 1.13, + "learning_rate": 1.9364986806882698e-05, + "loss": 0.0897, + "step": 8589 + }, + { + "epoch": 1.13, + "learning_rate": 1.9364800824039033e-05, + "loss": 0.0794, + "step": 8590 + }, + { + "epoch": 1.13, + "learning_rate": 1.9364614814857367e-05, + "loss": 0.0884, + "step": 8591 + }, + { + "epoch": 1.13, + "learning_rate": 1.9364428779338217e-05, + "loss": 0.0815, + "step": 8592 + }, + { + "epoch": 1.13, + "learning_rate": 1.9364242717482112e-05, + "loss": 0.0821, + "step": 8593 + }, + { + "epoch": 1.13, + "learning_rate": 1.9364056629289575e-05, + "loss": 0.0801, + "step": 8594 + }, + { + "epoch": 1.13, + "learning_rate": 1.9363870514761127e-05, + "loss": 0.079, + "step": 8595 + }, + { + "epoch": 1.13, + "learning_rate": 1.936368437389729e-05, + "loss": 0.079, + "step": 8596 + }, + { + "epoch": 1.13, + "learning_rate": 1.936349820669859e-05, + "loss": 0.0758, + "step": 8597 + }, + { + "epoch": 1.13, + "learning_rate": 1.936331201316555e-05, + "loss": 0.0836, + "step": 8598 + }, + { + "epoch": 1.13, + "learning_rate": 1.936312579329869e-05, + "loss": 0.0816, + "step": 8599 + }, + { + "epoch": 1.13, + "learning_rate": 1.9362939547098544e-05, + "loss": 0.0812, + "step": 8600 + }, + { + "epoch": 1.13, + "learning_rate": 1.9362753274565627e-05, + "loss": 0.0866, + "step": 8601 + }, + { + "epoch": 1.13, + "learning_rate": 1.9362566975700464e-05, + "loss": 0.0792, + "step": 8602 + }, + { + "epoch": 1.13, + "learning_rate": 1.9362380650503578e-05, + "loss": 0.0693, + "step": 8603 + }, + { + "epoch": 1.13, + "learning_rate": 1.9362194298975498e-05, + "loss": 0.0814, + "step": 8604 + }, + { + "epoch": 1.13, + "learning_rate": 1.9362007921116744e-05, + "loss": 0.0833, + "step": 8605 + }, + { + "epoch": 1.13, + "learning_rate": 1.936182151692784e-05, + "loss": 0.0685, + "step": 8606 + }, + { + "epoch": 1.13, + "learning_rate": 1.9361635086409314e-05, + "loss": 0.0796, + "step": 8607 + }, + { + "epoch": 1.13, + "learning_rate": 1.9361448629561685e-05, + "loss": 0.0842, + "step": 8608 + }, + { + "epoch": 1.13, + "learning_rate": 1.936126214638548e-05, + "loss": 0.0877, + "step": 8609 + }, + { + "epoch": 1.13, + "learning_rate": 1.936107563688123e-05, + "loss": 0.0857, + "step": 8610 + }, + { + "epoch": 1.13, + "learning_rate": 1.9360889101049448e-05, + "loss": 0.0929, + "step": 8611 + }, + { + "epoch": 1.13, + "learning_rate": 1.9360702538890663e-05, + "loss": 0.078, + "step": 8612 + }, + { + "epoch": 1.13, + "learning_rate": 1.9360515950405404e-05, + "loss": 0.0789, + "step": 8613 + }, + { + "epoch": 1.13, + "learning_rate": 1.936032933559419e-05, + "loss": 0.0811, + "step": 8614 + }, + { + "epoch": 1.13, + "learning_rate": 1.936014269445755e-05, + "loss": 0.0859, + "step": 8615 + }, + { + "epoch": 1.13, + "learning_rate": 1.9359956026996005e-05, + "loss": 0.0815, + "step": 8616 + }, + { + "epoch": 1.13, + "learning_rate": 1.9359769333210082e-05, + "loss": 0.0946, + "step": 8617 + }, + { + "epoch": 1.13, + "learning_rate": 1.935958261310031e-05, + "loss": 0.0867, + "step": 8618 + }, + { + "epoch": 1.13, + "learning_rate": 1.9359395866667204e-05, + "loss": 0.0948, + "step": 8619 + }, + { + "epoch": 1.13, + "learning_rate": 1.9359209093911302e-05, + "loss": 0.0848, + "step": 8620 + }, + { + "epoch": 1.13, + "learning_rate": 1.935902229483312e-05, + "loss": 0.0857, + "step": 8621 + }, + { + "epoch": 1.13, + "learning_rate": 1.9358835469433185e-05, + "loss": 0.0738, + "step": 8622 + }, + { + "epoch": 1.13, + "learning_rate": 1.9358648617712027e-05, + "loss": 0.0892, + "step": 8623 + }, + { + "epoch": 1.13, + "learning_rate": 1.9358461739670167e-05, + "loss": 0.0945, + "step": 8624 + }, + { + "epoch": 1.13, + "learning_rate": 1.9358274835308127e-05, + "loss": 0.088, + "step": 8625 + }, + { + "epoch": 1.13, + "learning_rate": 1.9358087904626442e-05, + "loss": 0.0883, + "step": 8626 + }, + { + "epoch": 1.13, + "learning_rate": 1.935790094762563e-05, + "loss": 0.09, + "step": 8627 + }, + { + "epoch": 1.13, + "learning_rate": 1.9357713964306226e-05, + "loss": 0.0805, + "step": 8628 + }, + { + "epoch": 1.13, + "learning_rate": 1.9357526954668744e-05, + "loss": 0.092, + "step": 8629 + }, + { + "epoch": 1.13, + "learning_rate": 1.935733991871372e-05, + "loss": 0.0793, + "step": 8630 + }, + { + "epoch": 1.13, + "learning_rate": 1.935715285644167e-05, + "loss": 0.082, + "step": 8631 + }, + { + "epoch": 1.13, + "learning_rate": 1.935696576785313e-05, + "loss": 0.0987, + "step": 8632 + }, + { + "epoch": 1.13, + "learning_rate": 1.9356778652948623e-05, + "loss": 0.0884, + "step": 8633 + }, + { + "epoch": 1.13, + "learning_rate": 1.9356591511728676e-05, + "loss": 0.0859, + "step": 8634 + }, + { + "epoch": 1.13, + "learning_rate": 1.9356404344193807e-05, + "loss": 0.0876, + "step": 8635 + }, + { + "epoch": 1.13, + "learning_rate": 1.9356217150344553e-05, + "loss": 0.0779, + "step": 8636 + }, + { + "epoch": 1.13, + "learning_rate": 1.9356029930181437e-05, + "loss": 0.0809, + "step": 8637 + }, + { + "epoch": 1.13, + "learning_rate": 1.9355842683704985e-05, + "loss": 0.0884, + "step": 8638 + }, + { + "epoch": 1.13, + "learning_rate": 1.9355655410915722e-05, + "loss": 0.0807, + "step": 8639 + }, + { + "epoch": 1.13, + "learning_rate": 1.9355468111814177e-05, + "loss": 0.09, + "step": 8640 + }, + { + "epoch": 1.13, + "learning_rate": 1.9355280786400875e-05, + "loss": 0.0814, + "step": 8641 + }, + { + "epoch": 1.13, + "learning_rate": 1.9355093434676347e-05, + "loss": 0.0913, + "step": 8642 + }, + { + "epoch": 1.13, + "learning_rate": 1.9354906056641113e-05, + "loss": 0.0706, + "step": 8643 + }, + { + "epoch": 1.13, + "learning_rate": 1.9354718652295708e-05, + "loss": 0.0827, + "step": 8644 + }, + { + "epoch": 1.13, + "learning_rate": 1.9354531221640652e-05, + "loss": 0.082, + "step": 8645 + }, + { + "epoch": 1.13, + "learning_rate": 1.935434376467648e-05, + "loss": 0.0972, + "step": 8646 + }, + { + "epoch": 1.13, + "learning_rate": 1.935415628140371e-05, + "loss": 0.0823, + "step": 8647 + }, + { + "epoch": 1.13, + "learning_rate": 1.935396877182287e-05, + "loss": 0.076, + "step": 8648 + }, + { + "epoch": 1.13, + "learning_rate": 1.9353781235934494e-05, + "loss": 0.0851, + "step": 8649 + }, + { + "epoch": 1.13, + "learning_rate": 1.935359367373911e-05, + "loss": 0.0698, + "step": 8650 + }, + { + "epoch": 1.13, + "learning_rate": 1.9353406085237237e-05, + "loss": 0.0846, + "step": 8651 + }, + { + "epoch": 1.13, + "learning_rate": 1.9353218470429413e-05, + "loss": 0.0851, + "step": 8652 + }, + { + "epoch": 1.13, + "learning_rate": 1.9353030829316157e-05, + "loss": 0.0867, + "step": 8653 + }, + { + "epoch": 1.13, + "learning_rate": 1.9352843161897998e-05, + "loss": 0.0868, + "step": 8654 + }, + { + "epoch": 1.13, + "learning_rate": 1.9352655468175468e-05, + "loss": 0.0789, + "step": 8655 + }, + { + "epoch": 1.13, + "learning_rate": 1.9352467748149092e-05, + "loss": 0.0775, + "step": 8656 + }, + { + "epoch": 1.13, + "learning_rate": 1.93522800018194e-05, + "loss": 0.0748, + "step": 8657 + }, + { + "epoch": 1.13, + "learning_rate": 1.9352092229186918e-05, + "loss": 0.0704, + "step": 8658 + }, + { + "epoch": 1.13, + "learning_rate": 1.9351904430252172e-05, + "loss": 0.0834, + "step": 8659 + }, + { + "epoch": 1.13, + "learning_rate": 1.9351716605015692e-05, + "loss": 0.0884, + "step": 8660 + }, + { + "epoch": 1.13, + "learning_rate": 1.9351528753478012e-05, + "loss": 0.0798, + "step": 8661 + }, + { + "epoch": 1.13, + "learning_rate": 1.9351340875639654e-05, + "loss": 0.0943, + "step": 8662 + }, + { + "epoch": 1.13, + "learning_rate": 1.9351152971501143e-05, + "loss": 0.0854, + "step": 8663 + }, + { + "epoch": 1.13, + "learning_rate": 1.9350965041063018e-05, + "loss": 0.0864, + "step": 8664 + }, + { + "epoch": 1.14, + "learning_rate": 1.93507770843258e-05, + "loss": 0.0841, + "step": 8665 + }, + { + "epoch": 1.14, + "learning_rate": 1.935058910129002e-05, + "loss": 0.0882, + "step": 8666 + }, + { + "epoch": 1.14, + "learning_rate": 1.9350401091956204e-05, + "loss": 0.0805, + "step": 8667 + }, + { + "epoch": 1.14, + "learning_rate": 1.935021305632488e-05, + "loss": 0.0676, + "step": 8668 + }, + { + "epoch": 1.14, + "learning_rate": 1.9350024994396587e-05, + "loss": 0.0862, + "step": 8669 + }, + { + "epoch": 1.14, + "learning_rate": 1.9349836906171843e-05, + "loss": 0.0842, + "step": 8670 + }, + { + "epoch": 1.14, + "learning_rate": 1.9349648791651182e-05, + "loss": 0.0917, + "step": 8671 + }, + { + "epoch": 1.14, + "learning_rate": 1.934946065083513e-05, + "loss": 0.0863, + "step": 8672 + }, + { + "epoch": 1.14, + "learning_rate": 1.934927248372422e-05, + "loss": 0.0825, + "step": 8673 + }, + { + "epoch": 1.14, + "learning_rate": 1.9349084290318978e-05, + "loss": 0.0687, + "step": 8674 + }, + { + "epoch": 1.14, + "learning_rate": 1.9348896070619935e-05, + "loss": 0.0792, + "step": 8675 + }, + { + "epoch": 1.14, + "learning_rate": 1.9348707824627614e-05, + "loss": 0.0866, + "step": 8676 + }, + { + "epoch": 1.14, + "learning_rate": 1.9348519552342556e-05, + "loss": 0.0836, + "step": 8677 + }, + { + "epoch": 1.14, + "learning_rate": 1.934833125376528e-05, + "loss": 0.0829, + "step": 8678 + }, + { + "epoch": 1.14, + "learning_rate": 1.9348142928896327e-05, + "loss": 0.0775, + "step": 8679 + }, + { + "epoch": 1.14, + "learning_rate": 1.9347954577736216e-05, + "loss": 0.0848, + "step": 8680 + }, + { + "epoch": 1.14, + "learning_rate": 1.9347766200285482e-05, + "loss": 0.0798, + "step": 8681 + }, + { + "epoch": 1.14, + "learning_rate": 1.934757779654465e-05, + "loss": 0.0783, + "step": 8682 + }, + { + "epoch": 1.14, + "learning_rate": 1.9347389366514257e-05, + "loss": 0.0845, + "step": 8683 + }, + { + "epoch": 1.14, + "learning_rate": 1.934720091019483e-05, + "loss": 0.0943, + "step": 8684 + }, + { + "epoch": 1.14, + "learning_rate": 1.9347012427586893e-05, + "loss": 0.0803, + "step": 8685 + }, + { + "epoch": 1.14, + "learning_rate": 1.934682391869098e-05, + "loss": 0.0889, + "step": 8686 + }, + { + "epoch": 1.14, + "learning_rate": 1.934663538350763e-05, + "loss": 0.0807, + "step": 8687 + }, + { + "epoch": 1.14, + "learning_rate": 1.9346446822037365e-05, + "loss": 0.0883, + "step": 8688 + }, + { + "epoch": 1.14, + "learning_rate": 1.934625823428071e-05, + "loss": 0.073, + "step": 8689 + }, + { + "epoch": 1.14, + "learning_rate": 1.9346069620238205e-05, + "loss": 0.0924, + "step": 8690 + }, + { + "epoch": 1.14, + "learning_rate": 1.9345880979910376e-05, + "loss": 0.0773, + "step": 8691 + }, + { + "epoch": 1.14, + "learning_rate": 1.9345692313297756e-05, + "loss": 0.0818, + "step": 8692 + }, + { + "epoch": 1.14, + "learning_rate": 1.934550362040087e-05, + "loss": 0.0795, + "step": 8693 + }, + { + "epoch": 1.14, + "learning_rate": 1.934531490122026e-05, + "loss": 0.0812, + "step": 8694 + }, + { + "epoch": 1.14, + "learning_rate": 1.9345126155756443e-05, + "loss": 0.0918, + "step": 8695 + }, + { + "epoch": 1.14, + "learning_rate": 1.934493738400996e-05, + "loss": 0.0782, + "step": 8696 + }, + { + "epoch": 1.14, + "learning_rate": 1.9344748585981336e-05, + "loss": 0.0841, + "step": 8697 + }, + { + "epoch": 1.14, + "learning_rate": 1.9344559761671107e-05, + "loss": 0.081, + "step": 8698 + }, + { + "epoch": 1.14, + "learning_rate": 1.93443709110798e-05, + "loss": 0.0806, + "step": 8699 + }, + { + "epoch": 1.14, + "learning_rate": 1.9344182034207945e-05, + "loss": 0.0894, + "step": 8700 + }, + { + "epoch": 1.14, + "learning_rate": 1.934399313105608e-05, + "loss": 0.0899, + "step": 8701 + }, + { + "epoch": 1.14, + "learning_rate": 1.9343804201624728e-05, + "loss": 0.0809, + "step": 8702 + }, + { + "epoch": 1.14, + "learning_rate": 1.9343615245914424e-05, + "loss": 0.0746, + "step": 8703 + }, + { + "epoch": 1.14, + "learning_rate": 1.9343426263925705e-05, + "loss": 0.0807, + "step": 8704 + }, + { + "epoch": 1.14, + "learning_rate": 1.934323725565909e-05, + "loss": 0.094, + "step": 8705 + }, + { + "epoch": 1.14, + "learning_rate": 1.9343048221115124e-05, + "loss": 0.0783, + "step": 8706 + }, + { + "epoch": 1.14, + "learning_rate": 1.934285916029433e-05, + "loss": 0.0813, + "step": 8707 + }, + { + "epoch": 1.14, + "learning_rate": 1.934267007319724e-05, + "loss": 0.0733, + "step": 8708 + }, + { + "epoch": 1.14, + "learning_rate": 1.9342480959824392e-05, + "loss": 0.0735, + "step": 8709 + }, + { + "epoch": 1.14, + "learning_rate": 1.9342291820176312e-05, + "loss": 0.0696, + "step": 8710 + }, + { + "epoch": 1.14, + "learning_rate": 1.9342102654253532e-05, + "loss": 0.0866, + "step": 8711 + }, + { + "epoch": 1.14, + "learning_rate": 1.934191346205659e-05, + "loss": 0.0849, + "step": 8712 + }, + { + "epoch": 1.14, + "learning_rate": 1.934172424358601e-05, + "loss": 0.0917, + "step": 8713 + }, + { + "epoch": 1.14, + "learning_rate": 1.9341534998842326e-05, + "loss": 0.0913, + "step": 8714 + }, + { + "epoch": 1.14, + "learning_rate": 1.9341345727826078e-05, + "loss": 0.0787, + "step": 8715 + }, + { + "epoch": 1.14, + "learning_rate": 1.934115643053779e-05, + "loss": 0.0781, + "step": 8716 + }, + { + "epoch": 1.14, + "learning_rate": 1.9340967106977998e-05, + "loss": 0.0897, + "step": 8717 + }, + { + "epoch": 1.14, + "learning_rate": 1.934077775714723e-05, + "loss": 0.0902, + "step": 8718 + }, + { + "epoch": 1.14, + "learning_rate": 1.9340588381046024e-05, + "loss": 0.0817, + "step": 8719 + }, + { + "epoch": 1.14, + "learning_rate": 1.934039897867491e-05, + "loss": 0.0951, + "step": 8720 + }, + { + "epoch": 1.14, + "learning_rate": 1.9340209550034427e-05, + "loss": 0.0852, + "step": 8721 + }, + { + "epoch": 1.14, + "learning_rate": 1.9340020095125096e-05, + "loss": 0.0765, + "step": 8722 + }, + { + "epoch": 1.14, + "learning_rate": 1.9339830613947457e-05, + "loss": 0.0704, + "step": 8723 + }, + { + "epoch": 1.14, + "learning_rate": 1.9339641106502042e-05, + "loss": 0.0792, + "step": 8724 + }, + { + "epoch": 1.14, + "learning_rate": 1.9339451572789384e-05, + "loss": 0.0927, + "step": 8725 + }, + { + "epoch": 1.14, + "learning_rate": 1.9339262012810014e-05, + "loss": 0.0943, + "step": 8726 + }, + { + "epoch": 1.14, + "learning_rate": 1.933907242656447e-05, + "loss": 0.0918, + "step": 8727 + }, + { + "epoch": 1.14, + "learning_rate": 1.9338882814053282e-05, + "loss": 0.0838, + "step": 8728 + }, + { + "epoch": 1.14, + "learning_rate": 1.933869317527698e-05, + "loss": 0.0796, + "step": 8729 + }, + { + "epoch": 1.14, + "learning_rate": 1.9338503510236103e-05, + "loss": 0.086, + "step": 8730 + }, + { + "epoch": 1.14, + "learning_rate": 1.9338313818931182e-05, + "loss": 0.0749, + "step": 8731 + }, + { + "epoch": 1.14, + "learning_rate": 1.9338124101362748e-05, + "loss": 0.0917, + "step": 8732 + }, + { + "epoch": 1.14, + "learning_rate": 1.9337934357531343e-05, + "loss": 0.0822, + "step": 8733 + }, + { + "epoch": 1.14, + "learning_rate": 1.933774458743749e-05, + "loss": 0.0826, + "step": 8734 + }, + { + "epoch": 1.14, + "learning_rate": 1.933755479108173e-05, + "loss": 0.0844, + "step": 8735 + }, + { + "epoch": 1.14, + "learning_rate": 1.9337364968464594e-05, + "loss": 0.0902, + "step": 8736 + }, + { + "epoch": 1.14, + "learning_rate": 1.933717511958662e-05, + "loss": 0.0812, + "step": 8737 + }, + { + "epoch": 1.14, + "learning_rate": 1.9336985244448332e-05, + "loss": 0.0809, + "step": 8738 + }, + { + "epoch": 1.14, + "learning_rate": 1.9336795343050272e-05, + "loss": 0.0786, + "step": 8739 + }, + { + "epoch": 1.14, + "learning_rate": 1.933660541539297e-05, + "loss": 0.079, + "step": 8740 + }, + { + "epoch": 1.15, + "learning_rate": 1.9336415461476968e-05, + "loss": 0.0818, + "step": 8741 + }, + { + "epoch": 1.15, + "learning_rate": 1.9336225481302794e-05, + "loss": 0.0862, + "step": 8742 + }, + { + "epoch": 1.15, + "learning_rate": 1.933603547487098e-05, + "loss": 0.0797, + "step": 8743 + }, + { + "epoch": 1.15, + "learning_rate": 1.9335845442182063e-05, + "loss": 0.0743, + "step": 8744 + }, + { + "epoch": 1.15, + "learning_rate": 1.933565538323658e-05, + "loss": 0.0848, + "step": 8745 + }, + { + "epoch": 1.15, + "learning_rate": 1.9335465298035064e-05, + "loss": 0.0764, + "step": 8746 + }, + { + "epoch": 1.15, + "learning_rate": 1.9335275186578047e-05, + "loss": 0.0753, + "step": 8747 + }, + { + "epoch": 1.15, + "learning_rate": 1.9335085048866066e-05, + "loss": 0.0882, + "step": 8748 + }, + { + "epoch": 1.15, + "learning_rate": 1.9334894884899657e-05, + "loss": 0.0737, + "step": 8749 + }, + { + "epoch": 1.15, + "learning_rate": 1.933470469467935e-05, + "loss": 0.0884, + "step": 8750 + }, + { + "epoch": 1.15, + "learning_rate": 1.9334514478205685e-05, + "loss": 0.0815, + "step": 8751 + }, + { + "epoch": 1.15, + "learning_rate": 1.93343242354792e-05, + "loss": 0.0779, + "step": 8752 + }, + { + "epoch": 1.15, + "learning_rate": 1.9334133966500418e-05, + "loss": 0.0834, + "step": 8753 + }, + { + "epoch": 1.15, + "learning_rate": 1.9333943671269883e-05, + "loss": 0.0933, + "step": 8754 + }, + { + "epoch": 1.15, + "learning_rate": 1.933375334978813e-05, + "loss": 0.0892, + "step": 8755 + }, + { + "epoch": 1.15, + "learning_rate": 1.933356300205569e-05, + "loss": 0.0862, + "step": 8756 + }, + { + "epoch": 1.15, + "learning_rate": 1.9333372628073105e-05, + "loss": 0.0722, + "step": 8757 + }, + { + "epoch": 1.15, + "learning_rate": 1.9333182227840903e-05, + "loss": 0.0889, + "step": 8758 + }, + { + "epoch": 1.15, + "learning_rate": 1.9332991801359628e-05, + "loss": 0.0777, + "step": 8759 + }, + { + "epoch": 1.15, + "learning_rate": 1.9332801348629805e-05, + "loss": 0.0789, + "step": 8760 + }, + { + "epoch": 1.15, + "learning_rate": 1.933261086965198e-05, + "loss": 0.0819, + "step": 8761 + }, + { + "epoch": 1.15, + "learning_rate": 1.9332420364426682e-05, + "loss": 0.0789, + "step": 8762 + }, + { + "epoch": 1.15, + "learning_rate": 1.9332229832954447e-05, + "loss": 0.0848, + "step": 8763 + }, + { + "epoch": 1.15, + "learning_rate": 1.933203927523581e-05, + "loss": 0.0832, + "step": 8764 + }, + { + "epoch": 1.15, + "learning_rate": 1.9331848691271315e-05, + "loss": 0.077, + "step": 8765 + }, + { + "epoch": 1.15, + "learning_rate": 1.9331658081061492e-05, + "loss": 0.0824, + "step": 8766 + }, + { + "epoch": 1.15, + "learning_rate": 1.933146744460687e-05, + "loss": 0.0742, + "step": 8767 + }, + { + "epoch": 1.15, + "learning_rate": 1.9331276781908003e-05, + "loss": 0.0735, + "step": 8768 + }, + { + "epoch": 1.15, + "learning_rate": 1.9331086092965414e-05, + "loss": 0.0782, + "step": 8769 + }, + { + "epoch": 1.15, + "learning_rate": 1.933089537777964e-05, + "loss": 0.0886, + "step": 8770 + }, + { + "epoch": 1.15, + "learning_rate": 1.9330704636351224e-05, + "loss": 0.0788, + "step": 8771 + }, + { + "epoch": 1.15, + "learning_rate": 1.933051386868069e-05, + "loss": 0.0801, + "step": 8772 + }, + { + "epoch": 1.15, + "learning_rate": 1.933032307476859e-05, + "loss": 0.0861, + "step": 8773 + }, + { + "epoch": 1.15, + "learning_rate": 1.9330132254615452e-05, + "loss": 0.0778, + "step": 8774 + }, + { + "epoch": 1.15, + "learning_rate": 1.932994140822181e-05, + "loss": 0.0843, + "step": 8775 + }, + { + "epoch": 1.15, + "learning_rate": 1.9329750535588208e-05, + "loss": 0.0766, + "step": 8776 + }, + { + "epoch": 1.15, + "learning_rate": 1.932955963671518e-05, + "loss": 0.0882, + "step": 8777 + }, + { + "epoch": 1.15, + "learning_rate": 1.932936871160326e-05, + "loss": 0.0829, + "step": 8778 + }, + { + "epoch": 1.15, + "learning_rate": 1.932917776025299e-05, + "loss": 0.0842, + "step": 8779 + }, + { + "epoch": 1.15, + "learning_rate": 1.9328986782664904e-05, + "loss": 0.0856, + "step": 8780 + }, + { + "epoch": 1.15, + "learning_rate": 1.9328795778839534e-05, + "loss": 0.0843, + "step": 8781 + }, + { + "epoch": 1.15, + "learning_rate": 1.9328604748777428e-05, + "loss": 0.0843, + "step": 8782 + }, + { + "epoch": 1.15, + "learning_rate": 1.9328413692479116e-05, + "loss": 0.0797, + "step": 8783 + }, + { + "epoch": 1.15, + "learning_rate": 1.932822260994514e-05, + "loss": 0.0734, + "step": 8784 + }, + { + "epoch": 1.15, + "learning_rate": 1.9328031501176032e-05, + "loss": 0.0843, + "step": 8785 + }, + { + "epoch": 1.15, + "learning_rate": 1.9327840366172334e-05, + "loss": 0.0834, + "step": 8786 + }, + { + "epoch": 1.15, + "learning_rate": 1.932764920493458e-05, + "loss": 0.0766, + "step": 8787 + }, + { + "epoch": 1.15, + "learning_rate": 1.9327458017463308e-05, + "loss": 0.0867, + "step": 8788 + }, + { + "epoch": 1.15, + "learning_rate": 1.932726680375906e-05, + "loss": 0.0825, + "step": 8789 + }, + { + "epoch": 1.15, + "learning_rate": 1.932707556382237e-05, + "loss": 0.0724, + "step": 8790 + }, + { + "epoch": 1.15, + "learning_rate": 1.9326884297653777e-05, + "loss": 0.0799, + "step": 8791 + }, + { + "epoch": 1.15, + "learning_rate": 1.9326693005253817e-05, + "loss": 0.0889, + "step": 8792 + }, + { + "epoch": 1.15, + "learning_rate": 1.932650168662303e-05, + "loss": 0.0799, + "step": 8793 + }, + { + "epoch": 1.15, + "learning_rate": 1.9326310341761957e-05, + "loss": 0.0778, + "step": 8794 + }, + { + "epoch": 1.15, + "learning_rate": 1.932611897067113e-05, + "loss": 0.0822, + "step": 8795 + }, + { + "epoch": 1.15, + "learning_rate": 1.932592757335109e-05, + "loss": 0.0773, + "step": 8796 + }, + { + "epoch": 1.15, + "learning_rate": 1.9325736149802377e-05, + "loss": 0.0816, + "step": 8797 + }, + { + "epoch": 1.15, + "learning_rate": 1.9325544700025524e-05, + "loss": 0.0803, + "step": 8798 + }, + { + "epoch": 1.15, + "learning_rate": 1.9325353224021076e-05, + "loss": 0.0894, + "step": 8799 + }, + { + "epoch": 1.15, + "learning_rate": 1.932516172178957e-05, + "loss": 0.0857, + "step": 8800 + }, + { + "epoch": 1.15, + "learning_rate": 1.932497019333154e-05, + "loss": 0.0774, + "step": 8801 + }, + { + "epoch": 1.15, + "learning_rate": 1.932477863864753e-05, + "loss": 0.079, + "step": 8802 + }, + { + "epoch": 1.15, + "learning_rate": 1.9324587057738074e-05, + "loss": 0.0867, + "step": 8803 + }, + { + "epoch": 1.15, + "learning_rate": 1.9324395450603717e-05, + "loss": 0.0752, + "step": 8804 + }, + { + "epoch": 1.15, + "learning_rate": 1.9324203817244993e-05, + "loss": 0.085, + "step": 8805 + }, + { + "epoch": 1.15, + "learning_rate": 1.9324012157662443e-05, + "loss": 0.0795, + "step": 8806 + }, + { + "epoch": 1.15, + "learning_rate": 1.93238204718566e-05, + "loss": 0.0842, + "step": 8807 + }, + { + "epoch": 1.15, + "learning_rate": 1.9323628759828014e-05, + "loss": 0.0803, + "step": 8808 + }, + { + "epoch": 1.15, + "learning_rate": 1.932343702157722e-05, + "loss": 0.0832, + "step": 8809 + }, + { + "epoch": 1.15, + "learning_rate": 1.9323245257104753e-05, + "loss": 0.0805, + "step": 8810 + }, + { + "epoch": 1.15, + "learning_rate": 1.9323053466411153e-05, + "loss": 0.0744, + "step": 8811 + }, + { + "epoch": 1.15, + "learning_rate": 1.9322861649496967e-05, + "loss": 0.0946, + "step": 8812 + }, + { + "epoch": 1.15, + "learning_rate": 1.9322669806362727e-05, + "loss": 0.0962, + "step": 8813 + }, + { + "epoch": 1.15, + "learning_rate": 1.9322477937008972e-05, + "loss": 0.0943, + "step": 8814 + }, + { + "epoch": 1.15, + "learning_rate": 1.9322286041436247e-05, + "loss": 0.0667, + "step": 8815 + }, + { + "epoch": 1.15, + "learning_rate": 1.9322094119645087e-05, + "loss": 0.065, + "step": 8816 + }, + { + "epoch": 1.15, + "learning_rate": 1.9321902171636035e-05, + "loss": 0.0774, + "step": 8817 + }, + { + "epoch": 1.16, + "learning_rate": 1.932171019740963e-05, + "loss": 0.0889, + "step": 8818 + }, + { + "epoch": 1.16, + "learning_rate": 1.9321518196966413e-05, + "loss": 0.0802, + "step": 8819 + }, + { + "epoch": 1.16, + "learning_rate": 1.9321326170306916e-05, + "loss": 0.0781, + "step": 8820 + }, + { + "epoch": 1.16, + "learning_rate": 1.9321134117431692e-05, + "loss": 0.0959, + "step": 8821 + }, + { + "epoch": 1.16, + "learning_rate": 1.9320942038341273e-05, + "loss": 0.0859, + "step": 8822 + }, + { + "epoch": 1.16, + "learning_rate": 1.93207499330362e-05, + "loss": 0.0725, + "step": 8823 + }, + { + "epoch": 1.16, + "learning_rate": 1.9320557801517014e-05, + "loss": 0.0786, + "step": 8824 + }, + { + "epoch": 1.16, + "learning_rate": 1.9320365643784257e-05, + "loss": 0.0843, + "step": 8825 + }, + { + "epoch": 1.16, + "learning_rate": 1.9320173459838467e-05, + "loss": 0.0767, + "step": 8826 + }, + { + "epoch": 1.16, + "learning_rate": 1.9319981249680186e-05, + "loss": 0.0732, + "step": 8827 + }, + { + "epoch": 1.16, + "learning_rate": 1.9319789013309955e-05, + "loss": 0.0793, + "step": 8828 + }, + { + "epoch": 1.16, + "learning_rate": 1.931959675072831e-05, + "loss": 0.0924, + "step": 8829 + }, + { + "epoch": 1.16, + "learning_rate": 1.9319404461935795e-05, + "loss": 0.0821, + "step": 8830 + }, + { + "epoch": 1.16, + "learning_rate": 1.9319212146932955e-05, + "loss": 0.0872, + "step": 8831 + }, + { + "epoch": 1.16, + "learning_rate": 1.9319019805720326e-05, + "loss": 0.0717, + "step": 8832 + }, + { + "epoch": 1.16, + "learning_rate": 1.931882743829845e-05, + "loss": 0.0729, + "step": 8833 + }, + { + "epoch": 1.16, + "learning_rate": 1.931863504466787e-05, + "loss": 0.0855, + "step": 8834 + }, + { + "epoch": 1.16, + "learning_rate": 1.9318442624829123e-05, + "loss": 0.0802, + "step": 8835 + }, + { + "epoch": 1.16, + "learning_rate": 1.9318250178782754e-05, + "loss": 0.0778, + "step": 8836 + }, + { + "epoch": 1.16, + "learning_rate": 1.93180577065293e-05, + "loss": 0.0891, + "step": 8837 + }, + { + "epoch": 1.16, + "learning_rate": 1.9317865208069308e-05, + "loss": 0.0854, + "step": 8838 + }, + { + "epoch": 1.16, + "learning_rate": 1.9317672683403312e-05, + "loss": 0.0716, + "step": 8839 + }, + { + "epoch": 1.16, + "learning_rate": 1.9317480132531862e-05, + "loss": 0.0802, + "step": 8840 + }, + { + "epoch": 1.16, + "learning_rate": 1.9317287555455492e-05, + "loss": 0.0761, + "step": 8841 + }, + { + "epoch": 1.16, + "learning_rate": 1.931709495217475e-05, + "loss": 0.0917, + "step": 8842 + }, + { + "epoch": 1.16, + "learning_rate": 1.931690232269017e-05, + "loss": 0.0702, + "step": 8843 + }, + { + "epoch": 1.16, + "learning_rate": 1.9316709667002304e-05, + "loss": 0.0754, + "step": 8844 + }, + { + "epoch": 1.16, + "learning_rate": 1.9316516985111685e-05, + "loss": 0.0826, + "step": 8845 + }, + { + "epoch": 1.16, + "learning_rate": 1.931632427701886e-05, + "loss": 0.0895, + "step": 8846 + }, + { + "epoch": 1.16, + "learning_rate": 1.9316131542724368e-05, + "loss": 0.0697, + "step": 8847 + }, + { + "epoch": 1.16, + "learning_rate": 1.931593878222875e-05, + "loss": 0.0931, + "step": 8848 + }, + { + "epoch": 1.16, + "learning_rate": 1.9315745995532556e-05, + "loss": 0.0737, + "step": 8849 + }, + { + "epoch": 1.16, + "learning_rate": 1.931555318263632e-05, + "loss": 0.0829, + "step": 8850 + }, + { + "epoch": 1.16, + "learning_rate": 1.9315360343540584e-05, + "loss": 0.0829, + "step": 8851 + }, + { + "epoch": 1.16, + "learning_rate": 1.9315167478245892e-05, + "loss": 0.0757, + "step": 8852 + }, + { + "epoch": 1.16, + "learning_rate": 1.931497458675279e-05, + "loss": 0.0754, + "step": 8853 + }, + { + "epoch": 1.16, + "learning_rate": 1.931478166906182e-05, + "loss": 0.0813, + "step": 8854 + }, + { + "epoch": 1.16, + "learning_rate": 1.931458872517352e-05, + "loss": 0.0902, + "step": 8855 + }, + { + "epoch": 1.16, + "learning_rate": 1.9314395755088438e-05, + "loss": 0.0783, + "step": 8856 + }, + { + "epoch": 1.16, + "learning_rate": 1.931420275880711e-05, + "loss": 0.07, + "step": 8857 + }, + { + "epoch": 1.16, + "learning_rate": 1.9314009736330088e-05, + "loss": 0.0733, + "step": 8858 + }, + { + "epoch": 1.16, + "learning_rate": 1.9313816687657906e-05, + "loss": 0.0816, + "step": 8859 + }, + { + "epoch": 1.16, + "learning_rate": 1.931362361279111e-05, + "loss": 0.0816, + "step": 8860 + }, + { + "epoch": 1.16, + "learning_rate": 1.9313430511730245e-05, + "loss": 0.0943, + "step": 8861 + }, + { + "epoch": 1.16, + "learning_rate": 1.9313237384475852e-05, + "loss": 0.0869, + "step": 8862 + }, + { + "epoch": 1.16, + "learning_rate": 1.9313044231028473e-05, + "loss": 0.0808, + "step": 8863 + }, + { + "epoch": 1.16, + "learning_rate": 1.9312851051388657e-05, + "loss": 0.0792, + "step": 8864 + }, + { + "epoch": 1.16, + "learning_rate": 1.9312657845556943e-05, + "loss": 0.072, + "step": 8865 + }, + { + "epoch": 1.16, + "learning_rate": 1.931246461353387e-05, + "loss": 0.0895, + "step": 8866 + }, + { + "epoch": 1.16, + "learning_rate": 1.931227135531999e-05, + "loss": 0.0816, + "step": 8867 + }, + { + "epoch": 1.16, + "learning_rate": 1.9312078070915842e-05, + "loss": 0.0736, + "step": 8868 + }, + { + "epoch": 1.16, + "learning_rate": 1.931188476032197e-05, + "loss": 0.0828, + "step": 8869 + }, + { + "epoch": 1.16, + "learning_rate": 1.9311691423538922e-05, + "loss": 0.0816, + "step": 8870 + }, + { + "epoch": 1.16, + "learning_rate": 1.9311498060567234e-05, + "loss": 0.0929, + "step": 8871 + }, + { + "epoch": 1.16, + "learning_rate": 1.9311304671407452e-05, + "loss": 0.078, + "step": 8872 + }, + { + "epoch": 1.16, + "learning_rate": 1.9311111256060126e-05, + "loss": 0.0809, + "step": 8873 + }, + { + "epoch": 1.16, + "learning_rate": 1.931091781452579e-05, + "loss": 0.0853, + "step": 8874 + }, + { + "epoch": 1.16, + "learning_rate": 1.9310724346804996e-05, + "loss": 0.0831, + "step": 8875 + }, + { + "epoch": 1.16, + "learning_rate": 1.9310530852898287e-05, + "loss": 0.0847, + "step": 8876 + }, + { + "epoch": 1.16, + "learning_rate": 1.9310337332806202e-05, + "loss": 0.0737, + "step": 8877 + }, + { + "epoch": 1.16, + "learning_rate": 1.9310143786529293e-05, + "loss": 0.0876, + "step": 8878 + }, + { + "epoch": 1.16, + "learning_rate": 1.93099502140681e-05, + "loss": 0.0834, + "step": 8879 + }, + { + "epoch": 1.16, + "learning_rate": 1.9309756615423164e-05, + "loss": 0.0856, + "step": 8880 + }, + { + "epoch": 1.16, + "learning_rate": 1.9309562990595035e-05, + "loss": 0.0701, + "step": 8881 + }, + { + "epoch": 1.16, + "learning_rate": 1.9309369339584257e-05, + "loss": 0.0817, + "step": 8882 + }, + { + "epoch": 1.16, + "learning_rate": 1.9309175662391376e-05, + "loss": 0.0884, + "step": 8883 + }, + { + "epoch": 1.16, + "learning_rate": 1.9308981959016933e-05, + "loss": 0.076, + "step": 8884 + }, + { + "epoch": 1.16, + "learning_rate": 1.9308788229461474e-05, + "loss": 0.0844, + "step": 8885 + }, + { + "epoch": 1.16, + "learning_rate": 1.9308594473725538e-05, + "loss": 0.0729, + "step": 8886 + }, + { + "epoch": 1.16, + "learning_rate": 1.930840069180968e-05, + "loss": 0.0738, + "step": 8887 + }, + { + "epoch": 1.16, + "learning_rate": 1.9308206883714443e-05, + "loss": 0.0945, + "step": 8888 + }, + { + "epoch": 1.16, + "learning_rate": 1.9308013049440365e-05, + "loss": 0.0753, + "step": 8889 + }, + { + "epoch": 1.16, + "learning_rate": 1.9307819188988003e-05, + "loss": 0.0688, + "step": 8890 + }, + { + "epoch": 1.16, + "learning_rate": 1.930762530235789e-05, + "loss": 0.0826, + "step": 8891 + }, + { + "epoch": 1.16, + "learning_rate": 1.9307431389550573e-05, + "loss": 0.0886, + "step": 8892 + }, + { + "epoch": 1.16, + "learning_rate": 1.9307237450566608e-05, + "loss": 0.081, + "step": 8893 + }, + { + "epoch": 1.17, + "learning_rate": 1.9307043485406533e-05, + "loss": 0.0684, + "step": 8894 + }, + { + "epoch": 1.17, + "learning_rate": 1.930684949407089e-05, + "loss": 0.0858, + "step": 8895 + }, + { + "epoch": 1.17, + "learning_rate": 1.930665547656023e-05, + "loss": 0.0862, + "step": 8896 + }, + { + "epoch": 1.17, + "learning_rate": 1.9306461432875096e-05, + "loss": 0.0938, + "step": 8897 + }, + { + "epoch": 1.17, + "learning_rate": 1.9306267363016036e-05, + "loss": 0.0884, + "step": 8898 + }, + { + "epoch": 1.17, + "learning_rate": 1.9306073266983593e-05, + "loss": 0.0764, + "step": 8899 + }, + { + "epoch": 1.17, + "learning_rate": 1.930587914477832e-05, + "loss": 0.0966, + "step": 8900 + }, + { + "epoch": 1.17, + "learning_rate": 1.930568499640075e-05, + "loss": 0.083, + "step": 8901 + }, + { + "epoch": 1.17, + "learning_rate": 1.9305490821851443e-05, + "loss": 0.0892, + "step": 8902 + }, + { + "epoch": 1.17, + "learning_rate": 1.9305296621130933e-05, + "loss": 0.0652, + "step": 8903 + }, + { + "epoch": 1.17, + "learning_rate": 1.9305102394239777e-05, + "loss": 0.0815, + "step": 8904 + }, + { + "epoch": 1.17, + "learning_rate": 1.9304908141178512e-05, + "loss": 0.0809, + "step": 8905 + }, + { + "epoch": 1.17, + "learning_rate": 1.930471386194769e-05, + "loss": 0.0824, + "step": 8906 + }, + { + "epoch": 1.17, + "learning_rate": 1.9304519556547853e-05, + "loss": 0.0834, + "step": 8907 + }, + { + "epoch": 1.17, + "learning_rate": 1.9304325224979553e-05, + "loss": 0.0933, + "step": 8908 + }, + { + "epoch": 1.17, + "learning_rate": 1.9304130867243332e-05, + "loss": 0.0687, + "step": 8909 + }, + { + "epoch": 1.17, + "learning_rate": 1.9303936483339743e-05, + "loss": 0.0764, + "step": 8910 + }, + { + "epoch": 1.17, + "learning_rate": 1.9303742073269323e-05, + "loss": 0.0712, + "step": 8911 + }, + { + "epoch": 1.17, + "learning_rate": 1.9303547637032627e-05, + "loss": 0.0756, + "step": 8912 + }, + { + "epoch": 1.17, + "learning_rate": 1.9303353174630198e-05, + "loss": 0.076, + "step": 8913 + }, + { + "epoch": 1.17, + "learning_rate": 1.9303158686062583e-05, + "loss": 0.086, + "step": 8914 + }, + { + "epoch": 1.17, + "learning_rate": 1.930296417133033e-05, + "loss": 0.0902, + "step": 8915 + }, + { + "epoch": 1.17, + "learning_rate": 1.9302769630433988e-05, + "loss": 0.0746, + "step": 8916 + }, + { + "epoch": 1.17, + "learning_rate": 1.9302575063374096e-05, + "loss": 0.0886, + "step": 8917 + }, + { + "epoch": 1.17, + "learning_rate": 1.930238047015121e-05, + "loss": 0.0831, + "step": 8918 + }, + { + "epoch": 1.17, + "learning_rate": 1.9302185850765874e-05, + "loss": 0.0833, + "step": 8919 + }, + { + "epoch": 1.17, + "learning_rate": 1.930199120521864e-05, + "loss": 0.0785, + "step": 8920 + }, + { + "epoch": 1.17, + "learning_rate": 1.9301796533510046e-05, + "loss": 0.0844, + "step": 8921 + }, + { + "epoch": 1.17, + "learning_rate": 1.9301601835640643e-05, + "loss": 0.0933, + "step": 8922 + }, + { + "epoch": 1.17, + "learning_rate": 1.9301407111610987e-05, + "loss": 0.0842, + "step": 8923 + }, + { + "epoch": 1.17, + "learning_rate": 1.9301212361421613e-05, + "loss": 0.0807, + "step": 8924 + }, + { + "epoch": 1.17, + "learning_rate": 1.9301017585073077e-05, + "loss": 0.0846, + "step": 8925 + }, + { + "epoch": 1.17, + "learning_rate": 1.9300822782565924e-05, + "loss": 0.0877, + "step": 8926 + }, + { + "epoch": 1.17, + "learning_rate": 1.9300627953900702e-05, + "loss": 0.0803, + "step": 8927 + }, + { + "epoch": 1.17, + "learning_rate": 1.9300433099077958e-05, + "loss": 0.0757, + "step": 8928 + }, + { + "epoch": 1.17, + "learning_rate": 1.9300238218098244e-05, + "loss": 0.0837, + "step": 8929 + }, + { + "epoch": 1.17, + "learning_rate": 1.9300043310962105e-05, + "loss": 0.079, + "step": 8930 + }, + { + "epoch": 1.17, + "learning_rate": 1.9299848377670086e-05, + "loss": 0.0741, + "step": 8931 + }, + { + "epoch": 1.17, + "learning_rate": 1.9299653418222744e-05, + "loss": 0.07, + "step": 8932 + }, + { + "epoch": 1.17, + "learning_rate": 1.9299458432620616e-05, + "loss": 0.0786, + "step": 8933 + }, + { + "epoch": 1.17, + "learning_rate": 1.929926342086426e-05, + "loss": 0.0826, + "step": 8934 + }, + { + "epoch": 1.17, + "learning_rate": 1.929906838295422e-05, + "loss": 0.0817, + "step": 8935 + }, + { + "epoch": 1.17, + "learning_rate": 1.9298873318891047e-05, + "loss": 0.0769, + "step": 8936 + }, + { + "epoch": 1.17, + "learning_rate": 1.9298678228675286e-05, + "loss": 0.0721, + "step": 8937 + }, + { + "epoch": 1.17, + "learning_rate": 1.9298483112307487e-05, + "loss": 0.0823, + "step": 8938 + }, + { + "epoch": 1.17, + "learning_rate": 1.92982879697882e-05, + "loss": 0.0786, + "step": 8939 + }, + { + "epoch": 1.17, + "learning_rate": 1.9298092801117977e-05, + "loss": 0.0858, + "step": 8940 + }, + { + "epoch": 1.17, + "learning_rate": 1.9297897606297356e-05, + "loss": 0.0691, + "step": 8941 + }, + { + "epoch": 1.17, + "learning_rate": 1.92977023853269e-05, + "loss": 0.0821, + "step": 8942 + }, + { + "epoch": 1.17, + "learning_rate": 1.929750713820715e-05, + "loss": 0.0767, + "step": 8943 + }, + { + "epoch": 1.17, + "learning_rate": 1.9297311864938654e-05, + "loss": 0.0794, + "step": 8944 + }, + { + "epoch": 1.17, + "learning_rate": 1.9297116565521963e-05, + "loss": 0.0855, + "step": 8945 + }, + { + "epoch": 1.17, + "learning_rate": 1.9296921239957626e-05, + "loss": 0.0764, + "step": 8946 + }, + { + "epoch": 1.17, + "learning_rate": 1.9296725888246195e-05, + "loss": 0.0778, + "step": 8947 + }, + { + "epoch": 1.17, + "learning_rate": 1.929653051038822e-05, + "loss": 0.0802, + "step": 8948 + }, + { + "epoch": 1.17, + "learning_rate": 1.9296335106384244e-05, + "loss": 0.0804, + "step": 8949 + }, + { + "epoch": 1.17, + "learning_rate": 1.929613967623482e-05, + "loss": 0.0681, + "step": 8950 + }, + { + "epoch": 1.17, + "learning_rate": 1.9295944219940504e-05, + "loss": 0.0772, + "step": 8951 + }, + { + "epoch": 1.17, + "learning_rate": 1.9295748737501836e-05, + "loss": 0.0917, + "step": 8952 + }, + { + "epoch": 1.17, + "learning_rate": 1.929555322891937e-05, + "loss": 0.0812, + "step": 8953 + }, + { + "epoch": 1.17, + "learning_rate": 1.9295357694193652e-05, + "loss": 0.081, + "step": 8954 + }, + { + "epoch": 1.17, + "learning_rate": 1.929516213332524e-05, + "loss": 0.0858, + "step": 8955 + }, + { + "epoch": 1.17, + "learning_rate": 1.929496654631468e-05, + "loss": 0.0716, + "step": 8956 + }, + { + "epoch": 1.17, + "learning_rate": 1.929477093316252e-05, + "loss": 0.0793, + "step": 8957 + }, + { + "epoch": 1.17, + "learning_rate": 1.929457529386931e-05, + "loss": 0.0819, + "step": 8958 + }, + { + "epoch": 1.17, + "learning_rate": 1.9294379628435606e-05, + "loss": 0.0806, + "step": 8959 + }, + { + "epoch": 1.17, + "learning_rate": 1.929418393686195e-05, + "loss": 0.0982, + "step": 8960 + }, + { + "epoch": 1.17, + "learning_rate": 1.92939882191489e-05, + "loss": 0.0838, + "step": 8961 + }, + { + "epoch": 1.17, + "learning_rate": 1.9293792475297e-05, + "loss": 0.0875, + "step": 8962 + }, + { + "epoch": 1.17, + "learning_rate": 1.929359670530681e-05, + "loss": 0.0816, + "step": 8963 + }, + { + "epoch": 1.17, + "learning_rate": 1.9293400909178864e-05, + "loss": 0.0907, + "step": 8964 + }, + { + "epoch": 1.17, + "learning_rate": 1.929320508691373e-05, + "loss": 0.0899, + "step": 8965 + }, + { + "epoch": 1.17, + "learning_rate": 1.929300923851195e-05, + "loss": 0.0859, + "step": 8966 + }, + { + "epoch": 1.17, + "learning_rate": 1.9292813363974074e-05, + "loss": 0.0784, + "step": 8967 + }, + { + "epoch": 1.17, + "learning_rate": 1.9292617463300658e-05, + "loss": 0.0775, + "step": 8968 + }, + { + "epoch": 1.17, + "learning_rate": 1.929242153649225e-05, + "loss": 0.0928, + "step": 8969 + }, + { + "epoch": 1.18, + "learning_rate": 1.9292225583549403e-05, + "loss": 0.0784, + "step": 8970 + }, + { + "epoch": 1.18, + "learning_rate": 1.9292029604472663e-05, + "loss": 0.0883, + "step": 8971 + }, + { + "epoch": 1.18, + "learning_rate": 1.9291833599262587e-05, + "loss": 0.078, + "step": 8972 + }, + { + "epoch": 1.18, + "learning_rate": 1.9291637567919722e-05, + "loss": 0.0805, + "step": 8973 + }, + { + "epoch": 1.18, + "learning_rate": 1.929144151044462e-05, + "loss": 0.0825, + "step": 8974 + }, + { + "epoch": 1.18, + "learning_rate": 1.9291245426837838e-05, + "loss": 0.0827, + "step": 8975 + }, + { + "epoch": 1.18, + "learning_rate": 1.929104931709992e-05, + "loss": 0.0875, + "step": 8976 + }, + { + "epoch": 1.18, + "learning_rate": 1.929085318123142e-05, + "loss": 0.083, + "step": 8977 + }, + { + "epoch": 1.18, + "learning_rate": 1.9290657019232893e-05, + "loss": 0.083, + "step": 8978 + }, + { + "epoch": 1.18, + "learning_rate": 1.9290460831104884e-05, + "loss": 0.0864, + "step": 8979 + }, + { + "epoch": 1.18, + "learning_rate": 1.9290264616847955e-05, + "loss": 0.071, + "step": 8980 + }, + { + "epoch": 1.18, + "learning_rate": 1.9290068376462644e-05, + "loss": 0.0925, + "step": 8981 + }, + { + "epoch": 1.18, + "learning_rate": 1.9289872109949515e-05, + "loss": 0.0752, + "step": 8982 + }, + { + "epoch": 1.18, + "learning_rate": 1.9289675817309112e-05, + "loss": 0.0826, + "step": 8983 + }, + { + "epoch": 1.18, + "learning_rate": 1.9289479498541995e-05, + "loss": 0.0885, + "step": 8984 + }, + { + "epoch": 1.18, + "learning_rate": 1.928928315364871e-05, + "loss": 0.0761, + "step": 8985 + }, + { + "epoch": 1.18, + "learning_rate": 1.928908678262981e-05, + "loss": 0.0759, + "step": 8986 + }, + { + "epoch": 1.18, + "learning_rate": 1.9288890385485847e-05, + "loss": 0.0926, + "step": 8987 + }, + { + "epoch": 1.18, + "learning_rate": 1.9288693962217374e-05, + "loss": 0.0709, + "step": 8988 + }, + { + "epoch": 1.18, + "learning_rate": 1.9288497512824946e-05, + "loss": 0.0877, + "step": 8989 + }, + { + "epoch": 1.18, + "learning_rate": 1.928830103730911e-05, + "loss": 0.0871, + "step": 8990 + }, + { + "epoch": 1.18, + "learning_rate": 1.9288104535670426e-05, + "loss": 0.082, + "step": 8991 + }, + { + "epoch": 1.18, + "learning_rate": 1.9287908007909445e-05, + "loss": 0.08, + "step": 8992 + }, + { + "epoch": 1.18, + "learning_rate": 1.928771145402671e-05, + "loss": 0.0735, + "step": 8993 + }, + { + "epoch": 1.18, + "learning_rate": 1.9287514874022788e-05, + "loss": 0.0837, + "step": 8994 + }, + { + "epoch": 1.18, + "learning_rate": 1.9287318267898222e-05, + "loss": 0.0864, + "step": 8995 + }, + { + "epoch": 1.18, + "learning_rate": 1.9287121635653566e-05, + "loss": 0.0769, + "step": 8996 + }, + { + "epoch": 1.18, + "learning_rate": 1.9286924977289375e-05, + "loss": 0.0807, + "step": 8997 + }, + { + "epoch": 1.18, + "learning_rate": 1.9286728292806208e-05, + "loss": 0.0758, + "step": 8998 + }, + { + "epoch": 1.18, + "learning_rate": 1.9286531582204607e-05, + "loss": 0.094, + "step": 8999 + }, + { + "epoch": 1.18, + "learning_rate": 1.9286334845485132e-05, + "loss": 0.0902, + "step": 9000 + }, + { + "epoch": 1.18, + "learning_rate": 1.9286138082648334e-05, + "loss": 0.0895, + "step": 9001 + }, + { + "epoch": 1.18, + "learning_rate": 1.9285941293694766e-05, + "loss": 0.0952, + "step": 9002 + }, + { + "epoch": 1.18, + "learning_rate": 1.9285744478624987e-05, + "loss": 0.0832, + "step": 9003 + }, + { + "epoch": 1.18, + "learning_rate": 1.9285547637439543e-05, + "loss": 0.0961, + "step": 9004 + }, + { + "epoch": 1.18, + "learning_rate": 1.928535077013899e-05, + "loss": 0.071, + "step": 9005 + }, + { + "epoch": 1.18, + "learning_rate": 1.9285153876723882e-05, + "loss": 0.0752, + "step": 9006 + }, + { + "epoch": 1.18, + "learning_rate": 1.9284956957194775e-05, + "loss": 0.0834, + "step": 9007 + }, + { + "epoch": 1.18, + "learning_rate": 1.928476001155222e-05, + "loss": 0.0899, + "step": 9008 + }, + { + "epoch": 1.18, + "learning_rate": 1.9284563039796773e-05, + "loss": 0.0865, + "step": 9009 + }, + { + "epoch": 1.18, + "learning_rate": 1.928436604192899e-05, + "loss": 0.0928, + "step": 9010 + }, + { + "epoch": 1.18, + "learning_rate": 1.9284169017949414e-05, + "loss": 0.0746, + "step": 9011 + }, + { + "epoch": 1.18, + "learning_rate": 1.9283971967858614e-05, + "loss": 0.091, + "step": 9012 + }, + { + "epoch": 1.18, + "learning_rate": 1.9283774891657135e-05, + "loss": 0.0834, + "step": 9013 + }, + { + "epoch": 1.18, + "learning_rate": 1.928357778934553e-05, + "loss": 0.0707, + "step": 9014 + }, + { + "epoch": 1.18, + "learning_rate": 1.9283380660924363e-05, + "loss": 0.0763, + "step": 9015 + }, + { + "epoch": 1.18, + "learning_rate": 1.9283183506394177e-05, + "loss": 0.0757, + "step": 9016 + }, + { + "epoch": 1.18, + "learning_rate": 1.9282986325755536e-05, + "loss": 0.0949, + "step": 9017 + }, + { + "epoch": 1.18, + "learning_rate": 1.9282789119008987e-05, + "loss": 0.0842, + "step": 9018 + }, + { + "epoch": 1.18, + "learning_rate": 1.9282591886155088e-05, + "loss": 0.0873, + "step": 9019 + }, + { + "epoch": 1.18, + "learning_rate": 1.92823946271944e-05, + "loss": 0.0771, + "step": 9020 + }, + { + "epoch": 1.18, + "learning_rate": 1.928219734212746e-05, + "loss": 0.0801, + "step": 9021 + }, + { + "epoch": 1.18, + "learning_rate": 1.9282000030954846e-05, + "loss": 0.0902, + "step": 9022 + }, + { + "epoch": 1.18, + "learning_rate": 1.9281802693677095e-05, + "loss": 0.0953, + "step": 9023 + }, + { + "epoch": 1.18, + "learning_rate": 1.9281605330294767e-05, + "loss": 0.0712, + "step": 9024 + }, + { + "epoch": 1.18, + "learning_rate": 1.928140794080842e-05, + "loss": 0.0817, + "step": 9025 + }, + { + "epoch": 1.18, + "learning_rate": 1.9281210525218607e-05, + "loss": 0.0759, + "step": 9026 + }, + { + "epoch": 1.18, + "learning_rate": 1.9281013083525888e-05, + "loss": 0.08, + "step": 9027 + }, + { + "epoch": 1.18, + "learning_rate": 1.928081561573081e-05, + "loss": 0.0937, + "step": 9028 + }, + { + "epoch": 1.18, + "learning_rate": 1.928061812183393e-05, + "loss": 0.0814, + "step": 9029 + }, + { + "epoch": 1.18, + "learning_rate": 1.928042060183581e-05, + "loss": 0.0821, + "step": 9030 + }, + { + "epoch": 1.18, + "learning_rate": 1.9280223055737e-05, + "loss": 0.0767, + "step": 9031 + }, + { + "epoch": 1.18, + "learning_rate": 1.9280025483538055e-05, + "loss": 0.0725, + "step": 9032 + }, + { + "epoch": 1.18, + "learning_rate": 1.9279827885239536e-05, + "loss": 0.0723, + "step": 9033 + }, + { + "epoch": 1.18, + "learning_rate": 1.9279630260841993e-05, + "loss": 0.0798, + "step": 9034 + }, + { + "epoch": 1.18, + "learning_rate": 1.9279432610345987e-05, + "loss": 0.0835, + "step": 9035 + }, + { + "epoch": 1.18, + "learning_rate": 1.9279234933752065e-05, + "loss": 0.0908, + "step": 9036 + }, + { + "epoch": 1.18, + "learning_rate": 1.9279037231060794e-05, + "loss": 0.0852, + "step": 9037 + }, + { + "epoch": 1.18, + "learning_rate": 1.9278839502272722e-05, + "loss": 0.0779, + "step": 9038 + }, + { + "epoch": 1.18, + "learning_rate": 1.927864174738841e-05, + "loss": 0.0875, + "step": 9039 + }, + { + "epoch": 1.18, + "learning_rate": 1.927844396640841e-05, + "loss": 0.0818, + "step": 9040 + }, + { + "epoch": 1.18, + "learning_rate": 1.927824615933328e-05, + "loss": 0.0958, + "step": 9041 + }, + { + "epoch": 1.18, + "learning_rate": 1.927804832616358e-05, + "loss": 0.0808, + "step": 9042 + }, + { + "epoch": 1.18, + "learning_rate": 1.9277850466899862e-05, + "loss": 0.0849, + "step": 9043 + }, + { + "epoch": 1.18, + "learning_rate": 1.9277652581542683e-05, + "loss": 0.0897, + "step": 9044 + }, + { + "epoch": 1.18, + "learning_rate": 1.92774546700926e-05, + "loss": 0.0902, + "step": 9045 + }, + { + "epoch": 1.18, + "learning_rate": 1.9277256732550175e-05, + "loss": 0.0832, + "step": 9046 + }, + { + "epoch": 1.19, + "learning_rate": 1.927705876891595e-05, + "loss": 0.0859, + "step": 9047 + }, + { + "epoch": 1.19, + "learning_rate": 1.92768607791905e-05, + "loss": 0.0759, + "step": 9048 + }, + { + "epoch": 1.19, + "learning_rate": 1.9276662763374367e-05, + "loss": 0.085, + "step": 9049 + }, + { + "epoch": 1.19, + "learning_rate": 1.9276464721468118e-05, + "loss": 0.0838, + "step": 9050 + }, + { + "epoch": 1.19, + "learning_rate": 1.9276266653472303e-05, + "loss": 0.0841, + "step": 9051 + }, + { + "epoch": 1.19, + "learning_rate": 1.9276068559387483e-05, + "loss": 0.0748, + "step": 9052 + }, + { + "epoch": 1.19, + "learning_rate": 1.9275870439214214e-05, + "loss": 0.081, + "step": 9053 + }, + { + "epoch": 1.19, + "learning_rate": 1.9275672292953052e-05, + "loss": 0.0875, + "step": 9054 + }, + { + "epoch": 1.19, + "learning_rate": 1.9275474120604557e-05, + "loss": 0.0839, + "step": 9055 + }, + { + "epoch": 1.19, + "learning_rate": 1.9275275922169286e-05, + "loss": 0.0745, + "step": 9056 + }, + { + "epoch": 1.19, + "learning_rate": 1.9275077697647796e-05, + "loss": 0.0795, + "step": 9057 + }, + { + "epoch": 1.19, + "learning_rate": 1.9274879447040642e-05, + "loss": 0.0815, + "step": 9058 + }, + { + "epoch": 1.19, + "learning_rate": 1.9274681170348387e-05, + "loss": 0.0787, + "step": 9059 + }, + { + "epoch": 1.19, + "learning_rate": 1.927448286757158e-05, + "loss": 0.0834, + "step": 9060 + }, + { + "epoch": 1.19, + "learning_rate": 1.9274284538710785e-05, + "loss": 0.0725, + "step": 9061 + }, + { + "epoch": 1.19, + "learning_rate": 1.927408618376656e-05, + "loss": 0.0961, + "step": 9062 + }, + { + "epoch": 1.19, + "learning_rate": 1.9273887802739463e-05, + "loss": 0.0871, + "step": 9063 + }, + { + "epoch": 1.19, + "learning_rate": 1.927368939563005e-05, + "loss": 0.0901, + "step": 9064 + }, + { + "epoch": 1.19, + "learning_rate": 1.9273490962438876e-05, + "loss": 0.0785, + "step": 9065 + }, + { + "epoch": 1.19, + "learning_rate": 1.9273292503166504e-05, + "loss": 0.0907, + "step": 9066 + }, + { + "epoch": 1.19, + "learning_rate": 1.927309401781349e-05, + "loss": 0.0812, + "step": 9067 + }, + { + "epoch": 1.19, + "learning_rate": 1.9272895506380395e-05, + "loss": 0.0813, + "step": 9068 + }, + { + "epoch": 1.19, + "learning_rate": 1.9272696968867776e-05, + "loss": 0.0844, + "step": 9069 + }, + { + "epoch": 1.19, + "learning_rate": 1.9272498405276185e-05, + "loss": 0.0813, + "step": 9070 + }, + { + "epoch": 1.19, + "learning_rate": 1.927229981560619e-05, + "loss": 0.0795, + "step": 9071 + }, + { + "epoch": 1.19, + "learning_rate": 1.927210119985835e-05, + "loss": 0.0838, + "step": 9072 + }, + { + "epoch": 1.19, + "learning_rate": 1.9271902558033213e-05, + "loss": 0.0897, + "step": 9073 + }, + { + "epoch": 1.19, + "learning_rate": 1.9271703890131344e-05, + "loss": 0.0866, + "step": 9074 + }, + { + "epoch": 1.19, + "learning_rate": 1.9271505196153302e-05, + "loss": 0.0906, + "step": 9075 + }, + { + "epoch": 1.19, + "learning_rate": 1.9271306476099648e-05, + "loss": 0.0767, + "step": 9076 + }, + { + "epoch": 1.19, + "learning_rate": 1.9271107729970936e-05, + "loss": 0.0752, + "step": 9077 + }, + { + "epoch": 1.19, + "learning_rate": 1.9270908957767727e-05, + "loss": 0.0667, + "step": 9078 + }, + { + "epoch": 1.19, + "learning_rate": 1.927071015949058e-05, + "loss": 0.0752, + "step": 9079 + }, + { + "epoch": 1.19, + "learning_rate": 1.927051133514006e-05, + "loss": 0.0757, + "step": 9080 + }, + { + "epoch": 1.19, + "learning_rate": 1.9270312484716712e-05, + "loss": 0.0896, + "step": 9081 + }, + { + "epoch": 1.19, + "learning_rate": 1.9270113608221108e-05, + "loss": 0.0754, + "step": 9082 + }, + { + "epoch": 1.19, + "learning_rate": 1.9269914705653802e-05, + "loss": 0.0854, + "step": 9083 + }, + { + "epoch": 1.19, + "learning_rate": 1.9269715777015358e-05, + "loss": 0.0726, + "step": 9084 + }, + { + "epoch": 1.19, + "learning_rate": 1.926951682230633e-05, + "loss": 0.0818, + "step": 9085 + }, + { + "epoch": 1.19, + "learning_rate": 1.926931784152728e-05, + "loss": 0.091, + "step": 9086 + }, + { + "epoch": 1.19, + "learning_rate": 1.9269118834678762e-05, + "loss": 0.0741, + "step": 9087 + }, + { + "epoch": 1.19, + "learning_rate": 1.9268919801761347e-05, + "loss": 0.0799, + "step": 9088 + }, + { + "epoch": 1.19, + "learning_rate": 1.9268720742775587e-05, + "loss": 0.0715, + "step": 9089 + }, + { + "epoch": 1.19, + "learning_rate": 1.9268521657722043e-05, + "loss": 0.0668, + "step": 9090 + }, + { + "epoch": 1.19, + "learning_rate": 1.9268322546601274e-05, + "loss": 0.0866, + "step": 9091 + }, + { + "epoch": 1.19, + "learning_rate": 1.9268123409413845e-05, + "loss": 0.078, + "step": 9092 + }, + { + "epoch": 1.19, + "learning_rate": 1.926792424616031e-05, + "loss": 0.0768, + "step": 9093 + }, + { + "epoch": 1.19, + "learning_rate": 1.9267725056841234e-05, + "loss": 0.0877, + "step": 9094 + }, + { + "epoch": 1.19, + "learning_rate": 1.9267525841457173e-05, + "loss": 0.0863, + "step": 9095 + }, + { + "epoch": 1.19, + "learning_rate": 1.9267326600008686e-05, + "loss": 0.0807, + "step": 9096 + }, + { + "epoch": 1.19, + "learning_rate": 1.9267127332496344e-05, + "loss": 0.0762, + "step": 9097 + }, + { + "epoch": 1.19, + "learning_rate": 1.9266928038920693e-05, + "loss": 0.0902, + "step": 9098 + }, + { + "epoch": 1.19, + "learning_rate": 1.9266728719282303e-05, + "loss": 0.0793, + "step": 9099 + }, + { + "epoch": 1.19, + "learning_rate": 1.9266529373581732e-05, + "loss": 0.0858, + "step": 9100 + }, + { + "epoch": 1.19, + "learning_rate": 1.926633000181954e-05, + "loss": 0.0791, + "step": 9101 + }, + { + "epoch": 1.19, + "learning_rate": 1.9266130603996288e-05, + "loss": 0.0855, + "step": 9102 + }, + { + "epoch": 1.19, + "learning_rate": 1.926593118011254e-05, + "loss": 0.0792, + "step": 9103 + }, + { + "epoch": 1.19, + "learning_rate": 1.9265731730168847e-05, + "loss": 0.0725, + "step": 9104 + }, + { + "epoch": 1.19, + "learning_rate": 1.9265532254165783e-05, + "loss": 0.0824, + "step": 9105 + }, + { + "epoch": 1.19, + "learning_rate": 1.9265332752103902e-05, + "loss": 0.0801, + "step": 9106 + }, + { + "epoch": 1.19, + "learning_rate": 1.9265133223983768e-05, + "loss": 0.0854, + "step": 9107 + }, + { + "epoch": 1.19, + "learning_rate": 1.926493366980594e-05, + "loss": 0.0815, + "step": 9108 + }, + { + "epoch": 1.19, + "learning_rate": 1.9264734089570975e-05, + "loss": 0.0849, + "step": 9109 + }, + { + "epoch": 1.19, + "learning_rate": 1.9264534483279444e-05, + "loss": 0.0822, + "step": 9110 + }, + { + "epoch": 1.19, + "learning_rate": 1.9264334850931895e-05, + "loss": 0.0799, + "step": 9111 + }, + { + "epoch": 1.19, + "learning_rate": 1.9264135192528905e-05, + "loss": 0.0763, + "step": 9112 + }, + { + "epoch": 1.19, + "learning_rate": 1.9263935508071025e-05, + "loss": 0.089, + "step": 9113 + }, + { + "epoch": 1.19, + "learning_rate": 1.926373579755882e-05, + "loss": 0.084, + "step": 9114 + }, + { + "epoch": 1.19, + "learning_rate": 1.9263536060992855e-05, + "loss": 0.0854, + "step": 9115 + }, + { + "epoch": 1.19, + "learning_rate": 1.9263336298373682e-05, + "loss": 0.0765, + "step": 9116 + }, + { + "epoch": 1.19, + "learning_rate": 1.9263136509701873e-05, + "loss": 0.0765, + "step": 9117 + }, + { + "epoch": 1.19, + "learning_rate": 1.9262936694977984e-05, + "loss": 0.0885, + "step": 9118 + }, + { + "epoch": 1.19, + "learning_rate": 1.9262736854202583e-05, + "loss": 0.0771, + "step": 9119 + }, + { + "epoch": 1.19, + "learning_rate": 1.9262536987376223e-05, + "loss": 0.0797, + "step": 9120 + }, + { + "epoch": 1.19, + "learning_rate": 1.9262337094499473e-05, + "loss": 0.079, + "step": 9121 + }, + { + "epoch": 1.19, + "learning_rate": 1.9262137175572895e-05, + "loss": 0.0857, + "step": 9122 + }, + { + "epoch": 1.2, + "learning_rate": 1.9261937230597046e-05, + "loss": 0.0955, + "step": 9123 + }, + { + "epoch": 1.2, + "learning_rate": 1.926173725957249e-05, + "loss": 0.0724, + "step": 9124 + }, + { + "epoch": 1.2, + "learning_rate": 1.9261537262499796e-05, + "loss": 0.0785, + "step": 9125 + }, + { + "epoch": 1.2, + "learning_rate": 1.926133723937952e-05, + "loss": 0.095, + "step": 9126 + }, + { + "epoch": 1.2, + "learning_rate": 1.9261137190212228e-05, + "loss": 0.0744, + "step": 9127 + }, + { + "epoch": 1.2, + "learning_rate": 1.926093711499848e-05, + "loss": 0.0843, + "step": 9128 + }, + { + "epoch": 1.2, + "learning_rate": 1.926073701373884e-05, + "loss": 0.0857, + "step": 9129 + }, + { + "epoch": 1.2, + "learning_rate": 1.9260536886433865e-05, + "loss": 0.0778, + "step": 9130 + }, + { + "epoch": 1.2, + "learning_rate": 1.9260336733084127e-05, + "loss": 0.0863, + "step": 9131 + }, + { + "epoch": 1.2, + "learning_rate": 1.9260136553690185e-05, + "loss": 0.0815, + "step": 9132 + }, + { + "epoch": 1.2, + "learning_rate": 1.9259936348252602e-05, + "loss": 0.0852, + "step": 9133 + }, + { + "epoch": 1.2, + "learning_rate": 1.9259736116771945e-05, + "loss": 0.0811, + "step": 9134 + }, + { + "epoch": 1.2, + "learning_rate": 1.9259535859248765e-05, + "loss": 0.0839, + "step": 9135 + }, + { + "epoch": 1.2, + "learning_rate": 1.9259335575683638e-05, + "loss": 0.0781, + "step": 9136 + }, + { + "epoch": 1.2, + "learning_rate": 1.9259135266077124e-05, + "loss": 0.0793, + "step": 9137 + }, + { + "epoch": 1.2, + "learning_rate": 1.9258934930429784e-05, + "loss": 0.0903, + "step": 9138 + }, + { + "epoch": 1.2, + "learning_rate": 1.9258734568742182e-05, + "loss": 0.0991, + "step": 9139 + }, + { + "epoch": 1.2, + "learning_rate": 1.9258534181014883e-05, + "loss": 0.0806, + "step": 9140 + }, + { + "epoch": 1.2, + "learning_rate": 1.925833376724845e-05, + "loss": 0.0719, + "step": 9141 + }, + { + "epoch": 1.2, + "learning_rate": 1.9258133327443444e-05, + "loss": 0.08, + "step": 9142 + }, + { + "epoch": 1.2, + "learning_rate": 1.9257932861600435e-05, + "loss": 0.0846, + "step": 9143 + }, + { + "epoch": 1.2, + "learning_rate": 1.925773236971998e-05, + "loss": 0.0738, + "step": 9144 + }, + { + "epoch": 1.2, + "learning_rate": 1.9257531851802645e-05, + "loss": 0.0872, + "step": 9145 + }, + { + "epoch": 1.2, + "learning_rate": 1.9257331307848996e-05, + "loss": 0.0698, + "step": 9146 + }, + { + "epoch": 1.2, + "learning_rate": 1.9257130737859597e-05, + "loss": 0.0927, + "step": 9147 + }, + { + "epoch": 1.2, + "learning_rate": 1.9256930141835007e-05, + "loss": 0.0798, + "step": 9148 + }, + { + "epoch": 1.2, + "learning_rate": 1.9256729519775796e-05, + "loss": 0.0858, + "step": 9149 + }, + { + "epoch": 1.2, + "learning_rate": 1.925652887168253e-05, + "loss": 0.0951, + "step": 9150 + }, + { + "epoch": 1.2, + "learning_rate": 1.925632819755576e-05, + "loss": 0.0818, + "step": 9151 + }, + { + "epoch": 1.2, + "learning_rate": 1.925612749739607e-05, + "loss": 0.0856, + "step": 9152 + }, + { + "epoch": 1.2, + "learning_rate": 1.925592677120401e-05, + "loss": 0.0862, + "step": 9153 + }, + { + "epoch": 1.2, + "learning_rate": 1.925572601898015e-05, + "loss": 0.0816, + "step": 9154 + }, + { + "epoch": 1.2, + "learning_rate": 1.9255525240725053e-05, + "loss": 0.0832, + "step": 9155 + }, + { + "epoch": 1.2, + "learning_rate": 1.925532443643928e-05, + "loss": 0.0819, + "step": 9156 + }, + { + "epoch": 1.2, + "learning_rate": 1.9255123606123405e-05, + "loss": 0.0809, + "step": 9157 + }, + { + "epoch": 1.2, + "learning_rate": 1.9254922749777987e-05, + "loss": 0.0816, + "step": 9158 + }, + { + "epoch": 1.2, + "learning_rate": 1.925472186740359e-05, + "loss": 0.0958, + "step": 9159 + }, + { + "epoch": 1.2, + "learning_rate": 1.925452095900078e-05, + "loss": 0.0853, + "step": 9160 + }, + { + "epoch": 1.2, + "learning_rate": 1.9254320024570126e-05, + "loss": 0.0849, + "step": 9161 + }, + { + "epoch": 1.2, + "learning_rate": 1.9254119064112186e-05, + "loss": 0.0856, + "step": 9162 + }, + { + "epoch": 1.2, + "learning_rate": 1.9253918077627532e-05, + "loss": 0.0867, + "step": 9163 + }, + { + "epoch": 1.2, + "learning_rate": 1.9253717065116727e-05, + "loss": 0.091, + "step": 9164 + }, + { + "epoch": 1.2, + "learning_rate": 1.925351602658033e-05, + "loss": 0.0792, + "step": 9165 + }, + { + "epoch": 1.2, + "learning_rate": 1.9253314962018916e-05, + "loss": 0.0774, + "step": 9166 + }, + { + "epoch": 1.2, + "learning_rate": 1.9253113871433045e-05, + "loss": 0.0891, + "step": 9167 + }, + { + "epoch": 1.2, + "learning_rate": 1.9252912754823287e-05, + "loss": 0.072, + "step": 9168 + }, + { + "epoch": 1.2, + "learning_rate": 1.9252711612190203e-05, + "loss": 0.0885, + "step": 9169 + }, + { + "epoch": 1.2, + "learning_rate": 1.925251044353436e-05, + "loss": 0.1035, + "step": 9170 + }, + { + "epoch": 1.2, + "learning_rate": 1.925230924885632e-05, + "loss": 0.0937, + "step": 9171 + }, + { + "epoch": 1.2, + "learning_rate": 1.925210802815666e-05, + "loss": 0.0818, + "step": 9172 + }, + { + "epoch": 1.2, + "learning_rate": 1.9251906781435935e-05, + "loss": 0.074, + "step": 9173 + }, + { + "epoch": 1.2, + "learning_rate": 1.9251705508694714e-05, + "loss": 0.091, + "step": 9174 + }, + { + "epoch": 1.2, + "learning_rate": 1.9251504209933566e-05, + "loss": 0.0783, + "step": 9175 + }, + { + "epoch": 1.2, + "learning_rate": 1.9251302885153057e-05, + "loss": 0.0804, + "step": 9176 + }, + { + "epoch": 1.2, + "learning_rate": 1.925110153435375e-05, + "loss": 0.0942, + "step": 9177 + }, + { + "epoch": 1.2, + "learning_rate": 1.925090015753621e-05, + "loss": 0.076, + "step": 9178 + }, + { + "epoch": 1.2, + "learning_rate": 1.9250698754701006e-05, + "loss": 0.0872, + "step": 9179 + }, + { + "epoch": 1.2, + "learning_rate": 1.9250497325848706e-05, + "loss": 0.0874, + "step": 9180 + }, + { + "epoch": 1.2, + "learning_rate": 1.9250295870979877e-05, + "loss": 0.0894, + "step": 9181 + }, + { + "epoch": 1.2, + "learning_rate": 1.925009439009508e-05, + "loss": 0.0831, + "step": 9182 + }, + { + "epoch": 1.2, + "learning_rate": 1.9249892883194886e-05, + "loss": 0.087, + "step": 9183 + }, + { + "epoch": 1.2, + "learning_rate": 1.9249691350279862e-05, + "loss": 0.087, + "step": 9184 + }, + { + "epoch": 1.2, + "learning_rate": 1.9249489791350573e-05, + "loss": 0.0819, + "step": 9185 + }, + { + "epoch": 1.2, + "learning_rate": 1.9249288206407587e-05, + "loss": 0.0681, + "step": 9186 + }, + { + "epoch": 1.2, + "learning_rate": 1.924908659545147e-05, + "loss": 0.0974, + "step": 9187 + }, + { + "epoch": 1.2, + "learning_rate": 1.924888495848279e-05, + "loss": 0.088, + "step": 9188 + }, + { + "epoch": 1.2, + "learning_rate": 1.9248683295502112e-05, + "loss": 0.0795, + "step": 9189 + }, + { + "epoch": 1.2, + "learning_rate": 1.9248481606510005e-05, + "loss": 0.0895, + "step": 9190 + }, + { + "epoch": 1.2, + "learning_rate": 1.9248279891507038e-05, + "loss": 0.0911, + "step": 9191 + }, + { + "epoch": 1.2, + "learning_rate": 1.9248078150493773e-05, + "loss": 0.0737, + "step": 9192 + }, + { + "epoch": 1.2, + "learning_rate": 1.9247876383470782e-05, + "loss": 0.0895, + "step": 9193 + }, + { + "epoch": 1.2, + "learning_rate": 1.924767459043863e-05, + "loss": 0.09, + "step": 9194 + }, + { + "epoch": 1.2, + "learning_rate": 1.924747277139789e-05, + "loss": 0.075, + "step": 9195 + }, + { + "epoch": 1.2, + "learning_rate": 1.9247270926349122e-05, + "loss": 0.0799, + "step": 9196 + }, + { + "epoch": 1.2, + "learning_rate": 1.92470690552929e-05, + "loss": 0.0878, + "step": 9197 + }, + { + "epoch": 1.2, + "learning_rate": 1.9246867158229786e-05, + "loss": 0.0858, + "step": 9198 + }, + { + "epoch": 1.21, + "learning_rate": 1.9246665235160352e-05, + "loss": 0.0806, + "step": 9199 + }, + { + "epoch": 1.21, + "learning_rate": 1.9246463286085163e-05, + "loss": 0.0724, + "step": 9200 + }, + { + "epoch": 1.21, + "learning_rate": 1.9246261311004788e-05, + "loss": 0.0892, + "step": 9201 + }, + { + "epoch": 1.21, + "learning_rate": 1.9246059309919797e-05, + "loss": 0.0799, + "step": 9202 + }, + { + "epoch": 1.21, + "learning_rate": 1.9245857282830756e-05, + "loss": 0.0759, + "step": 9203 + }, + { + "epoch": 1.21, + "learning_rate": 1.9245655229738234e-05, + "loss": 0.0886, + "step": 9204 + }, + { + "epoch": 1.21, + "learning_rate": 1.92454531506428e-05, + "loss": 0.0726, + "step": 9205 + }, + { + "epoch": 1.21, + "learning_rate": 1.924525104554502e-05, + "loss": 0.0816, + "step": 9206 + }, + { + "epoch": 1.21, + "learning_rate": 1.924504891444546e-05, + "loss": 0.0862, + "step": 9207 + }, + { + "epoch": 1.21, + "learning_rate": 1.9244846757344696e-05, + "loss": 0.0841, + "step": 9208 + }, + { + "epoch": 1.21, + "learning_rate": 1.9244644574243295e-05, + "loss": 0.0769, + "step": 9209 + }, + { + "epoch": 1.21, + "learning_rate": 1.924444236514182e-05, + "loss": 0.0825, + "step": 9210 + }, + { + "epoch": 1.21, + "learning_rate": 1.9244240130040846e-05, + "loss": 0.0776, + "step": 9211 + }, + { + "epoch": 1.21, + "learning_rate": 1.9244037868940938e-05, + "loss": 0.0713, + "step": 9212 + }, + { + "epoch": 1.21, + "learning_rate": 1.9243835581842662e-05, + "loss": 0.0709, + "step": 9213 + }, + { + "epoch": 1.21, + "learning_rate": 1.9243633268746597e-05, + "loss": 0.0786, + "step": 9214 + }, + { + "epoch": 1.21, + "learning_rate": 1.92434309296533e-05, + "loss": 0.0699, + "step": 9215 + }, + { + "epoch": 1.21, + "learning_rate": 1.924322856456335e-05, + "loss": 0.0896, + "step": 9216 + }, + { + "epoch": 1.21, + "learning_rate": 1.9243026173477308e-05, + "loss": 0.0854, + "step": 9217 + }, + { + "epoch": 1.21, + "learning_rate": 1.924282375639575e-05, + "loss": 0.0806, + "step": 9218 + }, + { + "epoch": 1.21, + "learning_rate": 1.9242621313319244e-05, + "loss": 0.0828, + "step": 9219 + }, + { + "epoch": 1.21, + "learning_rate": 1.9242418844248353e-05, + "loss": 0.0847, + "step": 9220 + }, + { + "epoch": 1.21, + "learning_rate": 1.9242216349183655e-05, + "loss": 0.0704, + "step": 9221 + }, + { + "epoch": 1.21, + "learning_rate": 1.9242013828125714e-05, + "loss": 0.0753, + "step": 9222 + }, + { + "epoch": 1.21, + "learning_rate": 1.92418112810751e-05, + "loss": 0.0713, + "step": 9223 + }, + { + "epoch": 1.21, + "learning_rate": 1.9241608708032385e-05, + "loss": 0.0906, + "step": 9224 + }, + { + "epoch": 1.21, + "learning_rate": 1.924140610899814e-05, + "loss": 0.0704, + "step": 9225 + }, + { + "epoch": 1.21, + "learning_rate": 1.924120348397293e-05, + "loss": 0.0881, + "step": 9226 + }, + { + "epoch": 1.21, + "learning_rate": 1.9241000832957327e-05, + "loss": 0.0795, + "step": 9227 + }, + { + "epoch": 1.21, + "learning_rate": 1.9240798155951906e-05, + "loss": 0.0809, + "step": 9228 + }, + { + "epoch": 1.21, + "learning_rate": 1.9240595452957228e-05, + "loss": 0.076, + "step": 9229 + }, + { + "epoch": 1.21, + "learning_rate": 1.9240392723973867e-05, + "loss": 0.0883, + "step": 9230 + }, + { + "epoch": 1.21, + "learning_rate": 1.9240189969002394e-05, + "loss": 0.0731, + "step": 9231 + }, + { + "epoch": 1.21, + "learning_rate": 1.9239987188043378e-05, + "loss": 0.0788, + "step": 9232 + }, + { + "epoch": 1.21, + "learning_rate": 1.9239784381097394e-05, + "loss": 0.0819, + "step": 9233 + }, + { + "epoch": 1.21, + "learning_rate": 1.9239581548165005e-05, + "loss": 0.0759, + "step": 9234 + }, + { + "epoch": 1.21, + "learning_rate": 1.9239378689246787e-05, + "loss": 0.0894, + "step": 9235 + }, + { + "epoch": 1.21, + "learning_rate": 1.9239175804343305e-05, + "loss": 0.0837, + "step": 9236 + }, + { + "epoch": 1.21, + "learning_rate": 1.9238972893455137e-05, + "loss": 0.0812, + "step": 9237 + }, + { + "epoch": 1.21, + "learning_rate": 1.9238769956582847e-05, + "loss": 0.083, + "step": 9238 + }, + { + "epoch": 1.21, + "learning_rate": 1.923856699372701e-05, + "loss": 0.0782, + "step": 9239 + }, + { + "epoch": 1.21, + "learning_rate": 1.9238364004888193e-05, + "loss": 0.0723, + "step": 9240 + }, + { + "epoch": 1.21, + "learning_rate": 1.923816099006697e-05, + "loss": 0.086, + "step": 9241 + }, + { + "epoch": 1.21, + "learning_rate": 1.9237957949263912e-05, + "loss": 0.0879, + "step": 9242 + }, + { + "epoch": 1.21, + "learning_rate": 1.923775488247959e-05, + "loss": 0.0815, + "step": 9243 + }, + { + "epoch": 1.21, + "learning_rate": 1.9237551789714576e-05, + "loss": 0.0829, + "step": 9244 + }, + { + "epoch": 1.21, + "learning_rate": 1.9237348670969435e-05, + "loss": 0.064, + "step": 9245 + }, + { + "epoch": 1.21, + "learning_rate": 1.9237145526244747e-05, + "loss": 0.0812, + "step": 9246 + }, + { + "epoch": 1.21, + "learning_rate": 1.9236942355541077e-05, + "loss": 0.0796, + "step": 9247 + }, + { + "epoch": 1.21, + "learning_rate": 1.9236739158858998e-05, + "loss": 0.0774, + "step": 9248 + }, + { + "epoch": 1.21, + "learning_rate": 1.9236535936199083e-05, + "loss": 0.0784, + "step": 9249 + }, + { + "epoch": 1.21, + "learning_rate": 1.92363326875619e-05, + "loss": 0.0804, + "step": 9250 + }, + { + "epoch": 1.21, + "learning_rate": 1.9236129412948026e-05, + "loss": 0.0831, + "step": 9251 + }, + { + "epoch": 1.21, + "learning_rate": 1.923592611235803e-05, + "loss": 0.0708, + "step": 9252 + }, + { + "epoch": 1.21, + "learning_rate": 1.9235722785792482e-05, + "loss": 0.0749, + "step": 9253 + }, + { + "epoch": 1.21, + "learning_rate": 1.9235519433251956e-05, + "loss": 0.0789, + "step": 9254 + }, + { + "epoch": 1.21, + "learning_rate": 1.9235316054737027e-05, + "loss": 0.0819, + "step": 9255 + }, + { + "epoch": 1.21, + "learning_rate": 1.923511265024826e-05, + "loss": 0.0817, + "step": 9256 + }, + { + "epoch": 1.21, + "learning_rate": 1.923490921978623e-05, + "loss": 0.0732, + "step": 9257 + }, + { + "epoch": 1.21, + "learning_rate": 1.9234705763351507e-05, + "loss": 0.0739, + "step": 9258 + }, + { + "epoch": 1.21, + "learning_rate": 1.923450228094467e-05, + "loss": 0.0747, + "step": 9259 + }, + { + "epoch": 1.21, + "learning_rate": 1.9234298772566287e-05, + "loss": 0.0734, + "step": 9260 + }, + { + "epoch": 1.21, + "learning_rate": 1.9234095238216933e-05, + "loss": 0.0886, + "step": 9261 + }, + { + "epoch": 1.21, + "learning_rate": 1.9233891677897172e-05, + "loss": 0.078, + "step": 9262 + }, + { + "epoch": 1.21, + "learning_rate": 1.9233688091607585e-05, + "loss": 0.0816, + "step": 9263 + }, + { + "epoch": 1.21, + "learning_rate": 1.923348447934874e-05, + "loss": 0.0814, + "step": 9264 + }, + { + "epoch": 1.21, + "learning_rate": 1.9233280841121216e-05, + "loss": 0.0704, + "step": 9265 + }, + { + "epoch": 1.21, + "learning_rate": 1.923307717692558e-05, + "loss": 0.0725, + "step": 9266 + }, + { + "epoch": 1.21, + "learning_rate": 1.92328734867624e-05, + "loss": 0.0826, + "step": 9267 + }, + { + "epoch": 1.21, + "learning_rate": 1.9232669770632264e-05, + "loss": 0.0853, + "step": 9268 + }, + { + "epoch": 1.21, + "learning_rate": 1.9232466028535733e-05, + "loss": 0.0862, + "step": 9269 + }, + { + "epoch": 1.21, + "learning_rate": 1.923226226047338e-05, + "loss": 0.0727, + "step": 9270 + }, + { + "epoch": 1.21, + "learning_rate": 1.9232058466445784e-05, + "loss": 0.0798, + "step": 9271 + }, + { + "epoch": 1.21, + "learning_rate": 1.9231854646453515e-05, + "loss": 0.0821, + "step": 9272 + }, + { + "epoch": 1.21, + "learning_rate": 1.9231650800497144e-05, + "loss": 0.0843, + "step": 9273 + }, + { + "epoch": 1.21, + "learning_rate": 1.923144692857725e-05, + "loss": 0.081, + "step": 9274 + }, + { + "epoch": 1.21, + "learning_rate": 1.92312430306944e-05, + "loss": 0.0756, + "step": 9275 + }, + { + "epoch": 1.22, + "learning_rate": 1.923103910684917e-05, + "loss": 0.0818, + "step": 9276 + }, + { + "epoch": 1.22, + "learning_rate": 1.923083515704214e-05, + "loss": 0.0785, + "step": 9277 + }, + { + "epoch": 1.22, + "learning_rate": 1.923063118127387e-05, + "loss": 0.0984, + "step": 9278 + }, + { + "epoch": 1.22, + "learning_rate": 1.923042717954495e-05, + "loss": 0.0757, + "step": 9279 + }, + { + "epoch": 1.22, + "learning_rate": 1.923022315185594e-05, + "loss": 0.0931, + "step": 9280 + }, + { + "epoch": 1.22, + "learning_rate": 1.9230019098207418e-05, + "loss": 0.0891, + "step": 9281 + }, + { + "epoch": 1.22, + "learning_rate": 1.9229815018599957e-05, + "loss": 0.0774, + "step": 9282 + }, + { + "epoch": 1.22, + "learning_rate": 1.9229610913034138e-05, + "loss": 0.0852, + "step": 9283 + }, + { + "epoch": 1.22, + "learning_rate": 1.9229406781510526e-05, + "loss": 0.0781, + "step": 9284 + }, + { + "epoch": 1.22, + "learning_rate": 1.9229202624029702e-05, + "loss": 0.073, + "step": 9285 + }, + { + "epoch": 1.22, + "learning_rate": 1.9228998440592234e-05, + "loss": 0.078, + "step": 9286 + }, + { + "epoch": 1.22, + "learning_rate": 1.92287942311987e-05, + "loss": 0.0776, + "step": 9287 + }, + { + "epoch": 1.22, + "learning_rate": 1.9228589995849674e-05, + "loss": 0.0876, + "step": 9288 + }, + { + "epoch": 1.22, + "learning_rate": 1.9228385734545735e-05, + "loss": 0.0851, + "step": 9289 + }, + { + "epoch": 1.22, + "learning_rate": 1.9228181447287445e-05, + "loss": 0.0938, + "step": 9290 + }, + { + "epoch": 1.22, + "learning_rate": 1.922797713407539e-05, + "loss": 0.0905, + "step": 9291 + }, + { + "epoch": 1.22, + "learning_rate": 1.922777279491014e-05, + "loss": 0.0814, + "step": 9292 + }, + { + "epoch": 1.22, + "learning_rate": 1.922756842979227e-05, + "loss": 0.091, + "step": 9293 + }, + { + "epoch": 1.22, + "learning_rate": 1.9227364038722358e-05, + "loss": 0.0788, + "step": 9294 + }, + { + "epoch": 1.22, + "learning_rate": 1.9227159621700973e-05, + "loss": 0.0752, + "step": 9295 + }, + { + "epoch": 1.22, + "learning_rate": 1.9226955178728693e-05, + "loss": 0.0877, + "step": 9296 + }, + { + "epoch": 1.22, + "learning_rate": 1.9226750709806095e-05, + "loss": 0.0825, + "step": 9297 + }, + { + "epoch": 1.22, + "learning_rate": 1.922654621493375e-05, + "loss": 0.0823, + "step": 9298 + }, + { + "epoch": 1.22, + "learning_rate": 1.9226341694112236e-05, + "loss": 0.0754, + "step": 9299 + }, + { + "epoch": 1.22, + "learning_rate": 1.9226137147342127e-05, + "loss": 0.0872, + "step": 9300 + }, + { + "epoch": 1.22, + "learning_rate": 1.9225932574623996e-05, + "loss": 0.0762, + "step": 9301 + }, + { + "epoch": 1.22, + "learning_rate": 1.9225727975958425e-05, + "loss": 0.0835, + "step": 9302 + }, + { + "epoch": 1.22, + "learning_rate": 1.9225523351345984e-05, + "loss": 0.0748, + "step": 9303 + }, + { + "epoch": 1.22, + "learning_rate": 1.922531870078725e-05, + "loss": 0.0732, + "step": 9304 + }, + { + "epoch": 1.22, + "learning_rate": 1.92251140242828e-05, + "loss": 0.0785, + "step": 9305 + }, + { + "epoch": 1.22, + "learning_rate": 1.9224909321833205e-05, + "loss": 0.0754, + "step": 9306 + }, + { + "epoch": 1.22, + "learning_rate": 1.9224704593439045e-05, + "loss": 0.0817, + "step": 9307 + }, + { + "epoch": 1.22, + "learning_rate": 1.9224499839100893e-05, + "loss": 0.0688, + "step": 9308 + }, + { + "epoch": 1.22, + "learning_rate": 1.922429505881933e-05, + "loss": 0.0775, + "step": 9309 + }, + { + "epoch": 1.22, + "learning_rate": 1.9224090252594926e-05, + "loss": 0.0818, + "step": 9310 + }, + { + "epoch": 1.22, + "learning_rate": 1.9223885420428262e-05, + "loss": 0.0788, + "step": 9311 + }, + { + "epoch": 1.22, + "learning_rate": 1.9223680562319907e-05, + "loss": 0.0943, + "step": 9312 + }, + { + "epoch": 1.22, + "learning_rate": 1.9223475678270443e-05, + "loss": 0.0874, + "step": 9313 + }, + { + "epoch": 1.22, + "learning_rate": 1.9223270768280446e-05, + "loss": 0.0801, + "step": 9314 + }, + { + "epoch": 1.22, + "learning_rate": 1.922306583235049e-05, + "loss": 0.0835, + "step": 9315 + }, + { + "epoch": 1.22, + "learning_rate": 1.9222860870481156e-05, + "loss": 0.0847, + "step": 9316 + }, + { + "epoch": 1.22, + "learning_rate": 1.9222655882673012e-05, + "loss": 0.084, + "step": 9317 + }, + { + "epoch": 1.22, + "learning_rate": 1.9222450868926646e-05, + "loss": 0.0874, + "step": 9318 + }, + { + "epoch": 1.22, + "learning_rate": 1.9222245829242623e-05, + "loss": 0.0877, + "step": 9319 + }, + { + "epoch": 1.22, + "learning_rate": 1.9222040763621526e-05, + "loss": 0.0816, + "step": 9320 + }, + { + "epoch": 1.22, + "learning_rate": 1.9221835672063928e-05, + "loss": 0.0862, + "step": 9321 + }, + { + "epoch": 1.22, + "learning_rate": 1.9221630554570412e-05, + "loss": 0.0755, + "step": 9322 + }, + { + "epoch": 1.22, + "learning_rate": 1.9221425411141546e-05, + "loss": 0.0818, + "step": 9323 + }, + { + "epoch": 1.22, + "learning_rate": 1.9221220241777918e-05, + "loss": 0.0781, + "step": 9324 + }, + { + "epoch": 1.22, + "learning_rate": 1.9221015046480095e-05, + "loss": 0.0846, + "step": 9325 + }, + { + "epoch": 1.22, + "learning_rate": 1.922080982524866e-05, + "loss": 0.0738, + "step": 9326 + }, + { + "epoch": 1.22, + "learning_rate": 1.922060457808419e-05, + "loss": 0.0791, + "step": 9327 + }, + { + "epoch": 1.22, + "learning_rate": 1.9220399304987257e-05, + "loss": 0.0797, + "step": 9328 + }, + { + "epoch": 1.22, + "learning_rate": 1.9220194005958446e-05, + "loss": 0.0797, + "step": 9329 + }, + { + "epoch": 1.22, + "learning_rate": 1.9219988680998328e-05, + "loss": 0.0852, + "step": 9330 + }, + { + "epoch": 1.22, + "learning_rate": 1.921978333010748e-05, + "loss": 0.0782, + "step": 9331 + }, + { + "epoch": 1.22, + "learning_rate": 1.9219577953286487e-05, + "loss": 0.0833, + "step": 9332 + }, + { + "epoch": 1.22, + "learning_rate": 1.9219372550535923e-05, + "loss": 0.0813, + "step": 9333 + }, + { + "epoch": 1.22, + "learning_rate": 1.9219167121856362e-05, + "loss": 0.0727, + "step": 9334 + }, + { + "epoch": 1.22, + "learning_rate": 1.9218961667248385e-05, + "loss": 0.0625, + "step": 9335 + }, + { + "epoch": 1.22, + "learning_rate": 1.9218756186712566e-05, + "loss": 0.088, + "step": 9336 + }, + { + "epoch": 1.22, + "learning_rate": 1.921855068024949e-05, + "loss": 0.0797, + "step": 9337 + }, + { + "epoch": 1.22, + "learning_rate": 1.921834514785973e-05, + "loss": 0.0708, + "step": 9338 + }, + { + "epoch": 1.22, + "learning_rate": 1.9218139589543864e-05, + "loss": 0.084, + "step": 9339 + }, + { + "epoch": 1.22, + "learning_rate": 1.9217934005302474e-05, + "loss": 0.0837, + "step": 9340 + }, + { + "epoch": 1.22, + "learning_rate": 1.9217728395136134e-05, + "loss": 0.0749, + "step": 9341 + }, + { + "epoch": 1.22, + "learning_rate": 1.9217522759045425e-05, + "loss": 0.0715, + "step": 9342 + }, + { + "epoch": 1.22, + "learning_rate": 1.921731709703092e-05, + "loss": 0.0817, + "step": 9343 + }, + { + "epoch": 1.22, + "learning_rate": 1.9217111409093206e-05, + "loss": 0.0669, + "step": 9344 + }, + { + "epoch": 1.22, + "learning_rate": 1.9216905695232855e-05, + "loss": 0.0801, + "step": 9345 + }, + { + "epoch": 1.22, + "learning_rate": 1.921669995545045e-05, + "loss": 0.0828, + "step": 9346 + }, + { + "epoch": 1.22, + "learning_rate": 1.9216494189746563e-05, + "loss": 0.0886, + "step": 9347 + }, + { + "epoch": 1.22, + "learning_rate": 1.9216288398121777e-05, + "loss": 0.0847, + "step": 9348 + }, + { + "epoch": 1.22, + "learning_rate": 1.9216082580576675e-05, + "loss": 0.0755, + "step": 9349 + }, + { + "epoch": 1.22, + "learning_rate": 1.921587673711183e-05, + "loss": 0.0887, + "step": 9350 + }, + { + "epoch": 1.22, + "learning_rate": 1.921567086772782e-05, + "loss": 0.0969, + "step": 9351 + }, + { + "epoch": 1.23, + "learning_rate": 1.9215464972425232e-05, + "loss": 0.0774, + "step": 9352 + }, + { + "epoch": 1.23, + "learning_rate": 1.9215259051204632e-05, + "loss": 0.0654, + "step": 9353 + }, + { + "epoch": 1.23, + "learning_rate": 1.9215053104066613e-05, + "loss": 0.0896, + "step": 9354 + }, + { + "epoch": 1.23, + "learning_rate": 1.921484713101174e-05, + "loss": 0.0788, + "step": 9355 + }, + { + "epoch": 1.23, + "learning_rate": 1.9214641132040608e-05, + "loss": 0.0694, + "step": 9356 + }, + { + "epoch": 1.23, + "learning_rate": 1.9214435107153787e-05, + "loss": 0.0849, + "step": 9357 + }, + { + "epoch": 1.23, + "learning_rate": 1.9214229056351858e-05, + "loss": 0.0869, + "step": 9358 + }, + { + "epoch": 1.23, + "learning_rate": 1.92140229796354e-05, + "loss": 0.0848, + "step": 9359 + }, + { + "epoch": 1.23, + "learning_rate": 1.921381687700499e-05, + "loss": 0.0838, + "step": 9360 + }, + { + "epoch": 1.23, + "learning_rate": 1.9213610748461215e-05, + "loss": 0.0786, + "step": 9361 + }, + { + "epoch": 1.23, + "learning_rate": 1.921340459400465e-05, + "loss": 0.0946, + "step": 9362 + }, + { + "epoch": 1.23, + "learning_rate": 1.921319841363587e-05, + "loss": 0.086, + "step": 9363 + }, + { + "epoch": 1.23, + "learning_rate": 1.9212992207355467e-05, + "loss": 0.0775, + "step": 9364 + }, + { + "epoch": 1.23, + "learning_rate": 1.921278597516401e-05, + "loss": 0.0795, + "step": 9365 + }, + { + "epoch": 1.23, + "learning_rate": 1.9212579717062083e-05, + "loss": 0.0872, + "step": 9366 + }, + { + "epoch": 1.23, + "learning_rate": 1.9212373433050265e-05, + "loss": 0.071, + "step": 9367 + }, + { + "epoch": 1.23, + "learning_rate": 1.9212167123129143e-05, + "loss": 0.0696, + "step": 9368 + }, + { + "epoch": 1.23, + "learning_rate": 1.9211960787299285e-05, + "loss": 0.092, + "step": 9369 + }, + { + "epoch": 1.23, + "learning_rate": 1.921175442556128e-05, + "loss": 0.0726, + "step": 9370 + }, + { + "epoch": 1.23, + "learning_rate": 1.9211548037915707e-05, + "loss": 0.0714, + "step": 9371 + }, + { + "epoch": 1.23, + "learning_rate": 1.9211341624363146e-05, + "loss": 0.084, + "step": 9372 + }, + { + "epoch": 1.23, + "learning_rate": 1.921113518490418e-05, + "loss": 0.0798, + "step": 9373 + }, + { + "epoch": 1.23, + "learning_rate": 1.9210928719539384e-05, + "loss": 0.0779, + "step": 9374 + }, + { + "epoch": 1.23, + "learning_rate": 1.921072222826934e-05, + "loss": 0.0764, + "step": 9375 + }, + { + "epoch": 1.23, + "learning_rate": 1.921051571109463e-05, + "loss": 0.0846, + "step": 9376 + }, + { + "epoch": 1.23, + "learning_rate": 1.921030916801584e-05, + "loss": 0.0805, + "step": 9377 + }, + { + "epoch": 1.23, + "learning_rate": 1.921010259903354e-05, + "loss": 0.0874, + "step": 9378 + }, + { + "epoch": 1.23, + "learning_rate": 1.920989600414832e-05, + "loss": 0.0843, + "step": 9379 + }, + { + "epoch": 1.23, + "learning_rate": 1.9209689383360756e-05, + "loss": 0.0795, + "step": 9380 + }, + { + "epoch": 1.23, + "learning_rate": 1.9209482736671436e-05, + "loss": 0.0885, + "step": 9381 + }, + { + "epoch": 1.23, + "learning_rate": 1.920927606408093e-05, + "loss": 0.073, + "step": 9382 + }, + { + "epoch": 1.23, + "learning_rate": 1.920906936558983e-05, + "loss": 0.0828, + "step": 9383 + }, + { + "epoch": 1.23, + "learning_rate": 1.9208862641198707e-05, + "loss": 0.0725, + "step": 9384 + }, + { + "epoch": 1.23, + "learning_rate": 1.9208655890908154e-05, + "loss": 0.0833, + "step": 9385 + }, + { + "epoch": 1.23, + "learning_rate": 1.9208449114718742e-05, + "loss": 0.0802, + "step": 9386 + }, + { + "epoch": 1.23, + "learning_rate": 1.9208242312631064e-05, + "loss": 0.082, + "step": 9387 + }, + { + "epoch": 1.23, + "learning_rate": 1.9208035484645688e-05, + "loss": 0.0856, + "step": 9388 + }, + { + "epoch": 1.23, + "learning_rate": 1.920782863076321e-05, + "loss": 0.0788, + "step": 9389 + }, + { + "epoch": 1.23, + "learning_rate": 1.9207621750984198e-05, + "loss": 0.0863, + "step": 9390 + }, + { + "epoch": 1.23, + "learning_rate": 1.920741484530924e-05, + "loss": 0.0849, + "step": 9391 + }, + { + "epoch": 1.23, + "learning_rate": 1.920720791373892e-05, + "loss": 0.0983, + "step": 9392 + }, + { + "epoch": 1.23, + "learning_rate": 1.920700095627382e-05, + "loss": 0.0873, + "step": 9393 + }, + { + "epoch": 1.23, + "learning_rate": 1.9206793972914516e-05, + "loss": 0.0876, + "step": 9394 + }, + { + "epoch": 1.23, + "learning_rate": 1.9206586963661597e-05, + "loss": 0.0771, + "step": 9395 + }, + { + "epoch": 1.23, + "learning_rate": 1.9206379928515642e-05, + "loss": 0.0842, + "step": 9396 + }, + { + "epoch": 1.23, + "learning_rate": 1.9206172867477235e-05, + "loss": 0.0827, + "step": 9397 + }, + { + "epoch": 1.23, + "learning_rate": 1.9205965780546956e-05, + "loss": 0.0738, + "step": 9398 + }, + { + "epoch": 1.23, + "learning_rate": 1.9205758667725386e-05, + "loss": 0.0838, + "step": 9399 + }, + { + "epoch": 1.23, + "learning_rate": 1.920555152901311e-05, + "loss": 0.0789, + "step": 9400 + }, + { + "epoch": 1.23, + "learning_rate": 1.9205344364410715e-05, + "loss": 0.0892, + "step": 9401 + }, + { + "epoch": 1.23, + "learning_rate": 1.9205137173918776e-05, + "loss": 0.0898, + "step": 9402 + }, + { + "epoch": 1.23, + "learning_rate": 1.920492995753788e-05, + "loss": 0.0823, + "step": 9403 + }, + { + "epoch": 1.23, + "learning_rate": 1.920472271526861e-05, + "loss": 0.0831, + "step": 9404 + }, + { + "epoch": 1.23, + "learning_rate": 1.9204515447111545e-05, + "loss": 0.0778, + "step": 9405 + }, + { + "epoch": 1.23, + "learning_rate": 1.9204308153067273e-05, + "loss": 0.0792, + "step": 9406 + }, + { + "epoch": 1.23, + "learning_rate": 1.9204100833136374e-05, + "loss": 0.0786, + "step": 9407 + }, + { + "epoch": 1.23, + "learning_rate": 1.920389348731943e-05, + "loss": 0.0752, + "step": 9408 + }, + { + "epoch": 1.23, + "learning_rate": 1.9203686115617025e-05, + "loss": 0.0955, + "step": 9409 + }, + { + "epoch": 1.23, + "learning_rate": 1.9203478718029745e-05, + "loss": 0.0872, + "step": 9410 + }, + { + "epoch": 1.23, + "learning_rate": 1.9203271294558167e-05, + "loss": 0.0777, + "step": 9411 + }, + { + "epoch": 1.23, + "learning_rate": 1.9203063845202883e-05, + "loss": 0.0803, + "step": 9412 + }, + { + "epoch": 1.23, + "learning_rate": 1.920285636996447e-05, + "loss": 0.0916, + "step": 9413 + }, + { + "epoch": 1.23, + "learning_rate": 1.9202648868843516e-05, + "loss": 0.0858, + "step": 9414 + }, + { + "epoch": 1.23, + "learning_rate": 1.92024413418406e-05, + "loss": 0.0812, + "step": 9415 + }, + { + "epoch": 1.23, + "learning_rate": 1.920223378895631e-05, + "loss": 0.0796, + "step": 9416 + }, + { + "epoch": 1.23, + "learning_rate": 1.9202026210191224e-05, + "loss": 0.0751, + "step": 9417 + }, + { + "epoch": 1.23, + "learning_rate": 1.9201818605545933e-05, + "loss": 0.079, + "step": 9418 + }, + { + "epoch": 1.23, + "learning_rate": 1.9201610975021015e-05, + "loss": 0.0768, + "step": 9419 + }, + { + "epoch": 1.23, + "learning_rate": 1.9201403318617058e-05, + "loss": 0.0854, + "step": 9420 + }, + { + "epoch": 1.23, + "learning_rate": 1.920119563633464e-05, + "loss": 0.0896, + "step": 9421 + }, + { + "epoch": 1.23, + "learning_rate": 1.9200987928174352e-05, + "loss": 0.0798, + "step": 9422 + }, + { + "epoch": 1.23, + "learning_rate": 1.9200780194136776e-05, + "loss": 0.0882, + "step": 9423 + }, + { + "epoch": 1.23, + "learning_rate": 1.9200572434222496e-05, + "loss": 0.0851, + "step": 9424 + }, + { + "epoch": 1.23, + "learning_rate": 1.9200364648432095e-05, + "loss": 0.0811, + "step": 9425 + }, + { + "epoch": 1.23, + "learning_rate": 1.920015683676616e-05, + "loss": 0.0719, + "step": 9426 + }, + { + "epoch": 1.23, + "learning_rate": 1.919994899922527e-05, + "loss": 0.0896, + "step": 9427 + }, + { + "epoch": 1.24, + "learning_rate": 1.9199741135810018e-05, + "loss": 0.0814, + "step": 9428 + }, + { + "epoch": 1.24, + "learning_rate": 1.919953324652098e-05, + "loss": 0.0705, + "step": 9429 + }, + { + "epoch": 1.24, + "learning_rate": 1.9199325331358746e-05, + "loss": 0.0779, + "step": 9430 + }, + { + "epoch": 1.24, + "learning_rate": 1.91991173903239e-05, + "loss": 0.0777, + "step": 9431 + }, + { + "epoch": 1.24, + "learning_rate": 1.9198909423417027e-05, + "loss": 0.0745, + "step": 9432 + }, + { + "epoch": 1.24, + "learning_rate": 1.9198701430638712e-05, + "loss": 0.0838, + "step": 9433 + }, + { + "epoch": 1.24, + "learning_rate": 1.9198493411989534e-05, + "loss": 0.0924, + "step": 9434 + }, + { + "epoch": 1.24, + "learning_rate": 1.9198285367470087e-05, + "loss": 0.066, + "step": 9435 + }, + { + "epoch": 1.24, + "learning_rate": 1.9198077297080952e-05, + "loss": 0.0798, + "step": 9436 + }, + { + "epoch": 1.24, + "learning_rate": 1.9197869200822713e-05, + "loss": 0.0792, + "step": 9437 + }, + { + "epoch": 1.24, + "learning_rate": 1.919766107869596e-05, + "loss": 0.0747, + "step": 9438 + }, + { + "epoch": 1.24, + "learning_rate": 1.9197452930701272e-05, + "loss": 0.0779, + "step": 9439 + }, + { + "epoch": 1.24, + "learning_rate": 1.9197244756839236e-05, + "loss": 0.0909, + "step": 9440 + }, + { + "epoch": 1.24, + "learning_rate": 1.9197036557110443e-05, + "loss": 0.0767, + "step": 9441 + }, + { + "epoch": 1.24, + "learning_rate": 1.919682833151547e-05, + "loss": 0.0861, + "step": 9442 + }, + { + "epoch": 1.24, + "learning_rate": 1.919662008005491e-05, + "loss": 0.0855, + "step": 9443 + }, + { + "epoch": 1.24, + "learning_rate": 1.9196411802729344e-05, + "loss": 0.0905, + "step": 9444 + }, + { + "epoch": 1.24, + "learning_rate": 1.9196203499539364e-05, + "loss": 0.0857, + "step": 9445 + }, + { + "epoch": 1.24, + "learning_rate": 1.9195995170485547e-05, + "loss": 0.0842, + "step": 9446 + }, + { + "epoch": 1.24, + "learning_rate": 1.9195786815568484e-05, + "loss": 0.0814, + "step": 9447 + }, + { + "epoch": 1.24, + "learning_rate": 1.919557843478876e-05, + "loss": 0.0859, + "step": 9448 + }, + { + "epoch": 1.24, + "learning_rate": 1.9195370028146963e-05, + "loss": 0.0844, + "step": 9449 + }, + { + "epoch": 1.24, + "learning_rate": 1.9195161595643677e-05, + "loss": 0.0797, + "step": 9450 + }, + { + "epoch": 1.24, + "learning_rate": 1.9194953137279487e-05, + "loss": 0.0759, + "step": 9451 + }, + { + "epoch": 1.24, + "learning_rate": 1.9194744653054984e-05, + "loss": 0.0738, + "step": 9452 + }, + { + "epoch": 1.24, + "learning_rate": 1.919453614297075e-05, + "loss": 0.0766, + "step": 9453 + }, + { + "epoch": 1.24, + "learning_rate": 1.9194327607027372e-05, + "loss": 0.0863, + "step": 9454 + }, + { + "epoch": 1.24, + "learning_rate": 1.9194119045225435e-05, + "loss": 0.0784, + "step": 9455 + }, + { + "epoch": 1.24, + "learning_rate": 1.919391045756553e-05, + "loss": 0.0794, + "step": 9456 + }, + { + "epoch": 1.24, + "learning_rate": 1.9193701844048245e-05, + "loss": 0.0775, + "step": 9457 + }, + { + "epoch": 1.24, + "learning_rate": 1.9193493204674157e-05, + "loss": 0.0873, + "step": 9458 + }, + { + "epoch": 1.24, + "learning_rate": 1.9193284539443864e-05, + "loss": 0.0777, + "step": 9459 + }, + { + "epoch": 1.24, + "learning_rate": 1.9193075848357947e-05, + "loss": 0.0729, + "step": 9460 + }, + { + "epoch": 1.24, + "learning_rate": 1.9192867131416993e-05, + "loss": 0.0837, + "step": 9461 + }, + { + "epoch": 1.24, + "learning_rate": 1.9192658388621588e-05, + "loss": 0.0861, + "step": 9462 + }, + { + "epoch": 1.24, + "learning_rate": 1.9192449619972322e-05, + "loss": 0.0828, + "step": 9463 + }, + { + "epoch": 1.24, + "learning_rate": 1.919224082546978e-05, + "loss": 0.0912, + "step": 9464 + }, + { + "epoch": 1.24, + "learning_rate": 1.919203200511455e-05, + "loss": 0.0881, + "step": 9465 + }, + { + "epoch": 1.24, + "learning_rate": 1.9191823158907224e-05, + "loss": 0.092, + "step": 9466 + }, + { + "epoch": 1.24, + "learning_rate": 1.9191614286848378e-05, + "loss": 0.0871, + "step": 9467 + }, + { + "epoch": 1.24, + "learning_rate": 1.9191405388938614e-05, + "loss": 0.0957, + "step": 9468 + }, + { + "epoch": 1.24, + "learning_rate": 1.9191196465178507e-05, + "loss": 0.0823, + "step": 9469 + }, + { + "epoch": 1.24, + "learning_rate": 1.9190987515568647e-05, + "loss": 0.0825, + "step": 9470 + }, + { + "epoch": 1.24, + "learning_rate": 1.9190778540109626e-05, + "loss": 0.0846, + "step": 9471 + }, + { + "epoch": 1.24, + "learning_rate": 1.9190569538802033e-05, + "loss": 0.0891, + "step": 9472 + }, + { + "epoch": 1.24, + "learning_rate": 1.919036051164645e-05, + "loss": 0.0725, + "step": 9473 + }, + { + "epoch": 1.24, + "learning_rate": 1.919015145864347e-05, + "loss": 0.0871, + "step": 9474 + }, + { + "epoch": 1.24, + "learning_rate": 1.9189942379793675e-05, + "loss": 0.0792, + "step": 9475 + }, + { + "epoch": 1.24, + "learning_rate": 1.918973327509766e-05, + "loss": 0.0821, + "step": 9476 + }, + { + "epoch": 1.24, + "learning_rate": 1.9189524144556007e-05, + "loss": 0.0837, + "step": 9477 + }, + { + "epoch": 1.24, + "learning_rate": 1.9189314988169305e-05, + "loss": 0.0855, + "step": 9478 + }, + { + "epoch": 1.24, + "learning_rate": 1.9189105805938147e-05, + "loss": 0.0855, + "step": 9479 + }, + { + "epoch": 1.24, + "learning_rate": 1.9188896597863116e-05, + "loss": 0.0771, + "step": 9480 + }, + { + "epoch": 1.24, + "learning_rate": 1.9188687363944805e-05, + "loss": 0.0844, + "step": 9481 + }, + { + "epoch": 1.24, + "learning_rate": 1.91884781041838e-05, + "loss": 0.0754, + "step": 9482 + }, + { + "epoch": 1.24, + "learning_rate": 1.9188268818580687e-05, + "loss": 0.085, + "step": 9483 + }, + { + "epoch": 1.24, + "learning_rate": 1.9188059507136062e-05, + "loss": 0.0773, + "step": 9484 + }, + { + "epoch": 1.24, + "learning_rate": 1.9187850169850502e-05, + "loss": 0.0767, + "step": 9485 + }, + { + "epoch": 1.24, + "learning_rate": 1.918764080672461e-05, + "loss": 0.0799, + "step": 9486 + }, + { + "epoch": 1.24, + "learning_rate": 1.918743141775896e-05, + "loss": 0.084, + "step": 9487 + }, + { + "epoch": 1.24, + "learning_rate": 1.9187222002954155e-05, + "loss": 0.0785, + "step": 9488 + }, + { + "epoch": 1.24, + "learning_rate": 1.9187012562310775e-05, + "loss": 0.0805, + "step": 9489 + }, + { + "epoch": 1.24, + "learning_rate": 1.918680309582941e-05, + "loss": 0.0799, + "step": 9490 + }, + { + "epoch": 1.24, + "learning_rate": 1.9186593603510654e-05, + "loss": 0.0846, + "step": 9491 + }, + { + "epoch": 1.24, + "learning_rate": 1.9186384085355086e-05, + "loss": 0.0906, + "step": 9492 + }, + { + "epoch": 1.24, + "learning_rate": 1.918617454136331e-05, + "loss": 0.0895, + "step": 9493 + }, + { + "epoch": 1.24, + "learning_rate": 1.9185964971535905e-05, + "loss": 0.0785, + "step": 9494 + }, + { + "epoch": 1.24, + "learning_rate": 1.918575537587346e-05, + "loss": 0.0703, + "step": 9495 + }, + { + "epoch": 1.24, + "learning_rate": 1.9185545754376566e-05, + "loss": 0.0822, + "step": 9496 + }, + { + "epoch": 1.24, + "learning_rate": 1.9185336107045817e-05, + "loss": 0.0816, + "step": 9497 + }, + { + "epoch": 1.24, + "learning_rate": 1.91851264338818e-05, + "loss": 0.0842, + "step": 9498 + }, + { + "epoch": 1.24, + "learning_rate": 1.9184916734885104e-05, + "loss": 0.0801, + "step": 9499 + }, + { + "epoch": 1.24, + "learning_rate": 1.9184707010056316e-05, + "loss": 0.0741, + "step": 9500 + }, + { + "epoch": 1.24, + "learning_rate": 1.918449725939603e-05, + "loss": 0.0905, + "step": 9501 + }, + { + "epoch": 1.24, + "learning_rate": 1.9184287482904837e-05, + "loss": 0.0744, + "step": 9502 + }, + { + "epoch": 1.24, + "learning_rate": 1.9184077680583318e-05, + "loss": 0.0812, + "step": 9503 + }, + { + "epoch": 1.24, + "learning_rate": 1.9183867852432075e-05, + "loss": 0.0806, + "step": 9504 + }, + { + "epoch": 1.25, + "learning_rate": 1.9183657998451694e-05, + "loss": 0.0688, + "step": 9505 + }, + { + "epoch": 1.25, + "learning_rate": 1.918344811864276e-05, + "loss": 0.0781, + "step": 9506 + }, + { + "epoch": 1.25, + "learning_rate": 1.918323821300587e-05, + "loss": 0.0796, + "step": 9507 + }, + { + "epoch": 1.25, + "learning_rate": 1.918302828154161e-05, + "loss": 0.0821, + "step": 9508 + }, + { + "epoch": 1.25, + "learning_rate": 1.9182818324250573e-05, + "loss": 0.0878, + "step": 9509 + }, + { + "epoch": 1.25, + "learning_rate": 1.9182608341133348e-05, + "loss": 0.0744, + "step": 9510 + }, + { + "epoch": 1.25, + "learning_rate": 1.9182398332190525e-05, + "loss": 0.0817, + "step": 9511 + }, + { + "epoch": 1.25, + "learning_rate": 1.91821882974227e-05, + "loss": 0.0834, + "step": 9512 + }, + { + "epoch": 1.25, + "learning_rate": 1.9181978236830453e-05, + "loss": 0.0786, + "step": 9513 + }, + { + "epoch": 1.25, + "learning_rate": 1.9181768150414385e-05, + "loss": 0.0819, + "step": 9514 + }, + { + "epoch": 1.25, + "learning_rate": 1.918155803817508e-05, + "loss": 0.0799, + "step": 9515 + }, + { + "epoch": 1.25, + "learning_rate": 1.9181347900113135e-05, + "loss": 0.0731, + "step": 9516 + }, + { + "epoch": 1.25, + "learning_rate": 1.9181137736229138e-05, + "loss": 0.0781, + "step": 9517 + }, + { + "epoch": 1.25, + "learning_rate": 1.9180927546523677e-05, + "loss": 0.0797, + "step": 9518 + }, + { + "epoch": 1.25, + "learning_rate": 1.918071733099735e-05, + "loss": 0.0855, + "step": 9519 + }, + { + "epoch": 1.25, + "learning_rate": 1.9180507089650743e-05, + "loss": 0.0743, + "step": 9520 + }, + { + "epoch": 1.25, + "learning_rate": 1.9180296822484447e-05, + "loss": 0.0808, + "step": 9521 + }, + { + "epoch": 1.25, + "learning_rate": 1.9180086529499057e-05, + "loss": 0.0935, + "step": 9522 + }, + { + "epoch": 1.25, + "learning_rate": 1.917987621069516e-05, + "loss": 0.0843, + "step": 9523 + }, + { + "epoch": 1.25, + "learning_rate": 1.9179665866073353e-05, + "loss": 0.0772, + "step": 9524 + }, + { + "epoch": 1.25, + "learning_rate": 1.9179455495634226e-05, + "loss": 0.0792, + "step": 9525 + }, + { + "epoch": 1.25, + "learning_rate": 1.9179245099378365e-05, + "loss": 0.0756, + "step": 9526 + }, + { + "epoch": 1.25, + "learning_rate": 1.917903467730637e-05, + "loss": 0.0849, + "step": 9527 + }, + { + "epoch": 1.25, + "learning_rate": 1.9178824229418825e-05, + "loss": 0.0725, + "step": 9528 + }, + { + "epoch": 1.25, + "learning_rate": 1.9178613755716326e-05, + "loss": 0.0803, + "step": 9529 + }, + { + "epoch": 1.25, + "learning_rate": 1.9178403256199468e-05, + "loss": 0.0852, + "step": 9530 + }, + { + "epoch": 1.25, + "learning_rate": 1.9178192730868837e-05, + "loss": 0.0838, + "step": 9531 + }, + { + "epoch": 1.25, + "learning_rate": 1.917798217972503e-05, + "loss": 0.0839, + "step": 9532 + }, + { + "epoch": 1.25, + "learning_rate": 1.9177771602768634e-05, + "loss": 0.0873, + "step": 9533 + }, + { + "epoch": 1.25, + "learning_rate": 1.9177561000000245e-05, + "loss": 0.0829, + "step": 9534 + }, + { + "epoch": 1.25, + "learning_rate": 1.9177350371420454e-05, + "loss": 0.0886, + "step": 9535 + }, + { + "epoch": 1.25, + "learning_rate": 1.9177139717029853e-05, + "loss": 0.0849, + "step": 9536 + }, + { + "epoch": 1.25, + "learning_rate": 1.917692903682904e-05, + "loss": 0.079, + "step": 9537 + }, + { + "epoch": 1.25, + "learning_rate": 1.9176718330818595e-05, + "loss": 0.085, + "step": 9538 + }, + { + "epoch": 1.25, + "learning_rate": 1.9176507598999126e-05, + "loss": 0.0761, + "step": 9539 + }, + { + "epoch": 1.25, + "learning_rate": 1.9176296841371213e-05, + "loss": 0.0891, + "step": 9540 + }, + { + "epoch": 1.25, + "learning_rate": 1.9176086057935457e-05, + "loss": 0.0716, + "step": 9541 + }, + { + "epoch": 1.25, + "learning_rate": 1.9175875248692447e-05, + "loss": 0.0832, + "step": 9542 + }, + { + "epoch": 1.25, + "learning_rate": 1.9175664413642775e-05, + "loss": 0.0953, + "step": 9543 + }, + { + "epoch": 1.25, + "learning_rate": 1.9175453552787035e-05, + "loss": 0.0804, + "step": 9544 + }, + { + "epoch": 1.25, + "learning_rate": 1.9175242666125823e-05, + "loss": 0.08, + "step": 9545 + }, + { + "epoch": 1.25, + "learning_rate": 1.9175031753659728e-05, + "loss": 0.0769, + "step": 9546 + }, + { + "epoch": 1.25, + "learning_rate": 1.9174820815389344e-05, + "loss": 0.0696, + "step": 9547 + }, + { + "epoch": 1.25, + "learning_rate": 1.917460985131527e-05, + "loss": 0.0785, + "step": 9548 + }, + { + "epoch": 1.25, + "learning_rate": 1.917439886143809e-05, + "loss": 0.0882, + "step": 9549 + }, + { + "epoch": 1.25, + "learning_rate": 1.91741878457584e-05, + "loss": 0.078, + "step": 9550 + }, + { + "epoch": 1.25, + "learning_rate": 1.91739768042768e-05, + "loss": 0.0803, + "step": 9551 + }, + { + "epoch": 1.25, + "learning_rate": 1.9173765736993874e-05, + "loss": 0.0763, + "step": 9552 + }, + { + "epoch": 1.25, + "learning_rate": 1.9173554643910225e-05, + "loss": 0.0897, + "step": 9553 + }, + { + "epoch": 1.25, + "learning_rate": 1.917334352502644e-05, + "loss": 0.0751, + "step": 9554 + }, + { + "epoch": 1.25, + "learning_rate": 1.9173132380343115e-05, + "loss": 0.0749, + "step": 9555 + }, + { + "epoch": 1.25, + "learning_rate": 1.917292120986084e-05, + "loss": 0.0778, + "step": 9556 + }, + { + "epoch": 1.25, + "learning_rate": 1.917271001358022e-05, + "loss": 0.0845, + "step": 9557 + }, + { + "epoch": 1.25, + "learning_rate": 1.9172498791501834e-05, + "loss": 0.072, + "step": 9558 + }, + { + "epoch": 1.25, + "learning_rate": 1.9172287543626288e-05, + "loss": 0.0684, + "step": 9559 + }, + { + "epoch": 1.25, + "learning_rate": 1.917207626995417e-05, + "loss": 0.0895, + "step": 9560 + }, + { + "epoch": 1.25, + "learning_rate": 1.9171864970486078e-05, + "loss": 0.0744, + "step": 9561 + }, + { + "epoch": 1.25, + "learning_rate": 1.9171653645222604e-05, + "loss": 0.092, + "step": 9562 + }, + { + "epoch": 1.25, + "learning_rate": 1.917144229416434e-05, + "loss": 0.0776, + "step": 9563 + }, + { + "epoch": 1.25, + "learning_rate": 1.9171230917311887e-05, + "loss": 0.082, + "step": 9564 + }, + { + "epoch": 1.25, + "learning_rate": 1.9171019514665833e-05, + "loss": 0.0706, + "step": 9565 + }, + { + "epoch": 1.25, + "learning_rate": 1.9170808086226773e-05, + "loss": 0.0862, + "step": 9566 + }, + { + "epoch": 1.25, + "learning_rate": 1.9170596631995305e-05, + "loss": 0.0867, + "step": 9567 + }, + { + "epoch": 1.25, + "learning_rate": 1.9170385151972023e-05, + "loss": 0.0793, + "step": 9568 + }, + { + "epoch": 1.25, + "learning_rate": 1.9170173646157523e-05, + "loss": 0.0823, + "step": 9569 + }, + { + "epoch": 1.25, + "learning_rate": 1.9169962114552394e-05, + "loss": 0.0932, + "step": 9570 + }, + { + "epoch": 1.25, + "learning_rate": 1.9169750557157236e-05, + "loss": 0.0745, + "step": 9571 + }, + { + "epoch": 1.25, + "learning_rate": 1.9169538973972646e-05, + "loss": 0.0791, + "step": 9572 + }, + { + "epoch": 1.25, + "learning_rate": 1.9169327364999213e-05, + "loss": 0.0862, + "step": 9573 + }, + { + "epoch": 1.25, + "learning_rate": 1.9169115730237536e-05, + "loss": 0.0676, + "step": 9574 + }, + { + "epoch": 1.25, + "learning_rate": 1.916890406968821e-05, + "loss": 0.0764, + "step": 9575 + }, + { + "epoch": 1.25, + "learning_rate": 1.9168692383351828e-05, + "loss": 0.0862, + "step": 9576 + }, + { + "epoch": 1.25, + "learning_rate": 1.916848067122899e-05, + "loss": 0.0923, + "step": 9577 + }, + { + "epoch": 1.25, + "learning_rate": 1.9168268933320285e-05, + "loss": 0.0874, + "step": 9578 + }, + { + "epoch": 1.25, + "learning_rate": 1.9168057169626313e-05, + "loss": 0.0786, + "step": 9579 + }, + { + "epoch": 1.25, + "learning_rate": 1.9167845380147664e-05, + "loss": 0.0891, + "step": 9580 + }, + { + "epoch": 1.26, + "learning_rate": 1.9167633564884944e-05, + "loss": 0.0803, + "step": 9581 + }, + { + "epoch": 1.26, + "learning_rate": 1.916742172383874e-05, + "loss": 0.0856, + "step": 9582 + }, + { + "epoch": 1.26, + "learning_rate": 1.9167209857009653e-05, + "loss": 0.0821, + "step": 9583 + }, + { + "epoch": 1.26, + "learning_rate": 1.9166997964398276e-05, + "loss": 0.0901, + "step": 9584 + }, + { + "epoch": 1.26, + "learning_rate": 1.9166786046005205e-05, + "loss": 0.0681, + "step": 9585 + }, + { + "epoch": 1.26, + "learning_rate": 1.9166574101831035e-05, + "loss": 0.0736, + "step": 9586 + }, + { + "epoch": 1.26, + "learning_rate": 1.9166362131876365e-05, + "loss": 0.0785, + "step": 9587 + }, + { + "epoch": 1.26, + "learning_rate": 1.916615013614179e-05, + "loss": 0.0767, + "step": 9588 + }, + { + "epoch": 1.26, + "learning_rate": 1.9165938114627904e-05, + "loss": 0.0823, + "step": 9589 + }, + { + "epoch": 1.26, + "learning_rate": 1.9165726067335307e-05, + "loss": 0.076, + "step": 9590 + }, + { + "epoch": 1.26, + "learning_rate": 1.9165513994264594e-05, + "loss": 0.0963, + "step": 9591 + }, + { + "epoch": 1.26, + "learning_rate": 1.9165301895416357e-05, + "loss": 0.0826, + "step": 9592 + }, + { + "epoch": 1.26, + "learning_rate": 1.9165089770791197e-05, + "loss": 0.0775, + "step": 9593 + }, + { + "epoch": 1.26, + "learning_rate": 1.9164877620389716e-05, + "loss": 0.0707, + "step": 9594 + }, + { + "epoch": 1.26, + "learning_rate": 1.91646654442125e-05, + "loss": 0.0864, + "step": 9595 + }, + { + "epoch": 1.26, + "learning_rate": 1.916445324226015e-05, + "loss": 0.0705, + "step": 9596 + }, + { + "epoch": 1.26, + "learning_rate": 1.9164241014533267e-05, + "loss": 0.0806, + "step": 9597 + }, + { + "epoch": 1.26, + "learning_rate": 1.916402876103244e-05, + "loss": 0.0937, + "step": 9598 + }, + { + "epoch": 1.26, + "learning_rate": 1.9163816481758274e-05, + "loss": 0.0685, + "step": 9599 + }, + { + "epoch": 1.26, + "learning_rate": 1.9163604176711363e-05, + "loss": 0.0871, + "step": 9600 + }, + { + "epoch": 1.26, + "learning_rate": 1.9163391845892298e-05, + "loss": 0.078, + "step": 9601 + }, + { + "epoch": 1.26, + "learning_rate": 1.9163179489301683e-05, + "loss": 0.082, + "step": 9602 + }, + { + "epoch": 1.26, + "learning_rate": 1.9162967106940114e-05, + "loss": 0.0721, + "step": 9603 + }, + { + "epoch": 1.26, + "learning_rate": 1.9162754698808192e-05, + "loss": 0.09, + "step": 9604 + }, + { + "epoch": 1.26, + "learning_rate": 1.91625422649065e-05, + "loss": 0.0843, + "step": 9605 + }, + { + "epoch": 1.26, + "learning_rate": 1.9162329805235656e-05, + "loss": 0.0676, + "step": 9606 + }, + { + "epoch": 1.26, + "learning_rate": 1.9162117319796246e-05, + "loss": 0.0813, + "step": 9607 + }, + { + "epoch": 1.26, + "learning_rate": 1.9161904808588864e-05, + "loss": 0.0824, + "step": 9608 + }, + { + "epoch": 1.26, + "learning_rate": 1.916169227161412e-05, + "loss": 0.0667, + "step": 9609 + }, + { + "epoch": 1.26, + "learning_rate": 1.9161479708872595e-05, + "loss": 0.0908, + "step": 9610 + }, + { + "epoch": 1.26, + "learning_rate": 1.9161267120364905e-05, + "loss": 0.0936, + "step": 9611 + }, + { + "epoch": 1.26, + "learning_rate": 1.9161054506091635e-05, + "loss": 0.0937, + "step": 9612 + }, + { + "epoch": 1.26, + "learning_rate": 1.9160841866053387e-05, + "loss": 0.0845, + "step": 9613 + }, + { + "epoch": 1.26, + "learning_rate": 1.916062920025076e-05, + "loss": 0.0933, + "step": 9614 + }, + { + "epoch": 1.26, + "learning_rate": 1.916041650868435e-05, + "loss": 0.082, + "step": 9615 + }, + { + "epoch": 1.26, + "learning_rate": 1.9160203791354756e-05, + "loss": 0.0807, + "step": 9616 + }, + { + "epoch": 1.26, + "learning_rate": 1.9159991048262577e-05, + "loss": 0.0878, + "step": 9617 + }, + { + "epoch": 1.26, + "learning_rate": 1.9159778279408413e-05, + "loss": 0.083, + "step": 9618 + }, + { + "epoch": 1.26, + "learning_rate": 1.915956548479286e-05, + "loss": 0.0901, + "step": 9619 + }, + { + "epoch": 1.26, + "learning_rate": 1.9159352664416517e-05, + "loss": 0.082, + "step": 9620 + }, + { + "epoch": 1.26, + "learning_rate": 1.9159139818279978e-05, + "loss": 0.0788, + "step": 9621 + }, + { + "epoch": 1.26, + "learning_rate": 1.9158926946383852e-05, + "loss": 0.0855, + "step": 9622 + }, + { + "epoch": 1.26, + "learning_rate": 1.915871404872873e-05, + "loss": 0.0849, + "step": 9623 + }, + { + "epoch": 1.26, + "learning_rate": 1.915850112531521e-05, + "loss": 0.0827, + "step": 9624 + }, + { + "epoch": 1.26, + "learning_rate": 1.9158288176143896e-05, + "loss": 0.07, + "step": 9625 + }, + { + "epoch": 1.26, + "learning_rate": 1.915807520121538e-05, + "loss": 0.0825, + "step": 9626 + }, + { + "epoch": 1.26, + "learning_rate": 1.9157862200530267e-05, + "loss": 0.0769, + "step": 9627 + }, + { + "epoch": 1.26, + "learning_rate": 1.915764917408916e-05, + "loss": 0.0781, + "step": 9628 + }, + { + "epoch": 1.26, + "learning_rate": 1.9157436121892645e-05, + "loss": 0.0795, + "step": 9629 + }, + { + "epoch": 1.26, + "learning_rate": 1.9157223043941335e-05, + "loss": 0.069, + "step": 9630 + }, + { + "epoch": 1.26, + "learning_rate": 1.9157009940235815e-05, + "loss": 0.0816, + "step": 9631 + }, + { + "epoch": 1.26, + "learning_rate": 1.91567968107767e-05, + "loss": 0.0707, + "step": 9632 + }, + { + "epoch": 1.26, + "learning_rate": 1.9156583655564577e-05, + "loss": 0.07, + "step": 9633 + }, + { + "epoch": 1.26, + "learning_rate": 1.9156370474600048e-05, + "loss": 0.0889, + "step": 9634 + }, + { + "epoch": 1.26, + "learning_rate": 1.9156157267883718e-05, + "loss": 0.08, + "step": 9635 + }, + { + "epoch": 1.26, + "learning_rate": 1.9155944035416187e-05, + "loss": 0.0814, + "step": 9636 + }, + { + "epoch": 1.26, + "learning_rate": 1.9155730777198044e-05, + "loss": 0.07, + "step": 9637 + }, + { + "epoch": 1.26, + "learning_rate": 1.9155517493229897e-05, + "loss": 0.0797, + "step": 9638 + }, + { + "epoch": 1.26, + "learning_rate": 1.9155304183512348e-05, + "loss": 0.0885, + "step": 9639 + }, + { + "epoch": 1.26, + "learning_rate": 1.9155090848045994e-05, + "loss": 0.0824, + "step": 9640 + }, + { + "epoch": 1.26, + "learning_rate": 1.915487748683143e-05, + "loss": 0.0861, + "step": 9641 + }, + { + "epoch": 1.26, + "learning_rate": 1.9154664099869263e-05, + "loss": 0.0817, + "step": 9642 + }, + { + "epoch": 1.26, + "learning_rate": 1.915445068716009e-05, + "loss": 0.1005, + "step": 9643 + }, + { + "epoch": 1.26, + "learning_rate": 1.9154237248704515e-05, + "loss": 0.0846, + "step": 9644 + }, + { + "epoch": 1.26, + "learning_rate": 1.915402378450313e-05, + "loss": 0.092, + "step": 9645 + }, + { + "epoch": 1.26, + "learning_rate": 1.9153810294556544e-05, + "loss": 0.082, + "step": 9646 + }, + { + "epoch": 1.26, + "learning_rate": 1.9153596778865357e-05, + "loss": 0.0757, + "step": 9647 + }, + { + "epoch": 1.26, + "learning_rate": 1.915338323743016e-05, + "loss": 0.071, + "step": 9648 + }, + { + "epoch": 1.26, + "learning_rate": 1.9153169670251564e-05, + "loss": 0.0905, + "step": 9649 + }, + { + "epoch": 1.26, + "learning_rate": 1.9152956077330165e-05, + "loss": 0.0842, + "step": 9650 + }, + { + "epoch": 1.26, + "learning_rate": 1.9152742458666564e-05, + "loss": 0.0894, + "step": 9651 + }, + { + "epoch": 1.26, + "learning_rate": 1.9152528814261364e-05, + "loss": 0.0846, + "step": 9652 + }, + { + "epoch": 1.26, + "learning_rate": 1.9152315144115162e-05, + "loss": 0.0828, + "step": 9653 + }, + { + "epoch": 1.26, + "learning_rate": 1.915210144822856e-05, + "loss": 0.0931, + "step": 9654 + }, + { + "epoch": 1.26, + "learning_rate": 1.9151887726602162e-05, + "loss": 0.0889, + "step": 9655 + }, + { + "epoch": 1.26, + "learning_rate": 1.9151673979236567e-05, + "loss": 0.0758, + "step": 9656 + }, + { + "epoch": 1.26, + "learning_rate": 1.9151460206132375e-05, + "loss": 0.0909, + "step": 9657 + }, + { + "epoch": 1.27, + "learning_rate": 1.9151246407290192e-05, + "loss": 0.0734, + "step": 9658 + }, + { + "epoch": 1.27, + "learning_rate": 1.9151032582710615e-05, + "loss": 0.0728, + "step": 9659 + }, + { + "epoch": 1.27, + "learning_rate": 1.9150818732394243e-05, + "loss": 0.0808, + "step": 9660 + }, + { + "epoch": 1.27, + "learning_rate": 1.915060485634168e-05, + "loss": 0.0699, + "step": 9661 + }, + { + "epoch": 1.27, + "learning_rate": 1.9150390954553532e-05, + "loss": 0.0772, + "step": 9662 + }, + { + "epoch": 1.27, + "learning_rate": 1.9150177027030396e-05, + "loss": 0.0813, + "step": 9663 + }, + { + "epoch": 1.27, + "learning_rate": 1.9149963073772873e-05, + "loss": 0.0799, + "step": 9664 + }, + { + "epoch": 1.27, + "learning_rate": 1.9149749094781565e-05, + "loss": 0.0829, + "step": 9665 + }, + { + "epoch": 1.27, + "learning_rate": 1.9149535090057078e-05, + "loss": 0.0827, + "step": 9666 + }, + { + "epoch": 1.27, + "learning_rate": 1.914932105960001e-05, + "loss": 0.0754, + "step": 9667 + }, + { + "epoch": 1.27, + "learning_rate": 1.9149107003410963e-05, + "loss": 0.0757, + "step": 9668 + }, + { + "epoch": 1.27, + "learning_rate": 1.9148892921490542e-05, + "loss": 0.081, + "step": 9669 + }, + { + "epoch": 1.27, + "learning_rate": 1.9148678813839344e-05, + "loss": 0.0702, + "step": 9670 + }, + { + "epoch": 1.27, + "learning_rate": 1.9148464680457975e-05, + "loss": 0.0768, + "step": 9671 + }, + { + "epoch": 1.27, + "learning_rate": 1.9148250521347035e-05, + "loss": 0.0754, + "step": 9672 + }, + { + "epoch": 1.27, + "learning_rate": 1.914803633650713e-05, + "loss": 0.077, + "step": 9673 + }, + { + "epoch": 1.27, + "learning_rate": 1.9147822125938858e-05, + "loss": 0.0819, + "step": 9674 + }, + { + "epoch": 1.27, + "learning_rate": 1.9147607889642823e-05, + "loss": 0.0758, + "step": 9675 + }, + { + "epoch": 1.27, + "learning_rate": 1.914739362761963e-05, + "loss": 0.0772, + "step": 9676 + }, + { + "epoch": 1.27, + "learning_rate": 1.914717933986988e-05, + "loss": 0.0815, + "step": 9677 + }, + { + "epoch": 1.27, + "learning_rate": 1.9146965026394172e-05, + "loss": 0.0925, + "step": 9678 + }, + { + "epoch": 1.27, + "learning_rate": 1.914675068719311e-05, + "loss": 0.0837, + "step": 9679 + }, + { + "epoch": 1.27, + "learning_rate": 1.9146536322267307e-05, + "loss": 0.0802, + "step": 9680 + }, + { + "epoch": 1.27, + "learning_rate": 1.9146321931617353e-05, + "loss": 0.0762, + "step": 9681 + }, + { + "epoch": 1.27, + "learning_rate": 1.9146107515243855e-05, + "loss": 0.0753, + "step": 9682 + }, + { + "epoch": 1.27, + "learning_rate": 1.9145893073147417e-05, + "loss": 0.0865, + "step": 9683 + }, + { + "epoch": 1.27, + "learning_rate": 1.9145678605328645e-05, + "loss": 0.08, + "step": 9684 + }, + { + "epoch": 1.27, + "learning_rate": 1.9145464111788135e-05, + "loss": 0.0783, + "step": 9685 + }, + { + "epoch": 1.27, + "learning_rate": 1.9145249592526497e-05, + "loss": 0.0689, + "step": 9686 + }, + { + "epoch": 1.27, + "learning_rate": 1.9145035047544328e-05, + "loss": 0.074, + "step": 9687 + }, + { + "epoch": 1.27, + "learning_rate": 1.9144820476842237e-05, + "loss": 0.0963, + "step": 9688 + }, + { + "epoch": 1.27, + "learning_rate": 1.9144605880420824e-05, + "loss": 0.0807, + "step": 9689 + }, + { + "epoch": 1.27, + "learning_rate": 1.9144391258280696e-05, + "loss": 0.0869, + "step": 9690 + }, + { + "epoch": 1.27, + "learning_rate": 1.9144176610422455e-05, + "loss": 0.0744, + "step": 9691 + }, + { + "epoch": 1.27, + "learning_rate": 1.9143961936846702e-05, + "loss": 0.0837, + "step": 9692 + }, + { + "epoch": 1.27, + "learning_rate": 1.9143747237554045e-05, + "loss": 0.0803, + "step": 9693 + }, + { + "epoch": 1.27, + "learning_rate": 1.9143532512545086e-05, + "loss": 0.0817, + "step": 9694 + }, + { + "epoch": 1.27, + "learning_rate": 1.9143317761820426e-05, + "loss": 0.0763, + "step": 9695 + }, + { + "epoch": 1.27, + "learning_rate": 1.9143102985380673e-05, + "loss": 0.0772, + "step": 9696 + }, + { + "epoch": 1.27, + "learning_rate": 1.914288818322643e-05, + "loss": 0.0882, + "step": 9697 + }, + { + "epoch": 1.27, + "learning_rate": 1.9142673355358302e-05, + "loss": 0.0789, + "step": 9698 + }, + { + "epoch": 1.27, + "learning_rate": 1.914245850177689e-05, + "loss": 0.0863, + "step": 9699 + }, + { + "epoch": 1.27, + "learning_rate": 1.9142243622482804e-05, + "loss": 0.0903, + "step": 9700 + }, + { + "epoch": 1.27, + "learning_rate": 1.914202871747664e-05, + "loss": 0.0793, + "step": 9701 + }, + { + "epoch": 1.27, + "learning_rate": 1.914181378675901e-05, + "loss": 0.0784, + "step": 9702 + }, + { + "epoch": 1.27, + "learning_rate": 1.9141598830330516e-05, + "loss": 0.0887, + "step": 9703 + }, + { + "epoch": 1.27, + "learning_rate": 1.914138384819176e-05, + "loss": 0.0694, + "step": 9704 + }, + { + "epoch": 1.27, + "learning_rate": 1.9141168840343347e-05, + "loss": 0.0785, + "step": 9705 + }, + { + "epoch": 1.27, + "learning_rate": 1.9140953806785885e-05, + "loss": 0.0901, + "step": 9706 + }, + { + "epoch": 1.27, + "learning_rate": 1.914073874751998e-05, + "loss": 0.076, + "step": 9707 + }, + { + "epoch": 1.27, + "learning_rate": 1.914052366254623e-05, + "loss": 0.0828, + "step": 9708 + }, + { + "epoch": 1.27, + "learning_rate": 1.9140308551865247e-05, + "loss": 0.0878, + "step": 9709 + }, + { + "epoch": 1.27, + "learning_rate": 1.9140093415477634e-05, + "loss": 0.0894, + "step": 9710 + }, + { + "epoch": 1.27, + "learning_rate": 1.9139878253383993e-05, + "loss": 0.0826, + "step": 9711 + }, + { + "epoch": 1.27, + "learning_rate": 1.9139663065584932e-05, + "loss": 0.0796, + "step": 9712 + }, + { + "epoch": 1.27, + "learning_rate": 1.9139447852081054e-05, + "loss": 0.0871, + "step": 9713 + }, + { + "epoch": 1.27, + "learning_rate": 1.9139232612872965e-05, + "loss": 0.0732, + "step": 9714 + }, + { + "epoch": 1.27, + "learning_rate": 1.9139017347961274e-05, + "loss": 0.0781, + "step": 9715 + }, + { + "epoch": 1.27, + "learning_rate": 1.9138802057346583e-05, + "loss": 0.0846, + "step": 9716 + }, + { + "epoch": 1.27, + "learning_rate": 1.9138586741029493e-05, + "loss": 0.086, + "step": 9717 + }, + { + "epoch": 1.27, + "learning_rate": 1.913837139901062e-05, + "loss": 0.083, + "step": 9718 + }, + { + "epoch": 1.27, + "learning_rate": 1.913815603129056e-05, + "loss": 0.0711, + "step": 9719 + }, + { + "epoch": 1.27, + "learning_rate": 1.9137940637869926e-05, + "loss": 0.0844, + "step": 9720 + }, + { + "epoch": 1.27, + "learning_rate": 1.9137725218749318e-05, + "loss": 0.0778, + "step": 9721 + }, + { + "epoch": 1.27, + "learning_rate": 1.9137509773929344e-05, + "loss": 0.0801, + "step": 9722 + }, + { + "epoch": 1.27, + "learning_rate": 1.9137294303410618e-05, + "loss": 0.0747, + "step": 9723 + }, + { + "epoch": 1.27, + "learning_rate": 1.9137078807193728e-05, + "loss": 0.0662, + "step": 9724 + }, + { + "epoch": 1.27, + "learning_rate": 1.9136863285279296e-05, + "loss": 0.0913, + "step": 9725 + }, + { + "epoch": 1.27, + "learning_rate": 1.913664773766792e-05, + "loss": 0.0885, + "step": 9726 + }, + { + "epoch": 1.27, + "learning_rate": 1.9136432164360214e-05, + "loss": 0.0749, + "step": 9727 + }, + { + "epoch": 1.27, + "learning_rate": 1.9136216565356774e-05, + "loss": 0.0775, + "step": 9728 + }, + { + "epoch": 1.27, + "learning_rate": 1.9136000940658214e-05, + "loss": 0.0785, + "step": 9729 + }, + { + "epoch": 1.27, + "learning_rate": 1.9135785290265136e-05, + "loss": 0.0661, + "step": 9730 + }, + { + "epoch": 1.27, + "learning_rate": 1.9135569614178147e-05, + "loss": 0.0766, + "step": 9731 + }, + { + "epoch": 1.27, + "learning_rate": 1.9135353912397858e-05, + "loss": 0.0792, + "step": 9732 + }, + { + "epoch": 1.27, + "learning_rate": 1.9135138184924873e-05, + "loss": 0.0794, + "step": 9733 + }, + { + "epoch": 1.28, + "learning_rate": 1.9134922431759797e-05, + "loss": 0.0813, + "step": 9734 + }, + { + "epoch": 1.28, + "learning_rate": 1.9134706652903236e-05, + "loss": 0.071, + "step": 9735 + }, + { + "epoch": 1.28, + "learning_rate": 1.9134490848355803e-05, + "loss": 0.0896, + "step": 9736 + }, + { + "epoch": 1.28, + "learning_rate": 1.9134275018118098e-05, + "loss": 0.0904, + "step": 9737 + }, + { + "epoch": 1.28, + "learning_rate": 1.913405916219073e-05, + "loss": 0.0787, + "step": 9738 + }, + { + "epoch": 1.28, + "learning_rate": 1.913384328057431e-05, + "loss": 0.0773, + "step": 9739 + }, + { + "epoch": 1.28, + "learning_rate": 1.9133627373269442e-05, + "loss": 0.0752, + "step": 9740 + }, + { + "epoch": 1.28, + "learning_rate": 1.9133411440276734e-05, + "loss": 0.0755, + "step": 9741 + }, + { + "epoch": 1.28, + "learning_rate": 1.9133195481596792e-05, + "loss": 0.083, + "step": 9742 + }, + { + "epoch": 1.28, + "learning_rate": 1.913297949723022e-05, + "loss": 0.0843, + "step": 9743 + }, + { + "epoch": 1.28, + "learning_rate": 1.9132763487177633e-05, + "loss": 0.0782, + "step": 9744 + }, + { + "epoch": 1.28, + "learning_rate": 1.913254745143964e-05, + "loss": 0.077, + "step": 9745 + }, + { + "epoch": 1.28, + "learning_rate": 1.9132331390016837e-05, + "loss": 0.0738, + "step": 9746 + }, + { + "epoch": 1.28, + "learning_rate": 1.9132115302909838e-05, + "loss": 0.0747, + "step": 9747 + }, + { + "epoch": 1.28, + "learning_rate": 1.9131899190119255e-05, + "loss": 0.0878, + "step": 9748 + }, + { + "epoch": 1.28, + "learning_rate": 1.9131683051645688e-05, + "loss": 0.0917, + "step": 9749 + }, + { + "epoch": 1.28, + "learning_rate": 1.913146688748975e-05, + "loss": 0.0749, + "step": 9750 + }, + { + "epoch": 1.28, + "learning_rate": 1.9131250697652047e-05, + "loss": 0.0766, + "step": 9751 + }, + { + "epoch": 1.28, + "learning_rate": 1.913103448213319e-05, + "loss": 0.0804, + "step": 9752 + }, + { + "epoch": 1.28, + "learning_rate": 1.913081824093378e-05, + "loss": 0.0924, + "step": 9753 + }, + { + "epoch": 1.28, + "learning_rate": 1.9130601974054437e-05, + "loss": 0.0752, + "step": 9754 + }, + { + "epoch": 1.28, + "learning_rate": 1.9130385681495758e-05, + "loss": 0.0758, + "step": 9755 + }, + { + "epoch": 1.28, + "learning_rate": 1.9130169363258354e-05, + "loss": 0.0822, + "step": 9756 + }, + { + "epoch": 1.28, + "learning_rate": 1.912995301934284e-05, + "loss": 0.0906, + "step": 9757 + }, + { + "epoch": 1.28, + "learning_rate": 1.9129736649749815e-05, + "loss": 0.0811, + "step": 9758 + }, + { + "epoch": 1.28, + "learning_rate": 1.9129520254479894e-05, + "loss": 0.085, + "step": 9759 + }, + { + "epoch": 1.28, + "learning_rate": 1.9129303833533683e-05, + "loss": 0.0772, + "step": 9760 + }, + { + "epoch": 1.28, + "learning_rate": 1.912908738691179e-05, + "loss": 0.0862, + "step": 9761 + }, + { + "epoch": 1.28, + "learning_rate": 1.9128870914614827e-05, + "loss": 0.0864, + "step": 9762 + }, + { + "epoch": 1.28, + "learning_rate": 1.91286544166434e-05, + "loss": 0.084, + "step": 9763 + }, + { + "epoch": 1.28, + "learning_rate": 1.9128437892998114e-05, + "loss": 0.0883, + "step": 9764 + }, + { + "epoch": 1.28, + "learning_rate": 1.9128221343679588e-05, + "loss": 0.0937, + "step": 9765 + }, + { + "epoch": 1.28, + "learning_rate": 1.9128004768688425e-05, + "loss": 0.0763, + "step": 9766 + }, + { + "epoch": 1.28, + "learning_rate": 1.9127788168025235e-05, + "loss": 0.0733, + "step": 9767 + }, + { + "epoch": 1.28, + "learning_rate": 1.9127571541690625e-05, + "loss": 0.0902, + "step": 9768 + }, + { + "epoch": 1.28, + "learning_rate": 1.9127354889685207e-05, + "loss": 0.0822, + "step": 9769 + }, + { + "epoch": 1.28, + "learning_rate": 1.9127138212009588e-05, + "loss": 0.0866, + "step": 9770 + }, + { + "epoch": 1.28, + "learning_rate": 1.912692150866438e-05, + "loss": 0.0806, + "step": 9771 + }, + { + "epoch": 1.28, + "learning_rate": 1.912670477965019e-05, + "loss": 0.0716, + "step": 9772 + }, + { + "epoch": 1.28, + "learning_rate": 1.9126488024967632e-05, + "loss": 0.0806, + "step": 9773 + }, + { + "epoch": 1.28, + "learning_rate": 1.912627124461731e-05, + "loss": 0.0794, + "step": 9774 + }, + { + "epoch": 1.28, + "learning_rate": 1.9126054438599837e-05, + "loss": 0.0765, + "step": 9775 + }, + { + "epoch": 1.28, + "learning_rate": 1.9125837606915824e-05, + "loss": 0.0693, + "step": 9776 + }, + { + "epoch": 1.28, + "learning_rate": 1.9125620749565876e-05, + "loss": 0.0822, + "step": 9777 + }, + { + "epoch": 1.28, + "learning_rate": 1.9125403866550605e-05, + "loss": 0.079, + "step": 9778 + }, + { + "epoch": 1.28, + "learning_rate": 1.9125186957870623e-05, + "loss": 0.0789, + "step": 9779 + }, + { + "epoch": 1.28, + "learning_rate": 1.9124970023526535e-05, + "loss": 0.0781, + "step": 9780 + }, + { + "epoch": 1.28, + "learning_rate": 1.912475306351896e-05, + "loss": 0.0847, + "step": 9781 + }, + { + "epoch": 1.28, + "learning_rate": 1.91245360778485e-05, + "loss": 0.0799, + "step": 9782 + }, + { + "epoch": 1.28, + "learning_rate": 1.9124319066515767e-05, + "loss": 0.0748, + "step": 9783 + }, + { + "epoch": 1.28, + "learning_rate": 1.9124102029521376e-05, + "loss": 0.0808, + "step": 9784 + }, + { + "epoch": 1.28, + "learning_rate": 1.9123884966865932e-05, + "loss": 0.073, + "step": 9785 + }, + { + "epoch": 1.28, + "learning_rate": 1.9123667878550048e-05, + "loss": 0.0843, + "step": 9786 + }, + { + "epoch": 1.28, + "learning_rate": 1.9123450764574335e-05, + "loss": 0.0765, + "step": 9787 + }, + { + "epoch": 1.28, + "learning_rate": 1.91232336249394e-05, + "loss": 0.0887, + "step": 9788 + }, + { + "epoch": 1.28, + "learning_rate": 1.9123016459645858e-05, + "loss": 0.0795, + "step": 9789 + }, + { + "epoch": 1.28, + "learning_rate": 1.9122799268694318e-05, + "loss": 0.0729, + "step": 9790 + }, + { + "epoch": 1.28, + "learning_rate": 1.9122582052085394e-05, + "loss": 0.0844, + "step": 9791 + }, + { + "epoch": 1.28, + "learning_rate": 1.9122364809819686e-05, + "loss": 0.0847, + "step": 9792 + }, + { + "epoch": 1.28, + "learning_rate": 1.912214754189782e-05, + "loss": 0.0803, + "step": 9793 + }, + { + "epoch": 1.28, + "learning_rate": 1.9121930248320396e-05, + "loss": 0.0763, + "step": 9794 + }, + { + "epoch": 1.28, + "learning_rate": 1.912171292908803e-05, + "loss": 0.0697, + "step": 9795 + }, + { + "epoch": 1.28, + "learning_rate": 1.9121495584201334e-05, + "loss": 0.0732, + "step": 9796 + }, + { + "epoch": 1.28, + "learning_rate": 1.9121278213660913e-05, + "loss": 0.0871, + "step": 9797 + }, + { + "epoch": 1.28, + "learning_rate": 1.9121060817467384e-05, + "loss": 0.0714, + "step": 9798 + }, + { + "epoch": 1.28, + "learning_rate": 1.912084339562136e-05, + "loss": 0.0775, + "step": 9799 + }, + { + "epoch": 1.28, + "learning_rate": 1.9120625948123446e-05, + "loss": 0.095, + "step": 9800 + }, + { + "epoch": 1.28, + "learning_rate": 1.912040847497426e-05, + "loss": 0.075, + "step": 9801 + }, + { + "epoch": 1.28, + "learning_rate": 1.9120190976174412e-05, + "loss": 0.0744, + "step": 9802 + }, + { + "epoch": 1.28, + "learning_rate": 1.911997345172451e-05, + "loss": 0.0697, + "step": 9803 + }, + { + "epoch": 1.28, + "learning_rate": 1.911975590162517e-05, + "loss": 0.0697, + "step": 9804 + }, + { + "epoch": 1.28, + "learning_rate": 1.9119538325877e-05, + "loss": 0.0837, + "step": 9805 + }, + { + "epoch": 1.28, + "learning_rate": 1.9119320724480615e-05, + "loss": 0.0893, + "step": 9806 + }, + { + "epoch": 1.28, + "learning_rate": 1.9119103097436625e-05, + "loss": 0.0801, + "step": 9807 + }, + { + "epoch": 1.28, + "learning_rate": 1.9118885444745647e-05, + "loss": 0.0777, + "step": 9808 + }, + { + "epoch": 1.28, + "learning_rate": 1.9118667766408286e-05, + "loss": 0.0755, + "step": 9809 + }, + { + "epoch": 1.29, + "learning_rate": 1.911845006242516e-05, + "loss": 0.0813, + "step": 9810 + }, + { + "epoch": 1.29, + "learning_rate": 1.9118232332796877e-05, + "loss": 0.0926, + "step": 9811 + }, + { + "epoch": 1.29, + "learning_rate": 1.911801457752405e-05, + "loss": 0.073, + "step": 9812 + }, + { + "epoch": 1.29, + "learning_rate": 1.9117796796607295e-05, + "loss": 0.0856, + "step": 9813 + }, + { + "epoch": 1.29, + "learning_rate": 1.9117578990047218e-05, + "loss": 0.0727, + "step": 9814 + }, + { + "epoch": 1.29, + "learning_rate": 1.9117361157844438e-05, + "loss": 0.0874, + "step": 9815 + }, + { + "epoch": 1.29, + "learning_rate": 1.911714329999957e-05, + "loss": 0.0738, + "step": 9816 + }, + { + "epoch": 1.29, + "learning_rate": 1.9116925416513214e-05, + "loss": 0.0787, + "step": 9817 + }, + { + "epoch": 1.29, + "learning_rate": 1.9116707507385995e-05, + "loss": 0.0886, + "step": 9818 + }, + { + "epoch": 1.29, + "learning_rate": 1.911648957261852e-05, + "loss": 0.0781, + "step": 9819 + }, + { + "epoch": 1.29, + "learning_rate": 1.911627161221141e-05, + "loss": 0.077, + "step": 9820 + }, + { + "epoch": 1.29, + "learning_rate": 1.9116053626165264e-05, + "loss": 0.0839, + "step": 9821 + }, + { + "epoch": 1.29, + "learning_rate": 1.9115835614480702e-05, + "loss": 0.0773, + "step": 9822 + }, + { + "epoch": 1.29, + "learning_rate": 1.911561757715834e-05, + "loss": 0.0757, + "step": 9823 + }, + { + "epoch": 1.29, + "learning_rate": 1.9115399514198795e-05, + "loss": 0.0756, + "step": 9824 + }, + { + "epoch": 1.29, + "learning_rate": 1.9115181425602666e-05, + "loss": 0.0845, + "step": 9825 + }, + { + "epoch": 1.29, + "learning_rate": 1.9114963311370576e-05, + "loss": 0.0793, + "step": 9826 + }, + { + "epoch": 1.29, + "learning_rate": 1.911474517150314e-05, + "loss": 0.0854, + "step": 9827 + }, + { + "epoch": 1.29, + "learning_rate": 1.911452700600097e-05, + "loss": 0.079, + "step": 9828 + }, + { + "epoch": 1.29, + "learning_rate": 1.911430881486467e-05, + "loss": 0.0742, + "step": 9829 + }, + { + "epoch": 1.29, + "learning_rate": 1.9114090598094873e-05, + "loss": 0.0881, + "step": 9830 + }, + { + "epoch": 1.29, + "learning_rate": 1.9113872355692174e-05, + "loss": 0.0868, + "step": 9831 + }, + { + "epoch": 1.29, + "learning_rate": 1.9113654087657197e-05, + "loss": 0.0779, + "step": 9832 + }, + { + "epoch": 1.29, + "learning_rate": 1.9113435793990553e-05, + "loss": 0.076, + "step": 9833 + }, + { + "epoch": 1.29, + "learning_rate": 1.9113217474692854e-05, + "loss": 0.0818, + "step": 9834 + }, + { + "epoch": 1.29, + "learning_rate": 1.9112999129764718e-05, + "loss": 0.0817, + "step": 9835 + }, + { + "epoch": 1.29, + "learning_rate": 1.9112780759206754e-05, + "loss": 0.0786, + "step": 9836 + }, + { + "epoch": 1.29, + "learning_rate": 1.9112562363019585e-05, + "loss": 0.0743, + "step": 9837 + }, + { + "epoch": 1.29, + "learning_rate": 1.9112343941203818e-05, + "loss": 0.079, + "step": 9838 + }, + { + "epoch": 1.29, + "learning_rate": 1.9112125493760067e-05, + "loss": 0.0821, + "step": 9839 + }, + { + "epoch": 1.29, + "learning_rate": 1.9111907020688947e-05, + "loss": 0.0759, + "step": 9840 + }, + { + "epoch": 1.29, + "learning_rate": 1.911168852199108e-05, + "loss": 0.0887, + "step": 9841 + }, + { + "epoch": 1.29, + "learning_rate": 1.911146999766707e-05, + "loss": 0.0712, + "step": 9842 + }, + { + "epoch": 1.29, + "learning_rate": 1.9111251447717534e-05, + "loss": 0.072, + "step": 9843 + }, + { + "epoch": 1.29, + "learning_rate": 1.911103287214309e-05, + "loss": 0.0893, + "step": 9844 + }, + { + "epoch": 1.29, + "learning_rate": 1.9110814270944354e-05, + "loss": 0.0914, + "step": 9845 + }, + { + "epoch": 1.29, + "learning_rate": 1.9110595644121935e-05, + "loss": 0.0845, + "step": 9846 + }, + { + "epoch": 1.29, + "learning_rate": 1.911037699167645e-05, + "loss": 0.0798, + "step": 9847 + }, + { + "epoch": 1.29, + "learning_rate": 1.9110158313608518e-05, + "loss": 0.0755, + "step": 9848 + }, + { + "epoch": 1.29, + "learning_rate": 1.910993960991875e-05, + "loss": 0.0839, + "step": 9849 + }, + { + "epoch": 1.29, + "learning_rate": 1.910972088060776e-05, + "loss": 0.0808, + "step": 9850 + }, + { + "epoch": 1.29, + "learning_rate": 1.910950212567617e-05, + "loss": 0.0733, + "step": 9851 + }, + { + "epoch": 1.29, + "learning_rate": 1.9109283345124584e-05, + "loss": 0.0719, + "step": 9852 + }, + { + "epoch": 1.29, + "learning_rate": 1.9109064538953628e-05, + "loss": 0.0735, + "step": 9853 + }, + { + "epoch": 1.29, + "learning_rate": 1.910884570716391e-05, + "loss": 0.0781, + "step": 9854 + }, + { + "epoch": 1.29, + "learning_rate": 1.910862684975605e-05, + "loss": 0.0766, + "step": 9855 + }, + { + "epoch": 1.29, + "learning_rate": 1.9108407966730664e-05, + "loss": 0.077, + "step": 9856 + }, + { + "epoch": 1.29, + "learning_rate": 1.9108189058088365e-05, + "loss": 0.0784, + "step": 9857 + }, + { + "epoch": 1.29, + "learning_rate": 1.9107970123829768e-05, + "loss": 0.0808, + "step": 9858 + }, + { + "epoch": 1.29, + "learning_rate": 1.910775116395549e-05, + "loss": 0.0782, + "step": 9859 + }, + { + "epoch": 1.29, + "learning_rate": 1.910753217846615e-05, + "loss": 0.0846, + "step": 9860 + }, + { + "epoch": 1.29, + "learning_rate": 1.9107313167362355e-05, + "loss": 0.0789, + "step": 9861 + }, + { + "epoch": 1.29, + "learning_rate": 1.9107094130644728e-05, + "loss": 0.0889, + "step": 9862 + }, + { + "epoch": 1.29, + "learning_rate": 1.9106875068313888e-05, + "loss": 0.0906, + "step": 9863 + }, + { + "epoch": 1.29, + "learning_rate": 1.9106655980370443e-05, + "loss": 0.0761, + "step": 9864 + }, + { + "epoch": 1.29, + "learning_rate": 1.9106436866815015e-05, + "loss": 0.0738, + "step": 9865 + }, + { + "epoch": 1.29, + "learning_rate": 1.9106217727648217e-05, + "loss": 0.0927, + "step": 9866 + }, + { + "epoch": 1.29, + "learning_rate": 1.9105998562870667e-05, + "loss": 0.0738, + "step": 9867 + }, + { + "epoch": 1.29, + "learning_rate": 1.910577937248298e-05, + "loss": 0.0853, + "step": 9868 + }, + { + "epoch": 1.29, + "learning_rate": 1.9105560156485775e-05, + "loss": 0.0844, + "step": 9869 + }, + { + "epoch": 1.29, + "learning_rate": 1.9105340914879667e-05, + "loss": 0.071, + "step": 9870 + }, + { + "epoch": 1.29, + "learning_rate": 1.9105121647665273e-05, + "loss": 0.0753, + "step": 9871 + }, + { + "epoch": 1.29, + "learning_rate": 1.910490235484321e-05, + "loss": 0.0708, + "step": 9872 + }, + { + "epoch": 1.29, + "learning_rate": 1.910468303641409e-05, + "loss": 0.0777, + "step": 9873 + }, + { + "epoch": 1.29, + "learning_rate": 1.9104463692378536e-05, + "loss": 0.084, + "step": 9874 + }, + { + "epoch": 1.29, + "learning_rate": 1.9104244322737164e-05, + "loss": 0.0787, + "step": 9875 + }, + { + "epoch": 1.29, + "learning_rate": 1.910402492749059e-05, + "loss": 0.082, + "step": 9876 + }, + { + "epoch": 1.29, + "learning_rate": 1.9103805506639427e-05, + "loss": 0.081, + "step": 9877 + }, + { + "epoch": 1.29, + "learning_rate": 1.91035860601843e-05, + "loss": 0.0793, + "step": 9878 + }, + { + "epoch": 1.29, + "learning_rate": 1.9103366588125815e-05, + "loss": 0.075, + "step": 9879 + }, + { + "epoch": 1.29, + "learning_rate": 1.9103147090464605e-05, + "loss": 0.0942, + "step": 9880 + }, + { + "epoch": 1.29, + "learning_rate": 1.910292756720127e-05, + "loss": 0.0739, + "step": 9881 + }, + { + "epoch": 1.29, + "learning_rate": 1.9102708018336445e-05, + "loss": 0.0869, + "step": 9882 + }, + { + "epoch": 1.29, + "learning_rate": 1.9102488443870736e-05, + "loss": 0.083, + "step": 9883 + }, + { + "epoch": 1.29, + "learning_rate": 1.910226884380476e-05, + "loss": 0.0712, + "step": 9884 + }, + { + "epoch": 1.29, + "learning_rate": 1.9102049218139138e-05, + "loss": 0.0707, + "step": 9885 + }, + { + "epoch": 1.29, + "learning_rate": 1.910182956687449e-05, + "loss": 0.0808, + "step": 9886 + }, + { + "epoch": 1.3, + "learning_rate": 1.9101609890011428e-05, + "loss": 0.0846, + "step": 9887 + }, + { + "epoch": 1.3, + "learning_rate": 1.9101390187550573e-05, + "loss": 0.0771, + "step": 9888 + }, + { + "epoch": 1.3, + "learning_rate": 1.9101170459492545e-05, + "loss": 0.0924, + "step": 9889 + }, + { + "epoch": 1.3, + "learning_rate": 1.9100950705837958e-05, + "loss": 0.0848, + "step": 9890 + }, + { + "epoch": 1.3, + "learning_rate": 1.910073092658743e-05, + "loss": 0.0826, + "step": 9891 + }, + { + "epoch": 1.3, + "learning_rate": 1.9100511121741584e-05, + "loss": 0.078, + "step": 9892 + }, + { + "epoch": 1.3, + "learning_rate": 1.9100291291301034e-05, + "loss": 0.0748, + "step": 9893 + }, + { + "epoch": 1.3, + "learning_rate": 1.9100071435266404e-05, + "loss": 0.0754, + "step": 9894 + }, + { + "epoch": 1.3, + "learning_rate": 1.90998515536383e-05, + "loss": 0.0796, + "step": 9895 + }, + { + "epoch": 1.3, + "learning_rate": 1.9099631646417355e-05, + "loss": 0.087, + "step": 9896 + }, + { + "epoch": 1.3, + "learning_rate": 1.9099411713604176e-05, + "loss": 0.0838, + "step": 9897 + }, + { + "epoch": 1.3, + "learning_rate": 1.9099191755199387e-05, + "loss": 0.0651, + "step": 9898 + }, + { + "epoch": 1.3, + "learning_rate": 1.9098971771203605e-05, + "loss": 0.0917, + "step": 9899 + }, + { + "epoch": 1.3, + "learning_rate": 1.909875176161745e-05, + "loss": 0.0813, + "step": 9900 + }, + { + "epoch": 1.3, + "learning_rate": 1.909853172644154e-05, + "loss": 0.08, + "step": 9901 + }, + { + "epoch": 1.3, + "learning_rate": 1.9098311665676498e-05, + "loss": 0.0845, + "step": 9902 + }, + { + "epoch": 1.3, + "learning_rate": 1.9098091579322934e-05, + "loss": 0.0845, + "step": 9903 + }, + { + "epoch": 1.3, + "learning_rate": 1.9097871467381474e-05, + "loss": 0.0766, + "step": 9904 + }, + { + "epoch": 1.3, + "learning_rate": 1.9097651329852734e-05, + "loss": 0.0711, + "step": 9905 + }, + { + "epoch": 1.3, + "learning_rate": 1.9097431166737334e-05, + "loss": 0.0777, + "step": 9906 + }, + { + "epoch": 1.3, + "learning_rate": 1.9097210978035894e-05, + "loss": 0.0822, + "step": 9907 + }, + { + "epoch": 1.3, + "learning_rate": 1.9096990763749032e-05, + "loss": 0.081, + "step": 9908 + }, + { + "epoch": 1.3, + "learning_rate": 1.9096770523877367e-05, + "loss": 0.0734, + "step": 9909 + }, + { + "epoch": 1.3, + "learning_rate": 1.9096550258421522e-05, + "loss": 0.0766, + "step": 9910 + }, + { + "epoch": 1.3, + "learning_rate": 1.9096329967382114e-05, + "loss": 0.0825, + "step": 9911 + }, + { + "epoch": 1.3, + "learning_rate": 1.909610965075976e-05, + "loss": 0.0837, + "step": 9912 + }, + { + "epoch": 1.3, + "learning_rate": 1.9095889308555083e-05, + "loss": 0.0688, + "step": 9913 + }, + { + "epoch": 1.3, + "learning_rate": 1.9095668940768703e-05, + "loss": 0.0782, + "step": 9914 + }, + { + "epoch": 1.3, + "learning_rate": 1.9095448547401237e-05, + "loss": 0.0678, + "step": 9915 + }, + { + "epoch": 1.3, + "learning_rate": 1.909522812845331e-05, + "loss": 0.0812, + "step": 9916 + }, + { + "epoch": 1.3, + "learning_rate": 1.909500768392553e-05, + "loss": 0.0821, + "step": 9917 + }, + { + "epoch": 1.3, + "learning_rate": 1.909478721381853e-05, + "loss": 0.0877, + "step": 9918 + }, + { + "epoch": 1.3, + "learning_rate": 1.9094566718132926e-05, + "loss": 0.0868, + "step": 9919 + }, + { + "epoch": 1.3, + "learning_rate": 1.9094346196869336e-05, + "loss": 0.0758, + "step": 9920 + }, + { + "epoch": 1.3, + "learning_rate": 1.9094125650028384e-05, + "loss": 0.0902, + "step": 9921 + }, + { + "epoch": 1.3, + "learning_rate": 1.9093905077610685e-05, + "loss": 0.083, + "step": 9922 + }, + { + "epoch": 1.3, + "learning_rate": 1.9093684479616864e-05, + "loss": 0.0819, + "step": 9923 + }, + { + "epoch": 1.3, + "learning_rate": 1.9093463856047538e-05, + "loss": 0.0966, + "step": 9924 + }, + { + "epoch": 1.3, + "learning_rate": 1.909324320690333e-05, + "loss": 0.0803, + "step": 9925 + }, + { + "epoch": 1.3, + "learning_rate": 1.9093022532184863e-05, + "loss": 0.0837, + "step": 9926 + }, + { + "epoch": 1.3, + "learning_rate": 1.909280183189275e-05, + "loss": 0.0772, + "step": 9927 + }, + { + "epoch": 1.3, + "learning_rate": 1.9092581106027616e-05, + "loss": 0.0852, + "step": 9928 + }, + { + "epoch": 1.3, + "learning_rate": 1.9092360354590083e-05, + "loss": 0.072, + "step": 9929 + }, + { + "epoch": 1.3, + "learning_rate": 1.9092139577580772e-05, + "loss": 0.0829, + "step": 9930 + }, + { + "epoch": 1.3, + "learning_rate": 1.90919187750003e-05, + "loss": 0.0795, + "step": 9931 + }, + { + "epoch": 1.3, + "learning_rate": 1.9091697946849296e-05, + "loss": 0.0847, + "step": 9932 + }, + { + "epoch": 1.3, + "learning_rate": 1.909147709312837e-05, + "loss": 0.0751, + "step": 9933 + }, + { + "epoch": 1.3, + "learning_rate": 1.909125621383815e-05, + "loss": 0.0874, + "step": 9934 + }, + { + "epoch": 1.3, + "learning_rate": 1.909103530897926e-05, + "loss": 0.0838, + "step": 9935 + }, + { + "epoch": 1.3, + "learning_rate": 1.909081437855231e-05, + "loss": 0.079, + "step": 9936 + }, + { + "epoch": 1.3, + "learning_rate": 1.9090593422557935e-05, + "loss": 0.0725, + "step": 9937 + }, + { + "epoch": 1.3, + "learning_rate": 1.9090372440996746e-05, + "loss": 0.0772, + "step": 9938 + }, + { + "epoch": 1.3, + "learning_rate": 1.9090151433869373e-05, + "loss": 0.0937, + "step": 9939 + }, + { + "epoch": 1.3, + "learning_rate": 1.9089930401176427e-05, + "loss": 0.0791, + "step": 9940 + }, + { + "epoch": 1.3, + "learning_rate": 1.9089709342918543e-05, + "loss": 0.079, + "step": 9941 + }, + { + "epoch": 1.3, + "learning_rate": 1.9089488259096328e-05, + "loss": 0.0991, + "step": 9942 + }, + { + "epoch": 1.3, + "learning_rate": 1.9089267149710417e-05, + "loss": 0.076, + "step": 9943 + }, + { + "epoch": 1.3, + "learning_rate": 1.9089046014761425e-05, + "loss": 0.0899, + "step": 9944 + }, + { + "epoch": 1.3, + "learning_rate": 1.9088824854249975e-05, + "loss": 0.0781, + "step": 9945 + }, + { + "epoch": 1.3, + "learning_rate": 1.908860366817669e-05, + "loss": 0.0821, + "step": 9946 + }, + { + "epoch": 1.3, + "learning_rate": 1.9088382456542186e-05, + "loss": 0.086, + "step": 9947 + }, + { + "epoch": 1.3, + "learning_rate": 1.9088161219347098e-05, + "loss": 0.0804, + "step": 9948 + }, + { + "epoch": 1.3, + "learning_rate": 1.9087939956592035e-05, + "loss": 0.081, + "step": 9949 + }, + { + "epoch": 1.3, + "learning_rate": 1.9087718668277628e-05, + "loss": 0.0785, + "step": 9950 + }, + { + "epoch": 1.3, + "learning_rate": 1.9087497354404492e-05, + "loss": 0.0873, + "step": 9951 + }, + { + "epoch": 1.3, + "learning_rate": 1.9087276014973256e-05, + "loss": 0.0825, + "step": 9952 + }, + { + "epoch": 1.3, + "learning_rate": 1.9087054649984545e-05, + "loss": 0.071, + "step": 9953 + }, + { + "epoch": 1.3, + "learning_rate": 1.908683325943897e-05, + "loss": 0.0744, + "step": 9954 + }, + { + "epoch": 1.3, + "learning_rate": 1.9086611843337164e-05, + "loss": 0.0828, + "step": 9955 + }, + { + "epoch": 1.3, + "learning_rate": 1.9086390401679744e-05, + "loss": 0.0833, + "step": 9956 + }, + { + "epoch": 1.3, + "learning_rate": 1.9086168934467337e-05, + "loss": 0.0806, + "step": 9957 + }, + { + "epoch": 1.3, + "learning_rate": 1.9085947441700558e-05, + "loss": 0.0811, + "step": 9958 + }, + { + "epoch": 1.3, + "learning_rate": 1.908572592338004e-05, + "loss": 0.0751, + "step": 9959 + }, + { + "epoch": 1.3, + "learning_rate": 1.9085504379506403e-05, + "loss": 0.0819, + "step": 9960 + }, + { + "epoch": 1.3, + "learning_rate": 1.9085282810080267e-05, + "loss": 0.0875, + "step": 9961 + }, + { + "epoch": 1.3, + "learning_rate": 1.908506121510226e-05, + "loss": 0.0791, + "step": 9962 + }, + { + "epoch": 1.31, + "learning_rate": 1.9084839594573e-05, + "loss": 0.0712, + "step": 9963 + }, + { + "epoch": 1.31, + "learning_rate": 1.908461794849311e-05, + "loss": 0.0745, + "step": 9964 + }, + { + "epoch": 1.31, + "learning_rate": 1.9084396276863217e-05, + "loss": 0.0786, + "step": 9965 + }, + { + "epoch": 1.31, + "learning_rate": 1.9084174579683946e-05, + "loss": 0.0839, + "step": 9966 + }, + { + "epoch": 1.31, + "learning_rate": 1.9083952856955915e-05, + "loss": 0.0786, + "step": 9967 + }, + { + "epoch": 1.31, + "learning_rate": 1.908373110867975e-05, + "loss": 0.0842, + "step": 9968 + }, + { + "epoch": 1.31, + "learning_rate": 1.9083509334856077e-05, + "loss": 0.0739, + "step": 9969 + }, + { + "epoch": 1.31, + "learning_rate": 1.9083287535485517e-05, + "loss": 0.0942, + "step": 9970 + }, + { + "epoch": 1.31, + "learning_rate": 1.9083065710568693e-05, + "loss": 0.0836, + "step": 9971 + }, + { + "epoch": 1.31, + "learning_rate": 1.9082843860106234e-05, + "loss": 0.0796, + "step": 9972 + }, + { + "epoch": 1.31, + "learning_rate": 1.9082621984098755e-05, + "loss": 0.0852, + "step": 9973 + }, + { + "epoch": 1.31, + "learning_rate": 1.9082400082546887e-05, + "loss": 0.0709, + "step": 9974 + }, + { + "epoch": 1.31, + "learning_rate": 1.9082178155451256e-05, + "loss": 0.0713, + "step": 9975 + }, + { + "epoch": 1.31, + "learning_rate": 1.908195620281248e-05, + "loss": 0.0807, + "step": 9976 + }, + { + "epoch": 1.31, + "learning_rate": 1.9081734224631186e-05, + "loss": 0.0744, + "step": 9977 + }, + { + "epoch": 1.31, + "learning_rate": 1.9081512220907998e-05, + "loss": 0.0863, + "step": 9978 + }, + { + "epoch": 1.31, + "learning_rate": 1.908129019164354e-05, + "loss": 0.0847, + "step": 9979 + }, + { + "epoch": 1.31, + "learning_rate": 1.9081068136838442e-05, + "loss": 0.0797, + "step": 9980 + }, + { + "epoch": 1.31, + "learning_rate": 1.9080846056493318e-05, + "loss": 0.0874, + "step": 9981 + }, + { + "epoch": 1.31, + "learning_rate": 1.9080623950608798e-05, + "loss": 0.0763, + "step": 9982 + }, + { + "epoch": 1.31, + "learning_rate": 1.9080401819185506e-05, + "loss": 0.075, + "step": 9983 + }, + { + "epoch": 1.31, + "learning_rate": 1.9080179662224073e-05, + "loss": 0.0754, + "step": 9984 + }, + { + "epoch": 1.31, + "learning_rate": 1.9079957479725112e-05, + "loss": 0.0883, + "step": 9985 + }, + { + "epoch": 1.31, + "learning_rate": 1.9079735271689255e-05, + "loss": 0.0892, + "step": 9986 + }, + { + "epoch": 1.31, + "learning_rate": 1.907951303811713e-05, + "loss": 0.0905, + "step": 9987 + }, + { + "epoch": 1.31, + "learning_rate": 1.907929077900936e-05, + "loss": 0.0933, + "step": 9988 + }, + { + "epoch": 1.31, + "learning_rate": 1.9079068494366562e-05, + "loss": 0.0871, + "step": 9989 + }, + { + "epoch": 1.31, + "learning_rate": 1.9078846184189368e-05, + "loss": 0.0752, + "step": 9990 + }, + { + "epoch": 1.31, + "learning_rate": 1.9078623848478406e-05, + "loss": 0.0837, + "step": 9991 + }, + { + "epoch": 1.31, + "learning_rate": 1.9078401487234296e-05, + "loss": 0.0809, + "step": 9992 + }, + { + "epoch": 1.31, + "learning_rate": 1.9078179100457666e-05, + "loss": 0.0832, + "step": 9993 + }, + { + "epoch": 1.31, + "learning_rate": 1.907795668814914e-05, + "loss": 0.0768, + "step": 9994 + }, + { + "epoch": 1.31, + "learning_rate": 1.9077734250309344e-05, + "loss": 0.0806, + "step": 9995 + }, + { + "epoch": 1.31, + "learning_rate": 1.9077511786938902e-05, + "loss": 0.0774, + "step": 9996 + }, + { + "epoch": 1.31, + "learning_rate": 1.9077289298038444e-05, + "loss": 0.0803, + "step": 9997 + }, + { + "epoch": 1.31, + "learning_rate": 1.907706678360859e-05, + "loss": 0.0859, + "step": 9998 + }, + { + "epoch": 1.31, + "learning_rate": 1.9076844243649973e-05, + "loss": 0.072, + "step": 9999 + }, + { + "epoch": 1.31, + "learning_rate": 1.9076621678163218e-05, + "loss": 0.0875, + "step": 10000 + }, + { + "epoch": 1.31, + "learning_rate": 1.9076399087148943e-05, + "loss": 0.0785, + "step": 10001 + }, + { + "epoch": 1.31, + "learning_rate": 1.9076176470607778e-05, + "loss": 0.0698, + "step": 10002 + }, + { + "epoch": 1.31, + "learning_rate": 1.907595382854035e-05, + "loss": 0.0829, + "step": 10003 + }, + { + "epoch": 1.31, + "learning_rate": 1.9075731160947287e-05, + "loss": 0.0899, + "step": 10004 + }, + { + "epoch": 1.31, + "learning_rate": 1.9075508467829215e-05, + "loss": 0.092, + "step": 10005 + }, + { + "epoch": 1.31, + "learning_rate": 1.9075285749186756e-05, + "loss": 0.0724, + "step": 10006 + }, + { + "epoch": 1.31, + "learning_rate": 1.907506300502054e-05, + "loss": 0.0739, + "step": 10007 + }, + { + "epoch": 1.31, + "learning_rate": 1.9074840235331193e-05, + "loss": 0.0721, + "step": 10008 + }, + { + "epoch": 1.31, + "learning_rate": 1.907461744011934e-05, + "loss": 0.0733, + "step": 10009 + }, + { + "epoch": 1.31, + "learning_rate": 1.907439461938561e-05, + "loss": 0.0943, + "step": 10010 + }, + { + "epoch": 1.31, + "learning_rate": 1.907417177313063e-05, + "loss": 0.0832, + "step": 10011 + }, + { + "epoch": 1.31, + "learning_rate": 1.907394890135502e-05, + "loss": 0.0811, + "step": 10012 + }, + { + "epoch": 1.31, + "learning_rate": 1.9073726004059417e-05, + "loss": 0.0632, + "step": 10013 + }, + { + "epoch": 1.31, + "learning_rate": 1.907350308124444e-05, + "loss": 0.0812, + "step": 10014 + }, + { + "epoch": 1.31, + "learning_rate": 1.907328013291072e-05, + "loss": 0.0795, + "step": 10015 + }, + { + "epoch": 1.31, + "learning_rate": 1.907305715905888e-05, + "loss": 0.0696, + "step": 10016 + }, + { + "epoch": 1.31, + "learning_rate": 1.9072834159689554e-05, + "loss": 0.0916, + "step": 10017 + }, + { + "epoch": 1.31, + "learning_rate": 1.9072611134803363e-05, + "loss": 0.0871, + "step": 10018 + }, + { + "epoch": 1.31, + "learning_rate": 1.9072388084400936e-05, + "loss": 0.0696, + "step": 10019 + }, + { + "epoch": 1.31, + "learning_rate": 1.90721650084829e-05, + "loss": 0.0806, + "step": 10020 + }, + { + "epoch": 1.31, + "learning_rate": 1.9071941907049885e-05, + "loss": 0.0893, + "step": 10021 + }, + { + "epoch": 1.31, + "learning_rate": 1.9071718780102516e-05, + "loss": 0.0787, + "step": 10022 + }, + { + "epoch": 1.31, + "learning_rate": 1.907149562764142e-05, + "loss": 0.0848, + "step": 10023 + }, + { + "epoch": 1.31, + "learning_rate": 1.9071272449667225e-05, + "loss": 0.0816, + "step": 10024 + }, + { + "epoch": 1.31, + "learning_rate": 1.9071049246180563e-05, + "loss": 0.0731, + "step": 10025 + }, + { + "epoch": 1.31, + "learning_rate": 1.9070826017182055e-05, + "loss": 0.0772, + "step": 10026 + }, + { + "epoch": 1.31, + "learning_rate": 1.9070602762672334e-05, + "loss": 0.0751, + "step": 10027 + }, + { + "epoch": 1.31, + "learning_rate": 1.9070379482652023e-05, + "loss": 0.0813, + "step": 10028 + }, + { + "epoch": 1.31, + "learning_rate": 1.9070156177121754e-05, + "loss": 0.0732, + "step": 10029 + }, + { + "epoch": 1.31, + "learning_rate": 1.9069932846082156e-05, + "loss": 0.081, + "step": 10030 + }, + { + "epoch": 1.31, + "learning_rate": 1.906970948953385e-05, + "loss": 0.0758, + "step": 10031 + }, + { + "epoch": 1.31, + "learning_rate": 1.9069486107477476e-05, + "loss": 0.0828, + "step": 10032 + }, + { + "epoch": 1.31, + "learning_rate": 1.9069262699913648e-05, + "loss": 0.0772, + "step": 10033 + }, + { + "epoch": 1.31, + "learning_rate": 1.9069039266843005e-05, + "loss": 0.076, + "step": 10034 + }, + { + "epoch": 1.31, + "learning_rate": 1.906881580826617e-05, + "loss": 0.073, + "step": 10035 + }, + { + "epoch": 1.31, + "learning_rate": 1.906859232418378e-05, + "loss": 0.0829, + "step": 10036 + }, + { + "epoch": 1.31, + "learning_rate": 1.906836881459645e-05, + "loss": 0.0692, + "step": 10037 + }, + { + "epoch": 1.31, + "learning_rate": 1.906814527950482e-05, + "loss": 0.0777, + "step": 10038 + }, + { + "epoch": 1.32, + "learning_rate": 1.906792171890951e-05, + "loss": 0.0802, + "step": 10039 + }, + { + "epoch": 1.32, + "learning_rate": 1.9067698132811156e-05, + "loss": 0.0815, + "step": 10040 + }, + { + "epoch": 1.32, + "learning_rate": 1.9067474521210383e-05, + "loss": 0.0901, + "step": 10041 + }, + { + "epoch": 1.32, + "learning_rate": 1.9067250884107823e-05, + "loss": 0.0817, + "step": 10042 + }, + { + "epoch": 1.32, + "learning_rate": 1.90670272215041e-05, + "loss": 0.0824, + "step": 10043 + }, + { + "epoch": 1.32, + "learning_rate": 1.9066803533399845e-05, + "loss": 0.085, + "step": 10044 + }, + { + "epoch": 1.32, + "learning_rate": 1.906657981979569e-05, + "loss": 0.0758, + "step": 10045 + }, + { + "epoch": 1.32, + "learning_rate": 1.9066356080692262e-05, + "loss": 0.0842, + "step": 10046 + }, + { + "epoch": 1.32, + "learning_rate": 1.906613231609019e-05, + "loss": 0.0841, + "step": 10047 + }, + { + "epoch": 1.32, + "learning_rate": 1.90659085259901e-05, + "loss": 0.0806, + "step": 10048 + }, + { + "epoch": 1.32, + "learning_rate": 1.906568471039263e-05, + "loss": 0.0814, + "step": 10049 + }, + { + "epoch": 1.32, + "learning_rate": 1.9065460869298404e-05, + "loss": 0.0722, + "step": 10050 + }, + { + "epoch": 1.32, + "learning_rate": 1.906523700270805e-05, + "loss": 0.0793, + "step": 10051 + }, + { + "epoch": 1.32, + "learning_rate": 1.90650131106222e-05, + "loss": 0.0811, + "step": 10052 + }, + { + "epoch": 1.32, + "learning_rate": 1.9064789193041483e-05, + "loss": 0.0787, + "step": 10053 + }, + { + "epoch": 1.32, + "learning_rate": 1.9064565249966532e-05, + "loss": 0.0775, + "step": 10054 + }, + { + "epoch": 1.32, + "learning_rate": 1.9064341281397974e-05, + "loss": 0.0757, + "step": 10055 + }, + { + "epoch": 1.32, + "learning_rate": 1.9064117287336438e-05, + "loss": 0.0756, + "step": 10056 + }, + { + "epoch": 1.32, + "learning_rate": 1.906389326778255e-05, + "loss": 0.0813, + "step": 10057 + }, + { + "epoch": 1.32, + "learning_rate": 1.906366922273695e-05, + "loss": 0.073, + "step": 10058 + }, + { + "epoch": 1.32, + "learning_rate": 1.906344515220026e-05, + "loss": 0.0826, + "step": 10059 + }, + { + "epoch": 1.32, + "learning_rate": 1.906322105617312e-05, + "loss": 0.0648, + "step": 10060 + }, + { + "epoch": 1.32, + "learning_rate": 1.9062996934656147e-05, + "loss": 0.0678, + "step": 10061 + }, + { + "epoch": 1.32, + "learning_rate": 1.9062772787649978e-05, + "loss": 0.0841, + "step": 10062 + }, + { + "epoch": 1.32, + "learning_rate": 1.9062548615155246e-05, + "loss": 0.0793, + "step": 10063 + }, + { + "epoch": 1.32, + "learning_rate": 1.9062324417172574e-05, + "loss": 0.0845, + "step": 10064 + }, + { + "epoch": 1.32, + "learning_rate": 1.9062100193702603e-05, + "loss": 0.0758, + "step": 10065 + }, + { + "epoch": 1.32, + "learning_rate": 1.9061875944745953e-05, + "loss": 0.078, + "step": 10066 + }, + { + "epoch": 1.32, + "learning_rate": 1.9061651670303263e-05, + "loss": 0.0809, + "step": 10067 + }, + { + "epoch": 1.32, + "learning_rate": 1.9061427370375155e-05, + "loss": 0.0797, + "step": 10068 + }, + { + "epoch": 1.32, + "learning_rate": 1.9061203044962272e-05, + "loss": 0.0869, + "step": 10069 + }, + { + "epoch": 1.32, + "learning_rate": 1.9060978694065234e-05, + "loss": 0.082, + "step": 10070 + }, + { + "epoch": 1.32, + "learning_rate": 1.9060754317684677e-05, + "loss": 0.0855, + "step": 10071 + }, + { + "epoch": 1.32, + "learning_rate": 1.9060529915821232e-05, + "loss": 0.0984, + "step": 10072 + }, + { + "epoch": 1.32, + "learning_rate": 1.9060305488475527e-05, + "loss": 0.078, + "step": 10073 + }, + { + "epoch": 1.32, + "learning_rate": 1.9060081035648193e-05, + "loss": 0.0958, + "step": 10074 + }, + { + "epoch": 1.32, + "learning_rate": 1.9059856557339866e-05, + "loss": 0.0761, + "step": 10075 + }, + { + "epoch": 1.32, + "learning_rate": 1.905963205355118e-05, + "loss": 0.0807, + "step": 10076 + }, + { + "epoch": 1.32, + "learning_rate": 1.9059407524282754e-05, + "loss": 0.079, + "step": 10077 + }, + { + "epoch": 1.32, + "learning_rate": 1.905918296953523e-05, + "loss": 0.0693, + "step": 10078 + }, + { + "epoch": 1.32, + "learning_rate": 1.9058958389309238e-05, + "loss": 0.0899, + "step": 10079 + }, + { + "epoch": 1.32, + "learning_rate": 1.9058733783605404e-05, + "loss": 0.074, + "step": 10080 + }, + { + "epoch": 1.32, + "learning_rate": 1.9058509152424363e-05, + "loss": 0.0715, + "step": 10081 + }, + { + "epoch": 1.32, + "learning_rate": 1.905828449576675e-05, + "loss": 0.084, + "step": 10082 + }, + { + "epoch": 1.32, + "learning_rate": 1.9058059813633196e-05, + "loss": 0.0741, + "step": 10083 + }, + { + "epoch": 1.32, + "learning_rate": 1.905783510602433e-05, + "loss": 0.0736, + "step": 10084 + }, + { + "epoch": 1.32, + "learning_rate": 1.9057610372940783e-05, + "loss": 0.086, + "step": 10085 + }, + { + "epoch": 1.32, + "learning_rate": 1.905738561438319e-05, + "loss": 0.0777, + "step": 10086 + }, + { + "epoch": 1.32, + "learning_rate": 1.9057160830352183e-05, + "loss": 0.0862, + "step": 10087 + }, + { + "epoch": 1.32, + "learning_rate": 1.9056936020848394e-05, + "loss": 0.0774, + "step": 10088 + }, + { + "epoch": 1.32, + "learning_rate": 1.9056711185872453e-05, + "loss": 0.0672, + "step": 10089 + }, + { + "epoch": 1.32, + "learning_rate": 1.9056486325424998e-05, + "loss": 0.0883, + "step": 10090 + }, + { + "epoch": 1.32, + "learning_rate": 1.9056261439506652e-05, + "loss": 0.0822, + "step": 10091 + }, + { + "epoch": 1.32, + "learning_rate": 1.9056036528118057e-05, + "loss": 0.0821, + "step": 10092 + }, + { + "epoch": 1.32, + "learning_rate": 1.9055811591259837e-05, + "loss": 0.0837, + "step": 10093 + }, + { + "epoch": 1.32, + "learning_rate": 1.9055586628932635e-05, + "loss": 0.0863, + "step": 10094 + }, + { + "epoch": 1.32, + "learning_rate": 1.9055361641137073e-05, + "loss": 0.0737, + "step": 10095 + }, + { + "epoch": 1.32, + "learning_rate": 1.9055136627873792e-05, + "loss": 0.0756, + "step": 10096 + }, + { + "epoch": 1.32, + "learning_rate": 1.905491158914342e-05, + "loss": 0.0866, + "step": 10097 + }, + { + "epoch": 1.32, + "learning_rate": 1.9054686524946593e-05, + "loss": 0.0838, + "step": 10098 + }, + { + "epoch": 1.32, + "learning_rate": 1.905446143528394e-05, + "loss": 0.0729, + "step": 10099 + }, + { + "epoch": 1.32, + "learning_rate": 1.9054236320156096e-05, + "loss": 0.0775, + "step": 10100 + }, + { + "epoch": 1.32, + "learning_rate": 1.9054011179563695e-05, + "loss": 0.0744, + "step": 10101 + }, + { + "epoch": 1.32, + "learning_rate": 1.9053786013507372e-05, + "loss": 0.0734, + "step": 10102 + }, + { + "epoch": 1.32, + "learning_rate": 1.905356082198776e-05, + "loss": 0.0794, + "step": 10103 + }, + { + "epoch": 1.32, + "learning_rate": 1.9053335605005485e-05, + "loss": 0.0795, + "step": 10104 + }, + { + "epoch": 1.32, + "learning_rate": 1.9053110362561184e-05, + "loss": 0.0817, + "step": 10105 + }, + { + "epoch": 1.32, + "learning_rate": 1.90528850946555e-05, + "loss": 0.0719, + "step": 10106 + }, + { + "epoch": 1.32, + "learning_rate": 1.9052659801289054e-05, + "loss": 0.0751, + "step": 10107 + }, + { + "epoch": 1.32, + "learning_rate": 1.9052434482462483e-05, + "loss": 0.0835, + "step": 10108 + }, + { + "epoch": 1.32, + "learning_rate": 1.9052209138176425e-05, + "loss": 0.078, + "step": 10109 + }, + { + "epoch": 1.32, + "learning_rate": 1.905198376843151e-05, + "loss": 0.0955, + "step": 10110 + }, + { + "epoch": 1.32, + "learning_rate": 1.9051758373228373e-05, + "loss": 0.0865, + "step": 10111 + }, + { + "epoch": 1.32, + "learning_rate": 1.9051532952567645e-05, + "loss": 0.0804, + "step": 10112 + }, + { + "epoch": 1.32, + "learning_rate": 1.9051307506449965e-05, + "loss": 0.083, + "step": 10113 + }, + { + "epoch": 1.32, + "learning_rate": 1.9051082034875964e-05, + "loss": 0.0771, + "step": 10114 + }, + { + "epoch": 1.32, + "learning_rate": 1.9050856537846278e-05, + "loss": 0.0769, + "step": 10115 + }, + { + "epoch": 1.33, + "learning_rate": 1.905063101536154e-05, + "loss": 0.0611, + "step": 10116 + }, + { + "epoch": 1.33, + "learning_rate": 1.9050405467422383e-05, + "loss": 0.0692, + "step": 10117 + }, + { + "epoch": 1.33, + "learning_rate": 1.905017989402944e-05, + "loss": 0.0804, + "step": 10118 + }, + { + "epoch": 1.33, + "learning_rate": 1.9049954295183352e-05, + "loss": 0.0742, + "step": 10119 + }, + { + "epoch": 1.33, + "learning_rate": 1.904972867088475e-05, + "loss": 0.0784, + "step": 10120 + }, + { + "epoch": 1.33, + "learning_rate": 1.9049503021134268e-05, + "loss": 0.0803, + "step": 10121 + }, + { + "epoch": 1.33, + "learning_rate": 1.9049277345932536e-05, + "loss": 0.0858, + "step": 10122 + }, + { + "epoch": 1.33, + "learning_rate": 1.90490516452802e-05, + "loss": 0.0919, + "step": 10123 + }, + { + "epoch": 1.33, + "learning_rate": 1.9048825919177886e-05, + "loss": 0.0742, + "step": 10124 + }, + { + "epoch": 1.33, + "learning_rate": 1.9048600167626227e-05, + "loss": 0.0882, + "step": 10125 + }, + { + "epoch": 1.33, + "learning_rate": 1.9048374390625867e-05, + "loss": 0.0853, + "step": 10126 + }, + { + "epoch": 1.33, + "learning_rate": 1.904814858817743e-05, + "loss": 0.0798, + "step": 10127 + }, + { + "epoch": 1.33, + "learning_rate": 1.9047922760281562e-05, + "loss": 0.0791, + "step": 10128 + }, + { + "epoch": 1.33, + "learning_rate": 1.904769690693889e-05, + "loss": 0.0709, + "step": 10129 + }, + { + "epoch": 1.33, + "learning_rate": 1.9047471028150057e-05, + "loss": 0.0737, + "step": 10130 + }, + { + "epoch": 1.33, + "learning_rate": 1.9047245123915688e-05, + "loss": 0.0727, + "step": 10131 + }, + { + "epoch": 1.33, + "learning_rate": 1.9047019194236426e-05, + "loss": 0.0821, + "step": 10132 + }, + { + "epoch": 1.33, + "learning_rate": 1.9046793239112904e-05, + "loss": 0.0698, + "step": 10133 + }, + { + "epoch": 1.33, + "learning_rate": 1.904656725854576e-05, + "loss": 0.0895, + "step": 10134 + }, + { + "epoch": 1.33, + "learning_rate": 1.9046341252535626e-05, + "loss": 0.0821, + "step": 10135 + }, + { + "epoch": 1.33, + "learning_rate": 1.9046115221083136e-05, + "loss": 0.0788, + "step": 10136 + }, + { + "epoch": 1.33, + "learning_rate": 1.9045889164188934e-05, + "loss": 0.0777, + "step": 10137 + }, + { + "epoch": 1.33, + "learning_rate": 1.9045663081853644e-05, + "loss": 0.0891, + "step": 10138 + }, + { + "epoch": 1.33, + "learning_rate": 1.9045436974077912e-05, + "loss": 0.0811, + "step": 10139 + }, + { + "epoch": 1.33, + "learning_rate": 1.9045210840862372e-05, + "loss": 0.0933, + "step": 10140 + }, + { + "epoch": 1.33, + "learning_rate": 1.9044984682207655e-05, + "loss": 0.0776, + "step": 10141 + }, + { + "epoch": 1.33, + "learning_rate": 1.90447584981144e-05, + "loss": 0.0716, + "step": 10142 + }, + { + "epoch": 1.33, + "learning_rate": 1.9044532288583245e-05, + "loss": 0.0738, + "step": 10143 + }, + { + "epoch": 1.33, + "learning_rate": 1.9044306053614824e-05, + "loss": 0.0778, + "step": 10144 + }, + { + "epoch": 1.33, + "learning_rate": 1.904407979320978e-05, + "loss": 0.0866, + "step": 10145 + }, + { + "epoch": 1.33, + "learning_rate": 1.9043853507368735e-05, + "loss": 0.0849, + "step": 10146 + }, + { + "epoch": 1.33, + "learning_rate": 1.9043627196092334e-05, + "loss": 0.0737, + "step": 10147 + }, + { + "epoch": 1.33, + "learning_rate": 1.904340085938122e-05, + "loss": 0.0795, + "step": 10148 + }, + { + "epoch": 1.33, + "learning_rate": 1.9043174497236014e-05, + "loss": 0.0732, + "step": 10149 + }, + { + "epoch": 1.33, + "learning_rate": 1.9042948109657368e-05, + "loss": 0.0773, + "step": 10150 + }, + { + "epoch": 1.33, + "learning_rate": 1.9042721696645907e-05, + "loss": 0.0755, + "step": 10151 + }, + { + "epoch": 1.33, + "learning_rate": 1.9042495258202275e-05, + "loss": 0.0784, + "step": 10152 + }, + { + "epoch": 1.33, + "learning_rate": 1.9042268794327106e-05, + "loss": 0.0765, + "step": 10153 + }, + { + "epoch": 1.33, + "learning_rate": 1.904204230502104e-05, + "loss": 0.0689, + "step": 10154 + }, + { + "epoch": 1.33, + "learning_rate": 1.904181579028471e-05, + "loss": 0.0787, + "step": 10155 + }, + { + "epoch": 1.33, + "learning_rate": 1.9041589250118754e-05, + "loss": 0.0837, + "step": 10156 + }, + { + "epoch": 1.33, + "learning_rate": 1.904136268452381e-05, + "loss": 0.0747, + "step": 10157 + }, + { + "epoch": 1.33, + "learning_rate": 1.9041136093500516e-05, + "loss": 0.0879, + "step": 10158 + }, + { + "epoch": 1.33, + "learning_rate": 1.904090947704951e-05, + "loss": 0.0816, + "step": 10159 + }, + { + "epoch": 1.33, + "learning_rate": 1.9040682835171423e-05, + "loss": 0.0826, + "step": 10160 + }, + { + "epoch": 1.33, + "learning_rate": 1.90404561678669e-05, + "loss": 0.0916, + "step": 10161 + }, + { + "epoch": 1.33, + "learning_rate": 1.9040229475136573e-05, + "loss": 0.0689, + "step": 10162 + }, + { + "epoch": 1.33, + "learning_rate": 1.9040002756981083e-05, + "loss": 0.0763, + "step": 10163 + }, + { + "epoch": 1.33, + "learning_rate": 1.9039776013401065e-05, + "loss": 0.0859, + "step": 10164 + }, + { + "epoch": 1.33, + "learning_rate": 1.903954924439716e-05, + "loss": 0.0921, + "step": 10165 + }, + { + "epoch": 1.33, + "learning_rate": 1.9039322449970006e-05, + "loss": 0.0846, + "step": 10166 + }, + { + "epoch": 1.33, + "learning_rate": 1.903909563012024e-05, + "loss": 0.0821, + "step": 10167 + }, + { + "epoch": 1.33, + "learning_rate": 1.9038868784848495e-05, + "loss": 0.085, + "step": 10168 + }, + { + "epoch": 1.33, + "learning_rate": 1.9038641914155412e-05, + "loss": 0.0778, + "step": 10169 + }, + { + "epoch": 1.33, + "learning_rate": 1.903841501804163e-05, + "loss": 0.0787, + "step": 10170 + }, + { + "epoch": 1.33, + "learning_rate": 1.9038188096507794e-05, + "loss": 0.0723, + "step": 10171 + }, + { + "epoch": 1.33, + "learning_rate": 1.903796114955453e-05, + "loss": 0.0733, + "step": 10172 + }, + { + "epoch": 1.33, + "learning_rate": 1.903773417718248e-05, + "loss": 0.0819, + "step": 10173 + }, + { + "epoch": 1.33, + "learning_rate": 1.9037507179392286e-05, + "loss": 0.0931, + "step": 10174 + }, + { + "epoch": 1.33, + "learning_rate": 1.9037280156184585e-05, + "loss": 0.0809, + "step": 10175 + }, + { + "epoch": 1.33, + "learning_rate": 1.9037053107560014e-05, + "loss": 0.0773, + "step": 10176 + }, + { + "epoch": 1.33, + "learning_rate": 1.9036826033519213e-05, + "loss": 0.083, + "step": 10177 + }, + { + "epoch": 1.33, + "learning_rate": 1.903659893406282e-05, + "loss": 0.0757, + "step": 10178 + }, + { + "epoch": 1.33, + "learning_rate": 1.9036371809191473e-05, + "loss": 0.0826, + "step": 10179 + }, + { + "epoch": 1.33, + "learning_rate": 1.903614465890581e-05, + "loss": 0.0771, + "step": 10180 + }, + { + "epoch": 1.33, + "learning_rate": 1.9035917483206474e-05, + "loss": 0.0745, + "step": 10181 + }, + { + "epoch": 1.33, + "learning_rate": 1.9035690282094102e-05, + "loss": 0.0749, + "step": 10182 + }, + { + "epoch": 1.33, + "learning_rate": 1.9035463055569332e-05, + "loss": 0.0838, + "step": 10183 + }, + { + "epoch": 1.33, + "learning_rate": 1.90352358036328e-05, + "loss": 0.0816, + "step": 10184 + }, + { + "epoch": 1.33, + "learning_rate": 1.9035008526285153e-05, + "loss": 0.078, + "step": 10185 + }, + { + "epoch": 1.33, + "learning_rate": 1.9034781223527022e-05, + "loss": 0.0793, + "step": 10186 + }, + { + "epoch": 1.33, + "learning_rate": 1.903455389535905e-05, + "loss": 0.0865, + "step": 10187 + }, + { + "epoch": 1.33, + "learning_rate": 1.9034326541781876e-05, + "loss": 0.079, + "step": 10188 + }, + { + "epoch": 1.33, + "learning_rate": 1.9034099162796143e-05, + "loss": 0.0814, + "step": 10189 + }, + { + "epoch": 1.33, + "learning_rate": 1.9033871758402484e-05, + "loss": 0.0705, + "step": 10190 + }, + { + "epoch": 1.33, + "learning_rate": 1.9033644328601543e-05, + "loss": 0.0804, + "step": 10191 + }, + { + "epoch": 1.34, + "learning_rate": 1.9033416873393958e-05, + "loss": 0.0797, + "step": 10192 + }, + { + "epoch": 1.34, + "learning_rate": 1.903318939278037e-05, + "loss": 0.087, + "step": 10193 + }, + { + "epoch": 1.34, + "learning_rate": 1.9032961886761417e-05, + "loss": 0.0805, + "step": 10194 + }, + { + "epoch": 1.34, + "learning_rate": 1.903273435533774e-05, + "loss": 0.0675, + "step": 10195 + }, + { + "epoch": 1.34, + "learning_rate": 1.903250679850998e-05, + "loss": 0.0772, + "step": 10196 + }, + { + "epoch": 1.34, + "learning_rate": 1.9032279216278774e-05, + "loss": 0.086, + "step": 10197 + }, + { + "epoch": 1.34, + "learning_rate": 1.9032051608644763e-05, + "loss": 0.0808, + "step": 10198 + }, + { + "epoch": 1.34, + "learning_rate": 1.903182397560859e-05, + "loss": 0.0812, + "step": 10199 + }, + { + "epoch": 1.34, + "learning_rate": 1.9031596317170892e-05, + "loss": 0.0704, + "step": 10200 + }, + { + "epoch": 1.34, + "learning_rate": 1.9031368633332313e-05, + "loss": 0.0849, + "step": 10201 + }, + { + "epoch": 1.34, + "learning_rate": 1.9031140924093488e-05, + "loss": 0.0812, + "step": 10202 + }, + { + "epoch": 1.34, + "learning_rate": 1.903091318945506e-05, + "loss": 0.0789, + "step": 10203 + }, + { + "epoch": 1.34, + "learning_rate": 1.9030685429417667e-05, + "loss": 0.0812, + "step": 10204 + }, + { + "epoch": 1.34, + "learning_rate": 1.9030457643981958e-05, + "loss": 0.0777, + "step": 10205 + }, + { + "epoch": 1.34, + "learning_rate": 1.9030229833148565e-05, + "loss": 0.0917, + "step": 10206 + }, + { + "epoch": 1.34, + "learning_rate": 1.903000199691813e-05, + "loss": 0.074, + "step": 10207 + }, + { + "epoch": 1.34, + "learning_rate": 1.9029774135291295e-05, + "loss": 0.075, + "step": 10208 + }, + { + "epoch": 1.34, + "learning_rate": 1.90295462482687e-05, + "loss": 0.0836, + "step": 10209 + }, + { + "epoch": 1.34, + "learning_rate": 1.902931833585099e-05, + "loss": 0.0748, + "step": 10210 + }, + { + "epoch": 1.34, + "learning_rate": 1.9029090398038807e-05, + "loss": 0.074, + "step": 10211 + }, + { + "epoch": 1.34, + "learning_rate": 1.9028862434832782e-05, + "loss": 0.0816, + "step": 10212 + }, + { + "epoch": 1.34, + "learning_rate": 1.9028634446233564e-05, + "loss": 0.0815, + "step": 10213 + }, + { + "epoch": 1.34, + "learning_rate": 1.902840643224179e-05, + "loss": 0.0841, + "step": 10214 + }, + { + "epoch": 1.34, + "learning_rate": 1.9028178392858104e-05, + "loss": 0.0754, + "step": 10215 + }, + { + "epoch": 1.34, + "learning_rate": 1.902795032808315e-05, + "loss": 0.0889, + "step": 10216 + }, + { + "epoch": 1.34, + "learning_rate": 1.9027722237917566e-05, + "loss": 0.0826, + "step": 10217 + }, + { + "epoch": 1.34, + "learning_rate": 1.9027494122361997e-05, + "loss": 0.0948, + "step": 10218 + }, + { + "epoch": 1.34, + "learning_rate": 1.9027265981417077e-05, + "loss": 0.0833, + "step": 10219 + }, + { + "epoch": 1.34, + "learning_rate": 1.902703781508345e-05, + "loss": 0.0734, + "step": 10220 + }, + { + "epoch": 1.34, + "learning_rate": 1.9026809623361765e-05, + "loss": 0.0731, + "step": 10221 + }, + { + "epoch": 1.34, + "learning_rate": 1.9026581406252654e-05, + "loss": 0.0797, + "step": 10222 + }, + { + "epoch": 1.34, + "learning_rate": 1.9026353163756767e-05, + "loss": 0.0784, + "step": 10223 + }, + { + "epoch": 1.34, + "learning_rate": 1.9026124895874744e-05, + "loss": 0.0963, + "step": 10224 + }, + { + "epoch": 1.34, + "learning_rate": 1.902589660260722e-05, + "loss": 0.0829, + "step": 10225 + }, + { + "epoch": 1.34, + "learning_rate": 1.9025668283954845e-05, + "loss": 0.0805, + "step": 10226 + }, + { + "epoch": 1.34, + "learning_rate": 1.902543993991826e-05, + "loss": 0.0831, + "step": 10227 + }, + { + "epoch": 1.34, + "learning_rate": 1.9025211570498105e-05, + "loss": 0.0744, + "step": 10228 + }, + { + "epoch": 1.34, + "learning_rate": 1.9024983175695025e-05, + "loss": 0.0853, + "step": 10229 + }, + { + "epoch": 1.34, + "learning_rate": 1.9024754755509657e-05, + "loss": 0.0745, + "step": 10230 + }, + { + "epoch": 1.34, + "learning_rate": 1.902452630994265e-05, + "loss": 0.0784, + "step": 10231 + }, + { + "epoch": 1.34, + "learning_rate": 1.9024297838994637e-05, + "loss": 0.0746, + "step": 10232 + }, + { + "epoch": 1.34, + "learning_rate": 1.9024069342666272e-05, + "loss": 0.0749, + "step": 10233 + }, + { + "epoch": 1.34, + "learning_rate": 1.9023840820958192e-05, + "loss": 0.0889, + "step": 10234 + }, + { + "epoch": 1.34, + "learning_rate": 1.902361227387104e-05, + "loss": 0.0867, + "step": 10235 + }, + { + "epoch": 1.34, + "learning_rate": 1.902338370140546e-05, + "loss": 0.0709, + "step": 10236 + }, + { + "epoch": 1.34, + "learning_rate": 1.902315510356209e-05, + "loss": 0.0774, + "step": 10237 + }, + { + "epoch": 1.34, + "learning_rate": 1.902292648034158e-05, + "loss": 0.087, + "step": 10238 + }, + { + "epoch": 1.34, + "learning_rate": 1.9022697831744574e-05, + "loss": 0.0749, + "step": 10239 + }, + { + "epoch": 1.34, + "learning_rate": 1.9022469157771706e-05, + "loss": 0.0753, + "step": 10240 + }, + { + "epoch": 1.34, + "learning_rate": 1.9022240458423625e-05, + "loss": 0.078, + "step": 10241 + }, + { + "epoch": 1.34, + "learning_rate": 1.9022011733700967e-05, + "loss": 0.0762, + "step": 10242 + }, + { + "epoch": 1.34, + "learning_rate": 1.902178298360439e-05, + "loss": 0.0788, + "step": 10243 + }, + { + "epoch": 1.34, + "learning_rate": 1.9021554208134527e-05, + "loss": 0.071, + "step": 10244 + }, + { + "epoch": 1.34, + "learning_rate": 1.902132540729202e-05, + "loss": 0.0843, + "step": 10245 + }, + { + "epoch": 1.34, + "learning_rate": 1.902109658107752e-05, + "loss": 0.0769, + "step": 10246 + }, + { + "epoch": 1.34, + "learning_rate": 1.9020867729491665e-05, + "loss": 0.0855, + "step": 10247 + }, + { + "epoch": 1.34, + "learning_rate": 1.9020638852535097e-05, + "loss": 0.0807, + "step": 10248 + }, + { + "epoch": 1.34, + "learning_rate": 1.902040995020846e-05, + "loss": 0.0839, + "step": 10249 + }, + { + "epoch": 1.34, + "learning_rate": 1.902018102251241e-05, + "loss": 0.0867, + "step": 10250 + }, + { + "epoch": 1.34, + "learning_rate": 1.9019952069447574e-05, + "loss": 0.0827, + "step": 10251 + }, + { + "epoch": 1.34, + "learning_rate": 1.9019723091014605e-05, + "loss": 0.0755, + "step": 10252 + }, + { + "epoch": 1.34, + "learning_rate": 1.9019494087214148e-05, + "loss": 0.0749, + "step": 10253 + }, + { + "epoch": 1.34, + "learning_rate": 1.901926505804684e-05, + "loss": 0.0828, + "step": 10254 + }, + { + "epoch": 1.34, + "learning_rate": 1.901903600351333e-05, + "loss": 0.0801, + "step": 10255 + }, + { + "epoch": 1.34, + "learning_rate": 1.9018806923614265e-05, + "loss": 0.0759, + "step": 10256 + }, + { + "epoch": 1.34, + "learning_rate": 1.9018577818350282e-05, + "loss": 0.08, + "step": 10257 + }, + { + "epoch": 1.34, + "learning_rate": 1.9018348687722027e-05, + "loss": 0.0812, + "step": 10258 + }, + { + "epoch": 1.34, + "learning_rate": 1.9018119531730152e-05, + "loss": 0.0789, + "step": 10259 + }, + { + "epoch": 1.34, + "learning_rate": 1.9017890350375293e-05, + "loss": 0.0875, + "step": 10260 + }, + { + "epoch": 1.34, + "learning_rate": 1.90176611436581e-05, + "loss": 0.0706, + "step": 10261 + }, + { + "epoch": 1.34, + "learning_rate": 1.9017431911579213e-05, + "loss": 0.0672, + "step": 10262 + }, + { + "epoch": 1.34, + "learning_rate": 1.901720265413928e-05, + "loss": 0.0861, + "step": 10263 + }, + { + "epoch": 1.34, + "learning_rate": 1.9016973371338946e-05, + "loss": 0.0825, + "step": 10264 + }, + { + "epoch": 1.34, + "learning_rate": 1.901674406317885e-05, + "loss": 0.096, + "step": 10265 + }, + { + "epoch": 1.34, + "learning_rate": 1.9016514729659644e-05, + "loss": 0.0838, + "step": 10266 + }, + { + "epoch": 1.34, + "learning_rate": 1.9016285370781973e-05, + "loss": 0.0724, + "step": 10267 + }, + { + "epoch": 1.35, + "learning_rate": 1.9016055986546476e-05, + "loss": 0.0729, + "step": 10268 + }, + { + "epoch": 1.35, + "learning_rate": 1.9015826576953805e-05, + "loss": 0.0709, + "step": 10269 + }, + { + "epoch": 1.35, + "learning_rate": 1.90155971420046e-05, + "loss": 0.0757, + "step": 10270 + }, + { + "epoch": 1.35, + "learning_rate": 1.901536768169951e-05, + "loss": 0.0796, + "step": 10271 + }, + { + "epoch": 1.35, + "learning_rate": 1.9015138196039176e-05, + "loss": 0.0742, + "step": 10272 + }, + { + "epoch": 1.35, + "learning_rate": 1.9014908685024247e-05, + "loss": 0.0831, + "step": 10273 + }, + { + "epoch": 1.35, + "learning_rate": 1.9014679148655365e-05, + "loss": 0.0928, + "step": 10274 + }, + { + "epoch": 1.35, + "learning_rate": 1.9014449586933184e-05, + "loss": 0.0733, + "step": 10275 + }, + { + "epoch": 1.35, + "learning_rate": 1.901421999985834e-05, + "loss": 0.0854, + "step": 10276 + }, + { + "epoch": 1.35, + "learning_rate": 1.901399038743148e-05, + "loss": 0.085, + "step": 10277 + }, + { + "epoch": 1.35, + "learning_rate": 1.9013760749653254e-05, + "loss": 0.0818, + "step": 10278 + }, + { + "epoch": 1.35, + "learning_rate": 1.9013531086524305e-05, + "loss": 0.0762, + "step": 10279 + }, + { + "epoch": 1.35, + "learning_rate": 1.901330139804528e-05, + "loss": 0.0746, + "step": 10280 + }, + { + "epoch": 1.35, + "learning_rate": 1.9013071684216826e-05, + "loss": 0.0738, + "step": 10281 + }, + { + "epoch": 1.35, + "learning_rate": 1.901284194503959e-05, + "loss": 0.0909, + "step": 10282 + }, + { + "epoch": 1.35, + "learning_rate": 1.901261218051421e-05, + "loss": 0.0803, + "step": 10283 + }, + { + "epoch": 1.35, + "learning_rate": 1.901238239064134e-05, + "loss": 0.0799, + "step": 10284 + }, + { + "epoch": 1.35, + "learning_rate": 1.901215257542163e-05, + "loss": 0.0702, + "step": 10285 + }, + { + "epoch": 1.35, + "learning_rate": 1.9011922734855716e-05, + "loss": 0.0868, + "step": 10286 + }, + { + "epoch": 1.35, + "learning_rate": 1.901169286894425e-05, + "loss": 0.0719, + "step": 10287 + }, + { + "epoch": 1.35, + "learning_rate": 1.901146297768788e-05, + "loss": 0.0784, + "step": 10288 + }, + { + "epoch": 1.35, + "learning_rate": 1.9011233061087246e-05, + "loss": 0.0815, + "step": 10289 + }, + { + "epoch": 1.35, + "learning_rate": 1.9011003119143e-05, + "loss": 0.069, + "step": 10290 + }, + { + "epoch": 1.35, + "learning_rate": 1.901077315185579e-05, + "loss": 0.0828, + "step": 10291 + }, + { + "epoch": 1.35, + "learning_rate": 1.9010543159226258e-05, + "loss": 0.0726, + "step": 10292 + }, + { + "epoch": 1.35, + "learning_rate": 1.9010313141255055e-05, + "loss": 0.0922, + "step": 10293 + }, + { + "epoch": 1.35, + "learning_rate": 1.9010083097942823e-05, + "loss": 0.0649, + "step": 10294 + }, + { + "epoch": 1.35, + "learning_rate": 1.9009853029290213e-05, + "loss": 0.0777, + "step": 10295 + }, + { + "epoch": 1.35, + "learning_rate": 1.9009622935297875e-05, + "loss": 0.0867, + "step": 10296 + }, + { + "epoch": 1.35, + "learning_rate": 1.900939281596645e-05, + "loss": 0.0769, + "step": 10297 + }, + { + "epoch": 1.35, + "learning_rate": 1.9009162671296586e-05, + "loss": 0.0812, + "step": 10298 + }, + { + "epoch": 1.35, + "learning_rate": 1.9008932501288933e-05, + "loss": 0.075, + "step": 10299 + }, + { + "epoch": 1.35, + "learning_rate": 1.900870230594414e-05, + "loss": 0.0607, + "step": 10300 + }, + { + "epoch": 1.35, + "learning_rate": 1.9008472085262844e-05, + "loss": 0.0739, + "step": 10301 + }, + { + "epoch": 1.35, + "learning_rate": 1.9008241839245707e-05, + "loss": 0.0906, + "step": 10302 + }, + { + "epoch": 1.35, + "learning_rate": 1.9008011567893367e-05, + "loss": 0.0857, + "step": 10303 + }, + { + "epoch": 1.35, + "learning_rate": 1.9007781271206474e-05, + "loss": 0.0746, + "step": 10304 + }, + { + "epoch": 1.35, + "learning_rate": 1.9007550949185677e-05, + "loss": 0.0748, + "step": 10305 + }, + { + "epoch": 1.35, + "learning_rate": 1.9007320601831622e-05, + "loss": 0.0736, + "step": 10306 + }, + { + "epoch": 1.35, + "learning_rate": 1.900709022914496e-05, + "loss": 0.0756, + "step": 10307 + }, + { + "epoch": 1.35, + "learning_rate": 1.900685983112633e-05, + "loss": 0.0747, + "step": 10308 + }, + { + "epoch": 1.35, + "learning_rate": 1.900662940777639e-05, + "loss": 0.0716, + "step": 10309 + }, + { + "epoch": 1.35, + "learning_rate": 1.9006398959095785e-05, + "loss": 0.0854, + "step": 10310 + }, + { + "epoch": 1.35, + "learning_rate": 1.900616848508516e-05, + "loss": 0.0782, + "step": 10311 + }, + { + "epoch": 1.35, + "learning_rate": 1.9005937985745172e-05, + "loss": 0.0826, + "step": 10312 + }, + { + "epoch": 1.35, + "learning_rate": 1.9005707461076458e-05, + "loss": 0.0732, + "step": 10313 + }, + { + "epoch": 1.35, + "learning_rate": 1.900547691107967e-05, + "loss": 0.0728, + "step": 10314 + }, + { + "epoch": 1.35, + "learning_rate": 1.900524633575546e-05, + "loss": 0.0742, + "step": 10315 + }, + { + "epoch": 1.35, + "learning_rate": 1.9005015735104475e-05, + "loss": 0.0811, + "step": 10316 + }, + { + "epoch": 1.35, + "learning_rate": 1.900478510912736e-05, + "loss": 0.0769, + "step": 10317 + }, + { + "epoch": 1.35, + "learning_rate": 1.900455445782477e-05, + "loss": 0.083, + "step": 10318 + }, + { + "epoch": 1.35, + "learning_rate": 1.9004323781197345e-05, + "loss": 0.0861, + "step": 10319 + }, + { + "epoch": 1.35, + "learning_rate": 1.9004093079245744e-05, + "loss": 0.0779, + "step": 10320 + }, + { + "epoch": 1.35, + "learning_rate": 1.900386235197061e-05, + "loss": 0.08, + "step": 10321 + }, + { + "epoch": 1.35, + "learning_rate": 1.9003631599372592e-05, + "loss": 0.0736, + "step": 10322 + }, + { + "epoch": 1.35, + "learning_rate": 1.9003400821452337e-05, + "loss": 0.0793, + "step": 10323 + }, + { + "epoch": 1.35, + "learning_rate": 1.90031700182105e-05, + "loss": 0.0755, + "step": 10324 + }, + { + "epoch": 1.35, + "learning_rate": 1.9002939189647726e-05, + "loss": 0.0843, + "step": 10325 + }, + { + "epoch": 1.35, + "learning_rate": 1.9002708335764665e-05, + "loss": 0.086, + "step": 10326 + }, + { + "epoch": 1.35, + "learning_rate": 1.9002477456561967e-05, + "loss": 0.0844, + "step": 10327 + }, + { + "epoch": 1.35, + "learning_rate": 1.9002246552040278e-05, + "loss": 0.0815, + "step": 10328 + }, + { + "epoch": 1.35, + "learning_rate": 1.900201562220025e-05, + "loss": 0.0914, + "step": 10329 + }, + { + "epoch": 1.35, + "learning_rate": 1.9001784667042537e-05, + "loss": 0.0894, + "step": 10330 + }, + { + "epoch": 1.35, + "learning_rate": 1.900155368656778e-05, + "loss": 0.0715, + "step": 10331 + }, + { + "epoch": 1.35, + "learning_rate": 1.900132268077663e-05, + "loss": 0.08, + "step": 10332 + }, + { + "epoch": 1.35, + "learning_rate": 1.9001091649669746e-05, + "loss": 0.081, + "step": 10333 + }, + { + "epoch": 1.35, + "learning_rate": 1.9000860593247768e-05, + "loss": 0.0848, + "step": 10334 + }, + { + "epoch": 1.35, + "learning_rate": 1.9000629511511347e-05, + "loss": 0.0801, + "step": 10335 + }, + { + "epoch": 1.35, + "learning_rate": 1.900039840446114e-05, + "loss": 0.0832, + "step": 10336 + }, + { + "epoch": 1.35, + "learning_rate": 1.9000167272097785e-05, + "loss": 0.0782, + "step": 10337 + }, + { + "epoch": 1.35, + "learning_rate": 1.8999936114421943e-05, + "loss": 0.0847, + "step": 10338 + }, + { + "epoch": 1.35, + "learning_rate": 1.899970493143426e-05, + "loss": 0.088, + "step": 10339 + }, + { + "epoch": 1.35, + "learning_rate": 1.8999473723135384e-05, + "loss": 0.0808, + "step": 10340 + }, + { + "epoch": 1.35, + "learning_rate": 1.899924248952597e-05, + "loss": 0.0771, + "step": 10341 + }, + { + "epoch": 1.35, + "learning_rate": 1.899901123060666e-05, + "loss": 0.074, + "step": 10342 + }, + { + "epoch": 1.35, + "learning_rate": 1.8998779946378116e-05, + "loss": 0.0648, + "step": 10343 + }, + { + "epoch": 1.35, + "learning_rate": 1.8998548636840982e-05, + "loss": 0.0773, + "step": 10344 + }, + { + "epoch": 1.36, + "learning_rate": 1.8998317301995906e-05, + "loss": 0.0976, + "step": 10345 + }, + { + "epoch": 1.36, + "learning_rate": 1.8998085941843543e-05, + "loss": 0.0791, + "step": 10346 + }, + { + "epoch": 1.36, + "learning_rate": 1.8997854556384546e-05, + "loss": 0.0742, + "step": 10347 + }, + { + "epoch": 1.36, + "learning_rate": 1.8997623145619562e-05, + "loss": 0.0793, + "step": 10348 + }, + { + "epoch": 1.36, + "learning_rate": 1.899739170954924e-05, + "loss": 0.0732, + "step": 10349 + }, + { + "epoch": 1.36, + "learning_rate": 1.8997160248174232e-05, + "loss": 0.0725, + "step": 10350 + }, + { + "epoch": 1.36, + "learning_rate": 1.8996928761495187e-05, + "loss": 0.0793, + "step": 10351 + }, + { + "epoch": 1.36, + "learning_rate": 1.8996697249512764e-05, + "loss": 0.0784, + "step": 10352 + }, + { + "epoch": 1.36, + "learning_rate": 1.899646571222761e-05, + "loss": 0.0766, + "step": 10353 + }, + { + "epoch": 1.36, + "learning_rate": 1.899623414964037e-05, + "loss": 0.0798, + "step": 10354 + }, + { + "epoch": 1.36, + "learning_rate": 1.8996002561751706e-05, + "loss": 0.0733, + "step": 10355 + }, + { + "epoch": 1.36, + "learning_rate": 1.899577094856226e-05, + "loss": 0.0661, + "step": 10356 + }, + { + "epoch": 1.36, + "learning_rate": 1.8995539310072688e-05, + "loss": 0.0865, + "step": 10357 + }, + { + "epoch": 1.36, + "learning_rate": 1.8995307646283642e-05, + "loss": 0.0785, + "step": 10358 + }, + { + "epoch": 1.36, + "learning_rate": 1.899507595719577e-05, + "loss": 0.0894, + "step": 10359 + }, + { + "epoch": 1.36, + "learning_rate": 1.899484424280973e-05, + "loss": 0.0804, + "step": 10360 + }, + { + "epoch": 1.36, + "learning_rate": 1.8994612503126165e-05, + "loss": 0.0811, + "step": 10361 + }, + { + "epoch": 1.36, + "learning_rate": 1.8994380738145738e-05, + "loss": 0.0719, + "step": 10362 + }, + { + "epoch": 1.36, + "learning_rate": 1.899414894786909e-05, + "loss": 0.0917, + "step": 10363 + }, + { + "epoch": 1.36, + "learning_rate": 1.8993917132296874e-05, + "loss": 0.0819, + "step": 10364 + }, + { + "epoch": 1.36, + "learning_rate": 1.899368529142975e-05, + "loss": 0.0813, + "step": 10365 + }, + { + "epoch": 1.36, + "learning_rate": 1.899345342526836e-05, + "loss": 0.083, + "step": 10366 + }, + { + "epoch": 1.36, + "learning_rate": 1.8993221533813364e-05, + "loss": 0.0717, + "step": 10367 + }, + { + "epoch": 1.36, + "learning_rate": 1.8992989617065412e-05, + "loss": 0.0669, + "step": 10368 + }, + { + "epoch": 1.36, + "learning_rate": 1.899275767502515e-05, + "loss": 0.0755, + "step": 10369 + }, + { + "epoch": 1.36, + "learning_rate": 1.8992525707693246e-05, + "loss": 0.0832, + "step": 10370 + }, + { + "epoch": 1.36, + "learning_rate": 1.899229371507034e-05, + "loss": 0.0853, + "step": 10371 + }, + { + "epoch": 1.36, + "learning_rate": 1.899206169715708e-05, + "loss": 0.077, + "step": 10372 + }, + { + "epoch": 1.36, + "learning_rate": 1.8991829653954127e-05, + "loss": 0.08, + "step": 10373 + }, + { + "epoch": 1.36, + "learning_rate": 1.8991597585462137e-05, + "loss": 0.0693, + "step": 10374 + }, + { + "epoch": 1.36, + "learning_rate": 1.8991365491681755e-05, + "loss": 0.0831, + "step": 10375 + }, + { + "epoch": 1.36, + "learning_rate": 1.8991133372613638e-05, + "loss": 0.0722, + "step": 10376 + }, + { + "epoch": 1.36, + "learning_rate": 1.8990901228258433e-05, + "loss": 0.0747, + "step": 10377 + }, + { + "epoch": 1.36, + "learning_rate": 1.89906690586168e-05, + "loss": 0.0966, + "step": 10378 + }, + { + "epoch": 1.36, + "learning_rate": 1.8990436863689388e-05, + "loss": 0.0682, + "step": 10379 + }, + { + "epoch": 1.36, + "learning_rate": 1.899020464347685e-05, + "loss": 0.0746, + "step": 10380 + }, + { + "epoch": 1.36, + "learning_rate": 1.8989972397979843e-05, + "loss": 0.0734, + "step": 10381 + }, + { + "epoch": 1.36, + "learning_rate": 1.8989740127199017e-05, + "loss": 0.0823, + "step": 10382 + }, + { + "epoch": 1.36, + "learning_rate": 1.8989507831135024e-05, + "loss": 0.0819, + "step": 10383 + }, + { + "epoch": 1.36, + "learning_rate": 1.898927550978852e-05, + "loss": 0.0754, + "step": 10384 + }, + { + "epoch": 1.36, + "learning_rate": 1.8989043163160157e-05, + "loss": 0.0681, + "step": 10385 + }, + { + "epoch": 1.36, + "learning_rate": 1.898881079125059e-05, + "loss": 0.0774, + "step": 10386 + }, + { + "epoch": 1.36, + "learning_rate": 1.8988578394060472e-05, + "loss": 0.0736, + "step": 10387 + }, + { + "epoch": 1.36, + "learning_rate": 1.8988345971590452e-05, + "loss": 0.0769, + "step": 10388 + }, + { + "epoch": 1.36, + "learning_rate": 1.8988113523841188e-05, + "loss": 0.0836, + "step": 10389 + }, + { + "epoch": 1.36, + "learning_rate": 1.8987881050813337e-05, + "loss": 0.0766, + "step": 10390 + }, + { + "epoch": 1.36, + "learning_rate": 1.8987648552507546e-05, + "loss": 0.0729, + "step": 10391 + }, + { + "epoch": 1.36, + "learning_rate": 1.8987416028924475e-05, + "loss": 0.071, + "step": 10392 + }, + { + "epoch": 1.36, + "learning_rate": 1.8987183480064773e-05, + "loss": 0.0802, + "step": 10393 + }, + { + "epoch": 1.36, + "learning_rate": 1.8986950905929096e-05, + "loss": 0.0812, + "step": 10394 + }, + { + "epoch": 1.36, + "learning_rate": 1.8986718306518098e-05, + "loss": 0.0968, + "step": 10395 + }, + { + "epoch": 1.36, + "learning_rate": 1.8986485681832433e-05, + "loss": 0.0854, + "step": 10396 + }, + { + "epoch": 1.36, + "learning_rate": 1.898625303187276e-05, + "loss": 0.0718, + "step": 10397 + }, + { + "epoch": 1.36, + "learning_rate": 1.898602035663972e-05, + "loss": 0.0878, + "step": 10398 + }, + { + "epoch": 1.36, + "learning_rate": 1.8985787656133986e-05, + "loss": 0.0812, + "step": 10399 + }, + { + "epoch": 1.36, + "learning_rate": 1.8985554930356196e-05, + "loss": 0.082, + "step": 10400 + }, + { + "epoch": 1.36, + "learning_rate": 1.8985322179307015e-05, + "loss": 0.0806, + "step": 10401 + }, + { + "epoch": 1.36, + "learning_rate": 1.8985089402987094e-05, + "loss": 0.0768, + "step": 10402 + }, + { + "epoch": 1.36, + "learning_rate": 1.8984856601397084e-05, + "loss": 0.0739, + "step": 10403 + }, + { + "epoch": 1.36, + "learning_rate": 1.8984623774537648e-05, + "loss": 0.0758, + "step": 10404 + }, + { + "epoch": 1.36, + "learning_rate": 1.8984390922409433e-05, + "loss": 0.0758, + "step": 10405 + }, + { + "epoch": 1.36, + "learning_rate": 1.8984158045013098e-05, + "loss": 0.0828, + "step": 10406 + }, + { + "epoch": 1.36, + "learning_rate": 1.8983925142349292e-05, + "loss": 0.0715, + "step": 10407 + }, + { + "epoch": 1.36, + "learning_rate": 1.8983692214418683e-05, + "loss": 0.076, + "step": 10408 + }, + { + "epoch": 1.36, + "learning_rate": 1.898345926122191e-05, + "loss": 0.0847, + "step": 10409 + }, + { + "epoch": 1.36, + "learning_rate": 1.8983226282759645e-05, + "loss": 0.0785, + "step": 10410 + }, + { + "epoch": 1.36, + "learning_rate": 1.898299327903253e-05, + "loss": 0.0837, + "step": 10411 + }, + { + "epoch": 1.36, + "learning_rate": 1.8982760250041224e-05, + "loss": 0.082, + "step": 10412 + }, + { + "epoch": 1.36, + "learning_rate": 1.8982527195786384e-05, + "loss": 0.0746, + "step": 10413 + }, + { + "epoch": 1.36, + "learning_rate": 1.8982294116268665e-05, + "loss": 0.0745, + "step": 10414 + }, + { + "epoch": 1.36, + "learning_rate": 1.898206101148872e-05, + "loss": 0.0796, + "step": 10415 + }, + { + "epoch": 1.36, + "learning_rate": 1.898182788144721e-05, + "loss": 0.0732, + "step": 10416 + }, + { + "epoch": 1.36, + "learning_rate": 1.8981594726144786e-05, + "loss": 0.0772, + "step": 10417 + }, + { + "epoch": 1.36, + "learning_rate": 1.8981361545582105e-05, + "loss": 0.0706, + "step": 10418 + }, + { + "epoch": 1.36, + "learning_rate": 1.8981128339759823e-05, + "loss": 0.0793, + "step": 10419 + }, + { + "epoch": 1.36, + "learning_rate": 1.8980895108678596e-05, + "loss": 0.0835, + "step": 10420 + }, + { + "epoch": 1.37, + "learning_rate": 1.898066185233908e-05, + "loss": 0.0832, + "step": 10421 + }, + { + "epoch": 1.37, + "learning_rate": 1.898042857074193e-05, + "loss": 0.0793, + "step": 10422 + }, + { + "epoch": 1.37, + "learning_rate": 1.8980195263887803e-05, + "loss": 0.0931, + "step": 10423 + }, + { + "epoch": 1.37, + "learning_rate": 1.8979961931777355e-05, + "loss": 0.0762, + "step": 10424 + }, + { + "epoch": 1.37, + "learning_rate": 1.8979728574411243e-05, + "loss": 0.0856, + "step": 10425 + }, + { + "epoch": 1.37, + "learning_rate": 1.897949519179012e-05, + "loss": 0.0688, + "step": 10426 + }, + { + "epoch": 1.37, + "learning_rate": 1.8979261783914647e-05, + "loss": 0.0693, + "step": 10427 + }, + { + "epoch": 1.37, + "learning_rate": 1.897902835078548e-05, + "loss": 0.0862, + "step": 10428 + }, + { + "epoch": 1.37, + "learning_rate": 1.8978794892403273e-05, + "loss": 0.0781, + "step": 10429 + }, + { + "epoch": 1.37, + "learning_rate": 1.897856140876868e-05, + "loss": 0.0871, + "step": 10430 + }, + { + "epoch": 1.37, + "learning_rate": 1.8978327899882367e-05, + "loss": 0.0716, + "step": 10431 + }, + { + "epoch": 1.37, + "learning_rate": 1.8978094365744982e-05, + "loss": 0.0789, + "step": 10432 + }, + { + "epoch": 1.37, + "learning_rate": 1.8977860806357186e-05, + "loss": 0.0785, + "step": 10433 + }, + { + "epoch": 1.37, + "learning_rate": 1.897762722171963e-05, + "loss": 0.0808, + "step": 10434 + }, + { + "epoch": 1.37, + "learning_rate": 1.8977393611832982e-05, + "loss": 0.072, + "step": 10435 + }, + { + "epoch": 1.37, + "learning_rate": 1.897715997669789e-05, + "loss": 0.0871, + "step": 10436 + }, + { + "epoch": 1.37, + "learning_rate": 1.897692631631501e-05, + "loss": 0.0801, + "step": 10437 + }, + { + "epoch": 1.37, + "learning_rate": 1.8976692630685004e-05, + "loss": 0.0852, + "step": 10438 + }, + { + "epoch": 1.37, + "learning_rate": 1.8976458919808533e-05, + "loss": 0.0813, + "step": 10439 + }, + { + "epoch": 1.37, + "learning_rate": 1.8976225183686245e-05, + "loss": 0.0846, + "step": 10440 + }, + { + "epoch": 1.37, + "learning_rate": 1.89759914223188e-05, + "loss": 0.0723, + "step": 10441 + }, + { + "epoch": 1.37, + "learning_rate": 1.897575763570686e-05, + "loss": 0.0911, + "step": 10442 + }, + { + "epoch": 1.37, + "learning_rate": 1.8975523823851083e-05, + "loss": 0.0777, + "step": 10443 + }, + { + "epoch": 1.37, + "learning_rate": 1.897528998675212e-05, + "loss": 0.0781, + "step": 10444 + }, + { + "epoch": 1.37, + "learning_rate": 1.897505612441063e-05, + "loss": 0.0684, + "step": 10445 + }, + { + "epoch": 1.37, + "learning_rate": 1.8974822236827275e-05, + "loss": 0.0802, + "step": 10446 + }, + { + "epoch": 1.37, + "learning_rate": 1.897458832400271e-05, + "loss": 0.0817, + "step": 10447 + }, + { + "epoch": 1.37, + "learning_rate": 1.8974354385937593e-05, + "loss": 0.0601, + "step": 10448 + }, + { + "epoch": 1.37, + "learning_rate": 1.897412042263258e-05, + "loss": 0.0783, + "step": 10449 + }, + { + "epoch": 1.37, + "learning_rate": 1.8973886434088336e-05, + "loss": 0.0816, + "step": 10450 + }, + { + "epoch": 1.37, + "learning_rate": 1.897365242030551e-05, + "loss": 0.075, + "step": 10451 + }, + { + "epoch": 1.37, + "learning_rate": 1.8973418381284766e-05, + "loss": 0.0921, + "step": 10452 + }, + { + "epoch": 1.37, + "learning_rate": 1.8973184317026762e-05, + "loss": 0.0756, + "step": 10453 + }, + { + "epoch": 1.37, + "learning_rate": 1.897295022753215e-05, + "loss": 0.0913, + "step": 10454 + }, + { + "epoch": 1.37, + "learning_rate": 1.89727161128016e-05, + "loss": 0.0843, + "step": 10455 + }, + { + "epoch": 1.37, + "learning_rate": 1.897248197283576e-05, + "loss": 0.078, + "step": 10456 + }, + { + "epoch": 1.37, + "learning_rate": 1.8972247807635293e-05, + "loss": 0.089, + "step": 10457 + }, + { + "epoch": 1.37, + "learning_rate": 1.8972013617200856e-05, + "loss": 0.0749, + "step": 10458 + }, + { + "epoch": 1.37, + "learning_rate": 1.8971779401533112e-05, + "loss": 0.0892, + "step": 10459 + }, + { + "epoch": 1.37, + "learning_rate": 1.8971545160632713e-05, + "loss": 0.0882, + "step": 10460 + }, + { + "epoch": 1.37, + "learning_rate": 1.897131089450032e-05, + "loss": 0.0777, + "step": 10461 + }, + { + "epoch": 1.37, + "learning_rate": 1.8971076603136596e-05, + "loss": 0.0872, + "step": 10462 + }, + { + "epoch": 1.37, + "learning_rate": 1.8970842286542196e-05, + "loss": 0.0818, + "step": 10463 + }, + { + "epoch": 1.37, + "learning_rate": 1.8970607944717778e-05, + "loss": 0.0793, + "step": 10464 + }, + { + "epoch": 1.37, + "learning_rate": 1.8970373577664e-05, + "loss": 0.0753, + "step": 10465 + }, + { + "epoch": 1.37, + "learning_rate": 1.8970139185381532e-05, + "loss": 0.0858, + "step": 10466 + }, + { + "epoch": 1.37, + "learning_rate": 1.896990476787102e-05, + "loss": 0.0836, + "step": 10467 + }, + { + "epoch": 1.37, + "learning_rate": 1.8969670325133133e-05, + "loss": 0.0746, + "step": 10468 + }, + { + "epoch": 1.37, + "learning_rate": 1.896943585716852e-05, + "loss": 0.0792, + "step": 10469 + }, + { + "epoch": 1.37, + "learning_rate": 1.896920136397785e-05, + "loss": 0.0886, + "step": 10470 + }, + { + "epoch": 1.37, + "learning_rate": 1.8968966845561778e-05, + "loss": 0.0729, + "step": 10471 + }, + { + "epoch": 1.37, + "learning_rate": 1.896873230192096e-05, + "loss": 0.0874, + "step": 10472 + }, + { + "epoch": 1.37, + "learning_rate": 1.8968497733056067e-05, + "loss": 0.0697, + "step": 10473 + }, + { + "epoch": 1.37, + "learning_rate": 1.8968263138967747e-05, + "loss": 0.0801, + "step": 10474 + }, + { + "epoch": 1.37, + "learning_rate": 1.8968028519656667e-05, + "loss": 0.0771, + "step": 10475 + }, + { + "epoch": 1.37, + "learning_rate": 1.8967793875123483e-05, + "loss": 0.079, + "step": 10476 + }, + { + "epoch": 1.37, + "learning_rate": 1.8967559205368856e-05, + "loss": 0.0834, + "step": 10477 + }, + { + "epoch": 1.37, + "learning_rate": 1.8967324510393447e-05, + "loss": 0.0742, + "step": 10478 + }, + { + "epoch": 1.37, + "learning_rate": 1.8967089790197917e-05, + "loss": 0.0832, + "step": 10479 + }, + { + "epoch": 1.37, + "learning_rate": 1.8966855044782923e-05, + "loss": 0.0799, + "step": 10480 + }, + { + "epoch": 1.37, + "learning_rate": 1.896662027414913e-05, + "loss": 0.084, + "step": 10481 + }, + { + "epoch": 1.37, + "learning_rate": 1.896638547829719e-05, + "loss": 0.0816, + "step": 10482 + }, + { + "epoch": 1.37, + "learning_rate": 1.896615065722777e-05, + "loss": 0.0741, + "step": 10483 + }, + { + "epoch": 1.37, + "learning_rate": 1.896591581094153e-05, + "loss": 0.0793, + "step": 10484 + }, + { + "epoch": 1.37, + "learning_rate": 1.8965680939439125e-05, + "loss": 0.074, + "step": 10485 + }, + { + "epoch": 1.37, + "learning_rate": 1.8965446042721224e-05, + "loss": 0.079, + "step": 10486 + }, + { + "epoch": 1.37, + "learning_rate": 1.896521112078848e-05, + "loss": 0.0788, + "step": 10487 + }, + { + "epoch": 1.37, + "learning_rate": 1.896497617364156e-05, + "loss": 0.0773, + "step": 10488 + }, + { + "epoch": 1.37, + "learning_rate": 1.8964741201281124e-05, + "loss": 0.0749, + "step": 10489 + }, + { + "epoch": 1.37, + "learning_rate": 1.8964506203707826e-05, + "loss": 0.077, + "step": 10490 + }, + { + "epoch": 1.37, + "learning_rate": 1.8964271180922336e-05, + "loss": 0.0932, + "step": 10491 + }, + { + "epoch": 1.37, + "learning_rate": 1.8964036132925306e-05, + "loss": 0.0702, + "step": 10492 + }, + { + "epoch": 1.37, + "learning_rate": 1.8963801059717407e-05, + "loss": 0.0855, + "step": 10493 + }, + { + "epoch": 1.37, + "learning_rate": 1.896356596129929e-05, + "loss": 0.0744, + "step": 10494 + }, + { + "epoch": 1.37, + "learning_rate": 1.8963330837671624e-05, + "loss": 0.085, + "step": 10495 + }, + { + "epoch": 1.37, + "learning_rate": 1.8963095688835068e-05, + "loss": 0.075, + "step": 10496 + }, + { + "epoch": 1.38, + "learning_rate": 1.896286051479028e-05, + "loss": 0.0921, + "step": 10497 + }, + { + "epoch": 1.38, + "learning_rate": 1.8962625315537925e-05, + "loss": 0.0837, + "step": 10498 + }, + { + "epoch": 1.38, + "learning_rate": 1.896239009107866e-05, + "loss": 0.0804, + "step": 10499 + }, + { + "epoch": 1.38, + "learning_rate": 1.896215484141316e-05, + "loss": 0.0853, + "step": 10500 + }, + { + "epoch": 1.38, + "learning_rate": 1.8961919566542067e-05, + "loss": 0.088, + "step": 10501 + }, + { + "epoch": 1.38, + "learning_rate": 1.8961684266466057e-05, + "loss": 0.0747, + "step": 10502 + }, + { + "epoch": 1.38, + "learning_rate": 1.8961448941185785e-05, + "loss": 0.0698, + "step": 10503 + }, + { + "epoch": 1.38, + "learning_rate": 1.896121359070192e-05, + "loss": 0.0778, + "step": 10504 + }, + { + "epoch": 1.38, + "learning_rate": 1.8960978215015112e-05, + "loss": 0.0785, + "step": 10505 + }, + { + "epoch": 1.38, + "learning_rate": 1.8960742814126037e-05, + "loss": 0.0803, + "step": 10506 + }, + { + "epoch": 1.38, + "learning_rate": 1.8960507388035346e-05, + "loss": 0.0736, + "step": 10507 + }, + { + "epoch": 1.38, + "learning_rate": 1.8960271936743707e-05, + "loss": 0.082, + "step": 10508 + }, + { + "epoch": 1.38, + "learning_rate": 1.8960036460251777e-05, + "loss": 0.0827, + "step": 10509 + }, + { + "epoch": 1.38, + "learning_rate": 1.8959800958560225e-05, + "loss": 0.071, + "step": 10510 + }, + { + "epoch": 1.38, + "learning_rate": 1.8959565431669706e-05, + "loss": 0.0777, + "step": 10511 + }, + { + "epoch": 1.38, + "learning_rate": 1.895932987958089e-05, + "loss": 0.0754, + "step": 10512 + }, + { + "epoch": 1.38, + "learning_rate": 1.8959094302294434e-05, + "loss": 0.0833, + "step": 10513 + }, + { + "epoch": 1.38, + "learning_rate": 1.8958858699811005e-05, + "loss": 0.0698, + "step": 10514 + }, + { + "epoch": 1.38, + "learning_rate": 1.895862307213126e-05, + "loss": 0.0833, + "step": 10515 + }, + { + "epoch": 1.38, + "learning_rate": 1.8958387419255866e-05, + "loss": 0.078, + "step": 10516 + }, + { + "epoch": 1.38, + "learning_rate": 1.8958151741185484e-05, + "loss": 0.0802, + "step": 10517 + }, + { + "epoch": 1.38, + "learning_rate": 1.8957916037920778e-05, + "loss": 0.087, + "step": 10518 + }, + { + "epoch": 1.38, + "learning_rate": 1.8957680309462408e-05, + "loss": 0.0826, + "step": 10519 + }, + { + "epoch": 1.38, + "learning_rate": 1.8957444555811042e-05, + "loss": 0.0815, + "step": 10520 + }, + { + "epoch": 1.38, + "learning_rate": 1.895720877696734e-05, + "loss": 0.0921, + "step": 10521 + }, + { + "epoch": 1.38, + "learning_rate": 1.8956972972931966e-05, + "loss": 0.0761, + "step": 10522 + }, + { + "epoch": 1.38, + "learning_rate": 1.895673714370558e-05, + "loss": 0.0733, + "step": 10523 + }, + { + "epoch": 1.38, + "learning_rate": 1.8956501289288846e-05, + "loss": 0.0899, + "step": 10524 + }, + { + "epoch": 1.38, + "learning_rate": 1.8956265409682434e-05, + "loss": 0.0867, + "step": 10525 + }, + { + "epoch": 1.38, + "learning_rate": 1.8956029504887e-05, + "loss": 0.0762, + "step": 10526 + }, + { + "epoch": 1.38, + "learning_rate": 1.895579357490321e-05, + "loss": 0.0776, + "step": 10527 + }, + { + "epoch": 1.38, + "learning_rate": 1.8955557619731725e-05, + "loss": 0.0719, + "step": 10528 + }, + { + "epoch": 1.38, + "learning_rate": 1.8955321639373214e-05, + "loss": 0.0778, + "step": 10529 + }, + { + "epoch": 1.38, + "learning_rate": 1.8955085633828337e-05, + "loss": 0.0795, + "step": 10530 + }, + { + "epoch": 1.38, + "learning_rate": 1.895484960309776e-05, + "loss": 0.0828, + "step": 10531 + }, + { + "epoch": 1.38, + "learning_rate": 1.8954613547182144e-05, + "loss": 0.0694, + "step": 10532 + }, + { + "epoch": 1.38, + "learning_rate": 1.8954377466082156e-05, + "loss": 0.0827, + "step": 10533 + }, + { + "epoch": 1.38, + "learning_rate": 1.8954141359798458e-05, + "loss": 0.069, + "step": 10534 + }, + { + "epoch": 1.38, + "learning_rate": 1.8953905228331712e-05, + "loss": 0.0758, + "step": 10535 + }, + { + "epoch": 1.38, + "learning_rate": 1.8953669071682585e-05, + "loss": 0.1, + "step": 10536 + }, + { + "epoch": 1.38, + "learning_rate": 1.895343288985174e-05, + "loss": 0.0715, + "step": 10537 + }, + { + "epoch": 1.38, + "learning_rate": 1.8953196682839845e-05, + "loss": 0.0812, + "step": 10538 + }, + { + "epoch": 1.38, + "learning_rate": 1.895296045064756e-05, + "loss": 0.0756, + "step": 10539 + }, + { + "epoch": 1.38, + "learning_rate": 1.895272419327555e-05, + "loss": 0.0805, + "step": 10540 + }, + { + "epoch": 1.38, + "learning_rate": 1.8952487910724478e-05, + "loss": 0.084, + "step": 10541 + }, + { + "epoch": 1.38, + "learning_rate": 1.8952251602995014e-05, + "loss": 0.075, + "step": 10542 + }, + { + "epoch": 1.38, + "learning_rate": 1.8952015270087814e-05, + "loss": 0.0755, + "step": 10543 + }, + { + "epoch": 1.38, + "learning_rate": 1.8951778912003553e-05, + "loss": 0.0872, + "step": 10544 + }, + { + "epoch": 1.38, + "learning_rate": 1.8951542528742887e-05, + "loss": 0.0778, + "step": 10545 + }, + { + "epoch": 1.38, + "learning_rate": 1.8951306120306487e-05, + "loss": 0.0694, + "step": 10546 + }, + { + "epoch": 1.38, + "learning_rate": 1.8951069686695015e-05, + "loss": 0.0938, + "step": 10547 + }, + { + "epoch": 1.38, + "learning_rate": 1.8950833227909134e-05, + "loss": 0.0819, + "step": 10548 + }, + { + "epoch": 1.38, + "learning_rate": 1.8950596743949515e-05, + "loss": 0.0729, + "step": 10549 + }, + { + "epoch": 1.38, + "learning_rate": 1.8950360234816813e-05, + "loss": 0.07, + "step": 10550 + }, + { + "epoch": 1.38, + "learning_rate": 1.8950123700511704e-05, + "loss": 0.086, + "step": 10551 + }, + { + "epoch": 1.38, + "learning_rate": 1.894988714103485e-05, + "loss": 0.0889, + "step": 10552 + }, + { + "epoch": 1.38, + "learning_rate": 1.8949650556386913e-05, + "loss": 0.0768, + "step": 10553 + }, + { + "epoch": 1.38, + "learning_rate": 1.894941394656856e-05, + "loss": 0.0798, + "step": 10554 + }, + { + "epoch": 1.38, + "learning_rate": 1.8949177311580455e-05, + "loss": 0.0869, + "step": 10555 + }, + { + "epoch": 1.38, + "learning_rate": 1.8948940651423265e-05, + "loss": 0.0721, + "step": 10556 + }, + { + "epoch": 1.38, + "learning_rate": 1.894870396609766e-05, + "loss": 0.0731, + "step": 10557 + }, + { + "epoch": 1.38, + "learning_rate": 1.89484672556043e-05, + "loss": 0.0858, + "step": 10558 + }, + { + "epoch": 1.38, + "learning_rate": 1.8948230519943854e-05, + "loss": 0.0808, + "step": 10559 + }, + { + "epoch": 1.38, + "learning_rate": 1.8947993759116985e-05, + "loss": 0.0804, + "step": 10560 + }, + { + "epoch": 1.38, + "learning_rate": 1.894775697312436e-05, + "loss": 0.0904, + "step": 10561 + }, + { + "epoch": 1.38, + "learning_rate": 1.8947520161966648e-05, + "loss": 0.0739, + "step": 10562 + }, + { + "epoch": 1.38, + "learning_rate": 1.8947283325644504e-05, + "loss": 0.0808, + "step": 10563 + }, + { + "epoch": 1.38, + "learning_rate": 1.8947046464158608e-05, + "loss": 0.0807, + "step": 10564 + }, + { + "epoch": 1.38, + "learning_rate": 1.894680957750962e-05, + "loss": 0.0826, + "step": 10565 + }, + { + "epoch": 1.38, + "learning_rate": 1.8946572665698205e-05, + "loss": 0.0791, + "step": 10566 + }, + { + "epoch": 1.38, + "learning_rate": 1.894633572872503e-05, + "loss": 0.0742, + "step": 10567 + }, + { + "epoch": 1.38, + "learning_rate": 1.8946098766590766e-05, + "loss": 0.0773, + "step": 10568 + }, + { + "epoch": 1.38, + "learning_rate": 1.8945861779296072e-05, + "loss": 0.0822, + "step": 10569 + }, + { + "epoch": 1.38, + "learning_rate": 1.894562476684162e-05, + "loss": 0.0756, + "step": 10570 + }, + { + "epoch": 1.38, + "learning_rate": 1.8945387729228074e-05, + "loss": 0.0803, + "step": 10571 + }, + { + "epoch": 1.38, + "learning_rate": 1.8945150666456103e-05, + "loss": 0.0693, + "step": 10572 + }, + { + "epoch": 1.38, + "learning_rate": 1.894491357852637e-05, + "loss": 0.0792, + "step": 10573 + }, + { + "epoch": 1.39, + "learning_rate": 1.8944676465439545e-05, + "loss": 0.0745, + "step": 10574 + }, + { + "epoch": 1.39, + "learning_rate": 1.8944439327196293e-05, + "loss": 0.0845, + "step": 10575 + }, + { + "epoch": 1.39, + "learning_rate": 1.8944202163797283e-05, + "loss": 0.0893, + "step": 10576 + }, + { + "epoch": 1.39, + "learning_rate": 1.894396497524318e-05, + "loss": 0.0823, + "step": 10577 + }, + { + "epoch": 1.39, + "learning_rate": 1.8943727761534648e-05, + "loss": 0.0794, + "step": 10578 + }, + { + "epoch": 1.39, + "learning_rate": 1.8943490522672365e-05, + "loss": 0.0776, + "step": 10579 + }, + { + "epoch": 1.39, + "learning_rate": 1.8943253258656988e-05, + "loss": 0.0735, + "step": 10580 + }, + { + "epoch": 1.39, + "learning_rate": 1.8943015969489184e-05, + "loss": 0.0808, + "step": 10581 + }, + { + "epoch": 1.39, + "learning_rate": 1.8942778655169628e-05, + "loss": 0.0783, + "step": 10582 + }, + { + "epoch": 1.39, + "learning_rate": 1.8942541315698984e-05, + "loss": 0.0612, + "step": 10583 + }, + { + "epoch": 1.39, + "learning_rate": 1.8942303951077916e-05, + "loss": 0.0699, + "step": 10584 + }, + { + "epoch": 1.39, + "learning_rate": 1.8942066561307094e-05, + "loss": 0.0709, + "step": 10585 + }, + { + "epoch": 1.39, + "learning_rate": 1.894182914638719e-05, + "loss": 0.0767, + "step": 10586 + }, + { + "epoch": 1.39, + "learning_rate": 1.8941591706318864e-05, + "loss": 0.0789, + "step": 10587 + }, + { + "epoch": 1.39, + "learning_rate": 1.894135424110279e-05, + "loss": 0.0732, + "step": 10588 + }, + { + "epoch": 1.39, + "learning_rate": 1.894111675073963e-05, + "loss": 0.0924, + "step": 10589 + }, + { + "epoch": 1.39, + "learning_rate": 1.8940879235230058e-05, + "loss": 0.0804, + "step": 10590 + }, + { + "epoch": 1.39, + "learning_rate": 1.8940641694574737e-05, + "loss": 0.0785, + "step": 10591 + }, + { + "epoch": 1.39, + "learning_rate": 1.894040412877434e-05, + "loss": 0.0882, + "step": 10592 + }, + { + "epoch": 1.39, + "learning_rate": 1.8940166537829532e-05, + "loss": 0.0774, + "step": 10593 + }, + { + "epoch": 1.39, + "learning_rate": 1.893992892174098e-05, + "loss": 0.0764, + "step": 10594 + }, + { + "epoch": 1.39, + "learning_rate": 1.8939691280509356e-05, + "loss": 0.0841, + "step": 10595 + }, + { + "epoch": 1.39, + "learning_rate": 1.8939453614135325e-05, + "loss": 0.0681, + "step": 10596 + }, + { + "epoch": 1.39, + "learning_rate": 1.893921592261956e-05, + "loss": 0.0792, + "step": 10597 + }, + { + "epoch": 1.39, + "learning_rate": 1.8938978205962718e-05, + "loss": 0.078, + "step": 10598 + }, + { + "epoch": 1.39, + "learning_rate": 1.8938740464165485e-05, + "loss": 0.079, + "step": 10599 + }, + { + "epoch": 1.39, + "learning_rate": 1.8938502697228513e-05, + "loss": 0.0746, + "step": 10600 + }, + { + "epoch": 1.39, + "learning_rate": 1.8938264905152485e-05, + "loss": 0.0784, + "step": 10601 + }, + { + "epoch": 1.39, + "learning_rate": 1.893802708793806e-05, + "loss": 0.0792, + "step": 10602 + }, + { + "epoch": 1.39, + "learning_rate": 1.893778924558591e-05, + "loss": 0.0678, + "step": 10603 + }, + { + "epoch": 1.39, + "learning_rate": 1.8937551378096703e-05, + "loss": 0.0851, + "step": 10604 + }, + { + "epoch": 1.39, + "learning_rate": 1.8937313485471107e-05, + "loss": 0.0792, + "step": 10605 + }, + { + "epoch": 1.39, + "learning_rate": 1.8937075567709796e-05, + "loss": 0.0807, + "step": 10606 + }, + { + "epoch": 1.39, + "learning_rate": 1.8936837624813433e-05, + "loss": 0.0762, + "step": 10607 + }, + { + "epoch": 1.39, + "learning_rate": 1.893659965678269e-05, + "loss": 0.0826, + "step": 10608 + }, + { + "epoch": 1.39, + "learning_rate": 1.8936361663618243e-05, + "loss": 0.085, + "step": 10609 + }, + { + "epoch": 1.39, + "learning_rate": 1.8936123645320748e-05, + "loss": 0.0771, + "step": 10610 + }, + { + "epoch": 1.39, + "learning_rate": 1.8935885601890884e-05, + "loss": 0.0813, + "step": 10611 + }, + { + "epoch": 1.39, + "learning_rate": 1.8935647533329315e-05, + "loss": 0.0765, + "step": 10612 + }, + { + "epoch": 1.39, + "learning_rate": 1.8935409439636716e-05, + "loss": 0.0793, + "step": 10613 + }, + { + "epoch": 1.39, + "learning_rate": 1.893517132081375e-05, + "loss": 0.0694, + "step": 10614 + }, + { + "epoch": 1.39, + "learning_rate": 1.8934933176861094e-05, + "loss": 0.0699, + "step": 10615 + }, + { + "epoch": 1.39, + "learning_rate": 1.8934695007779415e-05, + "loss": 0.0864, + "step": 10616 + }, + { + "epoch": 1.39, + "learning_rate": 1.893445681356938e-05, + "loss": 0.0786, + "step": 10617 + }, + { + "epoch": 1.39, + "learning_rate": 1.8934218594231663e-05, + "loss": 0.0796, + "step": 10618 + }, + { + "epoch": 1.39, + "learning_rate": 1.8933980349766928e-05, + "loss": 0.0773, + "step": 10619 + }, + { + "epoch": 1.39, + "learning_rate": 1.8933742080175852e-05, + "loss": 0.0738, + "step": 10620 + }, + { + "epoch": 1.39, + "learning_rate": 1.8933503785459103e-05, + "loss": 0.0796, + "step": 10621 + }, + { + "epoch": 1.39, + "learning_rate": 1.8933265465617346e-05, + "loss": 0.0856, + "step": 10622 + }, + { + "epoch": 1.39, + "learning_rate": 1.8933027120651258e-05, + "loss": 0.0787, + "step": 10623 + }, + { + "epoch": 1.39, + "learning_rate": 1.893278875056151e-05, + "loss": 0.0733, + "step": 10624 + }, + { + "epoch": 1.39, + "learning_rate": 1.8932550355348765e-05, + "loss": 0.0828, + "step": 10625 + }, + { + "epoch": 1.39, + "learning_rate": 1.89323119350137e-05, + "loss": 0.0883, + "step": 10626 + }, + { + "epoch": 1.39, + "learning_rate": 1.893207348955698e-05, + "loss": 0.0888, + "step": 10627 + }, + { + "epoch": 1.39, + "learning_rate": 1.8931835018979284e-05, + "loss": 0.0789, + "step": 10628 + }, + { + "epoch": 1.39, + "learning_rate": 1.8931596523281274e-05, + "loss": 0.0733, + "step": 10629 + }, + { + "epoch": 1.39, + "learning_rate": 1.8931358002463627e-05, + "loss": 0.0859, + "step": 10630 + }, + { + "epoch": 1.39, + "learning_rate": 1.893111945652701e-05, + "loss": 0.0762, + "step": 10631 + }, + { + "epoch": 1.39, + "learning_rate": 1.8930880885472096e-05, + "loss": 0.0777, + "step": 10632 + }, + { + "epoch": 1.39, + "learning_rate": 1.8930642289299556e-05, + "loss": 0.077, + "step": 10633 + }, + { + "epoch": 1.39, + "learning_rate": 1.8930403668010056e-05, + "loss": 0.0759, + "step": 10634 + }, + { + "epoch": 1.39, + "learning_rate": 1.8930165021604275e-05, + "loss": 0.0726, + "step": 10635 + }, + { + "epoch": 1.39, + "learning_rate": 1.892992635008288e-05, + "loss": 0.0854, + "step": 10636 + }, + { + "epoch": 1.39, + "learning_rate": 1.8929687653446542e-05, + "loss": 0.0784, + "step": 10637 + }, + { + "epoch": 1.39, + "learning_rate": 1.892944893169593e-05, + "loss": 0.0776, + "step": 10638 + }, + { + "epoch": 1.39, + "learning_rate": 1.8929210184831722e-05, + "loss": 0.0805, + "step": 10639 + }, + { + "epoch": 1.39, + "learning_rate": 1.8928971412854587e-05, + "loss": 0.0899, + "step": 10640 + }, + { + "epoch": 1.39, + "learning_rate": 1.8928732615765192e-05, + "loss": 0.0826, + "step": 10641 + }, + { + "epoch": 1.39, + "learning_rate": 1.8928493793564217e-05, + "loss": 0.0879, + "step": 10642 + }, + { + "epoch": 1.39, + "learning_rate": 1.8928254946252324e-05, + "loss": 0.0786, + "step": 10643 + }, + { + "epoch": 1.39, + "learning_rate": 1.892801607383019e-05, + "loss": 0.0744, + "step": 10644 + }, + { + "epoch": 1.39, + "learning_rate": 1.892777717629849e-05, + "loss": 0.0765, + "step": 10645 + }, + { + "epoch": 1.39, + "learning_rate": 1.8927538253657886e-05, + "loss": 0.0658, + "step": 10646 + }, + { + "epoch": 1.39, + "learning_rate": 1.892729930590906e-05, + "loss": 0.0673, + "step": 10647 + }, + { + "epoch": 1.39, + "learning_rate": 1.8927060333052678e-05, + "loss": 0.0828, + "step": 10648 + }, + { + "epoch": 1.39, + "learning_rate": 1.8926821335089416e-05, + "loss": 0.082, + "step": 10649 + }, + { + "epoch": 1.4, + "learning_rate": 1.8926582312019945e-05, + "loss": 0.0768, + "step": 10650 + }, + { + "epoch": 1.4, + "learning_rate": 1.8926343263844935e-05, + "loss": 0.0682, + "step": 10651 + }, + { + "epoch": 1.4, + "learning_rate": 1.892610419056506e-05, + "loss": 0.084, + "step": 10652 + }, + { + "epoch": 1.4, + "learning_rate": 1.8925865092180992e-05, + "loss": 0.0794, + "step": 10653 + }, + { + "epoch": 1.4, + "learning_rate": 1.8925625968693403e-05, + "loss": 0.0769, + "step": 10654 + }, + { + "epoch": 1.4, + "learning_rate": 1.892538682010297e-05, + "loss": 0.0872, + "step": 10655 + }, + { + "epoch": 1.4, + "learning_rate": 1.892514764641036e-05, + "loss": 0.0805, + "step": 10656 + }, + { + "epoch": 1.4, + "learning_rate": 1.8924908447616246e-05, + "loss": 0.0751, + "step": 10657 + }, + { + "epoch": 1.4, + "learning_rate": 1.8924669223721304e-05, + "loss": 0.0685, + "step": 10658 + }, + { + "epoch": 1.4, + "learning_rate": 1.8924429974726204e-05, + "loss": 0.0838, + "step": 10659 + }, + { + "epoch": 1.4, + "learning_rate": 1.892419070063162e-05, + "loss": 0.0958, + "step": 10660 + }, + { + "epoch": 1.4, + "learning_rate": 1.8923951401438224e-05, + "loss": 0.0838, + "step": 10661 + }, + { + "epoch": 1.4, + "learning_rate": 1.8923712077146693e-05, + "loss": 0.0773, + "step": 10662 + }, + { + "epoch": 1.4, + "learning_rate": 1.8923472727757697e-05, + "loss": 0.0755, + "step": 10663 + }, + { + "epoch": 1.4, + "learning_rate": 1.8923233353271905e-05, + "loss": 0.0782, + "step": 10664 + }, + { + "epoch": 1.4, + "learning_rate": 1.892299395369e-05, + "loss": 0.0752, + "step": 10665 + }, + { + "epoch": 1.4, + "learning_rate": 1.8922754529012644e-05, + "loss": 0.0876, + "step": 10666 + }, + { + "epoch": 1.4, + "learning_rate": 1.8922515079240517e-05, + "loss": 0.087, + "step": 10667 + }, + { + "epoch": 1.4, + "learning_rate": 1.8922275604374296e-05, + "loss": 0.0731, + "step": 10668 + }, + { + "epoch": 1.4, + "learning_rate": 1.8922036104414644e-05, + "loss": 0.08, + "step": 10669 + }, + { + "epoch": 1.4, + "learning_rate": 1.8921796579362243e-05, + "loss": 0.078, + "step": 10670 + }, + { + "epoch": 1.4, + "learning_rate": 1.892155702921777e-05, + "loss": 0.0828, + "step": 10671 + }, + { + "epoch": 1.4, + "learning_rate": 1.8921317453981885e-05, + "loss": 0.0854, + "step": 10672 + }, + { + "epoch": 1.4, + "learning_rate": 1.8921077853655272e-05, + "loss": 0.0743, + "step": 10673 + }, + { + "epoch": 1.4, + "learning_rate": 1.8920838228238602e-05, + "loss": 0.0976, + "step": 10674 + }, + { + "epoch": 1.4, + "learning_rate": 1.892059857773255e-05, + "loss": 0.079, + "step": 10675 + }, + { + "epoch": 1.4, + "learning_rate": 1.892035890213779e-05, + "loss": 0.0856, + "step": 10676 + }, + { + "epoch": 1.4, + "learning_rate": 1.8920119201455e-05, + "loss": 0.0739, + "step": 10677 + }, + { + "epoch": 1.4, + "learning_rate": 1.8919879475684844e-05, + "loss": 0.0745, + "step": 10678 + }, + { + "epoch": 1.4, + "learning_rate": 1.8919639724828e-05, + "loss": 0.0746, + "step": 10679 + }, + { + "epoch": 1.4, + "learning_rate": 1.891939994888515e-05, + "loss": 0.0645, + "step": 10680 + }, + { + "epoch": 1.4, + "learning_rate": 1.891916014785696e-05, + "loss": 0.0812, + "step": 10681 + }, + { + "epoch": 1.4, + "learning_rate": 1.891892032174411e-05, + "loss": 0.0825, + "step": 10682 + }, + { + "epoch": 1.4, + "learning_rate": 1.8918680470547267e-05, + "loss": 0.083, + "step": 10683 + }, + { + "epoch": 1.4, + "learning_rate": 1.8918440594267112e-05, + "loss": 0.0761, + "step": 10684 + }, + { + "epoch": 1.4, + "learning_rate": 1.891820069290432e-05, + "loss": 0.0737, + "step": 10685 + }, + { + "epoch": 1.4, + "learning_rate": 1.8917960766459563e-05, + "loss": 0.078, + "step": 10686 + }, + { + "epoch": 1.4, + "learning_rate": 1.8917720814933514e-05, + "loss": 0.0723, + "step": 10687 + }, + { + "epoch": 1.4, + "learning_rate": 1.8917480838326853e-05, + "loss": 0.0736, + "step": 10688 + }, + { + "epoch": 1.4, + "learning_rate": 1.8917240836640248e-05, + "loss": 0.0705, + "step": 10689 + }, + { + "epoch": 1.4, + "learning_rate": 1.891700080987438e-05, + "loss": 0.0685, + "step": 10690 + }, + { + "epoch": 1.4, + "learning_rate": 1.891676075802992e-05, + "loss": 0.0698, + "step": 10691 + }, + { + "epoch": 1.4, + "learning_rate": 1.8916520681107547e-05, + "loss": 0.0746, + "step": 10692 + }, + { + "epoch": 1.4, + "learning_rate": 1.8916280579107934e-05, + "loss": 0.0746, + "step": 10693 + }, + { + "epoch": 1.4, + "learning_rate": 1.891604045203176e-05, + "loss": 0.071, + "step": 10694 + }, + { + "epoch": 1.4, + "learning_rate": 1.8915800299879695e-05, + "loss": 0.0815, + "step": 10695 + }, + { + "epoch": 1.4, + "learning_rate": 1.8915560122652413e-05, + "loss": 0.0756, + "step": 10696 + }, + { + "epoch": 1.4, + "learning_rate": 1.8915319920350598e-05, + "loss": 0.0682, + "step": 10697 + }, + { + "epoch": 1.4, + "learning_rate": 1.8915079692974917e-05, + "loss": 0.0795, + "step": 10698 + }, + { + "epoch": 1.4, + "learning_rate": 1.891483944052605e-05, + "loss": 0.0752, + "step": 10699 + }, + { + "epoch": 1.4, + "learning_rate": 1.891459916300467e-05, + "loss": 0.084, + "step": 10700 + }, + { + "epoch": 1.4, + "learning_rate": 1.8914358860411458e-05, + "loss": 0.0853, + "step": 10701 + }, + { + "epoch": 1.4, + "learning_rate": 1.8914118532747086e-05, + "loss": 0.0945, + "step": 10702 + }, + { + "epoch": 1.4, + "learning_rate": 1.891387818001223e-05, + "loss": 0.074, + "step": 10703 + }, + { + "epoch": 1.4, + "learning_rate": 1.8913637802207566e-05, + "loss": 0.0885, + "step": 10704 + }, + { + "epoch": 1.4, + "learning_rate": 1.8913397399333767e-05, + "loss": 0.087, + "step": 10705 + }, + { + "epoch": 1.4, + "learning_rate": 1.891315697139152e-05, + "loss": 0.0777, + "step": 10706 + }, + { + "epoch": 1.4, + "learning_rate": 1.8912916518381487e-05, + "loss": 0.0802, + "step": 10707 + }, + { + "epoch": 1.4, + "learning_rate": 1.8912676040304356e-05, + "loss": 0.0851, + "step": 10708 + }, + { + "epoch": 1.4, + "learning_rate": 1.8912435537160795e-05, + "loss": 0.079, + "step": 10709 + }, + { + "epoch": 1.4, + "learning_rate": 1.8912195008951483e-05, + "loss": 0.0851, + "step": 10710 + }, + { + "epoch": 1.4, + "learning_rate": 1.8911954455677097e-05, + "loss": 0.0736, + "step": 10711 + }, + { + "epoch": 1.4, + "learning_rate": 1.8911713877338315e-05, + "loss": 0.0713, + "step": 10712 + }, + { + "epoch": 1.4, + "learning_rate": 1.8911473273935815e-05, + "loss": 0.0828, + "step": 10713 + }, + { + "epoch": 1.4, + "learning_rate": 1.8911232645470267e-05, + "loss": 0.0764, + "step": 10714 + }, + { + "epoch": 1.4, + "learning_rate": 1.8910991991942352e-05, + "loss": 0.0677, + "step": 10715 + }, + { + "epoch": 1.4, + "learning_rate": 1.8910751313352745e-05, + "loss": 0.0802, + "step": 10716 + }, + { + "epoch": 1.4, + "learning_rate": 1.891051060970213e-05, + "loss": 0.0899, + "step": 10717 + }, + { + "epoch": 1.4, + "learning_rate": 1.8910269880991175e-05, + "loss": 0.0796, + "step": 10718 + }, + { + "epoch": 1.4, + "learning_rate": 1.891002912722056e-05, + "loss": 0.073, + "step": 10719 + }, + { + "epoch": 1.4, + "learning_rate": 1.8909788348390964e-05, + "loss": 0.0823, + "step": 10720 + }, + { + "epoch": 1.4, + "learning_rate": 1.8909547544503063e-05, + "loss": 0.077, + "step": 10721 + }, + { + "epoch": 1.4, + "learning_rate": 1.8909306715557535e-05, + "loss": 0.0851, + "step": 10722 + }, + { + "epoch": 1.4, + "learning_rate": 1.8909065861555055e-05, + "loss": 0.0758, + "step": 10723 + }, + { + "epoch": 1.4, + "learning_rate": 1.8908824982496297e-05, + "loss": 0.0794, + "step": 10724 + }, + { + "epoch": 1.4, + "learning_rate": 1.8908584078381952e-05, + "loss": 0.082, + "step": 10725 + }, + { + "epoch": 1.41, + "learning_rate": 1.890834314921268e-05, + "loss": 0.0833, + "step": 10726 + }, + { + "epoch": 1.41, + "learning_rate": 1.890810219498917e-05, + "loss": 0.0753, + "step": 10727 + }, + { + "epoch": 1.41, + "learning_rate": 1.89078612157121e-05, + "loss": 0.0799, + "step": 10728 + }, + { + "epoch": 1.41, + "learning_rate": 1.8907620211382144e-05, + "loss": 0.0862, + "step": 10729 + }, + { + "epoch": 1.41, + "learning_rate": 1.890737918199998e-05, + "loss": 0.0733, + "step": 10730 + }, + { + "epoch": 1.41, + "learning_rate": 1.8907138127566287e-05, + "loss": 0.0791, + "step": 10731 + }, + { + "epoch": 1.41, + "learning_rate": 1.890689704808174e-05, + "loss": 0.0904, + "step": 10732 + }, + { + "epoch": 1.41, + "learning_rate": 1.8906655943547022e-05, + "loss": 0.0804, + "step": 10733 + }, + { + "epoch": 1.41, + "learning_rate": 1.8906414813962806e-05, + "loss": 0.0739, + "step": 10734 + }, + { + "epoch": 1.41, + "learning_rate": 1.8906173659329773e-05, + "loss": 0.0623, + "step": 10735 + }, + { + "epoch": 1.41, + "learning_rate": 1.8905932479648604e-05, + "loss": 0.0763, + "step": 10736 + }, + { + "epoch": 1.41, + "learning_rate": 1.8905691274919972e-05, + "loss": 0.0765, + "step": 10737 + }, + { + "epoch": 1.41, + "learning_rate": 1.8905450045144556e-05, + "loss": 0.0807, + "step": 10738 + }, + { + "epoch": 1.41, + "learning_rate": 1.8905208790323036e-05, + "loss": 0.0816, + "step": 10739 + }, + { + "epoch": 1.41, + "learning_rate": 1.890496751045609e-05, + "loss": 0.066, + "step": 10740 + }, + { + "epoch": 1.41, + "learning_rate": 1.8904726205544403e-05, + "loss": 0.0715, + "step": 10741 + }, + { + "epoch": 1.41, + "learning_rate": 1.890448487558864e-05, + "loss": 0.079, + "step": 10742 + }, + { + "epoch": 1.41, + "learning_rate": 1.8904243520589494e-05, + "loss": 0.0724, + "step": 10743 + }, + { + "epoch": 1.41, + "learning_rate": 1.8904002140547634e-05, + "loss": 0.0746, + "step": 10744 + }, + { + "epoch": 1.41, + "learning_rate": 1.8903760735463743e-05, + "loss": 0.0861, + "step": 10745 + }, + { + "epoch": 1.41, + "learning_rate": 1.89035193053385e-05, + "loss": 0.0817, + "step": 10746 + }, + { + "epoch": 1.41, + "learning_rate": 1.8903277850172583e-05, + "loss": 0.0891, + "step": 10747 + }, + { + "epoch": 1.41, + "learning_rate": 1.8903036369966665e-05, + "loss": 0.0805, + "step": 10748 + }, + { + "epoch": 1.41, + "learning_rate": 1.8902794864721436e-05, + "loss": 0.074, + "step": 10749 + }, + { + "epoch": 1.41, + "learning_rate": 1.8902553334437575e-05, + "loss": 0.098, + "step": 10750 + }, + { + "epoch": 1.41, + "learning_rate": 1.890231177911575e-05, + "loss": 0.0655, + "step": 10751 + }, + { + "epoch": 1.41, + "learning_rate": 1.890207019875665e-05, + "loss": 0.0751, + "step": 10752 + }, + { + "epoch": 1.41, + "learning_rate": 1.890182859336095e-05, + "loss": 0.0739, + "step": 10753 + }, + { + "epoch": 1.41, + "learning_rate": 1.8901586962929335e-05, + "loss": 0.0845, + "step": 10754 + }, + { + "epoch": 1.41, + "learning_rate": 1.8901345307462478e-05, + "loss": 0.0817, + "step": 10755 + }, + { + "epoch": 1.41, + "learning_rate": 1.890110362696106e-05, + "loss": 0.0741, + "step": 10756 + }, + { + "epoch": 1.41, + "learning_rate": 1.8900861921425764e-05, + "loss": 0.0822, + "step": 10757 + }, + { + "epoch": 1.41, + "learning_rate": 1.890062019085727e-05, + "loss": 0.0844, + "step": 10758 + }, + { + "epoch": 1.41, + "learning_rate": 1.890037843525625e-05, + "loss": 0.0745, + "step": 10759 + }, + { + "epoch": 1.41, + "learning_rate": 1.8900136654623393e-05, + "loss": 0.0719, + "step": 10760 + }, + { + "epoch": 1.41, + "learning_rate": 1.8899894848959377e-05, + "loss": 0.083, + "step": 10761 + }, + { + "epoch": 1.41, + "learning_rate": 1.889965301826488e-05, + "loss": 0.0768, + "step": 10762 + }, + { + "epoch": 1.41, + "learning_rate": 1.8899411162540582e-05, + "loss": 0.0757, + "step": 10763 + }, + { + "epoch": 1.41, + "learning_rate": 1.8899169281787166e-05, + "loss": 0.0835, + "step": 10764 + }, + { + "epoch": 1.41, + "learning_rate": 1.8898927376005308e-05, + "loss": 0.0832, + "step": 10765 + }, + { + "epoch": 1.41, + "learning_rate": 1.889868544519569e-05, + "loss": 0.0791, + "step": 10766 + }, + { + "epoch": 1.41, + "learning_rate": 1.8898443489358994e-05, + "loss": 0.0798, + "step": 10767 + }, + { + "epoch": 1.41, + "learning_rate": 1.88982015084959e-05, + "loss": 0.0855, + "step": 10768 + }, + { + "epoch": 1.41, + "learning_rate": 1.8897959502607086e-05, + "loss": 0.078, + "step": 10769 + }, + { + "epoch": 1.41, + "learning_rate": 1.8897717471693235e-05, + "loss": 0.0762, + "step": 10770 + }, + { + "epoch": 1.41, + "learning_rate": 1.889747541575503e-05, + "loss": 0.0793, + "step": 10771 + }, + { + "epoch": 1.41, + "learning_rate": 1.889723333479315e-05, + "loss": 0.0807, + "step": 10772 + }, + { + "epoch": 1.41, + "learning_rate": 1.8896991228808272e-05, + "loss": 0.0784, + "step": 10773 + }, + { + "epoch": 1.41, + "learning_rate": 1.8896749097801083e-05, + "loss": 0.0889, + "step": 10774 + }, + { + "epoch": 1.41, + "learning_rate": 1.889650694177226e-05, + "loss": 0.0829, + "step": 10775 + }, + { + "epoch": 1.41, + "learning_rate": 1.889626476072248e-05, + "loss": 0.0745, + "step": 10776 + }, + { + "epoch": 1.41, + "learning_rate": 1.8896022554652434e-05, + "loss": 0.0848, + "step": 10777 + }, + { + "epoch": 1.41, + "learning_rate": 1.88957803235628e-05, + "loss": 0.0769, + "step": 10778 + }, + { + "epoch": 1.41, + "learning_rate": 1.889553806745425e-05, + "loss": 0.0887, + "step": 10779 + }, + { + "epoch": 1.41, + "learning_rate": 1.889529578632748e-05, + "loss": 0.0809, + "step": 10780 + }, + { + "epoch": 1.41, + "learning_rate": 1.8895053480183164e-05, + "loss": 0.0784, + "step": 10781 + }, + { + "epoch": 1.41, + "learning_rate": 1.889481114902198e-05, + "loss": 0.0643, + "step": 10782 + }, + { + "epoch": 1.41, + "learning_rate": 1.8894568792844617e-05, + "loss": 0.08, + "step": 10783 + }, + { + "epoch": 1.41, + "learning_rate": 1.8894326411651752e-05, + "loss": 0.0888, + "step": 10784 + }, + { + "epoch": 1.41, + "learning_rate": 1.8894084005444066e-05, + "loss": 0.0761, + "step": 10785 + }, + { + "epoch": 1.41, + "learning_rate": 1.8893841574222243e-05, + "loss": 0.0757, + "step": 10786 + }, + { + "epoch": 1.41, + "learning_rate": 1.8893599117986965e-05, + "loss": 0.0741, + "step": 10787 + }, + { + "epoch": 1.41, + "learning_rate": 1.8893356636738913e-05, + "loss": 0.0754, + "step": 10788 + }, + { + "epoch": 1.41, + "learning_rate": 1.8893114130478767e-05, + "loss": 0.0741, + "step": 10789 + }, + { + "epoch": 1.41, + "learning_rate": 1.889287159920721e-05, + "loss": 0.0852, + "step": 10790 + }, + { + "epoch": 1.41, + "learning_rate": 1.889262904292493e-05, + "loss": 0.0763, + "step": 10791 + }, + { + "epoch": 1.41, + "learning_rate": 1.8892386461632605e-05, + "loss": 0.0823, + "step": 10792 + }, + { + "epoch": 1.41, + "learning_rate": 1.8892143855330915e-05, + "loss": 0.0892, + "step": 10793 + }, + { + "epoch": 1.41, + "learning_rate": 1.889190122402054e-05, + "loss": 0.0811, + "step": 10794 + }, + { + "epoch": 1.41, + "learning_rate": 1.889165856770217e-05, + "loss": 0.0741, + "step": 10795 + }, + { + "epoch": 1.41, + "learning_rate": 1.8891415886376484e-05, + "loss": 0.0807, + "step": 10796 + }, + { + "epoch": 1.41, + "learning_rate": 1.8891173180044164e-05, + "loss": 0.0859, + "step": 10797 + }, + { + "epoch": 1.41, + "learning_rate": 1.889093044870589e-05, + "loss": 0.0807, + "step": 10798 + }, + { + "epoch": 1.41, + "learning_rate": 1.889068769236235e-05, + "loss": 0.0756, + "step": 10799 + }, + { + "epoch": 1.41, + "learning_rate": 1.8890444911014226e-05, + "loss": 0.0836, + "step": 10800 + }, + { + "epoch": 1.41, + "learning_rate": 1.8890202104662196e-05, + "loss": 0.0816, + "step": 10801 + }, + { + "epoch": 1.41, + "learning_rate": 1.8889959273306946e-05, + "loss": 0.0835, + "step": 10802 + }, + { + "epoch": 1.42, + "learning_rate": 1.8889716416949163e-05, + "loss": 0.0836, + "step": 10803 + }, + { + "epoch": 1.42, + "learning_rate": 1.888947353558952e-05, + "loss": 0.0792, + "step": 10804 + }, + { + "epoch": 1.42, + "learning_rate": 1.8889230629228712e-05, + "loss": 0.099, + "step": 10805 + }, + { + "epoch": 1.42, + "learning_rate": 1.8888987697867412e-05, + "loss": 0.0687, + "step": 10806 + }, + { + "epoch": 1.42, + "learning_rate": 1.8888744741506308e-05, + "loss": 0.0751, + "step": 10807 + }, + { + "epoch": 1.42, + "learning_rate": 1.8888501760146086e-05, + "loss": 0.0956, + "step": 10808 + }, + { + "epoch": 1.42, + "learning_rate": 1.8888258753787422e-05, + "loss": 0.0723, + "step": 10809 + }, + { + "epoch": 1.42, + "learning_rate": 1.8888015722431006e-05, + "loss": 0.0791, + "step": 10810 + }, + { + "epoch": 1.42, + "learning_rate": 1.8887772666077517e-05, + "loss": 0.0669, + "step": 10811 + }, + { + "epoch": 1.42, + "learning_rate": 1.888752958472764e-05, + "loss": 0.0667, + "step": 10812 + }, + { + "epoch": 1.42, + "learning_rate": 1.8887286478382067e-05, + "loss": 0.0782, + "step": 10813 + }, + { + "epoch": 1.42, + "learning_rate": 1.888704334704147e-05, + "loss": 0.0811, + "step": 10814 + }, + { + "epoch": 1.42, + "learning_rate": 1.888680019070653e-05, + "loss": 0.0911, + "step": 10815 + }, + { + "epoch": 1.42, + "learning_rate": 1.8886557009377944e-05, + "loss": 0.0829, + "step": 10816 + }, + { + "epoch": 1.42, + "learning_rate": 1.8886313803056387e-05, + "loss": 0.0786, + "step": 10817 + }, + { + "epoch": 1.42, + "learning_rate": 1.8886070571742554e-05, + "loss": 0.0734, + "step": 10818 + }, + { + "epoch": 1.42, + "learning_rate": 1.888582731543711e-05, + "loss": 0.0879, + "step": 10819 + }, + { + "epoch": 1.42, + "learning_rate": 1.8885584034140754e-05, + "loss": 0.0875, + "step": 10820 + }, + { + "epoch": 1.42, + "learning_rate": 1.8885340727854168e-05, + "loss": 0.0786, + "step": 10821 + }, + { + "epoch": 1.42, + "learning_rate": 1.8885097396578032e-05, + "loss": 0.0831, + "step": 10822 + }, + { + "epoch": 1.42, + "learning_rate": 1.8884854040313036e-05, + "loss": 0.0795, + "step": 10823 + }, + { + "epoch": 1.42, + "learning_rate": 1.8884610659059856e-05, + "loss": 0.0822, + "step": 10824 + }, + { + "epoch": 1.42, + "learning_rate": 1.8884367252819184e-05, + "loss": 0.0824, + "step": 10825 + }, + { + "epoch": 1.42, + "learning_rate": 1.8884123821591703e-05, + "loss": 0.0788, + "step": 10826 + }, + { + "epoch": 1.42, + "learning_rate": 1.8883880365378096e-05, + "loss": 0.0822, + "step": 10827 + }, + { + "epoch": 1.42, + "learning_rate": 1.8883636884179047e-05, + "loss": 0.0748, + "step": 10828 + }, + { + "epoch": 1.42, + "learning_rate": 1.8883393377995246e-05, + "loss": 0.0652, + "step": 10829 + }, + { + "epoch": 1.42, + "learning_rate": 1.8883149846827372e-05, + "loss": 0.0831, + "step": 10830 + }, + { + "epoch": 1.42, + "learning_rate": 1.888290629067611e-05, + "loss": 0.0845, + "step": 10831 + }, + { + "epoch": 1.42, + "learning_rate": 1.888266270954215e-05, + "loss": 0.0774, + "step": 10832 + }, + { + "epoch": 1.42, + "learning_rate": 1.8882419103426175e-05, + "loss": 0.0679, + "step": 10833 + }, + { + "epoch": 1.42, + "learning_rate": 1.8882175472328867e-05, + "loss": 0.0788, + "step": 10834 + }, + { + "epoch": 1.42, + "learning_rate": 1.8881931816250917e-05, + "loss": 0.0835, + "step": 10835 + }, + { + "epoch": 1.42, + "learning_rate": 1.8881688135193004e-05, + "loss": 0.0767, + "step": 10836 + }, + { + "epoch": 1.42, + "learning_rate": 1.8881444429155815e-05, + "loss": 0.073, + "step": 10837 + }, + { + "epoch": 1.42, + "learning_rate": 1.888120069814004e-05, + "loss": 0.0768, + "step": 10838 + }, + { + "epoch": 1.42, + "learning_rate": 1.8880956942146355e-05, + "loss": 0.0808, + "step": 10839 + }, + { + "epoch": 1.42, + "learning_rate": 1.8880713161175458e-05, + "loss": 0.0919, + "step": 10840 + }, + { + "epoch": 1.42, + "learning_rate": 1.8880469355228025e-05, + "loss": 0.088, + "step": 10841 + }, + { + "epoch": 1.42, + "learning_rate": 1.8880225524304742e-05, + "loss": 0.0655, + "step": 10842 + }, + { + "epoch": 1.42, + "learning_rate": 1.8879981668406303e-05, + "loss": 0.0684, + "step": 10843 + }, + { + "epoch": 1.42, + "learning_rate": 1.887973778753339e-05, + "loss": 0.0799, + "step": 10844 + }, + { + "epoch": 1.42, + "learning_rate": 1.887949388168668e-05, + "loss": 0.0609, + "step": 10845 + }, + { + "epoch": 1.42, + "learning_rate": 1.8879249950866872e-05, + "loss": 0.0805, + "step": 10846 + }, + { + "epoch": 1.42, + "learning_rate": 1.8879005995074645e-05, + "loss": 0.0783, + "step": 10847 + }, + { + "epoch": 1.42, + "learning_rate": 1.887876201431068e-05, + "loss": 0.0777, + "step": 10848 + }, + { + "epoch": 1.42, + "learning_rate": 1.8878518008575678e-05, + "loss": 0.0775, + "step": 10849 + }, + { + "epoch": 1.42, + "learning_rate": 1.8878273977870316e-05, + "loss": 0.0839, + "step": 10850 + }, + { + "epoch": 1.42, + "learning_rate": 1.887802992219528e-05, + "loss": 0.0833, + "step": 10851 + }, + { + "epoch": 1.42, + "learning_rate": 1.8877785841551258e-05, + "loss": 0.0775, + "step": 10852 + }, + { + "epoch": 1.42, + "learning_rate": 1.8877541735938934e-05, + "loss": 0.0699, + "step": 10853 + }, + { + "epoch": 1.42, + "learning_rate": 1.8877297605359e-05, + "loss": 0.0716, + "step": 10854 + }, + { + "epoch": 1.42, + "learning_rate": 1.8877053449812134e-05, + "loss": 0.0889, + "step": 10855 + }, + { + "epoch": 1.42, + "learning_rate": 1.887680926929903e-05, + "loss": 0.0774, + "step": 10856 + }, + { + "epoch": 1.42, + "learning_rate": 1.8876565063820374e-05, + "loss": 0.0796, + "step": 10857 + }, + { + "epoch": 1.42, + "learning_rate": 1.8876320833376854e-05, + "loss": 0.0862, + "step": 10858 + }, + { + "epoch": 1.42, + "learning_rate": 1.8876076577969152e-05, + "loss": 0.0736, + "step": 10859 + }, + { + "epoch": 1.42, + "learning_rate": 1.8875832297597956e-05, + "loss": 0.0809, + "step": 10860 + }, + { + "epoch": 1.42, + "learning_rate": 1.8875587992263956e-05, + "loss": 0.0757, + "step": 10861 + }, + { + "epoch": 1.42, + "learning_rate": 1.887534366196784e-05, + "loss": 0.0861, + "step": 10862 + }, + { + "epoch": 1.42, + "learning_rate": 1.8875099306710283e-05, + "loss": 0.0763, + "step": 10863 + }, + { + "epoch": 1.42, + "learning_rate": 1.8874854926491992e-05, + "loss": 0.0782, + "step": 10864 + }, + { + "epoch": 1.42, + "learning_rate": 1.887461052131364e-05, + "loss": 0.0822, + "step": 10865 + }, + { + "epoch": 1.42, + "learning_rate": 1.8874366091175918e-05, + "loss": 0.0848, + "step": 10866 + }, + { + "epoch": 1.42, + "learning_rate": 1.8874121636079514e-05, + "loss": 0.0784, + "step": 10867 + }, + { + "epoch": 1.42, + "learning_rate": 1.8873877156025115e-05, + "loss": 0.0771, + "step": 10868 + }, + { + "epoch": 1.42, + "learning_rate": 1.8873632651013412e-05, + "loss": 0.0822, + "step": 10869 + }, + { + "epoch": 1.42, + "learning_rate": 1.887338812104509e-05, + "loss": 0.0783, + "step": 10870 + }, + { + "epoch": 1.42, + "learning_rate": 1.887314356612083e-05, + "loss": 0.0885, + "step": 10871 + }, + { + "epoch": 1.42, + "learning_rate": 1.8872898986241333e-05, + "loss": 0.0754, + "step": 10872 + }, + { + "epoch": 1.42, + "learning_rate": 1.8872654381407278e-05, + "loss": 0.0717, + "step": 10873 + }, + { + "epoch": 1.42, + "learning_rate": 1.8872409751619355e-05, + "loss": 0.085, + "step": 10874 + }, + { + "epoch": 1.42, + "learning_rate": 1.887216509687825e-05, + "loss": 0.069, + "step": 10875 + }, + { + "epoch": 1.42, + "learning_rate": 1.8871920417184658e-05, + "loss": 0.0709, + "step": 10876 + }, + { + "epoch": 1.42, + "learning_rate": 1.887167571253926e-05, + "loss": 0.0645, + "step": 10877 + }, + { + "epoch": 1.42, + "learning_rate": 1.8871430982942744e-05, + "loss": 0.0758, + "step": 10878 + }, + { + "epoch": 1.43, + "learning_rate": 1.8871186228395802e-05, + "loss": 0.0848, + "step": 10879 + }, + { + "epoch": 1.43, + "learning_rate": 1.8870941448899122e-05, + "loss": 0.079, + "step": 10880 + }, + { + "epoch": 1.43, + "learning_rate": 1.887069664445339e-05, + "loss": 0.084, + "step": 10881 + }, + { + "epoch": 1.43, + "learning_rate": 1.8870451815059298e-05, + "loss": 0.0715, + "step": 10882 + }, + { + "epoch": 1.43, + "learning_rate": 1.8870206960717533e-05, + "loss": 0.0772, + "step": 10883 + }, + { + "epoch": 1.43, + "learning_rate": 1.886996208142878e-05, + "loss": 0.0789, + "step": 10884 + }, + { + "epoch": 1.43, + "learning_rate": 1.8869717177193732e-05, + "loss": 0.0735, + "step": 10885 + }, + { + "epoch": 1.43, + "learning_rate": 1.886947224801308e-05, + "loss": 0.0859, + "step": 10886 + }, + { + "epoch": 1.43, + "learning_rate": 1.886922729388751e-05, + "loss": 0.0835, + "step": 10887 + }, + { + "epoch": 1.43, + "learning_rate": 1.8868982314817707e-05, + "loss": 0.0819, + "step": 10888 + }, + { + "epoch": 1.43, + "learning_rate": 1.8868737310804363e-05, + "loss": 0.0815, + "step": 10889 + }, + { + "epoch": 1.43, + "learning_rate": 1.8868492281848172e-05, + "loss": 0.0686, + "step": 10890 + }, + { + "epoch": 1.43, + "learning_rate": 1.8868247227949816e-05, + "loss": 0.0822, + "step": 10891 + }, + { + "epoch": 1.43, + "learning_rate": 1.8868002149109988e-05, + "loss": 0.071, + "step": 10892 + }, + { + "epoch": 1.43, + "learning_rate": 1.8867757045329376e-05, + "loss": 0.0805, + "step": 10893 + }, + { + "epoch": 1.43, + "learning_rate": 1.8867511916608666e-05, + "loss": 0.0878, + "step": 10894 + }, + { + "epoch": 1.43, + "learning_rate": 1.8867266762948554e-05, + "loss": 0.0697, + "step": 10895 + }, + { + "epoch": 1.43, + "learning_rate": 1.8867021584349726e-05, + "loss": 0.0839, + "step": 10896 + }, + { + "epoch": 1.43, + "learning_rate": 1.8866776380812872e-05, + "loss": 0.0758, + "step": 10897 + }, + { + "epoch": 1.43, + "learning_rate": 1.8866531152338684e-05, + "loss": 0.0759, + "step": 10898 + }, + { + "epoch": 1.43, + "learning_rate": 1.8866285898927845e-05, + "loss": 0.0889, + "step": 10899 + }, + { + "epoch": 1.43, + "learning_rate": 1.8866040620581052e-05, + "loss": 0.0718, + "step": 10900 + }, + { + "epoch": 1.43, + "learning_rate": 1.886579531729899e-05, + "loss": 0.0817, + "step": 10901 + }, + { + "epoch": 1.43, + "learning_rate": 1.8865549989082353e-05, + "loss": 0.0867, + "step": 10902 + }, + { + "epoch": 1.43, + "learning_rate": 1.8865304635931826e-05, + "loss": 0.0903, + "step": 10903 + }, + { + "epoch": 1.43, + "learning_rate": 1.8865059257848106e-05, + "loss": 0.0786, + "step": 10904 + }, + { + "epoch": 1.43, + "learning_rate": 1.8864813854831876e-05, + "loss": 0.0743, + "step": 10905 + }, + { + "epoch": 1.43, + "learning_rate": 1.886456842688383e-05, + "loss": 0.0764, + "step": 10906 + }, + { + "epoch": 1.43, + "learning_rate": 1.8864322974004655e-05, + "loss": 0.0799, + "step": 10907 + }, + { + "epoch": 1.43, + "learning_rate": 1.8864077496195046e-05, + "loss": 0.0796, + "step": 10908 + }, + { + "epoch": 1.43, + "learning_rate": 1.8863831993455693e-05, + "loss": 0.0815, + "step": 10909 + }, + { + "epoch": 1.43, + "learning_rate": 1.886358646578728e-05, + "loss": 0.0883, + "step": 10910 + }, + { + "epoch": 1.43, + "learning_rate": 1.8863340913190505e-05, + "loss": 0.0785, + "step": 10911 + }, + { + "epoch": 1.43, + "learning_rate": 1.8863095335666054e-05, + "loss": 0.0827, + "step": 10912 + }, + { + "epoch": 1.43, + "learning_rate": 1.8862849733214622e-05, + "loss": 0.0746, + "step": 10913 + }, + { + "epoch": 1.43, + "learning_rate": 1.8862604105836894e-05, + "loss": 0.0783, + "step": 10914 + }, + { + "epoch": 1.43, + "learning_rate": 1.8862358453533566e-05, + "loss": 0.0689, + "step": 10915 + }, + { + "epoch": 1.43, + "learning_rate": 1.8862112776305327e-05, + "loss": 0.0806, + "step": 10916 + }, + { + "epoch": 1.43, + "learning_rate": 1.8861867074152865e-05, + "loss": 0.0822, + "step": 10917 + }, + { + "epoch": 1.43, + "learning_rate": 1.8861621347076876e-05, + "loss": 0.0862, + "step": 10918 + }, + { + "epoch": 1.43, + "learning_rate": 1.8861375595078047e-05, + "loss": 0.0904, + "step": 10919 + }, + { + "epoch": 1.43, + "learning_rate": 1.8861129818157076e-05, + "loss": 0.0818, + "step": 10920 + }, + { + "epoch": 1.43, + "learning_rate": 1.8860884016314645e-05, + "loss": 0.0928, + "step": 10921 + }, + { + "epoch": 1.43, + "learning_rate": 1.886063818955145e-05, + "loss": 0.0759, + "step": 10922 + }, + { + "epoch": 1.43, + "learning_rate": 1.8860392337868183e-05, + "loss": 0.0725, + "step": 10923 + }, + { + "epoch": 1.43, + "learning_rate": 1.8860146461265535e-05, + "loss": 0.0712, + "step": 10924 + }, + { + "epoch": 1.43, + "learning_rate": 1.8859900559744195e-05, + "loss": 0.086, + "step": 10925 + }, + { + "epoch": 1.43, + "learning_rate": 1.8859654633304857e-05, + "loss": 0.0929, + "step": 10926 + }, + { + "epoch": 1.43, + "learning_rate": 1.885940868194821e-05, + "loss": 0.0762, + "step": 10927 + }, + { + "epoch": 1.43, + "learning_rate": 1.8859162705674953e-05, + "loss": 0.0863, + "step": 10928 + }, + { + "epoch": 1.43, + "learning_rate": 1.885891670448577e-05, + "loss": 0.0818, + "step": 10929 + }, + { + "epoch": 1.43, + "learning_rate": 1.8858670678381353e-05, + "loss": 0.0809, + "step": 10930 + }, + { + "epoch": 1.43, + "learning_rate": 1.88584246273624e-05, + "loss": 0.0719, + "step": 10931 + }, + { + "epoch": 1.43, + "learning_rate": 1.8858178551429597e-05, + "loss": 0.0893, + "step": 10932 + }, + { + "epoch": 1.43, + "learning_rate": 1.885793245058364e-05, + "loss": 0.073, + "step": 10933 + }, + { + "epoch": 1.43, + "learning_rate": 1.885768632482522e-05, + "loss": 0.075, + "step": 10934 + }, + { + "epoch": 1.43, + "learning_rate": 1.885744017415503e-05, + "loss": 0.0793, + "step": 10935 + }, + { + "epoch": 1.43, + "learning_rate": 1.885719399857376e-05, + "loss": 0.0808, + "step": 10936 + }, + { + "epoch": 1.43, + "learning_rate": 1.88569477980821e-05, + "loss": 0.0775, + "step": 10937 + }, + { + "epoch": 1.43, + "learning_rate": 1.8856701572680752e-05, + "loss": 0.0815, + "step": 10938 + }, + { + "epoch": 1.43, + "learning_rate": 1.88564553223704e-05, + "loss": 0.0753, + "step": 10939 + }, + { + "epoch": 1.43, + "learning_rate": 1.8856209047151738e-05, + "loss": 0.0682, + "step": 10940 + }, + { + "epoch": 1.43, + "learning_rate": 1.885596274702546e-05, + "loss": 0.078, + "step": 10941 + }, + { + "epoch": 1.43, + "learning_rate": 1.8855716421992262e-05, + "loss": 0.0762, + "step": 10942 + }, + { + "epoch": 1.43, + "learning_rate": 1.8855470072052828e-05, + "loss": 0.0713, + "step": 10943 + }, + { + "epoch": 1.43, + "learning_rate": 1.8855223697207857e-05, + "loss": 0.0733, + "step": 10944 + }, + { + "epoch": 1.43, + "learning_rate": 1.8854977297458044e-05, + "loss": 0.0869, + "step": 10945 + }, + { + "epoch": 1.43, + "learning_rate": 1.8854730872804073e-05, + "loss": 0.0769, + "step": 10946 + }, + { + "epoch": 1.43, + "learning_rate": 1.885448442324665e-05, + "loss": 0.0912, + "step": 10947 + }, + { + "epoch": 1.43, + "learning_rate": 1.8854237948786457e-05, + "loss": 0.0757, + "step": 10948 + }, + { + "epoch": 1.43, + "learning_rate": 1.885399144942419e-05, + "loss": 0.0803, + "step": 10949 + }, + { + "epoch": 1.43, + "learning_rate": 1.8853744925160548e-05, + "loss": 0.0733, + "step": 10950 + }, + { + "epoch": 1.43, + "learning_rate": 1.8853498375996214e-05, + "loss": 0.0829, + "step": 10951 + }, + { + "epoch": 1.43, + "learning_rate": 1.8853251801931892e-05, + "loss": 0.0764, + "step": 10952 + }, + { + "epoch": 1.43, + "learning_rate": 1.8853005202968272e-05, + "loss": 0.0899, + "step": 10953 + }, + { + "epoch": 1.43, + "learning_rate": 1.8852758579106043e-05, + "loss": 0.0789, + "step": 10954 + }, + { + "epoch": 1.44, + "learning_rate": 1.8852511930345905e-05, + "loss": 0.0794, + "step": 10955 + }, + { + "epoch": 1.44, + "learning_rate": 1.8852265256688544e-05, + "loss": 0.0743, + "step": 10956 + }, + { + "epoch": 1.44, + "learning_rate": 1.8852018558134662e-05, + "loss": 0.0682, + "step": 10957 + }, + { + "epoch": 1.44, + "learning_rate": 1.885177183468495e-05, + "loss": 0.0872, + "step": 10958 + }, + { + "epoch": 1.44, + "learning_rate": 1.8851525086340097e-05, + "loss": 0.0689, + "step": 10959 + }, + { + "epoch": 1.44, + "learning_rate": 1.8851278313100803e-05, + "loss": 0.0876, + "step": 10960 + }, + { + "epoch": 1.44, + "learning_rate": 1.8851031514967762e-05, + "loss": 0.0838, + "step": 10961 + }, + { + "epoch": 1.44, + "learning_rate": 1.8850784691941664e-05, + "loss": 0.0705, + "step": 10962 + }, + { + "epoch": 1.44, + "learning_rate": 1.8850537844023205e-05, + "loss": 0.077, + "step": 10963 + }, + { + "epoch": 1.44, + "learning_rate": 1.885029097121308e-05, + "loss": 0.0847, + "step": 10964 + }, + { + "epoch": 1.44, + "learning_rate": 1.8850044073511985e-05, + "loss": 0.0768, + "step": 10965 + }, + { + "epoch": 1.44, + "learning_rate": 1.884979715092061e-05, + "loss": 0.0789, + "step": 10966 + }, + { + "epoch": 1.44, + "learning_rate": 1.884955020343965e-05, + "loss": 0.0759, + "step": 10967 + }, + { + "epoch": 1.44, + "learning_rate": 1.8849303231069806e-05, + "loss": 0.0787, + "step": 10968 + }, + { + "epoch": 1.44, + "learning_rate": 1.8849056233811768e-05, + "loss": 0.0742, + "step": 10969 + }, + { + "epoch": 1.44, + "learning_rate": 1.8848809211666225e-05, + "loss": 0.0711, + "step": 10970 + }, + { + "epoch": 1.44, + "learning_rate": 1.8848562164633885e-05, + "loss": 0.0801, + "step": 10971 + }, + { + "epoch": 1.44, + "learning_rate": 1.8848315092715428e-05, + "loss": 0.0816, + "step": 10972 + }, + { + "epoch": 1.44, + "learning_rate": 1.8848067995911557e-05, + "loss": 0.0666, + "step": 10973 + }, + { + "epoch": 1.44, + "learning_rate": 1.884782087422297e-05, + "loss": 0.068, + "step": 10974 + }, + { + "epoch": 1.44, + "learning_rate": 1.8847573727650353e-05, + "loss": 0.0683, + "step": 10975 + }, + { + "epoch": 1.44, + "learning_rate": 1.8847326556194413e-05, + "loss": 0.0827, + "step": 10976 + }, + { + "epoch": 1.44, + "learning_rate": 1.8847079359855832e-05, + "loss": 0.0945, + "step": 10977 + }, + { + "epoch": 1.44, + "learning_rate": 1.8846832138635314e-05, + "loss": 0.0856, + "step": 10978 + }, + { + "epoch": 1.44, + "learning_rate": 1.884658489253355e-05, + "loss": 0.0689, + "step": 10979 + }, + { + "epoch": 1.44, + "learning_rate": 1.884633762155124e-05, + "loss": 0.0778, + "step": 10980 + }, + { + "epoch": 1.44, + "learning_rate": 1.8846090325689076e-05, + "loss": 0.0774, + "step": 10981 + }, + { + "epoch": 1.44, + "learning_rate": 1.8845843004947753e-05, + "loss": 0.083, + "step": 10982 + }, + { + "epoch": 1.44, + "learning_rate": 1.8845595659327967e-05, + "loss": 0.0762, + "step": 10983 + }, + { + "epoch": 1.44, + "learning_rate": 1.8845348288830414e-05, + "loss": 0.0782, + "step": 10984 + }, + { + "epoch": 1.44, + "learning_rate": 1.884510089345579e-05, + "loss": 0.0725, + "step": 10985 + }, + { + "epoch": 1.44, + "learning_rate": 1.8844853473204793e-05, + "loss": 0.0756, + "step": 10986 + }, + { + "epoch": 1.44, + "learning_rate": 1.8844606028078118e-05, + "loss": 0.0798, + "step": 10987 + }, + { + "epoch": 1.44, + "learning_rate": 1.8844358558076453e-05, + "loss": 0.0795, + "step": 10988 + }, + { + "epoch": 1.44, + "learning_rate": 1.8844111063200504e-05, + "loss": 0.0829, + "step": 10989 + }, + { + "epoch": 1.44, + "learning_rate": 1.8843863543450965e-05, + "loss": 0.0834, + "step": 10990 + }, + { + "epoch": 1.44, + "learning_rate": 1.884361599882853e-05, + "loss": 0.0844, + "step": 10991 + }, + { + "epoch": 1.44, + "learning_rate": 1.8843368429333895e-05, + "loss": 0.0783, + "step": 10992 + }, + { + "epoch": 1.44, + "learning_rate": 1.884312083496776e-05, + "loss": 0.0858, + "step": 10993 + }, + { + "epoch": 1.44, + "learning_rate": 1.8842873215730813e-05, + "loss": 0.0879, + "step": 10994 + }, + { + "epoch": 1.44, + "learning_rate": 1.8842625571623762e-05, + "loss": 0.0872, + "step": 10995 + }, + { + "epoch": 1.44, + "learning_rate": 1.8842377902647292e-05, + "loss": 0.086, + "step": 10996 + }, + { + "epoch": 1.44, + "learning_rate": 1.884213020880211e-05, + "loss": 0.0797, + "step": 10997 + }, + { + "epoch": 1.44, + "learning_rate": 1.8841882490088907e-05, + "loss": 0.0791, + "step": 10998 + }, + { + "epoch": 1.44, + "learning_rate": 1.8841634746508374e-05, + "loss": 0.0731, + "step": 10999 + }, + { + "epoch": 1.44, + "learning_rate": 1.884138697806122e-05, + "loss": 0.0747, + "step": 11000 + }, + { + "epoch": 1.44, + "learning_rate": 1.8841139184748137e-05, + "loss": 0.0737, + "step": 11001 + }, + { + "epoch": 1.44, + "learning_rate": 1.8840891366569818e-05, + "loss": 0.0772, + "step": 11002 + }, + { + "epoch": 1.44, + "learning_rate": 1.8840643523526963e-05, + "loss": 0.0652, + "step": 11003 + }, + { + "epoch": 1.44, + "learning_rate": 1.8840395655620272e-05, + "loss": 0.0843, + "step": 11004 + }, + { + "epoch": 1.44, + "learning_rate": 1.8840147762850437e-05, + "loss": 0.0741, + "step": 11005 + }, + { + "epoch": 1.44, + "learning_rate": 1.8839899845218155e-05, + "loss": 0.07, + "step": 11006 + }, + { + "epoch": 1.44, + "learning_rate": 1.8839651902724128e-05, + "loss": 0.0824, + "step": 11007 + }, + { + "epoch": 1.44, + "learning_rate": 1.883940393536905e-05, + "loss": 0.0762, + "step": 11008 + }, + { + "epoch": 1.44, + "learning_rate": 1.8839155943153617e-05, + "loss": 0.0768, + "step": 11009 + }, + { + "epoch": 1.44, + "learning_rate": 1.8838907926078532e-05, + "loss": 0.0719, + "step": 11010 + }, + { + "epoch": 1.44, + "learning_rate": 1.8838659884144487e-05, + "loss": 0.0964, + "step": 11011 + }, + { + "epoch": 1.44, + "learning_rate": 1.8838411817352183e-05, + "loss": 0.0847, + "step": 11012 + }, + { + "epoch": 1.44, + "learning_rate": 1.8838163725702316e-05, + "loss": 0.0758, + "step": 11013 + }, + { + "epoch": 1.44, + "learning_rate": 1.8837915609195585e-05, + "loss": 0.0846, + "step": 11014 + }, + { + "epoch": 1.44, + "learning_rate": 1.8837667467832686e-05, + "loss": 0.0787, + "step": 11015 + }, + { + "epoch": 1.44, + "learning_rate": 1.8837419301614316e-05, + "loss": 0.0718, + "step": 11016 + }, + { + "epoch": 1.44, + "learning_rate": 1.8837171110541178e-05, + "loss": 0.0773, + "step": 11017 + }, + { + "epoch": 1.44, + "learning_rate": 1.883692289461397e-05, + "loss": 0.0767, + "step": 11018 + }, + { + "epoch": 1.44, + "learning_rate": 1.883667465383338e-05, + "loss": 0.0852, + "step": 11019 + }, + { + "epoch": 1.44, + "learning_rate": 1.8836426388200115e-05, + "loss": 0.0778, + "step": 11020 + }, + { + "epoch": 1.44, + "learning_rate": 1.8836178097714874e-05, + "loss": 0.0745, + "step": 11021 + }, + { + "epoch": 1.44, + "learning_rate": 1.883592978237835e-05, + "loss": 0.0757, + "step": 11022 + }, + { + "epoch": 1.44, + "learning_rate": 1.8835681442191244e-05, + "loss": 0.0839, + "step": 11023 + }, + { + "epoch": 1.44, + "learning_rate": 1.8835433077154255e-05, + "loss": 0.0798, + "step": 11024 + }, + { + "epoch": 1.44, + "learning_rate": 1.883518468726808e-05, + "loss": 0.0698, + "step": 11025 + }, + { + "epoch": 1.44, + "learning_rate": 1.8834936272533423e-05, + "loss": 0.0734, + "step": 11026 + }, + { + "epoch": 1.44, + "learning_rate": 1.8834687832950976e-05, + "loss": 0.0839, + "step": 11027 + }, + { + "epoch": 1.44, + "learning_rate": 1.8834439368521437e-05, + "loss": 0.0696, + "step": 11028 + }, + { + "epoch": 1.44, + "learning_rate": 1.883419087924551e-05, + "loss": 0.0892, + "step": 11029 + }, + { + "epoch": 1.44, + "learning_rate": 1.883394236512389e-05, + "loss": 0.0745, + "step": 11030 + }, + { + "epoch": 1.44, + "learning_rate": 1.883369382615728e-05, + "loss": 0.0748, + "step": 11031 + }, + { + "epoch": 1.45, + "learning_rate": 1.8833445262346376e-05, + "loss": 0.0706, + "step": 11032 + }, + { + "epoch": 1.45, + "learning_rate": 1.8833196673691876e-05, + "loss": 0.0874, + "step": 11033 + }, + { + "epoch": 1.45, + "learning_rate": 1.8832948060194486e-05, + "loss": 0.079, + "step": 11034 + }, + { + "epoch": 1.45, + "learning_rate": 1.883269942185489e-05, + "loss": 0.0818, + "step": 11035 + }, + { + "epoch": 1.45, + "learning_rate": 1.8832450758673807e-05, + "loss": 0.074, + "step": 11036 + }, + { + "epoch": 1.45, + "learning_rate": 1.883220207065192e-05, + "loss": 0.0802, + "step": 11037 + }, + { + "epoch": 1.45, + "learning_rate": 1.883195335778994e-05, + "loss": 0.0837, + "step": 11038 + }, + { + "epoch": 1.45, + "learning_rate": 1.883170462008856e-05, + "loss": 0.0693, + "step": 11039 + }, + { + "epoch": 1.45, + "learning_rate": 1.883145585754848e-05, + "loss": 0.0797, + "step": 11040 + }, + { + "epoch": 1.45, + "learning_rate": 1.88312070701704e-05, + "loss": 0.0838, + "step": 11041 + }, + { + "epoch": 1.45, + "learning_rate": 1.8830958257955022e-05, + "loss": 0.0781, + "step": 11042 + }, + { + "epoch": 1.45, + "learning_rate": 1.8830709420903042e-05, + "loss": 0.0821, + "step": 11043 + }, + { + "epoch": 1.45, + "learning_rate": 1.8830460559015165e-05, + "loss": 0.0824, + "step": 11044 + }, + { + "epoch": 1.45, + "learning_rate": 1.8830211672292088e-05, + "loss": 0.0755, + "step": 11045 + }, + { + "epoch": 1.45, + "learning_rate": 1.8829962760734507e-05, + "loss": 0.0787, + "step": 11046 + }, + { + "epoch": 1.45, + "learning_rate": 1.882971382434313e-05, + "loss": 0.0914, + "step": 11047 + }, + { + "epoch": 1.45, + "learning_rate": 1.8829464863118653e-05, + "loss": 0.0801, + "step": 11048 + }, + { + "epoch": 1.45, + "learning_rate": 1.8829215877061773e-05, + "loss": 0.0705, + "step": 11049 + }, + { + "epoch": 1.45, + "learning_rate": 1.8828966866173194e-05, + "loss": 0.0847, + "step": 11050 + }, + { + "epoch": 1.45, + "learning_rate": 1.8828717830453617e-05, + "loss": 0.0847, + "step": 11051 + }, + { + "epoch": 1.45, + "learning_rate": 1.882846876990374e-05, + "loss": 0.0723, + "step": 11052 + }, + { + "epoch": 1.45, + "learning_rate": 1.8828219684524268e-05, + "loss": 0.0826, + "step": 11053 + }, + { + "epoch": 1.45, + "learning_rate": 1.88279705743159e-05, + "loss": 0.0711, + "step": 11054 + }, + { + "epoch": 1.45, + "learning_rate": 1.8827721439279327e-05, + "loss": 0.0825, + "step": 11055 + }, + { + "epoch": 1.45, + "learning_rate": 1.8827472279415263e-05, + "loss": 0.0751, + "step": 11056 + }, + { + "epoch": 1.45, + "learning_rate": 1.88272230947244e-05, + "loss": 0.0724, + "step": 11057 + }, + { + "epoch": 1.45, + "learning_rate": 1.8826973885207446e-05, + "loss": 0.0841, + "step": 11058 + }, + { + "epoch": 1.45, + "learning_rate": 1.8826724650865097e-05, + "loss": 0.0816, + "step": 11059 + }, + { + "epoch": 1.45, + "learning_rate": 1.8826475391698052e-05, + "loss": 0.0827, + "step": 11060 + }, + { + "epoch": 1.45, + "learning_rate": 1.8826226107707016e-05, + "loss": 0.0706, + "step": 11061 + }, + { + "epoch": 1.45, + "learning_rate": 1.8825976798892694e-05, + "loss": 0.0807, + "step": 11062 + }, + { + "epoch": 1.45, + "learning_rate": 1.8825727465255776e-05, + "loss": 0.0741, + "step": 11063 + }, + { + "epoch": 1.45, + "learning_rate": 1.8825478106796975e-05, + "loss": 0.0805, + "step": 11064 + }, + { + "epoch": 1.45, + "learning_rate": 1.8825228723516984e-05, + "loss": 0.0764, + "step": 11065 + }, + { + "epoch": 1.45, + "learning_rate": 1.8824979315416505e-05, + "loss": 0.0877, + "step": 11066 + }, + { + "epoch": 1.45, + "learning_rate": 1.8824729882496244e-05, + "loss": 0.0725, + "step": 11067 + }, + { + "epoch": 1.45, + "learning_rate": 1.88244804247569e-05, + "loss": 0.0833, + "step": 11068 + }, + { + "epoch": 1.45, + "learning_rate": 1.8824230942199176e-05, + "loss": 0.0765, + "step": 11069 + }, + { + "epoch": 1.45, + "learning_rate": 1.882398143482377e-05, + "loss": 0.072, + "step": 11070 + }, + { + "epoch": 1.45, + "learning_rate": 1.8823731902631386e-05, + "loss": 0.075, + "step": 11071 + }, + { + "epoch": 1.45, + "learning_rate": 1.882348234562273e-05, + "loss": 0.0963, + "step": 11072 + }, + { + "epoch": 1.45, + "learning_rate": 1.8823232763798493e-05, + "loss": 0.0736, + "step": 11073 + }, + { + "epoch": 1.45, + "learning_rate": 1.882298315715939e-05, + "loss": 0.0696, + "step": 11074 + }, + { + "epoch": 1.45, + "learning_rate": 1.8822733525706112e-05, + "loss": 0.0739, + "step": 11075 + }, + { + "epoch": 1.45, + "learning_rate": 1.882248386943937e-05, + "loss": 0.0811, + "step": 11076 + }, + { + "epoch": 1.45, + "learning_rate": 1.8822234188359857e-05, + "loss": 0.0787, + "step": 11077 + }, + { + "epoch": 1.45, + "learning_rate": 1.8821984482468286e-05, + "loss": 0.0698, + "step": 11078 + }, + { + "epoch": 1.45, + "learning_rate": 1.8821734751765352e-05, + "loss": 0.0881, + "step": 11079 + }, + { + "epoch": 1.45, + "learning_rate": 1.8821484996251755e-05, + "loss": 0.0859, + "step": 11080 + }, + { + "epoch": 1.45, + "learning_rate": 1.8821235215928205e-05, + "loss": 0.0751, + "step": 11081 + }, + { + "epoch": 1.45, + "learning_rate": 1.8820985410795398e-05, + "loss": 0.0725, + "step": 11082 + }, + { + "epoch": 1.45, + "learning_rate": 1.882073558085404e-05, + "loss": 0.0801, + "step": 11083 + }, + { + "epoch": 1.45, + "learning_rate": 1.8820485726104834e-05, + "loss": 0.0898, + "step": 11084 + }, + { + "epoch": 1.45, + "learning_rate": 1.882023584654848e-05, + "loss": 0.0825, + "step": 11085 + }, + { + "epoch": 1.45, + "learning_rate": 1.8819985942185686e-05, + "loss": 0.0707, + "step": 11086 + }, + { + "epoch": 1.45, + "learning_rate": 1.8819736013017145e-05, + "loss": 0.0693, + "step": 11087 + }, + { + "epoch": 1.45, + "learning_rate": 1.8819486059043573e-05, + "loss": 0.0698, + "step": 11088 + }, + { + "epoch": 1.45, + "learning_rate": 1.881923608026566e-05, + "loss": 0.0715, + "step": 11089 + }, + { + "epoch": 1.45, + "learning_rate": 1.881898607668412e-05, + "loss": 0.09, + "step": 11090 + }, + { + "epoch": 1.45, + "learning_rate": 1.881873604829965e-05, + "loss": 0.0753, + "step": 11091 + }, + { + "epoch": 1.45, + "learning_rate": 1.8818485995112953e-05, + "loss": 0.0646, + "step": 11092 + }, + { + "epoch": 1.45, + "learning_rate": 1.8818235917124735e-05, + "loss": 0.0791, + "step": 11093 + }, + { + "epoch": 1.45, + "learning_rate": 1.8817985814335697e-05, + "loss": 0.075, + "step": 11094 + }, + { + "epoch": 1.45, + "learning_rate": 1.8817735686746545e-05, + "loss": 0.0717, + "step": 11095 + }, + { + "epoch": 1.45, + "learning_rate": 1.881748553435798e-05, + "loss": 0.0799, + "step": 11096 + }, + { + "epoch": 1.45, + "learning_rate": 1.881723535717071e-05, + "loss": 0.0845, + "step": 11097 + }, + { + "epoch": 1.45, + "learning_rate": 1.8816985155185427e-05, + "loss": 0.0673, + "step": 11098 + }, + { + "epoch": 1.45, + "learning_rate": 1.8816734928402847e-05, + "loss": 0.0749, + "step": 11099 + }, + { + "epoch": 1.45, + "learning_rate": 1.8816484676823674e-05, + "loss": 0.0821, + "step": 11100 + }, + { + "epoch": 1.45, + "learning_rate": 1.8816234400448603e-05, + "loss": 0.0812, + "step": 11101 + }, + { + "epoch": 1.45, + "learning_rate": 1.8815984099278343e-05, + "loss": 0.077, + "step": 11102 + }, + { + "epoch": 1.45, + "learning_rate": 1.8815733773313597e-05, + "loss": 0.0814, + "step": 11103 + }, + { + "epoch": 1.45, + "learning_rate": 1.8815483422555067e-05, + "loss": 0.0915, + "step": 11104 + }, + { + "epoch": 1.45, + "learning_rate": 1.8815233047003464e-05, + "loss": 0.0795, + "step": 11105 + }, + { + "epoch": 1.45, + "learning_rate": 1.8814982646659487e-05, + "loss": 0.0755, + "step": 11106 + }, + { + "epoch": 1.45, + "learning_rate": 1.8814732221523834e-05, + "loss": 0.079, + "step": 11107 + }, + { + "epoch": 1.46, + "learning_rate": 1.8814481771597225e-05, + "loss": 0.0813, + "step": 11108 + }, + { + "epoch": 1.46, + "learning_rate": 1.881423129688035e-05, + "loss": 0.0872, + "step": 11109 + }, + { + "epoch": 1.46, + "learning_rate": 1.8813980797373918e-05, + "loss": 0.0719, + "step": 11110 + }, + { + "epoch": 1.46, + "learning_rate": 1.8813730273078636e-05, + "loss": 0.0797, + "step": 11111 + }, + { + "epoch": 1.46, + "learning_rate": 1.881347972399521e-05, + "loss": 0.0713, + "step": 11112 + }, + { + "epoch": 1.46, + "learning_rate": 1.881322915012434e-05, + "loss": 0.0863, + "step": 11113 + }, + { + "epoch": 1.46, + "learning_rate": 1.8812978551466727e-05, + "loss": 0.079, + "step": 11114 + }, + { + "epoch": 1.46, + "learning_rate": 1.8812727928023088e-05, + "loss": 0.0772, + "step": 11115 + }, + { + "epoch": 1.46, + "learning_rate": 1.8812477279794116e-05, + "loss": 0.0871, + "step": 11116 + }, + { + "epoch": 1.46, + "learning_rate": 1.8812226606780526e-05, + "loss": 0.0749, + "step": 11117 + }, + { + "epoch": 1.46, + "learning_rate": 1.8811975908983013e-05, + "loss": 0.0732, + "step": 11118 + }, + { + "epoch": 1.46, + "learning_rate": 1.881172518640229e-05, + "loss": 0.0672, + "step": 11119 + }, + { + "epoch": 1.46, + "learning_rate": 1.881147443903906e-05, + "loss": 0.082, + "step": 11120 + }, + { + "epoch": 1.46, + "learning_rate": 1.8811223666894023e-05, + "loss": 0.0823, + "step": 11121 + }, + { + "epoch": 1.46, + "learning_rate": 1.8810972869967892e-05, + "loss": 0.0838, + "step": 11122 + }, + { + "epoch": 1.46, + "learning_rate": 1.8810722048261372e-05, + "loss": 0.0786, + "step": 11123 + }, + { + "epoch": 1.46, + "learning_rate": 1.8810471201775164e-05, + "loss": 0.0743, + "step": 11124 + }, + { + "epoch": 1.46, + "learning_rate": 1.881022033050997e-05, + "loss": 0.079, + "step": 11125 + }, + { + "epoch": 1.46, + "learning_rate": 1.8809969434466505e-05, + "loss": 0.0724, + "step": 11126 + }, + { + "epoch": 1.46, + "learning_rate": 1.8809718513645468e-05, + "loss": 0.0835, + "step": 11127 + }, + { + "epoch": 1.46, + "learning_rate": 1.8809467568047565e-05, + "loss": 0.0827, + "step": 11128 + }, + { + "epoch": 1.46, + "learning_rate": 1.880921659767351e-05, + "loss": 0.0771, + "step": 11129 + }, + { + "epoch": 1.46, + "learning_rate": 1.8808965602523995e-05, + "loss": 0.0747, + "step": 11130 + }, + { + "epoch": 1.46, + "learning_rate": 1.8808714582599737e-05, + "loss": 0.0751, + "step": 11131 + }, + { + "epoch": 1.46, + "learning_rate": 1.880846353790144e-05, + "loss": 0.0825, + "step": 11132 + }, + { + "epoch": 1.46, + "learning_rate": 1.8808212468429806e-05, + "loss": 0.0813, + "step": 11133 + }, + { + "epoch": 1.46, + "learning_rate": 1.8807961374185543e-05, + "loss": 0.07, + "step": 11134 + }, + { + "epoch": 1.46, + "learning_rate": 1.880771025516936e-05, + "loss": 0.0845, + "step": 11135 + }, + { + "epoch": 1.46, + "learning_rate": 1.880745911138196e-05, + "loss": 0.0694, + "step": 11136 + }, + { + "epoch": 1.46, + "learning_rate": 1.880720794282405e-05, + "loss": 0.0734, + "step": 11137 + }, + { + "epoch": 1.46, + "learning_rate": 1.8806956749496335e-05, + "loss": 0.0825, + "step": 11138 + }, + { + "epoch": 1.46, + "learning_rate": 1.8806705531399524e-05, + "loss": 0.0726, + "step": 11139 + }, + { + "epoch": 1.46, + "learning_rate": 1.8806454288534324e-05, + "loss": 0.0825, + "step": 11140 + }, + { + "epoch": 1.46, + "learning_rate": 1.880620302090144e-05, + "loss": 0.0753, + "step": 11141 + }, + { + "epoch": 1.46, + "learning_rate": 1.8805951728501578e-05, + "loss": 0.0702, + "step": 11142 + }, + { + "epoch": 1.46, + "learning_rate": 1.8805700411335447e-05, + "loss": 0.0786, + "step": 11143 + }, + { + "epoch": 1.46, + "learning_rate": 1.880544906940375e-05, + "loss": 0.074, + "step": 11144 + }, + { + "epoch": 1.46, + "learning_rate": 1.88051977027072e-05, + "loss": 0.0798, + "step": 11145 + }, + { + "epoch": 1.46, + "learning_rate": 1.8804946311246498e-05, + "loss": 0.0774, + "step": 11146 + }, + { + "epoch": 1.46, + "learning_rate": 1.8804694895022354e-05, + "loss": 0.0847, + "step": 11147 + }, + { + "epoch": 1.46, + "learning_rate": 1.8804443454035474e-05, + "loss": 0.085, + "step": 11148 + }, + { + "epoch": 1.46, + "learning_rate": 1.8804191988286568e-05, + "loss": 0.0718, + "step": 11149 + }, + { + "epoch": 1.46, + "learning_rate": 1.8803940497776337e-05, + "loss": 0.084, + "step": 11150 + }, + { + "epoch": 1.46, + "learning_rate": 1.8803688982505494e-05, + "loss": 0.0777, + "step": 11151 + }, + { + "epoch": 1.46, + "learning_rate": 1.8803437442474746e-05, + "loss": 0.0717, + "step": 11152 + }, + { + "epoch": 1.46, + "learning_rate": 1.88031858776848e-05, + "loss": 0.0747, + "step": 11153 + }, + { + "epoch": 1.46, + "learning_rate": 1.880293428813636e-05, + "loss": 0.0825, + "step": 11154 + }, + { + "epoch": 1.46, + "learning_rate": 1.880268267383014e-05, + "loss": 0.0811, + "step": 11155 + }, + { + "epoch": 1.46, + "learning_rate": 1.8802431034766836e-05, + "loss": 0.084, + "step": 11156 + }, + { + "epoch": 1.46, + "learning_rate": 1.8802179370947166e-05, + "loss": 0.0705, + "step": 11157 + }, + { + "epoch": 1.46, + "learning_rate": 1.880192768237184e-05, + "loss": 0.0699, + "step": 11158 + }, + { + "epoch": 1.46, + "learning_rate": 1.8801675969041557e-05, + "loss": 0.0765, + "step": 11159 + }, + { + "epoch": 1.46, + "learning_rate": 1.880142423095703e-05, + "loss": 0.073, + "step": 11160 + }, + { + "epoch": 1.46, + "learning_rate": 1.8801172468118968e-05, + "loss": 0.0768, + "step": 11161 + }, + { + "epoch": 1.46, + "learning_rate": 1.8800920680528075e-05, + "loss": 0.0815, + "step": 11162 + }, + { + "epoch": 1.46, + "learning_rate": 1.8800668868185058e-05, + "loss": 0.09, + "step": 11163 + }, + { + "epoch": 1.46, + "learning_rate": 1.8800417031090633e-05, + "loss": 0.077, + "step": 11164 + }, + { + "epoch": 1.46, + "learning_rate": 1.8800165169245505e-05, + "loss": 0.0679, + "step": 11165 + }, + { + "epoch": 1.46, + "learning_rate": 1.8799913282650376e-05, + "loss": 0.0862, + "step": 11166 + }, + { + "epoch": 1.46, + "learning_rate": 1.8799661371305965e-05, + "loss": 0.0826, + "step": 11167 + }, + { + "epoch": 1.46, + "learning_rate": 1.879940943521297e-05, + "loss": 0.0777, + "step": 11168 + }, + { + "epoch": 1.46, + "learning_rate": 1.879915747437211e-05, + "loss": 0.075, + "step": 11169 + }, + { + "epoch": 1.46, + "learning_rate": 1.8798905488784083e-05, + "loss": 0.0793, + "step": 11170 + }, + { + "epoch": 1.46, + "learning_rate": 1.8798653478449604e-05, + "loss": 0.0653, + "step": 11171 + }, + { + "epoch": 1.46, + "learning_rate": 1.8798401443369383e-05, + "loss": 0.0738, + "step": 11172 + }, + { + "epoch": 1.46, + "learning_rate": 1.8798149383544124e-05, + "loss": 0.0788, + "step": 11173 + }, + { + "epoch": 1.46, + "learning_rate": 1.879789729897454e-05, + "loss": 0.0914, + "step": 11174 + }, + { + "epoch": 1.46, + "learning_rate": 1.8797645189661337e-05, + "loss": 0.076, + "step": 11175 + }, + { + "epoch": 1.46, + "learning_rate": 1.8797393055605228e-05, + "loss": 0.0789, + "step": 11176 + }, + { + "epoch": 1.46, + "learning_rate": 1.8797140896806917e-05, + "loss": 0.0725, + "step": 11177 + }, + { + "epoch": 1.46, + "learning_rate": 1.8796888713267116e-05, + "loss": 0.0807, + "step": 11178 + }, + { + "epoch": 1.46, + "learning_rate": 1.8796636504986536e-05, + "loss": 0.0734, + "step": 11179 + }, + { + "epoch": 1.46, + "learning_rate": 1.8796384271965882e-05, + "loss": 0.0694, + "step": 11180 + }, + { + "epoch": 1.46, + "learning_rate": 1.8796132014205866e-05, + "loss": 0.0725, + "step": 11181 + }, + { + "epoch": 1.46, + "learning_rate": 1.87958797317072e-05, + "loss": 0.0826, + "step": 11182 + }, + { + "epoch": 1.46, + "learning_rate": 1.8795627424470586e-05, + "loss": 0.0824, + "step": 11183 + }, + { + "epoch": 1.47, + "learning_rate": 1.879537509249674e-05, + "loss": 0.0778, + "step": 11184 + }, + { + "epoch": 1.47, + "learning_rate": 1.879512273578637e-05, + "loss": 0.0793, + "step": 11185 + }, + { + "epoch": 1.47, + "learning_rate": 1.8794870354340187e-05, + "loss": 0.078, + "step": 11186 + }, + { + "epoch": 1.47, + "learning_rate": 1.8794617948158897e-05, + "loss": 0.0776, + "step": 11187 + }, + { + "epoch": 1.47, + "learning_rate": 1.8794365517243216e-05, + "loss": 0.0706, + "step": 11188 + }, + { + "epoch": 1.47, + "learning_rate": 1.879411306159385e-05, + "loss": 0.0708, + "step": 11189 + }, + { + "epoch": 1.47, + "learning_rate": 1.8793860581211505e-05, + "loss": 0.0734, + "step": 11190 + }, + { + "epoch": 1.47, + "learning_rate": 1.8793608076096895e-05, + "loss": 0.0886, + "step": 11191 + }, + { + "epoch": 1.47, + "learning_rate": 1.8793355546250737e-05, + "loss": 0.0797, + "step": 11192 + }, + { + "epoch": 1.47, + "learning_rate": 1.879310299167373e-05, + "loss": 0.085, + "step": 11193 + }, + { + "epoch": 1.47, + "learning_rate": 1.879285041236659e-05, + "loss": 0.0821, + "step": 11194 + }, + { + "epoch": 1.47, + "learning_rate": 1.879259780833003e-05, + "loss": 0.0789, + "step": 11195 + }, + { + "epoch": 1.47, + "learning_rate": 1.8792345179564753e-05, + "loss": 0.0761, + "step": 11196 + }, + { + "epoch": 1.47, + "learning_rate": 1.8792092526071473e-05, + "loss": 0.0784, + "step": 11197 + }, + { + "epoch": 1.47, + "learning_rate": 1.8791839847850902e-05, + "loss": 0.0686, + "step": 11198 + }, + { + "epoch": 1.47, + "learning_rate": 1.879158714490375e-05, + "loss": 0.0794, + "step": 11199 + }, + { + "epoch": 1.47, + "learning_rate": 1.8791334417230727e-05, + "loss": 0.0605, + "step": 11200 + }, + { + "epoch": 1.47, + "learning_rate": 1.8791081664832544e-05, + "loss": 0.0811, + "step": 11201 + }, + { + "epoch": 1.47, + "learning_rate": 1.8790828887709912e-05, + "loss": 0.07, + "step": 11202 + }, + { + "epoch": 1.47, + "learning_rate": 1.8790576085863544e-05, + "loss": 0.0819, + "step": 11203 + }, + { + "epoch": 1.47, + "learning_rate": 1.8790323259294148e-05, + "loss": 0.0859, + "step": 11204 + }, + { + "epoch": 1.47, + "learning_rate": 1.8790070408002434e-05, + "loss": 0.0807, + "step": 11205 + }, + { + "epoch": 1.47, + "learning_rate": 1.8789817531989113e-05, + "loss": 0.0756, + "step": 11206 + }, + { + "epoch": 1.47, + "learning_rate": 1.8789564631254904e-05, + "loss": 0.0742, + "step": 11207 + }, + { + "epoch": 1.47, + "learning_rate": 1.8789311705800507e-05, + "loss": 0.079, + "step": 11208 + }, + { + "epoch": 1.47, + "learning_rate": 1.8789058755626645e-05, + "loss": 0.0825, + "step": 11209 + }, + { + "epoch": 1.47, + "learning_rate": 1.8788805780734017e-05, + "loss": 0.072, + "step": 11210 + }, + { + "epoch": 1.47, + "learning_rate": 1.8788552781123345e-05, + "loss": 0.0732, + "step": 11211 + }, + { + "epoch": 1.47, + "learning_rate": 1.8788299756795337e-05, + "loss": 0.0793, + "step": 11212 + }, + { + "epoch": 1.47, + "learning_rate": 1.87880467077507e-05, + "loss": 0.0792, + "step": 11213 + }, + { + "epoch": 1.47, + "learning_rate": 1.8787793633990152e-05, + "loss": 0.0719, + "step": 11214 + }, + { + "epoch": 1.47, + "learning_rate": 1.8787540535514404e-05, + "loss": 0.0714, + "step": 11215 + }, + { + "epoch": 1.47, + "learning_rate": 1.8787287412324163e-05, + "loss": 0.0897, + "step": 11216 + }, + { + "epoch": 1.47, + "learning_rate": 1.8787034264420147e-05, + "loss": 0.0829, + "step": 11217 + }, + { + "epoch": 1.47, + "learning_rate": 1.878678109180306e-05, + "loss": 0.069, + "step": 11218 + }, + { + "epoch": 1.47, + "learning_rate": 1.8786527894473624e-05, + "loss": 0.0822, + "step": 11219 + }, + { + "epoch": 1.47, + "learning_rate": 1.8786274672432544e-05, + "loss": 0.0779, + "step": 11220 + }, + { + "epoch": 1.47, + "learning_rate": 1.8786021425680535e-05, + "loss": 0.0774, + "step": 11221 + }, + { + "epoch": 1.47, + "learning_rate": 1.8785768154218308e-05, + "loss": 0.0706, + "step": 11222 + }, + { + "epoch": 1.47, + "learning_rate": 1.8785514858046578e-05, + "loss": 0.0837, + "step": 11223 + }, + { + "epoch": 1.47, + "learning_rate": 1.8785261537166057e-05, + "loss": 0.0847, + "step": 11224 + }, + { + "epoch": 1.47, + "learning_rate": 1.8785008191577452e-05, + "loss": 0.0727, + "step": 11225 + }, + { + "epoch": 1.47, + "learning_rate": 1.878475482128148e-05, + "loss": 0.0841, + "step": 11226 + }, + { + "epoch": 1.47, + "learning_rate": 1.878450142627885e-05, + "loss": 0.0863, + "step": 11227 + }, + { + "epoch": 1.47, + "learning_rate": 1.8784248006570282e-05, + "loss": 0.0797, + "step": 11228 + }, + { + "epoch": 1.47, + "learning_rate": 1.8783994562156484e-05, + "loss": 0.0819, + "step": 11229 + }, + { + "epoch": 1.47, + "learning_rate": 1.8783741093038167e-05, + "loss": 0.0745, + "step": 11230 + }, + { + "epoch": 1.47, + "learning_rate": 1.878348759921605e-05, + "loss": 0.0829, + "step": 11231 + }, + { + "epoch": 1.47, + "learning_rate": 1.8783234080690835e-05, + "loss": 0.0794, + "step": 11232 + }, + { + "epoch": 1.47, + "learning_rate": 1.8782980537463244e-05, + "loss": 0.0793, + "step": 11233 + }, + { + "epoch": 1.47, + "learning_rate": 1.878272696953399e-05, + "loss": 0.0851, + "step": 11234 + }, + { + "epoch": 1.47, + "learning_rate": 1.8782473376903784e-05, + "loss": 0.0688, + "step": 11235 + }, + { + "epoch": 1.47, + "learning_rate": 1.8782219759573343e-05, + "loss": 0.0887, + "step": 11236 + }, + { + "epoch": 1.47, + "learning_rate": 1.878196611754337e-05, + "loss": 0.0789, + "step": 11237 + }, + { + "epoch": 1.47, + "learning_rate": 1.878171245081459e-05, + "loss": 0.0778, + "step": 11238 + }, + { + "epoch": 1.47, + "learning_rate": 1.8781458759387705e-05, + "loss": 0.0804, + "step": 11239 + }, + { + "epoch": 1.47, + "learning_rate": 1.878120504326344e-05, + "loss": 0.0797, + "step": 11240 + }, + { + "epoch": 1.47, + "learning_rate": 1.8780951302442504e-05, + "loss": 0.0835, + "step": 11241 + }, + { + "epoch": 1.47, + "learning_rate": 1.8780697536925606e-05, + "loss": 0.0711, + "step": 11242 + }, + { + "epoch": 1.47, + "learning_rate": 1.8780443746713467e-05, + "loss": 0.0851, + "step": 11243 + }, + { + "epoch": 1.47, + "learning_rate": 1.8780189931806797e-05, + "loss": 0.0664, + "step": 11244 + }, + { + "epoch": 1.47, + "learning_rate": 1.877993609220631e-05, + "loss": 0.082, + "step": 11245 + }, + { + "epoch": 1.47, + "learning_rate": 1.877968222791272e-05, + "loss": 0.0722, + "step": 11246 + }, + { + "epoch": 1.47, + "learning_rate": 1.877942833892674e-05, + "loss": 0.0642, + "step": 11247 + }, + { + "epoch": 1.47, + "learning_rate": 1.877917442524909e-05, + "loss": 0.0781, + "step": 11248 + }, + { + "epoch": 1.47, + "learning_rate": 1.8778920486880473e-05, + "loss": 0.0927, + "step": 11249 + }, + { + "epoch": 1.47, + "learning_rate": 1.8778666523821614e-05, + "loss": 0.082, + "step": 11250 + }, + { + "epoch": 1.47, + "learning_rate": 1.877841253607322e-05, + "loss": 0.0655, + "step": 11251 + }, + { + "epoch": 1.47, + "learning_rate": 1.8778158523636013e-05, + "loss": 0.0742, + "step": 11252 + }, + { + "epoch": 1.47, + "learning_rate": 1.87779044865107e-05, + "loss": 0.0765, + "step": 11253 + }, + { + "epoch": 1.47, + "learning_rate": 1.8777650424697994e-05, + "loss": 0.0847, + "step": 11254 + }, + { + "epoch": 1.47, + "learning_rate": 1.8777396338198622e-05, + "loss": 0.0778, + "step": 11255 + }, + { + "epoch": 1.47, + "learning_rate": 1.877714222701328e-05, + "loss": 0.0852, + "step": 11256 + }, + { + "epoch": 1.47, + "learning_rate": 1.8776888091142704e-05, + "loss": 0.0867, + "step": 11257 + }, + { + "epoch": 1.47, + "learning_rate": 1.877663393058759e-05, + "loss": 0.0809, + "step": 11258 + }, + { + "epoch": 1.47, + "learning_rate": 1.8776379745348663e-05, + "loss": 0.0975, + "step": 11259 + }, + { + "epoch": 1.47, + "learning_rate": 1.8776125535426636e-05, + "loss": 0.0903, + "step": 11260 + }, + { + "epoch": 1.48, + "learning_rate": 1.877587130082222e-05, + "loss": 0.0865, + "step": 11261 + }, + { + "epoch": 1.48, + "learning_rate": 1.8775617041536138e-05, + "loss": 0.0757, + "step": 11262 + }, + { + "epoch": 1.48, + "learning_rate": 1.8775362757569096e-05, + "loss": 0.0693, + "step": 11263 + }, + { + "epoch": 1.48, + "learning_rate": 1.877510844892182e-05, + "loss": 0.0749, + "step": 11264 + }, + { + "epoch": 1.48, + "learning_rate": 1.877485411559501e-05, + "loss": 0.0812, + "step": 11265 + }, + { + "epoch": 1.48, + "learning_rate": 1.8774599757589397e-05, + "loss": 0.0715, + "step": 11266 + }, + { + "epoch": 1.48, + "learning_rate": 1.8774345374905684e-05, + "loss": 0.0737, + "step": 11267 + }, + { + "epoch": 1.48, + "learning_rate": 1.8774090967544597e-05, + "loss": 0.0695, + "step": 11268 + }, + { + "epoch": 1.48, + "learning_rate": 1.8773836535506845e-05, + "loss": 0.0864, + "step": 11269 + }, + { + "epoch": 1.48, + "learning_rate": 1.8773582078793143e-05, + "loss": 0.0875, + "step": 11270 + }, + { + "epoch": 1.48, + "learning_rate": 1.877332759740421e-05, + "loss": 0.0844, + "step": 11271 + }, + { + "epoch": 1.48, + "learning_rate": 1.8773073091340758e-05, + "loss": 0.0834, + "step": 11272 + }, + { + "epoch": 1.48, + "learning_rate": 1.8772818560603508e-05, + "loss": 0.0897, + "step": 11273 + }, + { + "epoch": 1.48, + "learning_rate": 1.877256400519317e-05, + "loss": 0.0702, + "step": 11274 + }, + { + "epoch": 1.48, + "learning_rate": 1.8772309425110465e-05, + "loss": 0.0705, + "step": 11275 + }, + { + "epoch": 1.48, + "learning_rate": 1.877205482035611e-05, + "loss": 0.0786, + "step": 11276 + }, + { + "epoch": 1.48, + "learning_rate": 1.8771800190930817e-05, + "loss": 0.0918, + "step": 11277 + }, + { + "epoch": 1.48, + "learning_rate": 1.8771545536835297e-05, + "loss": 0.0778, + "step": 11278 + }, + { + "epoch": 1.48, + "learning_rate": 1.8771290858070278e-05, + "loss": 0.0791, + "step": 11279 + }, + { + "epoch": 1.48, + "learning_rate": 1.8771036154636468e-05, + "loss": 0.077, + "step": 11280 + }, + { + "epoch": 1.48, + "learning_rate": 1.8770781426534585e-05, + "loss": 0.0793, + "step": 11281 + }, + { + "epoch": 1.48, + "learning_rate": 1.877052667376535e-05, + "loss": 0.0813, + "step": 11282 + }, + { + "epoch": 1.48, + "learning_rate": 1.8770271896329475e-05, + "loss": 0.0787, + "step": 11283 + }, + { + "epoch": 1.48, + "learning_rate": 1.8770017094227677e-05, + "loss": 0.0815, + "step": 11284 + }, + { + "epoch": 1.48, + "learning_rate": 1.876976226746067e-05, + "loss": 0.0769, + "step": 11285 + }, + { + "epoch": 1.48, + "learning_rate": 1.8769507416029176e-05, + "loss": 0.0785, + "step": 11286 + }, + { + "epoch": 1.48, + "learning_rate": 1.8769252539933907e-05, + "loss": 0.0764, + "step": 11287 + }, + { + "epoch": 1.48, + "learning_rate": 1.8768997639175586e-05, + "loss": 0.0818, + "step": 11288 + }, + { + "epoch": 1.48, + "learning_rate": 1.8768742713754926e-05, + "loss": 0.0912, + "step": 11289 + }, + { + "epoch": 1.48, + "learning_rate": 1.8768487763672643e-05, + "loss": 0.0723, + "step": 11290 + }, + { + "epoch": 1.48, + "learning_rate": 1.8768232788929454e-05, + "loss": 0.0852, + "step": 11291 + }, + { + "epoch": 1.48, + "learning_rate": 1.876797778952608e-05, + "loss": 0.0856, + "step": 11292 + }, + { + "epoch": 1.48, + "learning_rate": 1.876772276546323e-05, + "loss": 0.074, + "step": 11293 + }, + { + "epoch": 1.48, + "learning_rate": 1.876746771674163e-05, + "loss": 0.0846, + "step": 11294 + }, + { + "epoch": 1.48, + "learning_rate": 1.8767212643361996e-05, + "loss": 0.0871, + "step": 11295 + }, + { + "epoch": 1.48, + "learning_rate": 1.876695754532504e-05, + "loss": 0.0804, + "step": 11296 + }, + { + "epoch": 1.48, + "learning_rate": 1.8766702422631483e-05, + "loss": 0.0772, + "step": 11297 + }, + { + "epoch": 1.48, + "learning_rate": 1.8766447275282044e-05, + "loss": 0.0762, + "step": 11298 + }, + { + "epoch": 1.48, + "learning_rate": 1.876619210327744e-05, + "loss": 0.0818, + "step": 11299 + }, + { + "epoch": 1.48, + "learning_rate": 1.8765936906618386e-05, + "loss": 0.0675, + "step": 11300 + }, + { + "epoch": 1.48, + "learning_rate": 1.8765681685305603e-05, + "loss": 0.0792, + "step": 11301 + }, + { + "epoch": 1.48, + "learning_rate": 1.87654264393398e-05, + "loss": 0.0744, + "step": 11302 + }, + { + "epoch": 1.48, + "learning_rate": 1.876517116872171e-05, + "loss": 0.0829, + "step": 11303 + }, + { + "epoch": 1.48, + "learning_rate": 1.876491587345204e-05, + "loss": 0.0766, + "step": 11304 + }, + { + "epoch": 1.48, + "learning_rate": 1.876466055353151e-05, + "loss": 0.074, + "step": 11305 + }, + { + "epoch": 1.48, + "learning_rate": 1.876440520896084e-05, + "loss": 0.0769, + "step": 11306 + }, + { + "epoch": 1.48, + "learning_rate": 1.8764149839740745e-05, + "loss": 0.0838, + "step": 11307 + }, + { + "epoch": 1.48, + "learning_rate": 1.876389444587195e-05, + "loss": 0.083, + "step": 11308 + }, + { + "epoch": 1.48, + "learning_rate": 1.8763639027355164e-05, + "loss": 0.0884, + "step": 11309 + }, + { + "epoch": 1.48, + "learning_rate": 1.876338358419111e-05, + "loss": 0.0734, + "step": 11310 + }, + { + "epoch": 1.48, + "learning_rate": 1.8763128116380507e-05, + "loss": 0.0878, + "step": 11311 + }, + { + "epoch": 1.48, + "learning_rate": 1.8762872623924076e-05, + "loss": 0.0769, + "step": 11312 + }, + { + "epoch": 1.48, + "learning_rate": 1.876261710682253e-05, + "loss": 0.0721, + "step": 11313 + }, + { + "epoch": 1.48, + "learning_rate": 1.876236156507659e-05, + "loss": 0.0685, + "step": 11314 + }, + { + "epoch": 1.48, + "learning_rate": 1.8762105998686974e-05, + "loss": 0.0711, + "step": 11315 + }, + { + "epoch": 1.48, + "learning_rate": 1.8761850407654403e-05, + "loss": 0.0725, + "step": 11316 + }, + { + "epoch": 1.48, + "learning_rate": 1.8761594791979594e-05, + "loss": 0.0769, + "step": 11317 + }, + { + "epoch": 1.48, + "learning_rate": 1.8761339151663268e-05, + "loss": 0.0798, + "step": 11318 + }, + { + "epoch": 1.48, + "learning_rate": 1.8761083486706137e-05, + "loss": 0.0736, + "step": 11319 + }, + { + "epoch": 1.48, + "learning_rate": 1.876082779710893e-05, + "loss": 0.078, + "step": 11320 + }, + { + "epoch": 1.48, + "learning_rate": 1.876057208287236e-05, + "loss": 0.0776, + "step": 11321 + }, + { + "epoch": 1.48, + "learning_rate": 1.8760316343997144e-05, + "loss": 0.0693, + "step": 11322 + }, + { + "epoch": 1.48, + "learning_rate": 1.876006058048401e-05, + "loss": 0.0676, + "step": 11323 + }, + { + "epoch": 1.48, + "learning_rate": 1.875980479233367e-05, + "loss": 0.0727, + "step": 11324 + }, + { + "epoch": 1.48, + "learning_rate": 1.8759548979546847e-05, + "loss": 0.0642, + "step": 11325 + }, + { + "epoch": 1.48, + "learning_rate": 1.8759293142124257e-05, + "loss": 0.0699, + "step": 11326 + }, + { + "epoch": 1.48, + "learning_rate": 1.8759037280066622e-05, + "loss": 0.0683, + "step": 11327 + }, + { + "epoch": 1.48, + "learning_rate": 1.8758781393374663e-05, + "loss": 0.0745, + "step": 11328 + }, + { + "epoch": 1.48, + "learning_rate": 1.875852548204909e-05, + "loss": 0.0892, + "step": 11329 + }, + { + "epoch": 1.48, + "learning_rate": 1.875826954609064e-05, + "loss": 0.0864, + "step": 11330 + }, + { + "epoch": 1.48, + "learning_rate": 1.875801358550002e-05, + "loss": 0.0741, + "step": 11331 + }, + { + "epoch": 1.48, + "learning_rate": 1.8757757600277953e-05, + "loss": 0.0834, + "step": 11332 + }, + { + "epoch": 1.48, + "learning_rate": 1.8757501590425157e-05, + "loss": 0.0806, + "step": 11333 + }, + { + "epoch": 1.48, + "learning_rate": 1.8757245555942357e-05, + "loss": 0.0816, + "step": 11334 + }, + { + "epoch": 1.48, + "learning_rate": 1.8756989496830272e-05, + "loss": 0.0744, + "step": 11335 + }, + { + "epoch": 1.48, + "learning_rate": 1.8756733413089616e-05, + "loss": 0.0834, + "step": 11336 + }, + { + "epoch": 1.49, + "learning_rate": 1.8756477304721115e-05, + "loss": 0.0751, + "step": 11337 + }, + { + "epoch": 1.49, + "learning_rate": 1.875622117172549e-05, + "loss": 0.0747, + "step": 11338 + }, + { + "epoch": 1.49, + "learning_rate": 1.8755965014103457e-05, + "loss": 0.0798, + "step": 11339 + }, + { + "epoch": 1.49, + "learning_rate": 1.875570883185574e-05, + "loss": 0.0741, + "step": 11340 + }, + { + "epoch": 1.49, + "learning_rate": 1.8755452624983056e-05, + "loss": 0.0904, + "step": 11341 + }, + { + "epoch": 1.49, + "learning_rate": 1.875519639348613e-05, + "loss": 0.0768, + "step": 11342 + }, + { + "epoch": 1.49, + "learning_rate": 1.875494013736568e-05, + "loss": 0.0702, + "step": 11343 + }, + { + "epoch": 1.49, + "learning_rate": 1.8754683856622426e-05, + "loss": 0.0723, + "step": 11344 + }, + { + "epoch": 1.49, + "learning_rate": 1.875442755125709e-05, + "loss": 0.0653, + "step": 11345 + }, + { + "epoch": 1.49, + "learning_rate": 1.8754171221270395e-05, + "loss": 0.0793, + "step": 11346 + }, + { + "epoch": 1.49, + "learning_rate": 1.8753914866663056e-05, + "loss": 0.0922, + "step": 11347 + }, + { + "epoch": 1.49, + "learning_rate": 1.87536584874358e-05, + "loss": 0.0706, + "step": 11348 + }, + { + "epoch": 1.49, + "learning_rate": 1.8753402083589346e-05, + "loss": 0.0615, + "step": 11349 + }, + { + "epoch": 1.49, + "learning_rate": 1.8753145655124414e-05, + "loss": 0.0799, + "step": 11350 + }, + { + "epoch": 1.49, + "learning_rate": 1.8752889202041727e-05, + "loss": 0.0748, + "step": 11351 + }, + { + "epoch": 1.49, + "learning_rate": 1.8752632724342002e-05, + "loss": 0.0757, + "step": 11352 + }, + { + "epoch": 1.49, + "learning_rate": 1.8752376222025965e-05, + "loss": 0.0755, + "step": 11353 + }, + { + "epoch": 1.49, + "learning_rate": 1.875211969509434e-05, + "loss": 0.0881, + "step": 11354 + }, + { + "epoch": 1.49, + "learning_rate": 1.875186314354784e-05, + "loss": 0.0907, + "step": 11355 + }, + { + "epoch": 1.49, + "learning_rate": 1.8751606567387193e-05, + "loss": 0.0875, + "step": 11356 + }, + { + "epoch": 1.49, + "learning_rate": 1.8751349966613113e-05, + "loss": 0.0887, + "step": 11357 + }, + { + "epoch": 1.49, + "learning_rate": 1.8751093341226333e-05, + "loss": 0.0733, + "step": 11358 + }, + { + "epoch": 1.49, + "learning_rate": 1.875083669122757e-05, + "loss": 0.08, + "step": 11359 + }, + { + "epoch": 1.49, + "learning_rate": 1.875058001661754e-05, + "loss": 0.0664, + "step": 11360 + }, + { + "epoch": 1.49, + "learning_rate": 1.8750323317396976e-05, + "loss": 0.0762, + "step": 11361 + }, + { + "epoch": 1.49, + "learning_rate": 1.875006659356659e-05, + "loss": 0.0808, + "step": 11362 + }, + { + "epoch": 1.49, + "learning_rate": 1.874980984512711e-05, + "loss": 0.0953, + "step": 11363 + }, + { + "epoch": 1.49, + "learning_rate": 1.874955307207925e-05, + "loss": 0.0796, + "step": 11364 + }, + { + "epoch": 1.49, + "learning_rate": 1.8749296274423743e-05, + "loss": 0.0737, + "step": 11365 + }, + { + "epoch": 1.49, + "learning_rate": 1.8749039452161305e-05, + "loss": 0.0833, + "step": 11366 + }, + { + "epoch": 1.49, + "learning_rate": 1.8748782605292658e-05, + "loss": 0.0834, + "step": 11367 + }, + { + "epoch": 1.49, + "learning_rate": 1.8748525733818527e-05, + "loss": 0.0761, + "step": 11368 + }, + { + "epoch": 1.49, + "learning_rate": 1.874826883773963e-05, + "loss": 0.0798, + "step": 11369 + }, + { + "epoch": 1.49, + "learning_rate": 1.8748011917056698e-05, + "loss": 0.0828, + "step": 11370 + }, + { + "epoch": 1.49, + "learning_rate": 1.8747754971770446e-05, + "loss": 0.0767, + "step": 11371 + }, + { + "epoch": 1.49, + "learning_rate": 1.87474980018816e-05, + "loss": 0.0707, + "step": 11372 + }, + { + "epoch": 1.49, + "learning_rate": 1.8747241007390882e-05, + "loss": 0.0748, + "step": 11373 + }, + { + "epoch": 1.49, + "learning_rate": 1.8746983988299014e-05, + "loss": 0.0791, + "step": 11374 + }, + { + "epoch": 1.49, + "learning_rate": 1.8746726944606717e-05, + "loss": 0.0688, + "step": 11375 + }, + { + "epoch": 1.49, + "learning_rate": 1.874646987631472e-05, + "loss": 0.0703, + "step": 11376 + }, + { + "epoch": 1.49, + "learning_rate": 1.8746212783423737e-05, + "loss": 0.0794, + "step": 11377 + }, + { + "epoch": 1.49, + "learning_rate": 1.87459556659345e-05, + "loss": 0.0849, + "step": 11378 + }, + { + "epoch": 1.49, + "learning_rate": 1.874569852384773e-05, + "loss": 0.0796, + "step": 11379 + }, + { + "epoch": 1.49, + "learning_rate": 1.8745441357164145e-05, + "loss": 0.0858, + "step": 11380 + }, + { + "epoch": 1.49, + "learning_rate": 1.8745184165884475e-05, + "loss": 0.0682, + "step": 11381 + }, + { + "epoch": 1.49, + "learning_rate": 1.8744926950009438e-05, + "loss": 0.0804, + "step": 11382 + }, + { + "epoch": 1.49, + "learning_rate": 1.874466970953976e-05, + "loss": 0.0783, + "step": 11383 + }, + { + "epoch": 1.49, + "learning_rate": 1.8744412444476164e-05, + "loss": 0.0686, + "step": 11384 + }, + { + "epoch": 1.49, + "learning_rate": 1.8744155154819377e-05, + "loss": 0.0801, + "step": 11385 + }, + { + "epoch": 1.49, + "learning_rate": 1.8743897840570117e-05, + "loss": 0.0851, + "step": 11386 + }, + { + "epoch": 1.49, + "learning_rate": 1.8743640501729108e-05, + "loss": 0.0777, + "step": 11387 + }, + { + "epoch": 1.49, + "learning_rate": 1.8743383138297076e-05, + "loss": 0.0806, + "step": 11388 + }, + { + "epoch": 1.49, + "learning_rate": 1.874312575027475e-05, + "loss": 0.0907, + "step": 11389 + }, + { + "epoch": 1.49, + "learning_rate": 1.8742868337662843e-05, + "loss": 0.0679, + "step": 11390 + }, + { + "epoch": 1.49, + "learning_rate": 1.8742610900462084e-05, + "loss": 0.0834, + "step": 11391 + }, + { + "epoch": 1.49, + "learning_rate": 1.8742353438673197e-05, + "loss": 0.0802, + "step": 11392 + }, + { + "epoch": 1.49, + "learning_rate": 1.874209595229691e-05, + "loss": 0.0783, + "step": 11393 + }, + { + "epoch": 1.49, + "learning_rate": 1.8741838441333945e-05, + "loss": 0.0859, + "step": 11394 + }, + { + "epoch": 1.49, + "learning_rate": 1.8741580905785018e-05, + "loss": 0.0766, + "step": 11395 + }, + { + "epoch": 1.49, + "learning_rate": 1.8741323345650865e-05, + "loss": 0.0816, + "step": 11396 + }, + { + "epoch": 1.49, + "learning_rate": 1.8741065760932208e-05, + "loss": 0.0801, + "step": 11397 + }, + { + "epoch": 1.49, + "learning_rate": 1.8740808151629764e-05, + "loss": 0.0685, + "step": 11398 + }, + { + "epoch": 1.49, + "learning_rate": 1.8740550517744266e-05, + "loss": 0.085, + "step": 11399 + }, + { + "epoch": 1.49, + "learning_rate": 1.874029285927643e-05, + "loss": 0.0643, + "step": 11400 + }, + { + "epoch": 1.49, + "learning_rate": 1.8740035176226993e-05, + "loss": 0.0912, + "step": 11401 + }, + { + "epoch": 1.49, + "learning_rate": 1.873977746859667e-05, + "loss": 0.0844, + "step": 11402 + }, + { + "epoch": 1.49, + "learning_rate": 1.8739519736386186e-05, + "loss": 0.0755, + "step": 11403 + }, + { + "epoch": 1.49, + "learning_rate": 1.8739261979596265e-05, + "loss": 0.0857, + "step": 11404 + }, + { + "epoch": 1.49, + "learning_rate": 1.873900419822764e-05, + "loss": 0.0732, + "step": 11405 + }, + { + "epoch": 1.49, + "learning_rate": 1.8738746392281032e-05, + "loss": 0.076, + "step": 11406 + }, + { + "epoch": 1.49, + "learning_rate": 1.8738488561757164e-05, + "loss": 0.0834, + "step": 11407 + }, + { + "epoch": 1.49, + "learning_rate": 1.8738230706656764e-05, + "loss": 0.0763, + "step": 11408 + }, + { + "epoch": 1.49, + "learning_rate": 1.873797282698055e-05, + "loss": 0.0834, + "step": 11409 + }, + { + "epoch": 1.49, + "learning_rate": 1.8737714922729256e-05, + "loss": 0.0724, + "step": 11410 + }, + { + "epoch": 1.49, + "learning_rate": 1.8737456993903603e-05, + "loss": 0.0789, + "step": 11411 + }, + { + "epoch": 1.49, + "learning_rate": 1.873719904050432e-05, + "loss": 0.0774, + "step": 11412 + }, + { + "epoch": 1.5, + "learning_rate": 1.8736941062532132e-05, + "loss": 0.0844, + "step": 11413 + }, + { + "epoch": 1.5, + "learning_rate": 1.8736683059987757e-05, + "loss": 0.0817, + "step": 11414 + }, + { + "epoch": 1.5, + "learning_rate": 1.8736425032871926e-05, + "loss": 0.0804, + "step": 11415 + }, + { + "epoch": 1.5, + "learning_rate": 1.873616698118537e-05, + "loss": 0.0733, + "step": 11416 + }, + { + "epoch": 1.5, + "learning_rate": 1.8735908904928804e-05, + "loss": 0.0755, + "step": 11417 + }, + { + "epoch": 1.5, + "learning_rate": 1.8735650804102964e-05, + "loss": 0.0773, + "step": 11418 + }, + { + "epoch": 1.5, + "learning_rate": 1.873539267870857e-05, + "loss": 0.0793, + "step": 11419 + }, + { + "epoch": 1.5, + "learning_rate": 1.8735134528746347e-05, + "loss": 0.0826, + "step": 11420 + }, + { + "epoch": 1.5, + "learning_rate": 1.873487635421703e-05, + "loss": 0.0749, + "step": 11421 + }, + { + "epoch": 1.5, + "learning_rate": 1.8734618155121332e-05, + "loss": 0.075, + "step": 11422 + }, + { + "epoch": 1.5, + "learning_rate": 1.8734359931459986e-05, + "loss": 0.0915, + "step": 11423 + }, + { + "epoch": 1.5, + "learning_rate": 1.873410168323372e-05, + "loss": 0.0775, + "step": 11424 + }, + { + "epoch": 1.5, + "learning_rate": 1.8733843410443256e-05, + "loss": 0.0726, + "step": 11425 + }, + { + "epoch": 1.5, + "learning_rate": 1.8733585113089322e-05, + "loss": 0.0772, + "step": 11426 + }, + { + "epoch": 1.5, + "learning_rate": 1.873332679117265e-05, + "loss": 0.0729, + "step": 11427 + }, + { + "epoch": 1.5, + "learning_rate": 1.873306844469396e-05, + "loss": 0.0879, + "step": 11428 + }, + { + "epoch": 1.5, + "learning_rate": 1.8732810073653972e-05, + "loss": 0.0787, + "step": 11429 + }, + { + "epoch": 1.5, + "learning_rate": 1.873255167805343e-05, + "loss": 0.0742, + "step": 11430 + }, + { + "epoch": 1.5, + "learning_rate": 1.873229325789305e-05, + "loss": 0.0812, + "step": 11431 + }, + { + "epoch": 1.5, + "learning_rate": 1.8732034813173557e-05, + "loss": 0.0766, + "step": 11432 + }, + { + "epoch": 1.5, + "learning_rate": 1.873177634389568e-05, + "loss": 0.081, + "step": 11433 + }, + { + "epoch": 1.5, + "learning_rate": 1.873151785006015e-05, + "loss": 0.0785, + "step": 11434 + }, + { + "epoch": 1.5, + "learning_rate": 1.8731259331667693e-05, + "loss": 0.0766, + "step": 11435 + }, + { + "epoch": 1.5, + "learning_rate": 1.8731000788719033e-05, + "loss": 0.0796, + "step": 11436 + }, + { + "epoch": 1.5, + "learning_rate": 1.8730742221214894e-05, + "loss": 0.0724, + "step": 11437 + }, + { + "epoch": 1.5, + "learning_rate": 1.8730483629156012e-05, + "loss": 0.0818, + "step": 11438 + }, + { + "epoch": 1.5, + "learning_rate": 1.873022501254311e-05, + "loss": 0.0871, + "step": 11439 + }, + { + "epoch": 1.5, + "learning_rate": 1.872996637137691e-05, + "loss": 0.0764, + "step": 11440 + }, + { + "epoch": 1.5, + "learning_rate": 1.8729707705658148e-05, + "loss": 0.072, + "step": 11441 + }, + { + "epoch": 1.5, + "learning_rate": 1.8729449015387547e-05, + "loss": 0.0852, + "step": 11442 + }, + { + "epoch": 1.5, + "learning_rate": 1.8729190300565836e-05, + "loss": 0.0704, + "step": 11443 + }, + { + "epoch": 1.5, + "learning_rate": 1.8728931561193745e-05, + "loss": 0.0723, + "step": 11444 + }, + { + "epoch": 1.5, + "learning_rate": 1.8728672797271994e-05, + "loss": 0.0803, + "step": 11445 + }, + { + "epoch": 1.5, + "learning_rate": 1.8728414008801316e-05, + "loss": 0.0741, + "step": 11446 + }, + { + "epoch": 1.5, + "learning_rate": 1.872815519578244e-05, + "loss": 0.0827, + "step": 11447 + }, + { + "epoch": 1.5, + "learning_rate": 1.8727896358216095e-05, + "loss": 0.0771, + "step": 11448 + }, + { + "epoch": 1.5, + "learning_rate": 1.8727637496103e-05, + "loss": 0.0751, + "step": 11449 + }, + { + "epoch": 1.5, + "learning_rate": 1.8727378609443895e-05, + "loss": 0.0849, + "step": 11450 + }, + { + "epoch": 1.5, + "learning_rate": 1.8727119698239502e-05, + "loss": 0.0788, + "step": 11451 + }, + { + "epoch": 1.5, + "learning_rate": 1.8726860762490547e-05, + "loss": 0.0701, + "step": 11452 + }, + { + "epoch": 1.5, + "learning_rate": 1.8726601802197763e-05, + "loss": 0.0809, + "step": 11453 + }, + { + "epoch": 1.5, + "learning_rate": 1.8726342817361877e-05, + "loss": 0.0828, + "step": 11454 + }, + { + "epoch": 1.5, + "learning_rate": 1.8726083807983614e-05, + "loss": 0.0745, + "step": 11455 + }, + { + "epoch": 1.5, + "learning_rate": 1.8725824774063703e-05, + "loss": 0.0679, + "step": 11456 + }, + { + "epoch": 1.5, + "learning_rate": 1.8725565715602877e-05, + "loss": 0.0818, + "step": 11457 + }, + { + "epoch": 1.5, + "learning_rate": 1.8725306632601862e-05, + "loss": 0.0658, + "step": 11458 + }, + { + "epoch": 1.5, + "learning_rate": 1.872504752506139e-05, + "loss": 0.0654, + "step": 11459 + }, + { + "epoch": 1.5, + "learning_rate": 1.872478839298218e-05, + "loss": 0.0657, + "step": 11460 + }, + { + "epoch": 1.5, + "learning_rate": 1.8724529236364974e-05, + "loss": 0.079, + "step": 11461 + }, + { + "epoch": 1.5, + "learning_rate": 1.8724270055210486e-05, + "loss": 0.0931, + "step": 11462 + }, + { + "epoch": 1.5, + "learning_rate": 1.872401084951946e-05, + "loss": 0.0595, + "step": 11463 + }, + { + "epoch": 1.5, + "learning_rate": 1.8723751619292616e-05, + "loss": 0.0777, + "step": 11464 + }, + { + "epoch": 1.5, + "learning_rate": 1.8723492364530688e-05, + "loss": 0.0662, + "step": 11465 + }, + { + "epoch": 1.5, + "learning_rate": 1.8723233085234395e-05, + "loss": 0.0844, + "step": 11466 + }, + { + "epoch": 1.5, + "learning_rate": 1.872297378140448e-05, + "loss": 0.0746, + "step": 11467 + }, + { + "epoch": 1.5, + "learning_rate": 1.872271445304166e-05, + "loss": 0.0779, + "step": 11468 + }, + { + "epoch": 1.5, + "learning_rate": 1.8722455100146677e-05, + "loss": 0.076, + "step": 11469 + }, + { + "epoch": 1.5, + "learning_rate": 1.872219572272025e-05, + "loss": 0.0806, + "step": 11470 + }, + { + "epoch": 1.5, + "learning_rate": 1.8721936320763113e-05, + "loss": 0.0712, + "step": 11471 + }, + { + "epoch": 1.5, + "learning_rate": 1.872167689427599e-05, + "loss": 0.083, + "step": 11472 + }, + { + "epoch": 1.5, + "learning_rate": 1.8721417443259622e-05, + "loss": 0.092, + "step": 11473 + }, + { + "epoch": 1.5, + "learning_rate": 1.872115796771473e-05, + "loss": 0.0754, + "step": 11474 + }, + { + "epoch": 1.5, + "learning_rate": 1.8720898467642042e-05, + "loss": 0.0729, + "step": 11475 + }, + { + "epoch": 1.5, + "learning_rate": 1.8720638943042294e-05, + "loss": 0.0797, + "step": 11476 + }, + { + "epoch": 1.5, + "learning_rate": 1.8720379393916217e-05, + "loss": 0.0799, + "step": 11477 + }, + { + "epoch": 1.5, + "learning_rate": 1.872011982026453e-05, + "loss": 0.0672, + "step": 11478 + }, + { + "epoch": 1.5, + "learning_rate": 1.8719860222087977e-05, + "loss": 0.0773, + "step": 11479 + }, + { + "epoch": 1.5, + "learning_rate": 1.8719600599387276e-05, + "loss": 0.0727, + "step": 11480 + }, + { + "epoch": 1.5, + "learning_rate": 1.8719340952163165e-05, + "loss": 0.0922, + "step": 11481 + }, + { + "epoch": 1.5, + "learning_rate": 1.8719081280416373e-05, + "loss": 0.0853, + "step": 11482 + }, + { + "epoch": 1.5, + "learning_rate": 1.8718821584147627e-05, + "loss": 0.0745, + "step": 11483 + }, + { + "epoch": 1.5, + "learning_rate": 1.871856186335766e-05, + "loss": 0.0831, + "step": 11484 + }, + { + "epoch": 1.5, + "learning_rate": 1.8718302118047206e-05, + "loss": 0.0782, + "step": 11485 + }, + { + "epoch": 1.5, + "learning_rate": 1.8718042348216987e-05, + "loss": 0.0771, + "step": 11486 + }, + { + "epoch": 1.5, + "learning_rate": 1.871778255386774e-05, + "loss": 0.0777, + "step": 11487 + }, + { + "epoch": 1.5, + "learning_rate": 1.8717522735000194e-05, + "loss": 0.0676, + "step": 11488 + }, + { + "epoch": 1.5, + "learning_rate": 1.871726289161508e-05, + "loss": 0.0705, + "step": 11489 + }, + { + "epoch": 1.51, + "learning_rate": 1.8717003023713127e-05, + "loss": 0.0726, + "step": 11490 + }, + { + "epoch": 1.51, + "learning_rate": 1.8716743131295065e-05, + "loss": 0.0869, + "step": 11491 + }, + { + "epoch": 1.51, + "learning_rate": 1.8716483214361633e-05, + "loss": 0.0824, + "step": 11492 + }, + { + "epoch": 1.51, + "learning_rate": 1.8716223272913553e-05, + "loss": 0.0784, + "step": 11493 + }, + { + "epoch": 1.51, + "learning_rate": 1.8715963306951557e-05, + "loss": 0.0715, + "step": 11494 + }, + { + "epoch": 1.51, + "learning_rate": 1.871570331647638e-05, + "loss": 0.0726, + "step": 11495 + }, + { + "epoch": 1.51, + "learning_rate": 1.8715443301488755e-05, + "loss": 0.0688, + "step": 11496 + }, + { + "epoch": 1.51, + "learning_rate": 1.8715183261989405e-05, + "loss": 0.0752, + "step": 11497 + }, + { + "epoch": 1.51, + "learning_rate": 1.871492319797907e-05, + "loss": 0.0713, + "step": 11498 + }, + { + "epoch": 1.51, + "learning_rate": 1.8714663109458474e-05, + "loss": 0.071, + "step": 11499 + }, + { + "epoch": 1.51, + "learning_rate": 1.8714402996428353e-05, + "loss": 0.0799, + "step": 11500 + }, + { + "epoch": 1.51, + "learning_rate": 1.871414285888944e-05, + "loss": 0.074, + "step": 11501 + }, + { + "epoch": 1.51, + "learning_rate": 1.8713882696842464e-05, + "loss": 0.0862, + "step": 11502 + }, + { + "epoch": 1.51, + "learning_rate": 1.8713622510288152e-05, + "loss": 0.0809, + "step": 11503 + }, + { + "epoch": 1.51, + "learning_rate": 1.8713362299227245e-05, + "loss": 0.0831, + "step": 11504 + }, + { + "epoch": 1.51, + "learning_rate": 1.871310206366047e-05, + "loss": 0.0782, + "step": 11505 + }, + { + "epoch": 1.51, + "learning_rate": 1.871284180358856e-05, + "loss": 0.0695, + "step": 11506 + }, + { + "epoch": 1.51, + "learning_rate": 1.8712581519012244e-05, + "loss": 0.0736, + "step": 11507 + }, + { + "epoch": 1.51, + "learning_rate": 1.871232120993226e-05, + "loss": 0.0785, + "step": 11508 + }, + { + "epoch": 1.51, + "learning_rate": 1.8712060876349334e-05, + "loss": 0.0858, + "step": 11509 + }, + { + "epoch": 1.51, + "learning_rate": 1.87118005182642e-05, + "loss": 0.0862, + "step": 11510 + }, + { + "epoch": 1.51, + "learning_rate": 1.8711540135677593e-05, + "loss": 0.0673, + "step": 11511 + }, + { + "epoch": 1.51, + "learning_rate": 1.8711279728590244e-05, + "loss": 0.0797, + "step": 11512 + }, + { + "epoch": 1.51, + "learning_rate": 1.8711019297002883e-05, + "loss": 0.0816, + "step": 11513 + }, + { + "epoch": 1.51, + "learning_rate": 1.8710758840916243e-05, + "loss": 0.0704, + "step": 11514 + }, + { + "epoch": 1.51, + "learning_rate": 1.8710498360331063e-05, + "loss": 0.0784, + "step": 11515 + }, + { + "epoch": 1.51, + "learning_rate": 1.8710237855248063e-05, + "loss": 0.0657, + "step": 11516 + }, + { + "epoch": 1.51, + "learning_rate": 1.8709977325667986e-05, + "loss": 0.0727, + "step": 11517 + }, + { + "epoch": 1.51, + "learning_rate": 1.8709716771591563e-05, + "loss": 0.0967, + "step": 11518 + }, + { + "epoch": 1.51, + "learning_rate": 1.8709456193019524e-05, + "loss": 0.0858, + "step": 11519 + }, + { + "epoch": 1.51, + "learning_rate": 1.8709195589952603e-05, + "loss": 0.077, + "step": 11520 + }, + { + "epoch": 1.51, + "learning_rate": 1.8708934962391536e-05, + "loss": 0.0688, + "step": 11521 + }, + { + "epoch": 1.51, + "learning_rate": 1.870867431033705e-05, + "loss": 0.0709, + "step": 11522 + }, + { + "epoch": 1.51, + "learning_rate": 1.8708413633789883e-05, + "loss": 0.0836, + "step": 11523 + }, + { + "epoch": 1.51, + "learning_rate": 1.8708152932750764e-05, + "loss": 0.0832, + "step": 11524 + }, + { + "epoch": 1.51, + "learning_rate": 1.870789220722043e-05, + "loss": 0.0874, + "step": 11525 + }, + { + "epoch": 1.51, + "learning_rate": 1.8707631457199615e-05, + "loss": 0.0756, + "step": 11526 + }, + { + "epoch": 1.51, + "learning_rate": 1.8707370682689044e-05, + "loss": 0.0743, + "step": 11527 + }, + { + "epoch": 1.51, + "learning_rate": 1.8707109883689464e-05, + "loss": 0.0758, + "step": 11528 + }, + { + "epoch": 1.51, + "learning_rate": 1.87068490602016e-05, + "loss": 0.0791, + "step": 11529 + }, + { + "epoch": 1.51, + "learning_rate": 1.8706588212226184e-05, + "loss": 0.0743, + "step": 11530 + }, + { + "epoch": 1.51, + "learning_rate": 1.870632733976395e-05, + "loss": 0.0783, + "step": 11531 + }, + { + "epoch": 1.51, + "learning_rate": 1.870606644281564e-05, + "loss": 0.0778, + "step": 11532 + }, + { + "epoch": 1.51, + "learning_rate": 1.870580552138198e-05, + "loss": 0.0756, + "step": 11533 + }, + { + "epoch": 1.51, + "learning_rate": 1.87055445754637e-05, + "loss": 0.0821, + "step": 11534 + }, + { + "epoch": 1.51, + "learning_rate": 1.8705283605061545e-05, + "loss": 0.0886, + "step": 11535 + }, + { + "epoch": 1.51, + "learning_rate": 1.870502261017624e-05, + "loss": 0.0793, + "step": 11536 + }, + { + "epoch": 1.51, + "learning_rate": 1.8704761590808528e-05, + "loss": 0.0819, + "step": 11537 + }, + { + "epoch": 1.51, + "learning_rate": 1.8704500546959136e-05, + "loss": 0.0776, + "step": 11538 + }, + { + "epoch": 1.51, + "learning_rate": 1.8704239478628797e-05, + "loss": 0.0812, + "step": 11539 + }, + { + "epoch": 1.51, + "learning_rate": 1.8703978385818248e-05, + "loss": 0.0716, + "step": 11540 + }, + { + "epoch": 1.51, + "learning_rate": 1.8703717268528226e-05, + "loss": 0.0625, + "step": 11541 + }, + { + "epoch": 1.51, + "learning_rate": 1.870345612675946e-05, + "loss": 0.0713, + "step": 11542 + }, + { + "epoch": 1.51, + "learning_rate": 1.870319496051269e-05, + "loss": 0.0793, + "step": 11543 + }, + { + "epoch": 1.51, + "learning_rate": 1.8702933769788645e-05, + "loss": 0.0768, + "step": 11544 + }, + { + "epoch": 1.51, + "learning_rate": 1.8702672554588063e-05, + "loss": 0.0622, + "step": 11545 + }, + { + "epoch": 1.51, + "learning_rate": 1.8702411314911677e-05, + "loss": 0.0777, + "step": 11546 + }, + { + "epoch": 1.51, + "learning_rate": 1.8702150050760223e-05, + "loss": 0.0745, + "step": 11547 + }, + { + "epoch": 1.51, + "learning_rate": 1.8701888762134437e-05, + "loss": 0.0863, + "step": 11548 + }, + { + "epoch": 1.51, + "learning_rate": 1.870162744903505e-05, + "loss": 0.0782, + "step": 11549 + }, + { + "epoch": 1.51, + "learning_rate": 1.87013661114628e-05, + "loss": 0.0833, + "step": 11550 + }, + { + "epoch": 1.51, + "learning_rate": 1.8701104749418425e-05, + "loss": 0.0755, + "step": 11551 + }, + { + "epoch": 1.51, + "learning_rate": 1.870084336290265e-05, + "loss": 0.0836, + "step": 11552 + }, + { + "epoch": 1.51, + "learning_rate": 1.8700581951916222e-05, + "loss": 0.0712, + "step": 11553 + }, + { + "epoch": 1.51, + "learning_rate": 1.8700320516459867e-05, + "loss": 0.0813, + "step": 11554 + }, + { + "epoch": 1.51, + "learning_rate": 1.870005905653432e-05, + "loss": 0.0708, + "step": 11555 + }, + { + "epoch": 1.51, + "learning_rate": 1.869979757214033e-05, + "loss": 0.0701, + "step": 11556 + }, + { + "epoch": 1.51, + "learning_rate": 1.8699536063278617e-05, + "loss": 0.07, + "step": 11557 + }, + { + "epoch": 1.51, + "learning_rate": 1.8699274529949922e-05, + "loss": 0.0876, + "step": 11558 + }, + { + "epoch": 1.51, + "learning_rate": 1.8699012972154982e-05, + "loss": 0.081, + "step": 11559 + }, + { + "epoch": 1.51, + "learning_rate": 1.8698751389894528e-05, + "loss": 0.0767, + "step": 11560 + }, + { + "epoch": 1.51, + "learning_rate": 1.8698489783169306e-05, + "loss": 0.0757, + "step": 11561 + }, + { + "epoch": 1.51, + "learning_rate": 1.869822815198004e-05, + "loss": 0.0749, + "step": 11562 + }, + { + "epoch": 1.51, + "learning_rate": 1.869796649632747e-05, + "loss": 0.0765, + "step": 11563 + }, + { + "epoch": 1.51, + "learning_rate": 1.8697704816212332e-05, + "loss": 0.0835, + "step": 11564 + }, + { + "epoch": 1.51, + "learning_rate": 1.8697443111635364e-05, + "loss": 0.0787, + "step": 11565 + }, + { + "epoch": 1.52, + "learning_rate": 1.8697181382597297e-05, + "loss": 0.0761, + "step": 11566 + }, + { + "epoch": 1.52, + "learning_rate": 1.8696919629098874e-05, + "loss": 0.0741, + "step": 11567 + }, + { + "epoch": 1.52, + "learning_rate": 1.869665785114083e-05, + "loss": 0.076, + "step": 11568 + }, + { + "epoch": 1.52, + "learning_rate": 1.8696396048723895e-05, + "loss": 0.0822, + "step": 11569 + }, + { + "epoch": 1.52, + "learning_rate": 1.8696134221848813e-05, + "loss": 0.0778, + "step": 11570 + }, + { + "epoch": 1.52, + "learning_rate": 1.869587237051631e-05, + "loss": 0.0836, + "step": 11571 + }, + { + "epoch": 1.52, + "learning_rate": 1.8695610494727135e-05, + "loss": 0.0805, + "step": 11572 + }, + { + "epoch": 1.52, + "learning_rate": 1.8695348594482015e-05, + "loss": 0.0825, + "step": 11573 + }, + { + "epoch": 1.52, + "learning_rate": 1.8695086669781696e-05, + "loss": 0.0807, + "step": 11574 + }, + { + "epoch": 1.52, + "learning_rate": 1.8694824720626902e-05, + "loss": 0.0902, + "step": 11575 + }, + { + "epoch": 1.52, + "learning_rate": 1.8694562747018378e-05, + "loss": 0.0826, + "step": 11576 + }, + { + "epoch": 1.52, + "learning_rate": 1.869430074895686e-05, + "loss": 0.0785, + "step": 11577 + }, + { + "epoch": 1.52, + "learning_rate": 1.8694038726443087e-05, + "loss": 0.0802, + "step": 11578 + }, + { + "epoch": 1.52, + "learning_rate": 1.8693776679477787e-05, + "loss": 0.0759, + "step": 11579 + }, + { + "epoch": 1.52, + "learning_rate": 1.869351460806171e-05, + "loss": 0.0753, + "step": 11580 + }, + { + "epoch": 1.52, + "learning_rate": 1.869325251219558e-05, + "loss": 0.088, + "step": 11581 + }, + { + "epoch": 1.52, + "learning_rate": 1.8692990391880143e-05, + "loss": 0.0663, + "step": 11582 + }, + { + "epoch": 1.52, + "learning_rate": 1.8692728247116134e-05, + "loss": 0.0759, + "step": 11583 + }, + { + "epoch": 1.52, + "learning_rate": 1.8692466077904288e-05, + "loss": 0.078, + "step": 11584 + }, + { + "epoch": 1.52, + "learning_rate": 1.869220388424535e-05, + "loss": 0.0692, + "step": 11585 + }, + { + "epoch": 1.52, + "learning_rate": 1.8691941666140044e-05, + "loss": 0.0698, + "step": 11586 + }, + { + "epoch": 1.52, + "learning_rate": 1.8691679423589114e-05, + "loss": 0.0714, + "step": 11587 + }, + { + "epoch": 1.52, + "learning_rate": 1.8691417156593307e-05, + "loss": 0.0784, + "step": 11588 + }, + { + "epoch": 1.52, + "learning_rate": 1.8691154865153346e-05, + "loss": 0.0786, + "step": 11589 + }, + { + "epoch": 1.52, + "learning_rate": 1.8690892549269976e-05, + "loss": 0.0687, + "step": 11590 + }, + { + "epoch": 1.52, + "learning_rate": 1.869063020894393e-05, + "loss": 0.081, + "step": 11591 + }, + { + "epoch": 1.52, + "learning_rate": 1.8690367844175955e-05, + "loss": 0.0852, + "step": 11592 + }, + { + "epoch": 1.52, + "learning_rate": 1.869010545496678e-05, + "loss": 0.0695, + "step": 11593 + }, + { + "epoch": 1.52, + "learning_rate": 1.868984304131715e-05, + "loss": 0.0664, + "step": 11594 + }, + { + "epoch": 1.52, + "learning_rate": 1.8689580603227795e-05, + "loss": 0.0825, + "step": 11595 + }, + { + "epoch": 1.52, + "learning_rate": 1.8689318140699457e-05, + "loss": 0.0738, + "step": 11596 + }, + { + "epoch": 1.52, + "learning_rate": 1.8689055653732877e-05, + "loss": 0.0904, + "step": 11597 + }, + { + "epoch": 1.52, + "learning_rate": 1.868879314232879e-05, + "loss": 0.0777, + "step": 11598 + }, + { + "epoch": 1.52, + "learning_rate": 1.8688530606487934e-05, + "loss": 0.0788, + "step": 11599 + }, + { + "epoch": 1.52, + "learning_rate": 1.868826804621105e-05, + "loss": 0.078, + "step": 11600 + }, + { + "epoch": 1.52, + "learning_rate": 1.8688005461498876e-05, + "loss": 0.0768, + "step": 11601 + }, + { + "epoch": 1.52, + "learning_rate": 1.8687742852352144e-05, + "loss": 0.0733, + "step": 11602 + }, + { + "epoch": 1.52, + "learning_rate": 1.8687480218771604e-05, + "loss": 0.0761, + "step": 11603 + }, + { + "epoch": 1.52, + "learning_rate": 1.8687217560757986e-05, + "loss": 0.0768, + "step": 11604 + }, + { + "epoch": 1.52, + "learning_rate": 1.868695487831203e-05, + "loss": 0.0775, + "step": 11605 + }, + { + "epoch": 1.52, + "learning_rate": 1.8686692171434474e-05, + "loss": 0.0817, + "step": 11606 + }, + { + "epoch": 1.52, + "learning_rate": 1.8686429440126065e-05, + "loss": 0.0778, + "step": 11607 + }, + { + "epoch": 1.52, + "learning_rate": 1.868616668438753e-05, + "loss": 0.083, + "step": 11608 + }, + { + "epoch": 1.52, + "learning_rate": 1.868590390421962e-05, + "loss": 0.079, + "step": 11609 + }, + { + "epoch": 1.52, + "learning_rate": 1.8685641099623063e-05, + "loss": 0.0756, + "step": 11610 + }, + { + "epoch": 1.52, + "learning_rate": 1.8685378270598602e-05, + "loss": 0.0836, + "step": 11611 + }, + { + "epoch": 1.52, + "learning_rate": 1.868511541714698e-05, + "loss": 0.0894, + "step": 11612 + }, + { + "epoch": 1.52, + "learning_rate": 1.8684852539268933e-05, + "loss": 0.0668, + "step": 11613 + }, + { + "epoch": 1.52, + "learning_rate": 1.86845896369652e-05, + "loss": 0.068, + "step": 11614 + }, + { + "epoch": 1.52, + "learning_rate": 1.868432671023652e-05, + "loss": 0.073, + "step": 11615 + }, + { + "epoch": 1.52, + "learning_rate": 1.8684063759083634e-05, + "loss": 0.0738, + "step": 11616 + }, + { + "epoch": 1.52, + "learning_rate": 1.8683800783507284e-05, + "loss": 0.0772, + "step": 11617 + }, + { + "epoch": 1.52, + "learning_rate": 1.8683537783508203e-05, + "loss": 0.0795, + "step": 11618 + }, + { + "epoch": 1.52, + "learning_rate": 1.868327475908714e-05, + "loss": 0.0744, + "step": 11619 + }, + { + "epoch": 1.52, + "learning_rate": 1.868301171024482e-05, + "loss": 0.0722, + "step": 11620 + }, + { + "epoch": 1.52, + "learning_rate": 1.8682748636982e-05, + "loss": 0.0707, + "step": 11621 + }, + { + "epoch": 1.52, + "learning_rate": 1.868248553929941e-05, + "loss": 0.0821, + "step": 11622 + }, + { + "epoch": 1.52, + "learning_rate": 1.868222241719779e-05, + "loss": 0.0835, + "step": 11623 + }, + { + "epoch": 1.52, + "learning_rate": 1.8681959270677885e-05, + "loss": 0.0765, + "step": 11624 + }, + { + "epoch": 1.52, + "learning_rate": 1.868169609974043e-05, + "loss": 0.0691, + "step": 11625 + }, + { + "epoch": 1.52, + "learning_rate": 1.8681432904386166e-05, + "loss": 0.0745, + "step": 11626 + }, + { + "epoch": 1.52, + "learning_rate": 1.8681169684615835e-05, + "loss": 0.0893, + "step": 11627 + }, + { + "epoch": 1.52, + "learning_rate": 1.868090644043018e-05, + "loss": 0.0786, + "step": 11628 + }, + { + "epoch": 1.52, + "learning_rate": 1.868064317182993e-05, + "loss": 0.0705, + "step": 11629 + }, + { + "epoch": 1.52, + "learning_rate": 1.868037987881584e-05, + "loss": 0.0878, + "step": 11630 + }, + { + "epoch": 1.52, + "learning_rate": 1.8680116561388638e-05, + "loss": 0.0864, + "step": 11631 + }, + { + "epoch": 1.52, + "learning_rate": 1.867985321954908e-05, + "loss": 0.0762, + "step": 11632 + }, + { + "epoch": 1.52, + "learning_rate": 1.8679589853297887e-05, + "loss": 0.0871, + "step": 11633 + }, + { + "epoch": 1.52, + "learning_rate": 1.8679326462635816e-05, + "loss": 0.0738, + "step": 11634 + }, + { + "epoch": 1.52, + "learning_rate": 1.8679063047563598e-05, + "loss": 0.0768, + "step": 11635 + }, + { + "epoch": 1.52, + "learning_rate": 1.867879960808198e-05, + "loss": 0.0714, + "step": 11636 + }, + { + "epoch": 1.52, + "learning_rate": 1.8678536144191695e-05, + "loss": 0.0677, + "step": 11637 + }, + { + "epoch": 1.52, + "learning_rate": 1.8678272655893495e-05, + "loss": 0.0843, + "step": 11638 + }, + { + "epoch": 1.52, + "learning_rate": 1.8678009143188115e-05, + "loss": 0.0744, + "step": 11639 + }, + { + "epoch": 1.52, + "learning_rate": 1.8677745606076298e-05, + "loss": 0.0813, + "step": 11640 + }, + { + "epoch": 1.52, + "learning_rate": 1.8677482044558777e-05, + "loss": 0.0742, + "step": 11641 + }, + { + "epoch": 1.53, + "learning_rate": 1.8677218458636303e-05, + "loss": 0.0738, + "step": 11642 + }, + { + "epoch": 1.53, + "learning_rate": 1.8676954848309617e-05, + "loss": 0.0967, + "step": 11643 + }, + { + "epoch": 1.53, + "learning_rate": 1.8676691213579455e-05, + "loss": 0.0688, + "step": 11644 + }, + { + "epoch": 1.53, + "learning_rate": 1.867642755444656e-05, + "loss": 0.0773, + "step": 11645 + }, + { + "epoch": 1.53, + "learning_rate": 1.8676163870911678e-05, + "loss": 0.0785, + "step": 11646 + }, + { + "epoch": 1.53, + "learning_rate": 1.8675900162975545e-05, + "loss": 0.0807, + "step": 11647 + }, + { + "epoch": 1.53, + "learning_rate": 1.8675636430638902e-05, + "loss": 0.0781, + "step": 11648 + }, + { + "epoch": 1.53, + "learning_rate": 1.86753726739025e-05, + "loss": 0.0682, + "step": 11649 + }, + { + "epoch": 1.53, + "learning_rate": 1.867510889276707e-05, + "loss": 0.081, + "step": 11650 + }, + { + "epoch": 1.53, + "learning_rate": 1.8674845087233358e-05, + "loss": 0.0753, + "step": 11651 + }, + { + "epoch": 1.53, + "learning_rate": 1.8674581257302108e-05, + "loss": 0.076, + "step": 11652 + }, + { + "epoch": 1.53, + "learning_rate": 1.867431740297406e-05, + "loss": 0.0724, + "step": 11653 + }, + { + "epoch": 1.53, + "learning_rate": 1.8674053524249954e-05, + "loss": 0.0825, + "step": 11654 + }, + { + "epoch": 1.53, + "learning_rate": 1.8673789621130535e-05, + "loss": 0.0817, + "step": 11655 + }, + { + "epoch": 1.53, + "learning_rate": 1.867352569361655e-05, + "loss": 0.0763, + "step": 11656 + }, + { + "epoch": 1.53, + "learning_rate": 1.8673261741708726e-05, + "loss": 0.0686, + "step": 11657 + }, + { + "epoch": 1.53, + "learning_rate": 1.8672997765407822e-05, + "loss": 0.0752, + "step": 11658 + }, + { + "epoch": 1.53, + "learning_rate": 1.8672733764714574e-05, + "loss": 0.0744, + "step": 11659 + }, + { + "epoch": 1.53, + "learning_rate": 1.867246973962972e-05, + "loss": 0.0707, + "step": 11660 + }, + { + "epoch": 1.53, + "learning_rate": 1.867220569015401e-05, + "loss": 0.086, + "step": 11661 + }, + { + "epoch": 1.53, + "learning_rate": 1.8671941616288184e-05, + "loss": 0.0748, + "step": 11662 + }, + { + "epoch": 1.53, + "learning_rate": 1.8671677518032984e-05, + "loss": 0.0847, + "step": 11663 + }, + { + "epoch": 1.53, + "learning_rate": 1.867141339538915e-05, + "loss": 0.0557, + "step": 11664 + }, + { + "epoch": 1.53, + "learning_rate": 1.8671149248357428e-05, + "loss": 0.0712, + "step": 11665 + }, + { + "epoch": 1.53, + "learning_rate": 1.8670885076938557e-05, + "loss": 0.0712, + "step": 11666 + }, + { + "epoch": 1.53, + "learning_rate": 1.867062088113329e-05, + "loss": 0.0855, + "step": 11667 + }, + { + "epoch": 1.53, + "learning_rate": 1.8670356660942363e-05, + "loss": 0.0831, + "step": 11668 + }, + { + "epoch": 1.53, + "learning_rate": 1.8670092416366517e-05, + "loss": 0.0853, + "step": 11669 + }, + { + "epoch": 1.53, + "learning_rate": 1.86698281474065e-05, + "loss": 0.0703, + "step": 11670 + }, + { + "epoch": 1.53, + "learning_rate": 1.866956385406305e-05, + "loss": 0.0724, + "step": 11671 + }, + { + "epoch": 1.53, + "learning_rate": 1.8669299536336917e-05, + "loss": 0.0717, + "step": 11672 + }, + { + "epoch": 1.53, + "learning_rate": 1.866903519422884e-05, + "loss": 0.0718, + "step": 11673 + }, + { + "epoch": 1.53, + "learning_rate": 1.866877082773956e-05, + "loss": 0.0714, + "step": 11674 + }, + { + "epoch": 1.53, + "learning_rate": 1.8668506436869825e-05, + "loss": 0.0858, + "step": 11675 + }, + { + "epoch": 1.53, + "learning_rate": 1.8668242021620377e-05, + "loss": 0.0804, + "step": 11676 + }, + { + "epoch": 1.53, + "learning_rate": 1.866797758199196e-05, + "loss": 0.08, + "step": 11677 + }, + { + "epoch": 1.53, + "learning_rate": 1.866771311798532e-05, + "loss": 0.0762, + "step": 11678 + }, + { + "epoch": 1.53, + "learning_rate": 1.8667448629601197e-05, + "loss": 0.0813, + "step": 11679 + }, + { + "epoch": 1.53, + "learning_rate": 1.8667184116840334e-05, + "loss": 0.0743, + "step": 11680 + }, + { + "epoch": 1.53, + "learning_rate": 1.8666919579703482e-05, + "loss": 0.0716, + "step": 11681 + }, + { + "epoch": 1.53, + "learning_rate": 1.8666655018191377e-05, + "loss": 0.0822, + "step": 11682 + }, + { + "epoch": 1.53, + "learning_rate": 1.866639043230477e-05, + "loss": 0.0782, + "step": 11683 + }, + { + "epoch": 1.53, + "learning_rate": 1.8666125822044394e-05, + "loss": 0.0785, + "step": 11684 + }, + { + "epoch": 1.53, + "learning_rate": 1.8665861187411007e-05, + "loss": 0.0865, + "step": 11685 + }, + { + "epoch": 1.53, + "learning_rate": 1.866559652840534e-05, + "loss": 0.0854, + "step": 11686 + }, + { + "epoch": 1.53, + "learning_rate": 1.8665331845028153e-05, + "loss": 0.0962, + "step": 11687 + }, + { + "epoch": 1.53, + "learning_rate": 1.866506713728018e-05, + "loss": 0.083, + "step": 11688 + }, + { + "epoch": 1.53, + "learning_rate": 1.8664802405162164e-05, + "loss": 0.0695, + "step": 11689 + }, + { + "epoch": 1.53, + "learning_rate": 1.8664537648674855e-05, + "loss": 0.0684, + "step": 11690 + }, + { + "epoch": 1.53, + "learning_rate": 1.866427286781899e-05, + "loss": 0.0717, + "step": 11691 + }, + { + "epoch": 1.53, + "learning_rate": 1.8664008062595326e-05, + "loss": 0.0748, + "step": 11692 + }, + { + "epoch": 1.53, + "learning_rate": 1.8663743233004598e-05, + "loss": 0.0793, + "step": 11693 + }, + { + "epoch": 1.53, + "learning_rate": 1.866347837904755e-05, + "loss": 0.084, + "step": 11694 + }, + { + "epoch": 1.53, + "learning_rate": 1.8663213500724936e-05, + "loss": 0.0798, + "step": 11695 + }, + { + "epoch": 1.53, + "learning_rate": 1.866294859803749e-05, + "loss": 0.0676, + "step": 11696 + }, + { + "epoch": 1.53, + "learning_rate": 1.8662683670985965e-05, + "loss": 0.0776, + "step": 11697 + }, + { + "epoch": 1.53, + "learning_rate": 1.8662418719571106e-05, + "loss": 0.0821, + "step": 11698 + }, + { + "epoch": 1.53, + "learning_rate": 1.866215374379365e-05, + "loss": 0.0731, + "step": 11699 + }, + { + "epoch": 1.53, + "learning_rate": 1.8661888743654354e-05, + "loss": 0.0792, + "step": 11700 + }, + { + "epoch": 1.53, + "learning_rate": 1.8661623719153955e-05, + "loss": 0.0789, + "step": 11701 + }, + { + "epoch": 1.53, + "learning_rate": 1.86613586702932e-05, + "loss": 0.0853, + "step": 11702 + }, + { + "epoch": 1.53, + "learning_rate": 1.8661093597072834e-05, + "loss": 0.0849, + "step": 11703 + }, + { + "epoch": 1.53, + "learning_rate": 1.8660828499493604e-05, + "loss": 0.0759, + "step": 11704 + }, + { + "epoch": 1.53, + "learning_rate": 1.8660563377556257e-05, + "loss": 0.0756, + "step": 11705 + }, + { + "epoch": 1.53, + "learning_rate": 1.8660298231261532e-05, + "loss": 0.0786, + "step": 11706 + }, + { + "epoch": 1.53, + "learning_rate": 1.8660033060610186e-05, + "loss": 0.0704, + "step": 11707 + }, + { + "epoch": 1.53, + "learning_rate": 1.8659767865602952e-05, + "loss": 0.0682, + "step": 11708 + }, + { + "epoch": 1.53, + "learning_rate": 1.8659502646240587e-05, + "loss": 0.084, + "step": 11709 + }, + { + "epoch": 1.53, + "learning_rate": 1.865923740252383e-05, + "loss": 0.0774, + "step": 11710 + }, + { + "epoch": 1.53, + "learning_rate": 1.8658972134453428e-05, + "loss": 0.0796, + "step": 11711 + }, + { + "epoch": 1.53, + "learning_rate": 1.8658706842030128e-05, + "loss": 0.0683, + "step": 11712 + }, + { + "epoch": 1.53, + "learning_rate": 1.8658441525254678e-05, + "loss": 0.0717, + "step": 11713 + }, + { + "epoch": 1.53, + "learning_rate": 1.865817618412782e-05, + "loss": 0.0733, + "step": 11714 + }, + { + "epoch": 1.53, + "learning_rate": 1.8657910818650303e-05, + "loss": 0.062, + "step": 11715 + }, + { + "epoch": 1.53, + "learning_rate": 1.8657645428822876e-05, + "loss": 0.0698, + "step": 11716 + }, + { + "epoch": 1.53, + "learning_rate": 1.8657380014646278e-05, + "loss": 0.0791, + "step": 11717 + }, + { + "epoch": 1.53, + "learning_rate": 1.8657114576121263e-05, + "loss": 0.0733, + "step": 11718 + }, + { + "epoch": 1.54, + "learning_rate": 1.865684911324857e-05, + "loss": 0.0737, + "step": 11719 + }, + { + "epoch": 1.54, + "learning_rate": 1.8656583626028953e-05, + "loss": 0.0756, + "step": 11720 + }, + { + "epoch": 1.54, + "learning_rate": 1.8656318114463156e-05, + "loss": 0.0757, + "step": 11721 + }, + { + "epoch": 1.54, + "learning_rate": 1.8656052578551924e-05, + "loss": 0.0744, + "step": 11722 + }, + { + "epoch": 1.54, + "learning_rate": 1.8655787018296006e-05, + "loss": 0.0752, + "step": 11723 + }, + { + "epoch": 1.54, + "learning_rate": 1.8655521433696148e-05, + "loss": 0.0641, + "step": 11724 + }, + { + "epoch": 1.54, + "learning_rate": 1.86552558247531e-05, + "loss": 0.0727, + "step": 11725 + }, + { + "epoch": 1.54, + "learning_rate": 1.86549901914676e-05, + "loss": 0.0725, + "step": 11726 + }, + { + "epoch": 1.54, + "learning_rate": 1.8654724533840403e-05, + "loss": 0.0806, + "step": 11727 + }, + { + "epoch": 1.54, + "learning_rate": 1.8654458851872254e-05, + "loss": 0.0776, + "step": 11728 + }, + { + "epoch": 1.54, + "learning_rate": 1.8654193145563902e-05, + "loss": 0.0656, + "step": 11729 + }, + { + "epoch": 1.54, + "learning_rate": 1.8653927414916092e-05, + "loss": 0.0819, + "step": 11730 + }, + { + "epoch": 1.54, + "learning_rate": 1.865366165992957e-05, + "loss": 0.0723, + "step": 11731 + }, + { + "epoch": 1.54, + "learning_rate": 1.8653395880605088e-05, + "loss": 0.0777, + "step": 11732 + }, + { + "epoch": 1.54, + "learning_rate": 1.865313007694339e-05, + "loss": 0.0832, + "step": 11733 + }, + { + "epoch": 1.54, + "learning_rate": 1.8652864248945223e-05, + "loss": 0.0727, + "step": 11734 + }, + { + "epoch": 1.54, + "learning_rate": 1.8652598396611335e-05, + "loss": 0.0777, + "step": 11735 + }, + { + "epoch": 1.54, + "learning_rate": 1.865233251994248e-05, + "loss": 0.0679, + "step": 11736 + }, + { + "epoch": 1.54, + "learning_rate": 1.8652066618939396e-05, + "loss": 0.0774, + "step": 11737 + }, + { + "epoch": 1.54, + "learning_rate": 1.865180069360284e-05, + "loss": 0.0857, + "step": 11738 + }, + { + "epoch": 1.54, + "learning_rate": 1.8651534743933553e-05, + "loss": 0.0803, + "step": 11739 + }, + { + "epoch": 1.54, + "learning_rate": 1.8651268769932283e-05, + "loss": 0.0797, + "step": 11740 + }, + { + "epoch": 1.54, + "learning_rate": 1.865100277159978e-05, + "loss": 0.0841, + "step": 11741 + }, + { + "epoch": 1.54, + "learning_rate": 1.8650736748936794e-05, + "loss": 0.0773, + "step": 11742 + }, + { + "epoch": 1.54, + "learning_rate": 1.8650470701944076e-05, + "loss": 0.0746, + "step": 11743 + }, + { + "epoch": 1.54, + "learning_rate": 1.8650204630622362e-05, + "loss": 0.0844, + "step": 11744 + }, + { + "epoch": 1.54, + "learning_rate": 1.8649938534972413e-05, + "loss": 0.0841, + "step": 11745 + }, + { + "epoch": 1.54, + "learning_rate": 1.8649672414994972e-05, + "loss": 0.0727, + "step": 11746 + }, + { + "epoch": 1.54, + "learning_rate": 1.864940627069079e-05, + "loss": 0.0867, + "step": 11747 + }, + { + "epoch": 1.54, + "learning_rate": 1.864914010206061e-05, + "loss": 0.0831, + "step": 11748 + }, + { + "epoch": 1.54, + "learning_rate": 1.864887390910518e-05, + "loss": 0.0744, + "step": 11749 + }, + { + "epoch": 1.54, + "learning_rate": 1.864860769182526e-05, + "loss": 0.0764, + "step": 11750 + }, + { + "epoch": 1.54, + "learning_rate": 1.864834145022159e-05, + "loss": 0.0794, + "step": 11751 + }, + { + "epoch": 1.54, + "learning_rate": 1.8648075184294918e-05, + "loss": 0.0829, + "step": 11752 + }, + { + "epoch": 1.54, + "learning_rate": 1.8647808894045996e-05, + "loss": 0.0744, + "step": 11753 + }, + { + "epoch": 1.54, + "learning_rate": 1.8647542579475574e-05, + "loss": 0.0872, + "step": 11754 + }, + { + "epoch": 1.54, + "learning_rate": 1.8647276240584396e-05, + "loss": 0.0763, + "step": 11755 + }, + { + "epoch": 1.54, + "learning_rate": 1.8647009877373218e-05, + "loss": 0.08, + "step": 11756 + }, + { + "epoch": 1.54, + "learning_rate": 1.864674348984278e-05, + "loss": 0.0867, + "step": 11757 + }, + { + "epoch": 1.54, + "learning_rate": 1.8646477077993837e-05, + "loss": 0.0753, + "step": 11758 + }, + { + "epoch": 1.54, + "learning_rate": 1.8646210641827142e-05, + "loss": 0.0942, + "step": 11759 + }, + { + "epoch": 1.54, + "learning_rate": 1.864594418134344e-05, + "loss": 0.063, + "step": 11760 + }, + { + "epoch": 1.54, + "learning_rate": 1.8645677696543477e-05, + "loss": 0.083, + "step": 11761 + }, + { + "epoch": 1.54, + "learning_rate": 1.8645411187428008e-05, + "loss": 0.0854, + "step": 11762 + }, + { + "epoch": 1.54, + "learning_rate": 1.8645144653997777e-05, + "loss": 0.0854, + "step": 11763 + }, + { + "epoch": 1.54, + "learning_rate": 1.864487809625354e-05, + "loss": 0.0824, + "step": 11764 + }, + { + "epoch": 1.54, + "learning_rate": 1.8644611514196044e-05, + "loss": 0.0636, + "step": 11765 + }, + { + "epoch": 1.54, + "learning_rate": 1.8644344907826037e-05, + "loss": 0.0801, + "step": 11766 + }, + { + "epoch": 1.54, + "learning_rate": 1.864407827714427e-05, + "loss": 0.0854, + "step": 11767 + }, + { + "epoch": 1.54, + "learning_rate": 1.8643811622151496e-05, + "loss": 0.0842, + "step": 11768 + }, + { + "epoch": 1.54, + "learning_rate": 1.864354494284846e-05, + "loss": 0.084, + "step": 11769 + }, + { + "epoch": 1.54, + "learning_rate": 1.8643278239235912e-05, + "loss": 0.079, + "step": 11770 + }, + { + "epoch": 1.54, + "learning_rate": 1.8643011511314608e-05, + "loss": 0.074, + "step": 11771 + }, + { + "epoch": 1.54, + "learning_rate": 1.8642744759085293e-05, + "loss": 0.0825, + "step": 11772 + }, + { + "epoch": 1.54, + "learning_rate": 1.864247798254872e-05, + "loss": 0.0759, + "step": 11773 + }, + { + "epoch": 1.54, + "learning_rate": 1.8642211181705636e-05, + "loss": 0.0796, + "step": 11774 + }, + { + "epoch": 1.54, + "learning_rate": 1.8641944356556793e-05, + "loss": 0.0883, + "step": 11775 + }, + { + "epoch": 1.54, + "learning_rate": 1.8641677507102944e-05, + "loss": 0.0773, + "step": 11776 + }, + { + "epoch": 1.54, + "learning_rate": 1.8641410633344835e-05, + "loss": 0.076, + "step": 11777 + }, + { + "epoch": 1.54, + "learning_rate": 1.864114373528322e-05, + "loss": 0.0717, + "step": 11778 + }, + { + "epoch": 1.54, + "learning_rate": 1.8640876812918845e-05, + "loss": 0.0779, + "step": 11779 + }, + { + "epoch": 1.54, + "learning_rate": 1.8640609866252467e-05, + "loss": 0.0877, + "step": 11780 + }, + { + "epoch": 1.54, + "learning_rate": 1.8640342895284836e-05, + "loss": 0.075, + "step": 11781 + }, + { + "epoch": 1.54, + "learning_rate": 1.8640075900016697e-05, + "loss": 0.0813, + "step": 11782 + }, + { + "epoch": 1.54, + "learning_rate": 1.8639808880448806e-05, + "loss": 0.0724, + "step": 11783 + }, + { + "epoch": 1.54, + "learning_rate": 1.8639541836581915e-05, + "loss": 0.08, + "step": 11784 + }, + { + "epoch": 1.54, + "learning_rate": 1.863927476841677e-05, + "loss": 0.0955, + "step": 11785 + }, + { + "epoch": 1.54, + "learning_rate": 1.8639007675954123e-05, + "loss": 0.0721, + "step": 11786 + }, + { + "epoch": 1.54, + "learning_rate": 1.8638740559194734e-05, + "loss": 0.0701, + "step": 11787 + }, + { + "epoch": 1.54, + "learning_rate": 1.863847341813934e-05, + "loss": 0.0798, + "step": 11788 + }, + { + "epoch": 1.54, + "learning_rate": 1.8638206252788703e-05, + "loss": 0.0789, + "step": 11789 + }, + { + "epoch": 1.54, + "learning_rate": 1.863793906314357e-05, + "loss": 0.0776, + "step": 11790 + }, + { + "epoch": 1.54, + "learning_rate": 1.8637671849204692e-05, + "loss": 0.0828, + "step": 11791 + }, + { + "epoch": 1.54, + "learning_rate": 1.8637404610972823e-05, + "loss": 0.0682, + "step": 11792 + }, + { + "epoch": 1.54, + "learning_rate": 1.8637137348448713e-05, + "loss": 0.0747, + "step": 11793 + }, + { + "epoch": 1.54, + "learning_rate": 1.8636870061633118e-05, + "loss": 0.0788, + "step": 11794 + }, + { + "epoch": 1.55, + "learning_rate": 1.8636602750526783e-05, + "loss": 0.0745, + "step": 11795 + }, + { + "epoch": 1.55, + "learning_rate": 1.8636335415130463e-05, + "loss": 0.0655, + "step": 11796 + }, + { + "epoch": 1.55, + "learning_rate": 1.863606805544491e-05, + "loss": 0.0791, + "step": 11797 + }, + { + "epoch": 1.55, + "learning_rate": 1.8635800671470873e-05, + "loss": 0.0719, + "step": 11798 + }, + { + "epoch": 1.55, + "learning_rate": 1.8635533263209107e-05, + "loss": 0.0597, + "step": 11799 + }, + { + "epoch": 1.55, + "learning_rate": 1.863526583066036e-05, + "loss": 0.083, + "step": 11800 + }, + { + "epoch": 1.55, + "learning_rate": 1.8634998373825396e-05, + "loss": 0.0748, + "step": 11801 + }, + { + "epoch": 1.55, + "learning_rate": 1.8634730892704955e-05, + "loss": 0.0818, + "step": 11802 + }, + { + "epoch": 1.55, + "learning_rate": 1.8634463387299796e-05, + "loss": 0.071, + "step": 11803 + }, + { + "epoch": 1.55, + "learning_rate": 1.8634195857610662e-05, + "loss": 0.0617, + "step": 11804 + }, + { + "epoch": 1.55, + "learning_rate": 1.8633928303638318e-05, + "loss": 0.0809, + "step": 11805 + }, + { + "epoch": 1.55, + "learning_rate": 1.863366072538351e-05, + "loss": 0.081, + "step": 11806 + }, + { + "epoch": 1.55, + "learning_rate": 1.863339312284699e-05, + "loss": 0.0728, + "step": 11807 + }, + { + "epoch": 1.55, + "learning_rate": 1.863312549602951e-05, + "loss": 0.0844, + "step": 11808 + }, + { + "epoch": 1.55, + "learning_rate": 1.8632857844931826e-05, + "loss": 0.0733, + "step": 11809 + }, + { + "epoch": 1.55, + "learning_rate": 1.8632590169554686e-05, + "loss": 0.0643, + "step": 11810 + }, + { + "epoch": 1.55, + "learning_rate": 1.8632322469898847e-05, + "loss": 0.0854, + "step": 11811 + }, + { + "epoch": 1.55, + "learning_rate": 1.8632054745965066e-05, + "loss": 0.0689, + "step": 11812 + }, + { + "epoch": 1.55, + "learning_rate": 1.8631786997754083e-05, + "loss": 0.0773, + "step": 11813 + }, + { + "epoch": 1.55, + "learning_rate": 1.8631519225266662e-05, + "loss": 0.0714, + "step": 11814 + }, + { + "epoch": 1.55, + "learning_rate": 1.8631251428503555e-05, + "loss": 0.0748, + "step": 11815 + }, + { + "epoch": 1.55, + "learning_rate": 1.863098360746551e-05, + "loss": 0.0759, + "step": 11816 + }, + { + "epoch": 1.55, + "learning_rate": 1.8630715762153285e-05, + "loss": 0.0887, + "step": 11817 + }, + { + "epoch": 1.55, + "learning_rate": 1.863044789256763e-05, + "loss": 0.0714, + "step": 11818 + }, + { + "epoch": 1.55, + "learning_rate": 1.8630179998709303e-05, + "loss": 0.0745, + "step": 11819 + }, + { + "epoch": 1.55, + "learning_rate": 1.8629912080579053e-05, + "loss": 0.0733, + "step": 11820 + }, + { + "epoch": 1.55, + "learning_rate": 1.8629644138177632e-05, + "loss": 0.0783, + "step": 11821 + }, + { + "epoch": 1.55, + "learning_rate": 1.8629376171505797e-05, + "loss": 0.0739, + "step": 11822 + }, + { + "epoch": 1.55, + "learning_rate": 1.8629108180564303e-05, + "loss": 0.0732, + "step": 11823 + }, + { + "epoch": 1.55, + "learning_rate": 1.8628840165353902e-05, + "loss": 0.0743, + "step": 11824 + }, + { + "epoch": 1.55, + "learning_rate": 1.8628572125875347e-05, + "loss": 0.0816, + "step": 11825 + }, + { + "epoch": 1.55, + "learning_rate": 1.8628304062129392e-05, + "loss": 0.0764, + "step": 11826 + }, + { + "epoch": 1.55, + "learning_rate": 1.862803597411679e-05, + "loss": 0.0743, + "step": 11827 + }, + { + "epoch": 1.55, + "learning_rate": 1.86277678618383e-05, + "loss": 0.0763, + "step": 11828 + }, + { + "epoch": 1.55, + "learning_rate": 1.862749972529467e-05, + "loss": 0.0823, + "step": 11829 + }, + { + "epoch": 1.55, + "learning_rate": 1.8627231564486658e-05, + "loss": 0.0813, + "step": 11830 + }, + { + "epoch": 1.55, + "learning_rate": 1.8626963379415014e-05, + "loss": 0.0807, + "step": 11831 + }, + { + "epoch": 1.55, + "learning_rate": 1.8626695170080494e-05, + "loss": 0.072, + "step": 11832 + }, + { + "epoch": 1.55, + "learning_rate": 1.8626426936483858e-05, + "loss": 0.0806, + "step": 11833 + }, + { + "epoch": 1.55, + "learning_rate": 1.8626158678625853e-05, + "loss": 0.0856, + "step": 11834 + }, + { + "epoch": 1.55, + "learning_rate": 1.8625890396507236e-05, + "loss": 0.0884, + "step": 11835 + }, + { + "epoch": 1.55, + "learning_rate": 1.862562209012876e-05, + "loss": 0.0832, + "step": 11836 + }, + { + "epoch": 1.55, + "learning_rate": 1.8625353759491185e-05, + "loss": 0.0697, + "step": 11837 + }, + { + "epoch": 1.55, + "learning_rate": 1.8625085404595256e-05, + "loss": 0.071, + "step": 11838 + }, + { + "epoch": 1.55, + "learning_rate": 1.862481702544174e-05, + "loss": 0.0835, + "step": 11839 + }, + { + "epoch": 1.55, + "learning_rate": 1.8624548622031383e-05, + "loss": 0.0837, + "step": 11840 + }, + { + "epoch": 1.55, + "learning_rate": 1.8624280194364943e-05, + "loss": 0.0679, + "step": 11841 + }, + { + "epoch": 1.55, + "learning_rate": 1.8624011742443172e-05, + "loss": 0.0831, + "step": 11842 + }, + { + "epoch": 1.55, + "learning_rate": 1.8623743266266827e-05, + "loss": 0.0699, + "step": 11843 + }, + { + "epoch": 1.55, + "learning_rate": 1.8623474765836668e-05, + "loss": 0.0708, + "step": 11844 + }, + { + "epoch": 1.55, + "learning_rate": 1.862320624115344e-05, + "loss": 0.0767, + "step": 11845 + }, + { + "epoch": 1.55, + "learning_rate": 1.8622937692217905e-05, + "loss": 0.0767, + "step": 11846 + }, + { + "epoch": 1.55, + "learning_rate": 1.8622669119030822e-05, + "loss": 0.0671, + "step": 11847 + }, + { + "epoch": 1.55, + "learning_rate": 1.8622400521592934e-05, + "loss": 0.0773, + "step": 11848 + }, + { + "epoch": 1.55, + "learning_rate": 1.862213189990501e-05, + "loss": 0.0717, + "step": 11849 + }, + { + "epoch": 1.55, + "learning_rate": 1.8621863253967794e-05, + "loss": 0.0779, + "step": 11850 + }, + { + "epoch": 1.55, + "learning_rate": 1.862159458378205e-05, + "loss": 0.0854, + "step": 11851 + }, + { + "epoch": 1.55, + "learning_rate": 1.862132588934853e-05, + "loss": 0.0849, + "step": 11852 + }, + { + "epoch": 1.55, + "learning_rate": 1.8621057170667987e-05, + "loss": 0.081, + "step": 11853 + }, + { + "epoch": 1.55, + "learning_rate": 1.8620788427741183e-05, + "loss": 0.0822, + "step": 11854 + }, + { + "epoch": 1.55, + "learning_rate": 1.8620519660568867e-05, + "loss": 0.0698, + "step": 11855 + }, + { + "epoch": 1.55, + "learning_rate": 1.8620250869151805e-05, + "loss": 0.0864, + "step": 11856 + }, + { + "epoch": 1.55, + "learning_rate": 1.8619982053490743e-05, + "loss": 0.07, + "step": 11857 + }, + { + "epoch": 1.55, + "learning_rate": 1.861971321358644e-05, + "loss": 0.0705, + "step": 11858 + }, + { + "epoch": 1.55, + "learning_rate": 1.861944434943965e-05, + "loss": 0.068, + "step": 11859 + }, + { + "epoch": 1.55, + "learning_rate": 1.8619175461051134e-05, + "loss": 0.0778, + "step": 11860 + }, + { + "epoch": 1.55, + "learning_rate": 1.8618906548421648e-05, + "loss": 0.0706, + "step": 11861 + }, + { + "epoch": 1.55, + "learning_rate": 1.8618637611551946e-05, + "loss": 0.0835, + "step": 11862 + }, + { + "epoch": 1.55, + "learning_rate": 1.861836865044278e-05, + "loss": 0.0727, + "step": 11863 + }, + { + "epoch": 1.55, + "learning_rate": 1.8618099665094916e-05, + "loss": 0.0705, + "step": 11864 + }, + { + "epoch": 1.55, + "learning_rate": 1.8617830655509106e-05, + "loss": 0.0836, + "step": 11865 + }, + { + "epoch": 1.55, + "learning_rate": 1.8617561621686106e-05, + "loss": 0.075, + "step": 11866 + }, + { + "epoch": 1.55, + "learning_rate": 1.861729256362667e-05, + "loss": 0.0807, + "step": 11867 + }, + { + "epoch": 1.55, + "learning_rate": 1.861702348133156e-05, + "loss": 0.0703, + "step": 11868 + }, + { + "epoch": 1.55, + "learning_rate": 1.8616754374801527e-05, + "loss": 0.0719, + "step": 11869 + }, + { + "epoch": 1.55, + "learning_rate": 1.8616485244037334e-05, + "loss": 0.0868, + "step": 11870 + }, + { + "epoch": 1.56, + "learning_rate": 1.8616216089039734e-05, + "loss": 0.0685, + "step": 11871 + }, + { + "epoch": 1.56, + "learning_rate": 1.8615946909809488e-05, + "loss": 0.0693, + "step": 11872 + }, + { + "epoch": 1.56, + "learning_rate": 1.8615677706347347e-05, + "loss": 0.0733, + "step": 11873 + }, + { + "epoch": 1.56, + "learning_rate": 1.861540847865407e-05, + "loss": 0.0764, + "step": 11874 + }, + { + "epoch": 1.56, + "learning_rate": 1.8615139226730422e-05, + "loss": 0.0723, + "step": 11875 + }, + { + "epoch": 1.56, + "learning_rate": 1.8614869950577147e-05, + "loss": 0.0806, + "step": 11876 + }, + { + "epoch": 1.56, + "learning_rate": 1.8614600650195012e-05, + "loss": 0.0766, + "step": 11877 + }, + { + "epoch": 1.56, + "learning_rate": 1.861433132558477e-05, + "loss": 0.0756, + "step": 11878 + }, + { + "epoch": 1.56, + "learning_rate": 1.861406197674718e-05, + "loss": 0.0811, + "step": 11879 + }, + { + "epoch": 1.56, + "learning_rate": 1.8613792603683e-05, + "loss": 0.0723, + "step": 11880 + }, + { + "epoch": 1.56, + "learning_rate": 1.8613523206392988e-05, + "loss": 0.0782, + "step": 11881 + }, + { + "epoch": 1.56, + "learning_rate": 1.86132537848779e-05, + "loss": 0.0835, + "step": 11882 + }, + { + "epoch": 1.56, + "learning_rate": 1.8612984339138493e-05, + "loss": 0.0813, + "step": 11883 + }, + { + "epoch": 1.56, + "learning_rate": 1.8612714869175527e-05, + "loss": 0.0767, + "step": 11884 + }, + { + "epoch": 1.56, + "learning_rate": 1.861244537498976e-05, + "loss": 0.0715, + "step": 11885 + }, + { + "epoch": 1.56, + "learning_rate": 1.861217585658195e-05, + "loss": 0.0737, + "step": 11886 + }, + { + "epoch": 1.56, + "learning_rate": 1.8611906313952854e-05, + "loss": 0.0768, + "step": 11887 + }, + { + "epoch": 1.56, + "learning_rate": 1.8611636747103224e-05, + "loss": 0.0874, + "step": 11888 + }, + { + "epoch": 1.56, + "learning_rate": 1.861136715603383e-05, + "loss": 0.0823, + "step": 11889 + }, + { + "epoch": 1.56, + "learning_rate": 1.8611097540745424e-05, + "loss": 0.0864, + "step": 11890 + }, + { + "epoch": 1.56, + "learning_rate": 1.8610827901238765e-05, + "loss": 0.0693, + "step": 11891 + }, + { + "epoch": 1.56, + "learning_rate": 1.8610558237514607e-05, + "loss": 0.0794, + "step": 11892 + }, + { + "epoch": 1.56, + "learning_rate": 1.8610288549573717e-05, + "loss": 0.0792, + "step": 11893 + }, + { + "epoch": 1.56, + "learning_rate": 1.861001883741685e-05, + "loss": 0.0812, + "step": 11894 + }, + { + "epoch": 1.56, + "learning_rate": 1.860974910104476e-05, + "loss": 0.0725, + "step": 11895 + }, + { + "epoch": 1.56, + "learning_rate": 1.860947934045821e-05, + "loss": 0.0765, + "step": 11896 + }, + { + "epoch": 1.56, + "learning_rate": 1.860920955565796e-05, + "loss": 0.0752, + "step": 11897 + }, + { + "epoch": 1.56, + "learning_rate": 1.8608939746644765e-05, + "loss": 0.065, + "step": 11898 + }, + { + "epoch": 1.56, + "learning_rate": 1.8608669913419384e-05, + "loss": 0.0787, + "step": 11899 + }, + { + "epoch": 1.56, + "learning_rate": 1.860840005598258e-05, + "loss": 0.0819, + "step": 11900 + }, + { + "epoch": 1.56, + "learning_rate": 1.8608130174335106e-05, + "loss": 0.0795, + "step": 11901 + }, + { + "epoch": 1.56, + "learning_rate": 1.8607860268477727e-05, + "loss": 0.0697, + "step": 11902 + }, + { + "epoch": 1.56, + "learning_rate": 1.8607590338411198e-05, + "loss": 0.0835, + "step": 11903 + }, + { + "epoch": 1.56, + "learning_rate": 1.860732038413628e-05, + "loss": 0.0879, + "step": 11904 + }, + { + "epoch": 1.56, + "learning_rate": 1.860705040565373e-05, + "loss": 0.0882, + "step": 11905 + }, + { + "epoch": 1.56, + "learning_rate": 1.860678040296431e-05, + "loss": 0.0918, + "step": 11906 + }, + { + "epoch": 1.56, + "learning_rate": 1.860651037606878e-05, + "loss": 0.0743, + "step": 11907 + }, + { + "epoch": 1.56, + "learning_rate": 1.8606240324967898e-05, + "loss": 0.0797, + "step": 11908 + }, + { + "epoch": 1.56, + "learning_rate": 1.860597024966242e-05, + "loss": 0.0806, + "step": 11909 + }, + { + "epoch": 1.56, + "learning_rate": 1.8605700150153114e-05, + "loss": 0.0712, + "step": 11910 + }, + { + "epoch": 1.56, + "learning_rate": 1.860543002644073e-05, + "loss": 0.081, + "step": 11911 + }, + { + "epoch": 1.56, + "learning_rate": 1.8605159878526036e-05, + "loss": 0.0821, + "step": 11912 + }, + { + "epoch": 1.56, + "learning_rate": 1.8604889706409786e-05, + "loss": 0.0619, + "step": 11913 + }, + { + "epoch": 1.56, + "learning_rate": 1.8604619510092745e-05, + "loss": 0.0762, + "step": 11914 + }, + { + "epoch": 1.56, + "learning_rate": 1.8604349289575663e-05, + "loss": 0.0853, + "step": 11915 + }, + { + "epoch": 1.56, + "learning_rate": 1.8604079044859315e-05, + "loss": 0.0887, + "step": 11916 + }, + { + "epoch": 1.56, + "learning_rate": 1.860380877594445e-05, + "loss": 0.0662, + "step": 11917 + }, + { + "epoch": 1.56, + "learning_rate": 1.860353848283183e-05, + "loss": 0.0852, + "step": 11918 + }, + { + "epoch": 1.56, + "learning_rate": 1.8603268165522216e-05, + "loss": 0.0848, + "step": 11919 + }, + { + "epoch": 1.56, + "learning_rate": 1.8602997824016368e-05, + "loss": 0.0884, + "step": 11920 + }, + { + "epoch": 1.56, + "learning_rate": 1.8602727458315045e-05, + "loss": 0.087, + "step": 11921 + }, + { + "epoch": 1.56, + "learning_rate": 1.8602457068419014e-05, + "loss": 0.0757, + "step": 11922 + }, + { + "epoch": 1.56, + "learning_rate": 1.860218665432903e-05, + "loss": 0.0821, + "step": 11923 + }, + { + "epoch": 1.56, + "learning_rate": 1.860191621604585e-05, + "loss": 0.0759, + "step": 11924 + }, + { + "epoch": 1.56, + "learning_rate": 1.8601645753570242e-05, + "loss": 0.076, + "step": 11925 + }, + { + "epoch": 1.56, + "learning_rate": 1.8601375266902958e-05, + "loss": 0.0776, + "step": 11926 + }, + { + "epoch": 1.56, + "learning_rate": 1.860110475604477e-05, + "loss": 0.0778, + "step": 11927 + }, + { + "epoch": 1.56, + "learning_rate": 1.860083422099643e-05, + "loss": 0.071, + "step": 11928 + }, + { + "epoch": 1.56, + "learning_rate": 1.8600563661758705e-05, + "loss": 0.0795, + "step": 11929 + }, + { + "epoch": 1.56, + "learning_rate": 1.8600293078332353e-05, + "loss": 0.0681, + "step": 11930 + }, + { + "epoch": 1.56, + "learning_rate": 1.860002247071813e-05, + "loss": 0.072, + "step": 11931 + }, + { + "epoch": 1.56, + "learning_rate": 1.8599751838916807e-05, + "loss": 0.0848, + "step": 11932 + }, + { + "epoch": 1.56, + "learning_rate": 1.8599481182929137e-05, + "loss": 0.0772, + "step": 11933 + }, + { + "epoch": 1.56, + "learning_rate": 1.859921050275588e-05, + "loss": 0.0836, + "step": 11934 + }, + { + "epoch": 1.56, + "learning_rate": 1.8598939798397812e-05, + "loss": 0.0832, + "step": 11935 + }, + { + "epoch": 1.56, + "learning_rate": 1.8598669069855675e-05, + "loss": 0.0753, + "step": 11936 + }, + { + "epoch": 1.56, + "learning_rate": 1.8598398317130243e-05, + "loss": 0.0792, + "step": 11937 + }, + { + "epoch": 1.56, + "learning_rate": 1.859812754022227e-05, + "loss": 0.0822, + "step": 11938 + }, + { + "epoch": 1.56, + "learning_rate": 1.8597856739132524e-05, + "loss": 0.0859, + "step": 11939 + }, + { + "epoch": 1.56, + "learning_rate": 1.8597585913861765e-05, + "loss": 0.0701, + "step": 11940 + }, + { + "epoch": 1.56, + "learning_rate": 1.859731506441075e-05, + "loss": 0.0634, + "step": 11941 + }, + { + "epoch": 1.56, + "learning_rate": 1.8597044190780248e-05, + "loss": 0.0701, + "step": 11942 + }, + { + "epoch": 1.56, + "learning_rate": 1.8596773292971015e-05, + "loss": 0.0721, + "step": 11943 + }, + { + "epoch": 1.56, + "learning_rate": 1.8596502370983815e-05, + "loss": 0.0801, + "step": 11944 + }, + { + "epoch": 1.56, + "learning_rate": 1.8596231424819405e-05, + "loss": 0.0718, + "step": 11945 + }, + { + "epoch": 1.56, + "learning_rate": 1.859596045447856e-05, + "loss": 0.0728, + "step": 11946 + }, + { + "epoch": 1.56, + "learning_rate": 1.8595689459962033e-05, + "loss": 0.075, + "step": 11947 + }, + { + "epoch": 1.57, + "learning_rate": 1.8595418441270583e-05, + "loss": 0.0825, + "step": 11948 + }, + { + "epoch": 1.57, + "learning_rate": 1.859514739840498e-05, + "loss": 0.0923, + "step": 11949 + }, + { + "epoch": 1.57, + "learning_rate": 1.8594876331365975e-05, + "loss": 0.062, + "step": 11950 + }, + { + "epoch": 1.57, + "learning_rate": 1.8594605240154346e-05, + "loss": 0.0771, + "step": 11951 + }, + { + "epoch": 1.57, + "learning_rate": 1.859433412477085e-05, + "loss": 0.0704, + "step": 11952 + }, + { + "epoch": 1.57, + "learning_rate": 1.859406298521624e-05, + "loss": 0.0843, + "step": 11953 + }, + { + "epoch": 1.57, + "learning_rate": 1.8593791821491286e-05, + "loss": 0.0743, + "step": 11954 + }, + { + "epoch": 1.57, + "learning_rate": 1.8593520633596754e-05, + "loss": 0.0745, + "step": 11955 + }, + { + "epoch": 1.57, + "learning_rate": 1.8593249421533398e-05, + "loss": 0.0752, + "step": 11956 + }, + { + "epoch": 1.57, + "learning_rate": 1.8592978185301992e-05, + "loss": 0.0832, + "step": 11957 + }, + { + "epoch": 1.57, + "learning_rate": 1.8592706924903287e-05, + "loss": 0.0739, + "step": 11958 + }, + { + "epoch": 1.57, + "learning_rate": 1.8592435640338054e-05, + "loss": 0.0815, + "step": 11959 + }, + { + "epoch": 1.57, + "learning_rate": 1.8592164331607052e-05, + "loss": 0.0807, + "step": 11960 + }, + { + "epoch": 1.57, + "learning_rate": 1.8591892998711044e-05, + "loss": 0.0693, + "step": 11961 + }, + { + "epoch": 1.57, + "learning_rate": 1.85916216416508e-05, + "loss": 0.0709, + "step": 11962 + }, + { + "epoch": 1.57, + "learning_rate": 1.8591350260427072e-05, + "loss": 0.0707, + "step": 11963 + }, + { + "epoch": 1.57, + "learning_rate": 1.859107885504063e-05, + "loss": 0.0598, + "step": 11964 + }, + { + "epoch": 1.57, + "learning_rate": 1.8590807425492236e-05, + "loss": 0.0739, + "step": 11965 + }, + { + "epoch": 1.57, + "learning_rate": 1.8590535971782655e-05, + "loss": 0.0746, + "step": 11966 + }, + { + "epoch": 1.57, + "learning_rate": 1.8590264493912652e-05, + "loss": 0.0814, + "step": 11967 + }, + { + "epoch": 1.57, + "learning_rate": 1.8589992991882982e-05, + "loss": 0.0886, + "step": 11968 + }, + { + "epoch": 1.57, + "learning_rate": 1.8589721465694418e-05, + "loss": 0.0856, + "step": 11969 + }, + { + "epoch": 1.57, + "learning_rate": 1.8589449915347716e-05, + "loss": 0.0725, + "step": 11970 + }, + { + "epoch": 1.57, + "learning_rate": 1.8589178340843647e-05, + "loss": 0.0732, + "step": 11971 + }, + { + "epoch": 1.57, + "learning_rate": 1.858890674218297e-05, + "loss": 0.0711, + "step": 11972 + }, + { + "epoch": 1.57, + "learning_rate": 1.858863511936645e-05, + "loss": 0.0686, + "step": 11973 + }, + { + "epoch": 1.57, + "learning_rate": 1.8588363472394854e-05, + "loss": 0.0856, + "step": 11974 + }, + { + "epoch": 1.57, + "learning_rate": 1.858809180126894e-05, + "loss": 0.0697, + "step": 11975 + }, + { + "epoch": 1.57, + "learning_rate": 1.8587820105989473e-05, + "loss": 0.0829, + "step": 11976 + }, + { + "epoch": 1.57, + "learning_rate": 1.8587548386557223e-05, + "loss": 0.0715, + "step": 11977 + }, + { + "epoch": 1.57, + "learning_rate": 1.858727664297295e-05, + "loss": 0.0761, + "step": 11978 + }, + { + "epoch": 1.57, + "learning_rate": 1.8587004875237417e-05, + "loss": 0.0768, + "step": 11979 + }, + { + "epoch": 1.57, + "learning_rate": 1.8586733083351393e-05, + "loss": 0.07, + "step": 11980 + }, + { + "epoch": 1.57, + "learning_rate": 1.858646126731564e-05, + "loss": 0.0713, + "step": 11981 + }, + { + "epoch": 1.57, + "learning_rate": 1.8586189427130917e-05, + "loss": 0.0692, + "step": 11982 + }, + { + "epoch": 1.57, + "learning_rate": 1.8585917562798e-05, + "loss": 0.0743, + "step": 11983 + }, + { + "epoch": 1.57, + "learning_rate": 1.858564567431764e-05, + "loss": 0.078, + "step": 11984 + }, + { + "epoch": 1.57, + "learning_rate": 1.8585373761690613e-05, + "loss": 0.0801, + "step": 11985 + }, + { + "epoch": 1.57, + "learning_rate": 1.8585101824917678e-05, + "loss": 0.0935, + "step": 11986 + }, + { + "epoch": 1.57, + "learning_rate": 1.85848298639996e-05, + "loss": 0.0752, + "step": 11987 + }, + { + "epoch": 1.57, + "learning_rate": 1.858455787893715e-05, + "loss": 0.0768, + "step": 11988 + }, + { + "epoch": 1.57, + "learning_rate": 1.8584285869731086e-05, + "loss": 0.0761, + "step": 11989 + }, + { + "epoch": 1.57, + "learning_rate": 1.858401383638217e-05, + "loss": 0.0797, + "step": 11990 + }, + { + "epoch": 1.57, + "learning_rate": 1.8583741778891178e-05, + "loss": 0.0642, + "step": 11991 + }, + { + "epoch": 1.57, + "learning_rate": 1.8583469697258868e-05, + "loss": 0.0763, + "step": 11992 + }, + { + "epoch": 1.57, + "learning_rate": 1.8583197591486008e-05, + "loss": 0.0867, + "step": 11993 + }, + { + "epoch": 1.57, + "learning_rate": 1.858292546157336e-05, + "loss": 0.0782, + "step": 11994 + }, + { + "epoch": 1.57, + "learning_rate": 1.8582653307521694e-05, + "loss": 0.0717, + "step": 11995 + }, + { + "epoch": 1.57, + "learning_rate": 1.8582381129331768e-05, + "loss": 0.084, + "step": 11996 + }, + { + "epoch": 1.57, + "learning_rate": 1.858210892700435e-05, + "loss": 0.0734, + "step": 11997 + }, + { + "epoch": 1.57, + "learning_rate": 1.8581836700540216e-05, + "loss": 0.0755, + "step": 11998 + }, + { + "epoch": 1.57, + "learning_rate": 1.8581564449940117e-05, + "loss": 0.0815, + "step": 11999 + }, + { + "epoch": 1.57, + "learning_rate": 1.8581292175204827e-05, + "loss": 0.086, + "step": 12000 + }, + { + "epoch": 1.57, + "learning_rate": 1.8581019876335113e-05, + "loss": 0.0813, + "step": 12001 + }, + { + "epoch": 1.57, + "learning_rate": 1.8580747553331732e-05, + "loss": 0.0737, + "step": 12002 + }, + { + "epoch": 1.57, + "learning_rate": 1.8580475206195457e-05, + "loss": 0.0907, + "step": 12003 + }, + { + "epoch": 1.57, + "learning_rate": 1.858020283492706e-05, + "loss": 0.0836, + "step": 12004 + }, + { + "epoch": 1.57, + "learning_rate": 1.8579930439527287e-05, + "loss": 0.078, + "step": 12005 + }, + { + "epoch": 1.57, + "learning_rate": 1.8579658019996923e-05, + "loss": 0.0878, + "step": 12006 + }, + { + "epoch": 1.57, + "learning_rate": 1.8579385576336727e-05, + "loss": 0.0729, + "step": 12007 + }, + { + "epoch": 1.57, + "learning_rate": 1.8579113108547467e-05, + "loss": 0.0808, + "step": 12008 + }, + { + "epoch": 1.57, + "learning_rate": 1.8578840616629908e-05, + "loss": 0.0788, + "step": 12009 + }, + { + "epoch": 1.57, + "learning_rate": 1.857856810058481e-05, + "loss": 0.0763, + "step": 12010 + }, + { + "epoch": 1.57, + "learning_rate": 1.8578295560412957e-05, + "loss": 0.0847, + "step": 12011 + }, + { + "epoch": 1.57, + "learning_rate": 1.85780229961151e-05, + "loss": 0.0736, + "step": 12012 + }, + { + "epoch": 1.57, + "learning_rate": 1.8577750407692008e-05, + "loss": 0.0785, + "step": 12013 + }, + { + "epoch": 1.57, + "learning_rate": 1.857747779514445e-05, + "loss": 0.0654, + "step": 12014 + }, + { + "epoch": 1.57, + "learning_rate": 1.8577205158473196e-05, + "loss": 0.0667, + "step": 12015 + }, + { + "epoch": 1.57, + "learning_rate": 1.8576932497679005e-05, + "loss": 0.0713, + "step": 12016 + }, + { + "epoch": 1.57, + "learning_rate": 1.857665981276265e-05, + "loss": 0.0732, + "step": 12017 + }, + { + "epoch": 1.57, + "learning_rate": 1.85763871037249e-05, + "loss": 0.0697, + "step": 12018 + }, + { + "epoch": 1.57, + "learning_rate": 1.857611437056651e-05, + "loss": 0.0759, + "step": 12019 + }, + { + "epoch": 1.57, + "learning_rate": 1.857584161328826e-05, + "loss": 0.0754, + "step": 12020 + }, + { + "epoch": 1.57, + "learning_rate": 1.857556883189091e-05, + "loss": 0.0761, + "step": 12021 + }, + { + "epoch": 1.57, + "learning_rate": 1.8575296026375234e-05, + "loss": 0.0712, + "step": 12022 + }, + { + "epoch": 1.57, + "learning_rate": 1.857502319674199e-05, + "loss": 0.0784, + "step": 12023 + }, + { + "epoch": 1.58, + "learning_rate": 1.8574750342991952e-05, + "loss": 0.0784, + "step": 12024 + }, + { + "epoch": 1.58, + "learning_rate": 1.8574477465125883e-05, + "loss": 0.0936, + "step": 12025 + }, + { + "epoch": 1.58, + "learning_rate": 1.8574204563144556e-05, + "loss": 0.0858, + "step": 12026 + }, + { + "epoch": 1.58, + "learning_rate": 1.8573931637048733e-05, + "loss": 0.0729, + "step": 12027 + }, + { + "epoch": 1.58, + "learning_rate": 1.8573658686839186e-05, + "loss": 0.0862, + "step": 12028 + }, + { + "epoch": 1.58, + "learning_rate": 1.8573385712516678e-05, + "loss": 0.0755, + "step": 12029 + }, + { + "epoch": 1.58, + "learning_rate": 1.857311271408198e-05, + "loss": 0.0657, + "step": 12030 + }, + { + "epoch": 1.58, + "learning_rate": 1.8572839691535858e-05, + "loss": 0.0662, + "step": 12031 + }, + { + "epoch": 1.58, + "learning_rate": 1.8572566644879084e-05, + "loss": 0.0769, + "step": 12032 + }, + { + "epoch": 1.58, + "learning_rate": 1.857229357411242e-05, + "loss": 0.0756, + "step": 12033 + }, + { + "epoch": 1.58, + "learning_rate": 1.8572020479236638e-05, + "loss": 0.0898, + "step": 12034 + }, + { + "epoch": 1.58, + "learning_rate": 1.8571747360252503e-05, + "loss": 0.0721, + "step": 12035 + }, + { + "epoch": 1.58, + "learning_rate": 1.8571474217160788e-05, + "loss": 0.07, + "step": 12036 + }, + { + "epoch": 1.58, + "learning_rate": 1.8571201049962257e-05, + "loss": 0.0661, + "step": 12037 + }, + { + "epoch": 1.58, + "learning_rate": 1.8570927858657677e-05, + "loss": 0.0719, + "step": 12038 + }, + { + "epoch": 1.58, + "learning_rate": 1.8570654643247824e-05, + "loss": 0.0669, + "step": 12039 + }, + { + "epoch": 1.58, + "learning_rate": 1.8570381403733456e-05, + "loss": 0.0738, + "step": 12040 + }, + { + "epoch": 1.58, + "learning_rate": 1.8570108140115352e-05, + "loss": 0.0861, + "step": 12041 + }, + { + "epoch": 1.58, + "learning_rate": 1.8569834852394272e-05, + "loss": 0.072, + "step": 12042 + }, + { + "epoch": 1.58, + "learning_rate": 1.8569561540570985e-05, + "loss": 0.0801, + "step": 12043 + }, + { + "epoch": 1.58, + "learning_rate": 1.8569288204646266e-05, + "loss": 0.0746, + "step": 12044 + }, + { + "epoch": 1.58, + "learning_rate": 1.856901484462088e-05, + "loss": 0.0819, + "step": 12045 + }, + { + "epoch": 1.58, + "learning_rate": 1.8568741460495597e-05, + "loss": 0.081, + "step": 12046 + }, + { + "epoch": 1.58, + "learning_rate": 1.8568468052271185e-05, + "loss": 0.0741, + "step": 12047 + }, + { + "epoch": 1.58, + "learning_rate": 1.856819461994841e-05, + "loss": 0.0719, + "step": 12048 + }, + { + "epoch": 1.58, + "learning_rate": 1.856792116352805e-05, + "loss": 0.0866, + "step": 12049 + }, + { + "epoch": 1.58, + "learning_rate": 1.856764768301086e-05, + "loss": 0.0817, + "step": 12050 + }, + { + "epoch": 1.58, + "learning_rate": 1.8567374178397624e-05, + "loss": 0.0866, + "step": 12051 + }, + { + "epoch": 1.58, + "learning_rate": 1.85671006496891e-05, + "loss": 0.0755, + "step": 12052 + }, + { + "epoch": 1.58, + "learning_rate": 1.8566827096886065e-05, + "loss": 0.0835, + "step": 12053 + }, + { + "epoch": 1.58, + "learning_rate": 1.856655351998928e-05, + "loss": 0.0733, + "step": 12054 + }, + { + "epoch": 1.58, + "learning_rate": 1.8566279918999524e-05, + "loss": 0.0927, + "step": 12055 + }, + { + "epoch": 1.58, + "learning_rate": 1.856600629391756e-05, + "loss": 0.0866, + "step": 12056 + }, + { + "epoch": 1.58, + "learning_rate": 1.8565732644744158e-05, + "loss": 0.0621, + "step": 12057 + }, + { + "epoch": 1.58, + "learning_rate": 1.8565458971480094e-05, + "loss": 0.0756, + "step": 12058 + }, + { + "epoch": 1.58, + "learning_rate": 1.856518527412613e-05, + "loss": 0.0704, + "step": 12059 + }, + { + "epoch": 1.58, + "learning_rate": 1.8564911552683038e-05, + "loss": 0.0716, + "step": 12060 + }, + { + "epoch": 1.58, + "learning_rate": 1.8564637807151586e-05, + "loss": 0.0812, + "step": 12061 + }, + { + "epoch": 1.58, + "learning_rate": 1.856436403753255e-05, + "loss": 0.0644, + "step": 12062 + }, + { + "epoch": 1.58, + "learning_rate": 1.8564090243826694e-05, + "loss": 0.0805, + "step": 12063 + }, + { + "epoch": 1.58, + "learning_rate": 1.8563816426034795e-05, + "loss": 0.08, + "step": 12064 + }, + { + "epoch": 1.58, + "learning_rate": 1.856354258415761e-05, + "loss": 0.0775, + "step": 12065 + }, + { + "epoch": 1.58, + "learning_rate": 1.8563268718195924e-05, + "loss": 0.0797, + "step": 12066 + }, + { + "epoch": 1.58, + "learning_rate": 1.8562994828150498e-05, + "loss": 0.0813, + "step": 12067 + }, + { + "epoch": 1.58, + "learning_rate": 1.8562720914022105e-05, + "loss": 0.0816, + "step": 12068 + }, + { + "epoch": 1.58, + "learning_rate": 1.856244697581152e-05, + "loss": 0.0746, + "step": 12069 + }, + { + "epoch": 1.58, + "learning_rate": 1.8562173013519502e-05, + "loss": 0.0871, + "step": 12070 + }, + { + "epoch": 1.58, + "learning_rate": 1.8561899027146833e-05, + "loss": 0.0648, + "step": 12071 + }, + { + "epoch": 1.58, + "learning_rate": 1.8561625016694278e-05, + "loss": 0.0777, + "step": 12072 + }, + { + "epoch": 1.58, + "learning_rate": 1.856135098216261e-05, + "loss": 0.0774, + "step": 12073 + }, + { + "epoch": 1.58, + "learning_rate": 1.856107692355259e-05, + "loss": 0.0688, + "step": 12074 + }, + { + "epoch": 1.58, + "learning_rate": 1.8560802840865005e-05, + "loss": 0.0777, + "step": 12075 + }, + { + "epoch": 1.58, + "learning_rate": 1.856052873410062e-05, + "loss": 0.0604, + "step": 12076 + }, + { + "epoch": 1.58, + "learning_rate": 1.85602546032602e-05, + "loss": 0.0698, + "step": 12077 + }, + { + "epoch": 1.58, + "learning_rate": 1.8559980448344518e-05, + "loss": 0.0913, + "step": 12078 + }, + { + "epoch": 1.58, + "learning_rate": 1.855970626935435e-05, + "loss": 0.0707, + "step": 12079 + }, + { + "epoch": 1.58, + "learning_rate": 1.8559432066290463e-05, + "loss": 0.0725, + "step": 12080 + }, + { + "epoch": 1.58, + "learning_rate": 1.855915783915363e-05, + "loss": 0.0773, + "step": 12081 + }, + { + "epoch": 1.58, + "learning_rate": 1.855888358794462e-05, + "loss": 0.0728, + "step": 12082 + }, + { + "epoch": 1.58, + "learning_rate": 1.8558609312664205e-05, + "loss": 0.0772, + "step": 12083 + }, + { + "epoch": 1.58, + "learning_rate": 1.8558335013313156e-05, + "loss": 0.0798, + "step": 12084 + }, + { + "epoch": 1.58, + "learning_rate": 1.855806068989225e-05, + "loss": 0.0787, + "step": 12085 + }, + { + "epoch": 1.58, + "learning_rate": 1.855778634240225e-05, + "loss": 0.072, + "step": 12086 + }, + { + "epoch": 1.58, + "learning_rate": 1.8557511970843933e-05, + "loss": 0.087, + "step": 12087 + }, + { + "epoch": 1.58, + "learning_rate": 1.855723757521807e-05, + "loss": 0.0805, + "step": 12088 + }, + { + "epoch": 1.58, + "learning_rate": 1.855696315552543e-05, + "loss": 0.0768, + "step": 12089 + }, + { + "epoch": 1.58, + "learning_rate": 1.8556688711766792e-05, + "loss": 0.0764, + "step": 12090 + }, + { + "epoch": 1.58, + "learning_rate": 1.8556414243942916e-05, + "loss": 0.0688, + "step": 12091 + }, + { + "epoch": 1.58, + "learning_rate": 1.8556139752054584e-05, + "loss": 0.066, + "step": 12092 + }, + { + "epoch": 1.58, + "learning_rate": 1.8555865236102567e-05, + "loss": 0.0804, + "step": 12093 + }, + { + "epoch": 1.58, + "learning_rate": 1.855559069608763e-05, + "loss": 0.0809, + "step": 12094 + }, + { + "epoch": 1.58, + "learning_rate": 1.855531613201055e-05, + "loss": 0.0779, + "step": 12095 + }, + { + "epoch": 1.58, + "learning_rate": 1.85550415438721e-05, + "loss": 0.0815, + "step": 12096 + }, + { + "epoch": 1.58, + "learning_rate": 1.8554766931673054e-05, + "loss": 0.0781, + "step": 12097 + }, + { + "epoch": 1.58, + "learning_rate": 1.855449229541418e-05, + "loss": 0.0831, + "step": 12098 + }, + { + "epoch": 1.58, + "learning_rate": 1.8554217635096253e-05, + "loss": 0.0708, + "step": 12099 + }, + { + "epoch": 1.59, + "learning_rate": 1.855394295072004e-05, + "loss": 0.0736, + "step": 12100 + }, + { + "epoch": 1.59, + "learning_rate": 1.855366824228632e-05, + "loss": 0.0774, + "step": 12101 + }, + { + "epoch": 1.59, + "learning_rate": 1.8553393509795868e-05, + "loss": 0.0724, + "step": 12102 + }, + { + "epoch": 1.59, + "learning_rate": 1.8553118753249445e-05, + "loss": 0.0741, + "step": 12103 + }, + { + "epoch": 1.59, + "learning_rate": 1.8552843972647836e-05, + "loss": 0.0759, + "step": 12104 + }, + { + "epoch": 1.59, + "learning_rate": 1.855256916799181e-05, + "loss": 0.0848, + "step": 12105 + }, + { + "epoch": 1.59, + "learning_rate": 1.8552294339282136e-05, + "loss": 0.0734, + "step": 12106 + }, + { + "epoch": 1.59, + "learning_rate": 1.8552019486519587e-05, + "loss": 0.0683, + "step": 12107 + }, + { + "epoch": 1.59, + "learning_rate": 1.855174460970494e-05, + "loss": 0.0783, + "step": 12108 + }, + { + "epoch": 1.59, + "learning_rate": 1.855146970883897e-05, + "loss": 0.0771, + "step": 12109 + }, + { + "epoch": 1.59, + "learning_rate": 1.8551194783922443e-05, + "loss": 0.0889, + "step": 12110 + }, + { + "epoch": 1.59, + "learning_rate": 1.855091983495614e-05, + "loss": 0.0819, + "step": 12111 + }, + { + "epoch": 1.59, + "learning_rate": 1.855064486194083e-05, + "loss": 0.0748, + "step": 12112 + }, + { + "epoch": 1.59, + "learning_rate": 1.855036986487728e-05, + "loss": 0.0781, + "step": 12113 + }, + { + "epoch": 1.59, + "learning_rate": 1.8550094843766277e-05, + "loss": 0.0668, + "step": 12114 + }, + { + "epoch": 1.59, + "learning_rate": 1.8549819798608587e-05, + "loss": 0.0714, + "step": 12115 + }, + { + "epoch": 1.59, + "learning_rate": 1.8549544729404983e-05, + "loss": 0.0778, + "step": 12116 + }, + { + "epoch": 1.59, + "learning_rate": 1.854926963615624e-05, + "loss": 0.0856, + "step": 12117 + }, + { + "epoch": 1.59, + "learning_rate": 1.8548994518863132e-05, + "loss": 0.08, + "step": 12118 + }, + { + "epoch": 1.59, + "learning_rate": 1.854871937752643e-05, + "loss": 0.0744, + "step": 12119 + }, + { + "epoch": 1.59, + "learning_rate": 1.854844421214691e-05, + "loss": 0.075, + "step": 12120 + }, + { + "epoch": 1.59, + "learning_rate": 1.8548169022725347e-05, + "loss": 0.0729, + "step": 12121 + }, + { + "epoch": 1.59, + "learning_rate": 1.8547893809262515e-05, + "loss": 0.0712, + "step": 12122 + }, + { + "epoch": 1.59, + "learning_rate": 1.8547618571759184e-05, + "loss": 0.0782, + "step": 12123 + }, + { + "epoch": 1.59, + "learning_rate": 1.8547343310216132e-05, + "loss": 0.0774, + "step": 12124 + }, + { + "epoch": 1.59, + "learning_rate": 1.8547068024634134e-05, + "loss": 0.0814, + "step": 12125 + }, + { + "epoch": 1.59, + "learning_rate": 1.854679271501396e-05, + "loss": 0.0754, + "step": 12126 + }, + { + "epoch": 1.59, + "learning_rate": 1.8546517381356387e-05, + "loss": 0.0692, + "step": 12127 + }, + { + "epoch": 1.59, + "learning_rate": 1.854624202366219e-05, + "loss": 0.0753, + "step": 12128 + }, + { + "epoch": 1.59, + "learning_rate": 1.8545966641932144e-05, + "loss": 0.0714, + "step": 12129 + }, + { + "epoch": 1.59, + "learning_rate": 1.854569123616702e-05, + "loss": 0.0801, + "step": 12130 + }, + { + "epoch": 1.59, + "learning_rate": 1.854541580636759e-05, + "loss": 0.0823, + "step": 12131 + }, + { + "epoch": 1.59, + "learning_rate": 1.854514035253464e-05, + "loss": 0.0767, + "step": 12132 + }, + { + "epoch": 1.59, + "learning_rate": 1.8544864874668936e-05, + "loss": 0.0919, + "step": 12133 + }, + { + "epoch": 1.59, + "learning_rate": 1.8544589372771257e-05, + "loss": 0.0772, + "step": 12134 + }, + { + "epoch": 1.59, + "learning_rate": 1.854431384684237e-05, + "loss": 0.082, + "step": 12135 + }, + { + "epoch": 1.59, + "learning_rate": 1.8544038296883057e-05, + "loss": 0.0793, + "step": 12136 + }, + { + "epoch": 1.59, + "learning_rate": 1.8543762722894092e-05, + "loss": 0.0731, + "step": 12137 + }, + { + "epoch": 1.59, + "learning_rate": 1.854348712487625e-05, + "loss": 0.0791, + "step": 12138 + }, + { + "epoch": 1.59, + "learning_rate": 1.8543211502830303e-05, + "loss": 0.0806, + "step": 12139 + }, + { + "epoch": 1.59, + "learning_rate": 1.8542935856757032e-05, + "loss": 0.0669, + "step": 12140 + }, + { + "epoch": 1.59, + "learning_rate": 1.8542660186657207e-05, + "loss": 0.075, + "step": 12141 + }, + { + "epoch": 1.59, + "learning_rate": 1.8542384492531607e-05, + "loss": 0.0784, + "step": 12142 + }, + { + "epoch": 1.59, + "learning_rate": 1.8542108774381002e-05, + "loss": 0.0621, + "step": 12143 + }, + { + "epoch": 1.59, + "learning_rate": 1.854183303220617e-05, + "loss": 0.0782, + "step": 12144 + }, + { + "epoch": 1.59, + "learning_rate": 1.8541557266007896e-05, + "loss": 0.086, + "step": 12145 + }, + { + "epoch": 1.59, + "learning_rate": 1.8541281475786937e-05, + "loss": 0.0744, + "step": 12146 + }, + { + "epoch": 1.59, + "learning_rate": 1.8541005661544083e-05, + "loss": 0.0785, + "step": 12147 + }, + { + "epoch": 1.59, + "learning_rate": 1.8540729823280107e-05, + "loss": 0.0861, + "step": 12148 + }, + { + "epoch": 1.59, + "learning_rate": 1.854045396099578e-05, + "loss": 0.0812, + "step": 12149 + }, + { + "epoch": 1.59, + "learning_rate": 1.8540178074691883e-05, + "loss": 0.0833, + "step": 12150 + }, + { + "epoch": 1.59, + "learning_rate": 1.8539902164369194e-05, + "loss": 0.0724, + "step": 12151 + }, + { + "epoch": 1.59, + "learning_rate": 1.853962623002848e-05, + "loss": 0.0831, + "step": 12152 + }, + { + "epoch": 1.59, + "learning_rate": 1.853935027167052e-05, + "loss": 0.0793, + "step": 12153 + }, + { + "epoch": 1.59, + "learning_rate": 1.8539074289296096e-05, + "loss": 0.0823, + "step": 12154 + }, + { + "epoch": 1.59, + "learning_rate": 1.8538798282905976e-05, + "loss": 0.073, + "step": 12155 + }, + { + "epoch": 1.59, + "learning_rate": 1.8538522252500944e-05, + "loss": 0.0827, + "step": 12156 + }, + { + "epoch": 1.59, + "learning_rate": 1.8538246198081772e-05, + "loss": 0.0845, + "step": 12157 + }, + { + "epoch": 1.59, + "learning_rate": 1.853797011964924e-05, + "loss": 0.0798, + "step": 12158 + }, + { + "epoch": 1.59, + "learning_rate": 1.853769401720412e-05, + "loss": 0.071, + "step": 12159 + }, + { + "epoch": 1.59, + "learning_rate": 1.853741789074719e-05, + "loss": 0.0744, + "step": 12160 + }, + { + "epoch": 1.59, + "learning_rate": 1.8537141740279225e-05, + "loss": 0.0779, + "step": 12161 + }, + { + "epoch": 1.59, + "learning_rate": 1.8536865565801007e-05, + "loss": 0.0813, + "step": 12162 + }, + { + "epoch": 1.59, + "learning_rate": 1.8536589367313305e-05, + "loss": 0.0896, + "step": 12163 + }, + { + "epoch": 1.59, + "learning_rate": 1.85363131448169e-05, + "loss": 0.0963, + "step": 12164 + }, + { + "epoch": 1.59, + "learning_rate": 1.853603689831257e-05, + "loss": 0.0749, + "step": 12165 + }, + { + "epoch": 1.59, + "learning_rate": 1.8535760627801095e-05, + "loss": 0.0717, + "step": 12166 + }, + { + "epoch": 1.59, + "learning_rate": 1.8535484333283245e-05, + "loss": 0.0787, + "step": 12167 + }, + { + "epoch": 1.59, + "learning_rate": 1.85352080147598e-05, + "loss": 0.0792, + "step": 12168 + }, + { + "epoch": 1.59, + "learning_rate": 1.8534931672231536e-05, + "loss": 0.0721, + "step": 12169 + }, + { + "epoch": 1.59, + "learning_rate": 1.8534655305699233e-05, + "loss": 0.0748, + "step": 12170 + }, + { + "epoch": 1.59, + "learning_rate": 1.8534378915163663e-05, + "loss": 0.0728, + "step": 12171 + }, + { + "epoch": 1.59, + "learning_rate": 1.8534102500625608e-05, + "loss": 0.071, + "step": 12172 + }, + { + "epoch": 1.59, + "learning_rate": 1.8533826062085848e-05, + "loss": 0.0863, + "step": 12173 + }, + { + "epoch": 1.59, + "learning_rate": 1.8533549599545152e-05, + "loss": 0.0853, + "step": 12174 + }, + { + "epoch": 1.59, + "learning_rate": 1.8533273113004306e-05, + "loss": 0.0724, + "step": 12175 + }, + { + "epoch": 1.59, + "learning_rate": 1.853299660246408e-05, + "loss": 0.0748, + "step": 12176 + }, + { + "epoch": 1.6, + "learning_rate": 1.8532720067925257e-05, + "loss": 0.0886, + "step": 12177 + }, + { + "epoch": 1.6, + "learning_rate": 1.8532443509388613e-05, + "loss": 0.0831, + "step": 12178 + }, + { + "epoch": 1.6, + "learning_rate": 1.8532166926854925e-05, + "loss": 0.0803, + "step": 12179 + }, + { + "epoch": 1.6, + "learning_rate": 1.8531890320324974e-05, + "loss": 0.0742, + "step": 12180 + }, + { + "epoch": 1.6, + "learning_rate": 1.8531613689799535e-05, + "loss": 0.0729, + "step": 12181 + }, + { + "epoch": 1.6, + "learning_rate": 1.8531337035279388e-05, + "loss": 0.0778, + "step": 12182 + }, + { + "epoch": 1.6, + "learning_rate": 1.8531060356765307e-05, + "loss": 0.0772, + "step": 12183 + }, + { + "epoch": 1.6, + "learning_rate": 1.8530783654258076e-05, + "loss": 0.0769, + "step": 12184 + }, + { + "epoch": 1.6, + "learning_rate": 1.8530506927758468e-05, + "loss": 0.0787, + "step": 12185 + }, + { + "epoch": 1.6, + "learning_rate": 1.8530230177267265e-05, + "loss": 0.0621, + "step": 12186 + }, + { + "epoch": 1.6, + "learning_rate": 1.8529953402785243e-05, + "loss": 0.0759, + "step": 12187 + }, + { + "epoch": 1.6, + "learning_rate": 1.8529676604313184e-05, + "loss": 0.0827, + "step": 12188 + }, + { + "epoch": 1.6, + "learning_rate": 1.852939978185186e-05, + "loss": 0.0949, + "step": 12189 + }, + { + "epoch": 1.6, + "learning_rate": 1.8529122935402054e-05, + "loss": 0.0714, + "step": 12190 + }, + { + "epoch": 1.6, + "learning_rate": 1.8528846064964544e-05, + "loss": 0.0749, + "step": 12191 + }, + { + "epoch": 1.6, + "learning_rate": 1.8528569170540112e-05, + "loss": 0.0821, + "step": 12192 + }, + { + "epoch": 1.6, + "learning_rate": 1.8528292252129533e-05, + "loss": 0.0646, + "step": 12193 + }, + { + "epoch": 1.6, + "learning_rate": 1.8528015309733583e-05, + "loss": 0.0795, + "step": 12194 + }, + { + "epoch": 1.6, + "learning_rate": 1.8527738343353045e-05, + "loss": 0.0727, + "step": 12195 + }, + { + "epoch": 1.6, + "learning_rate": 1.8527461352988697e-05, + "loss": 0.0808, + "step": 12196 + }, + { + "epoch": 1.6, + "learning_rate": 1.8527184338641317e-05, + "loss": 0.0832, + "step": 12197 + }, + { + "epoch": 1.6, + "learning_rate": 1.852690730031169e-05, + "loss": 0.0829, + "step": 12198 + }, + { + "epoch": 1.6, + "learning_rate": 1.8526630238000583e-05, + "loss": 0.0833, + "step": 12199 + }, + { + "epoch": 1.6, + "learning_rate": 1.8526353151708785e-05, + "loss": 0.0771, + "step": 12200 + }, + { + "epoch": 1.6, + "learning_rate": 1.8526076041437075e-05, + "loss": 0.0815, + "step": 12201 + }, + { + "epoch": 1.6, + "learning_rate": 1.852579890718623e-05, + "loss": 0.0769, + "step": 12202 + }, + { + "epoch": 1.6, + "learning_rate": 1.852552174895703e-05, + "loss": 0.0794, + "step": 12203 + }, + { + "epoch": 1.6, + "learning_rate": 1.8525244566750254e-05, + "loss": 0.0781, + "step": 12204 + }, + { + "epoch": 1.6, + "learning_rate": 1.852496736056668e-05, + "loss": 0.065, + "step": 12205 + }, + { + "epoch": 1.6, + "learning_rate": 1.8524690130407087e-05, + "loss": 0.0784, + "step": 12206 + }, + { + "epoch": 1.6, + "learning_rate": 1.852441287627226e-05, + "loss": 0.0814, + "step": 12207 + }, + { + "epoch": 1.6, + "learning_rate": 1.852413559816298e-05, + "loss": 0.0704, + "step": 12208 + }, + { + "epoch": 1.6, + "learning_rate": 1.8523858296080013e-05, + "loss": 0.0764, + "step": 12209 + }, + { + "epoch": 1.6, + "learning_rate": 1.8523580970024156e-05, + "loss": 0.0918, + "step": 12210 + }, + { + "epoch": 1.6, + "learning_rate": 1.852330361999618e-05, + "loss": 0.0802, + "step": 12211 + }, + { + "epoch": 1.6, + "learning_rate": 1.8523026245996864e-05, + "loss": 0.0787, + "step": 12212 + }, + { + "epoch": 1.6, + "learning_rate": 1.8522748848026994e-05, + "loss": 0.0861, + "step": 12213 + }, + { + "epoch": 1.6, + "learning_rate": 1.8522471426087346e-05, + "loss": 0.0803, + "step": 12214 + }, + { + "epoch": 1.6, + "learning_rate": 1.8522193980178698e-05, + "loss": 0.0788, + "step": 12215 + }, + { + "epoch": 1.6, + "learning_rate": 1.8521916510301836e-05, + "loss": 0.0751, + "step": 12216 + }, + { + "epoch": 1.6, + "learning_rate": 1.8521639016457534e-05, + "loss": 0.0755, + "step": 12217 + }, + { + "epoch": 1.6, + "learning_rate": 1.852136149864658e-05, + "loss": 0.0866, + "step": 12218 + }, + { + "epoch": 1.6, + "learning_rate": 1.852108395686975e-05, + "loss": 0.0819, + "step": 12219 + }, + { + "epoch": 1.6, + "learning_rate": 1.8520806391127825e-05, + "loss": 0.084, + "step": 12220 + }, + { + "epoch": 1.6, + "learning_rate": 1.8520528801421585e-05, + "loss": 0.0734, + "step": 12221 + }, + { + "epoch": 1.6, + "learning_rate": 1.852025118775181e-05, + "loss": 0.0716, + "step": 12222 + }, + { + "epoch": 1.6, + "learning_rate": 1.8519973550119285e-05, + "loss": 0.0705, + "step": 12223 + }, + { + "epoch": 1.6, + "learning_rate": 1.8519695888524786e-05, + "loss": 0.08, + "step": 12224 + }, + { + "epoch": 1.6, + "learning_rate": 1.8519418202969098e-05, + "loss": 0.0745, + "step": 12225 + }, + { + "epoch": 1.6, + "learning_rate": 1.8519140493453e-05, + "loss": 0.0739, + "step": 12226 + }, + { + "epoch": 1.6, + "learning_rate": 1.851886275997727e-05, + "loss": 0.0728, + "step": 12227 + }, + { + "epoch": 1.6, + "learning_rate": 1.8518585002542696e-05, + "loss": 0.0835, + "step": 12228 + }, + { + "epoch": 1.6, + "learning_rate": 1.8518307221150054e-05, + "loss": 0.077, + "step": 12229 + }, + { + "epoch": 1.6, + "learning_rate": 1.8518029415800126e-05, + "loss": 0.0713, + "step": 12230 + }, + { + "epoch": 1.6, + "learning_rate": 1.8517751586493693e-05, + "loss": 0.0722, + "step": 12231 + }, + { + "epoch": 1.6, + "learning_rate": 1.851747373323154e-05, + "loss": 0.0717, + "step": 12232 + }, + { + "epoch": 1.6, + "learning_rate": 1.8517195856014443e-05, + "loss": 0.0808, + "step": 12233 + }, + { + "epoch": 1.6, + "learning_rate": 1.8516917954843188e-05, + "loss": 0.0774, + "step": 12234 + }, + { + "epoch": 1.6, + "learning_rate": 1.8516640029718552e-05, + "loss": 0.0781, + "step": 12235 + }, + { + "epoch": 1.6, + "learning_rate": 1.8516362080641325e-05, + "loss": 0.0732, + "step": 12236 + }, + { + "epoch": 1.6, + "learning_rate": 1.8516084107612277e-05, + "loss": 0.0818, + "step": 12237 + }, + { + "epoch": 1.6, + "learning_rate": 1.85158061106322e-05, + "loss": 0.0818, + "step": 12238 + }, + { + "epoch": 1.6, + "learning_rate": 1.851552808970187e-05, + "loss": 0.0763, + "step": 12239 + }, + { + "epoch": 1.6, + "learning_rate": 1.8515250044822072e-05, + "loss": 0.0881, + "step": 12240 + }, + { + "epoch": 1.6, + "learning_rate": 1.8514971975993587e-05, + "loss": 0.0759, + "step": 12241 + }, + { + "epoch": 1.6, + "learning_rate": 1.8514693883217194e-05, + "loss": 0.0819, + "step": 12242 + }, + { + "epoch": 1.6, + "learning_rate": 1.8514415766493678e-05, + "loss": 0.08, + "step": 12243 + }, + { + "epoch": 1.6, + "learning_rate": 1.8514137625823822e-05, + "loss": 0.0759, + "step": 12244 + }, + { + "epoch": 1.6, + "learning_rate": 1.8513859461208407e-05, + "loss": 0.0873, + "step": 12245 + }, + { + "epoch": 1.6, + "learning_rate": 1.8513581272648216e-05, + "loss": 0.0877, + "step": 12246 + }, + { + "epoch": 1.6, + "learning_rate": 1.8513303060144034e-05, + "loss": 0.0766, + "step": 12247 + }, + { + "epoch": 1.6, + "learning_rate": 1.851302482369664e-05, + "loss": 0.0733, + "step": 12248 + }, + { + "epoch": 1.6, + "learning_rate": 1.8512746563306813e-05, + "loss": 0.085, + "step": 12249 + }, + { + "epoch": 1.6, + "learning_rate": 1.851246827897534e-05, + "loss": 0.074, + "step": 12250 + }, + { + "epoch": 1.6, + "learning_rate": 1.851218997070301e-05, + "loss": 0.0783, + "step": 12251 + }, + { + "epoch": 1.6, + "learning_rate": 1.851191163849059e-05, + "loss": 0.0752, + "step": 12252 + }, + { + "epoch": 1.61, + "learning_rate": 1.8511633282338878e-05, + "loss": 0.0665, + "step": 12253 + }, + { + "epoch": 1.61, + "learning_rate": 1.8511354902248644e-05, + "loss": 0.0814, + "step": 12254 + }, + { + "epoch": 1.61, + "learning_rate": 1.8511076498220687e-05, + "loss": 0.0698, + "step": 12255 + }, + { + "epoch": 1.61, + "learning_rate": 1.8510798070255776e-05, + "loss": 0.0846, + "step": 12256 + }, + { + "epoch": 1.61, + "learning_rate": 1.8510519618354696e-05, + "loss": 0.0702, + "step": 12257 + }, + { + "epoch": 1.61, + "learning_rate": 1.8510241142518237e-05, + "loss": 0.0805, + "step": 12258 + }, + { + "epoch": 1.61, + "learning_rate": 1.8509962642747174e-05, + "loss": 0.0777, + "step": 12259 + }, + { + "epoch": 1.61, + "learning_rate": 1.8509684119042298e-05, + "loss": 0.0907, + "step": 12260 + }, + { + "epoch": 1.61, + "learning_rate": 1.8509405571404387e-05, + "loss": 0.0715, + "step": 12261 + }, + { + "epoch": 1.61, + "learning_rate": 1.8509126999834225e-05, + "loss": 0.0718, + "step": 12262 + }, + { + "epoch": 1.61, + "learning_rate": 1.8508848404332596e-05, + "loss": 0.0704, + "step": 12263 + }, + { + "epoch": 1.61, + "learning_rate": 1.8508569784900288e-05, + "loss": 0.0807, + "step": 12264 + }, + { + "epoch": 1.61, + "learning_rate": 1.8508291141538076e-05, + "loss": 0.0763, + "step": 12265 + }, + { + "epoch": 1.61, + "learning_rate": 1.8508012474246753e-05, + "loss": 0.0665, + "step": 12266 + }, + { + "epoch": 1.61, + "learning_rate": 1.850773378302709e-05, + "loss": 0.0877, + "step": 12267 + }, + { + "epoch": 1.61, + "learning_rate": 1.8507455067879888e-05, + "loss": 0.0706, + "step": 12268 + }, + { + "epoch": 1.61, + "learning_rate": 1.8507176328805917e-05, + "loss": 0.0913, + "step": 12269 + }, + { + "epoch": 1.61, + "learning_rate": 1.8506897565805966e-05, + "loss": 0.0799, + "step": 12270 + }, + { + "epoch": 1.61, + "learning_rate": 1.850661877888082e-05, + "loss": 0.0837, + "step": 12271 + }, + { + "epoch": 1.61, + "learning_rate": 1.850633996803126e-05, + "loss": 0.084, + "step": 12272 + }, + { + "epoch": 1.61, + "learning_rate": 1.850606113325807e-05, + "loss": 0.0821, + "step": 12273 + }, + { + "epoch": 1.61, + "learning_rate": 1.8505782274562043e-05, + "loss": 0.0633, + "step": 12274 + }, + { + "epoch": 1.61, + "learning_rate": 1.850550339194395e-05, + "loss": 0.0819, + "step": 12275 + }, + { + "epoch": 1.61, + "learning_rate": 1.8505224485404586e-05, + "loss": 0.0798, + "step": 12276 + }, + { + "epoch": 1.61, + "learning_rate": 1.8504945554944728e-05, + "loss": 0.0667, + "step": 12277 + }, + { + "epoch": 1.61, + "learning_rate": 1.8504666600565163e-05, + "loss": 0.0782, + "step": 12278 + }, + { + "epoch": 1.61, + "learning_rate": 1.8504387622266676e-05, + "loss": 0.0744, + "step": 12279 + }, + { + "epoch": 1.61, + "learning_rate": 1.8504108620050055e-05, + "loss": 0.0924, + "step": 12280 + }, + { + "epoch": 1.61, + "learning_rate": 1.8503829593916077e-05, + "loss": 0.073, + "step": 12281 + }, + { + "epoch": 1.61, + "learning_rate": 1.8503550543865534e-05, + "loss": 0.0798, + "step": 12282 + }, + { + "epoch": 1.61, + "learning_rate": 1.8503271469899205e-05, + "loss": 0.0675, + "step": 12283 + }, + { + "epoch": 1.61, + "learning_rate": 1.8502992372017884e-05, + "loss": 0.0797, + "step": 12284 + }, + { + "epoch": 1.61, + "learning_rate": 1.8502713250222342e-05, + "loss": 0.0758, + "step": 12285 + }, + { + "epoch": 1.61, + "learning_rate": 1.8502434104513377e-05, + "loss": 0.0799, + "step": 12286 + }, + { + "epoch": 1.61, + "learning_rate": 1.8502154934891766e-05, + "loss": 0.0703, + "step": 12287 + }, + { + "epoch": 1.61, + "learning_rate": 1.85018757413583e-05, + "loss": 0.0792, + "step": 12288 + }, + { + "epoch": 1.61, + "learning_rate": 1.850159652391376e-05, + "loss": 0.0746, + "step": 12289 + }, + { + "epoch": 1.61, + "learning_rate": 1.8501317282558934e-05, + "loss": 0.0769, + "step": 12290 + }, + { + "epoch": 1.61, + "learning_rate": 1.8501038017294602e-05, + "loss": 0.0782, + "step": 12291 + }, + { + "epoch": 1.61, + "learning_rate": 1.8500758728121555e-05, + "loss": 0.0796, + "step": 12292 + }, + { + "epoch": 1.61, + "learning_rate": 1.850047941504058e-05, + "loss": 0.0702, + "step": 12293 + }, + { + "epoch": 1.61, + "learning_rate": 1.8500200078052456e-05, + "loss": 0.0783, + "step": 12294 + }, + { + "epoch": 1.61, + "learning_rate": 1.8499920717157972e-05, + "loss": 0.082, + "step": 12295 + }, + { + "epoch": 1.61, + "learning_rate": 1.8499641332357914e-05, + "loss": 0.08, + "step": 12296 + }, + { + "epoch": 1.61, + "learning_rate": 1.8499361923653065e-05, + "loss": 0.0692, + "step": 12297 + }, + { + "epoch": 1.61, + "learning_rate": 1.8499082491044218e-05, + "loss": 0.0681, + "step": 12298 + }, + { + "epoch": 1.61, + "learning_rate": 1.8498803034532152e-05, + "loss": 0.0754, + "step": 12299 + }, + { + "epoch": 1.61, + "learning_rate": 1.8498523554117653e-05, + "loss": 0.0801, + "step": 12300 + }, + { + "epoch": 1.61, + "learning_rate": 1.849824404980151e-05, + "loss": 0.0767, + "step": 12301 + }, + { + "epoch": 1.61, + "learning_rate": 1.849796452158451e-05, + "loss": 0.0851, + "step": 12302 + }, + { + "epoch": 1.61, + "learning_rate": 1.8497684969467437e-05, + "loss": 0.0843, + "step": 12303 + }, + { + "epoch": 1.61, + "learning_rate": 1.8497405393451077e-05, + "loss": 0.0846, + "step": 12304 + }, + { + "epoch": 1.61, + "learning_rate": 1.8497125793536218e-05, + "loss": 0.0759, + "step": 12305 + }, + { + "epoch": 1.61, + "learning_rate": 1.8496846169723642e-05, + "loss": 0.0731, + "step": 12306 + }, + { + "epoch": 1.61, + "learning_rate": 1.8496566522014142e-05, + "loss": 0.0762, + "step": 12307 + }, + { + "epoch": 1.61, + "learning_rate": 1.8496286850408502e-05, + "loss": 0.0737, + "step": 12308 + }, + { + "epoch": 1.61, + "learning_rate": 1.8496007154907503e-05, + "loss": 0.0709, + "step": 12309 + }, + { + "epoch": 1.61, + "learning_rate": 1.8495727435511938e-05, + "loss": 0.0627, + "step": 12310 + }, + { + "epoch": 1.61, + "learning_rate": 1.8495447692222593e-05, + "loss": 0.0702, + "step": 12311 + }, + { + "epoch": 1.61, + "learning_rate": 1.8495167925040255e-05, + "loss": 0.0787, + "step": 12312 + }, + { + "epoch": 1.61, + "learning_rate": 1.8494888133965706e-05, + "loss": 0.0799, + "step": 12313 + }, + { + "epoch": 1.61, + "learning_rate": 1.849460831899974e-05, + "loss": 0.0729, + "step": 12314 + }, + { + "epoch": 1.61, + "learning_rate": 1.8494328480143137e-05, + "loss": 0.0737, + "step": 12315 + }, + { + "epoch": 1.61, + "learning_rate": 1.849404861739669e-05, + "loss": 0.0842, + "step": 12316 + }, + { + "epoch": 1.61, + "learning_rate": 1.8493768730761187e-05, + "loss": 0.0701, + "step": 12317 + }, + { + "epoch": 1.61, + "learning_rate": 1.8493488820237404e-05, + "loss": 0.076, + "step": 12318 + }, + { + "epoch": 1.61, + "learning_rate": 1.849320888582614e-05, + "loss": 0.0732, + "step": 12319 + }, + { + "epoch": 1.61, + "learning_rate": 1.849292892752818e-05, + "loss": 0.0852, + "step": 12320 + }, + { + "epoch": 1.61, + "learning_rate": 1.849264894534431e-05, + "loss": 0.0886, + "step": 12321 + }, + { + "epoch": 1.61, + "learning_rate": 1.849236893927531e-05, + "loss": 0.0801, + "step": 12322 + }, + { + "epoch": 1.61, + "learning_rate": 1.8492088909321984e-05, + "loss": 0.0771, + "step": 12323 + }, + { + "epoch": 1.61, + "learning_rate": 1.8491808855485103e-05, + "loss": 0.0859, + "step": 12324 + }, + { + "epoch": 1.61, + "learning_rate": 1.8491528777765467e-05, + "loss": 0.068, + "step": 12325 + }, + { + "epoch": 1.61, + "learning_rate": 1.8491248676163852e-05, + "loss": 0.0782, + "step": 12326 + }, + { + "epoch": 1.61, + "learning_rate": 1.8490968550681058e-05, + "loss": 0.08, + "step": 12327 + }, + { + "epoch": 1.61, + "learning_rate": 1.8490688401317867e-05, + "loss": 0.079, + "step": 12328 + }, + { + "epoch": 1.62, + "learning_rate": 1.8490408228075066e-05, + "loss": 0.0624, + "step": 12329 + }, + { + "epoch": 1.62, + "learning_rate": 1.849012803095344e-05, + "loss": 0.0756, + "step": 12330 + }, + { + "epoch": 1.62, + "learning_rate": 1.8489847809953788e-05, + "loss": 0.07, + "step": 12331 + }, + { + "epoch": 1.62, + "learning_rate": 1.848956756507689e-05, + "loss": 0.0737, + "step": 12332 + }, + { + "epoch": 1.62, + "learning_rate": 1.848928729632353e-05, + "loss": 0.0778, + "step": 12333 + }, + { + "epoch": 1.62, + "learning_rate": 1.8489007003694505e-05, + "loss": 0.0791, + "step": 12334 + }, + { + "epoch": 1.62, + "learning_rate": 1.84887266871906e-05, + "loss": 0.0779, + "step": 12335 + }, + { + "epoch": 1.62, + "learning_rate": 1.8488446346812602e-05, + "loss": 0.0715, + "step": 12336 + }, + { + "epoch": 1.62, + "learning_rate": 1.84881659825613e-05, + "loss": 0.0769, + "step": 12337 + }, + { + "epoch": 1.62, + "learning_rate": 1.8487885594437485e-05, + "loss": 0.0823, + "step": 12338 + }, + { + "epoch": 1.62, + "learning_rate": 1.8487605182441945e-05, + "loss": 0.0708, + "step": 12339 + }, + { + "epoch": 1.62, + "learning_rate": 1.8487324746575464e-05, + "loss": 0.0773, + "step": 12340 + }, + { + "epoch": 1.62, + "learning_rate": 1.848704428683884e-05, + "loss": 0.0712, + "step": 12341 + }, + { + "epoch": 1.62, + "learning_rate": 1.848676380323285e-05, + "loss": 0.0852, + "step": 12342 + }, + { + "epoch": 1.62, + "learning_rate": 1.8486483295758288e-05, + "loss": 0.0765, + "step": 12343 + }, + { + "epoch": 1.62, + "learning_rate": 1.8486202764415947e-05, + "loss": 0.0718, + "step": 12344 + }, + { + "epoch": 1.62, + "learning_rate": 1.8485922209206613e-05, + "loss": 0.0767, + "step": 12345 + }, + { + "epoch": 1.62, + "learning_rate": 1.8485641630131073e-05, + "loss": 0.0714, + "step": 12346 + }, + { + "epoch": 1.62, + "learning_rate": 1.8485361027190118e-05, + "loss": 0.0823, + "step": 12347 + }, + { + "epoch": 1.62, + "learning_rate": 1.8485080400384538e-05, + "loss": 0.0767, + "step": 12348 + }, + { + "epoch": 1.62, + "learning_rate": 1.848479974971512e-05, + "loss": 0.0847, + "step": 12349 + }, + { + "epoch": 1.62, + "learning_rate": 1.8484519075182656e-05, + "loss": 0.0677, + "step": 12350 + }, + { + "epoch": 1.62, + "learning_rate": 1.848423837678793e-05, + "loss": 0.0873, + "step": 12351 + }, + { + "epoch": 1.62, + "learning_rate": 1.848395765453174e-05, + "loss": 0.0691, + "step": 12352 + }, + { + "epoch": 1.62, + "learning_rate": 1.848367690841487e-05, + "loss": 0.0831, + "step": 12353 + }, + { + "epoch": 1.62, + "learning_rate": 1.848339613843811e-05, + "loss": 0.0824, + "step": 12354 + }, + { + "epoch": 1.62, + "learning_rate": 1.848311534460225e-05, + "loss": 0.0757, + "step": 12355 + }, + { + "epoch": 1.62, + "learning_rate": 1.848283452690808e-05, + "loss": 0.0755, + "step": 12356 + }, + { + "epoch": 1.62, + "learning_rate": 1.848255368535639e-05, + "loss": 0.0669, + "step": 12357 + }, + { + "epoch": 1.62, + "learning_rate": 1.848227281994797e-05, + "loss": 0.074, + "step": 12358 + }, + { + "epoch": 1.62, + "learning_rate": 1.8481991930683606e-05, + "loss": 0.0721, + "step": 12359 + }, + { + "epoch": 1.62, + "learning_rate": 1.8481711017564095e-05, + "loss": 0.0828, + "step": 12360 + }, + { + "epoch": 1.62, + "learning_rate": 1.8481430080590222e-05, + "loss": 0.0728, + "step": 12361 + }, + { + "epoch": 1.62, + "learning_rate": 1.8481149119762777e-05, + "loss": 0.0645, + "step": 12362 + }, + { + "epoch": 1.62, + "learning_rate": 1.8480868135082555e-05, + "loss": 0.0778, + "step": 12363 + }, + { + "epoch": 1.62, + "learning_rate": 1.8480587126550343e-05, + "loss": 0.0816, + "step": 12364 + }, + { + "epoch": 1.62, + "learning_rate": 1.848030609416693e-05, + "loss": 0.0778, + "step": 12365 + }, + { + "epoch": 1.62, + "learning_rate": 1.8480025037933104e-05, + "loss": 0.0772, + "step": 12366 + }, + { + "epoch": 1.62, + "learning_rate": 1.8479743957849664e-05, + "loss": 0.0866, + "step": 12367 + }, + { + "epoch": 1.62, + "learning_rate": 1.8479462853917395e-05, + "loss": 0.0653, + "step": 12368 + }, + { + "epoch": 1.62, + "learning_rate": 1.847918172613709e-05, + "loss": 0.0837, + "step": 12369 + }, + { + "epoch": 1.62, + "learning_rate": 1.8478900574509533e-05, + "loss": 0.066, + "step": 12370 + }, + { + "epoch": 1.62, + "learning_rate": 1.847861939903552e-05, + "loss": 0.0826, + "step": 12371 + }, + { + "epoch": 1.62, + "learning_rate": 1.8478338199715847e-05, + "loss": 0.078, + "step": 12372 + }, + { + "epoch": 1.62, + "learning_rate": 1.8478056976551296e-05, + "loss": 0.0929, + "step": 12373 + }, + { + "epoch": 1.62, + "learning_rate": 1.847777572954266e-05, + "loss": 0.0806, + "step": 12374 + }, + { + "epoch": 1.62, + "learning_rate": 1.8477494458690734e-05, + "loss": 0.0721, + "step": 12375 + }, + { + "epoch": 1.62, + "learning_rate": 1.84772131639963e-05, + "loss": 0.0742, + "step": 12376 + }, + { + "epoch": 1.62, + "learning_rate": 1.8476931845460164e-05, + "loss": 0.0711, + "step": 12377 + }, + { + "epoch": 1.62, + "learning_rate": 1.8476650503083105e-05, + "loss": 0.0793, + "step": 12378 + }, + { + "epoch": 1.62, + "learning_rate": 1.8476369136865918e-05, + "loss": 0.0757, + "step": 12379 + }, + { + "epoch": 1.62, + "learning_rate": 1.8476087746809392e-05, + "loss": 0.0772, + "step": 12380 + }, + { + "epoch": 1.62, + "learning_rate": 1.8475806332914323e-05, + "loss": 0.0696, + "step": 12381 + }, + { + "epoch": 1.62, + "learning_rate": 1.84755248951815e-05, + "loss": 0.0844, + "step": 12382 + }, + { + "epoch": 1.62, + "learning_rate": 1.8475243433611714e-05, + "loss": 0.0752, + "step": 12383 + }, + { + "epoch": 1.62, + "learning_rate": 1.8474961948205756e-05, + "loss": 0.0773, + "step": 12384 + }, + { + "epoch": 1.62, + "learning_rate": 1.847468043896442e-05, + "loss": 0.0761, + "step": 12385 + }, + { + "epoch": 1.62, + "learning_rate": 1.8474398905888498e-05, + "loss": 0.0741, + "step": 12386 + }, + { + "epoch": 1.62, + "learning_rate": 1.847411734897878e-05, + "loss": 0.0747, + "step": 12387 + }, + { + "epoch": 1.62, + "learning_rate": 1.8473835768236056e-05, + "loss": 0.0821, + "step": 12388 + }, + { + "epoch": 1.62, + "learning_rate": 1.8473554163661124e-05, + "loss": 0.0714, + "step": 12389 + }, + { + "epoch": 1.62, + "learning_rate": 1.8473272535254768e-05, + "loss": 0.0723, + "step": 12390 + }, + { + "epoch": 1.62, + "learning_rate": 1.8472990883017784e-05, + "loss": 0.0753, + "step": 12391 + }, + { + "epoch": 1.62, + "learning_rate": 1.8472709206950967e-05, + "loss": 0.0785, + "step": 12392 + }, + { + "epoch": 1.62, + "learning_rate": 1.8472427507055105e-05, + "loss": 0.0816, + "step": 12393 + }, + { + "epoch": 1.62, + "learning_rate": 1.8472145783330993e-05, + "loss": 0.0757, + "step": 12394 + }, + { + "epoch": 1.62, + "learning_rate": 1.847186403577942e-05, + "loss": 0.0896, + "step": 12395 + }, + { + "epoch": 1.62, + "learning_rate": 1.847158226440118e-05, + "loss": 0.0775, + "step": 12396 + }, + { + "epoch": 1.62, + "learning_rate": 1.8471300469197073e-05, + "loss": 0.0783, + "step": 12397 + }, + { + "epoch": 1.62, + "learning_rate": 1.8471018650167876e-05, + "loss": 0.0766, + "step": 12398 + }, + { + "epoch": 1.62, + "learning_rate": 1.8470736807314398e-05, + "loss": 0.0611, + "step": 12399 + }, + { + "epoch": 1.62, + "learning_rate": 1.8470454940637416e-05, + "loss": 0.0707, + "step": 12400 + }, + { + "epoch": 1.62, + "learning_rate": 1.8470173050137738e-05, + "loss": 0.0813, + "step": 12401 + }, + { + "epoch": 1.62, + "learning_rate": 1.8469891135816146e-05, + "loss": 0.0687, + "step": 12402 + }, + { + "epoch": 1.62, + "learning_rate": 1.8469609197673436e-05, + "loss": 0.0892, + "step": 12403 + }, + { + "epoch": 1.62, + "learning_rate": 1.84693272357104e-05, + "loss": 0.0759, + "step": 12404 + }, + { + "epoch": 1.62, + "learning_rate": 1.8469045249927834e-05, + "loss": 0.0786, + "step": 12405 + }, + { + "epoch": 1.63, + "learning_rate": 1.8468763240326528e-05, + "loss": 0.0694, + "step": 12406 + }, + { + "epoch": 1.63, + "learning_rate": 1.8468481206907275e-05, + "loss": 0.0797, + "step": 12407 + }, + { + "epoch": 1.63, + "learning_rate": 1.8468199149670872e-05, + "loss": 0.0786, + "step": 12408 + }, + { + "epoch": 1.63, + "learning_rate": 1.846791706861811e-05, + "loss": 0.0851, + "step": 12409 + }, + { + "epoch": 1.63, + "learning_rate": 1.846763496374978e-05, + "loss": 0.0875, + "step": 12410 + }, + { + "epoch": 1.63, + "learning_rate": 1.846735283506668e-05, + "loss": 0.0802, + "step": 12411 + }, + { + "epoch": 1.63, + "learning_rate": 1.84670706825696e-05, + "loss": 0.0726, + "step": 12412 + }, + { + "epoch": 1.63, + "learning_rate": 1.8466788506259336e-05, + "loss": 0.0689, + "step": 12413 + }, + { + "epoch": 1.63, + "learning_rate": 1.8466506306136677e-05, + "loss": 0.0662, + "step": 12414 + }, + { + "epoch": 1.63, + "learning_rate": 1.846622408220242e-05, + "loss": 0.0742, + "step": 12415 + }, + { + "epoch": 1.63, + "learning_rate": 1.8465941834457362e-05, + "loss": 0.0732, + "step": 12416 + }, + { + "epoch": 1.63, + "learning_rate": 1.8465659562902294e-05, + "loss": 0.0745, + "step": 12417 + }, + { + "epoch": 1.63, + "learning_rate": 1.8465377267538004e-05, + "loss": 0.065, + "step": 12418 + }, + { + "epoch": 1.63, + "learning_rate": 1.8465094948365295e-05, + "loss": 0.0777, + "step": 12419 + }, + { + "epoch": 1.63, + "learning_rate": 1.846481260538496e-05, + "loss": 0.0729, + "step": 12420 + }, + { + "epoch": 1.63, + "learning_rate": 1.8464530238597785e-05, + "loss": 0.088, + "step": 12421 + }, + { + "epoch": 1.63, + "learning_rate": 1.846424784800457e-05, + "loss": 0.07, + "step": 12422 + }, + { + "epoch": 1.63, + "learning_rate": 1.8463965433606112e-05, + "loss": 0.075, + "step": 12423 + }, + { + "epoch": 1.63, + "learning_rate": 1.84636829954032e-05, + "loss": 0.0753, + "step": 12424 + }, + { + "epoch": 1.63, + "learning_rate": 1.846340053339663e-05, + "loss": 0.0819, + "step": 12425 + }, + { + "epoch": 1.63, + "learning_rate": 1.8463118047587197e-05, + "loss": 0.0806, + "step": 12426 + }, + { + "epoch": 1.63, + "learning_rate": 1.8462835537975695e-05, + "loss": 0.0696, + "step": 12427 + }, + { + "epoch": 1.63, + "learning_rate": 1.846255300456292e-05, + "loss": 0.0765, + "step": 12428 + }, + { + "epoch": 1.63, + "learning_rate": 1.8462270447349662e-05, + "loss": 0.0632, + "step": 12429 + }, + { + "epoch": 1.63, + "learning_rate": 1.8461987866336725e-05, + "loss": 0.071, + "step": 12430 + }, + { + "epoch": 1.63, + "learning_rate": 1.8461705261524892e-05, + "loss": 0.0859, + "step": 12431 + }, + { + "epoch": 1.63, + "learning_rate": 1.8461422632914967e-05, + "loss": 0.0735, + "step": 12432 + }, + { + "epoch": 1.63, + "learning_rate": 1.846113998050774e-05, + "loss": 0.0803, + "step": 12433 + }, + { + "epoch": 1.63, + "learning_rate": 1.8460857304304007e-05, + "loss": 0.0726, + "step": 12434 + }, + { + "epoch": 1.63, + "learning_rate": 1.8460574604304563e-05, + "loss": 0.0894, + "step": 12435 + }, + { + "epoch": 1.63, + "learning_rate": 1.8460291880510206e-05, + "loss": 0.0807, + "step": 12436 + }, + { + "epoch": 1.63, + "learning_rate": 1.8460009132921727e-05, + "loss": 0.0765, + "step": 12437 + }, + { + "epoch": 1.63, + "learning_rate": 1.845972636153992e-05, + "loss": 0.0676, + "step": 12438 + }, + { + "epoch": 1.63, + "learning_rate": 1.8459443566365588e-05, + "loss": 0.0802, + "step": 12439 + }, + { + "epoch": 1.63, + "learning_rate": 1.845916074739952e-05, + "loss": 0.0794, + "step": 12440 + }, + { + "epoch": 1.63, + "learning_rate": 1.8458877904642508e-05, + "loss": 0.0801, + "step": 12441 + }, + { + "epoch": 1.63, + "learning_rate": 1.845859503809536e-05, + "loss": 0.0696, + "step": 12442 + }, + { + "epoch": 1.63, + "learning_rate": 1.8458312147758862e-05, + "loss": 0.0879, + "step": 12443 + }, + { + "epoch": 1.63, + "learning_rate": 1.8458029233633806e-05, + "loss": 0.0771, + "step": 12444 + }, + { + "epoch": 1.63, + "learning_rate": 1.8457746295720998e-05, + "loss": 0.0796, + "step": 12445 + }, + { + "epoch": 1.63, + "learning_rate": 1.8457463334021232e-05, + "loss": 0.075, + "step": 12446 + }, + { + "epoch": 1.63, + "learning_rate": 1.8457180348535293e-05, + "loss": 0.0659, + "step": 12447 + }, + { + "epoch": 1.63, + "learning_rate": 1.845689733926399e-05, + "loss": 0.0777, + "step": 12448 + }, + { + "epoch": 1.63, + "learning_rate": 1.8456614306208113e-05, + "loss": 0.074, + "step": 12449 + }, + { + "epoch": 1.63, + "learning_rate": 1.8456331249368458e-05, + "loss": 0.081, + "step": 12450 + }, + { + "epoch": 1.63, + "learning_rate": 1.845604816874582e-05, + "loss": 0.0793, + "step": 12451 + }, + { + "epoch": 1.63, + "learning_rate": 1.8455765064341e-05, + "loss": 0.0788, + "step": 12452 + }, + { + "epoch": 1.63, + "learning_rate": 1.845548193615479e-05, + "loss": 0.0798, + "step": 12453 + }, + { + "epoch": 1.63, + "learning_rate": 1.8455198784187985e-05, + "loss": 0.0666, + "step": 12454 + }, + { + "epoch": 1.63, + "learning_rate": 1.8454915608441384e-05, + "loss": 0.0726, + "step": 12455 + }, + { + "epoch": 1.63, + "learning_rate": 1.8454632408915786e-05, + "loss": 0.0736, + "step": 12456 + }, + { + "epoch": 1.63, + "learning_rate": 1.8454349185611984e-05, + "loss": 0.0815, + "step": 12457 + }, + { + "epoch": 1.63, + "learning_rate": 1.8454065938530772e-05, + "loss": 0.0824, + "step": 12458 + }, + { + "epoch": 1.63, + "learning_rate": 1.8453782667672952e-05, + "loss": 0.0833, + "step": 12459 + }, + { + "epoch": 1.63, + "learning_rate": 1.845349937303932e-05, + "loss": 0.0813, + "step": 12460 + }, + { + "epoch": 1.63, + "learning_rate": 1.8453216054630666e-05, + "loss": 0.0656, + "step": 12461 + }, + { + "epoch": 1.63, + "learning_rate": 1.84529327124478e-05, + "loss": 0.0805, + "step": 12462 + }, + { + "epoch": 1.63, + "learning_rate": 1.8452649346491503e-05, + "loss": 0.0605, + "step": 12463 + }, + { + "epoch": 1.63, + "learning_rate": 1.8452365956762587e-05, + "loss": 0.0699, + "step": 12464 + }, + { + "epoch": 1.63, + "learning_rate": 1.8452082543261836e-05, + "loss": 0.0747, + "step": 12465 + }, + { + "epoch": 1.63, + "learning_rate": 1.8451799105990055e-05, + "loss": 0.0875, + "step": 12466 + }, + { + "epoch": 1.63, + "learning_rate": 1.845151564494804e-05, + "loss": 0.0844, + "step": 12467 + }, + { + "epoch": 1.63, + "learning_rate": 1.8451232160136588e-05, + "loss": 0.0754, + "step": 12468 + }, + { + "epoch": 1.63, + "learning_rate": 1.845094865155649e-05, + "loss": 0.0749, + "step": 12469 + }, + { + "epoch": 1.63, + "learning_rate": 1.8450665119208554e-05, + "loss": 0.0798, + "step": 12470 + }, + { + "epoch": 1.63, + "learning_rate": 1.8450381563093575e-05, + "loss": 0.0886, + "step": 12471 + }, + { + "epoch": 1.63, + "learning_rate": 1.845009798321234e-05, + "loss": 0.0828, + "step": 12472 + }, + { + "epoch": 1.63, + "learning_rate": 1.844981437956566e-05, + "loss": 0.0822, + "step": 12473 + }, + { + "epoch": 1.63, + "learning_rate": 1.8449530752154326e-05, + "loss": 0.0708, + "step": 12474 + }, + { + "epoch": 1.63, + "learning_rate": 1.8449247100979136e-05, + "loss": 0.0736, + "step": 12475 + }, + { + "epoch": 1.63, + "learning_rate": 1.844896342604089e-05, + "loss": 0.0733, + "step": 12476 + }, + { + "epoch": 1.63, + "learning_rate": 1.8448679727340383e-05, + "loss": 0.0875, + "step": 12477 + }, + { + "epoch": 1.63, + "learning_rate": 1.8448396004878415e-05, + "loss": 0.081, + "step": 12478 + }, + { + "epoch": 1.63, + "learning_rate": 1.8448112258655785e-05, + "loss": 0.0741, + "step": 12479 + }, + { + "epoch": 1.63, + "learning_rate": 1.8447828488673283e-05, + "loss": 0.0839, + "step": 12480 + }, + { + "epoch": 1.63, + "learning_rate": 1.8447544694931718e-05, + "loss": 0.0752, + "step": 12481 + }, + { + "epoch": 1.64, + "learning_rate": 1.844726087743188e-05, + "loss": 0.0797, + "step": 12482 + }, + { + "epoch": 1.64, + "learning_rate": 1.8446977036174577e-05, + "loss": 0.0795, + "step": 12483 + }, + { + "epoch": 1.64, + "learning_rate": 1.8446693171160593e-05, + "loss": 0.0768, + "step": 12484 + }, + { + "epoch": 1.64, + "learning_rate": 1.844640928239074e-05, + "loss": 0.083, + "step": 12485 + }, + { + "epoch": 1.64, + "learning_rate": 1.8446125369865808e-05, + "loss": 0.0829, + "step": 12486 + }, + { + "epoch": 1.64, + "learning_rate": 1.84458414335866e-05, + "loss": 0.073, + "step": 12487 + }, + { + "epoch": 1.64, + "learning_rate": 1.8445557473553912e-05, + "loss": 0.0748, + "step": 12488 + }, + { + "epoch": 1.64, + "learning_rate": 1.8445273489768545e-05, + "loss": 0.086, + "step": 12489 + }, + { + "epoch": 1.64, + "learning_rate": 1.8444989482231294e-05, + "loss": 0.0778, + "step": 12490 + }, + { + "epoch": 1.64, + "learning_rate": 1.844470545094296e-05, + "loss": 0.0832, + "step": 12491 + }, + { + "epoch": 1.64, + "learning_rate": 1.844442139590434e-05, + "loss": 0.084, + "step": 12492 + }, + { + "epoch": 1.64, + "learning_rate": 1.8444137317116237e-05, + "loss": 0.084, + "step": 12493 + }, + { + "epoch": 1.64, + "learning_rate": 1.844385321457945e-05, + "loss": 0.0812, + "step": 12494 + }, + { + "epoch": 1.64, + "learning_rate": 1.844356908829477e-05, + "loss": 0.0906, + "step": 12495 + }, + { + "epoch": 1.64, + "learning_rate": 1.8443284938263003e-05, + "loss": 0.0618, + "step": 12496 + }, + { + "epoch": 1.64, + "learning_rate": 1.8443000764484948e-05, + "loss": 0.082, + "step": 12497 + }, + { + "epoch": 1.64, + "learning_rate": 1.8442716566961404e-05, + "loss": 0.0781, + "step": 12498 + }, + { + "epoch": 1.64, + "learning_rate": 1.8442432345693166e-05, + "loss": 0.0812, + "step": 12499 + }, + { + "epoch": 1.64, + "learning_rate": 1.844214810068104e-05, + "loss": 0.0753, + "step": 12500 + }, + { + "epoch": 1.64, + "learning_rate": 1.8441863831925816e-05, + "loss": 0.0719, + "step": 12501 + }, + { + "epoch": 1.64, + "learning_rate": 1.8441579539428307e-05, + "loss": 0.0804, + "step": 12502 + }, + { + "epoch": 1.64, + "learning_rate": 1.84412952231893e-05, + "loss": 0.0913, + "step": 12503 + }, + { + "epoch": 1.64, + "learning_rate": 1.84410108832096e-05, + "loss": 0.0809, + "step": 12504 + }, + { + "epoch": 1.64, + "learning_rate": 1.844072651949001e-05, + "loss": 0.0721, + "step": 12505 + }, + { + "epoch": 1.64, + "learning_rate": 1.8440442132031324e-05, + "loss": 0.0759, + "step": 12506 + }, + { + "epoch": 1.64, + "learning_rate": 1.8440157720834344e-05, + "loss": 0.0817, + "step": 12507 + }, + { + "epoch": 1.64, + "learning_rate": 1.8439873285899873e-05, + "loss": 0.0835, + "step": 12508 + }, + { + "epoch": 1.64, + "learning_rate": 1.8439588827228703e-05, + "loss": 0.0855, + "step": 12509 + }, + { + "epoch": 1.64, + "learning_rate": 1.843930434482164e-05, + "loss": 0.081, + "step": 12510 + }, + { + "epoch": 1.64, + "learning_rate": 1.8439019838679484e-05, + "loss": 0.0783, + "step": 12511 + }, + { + "epoch": 1.64, + "learning_rate": 1.8438735308803035e-05, + "loss": 0.0767, + "step": 12512 + }, + { + "epoch": 1.64, + "learning_rate": 1.843845075519309e-05, + "loss": 0.0715, + "step": 12513 + }, + { + "epoch": 1.64, + "learning_rate": 1.8438166177850454e-05, + "loss": 0.0748, + "step": 12514 + }, + { + "epoch": 1.64, + "learning_rate": 1.8437881576775925e-05, + "loss": 0.0694, + "step": 12515 + }, + { + "epoch": 1.64, + "learning_rate": 1.84375969519703e-05, + "loss": 0.078, + "step": 12516 + }, + { + "epoch": 1.64, + "learning_rate": 1.8437312303434385e-05, + "loss": 0.0894, + "step": 12517 + }, + { + "epoch": 1.64, + "learning_rate": 1.8437027631168976e-05, + "loss": 0.0747, + "step": 12518 + }, + { + "epoch": 1.64, + "learning_rate": 1.843674293517488e-05, + "loss": 0.0741, + "step": 12519 + }, + { + "epoch": 1.64, + "learning_rate": 1.8436458215452893e-05, + "loss": 0.0762, + "step": 12520 + }, + { + "epoch": 1.64, + "learning_rate": 1.8436173472003815e-05, + "loss": 0.0863, + "step": 12521 + }, + { + "epoch": 1.64, + "learning_rate": 1.843588870482845e-05, + "loss": 0.07, + "step": 12522 + }, + { + "epoch": 1.64, + "learning_rate": 1.8435603913927596e-05, + "loss": 0.0808, + "step": 12523 + }, + { + "epoch": 1.64, + "learning_rate": 1.8435319099302058e-05, + "loss": 0.0693, + "step": 12524 + }, + { + "epoch": 1.64, + "learning_rate": 1.8435034260952635e-05, + "loss": 0.0834, + "step": 12525 + }, + { + "epoch": 1.64, + "learning_rate": 1.843474939888012e-05, + "loss": 0.0759, + "step": 12526 + }, + { + "epoch": 1.64, + "learning_rate": 1.8434464513085325e-05, + "loss": 0.0809, + "step": 12527 + }, + { + "epoch": 1.64, + "learning_rate": 1.843417960356905e-05, + "loss": 0.0733, + "step": 12528 + }, + { + "epoch": 1.64, + "learning_rate": 1.8433894670332093e-05, + "loss": 0.0755, + "step": 12529 + }, + { + "epoch": 1.64, + "learning_rate": 1.843360971337526e-05, + "loss": 0.0828, + "step": 12530 + }, + { + "epoch": 1.64, + "learning_rate": 1.843332473269934e-05, + "loss": 0.0699, + "step": 12531 + }, + { + "epoch": 1.64, + "learning_rate": 1.843303972830515e-05, + "loss": 0.079, + "step": 12532 + }, + { + "epoch": 1.64, + "learning_rate": 1.8432754700193482e-05, + "loss": 0.0728, + "step": 12533 + }, + { + "epoch": 1.64, + "learning_rate": 1.8432469648365142e-05, + "loss": 0.0664, + "step": 12534 + }, + { + "epoch": 1.64, + "learning_rate": 1.843218457282093e-05, + "loss": 0.0692, + "step": 12535 + }, + { + "epoch": 1.64, + "learning_rate": 1.8431899473561646e-05, + "loss": 0.0759, + "step": 12536 + }, + { + "epoch": 1.64, + "learning_rate": 1.8431614350588095e-05, + "loss": 0.0661, + "step": 12537 + }, + { + "epoch": 1.64, + "learning_rate": 1.8431329203901078e-05, + "loss": 0.0748, + "step": 12538 + }, + { + "epoch": 1.64, + "learning_rate": 1.8431044033501394e-05, + "loss": 0.0774, + "step": 12539 + }, + { + "epoch": 1.64, + "learning_rate": 1.8430758839389852e-05, + "loss": 0.0671, + "step": 12540 + }, + { + "epoch": 1.64, + "learning_rate": 1.843047362156725e-05, + "loss": 0.0833, + "step": 12541 + }, + { + "epoch": 1.64, + "learning_rate": 1.8430188380034385e-05, + "loss": 0.0783, + "step": 12542 + }, + { + "epoch": 1.64, + "learning_rate": 1.842990311479207e-05, + "loss": 0.0728, + "step": 12543 + }, + { + "epoch": 1.64, + "learning_rate": 1.8429617825841095e-05, + "loss": 0.0832, + "step": 12544 + }, + { + "epoch": 1.64, + "learning_rate": 1.842933251318227e-05, + "loss": 0.0721, + "step": 12545 + }, + { + "epoch": 1.64, + "learning_rate": 1.84290471768164e-05, + "loss": 0.073, + "step": 12546 + }, + { + "epoch": 1.64, + "learning_rate": 1.8428761816744285e-05, + "loss": 0.0733, + "step": 12547 + }, + { + "epoch": 1.64, + "learning_rate": 1.842847643296672e-05, + "loss": 0.0662, + "step": 12548 + }, + { + "epoch": 1.64, + "learning_rate": 1.8428191025484517e-05, + "loss": 0.0791, + "step": 12549 + }, + { + "epoch": 1.64, + "learning_rate": 1.842790559429848e-05, + "loss": 0.0855, + "step": 12550 + }, + { + "epoch": 1.64, + "learning_rate": 1.84276201394094e-05, + "loss": 0.0584, + "step": 12551 + }, + { + "epoch": 1.64, + "learning_rate": 1.842733466081809e-05, + "loss": 0.0713, + "step": 12552 + }, + { + "epoch": 1.64, + "learning_rate": 1.8427049158525352e-05, + "loss": 0.0745, + "step": 12553 + }, + { + "epoch": 1.64, + "learning_rate": 1.8426763632531984e-05, + "loss": 0.0802, + "step": 12554 + }, + { + "epoch": 1.64, + "learning_rate": 1.8426478082838796e-05, + "loss": 0.0677, + "step": 12555 + }, + { + "epoch": 1.64, + "learning_rate": 1.8426192509446582e-05, + "loss": 0.0784, + "step": 12556 + }, + { + "epoch": 1.64, + "learning_rate": 1.842590691235615e-05, + "loss": 0.0677, + "step": 12557 + }, + { + "epoch": 1.65, + "learning_rate": 1.8425621291568306e-05, + "loss": 0.0891, + "step": 12558 + }, + { + "epoch": 1.65, + "learning_rate": 1.8425335647083854e-05, + "loss": 0.0689, + "step": 12559 + }, + { + "epoch": 1.65, + "learning_rate": 1.8425049978903588e-05, + "loss": 0.081, + "step": 12560 + }, + { + "epoch": 1.65, + "learning_rate": 1.8424764287028324e-05, + "loss": 0.0704, + "step": 12561 + }, + { + "epoch": 1.65, + "learning_rate": 1.8424478571458855e-05, + "loss": 0.083, + "step": 12562 + }, + { + "epoch": 1.65, + "learning_rate": 1.842419283219599e-05, + "loss": 0.0813, + "step": 12563 + }, + { + "epoch": 1.65, + "learning_rate": 1.842390706924053e-05, + "loss": 0.0776, + "step": 12564 + }, + { + "epoch": 1.65, + "learning_rate": 1.842362128259328e-05, + "loss": 0.0684, + "step": 12565 + }, + { + "epoch": 1.65, + "learning_rate": 1.8423335472255047e-05, + "loss": 0.0775, + "step": 12566 + }, + { + "epoch": 1.65, + "learning_rate": 1.8423049638226627e-05, + "loss": 0.0805, + "step": 12567 + }, + { + "epoch": 1.65, + "learning_rate": 1.842276378050883e-05, + "loss": 0.0769, + "step": 12568 + }, + { + "epoch": 1.65, + "learning_rate": 1.8422477899102457e-05, + "loss": 0.0781, + "step": 12569 + }, + { + "epoch": 1.65, + "learning_rate": 1.8422191994008315e-05, + "loss": 0.0618, + "step": 12570 + }, + { + "epoch": 1.65, + "learning_rate": 1.842190606522721e-05, + "loss": 0.0789, + "step": 12571 + }, + { + "epoch": 1.65, + "learning_rate": 1.8421620112759938e-05, + "loss": 0.0841, + "step": 12572 + }, + { + "epoch": 1.65, + "learning_rate": 1.842133413660731e-05, + "loss": 0.0787, + "step": 12573 + }, + { + "epoch": 1.65, + "learning_rate": 1.842104813677013e-05, + "loss": 0.0765, + "step": 12574 + }, + { + "epoch": 1.65, + "learning_rate": 1.8420762113249197e-05, + "loss": 0.0772, + "step": 12575 + }, + { + "epoch": 1.65, + "learning_rate": 1.842047606604532e-05, + "loss": 0.0806, + "step": 12576 + }, + { + "epoch": 1.65, + "learning_rate": 1.84201899951593e-05, + "loss": 0.0912, + "step": 12577 + }, + { + "epoch": 1.65, + "learning_rate": 1.841990390059195e-05, + "loss": 0.0701, + "step": 12578 + }, + { + "epoch": 1.65, + "learning_rate": 1.8419617782344064e-05, + "loss": 0.0739, + "step": 12579 + }, + { + "epoch": 1.65, + "learning_rate": 1.841933164041645e-05, + "loss": 0.0865, + "step": 12580 + }, + { + "epoch": 1.65, + "learning_rate": 1.8419045474809917e-05, + "loss": 0.0817, + "step": 12581 + }, + { + "epoch": 1.65, + "learning_rate": 1.8418759285525267e-05, + "loss": 0.0802, + "step": 12582 + }, + { + "epoch": 1.65, + "learning_rate": 1.8418473072563306e-05, + "loss": 0.0856, + "step": 12583 + }, + { + "epoch": 1.65, + "learning_rate": 1.8418186835924836e-05, + "loss": 0.0705, + "step": 12584 + }, + { + "epoch": 1.65, + "learning_rate": 1.8417900575610664e-05, + "loss": 0.0805, + "step": 12585 + }, + { + "epoch": 1.65, + "learning_rate": 1.8417614291621595e-05, + "loss": 0.0812, + "step": 12586 + }, + { + "epoch": 1.65, + "learning_rate": 1.8417327983958433e-05, + "loss": 0.0907, + "step": 12587 + }, + { + "epoch": 1.65, + "learning_rate": 1.8417041652621988e-05, + "loss": 0.0702, + "step": 12588 + }, + { + "epoch": 1.65, + "learning_rate": 1.8416755297613056e-05, + "loss": 0.071, + "step": 12589 + }, + { + "epoch": 1.65, + "learning_rate": 1.841646891893245e-05, + "loss": 0.0741, + "step": 12590 + }, + { + "epoch": 1.65, + "learning_rate": 1.8416182516580975e-05, + "loss": 0.08, + "step": 12591 + }, + { + "epoch": 1.65, + "learning_rate": 1.8415896090559437e-05, + "loss": 0.0715, + "step": 12592 + }, + { + "epoch": 1.65, + "learning_rate": 1.8415609640868636e-05, + "loss": 0.0708, + "step": 12593 + }, + { + "epoch": 1.65, + "learning_rate": 1.841532316750938e-05, + "loss": 0.079, + "step": 12594 + }, + { + "epoch": 1.65, + "learning_rate": 1.841503667048248e-05, + "loss": 0.0699, + "step": 12595 + }, + { + "epoch": 1.65, + "learning_rate": 1.8414750149788732e-05, + "loss": 0.0811, + "step": 12596 + }, + { + "epoch": 1.65, + "learning_rate": 1.8414463605428952e-05, + "loss": 0.084, + "step": 12597 + }, + { + "epoch": 1.65, + "learning_rate": 1.841417703740394e-05, + "loss": 0.0754, + "step": 12598 + }, + { + "epoch": 1.65, + "learning_rate": 1.8413890445714503e-05, + "loss": 0.0783, + "step": 12599 + }, + { + "epoch": 1.65, + "learning_rate": 1.8413603830361446e-05, + "loss": 0.0826, + "step": 12600 + }, + { + "epoch": 1.65, + "learning_rate": 1.8413317191345578e-05, + "loss": 0.0709, + "step": 12601 + }, + { + "epoch": 1.65, + "learning_rate": 1.8413030528667703e-05, + "loss": 0.0773, + "step": 12602 + }, + { + "epoch": 1.65, + "learning_rate": 1.8412743842328623e-05, + "loss": 0.074, + "step": 12603 + }, + { + "epoch": 1.65, + "learning_rate": 1.8412457132329155e-05, + "loss": 0.0713, + "step": 12604 + }, + { + "epoch": 1.65, + "learning_rate": 1.8412170398670098e-05, + "loss": 0.0616, + "step": 12605 + }, + { + "epoch": 1.65, + "learning_rate": 1.841188364135226e-05, + "loss": 0.0726, + "step": 12606 + }, + { + "epoch": 1.65, + "learning_rate": 1.8411596860376444e-05, + "loss": 0.0771, + "step": 12607 + }, + { + "epoch": 1.65, + "learning_rate": 1.8411310055743464e-05, + "loss": 0.0736, + "step": 12608 + }, + { + "epoch": 1.65, + "learning_rate": 1.841102322745412e-05, + "loss": 0.071, + "step": 12609 + }, + { + "epoch": 1.65, + "learning_rate": 1.8410736375509223e-05, + "loss": 0.0723, + "step": 12610 + }, + { + "epoch": 1.65, + "learning_rate": 1.8410449499909572e-05, + "loss": 0.0794, + "step": 12611 + }, + { + "epoch": 1.65, + "learning_rate": 1.8410162600655987e-05, + "loss": 0.0745, + "step": 12612 + }, + { + "epoch": 1.65, + "learning_rate": 1.8409875677749265e-05, + "loss": 0.0837, + "step": 12613 + }, + { + "epoch": 1.65, + "learning_rate": 1.8409588731190212e-05, + "loss": 0.0878, + "step": 12614 + }, + { + "epoch": 1.65, + "learning_rate": 1.8409301760979644e-05, + "loss": 0.0637, + "step": 12615 + }, + { + "epoch": 1.65, + "learning_rate": 1.840901476711836e-05, + "loss": 0.08, + "step": 12616 + }, + { + "epoch": 1.65, + "learning_rate": 1.8408727749607173e-05, + "loss": 0.085, + "step": 12617 + }, + { + "epoch": 1.65, + "learning_rate": 1.8408440708446882e-05, + "loss": 0.074, + "step": 12618 + }, + { + "epoch": 1.65, + "learning_rate": 1.84081536436383e-05, + "loss": 0.0705, + "step": 12619 + }, + { + "epoch": 1.65, + "learning_rate": 1.8407866555182234e-05, + "loss": 0.0686, + "step": 12620 + }, + { + "epoch": 1.65, + "learning_rate": 1.840757944307949e-05, + "loss": 0.0831, + "step": 12621 + }, + { + "epoch": 1.65, + "learning_rate": 1.8407292307330882e-05, + "loss": 0.084, + "step": 12622 + }, + { + "epoch": 1.65, + "learning_rate": 1.8407005147937207e-05, + "loss": 0.0935, + "step": 12623 + }, + { + "epoch": 1.65, + "learning_rate": 1.840671796489928e-05, + "loss": 0.0709, + "step": 12624 + }, + { + "epoch": 1.65, + "learning_rate": 1.8406430758217906e-05, + "loss": 0.0663, + "step": 12625 + }, + { + "epoch": 1.65, + "learning_rate": 1.8406143527893895e-05, + "loss": 0.0699, + "step": 12626 + }, + { + "epoch": 1.65, + "learning_rate": 1.840585627392805e-05, + "loss": 0.0812, + "step": 12627 + }, + { + "epoch": 1.65, + "learning_rate": 1.840556899632118e-05, + "loss": 0.0742, + "step": 12628 + }, + { + "epoch": 1.65, + "learning_rate": 1.84052816950741e-05, + "loss": 0.0712, + "step": 12629 + }, + { + "epoch": 1.65, + "learning_rate": 1.8404994370187612e-05, + "loss": 0.0825, + "step": 12630 + }, + { + "epoch": 1.65, + "learning_rate": 1.840470702166252e-05, + "loss": 0.0732, + "step": 12631 + }, + { + "epoch": 1.65, + "learning_rate": 1.8404419649499644e-05, + "loss": 0.0755, + "step": 12632 + }, + { + "epoch": 1.65, + "learning_rate": 1.840413225369978e-05, + "loss": 0.0694, + "step": 12633 + }, + { + "epoch": 1.65, + "learning_rate": 1.8403844834263745e-05, + "loss": 0.0791, + "step": 12634 + }, + { + "epoch": 1.66, + "learning_rate": 1.840355739119234e-05, + "loss": 0.074, + "step": 12635 + }, + { + "epoch": 1.66, + "learning_rate": 1.840326992448638e-05, + "loss": 0.0725, + "step": 12636 + }, + { + "epoch": 1.66, + "learning_rate": 1.840298243414667e-05, + "loss": 0.076, + "step": 12637 + }, + { + "epoch": 1.66, + "learning_rate": 1.840269492017402e-05, + "loss": 0.0713, + "step": 12638 + }, + { + "epoch": 1.66, + "learning_rate": 1.840240738256924e-05, + "loss": 0.0745, + "step": 12639 + }, + { + "epoch": 1.66, + "learning_rate": 1.8402119821333134e-05, + "loss": 0.0571, + "step": 12640 + }, + { + "epoch": 1.66, + "learning_rate": 1.840183223646651e-05, + "loss": 0.0801, + "step": 12641 + }, + { + "epoch": 1.66, + "learning_rate": 1.8401544627970187e-05, + "loss": 0.0851, + "step": 12642 + }, + { + "epoch": 1.66, + "learning_rate": 1.8401256995844966e-05, + "loss": 0.0816, + "step": 12643 + }, + { + "epoch": 1.66, + "learning_rate": 1.8400969340091657e-05, + "loss": 0.071, + "step": 12644 + }, + { + "epoch": 1.66, + "learning_rate": 1.8400681660711065e-05, + "loss": 0.0833, + "step": 12645 + }, + { + "epoch": 1.66, + "learning_rate": 1.8400393957704007e-05, + "loss": 0.0663, + "step": 12646 + }, + { + "epoch": 1.66, + "learning_rate": 1.8400106231071286e-05, + "loss": 0.0771, + "step": 12647 + }, + { + "epoch": 1.66, + "learning_rate": 1.8399818480813715e-05, + "loss": 0.0732, + "step": 12648 + }, + { + "epoch": 1.66, + "learning_rate": 1.83995307069321e-05, + "loss": 0.0667, + "step": 12649 + }, + { + "epoch": 1.66, + "learning_rate": 1.8399242909427254e-05, + "loss": 0.0727, + "step": 12650 + }, + { + "epoch": 1.66, + "learning_rate": 1.8398955088299984e-05, + "loss": 0.0861, + "step": 12651 + }, + { + "epoch": 1.66, + "learning_rate": 1.83986672435511e-05, + "loss": 0.0773, + "step": 12652 + }, + { + "epoch": 1.66, + "learning_rate": 1.8398379375181415e-05, + "loss": 0.0722, + "step": 12653 + }, + { + "epoch": 1.66, + "learning_rate": 1.839809148319173e-05, + "loss": 0.0854, + "step": 12654 + }, + { + "epoch": 1.66, + "learning_rate": 1.8397803567582863e-05, + "loss": 0.0826, + "step": 12655 + }, + { + "epoch": 1.66, + "learning_rate": 1.839751562835562e-05, + "loss": 0.0828, + "step": 12656 + }, + { + "epoch": 1.66, + "learning_rate": 1.839722766551081e-05, + "loss": 0.0817, + "step": 12657 + }, + { + "epoch": 1.66, + "learning_rate": 1.8396939679049248e-05, + "loss": 0.083, + "step": 12658 + }, + { + "epoch": 1.66, + "learning_rate": 1.8396651668971737e-05, + "loss": 0.0836, + "step": 12659 + }, + { + "epoch": 1.66, + "learning_rate": 1.8396363635279092e-05, + "loss": 0.0617, + "step": 12660 + }, + { + "epoch": 1.66, + "learning_rate": 1.839607557797212e-05, + "loss": 0.0816, + "step": 12661 + }, + { + "epoch": 1.66, + "learning_rate": 1.8395787497051634e-05, + "loss": 0.0886, + "step": 12662 + }, + { + "epoch": 1.66, + "learning_rate": 1.8395499392518443e-05, + "loss": 0.0723, + "step": 12663 + }, + { + "epoch": 1.66, + "learning_rate": 1.8395211264373355e-05, + "loss": 0.0786, + "step": 12664 + }, + { + "epoch": 1.66, + "learning_rate": 1.8394923112617186e-05, + "loss": 0.0715, + "step": 12665 + }, + { + "epoch": 1.66, + "learning_rate": 1.839463493725074e-05, + "loss": 0.0762, + "step": 12666 + }, + { + "epoch": 1.66, + "learning_rate": 1.8394346738274832e-05, + "loss": 0.0744, + "step": 12667 + }, + { + "epoch": 1.66, + "learning_rate": 1.8394058515690266e-05, + "loss": 0.0751, + "step": 12668 + }, + { + "epoch": 1.66, + "learning_rate": 1.8393770269497862e-05, + "loss": 0.0847, + "step": 12669 + }, + { + "epoch": 1.66, + "learning_rate": 1.8393481999698422e-05, + "loss": 0.0826, + "step": 12670 + }, + { + "epoch": 1.66, + "learning_rate": 1.8393193706292768e-05, + "loss": 0.0847, + "step": 12671 + }, + { + "epoch": 1.66, + "learning_rate": 1.8392905389281697e-05, + "loss": 0.0869, + "step": 12672 + }, + { + "epoch": 1.66, + "learning_rate": 1.839261704866603e-05, + "loss": 0.0718, + "step": 12673 + }, + { + "epoch": 1.66, + "learning_rate": 1.8392328684446576e-05, + "loss": 0.0819, + "step": 12674 + }, + { + "epoch": 1.66, + "learning_rate": 1.839204029662414e-05, + "loss": 0.0703, + "step": 12675 + }, + { + "epoch": 1.66, + "learning_rate": 1.839175188519954e-05, + "loss": 0.0777, + "step": 12676 + }, + { + "epoch": 1.66, + "learning_rate": 1.8391463450173584e-05, + "loss": 0.0678, + "step": 12677 + }, + { + "epoch": 1.66, + "learning_rate": 1.8391174991547085e-05, + "loss": 0.0653, + "step": 12678 + }, + { + "epoch": 1.66, + "learning_rate": 1.8390886509320852e-05, + "loss": 0.0708, + "step": 12679 + }, + { + "epoch": 1.66, + "learning_rate": 1.8390598003495696e-05, + "loss": 0.0794, + "step": 12680 + }, + { + "epoch": 1.66, + "learning_rate": 1.8390309474072435e-05, + "loss": 0.0745, + "step": 12681 + }, + { + "epoch": 1.66, + "learning_rate": 1.839002092105187e-05, + "loss": 0.0788, + "step": 12682 + }, + { + "epoch": 1.66, + "learning_rate": 1.8389732344434823e-05, + "loss": 0.0732, + "step": 12683 + }, + { + "epoch": 1.66, + "learning_rate": 1.8389443744222095e-05, + "loss": 0.0854, + "step": 12684 + }, + { + "epoch": 1.66, + "learning_rate": 1.838915512041451e-05, + "loss": 0.0745, + "step": 12685 + }, + { + "epoch": 1.66, + "learning_rate": 1.8388866473012867e-05, + "loss": 0.0871, + "step": 12686 + }, + { + "epoch": 1.66, + "learning_rate": 1.8388577802017984e-05, + "loss": 0.0681, + "step": 12687 + }, + { + "epoch": 1.66, + "learning_rate": 1.8388289107430677e-05, + "loss": 0.0702, + "step": 12688 + }, + { + "epoch": 1.66, + "learning_rate": 1.838800038925175e-05, + "loss": 0.0694, + "step": 12689 + }, + { + "epoch": 1.66, + "learning_rate": 1.8387711647482017e-05, + "loss": 0.0625, + "step": 12690 + }, + { + "epoch": 1.66, + "learning_rate": 1.8387422882122296e-05, + "loss": 0.0647, + "step": 12691 + }, + { + "epoch": 1.66, + "learning_rate": 1.8387134093173392e-05, + "loss": 0.0684, + "step": 12692 + }, + { + "epoch": 1.66, + "learning_rate": 1.838684528063612e-05, + "loss": 0.0692, + "step": 12693 + }, + { + "epoch": 1.66, + "learning_rate": 1.8386556444511297e-05, + "loss": 0.0686, + "step": 12694 + }, + { + "epoch": 1.66, + "learning_rate": 1.8386267584799726e-05, + "loss": 0.0714, + "step": 12695 + }, + { + "epoch": 1.66, + "learning_rate": 1.8385978701502224e-05, + "loss": 0.0645, + "step": 12696 + }, + { + "epoch": 1.66, + "learning_rate": 1.8385689794619605e-05, + "loss": 0.068, + "step": 12697 + }, + { + "epoch": 1.66, + "learning_rate": 1.838540086415268e-05, + "loss": 0.0885, + "step": 12698 + }, + { + "epoch": 1.66, + "learning_rate": 1.8385111910102263e-05, + "loss": 0.082, + "step": 12699 + }, + { + "epoch": 1.66, + "learning_rate": 1.8384822932469163e-05, + "loss": 0.0826, + "step": 12700 + }, + { + "epoch": 1.66, + "learning_rate": 1.8384533931254192e-05, + "loss": 0.0876, + "step": 12701 + }, + { + "epoch": 1.66, + "learning_rate": 1.8384244906458173e-05, + "loss": 0.0775, + "step": 12702 + }, + { + "epoch": 1.66, + "learning_rate": 1.838395585808191e-05, + "loss": 0.0796, + "step": 12703 + }, + { + "epoch": 1.66, + "learning_rate": 1.838366678612621e-05, + "loss": 0.0719, + "step": 12704 + }, + { + "epoch": 1.66, + "learning_rate": 1.8383377690591903e-05, + "loss": 0.085, + "step": 12705 + }, + { + "epoch": 1.66, + "learning_rate": 1.8383088571479788e-05, + "loss": 0.0781, + "step": 12706 + }, + { + "epoch": 1.66, + "learning_rate": 1.8382799428790684e-05, + "loss": 0.0656, + "step": 12707 + }, + { + "epoch": 1.66, + "learning_rate": 1.8382510262525403e-05, + "loss": 0.0731, + "step": 12708 + }, + { + "epoch": 1.66, + "learning_rate": 1.8382221072684757e-05, + "loss": 0.0706, + "step": 12709 + }, + { + "epoch": 1.66, + "learning_rate": 1.838193185926956e-05, + "loss": 0.0743, + "step": 12710 + }, + { + "epoch": 1.67, + "learning_rate": 1.8381642622280627e-05, + "loss": 0.0783, + "step": 12711 + }, + { + "epoch": 1.67, + "learning_rate": 1.8381353361718774e-05, + "loss": 0.0685, + "step": 12712 + }, + { + "epoch": 1.67, + "learning_rate": 1.8381064077584807e-05, + "loss": 0.0727, + "step": 12713 + }, + { + "epoch": 1.67, + "learning_rate": 1.8380774769879543e-05, + "loss": 0.072, + "step": 12714 + }, + { + "epoch": 1.67, + "learning_rate": 1.83804854386038e-05, + "loss": 0.0819, + "step": 12715 + }, + { + "epoch": 1.67, + "learning_rate": 1.838019608375839e-05, + "loss": 0.082, + "step": 12716 + }, + { + "epoch": 1.67, + "learning_rate": 1.8379906705344118e-05, + "loss": 0.0813, + "step": 12717 + }, + { + "epoch": 1.67, + "learning_rate": 1.8379617303361807e-05, + "loss": 0.0767, + "step": 12718 + }, + { + "epoch": 1.67, + "learning_rate": 1.837932787781227e-05, + "loss": 0.0704, + "step": 12719 + }, + { + "epoch": 1.67, + "learning_rate": 1.837903842869632e-05, + "loss": 0.0743, + "step": 12720 + }, + { + "epoch": 1.67, + "learning_rate": 1.8378748956014766e-05, + "loss": 0.0657, + "step": 12721 + }, + { + "epoch": 1.67, + "learning_rate": 1.8378459459768428e-05, + "loss": 0.0815, + "step": 12722 + }, + { + "epoch": 1.67, + "learning_rate": 1.837816993995812e-05, + "loss": 0.0732, + "step": 12723 + }, + { + "epoch": 1.67, + "learning_rate": 1.837788039658466e-05, + "loss": 0.0737, + "step": 12724 + }, + { + "epoch": 1.67, + "learning_rate": 1.8377590829648854e-05, + "loss": 0.082, + "step": 12725 + }, + { + "epoch": 1.67, + "learning_rate": 1.8377301239151517e-05, + "loss": 0.0726, + "step": 12726 + }, + { + "epoch": 1.67, + "learning_rate": 1.837701162509347e-05, + "loss": 0.0726, + "step": 12727 + }, + { + "epoch": 1.67, + "learning_rate": 1.8376721987475523e-05, + "loss": 0.068, + "step": 12728 + }, + { + "epoch": 1.67, + "learning_rate": 1.837643232629849e-05, + "loss": 0.0734, + "step": 12729 + }, + { + "epoch": 1.67, + "learning_rate": 1.837614264156319e-05, + "loss": 0.0743, + "step": 12730 + }, + { + "epoch": 1.67, + "learning_rate": 1.8375852933270432e-05, + "loss": 0.0771, + "step": 12731 + }, + { + "epoch": 1.67, + "learning_rate": 1.8375563201421036e-05, + "loss": 0.078, + "step": 12732 + }, + { + "epoch": 1.67, + "learning_rate": 1.8375273446015813e-05, + "loss": 0.0812, + "step": 12733 + }, + { + "epoch": 1.67, + "learning_rate": 1.8374983667055583e-05, + "loss": 0.0796, + "step": 12734 + }, + { + "epoch": 1.67, + "learning_rate": 1.8374693864541152e-05, + "loss": 0.0808, + "step": 12735 + }, + { + "epoch": 1.67, + "learning_rate": 1.8374404038473343e-05, + "loss": 0.0807, + "step": 12736 + }, + { + "epoch": 1.67, + "learning_rate": 1.8374114188852967e-05, + "loss": 0.0743, + "step": 12737 + }, + { + "epoch": 1.67, + "learning_rate": 1.8373824315680844e-05, + "loss": 0.0776, + "step": 12738 + }, + { + "epoch": 1.67, + "learning_rate": 1.8373534418957786e-05, + "loss": 0.0764, + "step": 12739 + }, + { + "epoch": 1.67, + "learning_rate": 1.8373244498684606e-05, + "loss": 0.0683, + "step": 12740 + }, + { + "epoch": 1.67, + "learning_rate": 1.837295455486212e-05, + "loss": 0.0821, + "step": 12741 + }, + { + "epoch": 1.67, + "learning_rate": 1.837266458749115e-05, + "loss": 0.0809, + "step": 12742 + }, + { + "epoch": 1.67, + "learning_rate": 1.8372374596572504e-05, + "loss": 0.0686, + "step": 12743 + }, + { + "epoch": 1.67, + "learning_rate": 1.8372084582107e-05, + "loss": 0.08, + "step": 12744 + }, + { + "epoch": 1.67, + "learning_rate": 1.8371794544095455e-05, + "loss": 0.0815, + "step": 12745 + }, + { + "epoch": 1.67, + "learning_rate": 1.837150448253868e-05, + "loss": 0.0757, + "step": 12746 + }, + { + "epoch": 1.67, + "learning_rate": 1.83712143974375e-05, + "loss": 0.0799, + "step": 12747 + }, + { + "epoch": 1.67, + "learning_rate": 1.8370924288792725e-05, + "loss": 0.0741, + "step": 12748 + }, + { + "epoch": 1.67, + "learning_rate": 1.837063415660517e-05, + "loss": 0.08, + "step": 12749 + }, + { + "epoch": 1.67, + "learning_rate": 1.8370344000875652e-05, + "loss": 0.0765, + "step": 12750 + }, + { + "epoch": 1.67, + "learning_rate": 1.8370053821604985e-05, + "loss": 0.0757, + "step": 12751 + }, + { + "epoch": 1.67, + "learning_rate": 1.836976361879399e-05, + "loss": 0.0772, + "step": 12752 + }, + { + "epoch": 1.67, + "learning_rate": 1.8369473392443477e-05, + "loss": 0.0744, + "step": 12753 + }, + { + "epoch": 1.67, + "learning_rate": 1.836918314255427e-05, + "loss": 0.0719, + "step": 12754 + }, + { + "epoch": 1.67, + "learning_rate": 1.836889286912718e-05, + "loss": 0.0606, + "step": 12755 + }, + { + "epoch": 1.67, + "learning_rate": 1.8368602572163022e-05, + "loss": 0.0737, + "step": 12756 + }, + { + "epoch": 1.67, + "learning_rate": 1.8368312251662617e-05, + "loss": 0.0727, + "step": 12757 + }, + { + "epoch": 1.67, + "learning_rate": 1.836802190762678e-05, + "loss": 0.0865, + "step": 12758 + }, + { + "epoch": 1.67, + "learning_rate": 1.8367731540056327e-05, + "loss": 0.0757, + "step": 12759 + }, + { + "epoch": 1.67, + "learning_rate": 1.836744114895207e-05, + "loss": 0.0782, + "step": 12760 + }, + { + "epoch": 1.67, + "learning_rate": 1.8367150734314837e-05, + "loss": 0.0764, + "step": 12761 + }, + { + "epoch": 1.67, + "learning_rate": 1.836686029614544e-05, + "loss": 0.0751, + "step": 12762 + }, + { + "epoch": 1.67, + "learning_rate": 1.8366569834444686e-05, + "loss": 0.072, + "step": 12763 + }, + { + "epoch": 1.67, + "learning_rate": 1.83662793492134e-05, + "loss": 0.0691, + "step": 12764 + }, + { + "epoch": 1.67, + "learning_rate": 1.8365988840452406e-05, + "loss": 0.0734, + "step": 12765 + }, + { + "epoch": 1.67, + "learning_rate": 1.836569830816251e-05, + "loss": 0.064, + "step": 12766 + }, + { + "epoch": 1.67, + "learning_rate": 1.836540775234453e-05, + "loss": 0.0796, + "step": 12767 + }, + { + "epoch": 1.67, + "learning_rate": 1.836511717299929e-05, + "loss": 0.0757, + "step": 12768 + }, + { + "epoch": 1.67, + "learning_rate": 1.8364826570127606e-05, + "loss": 0.0706, + "step": 12769 + }, + { + "epoch": 1.67, + "learning_rate": 1.836453594373029e-05, + "loss": 0.0718, + "step": 12770 + }, + { + "epoch": 1.67, + "learning_rate": 1.836424529380816e-05, + "loss": 0.0881, + "step": 12771 + }, + { + "epoch": 1.67, + "learning_rate": 1.8363954620362038e-05, + "loss": 0.0679, + "step": 12772 + }, + { + "epoch": 1.67, + "learning_rate": 1.8363663923392738e-05, + "loss": 0.0743, + "step": 12773 + }, + { + "epoch": 1.67, + "learning_rate": 1.8363373202901078e-05, + "loss": 0.0822, + "step": 12774 + }, + { + "epoch": 1.67, + "learning_rate": 1.8363082458887876e-05, + "loss": 0.0878, + "step": 12775 + }, + { + "epoch": 1.67, + "learning_rate": 1.836279169135395e-05, + "loss": 0.0779, + "step": 12776 + }, + { + "epoch": 1.67, + "learning_rate": 1.836250090030012e-05, + "loss": 0.0884, + "step": 12777 + }, + { + "epoch": 1.67, + "learning_rate": 1.83622100857272e-05, + "loss": 0.0748, + "step": 12778 + }, + { + "epoch": 1.67, + "learning_rate": 1.8361919247636008e-05, + "loss": 0.0687, + "step": 12779 + }, + { + "epoch": 1.67, + "learning_rate": 1.8361628386027365e-05, + "loss": 0.0828, + "step": 12780 + }, + { + "epoch": 1.67, + "learning_rate": 1.836133750090209e-05, + "loss": 0.0742, + "step": 12781 + }, + { + "epoch": 1.67, + "learning_rate": 1.8361046592260993e-05, + "loss": 0.0858, + "step": 12782 + }, + { + "epoch": 1.67, + "learning_rate": 1.8360755660104904e-05, + "loss": 0.0729, + "step": 12783 + }, + { + "epoch": 1.67, + "learning_rate": 1.8360464704434632e-05, + "loss": 0.0701, + "step": 12784 + }, + { + "epoch": 1.67, + "learning_rate": 1.8360173725250996e-05, + "loss": 0.0619, + "step": 12785 + }, + { + "epoch": 1.67, + "learning_rate": 1.835988272255482e-05, + "loss": 0.0784, + "step": 12786 + }, + { + "epoch": 1.68, + "learning_rate": 1.8359591696346916e-05, + "loss": 0.0716, + "step": 12787 + }, + { + "epoch": 1.68, + "learning_rate": 1.8359300646628105e-05, + "loss": 0.0632, + "step": 12788 + }, + { + "epoch": 1.68, + "learning_rate": 1.835900957339921e-05, + "loss": 0.0739, + "step": 12789 + }, + { + "epoch": 1.68, + "learning_rate": 1.8358718476661045e-05, + "loss": 0.0799, + "step": 12790 + }, + { + "epoch": 1.68, + "learning_rate": 1.8358427356414426e-05, + "loss": 0.0782, + "step": 12791 + }, + { + "epoch": 1.68, + "learning_rate": 1.835813621266018e-05, + "loss": 0.0744, + "step": 12792 + }, + { + "epoch": 1.68, + "learning_rate": 1.8357845045399115e-05, + "loss": 0.0695, + "step": 12793 + }, + { + "epoch": 1.68, + "learning_rate": 1.8357553854632063e-05, + "loss": 0.0703, + "step": 12794 + }, + { + "epoch": 1.68, + "learning_rate": 1.8357262640359833e-05, + "loss": 0.0882, + "step": 12795 + }, + { + "epoch": 1.68, + "learning_rate": 1.8356971402583245e-05, + "loss": 0.0733, + "step": 12796 + }, + { + "epoch": 1.68, + "learning_rate": 1.835668014130312e-05, + "loss": 0.0818, + "step": 12797 + }, + { + "epoch": 1.68, + "learning_rate": 1.835638885652028e-05, + "loss": 0.0798, + "step": 12798 + }, + { + "epoch": 1.68, + "learning_rate": 1.835609754823554e-05, + "loss": 0.0851, + "step": 12799 + }, + { + "epoch": 1.68, + "learning_rate": 1.835580621644972e-05, + "loss": 0.0744, + "step": 12800 + }, + { + "epoch": 1.68, + "learning_rate": 1.8355514861163638e-05, + "loss": 0.0805, + "step": 12801 + }, + { + "epoch": 1.68, + "learning_rate": 1.8355223482378118e-05, + "loss": 0.0737, + "step": 12802 + }, + { + "epoch": 1.68, + "learning_rate": 1.8354932080093975e-05, + "loss": 0.067, + "step": 12803 + }, + { + "epoch": 1.68, + "learning_rate": 1.835464065431203e-05, + "loss": 0.0769, + "step": 12804 + }, + { + "epoch": 1.68, + "learning_rate": 1.835434920503311e-05, + "loss": 0.0833, + "step": 12805 + }, + { + "epoch": 1.68, + "learning_rate": 1.835405773225802e-05, + "loss": 0.0816, + "step": 12806 + }, + { + "epoch": 1.68, + "learning_rate": 1.835376623598759e-05, + "loss": 0.0692, + "step": 12807 + }, + { + "epoch": 1.68, + "learning_rate": 1.835347471622264e-05, + "loss": 0.0775, + "step": 12808 + }, + { + "epoch": 1.68, + "learning_rate": 1.8353183172963985e-05, + "loss": 0.0662, + "step": 12809 + }, + { + "epoch": 1.68, + "learning_rate": 1.8352891606212445e-05, + "loss": 0.0721, + "step": 12810 + }, + { + "epoch": 1.68, + "learning_rate": 1.8352600015968846e-05, + "loss": 0.0744, + "step": 12811 + }, + { + "epoch": 1.68, + "learning_rate": 1.8352308402234e-05, + "loss": 0.0736, + "step": 12812 + }, + { + "epoch": 1.68, + "learning_rate": 1.8352016765008733e-05, + "loss": 0.0899, + "step": 12813 + }, + { + "epoch": 1.68, + "learning_rate": 1.8351725104293863e-05, + "loss": 0.0711, + "step": 12814 + }, + { + "epoch": 1.68, + "learning_rate": 1.8351433420090217e-05, + "loss": 0.078, + "step": 12815 + }, + { + "epoch": 1.68, + "learning_rate": 1.8351141712398602e-05, + "loss": 0.0725, + "step": 12816 + }, + { + "epoch": 1.68, + "learning_rate": 1.835084998121985e-05, + "loss": 0.0683, + "step": 12817 + }, + { + "epoch": 1.68, + "learning_rate": 1.8350558226554773e-05, + "loss": 0.0788, + "step": 12818 + }, + { + "epoch": 1.68, + "learning_rate": 1.83502664484042e-05, + "loss": 0.0743, + "step": 12819 + }, + { + "epoch": 1.68, + "learning_rate": 1.8349974646768945e-05, + "loss": 0.0674, + "step": 12820 + }, + { + "epoch": 1.68, + "learning_rate": 1.834968282164983e-05, + "loss": 0.078, + "step": 12821 + }, + { + "epoch": 1.68, + "learning_rate": 1.834939097304768e-05, + "loss": 0.0841, + "step": 12822 + }, + { + "epoch": 1.68, + "learning_rate": 1.834909910096331e-05, + "loss": 0.0711, + "step": 12823 + }, + { + "epoch": 1.68, + "learning_rate": 1.8348807205397543e-05, + "loss": 0.0684, + "step": 12824 + }, + { + "epoch": 1.68, + "learning_rate": 1.8348515286351204e-05, + "loss": 0.0809, + "step": 12825 + }, + { + "epoch": 1.68, + "learning_rate": 1.8348223343825105e-05, + "loss": 0.0796, + "step": 12826 + }, + { + "epoch": 1.68, + "learning_rate": 1.8347931377820075e-05, + "loss": 0.084, + "step": 12827 + }, + { + "epoch": 1.68, + "learning_rate": 1.8347639388336932e-05, + "loss": 0.0815, + "step": 12828 + }, + { + "epoch": 1.68, + "learning_rate": 1.83473473753765e-05, + "loss": 0.0745, + "step": 12829 + }, + { + "epoch": 1.68, + "learning_rate": 1.8347055338939593e-05, + "loss": 0.0788, + "step": 12830 + }, + { + "epoch": 1.68, + "learning_rate": 1.8346763279027043e-05, + "loss": 0.0855, + "step": 12831 + }, + { + "epoch": 1.68, + "learning_rate": 1.8346471195639663e-05, + "loss": 0.0718, + "step": 12832 + }, + { + "epoch": 1.68, + "learning_rate": 1.834617908877828e-05, + "loss": 0.0794, + "step": 12833 + }, + { + "epoch": 1.68, + "learning_rate": 1.8345886958443708e-05, + "loss": 0.0862, + "step": 12834 + }, + { + "epoch": 1.68, + "learning_rate": 1.8345594804636776e-05, + "loss": 0.0811, + "step": 12835 + }, + { + "epoch": 1.68, + "learning_rate": 1.83453026273583e-05, + "loss": 0.074, + "step": 12836 + }, + { + "epoch": 1.68, + "learning_rate": 1.834501042660911e-05, + "loss": 0.0786, + "step": 12837 + }, + { + "epoch": 1.68, + "learning_rate": 1.834471820239002e-05, + "loss": 0.066, + "step": 12838 + }, + { + "epoch": 1.68, + "learning_rate": 1.8344425954701855e-05, + "loss": 0.0859, + "step": 12839 + }, + { + "epoch": 1.68, + "learning_rate": 1.8344133683545434e-05, + "loss": 0.0736, + "step": 12840 + }, + { + "epoch": 1.68, + "learning_rate": 1.8343841388921586e-05, + "loss": 0.0876, + "step": 12841 + }, + { + "epoch": 1.68, + "learning_rate": 1.8343549070831127e-05, + "loss": 0.0789, + "step": 12842 + }, + { + "epoch": 1.68, + "learning_rate": 1.834325672927488e-05, + "loss": 0.0762, + "step": 12843 + }, + { + "epoch": 1.68, + "learning_rate": 1.8342964364253663e-05, + "loss": 0.0953, + "step": 12844 + }, + { + "epoch": 1.68, + "learning_rate": 1.8342671975768308e-05, + "loss": 0.093, + "step": 12845 + }, + { + "epoch": 1.68, + "learning_rate": 1.8342379563819636e-05, + "loss": 0.0738, + "step": 12846 + }, + { + "epoch": 1.68, + "learning_rate": 1.834208712840846e-05, + "loss": 0.0756, + "step": 12847 + }, + { + "epoch": 1.68, + "learning_rate": 1.8341794669535606e-05, + "loss": 0.0746, + "step": 12848 + }, + { + "epoch": 1.68, + "learning_rate": 1.8341502187201904e-05, + "loss": 0.0763, + "step": 12849 + }, + { + "epoch": 1.68, + "learning_rate": 1.834120968140817e-05, + "loss": 0.0842, + "step": 12850 + }, + { + "epoch": 1.68, + "learning_rate": 1.8340917152155228e-05, + "loss": 0.0803, + "step": 12851 + }, + { + "epoch": 1.68, + "learning_rate": 1.83406245994439e-05, + "loss": 0.0772, + "step": 12852 + }, + { + "epoch": 1.68, + "learning_rate": 1.834033202327501e-05, + "loss": 0.0878, + "step": 12853 + }, + { + "epoch": 1.68, + "learning_rate": 1.8340039423649377e-05, + "loss": 0.0711, + "step": 12854 + }, + { + "epoch": 1.68, + "learning_rate": 1.833974680056783e-05, + "loss": 0.0764, + "step": 12855 + }, + { + "epoch": 1.68, + "learning_rate": 1.833945415403119e-05, + "loss": 0.0653, + "step": 12856 + }, + { + "epoch": 1.68, + "learning_rate": 1.833916148404028e-05, + "loss": 0.0715, + "step": 12857 + }, + { + "epoch": 1.68, + "learning_rate": 1.8338868790595922e-05, + "loss": 0.0774, + "step": 12858 + }, + { + "epoch": 1.68, + "learning_rate": 1.8338576073698936e-05, + "loss": 0.0755, + "step": 12859 + }, + { + "epoch": 1.68, + "learning_rate": 1.8338283333350152e-05, + "loss": 0.0743, + "step": 12860 + }, + { + "epoch": 1.68, + "learning_rate": 1.833799056955039e-05, + "loss": 0.0747, + "step": 12861 + }, + { + "epoch": 1.68, + "learning_rate": 1.8337697782300474e-05, + "loss": 0.0734, + "step": 12862 + }, + { + "epoch": 1.68, + "learning_rate": 1.8337404971601227e-05, + "loss": 0.0821, + "step": 12863 + }, + { + "epoch": 1.69, + "learning_rate": 1.833711213745347e-05, + "loss": 0.0826, + "step": 12864 + }, + { + "epoch": 1.69, + "learning_rate": 1.833681927985803e-05, + "loss": 0.0807, + "step": 12865 + }, + { + "epoch": 1.69, + "learning_rate": 1.8336526398815733e-05, + "loss": 0.0697, + "step": 12866 + }, + { + "epoch": 1.69, + "learning_rate": 1.8336233494327394e-05, + "loss": 0.0765, + "step": 12867 + }, + { + "epoch": 1.69, + "learning_rate": 1.8335940566393847e-05, + "loss": 0.0734, + "step": 12868 + }, + { + "epoch": 1.69, + "learning_rate": 1.833564761501591e-05, + "loss": 0.0827, + "step": 12869 + }, + { + "epoch": 1.69, + "learning_rate": 1.8335354640194408e-05, + "loss": 0.0799, + "step": 12870 + }, + { + "epoch": 1.69, + "learning_rate": 1.8335061641930167e-05, + "loss": 0.0906, + "step": 12871 + }, + { + "epoch": 1.69, + "learning_rate": 1.8334768620224004e-05, + "loss": 0.0683, + "step": 12872 + }, + { + "epoch": 1.69, + "learning_rate": 1.833447557507675e-05, + "loss": 0.0781, + "step": 12873 + }, + { + "epoch": 1.69, + "learning_rate": 1.8334182506489228e-05, + "loss": 0.0887, + "step": 12874 + }, + { + "epoch": 1.69, + "learning_rate": 1.833388941446226e-05, + "loss": 0.0644, + "step": 12875 + }, + { + "epoch": 1.69, + "learning_rate": 1.8333596298996673e-05, + "loss": 0.0786, + "step": 12876 + }, + { + "epoch": 1.69, + "learning_rate": 1.833330316009329e-05, + "loss": 0.0678, + "step": 12877 + }, + { + "epoch": 1.69, + "learning_rate": 1.8333009997752936e-05, + "loss": 0.0774, + "step": 12878 + }, + { + "epoch": 1.69, + "learning_rate": 1.8332716811976437e-05, + "loss": 0.0734, + "step": 12879 + }, + { + "epoch": 1.69, + "learning_rate": 1.833242360276461e-05, + "loss": 0.0816, + "step": 12880 + }, + { + "epoch": 1.69, + "learning_rate": 1.833213037011829e-05, + "loss": 0.085, + "step": 12881 + }, + { + "epoch": 1.69, + "learning_rate": 1.8331837114038296e-05, + "loss": 0.0665, + "step": 12882 + }, + { + "epoch": 1.69, + "learning_rate": 1.8331543834525455e-05, + "loss": 0.0697, + "step": 12883 + }, + { + "epoch": 1.69, + "learning_rate": 1.833125053158059e-05, + "loss": 0.0792, + "step": 12884 + }, + { + "epoch": 1.69, + "learning_rate": 1.8330957205204526e-05, + "loss": 0.0844, + "step": 12885 + }, + { + "epoch": 1.69, + "learning_rate": 1.8330663855398088e-05, + "loss": 0.0742, + "step": 12886 + }, + { + "epoch": 1.69, + "learning_rate": 1.8330370482162102e-05, + "loss": 0.0781, + "step": 12887 + }, + { + "epoch": 1.69, + "learning_rate": 1.8330077085497395e-05, + "loss": 0.0694, + "step": 12888 + }, + { + "epoch": 1.69, + "learning_rate": 1.8329783665404786e-05, + "loss": 0.0781, + "step": 12889 + }, + { + "epoch": 1.69, + "learning_rate": 1.8329490221885106e-05, + "loss": 0.0812, + "step": 12890 + }, + { + "epoch": 1.69, + "learning_rate": 1.8329196754939178e-05, + "loss": 0.0755, + "step": 12891 + }, + { + "epoch": 1.69, + "learning_rate": 1.832890326456783e-05, + "loss": 0.0769, + "step": 12892 + }, + { + "epoch": 1.69, + "learning_rate": 1.8328609750771882e-05, + "loss": 0.0653, + "step": 12893 + }, + { + "epoch": 1.69, + "learning_rate": 1.8328316213552164e-05, + "loss": 0.0736, + "step": 12894 + }, + { + "epoch": 1.69, + "learning_rate": 1.8328022652909498e-05, + "loss": 0.0784, + "step": 12895 + }, + { + "epoch": 1.69, + "learning_rate": 1.8327729068844714e-05, + "loss": 0.0781, + "step": 12896 + }, + { + "epoch": 1.69, + "learning_rate": 1.8327435461358635e-05, + "loss": 0.0665, + "step": 12897 + }, + { + "epoch": 1.69, + "learning_rate": 1.8327141830452086e-05, + "loss": 0.0627, + "step": 12898 + }, + { + "epoch": 1.69, + "learning_rate": 1.8326848176125892e-05, + "loss": 0.0656, + "step": 12899 + }, + { + "epoch": 1.69, + "learning_rate": 1.8326554498380887e-05, + "loss": 0.0784, + "step": 12900 + }, + { + "epoch": 1.69, + "learning_rate": 1.8326260797217888e-05, + "loss": 0.0881, + "step": 12901 + }, + { + "epoch": 1.69, + "learning_rate": 1.8325967072637724e-05, + "loss": 0.082, + "step": 12902 + }, + { + "epoch": 1.69, + "learning_rate": 1.832567332464122e-05, + "loss": 0.082, + "step": 12903 + }, + { + "epoch": 1.69, + "learning_rate": 1.8325379553229203e-05, + "loss": 0.0678, + "step": 12904 + }, + { + "epoch": 1.69, + "learning_rate": 1.83250857584025e-05, + "loss": 0.0774, + "step": 12905 + }, + { + "epoch": 1.69, + "learning_rate": 1.8324791940161936e-05, + "loss": 0.0679, + "step": 12906 + }, + { + "epoch": 1.69, + "learning_rate": 1.832449809850834e-05, + "loss": 0.0763, + "step": 12907 + }, + { + "epoch": 1.69, + "learning_rate": 1.8324204233442535e-05, + "loss": 0.0747, + "step": 12908 + }, + { + "epoch": 1.69, + "learning_rate": 1.832391034496535e-05, + "loss": 0.0803, + "step": 12909 + }, + { + "epoch": 1.69, + "learning_rate": 1.832361643307761e-05, + "loss": 0.0693, + "step": 12910 + }, + { + "epoch": 1.69, + "learning_rate": 1.832332249778014e-05, + "loss": 0.0803, + "step": 12911 + }, + { + "epoch": 1.69, + "learning_rate": 1.8323028539073774e-05, + "loss": 0.0772, + "step": 12912 + }, + { + "epoch": 1.69, + "learning_rate": 1.8322734556959328e-05, + "loss": 0.0787, + "step": 12913 + }, + { + "epoch": 1.69, + "learning_rate": 1.8322440551437638e-05, + "loss": 0.082, + "step": 12914 + }, + { + "epoch": 1.69, + "learning_rate": 1.8322146522509524e-05, + "loss": 0.0682, + "step": 12915 + }, + { + "epoch": 1.69, + "learning_rate": 1.8321852470175816e-05, + "loss": 0.0672, + "step": 12916 + }, + { + "epoch": 1.69, + "learning_rate": 1.8321558394437344e-05, + "loss": 0.0726, + "step": 12917 + }, + { + "epoch": 1.69, + "learning_rate": 1.832126429529493e-05, + "loss": 0.0766, + "step": 12918 + }, + { + "epoch": 1.69, + "learning_rate": 1.8320970172749402e-05, + "loss": 0.0821, + "step": 12919 + }, + { + "epoch": 1.69, + "learning_rate": 1.8320676026801593e-05, + "loss": 0.0834, + "step": 12920 + }, + { + "epoch": 1.69, + "learning_rate": 1.832038185745232e-05, + "loss": 0.0873, + "step": 12921 + }, + { + "epoch": 1.69, + "learning_rate": 1.8320087664702423e-05, + "loss": 0.0794, + "step": 12922 + }, + { + "epoch": 1.69, + "learning_rate": 1.8319793448552716e-05, + "loss": 0.0779, + "step": 12923 + }, + { + "epoch": 1.69, + "learning_rate": 1.8319499209004035e-05, + "loss": 0.0693, + "step": 12924 + }, + { + "epoch": 1.69, + "learning_rate": 1.8319204946057208e-05, + "loss": 0.0837, + "step": 12925 + }, + { + "epoch": 1.69, + "learning_rate": 1.8318910659713057e-05, + "loss": 0.0874, + "step": 12926 + }, + { + "epoch": 1.69, + "learning_rate": 1.8318616349972416e-05, + "loss": 0.0764, + "step": 12927 + }, + { + "epoch": 1.69, + "learning_rate": 1.831832201683611e-05, + "loss": 0.0791, + "step": 12928 + }, + { + "epoch": 1.69, + "learning_rate": 1.831802766030496e-05, + "loss": 0.0697, + "step": 12929 + }, + { + "epoch": 1.69, + "learning_rate": 1.8317733280379807e-05, + "loss": 0.0715, + "step": 12930 + }, + { + "epoch": 1.69, + "learning_rate": 1.8317438877061467e-05, + "loss": 0.0676, + "step": 12931 + }, + { + "epoch": 1.69, + "learning_rate": 1.8317144450350777e-05, + "loss": 0.0784, + "step": 12932 + }, + { + "epoch": 1.69, + "learning_rate": 1.8316850000248557e-05, + "loss": 0.0762, + "step": 12933 + }, + { + "epoch": 1.69, + "learning_rate": 1.831655552675564e-05, + "loss": 0.0873, + "step": 12934 + }, + { + "epoch": 1.69, + "learning_rate": 1.831626102987286e-05, + "loss": 0.0755, + "step": 12935 + }, + { + "epoch": 1.69, + "learning_rate": 1.831596650960103e-05, + "loss": 0.0829, + "step": 12936 + }, + { + "epoch": 1.69, + "learning_rate": 1.8315671965940993e-05, + "loss": 0.0835, + "step": 12937 + }, + { + "epoch": 1.69, + "learning_rate": 1.831537739889357e-05, + "loss": 0.0839, + "step": 12938 + }, + { + "epoch": 1.69, + "learning_rate": 1.831508280845959e-05, + "loss": 0.079, + "step": 12939 + }, + { + "epoch": 1.7, + "learning_rate": 1.8314788194639884e-05, + "loss": 0.0831, + "step": 12940 + }, + { + "epoch": 1.7, + "learning_rate": 1.831449355743528e-05, + "loss": 0.0764, + "step": 12941 + }, + { + "epoch": 1.7, + "learning_rate": 1.83141988968466e-05, + "loss": 0.0763, + "step": 12942 + }, + { + "epoch": 1.7, + "learning_rate": 1.831390421287468e-05, + "loss": 0.0794, + "step": 12943 + }, + { + "epoch": 1.7, + "learning_rate": 1.831360950552035e-05, + "loss": 0.0717, + "step": 12944 + }, + { + "epoch": 1.7, + "learning_rate": 1.8313314774784436e-05, + "loss": 0.0727, + "step": 12945 + }, + { + "epoch": 1.7, + "learning_rate": 1.8313020020667764e-05, + "loss": 0.073, + "step": 12946 + }, + { + "epoch": 1.7, + "learning_rate": 1.831272524317117e-05, + "loss": 0.0643, + "step": 12947 + }, + { + "epoch": 1.7, + "learning_rate": 1.8312430442295475e-05, + "loss": 0.0686, + "step": 12948 + }, + { + "epoch": 1.7, + "learning_rate": 1.8312135618041513e-05, + "loss": 0.0754, + "step": 12949 + }, + { + "epoch": 1.7, + "learning_rate": 1.8311840770410116e-05, + "loss": 0.087, + "step": 12950 + }, + { + "epoch": 1.7, + "learning_rate": 1.8311545899402103e-05, + "loss": 0.0722, + "step": 12951 + }, + { + "epoch": 1.7, + "learning_rate": 1.8311251005018312e-05, + "loss": 0.0703, + "step": 12952 + }, + { + "epoch": 1.7, + "learning_rate": 1.8310956087259573e-05, + "loss": 0.0729, + "step": 12953 + }, + { + "epoch": 1.7, + "learning_rate": 1.831066114612671e-05, + "loss": 0.0741, + "step": 12954 + }, + { + "epoch": 1.7, + "learning_rate": 1.8310366181620554e-05, + "loss": 0.0802, + "step": 12955 + }, + { + "epoch": 1.7, + "learning_rate": 1.831007119374194e-05, + "loss": 0.0754, + "step": 12956 + }, + { + "epoch": 1.7, + "learning_rate": 1.8309776182491693e-05, + "loss": 0.0748, + "step": 12957 + }, + { + "epoch": 1.7, + "learning_rate": 1.8309481147870637e-05, + "loss": 0.0815, + "step": 12958 + }, + { + "epoch": 1.7, + "learning_rate": 1.8309186089879617e-05, + "loss": 0.0777, + "step": 12959 + }, + { + "epoch": 1.7, + "learning_rate": 1.8308891008519446e-05, + "loss": 0.0798, + "step": 12960 + }, + { + "epoch": 1.7, + "learning_rate": 1.830859590379096e-05, + "loss": 0.0749, + "step": 12961 + }, + { + "epoch": 1.7, + "learning_rate": 1.8308300775694994e-05, + "loss": 0.0788, + "step": 12962 + }, + { + "epoch": 1.7, + "learning_rate": 1.8308005624232378e-05, + "loss": 0.0823, + "step": 12963 + }, + { + "epoch": 1.7, + "learning_rate": 1.8307710449403933e-05, + "loss": 0.0677, + "step": 12964 + }, + { + "epoch": 1.7, + "learning_rate": 1.83074152512105e-05, + "loss": 0.0731, + "step": 12965 + }, + { + "epoch": 1.7, + "learning_rate": 1.83071200296529e-05, + "loss": 0.0625, + "step": 12966 + }, + { + "epoch": 1.7, + "learning_rate": 1.8306824784731968e-05, + "loss": 0.0728, + "step": 12967 + }, + { + "epoch": 1.7, + "learning_rate": 1.8306529516448535e-05, + "loss": 0.0986, + "step": 12968 + }, + { + "epoch": 1.7, + "learning_rate": 1.830623422480343e-05, + "loss": 0.0879, + "step": 12969 + }, + { + "epoch": 1.7, + "learning_rate": 1.8305938909797483e-05, + "loss": 0.0784, + "step": 12970 + }, + { + "epoch": 1.7, + "learning_rate": 1.8305643571431524e-05, + "loss": 0.0773, + "step": 12971 + }, + { + "epoch": 1.7, + "learning_rate": 1.8305348209706385e-05, + "loss": 0.0719, + "step": 12972 + }, + { + "epoch": 1.7, + "learning_rate": 1.83050528246229e-05, + "loss": 0.0553, + "step": 12973 + }, + { + "epoch": 1.7, + "learning_rate": 1.830475741618189e-05, + "loss": 0.0797, + "step": 12974 + }, + { + "epoch": 1.7, + "learning_rate": 1.8304461984384196e-05, + "loss": 0.0736, + "step": 12975 + }, + { + "epoch": 1.7, + "learning_rate": 1.8304166529230645e-05, + "loss": 0.0819, + "step": 12976 + }, + { + "epoch": 1.7, + "learning_rate": 1.8303871050722068e-05, + "loss": 0.0719, + "step": 12977 + }, + { + "epoch": 1.7, + "learning_rate": 1.8303575548859295e-05, + "loss": 0.0735, + "step": 12978 + }, + { + "epoch": 1.7, + "learning_rate": 1.8303280023643156e-05, + "loss": 0.0849, + "step": 12979 + }, + { + "epoch": 1.7, + "learning_rate": 1.8302984475074485e-05, + "loss": 0.0701, + "step": 12980 + }, + { + "epoch": 1.7, + "learning_rate": 1.8302688903154115e-05, + "loss": 0.0836, + "step": 12981 + }, + { + "epoch": 1.7, + "learning_rate": 1.8302393307882873e-05, + "loss": 0.075, + "step": 12982 + }, + { + "epoch": 1.7, + "learning_rate": 1.8302097689261593e-05, + "loss": 0.0782, + "step": 12983 + }, + { + "epoch": 1.7, + "learning_rate": 1.8301802047291103e-05, + "loss": 0.0766, + "step": 12984 + }, + { + "epoch": 1.7, + "learning_rate": 1.830150638197224e-05, + "loss": 0.0807, + "step": 12985 + }, + { + "epoch": 1.7, + "learning_rate": 1.830121069330583e-05, + "loss": 0.0804, + "step": 12986 + }, + { + "epoch": 1.7, + "learning_rate": 1.8300914981292705e-05, + "loss": 0.0782, + "step": 12987 + }, + { + "epoch": 1.7, + "learning_rate": 1.8300619245933703e-05, + "loss": 0.0709, + "step": 12988 + }, + { + "epoch": 1.7, + "learning_rate": 1.830032348722965e-05, + "loss": 0.0719, + "step": 12989 + }, + { + "epoch": 1.7, + "learning_rate": 1.8300027705181378e-05, + "loss": 0.0691, + "step": 12990 + }, + { + "epoch": 1.7, + "learning_rate": 1.8299731899789722e-05, + "loss": 0.07, + "step": 12991 + }, + { + "epoch": 1.7, + "learning_rate": 1.829943607105551e-05, + "loss": 0.0724, + "step": 12992 + }, + { + "epoch": 1.7, + "learning_rate": 1.8299140218979577e-05, + "loss": 0.0851, + "step": 12993 + }, + { + "epoch": 1.7, + "learning_rate": 1.8298844343562752e-05, + "loss": 0.0609, + "step": 12994 + }, + { + "epoch": 1.7, + "learning_rate": 1.8298548444805873e-05, + "loss": 0.0883, + "step": 12995 + }, + { + "epoch": 1.7, + "learning_rate": 1.8298252522709767e-05, + "loss": 0.076, + "step": 12996 + }, + { + "epoch": 1.7, + "learning_rate": 1.8297956577275267e-05, + "loss": 0.0814, + "step": 12997 + }, + { + "epoch": 1.7, + "learning_rate": 1.8297660608503207e-05, + "loss": 0.0861, + "step": 12998 + }, + { + "epoch": 1.7, + "learning_rate": 1.8297364616394417e-05, + "loss": 0.081, + "step": 12999 + }, + { + "epoch": 1.7, + "learning_rate": 1.829706860094973e-05, + "loss": 0.0792, + "step": 13000 + }, + { + "epoch": 1.7, + "learning_rate": 1.829677256216998e-05, + "loss": 0.0755, + "step": 13001 + }, + { + "epoch": 1.7, + "learning_rate": 1.8296476500056e-05, + "loss": 0.0754, + "step": 13002 + }, + { + "epoch": 1.7, + "learning_rate": 1.8296180414608623e-05, + "loss": 0.0742, + "step": 13003 + }, + { + "epoch": 1.7, + "learning_rate": 1.829588430582868e-05, + "loss": 0.0661, + "step": 13004 + }, + { + "epoch": 1.7, + "learning_rate": 1.8295588173717003e-05, + "loss": 0.0838, + "step": 13005 + }, + { + "epoch": 1.7, + "learning_rate": 1.8295292018274426e-05, + "loss": 0.0755, + "step": 13006 + }, + { + "epoch": 1.7, + "learning_rate": 1.8294995839501784e-05, + "loss": 0.0779, + "step": 13007 + }, + { + "epoch": 1.7, + "learning_rate": 1.8294699637399906e-05, + "loss": 0.0769, + "step": 13008 + }, + { + "epoch": 1.7, + "learning_rate": 1.8294403411969626e-05, + "loss": 0.0775, + "step": 13009 + }, + { + "epoch": 1.7, + "learning_rate": 1.8294107163211777e-05, + "loss": 0.0905, + "step": 13010 + }, + { + "epoch": 1.7, + "learning_rate": 1.8293810891127197e-05, + "loss": 0.0736, + "step": 13011 + }, + { + "epoch": 1.7, + "learning_rate": 1.8293514595716715e-05, + "loss": 0.0721, + "step": 13012 + }, + { + "epoch": 1.7, + "learning_rate": 1.8293218276981165e-05, + "loss": 0.082, + "step": 13013 + }, + { + "epoch": 1.7, + "learning_rate": 1.8292921934921377e-05, + "loss": 0.0886, + "step": 13014 + }, + { + "epoch": 1.7, + "learning_rate": 1.829262556953819e-05, + "loss": 0.0801, + "step": 13015 + }, + { + "epoch": 1.71, + "learning_rate": 1.8292329180832434e-05, + "loss": 0.0726, + "step": 13016 + }, + { + "epoch": 1.71, + "learning_rate": 1.829203276880495e-05, + "loss": 0.0779, + "step": 13017 + }, + { + "epoch": 1.71, + "learning_rate": 1.8291736333456558e-05, + "loss": 0.0677, + "step": 13018 + }, + { + "epoch": 1.71, + "learning_rate": 1.8291439874788102e-05, + "loss": 0.0717, + "step": 13019 + }, + { + "epoch": 1.71, + "learning_rate": 1.8291143392800412e-05, + "loss": 0.0811, + "step": 13020 + }, + { + "epoch": 1.71, + "learning_rate": 1.8290846887494322e-05, + "loss": 0.0745, + "step": 13021 + }, + { + "epoch": 1.71, + "learning_rate": 1.8290550358870666e-05, + "loss": 0.0784, + "step": 13022 + }, + { + "epoch": 1.71, + "learning_rate": 1.829025380693028e-05, + "loss": 0.0747, + "step": 13023 + }, + { + "epoch": 1.71, + "learning_rate": 1.8289957231674e-05, + "loss": 0.0793, + "step": 13024 + }, + { + "epoch": 1.71, + "learning_rate": 1.8289660633102653e-05, + "loss": 0.0907, + "step": 13025 + }, + { + "epoch": 1.71, + "learning_rate": 1.8289364011217077e-05, + "loss": 0.0804, + "step": 13026 + }, + { + "epoch": 1.71, + "learning_rate": 1.8289067366018106e-05, + "loss": 0.0651, + "step": 13027 + }, + { + "epoch": 1.71, + "learning_rate": 1.8288770697506578e-05, + "loss": 0.0728, + "step": 13028 + }, + { + "epoch": 1.71, + "learning_rate": 1.828847400568332e-05, + "loss": 0.0816, + "step": 13029 + }, + { + "epoch": 1.71, + "learning_rate": 1.828817729054917e-05, + "loss": 0.063, + "step": 13030 + }, + { + "epoch": 1.71, + "learning_rate": 1.8287880552104966e-05, + "loss": 0.0833, + "step": 13031 + }, + { + "epoch": 1.71, + "learning_rate": 1.8287583790351536e-05, + "loss": 0.0725, + "step": 13032 + }, + { + "epoch": 1.71, + "learning_rate": 1.828728700528972e-05, + "loss": 0.0819, + "step": 13033 + }, + { + "epoch": 1.71, + "learning_rate": 1.8286990196920346e-05, + "loss": 0.0763, + "step": 13034 + }, + { + "epoch": 1.71, + "learning_rate": 1.828669336524426e-05, + "loss": 0.0699, + "step": 13035 + }, + { + "epoch": 1.71, + "learning_rate": 1.8286396510262284e-05, + "loss": 0.0708, + "step": 13036 + }, + { + "epoch": 1.71, + "learning_rate": 1.8286099631975265e-05, + "loss": 0.0762, + "step": 13037 + }, + { + "epoch": 1.71, + "learning_rate": 1.828580273038403e-05, + "loss": 0.0827, + "step": 13038 + }, + { + "epoch": 1.71, + "learning_rate": 1.8285505805489413e-05, + "loss": 0.0798, + "step": 13039 + }, + { + "epoch": 1.71, + "learning_rate": 1.828520885729226e-05, + "loss": 0.0789, + "step": 13040 + }, + { + "epoch": 1.71, + "learning_rate": 1.828491188579339e-05, + "loss": 0.0825, + "step": 13041 + }, + { + "epoch": 1.71, + "learning_rate": 1.828461489099365e-05, + "loss": 0.076, + "step": 13042 + }, + { + "epoch": 1.71, + "learning_rate": 1.8284317872893873e-05, + "loss": 0.0879, + "step": 13043 + }, + { + "epoch": 1.71, + "learning_rate": 1.828402083149489e-05, + "loss": 0.0735, + "step": 13044 + }, + { + "epoch": 1.71, + "learning_rate": 1.828372376679754e-05, + "loss": 0.0732, + "step": 13045 + }, + { + "epoch": 1.71, + "learning_rate": 1.828342667880266e-05, + "loss": 0.0649, + "step": 13046 + }, + { + "epoch": 1.71, + "learning_rate": 1.8283129567511082e-05, + "loss": 0.0749, + "step": 13047 + }, + { + "epoch": 1.71, + "learning_rate": 1.8282832432923647e-05, + "loss": 0.0725, + "step": 13048 + }, + { + "epoch": 1.71, + "learning_rate": 1.8282535275041182e-05, + "loss": 0.0744, + "step": 13049 + }, + { + "epoch": 1.71, + "learning_rate": 1.828223809386453e-05, + "loss": 0.0728, + "step": 13050 + }, + { + "epoch": 1.71, + "learning_rate": 1.8281940889394523e-05, + "loss": 0.0835, + "step": 13051 + }, + { + "epoch": 1.71, + "learning_rate": 1.8281643661632e-05, + "loss": 0.0726, + "step": 13052 + }, + { + "epoch": 1.71, + "learning_rate": 1.8281346410577794e-05, + "loss": 0.0675, + "step": 13053 + }, + { + "epoch": 1.71, + "learning_rate": 1.828104913623274e-05, + "loss": 0.0752, + "step": 13054 + }, + { + "epoch": 1.71, + "learning_rate": 1.828075183859768e-05, + "loss": 0.0744, + "step": 13055 + }, + { + "epoch": 1.71, + "learning_rate": 1.8280454517673445e-05, + "loss": 0.0696, + "step": 13056 + }, + { + "epoch": 1.71, + "learning_rate": 1.8280157173460873e-05, + "loss": 0.0631, + "step": 13057 + }, + { + "epoch": 1.71, + "learning_rate": 1.82798598059608e-05, + "loss": 0.0735, + "step": 13058 + }, + { + "epoch": 1.71, + "learning_rate": 1.8279562415174063e-05, + "loss": 0.0802, + "step": 13059 + }, + { + "epoch": 1.71, + "learning_rate": 1.8279265001101496e-05, + "loss": 0.0738, + "step": 13060 + }, + { + "epoch": 1.71, + "learning_rate": 1.827896756374394e-05, + "loss": 0.0812, + "step": 13061 + }, + { + "epoch": 1.71, + "learning_rate": 1.8278670103102226e-05, + "loss": 0.0757, + "step": 13062 + }, + { + "epoch": 1.71, + "learning_rate": 1.8278372619177193e-05, + "loss": 0.0739, + "step": 13063 + }, + { + "epoch": 1.71, + "learning_rate": 1.8278075111969677e-05, + "loss": 0.0717, + "step": 13064 + }, + { + "epoch": 1.71, + "learning_rate": 1.827777758148052e-05, + "loss": 0.0882, + "step": 13065 + }, + { + "epoch": 1.71, + "learning_rate": 1.827748002771055e-05, + "loss": 0.0821, + "step": 13066 + }, + { + "epoch": 1.71, + "learning_rate": 1.827718245066061e-05, + "loss": 0.0753, + "step": 13067 + }, + { + "epoch": 1.71, + "learning_rate": 1.8276884850331535e-05, + "loss": 0.0647, + "step": 13068 + }, + { + "epoch": 1.71, + "learning_rate": 1.8276587226724162e-05, + "loss": 0.0871, + "step": 13069 + }, + { + "epoch": 1.71, + "learning_rate": 1.8276289579839324e-05, + "loss": 0.0659, + "step": 13070 + }, + { + "epoch": 1.71, + "learning_rate": 1.827599190967787e-05, + "loss": 0.0727, + "step": 13071 + }, + { + "epoch": 1.71, + "learning_rate": 1.8275694216240626e-05, + "loss": 0.0803, + "step": 13072 + }, + { + "epoch": 1.71, + "learning_rate": 1.827539649952843e-05, + "loss": 0.0794, + "step": 13073 + }, + { + "epoch": 1.71, + "learning_rate": 1.8275098759542123e-05, + "loss": 0.0808, + "step": 13074 + }, + { + "epoch": 1.71, + "learning_rate": 1.8274800996282545e-05, + "loss": 0.0725, + "step": 13075 + }, + { + "epoch": 1.71, + "learning_rate": 1.8274503209750526e-05, + "loss": 0.0807, + "step": 13076 + }, + { + "epoch": 1.71, + "learning_rate": 1.827420539994691e-05, + "loss": 0.0833, + "step": 13077 + }, + { + "epoch": 1.71, + "learning_rate": 1.827390756687253e-05, + "loss": 0.0684, + "step": 13078 + }, + { + "epoch": 1.71, + "learning_rate": 1.827360971052823e-05, + "loss": 0.0765, + "step": 13079 + }, + { + "epoch": 1.71, + "learning_rate": 1.827331183091484e-05, + "loss": 0.0759, + "step": 13080 + }, + { + "epoch": 1.71, + "learning_rate": 1.8273013928033202e-05, + "loss": 0.0765, + "step": 13081 + }, + { + "epoch": 1.71, + "learning_rate": 1.8272716001884153e-05, + "loss": 0.0743, + "step": 13082 + }, + { + "epoch": 1.71, + "learning_rate": 1.8272418052468528e-05, + "loss": 0.0794, + "step": 13083 + }, + { + "epoch": 1.71, + "learning_rate": 1.827212007978717e-05, + "loss": 0.0752, + "step": 13084 + }, + { + "epoch": 1.71, + "learning_rate": 1.827182208384091e-05, + "loss": 0.0809, + "step": 13085 + }, + { + "epoch": 1.71, + "learning_rate": 1.8271524064630598e-05, + "loss": 0.0818, + "step": 13086 + }, + { + "epoch": 1.71, + "learning_rate": 1.8271226022157064e-05, + "loss": 0.0744, + "step": 13087 + }, + { + "epoch": 1.71, + "learning_rate": 1.8270927956421146e-05, + "loss": 0.0709, + "step": 13088 + }, + { + "epoch": 1.71, + "learning_rate": 1.827062986742368e-05, + "loss": 0.0651, + "step": 13089 + }, + { + "epoch": 1.71, + "learning_rate": 1.827033175516551e-05, + "loss": 0.0901, + "step": 13090 + }, + { + "epoch": 1.71, + "learning_rate": 1.8270033619647478e-05, + "loss": 0.081, + "step": 13091 + }, + { + "epoch": 1.71, + "learning_rate": 1.8269735460870408e-05, + "loss": 0.0759, + "step": 13092 + }, + { + "epoch": 1.72, + "learning_rate": 1.8269437278835155e-05, + "loss": 0.0741, + "step": 13093 + }, + { + "epoch": 1.72, + "learning_rate": 1.8269139073542542e-05, + "loss": 0.0891, + "step": 13094 + }, + { + "epoch": 1.72, + "learning_rate": 1.8268840844993424e-05, + "loss": 0.0829, + "step": 13095 + }, + { + "epoch": 1.72, + "learning_rate": 1.8268542593188626e-05, + "loss": 0.082, + "step": 13096 + }, + { + "epoch": 1.72, + "learning_rate": 1.8268244318128996e-05, + "loss": 0.0761, + "step": 13097 + }, + { + "epoch": 1.72, + "learning_rate": 1.8267946019815364e-05, + "loss": 0.0834, + "step": 13098 + }, + { + "epoch": 1.72, + "learning_rate": 1.826764769824858e-05, + "loss": 0.0666, + "step": 13099 + }, + { + "epoch": 1.72, + "learning_rate": 1.8267349353429472e-05, + "loss": 0.085, + "step": 13100 + }, + { + "epoch": 1.72, + "learning_rate": 1.826705098535889e-05, + "loss": 0.0813, + "step": 13101 + }, + { + "epoch": 1.72, + "learning_rate": 1.8266752594037665e-05, + "loss": 0.0724, + "step": 13102 + }, + { + "epoch": 1.72, + "learning_rate": 1.8266454179466637e-05, + "loss": 0.085, + "step": 13103 + }, + { + "epoch": 1.72, + "learning_rate": 1.8266155741646648e-05, + "loss": 0.0756, + "step": 13104 + }, + { + "epoch": 1.72, + "learning_rate": 1.8265857280578535e-05, + "loss": 0.0759, + "step": 13105 + }, + { + "epoch": 1.72, + "learning_rate": 1.826555879626314e-05, + "loss": 0.0777, + "step": 13106 + }, + { + "epoch": 1.72, + "learning_rate": 1.82652602887013e-05, + "loss": 0.0775, + "step": 13107 + }, + { + "epoch": 1.72, + "learning_rate": 1.8264961757893855e-05, + "loss": 0.0939, + "step": 13108 + }, + { + "epoch": 1.72, + "learning_rate": 1.8264663203841647e-05, + "loss": 0.0889, + "step": 13109 + }, + { + "epoch": 1.72, + "learning_rate": 1.826436462654551e-05, + "loss": 0.0872, + "step": 13110 + }, + { + "epoch": 1.72, + "learning_rate": 1.826406602600629e-05, + "loss": 0.0776, + "step": 13111 + }, + { + "epoch": 1.72, + "learning_rate": 1.826376740222483e-05, + "loss": 0.0752, + "step": 13112 + }, + { + "epoch": 1.72, + "learning_rate": 1.8263468755201956e-05, + "loss": 0.0656, + "step": 13113 + }, + { + "epoch": 1.72, + "learning_rate": 1.826317008493852e-05, + "loss": 0.0654, + "step": 13114 + }, + { + "epoch": 1.72, + "learning_rate": 1.8262871391435356e-05, + "loss": 0.085, + "step": 13115 + }, + { + "epoch": 1.72, + "learning_rate": 1.826257267469331e-05, + "loss": 0.0819, + "step": 13116 + }, + { + "epoch": 1.72, + "learning_rate": 1.8262273934713213e-05, + "loss": 0.0785, + "step": 13117 + }, + { + "epoch": 1.72, + "learning_rate": 1.8261975171495912e-05, + "loss": 0.0826, + "step": 13118 + }, + { + "epoch": 1.72, + "learning_rate": 1.826167638504225e-05, + "loss": 0.0846, + "step": 13119 + }, + { + "epoch": 1.72, + "learning_rate": 1.826137757535306e-05, + "loss": 0.0802, + "step": 13120 + }, + { + "epoch": 1.72, + "learning_rate": 1.8261078742429186e-05, + "loss": 0.0827, + "step": 13121 + }, + { + "epoch": 1.72, + "learning_rate": 1.826077988627147e-05, + "loss": 0.0708, + "step": 13122 + }, + { + "epoch": 1.72, + "learning_rate": 1.8260481006880747e-05, + "loss": 0.0713, + "step": 13123 + }, + { + "epoch": 1.72, + "learning_rate": 1.8260182104257863e-05, + "loss": 0.0848, + "step": 13124 + }, + { + "epoch": 1.72, + "learning_rate": 1.8259883178403656e-05, + "loss": 0.0776, + "step": 13125 + }, + { + "epoch": 1.72, + "learning_rate": 1.8259584229318968e-05, + "loss": 0.079, + "step": 13126 + }, + { + "epoch": 1.72, + "learning_rate": 1.825928525700464e-05, + "loss": 0.0809, + "step": 13127 + }, + { + "epoch": 1.72, + "learning_rate": 1.825898626146151e-05, + "loss": 0.0808, + "step": 13128 + }, + { + "epoch": 1.72, + "learning_rate": 1.8258687242690423e-05, + "loss": 0.0866, + "step": 13129 + }, + { + "epoch": 1.72, + "learning_rate": 1.8258388200692217e-05, + "loss": 0.0712, + "step": 13130 + }, + { + "epoch": 1.72, + "learning_rate": 1.8258089135467735e-05, + "loss": 0.0641, + "step": 13131 + }, + { + "epoch": 1.72, + "learning_rate": 1.8257790047017812e-05, + "loss": 0.0742, + "step": 13132 + }, + { + "epoch": 1.72, + "learning_rate": 1.82574909353433e-05, + "loss": 0.0855, + "step": 13133 + }, + { + "epoch": 1.72, + "learning_rate": 1.8257191800445035e-05, + "loss": 0.0758, + "step": 13134 + }, + { + "epoch": 1.72, + "learning_rate": 1.8256892642323854e-05, + "loss": 0.0732, + "step": 13135 + }, + { + "epoch": 1.72, + "learning_rate": 1.8256593460980604e-05, + "loss": 0.0795, + "step": 13136 + }, + { + "epoch": 1.72, + "learning_rate": 1.825629425641612e-05, + "loss": 0.0796, + "step": 13137 + }, + { + "epoch": 1.72, + "learning_rate": 1.8255995028631254e-05, + "loss": 0.0748, + "step": 13138 + }, + { + "epoch": 1.72, + "learning_rate": 1.8255695777626842e-05, + "loss": 0.0695, + "step": 13139 + }, + { + "epoch": 1.72, + "learning_rate": 1.8255396503403722e-05, + "loss": 0.0947, + "step": 13140 + }, + { + "epoch": 1.72, + "learning_rate": 1.8255097205962742e-05, + "loss": 0.0791, + "step": 13141 + }, + { + "epoch": 1.72, + "learning_rate": 1.8254797885304737e-05, + "loss": 0.0906, + "step": 13142 + }, + { + "epoch": 1.72, + "learning_rate": 1.8254498541430555e-05, + "loss": 0.0733, + "step": 13143 + }, + { + "epoch": 1.72, + "learning_rate": 1.8254199174341037e-05, + "loss": 0.0767, + "step": 13144 + }, + { + "epoch": 1.72, + "learning_rate": 1.825389978403702e-05, + "loss": 0.091, + "step": 13145 + }, + { + "epoch": 1.72, + "learning_rate": 1.8253600370519348e-05, + "loss": 0.0928, + "step": 13146 + }, + { + "epoch": 1.72, + "learning_rate": 1.8253300933788867e-05, + "loss": 0.0762, + "step": 13147 + }, + { + "epoch": 1.72, + "learning_rate": 1.8253001473846418e-05, + "loss": 0.0795, + "step": 13148 + }, + { + "epoch": 1.72, + "learning_rate": 1.825270199069284e-05, + "loss": 0.0701, + "step": 13149 + }, + { + "epoch": 1.72, + "learning_rate": 1.825240248432898e-05, + "loss": 0.0839, + "step": 13150 + }, + { + "epoch": 1.72, + "learning_rate": 1.8252102954755673e-05, + "loss": 0.0812, + "step": 13151 + }, + { + "epoch": 1.72, + "learning_rate": 1.8251803401973768e-05, + "loss": 0.0783, + "step": 13152 + }, + { + "epoch": 1.72, + "learning_rate": 1.825150382598411e-05, + "loss": 0.0776, + "step": 13153 + }, + { + "epoch": 1.72, + "learning_rate": 1.825120422678753e-05, + "loss": 0.0818, + "step": 13154 + }, + { + "epoch": 1.72, + "learning_rate": 1.825090460438488e-05, + "loss": 0.0765, + "step": 13155 + }, + { + "epoch": 1.72, + "learning_rate": 1.8250604958777e-05, + "loss": 0.0764, + "step": 13156 + }, + { + "epoch": 1.72, + "learning_rate": 1.8250305289964732e-05, + "loss": 0.0668, + "step": 13157 + }, + { + "epoch": 1.72, + "learning_rate": 1.825000559794892e-05, + "loss": 0.0746, + "step": 13158 + }, + { + "epoch": 1.72, + "learning_rate": 1.8249705882730405e-05, + "loss": 0.0791, + "step": 13159 + }, + { + "epoch": 1.72, + "learning_rate": 1.8249406144310038e-05, + "loss": 0.0634, + "step": 13160 + }, + { + "epoch": 1.72, + "learning_rate": 1.824910638268865e-05, + "loss": 0.0733, + "step": 13161 + }, + { + "epoch": 1.72, + "learning_rate": 1.824880659786709e-05, + "loss": 0.0854, + "step": 13162 + }, + { + "epoch": 1.72, + "learning_rate": 1.8248506789846202e-05, + "loss": 0.0956, + "step": 13163 + }, + { + "epoch": 1.72, + "learning_rate": 1.824820695862683e-05, + "loss": 0.0786, + "step": 13164 + }, + { + "epoch": 1.72, + "learning_rate": 1.824790710420981e-05, + "loss": 0.0752, + "step": 13165 + }, + { + "epoch": 1.72, + "learning_rate": 1.8247607226595995e-05, + "loss": 0.0815, + "step": 13166 + }, + { + "epoch": 1.72, + "learning_rate": 1.8247307325786222e-05, + "loss": 0.0825, + "step": 13167 + }, + { + "epoch": 1.72, + "learning_rate": 1.8247007401781336e-05, + "loss": 0.0719, + "step": 13168 + }, + { + "epoch": 1.73, + "learning_rate": 1.824670745458218e-05, + "loss": 0.0767, + "step": 13169 + }, + { + "epoch": 1.73, + "learning_rate": 1.82464074841896e-05, + "loss": 0.0669, + "step": 13170 + }, + { + "epoch": 1.73, + "learning_rate": 1.8246107490604438e-05, + "loss": 0.0837, + "step": 13171 + }, + { + "epoch": 1.73, + "learning_rate": 1.824580747382754e-05, + "loss": 0.0738, + "step": 13172 + }, + { + "epoch": 1.73, + "learning_rate": 1.8245507433859745e-05, + "loss": 0.0809, + "step": 13173 + }, + { + "epoch": 1.73, + "learning_rate": 1.8245207370701898e-05, + "loss": 0.0755, + "step": 13174 + }, + { + "epoch": 1.73, + "learning_rate": 1.8244907284354847e-05, + "loss": 0.0696, + "step": 13175 + }, + { + "epoch": 1.73, + "learning_rate": 1.8244607174819432e-05, + "loss": 0.0828, + "step": 13176 + }, + { + "epoch": 1.73, + "learning_rate": 1.8244307042096497e-05, + "loss": 0.0859, + "step": 13177 + }, + { + "epoch": 1.73, + "learning_rate": 1.8244006886186893e-05, + "loss": 0.0774, + "step": 13178 + }, + { + "epoch": 1.73, + "learning_rate": 1.8243706707091452e-05, + "loss": 0.0734, + "step": 13179 + }, + { + "epoch": 1.73, + "learning_rate": 1.824340650481103e-05, + "loss": 0.0715, + "step": 13180 + }, + { + "epoch": 1.73, + "learning_rate": 1.8243106279346463e-05, + "loss": 0.0759, + "step": 13181 + }, + { + "epoch": 1.73, + "learning_rate": 1.82428060306986e-05, + "loss": 0.0771, + "step": 13182 + }, + { + "epoch": 1.73, + "learning_rate": 1.8242505758868283e-05, + "loss": 0.0798, + "step": 13183 + }, + { + "epoch": 1.73, + "learning_rate": 1.8242205463856357e-05, + "loss": 0.0814, + "step": 13184 + }, + { + "epoch": 1.73, + "learning_rate": 1.824190514566367e-05, + "loss": 0.0634, + "step": 13185 + }, + { + "epoch": 1.73, + "learning_rate": 1.824160480429106e-05, + "loss": 0.089, + "step": 13186 + }, + { + "epoch": 1.73, + "learning_rate": 1.824130443973938e-05, + "loss": 0.0793, + "step": 13187 + }, + { + "epoch": 1.73, + "learning_rate": 1.8241004052009464e-05, + "loss": 0.0752, + "step": 13188 + }, + { + "epoch": 1.73, + "learning_rate": 1.8240703641102166e-05, + "loss": 0.0788, + "step": 13189 + }, + { + "epoch": 1.73, + "learning_rate": 1.824040320701833e-05, + "loss": 0.0834, + "step": 13190 + }, + { + "epoch": 1.73, + "learning_rate": 1.8240102749758796e-05, + "loss": 0.0828, + "step": 13191 + }, + { + "epoch": 1.73, + "learning_rate": 1.8239802269324415e-05, + "loss": 0.0718, + "step": 13192 + }, + { + "epoch": 1.73, + "learning_rate": 1.8239501765716026e-05, + "loss": 0.0824, + "step": 13193 + }, + { + "epoch": 1.73, + "learning_rate": 1.8239201238934475e-05, + "loss": 0.0733, + "step": 13194 + }, + { + "epoch": 1.73, + "learning_rate": 1.8238900688980613e-05, + "loss": 0.0797, + "step": 13195 + }, + { + "epoch": 1.73, + "learning_rate": 1.823860011585528e-05, + "loss": 0.0784, + "step": 13196 + }, + { + "epoch": 1.73, + "learning_rate": 1.8238299519559322e-05, + "loss": 0.0674, + "step": 13197 + }, + { + "epoch": 1.73, + "learning_rate": 1.823799890009359e-05, + "loss": 0.0802, + "step": 13198 + }, + { + "epoch": 1.73, + "learning_rate": 1.823769825745892e-05, + "loss": 0.0787, + "step": 13199 + }, + { + "epoch": 1.73, + "learning_rate": 1.8237397591656164e-05, + "loss": 0.079, + "step": 13200 + }, + { + "epoch": 1.73, + "learning_rate": 1.8237096902686167e-05, + "loss": 0.0877, + "step": 13201 + }, + { + "epoch": 1.73, + "learning_rate": 1.8236796190549772e-05, + "loss": 0.0771, + "step": 13202 + }, + { + "epoch": 1.73, + "learning_rate": 1.8236495455247826e-05, + "loss": 0.0701, + "step": 13203 + }, + { + "epoch": 1.73, + "learning_rate": 1.8236194696781178e-05, + "loss": 0.0679, + "step": 13204 + }, + { + "epoch": 1.73, + "learning_rate": 1.823589391515067e-05, + "loss": 0.0835, + "step": 13205 + }, + { + "epoch": 1.73, + "learning_rate": 1.8235593110357147e-05, + "loss": 0.0814, + "step": 13206 + }, + { + "epoch": 1.73, + "learning_rate": 1.823529228240146e-05, + "loss": 0.0734, + "step": 13207 + }, + { + "epoch": 1.73, + "learning_rate": 1.8234991431284446e-05, + "loss": 0.0798, + "step": 13208 + }, + { + "epoch": 1.73, + "learning_rate": 1.823469055700696e-05, + "loss": 0.0833, + "step": 13209 + }, + { + "epoch": 1.73, + "learning_rate": 1.8234389659569847e-05, + "loss": 0.0765, + "step": 13210 + }, + { + "epoch": 1.73, + "learning_rate": 1.823408873897395e-05, + "loss": 0.0811, + "step": 13211 + }, + { + "epoch": 1.73, + "learning_rate": 1.823378779522012e-05, + "loss": 0.0716, + "step": 13212 + }, + { + "epoch": 1.73, + "learning_rate": 1.82334868283092e-05, + "loss": 0.077, + "step": 13213 + }, + { + "epoch": 1.73, + "learning_rate": 1.8233185838242032e-05, + "loss": 0.0793, + "step": 13214 + }, + { + "epoch": 1.73, + "learning_rate": 1.823288482501947e-05, + "loss": 0.0792, + "step": 13215 + }, + { + "epoch": 1.73, + "learning_rate": 1.823258378864236e-05, + "loss": 0.0732, + "step": 13216 + }, + { + "epoch": 1.73, + "learning_rate": 1.8232282729111547e-05, + "loss": 0.0718, + "step": 13217 + }, + { + "epoch": 1.73, + "learning_rate": 1.8231981646427873e-05, + "loss": 0.0857, + "step": 13218 + }, + { + "epoch": 1.73, + "learning_rate": 1.823168054059219e-05, + "loss": 0.0769, + "step": 13219 + }, + { + "epoch": 1.73, + "learning_rate": 1.8231379411605346e-05, + "loss": 0.0782, + "step": 13220 + }, + { + "epoch": 1.73, + "learning_rate": 1.8231078259468183e-05, + "loss": 0.0773, + "step": 13221 + }, + { + "epoch": 1.73, + "learning_rate": 1.823077708418155e-05, + "loss": 0.08, + "step": 13222 + }, + { + "epoch": 1.73, + "learning_rate": 1.82304758857463e-05, + "loss": 0.0799, + "step": 13223 + }, + { + "epoch": 1.73, + "learning_rate": 1.8230174664163273e-05, + "loss": 0.0681, + "step": 13224 + }, + { + "epoch": 1.73, + "learning_rate": 1.8229873419433313e-05, + "loss": 0.0732, + "step": 13225 + }, + { + "epoch": 1.73, + "learning_rate": 1.8229572151557277e-05, + "loss": 0.0748, + "step": 13226 + }, + { + "epoch": 1.73, + "learning_rate": 1.8229270860536007e-05, + "loss": 0.0735, + "step": 13227 + }, + { + "epoch": 1.73, + "learning_rate": 1.822896954637035e-05, + "loss": 0.0783, + "step": 13228 + }, + { + "epoch": 1.73, + "learning_rate": 1.8228668209061155e-05, + "loss": 0.0808, + "step": 13229 + }, + { + "epoch": 1.73, + "learning_rate": 1.822836684860927e-05, + "loss": 0.0706, + "step": 13230 + }, + { + "epoch": 1.73, + "learning_rate": 1.822806546501554e-05, + "loss": 0.0663, + "step": 13231 + }, + { + "epoch": 1.73, + "learning_rate": 1.8227764058280814e-05, + "loss": 0.0718, + "step": 13232 + }, + { + "epoch": 1.73, + "learning_rate": 1.8227462628405938e-05, + "loss": 0.0801, + "step": 13233 + }, + { + "epoch": 1.73, + "learning_rate": 1.8227161175391765e-05, + "loss": 0.0693, + "step": 13234 + }, + { + "epoch": 1.73, + "learning_rate": 1.822685969923914e-05, + "loss": 0.0886, + "step": 13235 + }, + { + "epoch": 1.73, + "learning_rate": 1.8226558199948905e-05, + "loss": 0.0806, + "step": 13236 + }, + { + "epoch": 1.73, + "learning_rate": 1.8226256677521917e-05, + "loss": 0.0716, + "step": 13237 + }, + { + "epoch": 1.73, + "learning_rate": 1.822595513195902e-05, + "loss": 0.0635, + "step": 13238 + }, + { + "epoch": 1.73, + "learning_rate": 1.8225653563261062e-05, + "loss": 0.0868, + "step": 13239 + }, + { + "epoch": 1.73, + "learning_rate": 1.822535197142889e-05, + "loss": 0.0789, + "step": 13240 + }, + { + "epoch": 1.73, + "learning_rate": 1.8225050356463353e-05, + "loss": 0.0891, + "step": 13241 + }, + { + "epoch": 1.73, + "learning_rate": 1.8224748718365305e-05, + "loss": 0.0739, + "step": 13242 + }, + { + "epoch": 1.73, + "learning_rate": 1.8224447057135584e-05, + "loss": 0.0701, + "step": 13243 + }, + { + "epoch": 1.73, + "learning_rate": 1.8224145372775047e-05, + "loss": 0.0758, + "step": 13244 + }, + { + "epoch": 1.74, + "learning_rate": 1.8223843665284536e-05, + "loss": 0.0651, + "step": 13245 + }, + { + "epoch": 1.74, + "learning_rate": 1.8223541934664904e-05, + "loss": 0.0825, + "step": 13246 + }, + { + "epoch": 1.74, + "learning_rate": 1.8223240180917e-05, + "loss": 0.0774, + "step": 13247 + }, + { + "epoch": 1.74, + "learning_rate": 1.8222938404041667e-05, + "loss": 0.0695, + "step": 13248 + }, + { + "epoch": 1.74, + "learning_rate": 1.822263660403976e-05, + "loss": 0.0684, + "step": 13249 + }, + { + "epoch": 1.74, + "learning_rate": 1.8222334780912125e-05, + "loss": 0.0838, + "step": 13250 + }, + { + "epoch": 1.74, + "learning_rate": 1.8222032934659614e-05, + "loss": 0.0851, + "step": 13251 + }, + { + "epoch": 1.74, + "learning_rate": 1.822173106528307e-05, + "loss": 0.0767, + "step": 13252 + }, + { + "epoch": 1.74, + "learning_rate": 1.8221429172783345e-05, + "loss": 0.0846, + "step": 13253 + }, + { + "epoch": 1.74, + "learning_rate": 1.8221127257161288e-05, + "loss": 0.0769, + "step": 13254 + }, + { + "epoch": 1.74, + "learning_rate": 1.8220825318417752e-05, + "loss": 0.0851, + "step": 13255 + }, + { + "epoch": 1.74, + "learning_rate": 1.8220523356553578e-05, + "loss": 0.0755, + "step": 13256 + }, + { + "epoch": 1.74, + "learning_rate": 1.8220221371569623e-05, + "loss": 0.0755, + "step": 13257 + }, + { + "epoch": 1.74, + "learning_rate": 1.8219919363466734e-05, + "loss": 0.0719, + "step": 13258 + }, + { + "epoch": 1.74, + "learning_rate": 1.8219617332245757e-05, + "loss": 0.0664, + "step": 13259 + }, + { + "epoch": 1.74, + "learning_rate": 1.8219315277907544e-05, + "loss": 0.082, + "step": 13260 + }, + { + "epoch": 1.74, + "learning_rate": 1.8219013200452944e-05, + "loss": 0.0701, + "step": 13261 + }, + { + "epoch": 1.74, + "learning_rate": 1.821871109988281e-05, + "loss": 0.0827, + "step": 13262 + }, + { + "epoch": 1.74, + "learning_rate": 1.8218408976197984e-05, + "loss": 0.0749, + "step": 13263 + }, + { + "epoch": 1.74, + "learning_rate": 1.8218106829399325e-05, + "loss": 0.0779, + "step": 13264 + }, + { + "epoch": 1.74, + "learning_rate": 1.8217804659487677e-05, + "loss": 0.0901, + "step": 13265 + }, + { + "epoch": 1.74, + "learning_rate": 1.821750246646389e-05, + "loss": 0.0865, + "step": 13266 + }, + { + "epoch": 1.74, + "learning_rate": 1.8217200250328816e-05, + "loss": 0.0631, + "step": 13267 + }, + { + "epoch": 1.74, + "learning_rate": 1.82168980110833e-05, + "loss": 0.0706, + "step": 13268 + }, + { + "epoch": 1.74, + "learning_rate": 1.82165957487282e-05, + "loss": 0.0712, + "step": 13269 + }, + { + "epoch": 1.74, + "learning_rate": 1.8216293463264363e-05, + "loss": 0.083, + "step": 13270 + }, + { + "epoch": 1.74, + "learning_rate": 1.8215991154692636e-05, + "loss": 0.0772, + "step": 13271 + }, + { + "epoch": 1.74, + "learning_rate": 1.8215688823013874e-05, + "loss": 0.0766, + "step": 13272 + }, + { + "epoch": 1.74, + "learning_rate": 1.821538646822892e-05, + "loss": 0.0782, + "step": 13273 + }, + { + "epoch": 1.74, + "learning_rate": 1.8215084090338636e-05, + "loss": 0.0678, + "step": 13274 + }, + { + "epoch": 1.74, + "learning_rate": 1.8214781689343858e-05, + "loss": 0.0794, + "step": 13275 + }, + { + "epoch": 1.74, + "learning_rate": 1.821447926524545e-05, + "loss": 0.0865, + "step": 13276 + }, + { + "epoch": 1.74, + "learning_rate": 1.821417681804425e-05, + "loss": 0.0666, + "step": 13277 + }, + { + "epoch": 1.74, + "learning_rate": 1.821387434774112e-05, + "loss": 0.0672, + "step": 13278 + }, + { + "epoch": 1.74, + "learning_rate": 1.8213571854336904e-05, + "loss": 0.0657, + "step": 13279 + }, + { + "epoch": 1.74, + "learning_rate": 1.8213269337832456e-05, + "loss": 0.0838, + "step": 13280 + }, + { + "epoch": 1.74, + "learning_rate": 1.8212966798228624e-05, + "loss": 0.0723, + "step": 13281 + }, + { + "epoch": 1.74, + "learning_rate": 1.821266423552626e-05, + "loss": 0.0727, + "step": 13282 + }, + { + "epoch": 1.74, + "learning_rate": 1.821236164972622e-05, + "loss": 0.0739, + "step": 13283 + }, + { + "epoch": 1.74, + "learning_rate": 1.8212059040829345e-05, + "loss": 0.0663, + "step": 13284 + }, + { + "epoch": 1.74, + "learning_rate": 1.821175640883649e-05, + "loss": 0.0869, + "step": 13285 + }, + { + "epoch": 1.74, + "learning_rate": 1.821145375374851e-05, + "loss": 0.0792, + "step": 13286 + }, + { + "epoch": 1.74, + "learning_rate": 1.8211151075566254e-05, + "loss": 0.0821, + "step": 13287 + }, + { + "epoch": 1.74, + "learning_rate": 1.8210848374290574e-05, + "loss": 0.0777, + "step": 13288 + }, + { + "epoch": 1.74, + "learning_rate": 1.8210545649922317e-05, + "loss": 0.0654, + "step": 13289 + }, + { + "epoch": 1.74, + "learning_rate": 1.8210242902462338e-05, + "loss": 0.0638, + "step": 13290 + }, + { + "epoch": 1.74, + "learning_rate": 1.820994013191149e-05, + "loss": 0.087, + "step": 13291 + }, + { + "epoch": 1.74, + "learning_rate": 1.8209637338270624e-05, + "loss": 0.0782, + "step": 13292 + }, + { + "epoch": 1.74, + "learning_rate": 1.8209334521540586e-05, + "loss": 0.0835, + "step": 13293 + }, + { + "epoch": 1.74, + "learning_rate": 1.8209031681722232e-05, + "loss": 0.0797, + "step": 13294 + }, + { + "epoch": 1.74, + "learning_rate": 1.8208728818816417e-05, + "loss": 0.077, + "step": 13295 + }, + { + "epoch": 1.74, + "learning_rate": 1.8208425932823987e-05, + "loss": 0.0603, + "step": 13296 + }, + { + "epoch": 1.74, + "learning_rate": 1.8208123023745795e-05, + "loss": 0.0692, + "step": 13297 + }, + { + "epoch": 1.74, + "learning_rate": 1.8207820091582697e-05, + "loss": 0.0727, + "step": 13298 + }, + { + "epoch": 1.74, + "learning_rate": 1.820751713633554e-05, + "loss": 0.0886, + "step": 13299 + }, + { + "epoch": 1.74, + "learning_rate": 1.820721415800518e-05, + "loss": 0.0735, + "step": 13300 + }, + { + "epoch": 1.74, + "learning_rate": 1.8206911156592465e-05, + "loss": 0.0765, + "step": 13301 + }, + { + "epoch": 1.74, + "learning_rate": 1.8206608132098252e-05, + "loss": 0.0659, + "step": 13302 + }, + { + "epoch": 1.74, + "learning_rate": 1.820630508452339e-05, + "loss": 0.0845, + "step": 13303 + }, + { + "epoch": 1.74, + "learning_rate": 1.8206002013868728e-05, + "loss": 0.0818, + "step": 13304 + }, + { + "epoch": 1.74, + "learning_rate": 1.820569892013513e-05, + "loss": 0.0764, + "step": 13305 + }, + { + "epoch": 1.74, + "learning_rate": 1.8205395803323435e-05, + "loss": 0.0775, + "step": 13306 + }, + { + "epoch": 1.74, + "learning_rate": 1.82050926634345e-05, + "loss": 0.073, + "step": 13307 + }, + { + "epoch": 1.74, + "learning_rate": 1.8204789500469185e-05, + "loss": 0.0694, + "step": 13308 + }, + { + "epoch": 1.74, + "learning_rate": 1.820448631442833e-05, + "loss": 0.0758, + "step": 13309 + }, + { + "epoch": 1.74, + "learning_rate": 1.82041831053128e-05, + "loss": 0.0803, + "step": 13310 + }, + { + "epoch": 1.74, + "learning_rate": 1.820387987312344e-05, + "loss": 0.0778, + "step": 13311 + }, + { + "epoch": 1.74, + "learning_rate": 1.82035766178611e-05, + "loss": 0.073, + "step": 13312 + }, + { + "epoch": 1.74, + "learning_rate": 1.8203273339526643e-05, + "loss": 0.0717, + "step": 13313 + }, + { + "epoch": 1.74, + "learning_rate": 1.8202970038120917e-05, + "loss": 0.073, + "step": 13314 + }, + { + "epoch": 1.74, + "learning_rate": 1.8202666713644772e-05, + "loss": 0.0756, + "step": 13315 + }, + { + "epoch": 1.74, + "learning_rate": 1.8202363366099064e-05, + "loss": 0.0749, + "step": 13316 + }, + { + "epoch": 1.74, + "learning_rate": 1.8202059995484647e-05, + "loss": 0.0844, + "step": 13317 + }, + { + "epoch": 1.74, + "learning_rate": 1.8201756601802373e-05, + "loss": 0.085, + "step": 13318 + }, + { + "epoch": 1.74, + "learning_rate": 1.8201453185053094e-05, + "loss": 0.0658, + "step": 13319 + }, + { + "epoch": 1.74, + "learning_rate": 1.8201149745237664e-05, + "loss": 0.0834, + "step": 13320 + }, + { + "epoch": 1.74, + "learning_rate": 1.8200846282356938e-05, + "loss": 0.0852, + "step": 13321 + }, + { + "epoch": 1.75, + "learning_rate": 1.820054279641177e-05, + "loss": 0.0719, + "step": 13322 + }, + { + "epoch": 1.75, + "learning_rate": 1.820023928740301e-05, + "loss": 0.0812, + "step": 13323 + }, + { + "epoch": 1.75, + "learning_rate": 1.819993575533151e-05, + "loss": 0.0791, + "step": 13324 + }, + { + "epoch": 1.75, + "learning_rate": 1.8199632200198136e-05, + "loss": 0.0834, + "step": 13325 + }, + { + "epoch": 1.75, + "learning_rate": 1.819932862200373e-05, + "loss": 0.0802, + "step": 13326 + }, + { + "epoch": 1.75, + "learning_rate": 1.8199025020749148e-05, + "loss": 0.0567, + "step": 13327 + }, + { + "epoch": 1.75, + "learning_rate": 1.8198721396435245e-05, + "loss": 0.0813, + "step": 13328 + }, + { + "epoch": 1.75, + "learning_rate": 1.8198417749062872e-05, + "loss": 0.0797, + "step": 13329 + }, + { + "epoch": 1.75, + "learning_rate": 1.819811407863289e-05, + "loss": 0.0731, + "step": 13330 + }, + { + "epoch": 1.75, + "learning_rate": 1.8197810385146147e-05, + "loss": 0.0716, + "step": 13331 + }, + { + "epoch": 1.75, + "learning_rate": 1.8197506668603498e-05, + "loss": 0.0745, + "step": 13332 + }, + { + "epoch": 1.75, + "learning_rate": 1.81972029290058e-05, + "loss": 0.0754, + "step": 13333 + }, + { + "epoch": 1.75, + "learning_rate": 1.8196899166353902e-05, + "loss": 0.0773, + "step": 13334 + }, + { + "epoch": 1.75, + "learning_rate": 1.8196595380648662e-05, + "loss": 0.0768, + "step": 13335 + }, + { + "epoch": 1.75, + "learning_rate": 1.819629157189094e-05, + "loss": 0.0733, + "step": 13336 + }, + { + "epoch": 1.75, + "learning_rate": 1.8195987740081578e-05, + "loss": 0.0826, + "step": 13337 + }, + { + "epoch": 1.75, + "learning_rate": 1.8195683885221437e-05, + "loss": 0.0763, + "step": 13338 + }, + { + "epoch": 1.75, + "learning_rate": 1.8195380007311376e-05, + "loss": 0.0803, + "step": 13339 + }, + { + "epoch": 1.75, + "learning_rate": 1.819507610635224e-05, + "loss": 0.0826, + "step": 13340 + }, + { + "epoch": 1.75, + "learning_rate": 1.8194772182344893e-05, + "loss": 0.077, + "step": 13341 + }, + { + "epoch": 1.75, + "learning_rate": 1.8194468235290187e-05, + "loss": 0.0723, + "step": 13342 + }, + { + "epoch": 1.75, + "learning_rate": 1.819416426518897e-05, + "loss": 0.0745, + "step": 13343 + }, + { + "epoch": 1.75, + "learning_rate": 1.8193860272042103e-05, + "loss": 0.0663, + "step": 13344 + }, + { + "epoch": 1.75, + "learning_rate": 1.8193556255850443e-05, + "loss": 0.0835, + "step": 13345 + }, + { + "epoch": 1.75, + "learning_rate": 1.8193252216614842e-05, + "loss": 0.0763, + "step": 13346 + }, + { + "epoch": 1.75, + "learning_rate": 1.8192948154336157e-05, + "loss": 0.0712, + "step": 13347 + }, + { + "epoch": 1.75, + "learning_rate": 1.8192644069015238e-05, + "loss": 0.0742, + "step": 13348 + }, + { + "epoch": 1.75, + "learning_rate": 1.8192339960652946e-05, + "loss": 0.0737, + "step": 13349 + }, + { + "epoch": 1.75, + "learning_rate": 1.819203582925013e-05, + "loss": 0.0706, + "step": 13350 + }, + { + "epoch": 1.75, + "learning_rate": 1.8191731674807655e-05, + "loss": 0.0812, + "step": 13351 + }, + { + "epoch": 1.75, + "learning_rate": 1.819142749732637e-05, + "loss": 0.0691, + "step": 13352 + }, + { + "epoch": 1.75, + "learning_rate": 1.819112329680713e-05, + "loss": 0.066, + "step": 13353 + }, + { + "epoch": 1.75, + "learning_rate": 1.8190819073250793e-05, + "loss": 0.0799, + "step": 13354 + }, + { + "epoch": 1.75, + "learning_rate": 1.8190514826658213e-05, + "loss": 0.0709, + "step": 13355 + }, + { + "epoch": 1.75, + "learning_rate": 1.8190210557030247e-05, + "loss": 0.0827, + "step": 13356 + }, + { + "epoch": 1.75, + "learning_rate": 1.8189906264367748e-05, + "loss": 0.0752, + "step": 13357 + }, + { + "epoch": 1.75, + "learning_rate": 1.8189601948671577e-05, + "loss": 0.0706, + "step": 13358 + }, + { + "epoch": 1.75, + "learning_rate": 1.818929760994258e-05, + "loss": 0.0695, + "step": 13359 + }, + { + "epoch": 1.75, + "learning_rate": 1.818899324818163e-05, + "loss": 0.0734, + "step": 13360 + }, + { + "epoch": 1.75, + "learning_rate": 1.818868886338956e-05, + "loss": 0.0719, + "step": 13361 + }, + { + "epoch": 1.75, + "learning_rate": 1.8188384455567248e-05, + "loss": 0.0672, + "step": 13362 + }, + { + "epoch": 1.75, + "learning_rate": 1.818808002471554e-05, + "loss": 0.0755, + "step": 13363 + }, + { + "epoch": 1.75, + "learning_rate": 1.818777557083529e-05, + "loss": 0.0796, + "step": 13364 + }, + { + "epoch": 1.75, + "learning_rate": 1.8187471093927357e-05, + "loss": 0.0853, + "step": 13365 + }, + { + "epoch": 1.75, + "learning_rate": 1.81871665939926e-05, + "loss": 0.0752, + "step": 13366 + }, + { + "epoch": 1.75, + "learning_rate": 1.818686207103187e-05, + "loss": 0.07, + "step": 13367 + }, + { + "epoch": 1.75, + "learning_rate": 1.818655752504603e-05, + "loss": 0.0721, + "step": 13368 + }, + { + "epoch": 1.75, + "learning_rate": 1.818625295603593e-05, + "loss": 0.0854, + "step": 13369 + }, + { + "epoch": 1.75, + "learning_rate": 1.8185948364002434e-05, + "loss": 0.07, + "step": 13370 + }, + { + "epoch": 1.75, + "learning_rate": 1.8185643748946388e-05, + "loss": 0.0728, + "step": 13371 + }, + { + "epoch": 1.75, + "learning_rate": 1.8185339110868658e-05, + "loss": 0.072, + "step": 13372 + }, + { + "epoch": 1.75, + "learning_rate": 1.81850344497701e-05, + "loss": 0.0796, + "step": 13373 + }, + { + "epoch": 1.75, + "learning_rate": 1.818472976565156e-05, + "loss": 0.0661, + "step": 13374 + }, + { + "epoch": 1.75, + "learning_rate": 1.8184425058513913e-05, + "loss": 0.0759, + "step": 13375 + }, + { + "epoch": 1.75, + "learning_rate": 1.8184120328358003e-05, + "loss": 0.0627, + "step": 13376 + }, + { + "epoch": 1.75, + "learning_rate": 1.818381557518469e-05, + "loss": 0.0812, + "step": 13377 + }, + { + "epoch": 1.75, + "learning_rate": 1.818351079899483e-05, + "loss": 0.0755, + "step": 13378 + }, + { + "epoch": 1.75, + "learning_rate": 1.8183205999789284e-05, + "loss": 0.0769, + "step": 13379 + }, + { + "epoch": 1.75, + "learning_rate": 1.8182901177568905e-05, + "loss": 0.0724, + "step": 13380 + }, + { + "epoch": 1.75, + "learning_rate": 1.8182596332334556e-05, + "loss": 0.0768, + "step": 13381 + }, + { + "epoch": 1.75, + "learning_rate": 1.8182291464087084e-05, + "loss": 0.0671, + "step": 13382 + }, + { + "epoch": 1.75, + "learning_rate": 1.818198657282736e-05, + "loss": 0.0792, + "step": 13383 + }, + { + "epoch": 1.75, + "learning_rate": 1.818168165855623e-05, + "loss": 0.0733, + "step": 13384 + }, + { + "epoch": 1.75, + "learning_rate": 1.8181376721274558e-05, + "loss": 0.074, + "step": 13385 + }, + { + "epoch": 1.75, + "learning_rate": 1.81810717609832e-05, + "loss": 0.0862, + "step": 13386 + }, + { + "epoch": 1.75, + "learning_rate": 1.8180766777683013e-05, + "loss": 0.0666, + "step": 13387 + }, + { + "epoch": 1.75, + "learning_rate": 1.818046177137485e-05, + "loss": 0.0801, + "step": 13388 + }, + { + "epoch": 1.75, + "learning_rate": 1.818015674205958e-05, + "loss": 0.0841, + "step": 13389 + }, + { + "epoch": 1.75, + "learning_rate": 1.8179851689738056e-05, + "loss": 0.079, + "step": 13390 + }, + { + "epoch": 1.75, + "learning_rate": 1.817954661441113e-05, + "loss": 0.077, + "step": 13391 + }, + { + "epoch": 1.75, + "learning_rate": 1.8179241516079667e-05, + "loss": 0.0754, + "step": 13392 + }, + { + "epoch": 1.75, + "learning_rate": 1.817893639474452e-05, + "loss": 0.0817, + "step": 13393 + }, + { + "epoch": 1.75, + "learning_rate": 1.817863125040655e-05, + "loss": 0.0721, + "step": 13394 + }, + { + "epoch": 1.75, + "learning_rate": 1.8178326083066623e-05, + "loss": 0.0705, + "step": 13395 + }, + { + "epoch": 1.75, + "learning_rate": 1.8178020892725583e-05, + "loss": 0.0778, + "step": 13396 + }, + { + "epoch": 1.75, + "learning_rate": 1.8177715679384295e-05, + "loss": 0.0718, + "step": 13397 + }, + { + "epoch": 1.76, + "learning_rate": 1.8177410443043616e-05, + "loss": 0.0723, + "step": 13398 + }, + { + "epoch": 1.76, + "learning_rate": 1.817710518370441e-05, + "loss": 0.068, + "step": 13399 + }, + { + "epoch": 1.76, + "learning_rate": 1.8176799901367528e-05, + "loss": 0.0755, + "step": 13400 + }, + { + "epoch": 1.76, + "learning_rate": 1.817649459603383e-05, + "loss": 0.0813, + "step": 13401 + }, + { + "epoch": 1.76, + "learning_rate": 1.8176189267704177e-05, + "loss": 0.0712, + "step": 13402 + }, + { + "epoch": 1.76, + "learning_rate": 1.8175883916379433e-05, + "loss": 0.0752, + "step": 13403 + }, + { + "epoch": 1.76, + "learning_rate": 1.8175578542060445e-05, + "loss": 0.0805, + "step": 13404 + }, + { + "epoch": 1.76, + "learning_rate": 1.817527314474808e-05, + "loss": 0.0799, + "step": 13405 + }, + { + "epoch": 1.76, + "learning_rate": 1.8174967724443195e-05, + "loss": 0.0737, + "step": 13406 + }, + { + "epoch": 1.76, + "learning_rate": 1.8174662281146647e-05, + "loss": 0.0656, + "step": 13407 + }, + { + "epoch": 1.76, + "learning_rate": 1.8174356814859298e-05, + "loss": 0.0705, + "step": 13408 + }, + { + "epoch": 1.76, + "learning_rate": 1.8174051325582e-05, + "loss": 0.0721, + "step": 13409 + }, + { + "epoch": 1.76, + "learning_rate": 1.8173745813315627e-05, + "loss": 0.0878, + "step": 13410 + }, + { + "epoch": 1.76, + "learning_rate": 1.8173440278061022e-05, + "loss": 0.0648, + "step": 13411 + }, + { + "epoch": 1.76, + "learning_rate": 1.8173134719819058e-05, + "loss": 0.0744, + "step": 13412 + }, + { + "epoch": 1.76, + "learning_rate": 1.817282913859058e-05, + "loss": 0.0678, + "step": 13413 + }, + { + "epoch": 1.76, + "learning_rate": 1.8172523534376463e-05, + "loss": 0.0656, + "step": 13414 + }, + { + "epoch": 1.76, + "learning_rate": 1.8172217907177553e-05, + "loss": 0.0794, + "step": 13415 + }, + { + "epoch": 1.76, + "learning_rate": 1.817191225699472e-05, + "loss": 0.0845, + "step": 13416 + }, + { + "epoch": 1.76, + "learning_rate": 1.8171606583828816e-05, + "loss": 0.0833, + "step": 13417 + }, + { + "epoch": 1.76, + "learning_rate": 1.8171300887680707e-05, + "loss": 0.0699, + "step": 13418 + }, + { + "epoch": 1.76, + "learning_rate": 1.8170995168551244e-05, + "loss": 0.0667, + "step": 13419 + }, + { + "epoch": 1.76, + "learning_rate": 1.8170689426441294e-05, + "loss": 0.0725, + "step": 13420 + }, + { + "epoch": 1.76, + "learning_rate": 1.817038366135172e-05, + "loss": 0.0677, + "step": 13421 + }, + { + "epoch": 1.76, + "learning_rate": 1.817007787328337e-05, + "loss": 0.0745, + "step": 13422 + }, + { + "epoch": 1.76, + "learning_rate": 1.8169772062237118e-05, + "loss": 0.0775, + "step": 13423 + }, + { + "epoch": 1.76, + "learning_rate": 1.816946622821381e-05, + "loss": 0.0658, + "step": 13424 + }, + { + "epoch": 1.76, + "learning_rate": 1.816916037121432e-05, + "loss": 0.0922, + "step": 13425 + }, + { + "epoch": 1.76, + "learning_rate": 1.8168854491239494e-05, + "loss": 0.0777, + "step": 13426 + }, + { + "epoch": 1.76, + "learning_rate": 1.8168548588290205e-05, + "loss": 0.0819, + "step": 13427 + }, + { + "epoch": 1.76, + "learning_rate": 1.8168242662367305e-05, + "loss": 0.0629, + "step": 13428 + }, + { + "epoch": 1.76, + "learning_rate": 1.816793671347166e-05, + "loss": 0.0711, + "step": 13429 + }, + { + "epoch": 1.76, + "learning_rate": 1.816763074160413e-05, + "loss": 0.0713, + "step": 13430 + }, + { + "epoch": 1.76, + "learning_rate": 1.8167324746765568e-05, + "loss": 0.0736, + "step": 13431 + }, + { + "epoch": 1.76, + "learning_rate": 1.8167018728956846e-05, + "loss": 0.071, + "step": 13432 + }, + { + "epoch": 1.76, + "learning_rate": 1.8166712688178812e-05, + "loss": 0.0755, + "step": 13433 + }, + { + "epoch": 1.76, + "learning_rate": 1.816640662443234e-05, + "loss": 0.0729, + "step": 13434 + }, + { + "epoch": 1.76, + "learning_rate": 1.8166100537718282e-05, + "loss": 0.0757, + "step": 13435 + }, + { + "epoch": 1.76, + "learning_rate": 1.81657944280375e-05, + "loss": 0.0743, + "step": 13436 + }, + { + "epoch": 1.76, + "learning_rate": 1.8165488295390856e-05, + "loss": 0.0754, + "step": 13437 + }, + { + "epoch": 1.76, + "learning_rate": 1.816518213977921e-05, + "loss": 0.0718, + "step": 13438 + }, + { + "epoch": 1.76, + "learning_rate": 1.816487596120343e-05, + "loss": 0.0784, + "step": 13439 + }, + { + "epoch": 1.76, + "learning_rate": 1.8164569759664362e-05, + "loss": 0.0684, + "step": 13440 + }, + { + "epoch": 1.76, + "learning_rate": 1.8164263535162884e-05, + "loss": 0.07, + "step": 13441 + }, + { + "epoch": 1.76, + "learning_rate": 1.8163957287699846e-05, + "loss": 0.0803, + "step": 13442 + }, + { + "epoch": 1.76, + "learning_rate": 1.8163651017276114e-05, + "loss": 0.0752, + "step": 13443 + }, + { + "epoch": 1.76, + "learning_rate": 1.8163344723892547e-05, + "loss": 0.0666, + "step": 13444 + }, + { + "epoch": 1.76, + "learning_rate": 1.816303840755001e-05, + "loss": 0.0794, + "step": 13445 + }, + { + "epoch": 1.76, + "learning_rate": 1.816273206824936e-05, + "loss": 0.079, + "step": 13446 + }, + { + "epoch": 1.76, + "learning_rate": 1.816242570599146e-05, + "loss": 0.0851, + "step": 13447 + }, + { + "epoch": 1.76, + "learning_rate": 1.8162119320777173e-05, + "loss": 0.0865, + "step": 13448 + }, + { + "epoch": 1.76, + "learning_rate": 1.8161812912607364e-05, + "loss": 0.071, + "step": 13449 + }, + { + "epoch": 1.76, + "learning_rate": 1.8161506481482887e-05, + "loss": 0.0839, + "step": 13450 + }, + { + "epoch": 1.76, + "learning_rate": 1.8161200027404607e-05, + "loss": 0.0755, + "step": 13451 + }, + { + "epoch": 1.76, + "learning_rate": 1.816089355037339e-05, + "loss": 0.0753, + "step": 13452 + }, + { + "epoch": 1.76, + "learning_rate": 1.8160587050390093e-05, + "loss": 0.07, + "step": 13453 + }, + { + "epoch": 1.76, + "learning_rate": 1.816028052745558e-05, + "loss": 0.0702, + "step": 13454 + }, + { + "epoch": 1.76, + "learning_rate": 1.815997398157071e-05, + "loss": 0.0798, + "step": 13455 + }, + { + "epoch": 1.76, + "learning_rate": 1.815966741273635e-05, + "loss": 0.0805, + "step": 13456 + }, + { + "epoch": 1.76, + "learning_rate": 1.8159360820953357e-05, + "loss": 0.0753, + "step": 13457 + }, + { + "epoch": 1.76, + "learning_rate": 1.81590542062226e-05, + "loss": 0.0691, + "step": 13458 + }, + { + "epoch": 1.76, + "learning_rate": 1.8158747568544934e-05, + "loss": 0.0794, + "step": 13459 + }, + { + "epoch": 1.76, + "learning_rate": 1.8158440907921228e-05, + "loss": 0.0728, + "step": 13460 + }, + { + "epoch": 1.76, + "learning_rate": 1.815813422435234e-05, + "loss": 0.0794, + "step": 13461 + }, + { + "epoch": 1.76, + "learning_rate": 1.8157827517839133e-05, + "loss": 0.0743, + "step": 13462 + }, + { + "epoch": 1.76, + "learning_rate": 1.8157520788382473e-05, + "loss": 0.0825, + "step": 13463 + }, + { + "epoch": 1.76, + "learning_rate": 1.8157214035983218e-05, + "loss": 0.0875, + "step": 13464 + }, + { + "epoch": 1.76, + "learning_rate": 1.8156907260642234e-05, + "loss": 0.0709, + "step": 13465 + }, + { + "epoch": 1.76, + "learning_rate": 1.8156600462360386e-05, + "loss": 0.0817, + "step": 13466 + }, + { + "epoch": 1.76, + "learning_rate": 1.8156293641138528e-05, + "loss": 0.079, + "step": 13467 + }, + { + "epoch": 1.76, + "learning_rate": 1.815598679697753e-05, + "loss": 0.0791, + "step": 13468 + }, + { + "epoch": 1.76, + "learning_rate": 1.8155679929878257e-05, + "loss": 0.0696, + "step": 13469 + }, + { + "epoch": 1.76, + "learning_rate": 1.8155373039841562e-05, + "loss": 0.0778, + "step": 13470 + }, + { + "epoch": 1.76, + "learning_rate": 1.8155066126868322e-05, + "loss": 0.0663, + "step": 13471 + }, + { + "epoch": 1.76, + "learning_rate": 1.815475919095939e-05, + "loss": 0.0643, + "step": 13472 + }, + { + "epoch": 1.76, + "learning_rate": 1.8154452232115626e-05, + "loss": 0.0766, + "step": 13473 + }, + { + "epoch": 1.76, + "learning_rate": 1.8154145250337903e-05, + "loss": 0.0876, + "step": 13474 + }, + { + "epoch": 1.77, + "learning_rate": 1.8153838245627086e-05, + "loss": 0.07, + "step": 13475 + }, + { + "epoch": 1.77, + "learning_rate": 1.8153531217984027e-05, + "loss": 0.0758, + "step": 13476 + }, + { + "epoch": 1.77, + "learning_rate": 1.8153224167409597e-05, + "loss": 0.0742, + "step": 13477 + }, + { + "epoch": 1.77, + "learning_rate": 1.815291709390466e-05, + "loss": 0.0707, + "step": 13478 + }, + { + "epoch": 1.77, + "learning_rate": 1.815260999747008e-05, + "loss": 0.0761, + "step": 13479 + }, + { + "epoch": 1.77, + "learning_rate": 1.8152302878106715e-05, + "loss": 0.0754, + "step": 13480 + }, + { + "epoch": 1.77, + "learning_rate": 1.8151995735815432e-05, + "loss": 0.0754, + "step": 13481 + }, + { + "epoch": 1.77, + "learning_rate": 1.8151688570597096e-05, + "loss": 0.072, + "step": 13482 + }, + { + "epoch": 1.77, + "learning_rate": 1.815138138245257e-05, + "loss": 0.0683, + "step": 13483 + }, + { + "epoch": 1.77, + "learning_rate": 1.815107417138272e-05, + "loss": 0.0767, + "step": 13484 + }, + { + "epoch": 1.77, + "learning_rate": 1.81507669373884e-05, + "loss": 0.08, + "step": 13485 + }, + { + "epoch": 1.77, + "learning_rate": 1.815045968047049e-05, + "loss": 0.0862, + "step": 13486 + }, + { + "epoch": 1.77, + "learning_rate": 1.8150152400629847e-05, + "loss": 0.0743, + "step": 13487 + }, + { + "epoch": 1.77, + "learning_rate": 1.814984509786733e-05, + "loss": 0.0727, + "step": 13488 + }, + { + "epoch": 1.77, + "learning_rate": 1.814953777218381e-05, + "loss": 0.0794, + "step": 13489 + }, + { + "epoch": 1.77, + "learning_rate": 1.8149230423580148e-05, + "loss": 0.0733, + "step": 13490 + }, + { + "epoch": 1.77, + "learning_rate": 1.814892305205721e-05, + "loss": 0.077, + "step": 13491 + }, + { + "epoch": 1.77, + "learning_rate": 1.8148615657615857e-05, + "loss": 0.0792, + "step": 13492 + }, + { + "epoch": 1.77, + "learning_rate": 1.814830824025696e-05, + "loss": 0.0763, + "step": 13493 + }, + { + "epoch": 1.77, + "learning_rate": 1.814800079998138e-05, + "loss": 0.0812, + "step": 13494 + }, + { + "epoch": 1.77, + "learning_rate": 1.814769333678998e-05, + "loss": 0.0756, + "step": 13495 + }, + { + "epoch": 1.77, + "learning_rate": 1.8147385850683626e-05, + "loss": 0.0686, + "step": 13496 + }, + { + "epoch": 1.77, + "learning_rate": 1.814707834166318e-05, + "loss": 0.0817, + "step": 13497 + }, + { + "epoch": 1.77, + "learning_rate": 1.8146770809729516e-05, + "loss": 0.0629, + "step": 13498 + }, + { + "epoch": 1.77, + "learning_rate": 1.814646325488349e-05, + "loss": 0.0733, + "step": 13499 + }, + { + "epoch": 1.77, + "learning_rate": 1.8146155677125974e-05, + "loss": 0.0731, + "step": 13500 + }, + { + "epoch": 1.77, + "learning_rate": 1.8145848076457825e-05, + "loss": 0.0726, + "step": 13501 + }, + { + "epoch": 1.77, + "learning_rate": 1.8145540452879915e-05, + "loss": 0.077, + "step": 13502 + }, + { + "epoch": 1.77, + "learning_rate": 1.8145232806393104e-05, + "loss": 0.0774, + "step": 13503 + }, + { + "epoch": 1.77, + "learning_rate": 1.814492513699826e-05, + "loss": 0.0836, + "step": 13504 + }, + { + "epoch": 1.77, + "learning_rate": 1.8144617444696247e-05, + "loss": 0.0637, + "step": 13505 + }, + { + "epoch": 1.77, + "learning_rate": 1.814430972948793e-05, + "loss": 0.0895, + "step": 13506 + }, + { + "epoch": 1.77, + "learning_rate": 1.814400199137418e-05, + "loss": 0.0834, + "step": 13507 + }, + { + "epoch": 1.77, + "learning_rate": 1.8143694230355856e-05, + "loss": 0.07, + "step": 13508 + }, + { + "epoch": 1.77, + "learning_rate": 1.8143386446433822e-05, + "loss": 0.074, + "step": 13509 + }, + { + "epoch": 1.77, + "learning_rate": 1.814307863960895e-05, + "loss": 0.0703, + "step": 13510 + }, + { + "epoch": 1.77, + "learning_rate": 1.8142770809882102e-05, + "loss": 0.0848, + "step": 13511 + }, + { + "epoch": 1.77, + "learning_rate": 1.8142462957254147e-05, + "loss": 0.0709, + "step": 13512 + }, + { + "epoch": 1.77, + "learning_rate": 1.8142155081725948e-05, + "loss": 0.0812, + "step": 13513 + }, + { + "epoch": 1.77, + "learning_rate": 1.8141847183298372e-05, + "loss": 0.0761, + "step": 13514 + }, + { + "epoch": 1.77, + "learning_rate": 1.814153926197228e-05, + "loss": 0.0733, + "step": 13515 + }, + { + "epoch": 1.77, + "learning_rate": 1.8141231317748547e-05, + "loss": 0.0695, + "step": 13516 + }, + { + "epoch": 1.77, + "learning_rate": 1.8140923350628036e-05, + "loss": 0.0694, + "step": 13517 + }, + { + "epoch": 1.77, + "learning_rate": 1.8140615360611607e-05, + "loss": 0.0749, + "step": 13518 + }, + { + "epoch": 1.77, + "learning_rate": 1.8140307347700132e-05, + "loss": 0.0744, + "step": 13519 + }, + { + "epoch": 1.77, + "learning_rate": 1.813999931189448e-05, + "loss": 0.0748, + "step": 13520 + }, + { + "epoch": 1.77, + "learning_rate": 1.8139691253195506e-05, + "loss": 0.0718, + "step": 13521 + }, + { + "epoch": 1.77, + "learning_rate": 1.813938317160409e-05, + "loss": 0.0807, + "step": 13522 + }, + { + "epoch": 1.77, + "learning_rate": 1.813907506712109e-05, + "loss": 0.0791, + "step": 13523 + }, + { + "epoch": 1.77, + "learning_rate": 1.8138766939747375e-05, + "loss": 0.0659, + "step": 13524 + }, + { + "epoch": 1.77, + "learning_rate": 1.8138458789483812e-05, + "loss": 0.0839, + "step": 13525 + }, + { + "epoch": 1.77, + "learning_rate": 1.8138150616331264e-05, + "loss": 0.0695, + "step": 13526 + }, + { + "epoch": 1.77, + "learning_rate": 1.8137842420290606e-05, + "loss": 0.0742, + "step": 13527 + }, + { + "epoch": 1.77, + "learning_rate": 1.8137534201362696e-05, + "loss": 0.0671, + "step": 13528 + }, + { + "epoch": 1.77, + "learning_rate": 1.8137225959548404e-05, + "loss": 0.0859, + "step": 13529 + }, + { + "epoch": 1.77, + "learning_rate": 1.81369176948486e-05, + "loss": 0.0705, + "step": 13530 + }, + { + "epoch": 1.77, + "learning_rate": 1.8136609407264144e-05, + "loss": 0.0833, + "step": 13531 + }, + { + "epoch": 1.77, + "learning_rate": 1.813630109679591e-05, + "loss": 0.0795, + "step": 13532 + }, + { + "epoch": 1.77, + "learning_rate": 1.8135992763444763e-05, + "loss": 0.0804, + "step": 13533 + }, + { + "epoch": 1.77, + "learning_rate": 1.813568440721157e-05, + "loss": 0.0737, + "step": 13534 + }, + { + "epoch": 1.77, + "learning_rate": 1.8135376028097198e-05, + "loss": 0.0778, + "step": 13535 + }, + { + "epoch": 1.77, + "learning_rate": 1.813506762610251e-05, + "loss": 0.0853, + "step": 13536 + }, + { + "epoch": 1.77, + "learning_rate": 1.8134759201228383e-05, + "loss": 0.0803, + "step": 13537 + }, + { + "epoch": 1.77, + "learning_rate": 1.8134450753475675e-05, + "loss": 0.0693, + "step": 13538 + }, + { + "epoch": 1.77, + "learning_rate": 1.8134142282845258e-05, + "loss": 0.0678, + "step": 13539 + }, + { + "epoch": 1.77, + "learning_rate": 1.8133833789338006e-05, + "loss": 0.0671, + "step": 13540 + }, + { + "epoch": 1.77, + "learning_rate": 1.813352527295477e-05, + "loss": 0.0775, + "step": 13541 + }, + { + "epoch": 1.77, + "learning_rate": 1.813321673369643e-05, + "loss": 0.068, + "step": 13542 + }, + { + "epoch": 1.77, + "learning_rate": 1.8132908171563856e-05, + "loss": 0.0636, + "step": 13543 + }, + { + "epoch": 1.77, + "learning_rate": 1.8132599586557906e-05, + "loss": 0.075, + "step": 13544 + }, + { + "epoch": 1.77, + "learning_rate": 1.8132290978679454e-05, + "loss": 0.0751, + "step": 13545 + }, + { + "epoch": 1.77, + "learning_rate": 1.8131982347929367e-05, + "loss": 0.0706, + "step": 13546 + }, + { + "epoch": 1.77, + "learning_rate": 1.813167369430851e-05, + "loss": 0.0839, + "step": 13547 + }, + { + "epoch": 1.77, + "learning_rate": 1.8131365017817756e-05, + "loss": 0.0799, + "step": 13548 + }, + { + "epoch": 1.77, + "learning_rate": 1.813105631845797e-05, + "loss": 0.081, + "step": 13549 + }, + { + "epoch": 1.77, + "learning_rate": 1.8130747596230023e-05, + "loss": 0.0804, + "step": 13550 + }, + { + "epoch": 1.78, + "learning_rate": 1.813043885113478e-05, + "loss": 0.0826, + "step": 13551 + }, + { + "epoch": 1.78, + "learning_rate": 1.8130130083173113e-05, + "loss": 0.0803, + "step": 13552 + }, + { + "epoch": 1.78, + "learning_rate": 1.8129821292345884e-05, + "loss": 0.0615, + "step": 13553 + }, + { + "epoch": 1.78, + "learning_rate": 1.8129512478653967e-05, + "loss": 0.0729, + "step": 13554 + }, + { + "epoch": 1.78, + "learning_rate": 1.812920364209823e-05, + "loss": 0.0755, + "step": 13555 + }, + { + "epoch": 1.78, + "learning_rate": 1.812889478267954e-05, + "loss": 0.0629, + "step": 13556 + }, + { + "epoch": 1.78, + "learning_rate": 1.8128585900398767e-05, + "loss": 0.0692, + "step": 13557 + }, + { + "epoch": 1.78, + "learning_rate": 1.812827699525678e-05, + "loss": 0.073, + "step": 13558 + }, + { + "epoch": 1.78, + "learning_rate": 1.8127968067254446e-05, + "loss": 0.0737, + "step": 13559 + }, + { + "epoch": 1.78, + "learning_rate": 1.812765911639263e-05, + "loss": 0.0736, + "step": 13560 + }, + { + "epoch": 1.78, + "learning_rate": 1.8127350142672212e-05, + "loss": 0.0893, + "step": 13561 + }, + { + "epoch": 1.78, + "learning_rate": 1.8127041146094054e-05, + "loss": 0.0755, + "step": 13562 + }, + { + "epoch": 1.78, + "learning_rate": 1.812673212665902e-05, + "loss": 0.0727, + "step": 13563 + }, + { + "epoch": 1.78, + "learning_rate": 1.812642308436799e-05, + "loss": 0.071, + "step": 13564 + }, + { + "epoch": 1.78, + "learning_rate": 1.8126114019221828e-05, + "loss": 0.0788, + "step": 13565 + }, + { + "epoch": 1.78, + "learning_rate": 1.81258049312214e-05, + "loss": 0.0768, + "step": 13566 + }, + { + "epoch": 1.78, + "learning_rate": 1.8125495820367582e-05, + "loss": 0.0729, + "step": 13567 + }, + { + "epoch": 1.78, + "learning_rate": 1.8125186686661237e-05, + "loss": 0.0638, + "step": 13568 + }, + { + "epoch": 1.78, + "learning_rate": 1.8124877530103234e-05, + "loss": 0.0782, + "step": 13569 + }, + { + "epoch": 1.78, + "learning_rate": 1.812456835069445e-05, + "loss": 0.0763, + "step": 13570 + }, + { + "epoch": 1.78, + "learning_rate": 1.812425914843575e-05, + "loss": 0.0755, + "step": 13571 + }, + { + "epoch": 1.78, + "learning_rate": 1.8123949923328005e-05, + "loss": 0.0836, + "step": 13572 + }, + { + "epoch": 1.78, + "learning_rate": 1.812364067537208e-05, + "loss": 0.0815, + "step": 13573 + }, + { + "epoch": 1.78, + "learning_rate": 1.8123331404568853e-05, + "loss": 0.0808, + "step": 13574 + }, + { + "epoch": 1.78, + "learning_rate": 1.812302211091918e-05, + "loss": 0.0787, + "step": 13575 + }, + { + "epoch": 1.78, + "learning_rate": 1.812271279442395e-05, + "loss": 0.0694, + "step": 13576 + }, + { + "epoch": 1.78, + "learning_rate": 1.8122403455084015e-05, + "loss": 0.0644, + "step": 13577 + }, + { + "epoch": 1.78, + "learning_rate": 1.8122094092900257e-05, + "loss": 0.0793, + "step": 13578 + }, + { + "epoch": 1.78, + "learning_rate": 1.8121784707873545e-05, + "loss": 0.0812, + "step": 13579 + }, + { + "epoch": 1.78, + "learning_rate": 1.812147530000474e-05, + "loss": 0.0743, + "step": 13580 + }, + { + "epoch": 1.78, + "learning_rate": 1.812116586929472e-05, + "loss": 0.0716, + "step": 13581 + }, + { + "epoch": 1.78, + "learning_rate": 1.8120856415744354e-05, + "loss": 0.0713, + "step": 13582 + }, + { + "epoch": 1.78, + "learning_rate": 1.8120546939354513e-05, + "loss": 0.0612, + "step": 13583 + }, + { + "epoch": 1.78, + "learning_rate": 1.8120237440126066e-05, + "loss": 0.0855, + "step": 13584 + }, + { + "epoch": 1.78, + "learning_rate": 1.811992791805988e-05, + "loss": 0.0614, + "step": 13585 + }, + { + "epoch": 1.78, + "learning_rate": 1.8119618373156832e-05, + "loss": 0.0731, + "step": 13586 + }, + { + "epoch": 1.78, + "learning_rate": 1.811930880541779e-05, + "loss": 0.073, + "step": 13587 + }, + { + "epoch": 1.78, + "learning_rate": 1.8118999214843623e-05, + "loss": 0.0743, + "step": 13588 + }, + { + "epoch": 1.78, + "learning_rate": 1.8118689601435205e-05, + "loss": 0.0789, + "step": 13589 + }, + { + "epoch": 1.78, + "learning_rate": 1.81183799651934e-05, + "loss": 0.0707, + "step": 13590 + }, + { + "epoch": 1.78, + "learning_rate": 1.8118070306119093e-05, + "loss": 0.0718, + "step": 13591 + }, + { + "epoch": 1.78, + "learning_rate": 1.811776062421314e-05, + "loss": 0.0844, + "step": 13592 + }, + { + "epoch": 1.78, + "learning_rate": 1.8117450919476416e-05, + "loss": 0.0569, + "step": 13593 + }, + { + "epoch": 1.78, + "learning_rate": 1.8117141191909794e-05, + "loss": 0.0746, + "step": 13594 + }, + { + "epoch": 1.78, + "learning_rate": 1.8116831441514146e-05, + "loss": 0.0758, + "step": 13595 + }, + { + "epoch": 1.78, + "learning_rate": 1.8116521668290343e-05, + "loss": 0.0823, + "step": 13596 + }, + { + "epoch": 1.78, + "learning_rate": 1.8116211872239253e-05, + "loss": 0.0804, + "step": 13597 + }, + { + "epoch": 1.78, + "learning_rate": 1.811590205336175e-05, + "loss": 0.0817, + "step": 13598 + }, + { + "epoch": 1.78, + "learning_rate": 1.8115592211658705e-05, + "loss": 0.0744, + "step": 13599 + }, + { + "epoch": 1.78, + "learning_rate": 1.8115282347130988e-05, + "loss": 0.0698, + "step": 13600 + }, + { + "epoch": 1.78, + "learning_rate": 1.811497245977947e-05, + "loss": 0.079, + "step": 13601 + }, + { + "epoch": 1.78, + "learning_rate": 1.811466254960503e-05, + "loss": 0.0785, + "step": 13602 + }, + { + "epoch": 1.78, + "learning_rate": 1.811435261660853e-05, + "loss": 0.0724, + "step": 13603 + }, + { + "epoch": 1.78, + "learning_rate": 1.8114042660790844e-05, + "loss": 0.0762, + "step": 13604 + }, + { + "epoch": 1.78, + "learning_rate": 1.8113732682152844e-05, + "loss": 0.0871, + "step": 13605 + }, + { + "epoch": 1.78, + "learning_rate": 1.8113422680695407e-05, + "loss": 0.0824, + "step": 13606 + }, + { + "epoch": 1.78, + "learning_rate": 1.81131126564194e-05, + "loss": 0.0752, + "step": 13607 + }, + { + "epoch": 1.78, + "learning_rate": 1.8112802609325696e-05, + "loss": 0.0684, + "step": 13608 + }, + { + "epoch": 1.78, + "learning_rate": 1.8112492539415166e-05, + "loss": 0.0652, + "step": 13609 + }, + { + "epoch": 1.78, + "learning_rate": 1.8112182446688683e-05, + "loss": 0.0678, + "step": 13610 + }, + { + "epoch": 1.78, + "learning_rate": 1.8111872331147116e-05, + "loss": 0.0736, + "step": 13611 + }, + { + "epoch": 1.78, + "learning_rate": 1.811156219279134e-05, + "loss": 0.0673, + "step": 13612 + }, + { + "epoch": 1.78, + "learning_rate": 1.8111252031622233e-05, + "loss": 0.0841, + "step": 13613 + }, + { + "epoch": 1.78, + "learning_rate": 1.8110941847640658e-05, + "loss": 0.0745, + "step": 13614 + }, + { + "epoch": 1.78, + "learning_rate": 1.811063164084749e-05, + "loss": 0.0802, + "step": 13615 + }, + { + "epoch": 1.78, + "learning_rate": 1.8110321411243602e-05, + "loss": 0.0685, + "step": 13616 + }, + { + "epoch": 1.78, + "learning_rate": 1.811001115882987e-05, + "loss": 0.0657, + "step": 13617 + }, + { + "epoch": 1.78, + "learning_rate": 1.8109700883607163e-05, + "loss": 0.0684, + "step": 13618 + }, + { + "epoch": 1.78, + "learning_rate": 1.8109390585576352e-05, + "loss": 0.0727, + "step": 13619 + }, + { + "epoch": 1.78, + "learning_rate": 1.8109080264738312e-05, + "loss": 0.0686, + "step": 13620 + }, + { + "epoch": 1.78, + "learning_rate": 1.8108769921093917e-05, + "loss": 0.0826, + "step": 13621 + }, + { + "epoch": 1.78, + "learning_rate": 1.8108459554644037e-05, + "loss": 0.0825, + "step": 13622 + }, + { + "epoch": 1.78, + "learning_rate": 1.8108149165389547e-05, + "loss": 0.0616, + "step": 13623 + }, + { + "epoch": 1.78, + "learning_rate": 1.810783875333132e-05, + "loss": 0.0677, + "step": 13624 + }, + { + "epoch": 1.78, + "learning_rate": 1.8107528318470228e-05, + "loss": 0.0764, + "step": 13625 + }, + { + "epoch": 1.78, + "learning_rate": 1.8107217860807145e-05, + "loss": 0.0882, + "step": 13626 + }, + { + "epoch": 1.79, + "learning_rate": 1.8106907380342944e-05, + "loss": 0.0765, + "step": 13627 + }, + { + "epoch": 1.79, + "learning_rate": 1.8106596877078496e-05, + "loss": 0.0777, + "step": 13628 + }, + { + "epoch": 1.79, + "learning_rate": 1.8106286351014676e-05, + "loss": 0.0778, + "step": 13629 + }, + { + "epoch": 1.79, + "learning_rate": 1.8105975802152362e-05, + "loss": 0.0788, + "step": 13630 + }, + { + "epoch": 1.79, + "learning_rate": 1.8105665230492416e-05, + "loss": 0.0757, + "step": 13631 + }, + { + "epoch": 1.79, + "learning_rate": 1.810535463603572e-05, + "loss": 0.0693, + "step": 13632 + }, + { + "epoch": 1.79, + "learning_rate": 1.8105044018783146e-05, + "loss": 0.0836, + "step": 13633 + }, + { + "epoch": 1.79, + "learning_rate": 1.810473337873557e-05, + "loss": 0.0699, + "step": 13634 + }, + { + "epoch": 1.79, + "learning_rate": 1.810442271589386e-05, + "loss": 0.0668, + "step": 13635 + }, + { + "epoch": 1.79, + "learning_rate": 1.8104112030258895e-05, + "loss": 0.0759, + "step": 13636 + }, + { + "epoch": 1.79, + "learning_rate": 1.810380132183154e-05, + "loss": 0.0789, + "step": 13637 + }, + { + "epoch": 1.79, + "learning_rate": 1.8103490590612682e-05, + "loss": 0.0763, + "step": 13638 + }, + { + "epoch": 1.79, + "learning_rate": 1.8103179836603186e-05, + "loss": 0.0788, + "step": 13639 + }, + { + "epoch": 1.79, + "learning_rate": 1.810286905980393e-05, + "loss": 0.0706, + "step": 13640 + }, + { + "epoch": 1.79, + "learning_rate": 1.8102558260215782e-05, + "loss": 0.0782, + "step": 13641 + }, + { + "epoch": 1.79, + "learning_rate": 1.8102247437839624e-05, + "loss": 0.0745, + "step": 13642 + }, + { + "epoch": 1.79, + "learning_rate": 1.8101936592676324e-05, + "loss": 0.0725, + "step": 13643 + }, + { + "epoch": 1.79, + "learning_rate": 1.8101625724726765e-05, + "loss": 0.0776, + "step": 13644 + }, + { + "epoch": 1.79, + "learning_rate": 1.810131483399181e-05, + "loss": 0.0877, + "step": 13645 + }, + { + "epoch": 1.79, + "learning_rate": 1.8101003920472337e-05, + "loss": 0.0814, + "step": 13646 + }, + { + "epoch": 1.79, + "learning_rate": 1.8100692984169227e-05, + "loss": 0.0819, + "step": 13647 + }, + { + "epoch": 1.79, + "learning_rate": 1.8100382025083344e-05, + "loss": 0.0803, + "step": 13648 + }, + { + "epoch": 1.79, + "learning_rate": 1.810007104321557e-05, + "loss": 0.0829, + "step": 13649 + }, + { + "epoch": 1.79, + "learning_rate": 1.8099760038566782e-05, + "loss": 0.0797, + "step": 13650 + }, + { + "epoch": 1.79, + "learning_rate": 1.8099449011137845e-05, + "loss": 0.097, + "step": 13651 + }, + { + "epoch": 1.79, + "learning_rate": 1.809913796092964e-05, + "loss": 0.0812, + "step": 13652 + }, + { + "epoch": 1.79, + "learning_rate": 1.809882688794304e-05, + "loss": 0.0686, + "step": 13653 + }, + { + "epoch": 1.79, + "learning_rate": 1.809851579217892e-05, + "loss": 0.0679, + "step": 13654 + }, + { + "epoch": 1.79, + "learning_rate": 1.809820467363816e-05, + "loss": 0.0744, + "step": 13655 + }, + { + "epoch": 1.79, + "learning_rate": 1.8097893532321626e-05, + "loss": 0.0838, + "step": 13656 + }, + { + "epoch": 1.79, + "learning_rate": 1.80975823682302e-05, + "loss": 0.0814, + "step": 13657 + }, + { + "epoch": 1.79, + "learning_rate": 1.8097271181364754e-05, + "loss": 0.0675, + "step": 13658 + }, + { + "epoch": 1.79, + "learning_rate": 1.8096959971726162e-05, + "loss": 0.0729, + "step": 13659 + }, + { + "epoch": 1.79, + "learning_rate": 1.8096648739315307e-05, + "loss": 0.0763, + "step": 13660 + }, + { + "epoch": 1.79, + "learning_rate": 1.8096337484133054e-05, + "loss": 0.0758, + "step": 13661 + }, + { + "epoch": 1.79, + "learning_rate": 1.8096026206180283e-05, + "loss": 0.0821, + "step": 13662 + }, + { + "epoch": 1.79, + "learning_rate": 1.809571490545787e-05, + "loss": 0.0824, + "step": 13663 + }, + { + "epoch": 1.79, + "learning_rate": 1.809540358196669e-05, + "loss": 0.0709, + "step": 13664 + }, + { + "epoch": 1.79, + "learning_rate": 1.809509223570762e-05, + "loss": 0.08, + "step": 13665 + }, + { + "epoch": 1.79, + "learning_rate": 1.8094780866681534e-05, + "loss": 0.0823, + "step": 13666 + }, + { + "epoch": 1.79, + "learning_rate": 1.8094469474889306e-05, + "loss": 0.0734, + "step": 13667 + }, + { + "epoch": 1.79, + "learning_rate": 1.8094158060331818e-05, + "loss": 0.0844, + "step": 13668 + }, + { + "epoch": 1.79, + "learning_rate": 1.8093846623009936e-05, + "loss": 0.0709, + "step": 13669 + }, + { + "epoch": 1.79, + "learning_rate": 1.8093535162924546e-05, + "loss": 0.0726, + "step": 13670 + }, + { + "epoch": 1.79, + "learning_rate": 1.8093223680076514e-05, + "loss": 0.0733, + "step": 13671 + }, + { + "epoch": 1.79, + "learning_rate": 1.8092912174466724e-05, + "loss": 0.0869, + "step": 13672 + }, + { + "epoch": 1.79, + "learning_rate": 1.809260064609605e-05, + "loss": 0.0855, + "step": 13673 + }, + { + "epoch": 1.79, + "learning_rate": 1.809228909496537e-05, + "loss": 0.078, + "step": 13674 + }, + { + "epoch": 1.79, + "learning_rate": 1.8091977521075553e-05, + "loss": 0.0818, + "step": 13675 + }, + { + "epoch": 1.79, + "learning_rate": 1.8091665924427482e-05, + "loss": 0.0727, + "step": 13676 + }, + { + "epoch": 1.79, + "learning_rate": 1.8091354305022035e-05, + "loss": 0.0706, + "step": 13677 + }, + { + "epoch": 1.79, + "learning_rate": 1.809104266286008e-05, + "loss": 0.08, + "step": 13678 + }, + { + "epoch": 1.79, + "learning_rate": 1.80907309979425e-05, + "loss": 0.076, + "step": 13679 + }, + { + "epoch": 1.79, + "learning_rate": 1.8090419310270172e-05, + "loss": 0.0706, + "step": 13680 + }, + { + "epoch": 1.79, + "learning_rate": 1.8090107599843967e-05, + "loss": 0.068, + "step": 13681 + }, + { + "epoch": 1.79, + "learning_rate": 1.808979586666477e-05, + "loss": 0.0846, + "step": 13682 + }, + { + "epoch": 1.79, + "learning_rate": 1.8089484110733447e-05, + "loss": 0.0716, + "step": 13683 + }, + { + "epoch": 1.79, + "learning_rate": 1.8089172332050884e-05, + "loss": 0.0756, + "step": 13684 + }, + { + "epoch": 1.79, + "learning_rate": 1.8088860530617955e-05, + "loss": 0.0692, + "step": 13685 + }, + { + "epoch": 1.79, + "learning_rate": 1.8088548706435537e-05, + "loss": 0.0793, + "step": 13686 + }, + { + "epoch": 1.79, + "learning_rate": 1.8088236859504504e-05, + "loss": 0.0809, + "step": 13687 + }, + { + "epoch": 1.79, + "learning_rate": 1.8087924989825737e-05, + "loss": 0.0774, + "step": 13688 + }, + { + "epoch": 1.79, + "learning_rate": 1.808761309740011e-05, + "loss": 0.0793, + "step": 13689 + }, + { + "epoch": 1.79, + "learning_rate": 1.8087301182228505e-05, + "loss": 0.0647, + "step": 13690 + }, + { + "epoch": 1.79, + "learning_rate": 1.808698924431179e-05, + "loss": 0.0787, + "step": 13691 + }, + { + "epoch": 1.79, + "learning_rate": 1.8086677283650854e-05, + "loss": 0.0867, + "step": 13692 + }, + { + "epoch": 1.79, + "learning_rate": 1.808636530024657e-05, + "loss": 0.0722, + "step": 13693 + }, + { + "epoch": 1.79, + "learning_rate": 1.808605329409981e-05, + "loss": 0.0696, + "step": 13694 + }, + { + "epoch": 1.79, + "learning_rate": 1.8085741265211458e-05, + "loss": 0.079, + "step": 13695 + }, + { + "epoch": 1.79, + "learning_rate": 1.8085429213582387e-05, + "loss": 0.0783, + "step": 13696 + }, + { + "epoch": 1.79, + "learning_rate": 1.808511713921348e-05, + "loss": 0.0725, + "step": 13697 + }, + { + "epoch": 1.79, + "learning_rate": 1.8084805042105612e-05, + "loss": 0.0696, + "step": 13698 + }, + { + "epoch": 1.79, + "learning_rate": 1.808449292225966e-05, + "loss": 0.0715, + "step": 13699 + }, + { + "epoch": 1.79, + "learning_rate": 1.80841807796765e-05, + "loss": 0.0684, + "step": 13700 + }, + { + "epoch": 1.79, + "learning_rate": 1.8083868614357012e-05, + "loss": 0.0812, + "step": 13701 + }, + { + "epoch": 1.79, + "learning_rate": 1.8083556426302077e-05, + "loss": 0.0694, + "step": 13702 + }, + { + "epoch": 1.79, + "learning_rate": 1.8083244215512565e-05, + "loss": 0.0741, + "step": 13703 + }, + { + "epoch": 1.8, + "learning_rate": 1.8082931981989366e-05, + "loss": 0.0718, + "step": 13704 + }, + { + "epoch": 1.8, + "learning_rate": 1.8082619725733344e-05, + "loss": 0.0782, + "step": 13705 + }, + { + "epoch": 1.8, + "learning_rate": 1.8082307446745387e-05, + "loss": 0.076, + "step": 13706 + }, + { + "epoch": 1.8, + "learning_rate": 1.8081995145026373e-05, + "loss": 0.0774, + "step": 13707 + }, + { + "epoch": 1.8, + "learning_rate": 1.8081682820577175e-05, + "loss": 0.082, + "step": 13708 + }, + { + "epoch": 1.8, + "learning_rate": 1.8081370473398675e-05, + "loss": 0.0655, + "step": 13709 + }, + { + "epoch": 1.8, + "learning_rate": 1.8081058103491753e-05, + "loss": 0.0755, + "step": 13710 + }, + { + "epoch": 1.8, + "learning_rate": 1.8080745710857288e-05, + "loss": 0.0744, + "step": 13711 + }, + { + "epoch": 1.8, + "learning_rate": 1.808043329549615e-05, + "loss": 0.0796, + "step": 13712 + }, + { + "epoch": 1.8, + "learning_rate": 1.8080120857409224e-05, + "loss": 0.0841, + "step": 13713 + }, + { + "epoch": 1.8, + "learning_rate": 1.807980839659739e-05, + "loss": 0.0695, + "step": 13714 + }, + { + "epoch": 1.8, + "learning_rate": 1.807949591306153e-05, + "loss": 0.0803, + "step": 13715 + }, + { + "epoch": 1.8, + "learning_rate": 1.8079183406802513e-05, + "loss": 0.0915, + "step": 13716 + }, + { + "epoch": 1.8, + "learning_rate": 1.807887087782122e-05, + "loss": 0.0718, + "step": 13717 + }, + { + "epoch": 1.8, + "learning_rate": 1.807855832611854e-05, + "loss": 0.0667, + "step": 13718 + }, + { + "epoch": 1.8, + "learning_rate": 1.807824575169534e-05, + "loss": 0.077, + "step": 13719 + }, + { + "epoch": 1.8, + "learning_rate": 1.8077933154552503e-05, + "loss": 0.0837, + "step": 13720 + }, + { + "epoch": 1.8, + "learning_rate": 1.8077620534690913e-05, + "loss": 0.0725, + "step": 13721 + }, + { + "epoch": 1.8, + "learning_rate": 1.8077307892111444e-05, + "loss": 0.0807, + "step": 13722 + }, + { + "epoch": 1.8, + "learning_rate": 1.807699522681498e-05, + "loss": 0.0689, + "step": 13723 + }, + { + "epoch": 1.8, + "learning_rate": 1.8076682538802388e-05, + "loss": 0.0829, + "step": 13724 + }, + { + "epoch": 1.8, + "learning_rate": 1.8076369828074565e-05, + "loss": 0.0791, + "step": 13725 + }, + { + "epoch": 1.8, + "learning_rate": 1.8076057094632378e-05, + "loss": 0.0789, + "step": 13726 + }, + { + "epoch": 1.8, + "learning_rate": 1.807574433847671e-05, + "loss": 0.0773, + "step": 13727 + }, + { + "epoch": 1.8, + "learning_rate": 1.8075431559608442e-05, + "loss": 0.0843, + "step": 13728 + }, + { + "epoch": 1.8, + "learning_rate": 1.8075118758028455e-05, + "loss": 0.0687, + "step": 13729 + }, + { + "epoch": 1.8, + "learning_rate": 1.8074805933737624e-05, + "loss": 0.0714, + "step": 13730 + }, + { + "epoch": 1.8, + "learning_rate": 1.8074493086736833e-05, + "loss": 0.0685, + "step": 13731 + }, + { + "epoch": 1.8, + "learning_rate": 1.807418021702696e-05, + "loss": 0.0808, + "step": 13732 + }, + { + "epoch": 1.8, + "learning_rate": 1.8073867324608883e-05, + "loss": 0.07, + "step": 13733 + }, + { + "epoch": 1.8, + "learning_rate": 1.8073554409483486e-05, + "loss": 0.0755, + "step": 13734 + }, + { + "epoch": 1.8, + "learning_rate": 1.8073241471651646e-05, + "loss": 0.0841, + "step": 13735 + }, + { + "epoch": 1.8, + "learning_rate": 1.8072928511114244e-05, + "loss": 0.0768, + "step": 13736 + }, + { + "epoch": 1.8, + "learning_rate": 1.8072615527872164e-05, + "loss": 0.0724, + "step": 13737 + }, + { + "epoch": 1.8, + "learning_rate": 1.8072302521926278e-05, + "loss": 0.086, + "step": 13738 + }, + { + "epoch": 1.8, + "learning_rate": 1.8071989493277476e-05, + "loss": 0.0736, + "step": 13739 + }, + { + "epoch": 1.8, + "learning_rate": 1.8071676441926633e-05, + "loss": 0.0838, + "step": 13740 + }, + { + "epoch": 1.8, + "learning_rate": 1.8071363367874625e-05, + "loss": 0.0678, + "step": 13741 + }, + { + "epoch": 1.8, + "learning_rate": 1.8071050271122338e-05, + "loss": 0.0723, + "step": 13742 + }, + { + "epoch": 1.8, + "learning_rate": 1.8070737151670656e-05, + "loss": 0.0864, + "step": 13743 + }, + { + "epoch": 1.8, + "learning_rate": 1.8070424009520452e-05, + "loss": 0.0741, + "step": 13744 + }, + { + "epoch": 1.8, + "learning_rate": 1.807011084467261e-05, + "loss": 0.0726, + "step": 13745 + }, + { + "epoch": 1.8, + "learning_rate": 1.8069797657128014e-05, + "loss": 0.0825, + "step": 13746 + }, + { + "epoch": 1.8, + "learning_rate": 1.8069484446887542e-05, + "loss": 0.0702, + "step": 13747 + }, + { + "epoch": 1.8, + "learning_rate": 1.8069171213952073e-05, + "loss": 0.0749, + "step": 13748 + }, + { + "epoch": 1.8, + "learning_rate": 1.806885795832249e-05, + "loss": 0.0716, + "step": 13749 + }, + { + "epoch": 1.8, + "learning_rate": 1.8068544679999675e-05, + "loss": 0.0705, + "step": 13750 + }, + { + "epoch": 1.8, + "learning_rate": 1.80682313789845e-05, + "loss": 0.0732, + "step": 13751 + }, + { + "epoch": 1.8, + "learning_rate": 1.8067918055277864e-05, + "loss": 0.0686, + "step": 13752 + }, + { + "epoch": 1.8, + "learning_rate": 1.8067604708880634e-05, + "loss": 0.071, + "step": 13753 + }, + { + "epoch": 1.8, + "learning_rate": 1.8067291339793695e-05, + "loss": 0.0764, + "step": 13754 + }, + { + "epoch": 1.8, + "learning_rate": 1.8066977948017932e-05, + "loss": 0.0731, + "step": 13755 + }, + { + "epoch": 1.8, + "learning_rate": 1.8066664533554218e-05, + "loss": 0.0778, + "step": 13756 + }, + { + "epoch": 1.8, + "learning_rate": 1.8066351096403442e-05, + "loss": 0.0826, + "step": 13757 + }, + { + "epoch": 1.8, + "learning_rate": 1.8066037636566484e-05, + "loss": 0.0693, + "step": 13758 + }, + { + "epoch": 1.8, + "learning_rate": 1.8065724154044225e-05, + "loss": 0.0666, + "step": 13759 + }, + { + "epoch": 1.8, + "learning_rate": 1.8065410648837543e-05, + "loss": 0.0699, + "step": 13760 + }, + { + "epoch": 1.8, + "learning_rate": 1.806509712094733e-05, + "loss": 0.078, + "step": 13761 + }, + { + "epoch": 1.8, + "learning_rate": 1.8064783570374454e-05, + "loss": 0.0694, + "step": 13762 + }, + { + "epoch": 1.8, + "learning_rate": 1.8064469997119804e-05, + "loss": 0.0687, + "step": 13763 + }, + { + "epoch": 1.8, + "learning_rate": 1.8064156401184266e-05, + "loss": 0.0695, + "step": 13764 + }, + { + "epoch": 1.8, + "learning_rate": 1.8063842782568715e-05, + "loss": 0.0611, + "step": 13765 + }, + { + "epoch": 1.8, + "learning_rate": 1.8063529141274033e-05, + "loss": 0.0635, + "step": 13766 + }, + { + "epoch": 1.8, + "learning_rate": 1.806321547730111e-05, + "loss": 0.0674, + "step": 13767 + }, + { + "epoch": 1.8, + "learning_rate": 1.8062901790650818e-05, + "loss": 0.0827, + "step": 13768 + }, + { + "epoch": 1.8, + "learning_rate": 1.8062588081324046e-05, + "loss": 0.0717, + "step": 13769 + }, + { + "epoch": 1.8, + "learning_rate": 1.8062274349321677e-05, + "loss": 0.0765, + "step": 13770 + }, + { + "epoch": 1.8, + "learning_rate": 1.8061960594644588e-05, + "loss": 0.0706, + "step": 13771 + }, + { + "epoch": 1.8, + "learning_rate": 1.8061646817293664e-05, + "loss": 0.0679, + "step": 13772 + }, + { + "epoch": 1.8, + "learning_rate": 1.8061333017269785e-05, + "loss": 0.0779, + "step": 13773 + }, + { + "epoch": 1.8, + "learning_rate": 1.806101919457384e-05, + "loss": 0.0823, + "step": 13774 + }, + { + "epoch": 1.8, + "learning_rate": 1.806070534920671e-05, + "loss": 0.0837, + "step": 13775 + }, + { + "epoch": 1.8, + "learning_rate": 1.806039148116927e-05, + "loss": 0.0724, + "step": 13776 + }, + { + "epoch": 1.8, + "learning_rate": 1.8060077590462413e-05, + "loss": 0.0734, + "step": 13777 + }, + { + "epoch": 1.8, + "learning_rate": 1.8059763677087013e-05, + "loss": 0.0796, + "step": 13778 + }, + { + "epoch": 1.8, + "learning_rate": 1.8059449741043958e-05, + "loss": 0.0751, + "step": 13779 + }, + { + "epoch": 1.81, + "learning_rate": 1.805913578233413e-05, + "loss": 0.0723, + "step": 13780 + }, + { + "epoch": 1.81, + "learning_rate": 1.8058821800958415e-05, + "loss": 0.0715, + "step": 13781 + }, + { + "epoch": 1.81, + "learning_rate": 1.8058507796917686e-05, + "loss": 0.0705, + "step": 13782 + }, + { + "epoch": 1.81, + "learning_rate": 1.805819377021284e-05, + "loss": 0.0716, + "step": 13783 + }, + { + "epoch": 1.81, + "learning_rate": 1.8057879720844748e-05, + "loss": 0.0768, + "step": 13784 + }, + { + "epoch": 1.81, + "learning_rate": 1.8057565648814304e-05, + "loss": 0.0699, + "step": 13785 + }, + { + "epoch": 1.81, + "learning_rate": 1.805725155412238e-05, + "loss": 0.0701, + "step": 13786 + }, + { + "epoch": 1.81, + "learning_rate": 1.8056937436769867e-05, + "loss": 0.0708, + "step": 13787 + }, + { + "epoch": 1.81, + "learning_rate": 1.805662329675765e-05, + "loss": 0.0892, + "step": 13788 + }, + { + "epoch": 1.81, + "learning_rate": 1.8056309134086603e-05, + "loss": 0.0769, + "step": 13789 + }, + { + "epoch": 1.81, + "learning_rate": 1.8055994948757618e-05, + "loss": 0.0746, + "step": 13790 + }, + { + "epoch": 1.81, + "learning_rate": 1.8055680740771576e-05, + "loss": 0.0791, + "step": 13791 + }, + { + "epoch": 1.81, + "learning_rate": 1.805536651012936e-05, + "loss": 0.0824, + "step": 13792 + }, + { + "epoch": 1.81, + "learning_rate": 1.8055052256831856e-05, + "loss": 0.0718, + "step": 13793 + }, + { + "epoch": 1.81, + "learning_rate": 1.805473798087995e-05, + "loss": 0.0749, + "step": 13794 + }, + { + "epoch": 1.81, + "learning_rate": 1.805442368227452e-05, + "loss": 0.0605, + "step": 13795 + }, + { + "epoch": 1.81, + "learning_rate": 1.805410936101645e-05, + "loss": 0.0773, + "step": 13796 + }, + { + "epoch": 1.81, + "learning_rate": 1.805379501710663e-05, + "loss": 0.0712, + "step": 13797 + }, + { + "epoch": 1.81, + "learning_rate": 1.8053480650545936e-05, + "loss": 0.0713, + "step": 13798 + }, + { + "epoch": 1.81, + "learning_rate": 1.805316626133526e-05, + "loss": 0.07, + "step": 13799 + }, + { + "epoch": 1.81, + "learning_rate": 1.8052851849475482e-05, + "loss": 0.068, + "step": 13800 + }, + { + "epoch": 1.81, + "learning_rate": 1.8052537414967487e-05, + "loss": 0.0739, + "step": 13801 + }, + { + "epoch": 1.81, + "learning_rate": 1.8052222957812158e-05, + "loss": 0.0882, + "step": 13802 + }, + { + "epoch": 1.81, + "learning_rate": 1.805190847801038e-05, + "loss": 0.0822, + "step": 13803 + }, + { + "epoch": 1.81, + "learning_rate": 1.805159397556304e-05, + "loss": 0.0765, + "step": 13804 + }, + { + "epoch": 1.81, + "learning_rate": 1.805127945047102e-05, + "loss": 0.0694, + "step": 13805 + }, + { + "epoch": 1.81, + "learning_rate": 1.8050964902735205e-05, + "loss": 0.0825, + "step": 13806 + }, + { + "epoch": 1.81, + "learning_rate": 1.805065033235648e-05, + "loss": 0.0796, + "step": 13807 + }, + { + "epoch": 1.81, + "learning_rate": 1.805033573933573e-05, + "loss": 0.0658, + "step": 13808 + }, + { + "epoch": 1.81, + "learning_rate": 1.8050021123673843e-05, + "loss": 0.0731, + "step": 13809 + }, + { + "epoch": 1.81, + "learning_rate": 1.8049706485371696e-05, + "loss": 0.0665, + "step": 13810 + }, + { + "epoch": 1.81, + "learning_rate": 1.8049391824430177e-05, + "loss": 0.0685, + "step": 13811 + }, + { + "epoch": 1.81, + "learning_rate": 1.8049077140850175e-05, + "loss": 0.0687, + "step": 13812 + }, + { + "epoch": 1.81, + "learning_rate": 1.8048762434632573e-05, + "loss": 0.083, + "step": 13813 + }, + { + "epoch": 1.81, + "learning_rate": 1.8048447705778253e-05, + "loss": 0.0687, + "step": 13814 + }, + { + "epoch": 1.81, + "learning_rate": 1.8048132954288103e-05, + "loss": 0.0717, + "step": 13815 + }, + { + "epoch": 1.81, + "learning_rate": 1.804781818016301e-05, + "loss": 0.0732, + "step": 13816 + }, + { + "epoch": 1.81, + "learning_rate": 1.8047503383403854e-05, + "loss": 0.078, + "step": 13817 + }, + { + "epoch": 1.81, + "learning_rate": 1.8047188564011524e-05, + "loss": 0.0759, + "step": 13818 + }, + { + "epoch": 1.81, + "learning_rate": 1.8046873721986908e-05, + "loss": 0.0719, + "step": 13819 + }, + { + "epoch": 1.81, + "learning_rate": 1.8046558857330882e-05, + "loss": 0.0707, + "step": 13820 + }, + { + "epoch": 1.81, + "learning_rate": 1.8046243970044338e-05, + "loss": 0.0807, + "step": 13821 + }, + { + "epoch": 1.81, + "learning_rate": 1.8045929060128167e-05, + "loss": 0.0776, + "step": 13822 + }, + { + "epoch": 1.81, + "learning_rate": 1.8045614127583247e-05, + "loss": 0.0729, + "step": 13823 + }, + { + "epoch": 1.81, + "learning_rate": 1.8045299172410465e-05, + "loss": 0.0841, + "step": 13824 + }, + { + "epoch": 1.81, + "learning_rate": 1.8044984194610706e-05, + "loss": 0.0746, + "step": 13825 + }, + { + "epoch": 1.81, + "learning_rate": 1.804466919418486e-05, + "loss": 0.0796, + "step": 13826 + }, + { + "epoch": 1.81, + "learning_rate": 1.8044354171133805e-05, + "loss": 0.0736, + "step": 13827 + }, + { + "epoch": 1.81, + "learning_rate": 1.804403912545844e-05, + "loss": 0.0725, + "step": 13828 + }, + { + "epoch": 1.81, + "learning_rate": 1.804372405715964e-05, + "loss": 0.0716, + "step": 13829 + }, + { + "epoch": 1.81, + "learning_rate": 1.8043408966238292e-05, + "loss": 0.0806, + "step": 13830 + }, + { + "epoch": 1.81, + "learning_rate": 1.804309385269529e-05, + "loss": 0.0765, + "step": 13831 + }, + { + "epoch": 1.81, + "learning_rate": 1.804277871653151e-05, + "loss": 0.0883, + "step": 13832 + }, + { + "epoch": 1.81, + "learning_rate": 1.8042463557747842e-05, + "loss": 0.073, + "step": 13833 + }, + { + "epoch": 1.81, + "learning_rate": 1.8042148376345178e-05, + "loss": 0.0818, + "step": 13834 + }, + { + "epoch": 1.81, + "learning_rate": 1.8041833172324395e-05, + "loss": 0.0738, + "step": 13835 + }, + { + "epoch": 1.81, + "learning_rate": 1.804151794568639e-05, + "loss": 0.0743, + "step": 13836 + }, + { + "epoch": 1.81, + "learning_rate": 1.8041202696432042e-05, + "loss": 0.0739, + "step": 13837 + }, + { + "epoch": 1.81, + "learning_rate": 1.804088742456224e-05, + "loss": 0.0661, + "step": 13838 + }, + { + "epoch": 1.81, + "learning_rate": 1.804057213007787e-05, + "loss": 0.0805, + "step": 13839 + }, + { + "epoch": 1.81, + "learning_rate": 1.8040256812979818e-05, + "loss": 0.0793, + "step": 13840 + }, + { + "epoch": 1.81, + "learning_rate": 1.8039941473268972e-05, + "loss": 0.0725, + "step": 13841 + }, + { + "epoch": 1.81, + "learning_rate": 1.803962611094622e-05, + "loss": 0.0775, + "step": 13842 + }, + { + "epoch": 1.81, + "learning_rate": 1.8039310726012447e-05, + "loss": 0.0673, + "step": 13843 + }, + { + "epoch": 1.81, + "learning_rate": 1.803899531846854e-05, + "loss": 0.0779, + "step": 13844 + }, + { + "epoch": 1.81, + "learning_rate": 1.8038679888315388e-05, + "loss": 0.0776, + "step": 13845 + }, + { + "epoch": 1.81, + "learning_rate": 1.8038364435553877e-05, + "loss": 0.0836, + "step": 13846 + }, + { + "epoch": 1.81, + "learning_rate": 1.8038048960184894e-05, + "loss": 0.0695, + "step": 13847 + }, + { + "epoch": 1.81, + "learning_rate": 1.8037733462209324e-05, + "loss": 0.0769, + "step": 13848 + }, + { + "epoch": 1.81, + "learning_rate": 1.803741794162806e-05, + "loss": 0.0703, + "step": 13849 + }, + { + "epoch": 1.81, + "learning_rate": 1.803710239844198e-05, + "loss": 0.058, + "step": 13850 + }, + { + "epoch": 1.81, + "learning_rate": 1.8036786832651986e-05, + "loss": 0.0717, + "step": 13851 + }, + { + "epoch": 1.81, + "learning_rate": 1.8036471244258952e-05, + "loss": 0.0729, + "step": 13852 + }, + { + "epoch": 1.81, + "learning_rate": 1.8036155633263768e-05, + "loss": 0.0755, + "step": 13853 + }, + { + "epoch": 1.81, + "learning_rate": 1.803583999966733e-05, + "loss": 0.0692, + "step": 13854 + }, + { + "epoch": 1.81, + "learning_rate": 1.8035524343470518e-05, + "loss": 0.0714, + "step": 13855 + }, + { + "epoch": 1.82, + "learning_rate": 1.8035208664674223e-05, + "loss": 0.0829, + "step": 13856 + }, + { + "epoch": 1.82, + "learning_rate": 1.803489296327933e-05, + "loss": 0.0718, + "step": 13857 + }, + { + "epoch": 1.82, + "learning_rate": 1.8034577239286725e-05, + "loss": 0.0735, + "step": 13858 + }, + { + "epoch": 1.82, + "learning_rate": 1.8034261492697305e-05, + "loss": 0.0702, + "step": 13859 + }, + { + "epoch": 1.82, + "learning_rate": 1.803394572351195e-05, + "loss": 0.0665, + "step": 13860 + }, + { + "epoch": 1.82, + "learning_rate": 1.8033629931731552e-05, + "loss": 0.0687, + "step": 13861 + }, + { + "epoch": 1.82, + "learning_rate": 1.8033314117356995e-05, + "loss": 0.0712, + "step": 13862 + }, + { + "epoch": 1.82, + "learning_rate": 1.8032998280389172e-05, + "loss": 0.0663, + "step": 13863 + }, + { + "epoch": 1.82, + "learning_rate": 1.803268242082897e-05, + "loss": 0.0607, + "step": 13864 + }, + { + "epoch": 1.82, + "learning_rate": 1.803236653867727e-05, + "loss": 0.0696, + "step": 13865 + }, + { + "epoch": 1.82, + "learning_rate": 1.8032050633934975e-05, + "loss": 0.0755, + "step": 13866 + }, + { + "epoch": 1.82, + "learning_rate": 1.803173470660296e-05, + "loss": 0.0755, + "step": 13867 + }, + { + "epoch": 1.82, + "learning_rate": 1.8031418756682125e-05, + "loss": 0.0735, + "step": 13868 + }, + { + "epoch": 1.82, + "learning_rate": 1.8031102784173344e-05, + "loss": 0.0685, + "step": 13869 + }, + { + "epoch": 1.82, + "learning_rate": 1.8030786789077522e-05, + "loss": 0.065, + "step": 13870 + }, + { + "epoch": 1.82, + "learning_rate": 1.8030470771395535e-05, + "loss": 0.0692, + "step": 13871 + }, + { + "epoch": 1.82, + "learning_rate": 1.803015473112828e-05, + "loss": 0.0596, + "step": 13872 + }, + { + "epoch": 1.82, + "learning_rate": 1.8029838668276637e-05, + "loss": 0.0712, + "step": 13873 + }, + { + "epoch": 1.82, + "learning_rate": 1.8029522582841505e-05, + "loss": 0.0842, + "step": 13874 + }, + { + "epoch": 1.82, + "learning_rate": 1.8029206474823766e-05, + "loss": 0.0737, + "step": 13875 + }, + { + "epoch": 1.82, + "learning_rate": 1.8028890344224312e-05, + "loss": 0.0726, + "step": 13876 + }, + { + "epoch": 1.82, + "learning_rate": 1.8028574191044034e-05, + "loss": 0.0736, + "step": 13877 + }, + { + "epoch": 1.82, + "learning_rate": 1.8028258015283815e-05, + "loss": 0.0731, + "step": 13878 + }, + { + "epoch": 1.82, + "learning_rate": 1.8027941816944547e-05, + "loss": 0.0595, + "step": 13879 + }, + { + "epoch": 1.82, + "learning_rate": 1.8027625596027122e-05, + "loss": 0.0686, + "step": 13880 + }, + { + "epoch": 1.82, + "learning_rate": 1.802730935253243e-05, + "loss": 0.0839, + "step": 13881 + }, + { + "epoch": 1.82, + "learning_rate": 1.8026993086461355e-05, + "loss": 0.0727, + "step": 13882 + }, + { + "epoch": 1.82, + "learning_rate": 1.802667679781479e-05, + "loss": 0.0767, + "step": 13883 + }, + { + "epoch": 1.82, + "learning_rate": 1.8026360486593626e-05, + "loss": 0.0739, + "step": 13884 + }, + { + "epoch": 1.82, + "learning_rate": 1.8026044152798747e-05, + "loss": 0.0744, + "step": 13885 + }, + { + "epoch": 1.82, + "learning_rate": 1.802572779643105e-05, + "loss": 0.0706, + "step": 13886 + }, + { + "epoch": 1.82, + "learning_rate": 1.802541141749142e-05, + "loss": 0.0719, + "step": 13887 + }, + { + "epoch": 1.82, + "learning_rate": 1.8025095015980745e-05, + "loss": 0.0781, + "step": 13888 + }, + { + "epoch": 1.82, + "learning_rate": 1.802477859189992e-05, + "loss": 0.0797, + "step": 13889 + }, + { + "epoch": 1.82, + "learning_rate": 1.802446214524983e-05, + "loss": 0.0776, + "step": 13890 + }, + { + "epoch": 1.82, + "learning_rate": 1.802414567603137e-05, + "loss": 0.074, + "step": 13891 + }, + { + "epoch": 1.82, + "learning_rate": 1.802382918424543e-05, + "loss": 0.0772, + "step": 13892 + }, + { + "epoch": 1.82, + "learning_rate": 1.8023512669892893e-05, + "loss": 0.0906, + "step": 13893 + }, + { + "epoch": 1.82, + "learning_rate": 1.8023196132974655e-05, + "loss": 0.0773, + "step": 13894 + }, + { + "epoch": 1.82, + "learning_rate": 1.8022879573491605e-05, + "loss": 0.0728, + "step": 13895 + }, + { + "epoch": 1.82, + "learning_rate": 1.8022562991444633e-05, + "loss": 0.0783, + "step": 13896 + }, + { + "epoch": 1.82, + "learning_rate": 1.8022246386834635e-05, + "loss": 0.0742, + "step": 13897 + }, + { + "epoch": 1.82, + "learning_rate": 1.8021929759662492e-05, + "loss": 0.0769, + "step": 13898 + }, + { + "epoch": 1.82, + "learning_rate": 1.8021613109929097e-05, + "loss": 0.0793, + "step": 13899 + }, + { + "epoch": 1.82, + "learning_rate": 1.8021296437635346e-05, + "loss": 0.0689, + "step": 13900 + }, + { + "epoch": 1.82, + "learning_rate": 1.8020979742782123e-05, + "loss": 0.0768, + "step": 13901 + }, + { + "epoch": 1.82, + "learning_rate": 1.8020663025370325e-05, + "loss": 0.0803, + "step": 13902 + }, + { + "epoch": 1.82, + "learning_rate": 1.8020346285400836e-05, + "loss": 0.0853, + "step": 13903 + }, + { + "epoch": 1.82, + "learning_rate": 1.8020029522874553e-05, + "loss": 0.0756, + "step": 13904 + }, + { + "epoch": 1.82, + "learning_rate": 1.8019712737792357e-05, + "loss": 0.0766, + "step": 13905 + }, + { + "epoch": 1.82, + "learning_rate": 1.8019395930155155e-05, + "loss": 0.0703, + "step": 13906 + }, + { + "epoch": 1.82, + "learning_rate": 1.8019079099963824e-05, + "loss": 0.0727, + "step": 13907 + }, + { + "epoch": 1.82, + "learning_rate": 1.801876224721926e-05, + "loss": 0.0671, + "step": 13908 + }, + { + "epoch": 1.82, + "learning_rate": 1.801844537192236e-05, + "loss": 0.0824, + "step": 13909 + }, + { + "epoch": 1.82, + "learning_rate": 1.8018128474074003e-05, + "loss": 0.078, + "step": 13910 + }, + { + "epoch": 1.82, + "learning_rate": 1.801781155367509e-05, + "loss": 0.0692, + "step": 13911 + }, + { + "epoch": 1.82, + "learning_rate": 1.8017494610726508e-05, + "loss": 0.071, + "step": 13912 + }, + { + "epoch": 1.82, + "learning_rate": 1.8017177645229148e-05, + "loss": 0.067, + "step": 13913 + }, + { + "epoch": 1.82, + "learning_rate": 1.8016860657183906e-05, + "loss": 0.0734, + "step": 13914 + }, + { + "epoch": 1.82, + "learning_rate": 1.8016543646591667e-05, + "loss": 0.0788, + "step": 13915 + }, + { + "epoch": 1.82, + "learning_rate": 1.8016226613453326e-05, + "loss": 0.0664, + "step": 13916 + }, + { + "epoch": 1.82, + "learning_rate": 1.801590955776978e-05, + "loss": 0.0835, + "step": 13917 + }, + { + "epoch": 1.82, + "learning_rate": 1.8015592479541908e-05, + "loss": 0.0769, + "step": 13918 + }, + { + "epoch": 1.82, + "learning_rate": 1.801527537877061e-05, + "loss": 0.0748, + "step": 13919 + }, + { + "epoch": 1.82, + "learning_rate": 1.801495825545678e-05, + "loss": 0.0772, + "step": 13920 + }, + { + "epoch": 1.82, + "learning_rate": 1.8014641109601307e-05, + "loss": 0.0685, + "step": 13921 + }, + { + "epoch": 1.82, + "learning_rate": 1.8014323941205082e-05, + "loss": 0.0744, + "step": 13922 + }, + { + "epoch": 1.82, + "learning_rate": 1.8014006750268995e-05, + "loss": 0.0754, + "step": 13923 + }, + { + "epoch": 1.82, + "learning_rate": 1.8013689536793944e-05, + "loss": 0.073, + "step": 13924 + }, + { + "epoch": 1.82, + "learning_rate": 1.8013372300780816e-05, + "loss": 0.07, + "step": 13925 + }, + { + "epoch": 1.82, + "learning_rate": 1.8013055042230504e-05, + "loss": 0.0673, + "step": 13926 + }, + { + "epoch": 1.82, + "learning_rate": 1.8012737761143904e-05, + "loss": 0.0739, + "step": 13927 + }, + { + "epoch": 1.82, + "learning_rate": 1.8012420457521908e-05, + "loss": 0.0693, + "step": 13928 + }, + { + "epoch": 1.82, + "learning_rate": 1.8012103131365403e-05, + "loss": 0.0748, + "step": 13929 + }, + { + "epoch": 1.82, + "learning_rate": 1.8011785782675284e-05, + "loss": 0.0652, + "step": 13930 + }, + { + "epoch": 1.82, + "learning_rate": 1.8011468411452445e-05, + "loss": 0.0696, + "step": 13931 + }, + { + "epoch": 1.82, + "learning_rate": 1.8011151017697778e-05, + "loss": 0.0785, + "step": 13932 + }, + { + "epoch": 1.83, + "learning_rate": 1.8010833601412174e-05, + "loss": 0.079, + "step": 13933 + }, + { + "epoch": 1.83, + "learning_rate": 1.8010516162596526e-05, + "loss": 0.0884, + "step": 13934 + }, + { + "epoch": 1.83, + "learning_rate": 1.801019870125173e-05, + "loss": 0.0586, + "step": 13935 + }, + { + "epoch": 1.83, + "learning_rate": 1.8009881217378676e-05, + "loss": 0.0744, + "step": 13936 + }, + { + "epoch": 1.83, + "learning_rate": 1.8009563710978263e-05, + "loss": 0.0839, + "step": 13937 + }, + { + "epoch": 1.83, + "learning_rate": 1.800924618205137e-05, + "loss": 0.076, + "step": 13938 + }, + { + "epoch": 1.83, + "learning_rate": 1.8008928630598905e-05, + "loss": 0.066, + "step": 13939 + }, + { + "epoch": 1.83, + "learning_rate": 1.800861105662175e-05, + "loss": 0.0619, + "step": 13940 + }, + { + "epoch": 1.83, + "learning_rate": 1.800829346012081e-05, + "loss": 0.0653, + "step": 13941 + }, + { + "epoch": 1.83, + "learning_rate": 1.8007975841096963e-05, + "loss": 0.066, + "step": 13942 + }, + { + "epoch": 1.83, + "learning_rate": 1.8007658199551112e-05, + "loss": 0.0942, + "step": 13943 + }, + { + "epoch": 1.83, + "learning_rate": 1.800734053548415e-05, + "loss": 0.0775, + "step": 13944 + }, + { + "epoch": 1.83, + "learning_rate": 1.800702284889697e-05, + "loss": 0.0829, + "step": 13945 + }, + { + "epoch": 1.83, + "learning_rate": 1.8006705139790464e-05, + "loss": 0.0638, + "step": 13946 + }, + { + "epoch": 1.83, + "learning_rate": 1.8006387408165528e-05, + "loss": 0.0769, + "step": 13947 + }, + { + "epoch": 1.83, + "learning_rate": 1.800606965402305e-05, + "loss": 0.0768, + "step": 13948 + }, + { + "epoch": 1.83, + "learning_rate": 1.8005751877363932e-05, + "loss": 0.0817, + "step": 13949 + }, + { + "epoch": 1.83, + "learning_rate": 1.800543407818906e-05, + "loss": 0.085, + "step": 13950 + }, + { + "epoch": 1.83, + "learning_rate": 1.8005116256499332e-05, + "loss": 0.0695, + "step": 13951 + }, + { + "epoch": 1.83, + "learning_rate": 1.800479841229564e-05, + "loss": 0.0756, + "step": 13952 + }, + { + "epoch": 1.83, + "learning_rate": 1.800448054557888e-05, + "loss": 0.0758, + "step": 13953 + }, + { + "epoch": 1.83, + "learning_rate": 1.8004162656349943e-05, + "loss": 0.0717, + "step": 13954 + }, + { + "epoch": 1.83, + "learning_rate": 1.8003844744609726e-05, + "loss": 0.0969, + "step": 13955 + }, + { + "epoch": 1.83, + "learning_rate": 1.8003526810359123e-05, + "loss": 0.081, + "step": 13956 + }, + { + "epoch": 1.83, + "learning_rate": 1.8003208853599027e-05, + "loss": 0.0733, + "step": 13957 + }, + { + "epoch": 1.83, + "learning_rate": 1.8002890874330328e-05, + "loss": 0.0735, + "step": 13958 + }, + { + "epoch": 1.83, + "learning_rate": 1.800257287255393e-05, + "loss": 0.0843, + "step": 13959 + }, + { + "epoch": 1.83, + "learning_rate": 1.8002254848270716e-05, + "loss": 0.0705, + "step": 13960 + }, + { + "epoch": 1.83, + "learning_rate": 1.8001936801481595e-05, + "loss": 0.0803, + "step": 13961 + }, + { + "epoch": 1.83, + "learning_rate": 1.8001618732187446e-05, + "loss": 0.0685, + "step": 13962 + }, + { + "epoch": 1.83, + "learning_rate": 1.8001300640389173e-05, + "loss": 0.0774, + "step": 13963 + }, + { + "epoch": 1.83, + "learning_rate": 1.8000982526087668e-05, + "loss": 0.0697, + "step": 13964 + }, + { + "epoch": 1.83, + "learning_rate": 1.8000664389283824e-05, + "loss": 0.0672, + "step": 13965 + }, + { + "epoch": 1.83, + "learning_rate": 1.800034622997854e-05, + "loss": 0.0785, + "step": 13966 + }, + { + "epoch": 1.83, + "learning_rate": 1.8000028048172705e-05, + "loss": 0.077, + "step": 13967 + }, + { + "epoch": 1.83, + "learning_rate": 1.7999709843867217e-05, + "loss": 0.0697, + "step": 13968 + }, + { + "epoch": 1.83, + "learning_rate": 1.7999391617062975e-05, + "loss": 0.07, + "step": 13969 + }, + { + "epoch": 1.83, + "learning_rate": 1.799907336776087e-05, + "loss": 0.086, + "step": 13970 + }, + { + "epoch": 1.83, + "learning_rate": 1.7998755095961794e-05, + "loss": 0.0748, + "step": 13971 + }, + { + "epoch": 1.83, + "learning_rate": 1.7998436801666644e-05, + "loss": 0.0805, + "step": 13972 + }, + { + "epoch": 1.83, + "learning_rate": 1.799811848487632e-05, + "loss": 0.0717, + "step": 13973 + }, + { + "epoch": 1.83, + "learning_rate": 1.799780014559171e-05, + "loss": 0.0824, + "step": 13974 + }, + { + "epoch": 1.83, + "learning_rate": 1.7997481783813718e-05, + "loss": 0.0772, + "step": 13975 + }, + { + "epoch": 1.83, + "learning_rate": 1.7997163399543232e-05, + "loss": 0.0705, + "step": 13976 + }, + { + "epoch": 1.83, + "learning_rate": 1.799684499278115e-05, + "loss": 0.067, + "step": 13977 + }, + { + "epoch": 1.83, + "learning_rate": 1.7996526563528368e-05, + "loss": 0.0724, + "step": 13978 + }, + { + "epoch": 1.83, + "learning_rate": 1.799620811178578e-05, + "loss": 0.0746, + "step": 13979 + }, + { + "epoch": 1.83, + "learning_rate": 1.799588963755428e-05, + "loss": 0.0693, + "step": 13980 + }, + { + "epoch": 1.83, + "learning_rate": 1.799557114083477e-05, + "loss": 0.061, + "step": 13981 + }, + { + "epoch": 1.83, + "learning_rate": 1.799525262162814e-05, + "loss": 0.0723, + "step": 13982 + }, + { + "epoch": 1.83, + "learning_rate": 1.7994934079935288e-05, + "loss": 0.0766, + "step": 13983 + }, + { + "epoch": 1.83, + "learning_rate": 1.799461551575711e-05, + "loss": 0.0723, + "step": 13984 + }, + { + "epoch": 1.83, + "learning_rate": 1.7994296929094503e-05, + "loss": 0.0664, + "step": 13985 + }, + { + "epoch": 1.83, + "learning_rate": 1.799397831994836e-05, + "loss": 0.0799, + "step": 13986 + }, + { + "epoch": 1.83, + "learning_rate": 1.7993659688319578e-05, + "loss": 0.0672, + "step": 13987 + }, + { + "epoch": 1.83, + "learning_rate": 1.7993341034209057e-05, + "loss": 0.082, + "step": 13988 + }, + { + "epoch": 1.83, + "learning_rate": 1.7993022357617685e-05, + "loss": 0.0684, + "step": 13989 + }, + { + "epoch": 1.83, + "learning_rate": 1.7992703658546368e-05, + "loss": 0.0759, + "step": 13990 + }, + { + "epoch": 1.83, + "learning_rate": 1.7992384936995995e-05, + "loss": 0.0648, + "step": 13991 + }, + { + "epoch": 1.83, + "learning_rate": 1.799206619296747e-05, + "loss": 0.0747, + "step": 13992 + }, + { + "epoch": 1.83, + "learning_rate": 1.7991747426461676e-05, + "loss": 0.0718, + "step": 13993 + }, + { + "epoch": 1.83, + "learning_rate": 1.7991428637479522e-05, + "loss": 0.0758, + "step": 13994 + }, + { + "epoch": 1.83, + "learning_rate": 1.79911098260219e-05, + "loss": 0.0728, + "step": 13995 + }, + { + "epoch": 1.83, + "learning_rate": 1.7990790992089708e-05, + "loss": 0.0669, + "step": 13996 + }, + { + "epoch": 1.83, + "learning_rate": 1.7990472135683843e-05, + "loss": 0.0808, + "step": 13997 + }, + { + "epoch": 1.83, + "learning_rate": 1.7990153256805198e-05, + "loss": 0.0688, + "step": 13998 + }, + { + "epoch": 1.83, + "learning_rate": 1.7989834355454674e-05, + "loss": 0.08, + "step": 13999 + }, + { + "epoch": 1.83, + "learning_rate": 1.7989515431633166e-05, + "loss": 0.0716, + "step": 14000 + }, + { + "epoch": 1.83, + "learning_rate": 1.798919648534157e-05, + "loss": 0.0781, + "step": 14001 + }, + { + "epoch": 1.83, + "learning_rate": 1.7988877516580787e-05, + "loss": 0.0848, + "step": 14002 + }, + { + "epoch": 1.83, + "learning_rate": 1.7988558525351712e-05, + "loss": 0.0582, + "step": 14003 + }, + { + "epoch": 1.83, + "learning_rate": 1.7988239511655237e-05, + "loss": 0.0875, + "step": 14004 + }, + { + "epoch": 1.83, + "learning_rate": 1.7987920475492268e-05, + "loss": 0.0755, + "step": 14005 + }, + { + "epoch": 1.83, + "learning_rate": 1.7987601416863697e-05, + "loss": 0.0728, + "step": 14006 + }, + { + "epoch": 1.83, + "learning_rate": 1.798728233577042e-05, + "loss": 0.081, + "step": 14007 + }, + { + "epoch": 1.83, + "learning_rate": 1.798696323221334e-05, + "loss": 0.0764, + "step": 14008 + }, + { + "epoch": 1.84, + "learning_rate": 1.798664410619335e-05, + "loss": 0.0839, + "step": 14009 + }, + { + "epoch": 1.84, + "learning_rate": 1.798632495771135e-05, + "loss": 0.0842, + "step": 14010 + }, + { + "epoch": 1.84, + "learning_rate": 1.7986005786768234e-05, + "loss": 0.0733, + "step": 14011 + }, + { + "epoch": 1.84, + "learning_rate": 1.7985686593364904e-05, + "loss": 0.0777, + "step": 14012 + }, + { + "epoch": 1.84, + "learning_rate": 1.7985367377502254e-05, + "loss": 0.079, + "step": 14013 + }, + { + "epoch": 1.84, + "learning_rate": 1.7985048139181184e-05, + "loss": 0.0606, + "step": 14014 + }, + { + "epoch": 1.84, + "learning_rate": 1.798472887840259e-05, + "loss": 0.0699, + "step": 14015 + }, + { + "epoch": 1.84, + "learning_rate": 1.7984409595167373e-05, + "loss": 0.0802, + "step": 14016 + }, + { + "epoch": 1.84, + "learning_rate": 1.798409028947643e-05, + "loss": 0.0735, + "step": 14017 + }, + { + "epoch": 1.84, + "learning_rate": 1.798377096133066e-05, + "loss": 0.0706, + "step": 14018 + }, + { + "epoch": 1.84, + "learning_rate": 1.7983451610730952e-05, + "loss": 0.0761, + "step": 14019 + }, + { + "epoch": 1.84, + "learning_rate": 1.798313223767822e-05, + "loss": 0.0827, + "step": 14020 + }, + { + "epoch": 1.84, + "learning_rate": 1.7982812842173346e-05, + "loss": 0.074, + "step": 14021 + }, + { + "epoch": 1.84, + "learning_rate": 1.798249342421724e-05, + "loss": 0.0765, + "step": 14022 + }, + { + "epoch": 1.84, + "learning_rate": 1.7982173983810796e-05, + "loss": 0.0771, + "step": 14023 + }, + { + "epoch": 1.84, + "learning_rate": 1.7981854520954914e-05, + "loss": 0.072, + "step": 14024 + }, + { + "epoch": 1.84, + "learning_rate": 1.798153503565049e-05, + "loss": 0.0826, + "step": 14025 + }, + { + "epoch": 1.84, + "learning_rate": 1.7981215527898425e-05, + "loss": 0.0762, + "step": 14026 + }, + { + "epoch": 1.84, + "learning_rate": 1.7980895997699616e-05, + "loss": 0.076, + "step": 14027 + }, + { + "epoch": 1.84, + "learning_rate": 1.798057644505496e-05, + "loss": 0.0653, + "step": 14028 + }, + { + "epoch": 1.84, + "learning_rate": 1.798025686996536e-05, + "loss": 0.0802, + "step": 14029 + }, + { + "epoch": 1.84, + "learning_rate": 1.797993727243171e-05, + "loss": 0.0771, + "step": 14030 + }, + { + "epoch": 1.84, + "learning_rate": 1.7979617652454913e-05, + "loss": 0.0752, + "step": 14031 + }, + { + "epoch": 1.84, + "learning_rate": 1.7979298010035867e-05, + "loss": 0.0736, + "step": 14032 + }, + { + "epoch": 1.84, + "learning_rate": 1.797897834517547e-05, + "loss": 0.078, + "step": 14033 + }, + { + "epoch": 1.84, + "learning_rate": 1.797865865787462e-05, + "loss": 0.0665, + "step": 14034 + }, + { + "epoch": 1.84, + "learning_rate": 1.7978338948134215e-05, + "loss": 0.0733, + "step": 14035 + }, + { + "epoch": 1.84, + "learning_rate": 1.797801921595516e-05, + "loss": 0.0774, + "step": 14036 + }, + { + "epoch": 1.84, + "learning_rate": 1.797769946133835e-05, + "loss": 0.0749, + "step": 14037 + }, + { + "epoch": 1.84, + "learning_rate": 1.7977379684284688e-05, + "loss": 0.0747, + "step": 14038 + }, + { + "epoch": 1.84, + "learning_rate": 1.797705988479507e-05, + "loss": 0.0788, + "step": 14039 + }, + { + "epoch": 1.84, + "learning_rate": 1.7976740062870393e-05, + "loss": 0.0706, + "step": 14040 + }, + { + "epoch": 1.84, + "learning_rate": 1.797642021851156e-05, + "loss": 0.0751, + "step": 14041 + }, + { + "epoch": 1.84, + "learning_rate": 1.797610035171947e-05, + "loss": 0.0725, + "step": 14042 + }, + { + "epoch": 1.84, + "learning_rate": 1.7975780462495023e-05, + "loss": 0.0775, + "step": 14043 + }, + { + "epoch": 1.84, + "learning_rate": 1.7975460550839117e-05, + "loss": 0.0722, + "step": 14044 + }, + { + "epoch": 1.84, + "learning_rate": 1.7975140616752654e-05, + "loss": 0.0714, + "step": 14045 + }, + { + "epoch": 1.84, + "learning_rate": 1.797482066023653e-05, + "loss": 0.0674, + "step": 14046 + }, + { + "epoch": 1.84, + "learning_rate": 1.797450068129165e-05, + "loss": 0.0712, + "step": 14047 + }, + { + "epoch": 1.84, + "learning_rate": 1.7974180679918914e-05, + "loss": 0.0735, + "step": 14048 + }, + { + "epoch": 1.84, + "learning_rate": 1.7973860656119217e-05, + "loss": 0.069, + "step": 14049 + }, + { + "epoch": 1.84, + "learning_rate": 1.797354060989346e-05, + "loss": 0.078, + "step": 14050 + }, + { + "epoch": 1.84, + "learning_rate": 1.7973220541242548e-05, + "loss": 0.0704, + "step": 14051 + }, + { + "epoch": 1.84, + "learning_rate": 1.7972900450167378e-05, + "loss": 0.0715, + "step": 14052 + }, + { + "epoch": 1.84, + "learning_rate": 1.7972580336668847e-05, + "loss": 0.0724, + "step": 14053 + }, + { + "epoch": 1.84, + "learning_rate": 1.7972260200747857e-05, + "loss": 0.0774, + "step": 14054 + }, + { + "epoch": 1.84, + "learning_rate": 1.7971940042405315e-05, + "loss": 0.0751, + "step": 14055 + }, + { + "epoch": 1.84, + "learning_rate": 1.7971619861642112e-05, + "loss": 0.071, + "step": 14056 + }, + { + "epoch": 1.84, + "learning_rate": 1.7971299658459153e-05, + "loss": 0.08, + "step": 14057 + }, + { + "epoch": 1.84, + "learning_rate": 1.797097943285734e-05, + "loss": 0.0778, + "step": 14058 + }, + { + "epoch": 1.84, + "learning_rate": 1.7970659184837573e-05, + "loss": 0.0765, + "step": 14059 + }, + { + "epoch": 1.84, + "learning_rate": 1.7970338914400747e-05, + "loss": 0.0717, + "step": 14060 + }, + { + "epoch": 1.84, + "learning_rate": 1.7970018621547774e-05, + "loss": 0.0707, + "step": 14061 + }, + { + "epoch": 1.84, + "learning_rate": 1.7969698306279544e-05, + "loss": 0.0747, + "step": 14062 + }, + { + "epoch": 1.84, + "learning_rate": 1.796937796859696e-05, + "loss": 0.0768, + "step": 14063 + }, + { + "epoch": 1.84, + "learning_rate": 1.7969057608500927e-05, + "loss": 0.0716, + "step": 14064 + }, + { + "epoch": 1.84, + "learning_rate": 1.7968737225992342e-05, + "loss": 0.0751, + "step": 14065 + }, + { + "epoch": 1.84, + "learning_rate": 1.796841682107211e-05, + "loss": 0.0799, + "step": 14066 + }, + { + "epoch": 1.84, + "learning_rate": 1.7968096393741132e-05, + "loss": 0.0638, + "step": 14067 + }, + { + "epoch": 1.84, + "learning_rate": 1.7967775944000303e-05, + "loss": 0.0758, + "step": 14068 + }, + { + "epoch": 1.84, + "learning_rate": 1.796745547185053e-05, + "loss": 0.0639, + "step": 14069 + }, + { + "epoch": 1.84, + "learning_rate": 1.7967134977292712e-05, + "loss": 0.0826, + "step": 14070 + }, + { + "epoch": 1.84, + "learning_rate": 1.7966814460327753e-05, + "loss": 0.0673, + "step": 14071 + }, + { + "epoch": 1.84, + "learning_rate": 1.796649392095655e-05, + "loss": 0.0829, + "step": 14072 + }, + { + "epoch": 1.84, + "learning_rate": 1.7966173359180012e-05, + "loss": 0.0702, + "step": 14073 + }, + { + "epoch": 1.84, + "learning_rate": 1.796585277499903e-05, + "loss": 0.0834, + "step": 14074 + }, + { + "epoch": 1.84, + "learning_rate": 1.7965532168414514e-05, + "loss": 0.0764, + "step": 14075 + }, + { + "epoch": 1.84, + "learning_rate": 1.7965211539427365e-05, + "loss": 0.061, + "step": 14076 + }, + { + "epoch": 1.84, + "learning_rate": 1.796489088803848e-05, + "loss": 0.0665, + "step": 14077 + }, + { + "epoch": 1.84, + "learning_rate": 1.7964570214248763e-05, + "loss": 0.0776, + "step": 14078 + }, + { + "epoch": 1.84, + "learning_rate": 1.7964249518059117e-05, + "loss": 0.0709, + "step": 14079 + }, + { + "epoch": 1.84, + "learning_rate": 1.7963928799470444e-05, + "loss": 0.0724, + "step": 14080 + }, + { + "epoch": 1.84, + "learning_rate": 1.7963608058483644e-05, + "loss": 0.0705, + "step": 14081 + }, + { + "epoch": 1.84, + "learning_rate": 1.7963287295099624e-05, + "loss": 0.0676, + "step": 14082 + }, + { + "epoch": 1.84, + "learning_rate": 1.796296650931928e-05, + "loss": 0.0681, + "step": 14083 + }, + { + "epoch": 1.84, + "learning_rate": 1.796264570114352e-05, + "loss": 0.0667, + "step": 14084 + }, + { + "epoch": 1.85, + "learning_rate": 1.7962324870573238e-05, + "loss": 0.075, + "step": 14085 + }, + { + "epoch": 1.85, + "learning_rate": 1.7962004017609343e-05, + "loss": 0.0783, + "step": 14086 + }, + { + "epoch": 1.85, + "learning_rate": 1.7961683142252734e-05, + "loss": 0.0778, + "step": 14087 + }, + { + "epoch": 1.85, + "learning_rate": 1.7961362244504317e-05, + "loss": 0.0803, + "step": 14088 + }, + { + "epoch": 1.85, + "learning_rate": 1.7961041324364994e-05, + "loss": 0.0638, + "step": 14089 + }, + { + "epoch": 1.85, + "learning_rate": 1.7960720381835663e-05, + "loss": 0.0658, + "step": 14090 + }, + { + "epoch": 1.85, + "learning_rate": 1.7960399416917232e-05, + "loss": 0.0648, + "step": 14091 + }, + { + "epoch": 1.85, + "learning_rate": 1.7960078429610602e-05, + "loss": 0.0871, + "step": 14092 + }, + { + "epoch": 1.85, + "learning_rate": 1.795975741991667e-05, + "loss": 0.0741, + "step": 14093 + }, + { + "epoch": 1.85, + "learning_rate": 1.7959436387836354e-05, + "loss": 0.0644, + "step": 14094 + }, + { + "epoch": 1.85, + "learning_rate": 1.795911533337054e-05, + "loss": 0.074, + "step": 14095 + }, + { + "epoch": 1.85, + "learning_rate": 1.795879425652014e-05, + "loss": 0.0724, + "step": 14096 + }, + { + "epoch": 1.85, + "learning_rate": 1.7958473157286053e-05, + "loss": 0.0811, + "step": 14097 + }, + { + "epoch": 1.85, + "learning_rate": 1.7958152035669187e-05, + "loss": 0.0702, + "step": 14098 + }, + { + "epoch": 1.85, + "learning_rate": 1.7957830891670443e-05, + "loss": 0.0707, + "step": 14099 + }, + { + "epoch": 1.85, + "learning_rate": 1.795750972529072e-05, + "loss": 0.0738, + "step": 14100 + }, + { + "epoch": 1.85, + "learning_rate": 1.795718853653093e-05, + "loss": 0.0661, + "step": 14101 + }, + { + "epoch": 1.85, + "learning_rate": 1.7956867325391966e-05, + "loss": 0.0616, + "step": 14102 + }, + { + "epoch": 1.85, + "learning_rate": 1.7956546091874732e-05, + "loss": 0.0795, + "step": 14103 + }, + { + "epoch": 1.85, + "learning_rate": 1.7956224835980145e-05, + "loss": 0.0789, + "step": 14104 + }, + { + "epoch": 1.85, + "learning_rate": 1.7955903557709095e-05, + "loss": 0.0758, + "step": 14105 + }, + { + "epoch": 1.85, + "learning_rate": 1.795558225706249e-05, + "loss": 0.0817, + "step": 14106 + }, + { + "epoch": 1.85, + "learning_rate": 1.7955260934041233e-05, + "loss": 0.0776, + "step": 14107 + }, + { + "epoch": 1.85, + "learning_rate": 1.795493958864623e-05, + "loss": 0.0702, + "step": 14108 + }, + { + "epoch": 1.85, + "learning_rate": 1.795461822087838e-05, + "loss": 0.0694, + "step": 14109 + }, + { + "epoch": 1.85, + "learning_rate": 1.7954296830738592e-05, + "loss": 0.0877, + "step": 14110 + }, + { + "epoch": 1.85, + "learning_rate": 1.795397541822777e-05, + "loss": 0.0778, + "step": 14111 + }, + { + "epoch": 1.85, + "learning_rate": 1.7953653983346812e-05, + "loss": 0.0866, + "step": 14112 + }, + { + "epoch": 1.85, + "learning_rate": 1.795333252609663e-05, + "loss": 0.0685, + "step": 14113 + }, + { + "epoch": 1.85, + "learning_rate": 1.795301104647812e-05, + "loss": 0.0824, + "step": 14114 + }, + { + "epoch": 1.85, + "learning_rate": 1.795268954449219e-05, + "loss": 0.0675, + "step": 14115 + }, + { + "epoch": 1.85, + "learning_rate": 1.7952368020139747e-05, + "loss": 0.0849, + "step": 14116 + }, + { + "epoch": 1.85, + "learning_rate": 1.795204647342169e-05, + "loss": 0.0736, + "step": 14117 + }, + { + "epoch": 1.85, + "learning_rate": 1.7951724904338926e-05, + "loss": 0.0672, + "step": 14118 + }, + { + "epoch": 1.85, + "learning_rate": 1.7951403312892362e-05, + "loss": 0.0885, + "step": 14119 + }, + { + "epoch": 1.85, + "learning_rate": 1.7951081699082898e-05, + "loss": 0.076, + "step": 14120 + }, + { + "epoch": 1.85, + "learning_rate": 1.795076006291144e-05, + "loss": 0.0635, + "step": 14121 + }, + { + "epoch": 1.85, + "learning_rate": 1.795043840437889e-05, + "loss": 0.0672, + "step": 14122 + }, + { + "epoch": 1.85, + "learning_rate": 1.7950116723486158e-05, + "loss": 0.0799, + "step": 14123 + }, + { + "epoch": 1.85, + "learning_rate": 1.7949795020234145e-05, + "loss": 0.0725, + "step": 14124 + }, + { + "epoch": 1.85, + "learning_rate": 1.794947329462376e-05, + "loss": 0.0847, + "step": 14125 + }, + { + "epoch": 1.85, + "learning_rate": 1.7949151546655898e-05, + "loss": 0.0698, + "step": 14126 + }, + { + "epoch": 1.85, + "learning_rate": 1.794882977633148e-05, + "loss": 0.071, + "step": 14127 + }, + { + "epoch": 1.85, + "learning_rate": 1.7948507983651395e-05, + "loss": 0.0855, + "step": 14128 + }, + { + "epoch": 1.85, + "learning_rate": 1.7948186168616555e-05, + "loss": 0.0747, + "step": 14129 + }, + { + "epoch": 1.85, + "learning_rate": 1.7947864331227863e-05, + "loss": 0.0799, + "step": 14130 + }, + { + "epoch": 1.85, + "learning_rate": 1.794754247148623e-05, + "loss": 0.0726, + "step": 14131 + }, + { + "epoch": 1.85, + "learning_rate": 1.7947220589392556e-05, + "loss": 0.076, + "step": 14132 + }, + { + "epoch": 1.85, + "learning_rate": 1.7946898684947745e-05, + "loss": 0.0747, + "step": 14133 + }, + { + "epoch": 1.85, + "learning_rate": 1.7946576758152706e-05, + "loss": 0.0583, + "step": 14134 + }, + { + "epoch": 1.85, + "learning_rate": 1.7946254809008343e-05, + "loss": 0.077, + "step": 14135 + }, + { + "epoch": 1.85, + "learning_rate": 1.7945932837515564e-05, + "loss": 0.0733, + "step": 14136 + }, + { + "epoch": 1.85, + "learning_rate": 1.794561084367527e-05, + "loss": 0.0751, + "step": 14137 + }, + { + "epoch": 1.85, + "learning_rate": 1.794528882748837e-05, + "loss": 0.0642, + "step": 14138 + }, + { + "epoch": 1.85, + "learning_rate": 1.7944966788955763e-05, + "loss": 0.0726, + "step": 14139 + }, + { + "epoch": 1.85, + "learning_rate": 1.7944644728078366e-05, + "loss": 0.0659, + "step": 14140 + }, + { + "epoch": 1.85, + "learning_rate": 1.7944322644857076e-05, + "loss": 0.0775, + "step": 14141 + }, + { + "epoch": 1.85, + "learning_rate": 1.7944000539292802e-05, + "loss": 0.0719, + "step": 14142 + }, + { + "epoch": 1.85, + "learning_rate": 1.7943678411386452e-05, + "loss": 0.0843, + "step": 14143 + }, + { + "epoch": 1.85, + "learning_rate": 1.7943356261138927e-05, + "loss": 0.068, + "step": 14144 + }, + { + "epoch": 1.85, + "learning_rate": 1.7943034088551136e-05, + "loss": 0.0791, + "step": 14145 + }, + { + "epoch": 1.85, + "learning_rate": 1.7942711893623985e-05, + "loss": 0.0733, + "step": 14146 + }, + { + "epoch": 1.85, + "learning_rate": 1.794238967635838e-05, + "loss": 0.0622, + "step": 14147 + }, + { + "epoch": 1.85, + "learning_rate": 1.7942067436755224e-05, + "loss": 0.078, + "step": 14148 + }, + { + "epoch": 1.85, + "learning_rate": 1.7941745174815428e-05, + "loss": 0.067, + "step": 14149 + }, + { + "epoch": 1.85, + "learning_rate": 1.7941422890539898e-05, + "loss": 0.0848, + "step": 14150 + }, + { + "epoch": 1.85, + "learning_rate": 1.7941100583929535e-05, + "loss": 0.0859, + "step": 14151 + }, + { + "epoch": 1.85, + "learning_rate": 1.7940778254985255e-05, + "loss": 0.0649, + "step": 14152 + }, + { + "epoch": 1.85, + "learning_rate": 1.7940455903707955e-05, + "loss": 0.0702, + "step": 14153 + }, + { + "epoch": 1.85, + "learning_rate": 1.794013353009855e-05, + "loss": 0.074, + "step": 14154 + }, + { + "epoch": 1.85, + "learning_rate": 1.7939811134157937e-05, + "loss": 0.0765, + "step": 14155 + }, + { + "epoch": 1.85, + "learning_rate": 1.7939488715887033e-05, + "loss": 0.0723, + "step": 14156 + }, + { + "epoch": 1.85, + "learning_rate": 1.7939166275286737e-05, + "loss": 0.0708, + "step": 14157 + }, + { + "epoch": 1.85, + "learning_rate": 1.793884381235796e-05, + "loss": 0.0698, + "step": 14158 + }, + { + "epoch": 1.85, + "learning_rate": 1.7938521327101605e-05, + "loss": 0.0662, + "step": 14159 + }, + { + "epoch": 1.85, + "learning_rate": 1.7938198819518585e-05, + "loss": 0.0737, + "step": 14160 + }, + { + "epoch": 1.85, + "learning_rate": 1.7937876289609798e-05, + "loss": 0.0736, + "step": 14161 + }, + { + "epoch": 1.86, + "learning_rate": 1.793755373737616e-05, + "loss": 0.0672, + "step": 14162 + }, + { + "epoch": 1.86, + "learning_rate": 1.7937231162818577e-05, + "loss": 0.0813, + "step": 14163 + }, + { + "epoch": 1.86, + "learning_rate": 1.7936908565937955e-05, + "loss": 0.074, + "step": 14164 + }, + { + "epoch": 1.86, + "learning_rate": 1.7936585946735196e-05, + "loss": 0.0739, + "step": 14165 + }, + { + "epoch": 1.86, + "learning_rate": 1.793626330521121e-05, + "loss": 0.0745, + "step": 14166 + }, + { + "epoch": 1.86, + "learning_rate": 1.7935940641366912e-05, + "loss": 0.0944, + "step": 14167 + }, + { + "epoch": 1.86, + "learning_rate": 1.79356179552032e-05, + "loss": 0.065, + "step": 14168 + }, + { + "epoch": 1.86, + "learning_rate": 1.7935295246720986e-05, + "loss": 0.0664, + "step": 14169 + }, + { + "epoch": 1.86, + "learning_rate": 1.7934972515921175e-05, + "loss": 0.0845, + "step": 14170 + }, + { + "epoch": 1.86, + "learning_rate": 1.793464976280468e-05, + "loss": 0.0808, + "step": 14171 + }, + { + "epoch": 1.86, + "learning_rate": 1.79343269873724e-05, + "loss": 0.074, + "step": 14172 + }, + { + "epoch": 1.86, + "learning_rate": 1.7934004189625255e-05, + "loss": 0.0781, + "step": 14173 + }, + { + "epoch": 1.86, + "learning_rate": 1.793368136956414e-05, + "loss": 0.0707, + "step": 14174 + }, + { + "epoch": 1.86, + "learning_rate": 1.7933358527189968e-05, + "loss": 0.0689, + "step": 14175 + }, + { + "epoch": 1.86, + "learning_rate": 1.7933035662503653e-05, + "loss": 0.0651, + "step": 14176 + }, + { + "epoch": 1.86, + "learning_rate": 1.7932712775506092e-05, + "loss": 0.071, + "step": 14177 + }, + { + "epoch": 1.86, + "learning_rate": 1.7932389866198204e-05, + "loss": 0.0807, + "step": 14178 + }, + { + "epoch": 1.86, + "learning_rate": 1.7932066934580888e-05, + "loss": 0.0805, + "step": 14179 + }, + { + "epoch": 1.86, + "learning_rate": 1.7931743980655055e-05, + "loss": 0.0783, + "step": 14180 + }, + { + "epoch": 1.86, + "learning_rate": 1.7931421004421617e-05, + "loss": 0.0769, + "step": 14181 + }, + { + "epoch": 1.86, + "learning_rate": 1.793109800588148e-05, + "loss": 0.071, + "step": 14182 + }, + { + "epoch": 1.86, + "learning_rate": 1.7930774985035553e-05, + "loss": 0.0814, + "step": 14183 + }, + { + "epoch": 1.86, + "learning_rate": 1.7930451941884743e-05, + "loss": 0.0776, + "step": 14184 + }, + { + "epoch": 1.86, + "learning_rate": 1.7930128876429957e-05, + "loss": 0.0739, + "step": 14185 + }, + { + "epoch": 1.86, + "learning_rate": 1.792980578867211e-05, + "loss": 0.0839, + "step": 14186 + }, + { + "epoch": 1.86, + "learning_rate": 1.7929482678612103e-05, + "loss": 0.0782, + "step": 14187 + }, + { + "epoch": 1.86, + "learning_rate": 1.792915954625085e-05, + "loss": 0.0714, + "step": 14188 + }, + { + "epoch": 1.86, + "learning_rate": 1.7928836391589256e-05, + "loss": 0.0573, + "step": 14189 + }, + { + "epoch": 1.86, + "learning_rate": 1.7928513214628234e-05, + "loss": 0.0685, + "step": 14190 + }, + { + "epoch": 1.86, + "learning_rate": 1.7928190015368693e-05, + "loss": 0.0763, + "step": 14191 + }, + { + "epoch": 1.86, + "learning_rate": 1.792786679381153e-05, + "loss": 0.0863, + "step": 14192 + }, + { + "epoch": 1.86, + "learning_rate": 1.7927543549957674e-05, + "loss": 0.0618, + "step": 14193 + }, + { + "epoch": 1.86, + "learning_rate": 1.7927220283808023e-05, + "loss": 0.0754, + "step": 14194 + }, + { + "epoch": 1.86, + "learning_rate": 1.7926896995363482e-05, + "loss": 0.082, + "step": 14195 + }, + { + "epoch": 1.86, + "learning_rate": 1.792657368462497e-05, + "loss": 0.0769, + "step": 14196 + }, + { + "epoch": 1.86, + "learning_rate": 1.792625035159339e-05, + "loss": 0.0677, + "step": 14197 + }, + { + "epoch": 1.86, + "learning_rate": 1.792592699626965e-05, + "loss": 0.0725, + "step": 14198 + }, + { + "epoch": 1.86, + "learning_rate": 1.7925603618654665e-05, + "loss": 0.0727, + "step": 14199 + }, + { + "epoch": 1.86, + "learning_rate": 1.7925280218749343e-05, + "loss": 0.0744, + "step": 14200 + }, + { + "epoch": 1.86, + "learning_rate": 1.792495679655459e-05, + "loss": 0.0809, + "step": 14201 + }, + { + "epoch": 1.86, + "learning_rate": 1.792463335207132e-05, + "loss": 0.0673, + "step": 14202 + }, + { + "epoch": 1.86, + "learning_rate": 1.792430988530044e-05, + "loss": 0.0757, + "step": 14203 + }, + { + "epoch": 1.86, + "learning_rate": 1.792398639624286e-05, + "loss": 0.0673, + "step": 14204 + }, + { + "epoch": 1.86, + "learning_rate": 1.7923662884899493e-05, + "loss": 0.063, + "step": 14205 + }, + { + "epoch": 1.86, + "learning_rate": 1.7923339351271243e-05, + "loss": 0.0733, + "step": 14206 + }, + { + "epoch": 1.86, + "learning_rate": 1.7923015795359023e-05, + "loss": 0.0776, + "step": 14207 + }, + { + "epoch": 1.86, + "learning_rate": 1.7922692217163745e-05, + "loss": 0.0775, + "step": 14208 + }, + { + "epoch": 1.86, + "learning_rate": 1.7922368616686317e-05, + "loss": 0.085, + "step": 14209 + }, + { + "epoch": 1.86, + "learning_rate": 1.7922044993927646e-05, + "loss": 0.0752, + "step": 14210 + }, + { + "epoch": 1.86, + "learning_rate": 1.792172134888865e-05, + "loss": 0.0692, + "step": 14211 + }, + { + "epoch": 1.86, + "learning_rate": 1.792139768157023e-05, + "loss": 0.0888, + "step": 14212 + }, + { + "epoch": 1.86, + "learning_rate": 1.7921073991973305e-05, + "loss": 0.0678, + "step": 14213 + }, + { + "epoch": 1.86, + "learning_rate": 1.792075028009878e-05, + "loss": 0.0792, + "step": 14214 + }, + { + "epoch": 1.86, + "learning_rate": 1.792042654594756e-05, + "loss": 0.071, + "step": 14215 + }, + { + "epoch": 1.86, + "learning_rate": 1.792010278952057e-05, + "loss": 0.0719, + "step": 14216 + }, + { + "epoch": 1.86, + "learning_rate": 1.791977901081871e-05, + "loss": 0.074, + "step": 14217 + }, + { + "epoch": 1.86, + "learning_rate": 1.7919455209842892e-05, + "loss": 0.0788, + "step": 14218 + }, + { + "epoch": 1.86, + "learning_rate": 1.7919131386594034e-05, + "loss": 0.0715, + "step": 14219 + }, + { + "epoch": 1.86, + "learning_rate": 1.7918807541073037e-05, + "loss": 0.0819, + "step": 14220 + }, + { + "epoch": 1.86, + "learning_rate": 1.7918483673280813e-05, + "loss": 0.0684, + "step": 14221 + }, + { + "epoch": 1.86, + "learning_rate": 1.791815978321828e-05, + "loss": 0.0765, + "step": 14222 + }, + { + "epoch": 1.86, + "learning_rate": 1.791783587088634e-05, + "loss": 0.0799, + "step": 14223 + }, + { + "epoch": 1.86, + "learning_rate": 1.7917511936285906e-05, + "loss": 0.069, + "step": 14224 + }, + { + "epoch": 1.86, + "learning_rate": 1.7917187979417898e-05, + "loss": 0.0836, + "step": 14225 + }, + { + "epoch": 1.86, + "learning_rate": 1.7916864000283217e-05, + "loss": 0.0725, + "step": 14226 + }, + { + "epoch": 1.86, + "learning_rate": 1.7916539998882777e-05, + "loss": 0.0621, + "step": 14227 + }, + { + "epoch": 1.86, + "learning_rate": 1.7916215975217493e-05, + "loss": 0.0742, + "step": 14228 + }, + { + "epoch": 1.86, + "learning_rate": 1.791589192928827e-05, + "loss": 0.0722, + "step": 14229 + }, + { + "epoch": 1.86, + "learning_rate": 1.7915567861096023e-05, + "loss": 0.0764, + "step": 14230 + }, + { + "epoch": 1.86, + "learning_rate": 1.7915243770641667e-05, + "loss": 0.0701, + "step": 14231 + }, + { + "epoch": 1.86, + "learning_rate": 1.7914919657926105e-05, + "loss": 0.0719, + "step": 14232 + }, + { + "epoch": 1.86, + "learning_rate": 1.7914595522950253e-05, + "loss": 0.0691, + "step": 14233 + }, + { + "epoch": 1.86, + "learning_rate": 1.7914271365715023e-05, + "loss": 0.0693, + "step": 14234 + }, + { + "epoch": 1.86, + "learning_rate": 1.7913947186221325e-05, + "loss": 0.0743, + "step": 14235 + }, + { + "epoch": 1.86, + "learning_rate": 1.7913622984470075e-05, + "loss": 0.0775, + "step": 14236 + }, + { + "epoch": 1.86, + "learning_rate": 1.791329876046218e-05, + "loss": 0.0845, + "step": 14237 + }, + { + "epoch": 1.87, + "learning_rate": 1.7912974514198555e-05, + "loss": 0.0715, + "step": 14238 + }, + { + "epoch": 1.87, + "learning_rate": 1.791265024568011e-05, + "loss": 0.071, + "step": 14239 + }, + { + "epoch": 1.87, + "learning_rate": 1.7912325954907756e-05, + "loss": 0.0726, + "step": 14240 + }, + { + "epoch": 1.87, + "learning_rate": 1.7912001641882408e-05, + "loss": 0.0777, + "step": 14241 + }, + { + "epoch": 1.87, + "learning_rate": 1.791167730660498e-05, + "loss": 0.0681, + "step": 14242 + }, + { + "epoch": 1.87, + "learning_rate": 1.7911352949076377e-05, + "loss": 0.0743, + "step": 14243 + }, + { + "epoch": 1.87, + "learning_rate": 1.7911028569297513e-05, + "loss": 0.0722, + "step": 14244 + }, + { + "epoch": 1.87, + "learning_rate": 1.7910704167269304e-05, + "loss": 0.0713, + "step": 14245 + }, + { + "epoch": 1.87, + "learning_rate": 1.7910379742992666e-05, + "loss": 0.0665, + "step": 14246 + }, + { + "epoch": 1.87, + "learning_rate": 1.79100552964685e-05, + "loss": 0.0815, + "step": 14247 + }, + { + "epoch": 1.87, + "learning_rate": 1.7909730827697727e-05, + "loss": 0.0799, + "step": 14248 + }, + { + "epoch": 1.87, + "learning_rate": 1.7909406336681256e-05, + "loss": 0.0869, + "step": 14249 + }, + { + "epoch": 1.87, + "learning_rate": 1.7909081823420002e-05, + "loss": 0.0662, + "step": 14250 + }, + { + "epoch": 1.87, + "learning_rate": 1.7908757287914874e-05, + "loss": 0.083, + "step": 14251 + }, + { + "epoch": 1.87, + "learning_rate": 1.7908432730166792e-05, + "loss": 0.0684, + "step": 14252 + }, + { + "epoch": 1.87, + "learning_rate": 1.7908108150176665e-05, + "loss": 0.0786, + "step": 14253 + }, + { + "epoch": 1.87, + "learning_rate": 1.7907783547945402e-05, + "loss": 0.069, + "step": 14254 + }, + { + "epoch": 1.87, + "learning_rate": 1.7907458923473916e-05, + "loss": 0.0814, + "step": 14255 + }, + { + "epoch": 1.87, + "learning_rate": 1.7907134276763126e-05, + "loss": 0.0763, + "step": 14256 + }, + { + "epoch": 1.87, + "learning_rate": 1.7906809607813944e-05, + "loss": 0.0774, + "step": 14257 + }, + { + "epoch": 1.87, + "learning_rate": 1.7906484916627277e-05, + "loss": 0.065, + "step": 14258 + }, + { + "epoch": 1.87, + "learning_rate": 1.7906160203204043e-05, + "loss": 0.0785, + "step": 14259 + }, + { + "epoch": 1.87, + "learning_rate": 1.7905835467545154e-05, + "loss": 0.0816, + "step": 14260 + }, + { + "epoch": 1.87, + "learning_rate": 1.7905510709651526e-05, + "loss": 0.086, + "step": 14261 + }, + { + "epoch": 1.87, + "learning_rate": 1.7905185929524068e-05, + "loss": 0.0628, + "step": 14262 + }, + { + "epoch": 1.87, + "learning_rate": 1.7904861127163696e-05, + "loss": 0.0744, + "step": 14263 + }, + { + "epoch": 1.87, + "learning_rate": 1.7904536302571325e-05, + "loss": 0.0721, + "step": 14264 + }, + { + "epoch": 1.87, + "learning_rate": 1.7904211455747863e-05, + "loss": 0.0686, + "step": 14265 + }, + { + "epoch": 1.87, + "learning_rate": 1.790388658669423e-05, + "loss": 0.0732, + "step": 14266 + }, + { + "epoch": 1.87, + "learning_rate": 1.790356169541134e-05, + "loss": 0.0763, + "step": 14267 + }, + { + "epoch": 1.87, + "learning_rate": 1.7903236781900098e-05, + "loss": 0.0619, + "step": 14268 + }, + { + "epoch": 1.87, + "learning_rate": 1.7902911846161426e-05, + "loss": 0.0672, + "step": 14269 + }, + { + "epoch": 1.87, + "learning_rate": 1.7902586888196233e-05, + "loss": 0.0725, + "step": 14270 + }, + { + "epoch": 1.87, + "learning_rate": 1.7902261908005438e-05, + "loss": 0.0732, + "step": 14271 + }, + { + "epoch": 1.87, + "learning_rate": 1.790193690558995e-05, + "loss": 0.0673, + "step": 14272 + }, + { + "epoch": 1.87, + "learning_rate": 1.790161188095069e-05, + "loss": 0.0619, + "step": 14273 + }, + { + "epoch": 1.87, + "learning_rate": 1.7901286834088562e-05, + "loss": 0.0742, + "step": 14274 + }, + { + "epoch": 1.87, + "learning_rate": 1.7900961765004488e-05, + "loss": 0.0612, + "step": 14275 + }, + { + "epoch": 1.87, + "learning_rate": 1.790063667369938e-05, + "loss": 0.0728, + "step": 14276 + }, + { + "epoch": 1.87, + "learning_rate": 1.790031156017415e-05, + "loss": 0.0823, + "step": 14277 + }, + { + "epoch": 1.87, + "learning_rate": 1.7899986424429714e-05, + "loss": 0.0681, + "step": 14278 + }, + { + "epoch": 1.87, + "learning_rate": 1.7899661266466988e-05, + "loss": 0.0728, + "step": 14279 + }, + { + "epoch": 1.87, + "learning_rate": 1.7899336086286886e-05, + "loss": 0.0724, + "step": 14280 + }, + { + "epoch": 1.87, + "learning_rate": 1.7899010883890322e-05, + "loss": 0.0551, + "step": 14281 + }, + { + "epoch": 1.87, + "learning_rate": 1.789868565927821e-05, + "loss": 0.0724, + "step": 14282 + }, + { + "epoch": 1.87, + "learning_rate": 1.7898360412451467e-05, + "loss": 0.0684, + "step": 14283 + }, + { + "epoch": 1.87, + "learning_rate": 1.7898035143411005e-05, + "loss": 0.0765, + "step": 14284 + }, + { + "epoch": 1.87, + "learning_rate": 1.7897709852157738e-05, + "loss": 0.0718, + "step": 14285 + }, + { + "epoch": 1.87, + "learning_rate": 1.7897384538692584e-05, + "loss": 0.0582, + "step": 14286 + }, + { + "epoch": 1.87, + "learning_rate": 1.7897059203016457e-05, + "loss": 0.0673, + "step": 14287 + }, + { + "epoch": 1.87, + "learning_rate": 1.789673384513027e-05, + "loss": 0.0686, + "step": 14288 + }, + { + "epoch": 1.87, + "learning_rate": 1.789640846503494e-05, + "loss": 0.0796, + "step": 14289 + }, + { + "epoch": 1.87, + "learning_rate": 1.7896083062731385e-05, + "loss": 0.068, + "step": 14290 + }, + { + "epoch": 1.87, + "learning_rate": 1.7895757638220513e-05, + "loss": 0.0757, + "step": 14291 + }, + { + "epoch": 1.87, + "learning_rate": 1.7895432191503242e-05, + "loss": 0.0746, + "step": 14292 + }, + { + "epoch": 1.87, + "learning_rate": 1.7895106722580494e-05, + "loss": 0.0696, + "step": 14293 + }, + { + "epoch": 1.87, + "learning_rate": 1.7894781231453174e-05, + "loss": 0.0736, + "step": 14294 + }, + { + "epoch": 1.87, + "learning_rate": 1.7894455718122203e-05, + "loss": 0.0775, + "step": 14295 + }, + { + "epoch": 1.87, + "learning_rate": 1.78941301825885e-05, + "loss": 0.0675, + "step": 14296 + }, + { + "epoch": 1.87, + "learning_rate": 1.789380462485297e-05, + "loss": 0.0798, + "step": 14297 + }, + { + "epoch": 1.87, + "learning_rate": 1.7893479044916534e-05, + "loss": 0.0766, + "step": 14298 + }, + { + "epoch": 1.87, + "learning_rate": 1.7893153442780112e-05, + "loss": 0.0788, + "step": 14299 + }, + { + "epoch": 1.87, + "learning_rate": 1.789282781844462e-05, + "loss": 0.0798, + "step": 14300 + }, + { + "epoch": 1.87, + "learning_rate": 1.7892502171910965e-05, + "loss": 0.072, + "step": 14301 + }, + { + "epoch": 1.87, + "learning_rate": 1.789217650318007e-05, + "loss": 0.0746, + "step": 14302 + }, + { + "epoch": 1.87, + "learning_rate": 1.7891850812252847e-05, + "loss": 0.0644, + "step": 14303 + }, + { + "epoch": 1.87, + "learning_rate": 1.7891525099130215e-05, + "loss": 0.0755, + "step": 14304 + }, + { + "epoch": 1.87, + "learning_rate": 1.789119936381309e-05, + "loss": 0.0822, + "step": 14305 + }, + { + "epoch": 1.87, + "learning_rate": 1.7890873606302387e-05, + "loss": 0.0672, + "step": 14306 + }, + { + "epoch": 1.87, + "learning_rate": 1.7890547826599022e-05, + "loss": 0.0677, + "step": 14307 + }, + { + "epoch": 1.87, + "learning_rate": 1.789022202470391e-05, + "loss": 0.0812, + "step": 14308 + }, + { + "epoch": 1.87, + "learning_rate": 1.7889896200617967e-05, + "loss": 0.0824, + "step": 14309 + }, + { + "epoch": 1.87, + "learning_rate": 1.7889570354342115e-05, + "loss": 0.0579, + "step": 14310 + }, + { + "epoch": 1.87, + "learning_rate": 1.7889244485877264e-05, + "loss": 0.0635, + "step": 14311 + }, + { + "epoch": 1.87, + "learning_rate": 1.7888918595224333e-05, + "loss": 0.0778, + "step": 14312 + }, + { + "epoch": 1.87, + "learning_rate": 1.7888592682384242e-05, + "loss": 0.0851, + "step": 14313 + }, + { + "epoch": 1.88, + "learning_rate": 1.78882667473579e-05, + "loss": 0.0784, + "step": 14314 + }, + { + "epoch": 1.88, + "learning_rate": 1.788794079014623e-05, + "loss": 0.0739, + "step": 14315 + }, + { + "epoch": 1.88, + "learning_rate": 1.7887614810750148e-05, + "loss": 0.0701, + "step": 14316 + }, + { + "epoch": 1.88, + "learning_rate": 1.7887288809170566e-05, + "loss": 0.06, + "step": 14317 + }, + { + "epoch": 1.88, + "learning_rate": 1.7886962785408406e-05, + "loss": 0.0752, + "step": 14318 + }, + { + "epoch": 1.88, + "learning_rate": 1.7886636739464582e-05, + "loss": 0.0662, + "step": 14319 + }, + { + "epoch": 1.88, + "learning_rate": 1.7886310671340014e-05, + "loss": 0.0643, + "step": 14320 + }, + { + "epoch": 1.88, + "learning_rate": 1.7885984581035617e-05, + "loss": 0.0625, + "step": 14321 + }, + { + "epoch": 1.88, + "learning_rate": 1.7885658468552306e-05, + "loss": 0.0667, + "step": 14322 + }, + { + "epoch": 1.88, + "learning_rate": 1.7885332333891003e-05, + "loss": 0.0756, + "step": 14323 + }, + { + "epoch": 1.88, + "learning_rate": 1.7885006177052622e-05, + "loss": 0.073, + "step": 14324 + }, + { + "epoch": 1.88, + "learning_rate": 1.788467999803808e-05, + "loss": 0.0833, + "step": 14325 + }, + { + "epoch": 1.88, + "learning_rate": 1.7884353796848297e-05, + "loss": 0.0661, + "step": 14326 + }, + { + "epoch": 1.88, + "learning_rate": 1.788402757348419e-05, + "loss": 0.0636, + "step": 14327 + }, + { + "epoch": 1.88, + "learning_rate": 1.788370132794667e-05, + "loss": 0.077, + "step": 14328 + }, + { + "epoch": 1.88, + "learning_rate": 1.7883375060236664e-05, + "loss": 0.08, + "step": 14329 + }, + { + "epoch": 1.88, + "learning_rate": 1.7883048770355086e-05, + "loss": 0.0753, + "step": 14330 + }, + { + "epoch": 1.88, + "learning_rate": 1.788272245830285e-05, + "loss": 0.0715, + "step": 14331 + }, + { + "epoch": 1.88, + "learning_rate": 1.788239612408088e-05, + "loss": 0.065, + "step": 14332 + }, + { + "epoch": 1.88, + "learning_rate": 1.788206976769009e-05, + "loss": 0.0717, + "step": 14333 + }, + { + "epoch": 1.88, + "learning_rate": 1.7881743389131398e-05, + "loss": 0.0826, + "step": 14334 + }, + { + "epoch": 1.88, + "learning_rate": 1.7881416988405723e-05, + "loss": 0.0759, + "step": 14335 + }, + { + "epoch": 1.88, + "learning_rate": 1.7881090565513978e-05, + "loss": 0.0684, + "step": 14336 + }, + { + "epoch": 1.88, + "learning_rate": 1.788076412045709e-05, + "loss": 0.0657, + "step": 14337 + }, + { + "epoch": 1.88, + "learning_rate": 1.788043765323597e-05, + "loss": 0.0666, + "step": 14338 + }, + { + "epoch": 1.88, + "learning_rate": 1.7880111163851543e-05, + "loss": 0.0747, + "step": 14339 + }, + { + "epoch": 1.88, + "learning_rate": 1.787978465230472e-05, + "loss": 0.0673, + "step": 14340 + }, + { + "epoch": 1.88, + "learning_rate": 1.7879458118596423e-05, + "loss": 0.0744, + "step": 14341 + }, + { + "epoch": 1.88, + "learning_rate": 1.7879131562727566e-05, + "loss": 0.0806, + "step": 14342 + }, + { + "epoch": 1.88, + "learning_rate": 1.7878804984699078e-05, + "loss": 0.068, + "step": 14343 + }, + { + "epoch": 1.88, + "learning_rate": 1.7878478384511865e-05, + "loss": 0.07, + "step": 14344 + }, + { + "epoch": 1.88, + "learning_rate": 1.7878151762166857e-05, + "loss": 0.0808, + "step": 14345 + }, + { + "epoch": 1.88, + "learning_rate": 1.787782511766496e-05, + "loss": 0.0763, + "step": 14346 + }, + { + "epoch": 1.88, + "learning_rate": 1.78774984510071e-05, + "loss": 0.0816, + "step": 14347 + }, + { + "epoch": 1.88, + "learning_rate": 1.78771717621942e-05, + "loss": 0.0698, + "step": 14348 + }, + { + "epoch": 1.88, + "learning_rate": 1.787684505122717e-05, + "loss": 0.0893, + "step": 14349 + }, + { + "epoch": 1.88, + "learning_rate": 1.7876518318106937e-05, + "loss": 0.0768, + "step": 14350 + }, + { + "epoch": 1.88, + "learning_rate": 1.7876191562834412e-05, + "loss": 0.0759, + "step": 14351 + }, + { + "epoch": 1.88, + "learning_rate": 1.787586478541052e-05, + "loss": 0.0718, + "step": 14352 + }, + { + "epoch": 1.88, + "learning_rate": 1.7875537985836175e-05, + "loss": 0.0765, + "step": 14353 + }, + { + "epoch": 1.88, + "learning_rate": 1.78752111641123e-05, + "loss": 0.0756, + "step": 14354 + }, + { + "epoch": 1.88, + "learning_rate": 1.787488432023981e-05, + "loss": 0.0706, + "step": 14355 + }, + { + "epoch": 1.88, + "learning_rate": 1.787455745421963e-05, + "loss": 0.0819, + "step": 14356 + }, + { + "epoch": 1.88, + "learning_rate": 1.787423056605268e-05, + "loss": 0.0715, + "step": 14357 + }, + { + "epoch": 1.88, + "learning_rate": 1.787390365573987e-05, + "loss": 0.0729, + "step": 14358 + }, + { + "epoch": 1.88, + "learning_rate": 1.787357672328213e-05, + "loss": 0.0825, + "step": 14359 + }, + { + "epoch": 1.88, + "learning_rate": 1.787324976868037e-05, + "loss": 0.0693, + "step": 14360 + }, + { + "epoch": 1.88, + "learning_rate": 1.7872922791935517e-05, + "loss": 0.0805, + "step": 14361 + }, + { + "epoch": 1.88, + "learning_rate": 1.7872595793048487e-05, + "loss": 0.0756, + "step": 14362 + }, + { + "epoch": 1.88, + "learning_rate": 1.78722687720202e-05, + "loss": 0.0671, + "step": 14363 + }, + { + "epoch": 1.88, + "learning_rate": 1.787194172885158e-05, + "loss": 0.0755, + "step": 14364 + }, + { + "epoch": 1.88, + "learning_rate": 1.7871614663543542e-05, + "loss": 0.0747, + "step": 14365 + }, + { + "epoch": 1.88, + "learning_rate": 1.7871287576097005e-05, + "loss": 0.0782, + "step": 14366 + }, + { + "epoch": 1.88, + "learning_rate": 1.7870960466512892e-05, + "loss": 0.0638, + "step": 14367 + }, + { + "epoch": 1.88, + "learning_rate": 1.787063333479212e-05, + "loss": 0.0805, + "step": 14368 + }, + { + "epoch": 1.88, + "learning_rate": 1.7870306180935613e-05, + "loss": 0.0749, + "step": 14369 + }, + { + "epoch": 1.88, + "learning_rate": 1.786997900494429e-05, + "loss": 0.073, + "step": 14370 + }, + { + "epoch": 1.88, + "learning_rate": 1.7869651806819067e-05, + "loss": 0.0772, + "step": 14371 + }, + { + "epoch": 1.88, + "learning_rate": 1.7869324586560867e-05, + "loss": 0.0746, + "step": 14372 + }, + { + "epoch": 1.88, + "learning_rate": 1.7868997344170614e-05, + "loss": 0.0739, + "step": 14373 + }, + { + "epoch": 1.88, + "learning_rate": 1.786867007964922e-05, + "loss": 0.075, + "step": 14374 + }, + { + "epoch": 1.88, + "learning_rate": 1.7868342792997617e-05, + "loss": 0.0789, + "step": 14375 + }, + { + "epoch": 1.88, + "learning_rate": 1.7868015484216715e-05, + "loss": 0.0638, + "step": 14376 + }, + { + "epoch": 1.88, + "learning_rate": 1.786768815330744e-05, + "loss": 0.0708, + "step": 14377 + }, + { + "epoch": 1.88, + "learning_rate": 1.786736080027071e-05, + "loss": 0.0786, + "step": 14378 + }, + { + "epoch": 1.88, + "learning_rate": 1.7867033425107446e-05, + "loss": 0.0745, + "step": 14379 + }, + { + "epoch": 1.88, + "learning_rate": 1.786670602781857e-05, + "loss": 0.0778, + "step": 14380 + }, + { + "epoch": 1.88, + "learning_rate": 1.7866378608405e-05, + "loss": 0.0828, + "step": 14381 + }, + { + "epoch": 1.88, + "learning_rate": 1.7866051166867663e-05, + "loss": 0.0745, + "step": 14382 + }, + { + "epoch": 1.88, + "learning_rate": 1.7865723703207475e-05, + "loss": 0.0689, + "step": 14383 + }, + { + "epoch": 1.88, + "learning_rate": 1.786539621742536e-05, + "loss": 0.0671, + "step": 14384 + }, + { + "epoch": 1.88, + "learning_rate": 1.7865068709522232e-05, + "loss": 0.0757, + "step": 14385 + }, + { + "epoch": 1.88, + "learning_rate": 1.786474117949902e-05, + "loss": 0.0647, + "step": 14386 + }, + { + "epoch": 1.88, + "learning_rate": 1.7864413627356643e-05, + "loss": 0.0781, + "step": 14387 + }, + { + "epoch": 1.88, + "learning_rate": 1.786408605309602e-05, + "loss": 0.077, + "step": 14388 + }, + { + "epoch": 1.88, + "learning_rate": 1.7863758456718073e-05, + "loss": 0.0765, + "step": 14389 + }, + { + "epoch": 1.88, + "learning_rate": 1.7863430838223728e-05, + "loss": 0.0689, + "step": 14390 + }, + { + "epoch": 1.89, + "learning_rate": 1.7863103197613898e-05, + "loss": 0.0679, + "step": 14391 + }, + { + "epoch": 1.89, + "learning_rate": 1.786277553488951e-05, + "loss": 0.0778, + "step": 14392 + }, + { + "epoch": 1.89, + "learning_rate": 1.7862447850051487e-05, + "loss": 0.0719, + "step": 14393 + }, + { + "epoch": 1.89, + "learning_rate": 1.7862120143100746e-05, + "loss": 0.0658, + "step": 14394 + }, + { + "epoch": 1.89, + "learning_rate": 1.7861792414038213e-05, + "loss": 0.071, + "step": 14395 + }, + { + "epoch": 1.89, + "learning_rate": 1.7861464662864806e-05, + "loss": 0.0729, + "step": 14396 + }, + { + "epoch": 1.89, + "learning_rate": 1.7861136889581448e-05, + "loss": 0.0814, + "step": 14397 + }, + { + "epoch": 1.89, + "learning_rate": 1.786080909418906e-05, + "loss": 0.0724, + "step": 14398 + }, + { + "epoch": 1.89, + "learning_rate": 1.7860481276688565e-05, + "loss": 0.0676, + "step": 14399 + }, + { + "epoch": 1.89, + "learning_rate": 1.7860153437080887e-05, + "loss": 0.0719, + "step": 14400 + }, + { + "epoch": 1.89, + "learning_rate": 1.7859825575366946e-05, + "loss": 0.088, + "step": 14401 + }, + { + "epoch": 1.89, + "learning_rate": 1.7859497691547662e-05, + "loss": 0.0703, + "step": 14402 + }, + { + "epoch": 1.89, + "learning_rate": 1.785916978562396e-05, + "loss": 0.0832, + "step": 14403 + }, + { + "epoch": 1.89, + "learning_rate": 1.7858841857596764e-05, + "loss": 0.0713, + "step": 14404 + }, + { + "epoch": 1.89, + "learning_rate": 1.7858513907466994e-05, + "loss": 0.0798, + "step": 14405 + }, + { + "epoch": 1.89, + "learning_rate": 1.7858185935235566e-05, + "loss": 0.0733, + "step": 14406 + }, + { + "epoch": 1.89, + "learning_rate": 1.7857857940903415e-05, + "loss": 0.0763, + "step": 14407 + }, + { + "epoch": 1.89, + "learning_rate": 1.7857529924471456e-05, + "loss": 0.0738, + "step": 14408 + }, + { + "epoch": 1.89, + "learning_rate": 1.7857201885940608e-05, + "loss": 0.0833, + "step": 14409 + }, + { + "epoch": 1.89, + "learning_rate": 1.78568738253118e-05, + "loss": 0.0716, + "step": 14410 + }, + { + "epoch": 1.89, + "learning_rate": 1.7856545742585956e-05, + "loss": 0.07, + "step": 14411 + }, + { + "epoch": 1.89, + "learning_rate": 1.785621763776399e-05, + "loss": 0.0722, + "step": 14412 + }, + { + "epoch": 1.89, + "learning_rate": 1.7855889510846837e-05, + "loss": 0.0576, + "step": 14413 + }, + { + "epoch": 1.89, + "learning_rate": 1.7855561361835408e-05, + "loss": 0.0715, + "step": 14414 + }, + { + "epoch": 1.89, + "learning_rate": 1.7855233190730635e-05, + "loss": 0.0739, + "step": 14415 + }, + { + "epoch": 1.89, + "learning_rate": 1.7854904997533434e-05, + "loss": 0.0702, + "step": 14416 + }, + { + "epoch": 1.89, + "learning_rate": 1.7854576782244733e-05, + "loss": 0.086, + "step": 14417 + }, + { + "epoch": 1.89, + "learning_rate": 1.785424854486545e-05, + "loss": 0.0751, + "step": 14418 + }, + { + "epoch": 1.89, + "learning_rate": 1.7853920285396517e-05, + "loss": 0.0683, + "step": 14419 + }, + { + "epoch": 1.89, + "learning_rate": 1.7853592003838844e-05, + "loss": 0.0746, + "step": 14420 + }, + { + "epoch": 1.89, + "learning_rate": 1.7853263700193365e-05, + "loss": 0.0686, + "step": 14421 + }, + { + "epoch": 1.89, + "learning_rate": 1.7852935374461003e-05, + "loss": 0.0792, + "step": 14422 + }, + { + "epoch": 1.89, + "learning_rate": 1.7852607026642674e-05, + "loss": 0.0849, + "step": 14423 + }, + { + "epoch": 1.89, + "learning_rate": 1.785227865673931e-05, + "loss": 0.0658, + "step": 14424 + }, + { + "epoch": 1.89, + "learning_rate": 1.7851950264751824e-05, + "loss": 0.0774, + "step": 14425 + }, + { + "epoch": 1.89, + "learning_rate": 1.7851621850681153e-05, + "loss": 0.0679, + "step": 14426 + }, + { + "epoch": 1.89, + "learning_rate": 1.785129341452821e-05, + "loss": 0.0724, + "step": 14427 + }, + { + "epoch": 1.89, + "learning_rate": 1.785096495629392e-05, + "loss": 0.0763, + "step": 14428 + }, + { + "epoch": 1.89, + "learning_rate": 1.7850636475979214e-05, + "loss": 0.0684, + "step": 14429 + }, + { + "epoch": 1.89, + "learning_rate": 1.785030797358501e-05, + "loss": 0.0742, + "step": 14430 + }, + { + "epoch": 1.89, + "learning_rate": 1.7849979449112233e-05, + "loss": 0.0818, + "step": 14431 + }, + { + "epoch": 1.89, + "learning_rate": 1.7849650902561804e-05, + "loss": 0.0698, + "step": 14432 + }, + { + "epoch": 1.89, + "learning_rate": 1.7849322333934654e-05, + "loss": 0.0717, + "step": 14433 + }, + { + "epoch": 1.89, + "learning_rate": 1.78489937432317e-05, + "loss": 0.0798, + "step": 14434 + }, + { + "epoch": 1.89, + "learning_rate": 1.784866513045387e-05, + "loss": 0.0778, + "step": 14435 + }, + { + "epoch": 1.89, + "learning_rate": 1.7848336495602086e-05, + "loss": 0.0781, + "step": 14436 + }, + { + "epoch": 1.89, + "learning_rate": 1.7848007838677277e-05, + "loss": 0.0787, + "step": 14437 + }, + { + "epoch": 1.89, + "learning_rate": 1.7847679159680362e-05, + "loss": 0.0658, + "step": 14438 + }, + { + "epoch": 1.89, + "learning_rate": 1.7847350458612265e-05, + "loss": 0.0749, + "step": 14439 + }, + { + "epoch": 1.89, + "learning_rate": 1.7847021735473915e-05, + "loss": 0.07, + "step": 14440 + }, + { + "epoch": 1.89, + "learning_rate": 1.7846692990266237e-05, + "loss": 0.0668, + "step": 14441 + }, + { + "epoch": 1.89, + "learning_rate": 1.784636422299015e-05, + "loss": 0.0774, + "step": 14442 + }, + { + "epoch": 1.89, + "learning_rate": 1.7846035433646582e-05, + "loss": 0.0773, + "step": 14443 + }, + { + "epoch": 1.89, + "learning_rate": 1.7845706622236455e-05, + "loss": 0.0732, + "step": 14444 + }, + { + "epoch": 1.89, + "learning_rate": 1.7845377788760696e-05, + "loss": 0.0672, + "step": 14445 + }, + { + "epoch": 1.89, + "learning_rate": 1.784504893322023e-05, + "loss": 0.0787, + "step": 14446 + }, + { + "epoch": 1.89, + "learning_rate": 1.7844720055615984e-05, + "loss": 0.0764, + "step": 14447 + }, + { + "epoch": 1.89, + "learning_rate": 1.784439115594888e-05, + "loss": 0.0715, + "step": 14448 + }, + { + "epoch": 1.89, + "learning_rate": 1.784406223421984e-05, + "loss": 0.0678, + "step": 14449 + }, + { + "epoch": 1.89, + "learning_rate": 1.7843733290429797e-05, + "loss": 0.0592, + "step": 14450 + }, + { + "epoch": 1.89, + "learning_rate": 1.784340432457967e-05, + "loss": 0.0727, + "step": 14451 + }, + { + "epoch": 1.89, + "learning_rate": 1.784307533667039e-05, + "loss": 0.0821, + "step": 14452 + }, + { + "epoch": 1.89, + "learning_rate": 1.784274632670287e-05, + "loss": 0.081, + "step": 14453 + }, + { + "epoch": 1.89, + "learning_rate": 1.784241729467805e-05, + "loss": 0.066, + "step": 14454 + }, + { + "epoch": 1.89, + "learning_rate": 1.7842088240596846e-05, + "loss": 0.0708, + "step": 14455 + }, + { + "epoch": 1.89, + "learning_rate": 1.784175916446019e-05, + "loss": 0.0803, + "step": 14456 + }, + { + "epoch": 1.89, + "learning_rate": 1.7841430066269002e-05, + "loss": 0.0692, + "step": 14457 + }, + { + "epoch": 1.89, + "learning_rate": 1.7841100946024208e-05, + "loss": 0.073, + "step": 14458 + }, + { + "epoch": 1.89, + "learning_rate": 1.784077180372674e-05, + "loss": 0.0654, + "step": 14459 + }, + { + "epoch": 1.89, + "learning_rate": 1.7840442639377512e-05, + "loss": 0.0607, + "step": 14460 + }, + { + "epoch": 1.89, + "learning_rate": 1.784011345297746e-05, + "loss": 0.08, + "step": 14461 + }, + { + "epoch": 1.89, + "learning_rate": 1.7839784244527507e-05, + "loss": 0.0757, + "step": 14462 + }, + { + "epoch": 1.89, + "learning_rate": 1.7839455014028578e-05, + "loss": 0.0735, + "step": 14463 + }, + { + "epoch": 1.89, + "learning_rate": 1.7839125761481598e-05, + "loss": 0.0677, + "step": 14464 + }, + { + "epoch": 1.89, + "learning_rate": 1.7838796486887493e-05, + "loss": 0.0705, + "step": 14465 + }, + { + "epoch": 1.89, + "learning_rate": 1.7838467190247193e-05, + "loss": 0.0746, + "step": 14466 + }, + { + "epoch": 1.9, + "learning_rate": 1.783813787156162e-05, + "loss": 0.0807, + "step": 14467 + }, + { + "epoch": 1.9, + "learning_rate": 1.7837808530831705e-05, + "loss": 0.0741, + "step": 14468 + }, + { + "epoch": 1.9, + "learning_rate": 1.7837479168058366e-05, + "loss": 0.0736, + "step": 14469 + }, + { + "epoch": 1.9, + "learning_rate": 1.7837149783242537e-05, + "loss": 0.0643, + "step": 14470 + }, + { + "epoch": 1.9, + "learning_rate": 1.783682037638514e-05, + "loss": 0.0688, + "step": 14471 + }, + { + "epoch": 1.9, + "learning_rate": 1.7836490947487105e-05, + "loss": 0.0668, + "step": 14472 + }, + { + "epoch": 1.9, + "learning_rate": 1.7836161496549356e-05, + "loss": 0.0752, + "step": 14473 + }, + { + "epoch": 1.9, + "learning_rate": 1.7835832023572822e-05, + "loss": 0.0764, + "step": 14474 + }, + { + "epoch": 1.9, + "learning_rate": 1.7835502528558423e-05, + "loss": 0.0711, + "step": 14475 + }, + { + "epoch": 1.9, + "learning_rate": 1.7835173011507093e-05, + "loss": 0.0733, + "step": 14476 + }, + { + "epoch": 1.9, + "learning_rate": 1.783484347241976e-05, + "loss": 0.0662, + "step": 14477 + }, + { + "epoch": 1.9, + "learning_rate": 1.783451391129734e-05, + "loss": 0.0804, + "step": 14478 + }, + { + "epoch": 1.9, + "learning_rate": 1.783418432814077e-05, + "loss": 0.081, + "step": 14479 + }, + { + "epoch": 1.9, + "learning_rate": 1.7833854722950975e-05, + "loss": 0.0801, + "step": 14480 + }, + { + "epoch": 1.9, + "learning_rate": 1.7833525095728878e-05, + "loss": 0.0743, + "step": 14481 + }, + { + "epoch": 1.9, + "learning_rate": 1.783319544647541e-05, + "loss": 0.0721, + "step": 14482 + }, + { + "epoch": 1.9, + "learning_rate": 1.78328657751915e-05, + "loss": 0.071, + "step": 14483 + }, + { + "epoch": 1.9, + "learning_rate": 1.7832536081878068e-05, + "loss": 0.0728, + "step": 14484 + }, + { + "epoch": 1.9, + "learning_rate": 1.7832206366536047e-05, + "loss": 0.0648, + "step": 14485 + }, + { + "epoch": 1.9, + "learning_rate": 1.783187662916636e-05, + "loss": 0.0756, + "step": 14486 + }, + { + "epoch": 1.9, + "learning_rate": 1.783154686976994e-05, + "loss": 0.076, + "step": 14487 + }, + { + "epoch": 1.9, + "learning_rate": 1.783121708834771e-05, + "loss": 0.0725, + "step": 14488 + }, + { + "epoch": 1.9, + "learning_rate": 1.7830887284900603e-05, + "loss": 0.0784, + "step": 14489 + }, + { + "epoch": 1.9, + "learning_rate": 1.783055745942954e-05, + "loss": 0.054, + "step": 14490 + }, + { + "epoch": 1.9, + "learning_rate": 1.7830227611935447e-05, + "loss": 0.0808, + "step": 14491 + }, + { + "epoch": 1.9, + "learning_rate": 1.782989774241926e-05, + "loss": 0.0719, + "step": 14492 + }, + { + "epoch": 1.9, + "learning_rate": 1.78295678508819e-05, + "loss": 0.0715, + "step": 14493 + }, + { + "epoch": 1.9, + "learning_rate": 1.78292379373243e-05, + "loss": 0.0813, + "step": 14494 + }, + { + "epoch": 1.9, + "learning_rate": 1.7828908001747388e-05, + "loss": 0.0868, + "step": 14495 + }, + { + "epoch": 1.9, + "learning_rate": 1.7828578044152084e-05, + "loss": 0.0782, + "step": 14496 + }, + { + "epoch": 1.9, + "learning_rate": 1.7828248064539324e-05, + "loss": 0.0708, + "step": 14497 + }, + { + "epoch": 1.9, + "learning_rate": 1.782791806291003e-05, + "loss": 0.0857, + "step": 14498 + }, + { + "epoch": 1.9, + "learning_rate": 1.7827588039265136e-05, + "loss": 0.0781, + "step": 14499 + }, + { + "epoch": 1.9, + "learning_rate": 1.782725799360557e-05, + "loss": 0.0754, + "step": 14500 + }, + { + "epoch": 1.9, + "learning_rate": 1.7826927925932255e-05, + "loss": 0.0755, + "step": 14501 + }, + { + "epoch": 1.9, + "learning_rate": 1.7826597836246116e-05, + "loss": 0.0793, + "step": 14502 + }, + { + "epoch": 1.9, + "learning_rate": 1.7826267724548096e-05, + "loss": 0.0642, + "step": 14503 + }, + { + "epoch": 1.9, + "learning_rate": 1.7825937590839114e-05, + "loss": 0.076, + "step": 14504 + }, + { + "epoch": 1.9, + "learning_rate": 1.7825607435120097e-05, + "loss": 0.0756, + "step": 14505 + }, + { + "epoch": 1.9, + "learning_rate": 1.782527725739197e-05, + "loss": 0.0612, + "step": 14506 + }, + { + "epoch": 1.9, + "learning_rate": 1.7824947057655677e-05, + "loss": 0.069, + "step": 14507 + }, + { + "epoch": 1.9, + "learning_rate": 1.7824616835912136e-05, + "loss": 0.0755, + "step": 14508 + }, + { + "epoch": 1.9, + "learning_rate": 1.782428659216227e-05, + "loss": 0.0717, + "step": 14509 + }, + { + "epoch": 1.9, + "learning_rate": 1.782395632640702e-05, + "loss": 0.0726, + "step": 14510 + }, + { + "epoch": 1.9, + "learning_rate": 1.782362603864731e-05, + "loss": 0.0778, + "step": 14511 + }, + { + "epoch": 1.9, + "learning_rate": 1.7823295728884064e-05, + "loss": 0.062, + "step": 14512 + }, + { + "epoch": 1.9, + "learning_rate": 1.782296539711822e-05, + "loss": 0.0822, + "step": 14513 + }, + { + "epoch": 1.9, + "learning_rate": 1.78226350433507e-05, + "loss": 0.0676, + "step": 14514 + }, + { + "epoch": 1.9, + "learning_rate": 1.782230466758244e-05, + "loss": 0.0838, + "step": 14515 + }, + { + "epoch": 1.9, + "learning_rate": 1.7821974269814358e-05, + "loss": 0.0707, + "step": 14516 + }, + { + "epoch": 1.9, + "learning_rate": 1.7821643850047392e-05, + "loss": 0.0722, + "step": 14517 + }, + { + "epoch": 1.9, + "learning_rate": 1.782131340828247e-05, + "loss": 0.0815, + "step": 14518 + }, + { + "epoch": 1.9, + "learning_rate": 1.7820982944520522e-05, + "loss": 0.0706, + "step": 14519 + }, + { + "epoch": 1.9, + "learning_rate": 1.7820652458762472e-05, + "loss": 0.0805, + "step": 14520 + }, + { + "epoch": 1.9, + "learning_rate": 1.7820321951009257e-05, + "loss": 0.0612, + "step": 14521 + }, + { + "epoch": 1.9, + "learning_rate": 1.78199914212618e-05, + "loss": 0.0746, + "step": 14522 + }, + { + "epoch": 1.9, + "learning_rate": 1.7819660869521037e-05, + "loss": 0.0743, + "step": 14523 + }, + { + "epoch": 1.9, + "learning_rate": 1.7819330295787892e-05, + "loss": 0.0716, + "step": 14524 + }, + { + "epoch": 1.9, + "learning_rate": 1.78189997000633e-05, + "loss": 0.0714, + "step": 14525 + }, + { + "epoch": 1.9, + "learning_rate": 1.7818669082348182e-05, + "loss": 0.0769, + "step": 14526 + }, + { + "epoch": 1.9, + "learning_rate": 1.7818338442643476e-05, + "loss": 0.0642, + "step": 14527 + }, + { + "epoch": 1.9, + "learning_rate": 1.7818007780950114e-05, + "loss": 0.0666, + "step": 14528 + }, + { + "epoch": 1.9, + "learning_rate": 1.7817677097269016e-05, + "loss": 0.058, + "step": 14529 + }, + { + "epoch": 1.9, + "learning_rate": 1.781734639160112e-05, + "loss": 0.0686, + "step": 14530 + }, + { + "epoch": 1.9, + "learning_rate": 1.7817015663947352e-05, + "loss": 0.0622, + "step": 14531 + }, + { + "epoch": 1.9, + "learning_rate": 1.7816684914308645e-05, + "loss": 0.0805, + "step": 14532 + }, + { + "epoch": 1.9, + "learning_rate": 1.7816354142685927e-05, + "loss": 0.0649, + "step": 14533 + }, + { + "epoch": 1.9, + "learning_rate": 1.781602334908013e-05, + "loss": 0.065, + "step": 14534 + }, + { + "epoch": 1.9, + "learning_rate": 1.7815692533492183e-05, + "loss": 0.0658, + "step": 14535 + }, + { + "epoch": 1.9, + "learning_rate": 1.7815361695923017e-05, + "loss": 0.0759, + "step": 14536 + }, + { + "epoch": 1.9, + "learning_rate": 1.7815030836373563e-05, + "loss": 0.0681, + "step": 14537 + }, + { + "epoch": 1.9, + "learning_rate": 1.781469995484475e-05, + "loss": 0.074, + "step": 14538 + }, + { + "epoch": 1.9, + "learning_rate": 1.781436905133751e-05, + "loss": 0.0718, + "step": 14539 + }, + { + "epoch": 1.9, + "learning_rate": 1.7814038125852775e-05, + "loss": 0.0781, + "step": 14540 + }, + { + "epoch": 1.9, + "learning_rate": 1.7813707178391472e-05, + "loss": 0.084, + "step": 14541 + }, + { + "epoch": 1.9, + "learning_rate": 1.7813376208954533e-05, + "loss": 0.072, + "step": 14542 + }, + { + "epoch": 1.91, + "learning_rate": 1.781304521754289e-05, + "loss": 0.0752, + "step": 14543 + }, + { + "epoch": 1.91, + "learning_rate": 1.7812714204157476e-05, + "loss": 0.081, + "step": 14544 + }, + { + "epoch": 1.91, + "learning_rate": 1.7812383168799215e-05, + "loss": 0.0751, + "step": 14545 + }, + { + "epoch": 1.91, + "learning_rate": 1.7812052111469045e-05, + "loss": 0.0721, + "step": 14546 + }, + { + "epoch": 1.91, + "learning_rate": 1.7811721032167894e-05, + "loss": 0.072, + "step": 14547 + }, + { + "epoch": 1.91, + "learning_rate": 1.7811389930896694e-05, + "loss": 0.0764, + "step": 14548 + }, + { + "epoch": 1.91, + "learning_rate": 1.7811058807656376e-05, + "loss": 0.0719, + "step": 14549 + }, + { + "epoch": 1.91, + "learning_rate": 1.781072766244787e-05, + "loss": 0.0671, + "step": 14550 + }, + { + "epoch": 1.91, + "learning_rate": 1.7810396495272108e-05, + "loss": 0.0703, + "step": 14551 + }, + { + "epoch": 1.91, + "learning_rate": 1.781006530613002e-05, + "loss": 0.0799, + "step": 14552 + }, + { + "epoch": 1.91, + "learning_rate": 1.7809734095022544e-05, + "loss": 0.0716, + "step": 14553 + }, + { + "epoch": 1.91, + "learning_rate": 1.7809402861950605e-05, + "loss": 0.0717, + "step": 14554 + }, + { + "epoch": 1.91, + "learning_rate": 1.7809071606915135e-05, + "loss": 0.069, + "step": 14555 + }, + { + "epoch": 1.91, + "learning_rate": 1.780874032991707e-05, + "loss": 0.0734, + "step": 14556 + }, + { + "epoch": 1.91, + "learning_rate": 1.7808409030957335e-05, + "loss": 0.0809, + "step": 14557 + }, + { + "epoch": 1.91, + "learning_rate": 1.7808077710036868e-05, + "loss": 0.0676, + "step": 14558 + }, + { + "epoch": 1.91, + "learning_rate": 1.7807746367156597e-05, + "loss": 0.0798, + "step": 14559 + }, + { + "epoch": 1.91, + "learning_rate": 1.7807415002317454e-05, + "loss": 0.0819, + "step": 14560 + }, + { + "epoch": 1.91, + "learning_rate": 1.7807083615520375e-05, + "loss": 0.069, + "step": 14561 + }, + { + "epoch": 1.91, + "learning_rate": 1.7806752206766288e-05, + "loss": 0.0662, + "step": 14562 + }, + { + "epoch": 1.91, + "learning_rate": 1.7806420776056127e-05, + "loss": 0.0782, + "step": 14563 + }, + { + "epoch": 1.91, + "learning_rate": 1.7806089323390822e-05, + "loss": 0.0651, + "step": 14564 + }, + { + "epoch": 1.91, + "learning_rate": 1.7805757848771303e-05, + "loss": 0.0702, + "step": 14565 + }, + { + "epoch": 1.91, + "learning_rate": 1.780542635219851e-05, + "loss": 0.0675, + "step": 14566 + }, + { + "epoch": 1.91, + "learning_rate": 1.780509483367337e-05, + "loss": 0.0795, + "step": 14567 + }, + { + "epoch": 1.91, + "learning_rate": 1.7804763293196817e-05, + "loss": 0.0796, + "step": 14568 + }, + { + "epoch": 1.91, + "learning_rate": 1.780443173076978e-05, + "loss": 0.0763, + "step": 14569 + }, + { + "epoch": 1.91, + "learning_rate": 1.78041001463932e-05, + "loss": 0.078, + "step": 14570 + }, + { + "epoch": 1.91, + "learning_rate": 1.7803768540068e-05, + "loss": 0.0846, + "step": 14571 + }, + { + "epoch": 1.91, + "learning_rate": 1.780343691179512e-05, + "loss": 0.0706, + "step": 14572 + }, + { + "epoch": 1.91, + "learning_rate": 1.7803105261575483e-05, + "loss": 0.0716, + "step": 14573 + }, + { + "epoch": 1.91, + "learning_rate": 1.7802773589410032e-05, + "loss": 0.0722, + "step": 14574 + }, + { + "epoch": 1.91, + "learning_rate": 1.7802441895299696e-05, + "loss": 0.0733, + "step": 14575 + }, + { + "epoch": 1.91, + "learning_rate": 1.7802110179245407e-05, + "loss": 0.0796, + "step": 14576 + }, + { + "epoch": 1.91, + "learning_rate": 1.7801778441248098e-05, + "loss": 0.0637, + "step": 14577 + }, + { + "epoch": 1.91, + "learning_rate": 1.7801446681308703e-05, + "loss": 0.0822, + "step": 14578 + }, + { + "epoch": 1.91, + "learning_rate": 1.7801114899428155e-05, + "loss": 0.077, + "step": 14579 + }, + { + "epoch": 1.91, + "learning_rate": 1.7800783095607385e-05, + "loss": 0.0598, + "step": 14580 + }, + { + "epoch": 1.91, + "learning_rate": 1.7800451269847332e-05, + "loss": 0.0722, + "step": 14581 + }, + { + "epoch": 1.91, + "learning_rate": 1.7800119422148923e-05, + "loss": 0.0769, + "step": 14582 + }, + { + "epoch": 1.91, + "learning_rate": 1.7799787552513096e-05, + "loss": 0.0574, + "step": 14583 + }, + { + "epoch": 1.91, + "learning_rate": 1.7799455660940776e-05, + "loss": 0.0724, + "step": 14584 + }, + { + "epoch": 1.91, + "learning_rate": 1.7799123747432908e-05, + "loss": 0.0738, + "step": 14585 + }, + { + "epoch": 1.91, + "learning_rate": 1.7798791811990413e-05, + "loss": 0.0786, + "step": 14586 + }, + { + "epoch": 1.91, + "learning_rate": 1.7798459854614235e-05, + "loss": 0.0756, + "step": 14587 + }, + { + "epoch": 1.91, + "learning_rate": 1.779812787530531e-05, + "loss": 0.0678, + "step": 14588 + }, + { + "epoch": 1.91, + "learning_rate": 1.7797795874064558e-05, + "loss": 0.0792, + "step": 14589 + }, + { + "epoch": 1.91, + "learning_rate": 1.7797463850892923e-05, + "loss": 0.0772, + "step": 14590 + }, + { + "epoch": 1.91, + "learning_rate": 1.7797131805791337e-05, + "loss": 0.0679, + "step": 14591 + }, + { + "epoch": 1.91, + "learning_rate": 1.779679973876073e-05, + "loss": 0.065, + "step": 14592 + }, + { + "epoch": 1.91, + "learning_rate": 1.779646764980204e-05, + "loss": 0.0698, + "step": 14593 + }, + { + "epoch": 1.91, + "learning_rate": 1.77961355389162e-05, + "loss": 0.0718, + "step": 14594 + }, + { + "epoch": 1.91, + "learning_rate": 1.7795803406104144e-05, + "loss": 0.0722, + "step": 14595 + }, + { + "epoch": 1.91, + "learning_rate": 1.779547125136681e-05, + "loss": 0.0833, + "step": 14596 + }, + { + "epoch": 1.91, + "learning_rate": 1.7795139074705123e-05, + "loss": 0.0679, + "step": 14597 + }, + { + "epoch": 1.91, + "learning_rate": 1.7794806876120026e-05, + "loss": 0.0755, + "step": 14598 + }, + { + "epoch": 1.91, + "learning_rate": 1.7794474655612445e-05, + "loss": 0.075, + "step": 14599 + }, + { + "epoch": 1.91, + "learning_rate": 1.7794142413183323e-05, + "loss": 0.0784, + "step": 14600 + }, + { + "epoch": 1.91, + "learning_rate": 1.779381014883359e-05, + "loss": 0.0699, + "step": 14601 + }, + { + "epoch": 1.91, + "learning_rate": 1.7793477862564183e-05, + "loss": 0.0695, + "step": 14602 + }, + { + "epoch": 1.91, + "learning_rate": 1.779314555437603e-05, + "loss": 0.0741, + "step": 14603 + }, + { + "epoch": 1.91, + "learning_rate": 1.7792813224270075e-05, + "loss": 0.0807, + "step": 14604 + }, + { + "epoch": 1.91, + "learning_rate": 1.7792480872247243e-05, + "loss": 0.0721, + "step": 14605 + }, + { + "epoch": 1.91, + "learning_rate": 1.7792148498308477e-05, + "loss": 0.0826, + "step": 14606 + }, + { + "epoch": 1.91, + "learning_rate": 1.7791816102454705e-05, + "loss": 0.0686, + "step": 14607 + }, + { + "epoch": 1.91, + "learning_rate": 1.7791483684686866e-05, + "loss": 0.07, + "step": 14608 + }, + { + "epoch": 1.91, + "learning_rate": 1.77911512450059e-05, + "loss": 0.0692, + "step": 14609 + }, + { + "epoch": 1.91, + "learning_rate": 1.7790818783412728e-05, + "loss": 0.0749, + "step": 14610 + }, + { + "epoch": 1.91, + "learning_rate": 1.77904862999083e-05, + "loss": 0.0744, + "step": 14611 + }, + { + "epoch": 1.91, + "learning_rate": 1.7790153794493536e-05, + "loss": 0.0802, + "step": 14612 + }, + { + "epoch": 1.91, + "learning_rate": 1.7789821267169385e-05, + "loss": 0.0824, + "step": 14613 + }, + { + "epoch": 1.91, + "learning_rate": 1.7789488717936775e-05, + "loss": 0.0701, + "step": 14614 + }, + { + "epoch": 1.91, + "learning_rate": 1.7789156146796643e-05, + "loss": 0.0696, + "step": 14615 + }, + { + "epoch": 1.91, + "learning_rate": 1.7788823553749925e-05, + "loss": 0.0836, + "step": 14616 + }, + { + "epoch": 1.91, + "learning_rate": 1.7788490938797553e-05, + "loss": 0.0687, + "step": 14617 + }, + { + "epoch": 1.91, + "learning_rate": 1.7788158301940468e-05, + "loss": 0.0722, + "step": 14618 + }, + { + "epoch": 1.91, + "learning_rate": 1.7787825643179602e-05, + "loss": 0.0823, + "step": 14619 + }, + { + "epoch": 1.92, + "learning_rate": 1.778749296251589e-05, + "loss": 0.0716, + "step": 14620 + }, + { + "epoch": 1.92, + "learning_rate": 1.778716025995027e-05, + "loss": 0.0687, + "step": 14621 + }, + { + "epoch": 1.92, + "learning_rate": 1.7786827535483676e-05, + "loss": 0.0801, + "step": 14622 + }, + { + "epoch": 1.92, + "learning_rate": 1.7786494789117043e-05, + "loss": 0.0674, + "step": 14623 + }, + { + "epoch": 1.92, + "learning_rate": 1.778616202085131e-05, + "loss": 0.0695, + "step": 14624 + }, + { + "epoch": 1.92, + "learning_rate": 1.7785829230687408e-05, + "loss": 0.0743, + "step": 14625 + }, + { + "epoch": 1.92, + "learning_rate": 1.778549641862628e-05, + "loss": 0.0815, + "step": 14626 + }, + { + "epoch": 1.92, + "learning_rate": 1.7785163584668855e-05, + "loss": 0.0753, + "step": 14627 + }, + { + "epoch": 1.92, + "learning_rate": 1.778483072881607e-05, + "loss": 0.0849, + "step": 14628 + }, + { + "epoch": 1.92, + "learning_rate": 1.778449785106887e-05, + "loss": 0.072, + "step": 14629 + }, + { + "epoch": 1.92, + "learning_rate": 1.778416495142818e-05, + "loss": 0.0862, + "step": 14630 + }, + { + "epoch": 1.92, + "learning_rate": 1.7783832029894942e-05, + "loss": 0.07, + "step": 14631 + }, + { + "epoch": 1.92, + "learning_rate": 1.7783499086470088e-05, + "loss": 0.0748, + "step": 14632 + }, + { + "epoch": 1.92, + "learning_rate": 1.7783166121154558e-05, + "loss": 0.0774, + "step": 14633 + }, + { + "epoch": 1.92, + "learning_rate": 1.7782833133949288e-05, + "loss": 0.0709, + "step": 14634 + }, + { + "epoch": 1.92, + "learning_rate": 1.7782500124855216e-05, + "loss": 0.0846, + "step": 14635 + }, + { + "epoch": 1.92, + "learning_rate": 1.7782167093873277e-05, + "loss": 0.0705, + "step": 14636 + }, + { + "epoch": 1.92, + "learning_rate": 1.7781834041004404e-05, + "loss": 0.0679, + "step": 14637 + }, + { + "epoch": 1.92, + "learning_rate": 1.778150096624954e-05, + "loss": 0.07, + "step": 14638 + }, + { + "epoch": 1.92, + "learning_rate": 1.778116786960962e-05, + "loss": 0.068, + "step": 14639 + }, + { + "epoch": 1.92, + "learning_rate": 1.7780834751085575e-05, + "loss": 0.0815, + "step": 14640 + }, + { + "epoch": 1.92, + "learning_rate": 1.778050161067835e-05, + "loss": 0.0822, + "step": 14641 + }, + { + "epoch": 1.92, + "learning_rate": 1.778016844838888e-05, + "loss": 0.0702, + "step": 14642 + }, + { + "epoch": 1.92, + "learning_rate": 1.77798352642181e-05, + "loss": 0.0637, + "step": 14643 + }, + { + "epoch": 1.92, + "learning_rate": 1.7779502058166945e-05, + "loss": 0.0872, + "step": 14644 + }, + { + "epoch": 1.92, + "learning_rate": 1.777916883023636e-05, + "loss": 0.0653, + "step": 14645 + }, + { + "epoch": 1.92, + "learning_rate": 1.777883558042727e-05, + "loss": 0.0698, + "step": 14646 + }, + { + "epoch": 1.92, + "learning_rate": 1.777850230874062e-05, + "loss": 0.0739, + "step": 14647 + }, + { + "epoch": 1.92, + "learning_rate": 1.7778169015177354e-05, + "loss": 0.0632, + "step": 14648 + }, + { + "epoch": 1.92, + "learning_rate": 1.7777835699738394e-05, + "loss": 0.0626, + "step": 14649 + }, + { + "epoch": 1.92, + "learning_rate": 1.777750236242469e-05, + "loss": 0.0723, + "step": 14650 + }, + { + "epoch": 1.92, + "learning_rate": 1.7777169003237172e-05, + "loss": 0.0663, + "step": 14651 + }, + { + "epoch": 1.92, + "learning_rate": 1.777683562217678e-05, + "loss": 0.0685, + "step": 14652 + }, + { + "epoch": 1.92, + "learning_rate": 1.7776502219244453e-05, + "loss": 0.0776, + "step": 14653 + }, + { + "epoch": 1.92, + "learning_rate": 1.777616879444113e-05, + "loss": 0.0658, + "step": 14654 + }, + { + "epoch": 1.92, + "learning_rate": 1.7775835347767743e-05, + "loss": 0.0772, + "step": 14655 + }, + { + "epoch": 1.92, + "learning_rate": 1.7775501879225233e-05, + "loss": 0.0717, + "step": 14656 + }, + { + "epoch": 1.92, + "learning_rate": 1.777516838881454e-05, + "loss": 0.0777, + "step": 14657 + }, + { + "epoch": 1.92, + "learning_rate": 1.77748348765366e-05, + "loss": 0.0801, + "step": 14658 + }, + { + "epoch": 1.92, + "learning_rate": 1.777450134239235e-05, + "loss": 0.0701, + "step": 14659 + }, + { + "epoch": 1.92, + "learning_rate": 1.7774167786382728e-05, + "loss": 0.0775, + "step": 14660 + }, + { + "epoch": 1.92, + "learning_rate": 1.7773834208508675e-05, + "loss": 0.0683, + "step": 14661 + }, + { + "epoch": 1.92, + "learning_rate": 1.7773500608771126e-05, + "loss": 0.0618, + "step": 14662 + }, + { + "epoch": 1.92, + "learning_rate": 1.777316698717102e-05, + "loss": 0.0649, + "step": 14663 + }, + { + "epoch": 1.92, + "learning_rate": 1.7772833343709298e-05, + "loss": 0.072, + "step": 14664 + }, + { + "epoch": 1.92, + "learning_rate": 1.7772499678386895e-05, + "loss": 0.0716, + "step": 14665 + }, + { + "epoch": 1.92, + "learning_rate": 1.777216599120475e-05, + "loss": 0.0785, + "step": 14666 + }, + { + "epoch": 1.92, + "learning_rate": 1.7771832282163804e-05, + "loss": 0.0722, + "step": 14667 + }, + { + "epoch": 1.92, + "learning_rate": 1.7771498551264992e-05, + "loss": 0.0758, + "step": 14668 + }, + { + "epoch": 1.92, + "learning_rate": 1.7771164798509254e-05, + "loss": 0.073, + "step": 14669 + }, + { + "epoch": 1.92, + "learning_rate": 1.777083102389753e-05, + "loss": 0.0727, + "step": 14670 + }, + { + "epoch": 1.92, + "learning_rate": 1.7770497227430756e-05, + "loss": 0.0649, + "step": 14671 + }, + { + "epoch": 1.92, + "learning_rate": 1.7770163409109873e-05, + "loss": 0.075, + "step": 14672 + }, + { + "epoch": 1.92, + "learning_rate": 1.776982956893582e-05, + "loss": 0.0749, + "step": 14673 + }, + { + "epoch": 1.92, + "learning_rate": 1.7769495706909533e-05, + "loss": 0.08, + "step": 14674 + }, + { + "epoch": 1.92, + "learning_rate": 1.7769161823031956e-05, + "loss": 0.0742, + "step": 14675 + }, + { + "epoch": 1.92, + "learning_rate": 1.7768827917304023e-05, + "loss": 0.0756, + "step": 14676 + }, + { + "epoch": 1.92, + "learning_rate": 1.7768493989726678e-05, + "loss": 0.0782, + "step": 14677 + }, + { + "epoch": 1.92, + "learning_rate": 1.7768160040300856e-05, + "loss": 0.0762, + "step": 14678 + }, + { + "epoch": 1.92, + "learning_rate": 1.7767826069027494e-05, + "loss": 0.0693, + "step": 14679 + }, + { + "epoch": 1.92, + "learning_rate": 1.776749207590754e-05, + "loss": 0.072, + "step": 14680 + }, + { + "epoch": 1.92, + "learning_rate": 1.7767158060941923e-05, + "loss": 0.0736, + "step": 14681 + }, + { + "epoch": 1.92, + "learning_rate": 1.776682402413159e-05, + "loss": 0.0713, + "step": 14682 + }, + { + "epoch": 1.92, + "learning_rate": 1.7766489965477478e-05, + "loss": 0.0759, + "step": 14683 + }, + { + "epoch": 1.92, + "learning_rate": 1.7766155884980528e-05, + "loss": 0.0844, + "step": 14684 + }, + { + "epoch": 1.92, + "learning_rate": 1.7765821782641678e-05, + "loss": 0.0706, + "step": 14685 + }, + { + "epoch": 1.92, + "learning_rate": 1.7765487658461867e-05, + "loss": 0.08, + "step": 14686 + }, + { + "epoch": 1.92, + "learning_rate": 1.7765153512442037e-05, + "loss": 0.068, + "step": 14687 + }, + { + "epoch": 1.92, + "learning_rate": 1.776481934458312e-05, + "loss": 0.0735, + "step": 14688 + }, + { + "epoch": 1.92, + "learning_rate": 1.7764485154886068e-05, + "loss": 0.0836, + "step": 14689 + }, + { + "epoch": 1.92, + "learning_rate": 1.7764150943351817e-05, + "loss": 0.0735, + "step": 14690 + }, + { + "epoch": 1.92, + "learning_rate": 1.77638167099813e-05, + "loss": 0.0734, + "step": 14691 + }, + { + "epoch": 1.92, + "learning_rate": 1.7763482454775463e-05, + "loss": 0.0775, + "step": 14692 + }, + { + "epoch": 1.92, + "learning_rate": 1.7763148177735248e-05, + "loss": 0.0744, + "step": 14693 + }, + { + "epoch": 1.92, + "learning_rate": 1.776281387886159e-05, + "loss": 0.0621, + "step": 14694 + }, + { + "epoch": 1.92, + "learning_rate": 1.776247955815543e-05, + "loss": 0.0697, + "step": 14695 + }, + { + "epoch": 1.93, + "learning_rate": 1.776214521561771e-05, + "loss": 0.0868, + "step": 14696 + }, + { + "epoch": 1.93, + "learning_rate": 1.7761810851249373e-05, + "loss": 0.0749, + "step": 14697 + }, + { + "epoch": 1.93, + "learning_rate": 1.776147646505135e-05, + "loss": 0.0628, + "step": 14698 + }, + { + "epoch": 1.93, + "learning_rate": 1.7761142057024592e-05, + "loss": 0.0735, + "step": 14699 + }, + { + "epoch": 1.93, + "learning_rate": 1.776080762717004e-05, + "loss": 0.0692, + "step": 14700 + }, + { + "epoch": 1.93, + "learning_rate": 1.776047317548862e-05, + "loss": 0.0867, + "step": 14701 + }, + { + "epoch": 1.93, + "learning_rate": 1.776013870198129e-05, + "loss": 0.0828, + "step": 14702 + }, + { + "epoch": 1.93, + "learning_rate": 1.7759804206648978e-05, + "loss": 0.0776, + "step": 14703 + }, + { + "epoch": 1.93, + "learning_rate": 1.7759469689492636e-05, + "loss": 0.0743, + "step": 14704 + }, + { + "epoch": 1.93, + "learning_rate": 1.775913515051319e-05, + "loss": 0.078, + "step": 14705 + }, + { + "epoch": 1.93, + "learning_rate": 1.77588005897116e-05, + "loss": 0.0698, + "step": 14706 + }, + { + "epoch": 1.93, + "learning_rate": 1.7758466007088788e-05, + "loss": 0.0717, + "step": 14707 + }, + { + "epoch": 1.93, + "learning_rate": 1.7758131402645707e-05, + "loss": 0.0651, + "step": 14708 + }, + { + "epoch": 1.93, + "learning_rate": 1.7757796776383295e-05, + "loss": 0.0756, + "step": 14709 + }, + { + "epoch": 1.93, + "learning_rate": 1.7757462128302493e-05, + "loss": 0.0645, + "step": 14710 + }, + { + "epoch": 1.93, + "learning_rate": 1.7757127458404243e-05, + "loss": 0.0669, + "step": 14711 + }, + { + "epoch": 1.93, + "learning_rate": 1.775679276668948e-05, + "loss": 0.0802, + "step": 14712 + }, + { + "epoch": 1.93, + "learning_rate": 1.7756458053159154e-05, + "loss": 0.0783, + "step": 14713 + }, + { + "epoch": 1.93, + "learning_rate": 1.7756123317814203e-05, + "loss": 0.07, + "step": 14714 + }, + { + "epoch": 1.93, + "learning_rate": 1.7755788560655567e-05, + "loss": 0.0732, + "step": 14715 + }, + { + "epoch": 1.93, + "learning_rate": 1.7755453781684187e-05, + "loss": 0.08, + "step": 14716 + }, + { + "epoch": 1.93, + "learning_rate": 1.775511898090101e-05, + "loss": 0.0795, + "step": 14717 + }, + { + "epoch": 1.93, + "learning_rate": 1.7754784158306973e-05, + "loss": 0.0747, + "step": 14718 + }, + { + "epoch": 1.93, + "learning_rate": 1.775444931390302e-05, + "loss": 0.0681, + "step": 14719 + }, + { + "epoch": 1.93, + "learning_rate": 1.7754114447690084e-05, + "loss": 0.0681, + "step": 14720 + }, + { + "epoch": 1.93, + "learning_rate": 1.775377955966912e-05, + "loss": 0.0828, + "step": 14721 + }, + { + "epoch": 1.93, + "learning_rate": 1.7753444649841064e-05, + "loss": 0.0729, + "step": 14722 + }, + { + "epoch": 1.93, + "learning_rate": 1.7753109718206855e-05, + "loss": 0.0767, + "step": 14723 + }, + { + "epoch": 1.93, + "learning_rate": 1.775277476476744e-05, + "loss": 0.0736, + "step": 14724 + }, + { + "epoch": 1.93, + "learning_rate": 1.775243978952376e-05, + "loss": 0.0709, + "step": 14725 + }, + { + "epoch": 1.93, + "learning_rate": 1.7752104792476754e-05, + "loss": 0.0789, + "step": 14726 + }, + { + "epoch": 1.93, + "learning_rate": 1.7751769773627368e-05, + "loss": 0.0624, + "step": 14727 + }, + { + "epoch": 1.93, + "learning_rate": 1.775143473297654e-05, + "loss": 0.0753, + "step": 14728 + }, + { + "epoch": 1.93, + "learning_rate": 1.7751099670525214e-05, + "loss": 0.0686, + "step": 14729 + }, + { + "epoch": 1.93, + "learning_rate": 1.7750764586274333e-05, + "loss": 0.0804, + "step": 14730 + }, + { + "epoch": 1.93, + "learning_rate": 1.7750429480224843e-05, + "loss": 0.0806, + "step": 14731 + }, + { + "epoch": 1.93, + "learning_rate": 1.775009435237768e-05, + "loss": 0.0787, + "step": 14732 + }, + { + "epoch": 1.93, + "learning_rate": 1.774975920273379e-05, + "loss": 0.08, + "step": 14733 + }, + { + "epoch": 1.93, + "learning_rate": 1.7749424031294116e-05, + "loss": 0.0726, + "step": 14734 + }, + { + "epoch": 1.93, + "learning_rate": 1.7749088838059595e-05, + "loss": 0.0703, + "step": 14735 + }, + { + "epoch": 1.93, + "learning_rate": 1.774875362303118e-05, + "loss": 0.0807, + "step": 14736 + }, + { + "epoch": 1.93, + "learning_rate": 1.7748418386209806e-05, + "loss": 0.0671, + "step": 14737 + }, + { + "epoch": 1.93, + "learning_rate": 1.7748083127596417e-05, + "loss": 0.0616, + "step": 14738 + }, + { + "epoch": 1.93, + "learning_rate": 1.7747747847191957e-05, + "loss": 0.0643, + "step": 14739 + }, + { + "epoch": 1.93, + "learning_rate": 1.774741254499737e-05, + "loss": 0.0774, + "step": 14740 + }, + { + "epoch": 1.93, + "learning_rate": 1.7747077221013595e-05, + "loss": 0.079, + "step": 14741 + }, + { + "epoch": 1.93, + "learning_rate": 1.7746741875241583e-05, + "loss": 0.0638, + "step": 14742 + }, + { + "epoch": 1.93, + "learning_rate": 1.7746406507682268e-05, + "loss": 0.0656, + "step": 14743 + }, + { + "epoch": 1.93, + "learning_rate": 1.7746071118336598e-05, + "loss": 0.0671, + "step": 14744 + }, + { + "epoch": 1.93, + "learning_rate": 1.7745735707205515e-05, + "loss": 0.0718, + "step": 14745 + }, + { + "epoch": 1.93, + "learning_rate": 1.7745400274289964e-05, + "loss": 0.0829, + "step": 14746 + }, + { + "epoch": 1.93, + "learning_rate": 1.7745064819590884e-05, + "loss": 0.0809, + "step": 14747 + }, + { + "epoch": 1.93, + "learning_rate": 1.7744729343109227e-05, + "loss": 0.0739, + "step": 14748 + }, + { + "epoch": 1.93, + "learning_rate": 1.774439384484593e-05, + "loss": 0.0801, + "step": 14749 + }, + { + "epoch": 1.93, + "learning_rate": 1.7744058324801933e-05, + "loss": 0.0638, + "step": 14750 + }, + { + "epoch": 1.93, + "learning_rate": 1.774372278297819e-05, + "loss": 0.0656, + "step": 14751 + }, + { + "epoch": 1.93, + "learning_rate": 1.7743387219375633e-05, + "loss": 0.0721, + "step": 14752 + }, + { + "epoch": 1.93, + "learning_rate": 1.7743051633995215e-05, + "loss": 0.0654, + "step": 14753 + }, + { + "epoch": 1.93, + "learning_rate": 1.7742716026837882e-05, + "loss": 0.0716, + "step": 14754 + }, + { + "epoch": 1.93, + "learning_rate": 1.7742380397904562e-05, + "loss": 0.0691, + "step": 14755 + }, + { + "epoch": 1.93, + "learning_rate": 1.7742044747196217e-05, + "loss": 0.0654, + "step": 14756 + }, + { + "epoch": 1.93, + "learning_rate": 1.7741709074713782e-05, + "loss": 0.0693, + "step": 14757 + }, + { + "epoch": 1.93, + "learning_rate": 1.7741373380458203e-05, + "loss": 0.0756, + "step": 14758 + }, + { + "epoch": 1.93, + "learning_rate": 1.7741037664430424e-05, + "loss": 0.0746, + "step": 14759 + }, + { + "epoch": 1.93, + "learning_rate": 1.7740701926631385e-05, + "loss": 0.083, + "step": 14760 + }, + { + "epoch": 1.93, + "learning_rate": 1.7740366167062037e-05, + "loss": 0.0775, + "step": 14761 + }, + { + "epoch": 1.93, + "learning_rate": 1.7740030385723322e-05, + "loss": 0.0665, + "step": 14762 + }, + { + "epoch": 1.93, + "learning_rate": 1.773969458261618e-05, + "loss": 0.0676, + "step": 14763 + }, + { + "epoch": 1.93, + "learning_rate": 1.7739358757741563e-05, + "loss": 0.0709, + "step": 14764 + }, + { + "epoch": 1.93, + "learning_rate": 1.7739022911100406e-05, + "loss": 0.0802, + "step": 14765 + }, + { + "epoch": 1.93, + "learning_rate": 1.7738687042693667e-05, + "loss": 0.0793, + "step": 14766 + }, + { + "epoch": 1.93, + "learning_rate": 1.773835115252228e-05, + "loss": 0.0859, + "step": 14767 + }, + { + "epoch": 1.93, + "learning_rate": 1.773801524058719e-05, + "loss": 0.077, + "step": 14768 + }, + { + "epoch": 1.93, + "learning_rate": 1.7737679306889344e-05, + "loss": 0.0691, + "step": 14769 + }, + { + "epoch": 1.93, + "learning_rate": 1.7737343351429692e-05, + "loss": 0.0715, + "step": 14770 + }, + { + "epoch": 1.93, + "learning_rate": 1.773700737420917e-05, + "loss": 0.0658, + "step": 14771 + }, + { + "epoch": 1.94, + "learning_rate": 1.7736671375228728e-05, + "loss": 0.0736, + "step": 14772 + }, + { + "epoch": 1.94, + "learning_rate": 1.773633535448931e-05, + "loss": 0.0657, + "step": 14773 + }, + { + "epoch": 1.94, + "learning_rate": 1.773599931199186e-05, + "loss": 0.0673, + "step": 14774 + }, + { + "epoch": 1.94, + "learning_rate": 1.7735663247737322e-05, + "loss": 0.0787, + "step": 14775 + }, + { + "epoch": 1.94, + "learning_rate": 1.7735327161726647e-05, + "loss": 0.0661, + "step": 14776 + }, + { + "epoch": 1.94, + "learning_rate": 1.7734991053960773e-05, + "loss": 0.0844, + "step": 14777 + }, + { + "epoch": 1.94, + "learning_rate": 1.773465492444065e-05, + "loss": 0.0774, + "step": 14778 + }, + { + "epoch": 1.94, + "learning_rate": 1.7734318773167222e-05, + "loss": 0.074, + "step": 14779 + }, + { + "epoch": 1.94, + "learning_rate": 1.7733982600141435e-05, + "loss": 0.076, + "step": 14780 + }, + { + "epoch": 1.94, + "learning_rate": 1.773364640536423e-05, + "loss": 0.0676, + "step": 14781 + }, + { + "epoch": 1.94, + "learning_rate": 1.7733310188836562e-05, + "loss": 0.0686, + "step": 14782 + }, + { + "epoch": 1.94, + "learning_rate": 1.7732973950559368e-05, + "loss": 0.0605, + "step": 14783 + }, + { + "epoch": 1.94, + "learning_rate": 1.7732637690533596e-05, + "loss": 0.0804, + "step": 14784 + }, + { + "epoch": 1.94, + "learning_rate": 1.7732301408760193e-05, + "loss": 0.078, + "step": 14785 + }, + { + "epoch": 1.94, + "learning_rate": 1.7731965105240103e-05, + "loss": 0.0765, + "step": 14786 + }, + { + "epoch": 1.94, + "learning_rate": 1.7731628779974275e-05, + "loss": 0.0767, + "step": 14787 + }, + { + "epoch": 1.94, + "learning_rate": 1.773129243296365e-05, + "loss": 0.0829, + "step": 14788 + }, + { + "epoch": 1.94, + "learning_rate": 1.773095606420918e-05, + "loss": 0.0619, + "step": 14789 + }, + { + "epoch": 1.94, + "learning_rate": 1.7730619673711806e-05, + "loss": 0.0666, + "step": 14790 + }, + { + "epoch": 1.94, + "learning_rate": 1.7730283261472476e-05, + "loss": 0.0707, + "step": 14791 + }, + { + "epoch": 1.94, + "learning_rate": 1.7729946827492135e-05, + "loss": 0.0844, + "step": 14792 + }, + { + "epoch": 1.94, + "learning_rate": 1.7729610371771733e-05, + "loss": 0.0681, + "step": 14793 + }, + { + "epoch": 1.94, + "learning_rate": 1.772927389431221e-05, + "loss": 0.0692, + "step": 14794 + }, + { + "epoch": 1.94, + "learning_rate": 1.772893739511452e-05, + "loss": 0.0732, + "step": 14795 + }, + { + "epoch": 1.94, + "learning_rate": 1.7728600874179603e-05, + "loss": 0.0672, + "step": 14796 + }, + { + "epoch": 1.94, + "learning_rate": 1.7728264331508403e-05, + "loss": 0.0742, + "step": 14797 + }, + { + "epoch": 1.94, + "learning_rate": 1.7727927767101877e-05, + "loss": 0.0673, + "step": 14798 + }, + { + "epoch": 1.94, + "learning_rate": 1.7727591180960967e-05, + "loss": 0.0827, + "step": 14799 + }, + { + "epoch": 1.94, + "learning_rate": 1.7727254573086615e-05, + "loss": 0.063, + "step": 14800 + }, + { + "epoch": 1.94, + "learning_rate": 1.7726917943479772e-05, + "loss": 0.0765, + "step": 14801 + }, + { + "epoch": 1.94, + "learning_rate": 1.772658129214138e-05, + "loss": 0.0661, + "step": 14802 + }, + { + "epoch": 1.94, + "learning_rate": 1.7726244619072393e-05, + "loss": 0.0593, + "step": 14803 + }, + { + "epoch": 1.94, + "learning_rate": 1.7725907924273755e-05, + "loss": 0.0755, + "step": 14804 + }, + { + "epoch": 1.94, + "learning_rate": 1.7725571207746413e-05, + "loss": 0.0773, + "step": 14805 + }, + { + "epoch": 1.94, + "learning_rate": 1.7725234469491312e-05, + "loss": 0.0783, + "step": 14806 + }, + { + "epoch": 1.94, + "learning_rate": 1.77248977095094e-05, + "loss": 0.0755, + "step": 14807 + }, + { + "epoch": 1.94, + "learning_rate": 1.7724560927801625e-05, + "loss": 0.0739, + "step": 14808 + }, + { + "epoch": 1.94, + "learning_rate": 1.7724224124368934e-05, + "loss": 0.0695, + "step": 14809 + }, + { + "epoch": 1.94, + "learning_rate": 1.7723887299212274e-05, + "loss": 0.0753, + "step": 14810 + }, + { + "epoch": 1.94, + "learning_rate": 1.7723550452332593e-05, + "loss": 0.0724, + "step": 14811 + }, + { + "epoch": 1.94, + "learning_rate": 1.7723213583730837e-05, + "loss": 0.0697, + "step": 14812 + }, + { + "epoch": 1.94, + "learning_rate": 1.7722876693407954e-05, + "loss": 0.0687, + "step": 14813 + }, + { + "epoch": 1.94, + "learning_rate": 1.772253978136489e-05, + "loss": 0.0657, + "step": 14814 + }, + { + "epoch": 1.94, + "learning_rate": 1.7722202847602594e-05, + "loss": 0.0733, + "step": 14815 + }, + { + "epoch": 1.94, + "learning_rate": 1.7721865892122017e-05, + "loss": 0.0763, + "step": 14816 + }, + { + "epoch": 1.94, + "learning_rate": 1.77215289149241e-05, + "loss": 0.068, + "step": 14817 + }, + { + "epoch": 1.94, + "learning_rate": 1.7721191916009796e-05, + "loss": 0.0658, + "step": 14818 + }, + { + "epoch": 1.94, + "learning_rate": 1.7720854895380052e-05, + "loss": 0.0725, + "step": 14819 + }, + { + "epoch": 1.94, + "learning_rate": 1.7720517853035813e-05, + "loss": 0.0747, + "step": 14820 + }, + { + "epoch": 1.94, + "learning_rate": 1.772018078897803e-05, + "loss": 0.0626, + "step": 14821 + }, + { + "epoch": 1.94, + "learning_rate": 1.7719843703207648e-05, + "loss": 0.0785, + "step": 14822 + }, + { + "epoch": 1.94, + "learning_rate": 1.771950659572562e-05, + "loss": 0.068, + "step": 14823 + }, + { + "epoch": 1.94, + "learning_rate": 1.7719169466532886e-05, + "loss": 0.0645, + "step": 14824 + }, + { + "epoch": 1.94, + "learning_rate": 1.77188323156304e-05, + "loss": 0.0716, + "step": 14825 + }, + { + "epoch": 1.94, + "learning_rate": 1.771849514301911e-05, + "loss": 0.0619, + "step": 14826 + }, + { + "epoch": 1.94, + "learning_rate": 1.7718157948699966e-05, + "loss": 0.0765, + "step": 14827 + }, + { + "epoch": 1.94, + "learning_rate": 1.771782073267391e-05, + "loss": 0.0771, + "step": 14828 + }, + { + "epoch": 1.94, + "learning_rate": 1.7717483494941895e-05, + "loss": 0.0632, + "step": 14829 + }, + { + "epoch": 1.94, + "learning_rate": 1.7717146235504873e-05, + "loss": 0.066, + "step": 14830 + }, + { + "epoch": 1.94, + "learning_rate": 1.7716808954363783e-05, + "loss": 0.0708, + "step": 14831 + }, + { + "epoch": 1.94, + "learning_rate": 1.771647165151958e-05, + "loss": 0.0684, + "step": 14832 + }, + { + "epoch": 1.94, + "learning_rate": 1.771613432697321e-05, + "loss": 0.0741, + "step": 14833 + }, + { + "epoch": 1.94, + "learning_rate": 1.7715796980725626e-05, + "loss": 0.0725, + "step": 14834 + }, + { + "epoch": 1.94, + "learning_rate": 1.771545961277777e-05, + "loss": 0.0754, + "step": 14835 + }, + { + "epoch": 1.94, + "learning_rate": 1.7715122223130602e-05, + "loss": 0.0684, + "step": 14836 + }, + { + "epoch": 1.94, + "learning_rate": 1.771478481178506e-05, + "loss": 0.0687, + "step": 14837 + }, + { + "epoch": 1.94, + "learning_rate": 1.771444737874209e-05, + "loss": 0.0764, + "step": 14838 + }, + { + "epoch": 1.94, + "learning_rate": 1.7714109924002657e-05, + "loss": 0.0805, + "step": 14839 + }, + { + "epoch": 1.94, + "learning_rate": 1.7713772447567696e-05, + "loss": 0.077, + "step": 14840 + }, + { + "epoch": 1.94, + "learning_rate": 1.7713434949438162e-05, + "loss": 0.0732, + "step": 14841 + }, + { + "epoch": 1.94, + "learning_rate": 1.7713097429615002e-05, + "loss": 0.0766, + "step": 14842 + }, + { + "epoch": 1.94, + "learning_rate": 1.771275988809917e-05, + "loss": 0.0811, + "step": 14843 + }, + { + "epoch": 1.94, + "learning_rate": 1.7712422324891606e-05, + "loss": 0.075, + "step": 14844 + }, + { + "epoch": 1.94, + "learning_rate": 1.7712084739993267e-05, + "loss": 0.0708, + "step": 14845 + }, + { + "epoch": 1.94, + "learning_rate": 1.77117471334051e-05, + "loss": 0.0689, + "step": 14846 + }, + { + "epoch": 1.94, + "learning_rate": 1.7711409505128057e-05, + "loss": 0.0697, + "step": 14847 + }, + { + "epoch": 1.94, + "learning_rate": 1.7711071855163086e-05, + "loss": 0.0791, + "step": 14848 + }, + { + "epoch": 1.95, + "learning_rate": 1.771073418351113e-05, + "loss": 0.0657, + "step": 14849 + }, + { + "epoch": 1.95, + "learning_rate": 1.771039649017315e-05, + "loss": 0.0803, + "step": 14850 + }, + { + "epoch": 1.95, + "learning_rate": 1.771005877515009e-05, + "loss": 0.0809, + "step": 14851 + }, + { + "epoch": 1.95, + "learning_rate": 1.7709721038442897e-05, + "loss": 0.0726, + "step": 14852 + }, + { + "epoch": 1.95, + "learning_rate": 1.770938328005253e-05, + "loss": 0.0777, + "step": 14853 + }, + { + "epoch": 1.95, + "learning_rate": 1.770904549997993e-05, + "loss": 0.0613, + "step": 14854 + }, + { + "epoch": 1.95, + "learning_rate": 1.7708707698226048e-05, + "loss": 0.0802, + "step": 14855 + }, + { + "epoch": 1.95, + "learning_rate": 1.7708369874791836e-05, + "loss": 0.0668, + "step": 14856 + }, + { + "epoch": 1.95, + "learning_rate": 1.7708032029678247e-05, + "loss": 0.0745, + "step": 14857 + }, + { + "epoch": 1.95, + "learning_rate": 1.7707694162886228e-05, + "loss": 0.0754, + "step": 14858 + }, + { + "epoch": 1.95, + "learning_rate": 1.770735627441673e-05, + "loss": 0.077, + "step": 14859 + }, + { + "epoch": 1.95, + "learning_rate": 1.77070183642707e-05, + "loss": 0.0801, + "step": 14860 + }, + { + "epoch": 1.95, + "learning_rate": 1.7706680432449096e-05, + "loss": 0.0749, + "step": 14861 + }, + { + "epoch": 1.95, + "learning_rate": 1.770634247895286e-05, + "loss": 0.0582, + "step": 14862 + }, + { + "epoch": 1.95, + "learning_rate": 1.7706004503782946e-05, + "loss": 0.0758, + "step": 14863 + }, + { + "epoch": 1.95, + "learning_rate": 1.7705666506940307e-05, + "loss": 0.0658, + "step": 14864 + }, + { + "epoch": 1.95, + "learning_rate": 1.770532848842589e-05, + "loss": 0.0722, + "step": 14865 + }, + { + "epoch": 1.95, + "learning_rate": 1.7704990448240647e-05, + "loss": 0.0737, + "step": 14866 + }, + { + "epoch": 1.95, + "learning_rate": 1.770465238638553e-05, + "loss": 0.0674, + "step": 14867 + }, + { + "epoch": 1.95, + "learning_rate": 1.7704314302861486e-05, + "loss": 0.0816, + "step": 14868 + }, + { + "epoch": 1.95, + "learning_rate": 1.770397619766947e-05, + "loss": 0.0707, + "step": 14869 + }, + { + "epoch": 1.95, + "learning_rate": 1.7703638070810428e-05, + "loss": 0.0766, + "step": 14870 + }, + { + "epoch": 1.95, + "learning_rate": 1.7703299922285317e-05, + "loss": 0.0778, + "step": 14871 + }, + { + "epoch": 1.95, + "learning_rate": 1.7702961752095085e-05, + "loss": 0.073, + "step": 14872 + }, + { + "epoch": 1.95, + "learning_rate": 1.7702623560240685e-05, + "loss": 0.0702, + "step": 14873 + }, + { + "epoch": 1.95, + "learning_rate": 1.7702285346723064e-05, + "loss": 0.0754, + "step": 14874 + }, + { + "epoch": 1.95, + "learning_rate": 1.770194711154318e-05, + "loss": 0.0754, + "step": 14875 + }, + { + "epoch": 1.95, + "learning_rate": 1.7701608854701972e-05, + "loss": 0.0617, + "step": 14876 + }, + { + "epoch": 1.95, + "learning_rate": 1.7701270576200404e-05, + "loss": 0.0794, + "step": 14877 + }, + { + "epoch": 1.95, + "learning_rate": 1.770093227603942e-05, + "loss": 0.068, + "step": 14878 + }, + { + "epoch": 1.95, + "learning_rate": 1.7700593954219978e-05, + "loss": 0.0798, + "step": 14879 + }, + { + "epoch": 1.95, + "learning_rate": 1.7700255610743028e-05, + "loss": 0.0778, + "step": 14880 + }, + { + "epoch": 1.95, + "learning_rate": 1.7699917245609512e-05, + "loss": 0.0705, + "step": 14881 + }, + { + "epoch": 1.95, + "learning_rate": 1.7699578858820392e-05, + "loss": 0.067, + "step": 14882 + }, + { + "epoch": 1.95, + "learning_rate": 1.7699240450376618e-05, + "loss": 0.0778, + "step": 14883 + }, + { + "epoch": 1.95, + "learning_rate": 1.7698902020279137e-05, + "loss": 0.0745, + "step": 14884 + }, + { + "epoch": 1.95, + "learning_rate": 1.7698563568528906e-05, + "loss": 0.0792, + "step": 14885 + }, + { + "epoch": 1.95, + "learning_rate": 1.7698225095126873e-05, + "loss": 0.0558, + "step": 14886 + }, + { + "epoch": 1.95, + "learning_rate": 1.7697886600073994e-05, + "loss": 0.0875, + "step": 14887 + }, + { + "epoch": 1.95, + "learning_rate": 1.769754808337122e-05, + "loss": 0.0812, + "step": 14888 + }, + { + "epoch": 1.95, + "learning_rate": 1.76972095450195e-05, + "loss": 0.0698, + "step": 14889 + }, + { + "epoch": 1.95, + "learning_rate": 1.7696870985019786e-05, + "loss": 0.0734, + "step": 14890 + }, + { + "epoch": 1.95, + "learning_rate": 1.7696532403373034e-05, + "loss": 0.0931, + "step": 14891 + }, + { + "epoch": 1.95, + "learning_rate": 1.7696193800080196e-05, + "loss": 0.0815, + "step": 14892 + }, + { + "epoch": 1.95, + "learning_rate": 1.7695855175142224e-05, + "loss": 0.0754, + "step": 14893 + }, + { + "epoch": 1.95, + "learning_rate": 1.7695516528560066e-05, + "loss": 0.0737, + "step": 14894 + }, + { + "epoch": 1.95, + "learning_rate": 1.7695177860334678e-05, + "loss": 0.0733, + "step": 14895 + }, + { + "epoch": 1.95, + "learning_rate": 1.7694839170467012e-05, + "loss": 0.0885, + "step": 14896 + }, + { + "epoch": 1.95, + "learning_rate": 1.769450045895802e-05, + "loss": 0.0712, + "step": 14897 + }, + { + "epoch": 1.95, + "learning_rate": 1.769416172580866e-05, + "loss": 0.0747, + "step": 14898 + }, + { + "epoch": 1.95, + "learning_rate": 1.7693822971019875e-05, + "loss": 0.0831, + "step": 14899 + }, + { + "epoch": 1.95, + "learning_rate": 1.769348419459262e-05, + "loss": 0.0717, + "step": 14900 + }, + { + "epoch": 1.95, + "learning_rate": 1.7693145396527857e-05, + "loss": 0.0755, + "step": 14901 + }, + { + "epoch": 1.95, + "learning_rate": 1.769280657682653e-05, + "loss": 0.0698, + "step": 14902 + }, + { + "epoch": 1.95, + "learning_rate": 1.7692467735489594e-05, + "loss": 0.0735, + "step": 14903 + }, + { + "epoch": 1.95, + "learning_rate": 1.7692128872518003e-05, + "loss": 0.0705, + "step": 14904 + }, + { + "epoch": 1.95, + "learning_rate": 1.7691789987912706e-05, + "loss": 0.067, + "step": 14905 + }, + { + "epoch": 1.95, + "learning_rate": 1.769145108167466e-05, + "loss": 0.0767, + "step": 14906 + }, + { + "epoch": 1.95, + "learning_rate": 1.7691112153804822e-05, + "loss": 0.0728, + "step": 14907 + }, + { + "epoch": 1.95, + "learning_rate": 1.7690773204304136e-05, + "loss": 0.0813, + "step": 14908 + }, + { + "epoch": 1.95, + "learning_rate": 1.769043423317356e-05, + "loss": 0.0828, + "step": 14909 + }, + { + "epoch": 1.95, + "learning_rate": 1.769009524041405e-05, + "loss": 0.0749, + "step": 14910 + }, + { + "epoch": 1.95, + "learning_rate": 1.7689756226026555e-05, + "loss": 0.0612, + "step": 14911 + }, + { + "epoch": 1.95, + "learning_rate": 1.768941719001203e-05, + "loss": 0.0768, + "step": 14912 + }, + { + "epoch": 1.95, + "learning_rate": 1.768907813237143e-05, + "loss": 0.0757, + "step": 14913 + }, + { + "epoch": 1.95, + "learning_rate": 1.7688739053105702e-05, + "loss": 0.0677, + "step": 14914 + }, + { + "epoch": 1.95, + "learning_rate": 1.768839995221581e-05, + "loss": 0.0785, + "step": 14915 + }, + { + "epoch": 1.95, + "learning_rate": 1.7688060829702706e-05, + "loss": 0.0716, + "step": 14916 + }, + { + "epoch": 1.95, + "learning_rate": 1.7687721685567333e-05, + "loss": 0.0808, + "step": 14917 + }, + { + "epoch": 1.95, + "learning_rate": 1.7687382519810655e-05, + "loss": 0.0675, + "step": 14918 + }, + { + "epoch": 1.95, + "learning_rate": 1.7687043332433624e-05, + "loss": 0.0786, + "step": 14919 + }, + { + "epoch": 1.95, + "learning_rate": 1.7686704123437192e-05, + "loss": 0.0628, + "step": 14920 + }, + { + "epoch": 1.95, + "learning_rate": 1.7686364892822314e-05, + "loss": 0.0769, + "step": 14921 + }, + { + "epoch": 1.95, + "learning_rate": 1.7686025640589942e-05, + "loss": 0.0665, + "step": 14922 + }, + { + "epoch": 1.95, + "learning_rate": 1.7685686366741035e-05, + "loss": 0.0716, + "step": 14923 + }, + { + "epoch": 1.95, + "learning_rate": 1.7685347071276545e-05, + "loss": 0.0617, + "step": 14924 + }, + { + "epoch": 1.96, + "learning_rate": 1.768500775419742e-05, + "loss": 0.0716, + "step": 14925 + }, + { + "epoch": 1.96, + "learning_rate": 1.768466841550463e-05, + "loss": 0.0746, + "step": 14926 + }, + { + "epoch": 1.96, + "learning_rate": 1.7684329055199113e-05, + "loss": 0.0704, + "step": 14927 + }, + { + "epoch": 1.96, + "learning_rate": 1.768398967328183e-05, + "loss": 0.0791, + "step": 14928 + }, + { + "epoch": 1.96, + "learning_rate": 1.7683650269753737e-05, + "loss": 0.0791, + "step": 14929 + }, + { + "epoch": 1.96, + "learning_rate": 1.7683310844615784e-05, + "loss": 0.0575, + "step": 14930 + }, + { + "epoch": 1.96, + "learning_rate": 1.768297139786893e-05, + "loss": 0.0748, + "step": 14931 + }, + { + "epoch": 1.96, + "learning_rate": 1.7682631929514126e-05, + "loss": 0.0672, + "step": 14932 + }, + { + "epoch": 1.96, + "learning_rate": 1.7682292439552332e-05, + "loss": 0.0741, + "step": 14933 + }, + { + "epoch": 1.96, + "learning_rate": 1.76819529279845e-05, + "loss": 0.0793, + "step": 14934 + }, + { + "epoch": 1.96, + "learning_rate": 1.768161339481158e-05, + "loss": 0.0776, + "step": 14935 + }, + { + "epoch": 1.96, + "learning_rate": 1.7681273840034533e-05, + "loss": 0.0688, + "step": 14936 + }, + { + "epoch": 1.96, + "learning_rate": 1.7680934263654314e-05, + "loss": 0.083, + "step": 14937 + }, + { + "epoch": 1.96, + "learning_rate": 1.768059466567188e-05, + "loss": 0.074, + "step": 14938 + }, + { + "epoch": 1.96, + "learning_rate": 1.768025504608818e-05, + "loss": 0.0583, + "step": 14939 + }, + { + "epoch": 1.96, + "learning_rate": 1.7679915404904166e-05, + "loss": 0.0759, + "step": 14940 + }, + { + "epoch": 1.96, + "learning_rate": 1.7679575742120803e-05, + "loss": 0.0678, + "step": 14941 + }, + { + "epoch": 1.96, + "learning_rate": 1.767923605773904e-05, + "loss": 0.0705, + "step": 14942 + }, + { + "epoch": 1.96, + "learning_rate": 1.767889635175984e-05, + "loss": 0.0636, + "step": 14943 + }, + { + "epoch": 1.96, + "learning_rate": 1.767855662418415e-05, + "loss": 0.0721, + "step": 14944 + }, + { + "epoch": 1.96, + "learning_rate": 1.767821687501293e-05, + "loss": 0.0703, + "step": 14945 + }, + { + "epoch": 1.96, + "learning_rate": 1.767787710424713e-05, + "loss": 0.0861, + "step": 14946 + }, + { + "epoch": 1.96, + "learning_rate": 1.7677537311887712e-05, + "loss": 0.0793, + "step": 14947 + }, + { + "epoch": 1.96, + "learning_rate": 1.767719749793563e-05, + "loss": 0.0715, + "step": 14948 + }, + { + "epoch": 1.96, + "learning_rate": 1.7676857662391836e-05, + "loss": 0.0683, + "step": 14949 + }, + { + "epoch": 1.96, + "learning_rate": 1.7676517805257288e-05, + "loss": 0.0729, + "step": 14950 + }, + { + "epoch": 1.96, + "learning_rate": 1.7676177926532948e-05, + "loss": 0.0661, + "step": 14951 + }, + { + "epoch": 1.96, + "learning_rate": 1.7675838026219764e-05, + "loss": 0.0676, + "step": 14952 + }, + { + "epoch": 1.96, + "learning_rate": 1.7675498104318692e-05, + "loss": 0.0783, + "step": 14953 + }, + { + "epoch": 1.96, + "learning_rate": 1.767515816083069e-05, + "loss": 0.0776, + "step": 14954 + }, + { + "epoch": 1.96, + "learning_rate": 1.767481819575672e-05, + "loss": 0.0642, + "step": 14955 + }, + { + "epoch": 1.96, + "learning_rate": 1.7674478209097725e-05, + "loss": 0.0763, + "step": 14956 + }, + { + "epoch": 1.96, + "learning_rate": 1.7674138200854674e-05, + "loss": 0.0739, + "step": 14957 + }, + { + "epoch": 1.96, + "learning_rate": 1.7673798171028517e-05, + "loss": 0.0796, + "step": 14958 + }, + { + "epoch": 1.96, + "learning_rate": 1.767345811962021e-05, + "loss": 0.0714, + "step": 14959 + }, + { + "epoch": 1.96, + "learning_rate": 1.767311804663071e-05, + "loss": 0.0785, + "step": 14960 + }, + { + "epoch": 1.96, + "learning_rate": 1.7672777952060975e-05, + "loss": 0.0734, + "step": 14961 + }, + { + "epoch": 1.96, + "learning_rate": 1.7672437835911962e-05, + "loss": 0.0742, + "step": 14962 + }, + { + "epoch": 1.96, + "learning_rate": 1.7672097698184625e-05, + "loss": 0.0676, + "step": 14963 + }, + { + "epoch": 1.96, + "learning_rate": 1.767175753887992e-05, + "loss": 0.0654, + "step": 14964 + }, + { + "epoch": 1.96, + "learning_rate": 1.767141735799881e-05, + "loss": 0.0612, + "step": 14965 + }, + { + "epoch": 1.96, + "learning_rate": 1.7671077155542244e-05, + "loss": 0.0683, + "step": 14966 + }, + { + "epoch": 1.96, + "learning_rate": 1.7670736931511184e-05, + "loss": 0.0727, + "step": 14967 + }, + { + "epoch": 1.96, + "learning_rate": 1.7670396685906584e-05, + "loss": 0.0677, + "step": 14968 + }, + { + "epoch": 1.96, + "learning_rate": 1.76700564187294e-05, + "loss": 0.0732, + "step": 14969 + }, + { + "epoch": 1.96, + "learning_rate": 1.7669716129980596e-05, + "loss": 0.0815, + "step": 14970 + }, + { + "epoch": 1.96, + "learning_rate": 1.7669375819661117e-05, + "loss": 0.0737, + "step": 14971 + }, + { + "epoch": 1.96, + "learning_rate": 1.7669035487771928e-05, + "loss": 0.0774, + "step": 14972 + }, + { + "epoch": 1.96, + "learning_rate": 1.766869513431399e-05, + "loss": 0.0688, + "step": 14973 + }, + { + "epoch": 1.96, + "learning_rate": 1.766835475928825e-05, + "loss": 0.0653, + "step": 14974 + }, + { + "epoch": 1.96, + "learning_rate": 1.7668014362695672e-05, + "loss": 0.066, + "step": 14975 + }, + { + "epoch": 1.96, + "learning_rate": 1.7667673944537217e-05, + "loss": 0.0639, + "step": 14976 + }, + { + "epoch": 1.96, + "learning_rate": 1.766733350481383e-05, + "loss": 0.0728, + "step": 14977 + }, + { + "epoch": 1.96, + "learning_rate": 1.766699304352648e-05, + "loss": 0.0694, + "step": 14978 + }, + { + "epoch": 1.96, + "learning_rate": 1.766665256067612e-05, + "loss": 0.0743, + "step": 14979 + }, + { + "epoch": 1.96, + "learning_rate": 1.7666312056263706e-05, + "loss": 0.0638, + "step": 14980 + }, + { + "epoch": 1.96, + "learning_rate": 1.7665971530290202e-05, + "loss": 0.0827, + "step": 14981 + }, + { + "epoch": 1.96, + "learning_rate": 1.7665630982756554e-05, + "loss": 0.0777, + "step": 14982 + }, + { + "epoch": 1.96, + "learning_rate": 1.7665290413663733e-05, + "loss": 0.061, + "step": 14983 + }, + { + "epoch": 1.96, + "learning_rate": 1.766494982301269e-05, + "loss": 0.0775, + "step": 14984 + }, + { + "epoch": 1.96, + "learning_rate": 1.766460921080438e-05, + "loss": 0.0608, + "step": 14985 + }, + { + "epoch": 1.96, + "learning_rate": 1.7664268577039768e-05, + "loss": 0.0759, + "step": 14986 + }, + { + "epoch": 1.96, + "learning_rate": 1.7663927921719812e-05, + "loss": 0.0734, + "step": 14987 + }, + { + "epoch": 1.96, + "learning_rate": 1.766358724484546e-05, + "loss": 0.0627, + "step": 14988 + }, + { + "epoch": 1.96, + "learning_rate": 1.7663246546417684e-05, + "loss": 0.066, + "step": 14989 + }, + { + "epoch": 1.96, + "learning_rate": 1.766290582643743e-05, + "loss": 0.0681, + "step": 14990 + }, + { + "epoch": 1.96, + "learning_rate": 1.7662565084905665e-05, + "loss": 0.0726, + "step": 14991 + }, + { + "epoch": 1.96, + "learning_rate": 1.766222432182334e-05, + "loss": 0.0796, + "step": 14992 + }, + { + "epoch": 1.96, + "learning_rate": 1.766188353719142e-05, + "loss": 0.0701, + "step": 14993 + }, + { + "epoch": 1.96, + "learning_rate": 1.766154273101086e-05, + "loss": 0.0747, + "step": 14994 + }, + { + "epoch": 1.96, + "learning_rate": 1.7661201903282615e-05, + "loss": 0.064, + "step": 14995 + }, + { + "epoch": 1.96, + "learning_rate": 1.7660861054007653e-05, + "loss": 0.075, + "step": 14996 + }, + { + "epoch": 1.96, + "learning_rate": 1.7660520183186927e-05, + "loss": 0.0689, + "step": 14997 + }, + { + "epoch": 1.96, + "learning_rate": 1.7660179290821393e-05, + "loss": 0.0756, + "step": 14998 + }, + { + "epoch": 1.96, + "learning_rate": 1.7659838376912013e-05, + "loss": 0.0738, + "step": 14999 + }, + { + "epoch": 1.96, + "learning_rate": 1.7659497441459746e-05, + "loss": 0.0721, + "step": 15000 + }, + { + "epoch": 1.97, + "learning_rate": 1.7659156484465554e-05, + "loss": 0.0765, + "step": 15001 + }, + { + "epoch": 1.97, + "learning_rate": 1.7658815505930386e-05, + "loss": 0.0697, + "step": 15002 + }, + { + "epoch": 1.97, + "learning_rate": 1.7658474505855214e-05, + "loss": 0.0592, + "step": 15003 + }, + { + "epoch": 1.97, + "learning_rate": 1.7658133484240984e-05, + "loss": 0.0741, + "step": 15004 + }, + { + "epoch": 1.97, + "learning_rate": 1.7657792441088664e-05, + "loss": 0.0647, + "step": 15005 + }, + { + "epoch": 1.97, + "learning_rate": 1.7657451376399213e-05, + "loss": 0.0791, + "step": 15006 + }, + { + "epoch": 1.97, + "learning_rate": 1.7657110290173584e-05, + "loss": 0.0699, + "step": 15007 + }, + { + "epoch": 1.97, + "learning_rate": 1.765676918241274e-05, + "loss": 0.081, + "step": 15008 + }, + { + "epoch": 1.97, + "learning_rate": 1.7656428053117645e-05, + "loss": 0.0731, + "step": 15009 + }, + { + "epoch": 1.97, + "learning_rate": 1.765608690228925e-05, + "loss": 0.0743, + "step": 15010 + }, + { + "epoch": 1.97, + "learning_rate": 1.765574572992852e-05, + "loss": 0.0721, + "step": 15011 + }, + { + "epoch": 1.97, + "learning_rate": 1.765540453603641e-05, + "loss": 0.0691, + "step": 15012 + }, + { + "epoch": 1.97, + "learning_rate": 1.7655063320613885e-05, + "loss": 0.0568, + "step": 15013 + }, + { + "epoch": 1.97, + "learning_rate": 1.76547220836619e-05, + "loss": 0.0722, + "step": 15014 + }, + { + "epoch": 1.97, + "learning_rate": 1.765438082518142e-05, + "loss": 0.0764, + "step": 15015 + }, + { + "epoch": 1.97, + "learning_rate": 1.76540395451734e-05, + "loss": 0.0628, + "step": 15016 + }, + { + "epoch": 1.97, + "learning_rate": 1.76536982436388e-05, + "loss": 0.066, + "step": 15017 + }, + { + "epoch": 1.97, + "learning_rate": 1.765335692057858e-05, + "loss": 0.0609, + "step": 15018 + }, + { + "epoch": 1.97, + "learning_rate": 1.7653015575993703e-05, + "loss": 0.0721, + "step": 15019 + }, + { + "epoch": 1.97, + "learning_rate": 1.7652674209885132e-05, + "loss": 0.0707, + "step": 15020 + }, + { + "epoch": 1.97, + "learning_rate": 1.7652332822253816e-05, + "loss": 0.0672, + "step": 15021 + }, + { + "epoch": 1.97, + "learning_rate": 1.7651991413100725e-05, + "loss": 0.0679, + "step": 15022 + }, + { + "epoch": 1.97, + "learning_rate": 1.7651649982426813e-05, + "loss": 0.0765, + "step": 15023 + }, + { + "epoch": 1.97, + "learning_rate": 1.7651308530233043e-05, + "loss": 0.0756, + "step": 15024 + }, + { + "epoch": 1.97, + "learning_rate": 1.7650967056520378e-05, + "loss": 0.071, + "step": 15025 + }, + { + "epoch": 1.97, + "learning_rate": 1.7650625561289772e-05, + "loss": 0.0647, + "step": 15026 + }, + { + "epoch": 1.97, + "learning_rate": 1.7650284044542192e-05, + "loss": 0.0608, + "step": 15027 + }, + { + "epoch": 1.97, + "learning_rate": 1.7649942506278596e-05, + "loss": 0.0799, + "step": 15028 + }, + { + "epoch": 1.97, + "learning_rate": 1.764960094649994e-05, + "loss": 0.0731, + "step": 15029 + }, + { + "epoch": 1.97, + "learning_rate": 1.764925936520719e-05, + "loss": 0.0688, + "step": 15030 + }, + { + "epoch": 1.97, + "learning_rate": 1.764891776240131e-05, + "loss": 0.0653, + "step": 15031 + }, + { + "epoch": 1.97, + "learning_rate": 1.7648576138083254e-05, + "loss": 0.0648, + "step": 15032 + }, + { + "epoch": 1.97, + "learning_rate": 1.764823449225398e-05, + "loss": 0.0824, + "step": 15033 + }, + { + "epoch": 1.97, + "learning_rate": 1.764789282491446e-05, + "loss": 0.0763, + "step": 15034 + }, + { + "epoch": 1.97, + "learning_rate": 1.7647551136065648e-05, + "loss": 0.0685, + "step": 15035 + }, + { + "epoch": 1.97, + "learning_rate": 1.76472094257085e-05, + "loss": 0.0762, + "step": 15036 + }, + { + "epoch": 1.97, + "learning_rate": 1.7646867693843988e-05, + "loss": 0.0727, + "step": 15037 + }, + { + "epoch": 1.97, + "learning_rate": 1.7646525940473068e-05, + "loss": 0.062, + "step": 15038 + }, + { + "epoch": 1.97, + "learning_rate": 1.76461841655967e-05, + "loss": 0.0836, + "step": 15039 + }, + { + "epoch": 1.97, + "learning_rate": 1.7645842369215846e-05, + "loss": 0.0592, + "step": 15040 + }, + { + "epoch": 1.97, + "learning_rate": 1.7645500551331464e-05, + "loss": 0.0574, + "step": 15041 + }, + { + "epoch": 1.97, + "learning_rate": 1.764515871194452e-05, + "loss": 0.0662, + "step": 15042 + }, + { + "epoch": 1.97, + "learning_rate": 1.7644816851055976e-05, + "loss": 0.0619, + "step": 15043 + }, + { + "epoch": 1.97, + "learning_rate": 1.7644474968666796e-05, + "loss": 0.0676, + "step": 15044 + }, + { + "epoch": 1.97, + "learning_rate": 1.7644133064777932e-05, + "loss": 0.0808, + "step": 15045 + }, + { + "epoch": 1.97, + "learning_rate": 1.764379113939035e-05, + "loss": 0.0731, + "step": 15046 + }, + { + "epoch": 1.97, + "learning_rate": 1.7643449192505017e-05, + "loss": 0.0556, + "step": 15047 + }, + { + "epoch": 1.97, + "learning_rate": 1.7643107224122885e-05, + "loss": 0.0665, + "step": 15048 + }, + { + "epoch": 1.97, + "learning_rate": 1.7642765234244926e-05, + "loss": 0.0755, + "step": 15049 + }, + { + "epoch": 1.97, + "learning_rate": 1.7642423222872094e-05, + "loss": 0.0694, + "step": 15050 + }, + { + "epoch": 1.97, + "learning_rate": 1.7642081190005353e-05, + "loss": 0.0621, + "step": 15051 + }, + { + "epoch": 1.97, + "learning_rate": 1.7641739135645666e-05, + "loss": 0.0652, + "step": 15052 + }, + { + "epoch": 1.97, + "learning_rate": 1.7641397059793996e-05, + "loss": 0.0597, + "step": 15053 + }, + { + "epoch": 1.97, + "learning_rate": 1.7641054962451304e-05, + "loss": 0.0749, + "step": 15054 + }, + { + "epoch": 1.97, + "learning_rate": 1.764071284361855e-05, + "loss": 0.0732, + "step": 15055 + }, + { + "epoch": 1.97, + "learning_rate": 1.76403707032967e-05, + "loss": 0.0662, + "step": 15056 + }, + { + "epoch": 1.97, + "learning_rate": 1.7640028541486713e-05, + "loss": 0.0793, + "step": 15057 + }, + { + "epoch": 1.97, + "learning_rate": 1.763968635818955e-05, + "loss": 0.0873, + "step": 15058 + }, + { + "epoch": 1.97, + "learning_rate": 1.763934415340618e-05, + "loss": 0.0736, + "step": 15059 + }, + { + "epoch": 1.97, + "learning_rate": 1.7639001927137562e-05, + "loss": 0.0741, + "step": 15060 + }, + { + "epoch": 1.97, + "learning_rate": 1.7638659679384655e-05, + "loss": 0.0691, + "step": 15061 + }, + { + "epoch": 1.97, + "learning_rate": 1.7638317410148427e-05, + "loss": 0.0716, + "step": 15062 + }, + { + "epoch": 1.97, + "learning_rate": 1.7637975119429834e-05, + "loss": 0.0724, + "step": 15063 + }, + { + "epoch": 1.97, + "learning_rate": 1.7637632807229845e-05, + "loss": 0.0734, + "step": 15064 + }, + { + "epoch": 1.97, + "learning_rate": 1.763729047354942e-05, + "loss": 0.0632, + "step": 15065 + }, + { + "epoch": 1.97, + "learning_rate": 1.7636948118389522e-05, + "loss": 0.0806, + "step": 15066 + }, + { + "epoch": 1.97, + "learning_rate": 1.763660574175112e-05, + "loss": 0.071, + "step": 15067 + }, + { + "epoch": 1.97, + "learning_rate": 1.7636263343635166e-05, + "loss": 0.0711, + "step": 15068 + }, + { + "epoch": 1.97, + "learning_rate": 1.7635920924042628e-05, + "loss": 0.0819, + "step": 15069 + }, + { + "epoch": 1.97, + "learning_rate": 1.7635578482974467e-05, + "loss": 0.0746, + "step": 15070 + }, + { + "epoch": 1.97, + "learning_rate": 1.763523602043165e-05, + "loss": 0.08, + "step": 15071 + }, + { + "epoch": 1.97, + "learning_rate": 1.7634893536415137e-05, + "loss": 0.0772, + "step": 15072 + }, + { + "epoch": 1.97, + "learning_rate": 1.7634551030925896e-05, + "loss": 0.0715, + "step": 15073 + }, + { + "epoch": 1.97, + "learning_rate": 1.7634208503964885e-05, + "loss": 0.0732, + "step": 15074 + }, + { + "epoch": 1.97, + "learning_rate": 1.763386595553307e-05, + "loss": 0.0678, + "step": 15075 + }, + { + "epoch": 1.97, + "learning_rate": 1.763352338563141e-05, + "loss": 0.0686, + "step": 15076 + }, + { + "epoch": 1.97, + "learning_rate": 1.7633180794260875e-05, + "loss": 0.0682, + "step": 15077 + }, + { + "epoch": 1.98, + "learning_rate": 1.7632838181422425e-05, + "loss": 0.082, + "step": 15078 + }, + { + "epoch": 1.98, + "learning_rate": 1.7632495547117023e-05, + "loss": 0.0605, + "step": 15079 + }, + { + "epoch": 1.98, + "learning_rate": 1.7632152891345634e-05, + "loss": 0.0725, + "step": 15080 + }, + { + "epoch": 1.98, + "learning_rate": 1.7631810214109224e-05, + "loss": 0.0754, + "step": 15081 + }, + { + "epoch": 1.98, + "learning_rate": 1.7631467515408752e-05, + "loss": 0.0714, + "step": 15082 + }, + { + "epoch": 1.98, + "learning_rate": 1.7631124795245184e-05, + "loss": 0.0779, + "step": 15083 + }, + { + "epoch": 1.98, + "learning_rate": 1.7630782053619484e-05, + "loss": 0.0763, + "step": 15084 + }, + { + "epoch": 1.98, + "learning_rate": 1.7630439290532616e-05, + "loss": 0.0576, + "step": 15085 + }, + { + "epoch": 1.98, + "learning_rate": 1.7630096505985544e-05, + "loss": 0.0584, + "step": 15086 + }, + { + "epoch": 1.98, + "learning_rate": 1.7629753699979232e-05, + "loss": 0.066, + "step": 15087 + }, + { + "epoch": 1.98, + "learning_rate": 1.7629410872514644e-05, + "loss": 0.0768, + "step": 15088 + }, + { + "epoch": 1.98, + "learning_rate": 1.7629068023592744e-05, + "loss": 0.0643, + "step": 15089 + }, + { + "epoch": 1.98, + "learning_rate": 1.7628725153214495e-05, + "loss": 0.0651, + "step": 15090 + }, + { + "epoch": 1.98, + "learning_rate": 1.7628382261380865e-05, + "loss": 0.0642, + "step": 15091 + }, + { + "epoch": 1.98, + "learning_rate": 1.7628039348092816e-05, + "loss": 0.0789, + "step": 15092 + }, + { + "epoch": 1.98, + "learning_rate": 1.7627696413351312e-05, + "loss": 0.0748, + "step": 15093 + }, + { + "epoch": 1.98, + "learning_rate": 1.762735345715732e-05, + "loss": 0.075, + "step": 15094 + }, + { + "epoch": 1.98, + "learning_rate": 1.76270104795118e-05, + "loss": 0.0717, + "step": 15095 + }, + { + "epoch": 1.98, + "learning_rate": 1.7626667480415717e-05, + "loss": 0.0717, + "step": 15096 + }, + { + "epoch": 1.98, + "learning_rate": 1.762632445987004e-05, + "loss": 0.0623, + "step": 15097 + }, + { + "epoch": 1.98, + "learning_rate": 1.7625981417875734e-05, + "loss": 0.0668, + "step": 15098 + }, + { + "epoch": 1.98, + "learning_rate": 1.762563835443376e-05, + "loss": 0.0745, + "step": 15099 + }, + { + "epoch": 1.98, + "learning_rate": 1.7625295269545083e-05, + "loss": 0.0633, + "step": 15100 + }, + { + "epoch": 1.98, + "learning_rate": 1.762495216321067e-05, + "loss": 0.0528, + "step": 15101 + }, + { + "epoch": 1.98, + "learning_rate": 1.7624609035431485e-05, + "loss": 0.0697, + "step": 15102 + }, + { + "epoch": 1.98, + "learning_rate": 1.7624265886208493e-05, + "loss": 0.0684, + "step": 15103 + }, + { + "epoch": 1.98, + "learning_rate": 1.7623922715542662e-05, + "loss": 0.0773, + "step": 15104 + }, + { + "epoch": 1.98, + "learning_rate": 1.7623579523434948e-05, + "loss": 0.0739, + "step": 15105 + }, + { + "epoch": 1.98, + "learning_rate": 1.7623236309886326e-05, + "loss": 0.0717, + "step": 15106 + }, + { + "epoch": 1.98, + "learning_rate": 1.762289307489776e-05, + "loss": 0.0675, + "step": 15107 + }, + { + "epoch": 1.98, + "learning_rate": 1.762254981847021e-05, + "loss": 0.0697, + "step": 15108 + }, + { + "epoch": 1.98, + "learning_rate": 1.762220654060465e-05, + "loss": 0.0779, + "step": 15109 + }, + { + "epoch": 1.98, + "learning_rate": 1.762186324130203e-05, + "loss": 0.0695, + "step": 15110 + }, + { + "epoch": 1.98, + "learning_rate": 1.7621519920563335e-05, + "loss": 0.0721, + "step": 15111 + }, + { + "epoch": 1.98, + "learning_rate": 1.7621176578389517e-05, + "loss": 0.0751, + "step": 15112 + }, + { + "epoch": 1.98, + "learning_rate": 1.7620833214781544e-05, + "loss": 0.0668, + "step": 15113 + }, + { + "epoch": 1.98, + "learning_rate": 1.762048982974039e-05, + "loss": 0.0646, + "step": 15114 + }, + { + "epoch": 1.98, + "learning_rate": 1.7620146423267006e-05, + "loss": 0.0718, + "step": 15115 + }, + { + "epoch": 1.98, + "learning_rate": 1.761980299536237e-05, + "loss": 0.061, + "step": 15116 + }, + { + "epoch": 1.98, + "learning_rate": 1.7619459546027444e-05, + "loss": 0.0746, + "step": 15117 + }, + { + "epoch": 1.98, + "learning_rate": 1.7619116075263193e-05, + "loss": 0.071, + "step": 15118 + }, + { + "epoch": 1.98, + "learning_rate": 1.7618772583070584e-05, + "loss": 0.0762, + "step": 15119 + }, + { + "epoch": 1.98, + "learning_rate": 1.7618429069450586e-05, + "loss": 0.0688, + "step": 15120 + }, + { + "epoch": 1.98, + "learning_rate": 1.7618085534404158e-05, + "loss": 0.0797, + "step": 15121 + }, + { + "epoch": 1.98, + "learning_rate": 1.761774197793227e-05, + "loss": 0.07, + "step": 15122 + }, + { + "epoch": 1.98, + "learning_rate": 1.761739840003589e-05, + "loss": 0.0805, + "step": 15123 + }, + { + "epoch": 1.98, + "learning_rate": 1.761705480071598e-05, + "loss": 0.0657, + "step": 15124 + }, + { + "epoch": 1.98, + "learning_rate": 1.7616711179973512e-05, + "loss": 0.0668, + "step": 15125 + }, + { + "epoch": 1.98, + "learning_rate": 1.7616367537809447e-05, + "loss": 0.0763, + "step": 15126 + }, + { + "epoch": 1.98, + "learning_rate": 1.7616023874224756e-05, + "loss": 0.0652, + "step": 15127 + }, + { + "epoch": 1.98, + "learning_rate": 1.7615680189220405e-05, + "loss": 0.0754, + "step": 15128 + }, + { + "epoch": 1.98, + "learning_rate": 1.7615336482797358e-05, + "loss": 0.0687, + "step": 15129 + }, + { + "epoch": 1.98, + "learning_rate": 1.7614992754956578e-05, + "loss": 0.0791, + "step": 15130 + }, + { + "epoch": 1.98, + "learning_rate": 1.7614649005699044e-05, + "loss": 0.0765, + "step": 15131 + }, + { + "epoch": 1.98, + "learning_rate": 1.761430523502571e-05, + "loss": 0.0704, + "step": 15132 + }, + { + "epoch": 1.98, + "learning_rate": 1.7613961442937548e-05, + "loss": 0.0779, + "step": 15133 + }, + { + "epoch": 1.98, + "learning_rate": 1.7613617629435523e-05, + "loss": 0.0661, + "step": 15134 + }, + { + "epoch": 1.98, + "learning_rate": 1.761327379452061e-05, + "loss": 0.0657, + "step": 15135 + }, + { + "epoch": 1.98, + "learning_rate": 1.761292993819377e-05, + "loss": 0.0662, + "step": 15136 + }, + { + "epoch": 1.98, + "learning_rate": 1.7612586060455963e-05, + "loss": 0.0892, + "step": 15137 + }, + { + "epoch": 1.98, + "learning_rate": 1.7612242161308167e-05, + "loss": 0.0799, + "step": 15138 + }, + { + "epoch": 1.98, + "learning_rate": 1.7611898240751342e-05, + "loss": 0.0719, + "step": 15139 + }, + { + "epoch": 1.98, + "learning_rate": 1.7611554298786468e-05, + "loss": 0.0783, + "step": 15140 + }, + { + "epoch": 1.98, + "learning_rate": 1.7611210335414496e-05, + "loss": 0.0813, + "step": 15141 + }, + { + "epoch": 1.98, + "learning_rate": 1.7610866350636398e-05, + "loss": 0.0658, + "step": 15142 + }, + { + "epoch": 1.98, + "learning_rate": 1.7610522344453148e-05, + "loss": 0.069, + "step": 15143 + }, + { + "epoch": 1.98, + "learning_rate": 1.7610178316865708e-05, + "loss": 0.069, + "step": 15144 + }, + { + "epoch": 1.98, + "learning_rate": 1.7609834267875045e-05, + "loss": 0.0777, + "step": 15145 + }, + { + "epoch": 1.98, + "learning_rate": 1.760949019748213e-05, + "loss": 0.076, + "step": 15146 + }, + { + "epoch": 1.98, + "learning_rate": 1.760914610568793e-05, + "loss": 0.0709, + "step": 15147 + }, + { + "epoch": 1.98, + "learning_rate": 1.760880199249341e-05, + "loss": 0.0713, + "step": 15148 + }, + { + "epoch": 1.98, + "learning_rate": 1.7608457857899538e-05, + "loss": 0.0734, + "step": 15149 + }, + { + "epoch": 1.98, + "learning_rate": 1.7608113701907286e-05, + "loss": 0.0607, + "step": 15150 + }, + { + "epoch": 1.98, + "learning_rate": 1.760776952451762e-05, + "loss": 0.0776, + "step": 15151 + }, + { + "epoch": 1.98, + "learning_rate": 1.7607425325731506e-05, + "loss": 0.0781, + "step": 15152 + }, + { + "epoch": 1.98, + "learning_rate": 1.7607081105549915e-05, + "loss": 0.0751, + "step": 15153 + }, + { + "epoch": 1.99, + "learning_rate": 1.760673686397381e-05, + "loss": 0.073, + "step": 15154 + }, + { + "epoch": 1.99, + "learning_rate": 1.7606392601004167e-05, + "loss": 0.0724, + "step": 15155 + }, + { + "epoch": 1.99, + "learning_rate": 1.7606048316641948e-05, + "loss": 0.0801, + "step": 15156 + }, + { + "epoch": 1.99, + "learning_rate": 1.760570401088812e-05, + "loss": 0.0863, + "step": 15157 + }, + { + "epoch": 1.99, + "learning_rate": 1.760535968374366e-05, + "loss": 0.0713, + "step": 15158 + }, + { + "epoch": 1.99, + "learning_rate": 1.760501533520953e-05, + "loss": 0.0759, + "step": 15159 + }, + { + "epoch": 1.99, + "learning_rate": 1.7604670965286698e-05, + "loss": 0.0737, + "step": 15160 + }, + { + "epoch": 1.99, + "learning_rate": 1.760432657397613e-05, + "loss": 0.0685, + "step": 15161 + }, + { + "epoch": 1.99, + "learning_rate": 1.7603982161278804e-05, + "loss": 0.0807, + "step": 15162 + }, + { + "epoch": 1.99, + "learning_rate": 1.760363772719568e-05, + "loss": 0.0828, + "step": 15163 + }, + { + "epoch": 1.99, + "learning_rate": 1.760329327172773e-05, + "loss": 0.0702, + "step": 15164 + }, + { + "epoch": 1.99, + "learning_rate": 1.7602948794875926e-05, + "loss": 0.0792, + "step": 15165 + }, + { + "epoch": 1.99, + "learning_rate": 1.760260429664123e-05, + "loss": 0.0686, + "step": 15166 + }, + { + "epoch": 1.99, + "learning_rate": 1.7602259777024614e-05, + "loss": 0.0737, + "step": 15167 + }, + { + "epoch": 1.99, + "learning_rate": 1.7601915236027048e-05, + "loss": 0.0837, + "step": 15168 + }, + { + "epoch": 1.99, + "learning_rate": 1.7601570673649503e-05, + "loss": 0.0793, + "step": 15169 + }, + { + "epoch": 1.99, + "learning_rate": 1.760122608989294e-05, + "loss": 0.0705, + "step": 15170 + }, + { + "epoch": 1.99, + "learning_rate": 1.760088148475834e-05, + "loss": 0.0755, + "step": 15171 + }, + { + "epoch": 1.99, + "learning_rate": 1.760053685824666e-05, + "loss": 0.0663, + "step": 15172 + }, + { + "epoch": 1.99, + "learning_rate": 1.7600192210358877e-05, + "loss": 0.0716, + "step": 15173 + }, + { + "epoch": 1.99, + "learning_rate": 1.759984754109596e-05, + "loss": 0.0661, + "step": 15174 + }, + { + "epoch": 1.99, + "learning_rate": 1.7599502850458872e-05, + "loss": 0.0755, + "step": 15175 + }, + { + "epoch": 1.99, + "learning_rate": 1.759915813844859e-05, + "loss": 0.0676, + "step": 15176 + }, + { + "epoch": 1.99, + "learning_rate": 1.759881340506608e-05, + "loss": 0.0693, + "step": 15177 + }, + { + "epoch": 1.99, + "learning_rate": 1.7598468650312313e-05, + "loss": 0.0689, + "step": 15178 + }, + { + "epoch": 1.99, + "learning_rate": 1.7598123874188254e-05, + "loss": 0.0727, + "step": 15179 + }, + { + "epoch": 1.99, + "learning_rate": 1.759777907669488e-05, + "loss": 0.0726, + "step": 15180 + }, + { + "epoch": 1.99, + "learning_rate": 1.7597434257833158e-05, + "loss": 0.0792, + "step": 15181 + }, + { + "epoch": 1.99, + "learning_rate": 1.7597089417604053e-05, + "loss": 0.0639, + "step": 15182 + }, + { + "epoch": 1.99, + "learning_rate": 1.759674455600854e-05, + "loss": 0.0712, + "step": 15183 + }, + { + "epoch": 1.99, + "learning_rate": 1.7596399673047586e-05, + "loss": 0.0696, + "step": 15184 + }, + { + "epoch": 1.99, + "learning_rate": 1.7596054768722163e-05, + "loss": 0.0802, + "step": 15185 + }, + { + "epoch": 1.99, + "learning_rate": 1.759570984303324e-05, + "loss": 0.0806, + "step": 15186 + }, + { + "epoch": 1.99, + "learning_rate": 1.759536489598179e-05, + "loss": 0.0741, + "step": 15187 + }, + { + "epoch": 1.99, + "learning_rate": 1.759501992756878e-05, + "loss": 0.0757, + "step": 15188 + }, + { + "epoch": 1.99, + "learning_rate": 1.7594674937795182e-05, + "loss": 0.0659, + "step": 15189 + }, + { + "epoch": 1.99, + "learning_rate": 1.7594329926661964e-05, + "loss": 0.0646, + "step": 15190 + }, + { + "epoch": 1.99, + "learning_rate": 1.7593984894170094e-05, + "loss": 0.0633, + "step": 15191 + }, + { + "epoch": 1.99, + "learning_rate": 1.759363984032055e-05, + "loss": 0.0854, + "step": 15192 + }, + { + "epoch": 1.99, + "learning_rate": 1.7593294765114297e-05, + "loss": 0.07, + "step": 15193 + }, + { + "epoch": 1.99, + "learning_rate": 1.7592949668552307e-05, + "loss": 0.0733, + "step": 15194 + }, + { + "epoch": 1.99, + "learning_rate": 1.7592604550635552e-05, + "loss": 0.0665, + "step": 15195 + }, + { + "epoch": 1.99, + "learning_rate": 1.7592259411365e-05, + "loss": 0.0718, + "step": 15196 + }, + { + "epoch": 1.99, + "learning_rate": 1.7591914250741622e-05, + "loss": 0.0749, + "step": 15197 + }, + { + "epoch": 1.99, + "learning_rate": 1.759156906876639e-05, + "loss": 0.0842, + "step": 15198 + }, + { + "epoch": 1.99, + "learning_rate": 1.7591223865440272e-05, + "loss": 0.0689, + "step": 15199 + }, + { + "epoch": 1.99, + "learning_rate": 1.7590878640764244e-05, + "loss": 0.0676, + "step": 15200 + }, + { + "epoch": 1.99, + "learning_rate": 1.7590533394739275e-05, + "loss": 0.07, + "step": 15201 + }, + { + "epoch": 1.99, + "learning_rate": 1.7590188127366335e-05, + "loss": 0.0779, + "step": 15202 + }, + { + "epoch": 1.99, + "learning_rate": 1.758984283864639e-05, + "loss": 0.062, + "step": 15203 + }, + { + "epoch": 1.99, + "learning_rate": 1.758949752858042e-05, + "loss": 0.0699, + "step": 15204 + }, + { + "epoch": 1.99, + "learning_rate": 1.758915219716939e-05, + "loss": 0.0778, + "step": 15205 + }, + { + "epoch": 1.99, + "learning_rate": 1.7588806844414277e-05, + "loss": 0.0775, + "step": 15206 + }, + { + "epoch": 1.99, + "learning_rate": 1.758846147031605e-05, + "loss": 0.0664, + "step": 15207 + }, + { + "epoch": 1.99, + "learning_rate": 1.7588116074875675e-05, + "loss": 0.0749, + "step": 15208 + }, + { + "epoch": 1.99, + "learning_rate": 1.758777065809413e-05, + "loss": 0.0682, + "step": 15209 + }, + { + "epoch": 1.99, + "learning_rate": 1.758742521997238e-05, + "loss": 0.0679, + "step": 15210 + }, + { + "epoch": 1.99, + "learning_rate": 1.7587079760511407e-05, + "loss": 0.0746, + "step": 15211 + }, + { + "epoch": 1.99, + "learning_rate": 1.7586734279712172e-05, + "loss": 0.0671, + "step": 15212 + }, + { + "epoch": 1.99, + "learning_rate": 1.7586388777575655e-05, + "loss": 0.0749, + "step": 15213 + }, + { + "epoch": 1.99, + "learning_rate": 1.7586043254102818e-05, + "loss": 0.0773, + "step": 15214 + }, + { + "epoch": 1.99, + "learning_rate": 1.758569770929464e-05, + "loss": 0.0752, + "step": 15215 + }, + { + "epoch": 1.99, + "learning_rate": 1.7585352143152093e-05, + "loss": 0.0633, + "step": 15216 + }, + { + "epoch": 1.99, + "learning_rate": 1.7585006555676146e-05, + "loss": 0.0655, + "step": 15217 + }, + { + "epoch": 1.99, + "learning_rate": 1.7584660946867773e-05, + "loss": 0.0701, + "step": 15218 + }, + { + "epoch": 1.99, + "learning_rate": 1.7584315316727945e-05, + "loss": 0.0683, + "step": 15219 + }, + { + "epoch": 1.99, + "learning_rate": 1.7583969665257632e-05, + "loss": 0.0833, + "step": 15220 + }, + { + "epoch": 1.99, + "learning_rate": 1.758362399245781e-05, + "loss": 0.0591, + "step": 15221 + }, + { + "epoch": 1.99, + "learning_rate": 1.758327829832945e-05, + "loss": 0.0727, + "step": 15222 + }, + { + "epoch": 1.99, + "learning_rate": 1.7582932582873524e-05, + "loss": 0.0728, + "step": 15223 + }, + { + "epoch": 1.99, + "learning_rate": 1.7582586846090998e-05, + "loss": 0.08, + "step": 15224 + }, + { + "epoch": 1.99, + "learning_rate": 1.7582241087982855e-05, + "loss": 0.0725, + "step": 15225 + }, + { + "epoch": 1.99, + "learning_rate": 1.7581895308550062e-05, + "loss": 0.0642, + "step": 15226 + }, + { + "epoch": 1.99, + "learning_rate": 1.7581549507793594e-05, + "loss": 0.0751, + "step": 15227 + }, + { + "epoch": 1.99, + "learning_rate": 1.758120368571442e-05, + "loss": 0.0771, + "step": 15228 + }, + { + "epoch": 1.99, + "learning_rate": 1.7580857842313516e-05, + "loss": 0.0789, + "step": 15229 + }, + { + "epoch": 2.0, + "learning_rate": 1.758051197759185e-05, + "loss": 0.0685, + "step": 15230 + }, + { + "epoch": 2.0, + "learning_rate": 1.75801660915504e-05, + "loss": 0.072, + "step": 15231 + }, + { + "epoch": 2.0, + "learning_rate": 1.7579820184190135e-05, + "loss": 0.0687, + "step": 15232 + }, + { + "epoch": 2.0, + "learning_rate": 1.757947425551203e-05, + "loss": 0.0749, + "step": 15233 + }, + { + "epoch": 2.0, + "learning_rate": 1.7579128305517058e-05, + "loss": 0.0763, + "step": 15234 + }, + { + "epoch": 2.0, + "learning_rate": 1.7578782334206193e-05, + "loss": 0.0635, + "step": 15235 + }, + { + "epoch": 2.0, + "learning_rate": 1.75784363415804e-05, + "loss": 0.0615, + "step": 15236 + }, + { + "epoch": 2.0, + "learning_rate": 1.7578090327640665e-05, + "loss": 0.0802, + "step": 15237 + }, + { + "epoch": 2.0, + "learning_rate": 1.7577744292387953e-05, + "loss": 0.0662, + "step": 15238 + }, + { + "epoch": 2.0, + "learning_rate": 1.757739823582324e-05, + "loss": 0.0765, + "step": 15239 + }, + { + "epoch": 2.0, + "learning_rate": 1.7577052157947494e-05, + "loss": 0.0734, + "step": 15240 + }, + { + "epoch": 2.0, + "learning_rate": 1.757670605876169e-05, + "loss": 0.0678, + "step": 15241 + }, + { + "epoch": 2.0, + "learning_rate": 1.7576359938266813e-05, + "loss": 0.0677, + "step": 15242 + }, + { + "epoch": 2.0, + "learning_rate": 1.7576013796463818e-05, + "loss": 0.0695, + "step": 15243 + }, + { + "epoch": 2.0, + "learning_rate": 1.7575667633353692e-05, + "loss": 0.0767, + "step": 15244 + }, + { + "epoch": 2.0, + "learning_rate": 1.7575321448937405e-05, + "loss": 0.0639, + "step": 15245 + }, + { + "epoch": 2.0, + "learning_rate": 1.7574975243215928e-05, + "loss": 0.0764, + "step": 15246 + }, + { + "epoch": 2.0, + "learning_rate": 1.7574629016190237e-05, + "loss": 0.075, + "step": 15247 + }, + { + "epoch": 2.0, + "learning_rate": 1.7574282767861306e-05, + "loss": 0.072, + "step": 15248 + }, + { + "epoch": 2.0, + "learning_rate": 1.7573936498230105e-05, + "loss": 0.0734, + "step": 15249 + }, + { + "epoch": 2.0, + "learning_rate": 1.7573590207297616e-05, + "loss": 0.0644, + "step": 15250 + }, + { + "epoch": 2.0, + "learning_rate": 1.75732438950648e-05, + "loss": 0.066, + "step": 15251 + }, + { + "epoch": 2.0, + "learning_rate": 1.7572897561532646e-05, + "loss": 0.0667, + "step": 15252 + }, + { + "epoch": 2.0, + "learning_rate": 1.757255120670212e-05, + "loss": 0.0765, + "step": 15253 + }, + { + "epoch": 2.0, + "learning_rate": 1.7572204830574198e-05, + "loss": 0.0789, + "step": 15254 + }, + { + "epoch": 2.0, + "learning_rate": 1.757185843314985e-05, + "loss": 0.0718, + "step": 15255 + }, + { + "epoch": 2.0, + "learning_rate": 1.7571512014430056e-05, + "loss": 0.0733, + "step": 15256 + }, + { + "epoch": 2.0, + "learning_rate": 1.7571165574415786e-05, + "loss": 0.0747, + "step": 15257 + }, + { + "epoch": 2.0, + "learning_rate": 1.7570819113108016e-05, + "loss": 0.0788, + "step": 15258 + }, + { + "epoch": 2.0, + "learning_rate": 1.757047263050772e-05, + "loss": 0.067, + "step": 15259 + }, + { + "epoch": 2.0, + "learning_rate": 1.7570126126615875e-05, + "loss": 0.0541, + "step": 15260 + }, + { + "epoch": 2.0, + "learning_rate": 1.756977960143345e-05, + "loss": 0.0885, + "step": 15261 + }, + { + "epoch": 2.0, + "learning_rate": 1.7569433054961425e-05, + "loss": 0.0764, + "step": 15262 + }, + { + "epoch": 2.0, + "learning_rate": 1.756908648720077e-05, + "loss": 0.073, + "step": 15263 + }, + { + "epoch": 2.0, + "learning_rate": 1.7568739898152466e-05, + "loss": 0.0715, + "step": 15264 + }, + { + "epoch": 2.0, + "learning_rate": 1.7568393287817486e-05, + "loss": 0.0656, + "step": 15265 + }, + { + "epoch": 2.0, + "learning_rate": 1.7568046656196798e-05, + "loss": 0.0905, + "step": 15266 + }, + { + "epoch": 2.0, + "learning_rate": 1.756770000329138e-05, + "loss": 0.0797, + "step": 15267 + }, + { + "epoch": 2.0, + "learning_rate": 1.7567353329102216e-05, + "loss": 0.0644, + "step": 15268 + }, + { + "epoch": 2.0, + "learning_rate": 1.7567006633630267e-05, + "loss": 0.0603, + "step": 15269 + }, + { + "epoch": 2.0, + "learning_rate": 1.756665991687652e-05, + "loss": 0.0687, + "step": 15270 + }, + { + "epoch": 2.0, + "learning_rate": 1.7566313178841944e-05, + "loss": 0.0748, + "step": 15271 + }, + { + "epoch": 2.0, + "learning_rate": 1.7565966419527512e-05, + "loss": 0.068, + "step": 15272 + }, + { + "epoch": 2.0, + "learning_rate": 1.7565619638934203e-05, + "loss": 0.0718, + "step": 15273 + }, + { + "epoch": 2.0, + "learning_rate": 1.7565272837062993e-05, + "loss": 0.0709, + "step": 15274 + }, + { + "epoch": 2.0, + "learning_rate": 1.7564926013914857e-05, + "loss": 0.0688, + "step": 15275 + }, + { + "epoch": 2.0, + "learning_rate": 1.7564579169490768e-05, + "loss": 0.0761, + "step": 15276 + }, + { + "epoch": 2.0, + "learning_rate": 1.7564232303791705e-05, + "loss": 0.0645, + "step": 15277 + }, + { + "epoch": 2.0, + "learning_rate": 1.7563885416818637e-05, + "loss": 0.0692, + "step": 15278 + }, + { + "epoch": 2.0, + "learning_rate": 1.7563538508572547e-05, + "loss": 0.0738, + "step": 15279 + }, + { + "epoch": 2.0, + "learning_rate": 1.756319157905441e-05, + "loss": 0.062, + "step": 15280 + }, + { + "epoch": 2.0, + "learning_rate": 1.7562844628265195e-05, + "loss": 0.0744, + "step": 15281 + }, + { + "epoch": 2.0, + "learning_rate": 1.756249765620589e-05, + "loss": 0.0636, + "step": 15282 + }, + { + "epoch": 2.0, + "learning_rate": 1.7562150662877453e-05, + "loss": 0.0651, + "step": 15283 + }, + { + "epoch": 2.0, + "learning_rate": 1.7561803648280874e-05, + "loss": 0.0642, + "step": 15284 + }, + { + "epoch": 2.0, + "learning_rate": 1.7561456612417127e-05, + "loss": 0.0607, + "step": 15285 + }, + { + "epoch": 2.0, + "learning_rate": 1.7561109555287185e-05, + "loss": 0.0677, + "step": 15286 + }, + { + "epoch": 2.0, + "learning_rate": 1.7560762476892023e-05, + "loss": 0.0674, + "step": 15287 + }, + { + "epoch": 2.0, + "learning_rate": 1.7560415377232618e-05, + "loss": 0.0623, + "step": 15288 + }, + { + "epoch": 2.0, + "learning_rate": 1.756006825630995e-05, + "loss": 0.0621, + "step": 15289 + }, + { + "epoch": 2.0, + "learning_rate": 1.7559721114124996e-05, + "loss": 0.0565, + "step": 15290 + }, + { + "epoch": 2.0, + "learning_rate": 1.755937395067872e-05, + "loss": 0.072, + "step": 15291 + }, + { + "epoch": 2.0, + "learning_rate": 1.7559026765972115e-05, + "loss": 0.0634, + "step": 15292 + }, + { + "epoch": 2.0, + "learning_rate": 1.7558679560006147e-05, + "loss": 0.0768, + "step": 15293 + }, + { + "epoch": 2.0, + "learning_rate": 1.7558332332781795e-05, + "loss": 0.059, + "step": 15294 + }, + { + "epoch": 2.0, + "learning_rate": 1.755798508430004e-05, + "loss": 0.0611, + "step": 15295 + }, + { + "epoch": 2.0, + "learning_rate": 1.755763781456185e-05, + "loss": 0.0633, + "step": 15296 + }, + { + "epoch": 2.0, + "learning_rate": 1.7557290523568203e-05, + "loss": 0.069, + "step": 15297 + }, + { + "epoch": 2.0, + "learning_rate": 1.755694321132009e-05, + "loss": 0.0693, + "step": 15298 + }, + { + "epoch": 2.0, + "learning_rate": 1.7556595877818465e-05, + "loss": 0.0684, + "step": 15299 + }, + { + "epoch": 2.0, + "learning_rate": 1.7556248523064323e-05, + "loss": 0.0611, + "step": 15300 + }, + { + "epoch": 2.0, + "learning_rate": 1.7555901147058633e-05, + "loss": 0.0679, + "step": 15301 + }, + { + "epoch": 2.0, + "learning_rate": 1.755555374980237e-05, + "loss": 0.0639, + "step": 15302 + }, + { + "epoch": 2.0, + "learning_rate": 1.7555206331296518e-05, + "loss": 0.0684, + "step": 15303 + }, + { + "epoch": 2.0, + "learning_rate": 1.7554858891542047e-05, + "loss": 0.0684, + "step": 15304 + }, + { + "epoch": 2.0, + "learning_rate": 1.7554511430539945e-05, + "loss": 0.0732, + "step": 15305 + }, + { + "epoch": 2.0, + "learning_rate": 1.7554163948291177e-05, + "loss": 0.0754, + "step": 15306 + }, + { + "epoch": 2.01, + "learning_rate": 1.7553816444796725e-05, + "loss": 0.0669, + "step": 15307 + }, + { + "epoch": 2.01, + "learning_rate": 1.7553468920057568e-05, + "loss": 0.0661, + "step": 15308 + }, + { + "epoch": 2.01, + "learning_rate": 1.755312137407468e-05, + "loss": 0.0712, + "step": 15309 + }, + { + "epoch": 2.01, + "learning_rate": 1.7552773806849042e-05, + "loss": 0.0702, + "step": 15310 + }, + { + "epoch": 2.01, + "learning_rate": 1.7552426218381627e-05, + "loss": 0.0719, + "step": 15311 + }, + { + "epoch": 2.01, + "learning_rate": 1.755207860867342e-05, + "loss": 0.0745, + "step": 15312 + }, + { + "epoch": 2.01, + "learning_rate": 1.7551730977725393e-05, + "loss": 0.0742, + "step": 15313 + }, + { + "epoch": 2.01, + "learning_rate": 1.7551383325538522e-05, + "loss": 0.0616, + "step": 15314 + }, + { + "epoch": 2.01, + "learning_rate": 1.7551035652113787e-05, + "loss": 0.0693, + "step": 15315 + }, + { + "epoch": 2.01, + "learning_rate": 1.755068795745217e-05, + "loss": 0.068, + "step": 15316 + }, + { + "epoch": 2.01, + "learning_rate": 1.7550340241554643e-05, + "loss": 0.0701, + "step": 15317 + }, + { + "epoch": 2.01, + "learning_rate": 1.7549992504422185e-05, + "loss": 0.0763, + "step": 15318 + }, + { + "epoch": 2.01, + "learning_rate": 1.7549644746055777e-05, + "loss": 0.0676, + "step": 15319 + }, + { + "epoch": 2.01, + "learning_rate": 1.7549296966456398e-05, + "loss": 0.071, + "step": 15320 + }, + { + "epoch": 2.01, + "learning_rate": 1.754894916562502e-05, + "loss": 0.0839, + "step": 15321 + }, + { + "epoch": 2.01, + "learning_rate": 1.7548601343562624e-05, + "loss": 0.0713, + "step": 15322 + }, + { + "epoch": 2.01, + "learning_rate": 1.754825350027019e-05, + "loss": 0.0598, + "step": 15323 + }, + { + "epoch": 2.01, + "learning_rate": 1.7547905635748695e-05, + "loss": 0.08, + "step": 15324 + }, + { + "epoch": 2.01, + "learning_rate": 1.7547557749999116e-05, + "loss": 0.0731, + "step": 15325 + }, + { + "epoch": 2.01, + "learning_rate": 1.7547209843022434e-05, + "loss": 0.0681, + "step": 15326 + }, + { + "epoch": 2.01, + "learning_rate": 1.7546861914819628e-05, + "loss": 0.0626, + "step": 15327 + }, + { + "epoch": 2.01, + "learning_rate": 1.754651396539167e-05, + "loss": 0.0757, + "step": 15328 + }, + { + "epoch": 2.01, + "learning_rate": 1.7546165994739548e-05, + "loss": 0.0726, + "step": 15329 + }, + { + "epoch": 2.01, + "learning_rate": 1.7545818002864235e-05, + "loss": 0.0702, + "step": 15330 + }, + { + "epoch": 2.01, + "learning_rate": 1.7545469989766708e-05, + "loss": 0.0756, + "step": 15331 + }, + { + "epoch": 2.01, + "learning_rate": 1.7545121955447953e-05, + "loss": 0.0771, + "step": 15332 + }, + { + "epoch": 2.01, + "learning_rate": 1.754477389990894e-05, + "loss": 0.0539, + "step": 15333 + }, + { + "epoch": 2.01, + "learning_rate": 1.7544425823150653e-05, + "loss": 0.0623, + "step": 15334 + }, + { + "epoch": 2.01, + "learning_rate": 1.7544077725174072e-05, + "loss": 0.0751, + "step": 15335 + }, + { + "epoch": 2.01, + "learning_rate": 1.7543729605980173e-05, + "loss": 0.078, + "step": 15336 + }, + { + "epoch": 2.01, + "learning_rate": 1.7543381465569935e-05, + "loss": 0.0647, + "step": 15337 + }, + { + "epoch": 2.01, + "learning_rate": 1.7543033303944337e-05, + "loss": 0.0798, + "step": 15338 + }, + { + "epoch": 2.01, + "learning_rate": 1.7542685121104366e-05, + "loss": 0.076, + "step": 15339 + }, + { + "epoch": 2.01, + "learning_rate": 1.754233691705099e-05, + "loss": 0.0685, + "step": 15340 + }, + { + "epoch": 2.01, + "learning_rate": 1.754198869178519e-05, + "loss": 0.0757, + "step": 15341 + }, + { + "epoch": 2.01, + "learning_rate": 1.7541640445307956e-05, + "loss": 0.0639, + "step": 15342 + }, + { + "epoch": 2.01, + "learning_rate": 1.7541292177620257e-05, + "loss": 0.075, + "step": 15343 + }, + { + "epoch": 2.01, + "learning_rate": 1.7540943888723073e-05, + "loss": 0.0654, + "step": 15344 + }, + { + "epoch": 2.01, + "learning_rate": 1.7540595578617387e-05, + "loss": 0.0692, + "step": 15345 + }, + { + "epoch": 2.01, + "learning_rate": 1.7540247247304178e-05, + "loss": 0.0685, + "step": 15346 + }, + { + "epoch": 2.01, + "learning_rate": 1.753989889478443e-05, + "loss": 0.0811, + "step": 15347 + }, + { + "epoch": 2.01, + "learning_rate": 1.753955052105911e-05, + "loss": 0.0711, + "step": 15348 + }, + { + "epoch": 2.01, + "learning_rate": 1.7539202126129208e-05, + "loss": 0.0693, + "step": 15349 + }, + { + "epoch": 2.01, + "learning_rate": 1.7538853709995702e-05, + "loss": 0.0674, + "step": 15350 + }, + { + "epoch": 2.01, + "learning_rate": 1.7538505272659572e-05, + "loss": 0.0712, + "step": 15351 + }, + { + "epoch": 2.01, + "learning_rate": 1.7538156814121798e-05, + "loss": 0.0621, + "step": 15352 + }, + { + "epoch": 2.01, + "learning_rate": 1.7537808334383358e-05, + "loss": 0.0709, + "step": 15353 + }, + { + "epoch": 2.01, + "learning_rate": 1.7537459833445235e-05, + "loss": 0.0666, + "step": 15354 + }, + { + "epoch": 2.01, + "learning_rate": 1.7537111311308405e-05, + "loss": 0.0692, + "step": 15355 + }, + { + "epoch": 2.01, + "learning_rate": 1.753676276797385e-05, + "loss": 0.0707, + "step": 15356 + }, + { + "epoch": 2.01, + "learning_rate": 1.7536414203442554e-05, + "loss": 0.0739, + "step": 15357 + }, + { + "epoch": 2.01, + "learning_rate": 1.7536065617715494e-05, + "loss": 0.0718, + "step": 15358 + }, + { + "epoch": 2.01, + "learning_rate": 1.753571701079365e-05, + "loss": 0.0654, + "step": 15359 + }, + { + "epoch": 2.01, + "learning_rate": 1.7535368382678007e-05, + "loss": 0.0718, + "step": 15360 + }, + { + "epoch": 2.01, + "learning_rate": 1.7535019733369538e-05, + "loss": 0.059, + "step": 15361 + }, + { + "epoch": 2.01, + "learning_rate": 1.7534671062869225e-05, + "loss": 0.0618, + "step": 15362 + }, + { + "epoch": 2.01, + "learning_rate": 1.7534322371178056e-05, + "loss": 0.0611, + "step": 15363 + }, + { + "epoch": 2.01, + "learning_rate": 1.7533973658297003e-05, + "loss": 0.0679, + "step": 15364 + }, + { + "epoch": 2.01, + "learning_rate": 1.753362492422705e-05, + "loss": 0.0499, + "step": 15365 + }, + { + "epoch": 2.01, + "learning_rate": 1.7533276168969183e-05, + "loss": 0.0686, + "step": 15366 + }, + { + "epoch": 2.01, + "learning_rate": 1.753292739252437e-05, + "loss": 0.0683, + "step": 15367 + }, + { + "epoch": 2.01, + "learning_rate": 1.753257859489361e-05, + "loss": 0.0653, + "step": 15368 + }, + { + "epoch": 2.01, + "learning_rate": 1.7532229776077865e-05, + "loss": 0.0803, + "step": 15369 + }, + { + "epoch": 2.01, + "learning_rate": 1.753188093607813e-05, + "loss": 0.0728, + "step": 15370 + }, + { + "epoch": 2.01, + "learning_rate": 1.7531532074895376e-05, + "loss": 0.0668, + "step": 15371 + }, + { + "epoch": 2.01, + "learning_rate": 1.7531183192530595e-05, + "loss": 0.0749, + "step": 15372 + }, + { + "epoch": 2.01, + "learning_rate": 1.7530834288984756e-05, + "loss": 0.0718, + "step": 15373 + }, + { + "epoch": 2.01, + "learning_rate": 1.753048536425885e-05, + "loss": 0.0632, + "step": 15374 + }, + { + "epoch": 2.01, + "learning_rate": 1.7530136418353857e-05, + "loss": 0.0654, + "step": 15375 + }, + { + "epoch": 2.01, + "learning_rate": 1.7529787451270756e-05, + "loss": 0.072, + "step": 15376 + }, + { + "epoch": 2.01, + "learning_rate": 1.7529438463010526e-05, + "loss": 0.0714, + "step": 15377 + }, + { + "epoch": 2.01, + "learning_rate": 1.752908945357415e-05, + "loss": 0.0605, + "step": 15378 + }, + { + "epoch": 2.01, + "learning_rate": 1.7528740422962612e-05, + "loss": 0.0717, + "step": 15379 + }, + { + "epoch": 2.01, + "learning_rate": 1.7528391371176897e-05, + "loss": 0.0645, + "step": 15380 + }, + { + "epoch": 2.01, + "learning_rate": 1.7528042298217977e-05, + "loss": 0.0643, + "step": 15381 + }, + { + "epoch": 2.01, + "learning_rate": 1.7527693204086838e-05, + "loss": 0.0588, + "step": 15382 + }, + { + "epoch": 2.02, + "learning_rate": 1.752734408878447e-05, + "loss": 0.0764, + "step": 15383 + }, + { + "epoch": 2.02, + "learning_rate": 1.752699495231184e-05, + "loss": 0.0677, + "step": 15384 + }, + { + "epoch": 2.02, + "learning_rate": 1.7526645794669942e-05, + "loss": 0.0528, + "step": 15385 + }, + { + "epoch": 2.02, + "learning_rate": 1.7526296615859754e-05, + "loss": 0.0702, + "step": 15386 + }, + { + "epoch": 2.02, + "learning_rate": 1.7525947415882253e-05, + "loss": 0.0713, + "step": 15387 + }, + { + "epoch": 2.02, + "learning_rate": 1.752559819473843e-05, + "loss": 0.0731, + "step": 15388 + }, + { + "epoch": 2.02, + "learning_rate": 1.752524895242926e-05, + "loss": 0.0634, + "step": 15389 + }, + { + "epoch": 2.02, + "learning_rate": 1.752489968895573e-05, + "loss": 0.0615, + "step": 15390 + }, + { + "epoch": 2.02, + "learning_rate": 1.7524550404318822e-05, + "loss": 0.0717, + "step": 15391 + }, + { + "epoch": 2.02, + "learning_rate": 1.752420109851951e-05, + "loss": 0.0659, + "step": 15392 + }, + { + "epoch": 2.02, + "learning_rate": 1.7523851771558787e-05, + "loss": 0.0585, + "step": 15393 + }, + { + "epoch": 2.02, + "learning_rate": 1.7523502423437637e-05, + "loss": 0.0676, + "step": 15394 + }, + { + "epoch": 2.02, + "learning_rate": 1.752315305415703e-05, + "loss": 0.0664, + "step": 15395 + }, + { + "epoch": 2.02, + "learning_rate": 1.752280366371796e-05, + "loss": 0.0574, + "step": 15396 + }, + { + "epoch": 2.02, + "learning_rate": 1.75224542521214e-05, + "loss": 0.0744, + "step": 15397 + }, + { + "epoch": 2.02, + "learning_rate": 1.7522104819368343e-05, + "loss": 0.0672, + "step": 15398 + }, + { + "epoch": 2.02, + "learning_rate": 1.7521755365459764e-05, + "loss": 0.0684, + "step": 15399 + }, + { + "epoch": 2.02, + "learning_rate": 1.752140589039665e-05, + "loss": 0.0688, + "step": 15400 + }, + { + "epoch": 2.02, + "learning_rate": 1.752105639417998e-05, + "loss": 0.074, + "step": 15401 + }, + { + "epoch": 2.02, + "learning_rate": 1.7520706876810742e-05, + "loss": 0.0703, + "step": 15402 + }, + { + "epoch": 2.02, + "learning_rate": 1.7520357338289917e-05, + "loss": 0.0709, + "step": 15403 + }, + { + "epoch": 2.02, + "learning_rate": 1.7520007778618487e-05, + "loss": 0.0658, + "step": 15404 + }, + { + "epoch": 2.02, + "learning_rate": 1.7519658197797432e-05, + "loss": 0.0625, + "step": 15405 + }, + { + "epoch": 2.02, + "learning_rate": 1.751930859582774e-05, + "loss": 0.0679, + "step": 15406 + }, + { + "epoch": 2.02, + "learning_rate": 1.7518958972710398e-05, + "loss": 0.0639, + "step": 15407 + }, + { + "epoch": 2.02, + "learning_rate": 1.751860932844638e-05, + "loss": 0.0625, + "step": 15408 + }, + { + "epoch": 2.02, + "learning_rate": 1.7518259663036675e-05, + "loss": 0.0634, + "step": 15409 + }, + { + "epoch": 2.02, + "learning_rate": 1.7517909976482267e-05, + "loss": 0.0662, + "step": 15410 + }, + { + "epoch": 2.02, + "learning_rate": 1.7517560268784134e-05, + "loss": 0.0771, + "step": 15411 + }, + { + "epoch": 2.02, + "learning_rate": 1.7517210539943263e-05, + "loss": 0.067, + "step": 15412 + }, + { + "epoch": 2.02, + "learning_rate": 1.751686078996064e-05, + "loss": 0.0766, + "step": 15413 + }, + { + "epoch": 2.02, + "learning_rate": 1.7516511018837244e-05, + "loss": 0.0645, + "step": 15414 + }, + { + "epoch": 2.02, + "learning_rate": 1.7516161226574063e-05, + "loss": 0.063, + "step": 15415 + }, + { + "epoch": 2.02, + "learning_rate": 1.7515811413172076e-05, + "loss": 0.0705, + "step": 15416 + }, + { + "epoch": 2.02, + "learning_rate": 1.7515461578632274e-05, + "loss": 0.0646, + "step": 15417 + }, + { + "epoch": 2.02, + "learning_rate": 1.7515111722955632e-05, + "loss": 0.0578, + "step": 15418 + }, + { + "epoch": 2.02, + "learning_rate": 1.7514761846143143e-05, + "loss": 0.0728, + "step": 15419 + }, + { + "epoch": 2.02, + "learning_rate": 1.7514411948195782e-05, + "loss": 0.0699, + "step": 15420 + }, + { + "epoch": 2.02, + "learning_rate": 1.7514062029114543e-05, + "loss": 0.0681, + "step": 15421 + }, + { + "epoch": 2.02, + "learning_rate": 1.7513712088900398e-05, + "loss": 0.0538, + "step": 15422 + }, + { + "epoch": 2.02, + "learning_rate": 1.7513362127554343e-05, + "loss": 0.0595, + "step": 15423 + }, + { + "epoch": 2.02, + "learning_rate": 1.7513012145077356e-05, + "loss": 0.0689, + "step": 15424 + }, + { + "epoch": 2.02, + "learning_rate": 1.751266214147042e-05, + "loss": 0.0717, + "step": 15425 + }, + { + "epoch": 2.02, + "learning_rate": 1.7512312116734524e-05, + "loss": 0.0662, + "step": 15426 + }, + { + "epoch": 2.02, + "learning_rate": 1.751196207087065e-05, + "loss": 0.0633, + "step": 15427 + }, + { + "epoch": 2.02, + "learning_rate": 1.7511612003879783e-05, + "loss": 0.072, + "step": 15428 + }, + { + "epoch": 2.02, + "learning_rate": 1.7511261915762908e-05, + "loss": 0.0613, + "step": 15429 + }, + { + "epoch": 2.02, + "learning_rate": 1.7510911806521007e-05, + "loss": 0.0683, + "step": 15430 + }, + { + "epoch": 2.02, + "learning_rate": 1.7510561676155066e-05, + "loss": 0.0741, + "step": 15431 + }, + { + "epoch": 2.02, + "learning_rate": 1.7510211524666072e-05, + "loss": 0.0701, + "step": 15432 + }, + { + "epoch": 2.02, + "learning_rate": 1.7509861352055007e-05, + "loss": 0.0684, + "step": 15433 + }, + { + "epoch": 2.02, + "learning_rate": 1.7509511158322857e-05, + "loss": 0.0705, + "step": 15434 + }, + { + "epoch": 2.02, + "learning_rate": 1.750916094347061e-05, + "loss": 0.0648, + "step": 15435 + }, + { + "epoch": 2.02, + "learning_rate": 1.7508810707499243e-05, + "loss": 0.0863, + "step": 15436 + }, + { + "epoch": 2.02, + "learning_rate": 1.7508460450409747e-05, + "loss": 0.0724, + "step": 15437 + }, + { + "epoch": 2.02, + "learning_rate": 1.750811017220311e-05, + "loss": 0.0533, + "step": 15438 + }, + { + "epoch": 2.02, + "learning_rate": 1.7507759872880307e-05, + "loss": 0.077, + "step": 15439 + }, + { + "epoch": 2.02, + "learning_rate": 1.750740955244233e-05, + "loss": 0.0759, + "step": 15440 + }, + { + "epoch": 2.02, + "learning_rate": 1.750705921089017e-05, + "loss": 0.0748, + "step": 15441 + }, + { + "epoch": 2.02, + "learning_rate": 1.75067088482248e-05, + "loss": 0.0641, + "step": 15442 + }, + { + "epoch": 2.02, + "learning_rate": 1.750635846444721e-05, + "loss": 0.0696, + "step": 15443 + }, + { + "epoch": 2.02, + "learning_rate": 1.7506008059558386e-05, + "loss": 0.0704, + "step": 15444 + }, + { + "epoch": 2.02, + "learning_rate": 1.750565763355932e-05, + "loss": 0.0719, + "step": 15445 + }, + { + "epoch": 2.02, + "learning_rate": 1.7505307186450985e-05, + "loss": 0.0703, + "step": 15446 + }, + { + "epoch": 2.02, + "learning_rate": 1.7504956718234377e-05, + "loss": 0.0596, + "step": 15447 + }, + { + "epoch": 2.02, + "learning_rate": 1.750460622891048e-05, + "loss": 0.0642, + "step": 15448 + }, + { + "epoch": 2.02, + "learning_rate": 1.7504255718480273e-05, + "loss": 0.0764, + "step": 15449 + }, + { + "epoch": 2.02, + "learning_rate": 1.750390518694475e-05, + "loss": 0.0642, + "step": 15450 + }, + { + "epoch": 2.02, + "learning_rate": 1.7503554634304892e-05, + "loss": 0.0686, + "step": 15451 + }, + { + "epoch": 2.02, + "learning_rate": 1.7503204060561686e-05, + "loss": 0.0837, + "step": 15452 + }, + { + "epoch": 2.02, + "learning_rate": 1.7502853465716115e-05, + "loss": 0.0743, + "step": 15453 + }, + { + "epoch": 2.02, + "learning_rate": 1.750250284976917e-05, + "loss": 0.0715, + "step": 15454 + }, + { + "epoch": 2.02, + "learning_rate": 1.750215221272184e-05, + "loss": 0.068, + "step": 15455 + }, + { + "epoch": 2.02, + "learning_rate": 1.7501801554575102e-05, + "loss": 0.0627, + "step": 15456 + }, + { + "epoch": 2.02, + "learning_rate": 1.7501450875329945e-05, + "loss": 0.0717, + "step": 15457 + }, + { + "epoch": 2.02, + "learning_rate": 1.7501100174987362e-05, + "loss": 0.0625, + "step": 15458 + }, + { + "epoch": 2.03, + "learning_rate": 1.7500749453548332e-05, + "loss": 0.0657, + "step": 15459 + }, + { + "epoch": 2.03, + "learning_rate": 1.7500398711013842e-05, + "loss": 0.0584, + "step": 15460 + }, + { + "epoch": 2.03, + "learning_rate": 1.750004794738488e-05, + "loss": 0.0586, + "step": 15461 + }, + { + "epoch": 2.03, + "learning_rate": 1.7499697162662435e-05, + "loss": 0.0678, + "step": 15462 + }, + { + "epoch": 2.03, + "learning_rate": 1.7499346356847492e-05, + "loss": 0.0586, + "step": 15463 + }, + { + "epoch": 2.03, + "learning_rate": 1.7498995529941033e-05, + "loss": 0.0591, + "step": 15464 + }, + { + "epoch": 2.03, + "learning_rate": 1.7498644681944046e-05, + "loss": 0.0694, + "step": 15465 + }, + { + "epoch": 2.03, + "learning_rate": 1.7498293812857525e-05, + "loss": 0.0696, + "step": 15466 + }, + { + "epoch": 2.03, + "learning_rate": 1.7497942922682455e-05, + "loss": 0.0626, + "step": 15467 + }, + { + "epoch": 2.03, + "learning_rate": 1.7497592011419815e-05, + "loss": 0.0815, + "step": 15468 + }, + { + "epoch": 2.03, + "learning_rate": 1.7497241079070595e-05, + "loss": 0.073, + "step": 15469 + }, + { + "epoch": 2.03, + "learning_rate": 1.7496890125635787e-05, + "loss": 0.0689, + "step": 15470 + }, + { + "epoch": 2.03, + "learning_rate": 1.7496539151116378e-05, + "loss": 0.0631, + "step": 15471 + }, + { + "epoch": 2.03, + "learning_rate": 1.7496188155513346e-05, + "loss": 0.0679, + "step": 15472 + }, + { + "epoch": 2.03, + "learning_rate": 1.7495837138827684e-05, + "loss": 0.0683, + "step": 15473 + }, + { + "epoch": 2.03, + "learning_rate": 1.7495486101060382e-05, + "loss": 0.0777, + "step": 15474 + }, + { + "epoch": 2.03, + "learning_rate": 1.7495135042212426e-05, + "loss": 0.0733, + "step": 15475 + }, + { + "epoch": 2.03, + "learning_rate": 1.7494783962284798e-05, + "loss": 0.0582, + "step": 15476 + }, + { + "epoch": 2.03, + "learning_rate": 1.749443286127849e-05, + "loss": 0.0734, + "step": 15477 + }, + { + "epoch": 2.03, + "learning_rate": 1.749408173919449e-05, + "loss": 0.0686, + "step": 15478 + }, + { + "epoch": 2.03, + "learning_rate": 1.7493730596033782e-05, + "loss": 0.0681, + "step": 15479 + }, + { + "epoch": 2.03, + "learning_rate": 1.749337943179736e-05, + "loss": 0.0668, + "step": 15480 + }, + { + "epoch": 2.03, + "learning_rate": 1.7493028246486205e-05, + "loss": 0.0639, + "step": 15481 + }, + { + "epoch": 2.03, + "learning_rate": 1.7492677040101304e-05, + "loss": 0.0646, + "step": 15482 + }, + { + "epoch": 2.03, + "learning_rate": 1.749232581264365e-05, + "loss": 0.075, + "step": 15483 + }, + { + "epoch": 2.03, + "learning_rate": 1.7491974564114233e-05, + "loss": 0.0694, + "step": 15484 + }, + { + "epoch": 2.03, + "learning_rate": 1.749162329451403e-05, + "loss": 0.0536, + "step": 15485 + }, + { + "epoch": 2.03, + "learning_rate": 1.749127200384404e-05, + "loss": 0.0759, + "step": 15486 + }, + { + "epoch": 2.03, + "learning_rate": 1.749092069210524e-05, + "loss": 0.0759, + "step": 15487 + }, + { + "epoch": 2.03, + "learning_rate": 1.7490569359298634e-05, + "loss": 0.0627, + "step": 15488 + }, + { + "epoch": 2.03, + "learning_rate": 1.7490218005425195e-05, + "loss": 0.0736, + "step": 15489 + }, + { + "epoch": 2.03, + "learning_rate": 1.7489866630485915e-05, + "loss": 0.0641, + "step": 15490 + }, + { + "epoch": 2.03, + "learning_rate": 1.748951523448179e-05, + "loss": 0.0621, + "step": 15491 + }, + { + "epoch": 2.03, + "learning_rate": 1.7489163817413798e-05, + "loss": 0.0679, + "step": 15492 + }, + { + "epoch": 2.03, + "learning_rate": 1.748881237928293e-05, + "loss": 0.0597, + "step": 15493 + }, + { + "epoch": 2.03, + "learning_rate": 1.748846092009018e-05, + "loss": 0.072, + "step": 15494 + }, + { + "epoch": 2.03, + "learning_rate": 1.7488109439836533e-05, + "loss": 0.0673, + "step": 15495 + }, + { + "epoch": 2.03, + "learning_rate": 1.7487757938522973e-05, + "loss": 0.068, + "step": 15496 + }, + { + "epoch": 2.03, + "learning_rate": 1.7487406416150493e-05, + "loss": 0.0653, + "step": 15497 + }, + { + "epoch": 2.03, + "learning_rate": 1.7487054872720083e-05, + "loss": 0.055, + "step": 15498 + }, + { + "epoch": 2.03, + "learning_rate": 1.748670330823273e-05, + "loss": 0.0697, + "step": 15499 + }, + { + "epoch": 2.03, + "learning_rate": 1.748635172268942e-05, + "loss": 0.0759, + "step": 15500 + }, + { + "epoch": 2.03, + "learning_rate": 1.748600011609115e-05, + "loss": 0.0605, + "step": 15501 + }, + { + "epoch": 2.03, + "learning_rate": 1.7485648488438898e-05, + "loss": 0.0668, + "step": 15502 + }, + { + "epoch": 2.03, + "learning_rate": 1.748529683973366e-05, + "loss": 0.0656, + "step": 15503 + }, + { + "epoch": 2.03, + "learning_rate": 1.748494516997642e-05, + "loss": 0.0743, + "step": 15504 + }, + { + "epoch": 2.03, + "learning_rate": 1.7484593479168174e-05, + "loss": 0.059, + "step": 15505 + }, + { + "epoch": 2.03, + "learning_rate": 1.7484241767309908e-05, + "loss": 0.0662, + "step": 15506 + }, + { + "epoch": 2.03, + "learning_rate": 1.7483890034402612e-05, + "loss": 0.0783, + "step": 15507 + }, + { + "epoch": 2.03, + "learning_rate": 1.748353828044727e-05, + "loss": 0.0669, + "step": 15508 + }, + { + "epoch": 2.03, + "learning_rate": 1.7483186505444875e-05, + "loss": 0.0676, + "step": 15509 + }, + { + "epoch": 2.03, + "learning_rate": 1.748283470939642e-05, + "loss": 0.0741, + "step": 15510 + }, + { + "epoch": 2.03, + "learning_rate": 1.748248289230289e-05, + "loss": 0.0634, + "step": 15511 + }, + { + "epoch": 2.03, + "learning_rate": 1.748213105416527e-05, + "loss": 0.0649, + "step": 15512 + }, + { + "epoch": 2.03, + "learning_rate": 1.748177919498456e-05, + "loss": 0.0724, + "step": 15513 + }, + { + "epoch": 2.03, + "learning_rate": 1.7481427314761745e-05, + "loss": 0.0625, + "step": 15514 + }, + { + "epoch": 2.03, + "learning_rate": 1.748107541349781e-05, + "loss": 0.064, + "step": 15515 + }, + { + "epoch": 2.03, + "learning_rate": 1.7480723491193755e-05, + "loss": 0.0667, + "step": 15516 + }, + { + "epoch": 2.03, + "learning_rate": 1.7480371547850556e-05, + "loss": 0.0709, + "step": 15517 + }, + { + "epoch": 2.03, + "learning_rate": 1.7480019583469214e-05, + "loss": 0.063, + "step": 15518 + }, + { + "epoch": 2.03, + "learning_rate": 1.7479667598050717e-05, + "loss": 0.0718, + "step": 15519 + }, + { + "epoch": 2.03, + "learning_rate": 1.7479315591596052e-05, + "loss": 0.0654, + "step": 15520 + }, + { + "epoch": 2.03, + "learning_rate": 1.747896356410621e-05, + "loss": 0.0735, + "step": 15521 + }, + { + "epoch": 2.03, + "learning_rate": 1.747861151558218e-05, + "loss": 0.067, + "step": 15522 + }, + { + "epoch": 2.03, + "learning_rate": 1.7478259446024957e-05, + "loss": 0.0643, + "step": 15523 + }, + { + "epoch": 2.03, + "learning_rate": 1.7477907355435523e-05, + "loss": 0.0638, + "step": 15524 + }, + { + "epoch": 2.03, + "learning_rate": 1.7477555243814873e-05, + "loss": 0.0677, + "step": 15525 + }, + { + "epoch": 2.03, + "learning_rate": 1.7477203111164002e-05, + "loss": 0.0712, + "step": 15526 + }, + { + "epoch": 2.03, + "learning_rate": 1.747685095748389e-05, + "loss": 0.0679, + "step": 15527 + }, + { + "epoch": 2.03, + "learning_rate": 1.7476498782775536e-05, + "loss": 0.0578, + "step": 15528 + }, + { + "epoch": 2.03, + "learning_rate": 1.7476146587039927e-05, + "loss": 0.067, + "step": 15529 + }, + { + "epoch": 2.03, + "learning_rate": 1.747579437027805e-05, + "loss": 0.0621, + "step": 15530 + }, + { + "epoch": 2.03, + "learning_rate": 1.7475442132490905e-05, + "loss": 0.0692, + "step": 15531 + }, + { + "epoch": 2.03, + "learning_rate": 1.7475089873679477e-05, + "loss": 0.0667, + "step": 15532 + }, + { + "epoch": 2.03, + "learning_rate": 1.7474737593844752e-05, + "loss": 0.069, + "step": 15533 + }, + { + "epoch": 2.03, + "learning_rate": 1.7474385292987726e-05, + "loss": 0.061, + "step": 15534 + }, + { + "epoch": 2.03, + "learning_rate": 1.7474032971109393e-05, + "loss": 0.0759, + "step": 15535 + }, + { + "epoch": 2.04, + "learning_rate": 1.747368062821074e-05, + "loss": 0.0712, + "step": 15536 + }, + { + "epoch": 2.04, + "learning_rate": 1.747332826429276e-05, + "loss": 0.0698, + "step": 15537 + }, + { + "epoch": 2.04, + "learning_rate": 1.7472975879356438e-05, + "loss": 0.0658, + "step": 15538 + }, + { + "epoch": 2.04, + "learning_rate": 1.747262347340277e-05, + "loss": 0.0614, + "step": 15539 + }, + { + "epoch": 2.04, + "learning_rate": 1.7472271046432746e-05, + "loss": 0.0608, + "step": 15540 + }, + { + "epoch": 2.04, + "learning_rate": 1.7471918598447358e-05, + "loss": 0.0674, + "step": 15541 + }, + { + "epoch": 2.04, + "learning_rate": 1.7471566129447598e-05, + "loss": 0.0603, + "step": 15542 + }, + { + "epoch": 2.04, + "learning_rate": 1.7471213639434457e-05, + "loss": 0.0628, + "step": 15543 + }, + { + "epoch": 2.04, + "learning_rate": 1.7470861128408927e-05, + "loss": 0.063, + "step": 15544 + }, + { + "epoch": 2.04, + "learning_rate": 1.7470508596371995e-05, + "loss": 0.0527, + "step": 15545 + }, + { + "epoch": 2.04, + "learning_rate": 1.747015604332465e-05, + "loss": 0.0649, + "step": 15546 + }, + { + "epoch": 2.04, + "learning_rate": 1.74698034692679e-05, + "loss": 0.0734, + "step": 15547 + }, + { + "epoch": 2.04, + "learning_rate": 1.7469450874202718e-05, + "loss": 0.0689, + "step": 15548 + }, + { + "epoch": 2.04, + "learning_rate": 1.7469098258130106e-05, + "loss": 0.074, + "step": 15549 + }, + { + "epoch": 2.04, + "learning_rate": 1.746874562105105e-05, + "loss": 0.0626, + "step": 15550 + }, + { + "epoch": 2.04, + "learning_rate": 1.746839296296655e-05, + "loss": 0.0711, + "step": 15551 + }, + { + "epoch": 2.04, + "learning_rate": 1.746804028387759e-05, + "loss": 0.0759, + "step": 15552 + }, + { + "epoch": 2.04, + "learning_rate": 1.746768758378516e-05, + "loss": 0.0641, + "step": 15553 + }, + { + "epoch": 2.04, + "learning_rate": 1.746733486269026e-05, + "loss": 0.0654, + "step": 15554 + }, + { + "epoch": 2.04, + "learning_rate": 1.7466982120593878e-05, + "loss": 0.0712, + "step": 15555 + }, + { + "epoch": 2.04, + "learning_rate": 1.7466629357497007e-05, + "loss": 0.0736, + "step": 15556 + }, + { + "epoch": 2.04, + "learning_rate": 1.746627657340064e-05, + "loss": 0.0654, + "step": 15557 + }, + { + "epoch": 2.04, + "learning_rate": 1.7465923768305763e-05, + "loss": 0.0692, + "step": 15558 + }, + { + "epoch": 2.04, + "learning_rate": 1.7465570942213378e-05, + "loss": 0.0624, + "step": 15559 + }, + { + "epoch": 2.04, + "learning_rate": 1.746521809512447e-05, + "loss": 0.072, + "step": 15560 + }, + { + "epoch": 2.04, + "learning_rate": 1.7464865227040033e-05, + "loss": 0.0678, + "step": 15561 + }, + { + "epoch": 2.04, + "learning_rate": 1.7464512337961065e-05, + "loss": 0.0593, + "step": 15562 + }, + { + "epoch": 2.04, + "learning_rate": 1.746415942788855e-05, + "loss": 0.0622, + "step": 15563 + }, + { + "epoch": 2.04, + "learning_rate": 1.7463806496823482e-05, + "loss": 0.0664, + "step": 15564 + }, + { + "epoch": 2.04, + "learning_rate": 1.7463453544766856e-05, + "loss": 0.0645, + "step": 15565 + }, + { + "epoch": 2.04, + "learning_rate": 1.7463100571719666e-05, + "loss": 0.069, + "step": 15566 + }, + { + "epoch": 2.04, + "learning_rate": 1.7462747577682904e-05, + "loss": 0.0555, + "step": 15567 + }, + { + "epoch": 2.04, + "learning_rate": 1.746239456265756e-05, + "loss": 0.0575, + "step": 15568 + }, + { + "epoch": 2.04, + "learning_rate": 1.746204152664463e-05, + "loss": 0.0652, + "step": 15569 + }, + { + "epoch": 2.04, + "learning_rate": 1.7461688469645105e-05, + "loss": 0.0571, + "step": 15570 + }, + { + "epoch": 2.04, + "learning_rate": 1.746133539165998e-05, + "loss": 0.0675, + "step": 15571 + }, + { + "epoch": 2.04, + "learning_rate": 1.746098229269024e-05, + "loss": 0.0663, + "step": 15572 + }, + { + "epoch": 2.04, + "learning_rate": 1.7460629172736892e-05, + "loss": 0.0703, + "step": 15573 + }, + { + "epoch": 2.04, + "learning_rate": 1.746027603180092e-05, + "loss": 0.0682, + "step": 15574 + }, + { + "epoch": 2.04, + "learning_rate": 1.745992286988332e-05, + "loss": 0.0645, + "step": 15575 + }, + { + "epoch": 2.04, + "learning_rate": 1.7459569686985082e-05, + "loss": 0.066, + "step": 15576 + }, + { + "epoch": 2.04, + "learning_rate": 1.74592164831072e-05, + "loss": 0.0693, + "step": 15577 + }, + { + "epoch": 2.04, + "learning_rate": 1.745886325825067e-05, + "loss": 0.0593, + "step": 15578 + }, + { + "epoch": 2.04, + "learning_rate": 1.7458510012416486e-05, + "loss": 0.0645, + "step": 15579 + }, + { + "epoch": 2.04, + "learning_rate": 1.7458156745605642e-05, + "loss": 0.0711, + "step": 15580 + }, + { + "epoch": 2.04, + "learning_rate": 1.7457803457819125e-05, + "loss": 0.0696, + "step": 15581 + }, + { + "epoch": 2.04, + "learning_rate": 1.7457450149057934e-05, + "loss": 0.0856, + "step": 15582 + }, + { + "epoch": 2.04, + "learning_rate": 1.7457096819323062e-05, + "loss": 0.0604, + "step": 15583 + }, + { + "epoch": 2.04, + "learning_rate": 1.7456743468615506e-05, + "loss": 0.0741, + "step": 15584 + }, + { + "epoch": 2.04, + "learning_rate": 1.745639009693625e-05, + "loss": 0.0679, + "step": 15585 + }, + { + "epoch": 2.04, + "learning_rate": 1.7456036704286295e-05, + "loss": 0.066, + "step": 15586 + }, + { + "epoch": 2.04, + "learning_rate": 1.7455683290666637e-05, + "loss": 0.0741, + "step": 15587 + }, + { + "epoch": 2.04, + "learning_rate": 1.7455329856078264e-05, + "loss": 0.0692, + "step": 15588 + }, + { + "epoch": 2.04, + "learning_rate": 1.7454976400522174e-05, + "loss": 0.0629, + "step": 15589 + }, + { + "epoch": 2.04, + "learning_rate": 1.745462292399936e-05, + "loss": 0.0707, + "step": 15590 + }, + { + "epoch": 2.04, + "learning_rate": 1.7454269426510815e-05, + "loss": 0.0585, + "step": 15591 + }, + { + "epoch": 2.04, + "learning_rate": 1.7453915908057534e-05, + "loss": 0.0609, + "step": 15592 + }, + { + "epoch": 2.04, + "learning_rate": 1.7453562368640514e-05, + "loss": 0.0635, + "step": 15593 + }, + { + "epoch": 2.04, + "learning_rate": 1.7453208808260742e-05, + "loss": 0.0713, + "step": 15594 + }, + { + "epoch": 2.04, + "learning_rate": 1.7452855226919222e-05, + "loss": 0.0634, + "step": 15595 + }, + { + "epoch": 2.04, + "learning_rate": 1.745250162461694e-05, + "loss": 0.0589, + "step": 15596 + }, + { + "epoch": 2.04, + "learning_rate": 1.7452148001354897e-05, + "loss": 0.066, + "step": 15597 + }, + { + "epoch": 2.04, + "learning_rate": 1.7451794357134083e-05, + "loss": 0.0613, + "step": 15598 + }, + { + "epoch": 2.04, + "learning_rate": 1.7451440691955492e-05, + "loss": 0.0577, + "step": 15599 + }, + { + "epoch": 2.04, + "learning_rate": 1.7451087005820127e-05, + "loss": 0.0685, + "step": 15600 + }, + { + "epoch": 2.04, + "learning_rate": 1.7450733298728973e-05, + "loss": 0.0739, + "step": 15601 + }, + { + "epoch": 2.04, + "learning_rate": 1.7450379570683025e-05, + "loss": 0.0722, + "step": 15602 + }, + { + "epoch": 2.04, + "learning_rate": 1.7450025821683286e-05, + "loss": 0.0764, + "step": 15603 + }, + { + "epoch": 2.04, + "learning_rate": 1.7449672051730745e-05, + "loss": 0.0677, + "step": 15604 + }, + { + "epoch": 2.04, + "learning_rate": 1.7449318260826396e-05, + "loss": 0.0623, + "step": 15605 + }, + { + "epoch": 2.04, + "learning_rate": 1.744896444897124e-05, + "loss": 0.063, + "step": 15606 + }, + { + "epoch": 2.04, + "learning_rate": 1.7448610616166265e-05, + "loss": 0.077, + "step": 15607 + }, + { + "epoch": 2.04, + "learning_rate": 1.7448256762412472e-05, + "loss": 0.081, + "step": 15608 + }, + { + "epoch": 2.04, + "learning_rate": 1.744790288771085e-05, + "loss": 0.0706, + "step": 15609 + }, + { + "epoch": 2.04, + "learning_rate": 1.74475489920624e-05, + "loss": 0.0506, + "step": 15610 + }, + { + "epoch": 2.04, + "learning_rate": 1.7447195075468112e-05, + "loss": 0.0658, + "step": 15611 + }, + { + "epoch": 2.05, + "learning_rate": 1.7446841137928988e-05, + "loss": 0.0679, + "step": 15612 + }, + { + "epoch": 2.05, + "learning_rate": 1.7446487179446018e-05, + "loss": 0.0586, + "step": 15613 + }, + { + "epoch": 2.05, + "learning_rate": 1.7446133200020202e-05, + "loss": 0.0581, + "step": 15614 + }, + { + "epoch": 2.05, + "learning_rate": 1.744577919965253e-05, + "loss": 0.0644, + "step": 15615 + }, + { + "epoch": 2.05, + "learning_rate": 1.7445425178344006e-05, + "loss": 0.0613, + "step": 15616 + }, + { + "epoch": 2.05, + "learning_rate": 1.7445071136095613e-05, + "loss": 0.0698, + "step": 15617 + }, + { + "epoch": 2.05, + "learning_rate": 1.7444717072908357e-05, + "loss": 0.0639, + "step": 15618 + }, + { + "epoch": 2.05, + "learning_rate": 1.744436298878323e-05, + "loss": 0.0824, + "step": 15619 + }, + { + "epoch": 2.05, + "learning_rate": 1.7444008883721227e-05, + "loss": 0.0799, + "step": 15620 + }, + { + "epoch": 2.05, + "learning_rate": 1.744365475772335e-05, + "loss": 0.0538, + "step": 15621 + }, + { + "epoch": 2.05, + "learning_rate": 1.744330061079059e-05, + "loss": 0.0684, + "step": 15622 + }, + { + "epoch": 2.05, + "learning_rate": 1.7442946442923938e-05, + "loss": 0.0581, + "step": 15623 + }, + { + "epoch": 2.05, + "learning_rate": 1.74425922541244e-05, + "loss": 0.0739, + "step": 15624 + }, + { + "epoch": 2.05, + "learning_rate": 1.7442238044392964e-05, + "loss": 0.0562, + "step": 15625 + }, + { + "epoch": 2.05, + "learning_rate": 1.7441883813730634e-05, + "loss": 0.0597, + "step": 15626 + }, + { + "epoch": 2.05, + "learning_rate": 1.74415295621384e-05, + "loss": 0.0676, + "step": 15627 + }, + { + "epoch": 2.05, + "learning_rate": 1.744117528961726e-05, + "loss": 0.0567, + "step": 15628 + }, + { + "epoch": 2.05, + "learning_rate": 1.7440820996168214e-05, + "loss": 0.0676, + "step": 15629 + }, + { + "epoch": 2.05, + "learning_rate": 1.744046668179225e-05, + "loss": 0.0681, + "step": 15630 + }, + { + "epoch": 2.05, + "learning_rate": 1.7440112346490373e-05, + "loss": 0.0745, + "step": 15631 + }, + { + "epoch": 2.05, + "learning_rate": 1.7439757990263574e-05, + "loss": 0.0553, + "step": 15632 + }, + { + "epoch": 2.05, + "learning_rate": 1.743940361311285e-05, + "loss": 0.0585, + "step": 15633 + }, + { + "epoch": 2.05, + "learning_rate": 1.7439049215039206e-05, + "loss": 0.0626, + "step": 15634 + }, + { + "epoch": 2.05, + "learning_rate": 1.743869479604363e-05, + "loss": 0.0684, + "step": 15635 + }, + { + "epoch": 2.05, + "learning_rate": 1.743834035612712e-05, + "loss": 0.0657, + "step": 15636 + }, + { + "epoch": 2.05, + "learning_rate": 1.7437985895290675e-05, + "loss": 0.0711, + "step": 15637 + }, + { + "epoch": 2.05, + "learning_rate": 1.743763141353529e-05, + "loss": 0.0729, + "step": 15638 + }, + { + "epoch": 2.05, + "learning_rate": 1.7437276910861963e-05, + "loss": 0.0596, + "step": 15639 + }, + { + "epoch": 2.05, + "learning_rate": 1.743692238727169e-05, + "loss": 0.0653, + "step": 15640 + }, + { + "epoch": 2.05, + "learning_rate": 1.7436567842765465e-05, + "loss": 0.0686, + "step": 15641 + }, + { + "epoch": 2.05, + "learning_rate": 1.7436213277344296e-05, + "loss": 0.0666, + "step": 15642 + }, + { + "epoch": 2.05, + "learning_rate": 1.7435858691009172e-05, + "loss": 0.0771, + "step": 15643 + }, + { + "epoch": 2.05, + "learning_rate": 1.743550408376109e-05, + "loss": 0.0738, + "step": 15644 + }, + { + "epoch": 2.05, + "learning_rate": 1.7435149455601047e-05, + "loss": 0.0727, + "step": 15645 + }, + { + "epoch": 2.05, + "learning_rate": 1.7434794806530042e-05, + "loss": 0.0688, + "step": 15646 + }, + { + "epoch": 2.05, + "learning_rate": 1.7434440136549073e-05, + "loss": 0.0686, + "step": 15647 + }, + { + "epoch": 2.05, + "learning_rate": 1.743408544565914e-05, + "loss": 0.0713, + "step": 15648 + }, + { + "epoch": 2.05, + "learning_rate": 1.7433730733861235e-05, + "loss": 0.0658, + "step": 15649 + }, + { + "epoch": 2.05, + "learning_rate": 1.743337600115636e-05, + "loss": 0.0605, + "step": 15650 + }, + { + "epoch": 2.05, + "learning_rate": 1.743302124754551e-05, + "loss": 0.0664, + "step": 15651 + }, + { + "epoch": 2.05, + "learning_rate": 1.743266647302968e-05, + "loss": 0.0657, + "step": 15652 + }, + { + "epoch": 2.05, + "learning_rate": 1.7432311677609877e-05, + "loss": 0.0748, + "step": 15653 + }, + { + "epoch": 2.05, + "learning_rate": 1.7431956861287087e-05, + "loss": 0.0693, + "step": 15654 + }, + { + "epoch": 2.05, + "learning_rate": 1.743160202406232e-05, + "loss": 0.0602, + "step": 15655 + }, + { + "epoch": 2.05, + "learning_rate": 1.7431247165936565e-05, + "loss": 0.0671, + "step": 15656 + }, + { + "epoch": 2.05, + "learning_rate": 1.7430892286910826e-05, + "loss": 0.0614, + "step": 15657 + }, + { + "epoch": 2.05, + "learning_rate": 1.7430537386986094e-05, + "loss": 0.0589, + "step": 15658 + }, + { + "epoch": 2.05, + "learning_rate": 1.7430182466163375e-05, + "loss": 0.071, + "step": 15659 + }, + { + "epoch": 2.05, + "learning_rate": 1.7429827524443662e-05, + "loss": 0.0678, + "step": 15660 + }, + { + "epoch": 2.05, + "learning_rate": 1.7429472561827955e-05, + "loss": 0.0738, + "step": 15661 + }, + { + "epoch": 2.05, + "learning_rate": 1.742911757831725e-05, + "loss": 0.064, + "step": 15662 + }, + { + "epoch": 2.05, + "learning_rate": 1.742876257391255e-05, + "loss": 0.0644, + "step": 15663 + }, + { + "epoch": 2.05, + "learning_rate": 1.742840754861485e-05, + "loss": 0.068, + "step": 15664 + }, + { + "epoch": 2.05, + "learning_rate": 1.7428052502425147e-05, + "loss": 0.0737, + "step": 15665 + }, + { + "epoch": 2.05, + "learning_rate": 1.7427697435344446e-05, + "loss": 0.0692, + "step": 15666 + }, + { + "epoch": 2.05, + "learning_rate": 1.7427342347373738e-05, + "loss": 0.0624, + "step": 15667 + }, + { + "epoch": 2.05, + "learning_rate": 1.7426987238514028e-05, + "loss": 0.0562, + "step": 15668 + }, + { + "epoch": 2.05, + "learning_rate": 1.7426632108766313e-05, + "loss": 0.068, + "step": 15669 + }, + { + "epoch": 2.05, + "learning_rate": 1.7426276958131585e-05, + "loss": 0.0755, + "step": 15670 + }, + { + "epoch": 2.05, + "learning_rate": 1.7425921786610854e-05, + "loss": 0.0679, + "step": 15671 + }, + { + "epoch": 2.05, + "learning_rate": 1.7425566594205113e-05, + "loss": 0.0712, + "step": 15672 + }, + { + "epoch": 2.05, + "learning_rate": 1.7425211380915355e-05, + "loss": 0.0611, + "step": 15673 + }, + { + "epoch": 2.05, + "learning_rate": 1.7424856146742593e-05, + "loss": 0.0665, + "step": 15674 + }, + { + "epoch": 2.05, + "learning_rate": 1.7424500891687815e-05, + "loss": 0.0663, + "step": 15675 + }, + { + "epoch": 2.05, + "learning_rate": 1.7424145615752025e-05, + "loss": 0.0604, + "step": 15676 + }, + { + "epoch": 2.05, + "learning_rate": 1.742379031893622e-05, + "loss": 0.0674, + "step": 15677 + }, + { + "epoch": 2.05, + "learning_rate": 1.74234350012414e-05, + "loss": 0.0766, + "step": 15678 + }, + { + "epoch": 2.05, + "learning_rate": 1.7423079662668564e-05, + "loss": 0.0712, + "step": 15679 + }, + { + "epoch": 2.05, + "learning_rate": 1.7422724303218714e-05, + "loss": 0.0792, + "step": 15680 + }, + { + "epoch": 2.05, + "learning_rate": 1.7422368922892846e-05, + "loss": 0.065, + "step": 15681 + }, + { + "epoch": 2.05, + "learning_rate": 1.742201352169196e-05, + "loss": 0.0576, + "step": 15682 + }, + { + "epoch": 2.05, + "learning_rate": 1.7421658099617056e-05, + "loss": 0.0596, + "step": 15683 + }, + { + "epoch": 2.05, + "learning_rate": 1.7421302656669134e-05, + "loss": 0.0706, + "step": 15684 + }, + { + "epoch": 2.05, + "learning_rate": 1.7420947192849195e-05, + "loss": 0.0662, + "step": 15685 + }, + { + "epoch": 2.05, + "learning_rate": 1.7420591708158236e-05, + "loss": 0.0737, + "step": 15686 + }, + { + "epoch": 2.05, + "learning_rate": 1.7420236202597255e-05, + "loss": 0.0635, + "step": 15687 + }, + { + "epoch": 2.06, + "learning_rate": 1.741988067616726e-05, + "loss": 0.0726, + "step": 15688 + }, + { + "epoch": 2.06, + "learning_rate": 1.7419525128869243e-05, + "loss": 0.0719, + "step": 15689 + }, + { + "epoch": 2.06, + "learning_rate": 1.741916956070421e-05, + "loss": 0.0674, + "step": 15690 + }, + { + "epoch": 2.06, + "learning_rate": 1.7418813971673153e-05, + "loss": 0.0681, + "step": 15691 + }, + { + "epoch": 2.06, + "learning_rate": 1.7418458361777077e-05, + "loss": 0.067, + "step": 15692 + }, + { + "epoch": 2.06, + "learning_rate": 1.7418102731016983e-05, + "loss": 0.0589, + "step": 15693 + }, + { + "epoch": 2.06, + "learning_rate": 1.741774707939387e-05, + "loss": 0.0667, + "step": 15694 + }, + { + "epoch": 2.06, + "learning_rate": 1.741739140690874e-05, + "loss": 0.069, + "step": 15695 + }, + { + "epoch": 2.06, + "learning_rate": 1.741703571356259e-05, + "loss": 0.0636, + "step": 15696 + }, + { + "epoch": 2.06, + "learning_rate": 1.7416679999356425e-05, + "loss": 0.0693, + "step": 15697 + }, + { + "epoch": 2.06, + "learning_rate": 1.741632426429124e-05, + "loss": 0.0789, + "step": 15698 + }, + { + "epoch": 2.06, + "learning_rate": 1.7415968508368036e-05, + "loss": 0.0653, + "step": 15699 + }, + { + "epoch": 2.06, + "learning_rate": 1.7415612731587817e-05, + "loss": 0.0674, + "step": 15700 + }, + { + "epoch": 2.06, + "learning_rate": 1.7415256933951584e-05, + "loss": 0.0656, + "step": 15701 + }, + { + "epoch": 2.06, + "learning_rate": 1.7414901115460336e-05, + "loss": 0.0632, + "step": 15702 + }, + { + "epoch": 2.06, + "learning_rate": 1.7414545276115072e-05, + "loss": 0.0683, + "step": 15703 + }, + { + "epoch": 2.06, + "learning_rate": 1.7414189415916792e-05, + "loss": 0.0547, + "step": 15704 + }, + { + "epoch": 2.06, + "learning_rate": 1.7413833534866503e-05, + "loss": 0.0664, + "step": 15705 + }, + { + "epoch": 2.06, + "learning_rate": 1.74134776329652e-05, + "loss": 0.0646, + "step": 15706 + }, + { + "epoch": 2.06, + "learning_rate": 1.741312171021389e-05, + "loss": 0.0688, + "step": 15707 + }, + { + "epoch": 2.06, + "learning_rate": 1.7412765766613564e-05, + "loss": 0.0623, + "step": 15708 + }, + { + "epoch": 2.06, + "learning_rate": 1.741240980216523e-05, + "loss": 0.0663, + "step": 15709 + }, + { + "epoch": 2.06, + "learning_rate": 1.741205381686989e-05, + "loss": 0.0673, + "step": 15710 + }, + { + "epoch": 2.06, + "learning_rate": 1.7411697810728542e-05, + "loss": 0.0668, + "step": 15711 + }, + { + "epoch": 2.06, + "learning_rate": 1.7411341783742194e-05, + "loss": 0.0631, + "step": 15712 + }, + { + "epoch": 2.06, + "learning_rate": 1.7410985735911835e-05, + "loss": 0.0691, + "step": 15713 + }, + { + "epoch": 2.06, + "learning_rate": 1.7410629667238476e-05, + "loss": 0.0709, + "step": 15714 + }, + { + "epoch": 2.06, + "learning_rate": 1.7410273577723112e-05, + "loss": 0.0755, + "step": 15715 + }, + { + "epoch": 2.06, + "learning_rate": 1.7409917467366753e-05, + "loss": 0.0717, + "step": 15716 + }, + { + "epoch": 2.06, + "learning_rate": 1.7409561336170396e-05, + "loss": 0.0681, + "step": 15717 + }, + { + "epoch": 2.06, + "learning_rate": 1.740920518413504e-05, + "loss": 0.0646, + "step": 15718 + }, + { + "epoch": 2.06, + "learning_rate": 1.740884901126169e-05, + "loss": 0.0664, + "step": 15719 + }, + { + "epoch": 2.06, + "learning_rate": 1.7408492817551345e-05, + "loss": 0.0683, + "step": 15720 + }, + { + "epoch": 2.06, + "learning_rate": 1.740813660300501e-05, + "loss": 0.0545, + "step": 15721 + }, + { + "epoch": 2.06, + "learning_rate": 1.7407780367623685e-05, + "loss": 0.0684, + "step": 15722 + }, + { + "epoch": 2.06, + "learning_rate": 1.740742411140837e-05, + "loss": 0.0618, + "step": 15723 + }, + { + "epoch": 2.06, + "learning_rate": 1.7407067834360075e-05, + "loss": 0.0614, + "step": 15724 + }, + { + "epoch": 2.06, + "learning_rate": 1.7406711536479793e-05, + "loss": 0.0616, + "step": 15725 + }, + { + "epoch": 2.06, + "learning_rate": 1.7406355217768525e-05, + "loss": 0.0691, + "step": 15726 + }, + { + "epoch": 2.06, + "learning_rate": 1.7405998878227284e-05, + "loss": 0.0506, + "step": 15727 + }, + { + "epoch": 2.06, + "learning_rate": 1.740564251785706e-05, + "loss": 0.067, + "step": 15728 + }, + { + "epoch": 2.06, + "learning_rate": 1.7405286136658865e-05, + "loss": 0.0577, + "step": 15729 + }, + { + "epoch": 2.06, + "learning_rate": 1.7404929734633694e-05, + "loss": 0.0611, + "step": 15730 + }, + { + "epoch": 2.06, + "learning_rate": 1.7404573311782557e-05, + "loss": 0.0633, + "step": 15731 + }, + { + "epoch": 2.06, + "learning_rate": 1.7404216868106448e-05, + "loss": 0.0807, + "step": 15732 + }, + { + "epoch": 2.06, + "learning_rate": 1.7403860403606373e-05, + "loss": 0.0674, + "step": 15733 + }, + { + "epoch": 2.06, + "learning_rate": 1.7403503918283338e-05, + "loss": 0.0679, + "step": 15734 + }, + { + "epoch": 2.06, + "learning_rate": 1.7403147412138342e-05, + "loss": 0.0735, + "step": 15735 + }, + { + "epoch": 2.06, + "learning_rate": 1.7402790885172386e-05, + "loss": 0.0724, + "step": 15736 + }, + { + "epoch": 2.06, + "learning_rate": 1.7402434337386475e-05, + "loss": 0.066, + "step": 15737 + }, + { + "epoch": 2.06, + "learning_rate": 1.7402077768781612e-05, + "loss": 0.071, + "step": 15738 + }, + { + "epoch": 2.06, + "learning_rate": 1.74017211793588e-05, + "loss": 0.0746, + "step": 15739 + }, + { + "epoch": 2.06, + "learning_rate": 1.7401364569119043e-05, + "loss": 0.0674, + "step": 15740 + }, + { + "epoch": 2.06, + "learning_rate": 1.7401007938063338e-05, + "loss": 0.074, + "step": 15741 + }, + { + "epoch": 2.06, + "learning_rate": 1.7400651286192697e-05, + "loss": 0.0717, + "step": 15742 + }, + { + "epoch": 2.06, + "learning_rate": 1.7400294613508112e-05, + "loss": 0.0544, + "step": 15743 + }, + { + "epoch": 2.06, + "learning_rate": 1.7399937920010596e-05, + "loss": 0.0738, + "step": 15744 + }, + { + "epoch": 2.06, + "learning_rate": 1.739958120570115e-05, + "loss": 0.0664, + "step": 15745 + }, + { + "epoch": 2.06, + "learning_rate": 1.739922447058078e-05, + "loss": 0.0741, + "step": 15746 + }, + { + "epoch": 2.06, + "learning_rate": 1.739886771465048e-05, + "loss": 0.0639, + "step": 15747 + }, + { + "epoch": 2.06, + "learning_rate": 1.7398510937911255e-05, + "loss": 0.0842, + "step": 15748 + }, + { + "epoch": 2.06, + "learning_rate": 1.7398154140364116e-05, + "loss": 0.0628, + "step": 15749 + }, + { + "epoch": 2.06, + "learning_rate": 1.739779732201006e-05, + "loss": 0.0707, + "step": 15750 + }, + { + "epoch": 2.06, + "learning_rate": 1.7397440482850098e-05, + "loss": 0.0613, + "step": 15751 + }, + { + "epoch": 2.06, + "learning_rate": 1.7397083622885227e-05, + "loss": 0.0724, + "step": 15752 + }, + { + "epoch": 2.06, + "learning_rate": 1.739672674211645e-05, + "loss": 0.0649, + "step": 15753 + }, + { + "epoch": 2.06, + "learning_rate": 1.7396369840544777e-05, + "loss": 0.0737, + "step": 15754 + }, + { + "epoch": 2.06, + "learning_rate": 1.7396012918171203e-05, + "loss": 0.0733, + "step": 15755 + }, + { + "epoch": 2.06, + "learning_rate": 1.7395655974996736e-05, + "loss": 0.06, + "step": 15756 + }, + { + "epoch": 2.06, + "learning_rate": 1.7395299011022385e-05, + "loss": 0.0616, + "step": 15757 + }, + { + "epoch": 2.06, + "learning_rate": 1.7394942026249148e-05, + "loss": 0.0716, + "step": 15758 + }, + { + "epoch": 2.06, + "learning_rate": 1.7394585020678026e-05, + "loss": 0.0693, + "step": 15759 + }, + { + "epoch": 2.06, + "learning_rate": 1.7394227994310033e-05, + "loss": 0.0578, + "step": 15760 + }, + { + "epoch": 2.06, + "learning_rate": 1.7393870947146164e-05, + "loss": 0.0692, + "step": 15761 + }, + { + "epoch": 2.06, + "learning_rate": 1.739351387918743e-05, + "loss": 0.0737, + "step": 15762 + }, + { + "epoch": 2.06, + "learning_rate": 1.739315679043483e-05, + "loss": 0.0592, + "step": 15763 + }, + { + "epoch": 2.06, + "learning_rate": 1.739279968088937e-05, + "loss": 0.0637, + "step": 15764 + }, + { + "epoch": 2.07, + "learning_rate": 1.7392442550552054e-05, + "loss": 0.0745, + "step": 15765 + }, + { + "epoch": 2.07, + "learning_rate": 1.7392085399423887e-05, + "loss": 0.0702, + "step": 15766 + }, + { + "epoch": 2.07, + "learning_rate": 1.7391728227505875e-05, + "loss": 0.0607, + "step": 15767 + }, + { + "epoch": 2.07, + "learning_rate": 1.7391371034799022e-05, + "loss": 0.0569, + "step": 15768 + }, + { + "epoch": 2.07, + "learning_rate": 1.7391013821304327e-05, + "loss": 0.0666, + "step": 15769 + }, + { + "epoch": 2.07, + "learning_rate": 1.7390656587022803e-05, + "loss": 0.0731, + "step": 15770 + }, + { + "epoch": 2.07, + "learning_rate": 1.739029933195545e-05, + "loss": 0.0682, + "step": 15771 + }, + { + "epoch": 2.07, + "learning_rate": 1.7389942056103274e-05, + "loss": 0.0638, + "step": 15772 + }, + { + "epoch": 2.07, + "learning_rate": 1.738958475946728e-05, + "loss": 0.0658, + "step": 15773 + }, + { + "epoch": 2.07, + "learning_rate": 1.738922744204847e-05, + "loss": 0.0825, + "step": 15774 + }, + { + "epoch": 2.07, + "learning_rate": 1.7388870103847852e-05, + "loss": 0.0661, + "step": 15775 + }, + { + "epoch": 2.07, + "learning_rate": 1.7388512744866433e-05, + "loss": 0.0688, + "step": 15776 + }, + { + "epoch": 2.07, + "learning_rate": 1.738815536510521e-05, + "loss": 0.061, + "step": 15777 + }, + { + "epoch": 2.07, + "learning_rate": 1.73877979645652e-05, + "loss": 0.0668, + "step": 15778 + }, + { + "epoch": 2.07, + "learning_rate": 1.73874405432474e-05, + "loss": 0.0722, + "step": 15779 + }, + { + "epoch": 2.07, + "learning_rate": 1.7387083101152815e-05, + "loss": 0.0707, + "step": 15780 + }, + { + "epoch": 2.07, + "learning_rate": 1.7386725638282452e-05, + "loss": 0.0696, + "step": 15781 + }, + { + "epoch": 2.07, + "learning_rate": 1.738636815463732e-05, + "loss": 0.0567, + "step": 15782 + }, + { + "epoch": 2.07, + "learning_rate": 1.7386010650218415e-05, + "loss": 0.0631, + "step": 15783 + }, + { + "epoch": 2.07, + "learning_rate": 1.7385653125026752e-05, + "loss": 0.0685, + "step": 15784 + }, + { + "epoch": 2.07, + "learning_rate": 1.738529557906333e-05, + "loss": 0.0648, + "step": 15785 + }, + { + "epoch": 2.07, + "learning_rate": 1.738493801232916e-05, + "loss": 0.0693, + "step": 15786 + }, + { + "epoch": 2.07, + "learning_rate": 1.738458042482525e-05, + "loss": 0.0661, + "step": 15787 + }, + { + "epoch": 2.07, + "learning_rate": 1.7384222816552594e-05, + "loss": 0.0613, + "step": 15788 + }, + { + "epoch": 2.07, + "learning_rate": 1.7383865187512206e-05, + "loss": 0.073, + "step": 15789 + }, + { + "epoch": 2.07, + "learning_rate": 1.738350753770509e-05, + "loss": 0.0679, + "step": 15790 + }, + { + "epoch": 2.07, + "learning_rate": 1.7383149867132253e-05, + "loss": 0.0682, + "step": 15791 + }, + { + "epoch": 2.07, + "learning_rate": 1.73827921757947e-05, + "loss": 0.0709, + "step": 15792 + }, + { + "epoch": 2.07, + "learning_rate": 1.738243446369344e-05, + "loss": 0.0705, + "step": 15793 + }, + { + "epoch": 2.07, + "learning_rate": 1.7382076730829475e-05, + "loss": 0.0732, + "step": 15794 + }, + { + "epoch": 2.07, + "learning_rate": 1.7381718977203808e-05, + "loss": 0.0713, + "step": 15795 + }, + { + "epoch": 2.07, + "learning_rate": 1.738136120281745e-05, + "loss": 0.0645, + "step": 15796 + }, + { + "epoch": 2.07, + "learning_rate": 1.738100340767141e-05, + "loss": 0.0644, + "step": 15797 + }, + { + "epoch": 2.07, + "learning_rate": 1.738064559176669e-05, + "loss": 0.0608, + "step": 15798 + }, + { + "epoch": 2.07, + "learning_rate": 1.7380287755104297e-05, + "loss": 0.0693, + "step": 15799 + }, + { + "epoch": 2.07, + "learning_rate": 1.737992989768524e-05, + "loss": 0.0658, + "step": 15800 + }, + { + "epoch": 2.07, + "learning_rate": 1.7379572019510522e-05, + "loss": 0.0608, + "step": 15801 + }, + { + "epoch": 2.07, + "learning_rate": 1.737921412058115e-05, + "loss": 0.0693, + "step": 15802 + }, + { + "epoch": 2.07, + "learning_rate": 1.737885620089813e-05, + "loss": 0.054, + "step": 15803 + }, + { + "epoch": 2.07, + "learning_rate": 1.737849826046247e-05, + "loss": 0.0773, + "step": 15804 + }, + { + "epoch": 2.07, + "learning_rate": 1.7378140299275177e-05, + "loss": 0.0838, + "step": 15805 + }, + { + "epoch": 2.07, + "learning_rate": 1.737778231733726e-05, + "loss": 0.0598, + "step": 15806 + }, + { + "epoch": 2.07, + "learning_rate": 1.737742431464972e-05, + "loss": 0.0569, + "step": 15807 + }, + { + "epoch": 2.07, + "learning_rate": 1.7377066291213566e-05, + "loss": 0.0702, + "step": 15808 + }, + { + "epoch": 2.07, + "learning_rate": 1.737670824702981e-05, + "loss": 0.0629, + "step": 15809 + }, + { + "epoch": 2.07, + "learning_rate": 1.7376350182099454e-05, + "loss": 0.0687, + "step": 15810 + }, + { + "epoch": 2.07, + "learning_rate": 1.7375992096423502e-05, + "loss": 0.06, + "step": 15811 + }, + { + "epoch": 2.07, + "learning_rate": 1.7375633990002968e-05, + "loss": 0.0821, + "step": 15812 + }, + { + "epoch": 2.07, + "learning_rate": 1.7375275862838858e-05, + "loss": 0.0795, + "step": 15813 + }, + { + "epoch": 2.07, + "learning_rate": 1.7374917714932174e-05, + "loss": 0.0635, + "step": 15814 + }, + { + "epoch": 2.07, + "learning_rate": 1.737455954628393e-05, + "loss": 0.0719, + "step": 15815 + }, + { + "epoch": 2.07, + "learning_rate": 1.737420135689513e-05, + "loss": 0.065, + "step": 15816 + }, + { + "epoch": 2.07, + "learning_rate": 1.7373843146766778e-05, + "loss": 0.0581, + "step": 15817 + }, + { + "epoch": 2.07, + "learning_rate": 1.7373484915899882e-05, + "loss": 0.0614, + "step": 15818 + }, + { + "epoch": 2.07, + "learning_rate": 1.737312666429546e-05, + "loss": 0.0661, + "step": 15819 + }, + { + "epoch": 2.07, + "learning_rate": 1.7372768391954508e-05, + "loss": 0.0661, + "step": 15820 + }, + { + "epoch": 2.07, + "learning_rate": 1.7372410098878036e-05, + "loss": 0.0733, + "step": 15821 + }, + { + "epoch": 2.07, + "learning_rate": 1.7372051785067058e-05, + "loss": 0.0583, + "step": 15822 + }, + { + "epoch": 2.07, + "learning_rate": 1.7371693450522573e-05, + "loss": 0.0641, + "step": 15823 + }, + { + "epoch": 2.07, + "learning_rate": 1.7371335095245595e-05, + "loss": 0.0768, + "step": 15824 + }, + { + "epoch": 2.07, + "learning_rate": 1.7370976719237127e-05, + "loss": 0.0603, + "step": 15825 + }, + { + "epoch": 2.07, + "learning_rate": 1.737061832249818e-05, + "loss": 0.0735, + "step": 15826 + }, + { + "epoch": 2.07, + "learning_rate": 1.7370259905029764e-05, + "loss": 0.0682, + "step": 15827 + }, + { + "epoch": 2.07, + "learning_rate": 1.7369901466832882e-05, + "loss": 0.0673, + "step": 15828 + }, + { + "epoch": 2.07, + "learning_rate": 1.7369543007908545e-05, + "loss": 0.0762, + "step": 15829 + }, + { + "epoch": 2.07, + "learning_rate": 1.7369184528257762e-05, + "loss": 0.0653, + "step": 15830 + }, + { + "epoch": 2.07, + "learning_rate": 1.736882602788154e-05, + "loss": 0.0699, + "step": 15831 + }, + { + "epoch": 2.07, + "learning_rate": 1.7368467506780888e-05, + "loss": 0.0646, + "step": 15832 + }, + { + "epoch": 2.07, + "learning_rate": 1.7368108964956812e-05, + "loss": 0.0789, + "step": 15833 + }, + { + "epoch": 2.07, + "learning_rate": 1.736775040241032e-05, + "loss": 0.0725, + "step": 15834 + }, + { + "epoch": 2.07, + "learning_rate": 1.7367391819142424e-05, + "loss": 0.0681, + "step": 15835 + }, + { + "epoch": 2.07, + "learning_rate": 1.736703321515413e-05, + "loss": 0.0717, + "step": 15836 + }, + { + "epoch": 2.07, + "learning_rate": 1.7366674590446447e-05, + "loss": 0.0594, + "step": 15837 + }, + { + "epoch": 2.07, + "learning_rate": 1.736631594502039e-05, + "loss": 0.0652, + "step": 15838 + }, + { + "epoch": 2.07, + "learning_rate": 1.7365957278876953e-05, + "loss": 0.0676, + "step": 15839 + }, + { + "epoch": 2.07, + "learning_rate": 1.7365598592017158e-05, + "loss": 0.0634, + "step": 15840 + }, + { + "epoch": 2.08, + "learning_rate": 1.7365239884442013e-05, + "loss": 0.0721, + "step": 15841 + }, + { + "epoch": 2.08, + "learning_rate": 1.7364881156152518e-05, + "loss": 0.0728, + "step": 15842 + }, + { + "epoch": 2.08, + "learning_rate": 1.7364522407149686e-05, + "loss": 0.0758, + "step": 15843 + }, + { + "epoch": 2.08, + "learning_rate": 1.7364163637434527e-05, + "loss": 0.0586, + "step": 15844 + }, + { + "epoch": 2.08, + "learning_rate": 1.7363804847008054e-05, + "loss": 0.0658, + "step": 15845 + }, + { + "epoch": 2.08, + "learning_rate": 1.736344603587127e-05, + "loss": 0.0638, + "step": 15846 + }, + { + "epoch": 2.08, + "learning_rate": 1.7363087204025183e-05, + "loss": 0.062, + "step": 15847 + }, + { + "epoch": 2.08, + "learning_rate": 1.736272835147081e-05, + "loss": 0.0686, + "step": 15848 + }, + { + "epoch": 2.08, + "learning_rate": 1.7362369478209154e-05, + "loss": 0.0742, + "step": 15849 + }, + { + "epoch": 2.08, + "learning_rate": 1.7362010584241224e-05, + "loss": 0.0641, + "step": 15850 + }, + { + "epoch": 2.08, + "learning_rate": 1.7361651669568035e-05, + "loss": 0.0584, + "step": 15851 + }, + { + "epoch": 2.08, + "learning_rate": 1.7361292734190592e-05, + "loss": 0.0757, + "step": 15852 + }, + { + "epoch": 2.08, + "learning_rate": 1.7360933778109902e-05, + "loss": 0.066, + "step": 15853 + }, + { + "epoch": 2.08, + "learning_rate": 1.7360574801326978e-05, + "loss": 0.0704, + "step": 15854 + }, + { + "epoch": 2.08, + "learning_rate": 1.7360215803842833e-05, + "loss": 0.0676, + "step": 15855 + }, + { + "epoch": 2.08, + "learning_rate": 1.735985678565847e-05, + "loss": 0.0608, + "step": 15856 + }, + { + "epoch": 2.08, + "learning_rate": 1.7359497746774903e-05, + "loss": 0.0564, + "step": 15857 + }, + { + "epoch": 2.08, + "learning_rate": 1.735913868719314e-05, + "loss": 0.073, + "step": 15858 + }, + { + "epoch": 2.08, + "learning_rate": 1.735877960691419e-05, + "loss": 0.0674, + "step": 15859 + }, + { + "epoch": 2.08, + "learning_rate": 1.735842050593907e-05, + "loss": 0.0631, + "step": 15860 + }, + { + "epoch": 2.08, + "learning_rate": 1.7358061384268778e-05, + "loss": 0.0794, + "step": 15861 + }, + { + "epoch": 2.08, + "learning_rate": 1.735770224190433e-05, + "loss": 0.0826, + "step": 15862 + }, + { + "epoch": 2.08, + "learning_rate": 1.7357343078846738e-05, + "loss": 0.0694, + "step": 15863 + }, + { + "epoch": 2.08, + "learning_rate": 1.7356983895097012e-05, + "loss": 0.0663, + "step": 15864 + }, + { + "epoch": 2.08, + "learning_rate": 1.735662469065616e-05, + "loss": 0.0682, + "step": 15865 + }, + { + "epoch": 2.08, + "learning_rate": 1.7356265465525194e-05, + "loss": 0.0653, + "step": 15866 + }, + { + "epoch": 2.08, + "learning_rate": 1.7355906219705118e-05, + "loss": 0.0752, + "step": 15867 + }, + { + "epoch": 2.08, + "learning_rate": 1.7355546953196953e-05, + "loss": 0.0677, + "step": 15868 + }, + { + "epoch": 2.08, + "learning_rate": 1.7355187666001702e-05, + "loss": 0.0521, + "step": 15869 + }, + { + "epoch": 2.08, + "learning_rate": 1.7354828358120375e-05, + "loss": 0.0666, + "step": 15870 + }, + { + "epoch": 2.08, + "learning_rate": 1.7354469029553986e-05, + "loss": 0.0708, + "step": 15871 + }, + { + "epoch": 2.08, + "learning_rate": 1.7354109680303546e-05, + "loss": 0.0659, + "step": 15872 + }, + { + "epoch": 2.08, + "learning_rate": 1.7353750310370063e-05, + "loss": 0.071, + "step": 15873 + }, + { + "epoch": 2.08, + "learning_rate": 1.7353390919754553e-05, + "loss": 0.0664, + "step": 15874 + }, + { + "epoch": 2.08, + "learning_rate": 1.7353031508458018e-05, + "loss": 0.0689, + "step": 15875 + }, + { + "epoch": 2.08, + "learning_rate": 1.7352672076481475e-05, + "loss": 0.0688, + "step": 15876 + }, + { + "epoch": 2.08, + "learning_rate": 1.7352312623825935e-05, + "loss": 0.0678, + "step": 15877 + }, + { + "epoch": 2.08, + "learning_rate": 1.7351953150492402e-05, + "loss": 0.0612, + "step": 15878 + }, + { + "epoch": 2.08, + "learning_rate": 1.7351593656481894e-05, + "loss": 0.0797, + "step": 15879 + }, + { + "epoch": 2.08, + "learning_rate": 1.735123414179542e-05, + "loss": 0.0676, + "step": 15880 + }, + { + "epoch": 2.08, + "learning_rate": 1.7350874606433997e-05, + "loss": 0.0614, + "step": 15881 + }, + { + "epoch": 2.08, + "learning_rate": 1.7350515050398627e-05, + "loss": 0.0617, + "step": 15882 + }, + { + "epoch": 2.08, + "learning_rate": 1.7350155473690323e-05, + "loss": 0.064, + "step": 15883 + }, + { + "epoch": 2.08, + "learning_rate": 1.73497958763101e-05, + "loss": 0.0669, + "step": 15884 + }, + { + "epoch": 2.08, + "learning_rate": 1.7349436258258965e-05, + "loss": 0.0699, + "step": 15885 + }, + { + "epoch": 2.08, + "learning_rate": 1.7349076619537937e-05, + "loss": 0.0672, + "step": 15886 + }, + { + "epoch": 2.08, + "learning_rate": 1.734871696014802e-05, + "loss": 0.0678, + "step": 15887 + }, + { + "epoch": 2.08, + "learning_rate": 1.734835728009023e-05, + "loss": 0.0602, + "step": 15888 + }, + { + "epoch": 2.08, + "learning_rate": 1.734799757936557e-05, + "loss": 0.0684, + "step": 15889 + }, + { + "epoch": 2.08, + "learning_rate": 1.7347637857975062e-05, + "loss": 0.0621, + "step": 15890 + }, + { + "epoch": 2.08, + "learning_rate": 1.7347278115919714e-05, + "loss": 0.0643, + "step": 15891 + }, + { + "epoch": 2.08, + "learning_rate": 1.734691835320054e-05, + "loss": 0.0668, + "step": 15892 + }, + { + "epoch": 2.08, + "learning_rate": 1.7346558569818545e-05, + "loss": 0.0666, + "step": 15893 + }, + { + "epoch": 2.08, + "learning_rate": 1.7346198765774747e-05, + "loss": 0.0628, + "step": 15894 + }, + { + "epoch": 2.08, + "learning_rate": 1.7345838941070155e-05, + "loss": 0.0779, + "step": 15895 + }, + { + "epoch": 2.08, + "learning_rate": 1.7345479095705786e-05, + "loss": 0.0699, + "step": 15896 + }, + { + "epoch": 2.08, + "learning_rate": 1.7345119229682646e-05, + "loss": 0.0546, + "step": 15897 + }, + { + "epoch": 2.08, + "learning_rate": 1.7344759343001748e-05, + "loss": 0.0624, + "step": 15898 + }, + { + "epoch": 2.08, + "learning_rate": 1.7344399435664105e-05, + "loss": 0.07, + "step": 15899 + }, + { + "epoch": 2.08, + "learning_rate": 1.734403950767073e-05, + "loss": 0.0629, + "step": 15900 + }, + { + "epoch": 2.08, + "learning_rate": 1.734367955902264e-05, + "loss": 0.0734, + "step": 15901 + }, + { + "epoch": 2.08, + "learning_rate": 1.7343319589720837e-05, + "loss": 0.0703, + "step": 15902 + }, + { + "epoch": 2.08, + "learning_rate": 1.734295959976634e-05, + "loss": 0.0674, + "step": 15903 + }, + { + "epoch": 2.08, + "learning_rate": 1.7342599589160163e-05, + "loss": 0.0756, + "step": 15904 + }, + { + "epoch": 2.08, + "learning_rate": 1.734223955790331e-05, + "loss": 0.0784, + "step": 15905 + }, + { + "epoch": 2.08, + "learning_rate": 1.7341879505996805e-05, + "loss": 0.0696, + "step": 15906 + }, + { + "epoch": 2.08, + "learning_rate": 1.734151943344165e-05, + "loss": 0.0536, + "step": 15907 + }, + { + "epoch": 2.08, + "learning_rate": 1.7341159340238865e-05, + "loss": 0.0629, + "step": 15908 + }, + { + "epoch": 2.08, + "learning_rate": 1.734079922638946e-05, + "loss": 0.0655, + "step": 15909 + }, + { + "epoch": 2.08, + "learning_rate": 1.7340439091894446e-05, + "loss": 0.0729, + "step": 15910 + }, + { + "epoch": 2.08, + "learning_rate": 1.734007893675484e-05, + "loss": 0.0642, + "step": 15911 + }, + { + "epoch": 2.08, + "learning_rate": 1.7339718760971654e-05, + "loss": 0.061, + "step": 15912 + }, + { + "epoch": 2.08, + "learning_rate": 1.7339358564545898e-05, + "loss": 0.0698, + "step": 15913 + }, + { + "epoch": 2.08, + "learning_rate": 1.733899834747859e-05, + "loss": 0.0598, + "step": 15914 + }, + { + "epoch": 2.08, + "learning_rate": 1.7338638109770734e-05, + "loss": 0.0673, + "step": 15915 + }, + { + "epoch": 2.08, + "learning_rate": 1.7338277851423356e-05, + "loss": 0.0684, + "step": 15916 + }, + { + "epoch": 2.09, + "learning_rate": 1.7337917572437458e-05, + "loss": 0.0605, + "step": 15917 + }, + { + "epoch": 2.09, + "learning_rate": 1.733755727281406e-05, + "loss": 0.0774, + "step": 15918 + }, + { + "epoch": 2.09, + "learning_rate": 1.733719695255417e-05, + "loss": 0.0642, + "step": 15919 + }, + { + "epoch": 2.09, + "learning_rate": 1.7336836611658807e-05, + "loss": 0.0712, + "step": 15920 + }, + { + "epoch": 2.09, + "learning_rate": 1.733647625012898e-05, + "loss": 0.0724, + "step": 15921 + }, + { + "epoch": 2.09, + "learning_rate": 1.7336115867965702e-05, + "loss": 0.0642, + "step": 15922 + }, + { + "epoch": 2.09, + "learning_rate": 1.7335755465169994e-05, + "loss": 0.0663, + "step": 15923 + }, + { + "epoch": 2.09, + "learning_rate": 1.7335395041742864e-05, + "loss": 0.0686, + "step": 15924 + }, + { + "epoch": 2.09, + "learning_rate": 1.7335034597685322e-05, + "loss": 0.0711, + "step": 15925 + }, + { + "epoch": 2.09, + "learning_rate": 1.733467413299839e-05, + "loss": 0.0697, + "step": 15926 + }, + { + "epoch": 2.09, + "learning_rate": 1.7334313647683076e-05, + "loss": 0.0688, + "step": 15927 + }, + { + "epoch": 2.09, + "learning_rate": 1.733395314174039e-05, + "loss": 0.064, + "step": 15928 + }, + { + "epoch": 2.09, + "learning_rate": 1.733359261517136e-05, + "loss": 0.0592, + "step": 15929 + }, + { + "epoch": 2.09, + "learning_rate": 1.7333232067976985e-05, + "loss": 0.0701, + "step": 15930 + }, + { + "epoch": 2.09, + "learning_rate": 1.733287150015829e-05, + "loss": 0.0747, + "step": 15931 + }, + { + "epoch": 2.09, + "learning_rate": 1.7332510911716284e-05, + "loss": 0.0767, + "step": 15932 + }, + { + "epoch": 2.09, + "learning_rate": 1.7332150302651977e-05, + "loss": 0.0625, + "step": 15933 + }, + { + "epoch": 2.09, + "learning_rate": 1.733178967296639e-05, + "loss": 0.0656, + "step": 15934 + }, + { + "epoch": 2.09, + "learning_rate": 1.7331429022660534e-05, + "loss": 0.0791, + "step": 15935 + }, + { + "epoch": 2.09, + "learning_rate": 1.733106835173543e-05, + "loss": 0.0732, + "step": 15936 + }, + { + "epoch": 2.09, + "learning_rate": 1.7330707660192084e-05, + "loss": 0.0642, + "step": 15937 + }, + { + "epoch": 2.09, + "learning_rate": 1.7330346948031507e-05, + "loss": 0.0663, + "step": 15938 + }, + { + "epoch": 2.09, + "learning_rate": 1.7329986215254728e-05, + "loss": 0.0635, + "step": 15939 + }, + { + "epoch": 2.09, + "learning_rate": 1.7329625461862745e-05, + "loss": 0.0621, + "step": 15940 + }, + { + "epoch": 2.09, + "learning_rate": 1.732926468785659e-05, + "loss": 0.0766, + "step": 15941 + }, + { + "epoch": 2.09, + "learning_rate": 1.732890389323726e-05, + "loss": 0.0586, + "step": 15942 + }, + { + "epoch": 2.09, + "learning_rate": 1.732854307800578e-05, + "loss": 0.0666, + "step": 15943 + }, + { + "epoch": 2.09, + "learning_rate": 1.7328182242163167e-05, + "loss": 0.0553, + "step": 15944 + }, + { + "epoch": 2.09, + "learning_rate": 1.732782138571043e-05, + "loss": 0.0756, + "step": 15945 + }, + { + "epoch": 2.09, + "learning_rate": 1.7327460508648584e-05, + "loss": 0.0749, + "step": 15946 + }, + { + "epoch": 2.09, + "learning_rate": 1.7327099610978647e-05, + "loss": 0.066, + "step": 15947 + }, + { + "epoch": 2.09, + "learning_rate": 1.732673869270163e-05, + "loss": 0.0678, + "step": 15948 + }, + { + "epoch": 2.09, + "learning_rate": 1.732637775381855e-05, + "loss": 0.0447, + "step": 15949 + }, + { + "epoch": 2.09, + "learning_rate": 1.7326016794330425e-05, + "loss": 0.0585, + "step": 15950 + }, + { + "epoch": 2.09, + "learning_rate": 1.732565581423827e-05, + "loss": 0.0662, + "step": 15951 + }, + { + "epoch": 2.09, + "learning_rate": 1.7325294813543096e-05, + "loss": 0.0603, + "step": 15952 + }, + { + "epoch": 2.09, + "learning_rate": 1.7324933792245922e-05, + "loss": 0.0682, + "step": 15953 + }, + { + "epoch": 2.09, + "learning_rate": 1.732457275034776e-05, + "loss": 0.0676, + "step": 15954 + }, + { + "epoch": 2.09, + "learning_rate": 1.7324211687849625e-05, + "loss": 0.0711, + "step": 15955 + }, + { + "epoch": 2.09, + "learning_rate": 1.7323850604752537e-05, + "loss": 0.0618, + "step": 15956 + }, + { + "epoch": 2.09, + "learning_rate": 1.732348950105751e-05, + "loss": 0.0733, + "step": 15957 + }, + { + "epoch": 2.09, + "learning_rate": 1.732312837676556e-05, + "loss": 0.069, + "step": 15958 + }, + { + "epoch": 2.09, + "learning_rate": 1.73227672318777e-05, + "loss": 0.0675, + "step": 15959 + }, + { + "epoch": 2.09, + "learning_rate": 1.7322406066394947e-05, + "loss": 0.0602, + "step": 15960 + }, + { + "epoch": 2.09, + "learning_rate": 1.7322044880318315e-05, + "loss": 0.0719, + "step": 15961 + }, + { + "epoch": 2.09, + "learning_rate": 1.7321683673648824e-05, + "loss": 0.062, + "step": 15962 + }, + { + "epoch": 2.09, + "learning_rate": 1.7321322446387485e-05, + "loss": 0.0615, + "step": 15963 + }, + { + "epoch": 2.09, + "learning_rate": 1.732096119853532e-05, + "loss": 0.0748, + "step": 15964 + }, + { + "epoch": 2.09, + "learning_rate": 1.732059993009334e-05, + "loss": 0.0674, + "step": 15965 + }, + { + "epoch": 2.09, + "learning_rate": 1.7320238641062562e-05, + "loss": 0.0637, + "step": 15966 + }, + { + "epoch": 2.09, + "learning_rate": 1.7319877331444003e-05, + "loss": 0.0679, + "step": 15967 + }, + { + "epoch": 2.09, + "learning_rate": 1.731951600123868e-05, + "loss": 0.0593, + "step": 15968 + }, + { + "epoch": 2.09, + "learning_rate": 1.7319154650447604e-05, + "loss": 0.064, + "step": 15969 + }, + { + "epoch": 2.09, + "learning_rate": 1.7318793279071798e-05, + "loss": 0.0638, + "step": 15970 + }, + { + "epoch": 2.09, + "learning_rate": 1.7318431887112278e-05, + "loss": 0.073, + "step": 15971 + }, + { + "epoch": 2.09, + "learning_rate": 1.7318070474570054e-05, + "loss": 0.0694, + "step": 15972 + }, + { + "epoch": 2.09, + "learning_rate": 1.731770904144615e-05, + "loss": 0.0633, + "step": 15973 + }, + { + "epoch": 2.09, + "learning_rate": 1.731734758774158e-05, + "loss": 0.0718, + "step": 15974 + }, + { + "epoch": 2.09, + "learning_rate": 1.731698611345735e-05, + "loss": 0.0619, + "step": 15975 + }, + { + "epoch": 2.09, + "learning_rate": 1.7316624618594496e-05, + "loss": 0.0635, + "step": 15976 + }, + { + "epoch": 2.09, + "learning_rate": 1.731626310315402e-05, + "loss": 0.0598, + "step": 15977 + }, + { + "epoch": 2.09, + "learning_rate": 1.7315901567136948e-05, + "loss": 0.0663, + "step": 15978 + }, + { + "epoch": 2.09, + "learning_rate": 1.7315540010544293e-05, + "loss": 0.0781, + "step": 15979 + }, + { + "epoch": 2.09, + "learning_rate": 1.7315178433377065e-05, + "loss": 0.061, + "step": 15980 + }, + { + "epoch": 2.09, + "learning_rate": 1.731481683563629e-05, + "loss": 0.0632, + "step": 15981 + }, + { + "epoch": 2.09, + "learning_rate": 1.7314455217322984e-05, + "loss": 0.0706, + "step": 15982 + }, + { + "epoch": 2.09, + "learning_rate": 1.731409357843816e-05, + "loss": 0.0699, + "step": 15983 + }, + { + "epoch": 2.09, + "learning_rate": 1.7313731918982838e-05, + "loss": 0.0725, + "step": 15984 + }, + { + "epoch": 2.09, + "learning_rate": 1.7313370238958037e-05, + "loss": 0.0717, + "step": 15985 + }, + { + "epoch": 2.09, + "learning_rate": 1.731300853836477e-05, + "loss": 0.0779, + "step": 15986 + }, + { + "epoch": 2.09, + "learning_rate": 1.7312646817204055e-05, + "loss": 0.0836, + "step": 15987 + }, + { + "epoch": 2.09, + "learning_rate": 1.731228507547691e-05, + "loss": 0.0659, + "step": 15988 + }, + { + "epoch": 2.09, + "learning_rate": 1.7311923313184353e-05, + "loss": 0.0747, + "step": 15989 + }, + { + "epoch": 2.09, + "learning_rate": 1.73115615303274e-05, + "loss": 0.0608, + "step": 15990 + }, + { + "epoch": 2.09, + "learning_rate": 1.7311199726907072e-05, + "loss": 0.0648, + "step": 15991 + }, + { + "epoch": 2.09, + "learning_rate": 1.731083790292438e-05, + "loss": 0.0758, + "step": 15992 + }, + { + "epoch": 2.09, + "learning_rate": 1.731047605838035e-05, + "loss": 0.0647, + "step": 15993 + }, + { + "epoch": 2.1, + "learning_rate": 1.7310114193275995e-05, + "loss": 0.078, + "step": 15994 + }, + { + "epoch": 2.1, + "learning_rate": 1.7309752307612334e-05, + "loss": 0.0759, + "step": 15995 + }, + { + "epoch": 2.1, + "learning_rate": 1.7309390401390384e-05, + "loss": 0.0697, + "step": 15996 + }, + { + "epoch": 2.1, + "learning_rate": 1.7309028474611158e-05, + "loss": 0.0726, + "step": 15997 + }, + { + "epoch": 2.1, + "learning_rate": 1.730866652727568e-05, + "loss": 0.0592, + "step": 15998 + }, + { + "epoch": 2.1, + "learning_rate": 1.730830455938497e-05, + "loss": 0.0636, + "step": 15999 + }, + { + "epoch": 2.1, + "learning_rate": 1.730794257094004e-05, + "loss": 0.063, + "step": 16000 + }, + { + "epoch": 2.1, + "learning_rate": 1.730758056194191e-05, + "loss": 0.0607, + "step": 16001 + }, + { + "epoch": 2.1, + "learning_rate": 1.7307218532391604e-05, + "loss": 0.067, + "step": 16002 + }, + { + "epoch": 2.1, + "learning_rate": 1.730685648229013e-05, + "loss": 0.0742, + "step": 16003 + }, + { + "epoch": 2.1, + "learning_rate": 1.730649441163851e-05, + "loss": 0.068, + "step": 16004 + }, + { + "epoch": 2.1, + "learning_rate": 1.7306132320437766e-05, + "loss": 0.0731, + "step": 16005 + }, + { + "epoch": 2.1, + "learning_rate": 1.7305770208688916e-05, + "loss": 0.0775, + "step": 16006 + }, + { + "epoch": 2.1, + "learning_rate": 1.7305408076392974e-05, + "loss": 0.0679, + "step": 16007 + }, + { + "epoch": 2.1, + "learning_rate": 1.730504592355096e-05, + "loss": 0.0714, + "step": 16008 + }, + { + "epoch": 2.1, + "learning_rate": 1.73046837501639e-05, + "loss": 0.0704, + "step": 16009 + }, + { + "epoch": 2.1, + "learning_rate": 1.7304321556232798e-05, + "loss": 0.0646, + "step": 16010 + }, + { + "epoch": 2.1, + "learning_rate": 1.7303959341758682e-05, + "loss": 0.0645, + "step": 16011 + }, + { + "epoch": 2.1, + "learning_rate": 1.730359710674257e-05, + "loss": 0.0699, + "step": 16012 + }, + { + "epoch": 2.1, + "learning_rate": 1.730323485118548e-05, + "loss": 0.0633, + "step": 16013 + }, + { + "epoch": 2.1, + "learning_rate": 1.7302872575088432e-05, + "loss": 0.0729, + "step": 16014 + }, + { + "epoch": 2.1, + "learning_rate": 1.7302510278452442e-05, + "loss": 0.0603, + "step": 16015 + }, + { + "epoch": 2.1, + "learning_rate": 1.7302147961278532e-05, + "loss": 0.0661, + "step": 16016 + }, + { + "epoch": 2.1, + "learning_rate": 1.730178562356772e-05, + "loss": 0.0723, + "step": 16017 + }, + { + "epoch": 2.1, + "learning_rate": 1.7301423265321026e-05, + "loss": 0.0639, + "step": 16018 + }, + { + "epoch": 2.1, + "learning_rate": 1.7301060886539465e-05, + "loss": 0.0624, + "step": 16019 + }, + { + "epoch": 2.1, + "learning_rate": 1.730069848722406e-05, + "loss": 0.0688, + "step": 16020 + }, + { + "epoch": 2.1, + "learning_rate": 1.730033606737583e-05, + "loss": 0.0627, + "step": 16021 + }, + { + "epoch": 2.1, + "learning_rate": 1.7299973626995793e-05, + "loss": 0.0623, + "step": 16022 + }, + { + "epoch": 2.1, + "learning_rate": 1.729961116608497e-05, + "loss": 0.0568, + "step": 16023 + }, + { + "epoch": 2.1, + "learning_rate": 1.7299248684644378e-05, + "loss": 0.0681, + "step": 16024 + }, + { + "epoch": 2.1, + "learning_rate": 1.729888618267504e-05, + "loss": 0.062, + "step": 16025 + }, + { + "epoch": 2.1, + "learning_rate": 1.7298523660177967e-05, + "loss": 0.0549, + "step": 16026 + }, + { + "epoch": 2.1, + "learning_rate": 1.7298161117154194e-05, + "loss": 0.0637, + "step": 16027 + }, + { + "epoch": 2.1, + "learning_rate": 1.7297798553604726e-05, + "loss": 0.0648, + "step": 16028 + }, + { + "epoch": 2.1, + "learning_rate": 1.729743596953059e-05, + "loss": 0.0687, + "step": 16029 + }, + { + "epoch": 2.1, + "learning_rate": 1.72970733649328e-05, + "loss": 0.0669, + "step": 16030 + }, + { + "epoch": 2.1, + "learning_rate": 1.7296710739812383e-05, + "loss": 0.0687, + "step": 16031 + }, + { + "epoch": 2.1, + "learning_rate": 1.7296348094170355e-05, + "loss": 0.0642, + "step": 16032 + }, + { + "epoch": 2.1, + "learning_rate": 1.7295985428007738e-05, + "loss": 0.0661, + "step": 16033 + }, + { + "epoch": 2.1, + "learning_rate": 1.7295622741325547e-05, + "loss": 0.0646, + "step": 16034 + }, + { + "epoch": 2.1, + "learning_rate": 1.729526003412481e-05, + "loss": 0.0639, + "step": 16035 + }, + { + "epoch": 2.1, + "learning_rate": 1.729489730640654e-05, + "loss": 0.0649, + "step": 16036 + }, + { + "epoch": 2.1, + "learning_rate": 1.729453455817176e-05, + "loss": 0.0691, + "step": 16037 + }, + { + "epoch": 2.1, + "learning_rate": 1.729417178942149e-05, + "loss": 0.0662, + "step": 16038 + }, + { + "epoch": 2.1, + "learning_rate": 1.729380900015675e-05, + "loss": 0.0668, + "step": 16039 + }, + { + "epoch": 2.1, + "learning_rate": 1.7293446190378565e-05, + "loss": 0.0643, + "step": 16040 + }, + { + "epoch": 2.1, + "learning_rate": 1.7293083360087945e-05, + "loss": 0.0661, + "step": 16041 + }, + { + "epoch": 2.1, + "learning_rate": 1.729272050928592e-05, + "loss": 0.0752, + "step": 16042 + }, + { + "epoch": 2.1, + "learning_rate": 1.7292357637973506e-05, + "loss": 0.0755, + "step": 16043 + }, + { + "epoch": 2.1, + "learning_rate": 1.7291994746151723e-05, + "loss": 0.0717, + "step": 16044 + }, + { + "epoch": 2.1, + "learning_rate": 1.7291631833821595e-05, + "loss": 0.0686, + "step": 16045 + }, + { + "epoch": 2.1, + "learning_rate": 1.729126890098414e-05, + "loss": 0.0639, + "step": 16046 + }, + { + "epoch": 2.1, + "learning_rate": 1.729090594764038e-05, + "loss": 0.0602, + "step": 16047 + }, + { + "epoch": 2.1, + "learning_rate": 1.7290542973791334e-05, + "loss": 0.0653, + "step": 16048 + }, + { + "epoch": 2.1, + "learning_rate": 1.7290179979438026e-05, + "loss": 0.0732, + "step": 16049 + }, + { + "epoch": 2.1, + "learning_rate": 1.7289816964581475e-05, + "loss": 0.0628, + "step": 16050 + }, + { + "epoch": 2.1, + "learning_rate": 1.7289453929222698e-05, + "loss": 0.0722, + "step": 16051 + }, + { + "epoch": 2.1, + "learning_rate": 1.7289090873362725e-05, + "loss": 0.0646, + "step": 16052 + }, + { + "epoch": 2.1, + "learning_rate": 1.728872779700257e-05, + "loss": 0.0736, + "step": 16053 + }, + { + "epoch": 2.1, + "learning_rate": 1.7288364700143255e-05, + "loss": 0.0706, + "step": 16054 + }, + { + "epoch": 2.1, + "learning_rate": 1.7288001582785804e-05, + "loss": 0.0564, + "step": 16055 + }, + { + "epoch": 2.1, + "learning_rate": 1.7287638444931236e-05, + "loss": 0.0554, + "step": 16056 + }, + { + "epoch": 2.1, + "learning_rate": 1.7287275286580575e-05, + "loss": 0.065, + "step": 16057 + }, + { + "epoch": 2.1, + "learning_rate": 1.7286912107734837e-05, + "loss": 0.072, + "step": 16058 + }, + { + "epoch": 2.1, + "learning_rate": 1.728654890839505e-05, + "loss": 0.0575, + "step": 16059 + }, + { + "epoch": 2.1, + "learning_rate": 1.728618568856223e-05, + "loss": 0.0689, + "step": 16060 + }, + { + "epoch": 2.1, + "learning_rate": 1.7285822448237398e-05, + "loss": 0.0707, + "step": 16061 + }, + { + "epoch": 2.1, + "learning_rate": 1.7285459187421585e-05, + "loss": 0.0712, + "step": 16062 + }, + { + "epoch": 2.1, + "learning_rate": 1.72850959061158e-05, + "loss": 0.0647, + "step": 16063 + }, + { + "epoch": 2.1, + "learning_rate": 1.7284732604321075e-05, + "loss": 0.0688, + "step": 16064 + }, + { + "epoch": 2.1, + "learning_rate": 1.7284369282038426e-05, + "loss": 0.0709, + "step": 16065 + }, + { + "epoch": 2.1, + "learning_rate": 1.7284005939268875e-05, + "loss": 0.0719, + "step": 16066 + }, + { + "epoch": 2.1, + "learning_rate": 1.728364257601345e-05, + "loss": 0.0654, + "step": 16067 + }, + { + "epoch": 2.1, + "learning_rate": 1.7283279192273162e-05, + "loss": 0.066, + "step": 16068 + }, + { + "epoch": 2.1, + "learning_rate": 1.728291578804904e-05, + "loss": 0.0608, + "step": 16069 + }, + { + "epoch": 2.11, + "learning_rate": 1.728255236334211e-05, + "loss": 0.0692, + "step": 16070 + }, + { + "epoch": 2.11, + "learning_rate": 1.7282188918153387e-05, + "loss": 0.0625, + "step": 16071 + }, + { + "epoch": 2.11, + "learning_rate": 1.7281825452483892e-05, + "loss": 0.056, + "step": 16072 + }, + { + "epoch": 2.11, + "learning_rate": 1.7281461966334655e-05, + "loss": 0.0688, + "step": 16073 + }, + { + "epoch": 2.11, + "learning_rate": 1.7281098459706692e-05, + "loss": 0.068, + "step": 16074 + }, + { + "epoch": 2.11, + "learning_rate": 1.7280734932601027e-05, + "loss": 0.0575, + "step": 16075 + }, + { + "epoch": 2.11, + "learning_rate": 1.7280371385018683e-05, + "loss": 0.0718, + "step": 16076 + }, + { + "epoch": 2.11, + "learning_rate": 1.7280007816960685e-05, + "loss": 0.0608, + "step": 16077 + }, + { + "epoch": 2.11, + "learning_rate": 1.727964422842805e-05, + "loss": 0.0597, + "step": 16078 + }, + { + "epoch": 2.11, + "learning_rate": 1.7279280619421802e-05, + "loss": 0.0661, + "step": 16079 + }, + { + "epoch": 2.11, + "learning_rate": 1.7278916989942966e-05, + "loss": 0.0642, + "step": 16080 + }, + { + "epoch": 2.11, + "learning_rate": 1.7278553339992566e-05, + "loss": 0.0624, + "step": 16081 + }, + { + "epoch": 2.11, + "learning_rate": 1.727818966957162e-05, + "loss": 0.0643, + "step": 16082 + }, + { + "epoch": 2.11, + "learning_rate": 1.7277825978681153e-05, + "loss": 0.0711, + "step": 16083 + }, + { + "epoch": 2.11, + "learning_rate": 1.7277462267322187e-05, + "loss": 0.0613, + "step": 16084 + }, + { + "epoch": 2.11, + "learning_rate": 1.727709853549575e-05, + "loss": 0.0718, + "step": 16085 + }, + { + "epoch": 2.11, + "learning_rate": 1.7276734783202858e-05, + "loss": 0.0694, + "step": 16086 + }, + { + "epoch": 2.11, + "learning_rate": 1.7276371010444536e-05, + "loss": 0.0657, + "step": 16087 + }, + { + "epoch": 2.11, + "learning_rate": 1.727600721722181e-05, + "loss": 0.0655, + "step": 16088 + }, + { + "epoch": 2.11, + "learning_rate": 1.72756434035357e-05, + "loss": 0.0657, + "step": 16089 + }, + { + "epoch": 2.11, + "learning_rate": 1.7275279569387232e-05, + "loss": 0.0606, + "step": 16090 + }, + { + "epoch": 2.11, + "learning_rate": 1.7274915714777427e-05, + "loss": 0.0683, + "step": 16091 + }, + { + "epoch": 2.11, + "learning_rate": 1.7274551839707306e-05, + "loss": 0.0739, + "step": 16092 + }, + { + "epoch": 2.11, + "learning_rate": 1.7274187944177898e-05, + "loss": 0.0718, + "step": 16093 + }, + { + "epoch": 2.11, + "learning_rate": 1.7273824028190225e-05, + "loss": 0.0637, + "step": 16094 + }, + { + "epoch": 2.11, + "learning_rate": 1.727346009174531e-05, + "loss": 0.0732, + "step": 16095 + }, + { + "epoch": 2.11, + "learning_rate": 1.727309613484417e-05, + "loss": 0.0493, + "step": 16096 + }, + { + "epoch": 2.11, + "learning_rate": 1.727273215748784e-05, + "loss": 0.0591, + "step": 16097 + }, + { + "epoch": 2.11, + "learning_rate": 1.7272368159677335e-05, + "loss": 0.0805, + "step": 16098 + }, + { + "epoch": 2.11, + "learning_rate": 1.7272004141413683e-05, + "loss": 0.0689, + "step": 16099 + }, + { + "epoch": 2.11, + "learning_rate": 1.7271640102697905e-05, + "loss": 0.0699, + "step": 16100 + }, + { + "epoch": 2.11, + "learning_rate": 1.727127604353103e-05, + "loss": 0.0696, + "step": 16101 + }, + { + "epoch": 2.11, + "learning_rate": 1.7270911963914076e-05, + "loss": 0.068, + "step": 16102 + }, + { + "epoch": 2.11, + "learning_rate": 1.727054786384807e-05, + "loss": 0.0754, + "step": 16103 + }, + { + "epoch": 2.11, + "learning_rate": 1.7270183743334034e-05, + "loss": 0.0585, + "step": 16104 + }, + { + "epoch": 2.11, + "learning_rate": 1.7269819602372993e-05, + "loss": 0.0669, + "step": 16105 + }, + { + "epoch": 2.11, + "learning_rate": 1.7269455440965974e-05, + "loss": 0.0595, + "step": 16106 + }, + { + "epoch": 2.11, + "learning_rate": 1.7269091259113998e-05, + "loss": 0.0793, + "step": 16107 + }, + { + "epoch": 2.11, + "learning_rate": 1.726872705681809e-05, + "loss": 0.0659, + "step": 16108 + }, + { + "epoch": 2.11, + "learning_rate": 1.7268362834079274e-05, + "loss": 0.0614, + "step": 16109 + }, + { + "epoch": 2.11, + "learning_rate": 1.7267998590898572e-05, + "loss": 0.0694, + "step": 16110 + }, + { + "epoch": 2.11, + "learning_rate": 1.7267634327277017e-05, + "loss": 0.0724, + "step": 16111 + }, + { + "epoch": 2.11, + "learning_rate": 1.7267270043215622e-05, + "loss": 0.077, + "step": 16112 + }, + { + "epoch": 2.11, + "learning_rate": 1.726690573871542e-05, + "loss": 0.0745, + "step": 16113 + }, + { + "epoch": 2.11, + "learning_rate": 1.7266541413777435e-05, + "loss": 0.0658, + "step": 16114 + }, + { + "epoch": 2.11, + "learning_rate": 1.7266177068402682e-05, + "loss": 0.0721, + "step": 16115 + }, + { + "epoch": 2.11, + "learning_rate": 1.7265812702592196e-05, + "loss": 0.0691, + "step": 16116 + }, + { + "epoch": 2.11, + "learning_rate": 1.7265448316347e-05, + "loss": 0.0642, + "step": 16117 + }, + { + "epoch": 2.11, + "learning_rate": 1.726508390966812e-05, + "loss": 0.0709, + "step": 16118 + }, + { + "epoch": 2.11, + "learning_rate": 1.7264719482556574e-05, + "loss": 0.0766, + "step": 16119 + }, + { + "epoch": 2.11, + "learning_rate": 1.726435503501339e-05, + "loss": 0.0637, + "step": 16120 + }, + { + "epoch": 2.11, + "learning_rate": 1.72639905670396e-05, + "loss": 0.0667, + "step": 16121 + }, + { + "epoch": 2.11, + "learning_rate": 1.7263626078636218e-05, + "loss": 0.0598, + "step": 16122 + }, + { + "epoch": 2.11, + "learning_rate": 1.7263261569804276e-05, + "loss": 0.0782, + "step": 16123 + }, + { + "epoch": 2.11, + "learning_rate": 1.72628970405448e-05, + "loss": 0.0654, + "step": 16124 + }, + { + "epoch": 2.11, + "learning_rate": 1.7262532490858812e-05, + "loss": 0.0721, + "step": 16125 + }, + { + "epoch": 2.11, + "learning_rate": 1.7262167920747333e-05, + "loss": 0.0663, + "step": 16126 + }, + { + "epoch": 2.11, + "learning_rate": 1.72618033302114e-05, + "loss": 0.0735, + "step": 16127 + }, + { + "epoch": 2.11, + "learning_rate": 1.726143871925203e-05, + "loss": 0.064, + "step": 16128 + }, + { + "epoch": 2.11, + "learning_rate": 1.726107408787025e-05, + "loss": 0.0624, + "step": 16129 + }, + { + "epoch": 2.11, + "learning_rate": 1.7260709436067086e-05, + "loss": 0.0535, + "step": 16130 + }, + { + "epoch": 2.11, + "learning_rate": 1.7260344763843563e-05, + "loss": 0.0531, + "step": 16131 + }, + { + "epoch": 2.11, + "learning_rate": 1.7259980071200705e-05, + "loss": 0.0732, + "step": 16132 + }, + { + "epoch": 2.11, + "learning_rate": 1.7259615358139544e-05, + "loss": 0.0742, + "step": 16133 + }, + { + "epoch": 2.11, + "learning_rate": 1.72592506246611e-05, + "loss": 0.0631, + "step": 16134 + }, + { + "epoch": 2.11, + "learning_rate": 1.7258885870766398e-05, + "loss": 0.0455, + "step": 16135 + }, + { + "epoch": 2.11, + "learning_rate": 1.7258521096456468e-05, + "loss": 0.0617, + "step": 16136 + }, + { + "epoch": 2.11, + "learning_rate": 1.7258156301732333e-05, + "loss": 0.0714, + "step": 16137 + }, + { + "epoch": 2.11, + "learning_rate": 1.7257791486595022e-05, + "loss": 0.0655, + "step": 16138 + }, + { + "epoch": 2.11, + "learning_rate": 1.7257426651045555e-05, + "loss": 0.0645, + "step": 16139 + }, + { + "epoch": 2.11, + "learning_rate": 1.7257061795084963e-05, + "loss": 0.071, + "step": 16140 + }, + { + "epoch": 2.11, + "learning_rate": 1.7256696918714272e-05, + "loss": 0.0717, + "step": 16141 + }, + { + "epoch": 2.11, + "learning_rate": 1.725633202193451e-05, + "loss": 0.0576, + "step": 16142 + }, + { + "epoch": 2.11, + "learning_rate": 1.7255967104746698e-05, + "loss": 0.0586, + "step": 16143 + }, + { + "epoch": 2.11, + "learning_rate": 1.7255602167151866e-05, + "loss": 0.0631, + "step": 16144 + }, + { + "epoch": 2.11, + "learning_rate": 1.7255237209151037e-05, + "loss": 0.065, + "step": 16145 + }, + { + "epoch": 2.12, + "learning_rate": 1.7254872230745242e-05, + "loss": 0.0615, + "step": 16146 + }, + { + "epoch": 2.12, + "learning_rate": 1.7254507231935505e-05, + "loss": 0.0613, + "step": 16147 + }, + { + "epoch": 2.12, + "learning_rate": 1.7254142212722854e-05, + "loss": 0.0707, + "step": 16148 + }, + { + "epoch": 2.12, + "learning_rate": 1.725377717310831e-05, + "loss": 0.0712, + "step": 16149 + }, + { + "epoch": 2.12, + "learning_rate": 1.7253412113092908e-05, + "loss": 0.0728, + "step": 16150 + }, + { + "epoch": 2.12, + "learning_rate": 1.7253047032677668e-05, + "loss": 0.0708, + "step": 16151 + }, + { + "epoch": 2.12, + "learning_rate": 1.7252681931863622e-05, + "loss": 0.0624, + "step": 16152 + }, + { + "epoch": 2.12, + "learning_rate": 1.7252316810651792e-05, + "loss": 0.0711, + "step": 16153 + }, + { + "epoch": 2.12, + "learning_rate": 1.725195166904321e-05, + "loss": 0.0699, + "step": 16154 + }, + { + "epoch": 2.12, + "learning_rate": 1.7251586507038897e-05, + "loss": 0.0735, + "step": 16155 + }, + { + "epoch": 2.12, + "learning_rate": 1.7251221324639888e-05, + "loss": 0.069, + "step": 16156 + }, + { + "epoch": 2.12, + "learning_rate": 1.7250856121847198e-05, + "loss": 0.0689, + "step": 16157 + }, + { + "epoch": 2.12, + "learning_rate": 1.7250490898661865e-05, + "loss": 0.0668, + "step": 16158 + }, + { + "epoch": 2.12, + "learning_rate": 1.725012565508491e-05, + "loss": 0.0714, + "step": 16159 + }, + { + "epoch": 2.12, + "learning_rate": 1.7249760391117368e-05, + "loss": 0.0584, + "step": 16160 + }, + { + "epoch": 2.12, + "learning_rate": 1.7249395106760257e-05, + "loss": 0.0726, + "step": 16161 + }, + { + "epoch": 2.12, + "learning_rate": 1.724902980201461e-05, + "loss": 0.0577, + "step": 16162 + }, + { + "epoch": 2.12, + "learning_rate": 1.724866447688145e-05, + "loss": 0.0723, + "step": 16163 + }, + { + "epoch": 2.12, + "learning_rate": 1.7248299131361813e-05, + "loss": 0.0582, + "step": 16164 + }, + { + "epoch": 2.12, + "learning_rate": 1.7247933765456713e-05, + "loss": 0.0554, + "step": 16165 + }, + { + "epoch": 2.12, + "learning_rate": 1.724756837916719e-05, + "loss": 0.0665, + "step": 16166 + }, + { + "epoch": 2.12, + "learning_rate": 1.7247202972494267e-05, + "loss": 0.0756, + "step": 16167 + }, + { + "epoch": 2.12, + "learning_rate": 1.724683754543897e-05, + "loss": 0.0662, + "step": 16168 + }, + { + "epoch": 2.12, + "learning_rate": 1.724647209800233e-05, + "loss": 0.0543, + "step": 16169 + }, + { + "epoch": 2.12, + "learning_rate": 1.7246106630185374e-05, + "loss": 0.069, + "step": 16170 + }, + { + "epoch": 2.12, + "learning_rate": 1.7245741141989124e-05, + "loss": 0.061, + "step": 16171 + }, + { + "epoch": 2.12, + "learning_rate": 1.7245375633414617e-05, + "loss": 0.0739, + "step": 16172 + }, + { + "epoch": 2.12, + "learning_rate": 1.7245010104462875e-05, + "loss": 0.0706, + "step": 16173 + }, + { + "epoch": 2.12, + "learning_rate": 1.724464455513493e-05, + "loss": 0.0648, + "step": 16174 + }, + { + "epoch": 2.12, + "learning_rate": 1.7244278985431805e-05, + "loss": 0.0698, + "step": 16175 + }, + { + "epoch": 2.12, + "learning_rate": 1.7243913395354534e-05, + "loss": 0.0606, + "step": 16176 + }, + { + "epoch": 2.12, + "learning_rate": 1.724354778490414e-05, + "loss": 0.0638, + "step": 16177 + }, + { + "epoch": 2.12, + "learning_rate": 1.7243182154081655e-05, + "loss": 0.0632, + "step": 16178 + }, + { + "epoch": 2.12, + "learning_rate": 1.7242816502888107e-05, + "loss": 0.068, + "step": 16179 + }, + { + "epoch": 2.12, + "learning_rate": 1.724245083132452e-05, + "loss": 0.0694, + "step": 16180 + }, + { + "epoch": 2.12, + "learning_rate": 1.724208513939193e-05, + "loss": 0.0737, + "step": 16181 + }, + { + "epoch": 2.12, + "learning_rate": 1.724171942709136e-05, + "loss": 0.0676, + "step": 16182 + }, + { + "epoch": 2.12, + "learning_rate": 1.7241353694423836e-05, + "loss": 0.0682, + "step": 16183 + }, + { + "epoch": 2.12, + "learning_rate": 1.7240987941390392e-05, + "loss": 0.0796, + "step": 16184 + }, + { + "epoch": 2.12, + "learning_rate": 1.7240622167992057e-05, + "loss": 0.0589, + "step": 16185 + }, + { + "epoch": 2.12, + "learning_rate": 1.7240256374229857e-05, + "loss": 0.0654, + "step": 16186 + }, + { + "epoch": 2.12, + "learning_rate": 1.7239890560104822e-05, + "loss": 0.0575, + "step": 16187 + }, + { + "epoch": 2.12, + "learning_rate": 1.723952472561798e-05, + "loss": 0.0475, + "step": 16188 + }, + { + "epoch": 2.12, + "learning_rate": 1.723915887077036e-05, + "loss": 0.0652, + "step": 16189 + }, + { + "epoch": 2.12, + "learning_rate": 1.7238792995562993e-05, + "loss": 0.0735, + "step": 16190 + }, + { + "epoch": 2.12, + "learning_rate": 1.7238427099996903e-05, + "loss": 0.0627, + "step": 16191 + }, + { + "epoch": 2.12, + "learning_rate": 1.7238061184073124e-05, + "loss": 0.0642, + "step": 16192 + }, + { + "epoch": 2.12, + "learning_rate": 1.723769524779268e-05, + "loss": 0.0608, + "step": 16193 + }, + { + "epoch": 2.12, + "learning_rate": 1.723732929115661e-05, + "loss": 0.075, + "step": 16194 + }, + { + "epoch": 2.12, + "learning_rate": 1.723696331416593e-05, + "loss": 0.0656, + "step": 16195 + }, + { + "epoch": 2.12, + "learning_rate": 1.7236597316821678e-05, + "loss": 0.081, + "step": 16196 + }, + { + "epoch": 2.12, + "learning_rate": 1.7236231299124885e-05, + "loss": 0.065, + "step": 16197 + }, + { + "epoch": 2.12, + "learning_rate": 1.7235865261076577e-05, + "loss": 0.0779, + "step": 16198 + }, + { + "epoch": 2.12, + "learning_rate": 1.7235499202677778e-05, + "loss": 0.0562, + "step": 16199 + }, + { + "epoch": 2.12, + "learning_rate": 1.7235133123929527e-05, + "loss": 0.0662, + "step": 16200 + }, + { + "epoch": 2.12, + "learning_rate": 1.7234767024832847e-05, + "loss": 0.0641, + "step": 16201 + }, + { + "epoch": 2.12, + "learning_rate": 1.723440090538877e-05, + "loss": 0.0733, + "step": 16202 + }, + { + "epoch": 2.12, + "learning_rate": 1.723403476559833e-05, + "loss": 0.071, + "step": 16203 + }, + { + "epoch": 2.12, + "learning_rate": 1.723366860546255e-05, + "loss": 0.067, + "step": 16204 + }, + { + "epoch": 2.12, + "learning_rate": 1.723330242498246e-05, + "loss": 0.0719, + "step": 16205 + }, + { + "epoch": 2.12, + "learning_rate": 1.723293622415909e-05, + "loss": 0.0747, + "step": 16206 + }, + { + "epoch": 2.12, + "learning_rate": 1.7232570002993475e-05, + "loss": 0.0653, + "step": 16207 + }, + { + "epoch": 2.12, + "learning_rate": 1.7232203761486644e-05, + "loss": 0.0634, + "step": 16208 + }, + { + "epoch": 2.12, + "learning_rate": 1.7231837499639624e-05, + "loss": 0.0653, + "step": 16209 + }, + { + "epoch": 2.12, + "learning_rate": 1.7231471217453446e-05, + "loss": 0.0724, + "step": 16210 + }, + { + "epoch": 2.12, + "learning_rate": 1.7231104914929143e-05, + "loss": 0.0636, + "step": 16211 + }, + { + "epoch": 2.12, + "learning_rate": 1.7230738592067738e-05, + "loss": 0.0603, + "step": 16212 + }, + { + "epoch": 2.12, + "learning_rate": 1.7230372248870267e-05, + "loss": 0.0649, + "step": 16213 + }, + { + "epoch": 2.12, + "learning_rate": 1.723000588533776e-05, + "loss": 0.0669, + "step": 16214 + }, + { + "epoch": 2.12, + "learning_rate": 1.7229639501471245e-05, + "loss": 0.0663, + "step": 16215 + }, + { + "epoch": 2.12, + "learning_rate": 1.7229273097271753e-05, + "loss": 0.0577, + "step": 16216 + }, + { + "epoch": 2.12, + "learning_rate": 1.722890667274032e-05, + "loss": 0.0792, + "step": 16217 + }, + { + "epoch": 2.12, + "learning_rate": 1.722854022787797e-05, + "loss": 0.0748, + "step": 16218 + }, + { + "epoch": 2.12, + "learning_rate": 1.7228173762685733e-05, + "loss": 0.0686, + "step": 16219 + }, + { + "epoch": 2.12, + "learning_rate": 1.7227807277164644e-05, + "loss": 0.0699, + "step": 16220 + }, + { + "epoch": 2.12, + "learning_rate": 1.7227440771315733e-05, + "loss": 0.0598, + "step": 16221 + }, + { + "epoch": 2.12, + "learning_rate": 1.7227074245140028e-05, + "loss": 0.0625, + "step": 16222 + }, + { + "epoch": 2.13, + "learning_rate": 1.7226707698638563e-05, + "loss": 0.0618, + "step": 16223 + }, + { + "epoch": 2.13, + "learning_rate": 1.7226341131812364e-05, + "loss": 0.0633, + "step": 16224 + }, + { + "epoch": 2.13, + "learning_rate": 1.722597454466247e-05, + "loss": 0.0658, + "step": 16225 + }, + { + "epoch": 2.13, + "learning_rate": 1.7225607937189907e-05, + "loss": 0.0651, + "step": 16226 + }, + { + "epoch": 2.13, + "learning_rate": 1.7225241309395706e-05, + "loss": 0.0644, + "step": 16227 + }, + { + "epoch": 2.13, + "learning_rate": 1.7224874661280893e-05, + "loss": 0.0707, + "step": 16228 + }, + { + "epoch": 2.13, + "learning_rate": 1.722450799284651e-05, + "loss": 0.0684, + "step": 16229 + }, + { + "epoch": 2.13, + "learning_rate": 1.7224141304093582e-05, + "loss": 0.0712, + "step": 16230 + }, + { + "epoch": 2.13, + "learning_rate": 1.7223774595023144e-05, + "loss": 0.0686, + "step": 16231 + }, + { + "epoch": 2.13, + "learning_rate": 1.7223407865636218e-05, + "loss": 0.0738, + "step": 16232 + }, + { + "epoch": 2.13, + "learning_rate": 1.7223041115933846e-05, + "loss": 0.0597, + "step": 16233 + }, + { + "epoch": 2.13, + "learning_rate": 1.7222674345917057e-05, + "loss": 0.0652, + "step": 16234 + }, + { + "epoch": 2.13, + "learning_rate": 1.722230755558688e-05, + "loss": 0.0703, + "step": 16235 + }, + { + "epoch": 2.13, + "learning_rate": 1.7221940744944347e-05, + "loss": 0.0589, + "step": 16236 + }, + { + "epoch": 2.13, + "learning_rate": 1.7221573913990493e-05, + "loss": 0.071, + "step": 16237 + }, + { + "epoch": 2.13, + "learning_rate": 1.7221207062726343e-05, + "loss": 0.0563, + "step": 16238 + }, + { + "epoch": 2.13, + "learning_rate": 1.7220840191152933e-05, + "loss": 0.0665, + "step": 16239 + }, + { + "epoch": 2.13, + "learning_rate": 1.7220473299271297e-05, + "loss": 0.068, + "step": 16240 + }, + { + "epoch": 2.13, + "learning_rate": 1.7220106387082464e-05, + "loss": 0.074, + "step": 16241 + }, + { + "epoch": 2.13, + "learning_rate": 1.7219739454587465e-05, + "loss": 0.0668, + "step": 16242 + }, + { + "epoch": 2.13, + "learning_rate": 1.7219372501787336e-05, + "loss": 0.07, + "step": 16243 + }, + { + "epoch": 2.13, + "learning_rate": 1.72190055286831e-05, + "loss": 0.0708, + "step": 16244 + }, + { + "epoch": 2.13, + "learning_rate": 1.72186385352758e-05, + "loss": 0.0791, + "step": 16245 + }, + { + "epoch": 2.13, + "learning_rate": 1.7218271521566465e-05, + "loss": 0.0685, + "step": 16246 + }, + { + "epoch": 2.13, + "learning_rate": 1.7217904487556123e-05, + "loss": 0.0608, + "step": 16247 + }, + { + "epoch": 2.13, + "learning_rate": 1.721753743324581e-05, + "loss": 0.0726, + "step": 16248 + }, + { + "epoch": 2.13, + "learning_rate": 1.7217170358636555e-05, + "loss": 0.0726, + "step": 16249 + }, + { + "epoch": 2.13, + "learning_rate": 1.7216803263729396e-05, + "loss": 0.0745, + "step": 16250 + }, + { + "epoch": 2.13, + "learning_rate": 1.721643614852536e-05, + "loss": 0.0748, + "step": 16251 + }, + { + "epoch": 2.13, + "learning_rate": 1.7216069013025484e-05, + "loss": 0.075, + "step": 16252 + }, + { + "epoch": 2.13, + "learning_rate": 1.7215701857230795e-05, + "loss": 0.0709, + "step": 16253 + }, + { + "epoch": 2.13, + "learning_rate": 1.7215334681142328e-05, + "loss": 0.0597, + "step": 16254 + }, + { + "epoch": 2.13, + "learning_rate": 1.721496748476112e-05, + "loss": 0.0694, + "step": 16255 + }, + { + "epoch": 2.13, + "learning_rate": 1.72146002680882e-05, + "loss": 0.0734, + "step": 16256 + }, + { + "epoch": 2.13, + "learning_rate": 1.7214233031124597e-05, + "loss": 0.0621, + "step": 16257 + }, + { + "epoch": 2.13, + "learning_rate": 1.7213865773871353e-05, + "loss": 0.0619, + "step": 16258 + }, + { + "epoch": 2.13, + "learning_rate": 1.7213498496329494e-05, + "loss": 0.0731, + "step": 16259 + }, + { + "epoch": 2.13, + "learning_rate": 1.7213131198500055e-05, + "loss": 0.0687, + "step": 16260 + }, + { + "epoch": 2.13, + "learning_rate": 1.7212763880384068e-05, + "loss": 0.0727, + "step": 16261 + }, + { + "epoch": 2.13, + "learning_rate": 1.7212396541982563e-05, + "loss": 0.0637, + "step": 16262 + }, + { + "epoch": 2.13, + "learning_rate": 1.721202918329658e-05, + "loss": 0.0657, + "step": 16263 + }, + { + "epoch": 2.13, + "learning_rate": 1.7211661804327152e-05, + "loss": 0.0603, + "step": 16264 + }, + { + "epoch": 2.13, + "learning_rate": 1.7211294405075305e-05, + "loss": 0.0597, + "step": 16265 + }, + { + "epoch": 2.13, + "learning_rate": 1.7210926985542078e-05, + "loss": 0.0663, + "step": 16266 + }, + { + "epoch": 2.13, + "learning_rate": 1.7210559545728504e-05, + "loss": 0.0636, + "step": 16267 + }, + { + "epoch": 2.13, + "learning_rate": 1.7210192085635612e-05, + "loss": 0.0631, + "step": 16268 + }, + { + "epoch": 2.13, + "learning_rate": 1.720982460526444e-05, + "loss": 0.0672, + "step": 16269 + }, + { + "epoch": 2.13, + "learning_rate": 1.720945710461602e-05, + "loss": 0.0652, + "step": 16270 + }, + { + "epoch": 2.13, + "learning_rate": 1.720908958369139e-05, + "loss": 0.0635, + "step": 16271 + }, + { + "epoch": 2.13, + "learning_rate": 1.7208722042491574e-05, + "loss": 0.0604, + "step": 16272 + }, + { + "epoch": 2.13, + "learning_rate": 1.7208354481017614e-05, + "loss": 0.0696, + "step": 16273 + }, + { + "epoch": 2.13, + "learning_rate": 1.720798689927054e-05, + "loss": 0.0666, + "step": 16274 + }, + { + "epoch": 2.13, + "learning_rate": 1.720761929725139e-05, + "loss": 0.0706, + "step": 16275 + }, + { + "epoch": 2.13, + "learning_rate": 1.720725167496119e-05, + "loss": 0.0613, + "step": 16276 + }, + { + "epoch": 2.13, + "learning_rate": 1.7206884032400978e-05, + "loss": 0.0637, + "step": 16277 + }, + { + "epoch": 2.13, + "learning_rate": 1.720651636957179e-05, + "loss": 0.0593, + "step": 16278 + }, + { + "epoch": 2.13, + "learning_rate": 1.720614868647466e-05, + "loss": 0.0768, + "step": 16279 + }, + { + "epoch": 2.13, + "learning_rate": 1.720578098311062e-05, + "loss": 0.0599, + "step": 16280 + }, + { + "epoch": 2.13, + "learning_rate": 1.7205413259480702e-05, + "loss": 0.0675, + "step": 16281 + }, + { + "epoch": 2.13, + "learning_rate": 1.7205045515585945e-05, + "loss": 0.0688, + "step": 16282 + }, + { + "epoch": 2.13, + "learning_rate": 1.7204677751427385e-05, + "loss": 0.0686, + "step": 16283 + }, + { + "epoch": 2.13, + "learning_rate": 1.720430996700605e-05, + "loss": 0.0683, + "step": 16284 + }, + { + "epoch": 2.13, + "learning_rate": 1.7203942162322973e-05, + "loss": 0.0627, + "step": 16285 + }, + { + "epoch": 2.13, + "learning_rate": 1.7203574337379194e-05, + "loss": 0.0659, + "step": 16286 + }, + { + "epoch": 2.13, + "learning_rate": 1.7203206492175746e-05, + "loss": 0.0645, + "step": 16287 + }, + { + "epoch": 2.13, + "learning_rate": 1.720283862671366e-05, + "loss": 0.0683, + "step": 16288 + }, + { + "epoch": 2.13, + "learning_rate": 1.720247074099398e-05, + "loss": 0.0635, + "step": 16289 + }, + { + "epoch": 2.13, + "learning_rate": 1.7202102835017735e-05, + "loss": 0.0778, + "step": 16290 + }, + { + "epoch": 2.13, + "learning_rate": 1.7201734908785957e-05, + "loss": 0.0752, + "step": 16291 + }, + { + "epoch": 2.13, + "learning_rate": 1.7201366962299682e-05, + "loss": 0.0616, + "step": 16292 + }, + { + "epoch": 2.13, + "learning_rate": 1.720099899555995e-05, + "loss": 0.0672, + "step": 16293 + }, + { + "epoch": 2.13, + "learning_rate": 1.7200631008567786e-05, + "loss": 0.0613, + "step": 16294 + }, + { + "epoch": 2.13, + "learning_rate": 1.7200263001324232e-05, + "loss": 0.0709, + "step": 16295 + }, + { + "epoch": 2.13, + "learning_rate": 1.7199894973830325e-05, + "loss": 0.0697, + "step": 16296 + }, + { + "epoch": 2.13, + "learning_rate": 1.7199526926087093e-05, + "loss": 0.0582, + "step": 16297 + }, + { + "epoch": 2.13, + "learning_rate": 1.719915885809558e-05, + "loss": 0.0585, + "step": 16298 + }, + { + "epoch": 2.14, + "learning_rate": 1.7198790769856812e-05, + "loss": 0.0549, + "step": 16299 + }, + { + "epoch": 2.14, + "learning_rate": 1.719842266137183e-05, + "loss": 0.0639, + "step": 16300 + }, + { + "epoch": 2.14, + "learning_rate": 1.7198054532641664e-05, + "loss": 0.0689, + "step": 16301 + }, + { + "epoch": 2.14, + "learning_rate": 1.719768638366736e-05, + "loss": 0.0695, + "step": 16302 + }, + { + "epoch": 2.14, + "learning_rate": 1.719731821444994e-05, + "loss": 0.0764, + "step": 16303 + }, + { + "epoch": 2.14, + "learning_rate": 1.7196950024990448e-05, + "loss": 0.0659, + "step": 16304 + }, + { + "epoch": 2.14, + "learning_rate": 1.719658181528992e-05, + "loss": 0.0677, + "step": 16305 + }, + { + "epoch": 2.14, + "learning_rate": 1.7196213585349384e-05, + "loss": 0.0715, + "step": 16306 + }, + { + "epoch": 2.14, + "learning_rate": 1.7195845335169882e-05, + "loss": 0.0601, + "step": 16307 + }, + { + "epoch": 2.14, + "learning_rate": 1.719547706475245e-05, + "loss": 0.0634, + "step": 16308 + }, + { + "epoch": 2.14, + "learning_rate": 1.719510877409812e-05, + "loss": 0.0698, + "step": 16309 + }, + { + "epoch": 2.14, + "learning_rate": 1.7194740463207932e-05, + "loss": 0.0698, + "step": 16310 + }, + { + "epoch": 2.14, + "learning_rate": 1.719437213208292e-05, + "loss": 0.0736, + "step": 16311 + }, + { + "epoch": 2.14, + "learning_rate": 1.7194003780724115e-05, + "loss": 0.0611, + "step": 16312 + }, + { + "epoch": 2.14, + "learning_rate": 1.719363540913256e-05, + "loss": 0.0588, + "step": 16313 + }, + { + "epoch": 2.14, + "learning_rate": 1.7193267017309288e-05, + "loss": 0.0661, + "step": 16314 + }, + { + "epoch": 2.14, + "learning_rate": 1.7192898605255335e-05, + "loss": 0.0627, + "step": 16315 + }, + { + "epoch": 2.14, + "learning_rate": 1.7192530172971738e-05, + "loss": 0.0727, + "step": 16316 + }, + { + "epoch": 2.14, + "learning_rate": 1.7192161720459534e-05, + "loss": 0.0719, + "step": 16317 + }, + { + "epoch": 2.14, + "learning_rate": 1.719179324771976e-05, + "loss": 0.0621, + "step": 16318 + }, + { + "epoch": 2.14, + "learning_rate": 1.7191424754753446e-05, + "loss": 0.0692, + "step": 16319 + }, + { + "epoch": 2.14, + "learning_rate": 1.719105624156164e-05, + "loss": 0.0645, + "step": 16320 + }, + { + "epoch": 2.14, + "learning_rate": 1.7190687708145364e-05, + "loss": 0.0708, + "step": 16321 + }, + { + "epoch": 2.14, + "learning_rate": 1.7190319154505663e-05, + "loss": 0.0679, + "step": 16322 + }, + { + "epoch": 2.14, + "learning_rate": 1.7189950580643575e-05, + "loss": 0.0698, + "step": 16323 + }, + { + "epoch": 2.14, + "learning_rate": 1.7189581986560135e-05, + "loss": 0.0708, + "step": 16324 + }, + { + "epoch": 2.14, + "learning_rate": 1.718921337225637e-05, + "loss": 0.0623, + "step": 16325 + }, + { + "epoch": 2.14, + "learning_rate": 1.7188844737733334e-05, + "loss": 0.0553, + "step": 16326 + }, + { + "epoch": 2.14, + "learning_rate": 1.7188476082992052e-05, + "loss": 0.0771, + "step": 16327 + }, + { + "epoch": 2.14, + "learning_rate": 1.7188107408033567e-05, + "loss": 0.0748, + "step": 16328 + }, + { + "epoch": 2.14, + "learning_rate": 1.718773871285891e-05, + "loss": 0.0695, + "step": 16329 + }, + { + "epoch": 2.14, + "learning_rate": 1.718736999746912e-05, + "loss": 0.0639, + "step": 16330 + }, + { + "epoch": 2.14, + "learning_rate": 1.7187001261865232e-05, + "loss": 0.0718, + "step": 16331 + }, + { + "epoch": 2.14, + "learning_rate": 1.7186632506048288e-05, + "loss": 0.068, + "step": 16332 + }, + { + "epoch": 2.14, + "learning_rate": 1.7186263730019327e-05, + "loss": 0.069, + "step": 16333 + }, + { + "epoch": 2.14, + "learning_rate": 1.7185894933779375e-05, + "loss": 0.0735, + "step": 16334 + }, + { + "epoch": 2.14, + "learning_rate": 1.718552611732948e-05, + "loss": 0.0688, + "step": 16335 + }, + { + "epoch": 2.14, + "learning_rate": 1.7185157280670678e-05, + "loss": 0.0678, + "step": 16336 + }, + { + "epoch": 2.14, + "learning_rate": 1.7184788423804e-05, + "loss": 0.0622, + "step": 16337 + }, + { + "epoch": 2.14, + "learning_rate": 1.7184419546730487e-05, + "loss": 0.0688, + "step": 16338 + }, + { + "epoch": 2.14, + "learning_rate": 1.7184050649451176e-05, + "loss": 0.0666, + "step": 16339 + }, + { + "epoch": 2.14, + "learning_rate": 1.7183681731967107e-05, + "loss": 0.0667, + "step": 16340 + }, + { + "epoch": 2.14, + "learning_rate": 1.7183312794279317e-05, + "loss": 0.0576, + "step": 16341 + }, + { + "epoch": 2.14, + "learning_rate": 1.7182943836388842e-05, + "loss": 0.0705, + "step": 16342 + }, + { + "epoch": 2.14, + "learning_rate": 1.7182574858296716e-05, + "loss": 0.0605, + "step": 16343 + }, + { + "epoch": 2.14, + "learning_rate": 1.7182205860003983e-05, + "loss": 0.0697, + "step": 16344 + }, + { + "epoch": 2.14, + "learning_rate": 1.7181836841511682e-05, + "loss": 0.0718, + "step": 16345 + }, + { + "epoch": 2.14, + "learning_rate": 1.7181467802820844e-05, + "loss": 0.0738, + "step": 16346 + }, + { + "epoch": 2.14, + "learning_rate": 1.718109874393251e-05, + "loss": 0.0693, + "step": 16347 + }, + { + "epoch": 2.14, + "learning_rate": 1.7180729664847716e-05, + "loss": 0.0613, + "step": 16348 + }, + { + "epoch": 2.14, + "learning_rate": 1.7180360565567507e-05, + "loss": 0.0761, + "step": 16349 + }, + { + "epoch": 2.14, + "learning_rate": 1.7179991446092917e-05, + "loss": 0.0685, + "step": 16350 + }, + { + "epoch": 2.14, + "learning_rate": 1.7179622306424978e-05, + "loss": 0.0623, + "step": 16351 + }, + { + "epoch": 2.14, + "learning_rate": 1.7179253146564735e-05, + "loss": 0.0709, + "step": 16352 + }, + { + "epoch": 2.14, + "learning_rate": 1.7178883966513225e-05, + "loss": 0.0631, + "step": 16353 + }, + { + "epoch": 2.14, + "learning_rate": 1.717851476627149e-05, + "loss": 0.0627, + "step": 16354 + }, + { + "epoch": 2.14, + "learning_rate": 1.717814554584056e-05, + "loss": 0.062, + "step": 16355 + }, + { + "epoch": 2.14, + "learning_rate": 1.717777630522148e-05, + "loss": 0.0719, + "step": 16356 + }, + { + "epoch": 2.14, + "learning_rate": 1.717740704441529e-05, + "loss": 0.0638, + "step": 16357 + }, + { + "epoch": 2.14, + "learning_rate": 1.717703776342302e-05, + "loss": 0.0735, + "step": 16358 + }, + { + "epoch": 2.14, + "learning_rate": 1.7176668462245714e-05, + "loss": 0.0652, + "step": 16359 + }, + { + "epoch": 2.14, + "learning_rate": 1.717629914088441e-05, + "loss": 0.0675, + "step": 16360 + }, + { + "epoch": 2.14, + "learning_rate": 1.7175929799340146e-05, + "loss": 0.0715, + "step": 16361 + }, + { + "epoch": 2.14, + "learning_rate": 1.7175560437613964e-05, + "loss": 0.0706, + "step": 16362 + }, + { + "epoch": 2.14, + "learning_rate": 1.71751910557069e-05, + "loss": 0.0674, + "step": 16363 + }, + { + "epoch": 2.14, + "learning_rate": 1.717482165361999e-05, + "loss": 0.0594, + "step": 16364 + }, + { + "epoch": 2.14, + "learning_rate": 1.717445223135428e-05, + "loss": 0.053, + "step": 16365 + }, + { + "epoch": 2.14, + "learning_rate": 1.7174082788910803e-05, + "loss": 0.0601, + "step": 16366 + }, + { + "epoch": 2.14, + "learning_rate": 1.7173713326290603e-05, + "loss": 0.069, + "step": 16367 + }, + { + "epoch": 2.14, + "learning_rate": 1.7173343843494716e-05, + "loss": 0.0522, + "step": 16368 + }, + { + "epoch": 2.14, + "learning_rate": 1.7172974340524174e-05, + "loss": 0.0678, + "step": 16369 + }, + { + "epoch": 2.14, + "learning_rate": 1.7172604817380035e-05, + "loss": 0.0719, + "step": 16370 + }, + { + "epoch": 2.14, + "learning_rate": 1.7172235274063318e-05, + "loss": 0.067, + "step": 16371 + }, + { + "epoch": 2.14, + "learning_rate": 1.7171865710575074e-05, + "loss": 0.0679, + "step": 16372 + }, + { + "epoch": 2.14, + "learning_rate": 1.717149612691634e-05, + "loss": 0.056, + "step": 16373 + }, + { + "epoch": 2.14, + "learning_rate": 1.7171126523088158e-05, + "loss": 0.0716, + "step": 16374 + }, + { + "epoch": 2.15, + "learning_rate": 1.717075689909156e-05, + "loss": 0.0723, + "step": 16375 + }, + { + "epoch": 2.15, + "learning_rate": 1.7170387254927595e-05, + "loss": 0.0635, + "step": 16376 + }, + { + "epoch": 2.15, + "learning_rate": 1.7170017590597293e-05, + "loss": 0.0661, + "step": 16377 + }, + { + "epoch": 2.15, + "learning_rate": 1.71696479061017e-05, + "loss": 0.0605, + "step": 16378 + }, + { + "epoch": 2.15, + "learning_rate": 1.716927820144185e-05, + "loss": 0.0661, + "step": 16379 + }, + { + "epoch": 2.15, + "learning_rate": 1.716890847661879e-05, + "loss": 0.0609, + "step": 16380 + }, + { + "epoch": 2.15, + "learning_rate": 1.7168538731633565e-05, + "loss": 0.0679, + "step": 16381 + }, + { + "epoch": 2.15, + "learning_rate": 1.7168168966487197e-05, + "loss": 0.0703, + "step": 16382 + }, + { + "epoch": 2.15, + "learning_rate": 1.7167799181180736e-05, + "loss": 0.0716, + "step": 16383 + }, + { + "epoch": 2.15, + "learning_rate": 1.7167429375715223e-05, + "loss": 0.0604, + "step": 16384 + }, + { + "epoch": 2.15, + "learning_rate": 1.71670595500917e-05, + "loss": 0.0701, + "step": 16385 + }, + { + "epoch": 2.15, + "learning_rate": 1.71666897043112e-05, + "loss": 0.0573, + "step": 16386 + }, + { + "epoch": 2.15, + "learning_rate": 1.7166319838374767e-05, + "loss": 0.0741, + "step": 16387 + }, + { + "epoch": 2.15, + "learning_rate": 1.716594995228344e-05, + "loss": 0.0676, + "step": 16388 + }, + { + "epoch": 2.15, + "learning_rate": 1.7165580046038262e-05, + "loss": 0.063, + "step": 16389 + }, + { + "epoch": 2.15, + "learning_rate": 1.7165210119640273e-05, + "loss": 0.0603, + "step": 16390 + }, + { + "epoch": 2.15, + "learning_rate": 1.716484017309051e-05, + "loss": 0.0539, + "step": 16391 + }, + { + "epoch": 2.15, + "learning_rate": 1.7164470206390012e-05, + "loss": 0.0604, + "step": 16392 + }, + { + "epoch": 2.15, + "learning_rate": 1.716410021953983e-05, + "loss": 0.0725, + "step": 16393 + }, + { + "epoch": 2.15, + "learning_rate": 1.7163730212540995e-05, + "loss": 0.077, + "step": 16394 + }, + { + "epoch": 2.15, + "learning_rate": 1.716336018539455e-05, + "loss": 0.072, + "step": 16395 + }, + { + "epoch": 2.15, + "learning_rate": 1.7162990138101533e-05, + "loss": 0.0665, + "step": 16396 + }, + { + "epoch": 2.15, + "learning_rate": 1.716262007066299e-05, + "loss": 0.0737, + "step": 16397 + }, + { + "epoch": 2.15, + "learning_rate": 1.7162249983079964e-05, + "loss": 0.0718, + "step": 16398 + }, + { + "epoch": 2.15, + "learning_rate": 1.7161879875353485e-05, + "loss": 0.0634, + "step": 16399 + }, + { + "epoch": 2.15, + "learning_rate": 1.7161509747484607e-05, + "loss": 0.0661, + "step": 16400 + }, + { + "epoch": 2.15, + "learning_rate": 1.716113959947436e-05, + "loss": 0.0729, + "step": 16401 + }, + { + "epoch": 2.15, + "learning_rate": 1.716076943132379e-05, + "loss": 0.0671, + "step": 16402 + }, + { + "epoch": 2.15, + "learning_rate": 1.7160399243033935e-05, + "loss": 0.0747, + "step": 16403 + }, + { + "epoch": 2.15, + "learning_rate": 1.7160029034605837e-05, + "loss": 0.065, + "step": 16404 + }, + { + "epoch": 2.15, + "learning_rate": 1.7159658806040543e-05, + "loss": 0.0592, + "step": 16405 + }, + { + "epoch": 2.15, + "learning_rate": 1.715928855733909e-05, + "loss": 0.0633, + "step": 16406 + }, + { + "epoch": 2.15, + "learning_rate": 1.7158918288502517e-05, + "loss": 0.065, + "step": 16407 + }, + { + "epoch": 2.15, + "learning_rate": 1.715854799953187e-05, + "loss": 0.0728, + "step": 16408 + }, + { + "epoch": 2.15, + "learning_rate": 1.7158177690428185e-05, + "loss": 0.0654, + "step": 16409 + }, + { + "epoch": 2.15, + "learning_rate": 1.7157807361192507e-05, + "loss": 0.0696, + "step": 16410 + }, + { + "epoch": 2.15, + "learning_rate": 1.7157437011825878e-05, + "loss": 0.0695, + "step": 16411 + }, + { + "epoch": 2.15, + "learning_rate": 1.715706664232934e-05, + "loss": 0.058, + "step": 16412 + }, + { + "epoch": 2.15, + "learning_rate": 1.7156696252703932e-05, + "loss": 0.0682, + "step": 16413 + }, + { + "epoch": 2.15, + "learning_rate": 1.7156325842950694e-05, + "loss": 0.0576, + "step": 16414 + }, + { + "epoch": 2.15, + "learning_rate": 1.7155955413070674e-05, + "loss": 0.0625, + "step": 16415 + }, + { + "epoch": 2.15, + "learning_rate": 1.715558496306491e-05, + "loss": 0.0618, + "step": 16416 + }, + { + "epoch": 2.15, + "learning_rate": 1.7155214492934442e-05, + "loss": 0.0674, + "step": 16417 + }, + { + "epoch": 2.15, + "learning_rate": 1.7154844002680315e-05, + "loss": 0.0518, + "step": 16418 + }, + { + "epoch": 2.15, + "learning_rate": 1.715447349230357e-05, + "loss": 0.0714, + "step": 16419 + }, + { + "epoch": 2.15, + "learning_rate": 1.7154102961805252e-05, + "loss": 0.056, + "step": 16420 + }, + { + "epoch": 2.15, + "learning_rate": 1.71537324111864e-05, + "loss": 0.0674, + "step": 16421 + }, + { + "epoch": 2.15, + "learning_rate": 1.7153361840448056e-05, + "loss": 0.0606, + "step": 16422 + }, + { + "epoch": 2.15, + "learning_rate": 1.7152991249591264e-05, + "loss": 0.0658, + "step": 16423 + }, + { + "epoch": 2.15, + "learning_rate": 1.715262063861706e-05, + "loss": 0.0694, + "step": 16424 + }, + { + "epoch": 2.15, + "learning_rate": 1.7152250007526492e-05, + "loss": 0.0592, + "step": 16425 + }, + { + "epoch": 2.15, + "learning_rate": 1.7151879356320608e-05, + "loss": 0.0596, + "step": 16426 + }, + { + "epoch": 2.15, + "learning_rate": 1.715150868500044e-05, + "loss": 0.062, + "step": 16427 + }, + { + "epoch": 2.15, + "learning_rate": 1.715113799356704e-05, + "loss": 0.0662, + "step": 16428 + }, + { + "epoch": 2.15, + "learning_rate": 1.7150767282021437e-05, + "loss": 0.0633, + "step": 16429 + }, + { + "epoch": 2.15, + "learning_rate": 1.7150396550364686e-05, + "loss": 0.0629, + "step": 16430 + }, + { + "epoch": 2.15, + "learning_rate": 1.7150025798597823e-05, + "loss": 0.0669, + "step": 16431 + }, + { + "epoch": 2.15, + "learning_rate": 1.7149655026721897e-05, + "loss": 0.0705, + "step": 16432 + }, + { + "epoch": 2.15, + "learning_rate": 1.7149284234737944e-05, + "loss": 0.06, + "step": 16433 + }, + { + "epoch": 2.15, + "learning_rate": 1.7148913422647008e-05, + "loss": 0.0614, + "step": 16434 + }, + { + "epoch": 2.15, + "learning_rate": 1.714854259045014e-05, + "loss": 0.0669, + "step": 16435 + }, + { + "epoch": 2.15, + "learning_rate": 1.714817173814837e-05, + "loss": 0.0589, + "step": 16436 + }, + { + "epoch": 2.15, + "learning_rate": 1.7147800865742752e-05, + "loss": 0.0682, + "step": 16437 + }, + { + "epoch": 2.15, + "learning_rate": 1.7147429973234325e-05, + "loss": 0.0626, + "step": 16438 + }, + { + "epoch": 2.15, + "learning_rate": 1.714705906062413e-05, + "loss": 0.068, + "step": 16439 + }, + { + "epoch": 2.15, + "learning_rate": 1.714668812791321e-05, + "loss": 0.0649, + "step": 16440 + }, + { + "epoch": 2.15, + "learning_rate": 1.7146317175102614e-05, + "loss": 0.0628, + "step": 16441 + }, + { + "epoch": 2.15, + "learning_rate": 1.7145946202193382e-05, + "loss": 0.074, + "step": 16442 + }, + { + "epoch": 2.15, + "learning_rate": 1.7145575209186552e-05, + "loss": 0.0732, + "step": 16443 + }, + { + "epoch": 2.15, + "learning_rate": 1.7145204196083175e-05, + "loss": 0.0783, + "step": 16444 + }, + { + "epoch": 2.15, + "learning_rate": 1.7144833162884293e-05, + "loss": 0.0738, + "step": 16445 + }, + { + "epoch": 2.15, + "learning_rate": 1.7144462109590947e-05, + "loss": 0.0661, + "step": 16446 + }, + { + "epoch": 2.15, + "learning_rate": 1.7144091036204183e-05, + "loss": 0.0701, + "step": 16447 + }, + { + "epoch": 2.15, + "learning_rate": 1.7143719942725043e-05, + "loss": 0.0603, + "step": 16448 + }, + { + "epoch": 2.15, + "learning_rate": 1.714334882915457e-05, + "loss": 0.0531, + "step": 16449 + }, + { + "epoch": 2.15, + "learning_rate": 1.714297769549381e-05, + "loss": 0.0625, + "step": 16450 + }, + { + "epoch": 2.15, + "learning_rate": 1.71426065417438e-05, + "loss": 0.066, + "step": 16451 + }, + { + "epoch": 2.16, + "learning_rate": 1.71422353679056e-05, + "loss": 0.0626, + "step": 16452 + }, + { + "epoch": 2.16, + "learning_rate": 1.7141864173980236e-05, + "loss": 0.0661, + "step": 16453 + }, + { + "epoch": 2.16, + "learning_rate": 1.7141492959968758e-05, + "loss": 0.0611, + "step": 16454 + }, + { + "epoch": 2.16, + "learning_rate": 1.7141121725872218e-05, + "loss": 0.069, + "step": 16455 + }, + { + "epoch": 2.16, + "learning_rate": 1.714075047169165e-05, + "loss": 0.0603, + "step": 16456 + }, + { + "epoch": 2.16, + "learning_rate": 1.7140379197428098e-05, + "loss": 0.0788, + "step": 16457 + }, + { + "epoch": 2.16, + "learning_rate": 1.7140007903082613e-05, + "loss": 0.0681, + "step": 16458 + }, + { + "epoch": 2.16, + "learning_rate": 1.7139636588656236e-05, + "loss": 0.0477, + "step": 16459 + }, + { + "epoch": 2.16, + "learning_rate": 1.713926525415001e-05, + "loss": 0.069, + "step": 16460 + }, + { + "epoch": 2.16, + "learning_rate": 1.7138893899564982e-05, + "loss": 0.058, + "step": 16461 + }, + { + "epoch": 2.16, + "learning_rate": 1.7138522524902196e-05, + "loss": 0.066, + "step": 16462 + }, + { + "epoch": 2.16, + "learning_rate": 1.7138151130162696e-05, + "loss": 0.0696, + "step": 16463 + }, + { + "epoch": 2.16, + "learning_rate": 1.7137779715347523e-05, + "loss": 0.0754, + "step": 16464 + }, + { + "epoch": 2.16, + "learning_rate": 1.7137408280457726e-05, + "loss": 0.0584, + "step": 16465 + }, + { + "epoch": 2.16, + "learning_rate": 1.713703682549435e-05, + "loss": 0.0738, + "step": 16466 + }, + { + "epoch": 2.16, + "learning_rate": 1.7136665350458438e-05, + "loss": 0.068, + "step": 16467 + }, + { + "epoch": 2.16, + "learning_rate": 1.7136293855351032e-05, + "loss": 0.0568, + "step": 16468 + }, + { + "epoch": 2.16, + "learning_rate": 1.713592234017318e-05, + "loss": 0.057, + "step": 16469 + }, + { + "epoch": 2.16, + "learning_rate": 1.713555080492593e-05, + "loss": 0.0639, + "step": 16470 + }, + { + "epoch": 2.16, + "learning_rate": 1.7135179249610317e-05, + "loss": 0.0601, + "step": 16471 + }, + { + "epoch": 2.16, + "learning_rate": 1.7134807674227397e-05, + "loss": 0.0647, + "step": 16472 + }, + { + "epoch": 2.16, + "learning_rate": 1.7134436078778213e-05, + "loss": 0.0553, + "step": 16473 + }, + { + "epoch": 2.16, + "learning_rate": 1.7134064463263804e-05, + "loss": 0.0689, + "step": 16474 + }, + { + "epoch": 2.16, + "learning_rate": 1.7133692827685217e-05, + "loss": 0.0681, + "step": 16475 + }, + { + "epoch": 2.16, + "learning_rate": 1.7133321172043503e-05, + "loss": 0.0705, + "step": 16476 + }, + { + "epoch": 2.16, + "learning_rate": 1.7132949496339698e-05, + "loss": 0.0614, + "step": 16477 + }, + { + "epoch": 2.16, + "learning_rate": 1.7132577800574853e-05, + "loss": 0.0684, + "step": 16478 + }, + { + "epoch": 2.16, + "learning_rate": 1.7132206084750015e-05, + "loss": 0.0637, + "step": 16479 + }, + { + "epoch": 2.16, + "learning_rate": 1.7131834348866227e-05, + "loss": 0.0599, + "step": 16480 + }, + { + "epoch": 2.16, + "learning_rate": 1.7131462592924537e-05, + "loss": 0.0686, + "step": 16481 + }, + { + "epoch": 2.16, + "learning_rate": 1.7131090816925983e-05, + "loss": 0.0664, + "step": 16482 + }, + { + "epoch": 2.16, + "learning_rate": 1.713071902087162e-05, + "loss": 0.0608, + "step": 16483 + }, + { + "epoch": 2.16, + "learning_rate": 1.7130347204762486e-05, + "loss": 0.0632, + "step": 16484 + }, + { + "epoch": 2.16, + "learning_rate": 1.7129975368599633e-05, + "loss": 0.0579, + "step": 16485 + }, + { + "epoch": 2.16, + "learning_rate": 1.7129603512384102e-05, + "loss": 0.0586, + "step": 16486 + }, + { + "epoch": 2.16, + "learning_rate": 1.7129231636116946e-05, + "loss": 0.0636, + "step": 16487 + }, + { + "epoch": 2.16, + "learning_rate": 1.71288597397992e-05, + "loss": 0.0688, + "step": 16488 + }, + { + "epoch": 2.16, + "learning_rate": 1.7128487823431917e-05, + "loss": 0.0688, + "step": 16489 + }, + { + "epoch": 2.16, + "learning_rate": 1.7128115887016142e-05, + "loss": 0.0653, + "step": 16490 + }, + { + "epoch": 2.16, + "learning_rate": 1.712774393055292e-05, + "loss": 0.078, + "step": 16491 + }, + { + "epoch": 2.16, + "learning_rate": 1.71273719540433e-05, + "loss": 0.0652, + "step": 16492 + }, + { + "epoch": 2.16, + "learning_rate": 1.7126999957488325e-05, + "loss": 0.0686, + "step": 16493 + }, + { + "epoch": 2.16, + "learning_rate": 1.712662794088904e-05, + "loss": 0.0656, + "step": 16494 + }, + { + "epoch": 2.16, + "learning_rate": 1.7126255904246495e-05, + "loss": 0.0702, + "step": 16495 + }, + { + "epoch": 2.16, + "learning_rate": 1.7125883847561735e-05, + "loss": 0.0622, + "step": 16496 + }, + { + "epoch": 2.16, + "learning_rate": 1.7125511770835802e-05, + "loss": 0.0713, + "step": 16497 + }, + { + "epoch": 2.16, + "learning_rate": 1.7125139674069754e-05, + "loss": 0.0621, + "step": 16498 + }, + { + "epoch": 2.16, + "learning_rate": 1.7124767557264625e-05, + "loss": 0.0639, + "step": 16499 + }, + { + "epoch": 2.16, + "learning_rate": 1.712439542042147e-05, + "loss": 0.0715, + "step": 16500 + }, + { + "epoch": 2.16, + "learning_rate": 1.712402326354133e-05, + "loss": 0.0714, + "step": 16501 + }, + { + "epoch": 2.16, + "learning_rate": 1.7123651086625256e-05, + "loss": 0.0743, + "step": 16502 + }, + { + "epoch": 2.16, + "learning_rate": 1.712327888967429e-05, + "loss": 0.0654, + "step": 16503 + }, + { + "epoch": 2.16, + "learning_rate": 1.7122906672689482e-05, + "loss": 0.0623, + "step": 16504 + }, + { + "epoch": 2.16, + "learning_rate": 1.7122534435671875e-05, + "loss": 0.0621, + "step": 16505 + }, + { + "epoch": 2.16, + "learning_rate": 1.7122162178622524e-05, + "loss": 0.0627, + "step": 16506 + }, + { + "epoch": 2.16, + "learning_rate": 1.712178990154247e-05, + "loss": 0.0596, + "step": 16507 + }, + { + "epoch": 2.16, + "learning_rate": 1.7121417604432765e-05, + "loss": 0.0699, + "step": 16508 + }, + { + "epoch": 2.16, + "learning_rate": 1.7121045287294447e-05, + "loss": 0.0696, + "step": 16509 + }, + { + "epoch": 2.16, + "learning_rate": 1.712067295012857e-05, + "loss": 0.0657, + "step": 16510 + }, + { + "epoch": 2.16, + "learning_rate": 1.7120300592936178e-05, + "loss": 0.067, + "step": 16511 + }, + { + "epoch": 2.16, + "learning_rate": 1.711992821571832e-05, + "loss": 0.0596, + "step": 16512 + }, + { + "epoch": 2.16, + "learning_rate": 1.7119555818476046e-05, + "loss": 0.0692, + "step": 16513 + }, + { + "epoch": 2.16, + "learning_rate": 1.7119183401210397e-05, + "loss": 0.0752, + "step": 16514 + }, + { + "epoch": 2.16, + "learning_rate": 1.7118810963922427e-05, + "loss": 0.0667, + "step": 16515 + }, + { + "epoch": 2.16, + "learning_rate": 1.711843850661318e-05, + "loss": 0.0538, + "step": 16516 + }, + { + "epoch": 2.16, + "learning_rate": 1.7118066029283703e-05, + "loss": 0.0682, + "step": 16517 + }, + { + "epoch": 2.16, + "learning_rate": 1.7117693531935042e-05, + "loss": 0.0663, + "step": 16518 + }, + { + "epoch": 2.16, + "learning_rate": 1.711732101456825e-05, + "loss": 0.0658, + "step": 16519 + }, + { + "epoch": 2.16, + "learning_rate": 1.711694847718437e-05, + "loss": 0.0714, + "step": 16520 + }, + { + "epoch": 2.16, + "learning_rate": 1.7116575919784454e-05, + "loss": 0.0634, + "step": 16521 + }, + { + "epoch": 2.16, + "learning_rate": 1.711620334236954e-05, + "loss": 0.0544, + "step": 16522 + }, + { + "epoch": 2.16, + "learning_rate": 1.7115830744940688e-05, + "loss": 0.0636, + "step": 16523 + }, + { + "epoch": 2.16, + "learning_rate": 1.7115458127498944e-05, + "loss": 0.0708, + "step": 16524 + }, + { + "epoch": 2.16, + "learning_rate": 1.711508549004535e-05, + "loss": 0.0751, + "step": 16525 + }, + { + "epoch": 2.16, + "learning_rate": 1.7114712832580956e-05, + "loss": 0.0747, + "step": 16526 + }, + { + "epoch": 2.16, + "learning_rate": 1.7114340155106812e-05, + "loss": 0.0798, + "step": 16527 + }, + { + "epoch": 2.17, + "learning_rate": 1.7113967457623965e-05, + "loss": 0.0655, + "step": 16528 + }, + { + "epoch": 2.17, + "learning_rate": 1.7113594740133464e-05, + "loss": 0.0788, + "step": 16529 + }, + { + "epoch": 2.17, + "learning_rate": 1.7113222002636357e-05, + "loss": 0.0695, + "step": 16530 + }, + { + "epoch": 2.17, + "learning_rate": 1.711284924513369e-05, + "loss": 0.0686, + "step": 16531 + }, + { + "epoch": 2.17, + "learning_rate": 1.7112476467626514e-05, + "loss": 0.0619, + "step": 16532 + }, + { + "epoch": 2.17, + "learning_rate": 1.711210367011588e-05, + "loss": 0.0714, + "step": 16533 + }, + { + "epoch": 2.17, + "learning_rate": 1.711173085260283e-05, + "loss": 0.0677, + "step": 16534 + }, + { + "epoch": 2.17, + "learning_rate": 1.7111358015088413e-05, + "loss": 0.0609, + "step": 16535 + }, + { + "epoch": 2.17, + "learning_rate": 1.7110985157573687e-05, + "loss": 0.0577, + "step": 16536 + }, + { + "epoch": 2.17, + "learning_rate": 1.711061228005969e-05, + "loss": 0.0656, + "step": 16537 + }, + { + "epoch": 2.17, + "learning_rate": 1.7110239382547473e-05, + "loss": 0.0807, + "step": 16538 + }, + { + "epoch": 2.17, + "learning_rate": 1.7109866465038087e-05, + "loss": 0.0584, + "step": 16539 + }, + { + "epoch": 2.17, + "learning_rate": 1.710949352753258e-05, + "loss": 0.0665, + "step": 16540 + }, + { + "epoch": 2.17, + "learning_rate": 1.7109120570032004e-05, + "loss": 0.0778, + "step": 16541 + }, + { + "epoch": 2.17, + "learning_rate": 1.7108747592537405e-05, + "loss": 0.0587, + "step": 16542 + }, + { + "epoch": 2.17, + "learning_rate": 1.7108374595049827e-05, + "loss": 0.0808, + "step": 16543 + }, + { + "epoch": 2.17, + "learning_rate": 1.710800157757033e-05, + "loss": 0.0643, + "step": 16544 + }, + { + "epoch": 2.17, + "learning_rate": 1.7107628540099956e-05, + "loss": 0.0784, + "step": 16545 + }, + { + "epoch": 2.17, + "learning_rate": 1.710725548263975e-05, + "loss": 0.0727, + "step": 16546 + }, + { + "epoch": 2.17, + "learning_rate": 1.710688240519077e-05, + "loss": 0.0648, + "step": 16547 + }, + { + "epoch": 2.17, + "learning_rate": 1.7106509307754064e-05, + "loss": 0.0622, + "step": 16548 + }, + { + "epoch": 2.17, + "learning_rate": 1.7106136190330676e-05, + "loss": 0.0647, + "step": 16549 + }, + { + "epoch": 2.17, + "learning_rate": 1.710576305292166e-05, + "loss": 0.073, + "step": 16550 + }, + { + "epoch": 2.17, + "learning_rate": 1.710538989552806e-05, + "loss": 0.0697, + "step": 16551 + }, + { + "epoch": 2.17, + "learning_rate": 1.7105016718150935e-05, + "loss": 0.0631, + "step": 16552 + }, + { + "epoch": 2.17, + "learning_rate": 1.7104643520791326e-05, + "loss": 0.078, + "step": 16553 + }, + { + "epoch": 2.17, + "learning_rate": 1.7104270303450285e-05, + "loss": 0.0742, + "step": 16554 + }, + { + "epoch": 2.17, + "learning_rate": 1.7103897066128864e-05, + "loss": 0.0617, + "step": 16555 + }, + { + "epoch": 2.17, + "learning_rate": 1.7103523808828112e-05, + "loss": 0.0626, + "step": 16556 + }, + { + "epoch": 2.17, + "learning_rate": 1.7103150531549074e-05, + "loss": 0.0592, + "step": 16557 + }, + { + "epoch": 2.17, + "learning_rate": 1.71027772342928e-05, + "loss": 0.0527, + "step": 16558 + }, + { + "epoch": 2.17, + "learning_rate": 1.710240391706035e-05, + "loss": 0.0682, + "step": 16559 + }, + { + "epoch": 2.17, + "learning_rate": 1.7102030579852767e-05, + "loss": 0.0712, + "step": 16560 + }, + { + "epoch": 2.17, + "learning_rate": 1.7101657222671096e-05, + "loss": 0.0627, + "step": 16561 + }, + { + "epoch": 2.17, + "learning_rate": 1.7101283845516398e-05, + "loss": 0.0702, + "step": 16562 + }, + { + "epoch": 2.17, + "learning_rate": 1.7100910448389713e-05, + "loss": 0.0753, + "step": 16563 + }, + { + "epoch": 2.17, + "learning_rate": 1.71005370312921e-05, + "loss": 0.0608, + "step": 16564 + }, + { + "epoch": 2.17, + "learning_rate": 1.71001635942246e-05, + "loss": 0.0672, + "step": 16565 + }, + { + "epoch": 2.17, + "learning_rate": 1.709979013718827e-05, + "loss": 0.0664, + "step": 16566 + }, + { + "epoch": 2.17, + "learning_rate": 1.709941666018416e-05, + "loss": 0.0522, + "step": 16567 + }, + { + "epoch": 2.17, + "learning_rate": 1.7099043163213313e-05, + "loss": 0.0592, + "step": 16568 + }, + { + "epoch": 2.17, + "learning_rate": 1.7098669646276792e-05, + "loss": 0.0726, + "step": 16569 + }, + { + "epoch": 2.17, + "learning_rate": 1.7098296109375638e-05, + "loss": 0.0646, + "step": 16570 + }, + { + "epoch": 2.17, + "learning_rate": 1.7097922552510902e-05, + "loss": 0.0687, + "step": 16571 + }, + { + "epoch": 2.17, + "learning_rate": 1.709754897568364e-05, + "loss": 0.0564, + "step": 16572 + }, + { + "epoch": 2.17, + "learning_rate": 1.7097175378894898e-05, + "loss": 0.0666, + "step": 16573 + }, + { + "epoch": 2.17, + "learning_rate": 1.709680176214573e-05, + "loss": 0.0697, + "step": 16574 + }, + { + "epoch": 2.17, + "learning_rate": 1.7096428125437185e-05, + "loss": 0.0553, + "step": 16575 + }, + { + "epoch": 2.17, + "learning_rate": 1.7096054468770308e-05, + "loss": 0.0658, + "step": 16576 + }, + { + "epoch": 2.17, + "learning_rate": 1.7095680792146163e-05, + "loss": 0.0594, + "step": 16577 + }, + { + "epoch": 2.17, + "learning_rate": 1.709530709556579e-05, + "loss": 0.0682, + "step": 16578 + }, + { + "epoch": 2.17, + "learning_rate": 1.7094933379030244e-05, + "loss": 0.0607, + "step": 16579 + }, + { + "epoch": 2.17, + "learning_rate": 1.7094559642540574e-05, + "loss": 0.0592, + "step": 16580 + }, + { + "epoch": 2.17, + "learning_rate": 1.7094185886097834e-05, + "loss": 0.0693, + "step": 16581 + }, + { + "epoch": 2.17, + "learning_rate": 1.7093812109703077e-05, + "loss": 0.0752, + "step": 16582 + }, + { + "epoch": 2.17, + "learning_rate": 1.709343831335735e-05, + "loss": 0.0688, + "step": 16583 + }, + { + "epoch": 2.17, + "learning_rate": 1.7093064497061704e-05, + "loss": 0.0698, + "step": 16584 + }, + { + "epoch": 2.17, + "learning_rate": 1.709269066081719e-05, + "loss": 0.0608, + "step": 16585 + }, + { + "epoch": 2.17, + "learning_rate": 1.7092316804624863e-05, + "loss": 0.0641, + "step": 16586 + }, + { + "epoch": 2.17, + "learning_rate": 1.7091942928485773e-05, + "loss": 0.0756, + "step": 16587 + }, + { + "epoch": 2.17, + "learning_rate": 1.709156903240097e-05, + "loss": 0.0636, + "step": 16588 + }, + { + "epoch": 2.17, + "learning_rate": 1.709119511637151e-05, + "loss": 0.0605, + "step": 16589 + }, + { + "epoch": 2.17, + "learning_rate": 1.709082118039844e-05, + "loss": 0.0625, + "step": 16590 + }, + { + "epoch": 2.17, + "learning_rate": 1.7090447224482813e-05, + "loss": 0.0651, + "step": 16591 + }, + { + "epoch": 2.17, + "learning_rate": 1.709007324862568e-05, + "loss": 0.066, + "step": 16592 + }, + { + "epoch": 2.17, + "learning_rate": 1.7089699252828094e-05, + "loss": 0.0627, + "step": 16593 + }, + { + "epoch": 2.17, + "learning_rate": 1.708932523709111e-05, + "loss": 0.0569, + "step": 16594 + }, + { + "epoch": 2.17, + "learning_rate": 1.708895120141577e-05, + "loss": 0.0667, + "step": 16595 + }, + { + "epoch": 2.17, + "learning_rate": 1.7088577145803134e-05, + "loss": 0.0707, + "step": 16596 + }, + { + "epoch": 2.17, + "learning_rate": 1.7088203070254256e-05, + "loss": 0.0465, + "step": 16597 + }, + { + "epoch": 2.17, + "learning_rate": 1.708782897477018e-05, + "loss": 0.0689, + "step": 16598 + }, + { + "epoch": 2.17, + "learning_rate": 1.708745485935196e-05, + "loss": 0.0627, + "step": 16599 + }, + { + "epoch": 2.17, + "learning_rate": 1.708708072400066e-05, + "loss": 0.0655, + "step": 16600 + }, + { + "epoch": 2.17, + "learning_rate": 1.7086706568717314e-05, + "loss": 0.0614, + "step": 16601 + }, + { + "epoch": 2.17, + "learning_rate": 1.7086332393502986e-05, + "loss": 0.0678, + "step": 16602 + }, + { + "epoch": 2.17, + "learning_rate": 1.708595819835873e-05, + "loss": 0.0609, + "step": 16603 + }, + { + "epoch": 2.18, + "learning_rate": 1.7085583983285588e-05, + "loss": 0.0663, + "step": 16604 + }, + { + "epoch": 2.18, + "learning_rate": 1.7085209748284618e-05, + "loss": 0.0721, + "step": 16605 + }, + { + "epoch": 2.18, + "learning_rate": 1.7084835493356876e-05, + "loss": 0.0762, + "step": 16606 + }, + { + "epoch": 2.18, + "learning_rate": 1.708446121850341e-05, + "loss": 0.0663, + "step": 16607 + }, + { + "epoch": 2.18, + "learning_rate": 1.7084086923725275e-05, + "loss": 0.0656, + "step": 16608 + }, + { + "epoch": 2.18, + "learning_rate": 1.7083712609023523e-05, + "loss": 0.0681, + "step": 16609 + }, + { + "epoch": 2.18, + "learning_rate": 1.7083338274399207e-05, + "loss": 0.0664, + "step": 16610 + }, + { + "epoch": 2.18, + "learning_rate": 1.708296391985338e-05, + "loss": 0.0731, + "step": 16611 + }, + { + "epoch": 2.18, + "learning_rate": 1.708258954538709e-05, + "loss": 0.0738, + "step": 16612 + }, + { + "epoch": 2.18, + "learning_rate": 1.7082215151001395e-05, + "loss": 0.0622, + "step": 16613 + }, + { + "epoch": 2.18, + "learning_rate": 1.708184073669735e-05, + "loss": 0.0694, + "step": 16614 + }, + { + "epoch": 2.18, + "learning_rate": 1.7081466302476004e-05, + "loss": 0.0632, + "step": 16615 + }, + { + "epoch": 2.18, + "learning_rate": 1.708109184833841e-05, + "loss": 0.0613, + "step": 16616 + }, + { + "epoch": 2.18, + "learning_rate": 1.708071737428562e-05, + "loss": 0.063, + "step": 16617 + }, + { + "epoch": 2.18, + "learning_rate": 1.7080342880318697e-05, + "loss": 0.0695, + "step": 16618 + }, + { + "epoch": 2.18, + "learning_rate": 1.7079968366438682e-05, + "loss": 0.064, + "step": 16619 + }, + { + "epoch": 2.18, + "learning_rate": 1.7079593832646634e-05, + "loss": 0.0603, + "step": 16620 + }, + { + "epoch": 2.18, + "learning_rate": 1.7079219278943605e-05, + "loss": 0.0657, + "step": 16621 + }, + { + "epoch": 2.18, + "learning_rate": 1.7078844705330647e-05, + "loss": 0.0653, + "step": 16622 + }, + { + "epoch": 2.18, + "learning_rate": 1.7078470111808817e-05, + "loss": 0.0616, + "step": 16623 + }, + { + "epoch": 2.18, + "learning_rate": 1.7078095498379168e-05, + "loss": 0.063, + "step": 16624 + }, + { + "epoch": 2.18, + "learning_rate": 1.7077720865042747e-05, + "loss": 0.0721, + "step": 16625 + }, + { + "epoch": 2.18, + "learning_rate": 1.707734621180062e-05, + "loss": 0.0691, + "step": 16626 + }, + { + "epoch": 2.18, + "learning_rate": 1.7076971538653828e-05, + "loss": 0.0621, + "step": 16627 + }, + { + "epoch": 2.18, + "learning_rate": 1.707659684560343e-05, + "loss": 0.061, + "step": 16628 + }, + { + "epoch": 2.18, + "learning_rate": 1.7076222132650487e-05, + "loss": 0.0754, + "step": 16629 + }, + { + "epoch": 2.18, + "learning_rate": 1.7075847399796043e-05, + "loss": 0.0656, + "step": 16630 + }, + { + "epoch": 2.18, + "learning_rate": 1.7075472647041153e-05, + "loss": 0.0535, + "step": 16631 + }, + { + "epoch": 2.18, + "learning_rate": 1.7075097874386874e-05, + "loss": 0.0675, + "step": 16632 + }, + { + "epoch": 2.18, + "learning_rate": 1.707472308183426e-05, + "loss": 0.063, + "step": 16633 + }, + { + "epoch": 2.18, + "learning_rate": 1.7074348269384363e-05, + "loss": 0.0645, + "step": 16634 + }, + { + "epoch": 2.18, + "learning_rate": 1.707397343703824e-05, + "loss": 0.0672, + "step": 16635 + }, + { + "epoch": 2.18, + "learning_rate": 1.707359858479694e-05, + "loss": 0.0506, + "step": 16636 + }, + { + "epoch": 2.18, + "learning_rate": 1.707322371266152e-05, + "loss": 0.0559, + "step": 16637 + }, + { + "epoch": 2.18, + "learning_rate": 1.707284882063304e-05, + "loss": 0.0667, + "step": 16638 + }, + { + "epoch": 2.18, + "learning_rate": 1.7072473908712547e-05, + "loss": 0.0608, + "step": 16639 + }, + { + "epoch": 2.18, + "learning_rate": 1.7072098976901096e-05, + "loss": 0.0648, + "step": 16640 + }, + { + "epoch": 2.18, + "learning_rate": 1.707172402519975e-05, + "loss": 0.0688, + "step": 16641 + }, + { + "epoch": 2.18, + "learning_rate": 1.7071349053609546e-05, + "loss": 0.0741, + "step": 16642 + }, + { + "epoch": 2.18, + "learning_rate": 1.707097406213156e-05, + "loss": 0.064, + "step": 16643 + }, + { + "epoch": 2.18, + "learning_rate": 1.7070599050766828e-05, + "loss": 0.0552, + "step": 16644 + }, + { + "epoch": 2.18, + "learning_rate": 1.7070224019516415e-05, + "loss": 0.0639, + "step": 16645 + }, + { + "epoch": 2.18, + "learning_rate": 1.7069848968381376e-05, + "loss": 0.0602, + "step": 16646 + }, + { + "epoch": 2.18, + "learning_rate": 1.706947389736276e-05, + "loss": 0.0591, + "step": 16647 + }, + { + "epoch": 2.18, + "learning_rate": 1.7069098806461624e-05, + "loss": 0.0751, + "step": 16648 + }, + { + "epoch": 2.18, + "learning_rate": 1.706872369567903e-05, + "loss": 0.0565, + "step": 16649 + }, + { + "epoch": 2.18, + "learning_rate": 1.7068348565016023e-05, + "loss": 0.0698, + "step": 16650 + }, + { + "epoch": 2.18, + "learning_rate": 1.706797341447366e-05, + "loss": 0.058, + "step": 16651 + }, + { + "epoch": 2.18, + "learning_rate": 1.7067598244053e-05, + "loss": 0.0652, + "step": 16652 + }, + { + "epoch": 2.18, + "learning_rate": 1.70672230537551e-05, + "loss": 0.0677, + "step": 16653 + }, + { + "epoch": 2.18, + "learning_rate": 1.706684784358101e-05, + "loss": 0.0657, + "step": 16654 + }, + { + "epoch": 2.18, + "learning_rate": 1.7066472613531785e-05, + "loss": 0.0838, + "step": 16655 + }, + { + "epoch": 2.18, + "learning_rate": 1.7066097363608484e-05, + "loss": 0.0659, + "step": 16656 + }, + { + "epoch": 2.18, + "learning_rate": 1.7065722093812157e-05, + "loss": 0.077, + "step": 16657 + }, + { + "epoch": 2.18, + "learning_rate": 1.7065346804143864e-05, + "loss": 0.0598, + "step": 16658 + }, + { + "epoch": 2.18, + "learning_rate": 1.706497149460466e-05, + "loss": 0.061, + "step": 16659 + }, + { + "epoch": 2.18, + "learning_rate": 1.70645961651956e-05, + "loss": 0.0675, + "step": 16660 + }, + { + "epoch": 2.18, + "learning_rate": 1.706422081591774e-05, + "loss": 0.0631, + "step": 16661 + }, + { + "epoch": 2.18, + "learning_rate": 1.7063845446772133e-05, + "loss": 0.0632, + "step": 16662 + }, + { + "epoch": 2.18, + "learning_rate": 1.706347005775984e-05, + "loss": 0.0665, + "step": 16663 + }, + { + "epoch": 2.18, + "learning_rate": 1.706309464888191e-05, + "loss": 0.0664, + "step": 16664 + }, + { + "epoch": 2.18, + "learning_rate": 1.7062719220139408e-05, + "loss": 0.0708, + "step": 16665 + }, + { + "epoch": 2.18, + "learning_rate": 1.706234377153338e-05, + "loss": 0.0652, + "step": 16666 + }, + { + "epoch": 2.18, + "learning_rate": 1.7061968303064888e-05, + "loss": 0.0658, + "step": 16667 + }, + { + "epoch": 2.18, + "learning_rate": 1.7061592814734986e-05, + "loss": 0.0731, + "step": 16668 + }, + { + "epoch": 2.18, + "learning_rate": 1.706121730654473e-05, + "loss": 0.0692, + "step": 16669 + }, + { + "epoch": 2.18, + "learning_rate": 1.7060841778495175e-05, + "loss": 0.0838, + "step": 16670 + }, + { + "epoch": 2.18, + "learning_rate": 1.7060466230587377e-05, + "loss": 0.0663, + "step": 16671 + }, + { + "epoch": 2.18, + "learning_rate": 1.7060090662822396e-05, + "loss": 0.0678, + "step": 16672 + }, + { + "epoch": 2.18, + "learning_rate": 1.705971507520129e-05, + "loss": 0.0668, + "step": 16673 + }, + { + "epoch": 2.18, + "learning_rate": 1.7059339467725106e-05, + "loss": 0.0623, + "step": 16674 + }, + { + "epoch": 2.18, + "learning_rate": 1.7058963840394905e-05, + "loss": 0.0631, + "step": 16675 + }, + { + "epoch": 2.18, + "learning_rate": 1.7058588193211746e-05, + "loss": 0.0677, + "step": 16676 + }, + { + "epoch": 2.18, + "learning_rate": 1.7058212526176685e-05, + "loss": 0.0628, + "step": 16677 + }, + { + "epoch": 2.18, + "learning_rate": 1.7057836839290778e-05, + "loss": 0.0664, + "step": 16678 + }, + { + "epoch": 2.18, + "learning_rate": 1.7057461132555077e-05, + "loss": 0.0594, + "step": 16679 + }, + { + "epoch": 2.18, + "learning_rate": 1.7057085405970646e-05, + "loss": 0.0653, + "step": 16680 + }, + { + "epoch": 2.19, + "learning_rate": 1.7056709659538535e-05, + "loss": 0.0673, + "step": 16681 + }, + { + "epoch": 2.19, + "learning_rate": 1.7056333893259806e-05, + "loss": 0.068, + "step": 16682 + }, + { + "epoch": 2.19, + "learning_rate": 1.705595810713551e-05, + "loss": 0.0564, + "step": 16683 + }, + { + "epoch": 2.19, + "learning_rate": 1.7055582301166712e-05, + "loss": 0.0642, + "step": 16684 + }, + { + "epoch": 2.19, + "learning_rate": 1.7055206475354465e-05, + "loss": 0.0613, + "step": 16685 + }, + { + "epoch": 2.19, + "learning_rate": 1.7054830629699824e-05, + "loss": 0.063, + "step": 16686 + }, + { + "epoch": 2.19, + "learning_rate": 1.7054454764203847e-05, + "loss": 0.0602, + "step": 16687 + }, + { + "epoch": 2.19, + "learning_rate": 1.705407887886759e-05, + "loss": 0.0698, + "step": 16688 + }, + { + "epoch": 2.19, + "learning_rate": 1.7053702973692116e-05, + "loss": 0.0543, + "step": 16689 + }, + { + "epoch": 2.19, + "learning_rate": 1.705332704867847e-05, + "loss": 0.062, + "step": 16690 + }, + { + "epoch": 2.19, + "learning_rate": 1.7052951103827727e-05, + "loss": 0.0631, + "step": 16691 + }, + { + "epoch": 2.19, + "learning_rate": 1.7052575139140928e-05, + "loss": 0.0675, + "step": 16692 + }, + { + "epoch": 2.19, + "learning_rate": 1.705219915461914e-05, + "loss": 0.074, + "step": 16693 + }, + { + "epoch": 2.19, + "learning_rate": 1.705182315026342e-05, + "loss": 0.0572, + "step": 16694 + }, + { + "epoch": 2.19, + "learning_rate": 1.705144712607482e-05, + "loss": 0.0707, + "step": 16695 + }, + { + "epoch": 2.19, + "learning_rate": 1.7051071082054396e-05, + "loss": 0.0584, + "step": 16696 + }, + { + "epoch": 2.19, + "learning_rate": 1.7050695018203215e-05, + "loss": 0.0723, + "step": 16697 + }, + { + "epoch": 2.19, + "learning_rate": 1.705031893452233e-05, + "loss": 0.0687, + "step": 16698 + }, + { + "epoch": 2.19, + "learning_rate": 1.70499428310128e-05, + "loss": 0.0646, + "step": 16699 + }, + { + "epoch": 2.19, + "learning_rate": 1.7049566707675675e-05, + "loss": 0.0679, + "step": 16700 + }, + { + "epoch": 2.19, + "learning_rate": 1.7049190564512025e-05, + "loss": 0.0599, + "step": 16701 + }, + { + "epoch": 2.19, + "learning_rate": 1.7048814401522897e-05, + "loss": 0.0685, + "step": 16702 + }, + { + "epoch": 2.19, + "learning_rate": 1.7048438218709357e-05, + "loss": 0.0728, + "step": 16703 + }, + { + "epoch": 2.19, + "learning_rate": 1.7048062016072463e-05, + "loss": 0.0683, + "step": 16704 + }, + { + "epoch": 2.19, + "learning_rate": 1.7047685793613264e-05, + "loss": 0.0655, + "step": 16705 + }, + { + "epoch": 2.19, + "learning_rate": 1.7047309551332824e-05, + "loss": 0.0595, + "step": 16706 + }, + { + "epoch": 2.19, + "learning_rate": 1.7046933289232205e-05, + "loss": 0.0684, + "step": 16707 + }, + { + "epoch": 2.19, + "learning_rate": 1.7046557007312462e-05, + "loss": 0.0708, + "step": 16708 + }, + { + "epoch": 2.19, + "learning_rate": 1.704618070557465e-05, + "loss": 0.0624, + "step": 16709 + }, + { + "epoch": 2.19, + "learning_rate": 1.704580438401983e-05, + "loss": 0.0694, + "step": 16710 + }, + { + "epoch": 2.19, + "learning_rate": 1.704542804264906e-05, + "loss": 0.069, + "step": 16711 + }, + { + "epoch": 2.19, + "learning_rate": 1.70450516814634e-05, + "loss": 0.0666, + "step": 16712 + }, + { + "epoch": 2.19, + "learning_rate": 1.7044675300463906e-05, + "loss": 0.0555, + "step": 16713 + }, + { + "epoch": 2.19, + "learning_rate": 1.704429889965164e-05, + "loss": 0.0647, + "step": 16714 + }, + { + "epoch": 2.19, + "learning_rate": 1.704392247902766e-05, + "loss": 0.0614, + "step": 16715 + }, + { + "epoch": 2.19, + "learning_rate": 1.7043546038593016e-05, + "loss": 0.061, + "step": 16716 + }, + { + "epoch": 2.19, + "learning_rate": 1.704316957834878e-05, + "loss": 0.0707, + "step": 16717 + }, + { + "epoch": 2.19, + "learning_rate": 1.7042793098296005e-05, + "loss": 0.0673, + "step": 16718 + }, + { + "epoch": 2.19, + "learning_rate": 1.704241659843575e-05, + "loss": 0.0532, + "step": 16719 + }, + { + "epoch": 2.19, + "learning_rate": 1.7042040078769066e-05, + "loss": 0.0643, + "step": 16720 + }, + { + "epoch": 2.19, + "learning_rate": 1.7041663539297023e-05, + "loss": 0.0551, + "step": 16721 + }, + { + "epoch": 2.19, + "learning_rate": 1.7041286980020678e-05, + "loss": 0.0591, + "step": 16722 + }, + { + "epoch": 2.19, + "learning_rate": 1.704091040094109e-05, + "loss": 0.0623, + "step": 16723 + }, + { + "epoch": 2.19, + "learning_rate": 1.7040533802059314e-05, + "loss": 0.0627, + "step": 16724 + }, + { + "epoch": 2.19, + "learning_rate": 1.704015718337641e-05, + "loss": 0.0698, + "step": 16725 + }, + { + "epoch": 2.19, + "learning_rate": 1.7039780544893443e-05, + "loss": 0.0598, + "step": 16726 + }, + { + "epoch": 2.19, + "learning_rate": 1.7039403886611466e-05, + "loss": 0.0692, + "step": 16727 + }, + { + "epoch": 2.19, + "learning_rate": 1.703902720853154e-05, + "loss": 0.0614, + "step": 16728 + }, + { + "epoch": 2.19, + "learning_rate": 1.7038650510654724e-05, + "loss": 0.0741, + "step": 16729 + }, + { + "epoch": 2.19, + "learning_rate": 1.703827379298208e-05, + "loss": 0.0658, + "step": 16730 + }, + { + "epoch": 2.19, + "learning_rate": 1.7037897055514665e-05, + "loss": 0.0617, + "step": 16731 + }, + { + "epoch": 2.19, + "learning_rate": 1.703752029825354e-05, + "loss": 0.0673, + "step": 16732 + }, + { + "epoch": 2.19, + "learning_rate": 1.7037143521199763e-05, + "loss": 0.0686, + "step": 16733 + }, + { + "epoch": 2.19, + "learning_rate": 1.7036766724354394e-05, + "loss": 0.0686, + "step": 16734 + }, + { + "epoch": 2.19, + "learning_rate": 1.7036389907718492e-05, + "loss": 0.0656, + "step": 16735 + }, + { + "epoch": 2.19, + "learning_rate": 1.7036013071293123e-05, + "loss": 0.0553, + "step": 16736 + }, + { + "epoch": 2.19, + "learning_rate": 1.703563621507934e-05, + "loss": 0.0697, + "step": 16737 + }, + { + "epoch": 2.19, + "learning_rate": 1.7035259339078202e-05, + "loss": 0.0649, + "step": 16738 + }, + { + "epoch": 2.19, + "learning_rate": 1.7034882443290776e-05, + "loss": 0.0678, + "step": 16739 + }, + { + "epoch": 2.19, + "learning_rate": 1.7034505527718112e-05, + "loss": 0.0557, + "step": 16740 + }, + { + "epoch": 2.19, + "learning_rate": 1.7034128592361284e-05, + "loss": 0.0665, + "step": 16741 + }, + { + "epoch": 2.19, + "learning_rate": 1.7033751637221337e-05, + "loss": 0.0556, + "step": 16742 + }, + { + "epoch": 2.19, + "learning_rate": 1.703337466229934e-05, + "loss": 0.0576, + "step": 16743 + }, + { + "epoch": 2.19, + "learning_rate": 1.7032997667596352e-05, + "loss": 0.0643, + "step": 16744 + }, + { + "epoch": 2.19, + "learning_rate": 1.7032620653113433e-05, + "loss": 0.0656, + "step": 16745 + }, + { + "epoch": 2.19, + "learning_rate": 1.7032243618851643e-05, + "loss": 0.07, + "step": 16746 + }, + { + "epoch": 2.19, + "learning_rate": 1.703186656481204e-05, + "loss": 0.0578, + "step": 16747 + }, + { + "epoch": 2.19, + "learning_rate": 1.703148949099569e-05, + "loss": 0.0631, + "step": 16748 + }, + { + "epoch": 2.19, + "learning_rate": 1.7031112397403647e-05, + "loss": 0.0638, + "step": 16749 + }, + { + "epoch": 2.19, + "learning_rate": 1.7030735284036976e-05, + "loss": 0.0611, + "step": 16750 + }, + { + "epoch": 2.19, + "learning_rate": 1.7030358150896738e-05, + "loss": 0.0588, + "step": 16751 + }, + { + "epoch": 2.19, + "learning_rate": 1.702998099798399e-05, + "loss": 0.0597, + "step": 16752 + }, + { + "epoch": 2.19, + "learning_rate": 1.7029603825299795e-05, + "loss": 0.0743, + "step": 16753 + }, + { + "epoch": 2.19, + "learning_rate": 1.7029226632845214e-05, + "loss": 0.0696, + "step": 16754 + }, + { + "epoch": 2.19, + "learning_rate": 1.7028849420621308e-05, + "loss": 0.0563, + "step": 16755 + }, + { + "epoch": 2.19, + "learning_rate": 1.702847218862914e-05, + "loss": 0.0609, + "step": 16756 + }, + { + "epoch": 2.2, + "learning_rate": 1.702809493686976e-05, + "loss": 0.0698, + "step": 16757 + }, + { + "epoch": 2.2, + "learning_rate": 1.7027717665344244e-05, + "loss": 0.0614, + "step": 16758 + }, + { + "epoch": 2.2, + "learning_rate": 1.7027340374053647e-05, + "loss": 0.0712, + "step": 16759 + }, + { + "epoch": 2.2, + "learning_rate": 1.7026963062999028e-05, + "loss": 0.0836, + "step": 16760 + }, + { + "epoch": 2.2, + "learning_rate": 1.7026585732181446e-05, + "loss": 0.056, + "step": 16761 + }, + { + "epoch": 2.2, + "learning_rate": 1.7026208381601968e-05, + "loss": 0.0761, + "step": 16762 + }, + { + "epoch": 2.2, + "learning_rate": 1.7025831011261656e-05, + "loss": 0.0728, + "step": 16763 + }, + { + "epoch": 2.2, + "learning_rate": 1.7025453621161566e-05, + "loss": 0.0648, + "step": 16764 + }, + { + "epoch": 2.2, + "learning_rate": 1.7025076211302762e-05, + "loss": 0.0646, + "step": 16765 + }, + { + "epoch": 2.2, + "learning_rate": 1.7024698781686308e-05, + "loss": 0.0725, + "step": 16766 + }, + { + "epoch": 2.2, + "learning_rate": 1.7024321332313254e-05, + "loss": 0.0687, + "step": 16767 + }, + { + "epoch": 2.2, + "learning_rate": 1.702394386318468e-05, + "loss": 0.0655, + "step": 16768 + }, + { + "epoch": 2.2, + "learning_rate": 1.7023566374301635e-05, + "loss": 0.0608, + "step": 16769 + }, + { + "epoch": 2.2, + "learning_rate": 1.7023188865665182e-05, + "loss": 0.0671, + "step": 16770 + }, + { + "epoch": 2.2, + "learning_rate": 1.7022811337276386e-05, + "loss": 0.0675, + "step": 16771 + }, + { + "epoch": 2.2, + "learning_rate": 1.7022433789136306e-05, + "loss": 0.0709, + "step": 16772 + }, + { + "epoch": 2.2, + "learning_rate": 1.7022056221246006e-05, + "loss": 0.0705, + "step": 16773 + }, + { + "epoch": 2.2, + "learning_rate": 1.7021678633606543e-05, + "loss": 0.0776, + "step": 16774 + }, + { + "epoch": 2.2, + "learning_rate": 1.702130102621899e-05, + "loss": 0.0709, + "step": 16775 + }, + { + "epoch": 2.2, + "learning_rate": 1.7020923399084396e-05, + "loss": 0.0643, + "step": 16776 + }, + { + "epoch": 2.2, + "learning_rate": 1.7020545752203828e-05, + "loss": 0.0781, + "step": 16777 + }, + { + "epoch": 2.2, + "learning_rate": 1.7020168085578354e-05, + "loss": 0.0616, + "step": 16778 + }, + { + "epoch": 2.2, + "learning_rate": 1.701979039920903e-05, + "loss": 0.0643, + "step": 16779 + }, + { + "epoch": 2.2, + "learning_rate": 1.7019412693096918e-05, + "loss": 0.0666, + "step": 16780 + }, + { + "epoch": 2.2, + "learning_rate": 1.701903496724308e-05, + "loss": 0.0722, + "step": 16781 + }, + { + "epoch": 2.2, + "learning_rate": 1.701865722164858e-05, + "loss": 0.0595, + "step": 16782 + }, + { + "epoch": 2.2, + "learning_rate": 1.7018279456314482e-05, + "loss": 0.0546, + "step": 16783 + }, + { + "epoch": 2.2, + "learning_rate": 1.7017901671241848e-05, + "loss": 0.0652, + "step": 16784 + }, + { + "epoch": 2.2, + "learning_rate": 1.7017523866431737e-05, + "loss": 0.0695, + "step": 16785 + }, + { + "epoch": 2.2, + "learning_rate": 1.7017146041885216e-05, + "loss": 0.0713, + "step": 16786 + }, + { + "epoch": 2.2, + "learning_rate": 1.7016768197603342e-05, + "loss": 0.07, + "step": 16787 + }, + { + "epoch": 2.2, + "learning_rate": 1.7016390333587183e-05, + "loss": 0.062, + "step": 16788 + }, + { + "epoch": 2.2, + "learning_rate": 1.70160124498378e-05, + "loss": 0.0601, + "step": 16789 + }, + { + "epoch": 2.2, + "learning_rate": 1.7015634546356254e-05, + "loss": 0.0604, + "step": 16790 + }, + { + "epoch": 2.2, + "learning_rate": 1.7015256623143616e-05, + "loss": 0.0668, + "step": 16791 + }, + { + "epoch": 2.2, + "learning_rate": 1.7014878680200933e-05, + "loss": 0.0751, + "step": 16792 + }, + { + "epoch": 2.2, + "learning_rate": 1.701450071752928e-05, + "loss": 0.0693, + "step": 16793 + }, + { + "epoch": 2.2, + "learning_rate": 1.701412273512972e-05, + "loss": 0.0551, + "step": 16794 + }, + { + "epoch": 2.2, + "learning_rate": 1.7013744733003315e-05, + "loss": 0.0587, + "step": 16795 + }, + { + "epoch": 2.2, + "learning_rate": 1.701336671115112e-05, + "loss": 0.0759, + "step": 16796 + }, + { + "epoch": 2.2, + "learning_rate": 1.701298866957421e-05, + "loss": 0.068, + "step": 16797 + }, + { + "epoch": 2.2, + "learning_rate": 1.7012610608273642e-05, + "loss": 0.0698, + "step": 16798 + }, + { + "epoch": 2.2, + "learning_rate": 1.7012232527250478e-05, + "loss": 0.0582, + "step": 16799 + }, + { + "epoch": 2.2, + "learning_rate": 1.7011854426505785e-05, + "loss": 0.0586, + "step": 16800 + }, + { + "epoch": 2.2, + "learning_rate": 1.7011476306040626e-05, + "loss": 0.0663, + "step": 16801 + }, + { + "epoch": 2.2, + "learning_rate": 1.7011098165856063e-05, + "loss": 0.0602, + "step": 16802 + }, + { + "epoch": 2.2, + "learning_rate": 1.7010720005953155e-05, + "loss": 0.0612, + "step": 16803 + }, + { + "epoch": 2.2, + "learning_rate": 1.7010341826332974e-05, + "loss": 0.0624, + "step": 16804 + }, + { + "epoch": 2.2, + "learning_rate": 1.700996362699658e-05, + "loss": 0.0631, + "step": 16805 + }, + { + "epoch": 2.2, + "learning_rate": 1.7009585407945036e-05, + "loss": 0.0712, + "step": 16806 + }, + { + "epoch": 2.2, + "learning_rate": 1.7009207169179407e-05, + "loss": 0.0658, + "step": 16807 + }, + { + "epoch": 2.2, + "learning_rate": 1.7008828910700757e-05, + "loss": 0.0624, + "step": 16808 + }, + { + "epoch": 2.2, + "learning_rate": 1.700845063251015e-05, + "loss": 0.0715, + "step": 16809 + }, + { + "epoch": 2.2, + "learning_rate": 1.7008072334608644e-05, + "loss": 0.0612, + "step": 16810 + }, + { + "epoch": 2.2, + "learning_rate": 1.7007694016997312e-05, + "loss": 0.0653, + "step": 16811 + }, + { + "epoch": 2.2, + "learning_rate": 1.700731567967721e-05, + "loss": 0.0621, + "step": 16812 + }, + { + "epoch": 2.2, + "learning_rate": 1.700693732264941e-05, + "loss": 0.0637, + "step": 16813 + }, + { + "epoch": 2.2, + "learning_rate": 1.700655894591497e-05, + "loss": 0.0559, + "step": 16814 + }, + { + "epoch": 2.2, + "learning_rate": 1.7006180549474954e-05, + "loss": 0.0629, + "step": 16815 + }, + { + "epoch": 2.2, + "learning_rate": 1.7005802133330427e-05, + "loss": 0.0632, + "step": 16816 + }, + { + "epoch": 2.2, + "learning_rate": 1.700542369748246e-05, + "loss": 0.0597, + "step": 16817 + }, + { + "epoch": 2.2, + "learning_rate": 1.7005045241932112e-05, + "loss": 0.0732, + "step": 16818 + }, + { + "epoch": 2.2, + "learning_rate": 1.7004666766680444e-05, + "loss": 0.0648, + "step": 16819 + }, + { + "epoch": 2.2, + "learning_rate": 1.7004288271728525e-05, + "loss": 0.0779, + "step": 16820 + }, + { + "epoch": 2.2, + "learning_rate": 1.7003909757077418e-05, + "loss": 0.0543, + "step": 16821 + }, + { + "epoch": 2.2, + "learning_rate": 1.7003531222728186e-05, + "loss": 0.0616, + "step": 16822 + }, + { + "epoch": 2.2, + "learning_rate": 1.70031526686819e-05, + "loss": 0.0727, + "step": 16823 + }, + { + "epoch": 2.2, + "learning_rate": 1.7002774094939615e-05, + "loss": 0.0737, + "step": 16824 + }, + { + "epoch": 2.2, + "learning_rate": 1.7002395501502405e-05, + "loss": 0.0628, + "step": 16825 + }, + { + "epoch": 2.2, + "learning_rate": 1.7002016888371328e-05, + "loss": 0.0572, + "step": 16826 + }, + { + "epoch": 2.2, + "learning_rate": 1.700163825554745e-05, + "loss": 0.0678, + "step": 16827 + }, + { + "epoch": 2.2, + "learning_rate": 1.700125960303184e-05, + "loss": 0.0565, + "step": 16828 + }, + { + "epoch": 2.2, + "learning_rate": 1.700088093082556e-05, + "loss": 0.0602, + "step": 16829 + }, + { + "epoch": 2.2, + "learning_rate": 1.7000502238929675e-05, + "loss": 0.0622, + "step": 16830 + }, + { + "epoch": 2.2, + "learning_rate": 1.700012352734525e-05, + "loss": 0.0666, + "step": 16831 + }, + { + "epoch": 2.2, + "learning_rate": 1.699974479607335e-05, + "loss": 0.0568, + "step": 16832 + }, + { + "epoch": 2.21, + "learning_rate": 1.699936604511504e-05, + "loss": 0.0623, + "step": 16833 + }, + { + "epoch": 2.21, + "learning_rate": 1.699898727447139e-05, + "loss": 0.0637, + "step": 16834 + }, + { + "epoch": 2.21, + "learning_rate": 1.6998608484143455e-05, + "loss": 0.0669, + "step": 16835 + }, + { + "epoch": 2.21, + "learning_rate": 1.6998229674132308e-05, + "loss": 0.0608, + "step": 16836 + }, + { + "epoch": 2.21, + "learning_rate": 1.6997850844439012e-05, + "loss": 0.0651, + "step": 16837 + }, + { + "epoch": 2.21, + "learning_rate": 1.6997471995064635e-05, + "loss": 0.0623, + "step": 16838 + }, + { + "epoch": 2.21, + "learning_rate": 1.6997093126010243e-05, + "loss": 0.0709, + "step": 16839 + }, + { + "epoch": 2.21, + "learning_rate": 1.6996714237276894e-05, + "loss": 0.0629, + "step": 16840 + }, + { + "epoch": 2.21, + "learning_rate": 1.6996335328865662e-05, + "loss": 0.0624, + "step": 16841 + }, + { + "epoch": 2.21, + "learning_rate": 1.6995956400777607e-05, + "loss": 0.062, + "step": 16842 + }, + { + "epoch": 2.21, + "learning_rate": 1.69955774530138e-05, + "loss": 0.0644, + "step": 16843 + }, + { + "epoch": 2.21, + "learning_rate": 1.6995198485575305e-05, + "loss": 0.0701, + "step": 16844 + }, + { + "epoch": 2.21, + "learning_rate": 1.6994819498463183e-05, + "loss": 0.0689, + "step": 16845 + }, + { + "epoch": 2.21, + "learning_rate": 1.6994440491678506e-05, + "loss": 0.0679, + "step": 16846 + }, + { + "epoch": 2.21, + "learning_rate": 1.6994061465222337e-05, + "loss": 0.0648, + "step": 16847 + }, + { + "epoch": 2.21, + "learning_rate": 1.699368241909574e-05, + "loss": 0.0776, + "step": 16848 + }, + { + "epoch": 2.21, + "learning_rate": 1.6993303353299787e-05, + "loss": 0.0754, + "step": 16849 + }, + { + "epoch": 2.21, + "learning_rate": 1.699292426783554e-05, + "loss": 0.0671, + "step": 16850 + }, + { + "epoch": 2.21, + "learning_rate": 1.699254516270407e-05, + "loss": 0.0772, + "step": 16851 + }, + { + "epoch": 2.21, + "learning_rate": 1.6992166037906432e-05, + "loss": 0.0652, + "step": 16852 + }, + { + "epoch": 2.21, + "learning_rate": 1.6991786893443702e-05, + "loss": 0.0647, + "step": 16853 + }, + { + "epoch": 2.21, + "learning_rate": 1.6991407729316944e-05, + "loss": 0.0721, + "step": 16854 + }, + { + "epoch": 2.21, + "learning_rate": 1.6991028545527223e-05, + "loss": 0.0598, + "step": 16855 + }, + { + "epoch": 2.21, + "learning_rate": 1.6990649342075608e-05, + "loss": 0.0735, + "step": 16856 + }, + { + "epoch": 2.21, + "learning_rate": 1.6990270118963165e-05, + "loss": 0.0599, + "step": 16857 + }, + { + "epoch": 2.21, + "learning_rate": 1.6989890876190956e-05, + "loss": 0.0625, + "step": 16858 + }, + { + "epoch": 2.21, + "learning_rate": 1.6989511613760054e-05, + "loss": 0.0715, + "step": 16859 + }, + { + "epoch": 2.21, + "learning_rate": 1.6989132331671523e-05, + "loss": 0.0708, + "step": 16860 + }, + { + "epoch": 2.21, + "learning_rate": 1.6988753029926427e-05, + "loss": 0.0659, + "step": 16861 + }, + { + "epoch": 2.21, + "learning_rate": 1.698837370852584e-05, + "loss": 0.0752, + "step": 16862 + }, + { + "epoch": 2.21, + "learning_rate": 1.698799436747082e-05, + "loss": 0.0587, + "step": 16863 + }, + { + "epoch": 2.21, + "learning_rate": 1.6987615006762437e-05, + "loss": 0.0776, + "step": 16864 + }, + { + "epoch": 2.21, + "learning_rate": 1.698723562640176e-05, + "loss": 0.0612, + "step": 16865 + }, + { + "epoch": 2.21, + "learning_rate": 1.6986856226389855e-05, + "loss": 0.0674, + "step": 16866 + }, + { + "epoch": 2.21, + "learning_rate": 1.6986476806727794e-05, + "loss": 0.0692, + "step": 16867 + }, + { + "epoch": 2.21, + "learning_rate": 1.6986097367416633e-05, + "loss": 0.0543, + "step": 16868 + }, + { + "epoch": 2.21, + "learning_rate": 1.6985717908457448e-05, + "loss": 0.0641, + "step": 16869 + }, + { + "epoch": 2.21, + "learning_rate": 1.69853384298513e-05, + "loss": 0.0602, + "step": 16870 + }, + { + "epoch": 2.21, + "learning_rate": 1.6984958931599262e-05, + "loss": 0.0644, + "step": 16871 + }, + { + "epoch": 2.21, + "learning_rate": 1.69845794137024e-05, + "loss": 0.0602, + "step": 16872 + }, + { + "epoch": 2.21, + "learning_rate": 1.698419987616178e-05, + "loss": 0.0524, + "step": 16873 + }, + { + "epoch": 2.21, + "learning_rate": 1.698382031897847e-05, + "loss": 0.0625, + "step": 16874 + }, + { + "epoch": 2.21, + "learning_rate": 1.6983440742153535e-05, + "loss": 0.0614, + "step": 16875 + }, + { + "epoch": 2.21, + "learning_rate": 1.6983061145688044e-05, + "loss": 0.0611, + "step": 16876 + }, + { + "epoch": 2.21, + "learning_rate": 1.6982681529583068e-05, + "loss": 0.0632, + "step": 16877 + }, + { + "epoch": 2.21, + "learning_rate": 1.6982301893839674e-05, + "loss": 0.0703, + "step": 16878 + }, + { + "epoch": 2.21, + "learning_rate": 1.6981922238458922e-05, + "loss": 0.0631, + "step": 16879 + }, + { + "epoch": 2.21, + "learning_rate": 1.698154256344189e-05, + "loss": 0.0683, + "step": 16880 + }, + { + "epoch": 2.21, + "learning_rate": 1.698116286878964e-05, + "loss": 0.0626, + "step": 16881 + }, + { + "epoch": 2.21, + "learning_rate": 1.6980783154503243e-05, + "loss": 0.0576, + "step": 16882 + }, + { + "epoch": 2.21, + "learning_rate": 1.6980403420583763e-05, + "loss": 0.0628, + "step": 16883 + }, + { + "epoch": 2.21, + "learning_rate": 1.698002366703227e-05, + "loss": 0.0682, + "step": 16884 + }, + { + "epoch": 2.21, + "learning_rate": 1.6979643893849834e-05, + "loss": 0.0715, + "step": 16885 + }, + { + "epoch": 2.21, + "learning_rate": 1.697926410103752e-05, + "loss": 0.0636, + "step": 16886 + }, + { + "epoch": 2.21, + "learning_rate": 1.6978884288596394e-05, + "loss": 0.0666, + "step": 16887 + }, + { + "epoch": 2.21, + "learning_rate": 1.6978504456527534e-05, + "loss": 0.0651, + "step": 16888 + }, + { + "epoch": 2.21, + "learning_rate": 1.6978124604832e-05, + "loss": 0.0564, + "step": 16889 + }, + { + "epoch": 2.21, + "learning_rate": 1.697774473351086e-05, + "loss": 0.0653, + "step": 16890 + }, + { + "epoch": 2.21, + "learning_rate": 1.6977364842565187e-05, + "loss": 0.0608, + "step": 16891 + }, + { + "epoch": 2.21, + "learning_rate": 1.6976984931996044e-05, + "loss": 0.0653, + "step": 16892 + }, + { + "epoch": 2.21, + "learning_rate": 1.6976605001804502e-05, + "loss": 0.0584, + "step": 16893 + }, + { + "epoch": 2.21, + "learning_rate": 1.697622505199163e-05, + "loss": 0.0717, + "step": 16894 + }, + { + "epoch": 2.21, + "learning_rate": 1.69758450825585e-05, + "loss": 0.0673, + "step": 16895 + }, + { + "epoch": 2.21, + "learning_rate": 1.6975465093506177e-05, + "loss": 0.0655, + "step": 16896 + }, + { + "epoch": 2.21, + "learning_rate": 1.697508508483573e-05, + "loss": 0.0708, + "step": 16897 + }, + { + "epoch": 2.21, + "learning_rate": 1.6974705056548223e-05, + "loss": 0.0708, + "step": 16898 + }, + { + "epoch": 2.21, + "learning_rate": 1.6974325008644733e-05, + "loss": 0.0594, + "step": 16899 + }, + { + "epoch": 2.21, + "learning_rate": 1.6973944941126324e-05, + "loss": 0.0686, + "step": 16900 + }, + { + "epoch": 2.21, + "learning_rate": 1.6973564853994065e-05, + "loss": 0.0596, + "step": 16901 + }, + { + "epoch": 2.21, + "learning_rate": 1.6973184747249032e-05, + "loss": 0.0655, + "step": 16902 + }, + { + "epoch": 2.21, + "learning_rate": 1.6972804620892286e-05, + "loss": 0.0569, + "step": 16903 + }, + { + "epoch": 2.21, + "learning_rate": 1.6972424474924895e-05, + "loss": 0.0651, + "step": 16904 + }, + { + "epoch": 2.21, + "learning_rate": 1.697204430934793e-05, + "loss": 0.0629, + "step": 16905 + }, + { + "epoch": 2.21, + "learning_rate": 1.6971664124162467e-05, + "loss": 0.0613, + "step": 16906 + }, + { + "epoch": 2.21, + "learning_rate": 1.697128391936957e-05, + "loss": 0.0658, + "step": 16907 + }, + { + "epoch": 2.21, + "learning_rate": 1.69709036949703e-05, + "loss": 0.0714, + "step": 16908 + }, + { + "epoch": 2.21, + "learning_rate": 1.6970523450965744e-05, + "loss": 0.0726, + "step": 16909 + }, + { + "epoch": 2.22, + "learning_rate": 1.697014318735696e-05, + "loss": 0.0555, + "step": 16910 + }, + { + "epoch": 2.22, + "learning_rate": 1.696976290414502e-05, + "loss": 0.0628, + "step": 16911 + }, + { + "epoch": 2.22, + "learning_rate": 1.696938260133099e-05, + "loss": 0.0604, + "step": 16912 + }, + { + "epoch": 2.22, + "learning_rate": 1.696900227891594e-05, + "loss": 0.0763, + "step": 16913 + }, + { + "epoch": 2.22, + "learning_rate": 1.6968621936900947e-05, + "loss": 0.0673, + "step": 16914 + }, + { + "epoch": 2.22, + "learning_rate": 1.6968241575287078e-05, + "loss": 0.0643, + "step": 16915 + }, + { + "epoch": 2.22, + "learning_rate": 1.6967861194075395e-05, + "loss": 0.0612, + "step": 16916 + }, + { + "epoch": 2.22, + "learning_rate": 1.6967480793266977e-05, + "loss": 0.0622, + "step": 16917 + }, + { + "epoch": 2.22, + "learning_rate": 1.696710037286289e-05, + "loss": 0.0697, + "step": 16918 + }, + { + "epoch": 2.22, + "learning_rate": 1.6966719932864202e-05, + "loss": 0.054, + "step": 16919 + }, + { + "epoch": 2.22, + "learning_rate": 1.6966339473271986e-05, + "loss": 0.0734, + "step": 16920 + }, + { + "epoch": 2.22, + "learning_rate": 1.6965958994087315e-05, + "loss": 0.0691, + "step": 16921 + }, + { + "epoch": 2.22, + "learning_rate": 1.696557849531125e-05, + "loss": 0.058, + "step": 16922 + }, + { + "epoch": 2.22, + "learning_rate": 1.696519797694487e-05, + "loss": 0.0635, + "step": 16923 + }, + { + "epoch": 2.22, + "learning_rate": 1.696481743898924e-05, + "loss": 0.069, + "step": 16924 + }, + { + "epoch": 2.22, + "learning_rate": 1.696443688144543e-05, + "loss": 0.0577, + "step": 16925 + }, + { + "epoch": 2.22, + "learning_rate": 1.6964056304314515e-05, + "loss": 0.0599, + "step": 16926 + }, + { + "epoch": 2.22, + "learning_rate": 1.696367570759756e-05, + "loss": 0.0584, + "step": 16927 + }, + { + "epoch": 2.22, + "learning_rate": 1.696329509129564e-05, + "loss": 0.0642, + "step": 16928 + }, + { + "epoch": 2.22, + "learning_rate": 1.6962914455409824e-05, + "loss": 0.065, + "step": 16929 + }, + { + "epoch": 2.22, + "learning_rate": 1.696253379994118e-05, + "loss": 0.0708, + "step": 16930 + }, + { + "epoch": 2.22, + "learning_rate": 1.6962153124890783e-05, + "loss": 0.0575, + "step": 16931 + }, + { + "epoch": 2.22, + "learning_rate": 1.6961772430259697e-05, + "loss": 0.0631, + "step": 16932 + }, + { + "epoch": 2.22, + "learning_rate": 1.6961391716049004e-05, + "loss": 0.0695, + "step": 16933 + }, + { + "epoch": 2.22, + "learning_rate": 1.696101098225976e-05, + "loss": 0.0596, + "step": 16934 + }, + { + "epoch": 2.22, + "learning_rate": 1.6960630228893046e-05, + "loss": 0.0682, + "step": 16935 + }, + { + "epoch": 2.22, + "learning_rate": 1.6960249455949932e-05, + "loss": 0.0657, + "step": 16936 + }, + { + "epoch": 2.22, + "learning_rate": 1.6959868663431484e-05, + "loss": 0.06, + "step": 16937 + }, + { + "epoch": 2.22, + "learning_rate": 1.6959487851338775e-05, + "loss": 0.0765, + "step": 16938 + }, + { + "epoch": 2.22, + "learning_rate": 1.695910701967288e-05, + "loss": 0.0737, + "step": 16939 + }, + { + "epoch": 2.22, + "learning_rate": 1.695872616843487e-05, + "loss": 0.0773, + "step": 16940 + }, + { + "epoch": 2.22, + "learning_rate": 1.6958345297625805e-05, + "loss": 0.0712, + "step": 16941 + }, + { + "epoch": 2.22, + "learning_rate": 1.695796440724677e-05, + "loss": 0.0606, + "step": 16942 + }, + { + "epoch": 2.22, + "learning_rate": 1.6957583497298832e-05, + "loss": 0.0647, + "step": 16943 + }, + { + "epoch": 2.22, + "learning_rate": 1.6957202567783054e-05, + "loss": 0.0666, + "step": 16944 + }, + { + "epoch": 2.22, + "learning_rate": 1.6956821618700523e-05, + "loss": 0.0712, + "step": 16945 + }, + { + "epoch": 2.22, + "learning_rate": 1.6956440650052297e-05, + "loss": 0.0685, + "step": 16946 + }, + { + "epoch": 2.22, + "learning_rate": 1.695605966183945e-05, + "loss": 0.0669, + "step": 16947 + }, + { + "epoch": 2.22, + "learning_rate": 1.695567865406306e-05, + "loss": 0.0754, + "step": 16948 + }, + { + "epoch": 2.22, + "learning_rate": 1.6955297626724193e-05, + "loss": 0.0703, + "step": 16949 + }, + { + "epoch": 2.22, + "learning_rate": 1.6954916579823922e-05, + "loss": 0.0674, + "step": 16950 + }, + { + "epoch": 2.22, + "learning_rate": 1.6954535513363316e-05, + "loss": 0.0589, + "step": 16951 + }, + { + "epoch": 2.22, + "learning_rate": 1.695415442734345e-05, + "loss": 0.0759, + "step": 16952 + }, + { + "epoch": 2.22, + "learning_rate": 1.6953773321765397e-05, + "loss": 0.0652, + "step": 16953 + }, + { + "epoch": 2.22, + "learning_rate": 1.6953392196630226e-05, + "loss": 0.0657, + "step": 16954 + }, + { + "epoch": 2.22, + "learning_rate": 1.695301105193901e-05, + "loss": 0.061, + "step": 16955 + }, + { + "epoch": 2.22, + "learning_rate": 1.6952629887692823e-05, + "loss": 0.0666, + "step": 16956 + }, + { + "epoch": 2.22, + "learning_rate": 1.695224870389273e-05, + "loss": 0.0675, + "step": 16957 + }, + { + "epoch": 2.22, + "learning_rate": 1.6951867500539808e-05, + "loss": 0.0725, + "step": 16958 + }, + { + "epoch": 2.22, + "learning_rate": 1.695148627763513e-05, + "loss": 0.0672, + "step": 16959 + }, + { + "epoch": 2.22, + "learning_rate": 1.695110503517977e-05, + "loss": 0.0748, + "step": 16960 + }, + { + "epoch": 2.22, + "learning_rate": 1.6950723773174794e-05, + "loss": 0.0646, + "step": 16961 + }, + { + "epoch": 2.22, + "learning_rate": 1.6950342491621278e-05, + "loss": 0.0668, + "step": 16962 + }, + { + "epoch": 2.22, + "learning_rate": 1.6949961190520294e-05, + "loss": 0.0728, + "step": 16963 + }, + { + "epoch": 2.22, + "learning_rate": 1.6949579869872913e-05, + "loss": 0.0721, + "step": 16964 + }, + { + "epoch": 2.22, + "learning_rate": 1.6949198529680214e-05, + "loss": 0.0605, + "step": 16965 + }, + { + "epoch": 2.22, + "learning_rate": 1.694881716994326e-05, + "loss": 0.0648, + "step": 16966 + }, + { + "epoch": 2.22, + "learning_rate": 1.6948435790663126e-05, + "loss": 0.0736, + "step": 16967 + }, + { + "epoch": 2.22, + "learning_rate": 1.694805439184089e-05, + "loss": 0.0618, + "step": 16968 + }, + { + "epoch": 2.22, + "learning_rate": 1.694767297347762e-05, + "loss": 0.0725, + "step": 16969 + }, + { + "epoch": 2.22, + "learning_rate": 1.6947291535574387e-05, + "loss": 0.0693, + "step": 16970 + }, + { + "epoch": 2.22, + "learning_rate": 1.6946910078132267e-05, + "loss": 0.0594, + "step": 16971 + }, + { + "epoch": 2.22, + "learning_rate": 1.6946528601152338e-05, + "loss": 0.0754, + "step": 16972 + }, + { + "epoch": 2.22, + "learning_rate": 1.6946147104635663e-05, + "loss": 0.0607, + "step": 16973 + }, + { + "epoch": 2.22, + "learning_rate": 1.6945765588583317e-05, + "loss": 0.0599, + "step": 16974 + }, + { + "epoch": 2.22, + "learning_rate": 1.694538405299638e-05, + "loss": 0.0635, + "step": 16975 + }, + { + "epoch": 2.22, + "learning_rate": 1.6945002497875917e-05, + "loss": 0.0691, + "step": 16976 + }, + { + "epoch": 2.22, + "learning_rate": 1.6944620923223007e-05, + "loss": 0.0612, + "step": 16977 + }, + { + "epoch": 2.22, + "learning_rate": 1.694423932903872e-05, + "loss": 0.0661, + "step": 16978 + }, + { + "epoch": 2.22, + "learning_rate": 1.6943857715324126e-05, + "loss": 0.0717, + "step": 16979 + }, + { + "epoch": 2.22, + "learning_rate": 1.6943476082080304e-05, + "loss": 0.0749, + "step": 16980 + }, + { + "epoch": 2.22, + "learning_rate": 1.6943094429308326e-05, + "loss": 0.0661, + "step": 16981 + }, + { + "epoch": 2.22, + "learning_rate": 1.6942712757009265e-05, + "loss": 0.0638, + "step": 16982 + }, + { + "epoch": 2.22, + "learning_rate": 1.6942331065184193e-05, + "loss": 0.0715, + "step": 16983 + }, + { + "epoch": 2.22, + "learning_rate": 1.6941949353834185e-05, + "loss": 0.0728, + "step": 16984 + }, + { + "epoch": 2.22, + "learning_rate": 1.6941567622960312e-05, + "loss": 0.0718, + "step": 16985 + }, + { + "epoch": 2.23, + "learning_rate": 1.6941185872563654e-05, + "loss": 0.0702, + "step": 16986 + }, + { + "epoch": 2.23, + "learning_rate": 1.694080410264528e-05, + "loss": 0.0712, + "step": 16987 + }, + { + "epoch": 2.23, + "learning_rate": 1.694042231320626e-05, + "loss": 0.0653, + "step": 16988 + }, + { + "epoch": 2.23, + "learning_rate": 1.6940040504247675e-05, + "loss": 0.062, + "step": 16989 + }, + { + "epoch": 2.23, + "learning_rate": 1.6939658675770595e-05, + "loss": 0.0788, + "step": 16990 + }, + { + "epoch": 2.23, + "learning_rate": 1.6939276827776094e-05, + "loss": 0.0628, + "step": 16991 + }, + { + "epoch": 2.23, + "learning_rate": 1.6938894960265245e-05, + "loss": 0.0687, + "step": 16992 + }, + { + "epoch": 2.23, + "learning_rate": 1.6938513073239127e-05, + "loss": 0.0762, + "step": 16993 + }, + { + "epoch": 2.23, + "learning_rate": 1.693813116669881e-05, + "loss": 0.067, + "step": 16994 + }, + { + "epoch": 2.23, + "learning_rate": 1.6937749240645363e-05, + "loss": 0.0665, + "step": 16995 + }, + { + "epoch": 2.23, + "learning_rate": 1.6937367295079873e-05, + "loss": 0.0712, + "step": 16996 + }, + { + "epoch": 2.23, + "learning_rate": 1.6936985330003402e-05, + "loss": 0.0741, + "step": 16997 + }, + { + "epoch": 2.23, + "learning_rate": 1.6936603345417032e-05, + "loss": 0.0498, + "step": 16998 + }, + { + "epoch": 2.23, + "learning_rate": 1.6936221341321836e-05, + "loss": 0.0673, + "step": 16999 + }, + { + "epoch": 2.23, + "learning_rate": 1.6935839317718885e-05, + "loss": 0.0601, + "step": 17000 + }, + { + "epoch": 2.23, + "learning_rate": 1.6935457274609254e-05, + "loss": 0.0606, + "step": 17001 + }, + { + "epoch": 2.23, + "learning_rate": 1.693507521199402e-05, + "loss": 0.0722, + "step": 17002 + }, + { + "epoch": 2.23, + "learning_rate": 1.6934693129874254e-05, + "loss": 0.0618, + "step": 17003 + }, + { + "epoch": 2.23, + "learning_rate": 1.6934311028251037e-05, + "loss": 0.0706, + "step": 17004 + }, + { + "epoch": 2.23, + "learning_rate": 1.6933928907125433e-05, + "loss": 0.0642, + "step": 17005 + }, + { + "epoch": 2.23, + "learning_rate": 1.693354676649853e-05, + "loss": 0.0695, + "step": 17006 + }, + { + "epoch": 2.23, + "learning_rate": 1.6933164606371392e-05, + "loss": 0.061, + "step": 17007 + }, + { + "epoch": 2.23, + "learning_rate": 1.69327824267451e-05, + "loss": 0.0744, + "step": 17008 + }, + { + "epoch": 2.23, + "learning_rate": 1.6932400227620725e-05, + "loss": 0.0664, + "step": 17009 + }, + { + "epoch": 2.23, + "learning_rate": 1.6932018008999345e-05, + "loss": 0.0556, + "step": 17010 + }, + { + "epoch": 2.23, + "learning_rate": 1.6931635770882034e-05, + "loss": 0.0684, + "step": 17011 + }, + { + "epoch": 2.23, + "learning_rate": 1.693125351326986e-05, + "loss": 0.0666, + "step": 17012 + }, + { + "epoch": 2.23, + "learning_rate": 1.6930871236163914e-05, + "loss": 0.0599, + "step": 17013 + }, + { + "epoch": 2.23, + "learning_rate": 1.6930488939565256e-05, + "loss": 0.0673, + "step": 17014 + }, + { + "epoch": 2.23, + "learning_rate": 1.6930106623474968e-05, + "loss": 0.0663, + "step": 17015 + }, + { + "epoch": 2.23, + "learning_rate": 1.6929724287894126e-05, + "loss": 0.0719, + "step": 17016 + }, + { + "epoch": 2.23, + "learning_rate": 1.69293419328238e-05, + "loss": 0.0702, + "step": 17017 + }, + { + "epoch": 2.23, + "learning_rate": 1.692895955826507e-05, + "loss": 0.0755, + "step": 17018 + }, + { + "epoch": 2.23, + "learning_rate": 1.6928577164219017e-05, + "loss": 0.0586, + "step": 17019 + }, + { + "epoch": 2.23, + "learning_rate": 1.69281947506867e-05, + "loss": 0.0651, + "step": 17020 + }, + { + "epoch": 2.23, + "learning_rate": 1.692781231766921e-05, + "loss": 0.0721, + "step": 17021 + }, + { + "epoch": 2.23, + "learning_rate": 1.6927429865167614e-05, + "loss": 0.0773, + "step": 17022 + }, + { + "epoch": 2.23, + "learning_rate": 1.692704739318299e-05, + "loss": 0.0697, + "step": 17023 + }, + { + "epoch": 2.23, + "learning_rate": 1.6926664901716415e-05, + "loss": 0.0683, + "step": 17024 + }, + { + "epoch": 2.23, + "learning_rate": 1.6926282390768965e-05, + "loss": 0.06, + "step": 17025 + }, + { + "epoch": 2.23, + "learning_rate": 1.6925899860341716e-05, + "loss": 0.0612, + "step": 17026 + }, + { + "epoch": 2.23, + "learning_rate": 1.692551731043574e-05, + "loss": 0.0547, + "step": 17027 + }, + { + "epoch": 2.23, + "learning_rate": 1.6925134741052115e-05, + "loss": 0.0639, + "step": 17028 + }, + { + "epoch": 2.23, + "learning_rate": 1.692475215219192e-05, + "loss": 0.0616, + "step": 17029 + }, + { + "epoch": 2.23, + "learning_rate": 1.6924369543856225e-05, + "loss": 0.0669, + "step": 17030 + }, + { + "epoch": 2.23, + "learning_rate": 1.692398691604611e-05, + "loss": 0.0618, + "step": 17031 + }, + { + "epoch": 2.23, + "learning_rate": 1.692360426876265e-05, + "loss": 0.0704, + "step": 17032 + }, + { + "epoch": 2.23, + "learning_rate": 1.6923221602006924e-05, + "loss": 0.075, + "step": 17033 + }, + { + "epoch": 2.23, + "learning_rate": 1.6922838915780005e-05, + "loss": 0.0749, + "step": 17034 + }, + { + "epoch": 2.23, + "learning_rate": 1.692245621008297e-05, + "loss": 0.0663, + "step": 17035 + }, + { + "epoch": 2.23, + "learning_rate": 1.6922073484916897e-05, + "loss": 0.0531, + "step": 17036 + }, + { + "epoch": 2.23, + "learning_rate": 1.6921690740282856e-05, + "loss": 0.0625, + "step": 17037 + }, + { + "epoch": 2.23, + "learning_rate": 1.6921307976181937e-05, + "loss": 0.0691, + "step": 17038 + }, + { + "epoch": 2.23, + "learning_rate": 1.6920925192615198e-05, + "loss": 0.0602, + "step": 17039 + }, + { + "epoch": 2.23, + "learning_rate": 1.692054238958373e-05, + "loss": 0.0722, + "step": 17040 + }, + { + "epoch": 2.23, + "learning_rate": 1.6920159567088608e-05, + "loss": 0.0596, + "step": 17041 + }, + { + "epoch": 2.23, + "learning_rate": 1.6919776725130904e-05, + "loss": 0.062, + "step": 17042 + }, + { + "epoch": 2.23, + "learning_rate": 1.6919393863711694e-05, + "loss": 0.0587, + "step": 17043 + }, + { + "epoch": 2.23, + "learning_rate": 1.6919010982832057e-05, + "loss": 0.067, + "step": 17044 + }, + { + "epoch": 2.23, + "learning_rate": 1.6918628082493075e-05, + "loss": 0.0592, + "step": 17045 + }, + { + "epoch": 2.23, + "learning_rate": 1.6918245162695818e-05, + "loss": 0.0721, + "step": 17046 + }, + { + "epoch": 2.23, + "learning_rate": 1.691786222344136e-05, + "loss": 0.0722, + "step": 17047 + }, + { + "epoch": 2.23, + "learning_rate": 1.691747926473079e-05, + "loss": 0.0611, + "step": 17048 + }, + { + "epoch": 2.23, + "learning_rate": 1.6917096286565175e-05, + "loss": 0.066, + "step": 17049 + }, + { + "epoch": 2.23, + "learning_rate": 1.691671328894559e-05, + "loss": 0.0639, + "step": 17050 + }, + { + "epoch": 2.23, + "learning_rate": 1.6916330271873125e-05, + "loss": 0.0629, + "step": 17051 + }, + { + "epoch": 2.23, + "learning_rate": 1.6915947235348847e-05, + "loss": 0.0716, + "step": 17052 + }, + { + "epoch": 2.23, + "learning_rate": 1.6915564179373832e-05, + "loss": 0.0786, + "step": 17053 + }, + { + "epoch": 2.23, + "learning_rate": 1.691518110394917e-05, + "loss": 0.0706, + "step": 17054 + }, + { + "epoch": 2.23, + "learning_rate": 1.6914798009075918e-05, + "loss": 0.0728, + "step": 17055 + }, + { + "epoch": 2.23, + "learning_rate": 1.691441489475517e-05, + "loss": 0.0634, + "step": 17056 + }, + { + "epoch": 2.23, + "learning_rate": 1.6914031760988e-05, + "loss": 0.0644, + "step": 17057 + }, + { + "epoch": 2.23, + "learning_rate": 1.6913648607775482e-05, + "loss": 0.0518, + "step": 17058 + }, + { + "epoch": 2.23, + "learning_rate": 1.6913265435118698e-05, + "loss": 0.0687, + "step": 17059 + }, + { + "epoch": 2.23, + "learning_rate": 1.691288224301872e-05, + "loss": 0.0669, + "step": 17060 + }, + { + "epoch": 2.23, + "learning_rate": 1.691249903147663e-05, + "loss": 0.0607, + "step": 17061 + }, + { + "epoch": 2.24, + "learning_rate": 1.6912115800493506e-05, + "loss": 0.0631, + "step": 17062 + }, + { + "epoch": 2.24, + "learning_rate": 1.691173255007042e-05, + "loss": 0.055, + "step": 17063 + }, + { + "epoch": 2.24, + "learning_rate": 1.691134928020846e-05, + "loss": 0.0682, + "step": 17064 + }, + { + "epoch": 2.24, + "learning_rate": 1.6910965990908695e-05, + "loss": 0.0667, + "step": 17065 + }, + { + "epoch": 2.24, + "learning_rate": 1.6910582682172205e-05, + "loss": 0.0698, + "step": 17066 + }, + { + "epoch": 2.24, + "learning_rate": 1.6910199354000073e-05, + "loss": 0.0711, + "step": 17067 + }, + { + "epoch": 2.24, + "learning_rate": 1.690981600639337e-05, + "loss": 0.0666, + "step": 17068 + }, + { + "epoch": 2.24, + "learning_rate": 1.6909432639353176e-05, + "loss": 0.061, + "step": 17069 + }, + { + "epoch": 2.24, + "learning_rate": 1.6909049252880573e-05, + "loss": 0.0662, + "step": 17070 + }, + { + "epoch": 2.24, + "learning_rate": 1.690866584697664e-05, + "loss": 0.0727, + "step": 17071 + }, + { + "epoch": 2.24, + "learning_rate": 1.6908282421642447e-05, + "loss": 0.0712, + "step": 17072 + }, + { + "epoch": 2.24, + "learning_rate": 1.690789897687908e-05, + "loss": 0.0634, + "step": 17073 + }, + { + "epoch": 2.24, + "learning_rate": 1.6907515512687614e-05, + "loss": 0.0687, + "step": 17074 + }, + { + "epoch": 2.24, + "learning_rate": 1.690713202906913e-05, + "loss": 0.0687, + "step": 17075 + }, + { + "epoch": 2.24, + "learning_rate": 1.6906748526024706e-05, + "loss": 0.0611, + "step": 17076 + }, + { + "epoch": 2.24, + "learning_rate": 1.6906365003555416e-05, + "loss": 0.0643, + "step": 17077 + }, + { + "epoch": 2.24, + "learning_rate": 1.6905981461662343e-05, + "loss": 0.0686, + "step": 17078 + }, + { + "epoch": 2.24, + "learning_rate": 1.6905597900346568e-05, + "loss": 0.0731, + "step": 17079 + }, + { + "epoch": 2.24, + "learning_rate": 1.6905214319609164e-05, + "loss": 0.0537, + "step": 17080 + }, + { + "epoch": 2.24, + "learning_rate": 1.6904830719451215e-05, + "loss": 0.0646, + "step": 17081 + }, + { + "epoch": 2.24, + "learning_rate": 1.6904447099873796e-05, + "loss": 0.0638, + "step": 17082 + }, + { + "epoch": 2.24, + "learning_rate": 1.6904063460877984e-05, + "loss": 0.0651, + "step": 17083 + }, + { + "epoch": 2.24, + "learning_rate": 1.6903679802464867e-05, + "loss": 0.0699, + "step": 17084 + }, + { + "epoch": 2.24, + "learning_rate": 1.6903296124635515e-05, + "loss": 0.0671, + "step": 17085 + }, + { + "epoch": 2.24, + "learning_rate": 1.690291242739101e-05, + "loss": 0.0625, + "step": 17086 + }, + { + "epoch": 2.24, + "learning_rate": 1.690252871073243e-05, + "loss": 0.0626, + "step": 17087 + }, + { + "epoch": 2.24, + "learning_rate": 1.6902144974660855e-05, + "loss": 0.0595, + "step": 17088 + }, + { + "epoch": 2.24, + "learning_rate": 1.690176121917737e-05, + "loss": 0.0692, + "step": 17089 + }, + { + "epoch": 2.24, + "learning_rate": 1.6901377444283046e-05, + "loss": 0.0782, + "step": 17090 + }, + { + "epoch": 2.24, + "learning_rate": 1.6900993649978966e-05, + "loss": 0.0646, + "step": 17091 + }, + { + "epoch": 2.24, + "learning_rate": 1.6900609836266206e-05, + "loss": 0.0679, + "step": 17092 + }, + { + "epoch": 2.24, + "learning_rate": 1.6900226003145852e-05, + "loss": 0.0773, + "step": 17093 + }, + { + "epoch": 2.24, + "learning_rate": 1.6899842150618977e-05, + "loss": 0.0628, + "step": 17094 + }, + { + "epoch": 2.24, + "learning_rate": 1.6899458278686665e-05, + "loss": 0.0747, + "step": 17095 + }, + { + "epoch": 2.24, + "learning_rate": 1.6899074387349994e-05, + "loss": 0.0724, + "step": 17096 + }, + { + "epoch": 2.24, + "learning_rate": 1.6898690476610044e-05, + "loss": 0.074, + "step": 17097 + }, + { + "epoch": 2.24, + "learning_rate": 1.6898306546467894e-05, + "loss": 0.0662, + "step": 17098 + }, + { + "epoch": 2.24, + "learning_rate": 1.6897922596924626e-05, + "loss": 0.0681, + "step": 17099 + }, + { + "epoch": 2.24, + "learning_rate": 1.6897538627981312e-05, + "loss": 0.058, + "step": 17100 + }, + { + "epoch": 2.24, + "learning_rate": 1.6897154639639043e-05, + "loss": 0.0611, + "step": 17101 + }, + { + "epoch": 2.24, + "learning_rate": 1.6896770631898896e-05, + "loss": 0.0609, + "step": 17102 + }, + { + "epoch": 2.24, + "learning_rate": 1.6896386604761946e-05, + "loss": 0.0492, + "step": 17103 + }, + { + "epoch": 2.24, + "learning_rate": 1.6896002558229276e-05, + "loss": 0.0672, + "step": 17104 + }, + { + "epoch": 2.24, + "learning_rate": 1.6895618492301963e-05, + "loss": 0.056, + "step": 17105 + }, + { + "epoch": 2.24, + "learning_rate": 1.6895234406981096e-05, + "loss": 0.0696, + "step": 17106 + }, + { + "epoch": 2.24, + "learning_rate": 1.6894850302267746e-05, + "loss": 0.0682, + "step": 17107 + }, + { + "epoch": 2.24, + "learning_rate": 1.6894466178163e-05, + "loss": 0.0566, + "step": 17108 + }, + { + "epoch": 2.24, + "learning_rate": 1.6894082034667932e-05, + "loss": 0.0694, + "step": 17109 + }, + { + "epoch": 2.24, + "learning_rate": 1.689369787178362e-05, + "loss": 0.0688, + "step": 17110 + }, + { + "epoch": 2.24, + "learning_rate": 1.689331368951116e-05, + "loss": 0.0649, + "step": 17111 + }, + { + "epoch": 2.24, + "learning_rate": 1.6892929487851616e-05, + "loss": 0.0652, + "step": 17112 + }, + { + "epoch": 2.24, + "learning_rate": 1.689254526680608e-05, + "loss": 0.0684, + "step": 17113 + }, + { + "epoch": 2.24, + "learning_rate": 1.689216102637562e-05, + "loss": 0.0604, + "step": 17114 + }, + { + "epoch": 2.24, + "learning_rate": 1.6891776766561328e-05, + "loss": 0.0556, + "step": 17115 + }, + { + "epoch": 2.24, + "learning_rate": 1.6891392487364283e-05, + "loss": 0.0587, + "step": 17116 + }, + { + "epoch": 2.24, + "learning_rate": 1.689100818878556e-05, + "loss": 0.053, + "step": 17117 + }, + { + "epoch": 2.24, + "learning_rate": 1.689062387082625e-05, + "loss": 0.0669, + "step": 17118 + }, + { + "epoch": 2.24, + "learning_rate": 1.689023953348742e-05, + "loss": 0.0634, + "step": 17119 + }, + { + "epoch": 2.24, + "learning_rate": 1.688985517677016e-05, + "loss": 0.0677, + "step": 17120 + }, + { + "epoch": 2.24, + "learning_rate": 1.6889470800675552e-05, + "loss": 0.0648, + "step": 17121 + }, + { + "epoch": 2.24, + "learning_rate": 1.688908640520467e-05, + "loss": 0.0755, + "step": 17122 + }, + { + "epoch": 2.24, + "learning_rate": 1.6888701990358605e-05, + "loss": 0.0619, + "step": 17123 + }, + { + "epoch": 2.24, + "learning_rate": 1.6888317556138427e-05, + "loss": 0.0556, + "step": 17124 + }, + { + "epoch": 2.24, + "learning_rate": 1.6887933102545226e-05, + "loss": 0.0665, + "step": 17125 + }, + { + "epoch": 2.24, + "learning_rate": 1.6887548629580076e-05, + "loss": 0.0611, + "step": 17126 + }, + { + "epoch": 2.24, + "learning_rate": 1.688716413724407e-05, + "loss": 0.0564, + "step": 17127 + }, + { + "epoch": 2.24, + "learning_rate": 1.6886779625538275e-05, + "loss": 0.0689, + "step": 17128 + }, + { + "epoch": 2.24, + "learning_rate": 1.688639509446378e-05, + "loss": 0.0627, + "step": 17129 + }, + { + "epoch": 2.24, + "learning_rate": 1.6886010544021667e-05, + "loss": 0.0579, + "step": 17130 + }, + { + "epoch": 2.24, + "learning_rate": 1.6885625974213015e-05, + "loss": 0.0628, + "step": 17131 + }, + { + "epoch": 2.24, + "learning_rate": 1.6885241385038905e-05, + "loss": 0.0659, + "step": 17132 + }, + { + "epoch": 2.24, + "learning_rate": 1.6884856776500423e-05, + "loss": 0.0673, + "step": 17133 + }, + { + "epoch": 2.24, + "learning_rate": 1.6884472148598647e-05, + "loss": 0.0515, + "step": 17134 + }, + { + "epoch": 2.24, + "learning_rate": 1.6884087501334662e-05, + "loss": 0.0634, + "step": 17135 + }, + { + "epoch": 2.24, + "learning_rate": 1.6883702834709544e-05, + "loss": 0.0678, + "step": 17136 + }, + { + "epoch": 2.24, + "learning_rate": 1.688331814872438e-05, + "loss": 0.0695, + "step": 17137 + }, + { + "epoch": 2.24, + "learning_rate": 1.688293344338025e-05, + "loss": 0.0577, + "step": 17138 + }, + { + "epoch": 2.25, + "learning_rate": 1.6882548718678234e-05, + "loss": 0.0671, + "step": 17139 + }, + { + "epoch": 2.25, + "learning_rate": 1.688216397461942e-05, + "loss": 0.0663, + "step": 17140 + }, + { + "epoch": 2.25, + "learning_rate": 1.6881779211204885e-05, + "loss": 0.0643, + "step": 17141 + }, + { + "epoch": 2.25, + "learning_rate": 1.6881394428435713e-05, + "loss": 0.0656, + "step": 17142 + }, + { + "epoch": 2.25, + "learning_rate": 1.6881009626312984e-05, + "loss": 0.0664, + "step": 17143 + }, + { + "epoch": 2.25, + "learning_rate": 1.6880624804837785e-05, + "loss": 0.0665, + "step": 17144 + }, + { + "epoch": 2.25, + "learning_rate": 1.688023996401119e-05, + "loss": 0.0578, + "step": 17145 + }, + { + "epoch": 2.25, + "learning_rate": 1.6879855103834294e-05, + "loss": 0.078, + "step": 17146 + }, + { + "epoch": 2.25, + "learning_rate": 1.6879470224308166e-05, + "loss": 0.0585, + "step": 17147 + }, + { + "epoch": 2.25, + "learning_rate": 1.687908532543389e-05, + "loss": 0.0641, + "step": 17148 + }, + { + "epoch": 2.25, + "learning_rate": 1.687870040721256e-05, + "loss": 0.0553, + "step": 17149 + }, + { + "epoch": 2.25, + "learning_rate": 1.6878315469645253e-05, + "loss": 0.0627, + "step": 17150 + }, + { + "epoch": 2.25, + "learning_rate": 1.6877930512733046e-05, + "loss": 0.0648, + "step": 17151 + }, + { + "epoch": 2.25, + "learning_rate": 1.687754553647703e-05, + "loss": 0.0671, + "step": 17152 + }, + { + "epoch": 2.25, + "learning_rate": 1.6877160540878278e-05, + "loss": 0.0628, + "step": 17153 + }, + { + "epoch": 2.25, + "learning_rate": 1.687677552593788e-05, + "loss": 0.0648, + "step": 17154 + }, + { + "epoch": 2.25, + "learning_rate": 1.6876390491656916e-05, + "loss": 0.0626, + "step": 17155 + }, + { + "epoch": 2.25, + "learning_rate": 1.6876005438036474e-05, + "loss": 0.0682, + "step": 17156 + }, + { + "epoch": 2.25, + "learning_rate": 1.687562036507763e-05, + "loss": 0.0644, + "step": 17157 + }, + { + "epoch": 2.25, + "learning_rate": 1.6875235272781468e-05, + "loss": 0.0624, + "step": 17158 + }, + { + "epoch": 2.25, + "learning_rate": 1.6874850161149075e-05, + "loss": 0.0682, + "step": 17159 + }, + { + "epoch": 2.25, + "learning_rate": 1.687446503018153e-05, + "loss": 0.0679, + "step": 17160 + }, + { + "epoch": 2.25, + "learning_rate": 1.6874079879879924e-05, + "loss": 0.0679, + "step": 17161 + }, + { + "epoch": 2.25, + "learning_rate": 1.6873694710245327e-05, + "loss": 0.06, + "step": 17162 + }, + { + "epoch": 2.25, + "learning_rate": 1.6873309521278836e-05, + "loss": 0.0586, + "step": 17163 + }, + { + "epoch": 2.25, + "learning_rate": 1.6872924312981528e-05, + "loss": 0.0618, + "step": 17164 + }, + { + "epoch": 2.25, + "learning_rate": 1.687253908535448e-05, + "loss": 0.0694, + "step": 17165 + }, + { + "epoch": 2.25, + "learning_rate": 1.6872153838398784e-05, + "loss": 0.0644, + "step": 17166 + }, + { + "epoch": 2.25, + "learning_rate": 1.6871768572115525e-05, + "loss": 0.0743, + "step": 17167 + }, + { + "epoch": 2.25, + "learning_rate": 1.687138328650578e-05, + "loss": 0.0647, + "step": 17168 + }, + { + "epoch": 2.25, + "learning_rate": 1.6870997981570636e-05, + "loss": 0.0753, + "step": 17169 + }, + { + "epoch": 2.25, + "learning_rate": 1.6870612657311177e-05, + "loss": 0.059, + "step": 17170 + }, + { + "epoch": 2.25, + "learning_rate": 1.6870227313728485e-05, + "loss": 0.0563, + "step": 17171 + }, + { + "epoch": 2.25, + "learning_rate": 1.6869841950823643e-05, + "loss": 0.0633, + "step": 17172 + }, + { + "epoch": 2.25, + "learning_rate": 1.686945656859774e-05, + "loss": 0.0667, + "step": 17173 + }, + { + "epoch": 2.25, + "learning_rate": 1.6869071167051853e-05, + "loss": 0.064, + "step": 17174 + }, + { + "epoch": 2.25, + "learning_rate": 1.6868685746187067e-05, + "loss": 0.0667, + "step": 17175 + }, + { + "epoch": 2.25, + "learning_rate": 1.686830030600447e-05, + "loss": 0.0599, + "step": 17176 + }, + { + "epoch": 2.25, + "learning_rate": 1.6867914846505147e-05, + "loss": 0.0742, + "step": 17177 + }, + { + "epoch": 2.25, + "learning_rate": 1.686752936769018e-05, + "loss": 0.058, + "step": 17178 + }, + { + "epoch": 2.25, + "learning_rate": 1.686714386956065e-05, + "loss": 0.0597, + "step": 17179 + }, + { + "epoch": 2.25, + "learning_rate": 1.686675835211764e-05, + "loss": 0.0699, + "step": 17180 + }, + { + "epoch": 2.25, + "learning_rate": 1.6866372815362245e-05, + "loss": 0.0715, + "step": 17181 + }, + { + "epoch": 2.25, + "learning_rate": 1.6865987259295537e-05, + "loss": 0.0638, + "step": 17182 + }, + { + "epoch": 2.25, + "learning_rate": 1.6865601683918606e-05, + "loss": 0.0651, + "step": 17183 + }, + { + "epoch": 2.25, + "learning_rate": 1.6865216089232535e-05, + "loss": 0.0571, + "step": 17184 + }, + { + "epoch": 2.25, + "learning_rate": 1.6864830475238413e-05, + "loss": 0.0669, + "step": 17185 + }, + { + "epoch": 2.25, + "learning_rate": 1.6864444841937317e-05, + "loss": 0.0673, + "step": 17186 + }, + { + "epoch": 2.25, + "learning_rate": 1.686405918933034e-05, + "loss": 0.0669, + "step": 17187 + }, + { + "epoch": 2.25, + "learning_rate": 1.686367351741856e-05, + "loss": 0.0639, + "step": 17188 + }, + { + "epoch": 2.25, + "learning_rate": 1.686328782620306e-05, + "loss": 0.0645, + "step": 17189 + }, + { + "epoch": 2.25, + "learning_rate": 1.6862902115684937e-05, + "loss": 0.0579, + "step": 17190 + }, + { + "epoch": 2.25, + "learning_rate": 1.686251638586526e-05, + "loss": 0.0776, + "step": 17191 + }, + { + "epoch": 2.25, + "learning_rate": 1.6862130636745124e-05, + "loss": 0.0637, + "step": 17192 + }, + { + "epoch": 2.25, + "learning_rate": 1.686174486832561e-05, + "loss": 0.0642, + "step": 17193 + }, + { + "epoch": 2.25, + "learning_rate": 1.68613590806078e-05, + "loss": 0.0547, + "step": 17194 + }, + { + "epoch": 2.25, + "learning_rate": 1.686097327359279e-05, + "loss": 0.0674, + "step": 17195 + }, + { + "epoch": 2.25, + "learning_rate": 1.6860587447281657e-05, + "loss": 0.0723, + "step": 17196 + }, + { + "epoch": 2.25, + "learning_rate": 1.6860201601675483e-05, + "loss": 0.0606, + "step": 17197 + }, + { + "epoch": 2.25, + "learning_rate": 1.685981573677536e-05, + "loss": 0.0675, + "step": 17198 + }, + { + "epoch": 2.25, + "learning_rate": 1.685942985258237e-05, + "loss": 0.0629, + "step": 17199 + }, + { + "epoch": 2.25, + "learning_rate": 1.68590439490976e-05, + "loss": 0.0567, + "step": 17200 + }, + { + "epoch": 2.25, + "learning_rate": 1.6858658026322132e-05, + "loss": 0.0705, + "step": 17201 + }, + { + "epoch": 2.25, + "learning_rate": 1.6858272084257055e-05, + "loss": 0.0766, + "step": 17202 + }, + { + "epoch": 2.25, + "learning_rate": 1.6857886122903452e-05, + "loss": 0.0616, + "step": 17203 + }, + { + "epoch": 2.25, + "learning_rate": 1.685750014226241e-05, + "loss": 0.0717, + "step": 17204 + }, + { + "epoch": 2.25, + "learning_rate": 1.6857114142335013e-05, + "loss": 0.066, + "step": 17205 + }, + { + "epoch": 2.25, + "learning_rate": 1.6856728123122352e-05, + "loss": 0.0651, + "step": 17206 + }, + { + "epoch": 2.25, + "learning_rate": 1.6856342084625505e-05, + "loss": 0.0488, + "step": 17207 + }, + { + "epoch": 2.25, + "learning_rate": 1.685595602684556e-05, + "loss": 0.0563, + "step": 17208 + }, + { + "epoch": 2.25, + "learning_rate": 1.6855569949783604e-05, + "loss": 0.0745, + "step": 17209 + }, + { + "epoch": 2.25, + "learning_rate": 1.6855183853440725e-05, + "loss": 0.0659, + "step": 17210 + }, + { + "epoch": 2.25, + "learning_rate": 1.6854797737818005e-05, + "loss": 0.0706, + "step": 17211 + }, + { + "epoch": 2.25, + "learning_rate": 1.685441160291653e-05, + "loss": 0.0743, + "step": 17212 + }, + { + "epoch": 2.25, + "learning_rate": 1.685402544873739e-05, + "loss": 0.0654, + "step": 17213 + }, + { + "epoch": 2.25, + "learning_rate": 1.6853639275281667e-05, + "loss": 0.0659, + "step": 17214 + }, + { + "epoch": 2.26, + "learning_rate": 1.685325308255045e-05, + "loss": 0.073, + "step": 17215 + }, + { + "epoch": 2.26, + "learning_rate": 1.685286687054482e-05, + "loss": 0.0617, + "step": 17216 + }, + { + "epoch": 2.26, + "learning_rate": 1.685248063926587e-05, + "loss": 0.0618, + "step": 17217 + }, + { + "epoch": 2.26, + "learning_rate": 1.6852094388714683e-05, + "loss": 0.0677, + "step": 17218 + }, + { + "epoch": 2.26, + "learning_rate": 1.6851708118892345e-05, + "loss": 0.06, + "step": 17219 + }, + { + "epoch": 2.26, + "learning_rate": 1.6851321829799943e-05, + "loss": 0.0657, + "step": 17220 + }, + { + "epoch": 2.26, + "learning_rate": 1.6850935521438562e-05, + "loss": 0.0741, + "step": 17221 + }, + { + "epoch": 2.26, + "learning_rate": 1.6850549193809288e-05, + "loss": 0.0652, + "step": 17222 + }, + { + "epoch": 2.26, + "learning_rate": 1.6850162846913213e-05, + "loss": 0.0619, + "step": 17223 + }, + { + "epoch": 2.26, + "learning_rate": 1.684977648075142e-05, + "loss": 0.0665, + "step": 17224 + }, + { + "epoch": 2.26, + "learning_rate": 1.684939009532499e-05, + "loss": 0.0608, + "step": 17225 + }, + { + "epoch": 2.26, + "learning_rate": 1.684900369063502e-05, + "loss": 0.0674, + "step": 17226 + }, + { + "epoch": 2.26, + "learning_rate": 1.684861726668259e-05, + "loss": 0.0592, + "step": 17227 + }, + { + "epoch": 2.26, + "learning_rate": 1.6848230823468793e-05, + "loss": 0.0679, + "step": 17228 + }, + { + "epoch": 2.26, + "learning_rate": 1.6847844360994707e-05, + "loss": 0.0647, + "step": 17229 + }, + { + "epoch": 2.26, + "learning_rate": 1.6847457879261423e-05, + "loss": 0.0553, + "step": 17230 + }, + { + "epoch": 2.26, + "learning_rate": 1.684707137827003e-05, + "loss": 0.0663, + "step": 17231 + }, + { + "epoch": 2.26, + "learning_rate": 1.6846684858021613e-05, + "loss": 0.0608, + "step": 17232 + }, + { + "epoch": 2.26, + "learning_rate": 1.684629831851726e-05, + "loss": 0.0563, + "step": 17233 + }, + { + "epoch": 2.26, + "learning_rate": 1.6845911759758058e-05, + "loss": 0.072, + "step": 17234 + }, + { + "epoch": 2.26, + "learning_rate": 1.6845525181745093e-05, + "loss": 0.0564, + "step": 17235 + }, + { + "epoch": 2.26, + "learning_rate": 1.6845138584479452e-05, + "loss": 0.0626, + "step": 17236 + }, + { + "epoch": 2.26, + "learning_rate": 1.6844751967962226e-05, + "loss": 0.0717, + "step": 17237 + }, + { + "epoch": 2.26, + "learning_rate": 1.6844365332194496e-05, + "loss": 0.0627, + "step": 17238 + }, + { + "epoch": 2.26, + "learning_rate": 1.6843978677177356e-05, + "loss": 0.0654, + "step": 17239 + }, + { + "epoch": 2.26, + "learning_rate": 1.684359200291189e-05, + "loss": 0.0661, + "step": 17240 + }, + { + "epoch": 2.26, + "learning_rate": 1.6843205309399183e-05, + "loss": 0.0537, + "step": 17241 + }, + { + "epoch": 2.26, + "learning_rate": 1.6842818596640328e-05, + "loss": 0.0693, + "step": 17242 + }, + { + "epoch": 2.26, + "learning_rate": 1.6842431864636407e-05, + "loss": 0.0697, + "step": 17243 + }, + { + "epoch": 2.26, + "learning_rate": 1.6842045113388515e-05, + "loss": 0.0705, + "step": 17244 + }, + { + "epoch": 2.26, + "learning_rate": 1.6841658342897735e-05, + "loss": 0.0581, + "step": 17245 + }, + { + "epoch": 2.26, + "learning_rate": 1.684127155316515e-05, + "loss": 0.0633, + "step": 17246 + }, + { + "epoch": 2.26, + "learning_rate": 1.6840884744191858e-05, + "loss": 0.0591, + "step": 17247 + }, + { + "epoch": 2.26, + "learning_rate": 1.6840497915978943e-05, + "loss": 0.0566, + "step": 17248 + }, + { + "epoch": 2.26, + "learning_rate": 1.6840111068527488e-05, + "loss": 0.0625, + "step": 17249 + }, + { + "epoch": 2.26, + "learning_rate": 1.6839724201838586e-05, + "loss": 0.0725, + "step": 17250 + }, + { + "epoch": 2.26, + "learning_rate": 1.6839337315913326e-05, + "loss": 0.0654, + "step": 17251 + }, + { + "epoch": 2.26, + "learning_rate": 1.6838950410752794e-05, + "loss": 0.0596, + "step": 17252 + }, + { + "epoch": 2.26, + "learning_rate": 1.6838563486358075e-05, + "loss": 0.0718, + "step": 17253 + }, + { + "epoch": 2.26, + "learning_rate": 1.6838176542730264e-05, + "loss": 0.0711, + "step": 17254 + }, + { + "epoch": 2.26, + "learning_rate": 1.6837789579870444e-05, + "loss": 0.0646, + "step": 17255 + }, + { + "epoch": 2.26, + "learning_rate": 1.6837402597779707e-05, + "loss": 0.073, + "step": 17256 + }, + { + "epoch": 2.26, + "learning_rate": 1.6837015596459134e-05, + "loss": 0.0669, + "step": 17257 + }, + { + "epoch": 2.26, + "learning_rate": 1.6836628575909824e-05, + "loss": 0.0551, + "step": 17258 + }, + { + "epoch": 2.26, + "learning_rate": 1.6836241536132857e-05, + "loss": 0.061, + "step": 17259 + }, + { + "epoch": 2.26, + "learning_rate": 1.683585447712933e-05, + "loss": 0.0662, + "step": 17260 + }, + { + "epoch": 2.26, + "learning_rate": 1.683546739890032e-05, + "loss": 0.0816, + "step": 17261 + }, + { + "epoch": 2.26, + "learning_rate": 1.6835080301446923e-05, + "loss": 0.0651, + "step": 17262 + }, + { + "epoch": 2.26, + "learning_rate": 1.6834693184770226e-05, + "loss": 0.0592, + "step": 17263 + }, + { + "epoch": 2.26, + "learning_rate": 1.683430604887132e-05, + "loss": 0.0682, + "step": 17264 + }, + { + "epoch": 2.26, + "learning_rate": 1.6833918893751294e-05, + "loss": 0.0677, + "step": 17265 + }, + { + "epoch": 2.26, + "learning_rate": 1.6833531719411234e-05, + "loss": 0.0641, + "step": 17266 + }, + { + "epoch": 2.26, + "learning_rate": 1.683314452585223e-05, + "loss": 0.0573, + "step": 17267 + }, + { + "epoch": 2.26, + "learning_rate": 1.683275731307537e-05, + "loss": 0.0476, + "step": 17268 + }, + { + "epoch": 2.26, + "learning_rate": 1.6832370081081744e-05, + "loss": 0.0663, + "step": 17269 + }, + { + "epoch": 2.26, + "learning_rate": 1.683198282987244e-05, + "loss": 0.066, + "step": 17270 + }, + { + "epoch": 2.26, + "learning_rate": 1.683159555944855e-05, + "loss": 0.0616, + "step": 17271 + }, + { + "epoch": 2.26, + "learning_rate": 1.6831208269811162e-05, + "loss": 0.065, + "step": 17272 + }, + { + "epoch": 2.26, + "learning_rate": 1.6830820960961362e-05, + "loss": 0.0652, + "step": 17273 + }, + { + "epoch": 2.26, + "learning_rate": 1.6830433632900244e-05, + "loss": 0.0665, + "step": 17274 + }, + { + "epoch": 2.26, + "learning_rate": 1.6830046285628894e-05, + "loss": 0.0662, + "step": 17275 + }, + { + "epoch": 2.26, + "learning_rate": 1.68296589191484e-05, + "loss": 0.0743, + "step": 17276 + }, + { + "epoch": 2.26, + "learning_rate": 1.682927153345986e-05, + "loss": 0.0622, + "step": 17277 + }, + { + "epoch": 2.26, + "learning_rate": 1.6828884128564352e-05, + "loss": 0.0496, + "step": 17278 + }, + { + "epoch": 2.26, + "learning_rate": 1.6828496704462973e-05, + "loss": 0.0651, + "step": 17279 + }, + { + "epoch": 2.26, + "learning_rate": 1.6828109261156812e-05, + "loss": 0.0699, + "step": 17280 + }, + { + "epoch": 2.26, + "learning_rate": 1.6827721798646958e-05, + "loss": 0.0831, + "step": 17281 + }, + { + "epoch": 2.26, + "learning_rate": 1.6827334316934494e-05, + "loss": 0.0585, + "step": 17282 + }, + { + "epoch": 2.26, + "learning_rate": 1.6826946816020522e-05, + "loss": 0.0745, + "step": 17283 + }, + { + "epoch": 2.26, + "learning_rate": 1.6826559295906122e-05, + "loss": 0.0582, + "step": 17284 + }, + { + "epoch": 2.26, + "learning_rate": 1.682617175659239e-05, + "loss": 0.0746, + "step": 17285 + }, + { + "epoch": 2.26, + "learning_rate": 1.6825784198080413e-05, + "loss": 0.0639, + "step": 17286 + }, + { + "epoch": 2.26, + "learning_rate": 1.682539662037128e-05, + "loss": 0.0547, + "step": 17287 + }, + { + "epoch": 2.26, + "learning_rate": 1.6825009023466082e-05, + "loss": 0.066, + "step": 17288 + }, + { + "epoch": 2.26, + "learning_rate": 1.6824621407365912e-05, + "loss": 0.0673, + "step": 17289 + }, + { + "epoch": 2.26, + "learning_rate": 1.6824233772071857e-05, + "loss": 0.0534, + "step": 17290 + }, + { + "epoch": 2.26, + "learning_rate": 1.6823846117585008e-05, + "loss": 0.0727, + "step": 17291 + }, + { + "epoch": 2.27, + "learning_rate": 1.6823458443906455e-05, + "loss": 0.0695, + "step": 17292 + }, + { + "epoch": 2.27, + "learning_rate": 1.682307075103729e-05, + "loss": 0.0632, + "step": 17293 + }, + { + "epoch": 2.27, + "learning_rate": 1.68226830389786e-05, + "loss": 0.0681, + "step": 17294 + }, + { + "epoch": 2.27, + "learning_rate": 1.6822295307731477e-05, + "loss": 0.0612, + "step": 17295 + }, + { + "epoch": 2.27, + "learning_rate": 1.682190755729701e-05, + "loss": 0.0616, + "step": 17296 + }, + { + "epoch": 2.27, + "learning_rate": 1.6821519787676294e-05, + "loss": 0.0568, + "step": 17297 + }, + { + "epoch": 2.27, + "learning_rate": 1.682113199887042e-05, + "loss": 0.0608, + "step": 17298 + }, + { + "epoch": 2.27, + "learning_rate": 1.682074419088047e-05, + "loss": 0.0563, + "step": 17299 + }, + { + "epoch": 2.27, + "learning_rate": 1.6820356363707544e-05, + "loss": 0.0644, + "step": 17300 + }, + { + "epoch": 2.27, + "learning_rate": 1.6819968517352726e-05, + "loss": 0.0708, + "step": 17301 + }, + { + "epoch": 2.27, + "learning_rate": 1.681958065181711e-05, + "loss": 0.0637, + "step": 17302 + }, + { + "epoch": 2.27, + "learning_rate": 1.6819192767101787e-05, + "loss": 0.0576, + "step": 17303 + }, + { + "epoch": 2.27, + "learning_rate": 1.6818804863207848e-05, + "loss": 0.0733, + "step": 17304 + }, + { + "epoch": 2.27, + "learning_rate": 1.6818416940136385e-05, + "loss": 0.0636, + "step": 17305 + }, + { + "epoch": 2.27, + "learning_rate": 1.6818028997888482e-05, + "loss": 0.0741, + "step": 17306 + }, + { + "epoch": 2.27, + "learning_rate": 1.681764103646524e-05, + "loss": 0.0643, + "step": 17307 + }, + { + "epoch": 2.27, + "learning_rate": 1.6817253055867746e-05, + "loss": 0.0623, + "step": 17308 + }, + { + "epoch": 2.27, + "learning_rate": 1.681686505609709e-05, + "loss": 0.0581, + "step": 17309 + }, + { + "epoch": 2.27, + "learning_rate": 1.6816477037154363e-05, + "loss": 0.0622, + "step": 17310 + }, + { + "epoch": 2.27, + "learning_rate": 1.6816088999040657e-05, + "loss": 0.0632, + "step": 17311 + }, + { + "epoch": 2.27, + "learning_rate": 1.6815700941757068e-05, + "loss": 0.0629, + "step": 17312 + }, + { + "epoch": 2.27, + "learning_rate": 1.681531286530468e-05, + "loss": 0.0577, + "step": 17313 + }, + { + "epoch": 2.27, + "learning_rate": 1.6814924769684585e-05, + "loss": 0.0579, + "step": 17314 + }, + { + "epoch": 2.27, + "learning_rate": 1.681453665489788e-05, + "loss": 0.0651, + "step": 17315 + }, + { + "epoch": 2.27, + "learning_rate": 1.6814148520945656e-05, + "loss": 0.0656, + "step": 17316 + }, + { + "epoch": 2.27, + "learning_rate": 1.6813760367828996e-05, + "loss": 0.0619, + "step": 17317 + }, + { + "epoch": 2.27, + "learning_rate": 1.6813372195549002e-05, + "loss": 0.0596, + "step": 17318 + }, + { + "epoch": 2.27, + "learning_rate": 1.6812984004106757e-05, + "loss": 0.0717, + "step": 17319 + }, + { + "epoch": 2.27, + "learning_rate": 1.6812595793503364e-05, + "loss": 0.0717, + "step": 17320 + }, + { + "epoch": 2.27, + "learning_rate": 1.6812207563739905e-05, + "loss": 0.0665, + "step": 17321 + }, + { + "epoch": 2.27, + "learning_rate": 1.6811819314817473e-05, + "loss": 0.0591, + "step": 17322 + }, + { + "epoch": 2.27, + "learning_rate": 1.6811431046737165e-05, + "loss": 0.0631, + "step": 17323 + }, + { + "epoch": 2.27, + "learning_rate": 1.6811042759500068e-05, + "loss": 0.0686, + "step": 17324 + }, + { + "epoch": 2.27, + "learning_rate": 1.6810654453107277e-05, + "loss": 0.0749, + "step": 17325 + }, + { + "epoch": 2.27, + "learning_rate": 1.6810266127559883e-05, + "loss": 0.0652, + "step": 17326 + }, + { + "epoch": 2.27, + "learning_rate": 1.680987778285898e-05, + "loss": 0.0699, + "step": 17327 + }, + { + "epoch": 2.27, + "learning_rate": 1.6809489419005656e-05, + "loss": 0.0475, + "step": 17328 + }, + { + "epoch": 2.27, + "learning_rate": 1.6809101036001002e-05, + "loss": 0.0574, + "step": 17329 + }, + { + "epoch": 2.27, + "learning_rate": 1.6808712633846118e-05, + "loss": 0.0538, + "step": 17330 + }, + { + "epoch": 2.27, + "learning_rate": 1.6808324212542097e-05, + "loss": 0.0629, + "step": 17331 + }, + { + "epoch": 2.27, + "learning_rate": 1.680793577209002e-05, + "loss": 0.0629, + "step": 17332 + }, + { + "epoch": 2.27, + "learning_rate": 1.6807547312490986e-05, + "loss": 0.0636, + "step": 17333 + }, + { + "epoch": 2.27, + "learning_rate": 1.680715883374609e-05, + "loss": 0.0561, + "step": 17334 + }, + { + "epoch": 2.27, + "learning_rate": 1.6806770335856422e-05, + "loss": 0.0717, + "step": 17335 + }, + { + "epoch": 2.27, + "learning_rate": 1.6806381818823078e-05, + "loss": 0.0709, + "step": 17336 + }, + { + "epoch": 2.27, + "learning_rate": 1.680599328264714e-05, + "loss": 0.0687, + "step": 17337 + }, + { + "epoch": 2.27, + "learning_rate": 1.6805604727329717e-05, + "loss": 0.0635, + "step": 17338 + }, + { + "epoch": 2.27, + "learning_rate": 1.6805216152871886e-05, + "loss": 0.0591, + "step": 17339 + }, + { + "epoch": 2.27, + "learning_rate": 1.680482755927475e-05, + "loss": 0.0716, + "step": 17340 + }, + { + "epoch": 2.27, + "learning_rate": 1.6804438946539398e-05, + "loss": 0.0725, + "step": 17341 + }, + { + "epoch": 2.27, + "learning_rate": 1.6804050314666926e-05, + "loss": 0.0626, + "step": 17342 + }, + { + "epoch": 2.27, + "learning_rate": 1.6803661663658425e-05, + "loss": 0.0607, + "step": 17343 + }, + { + "epoch": 2.27, + "learning_rate": 1.6803272993514985e-05, + "loss": 0.0672, + "step": 17344 + }, + { + "epoch": 2.27, + "learning_rate": 1.6802884304237705e-05, + "loss": 0.0625, + "step": 17345 + }, + { + "epoch": 2.27, + "learning_rate": 1.680249559582767e-05, + "loss": 0.0667, + "step": 17346 + }, + { + "epoch": 2.27, + "learning_rate": 1.6802106868285982e-05, + "loss": 0.0572, + "step": 17347 + }, + { + "epoch": 2.27, + "learning_rate": 1.6801718121613733e-05, + "loss": 0.0622, + "step": 17348 + }, + { + "epoch": 2.27, + "learning_rate": 1.680132935581201e-05, + "loss": 0.0681, + "step": 17349 + }, + { + "epoch": 2.27, + "learning_rate": 1.680094057088191e-05, + "loss": 0.0593, + "step": 17350 + }, + { + "epoch": 2.27, + "learning_rate": 1.680055176682453e-05, + "loss": 0.0679, + "step": 17351 + }, + { + "epoch": 2.27, + "learning_rate": 1.680016294364096e-05, + "loss": 0.0682, + "step": 17352 + }, + { + "epoch": 2.27, + "learning_rate": 1.6799774101332293e-05, + "loss": 0.0645, + "step": 17353 + }, + { + "epoch": 2.27, + "learning_rate": 1.6799385239899625e-05, + "loss": 0.0592, + "step": 17354 + }, + { + "epoch": 2.27, + "learning_rate": 1.6798996359344047e-05, + "loss": 0.0691, + "step": 17355 + }, + { + "epoch": 2.27, + "learning_rate": 1.6798607459666652e-05, + "loss": 0.0693, + "step": 17356 + }, + { + "epoch": 2.27, + "learning_rate": 1.679821854086854e-05, + "loss": 0.059, + "step": 17357 + }, + { + "epoch": 2.27, + "learning_rate": 1.6797829602950794e-05, + "loss": 0.0545, + "step": 17358 + }, + { + "epoch": 2.27, + "learning_rate": 1.6797440645914517e-05, + "loss": 0.0602, + "step": 17359 + }, + { + "epoch": 2.27, + "learning_rate": 1.67970516697608e-05, + "loss": 0.057, + "step": 17360 + }, + { + "epoch": 2.27, + "learning_rate": 1.6796662674490737e-05, + "loss": 0.0634, + "step": 17361 + }, + { + "epoch": 2.27, + "learning_rate": 1.6796273660105424e-05, + "loss": 0.0621, + "step": 17362 + }, + { + "epoch": 2.27, + "learning_rate": 1.6795884626605954e-05, + "loss": 0.0647, + "step": 17363 + }, + { + "epoch": 2.27, + "learning_rate": 1.679549557399342e-05, + "loss": 0.0654, + "step": 17364 + }, + { + "epoch": 2.27, + "learning_rate": 1.6795106502268916e-05, + "loss": 0.0702, + "step": 17365 + }, + { + "epoch": 2.27, + "learning_rate": 1.6794717411433535e-05, + "loss": 0.0698, + "step": 17366 + }, + { + "epoch": 2.27, + "learning_rate": 1.6794328301488373e-05, + "loss": 0.0755, + "step": 17367 + }, + { + "epoch": 2.28, + "learning_rate": 1.6793939172434524e-05, + "loss": 0.0598, + "step": 17368 + }, + { + "epoch": 2.28, + "learning_rate": 1.6793550024273086e-05, + "loss": 0.0755, + "step": 17369 + }, + { + "epoch": 2.28, + "learning_rate": 1.679316085700515e-05, + "loss": 0.0648, + "step": 17370 + }, + { + "epoch": 2.28, + "learning_rate": 1.6792771670631807e-05, + "loss": 0.0627, + "step": 17371 + }, + { + "epoch": 2.28, + "learning_rate": 1.6792382465154156e-05, + "loss": 0.0576, + "step": 17372 + }, + { + "epoch": 2.28, + "learning_rate": 1.6791993240573297e-05, + "loss": 0.0681, + "step": 17373 + }, + { + "epoch": 2.28, + "learning_rate": 1.6791603996890313e-05, + "loss": 0.059, + "step": 17374 + }, + { + "epoch": 2.28, + "learning_rate": 1.6791214734106306e-05, + "loss": 0.0551, + "step": 17375 + }, + { + "epoch": 2.28, + "learning_rate": 1.679082545222237e-05, + "loss": 0.0699, + "step": 17376 + }, + { + "epoch": 2.28, + "learning_rate": 1.67904361512396e-05, + "loss": 0.0629, + "step": 17377 + }, + { + "epoch": 2.28, + "learning_rate": 1.679004683115909e-05, + "loss": 0.0632, + "step": 17378 + }, + { + "epoch": 2.28, + "learning_rate": 1.6789657491981932e-05, + "loss": 0.0686, + "step": 17379 + }, + { + "epoch": 2.28, + "learning_rate": 1.6789268133709224e-05, + "loss": 0.0702, + "step": 17380 + }, + { + "epoch": 2.28, + "learning_rate": 1.6788878756342062e-05, + "loss": 0.0619, + "step": 17381 + }, + { + "epoch": 2.28, + "learning_rate": 1.678848935988154e-05, + "loss": 0.0621, + "step": 17382 + }, + { + "epoch": 2.28, + "learning_rate": 1.6788099944328752e-05, + "loss": 0.066, + "step": 17383 + }, + { + "epoch": 2.28, + "learning_rate": 1.6787710509684796e-05, + "loss": 0.0614, + "step": 17384 + }, + { + "epoch": 2.28, + "learning_rate": 1.6787321055950767e-05, + "loss": 0.0673, + "step": 17385 + }, + { + "epoch": 2.28, + "learning_rate": 1.678693158312776e-05, + "loss": 0.0715, + "step": 17386 + }, + { + "epoch": 2.28, + "learning_rate": 1.6786542091216864e-05, + "loss": 0.0655, + "step": 17387 + }, + { + "epoch": 2.28, + "learning_rate": 1.6786152580219182e-05, + "loss": 0.0414, + "step": 17388 + }, + { + "epoch": 2.28, + "learning_rate": 1.678576305013581e-05, + "loss": 0.0481, + "step": 17389 + }, + { + "epoch": 2.28, + "learning_rate": 1.6785373500967838e-05, + "loss": 0.0716, + "step": 17390 + }, + { + "epoch": 2.28, + "learning_rate": 1.6784983932716364e-05, + "loss": 0.0659, + "step": 17391 + }, + { + "epoch": 2.28, + "learning_rate": 1.6784594345382484e-05, + "loss": 0.0659, + "step": 17392 + }, + { + "epoch": 2.28, + "learning_rate": 1.6784204738967297e-05, + "loss": 0.065, + "step": 17393 + }, + { + "epoch": 2.28, + "learning_rate": 1.678381511347189e-05, + "loss": 0.0562, + "step": 17394 + }, + { + "epoch": 2.28, + "learning_rate": 1.6783425468897366e-05, + "loss": 0.0622, + "step": 17395 + }, + { + "epoch": 2.28, + "learning_rate": 1.678303580524482e-05, + "loss": 0.0667, + "step": 17396 + }, + { + "epoch": 2.28, + "learning_rate": 1.678264612251535e-05, + "loss": 0.0606, + "step": 17397 + }, + { + "epoch": 2.28, + "learning_rate": 1.6782256420710047e-05, + "loss": 0.0627, + "step": 17398 + }, + { + "epoch": 2.28, + "learning_rate": 1.6781866699830008e-05, + "loss": 0.0663, + "step": 17399 + }, + { + "epoch": 2.28, + "learning_rate": 1.678147695987633e-05, + "loss": 0.0652, + "step": 17400 + }, + { + "epoch": 2.28, + "learning_rate": 1.678108720085011e-05, + "loss": 0.0677, + "step": 17401 + }, + { + "epoch": 2.28, + "learning_rate": 1.678069742275244e-05, + "loss": 0.0689, + "step": 17402 + }, + { + "epoch": 2.28, + "learning_rate": 1.6780307625584428e-05, + "loss": 0.0709, + "step": 17403 + }, + { + "epoch": 2.28, + "learning_rate": 1.6779917809347155e-05, + "loss": 0.0588, + "step": 17404 + }, + { + "epoch": 2.28, + "learning_rate": 1.6779527974041724e-05, + "loss": 0.0757, + "step": 17405 + }, + { + "epoch": 2.28, + "learning_rate": 1.6779138119669232e-05, + "loss": 0.0647, + "step": 17406 + }, + { + "epoch": 2.28, + "learning_rate": 1.677874824623078e-05, + "loss": 0.0645, + "step": 17407 + }, + { + "epoch": 2.28, + "learning_rate": 1.6778358353727454e-05, + "loss": 0.0673, + "step": 17408 + }, + { + "epoch": 2.28, + "learning_rate": 1.6777968442160357e-05, + "loss": 0.0639, + "step": 17409 + }, + { + "epoch": 2.28, + "learning_rate": 1.6777578511530586e-05, + "loss": 0.0642, + "step": 17410 + }, + { + "epoch": 2.28, + "learning_rate": 1.6777188561839237e-05, + "loss": 0.0672, + "step": 17411 + }, + { + "epoch": 2.28, + "learning_rate": 1.6776798593087407e-05, + "loss": 0.0739, + "step": 17412 + }, + { + "epoch": 2.28, + "learning_rate": 1.6776408605276188e-05, + "loss": 0.061, + "step": 17413 + }, + { + "epoch": 2.28, + "learning_rate": 1.6776018598406682e-05, + "loss": 0.0677, + "step": 17414 + }, + { + "epoch": 2.28, + "learning_rate": 1.6775628572479986e-05, + "loss": 0.0676, + "step": 17415 + }, + { + "epoch": 2.28, + "learning_rate": 1.6775238527497194e-05, + "loss": 0.0668, + "step": 17416 + }, + { + "epoch": 2.28, + "learning_rate": 1.6774848463459405e-05, + "loss": 0.05, + "step": 17417 + }, + { + "epoch": 2.28, + "learning_rate": 1.6774458380367716e-05, + "loss": 0.069, + "step": 17418 + }, + { + "epoch": 2.28, + "learning_rate": 1.6774068278223223e-05, + "loss": 0.0623, + "step": 17419 + }, + { + "epoch": 2.28, + "learning_rate": 1.6773678157027028e-05, + "loss": 0.0626, + "step": 17420 + }, + { + "epoch": 2.28, + "learning_rate": 1.6773288016780215e-05, + "loss": 0.0651, + "step": 17421 + }, + { + "epoch": 2.28, + "learning_rate": 1.6772897857483896e-05, + "loss": 0.0549, + "step": 17422 + }, + { + "epoch": 2.28, + "learning_rate": 1.677250767913916e-05, + "loss": 0.0679, + "step": 17423 + }, + { + "epoch": 2.28, + "learning_rate": 1.677211748174711e-05, + "loss": 0.0643, + "step": 17424 + }, + { + "epoch": 2.28, + "learning_rate": 1.677172726530884e-05, + "loss": 0.0688, + "step": 17425 + }, + { + "epoch": 2.28, + "learning_rate": 1.6771337029825444e-05, + "loss": 0.0791, + "step": 17426 + }, + { + "epoch": 2.28, + "learning_rate": 1.6770946775298023e-05, + "loss": 0.0688, + "step": 17427 + }, + { + "epoch": 2.28, + "learning_rate": 1.677055650172768e-05, + "loss": 0.0737, + "step": 17428 + }, + { + "epoch": 2.28, + "learning_rate": 1.6770166209115505e-05, + "loss": 0.0604, + "step": 17429 + }, + { + "epoch": 2.28, + "learning_rate": 1.6769775897462594e-05, + "loss": 0.0657, + "step": 17430 + }, + { + "epoch": 2.28, + "learning_rate": 1.6769385566770055e-05, + "loss": 0.0534, + "step": 17431 + }, + { + "epoch": 2.28, + "learning_rate": 1.6768995217038977e-05, + "loss": 0.0604, + "step": 17432 + }, + { + "epoch": 2.28, + "learning_rate": 1.676860484827046e-05, + "loss": 0.069, + "step": 17433 + }, + { + "epoch": 2.28, + "learning_rate": 1.67682144604656e-05, + "loss": 0.0571, + "step": 17434 + }, + { + "epoch": 2.28, + "learning_rate": 1.67678240536255e-05, + "loss": 0.0616, + "step": 17435 + }, + { + "epoch": 2.28, + "learning_rate": 1.6767433627751257e-05, + "loss": 0.0652, + "step": 17436 + }, + { + "epoch": 2.28, + "learning_rate": 1.6767043182843966e-05, + "loss": 0.0586, + "step": 17437 + }, + { + "epoch": 2.28, + "learning_rate": 1.6766652718904725e-05, + "loss": 0.0687, + "step": 17438 + }, + { + "epoch": 2.28, + "learning_rate": 1.6766262235934636e-05, + "loss": 0.06, + "step": 17439 + }, + { + "epoch": 2.28, + "learning_rate": 1.6765871733934793e-05, + "loss": 0.0609, + "step": 17440 + }, + { + "epoch": 2.28, + "learning_rate": 1.6765481212906297e-05, + "loss": 0.0622, + "step": 17441 + }, + { + "epoch": 2.28, + "learning_rate": 1.6765090672850243e-05, + "loss": 0.0685, + "step": 17442 + }, + { + "epoch": 2.28, + "learning_rate": 1.6764700113767735e-05, + "loss": 0.0654, + "step": 17443 + }, + { + "epoch": 2.29, + "learning_rate": 1.6764309535659868e-05, + "loss": 0.0579, + "step": 17444 + }, + { + "epoch": 2.29, + "learning_rate": 1.676391893852774e-05, + "loss": 0.0642, + "step": 17445 + }, + { + "epoch": 2.29, + "learning_rate": 1.6763528322372445e-05, + "loss": 0.057, + "step": 17446 + }, + { + "epoch": 2.29, + "learning_rate": 1.6763137687195093e-05, + "loss": 0.0653, + "step": 17447 + }, + { + "epoch": 2.29, + "learning_rate": 1.6762747032996775e-05, + "loss": 0.0664, + "step": 17448 + }, + { + "epoch": 2.29, + "learning_rate": 1.676235635977859e-05, + "loss": 0.0769, + "step": 17449 + }, + { + "epoch": 2.29, + "learning_rate": 1.676196566754164e-05, + "loss": 0.0685, + "step": 17450 + }, + { + "epoch": 2.29, + "learning_rate": 1.6761574956287015e-05, + "loss": 0.0623, + "step": 17451 + }, + { + "epoch": 2.29, + "learning_rate": 1.676118422601583e-05, + "loss": 0.0539, + "step": 17452 + }, + { + "epoch": 2.29, + "learning_rate": 1.676079347672917e-05, + "loss": 0.0613, + "step": 17453 + }, + { + "epoch": 2.29, + "learning_rate": 1.6760402708428137e-05, + "loss": 0.0717, + "step": 17454 + }, + { + "epoch": 2.29, + "learning_rate": 1.676001192111383e-05, + "loss": 0.0674, + "step": 17455 + }, + { + "epoch": 2.29, + "learning_rate": 1.6759621114787356e-05, + "loss": 0.0588, + "step": 17456 + }, + { + "epoch": 2.29, + "learning_rate": 1.6759230289449805e-05, + "loss": 0.068, + "step": 17457 + }, + { + "epoch": 2.29, + "learning_rate": 1.6758839445102278e-05, + "loss": 0.071, + "step": 17458 + }, + { + "epoch": 2.29, + "learning_rate": 1.675844858174587e-05, + "loss": 0.0635, + "step": 17459 + }, + { + "epoch": 2.29, + "learning_rate": 1.6758057699381695e-05, + "loss": 0.0551, + "step": 17460 + }, + { + "epoch": 2.29, + "learning_rate": 1.6757666798010835e-05, + "loss": 0.0585, + "step": 17461 + }, + { + "epoch": 2.29, + "learning_rate": 1.67572758776344e-05, + "loss": 0.0617, + "step": 17462 + }, + { + "epoch": 2.29, + "learning_rate": 1.6756884938253485e-05, + "loss": 0.0778, + "step": 17463 + }, + { + "epoch": 2.29, + "learning_rate": 1.6756493979869196e-05, + "loss": 0.0634, + "step": 17464 + }, + { + "epoch": 2.29, + "learning_rate": 1.6756103002482618e-05, + "loss": 0.0673, + "step": 17465 + }, + { + "epoch": 2.29, + "learning_rate": 1.6755712006094865e-05, + "loss": 0.0568, + "step": 17466 + }, + { + "epoch": 2.29, + "learning_rate": 1.6755320990707035e-05, + "loss": 0.0671, + "step": 17467 + }, + { + "epoch": 2.29, + "learning_rate": 1.6754929956320222e-05, + "loss": 0.0684, + "step": 17468 + }, + { + "epoch": 2.29, + "learning_rate": 1.675453890293553e-05, + "loss": 0.0622, + "step": 17469 + }, + { + "epoch": 2.29, + "learning_rate": 1.675414783055405e-05, + "loss": 0.0691, + "step": 17470 + }, + { + "epoch": 2.29, + "learning_rate": 1.6753756739176896e-05, + "loss": 0.0641, + "step": 17471 + }, + { + "epoch": 2.29, + "learning_rate": 1.675336562880516e-05, + "loss": 0.0642, + "step": 17472 + }, + { + "epoch": 2.29, + "learning_rate": 1.6752974499439942e-05, + "loss": 0.0629, + "step": 17473 + }, + { + "epoch": 2.29, + "learning_rate": 1.6752583351082342e-05, + "loss": 0.0532, + "step": 17474 + }, + { + "epoch": 2.29, + "learning_rate": 1.675219218373346e-05, + "loss": 0.0831, + "step": 17475 + }, + { + "epoch": 2.29, + "learning_rate": 1.67518009973944e-05, + "loss": 0.0634, + "step": 17476 + }, + { + "epoch": 2.29, + "learning_rate": 1.6751409792066257e-05, + "loss": 0.07, + "step": 17477 + }, + { + "epoch": 2.29, + "learning_rate": 1.6751018567750134e-05, + "loss": 0.0517, + "step": 17478 + }, + { + "epoch": 2.29, + "learning_rate": 1.675062732444713e-05, + "loss": 0.0777, + "step": 17479 + }, + { + "epoch": 2.29, + "learning_rate": 1.675023606215835e-05, + "loss": 0.0727, + "step": 17480 + }, + { + "epoch": 2.29, + "learning_rate": 1.6749844780884882e-05, + "loss": 0.0558, + "step": 17481 + }, + { + "epoch": 2.29, + "learning_rate": 1.674945348062784e-05, + "loss": 0.0628, + "step": 17482 + }, + { + "epoch": 2.29, + "learning_rate": 1.6749062161388323e-05, + "loss": 0.0696, + "step": 17483 + }, + { + "epoch": 2.29, + "learning_rate": 1.6748670823167427e-05, + "loss": 0.0683, + "step": 17484 + }, + { + "epoch": 2.29, + "learning_rate": 1.674827946596625e-05, + "loss": 0.0624, + "step": 17485 + }, + { + "epoch": 2.29, + "learning_rate": 1.67478880897859e-05, + "loss": 0.064, + "step": 17486 + }, + { + "epoch": 2.29, + "learning_rate": 1.674749669462747e-05, + "loss": 0.0576, + "step": 17487 + }, + { + "epoch": 2.29, + "learning_rate": 1.674710528049207e-05, + "loss": 0.0668, + "step": 17488 + }, + { + "epoch": 2.29, + "learning_rate": 1.6746713847380792e-05, + "loss": 0.0677, + "step": 17489 + }, + { + "epoch": 2.29, + "learning_rate": 1.674632239529474e-05, + "loss": 0.0704, + "step": 17490 + }, + { + "epoch": 2.29, + "learning_rate": 1.674593092423502e-05, + "loss": 0.0638, + "step": 17491 + }, + { + "epoch": 2.29, + "learning_rate": 1.6745539434202723e-05, + "loss": 0.0598, + "step": 17492 + }, + { + "epoch": 2.29, + "learning_rate": 1.674514792519896e-05, + "loss": 0.0685, + "step": 17493 + }, + { + "epoch": 2.29, + "learning_rate": 1.6744756397224824e-05, + "loss": 0.0753, + "step": 17494 + }, + { + "epoch": 2.29, + "learning_rate": 1.6744364850281424e-05, + "loss": 0.0546, + "step": 17495 + }, + { + "epoch": 2.29, + "learning_rate": 1.6743973284369852e-05, + "loss": 0.0638, + "step": 17496 + }, + { + "epoch": 2.29, + "learning_rate": 1.6743581699491217e-05, + "loss": 0.0668, + "step": 17497 + }, + { + "epoch": 2.29, + "learning_rate": 1.6743190095646617e-05, + "loss": 0.0733, + "step": 17498 + }, + { + "epoch": 2.29, + "learning_rate": 1.6742798472837155e-05, + "loss": 0.0615, + "step": 17499 + }, + { + "epoch": 2.29, + "learning_rate": 1.674240683106393e-05, + "loss": 0.0622, + "step": 17500 + }, + { + "epoch": 2.29, + "learning_rate": 1.6742015170328047e-05, + "loss": 0.0723, + "step": 17501 + }, + { + "epoch": 2.29, + "learning_rate": 1.6741623490630603e-05, + "loss": 0.0635, + "step": 17502 + }, + { + "epoch": 2.29, + "learning_rate": 1.6741231791972702e-05, + "loss": 0.0788, + "step": 17503 + }, + { + "epoch": 2.29, + "learning_rate": 1.6740840074355446e-05, + "loss": 0.0528, + "step": 17504 + }, + { + "epoch": 2.29, + "learning_rate": 1.674044833777994e-05, + "loss": 0.0776, + "step": 17505 + }, + { + "epoch": 2.29, + "learning_rate": 1.6740056582247275e-05, + "loss": 0.0577, + "step": 17506 + }, + { + "epoch": 2.29, + "learning_rate": 1.6739664807758567e-05, + "loss": 0.0594, + "step": 17507 + }, + { + "epoch": 2.29, + "learning_rate": 1.6739273014314905e-05, + "loss": 0.0663, + "step": 17508 + }, + { + "epoch": 2.29, + "learning_rate": 1.67388812019174e-05, + "loss": 0.0552, + "step": 17509 + }, + { + "epoch": 2.29, + "learning_rate": 1.673848937056715e-05, + "loss": 0.0635, + "step": 17510 + }, + { + "epoch": 2.29, + "learning_rate": 1.6738097520265255e-05, + "loss": 0.0556, + "step": 17511 + }, + { + "epoch": 2.29, + "learning_rate": 1.673770565101282e-05, + "loss": 0.0639, + "step": 17512 + }, + { + "epoch": 2.29, + "learning_rate": 1.673731376281095e-05, + "loss": 0.0606, + "step": 17513 + }, + { + "epoch": 2.29, + "learning_rate": 1.673692185566074e-05, + "loss": 0.0636, + "step": 17514 + }, + { + "epoch": 2.29, + "learning_rate": 1.67365299295633e-05, + "loss": 0.0641, + "step": 17515 + }, + { + "epoch": 2.29, + "learning_rate": 1.6736137984519725e-05, + "loss": 0.0715, + "step": 17516 + }, + { + "epoch": 2.29, + "learning_rate": 1.6735746020531122e-05, + "loss": 0.0639, + "step": 17517 + }, + { + "epoch": 2.29, + "learning_rate": 1.6735354037598593e-05, + "loss": 0.0657, + "step": 17518 + }, + { + "epoch": 2.29, + "learning_rate": 1.6734962035723236e-05, + "loss": 0.0681, + "step": 17519 + }, + { + "epoch": 2.29, + "learning_rate": 1.673457001490616e-05, + "loss": 0.0601, + "step": 17520 + }, + { + "epoch": 2.3, + "learning_rate": 1.6734177975148463e-05, + "loss": 0.0669, + "step": 17521 + }, + { + "epoch": 2.3, + "learning_rate": 1.673378591645125e-05, + "loss": 0.0606, + "step": 17522 + }, + { + "epoch": 2.3, + "learning_rate": 1.6733393838815622e-05, + "loss": 0.0704, + "step": 17523 + }, + { + "epoch": 2.3, + "learning_rate": 1.6733001742242683e-05, + "loss": 0.0681, + "step": 17524 + }, + { + "epoch": 2.3, + "learning_rate": 1.6732609626733536e-05, + "loss": 0.0529, + "step": 17525 + }, + { + "epoch": 2.3, + "learning_rate": 1.6732217492289283e-05, + "loss": 0.0605, + "step": 17526 + }, + { + "epoch": 2.3, + "learning_rate": 1.6731825338911024e-05, + "loss": 0.0632, + "step": 17527 + }, + { + "epoch": 2.3, + "learning_rate": 1.6731433166599865e-05, + "loss": 0.0731, + "step": 17528 + }, + { + "epoch": 2.3, + "learning_rate": 1.6731040975356913e-05, + "loss": 0.0729, + "step": 17529 + }, + { + "epoch": 2.3, + "learning_rate": 1.6730648765183262e-05, + "loss": 0.0645, + "step": 17530 + }, + { + "epoch": 2.3, + "learning_rate": 1.673025653608002e-05, + "loss": 0.0705, + "step": 17531 + }, + { + "epoch": 2.3, + "learning_rate": 1.6729864288048294e-05, + "loss": 0.059, + "step": 17532 + }, + { + "epoch": 2.3, + "learning_rate": 1.672947202108918e-05, + "loss": 0.0667, + "step": 17533 + }, + { + "epoch": 2.3, + "learning_rate": 1.6729079735203786e-05, + "loss": 0.0633, + "step": 17534 + }, + { + "epoch": 2.3, + "learning_rate": 1.672868743039321e-05, + "loss": 0.0582, + "step": 17535 + }, + { + "epoch": 2.3, + "learning_rate": 1.672829510665856e-05, + "loss": 0.0651, + "step": 17536 + }, + { + "epoch": 2.3, + "learning_rate": 1.6727902764000942e-05, + "loss": 0.0647, + "step": 17537 + }, + { + "epoch": 2.3, + "learning_rate": 1.6727510402421453e-05, + "loss": 0.0711, + "step": 17538 + }, + { + "epoch": 2.3, + "learning_rate": 1.67271180219212e-05, + "loss": 0.0779, + "step": 17539 + }, + { + "epoch": 2.3, + "learning_rate": 1.6726725622501286e-05, + "loss": 0.0647, + "step": 17540 + }, + { + "epoch": 2.3, + "learning_rate": 1.6726333204162814e-05, + "loss": 0.0571, + "step": 17541 + }, + { + "epoch": 2.3, + "learning_rate": 1.6725940766906888e-05, + "loss": 0.0628, + "step": 17542 + }, + { + "epoch": 2.3, + "learning_rate": 1.672554831073461e-05, + "loss": 0.0608, + "step": 17543 + }, + { + "epoch": 2.3, + "learning_rate": 1.672515583564709e-05, + "loss": 0.0653, + "step": 17544 + }, + { + "epoch": 2.3, + "learning_rate": 1.6724763341645425e-05, + "loss": 0.0696, + "step": 17545 + }, + { + "epoch": 2.3, + "learning_rate": 1.672437082873072e-05, + "loss": 0.0648, + "step": 17546 + }, + { + "epoch": 2.3, + "learning_rate": 1.672397829690408e-05, + "loss": 0.0598, + "step": 17547 + }, + { + "epoch": 2.3, + "learning_rate": 1.672358574616661e-05, + "loss": 0.0774, + "step": 17548 + }, + { + "epoch": 2.3, + "learning_rate": 1.6723193176519413e-05, + "loss": 0.0677, + "step": 17549 + }, + { + "epoch": 2.3, + "learning_rate": 1.6722800587963596e-05, + "loss": 0.0566, + "step": 17550 + }, + { + "epoch": 2.3, + "learning_rate": 1.6722407980500258e-05, + "loss": 0.0672, + "step": 17551 + }, + { + "epoch": 2.3, + "learning_rate": 1.6722015354130504e-05, + "loss": 0.061, + "step": 17552 + }, + { + "epoch": 2.3, + "learning_rate": 1.6721622708855442e-05, + "loss": 0.0627, + "step": 17553 + }, + { + "epoch": 2.3, + "learning_rate": 1.672123004467617e-05, + "loss": 0.0707, + "step": 17554 + }, + { + "epoch": 2.3, + "learning_rate": 1.67208373615938e-05, + "loss": 0.0689, + "step": 17555 + }, + { + "epoch": 2.3, + "learning_rate": 1.6720444659609435e-05, + "loss": 0.0636, + "step": 17556 + }, + { + "epoch": 2.3, + "learning_rate": 1.6720051938724174e-05, + "loss": 0.0662, + "step": 17557 + }, + { + "epoch": 2.3, + "learning_rate": 1.6719659198939125e-05, + "loss": 0.064, + "step": 17558 + }, + { + "epoch": 2.3, + "learning_rate": 1.6719266440255394e-05, + "loss": 0.0697, + "step": 17559 + }, + { + "epoch": 2.3, + "learning_rate": 1.671887366267408e-05, + "loss": 0.0646, + "step": 17560 + }, + { + "epoch": 2.3, + "learning_rate": 1.6718480866196296e-05, + "loss": 0.0658, + "step": 17561 + }, + { + "epoch": 2.3, + "learning_rate": 1.671808805082314e-05, + "loss": 0.0615, + "step": 17562 + }, + { + "epoch": 2.3, + "learning_rate": 1.671769521655572e-05, + "loss": 0.0679, + "step": 17563 + }, + { + "epoch": 2.3, + "learning_rate": 1.671730236339514e-05, + "loss": 0.0541, + "step": 17564 + }, + { + "epoch": 2.3, + "learning_rate": 1.6716909491342505e-05, + "loss": 0.0582, + "step": 17565 + }, + { + "epoch": 2.3, + "learning_rate": 1.671651660039892e-05, + "loss": 0.0687, + "step": 17566 + }, + { + "epoch": 2.3, + "learning_rate": 1.671612369056549e-05, + "loss": 0.0595, + "step": 17567 + }, + { + "epoch": 2.3, + "learning_rate": 1.671573076184332e-05, + "loss": 0.0623, + "step": 17568 + }, + { + "epoch": 2.3, + "learning_rate": 1.671533781423351e-05, + "loss": 0.0642, + "step": 17569 + }, + { + "epoch": 2.3, + "learning_rate": 1.6714944847737176e-05, + "loss": 0.0658, + "step": 17570 + }, + { + "epoch": 2.3, + "learning_rate": 1.6714551862355414e-05, + "loss": 0.06, + "step": 17571 + }, + { + "epoch": 2.3, + "learning_rate": 1.6714158858089334e-05, + "loss": 0.0689, + "step": 17572 + }, + { + "epoch": 2.3, + "learning_rate": 1.6713765834940037e-05, + "loss": 0.0652, + "step": 17573 + }, + { + "epoch": 2.3, + "learning_rate": 1.6713372792908635e-05, + "loss": 0.0511, + "step": 17574 + }, + { + "epoch": 2.3, + "learning_rate": 1.6712979731996227e-05, + "loss": 0.0722, + "step": 17575 + }, + { + "epoch": 2.3, + "learning_rate": 1.6712586652203922e-05, + "loss": 0.0658, + "step": 17576 + }, + { + "epoch": 2.3, + "learning_rate": 1.6712193553532826e-05, + "loss": 0.0683, + "step": 17577 + }, + { + "epoch": 2.3, + "learning_rate": 1.671180043598404e-05, + "loss": 0.0574, + "step": 17578 + }, + { + "epoch": 2.3, + "learning_rate": 1.6711407299558677e-05, + "loss": 0.0689, + "step": 17579 + }, + { + "epoch": 2.3, + "learning_rate": 1.6711014144257834e-05, + "loss": 0.0609, + "step": 17580 + }, + { + "epoch": 2.3, + "learning_rate": 1.6710620970082627e-05, + "loss": 0.0587, + "step": 17581 + }, + { + "epoch": 2.3, + "learning_rate": 1.6710227777034146e-05, + "loss": 0.0547, + "step": 17582 + }, + { + "epoch": 2.3, + "learning_rate": 1.6709834565113515e-05, + "loss": 0.067, + "step": 17583 + }, + { + "epoch": 2.3, + "learning_rate": 1.670944133432183e-05, + "loss": 0.0565, + "step": 17584 + }, + { + "epoch": 2.3, + "learning_rate": 1.6709048084660194e-05, + "loss": 0.0571, + "step": 17585 + }, + { + "epoch": 2.3, + "learning_rate": 1.670865481612972e-05, + "loss": 0.0602, + "step": 17586 + }, + { + "epoch": 2.3, + "learning_rate": 1.6708261528731515e-05, + "loss": 0.0618, + "step": 17587 + }, + { + "epoch": 2.3, + "learning_rate": 1.670786822246668e-05, + "loss": 0.0663, + "step": 17588 + }, + { + "epoch": 2.3, + "learning_rate": 1.670747489733632e-05, + "loss": 0.0598, + "step": 17589 + }, + { + "epoch": 2.3, + "learning_rate": 1.6707081553341545e-05, + "loss": 0.0699, + "step": 17590 + }, + { + "epoch": 2.3, + "learning_rate": 1.6706688190483464e-05, + "loss": 0.0604, + "step": 17591 + }, + { + "epoch": 2.3, + "learning_rate": 1.6706294808763175e-05, + "loss": 0.0595, + "step": 17592 + }, + { + "epoch": 2.3, + "learning_rate": 1.6705901408181794e-05, + "loss": 0.0702, + "step": 17593 + }, + { + "epoch": 2.3, + "learning_rate": 1.6705507988740418e-05, + "loss": 0.0639, + "step": 17594 + }, + { + "epoch": 2.3, + "learning_rate": 1.670511455044016e-05, + "loss": 0.0581, + "step": 17595 + }, + { + "epoch": 2.3, + "learning_rate": 1.6704721093282123e-05, + "loss": 0.0554, + "step": 17596 + }, + { + "epoch": 2.31, + "learning_rate": 1.6704327617267416e-05, + "loss": 0.0613, + "step": 17597 + }, + { + "epoch": 2.31, + "learning_rate": 1.6703934122397143e-05, + "loss": 0.0694, + "step": 17598 + }, + { + "epoch": 2.31, + "learning_rate": 1.6703540608672414e-05, + "loss": 0.0631, + "step": 17599 + }, + { + "epoch": 2.31, + "learning_rate": 1.670314707609433e-05, + "loss": 0.067, + "step": 17600 + }, + { + "epoch": 2.31, + "learning_rate": 1.670275352466401e-05, + "loss": 0.0712, + "step": 17601 + }, + { + "epoch": 2.31, + "learning_rate": 1.6702359954382548e-05, + "loss": 0.0686, + "step": 17602 + }, + { + "epoch": 2.31, + "learning_rate": 1.6701966365251053e-05, + "loss": 0.0673, + "step": 17603 + }, + { + "epoch": 2.31, + "learning_rate": 1.670157275727064e-05, + "loss": 0.0671, + "step": 17604 + }, + { + "epoch": 2.31, + "learning_rate": 1.6701179130442408e-05, + "loss": 0.0618, + "step": 17605 + }, + { + "epoch": 2.31, + "learning_rate": 1.6700785484767466e-05, + "loss": 0.0521, + "step": 17606 + }, + { + "epoch": 2.31, + "learning_rate": 1.670039182024692e-05, + "loss": 0.0636, + "step": 17607 + }, + { + "epoch": 2.31, + "learning_rate": 1.669999813688188e-05, + "loss": 0.0693, + "step": 17608 + }, + { + "epoch": 2.31, + "learning_rate": 1.6699604434673452e-05, + "loss": 0.06, + "step": 17609 + }, + { + "epoch": 2.31, + "learning_rate": 1.6699210713622746e-05, + "loss": 0.0661, + "step": 17610 + }, + { + "epoch": 2.31, + "learning_rate": 1.6698816973730866e-05, + "loss": 0.0722, + "step": 17611 + }, + { + "epoch": 2.31, + "learning_rate": 1.6698423214998916e-05, + "loss": 0.0604, + "step": 17612 + }, + { + "epoch": 2.31, + "learning_rate": 1.669802943742801e-05, + "loss": 0.0732, + "step": 17613 + }, + { + "epoch": 2.31, + "learning_rate": 1.6697635641019254e-05, + "loss": 0.0602, + "step": 17614 + }, + { + "epoch": 2.31, + "learning_rate": 1.6697241825773754e-05, + "loss": 0.0654, + "step": 17615 + }, + { + "epoch": 2.31, + "learning_rate": 1.6696847991692616e-05, + "loss": 0.0747, + "step": 17616 + }, + { + "epoch": 2.31, + "learning_rate": 1.6696454138776952e-05, + "loss": 0.0589, + "step": 17617 + }, + { + "epoch": 2.31, + "learning_rate": 1.6696060267027863e-05, + "loss": 0.0672, + "step": 17618 + }, + { + "epoch": 2.31, + "learning_rate": 1.6695666376446467e-05, + "loss": 0.0737, + "step": 17619 + }, + { + "epoch": 2.31, + "learning_rate": 1.669527246703386e-05, + "loss": 0.059, + "step": 17620 + }, + { + "epoch": 2.31, + "learning_rate": 1.669487853879116e-05, + "loss": 0.0639, + "step": 17621 + }, + { + "epoch": 2.31, + "learning_rate": 1.6694484591719474e-05, + "loss": 0.055, + "step": 17622 + }, + { + "epoch": 2.31, + "learning_rate": 1.6694090625819898e-05, + "loss": 0.0594, + "step": 17623 + }, + { + "epoch": 2.31, + "learning_rate": 1.6693696641093554e-05, + "loss": 0.0641, + "step": 17624 + }, + { + "epoch": 2.31, + "learning_rate": 1.6693302637541544e-05, + "loss": 0.0606, + "step": 17625 + }, + { + "epoch": 2.31, + "learning_rate": 1.6692908615164975e-05, + "loss": 0.0754, + "step": 17626 + }, + { + "epoch": 2.31, + "learning_rate": 1.669251457396496e-05, + "loss": 0.056, + "step": 17627 + }, + { + "epoch": 2.31, + "learning_rate": 1.66921205139426e-05, + "loss": 0.0636, + "step": 17628 + }, + { + "epoch": 2.31, + "learning_rate": 1.669172643509901e-05, + "loss": 0.0532, + "step": 17629 + }, + { + "epoch": 2.31, + "learning_rate": 1.66913323374353e-05, + "loss": 0.0624, + "step": 17630 + }, + { + "epoch": 2.31, + "learning_rate": 1.6690938220952566e-05, + "loss": 0.0573, + "step": 17631 + }, + { + "epoch": 2.31, + "learning_rate": 1.669054408565193e-05, + "loss": 0.046, + "step": 17632 + }, + { + "epoch": 2.31, + "learning_rate": 1.6690149931534494e-05, + "loss": 0.0676, + "step": 17633 + }, + { + "epoch": 2.31, + "learning_rate": 1.6689755758601363e-05, + "loss": 0.0569, + "step": 17634 + }, + { + "epoch": 2.31, + "learning_rate": 1.6689361566853655e-05, + "loss": 0.0713, + "step": 17635 + }, + { + "epoch": 2.31, + "learning_rate": 1.6688967356292475e-05, + "loss": 0.0641, + "step": 17636 + }, + { + "epoch": 2.31, + "learning_rate": 1.668857312691893e-05, + "loss": 0.0612, + "step": 17637 + }, + { + "epoch": 2.31, + "learning_rate": 1.6688178878734125e-05, + "loss": 0.0645, + "step": 17638 + }, + { + "epoch": 2.31, + "learning_rate": 1.6687784611739174e-05, + "loss": 0.056, + "step": 17639 + }, + { + "epoch": 2.31, + "learning_rate": 1.668739032593519e-05, + "loss": 0.0624, + "step": 17640 + }, + { + "epoch": 2.31, + "learning_rate": 1.6686996021323276e-05, + "loss": 0.0688, + "step": 17641 + }, + { + "epoch": 2.31, + "learning_rate": 1.6686601697904536e-05, + "loss": 0.0591, + "step": 17642 + }, + { + "epoch": 2.31, + "learning_rate": 1.6686207355680087e-05, + "loss": 0.0618, + "step": 17643 + }, + { + "epoch": 2.31, + "learning_rate": 1.668581299465104e-05, + "loss": 0.0776, + "step": 17644 + }, + { + "epoch": 2.31, + "learning_rate": 1.6685418614818493e-05, + "loss": 0.0659, + "step": 17645 + }, + { + "epoch": 2.31, + "learning_rate": 1.6685024216183567e-05, + "loss": 0.0563, + "step": 17646 + }, + { + "epoch": 2.31, + "learning_rate": 1.6684629798747365e-05, + "loss": 0.0668, + "step": 17647 + }, + { + "epoch": 2.31, + "learning_rate": 1.6684235362511e-05, + "loss": 0.0584, + "step": 17648 + }, + { + "epoch": 2.31, + "learning_rate": 1.6683840907475576e-05, + "loss": 0.0738, + "step": 17649 + }, + { + "epoch": 2.31, + "learning_rate": 1.6683446433642206e-05, + "loss": 0.0826, + "step": 17650 + }, + { + "epoch": 2.31, + "learning_rate": 1.6683051941012e-05, + "loss": 0.065, + "step": 17651 + }, + { + "epoch": 2.31, + "learning_rate": 1.6682657429586065e-05, + "loss": 0.0683, + "step": 17652 + }, + { + "epoch": 2.31, + "learning_rate": 1.6682262899365516e-05, + "loss": 0.0632, + "step": 17653 + }, + { + "epoch": 2.31, + "learning_rate": 1.6681868350351454e-05, + "loss": 0.0508, + "step": 17654 + }, + { + "epoch": 2.31, + "learning_rate": 1.6681473782544994e-05, + "loss": 0.0588, + "step": 17655 + }, + { + "epoch": 2.31, + "learning_rate": 1.6681079195947247e-05, + "loss": 0.0772, + "step": 17656 + }, + { + "epoch": 2.31, + "learning_rate": 1.6680684590559317e-05, + "loss": 0.0674, + "step": 17657 + }, + { + "epoch": 2.31, + "learning_rate": 1.6680289966382316e-05, + "loss": 0.07, + "step": 17658 + }, + { + "epoch": 2.31, + "learning_rate": 1.667989532341736e-05, + "loss": 0.0582, + "step": 17659 + }, + { + "epoch": 2.31, + "learning_rate": 1.6679500661665552e-05, + "loss": 0.0638, + "step": 17660 + }, + { + "epoch": 2.31, + "learning_rate": 1.6679105981128e-05, + "loss": 0.0599, + "step": 17661 + }, + { + "epoch": 2.31, + "learning_rate": 1.6678711281805825e-05, + "loss": 0.061, + "step": 17662 + }, + { + "epoch": 2.31, + "learning_rate": 1.6678316563700128e-05, + "loss": 0.0706, + "step": 17663 + }, + { + "epoch": 2.31, + "learning_rate": 1.6677921826812024e-05, + "loss": 0.062, + "step": 17664 + }, + { + "epoch": 2.31, + "learning_rate": 1.6677527071142617e-05, + "loss": 0.0682, + "step": 17665 + }, + { + "epoch": 2.31, + "learning_rate": 1.667713229669302e-05, + "loss": 0.0699, + "step": 17666 + }, + { + "epoch": 2.31, + "learning_rate": 1.6676737503464343e-05, + "loss": 0.0653, + "step": 17667 + }, + { + "epoch": 2.31, + "learning_rate": 1.6676342691457698e-05, + "loss": 0.0636, + "step": 17668 + }, + { + "epoch": 2.31, + "learning_rate": 1.66759478606742e-05, + "loss": 0.0664, + "step": 17669 + }, + { + "epoch": 2.31, + "learning_rate": 1.6675553011114947e-05, + "loss": 0.0688, + "step": 17670 + }, + { + "epoch": 2.31, + "learning_rate": 1.6675158142781062e-05, + "loss": 0.068, + "step": 17671 + }, + { + "epoch": 2.31, + "learning_rate": 1.6674763255673646e-05, + "loss": 0.0598, + "step": 17672 + }, + { + "epoch": 2.32, + "learning_rate": 1.6674368349793816e-05, + "loss": 0.0652, + "step": 17673 + }, + { + "epoch": 2.32, + "learning_rate": 1.667397342514268e-05, + "loss": 0.0601, + "step": 17674 + }, + { + "epoch": 2.32, + "learning_rate": 1.6673578481721355e-05, + "loss": 0.0654, + "step": 17675 + }, + { + "epoch": 2.32, + "learning_rate": 1.6673183519530936e-05, + "loss": 0.062, + "step": 17676 + }, + { + "epoch": 2.32, + "learning_rate": 1.6672788538572548e-05, + "loss": 0.0631, + "step": 17677 + }, + { + "epoch": 2.32, + "learning_rate": 1.6672393538847298e-05, + "loss": 0.0708, + "step": 17678 + }, + { + "epoch": 2.32, + "learning_rate": 1.6671998520356297e-05, + "loss": 0.0617, + "step": 17679 + }, + { + "epoch": 2.32, + "learning_rate": 1.6671603483100654e-05, + "loss": 0.0635, + "step": 17680 + }, + { + "epoch": 2.32, + "learning_rate": 1.6671208427081485e-05, + "loss": 0.0696, + "step": 17681 + }, + { + "epoch": 2.32, + "learning_rate": 1.6670813352299894e-05, + "loss": 0.0722, + "step": 17682 + }, + { + "epoch": 2.32, + "learning_rate": 1.6670418258756993e-05, + "loss": 0.0692, + "step": 17683 + }, + { + "epoch": 2.32, + "learning_rate": 1.6670023146453898e-05, + "loss": 0.0636, + "step": 17684 + }, + { + "epoch": 2.32, + "learning_rate": 1.6669628015391724e-05, + "loss": 0.0701, + "step": 17685 + }, + { + "epoch": 2.32, + "learning_rate": 1.6669232865571568e-05, + "loss": 0.0687, + "step": 17686 + }, + { + "epoch": 2.32, + "learning_rate": 1.6668837696994556e-05, + "loss": 0.0615, + "step": 17687 + }, + { + "epoch": 2.32, + "learning_rate": 1.666844250966179e-05, + "loss": 0.0594, + "step": 17688 + }, + { + "epoch": 2.32, + "learning_rate": 1.6668047303574385e-05, + "loss": 0.0571, + "step": 17689 + }, + { + "epoch": 2.32, + "learning_rate": 1.6667652078733452e-05, + "loss": 0.0642, + "step": 17690 + }, + { + "epoch": 2.32, + "learning_rate": 1.6667256835140105e-05, + "loss": 0.0694, + "step": 17691 + }, + { + "epoch": 2.32, + "learning_rate": 1.666686157279545e-05, + "loss": 0.0631, + "step": 17692 + }, + { + "epoch": 2.32, + "learning_rate": 1.6666466291700604e-05, + "loss": 0.0714, + "step": 17693 + }, + { + "epoch": 2.32, + "learning_rate": 1.6666070991856673e-05, + "loss": 0.059, + "step": 17694 + }, + { + "epoch": 2.32, + "learning_rate": 1.6665675673264775e-05, + "loss": 0.0632, + "step": 17695 + }, + { + "epoch": 2.32, + "learning_rate": 1.6665280335926016e-05, + "loss": 0.0655, + "step": 17696 + }, + { + "epoch": 2.32, + "learning_rate": 1.6664884979841516e-05, + "loss": 0.0615, + "step": 17697 + }, + { + "epoch": 2.32, + "learning_rate": 1.666448960501238e-05, + "loss": 0.0673, + "step": 17698 + }, + { + "epoch": 2.32, + "learning_rate": 1.6664094211439724e-05, + "loss": 0.0742, + "step": 17699 + }, + { + "epoch": 2.32, + "learning_rate": 1.666369879912465e-05, + "loss": 0.0507, + "step": 17700 + }, + { + "epoch": 2.32, + "learning_rate": 1.6663303368068286e-05, + "loss": 0.0646, + "step": 17701 + }, + { + "epoch": 2.32, + "learning_rate": 1.6662907918271737e-05, + "loss": 0.0679, + "step": 17702 + }, + { + "epoch": 2.32, + "learning_rate": 1.666251244973611e-05, + "loss": 0.0778, + "step": 17703 + }, + { + "epoch": 2.32, + "learning_rate": 1.6662116962462523e-05, + "loss": 0.0591, + "step": 17704 + }, + { + "epoch": 2.32, + "learning_rate": 1.6661721456452083e-05, + "loss": 0.0612, + "step": 17705 + }, + { + "epoch": 2.32, + "learning_rate": 1.6661325931705912e-05, + "loss": 0.0666, + "step": 17706 + }, + { + "epoch": 2.32, + "learning_rate": 1.6660930388225114e-05, + "loss": 0.0554, + "step": 17707 + }, + { + "epoch": 2.32, + "learning_rate": 1.6660534826010803e-05, + "loss": 0.0695, + "step": 17708 + }, + { + "epoch": 2.32, + "learning_rate": 1.6660139245064094e-05, + "loss": 0.0653, + "step": 17709 + }, + { + "epoch": 2.32, + "learning_rate": 1.66597436453861e-05, + "loss": 0.0746, + "step": 17710 + }, + { + "epoch": 2.32, + "learning_rate": 1.6659348026977928e-05, + "loss": 0.055, + "step": 17711 + }, + { + "epoch": 2.32, + "learning_rate": 1.6658952389840693e-05, + "loss": 0.055, + "step": 17712 + }, + { + "epoch": 2.32, + "learning_rate": 1.6658556733975512e-05, + "loss": 0.0608, + "step": 17713 + }, + { + "epoch": 2.32, + "learning_rate": 1.6658161059383497e-05, + "loss": 0.0585, + "step": 17714 + }, + { + "epoch": 2.32, + "learning_rate": 1.6657765366065755e-05, + "loss": 0.0664, + "step": 17715 + }, + { + "epoch": 2.32, + "learning_rate": 1.6657369654023405e-05, + "loss": 0.0684, + "step": 17716 + }, + { + "epoch": 2.32, + "learning_rate": 1.6656973923257553e-05, + "loss": 0.0622, + "step": 17717 + }, + { + "epoch": 2.32, + "learning_rate": 1.665657817376932e-05, + "loss": 0.0772, + "step": 17718 + }, + { + "epoch": 2.32, + "learning_rate": 1.6656182405559815e-05, + "loss": 0.0585, + "step": 17719 + }, + { + "epoch": 2.32, + "learning_rate": 1.6655786618630153e-05, + "loss": 0.059, + "step": 17720 + }, + { + "epoch": 2.32, + "learning_rate": 1.665539081298144e-05, + "loss": 0.0517, + "step": 17721 + }, + { + "epoch": 2.32, + "learning_rate": 1.66549949886148e-05, + "loss": 0.0613, + "step": 17722 + }, + { + "epoch": 2.32, + "learning_rate": 1.6654599145531338e-05, + "loss": 0.0595, + "step": 17723 + }, + { + "epoch": 2.32, + "learning_rate": 1.6654203283732173e-05, + "loss": 0.0605, + "step": 17724 + }, + { + "epoch": 2.32, + "learning_rate": 1.6653807403218415e-05, + "loss": 0.0689, + "step": 17725 + }, + { + "epoch": 2.32, + "learning_rate": 1.6653411503991175e-05, + "loss": 0.0565, + "step": 17726 + }, + { + "epoch": 2.32, + "learning_rate": 1.665301558605157e-05, + "loss": 0.0626, + "step": 17727 + }, + { + "epoch": 2.32, + "learning_rate": 1.6652619649400715e-05, + "loss": 0.0795, + "step": 17728 + }, + { + "epoch": 2.32, + "learning_rate": 1.6652223694039724e-05, + "loss": 0.0613, + "step": 17729 + }, + { + "epoch": 2.32, + "learning_rate": 1.6651827719969703e-05, + "loss": 0.0699, + "step": 17730 + }, + { + "epoch": 2.32, + "learning_rate": 1.6651431727191774e-05, + "loss": 0.0706, + "step": 17731 + }, + { + "epoch": 2.32, + "learning_rate": 1.6651035715707048e-05, + "loss": 0.071, + "step": 17732 + }, + { + "epoch": 2.32, + "learning_rate": 1.6650639685516635e-05, + "loss": 0.0675, + "step": 17733 + }, + { + "epoch": 2.32, + "learning_rate": 1.665024363662165e-05, + "loss": 0.0669, + "step": 17734 + }, + { + "epoch": 2.32, + "learning_rate": 1.664984756902322e-05, + "loss": 0.0681, + "step": 17735 + }, + { + "epoch": 2.32, + "learning_rate": 1.664945148272244e-05, + "loss": 0.0583, + "step": 17736 + }, + { + "epoch": 2.32, + "learning_rate": 1.664905537772043e-05, + "loss": 0.0555, + "step": 17737 + }, + { + "epoch": 2.32, + "learning_rate": 1.6648659254018306e-05, + "loss": 0.0617, + "step": 17738 + }, + { + "epoch": 2.32, + "learning_rate": 1.6648263111617184e-05, + "loss": 0.0694, + "step": 17739 + }, + { + "epoch": 2.32, + "learning_rate": 1.664786695051818e-05, + "loss": 0.062, + "step": 17740 + }, + { + "epoch": 2.32, + "learning_rate": 1.66474707707224e-05, + "loss": 0.0582, + "step": 17741 + }, + { + "epoch": 2.32, + "learning_rate": 1.6647074572230964e-05, + "loss": 0.0622, + "step": 17742 + }, + { + "epoch": 2.32, + "learning_rate": 1.664667835504498e-05, + "loss": 0.0729, + "step": 17743 + }, + { + "epoch": 2.32, + "learning_rate": 1.664628211916557e-05, + "loss": 0.0627, + "step": 17744 + }, + { + "epoch": 2.32, + "learning_rate": 1.664588586459385e-05, + "loss": 0.0668, + "step": 17745 + }, + { + "epoch": 2.32, + "learning_rate": 1.6645489591330924e-05, + "loss": 0.0622, + "step": 17746 + }, + { + "epoch": 2.32, + "learning_rate": 1.6645093299377912e-05, + "loss": 0.0689, + "step": 17747 + }, + { + "epoch": 2.32, + "learning_rate": 1.6644696988735934e-05, + "loss": 0.0583, + "step": 17748 + }, + { + "epoch": 2.32, + "learning_rate": 1.6644300659406096e-05, + "loss": 0.0593, + "step": 17749 + }, + { + "epoch": 2.33, + "learning_rate": 1.6643904311389517e-05, + "loss": 0.0772, + "step": 17750 + }, + { + "epoch": 2.33, + "learning_rate": 1.664350794468731e-05, + "loss": 0.0774, + "step": 17751 + }, + { + "epoch": 2.33, + "learning_rate": 1.6643111559300592e-05, + "loss": 0.0699, + "step": 17752 + }, + { + "epoch": 2.33, + "learning_rate": 1.6642715155230477e-05, + "loss": 0.0635, + "step": 17753 + }, + { + "epoch": 2.33, + "learning_rate": 1.664231873247808e-05, + "loss": 0.0677, + "step": 17754 + }, + { + "epoch": 2.33, + "learning_rate": 1.6641922291044513e-05, + "loss": 0.0582, + "step": 17755 + }, + { + "epoch": 2.33, + "learning_rate": 1.6641525830930897e-05, + "loss": 0.0673, + "step": 17756 + }, + { + "epoch": 2.33, + "learning_rate": 1.664112935213834e-05, + "loss": 0.0624, + "step": 17757 + }, + { + "epoch": 2.33, + "learning_rate": 1.664073285466796e-05, + "loss": 0.0583, + "step": 17758 + }, + { + "epoch": 2.33, + "learning_rate": 1.6640336338520874e-05, + "loss": 0.0705, + "step": 17759 + }, + { + "epoch": 2.33, + "learning_rate": 1.6639939803698193e-05, + "loss": 0.0662, + "step": 17760 + }, + { + "epoch": 2.33, + "learning_rate": 1.6639543250201037e-05, + "loss": 0.0613, + "step": 17761 + }, + { + "epoch": 2.33, + "learning_rate": 1.663914667803052e-05, + "loss": 0.0825, + "step": 17762 + }, + { + "epoch": 2.33, + "learning_rate": 1.6638750087187755e-05, + "loss": 0.0561, + "step": 17763 + }, + { + "epoch": 2.33, + "learning_rate": 1.663835347767386e-05, + "loss": 0.0629, + "step": 17764 + }, + { + "epoch": 2.33, + "learning_rate": 1.6637956849489948e-05, + "loss": 0.0662, + "step": 17765 + }, + { + "epoch": 2.33, + "learning_rate": 1.6637560202637137e-05, + "loss": 0.0541, + "step": 17766 + }, + { + "epoch": 2.33, + "learning_rate": 1.6637163537116544e-05, + "loss": 0.0657, + "step": 17767 + }, + { + "epoch": 2.33, + "learning_rate": 1.663676685292928e-05, + "loss": 0.0679, + "step": 17768 + }, + { + "epoch": 2.33, + "learning_rate": 1.663637015007646e-05, + "loss": 0.0632, + "step": 17769 + }, + { + "epoch": 2.33, + "learning_rate": 1.6635973428559204e-05, + "loss": 0.0684, + "step": 17770 + }, + { + "epoch": 2.33, + "learning_rate": 1.6635576688378626e-05, + "loss": 0.0584, + "step": 17771 + }, + { + "epoch": 2.33, + "learning_rate": 1.6635179929535842e-05, + "loss": 0.0689, + "step": 17772 + }, + { + "epoch": 2.33, + "learning_rate": 1.6634783152031967e-05, + "loss": 0.0763, + "step": 17773 + }, + { + "epoch": 2.33, + "learning_rate": 1.663438635586812e-05, + "loss": 0.06, + "step": 17774 + }, + { + "epoch": 2.33, + "learning_rate": 1.6633989541045412e-05, + "loss": 0.0614, + "step": 17775 + }, + { + "epoch": 2.33, + "learning_rate": 1.6633592707564962e-05, + "loss": 0.0644, + "step": 17776 + }, + { + "epoch": 2.33, + "learning_rate": 1.663319585542789e-05, + "loss": 0.0522, + "step": 17777 + }, + { + "epoch": 2.33, + "learning_rate": 1.6632798984635302e-05, + "loss": 0.0618, + "step": 17778 + }, + { + "epoch": 2.33, + "learning_rate": 1.6632402095188325e-05, + "loss": 0.0625, + "step": 17779 + }, + { + "epoch": 2.33, + "learning_rate": 1.663200518708807e-05, + "loss": 0.0589, + "step": 17780 + }, + { + "epoch": 2.33, + "learning_rate": 1.6631608260335648e-05, + "loss": 0.0566, + "step": 17781 + }, + { + "epoch": 2.33, + "learning_rate": 1.6631211314932184e-05, + "loss": 0.0676, + "step": 17782 + }, + { + "epoch": 2.33, + "learning_rate": 1.6630814350878792e-05, + "loss": 0.0601, + "step": 17783 + }, + { + "epoch": 2.33, + "learning_rate": 1.663041736817659e-05, + "loss": 0.0608, + "step": 17784 + }, + { + "epoch": 2.33, + "learning_rate": 1.6630020366826693e-05, + "loss": 0.0567, + "step": 17785 + }, + { + "epoch": 2.33, + "learning_rate": 1.6629623346830214e-05, + "loss": 0.0668, + "step": 17786 + }, + { + "epoch": 2.33, + "learning_rate": 1.6629226308188267e-05, + "loss": 0.0657, + "step": 17787 + }, + { + "epoch": 2.33, + "learning_rate": 1.6628829250901982e-05, + "loss": 0.0669, + "step": 17788 + }, + { + "epoch": 2.33, + "learning_rate": 1.6628432174972466e-05, + "loss": 0.0634, + "step": 17789 + }, + { + "epoch": 2.33, + "learning_rate": 1.6628035080400835e-05, + "loss": 0.0569, + "step": 17790 + }, + { + "epoch": 2.33, + "learning_rate": 1.6627637967188207e-05, + "loss": 0.0599, + "step": 17791 + }, + { + "epoch": 2.33, + "learning_rate": 1.6627240835335702e-05, + "loss": 0.0682, + "step": 17792 + }, + { + "epoch": 2.33, + "learning_rate": 1.662684368484444e-05, + "loss": 0.0597, + "step": 17793 + }, + { + "epoch": 2.33, + "learning_rate": 1.6626446515715525e-05, + "loss": 0.0603, + "step": 17794 + }, + { + "epoch": 2.33, + "learning_rate": 1.662604932795009e-05, + "loss": 0.069, + "step": 17795 + }, + { + "epoch": 2.33, + "learning_rate": 1.662565212154924e-05, + "loss": 0.0636, + "step": 17796 + }, + { + "epoch": 2.33, + "learning_rate": 1.6625254896514097e-05, + "loss": 0.0609, + "step": 17797 + }, + { + "epoch": 2.33, + "learning_rate": 1.6624857652845775e-05, + "loss": 0.0615, + "step": 17798 + }, + { + "epoch": 2.33, + "learning_rate": 1.6624460390545395e-05, + "loss": 0.0748, + "step": 17799 + }, + { + "epoch": 2.33, + "learning_rate": 1.6624063109614074e-05, + "loss": 0.063, + "step": 17800 + }, + { + "epoch": 2.33, + "learning_rate": 1.6623665810052928e-05, + "loss": 0.0666, + "step": 17801 + }, + { + "epoch": 2.33, + "learning_rate": 1.6623268491863074e-05, + "loss": 0.0706, + "step": 17802 + }, + { + "epoch": 2.33, + "learning_rate": 1.662287115504563e-05, + "loss": 0.0554, + "step": 17803 + }, + { + "epoch": 2.33, + "learning_rate": 1.6622473799601713e-05, + "loss": 0.0574, + "step": 17804 + }, + { + "epoch": 2.33, + "learning_rate": 1.662207642553244e-05, + "loss": 0.0646, + "step": 17805 + }, + { + "epoch": 2.33, + "learning_rate": 1.6621679032838932e-05, + "loss": 0.0606, + "step": 17806 + }, + { + "epoch": 2.33, + "learning_rate": 1.6621281621522308e-05, + "loss": 0.061, + "step": 17807 + }, + { + "epoch": 2.33, + "learning_rate": 1.6620884191583675e-05, + "loss": 0.0634, + "step": 17808 + }, + { + "epoch": 2.33, + "learning_rate": 1.6620486743024162e-05, + "loss": 0.0662, + "step": 17809 + }, + { + "epoch": 2.33, + "learning_rate": 1.6620089275844882e-05, + "loss": 0.0724, + "step": 17810 + }, + { + "epoch": 2.33, + "learning_rate": 1.6619691790046954e-05, + "loss": 0.0723, + "step": 17811 + }, + { + "epoch": 2.33, + "learning_rate": 1.6619294285631492e-05, + "loss": 0.0598, + "step": 17812 + }, + { + "epoch": 2.33, + "learning_rate": 1.6618896762599622e-05, + "loss": 0.0627, + "step": 17813 + }, + { + "epoch": 2.33, + "learning_rate": 1.661849922095245e-05, + "loss": 0.0541, + "step": 17814 + }, + { + "epoch": 2.33, + "learning_rate": 1.6618101660691112e-05, + "loss": 0.0696, + "step": 17815 + }, + { + "epoch": 2.33, + "learning_rate": 1.6617704081816708e-05, + "loss": 0.0572, + "step": 17816 + }, + { + "epoch": 2.33, + "learning_rate": 1.6617306484330367e-05, + "loss": 0.0646, + "step": 17817 + }, + { + "epoch": 2.33, + "learning_rate": 1.66169088682332e-05, + "loss": 0.0684, + "step": 17818 + }, + { + "epoch": 2.33, + "learning_rate": 1.6616511233526332e-05, + "loss": 0.0692, + "step": 17819 + }, + { + "epoch": 2.33, + "learning_rate": 1.6616113580210878e-05, + "loss": 0.0583, + "step": 17820 + }, + { + "epoch": 2.33, + "learning_rate": 1.6615715908287956e-05, + "loss": 0.0631, + "step": 17821 + }, + { + "epoch": 2.33, + "learning_rate": 1.661531821775869e-05, + "loss": 0.0718, + "step": 17822 + }, + { + "epoch": 2.33, + "learning_rate": 1.661492050862419e-05, + "loss": 0.064, + "step": 17823 + }, + { + "epoch": 2.33, + "learning_rate": 1.6614522780885578e-05, + "loss": 0.0641, + "step": 17824 + }, + { + "epoch": 2.33, + "learning_rate": 1.6614125034543975e-05, + "loss": 0.051, + "step": 17825 + }, + { + "epoch": 2.34, + "learning_rate": 1.661372726960049e-05, + "loss": 0.0694, + "step": 17826 + }, + { + "epoch": 2.34, + "learning_rate": 1.661332948605626e-05, + "loss": 0.0637, + "step": 17827 + }, + { + "epoch": 2.34, + "learning_rate": 1.661293168391239e-05, + "loss": 0.0656, + "step": 17828 + }, + { + "epoch": 2.34, + "learning_rate": 1.6612533863169996e-05, + "loss": 0.0672, + "step": 17829 + }, + { + "epoch": 2.34, + "learning_rate": 1.6612136023830206e-05, + "loss": 0.0636, + "step": 17830 + }, + { + "epoch": 2.34, + "learning_rate": 1.6611738165894136e-05, + "loss": 0.0659, + "step": 17831 + }, + { + "epoch": 2.34, + "learning_rate": 1.661134028936291e-05, + "loss": 0.0663, + "step": 17832 + }, + { + "epoch": 2.34, + "learning_rate": 1.6610942394237633e-05, + "loss": 0.0615, + "step": 17833 + }, + { + "epoch": 2.34, + "learning_rate": 1.6610544480519436e-05, + "loss": 0.0736, + "step": 17834 + }, + { + "epoch": 2.34, + "learning_rate": 1.6610146548209435e-05, + "loss": 0.0534, + "step": 17835 + }, + { + "epoch": 2.34, + "learning_rate": 1.6609748597308745e-05, + "loss": 0.0636, + "step": 17836 + }, + { + "epoch": 2.34, + "learning_rate": 1.6609350627818493e-05, + "loss": 0.0598, + "step": 17837 + }, + { + "epoch": 2.34, + "learning_rate": 1.6608952639739796e-05, + "loss": 0.065, + "step": 17838 + }, + { + "epoch": 2.34, + "learning_rate": 1.660855463307377e-05, + "loss": 0.0575, + "step": 17839 + }, + { + "epoch": 2.34, + "learning_rate": 1.6608156607821537e-05, + "loss": 0.07, + "step": 17840 + }, + { + "epoch": 2.34, + "learning_rate": 1.660775856398421e-05, + "loss": 0.0531, + "step": 17841 + }, + { + "epoch": 2.34, + "learning_rate": 1.6607360501562916e-05, + "loss": 0.0556, + "step": 17842 + }, + { + "epoch": 2.34, + "learning_rate": 1.6606962420558775e-05, + "loss": 0.0678, + "step": 17843 + }, + { + "epoch": 2.34, + "learning_rate": 1.6606564320972906e-05, + "loss": 0.0704, + "step": 17844 + }, + { + "epoch": 2.34, + "learning_rate": 1.6606166202806424e-05, + "loss": 0.064, + "step": 17845 + }, + { + "epoch": 2.34, + "learning_rate": 1.660576806606045e-05, + "loss": 0.054, + "step": 17846 + }, + { + "epoch": 2.34, + "learning_rate": 1.6605369910736107e-05, + "loss": 0.0659, + "step": 17847 + }, + { + "epoch": 2.34, + "learning_rate": 1.660497173683451e-05, + "loss": 0.0565, + "step": 17848 + }, + { + "epoch": 2.34, + "learning_rate": 1.6604573544356785e-05, + "loss": 0.0573, + "step": 17849 + }, + { + "epoch": 2.34, + "learning_rate": 1.660417533330405e-05, + "loss": 0.0592, + "step": 17850 + }, + { + "epoch": 2.34, + "learning_rate": 1.660377710367742e-05, + "loss": 0.0618, + "step": 17851 + }, + { + "epoch": 2.34, + "learning_rate": 1.660337885547802e-05, + "loss": 0.0618, + "step": 17852 + }, + { + "epoch": 2.34, + "learning_rate": 1.660298058870697e-05, + "loss": 0.0675, + "step": 17853 + }, + { + "epoch": 2.34, + "learning_rate": 1.6602582303365385e-05, + "loss": 0.0534, + "step": 17854 + }, + { + "epoch": 2.34, + "learning_rate": 1.660218399945439e-05, + "loss": 0.0734, + "step": 17855 + }, + { + "epoch": 2.34, + "learning_rate": 1.6601785676975108e-05, + "loss": 0.0599, + "step": 17856 + }, + { + "epoch": 2.34, + "learning_rate": 1.660138733592865e-05, + "loss": 0.0725, + "step": 17857 + }, + { + "epoch": 2.34, + "learning_rate": 1.6600988976316145e-05, + "loss": 0.0566, + "step": 17858 + }, + { + "epoch": 2.34, + "learning_rate": 1.6600590598138707e-05, + "loss": 0.0703, + "step": 17859 + }, + { + "epoch": 2.34, + "learning_rate": 1.6600192201397464e-05, + "loss": 0.0681, + "step": 17860 + }, + { + "epoch": 2.34, + "learning_rate": 1.659979378609353e-05, + "loss": 0.0631, + "step": 17861 + }, + { + "epoch": 2.34, + "learning_rate": 1.6599395352228028e-05, + "loss": 0.0627, + "step": 17862 + }, + { + "epoch": 2.34, + "learning_rate": 1.6598996899802074e-05, + "loss": 0.0647, + "step": 17863 + }, + { + "epoch": 2.34, + "learning_rate": 1.6598598428816798e-05, + "loss": 0.0565, + "step": 17864 + }, + { + "epoch": 2.34, + "learning_rate": 1.659819993927331e-05, + "loss": 0.073, + "step": 17865 + }, + { + "epoch": 2.34, + "learning_rate": 1.659780143117274e-05, + "loss": 0.0528, + "step": 17866 + }, + { + "epoch": 2.34, + "learning_rate": 1.6597402904516205e-05, + "loss": 0.0515, + "step": 17867 + }, + { + "epoch": 2.34, + "learning_rate": 1.6597004359304825e-05, + "loss": 0.0573, + "step": 17868 + }, + { + "epoch": 2.34, + "learning_rate": 1.6596605795539717e-05, + "loss": 0.0646, + "step": 17869 + }, + { + "epoch": 2.34, + "learning_rate": 1.659620721322201e-05, + "loss": 0.0551, + "step": 17870 + }, + { + "epoch": 2.34, + "learning_rate": 1.6595808612352823e-05, + "loss": 0.0694, + "step": 17871 + }, + { + "epoch": 2.34, + "learning_rate": 1.6595409992933276e-05, + "loss": 0.0596, + "step": 17872 + }, + { + "epoch": 2.34, + "learning_rate": 1.6595011354964487e-05, + "loss": 0.0532, + "step": 17873 + }, + { + "epoch": 2.34, + "learning_rate": 1.659461269844758e-05, + "loss": 0.0667, + "step": 17874 + }, + { + "epoch": 2.34, + "learning_rate": 1.6594214023383675e-05, + "loss": 0.0566, + "step": 17875 + }, + { + "epoch": 2.34, + "learning_rate": 1.65938153297739e-05, + "loss": 0.0673, + "step": 17876 + }, + { + "epoch": 2.34, + "learning_rate": 1.6593416617619366e-05, + "loss": 0.0604, + "step": 17877 + }, + { + "epoch": 2.34, + "learning_rate": 1.6593017886921198e-05, + "loss": 0.0571, + "step": 17878 + }, + { + "epoch": 2.34, + "learning_rate": 1.6592619137680523e-05, + "loss": 0.0571, + "step": 17879 + }, + { + "epoch": 2.34, + "learning_rate": 1.659222036989845e-05, + "loss": 0.0522, + "step": 17880 + }, + { + "epoch": 2.34, + "learning_rate": 1.6591821583576116e-05, + "loss": 0.0605, + "step": 17881 + }, + { + "epoch": 2.34, + "learning_rate": 1.659142277871463e-05, + "loss": 0.0659, + "step": 17882 + }, + { + "epoch": 2.34, + "learning_rate": 1.659102395531512e-05, + "loss": 0.076, + "step": 17883 + }, + { + "epoch": 2.34, + "learning_rate": 1.6590625113378706e-05, + "loss": 0.0619, + "step": 17884 + }, + { + "epoch": 2.34, + "learning_rate": 1.6590226252906508e-05, + "loss": 0.0619, + "step": 17885 + }, + { + "epoch": 2.34, + "learning_rate": 1.658982737389965e-05, + "loss": 0.0596, + "step": 17886 + }, + { + "epoch": 2.34, + "learning_rate": 1.658942847635926e-05, + "loss": 0.0547, + "step": 17887 + }, + { + "epoch": 2.34, + "learning_rate": 1.6589029560286446e-05, + "loss": 0.0593, + "step": 17888 + }, + { + "epoch": 2.34, + "learning_rate": 1.658863062568234e-05, + "loss": 0.0503, + "step": 17889 + }, + { + "epoch": 2.34, + "learning_rate": 1.658823167254806e-05, + "loss": 0.0583, + "step": 17890 + }, + { + "epoch": 2.34, + "learning_rate": 1.6587832700884732e-05, + "loss": 0.0699, + "step": 17891 + }, + { + "epoch": 2.34, + "learning_rate": 1.6587433710693474e-05, + "loss": 0.056, + "step": 17892 + }, + { + "epoch": 2.34, + "learning_rate": 1.658703470197541e-05, + "loss": 0.0608, + "step": 17893 + }, + { + "epoch": 2.34, + "learning_rate": 1.658663567473166e-05, + "loss": 0.069, + "step": 17894 + }, + { + "epoch": 2.34, + "learning_rate": 1.658623662896335e-05, + "loss": 0.0696, + "step": 17895 + }, + { + "epoch": 2.34, + "learning_rate": 1.6585837564671595e-05, + "loss": 0.0634, + "step": 17896 + }, + { + "epoch": 2.34, + "learning_rate": 1.6585438481857527e-05, + "loss": 0.0613, + "step": 17897 + }, + { + "epoch": 2.34, + "learning_rate": 1.658503938052226e-05, + "loss": 0.0677, + "step": 17898 + }, + { + "epoch": 2.34, + "learning_rate": 1.6584640260666925e-05, + "loss": 0.0546, + "step": 17899 + }, + { + "epoch": 2.34, + "learning_rate": 1.658424112229264e-05, + "loss": 0.0652, + "step": 17900 + }, + { + "epoch": 2.34, + "learning_rate": 1.6583841965400527e-05, + "loss": 0.0713, + "step": 17901 + }, + { + "epoch": 2.35, + "learning_rate": 1.6583442789991705e-05, + "loss": 0.0605, + "step": 17902 + }, + { + "epoch": 2.35, + "learning_rate": 1.6583043596067305e-05, + "loss": 0.0638, + "step": 17903 + }, + { + "epoch": 2.35, + "learning_rate": 1.6582644383628445e-05, + "loss": 0.0647, + "step": 17904 + }, + { + "epoch": 2.35, + "learning_rate": 1.6582245152676247e-05, + "loss": 0.0612, + "step": 17905 + }, + { + "epoch": 2.35, + "learning_rate": 1.6581845903211836e-05, + "loss": 0.0605, + "step": 17906 + }, + { + "epoch": 2.35, + "learning_rate": 1.6581446635236332e-05, + "loss": 0.0745, + "step": 17907 + }, + { + "epoch": 2.35, + "learning_rate": 1.6581047348750865e-05, + "loss": 0.0568, + "step": 17908 + }, + { + "epoch": 2.35, + "learning_rate": 1.6580648043756546e-05, + "loss": 0.0585, + "step": 17909 + }, + { + "epoch": 2.35, + "learning_rate": 1.658024872025451e-05, + "loss": 0.0636, + "step": 17910 + }, + { + "epoch": 2.35, + "learning_rate": 1.6579849378245874e-05, + "loss": 0.0668, + "step": 17911 + }, + { + "epoch": 2.35, + "learning_rate": 1.6579450017731757e-05, + "loss": 0.0648, + "step": 17912 + }, + { + "epoch": 2.35, + "learning_rate": 1.6579050638713293e-05, + "loss": 0.0673, + "step": 17913 + }, + { + "epoch": 2.35, + "learning_rate": 1.6578651241191595e-05, + "loss": 0.0606, + "step": 17914 + }, + { + "epoch": 2.35, + "learning_rate": 1.6578251825167797e-05, + "loss": 0.0624, + "step": 17915 + }, + { + "epoch": 2.35, + "learning_rate": 1.657785239064301e-05, + "loss": 0.066, + "step": 17916 + }, + { + "epoch": 2.35, + "learning_rate": 1.6577452937618368e-05, + "loss": 0.0645, + "step": 17917 + }, + { + "epoch": 2.35, + "learning_rate": 1.6577053466094986e-05, + "loss": 0.0688, + "step": 17918 + }, + { + "epoch": 2.35, + "learning_rate": 1.6576653976073993e-05, + "loss": 0.0661, + "step": 17919 + }, + { + "epoch": 2.35, + "learning_rate": 1.6576254467556512e-05, + "loss": 0.0644, + "step": 17920 + }, + { + "epoch": 2.35, + "learning_rate": 1.6575854940543666e-05, + "loss": 0.0591, + "step": 17921 + }, + { + "epoch": 2.35, + "learning_rate": 1.6575455395036574e-05, + "loss": 0.0641, + "step": 17922 + }, + { + "epoch": 2.35, + "learning_rate": 1.6575055831036368e-05, + "loss": 0.0552, + "step": 17923 + }, + { + "epoch": 2.35, + "learning_rate": 1.6574656248544165e-05, + "loss": 0.0546, + "step": 17924 + }, + { + "epoch": 2.35, + "learning_rate": 1.6574256647561094e-05, + "loss": 0.0606, + "step": 17925 + }, + { + "epoch": 2.35, + "learning_rate": 1.6573857028088273e-05, + "loss": 0.0577, + "step": 17926 + }, + { + "epoch": 2.35, + "learning_rate": 1.657345739012683e-05, + "loss": 0.0688, + "step": 17927 + }, + { + "epoch": 2.35, + "learning_rate": 1.657305773367789e-05, + "loss": 0.0574, + "step": 17928 + }, + { + "epoch": 2.35, + "learning_rate": 1.6572658058742575e-05, + "loss": 0.0716, + "step": 17929 + }, + { + "epoch": 2.35, + "learning_rate": 1.657225836532201e-05, + "loss": 0.067, + "step": 17930 + }, + { + "epoch": 2.35, + "learning_rate": 1.6571858653417314e-05, + "loss": 0.0591, + "step": 17931 + }, + { + "epoch": 2.35, + "learning_rate": 1.6571458923029618e-05, + "loss": 0.056, + "step": 17932 + }, + { + "epoch": 2.35, + "learning_rate": 1.6571059174160046e-05, + "loss": 0.0741, + "step": 17933 + }, + { + "epoch": 2.35, + "learning_rate": 1.6570659406809716e-05, + "loss": 0.0671, + "step": 17934 + }, + { + "epoch": 2.35, + "learning_rate": 1.657025962097976e-05, + "loss": 0.0577, + "step": 17935 + }, + { + "epoch": 2.35, + "learning_rate": 1.6569859816671295e-05, + "loss": 0.0559, + "step": 17936 + }, + { + "epoch": 2.35, + "learning_rate": 1.6569459993885454e-05, + "loss": 0.0671, + "step": 17937 + }, + { + "epoch": 2.35, + "learning_rate": 1.6569060152623354e-05, + "loss": 0.0473, + "step": 17938 + }, + { + "epoch": 2.35, + "learning_rate": 1.656866029288612e-05, + "loss": 0.0637, + "step": 17939 + }, + { + "epoch": 2.35, + "learning_rate": 1.6568260414674882e-05, + "loss": 0.0587, + "step": 17940 + }, + { + "epoch": 2.35, + "learning_rate": 1.656786051799076e-05, + "loss": 0.0741, + "step": 17941 + }, + { + "epoch": 2.35, + "learning_rate": 1.656746060283488e-05, + "loss": 0.068, + "step": 17942 + }, + { + "epoch": 2.35, + "learning_rate": 1.6567060669208368e-05, + "loss": 0.0724, + "step": 17943 + }, + { + "epoch": 2.35, + "learning_rate": 1.6566660717112343e-05, + "loss": 0.063, + "step": 17944 + }, + { + "epoch": 2.35, + "learning_rate": 1.6566260746547936e-05, + "loss": 0.0719, + "step": 17945 + }, + { + "epoch": 2.35, + "learning_rate": 1.6565860757516277e-05, + "loss": 0.0707, + "step": 17946 + }, + { + "epoch": 2.35, + "learning_rate": 1.656546075001848e-05, + "loss": 0.0581, + "step": 17947 + }, + { + "epoch": 2.35, + "learning_rate": 1.656506072405567e-05, + "loss": 0.0625, + "step": 17948 + }, + { + "epoch": 2.35, + "learning_rate": 1.6564660679628982e-05, + "loss": 0.0616, + "step": 17949 + }, + { + "epoch": 2.35, + "learning_rate": 1.6564260616739533e-05, + "loss": 0.0663, + "step": 17950 + }, + { + "epoch": 2.35, + "learning_rate": 1.656386053538845e-05, + "loss": 0.0696, + "step": 17951 + }, + { + "epoch": 2.35, + "learning_rate": 1.6563460435576863e-05, + "loss": 0.0532, + "step": 17952 + }, + { + "epoch": 2.35, + "learning_rate": 1.656306031730589e-05, + "loss": 0.054, + "step": 17953 + }, + { + "epoch": 2.35, + "learning_rate": 1.6562660180576657e-05, + "loss": 0.0685, + "step": 17954 + }, + { + "epoch": 2.35, + "learning_rate": 1.6562260025390297e-05, + "loss": 0.0681, + "step": 17955 + }, + { + "epoch": 2.35, + "learning_rate": 1.6561859851747925e-05, + "loss": 0.067, + "step": 17956 + }, + { + "epoch": 2.35, + "learning_rate": 1.6561459659650674e-05, + "loss": 0.0686, + "step": 17957 + }, + { + "epoch": 2.35, + "learning_rate": 1.6561059449099668e-05, + "loss": 0.06, + "step": 17958 + }, + { + "epoch": 2.35, + "learning_rate": 1.656065922009603e-05, + "loss": 0.0526, + "step": 17959 + }, + { + "epoch": 2.35, + "learning_rate": 1.656025897264089e-05, + "loss": 0.0609, + "step": 17960 + }, + { + "epoch": 2.35, + "learning_rate": 1.655985870673537e-05, + "loss": 0.0537, + "step": 17961 + }, + { + "epoch": 2.35, + "learning_rate": 1.6559458422380594e-05, + "loss": 0.0669, + "step": 17962 + }, + { + "epoch": 2.35, + "learning_rate": 1.6559058119577688e-05, + "loss": 0.0641, + "step": 17963 + }, + { + "epoch": 2.35, + "learning_rate": 1.6558657798327787e-05, + "loss": 0.0672, + "step": 17964 + }, + { + "epoch": 2.35, + "learning_rate": 1.655825745863201e-05, + "loss": 0.0625, + "step": 17965 + }, + { + "epoch": 2.35, + "learning_rate": 1.655785710049148e-05, + "loss": 0.0583, + "step": 17966 + }, + { + "epoch": 2.35, + "learning_rate": 1.6557456723907325e-05, + "loss": 0.0669, + "step": 17967 + }, + { + "epoch": 2.35, + "learning_rate": 1.6557056328880675e-05, + "loss": 0.0619, + "step": 17968 + }, + { + "epoch": 2.35, + "learning_rate": 1.655665591541265e-05, + "loss": 0.0639, + "step": 17969 + }, + { + "epoch": 2.35, + "learning_rate": 1.6556255483504383e-05, + "loss": 0.0674, + "step": 17970 + }, + { + "epoch": 2.35, + "learning_rate": 1.6555855033157e-05, + "loss": 0.0727, + "step": 17971 + }, + { + "epoch": 2.35, + "learning_rate": 1.6555454564371617e-05, + "loss": 0.0652, + "step": 17972 + }, + { + "epoch": 2.35, + "learning_rate": 1.6555054077149368e-05, + "loss": 0.0675, + "step": 17973 + }, + { + "epoch": 2.35, + "learning_rate": 1.655465357149138e-05, + "loss": 0.0666, + "step": 17974 + }, + { + "epoch": 2.35, + "learning_rate": 1.6554253047398778e-05, + "loss": 0.0636, + "step": 17975 + }, + { + "epoch": 2.35, + "learning_rate": 1.6553852504872687e-05, + "loss": 0.0635, + "step": 17976 + }, + { + "epoch": 2.35, + "learning_rate": 1.6553451943914237e-05, + "loss": 0.0573, + "step": 17977 + }, + { + "epoch": 2.35, + "learning_rate": 1.655305136452455e-05, + "loss": 0.0589, + "step": 17978 + }, + { + "epoch": 2.36, + "learning_rate": 1.6552650766704755e-05, + "loss": 0.0615, + "step": 17979 + }, + { + "epoch": 2.36, + "learning_rate": 1.6552250150455983e-05, + "loss": 0.0666, + "step": 17980 + }, + { + "epoch": 2.36, + "learning_rate": 1.655184951577935e-05, + "loss": 0.0575, + "step": 17981 + }, + { + "epoch": 2.36, + "learning_rate": 1.6551448862675992e-05, + "loss": 0.0579, + "step": 17982 + }, + { + "epoch": 2.36, + "learning_rate": 1.6551048191147034e-05, + "loss": 0.0649, + "step": 17983 + }, + { + "epoch": 2.36, + "learning_rate": 1.6550647501193605e-05, + "loss": 0.0685, + "step": 17984 + }, + { + "epoch": 2.36, + "learning_rate": 1.6550246792816823e-05, + "loss": 0.0712, + "step": 17985 + }, + { + "epoch": 2.36, + "learning_rate": 1.6549846066017824e-05, + "loss": 0.0588, + "step": 17986 + }, + { + "epoch": 2.36, + "learning_rate": 1.654944532079773e-05, + "loss": 0.0587, + "step": 17987 + }, + { + "epoch": 2.36, + "learning_rate": 1.6549044557157674e-05, + "loss": 0.0664, + "step": 17988 + }, + { + "epoch": 2.36, + "learning_rate": 1.6548643775098773e-05, + "loss": 0.0683, + "step": 17989 + }, + { + "epoch": 2.36, + "learning_rate": 1.654824297462216e-05, + "loss": 0.0608, + "step": 17990 + }, + { + "epoch": 2.36, + "learning_rate": 1.6547842155728965e-05, + "loss": 0.0517, + "step": 17991 + }, + { + "epoch": 2.36, + "learning_rate": 1.654744131842031e-05, + "loss": 0.0666, + "step": 17992 + }, + { + "epoch": 2.36, + "learning_rate": 1.654704046269733e-05, + "loss": 0.0564, + "step": 17993 + }, + { + "epoch": 2.36, + "learning_rate": 1.6546639588561143e-05, + "loss": 0.0574, + "step": 17994 + }, + { + "epoch": 2.36, + "learning_rate": 1.6546238696012883e-05, + "loss": 0.069, + "step": 17995 + }, + { + "epoch": 2.36, + "learning_rate": 1.6545837785053674e-05, + "loss": 0.0536, + "step": 17996 + }, + { + "epoch": 2.36, + "learning_rate": 1.6545436855684648e-05, + "loss": 0.0624, + "step": 17997 + }, + { + "epoch": 2.36, + "learning_rate": 1.6545035907906926e-05, + "loss": 0.0621, + "step": 17998 + }, + { + "epoch": 2.36, + "learning_rate": 1.654463494172164e-05, + "loss": 0.065, + "step": 17999 + }, + { + "epoch": 2.36, + "learning_rate": 1.6544233957129917e-05, + "loss": 0.0592, + "step": 18000 + }, + { + "epoch": 2.36, + "learning_rate": 1.6543832954132884e-05, + "loss": 0.0555, + "step": 18001 + }, + { + "epoch": 2.36, + "learning_rate": 1.654343193273167e-05, + "loss": 0.0744, + "step": 18002 + }, + { + "epoch": 2.36, + "learning_rate": 1.65430308929274e-05, + "loss": 0.0637, + "step": 18003 + }, + { + "epoch": 2.36, + "learning_rate": 1.6542629834721207e-05, + "loss": 0.0681, + "step": 18004 + }, + { + "epoch": 2.36, + "learning_rate": 1.6542228758114216e-05, + "loss": 0.0546, + "step": 18005 + }, + { + "epoch": 2.36, + "learning_rate": 1.654182766310755e-05, + "loss": 0.0722, + "step": 18006 + }, + { + "epoch": 2.36, + "learning_rate": 1.6541426549702348e-05, + "loss": 0.0615, + "step": 18007 + }, + { + "epoch": 2.36, + "learning_rate": 1.654102541789973e-05, + "loss": 0.0641, + "step": 18008 + }, + { + "epoch": 2.36, + "learning_rate": 1.6540624267700827e-05, + "loss": 0.0643, + "step": 18009 + }, + { + "epoch": 2.36, + "learning_rate": 1.6540223099106766e-05, + "loss": 0.0677, + "step": 18010 + }, + { + "epoch": 2.36, + "learning_rate": 1.6539821912118676e-05, + "loss": 0.0623, + "step": 18011 + }, + { + "epoch": 2.36, + "learning_rate": 1.6539420706737685e-05, + "loss": 0.0575, + "step": 18012 + }, + { + "epoch": 2.36, + "learning_rate": 1.653901948296492e-05, + "loss": 0.0554, + "step": 18013 + }, + { + "epoch": 2.36, + "learning_rate": 1.6538618240801513e-05, + "loss": 0.0657, + "step": 18014 + }, + { + "epoch": 2.36, + "learning_rate": 1.653821698024859e-05, + "loss": 0.0575, + "step": 18015 + }, + { + "epoch": 2.36, + "learning_rate": 1.653781570130728e-05, + "loss": 0.0584, + "step": 18016 + }, + { + "epoch": 2.36, + "learning_rate": 1.653741440397871e-05, + "loss": 0.0626, + "step": 18017 + }, + { + "epoch": 2.36, + "learning_rate": 1.653701308826401e-05, + "loss": 0.0623, + "step": 18018 + }, + { + "epoch": 2.36, + "learning_rate": 1.6536611754164313e-05, + "loss": 0.0643, + "step": 18019 + }, + { + "epoch": 2.36, + "learning_rate": 1.653621040168074e-05, + "loss": 0.0586, + "step": 18020 + }, + { + "epoch": 2.36, + "learning_rate": 1.6535809030814425e-05, + "loss": 0.0645, + "step": 18021 + }, + { + "epoch": 2.36, + "learning_rate": 1.653540764156649e-05, + "loss": 0.0585, + "step": 18022 + }, + { + "epoch": 2.36, + "learning_rate": 1.6535006233938076e-05, + "loss": 0.0624, + "step": 18023 + }, + { + "epoch": 2.36, + "learning_rate": 1.6534604807930305e-05, + "loss": 0.065, + "step": 18024 + }, + { + "epoch": 2.36, + "learning_rate": 1.6534203363544298e-05, + "loss": 0.0564, + "step": 18025 + }, + { + "epoch": 2.36, + "learning_rate": 1.65338019007812e-05, + "loss": 0.0631, + "step": 18026 + }, + { + "epoch": 2.36, + "learning_rate": 1.6533400419642128e-05, + "loss": 0.0637, + "step": 18027 + }, + { + "epoch": 2.36, + "learning_rate": 1.6532998920128215e-05, + "loss": 0.062, + "step": 18028 + }, + { + "epoch": 2.36, + "learning_rate": 1.653259740224059e-05, + "loss": 0.0633, + "step": 18029 + }, + { + "epoch": 2.36, + "learning_rate": 1.6532195865980387e-05, + "loss": 0.0544, + "step": 18030 + }, + { + "epoch": 2.36, + "learning_rate": 1.653179431134873e-05, + "loss": 0.0528, + "step": 18031 + }, + { + "epoch": 2.36, + "learning_rate": 1.653139273834675e-05, + "loss": 0.0627, + "step": 18032 + }, + { + "epoch": 2.36, + "learning_rate": 1.653099114697557e-05, + "loss": 0.0609, + "step": 18033 + }, + { + "epoch": 2.36, + "learning_rate": 1.653058953723633e-05, + "loss": 0.0645, + "step": 18034 + }, + { + "epoch": 2.36, + "learning_rate": 1.653018790913015e-05, + "loss": 0.0688, + "step": 18035 + }, + { + "epoch": 2.36, + "learning_rate": 1.6529786262658172e-05, + "loss": 0.066, + "step": 18036 + }, + { + "epoch": 2.36, + "learning_rate": 1.6529384597821513e-05, + "loss": 0.0656, + "step": 18037 + }, + { + "epoch": 2.36, + "learning_rate": 1.6528982914621306e-05, + "loss": 0.068, + "step": 18038 + }, + { + "epoch": 2.36, + "learning_rate": 1.6528581213058684e-05, + "loss": 0.0807, + "step": 18039 + }, + { + "epoch": 2.36, + "learning_rate": 1.6528179493134777e-05, + "loss": 0.0654, + "step": 18040 + }, + { + "epoch": 2.36, + "learning_rate": 1.6527777754850712e-05, + "loss": 0.0617, + "step": 18041 + }, + { + "epoch": 2.36, + "learning_rate": 1.652737599820762e-05, + "loss": 0.0601, + "step": 18042 + }, + { + "epoch": 2.36, + "learning_rate": 1.6526974223206627e-05, + "loss": 0.0676, + "step": 18043 + }, + { + "epoch": 2.36, + "learning_rate": 1.652657242984887e-05, + "loss": 0.068, + "step": 18044 + }, + { + "epoch": 2.36, + "learning_rate": 1.6526170618135475e-05, + "loss": 0.0613, + "step": 18045 + }, + { + "epoch": 2.36, + "learning_rate": 1.652576878806757e-05, + "loss": 0.0538, + "step": 18046 + }, + { + "epoch": 2.36, + "learning_rate": 1.652536693964629e-05, + "loss": 0.0618, + "step": 18047 + }, + { + "epoch": 2.36, + "learning_rate": 1.6524965072872765e-05, + "loss": 0.0659, + "step": 18048 + }, + { + "epoch": 2.36, + "learning_rate": 1.652456318774812e-05, + "loss": 0.0507, + "step": 18049 + }, + { + "epoch": 2.36, + "learning_rate": 1.652416128427349e-05, + "loss": 0.0718, + "step": 18050 + }, + { + "epoch": 2.36, + "learning_rate": 1.6523759362450006e-05, + "loss": 0.0547, + "step": 18051 + }, + { + "epoch": 2.36, + "learning_rate": 1.6523357422278793e-05, + "loss": 0.066, + "step": 18052 + }, + { + "epoch": 2.36, + "learning_rate": 1.6522955463760985e-05, + "loss": 0.0615, + "step": 18053 + }, + { + "epoch": 2.36, + "learning_rate": 1.652255348689771e-05, + "loss": 0.0668, + "step": 18054 + }, + { + "epoch": 2.37, + "learning_rate": 1.6522151491690105e-05, + "loss": 0.0678, + "step": 18055 + }, + { + "epoch": 2.37, + "learning_rate": 1.6521749478139296e-05, + "loss": 0.0646, + "step": 18056 + }, + { + "epoch": 2.37, + "learning_rate": 1.6521347446246412e-05, + "loss": 0.0632, + "step": 18057 + }, + { + "epoch": 2.37, + "learning_rate": 1.6520945396012588e-05, + "loss": 0.0611, + "step": 18058 + }, + { + "epoch": 2.37, + "learning_rate": 1.652054332743895e-05, + "loss": 0.0715, + "step": 18059 + }, + { + "epoch": 2.37, + "learning_rate": 1.6520141240526627e-05, + "loss": 0.0675, + "step": 18060 + }, + { + "epoch": 2.37, + "learning_rate": 1.6519739135276763e-05, + "loss": 0.0594, + "step": 18061 + }, + { + "epoch": 2.37, + "learning_rate": 1.6519337011690475e-05, + "loss": 0.0609, + "step": 18062 + }, + { + "epoch": 2.37, + "learning_rate": 1.65189348697689e-05, + "loss": 0.0662, + "step": 18063 + }, + { + "epoch": 2.37, + "learning_rate": 1.6518532709513166e-05, + "loss": 0.0607, + "step": 18064 + }, + { + "epoch": 2.37, + "learning_rate": 1.6518130530924404e-05, + "loss": 0.0635, + "step": 18065 + }, + { + "epoch": 2.37, + "learning_rate": 1.6517728334003752e-05, + "loss": 0.0626, + "step": 18066 + }, + { + "epoch": 2.37, + "learning_rate": 1.651732611875233e-05, + "loss": 0.0682, + "step": 18067 + }, + { + "epoch": 2.37, + "learning_rate": 1.651692388517128e-05, + "loss": 0.0587, + "step": 18068 + }, + { + "epoch": 2.37, + "learning_rate": 1.651652163326173e-05, + "loss": 0.0665, + "step": 18069 + }, + { + "epoch": 2.37, + "learning_rate": 1.651611936302481e-05, + "loss": 0.0624, + "step": 18070 + }, + { + "epoch": 2.37, + "learning_rate": 1.6515717074461645e-05, + "loss": 0.0701, + "step": 18071 + }, + { + "epoch": 2.37, + "learning_rate": 1.6515314767573382e-05, + "loss": 0.0585, + "step": 18072 + }, + { + "epoch": 2.37, + "learning_rate": 1.6514912442361137e-05, + "loss": 0.0593, + "step": 18073 + }, + { + "epoch": 2.37, + "learning_rate": 1.6514510098826048e-05, + "loss": 0.0681, + "step": 18074 + }, + { + "epoch": 2.37, + "learning_rate": 1.6514107736969246e-05, + "loss": 0.0647, + "step": 18075 + }, + { + "epoch": 2.37, + "learning_rate": 1.6513705356791864e-05, + "loss": 0.0583, + "step": 18076 + }, + { + "epoch": 2.37, + "learning_rate": 1.6513302958295033e-05, + "loss": 0.0635, + "step": 18077 + }, + { + "epoch": 2.37, + "learning_rate": 1.6512900541479885e-05, + "loss": 0.062, + "step": 18078 + }, + { + "epoch": 2.37, + "learning_rate": 1.651249810634755e-05, + "loss": 0.0604, + "step": 18079 + }, + { + "epoch": 2.37, + "learning_rate": 1.651209565289916e-05, + "loss": 0.0627, + "step": 18080 + }, + { + "epoch": 2.37, + "learning_rate": 1.651169318113585e-05, + "loss": 0.0611, + "step": 18081 + }, + { + "epoch": 2.37, + "learning_rate": 1.6511290691058747e-05, + "loss": 0.0701, + "step": 18082 + }, + { + "epoch": 2.37, + "learning_rate": 1.6510888182668986e-05, + "loss": 0.0637, + "step": 18083 + }, + { + "epoch": 2.37, + "learning_rate": 1.65104856559677e-05, + "loss": 0.0587, + "step": 18084 + }, + { + "epoch": 2.37, + "learning_rate": 1.651008311095602e-05, + "loss": 0.068, + "step": 18085 + }, + { + "epoch": 2.37, + "learning_rate": 1.650968054763508e-05, + "loss": 0.0593, + "step": 18086 + }, + { + "epoch": 2.37, + "learning_rate": 1.6509277966006004e-05, + "loss": 0.0558, + "step": 18087 + }, + { + "epoch": 2.37, + "learning_rate": 1.6508875366069936e-05, + "loss": 0.0672, + "step": 18088 + }, + { + "epoch": 2.37, + "learning_rate": 1.6508472747828e-05, + "loss": 0.0739, + "step": 18089 + }, + { + "epoch": 2.37, + "learning_rate": 1.6508070111281332e-05, + "loss": 0.0679, + "step": 18090 + }, + { + "epoch": 2.37, + "learning_rate": 1.650766745643106e-05, + "loss": 0.0595, + "step": 18091 + }, + { + "epoch": 2.37, + "learning_rate": 1.6507264783278324e-05, + "loss": 0.0585, + "step": 18092 + }, + { + "epoch": 2.37, + "learning_rate": 1.650686209182425e-05, + "loss": 0.0647, + "step": 18093 + }, + { + "epoch": 2.37, + "learning_rate": 1.6506459382069976e-05, + "loss": 0.0768, + "step": 18094 + }, + { + "epoch": 2.37, + "learning_rate": 1.650605665401663e-05, + "loss": 0.0548, + "step": 18095 + }, + { + "epoch": 2.37, + "learning_rate": 1.6505653907665343e-05, + "loss": 0.0652, + "step": 18096 + }, + { + "epoch": 2.37, + "learning_rate": 1.6505251143017256e-05, + "loss": 0.0677, + "step": 18097 + }, + { + "epoch": 2.37, + "learning_rate": 1.6504848360073492e-05, + "loss": 0.0673, + "step": 18098 + }, + { + "epoch": 2.37, + "learning_rate": 1.6504445558835188e-05, + "loss": 0.0657, + "step": 18099 + }, + { + "epoch": 2.37, + "learning_rate": 1.6504042739303483e-05, + "loss": 0.0564, + "step": 18100 + }, + { + "epoch": 2.37, + "learning_rate": 1.65036399014795e-05, + "loss": 0.0664, + "step": 18101 + }, + { + "epoch": 2.37, + "learning_rate": 1.6503237045364377e-05, + "loss": 0.0535, + "step": 18102 + }, + { + "epoch": 2.37, + "learning_rate": 1.6502834170959245e-05, + "loss": 0.0627, + "step": 18103 + }, + { + "epoch": 2.37, + "learning_rate": 1.650243127826524e-05, + "loss": 0.0632, + "step": 18104 + }, + { + "epoch": 2.37, + "learning_rate": 1.650202836728349e-05, + "loss": 0.0549, + "step": 18105 + }, + { + "epoch": 2.37, + "learning_rate": 1.6501625438015136e-05, + "loss": 0.0626, + "step": 18106 + }, + { + "epoch": 2.37, + "learning_rate": 1.6501222490461307e-05, + "loss": 0.0573, + "step": 18107 + }, + { + "epoch": 2.37, + "learning_rate": 1.6500819524623132e-05, + "loss": 0.0718, + "step": 18108 + }, + { + "epoch": 2.37, + "learning_rate": 1.650041654050175e-05, + "loss": 0.0627, + "step": 18109 + }, + { + "epoch": 2.37, + "learning_rate": 1.6500013538098293e-05, + "loss": 0.0633, + "step": 18110 + }, + { + "epoch": 2.37, + "learning_rate": 1.6499610517413892e-05, + "loss": 0.0673, + "step": 18111 + }, + { + "epoch": 2.37, + "learning_rate": 1.6499207478449683e-05, + "loss": 0.059, + "step": 18112 + }, + { + "epoch": 2.37, + "learning_rate": 1.64988044212068e-05, + "loss": 0.0718, + "step": 18113 + }, + { + "epoch": 2.37, + "learning_rate": 1.6498401345686374e-05, + "loss": 0.0606, + "step": 18114 + }, + { + "epoch": 2.37, + "learning_rate": 1.6497998251889545e-05, + "loss": 0.065, + "step": 18115 + }, + { + "epoch": 2.37, + "learning_rate": 1.649759513981744e-05, + "loss": 0.0612, + "step": 18116 + }, + { + "epoch": 2.37, + "learning_rate": 1.6497192009471194e-05, + "loss": 0.0549, + "step": 18117 + }, + { + "epoch": 2.37, + "learning_rate": 1.649678886085194e-05, + "loss": 0.0568, + "step": 18118 + }, + { + "epoch": 2.37, + "learning_rate": 1.6496385693960815e-05, + "loss": 0.0639, + "step": 18119 + }, + { + "epoch": 2.37, + "learning_rate": 1.649598250879895e-05, + "loss": 0.0587, + "step": 18120 + }, + { + "epoch": 2.37, + "learning_rate": 1.649557930536748e-05, + "loss": 0.0596, + "step": 18121 + }, + { + "epoch": 2.37, + "learning_rate": 1.649517608366754e-05, + "loss": 0.0585, + "step": 18122 + }, + { + "epoch": 2.37, + "learning_rate": 1.6494772843700266e-05, + "loss": 0.0674, + "step": 18123 + }, + { + "epoch": 2.37, + "learning_rate": 1.6494369585466786e-05, + "loss": 0.0594, + "step": 18124 + }, + { + "epoch": 2.37, + "learning_rate": 1.6493966308968236e-05, + "loss": 0.0694, + "step": 18125 + }, + { + "epoch": 2.37, + "learning_rate": 1.6493563014205755e-05, + "loss": 0.0624, + "step": 18126 + }, + { + "epoch": 2.37, + "learning_rate": 1.649315970118047e-05, + "loss": 0.0523, + "step": 18127 + }, + { + "epoch": 2.37, + "learning_rate": 1.6492756369893524e-05, + "loss": 0.0714, + "step": 18128 + }, + { + "epoch": 2.37, + "learning_rate": 1.6492353020346043e-05, + "loss": 0.0592, + "step": 18129 + }, + { + "epoch": 2.37, + "learning_rate": 1.6491949652539164e-05, + "loss": 0.0574, + "step": 18130 + }, + { + "epoch": 2.38, + "learning_rate": 1.6491546266474023e-05, + "loss": 0.0652, + "step": 18131 + }, + { + "epoch": 2.38, + "learning_rate": 1.6491142862151757e-05, + "loss": 0.0552, + "step": 18132 + }, + { + "epoch": 2.38, + "learning_rate": 1.6490739439573492e-05, + "loss": 0.0662, + "step": 18133 + }, + { + "epoch": 2.38, + "learning_rate": 1.6490335998740373e-05, + "loss": 0.0632, + "step": 18134 + }, + { + "epoch": 2.38, + "learning_rate": 1.6489932539653527e-05, + "loss": 0.067, + "step": 18135 + }, + { + "epoch": 2.38, + "learning_rate": 1.648952906231409e-05, + "loss": 0.0611, + "step": 18136 + }, + { + "epoch": 2.38, + "learning_rate": 1.6489125566723202e-05, + "loss": 0.0679, + "step": 18137 + }, + { + "epoch": 2.38, + "learning_rate": 1.6488722052881995e-05, + "loss": 0.0673, + "step": 18138 + }, + { + "epoch": 2.38, + "learning_rate": 1.6488318520791596e-05, + "loss": 0.0734, + "step": 18139 + }, + { + "epoch": 2.38, + "learning_rate": 1.6487914970453152e-05, + "loss": 0.0605, + "step": 18140 + }, + { + "epoch": 2.38, + "learning_rate": 1.648751140186779e-05, + "loss": 0.0718, + "step": 18141 + }, + { + "epoch": 2.38, + "learning_rate": 1.648710781503665e-05, + "loss": 0.0711, + "step": 18142 + }, + { + "epoch": 2.38, + "learning_rate": 1.6486704209960864e-05, + "loss": 0.0616, + "step": 18143 + }, + { + "epoch": 2.38, + "learning_rate": 1.6486300586641567e-05, + "loss": 0.0623, + "step": 18144 + }, + { + "epoch": 2.38, + "learning_rate": 1.6485896945079894e-05, + "loss": 0.0813, + "step": 18145 + }, + { + "epoch": 2.38, + "learning_rate": 1.6485493285276984e-05, + "loss": 0.0553, + "step": 18146 + }, + { + "epoch": 2.38, + "learning_rate": 1.6485089607233965e-05, + "loss": 0.0649, + "step": 18147 + }, + { + "epoch": 2.38, + "learning_rate": 1.6484685910951982e-05, + "loss": 0.0559, + "step": 18148 + }, + { + "epoch": 2.38, + "learning_rate": 1.6484282196432163e-05, + "loss": 0.0658, + "step": 18149 + }, + { + "epoch": 2.38, + "learning_rate": 1.6483878463675647e-05, + "loss": 0.0659, + "step": 18150 + }, + { + "epoch": 2.38, + "learning_rate": 1.648347471268357e-05, + "loss": 0.0584, + "step": 18151 + }, + { + "epoch": 2.38, + "learning_rate": 1.648307094345706e-05, + "loss": 0.0648, + "step": 18152 + }, + { + "epoch": 2.38, + "learning_rate": 1.6482667155997262e-05, + "loss": 0.0599, + "step": 18153 + }, + { + "epoch": 2.38, + "learning_rate": 1.6482263350305304e-05, + "loss": 0.0704, + "step": 18154 + }, + { + "epoch": 2.38, + "learning_rate": 1.648185952638233e-05, + "loss": 0.0624, + "step": 18155 + }, + { + "epoch": 2.38, + "learning_rate": 1.648145568422947e-05, + "loss": 0.0651, + "step": 18156 + }, + { + "epoch": 2.38, + "learning_rate": 1.648105182384786e-05, + "loss": 0.0601, + "step": 18157 + }, + { + "epoch": 2.38, + "learning_rate": 1.6480647945238636e-05, + "loss": 0.0707, + "step": 18158 + }, + { + "epoch": 2.38, + "learning_rate": 1.6480244048402934e-05, + "loss": 0.0725, + "step": 18159 + }, + { + "epoch": 2.38, + "learning_rate": 1.6479840133341895e-05, + "loss": 0.0627, + "step": 18160 + }, + { + "epoch": 2.38, + "learning_rate": 1.6479436200056648e-05, + "loss": 0.0573, + "step": 18161 + }, + { + "epoch": 2.38, + "learning_rate": 1.6479032248548334e-05, + "loss": 0.0641, + "step": 18162 + }, + { + "epoch": 2.38, + "learning_rate": 1.6478628278818084e-05, + "loss": 0.0644, + "step": 18163 + }, + { + "epoch": 2.38, + "learning_rate": 1.6478224290867038e-05, + "loss": 0.0654, + "step": 18164 + }, + { + "epoch": 2.38, + "learning_rate": 1.6477820284696333e-05, + "loss": 0.063, + "step": 18165 + }, + { + "epoch": 2.38, + "learning_rate": 1.64774162603071e-05, + "loss": 0.0569, + "step": 18166 + }, + { + "epoch": 2.38, + "learning_rate": 1.6477012217700482e-05, + "loss": 0.0595, + "step": 18167 + }, + { + "epoch": 2.38, + "learning_rate": 1.6476608156877613e-05, + "loss": 0.0612, + "step": 18168 + }, + { + "epoch": 2.38, + "learning_rate": 1.647620407783962e-05, + "loss": 0.0577, + "step": 18169 + }, + { + "epoch": 2.38, + "learning_rate": 1.6475799980587654e-05, + "loss": 0.0581, + "step": 18170 + }, + { + "epoch": 2.38, + "learning_rate": 1.6475395865122848e-05, + "loss": 0.0572, + "step": 18171 + }, + { + "epoch": 2.38, + "learning_rate": 1.6474991731446333e-05, + "loss": 0.0551, + "step": 18172 + }, + { + "epoch": 2.38, + "learning_rate": 1.6474587579559246e-05, + "loss": 0.0567, + "step": 18173 + }, + { + "epoch": 2.38, + "learning_rate": 1.6474183409462732e-05, + "loss": 0.0583, + "step": 18174 + }, + { + "epoch": 2.38, + "learning_rate": 1.6473779221157918e-05, + "loss": 0.0534, + "step": 18175 + }, + { + "epoch": 2.38, + "learning_rate": 1.6473375014645944e-05, + "loss": 0.0556, + "step": 18176 + }, + { + "epoch": 2.38, + "learning_rate": 1.6472970789927952e-05, + "loss": 0.0617, + "step": 18177 + }, + { + "epoch": 2.38, + "learning_rate": 1.6472566547005067e-05, + "loss": 0.0636, + "step": 18178 + }, + { + "epoch": 2.38, + "learning_rate": 1.647216228587844e-05, + "loss": 0.0545, + "step": 18179 + }, + { + "epoch": 2.38, + "learning_rate": 1.64717580065492e-05, + "loss": 0.0568, + "step": 18180 + }, + { + "epoch": 2.38, + "learning_rate": 1.6471353709018484e-05, + "loss": 0.0701, + "step": 18181 + }, + { + "epoch": 2.38, + "learning_rate": 1.647094939328743e-05, + "loss": 0.0637, + "step": 18182 + }, + { + "epoch": 2.38, + "learning_rate": 1.6470545059357176e-05, + "loss": 0.0608, + "step": 18183 + }, + { + "epoch": 2.38, + "learning_rate": 1.6470140707228858e-05, + "loss": 0.0636, + "step": 18184 + }, + { + "epoch": 2.38, + "learning_rate": 1.6469736336903615e-05, + "loss": 0.0619, + "step": 18185 + }, + { + "epoch": 2.38, + "learning_rate": 1.6469331948382584e-05, + "loss": 0.076, + "step": 18186 + }, + { + "epoch": 2.38, + "learning_rate": 1.64689275416669e-05, + "loss": 0.0552, + "step": 18187 + }, + { + "epoch": 2.38, + "learning_rate": 1.64685231167577e-05, + "loss": 0.0634, + "step": 18188 + }, + { + "epoch": 2.38, + "learning_rate": 1.6468118673656123e-05, + "loss": 0.0543, + "step": 18189 + }, + { + "epoch": 2.38, + "learning_rate": 1.646771421236331e-05, + "loss": 0.0593, + "step": 18190 + }, + { + "epoch": 2.38, + "learning_rate": 1.6467309732880396e-05, + "loss": 0.07, + "step": 18191 + }, + { + "epoch": 2.38, + "learning_rate": 1.6466905235208514e-05, + "loss": 0.0666, + "step": 18192 + }, + { + "epoch": 2.38, + "learning_rate": 1.6466500719348804e-05, + "loss": 0.0762, + "step": 18193 + }, + { + "epoch": 2.38, + "learning_rate": 1.646609618530241e-05, + "loss": 0.0608, + "step": 18194 + }, + { + "epoch": 2.38, + "learning_rate": 1.6465691633070462e-05, + "loss": 0.0631, + "step": 18195 + }, + { + "epoch": 2.38, + "learning_rate": 1.6465287062654103e-05, + "loss": 0.0563, + "step": 18196 + }, + { + "epoch": 2.38, + "learning_rate": 1.6464882474054467e-05, + "loss": 0.0587, + "step": 18197 + }, + { + "epoch": 2.38, + "learning_rate": 1.6464477867272693e-05, + "loss": 0.0592, + "step": 18198 + }, + { + "epoch": 2.38, + "learning_rate": 1.646407324230992e-05, + "loss": 0.0629, + "step": 18199 + }, + { + "epoch": 2.38, + "learning_rate": 1.6463668599167284e-05, + "loss": 0.0564, + "step": 18200 + }, + { + "epoch": 2.38, + "learning_rate": 1.6463263937845922e-05, + "loss": 0.0683, + "step": 18201 + }, + { + "epoch": 2.38, + "learning_rate": 1.646285925834698e-05, + "loss": 0.072, + "step": 18202 + }, + { + "epoch": 2.38, + "learning_rate": 1.6462454560671587e-05, + "loss": 0.0636, + "step": 18203 + }, + { + "epoch": 2.38, + "learning_rate": 1.6462049844820883e-05, + "loss": 0.0629, + "step": 18204 + }, + { + "epoch": 2.38, + "learning_rate": 1.6461645110796014e-05, + "loss": 0.0612, + "step": 18205 + }, + { + "epoch": 2.38, + "learning_rate": 1.6461240358598107e-05, + "loss": 0.074, + "step": 18206 + }, + { + "epoch": 2.38, + "learning_rate": 1.6460835588228308e-05, + "loss": 0.062, + "step": 18207 + }, + { + "epoch": 2.39, + "learning_rate": 1.646043079968775e-05, + "loss": 0.0734, + "step": 18208 + }, + { + "epoch": 2.39, + "learning_rate": 1.646002599297758e-05, + "loss": 0.0678, + "step": 18209 + }, + { + "epoch": 2.39, + "learning_rate": 1.6459621168098928e-05, + "loss": 0.063, + "step": 18210 + }, + { + "epoch": 2.39, + "learning_rate": 1.6459216325052932e-05, + "loss": 0.0668, + "step": 18211 + }, + { + "epoch": 2.39, + "learning_rate": 1.6458811463840737e-05, + "loss": 0.0621, + "step": 18212 + }, + { + "epoch": 2.39, + "learning_rate": 1.645840658446348e-05, + "loss": 0.0655, + "step": 18213 + }, + { + "epoch": 2.39, + "learning_rate": 1.6458001686922297e-05, + "loss": 0.0601, + "step": 18214 + }, + { + "epoch": 2.39, + "learning_rate": 1.645759677121833e-05, + "loss": 0.0658, + "step": 18215 + }, + { + "epoch": 2.39, + "learning_rate": 1.6457191837352714e-05, + "loss": 0.0545, + "step": 18216 + }, + { + "epoch": 2.39, + "learning_rate": 1.6456786885326588e-05, + "loss": 0.0607, + "step": 18217 + }, + { + "epoch": 2.39, + "learning_rate": 1.6456381915141096e-05, + "loss": 0.0697, + "step": 18218 + }, + { + "epoch": 2.39, + "learning_rate": 1.6455976926797373e-05, + "loss": 0.0685, + "step": 18219 + }, + { + "epoch": 2.39, + "learning_rate": 1.6455571920296557e-05, + "loss": 0.0587, + "step": 18220 + }, + { + "epoch": 2.39, + "learning_rate": 1.6455166895639792e-05, + "loss": 0.0695, + "step": 18221 + }, + { + "epoch": 2.39, + "learning_rate": 1.645476185282821e-05, + "loss": 0.0603, + "step": 18222 + }, + { + "epoch": 2.39, + "learning_rate": 1.6454356791862956e-05, + "loss": 0.0619, + "step": 18223 + }, + { + "epoch": 2.39, + "learning_rate": 1.6453951712745165e-05, + "loss": 0.0728, + "step": 18224 + }, + { + "epoch": 2.39, + "learning_rate": 1.6453546615475983e-05, + "loss": 0.0599, + "step": 18225 + }, + { + "epoch": 2.39, + "learning_rate": 1.645314150005654e-05, + "loss": 0.0534, + "step": 18226 + }, + { + "epoch": 2.39, + "learning_rate": 1.6452736366487986e-05, + "loss": 0.0673, + "step": 18227 + }, + { + "epoch": 2.39, + "learning_rate": 1.645233121477145e-05, + "loss": 0.0658, + "step": 18228 + }, + { + "epoch": 2.39, + "learning_rate": 1.6451926044908076e-05, + "loss": 0.0534, + "step": 18229 + }, + { + "epoch": 2.39, + "learning_rate": 1.6451520856899005e-05, + "loss": 0.0647, + "step": 18230 + }, + { + "epoch": 2.39, + "learning_rate": 1.6451115650745374e-05, + "loss": 0.0597, + "step": 18231 + }, + { + "epoch": 2.39, + "learning_rate": 1.6450710426448328e-05, + "loss": 0.0658, + "step": 18232 + }, + { + "epoch": 2.39, + "learning_rate": 1.6450305184008998e-05, + "loss": 0.0643, + "step": 18233 + }, + { + "epoch": 2.39, + "learning_rate": 1.6449899923428526e-05, + "loss": 0.0668, + "step": 18234 + }, + { + "epoch": 2.39, + "learning_rate": 1.6449494644708058e-05, + "loss": 0.0607, + "step": 18235 + }, + { + "epoch": 2.39, + "learning_rate": 1.644908934784873e-05, + "loss": 0.0711, + "step": 18236 + }, + { + "epoch": 2.39, + "learning_rate": 1.6448684032851682e-05, + "loss": 0.0693, + "step": 18237 + }, + { + "epoch": 2.39, + "learning_rate": 1.644827869971805e-05, + "loss": 0.0557, + "step": 18238 + }, + { + "epoch": 2.39, + "learning_rate": 1.644787334844898e-05, + "loss": 0.0577, + "step": 18239 + }, + { + "epoch": 2.39, + "learning_rate": 1.644746797904561e-05, + "loss": 0.0641, + "step": 18240 + }, + { + "epoch": 2.39, + "learning_rate": 1.644706259150908e-05, + "loss": 0.0562, + "step": 18241 + }, + { + "epoch": 2.39, + "learning_rate": 1.644665718584053e-05, + "loss": 0.0674, + "step": 18242 + }, + { + "epoch": 2.39, + "learning_rate": 1.64462517620411e-05, + "loss": 0.0583, + "step": 18243 + }, + { + "epoch": 2.39, + "learning_rate": 1.6445846320111928e-05, + "loss": 0.0614, + "step": 18244 + }, + { + "epoch": 2.39, + "learning_rate": 1.6445440860054157e-05, + "loss": 0.0667, + "step": 18245 + }, + { + "epoch": 2.39, + "learning_rate": 1.644503538186893e-05, + "loss": 0.0598, + "step": 18246 + }, + { + "epoch": 2.39, + "learning_rate": 1.644462988555738e-05, + "loss": 0.0605, + "step": 18247 + }, + { + "epoch": 2.39, + "learning_rate": 1.6444224371120655e-05, + "loss": 0.0524, + "step": 18248 + }, + { + "epoch": 2.39, + "learning_rate": 1.644381883855989e-05, + "loss": 0.0608, + "step": 18249 + }, + { + "epoch": 2.39, + "learning_rate": 1.6443413287876227e-05, + "loss": 0.0609, + "step": 18250 + }, + { + "epoch": 2.39, + "learning_rate": 1.6443007719070807e-05, + "loss": 0.0617, + "step": 18251 + }, + { + "epoch": 2.39, + "learning_rate": 1.6442602132144773e-05, + "loss": 0.0621, + "step": 18252 + }, + { + "epoch": 2.39, + "learning_rate": 1.6442196527099264e-05, + "loss": 0.0638, + "step": 18253 + }, + { + "epoch": 2.39, + "learning_rate": 1.6441790903935416e-05, + "loss": 0.0619, + "step": 18254 + }, + { + "epoch": 2.39, + "learning_rate": 1.644138526265438e-05, + "loss": 0.0627, + "step": 18255 + }, + { + "epoch": 2.39, + "learning_rate": 1.6440979603257285e-05, + "loss": 0.0604, + "step": 18256 + }, + { + "epoch": 2.39, + "learning_rate": 1.6440573925745282e-05, + "loss": 0.0616, + "step": 18257 + }, + { + "epoch": 2.39, + "learning_rate": 1.644016823011951e-05, + "loss": 0.0634, + "step": 18258 + }, + { + "epoch": 2.39, + "learning_rate": 1.64397625163811e-05, + "loss": 0.0617, + "step": 18259 + }, + { + "epoch": 2.39, + "learning_rate": 1.643935678453121e-05, + "loss": 0.0687, + "step": 18260 + }, + { + "epoch": 2.39, + "learning_rate": 1.6438951034570964e-05, + "loss": 0.0662, + "step": 18261 + }, + { + "epoch": 2.39, + "learning_rate": 1.6438545266501514e-05, + "loss": 0.0684, + "step": 18262 + }, + { + "epoch": 2.39, + "learning_rate": 1.6438139480324e-05, + "loss": 0.0568, + "step": 18263 + }, + { + "epoch": 2.39, + "learning_rate": 1.643773367603956e-05, + "loss": 0.0708, + "step": 18264 + }, + { + "epoch": 2.39, + "learning_rate": 1.6437327853649338e-05, + "loss": 0.0622, + "step": 18265 + }, + { + "epoch": 2.39, + "learning_rate": 1.6436922013154474e-05, + "loss": 0.0695, + "step": 18266 + }, + { + "epoch": 2.39, + "learning_rate": 1.6436516154556108e-05, + "loss": 0.064, + "step": 18267 + }, + { + "epoch": 2.39, + "learning_rate": 1.6436110277855383e-05, + "loss": 0.0604, + "step": 18268 + }, + { + "epoch": 2.39, + "learning_rate": 1.6435704383053446e-05, + "loss": 0.0631, + "step": 18269 + }, + { + "epoch": 2.39, + "learning_rate": 1.6435298470151427e-05, + "loss": 0.0682, + "step": 18270 + }, + { + "epoch": 2.39, + "learning_rate": 1.6434892539150476e-05, + "loss": 0.066, + "step": 18271 + }, + { + "epoch": 2.39, + "learning_rate": 1.6434486590051735e-05, + "loss": 0.0593, + "step": 18272 + }, + { + "epoch": 2.39, + "learning_rate": 1.6434080622856338e-05, + "loss": 0.072, + "step": 18273 + }, + { + "epoch": 2.39, + "learning_rate": 1.6433674637565434e-05, + "loss": 0.0629, + "step": 18274 + }, + { + "epoch": 2.39, + "learning_rate": 1.6433268634180164e-05, + "loss": 0.0675, + "step": 18275 + }, + { + "epoch": 2.39, + "learning_rate": 1.6432862612701668e-05, + "loss": 0.058, + "step": 18276 + }, + { + "epoch": 2.39, + "learning_rate": 1.6432456573131087e-05, + "loss": 0.0707, + "step": 18277 + }, + { + "epoch": 2.39, + "learning_rate": 1.6432050515469565e-05, + "loss": 0.0576, + "step": 18278 + }, + { + "epoch": 2.39, + "learning_rate": 1.6431644439718243e-05, + "loss": 0.0648, + "step": 18279 + }, + { + "epoch": 2.39, + "learning_rate": 1.6431238345878264e-05, + "loss": 0.0574, + "step": 18280 + }, + { + "epoch": 2.39, + "learning_rate": 1.643083223395077e-05, + "loss": 0.0579, + "step": 18281 + }, + { + "epoch": 2.39, + "learning_rate": 1.64304261039369e-05, + "loss": 0.0703, + "step": 18282 + }, + { + "epoch": 2.39, + "learning_rate": 1.6430019955837804e-05, + "loss": 0.0644, + "step": 18283 + }, + { + "epoch": 2.4, + "learning_rate": 1.642961378965462e-05, + "loss": 0.0642, + "step": 18284 + }, + { + "epoch": 2.4, + "learning_rate": 1.6429207605388485e-05, + "loss": 0.0657, + "step": 18285 + }, + { + "epoch": 2.4, + "learning_rate": 1.6428801403040547e-05, + "loss": 0.0564, + "step": 18286 + }, + { + "epoch": 2.4, + "learning_rate": 1.6428395182611948e-05, + "loss": 0.051, + "step": 18287 + }, + { + "epoch": 2.4, + "learning_rate": 1.642798894410383e-05, + "loss": 0.061, + "step": 18288 + }, + { + "epoch": 2.4, + "learning_rate": 1.6427582687517337e-05, + "loss": 0.0604, + "step": 18289 + }, + { + "epoch": 2.4, + "learning_rate": 1.6427176412853608e-05, + "loss": 0.0661, + "step": 18290 + }, + { + "epoch": 2.4, + "learning_rate": 1.6426770120113786e-05, + "loss": 0.0616, + "step": 18291 + }, + { + "epoch": 2.4, + "learning_rate": 1.642636380929902e-05, + "loss": 0.0589, + "step": 18292 + }, + { + "epoch": 2.4, + "learning_rate": 1.6425957480410444e-05, + "loss": 0.0656, + "step": 18293 + }, + { + "epoch": 2.4, + "learning_rate": 1.6425551133449207e-05, + "loss": 0.0627, + "step": 18294 + }, + { + "epoch": 2.4, + "learning_rate": 1.6425144768416448e-05, + "loss": 0.0613, + "step": 18295 + }, + { + "epoch": 2.4, + "learning_rate": 1.6424738385313312e-05, + "loss": 0.059, + "step": 18296 + }, + { + "epoch": 2.4, + "learning_rate": 1.6424331984140943e-05, + "loss": 0.0678, + "step": 18297 + }, + { + "epoch": 2.4, + "learning_rate": 1.642392556490048e-05, + "loss": 0.0631, + "step": 18298 + }, + { + "epoch": 2.4, + "learning_rate": 1.6423519127593067e-05, + "loss": 0.0547, + "step": 18299 + }, + { + "epoch": 2.4, + "learning_rate": 1.6423112672219853e-05, + "loss": 0.0661, + "step": 18300 + }, + { + "epoch": 2.4, + "learning_rate": 1.642270619878197e-05, + "loss": 0.0577, + "step": 18301 + }, + { + "epoch": 2.4, + "learning_rate": 1.6422299707280576e-05, + "loss": 0.0587, + "step": 18302 + }, + { + "epoch": 2.4, + "learning_rate": 1.6421893197716803e-05, + "loss": 0.0518, + "step": 18303 + }, + { + "epoch": 2.4, + "learning_rate": 1.6421486670091797e-05, + "loss": 0.0703, + "step": 18304 + }, + { + "epoch": 2.4, + "learning_rate": 1.64210801244067e-05, + "loss": 0.0545, + "step": 18305 + }, + { + "epoch": 2.4, + "learning_rate": 1.6420673560662658e-05, + "loss": 0.0549, + "step": 18306 + }, + { + "epoch": 2.4, + "learning_rate": 1.6420266978860815e-05, + "loss": 0.0539, + "step": 18307 + }, + { + "epoch": 2.4, + "learning_rate": 1.6419860379002313e-05, + "loss": 0.0639, + "step": 18308 + }, + { + "epoch": 2.4, + "learning_rate": 1.6419453761088294e-05, + "loss": 0.0561, + "step": 18309 + }, + { + "epoch": 2.4, + "learning_rate": 1.6419047125119907e-05, + "loss": 0.0725, + "step": 18310 + }, + { + "epoch": 2.4, + "learning_rate": 1.6418640471098287e-05, + "loss": 0.0575, + "step": 18311 + }, + { + "epoch": 2.4, + "learning_rate": 1.6418233799024585e-05, + "loss": 0.0628, + "step": 18312 + }, + { + "epoch": 2.4, + "learning_rate": 1.641782710889994e-05, + "loss": 0.0609, + "step": 18313 + }, + { + "epoch": 2.4, + "learning_rate": 1.64174204007255e-05, + "loss": 0.0535, + "step": 18314 + }, + { + "epoch": 2.4, + "learning_rate": 1.641701367450241e-05, + "loss": 0.073, + "step": 18315 + }, + { + "epoch": 2.4, + "learning_rate": 1.64166069302318e-05, + "loss": 0.0591, + "step": 18316 + }, + { + "epoch": 2.4, + "learning_rate": 1.6416200167914836e-05, + "loss": 0.0608, + "step": 18317 + }, + { + "epoch": 2.4, + "learning_rate": 1.6415793387552644e-05, + "loss": 0.064, + "step": 18318 + }, + { + "epoch": 2.4, + "learning_rate": 1.641538658914638e-05, + "loss": 0.0628, + "step": 18319 + }, + { + "epoch": 2.4, + "learning_rate": 1.641497977269718e-05, + "loss": 0.0709, + "step": 18320 + }, + { + "epoch": 2.4, + "learning_rate": 1.6414572938206192e-05, + "loss": 0.0624, + "step": 18321 + }, + { + "epoch": 2.4, + "learning_rate": 1.6414166085674557e-05, + "loss": 0.0638, + "step": 18322 + }, + { + "epoch": 2.4, + "learning_rate": 1.641375921510342e-05, + "loss": 0.063, + "step": 18323 + }, + { + "epoch": 2.4, + "learning_rate": 1.641335232649393e-05, + "loss": 0.0559, + "step": 18324 + }, + { + "epoch": 2.4, + "learning_rate": 1.6412945419847232e-05, + "loss": 0.0664, + "step": 18325 + }, + { + "epoch": 2.4, + "learning_rate": 1.641253849516446e-05, + "loss": 0.0584, + "step": 18326 + }, + { + "epoch": 2.4, + "learning_rate": 1.6412131552446767e-05, + "loss": 0.0638, + "step": 18327 + }, + { + "epoch": 2.4, + "learning_rate": 1.64117245916953e-05, + "loss": 0.0621, + "step": 18328 + }, + { + "epoch": 2.4, + "learning_rate": 1.641131761291119e-05, + "loss": 0.0676, + "step": 18329 + }, + { + "epoch": 2.4, + "learning_rate": 1.6410910616095597e-05, + "loss": 0.0627, + "step": 18330 + }, + { + "epoch": 2.4, + "learning_rate": 1.6410503601249658e-05, + "loss": 0.0494, + "step": 18331 + }, + { + "epoch": 2.4, + "learning_rate": 1.641009656837452e-05, + "loss": 0.0657, + "step": 18332 + }, + { + "epoch": 2.4, + "learning_rate": 1.6409689517471323e-05, + "loss": 0.0724, + "step": 18333 + }, + { + "epoch": 2.4, + "learning_rate": 1.6409282448541218e-05, + "loss": 0.0623, + "step": 18334 + }, + { + "epoch": 2.4, + "learning_rate": 1.6408875361585348e-05, + "loss": 0.062, + "step": 18335 + }, + { + "epoch": 2.4, + "learning_rate": 1.6408468256604857e-05, + "loss": 0.0627, + "step": 18336 + }, + { + "epoch": 2.4, + "learning_rate": 1.640806113360089e-05, + "loss": 0.0624, + "step": 18337 + }, + { + "epoch": 2.4, + "learning_rate": 1.6407653992574594e-05, + "loss": 0.0657, + "step": 18338 + }, + { + "epoch": 2.4, + "learning_rate": 1.640724683352711e-05, + "loss": 0.0611, + "step": 18339 + }, + { + "epoch": 2.4, + "learning_rate": 1.6406839656459584e-05, + "loss": 0.0487, + "step": 18340 + }, + { + "epoch": 2.4, + "learning_rate": 1.6406432461373168e-05, + "loss": 0.0562, + "step": 18341 + }, + { + "epoch": 2.4, + "learning_rate": 1.6406025248268996e-05, + "loss": 0.0611, + "step": 18342 + }, + { + "epoch": 2.4, + "learning_rate": 1.6405618017148223e-05, + "loss": 0.0677, + "step": 18343 + }, + { + "epoch": 2.4, + "learning_rate": 1.640521076801199e-05, + "loss": 0.057, + "step": 18344 + }, + { + "epoch": 2.4, + "learning_rate": 1.6404803500861438e-05, + "loss": 0.0661, + "step": 18345 + }, + { + "epoch": 2.4, + "learning_rate": 1.6404396215697723e-05, + "loss": 0.058, + "step": 18346 + }, + { + "epoch": 2.4, + "learning_rate": 1.6403988912521985e-05, + "loss": 0.0528, + "step": 18347 + }, + { + "epoch": 2.4, + "learning_rate": 1.6403581591335366e-05, + "loss": 0.068, + "step": 18348 + }, + { + "epoch": 2.4, + "learning_rate": 1.6403174252139016e-05, + "loss": 0.0605, + "step": 18349 + }, + { + "epoch": 2.4, + "learning_rate": 1.640276689493408e-05, + "loss": 0.0597, + "step": 18350 + }, + { + "epoch": 2.4, + "learning_rate": 1.6402359519721702e-05, + "loss": 0.0609, + "step": 18351 + }, + { + "epoch": 2.4, + "learning_rate": 1.640195212650303e-05, + "loss": 0.0596, + "step": 18352 + }, + { + "epoch": 2.4, + "learning_rate": 1.6401544715279208e-05, + "loss": 0.0648, + "step": 18353 + }, + { + "epoch": 2.4, + "learning_rate": 1.6401137286051384e-05, + "loss": 0.0608, + "step": 18354 + }, + { + "epoch": 2.4, + "learning_rate": 1.6400729838820698e-05, + "loss": 0.0626, + "step": 18355 + }, + { + "epoch": 2.4, + "learning_rate": 1.6400322373588303e-05, + "loss": 0.0606, + "step": 18356 + }, + { + "epoch": 2.4, + "learning_rate": 1.6399914890355342e-05, + "loss": 0.0567, + "step": 18357 + }, + { + "epoch": 2.4, + "learning_rate": 1.6399507389122962e-05, + "loss": 0.0691, + "step": 18358 + }, + { + "epoch": 2.4, + "learning_rate": 1.6399099869892307e-05, + "loss": 0.0653, + "step": 18359 + }, + { + "epoch": 2.41, + "learning_rate": 1.6398692332664527e-05, + "loss": 0.0606, + "step": 18360 + }, + { + "epoch": 2.41, + "learning_rate": 1.639828477744076e-05, + "loss": 0.0636, + "step": 18361 + }, + { + "epoch": 2.41, + "learning_rate": 1.6397877204222162e-05, + "loss": 0.0525, + "step": 18362 + }, + { + "epoch": 2.41, + "learning_rate": 1.6397469613009872e-05, + "loss": 0.0666, + "step": 18363 + }, + { + "epoch": 2.41, + "learning_rate": 1.6397062003805043e-05, + "loss": 0.0543, + "step": 18364 + }, + { + "epoch": 2.41, + "learning_rate": 1.6396654376608818e-05, + "loss": 0.0572, + "step": 18365 + }, + { + "epoch": 2.41, + "learning_rate": 1.639624673142234e-05, + "loss": 0.0686, + "step": 18366 + }, + { + "epoch": 2.41, + "learning_rate": 1.639583906824676e-05, + "loss": 0.0618, + "step": 18367 + }, + { + "epoch": 2.41, + "learning_rate": 1.6395431387083225e-05, + "loss": 0.0588, + "step": 18368 + }, + { + "epoch": 2.41, + "learning_rate": 1.639502368793288e-05, + "loss": 0.0624, + "step": 18369 + }, + { + "epoch": 2.41, + "learning_rate": 1.6394615970796866e-05, + "loss": 0.0551, + "step": 18370 + }, + { + "epoch": 2.41, + "learning_rate": 1.6394208235676338e-05, + "loss": 0.0569, + "step": 18371 + }, + { + "epoch": 2.41, + "learning_rate": 1.6393800482572443e-05, + "loss": 0.0544, + "step": 18372 + }, + { + "epoch": 2.41, + "learning_rate": 1.6393392711486323e-05, + "loss": 0.0548, + "step": 18373 + }, + { + "epoch": 2.41, + "learning_rate": 1.6392984922419123e-05, + "loss": 0.0642, + "step": 18374 + }, + { + "epoch": 2.41, + "learning_rate": 1.6392577115371996e-05, + "loss": 0.0562, + "step": 18375 + }, + { + "epoch": 2.41, + "learning_rate": 1.6392169290346086e-05, + "loss": 0.0659, + "step": 18376 + }, + { + "epoch": 2.41, + "learning_rate": 1.6391761447342543e-05, + "loss": 0.066, + "step": 18377 + }, + { + "epoch": 2.41, + "learning_rate": 1.639135358636251e-05, + "loss": 0.0639, + "step": 18378 + }, + { + "epoch": 2.41, + "learning_rate": 1.6390945707407134e-05, + "loss": 0.0554, + "step": 18379 + }, + { + "epoch": 2.41, + "learning_rate": 1.6390537810477566e-05, + "loss": 0.0644, + "step": 18380 + }, + { + "epoch": 2.41, + "learning_rate": 1.6390129895574945e-05, + "loss": 0.062, + "step": 18381 + }, + { + "epoch": 2.41, + "learning_rate": 1.638972196270043e-05, + "loss": 0.0563, + "step": 18382 + }, + { + "epoch": 2.41, + "learning_rate": 1.638931401185516e-05, + "loss": 0.0649, + "step": 18383 + }, + { + "epoch": 2.41, + "learning_rate": 1.6388906043040283e-05, + "loss": 0.0701, + "step": 18384 + }, + { + "epoch": 2.41, + "learning_rate": 1.638849805625695e-05, + "loss": 0.0602, + "step": 18385 + }, + { + "epoch": 2.41, + "learning_rate": 1.6388090051506304e-05, + "loss": 0.0716, + "step": 18386 + }, + { + "epoch": 2.41, + "learning_rate": 1.6387682028789494e-05, + "loss": 0.0604, + "step": 18387 + }, + { + "epoch": 2.41, + "learning_rate": 1.6387273988107675e-05, + "loss": 0.0605, + "step": 18388 + }, + { + "epoch": 2.41, + "learning_rate": 1.6386865929461984e-05, + "loss": 0.0575, + "step": 18389 + }, + { + "epoch": 2.41, + "learning_rate": 1.6386457852853573e-05, + "loss": 0.0671, + "step": 18390 + }, + { + "epoch": 2.41, + "learning_rate": 1.6386049758283592e-05, + "loss": 0.0655, + "step": 18391 + }, + { + "epoch": 2.41, + "learning_rate": 1.6385641645753182e-05, + "loss": 0.0655, + "step": 18392 + }, + { + "epoch": 2.41, + "learning_rate": 1.6385233515263495e-05, + "loss": 0.062, + "step": 18393 + }, + { + "epoch": 2.41, + "learning_rate": 1.638482536681568e-05, + "loss": 0.061, + "step": 18394 + }, + { + "epoch": 2.41, + "learning_rate": 1.6384417200410884e-05, + "loss": 0.0595, + "step": 18395 + }, + { + "epoch": 2.41, + "learning_rate": 1.6384009016050257e-05, + "loss": 0.0548, + "step": 18396 + }, + { + "epoch": 2.41, + "learning_rate": 1.638360081373494e-05, + "loss": 0.0629, + "step": 18397 + }, + { + "epoch": 2.41, + "learning_rate": 1.6383192593466086e-05, + "loss": 0.0601, + "step": 18398 + }, + { + "epoch": 2.41, + "learning_rate": 1.6382784355244844e-05, + "loss": 0.0561, + "step": 18399 + }, + { + "epoch": 2.41, + "learning_rate": 1.6382376099072364e-05, + "loss": 0.0604, + "step": 18400 + }, + { + "epoch": 2.41, + "learning_rate": 1.638196782494979e-05, + "loss": 0.0574, + "step": 18401 + }, + { + "epoch": 2.41, + "learning_rate": 1.638155953287827e-05, + "loss": 0.0641, + "step": 18402 + }, + { + "epoch": 2.41, + "learning_rate": 1.638115122285895e-05, + "loss": 0.0718, + "step": 18403 + }, + { + "epoch": 2.41, + "learning_rate": 1.6380742894892984e-05, + "loss": 0.0632, + "step": 18404 + }, + { + "epoch": 2.41, + "learning_rate": 1.6380334548981524e-05, + "loss": 0.0556, + "step": 18405 + }, + { + "epoch": 2.41, + "learning_rate": 1.6379926185125708e-05, + "loss": 0.0541, + "step": 18406 + }, + { + "epoch": 2.41, + "learning_rate": 1.637951780332669e-05, + "loss": 0.052, + "step": 18407 + }, + { + "epoch": 2.41, + "learning_rate": 1.6379109403585617e-05, + "loss": 0.0585, + "step": 18408 + }, + { + "epoch": 2.41, + "learning_rate": 1.637870098590364e-05, + "loss": 0.0577, + "step": 18409 + }, + { + "epoch": 2.41, + "learning_rate": 1.6378292550281903e-05, + "loss": 0.0616, + "step": 18410 + }, + { + "epoch": 2.41, + "learning_rate": 1.6377884096721565e-05, + "loss": 0.0677, + "step": 18411 + }, + { + "epoch": 2.41, + "learning_rate": 1.637747562522376e-05, + "loss": 0.0677, + "step": 18412 + }, + { + "epoch": 2.41, + "learning_rate": 1.637706713578965e-05, + "loss": 0.0545, + "step": 18413 + }, + { + "epoch": 2.41, + "learning_rate": 1.6376658628420378e-05, + "loss": 0.0579, + "step": 18414 + }, + { + "epoch": 2.41, + "learning_rate": 1.6376250103117087e-05, + "loss": 0.0571, + "step": 18415 + }, + { + "epoch": 2.41, + "learning_rate": 1.6375841559880937e-05, + "loss": 0.0629, + "step": 18416 + }, + { + "epoch": 2.41, + "learning_rate": 1.637543299871307e-05, + "loss": 0.0693, + "step": 18417 + }, + { + "epoch": 2.41, + "learning_rate": 1.637502441961464e-05, + "loss": 0.0659, + "step": 18418 + }, + { + "epoch": 2.41, + "learning_rate": 1.6374615822586792e-05, + "loss": 0.0592, + "step": 18419 + }, + { + "epoch": 2.41, + "learning_rate": 1.6374207207630678e-05, + "loss": 0.0657, + "step": 18420 + }, + { + "epoch": 2.41, + "learning_rate": 1.637379857474744e-05, + "loss": 0.0715, + "step": 18421 + }, + { + "epoch": 2.41, + "learning_rate": 1.6373389923938238e-05, + "loss": 0.0682, + "step": 18422 + }, + { + "epoch": 2.41, + "learning_rate": 1.6372981255204215e-05, + "loss": 0.0554, + "step": 18423 + }, + { + "epoch": 2.41, + "learning_rate": 1.6372572568546522e-05, + "loss": 0.0573, + "step": 18424 + }, + { + "epoch": 2.41, + "learning_rate": 1.637216386396631e-05, + "loss": 0.0576, + "step": 18425 + }, + { + "epoch": 2.41, + "learning_rate": 1.6371755141464722e-05, + "loss": 0.0619, + "step": 18426 + }, + { + "epoch": 2.41, + "learning_rate": 1.6371346401042917e-05, + "loss": 0.0569, + "step": 18427 + }, + { + "epoch": 2.41, + "learning_rate": 1.6370937642702035e-05, + "loss": 0.0696, + "step": 18428 + }, + { + "epoch": 2.41, + "learning_rate": 1.637052886644323e-05, + "loss": 0.063, + "step": 18429 + }, + { + "epoch": 2.41, + "learning_rate": 1.6370120072267657e-05, + "loss": 0.0631, + "step": 18430 + }, + { + "epoch": 2.41, + "learning_rate": 1.6369711260176456e-05, + "loss": 0.0623, + "step": 18431 + }, + { + "epoch": 2.41, + "learning_rate": 1.6369302430170783e-05, + "loss": 0.0768, + "step": 18432 + }, + { + "epoch": 2.41, + "learning_rate": 1.6368893582251786e-05, + "loss": 0.056, + "step": 18433 + }, + { + "epoch": 2.41, + "learning_rate": 1.6368484716420613e-05, + "loss": 0.0645, + "step": 18434 + }, + { + "epoch": 2.41, + "learning_rate": 1.6368075832678417e-05, + "loss": 0.0567, + "step": 18435 + }, + { + "epoch": 2.41, + "learning_rate": 1.6367666931026347e-05, + "loss": 0.0678, + "step": 18436 + }, + { + "epoch": 2.42, + "learning_rate": 1.6367258011465553e-05, + "loss": 0.0629, + "step": 18437 + }, + { + "epoch": 2.42, + "learning_rate": 1.636684907399719e-05, + "loss": 0.0661, + "step": 18438 + }, + { + "epoch": 2.42, + "learning_rate": 1.6366440118622396e-05, + "loss": 0.0672, + "step": 18439 + }, + { + "epoch": 2.42, + "learning_rate": 1.636603114534233e-05, + "loss": 0.0636, + "step": 18440 + }, + { + "epoch": 2.42, + "learning_rate": 1.636562215415814e-05, + "loss": 0.0646, + "step": 18441 + }, + { + "epoch": 2.42, + "learning_rate": 1.6365213145070976e-05, + "loss": 0.0545, + "step": 18442 + }, + { + "epoch": 2.42, + "learning_rate": 1.6364804118081994e-05, + "loss": 0.0654, + "step": 18443 + }, + { + "epoch": 2.42, + "learning_rate": 1.6364395073192336e-05, + "loss": 0.0677, + "step": 18444 + }, + { + "epoch": 2.42, + "learning_rate": 1.6363986010403153e-05, + "loss": 0.0609, + "step": 18445 + }, + { + "epoch": 2.42, + "learning_rate": 1.6363576929715602e-05, + "loss": 0.0564, + "step": 18446 + }, + { + "epoch": 2.42, + "learning_rate": 1.6363167831130828e-05, + "loss": 0.055, + "step": 18447 + }, + { + "epoch": 2.42, + "learning_rate": 1.6362758714649985e-05, + "loss": 0.0496, + "step": 18448 + }, + { + "epoch": 2.42, + "learning_rate": 1.636234958027422e-05, + "loss": 0.074, + "step": 18449 + }, + { + "epoch": 2.42, + "learning_rate": 1.6361940428004687e-05, + "loss": 0.0607, + "step": 18450 + }, + { + "epoch": 2.42, + "learning_rate": 1.6361531257842535e-05, + "loss": 0.0635, + "step": 18451 + }, + { + "epoch": 2.42, + "learning_rate": 1.6361122069788917e-05, + "loss": 0.0631, + "step": 18452 + }, + { + "epoch": 2.42, + "learning_rate": 1.6360712863844976e-05, + "loss": 0.0554, + "step": 18453 + }, + { + "epoch": 2.42, + "learning_rate": 1.6360303640011876e-05, + "loss": 0.058, + "step": 18454 + }, + { + "epoch": 2.42, + "learning_rate": 1.6359894398290754e-05, + "loss": 0.0651, + "step": 18455 + }, + { + "epoch": 2.42, + "learning_rate": 1.6359485138682774e-05, + "loss": 0.0571, + "step": 18456 + }, + { + "epoch": 2.42, + "learning_rate": 1.635907586118908e-05, + "loss": 0.0605, + "step": 18457 + }, + { + "epoch": 2.42, + "learning_rate": 1.6358666565810816e-05, + "loss": 0.0585, + "step": 18458 + }, + { + "epoch": 2.42, + "learning_rate": 1.635825725254915e-05, + "loss": 0.0726, + "step": 18459 + }, + { + "epoch": 2.42, + "learning_rate": 1.6357847921405218e-05, + "loss": 0.0704, + "step": 18460 + }, + { + "epoch": 2.42, + "learning_rate": 1.635743857238018e-05, + "loss": 0.0704, + "step": 18461 + }, + { + "epoch": 2.42, + "learning_rate": 1.6357029205475186e-05, + "loss": 0.0592, + "step": 18462 + }, + { + "epoch": 2.42, + "learning_rate": 1.6356619820691385e-05, + "loss": 0.0676, + "step": 18463 + }, + { + "epoch": 2.42, + "learning_rate": 1.6356210418029928e-05, + "loss": 0.0536, + "step": 18464 + }, + { + "epoch": 2.42, + "learning_rate": 1.6355800997491966e-05, + "loss": 0.0645, + "step": 18465 + }, + { + "epoch": 2.42, + "learning_rate": 1.6355391559078653e-05, + "loss": 0.0626, + "step": 18466 + }, + { + "epoch": 2.42, + "learning_rate": 1.6354982102791144e-05, + "loss": 0.0587, + "step": 18467 + }, + { + "epoch": 2.42, + "learning_rate": 1.635457262863058e-05, + "loss": 0.0642, + "step": 18468 + }, + { + "epoch": 2.42, + "learning_rate": 1.6354163136598126e-05, + "loss": 0.0679, + "step": 18469 + }, + { + "epoch": 2.42, + "learning_rate": 1.635375362669492e-05, + "loss": 0.062, + "step": 18470 + }, + { + "epoch": 2.42, + "learning_rate": 1.6353344098922127e-05, + "loss": 0.0559, + "step": 18471 + }, + { + "epoch": 2.42, + "learning_rate": 1.6352934553280885e-05, + "loss": 0.0677, + "step": 18472 + }, + { + "epoch": 2.42, + "learning_rate": 1.6352524989772356e-05, + "loss": 0.0601, + "step": 18473 + }, + { + "epoch": 2.42, + "learning_rate": 1.635211540839769e-05, + "loss": 0.0607, + "step": 18474 + }, + { + "epoch": 2.42, + "learning_rate": 1.6351705809158037e-05, + "loss": 0.0516, + "step": 18475 + }, + { + "epoch": 2.42, + "learning_rate": 1.635129619205455e-05, + "loss": 0.0642, + "step": 18476 + }, + { + "epoch": 2.42, + "learning_rate": 1.635088655708838e-05, + "loss": 0.0682, + "step": 18477 + }, + { + "epoch": 2.42, + "learning_rate": 1.6350476904260678e-05, + "loss": 0.0617, + "step": 18478 + }, + { + "epoch": 2.42, + "learning_rate": 1.6350067233572602e-05, + "loss": 0.0587, + "step": 18479 + }, + { + "epoch": 2.42, + "learning_rate": 1.6349657545025294e-05, + "loss": 0.0634, + "step": 18480 + }, + { + "epoch": 2.42, + "learning_rate": 1.6349247838619916e-05, + "loss": 0.0681, + "step": 18481 + }, + { + "epoch": 2.42, + "learning_rate": 1.6348838114357616e-05, + "loss": 0.0654, + "step": 18482 + }, + { + "epoch": 2.42, + "learning_rate": 1.634842837223955e-05, + "loss": 0.0644, + "step": 18483 + }, + { + "epoch": 2.42, + "learning_rate": 1.6348018612266866e-05, + "loss": 0.0555, + "step": 18484 + }, + { + "epoch": 2.42, + "learning_rate": 1.6347608834440716e-05, + "loss": 0.0846, + "step": 18485 + }, + { + "epoch": 2.42, + "learning_rate": 1.6347199038762253e-05, + "loss": 0.0554, + "step": 18486 + }, + { + "epoch": 2.42, + "learning_rate": 1.6346789225232634e-05, + "loss": 0.0629, + "step": 18487 + }, + { + "epoch": 2.42, + "learning_rate": 1.6346379393853006e-05, + "loss": 0.057, + "step": 18488 + }, + { + "epoch": 2.42, + "learning_rate": 1.6345969544624524e-05, + "loss": 0.0606, + "step": 18489 + }, + { + "epoch": 2.42, + "learning_rate": 1.634555967754834e-05, + "loss": 0.0663, + "step": 18490 + }, + { + "epoch": 2.42, + "learning_rate": 1.6345149792625606e-05, + "loss": 0.0695, + "step": 18491 + }, + { + "epoch": 2.42, + "learning_rate": 1.634473988985748e-05, + "loss": 0.0615, + "step": 18492 + }, + { + "epoch": 2.42, + "learning_rate": 1.634432996924511e-05, + "loss": 0.0596, + "step": 18493 + }, + { + "epoch": 2.42, + "learning_rate": 1.6343920030789647e-05, + "loss": 0.0574, + "step": 18494 + }, + { + "epoch": 2.42, + "learning_rate": 1.634351007449225e-05, + "loss": 0.0675, + "step": 18495 + }, + { + "epoch": 2.42, + "learning_rate": 1.634310010035407e-05, + "loss": 0.0598, + "step": 18496 + }, + { + "epoch": 2.42, + "learning_rate": 1.6342690108376254e-05, + "loss": 0.0568, + "step": 18497 + }, + { + "epoch": 2.42, + "learning_rate": 1.6342280098559964e-05, + "loss": 0.0671, + "step": 18498 + }, + { + "epoch": 2.42, + "learning_rate": 1.6341870070906346e-05, + "loss": 0.061, + "step": 18499 + }, + { + "epoch": 2.42, + "learning_rate": 1.634146002541656e-05, + "loss": 0.07, + "step": 18500 + }, + { + "epoch": 2.42, + "learning_rate": 1.634104996209175e-05, + "loss": 0.0732, + "step": 18501 + }, + { + "epoch": 2.42, + "learning_rate": 1.6340639880933077e-05, + "loss": 0.0499, + "step": 18502 + }, + { + "epoch": 2.42, + "learning_rate": 1.6340229781941693e-05, + "loss": 0.0519, + "step": 18503 + }, + { + "epoch": 2.42, + "learning_rate": 1.6339819665118752e-05, + "loss": 0.0624, + "step": 18504 + }, + { + "epoch": 2.42, + "learning_rate": 1.63394095304654e-05, + "loss": 0.0638, + "step": 18505 + }, + { + "epoch": 2.42, + "learning_rate": 1.6338999377982803e-05, + "loss": 0.0505, + "step": 18506 + }, + { + "epoch": 2.42, + "learning_rate": 1.6338589207672102e-05, + "loss": 0.07, + "step": 18507 + }, + { + "epoch": 2.42, + "learning_rate": 1.6338179019534458e-05, + "loss": 0.0632, + "step": 18508 + }, + { + "epoch": 2.42, + "learning_rate": 1.6337768813571026e-05, + "loss": 0.0707, + "step": 18509 + }, + { + "epoch": 2.42, + "learning_rate": 1.6337358589782955e-05, + "loss": 0.0585, + "step": 18510 + }, + { + "epoch": 2.42, + "learning_rate": 1.6336948348171403e-05, + "loss": 0.0715, + "step": 18511 + }, + { + "epoch": 2.42, + "learning_rate": 1.6336538088737515e-05, + "loss": 0.0599, + "step": 18512 + }, + { + "epoch": 2.43, + "learning_rate": 1.6336127811482458e-05, + "loss": 0.0614, + "step": 18513 + }, + { + "epoch": 2.43, + "learning_rate": 1.6335717516407372e-05, + "loss": 0.0654, + "step": 18514 + }, + { + "epoch": 2.43, + "learning_rate": 1.6335307203513423e-05, + "loss": 0.0557, + "step": 18515 + }, + { + "epoch": 2.43, + "learning_rate": 1.633489687280176e-05, + "loss": 0.0656, + "step": 18516 + }, + { + "epoch": 2.43, + "learning_rate": 1.6334486524273536e-05, + "loss": 0.0717, + "step": 18517 + }, + { + "epoch": 2.43, + "learning_rate": 1.63340761579299e-05, + "loss": 0.0543, + "step": 18518 + }, + { + "epoch": 2.43, + "learning_rate": 1.6333665773772017e-05, + "loss": 0.0516, + "step": 18519 + }, + { + "epoch": 2.43, + "learning_rate": 1.6333255371801038e-05, + "loss": 0.0631, + "step": 18520 + }, + { + "epoch": 2.43, + "learning_rate": 1.6332844952018115e-05, + "loss": 0.062, + "step": 18521 + }, + { + "epoch": 2.43, + "learning_rate": 1.6332434514424403e-05, + "loss": 0.0693, + "step": 18522 + }, + { + "epoch": 2.43, + "learning_rate": 1.6332024059021057e-05, + "loss": 0.0617, + "step": 18523 + }, + { + "epoch": 2.43, + "learning_rate": 1.6331613585809227e-05, + "loss": 0.0619, + "step": 18524 + }, + { + "epoch": 2.43, + "learning_rate": 1.633120309479007e-05, + "loss": 0.0666, + "step": 18525 + }, + { + "epoch": 2.43, + "learning_rate": 1.6330792585964746e-05, + "loss": 0.0572, + "step": 18526 + }, + { + "epoch": 2.43, + "learning_rate": 1.6330382059334403e-05, + "loss": 0.061, + "step": 18527 + }, + { + "epoch": 2.43, + "learning_rate": 1.6329971514900197e-05, + "loss": 0.0608, + "step": 18528 + }, + { + "epoch": 2.43, + "learning_rate": 1.6329560952663283e-05, + "loss": 0.063, + "step": 18529 + }, + { + "epoch": 2.43, + "learning_rate": 1.632915037262482e-05, + "loss": 0.0675, + "step": 18530 + }, + { + "epoch": 2.43, + "learning_rate": 1.6328739774785955e-05, + "loss": 0.0571, + "step": 18531 + }, + { + "epoch": 2.43, + "learning_rate": 1.6328329159147845e-05, + "loss": 0.0646, + "step": 18532 + }, + { + "epoch": 2.43, + "learning_rate": 1.632791852571165e-05, + "loss": 0.0627, + "step": 18533 + }, + { + "epoch": 2.43, + "learning_rate": 1.6327507874478517e-05, + "loss": 0.0637, + "step": 18534 + }, + { + "epoch": 2.43, + "learning_rate": 1.632709720544961e-05, + "loss": 0.0627, + "step": 18535 + }, + { + "epoch": 2.43, + "learning_rate": 1.6326686518626072e-05, + "loss": 0.0545, + "step": 18536 + }, + { + "epoch": 2.43, + "learning_rate": 1.632627581400907e-05, + "loss": 0.0549, + "step": 18537 + }, + { + "epoch": 2.43, + "learning_rate": 1.6325865091599756e-05, + "loss": 0.0652, + "step": 18538 + }, + { + "epoch": 2.43, + "learning_rate": 1.632545435139928e-05, + "loss": 0.0479, + "step": 18539 + }, + { + "epoch": 2.43, + "learning_rate": 1.6325043593408797e-05, + "loss": 0.0652, + "step": 18540 + }, + { + "epoch": 2.43, + "learning_rate": 1.6324632817629473e-05, + "loss": 0.0644, + "step": 18541 + }, + { + "epoch": 2.43, + "learning_rate": 1.6324222024062453e-05, + "loss": 0.072, + "step": 18542 + }, + { + "epoch": 2.43, + "learning_rate": 1.6323811212708897e-05, + "loss": 0.0559, + "step": 18543 + }, + { + "epoch": 2.43, + "learning_rate": 1.6323400383569956e-05, + "loss": 0.0651, + "step": 18544 + }, + { + "epoch": 2.43, + "learning_rate": 1.632298953664679e-05, + "loss": 0.0585, + "step": 18545 + }, + { + "epoch": 2.43, + "learning_rate": 1.632257867194055e-05, + "loss": 0.0672, + "step": 18546 + }, + { + "epoch": 2.43, + "learning_rate": 1.6322167789452397e-05, + "loss": 0.0633, + "step": 18547 + }, + { + "epoch": 2.43, + "learning_rate": 1.6321756889183485e-05, + "loss": 0.0678, + "step": 18548 + }, + { + "epoch": 2.43, + "learning_rate": 1.6321345971134962e-05, + "loss": 0.0722, + "step": 18549 + }, + { + "epoch": 2.43, + "learning_rate": 1.6320935035307996e-05, + "loss": 0.0608, + "step": 18550 + }, + { + "epoch": 2.43, + "learning_rate": 1.6320524081703734e-05, + "loss": 0.0564, + "step": 18551 + }, + { + "epoch": 2.43, + "learning_rate": 1.6320113110323336e-05, + "loss": 0.0602, + "step": 18552 + }, + { + "epoch": 2.43, + "learning_rate": 1.6319702121167956e-05, + "loss": 0.0625, + "step": 18553 + }, + { + "epoch": 2.43, + "learning_rate": 1.631929111423875e-05, + "loss": 0.0599, + "step": 18554 + }, + { + "epoch": 2.43, + "learning_rate": 1.6318880089536877e-05, + "loss": 0.0523, + "step": 18555 + }, + { + "epoch": 2.43, + "learning_rate": 1.6318469047063487e-05, + "loss": 0.0584, + "step": 18556 + }, + { + "epoch": 2.43, + "learning_rate": 1.631805798681974e-05, + "loss": 0.0598, + "step": 18557 + }, + { + "epoch": 2.43, + "learning_rate": 1.6317646908806793e-05, + "loss": 0.0625, + "step": 18558 + }, + { + "epoch": 2.43, + "learning_rate": 1.63172358130258e-05, + "loss": 0.068, + "step": 18559 + }, + { + "epoch": 2.43, + "learning_rate": 1.6316824699477916e-05, + "loss": 0.0599, + "step": 18560 + }, + { + "epoch": 2.43, + "learning_rate": 1.63164135681643e-05, + "loss": 0.0574, + "step": 18561 + }, + { + "epoch": 2.43, + "learning_rate": 1.6316002419086105e-05, + "loss": 0.0635, + "step": 18562 + }, + { + "epoch": 2.43, + "learning_rate": 1.6315591252244493e-05, + "loss": 0.0578, + "step": 18563 + }, + { + "epoch": 2.43, + "learning_rate": 1.6315180067640617e-05, + "loss": 0.056, + "step": 18564 + }, + { + "epoch": 2.43, + "learning_rate": 1.6314768865275632e-05, + "loss": 0.0603, + "step": 18565 + }, + { + "epoch": 2.43, + "learning_rate": 1.6314357645150694e-05, + "loss": 0.0587, + "step": 18566 + }, + { + "epoch": 2.43, + "learning_rate": 1.6313946407266962e-05, + "loss": 0.0545, + "step": 18567 + }, + { + "epoch": 2.43, + "learning_rate": 1.6313535151625594e-05, + "loss": 0.0564, + "step": 18568 + }, + { + "epoch": 2.43, + "learning_rate": 1.6313123878227742e-05, + "loss": 0.0654, + "step": 18569 + }, + { + "epoch": 2.43, + "learning_rate": 1.6312712587074568e-05, + "loss": 0.0611, + "step": 18570 + }, + { + "epoch": 2.43, + "learning_rate": 1.6312301278167225e-05, + "loss": 0.0666, + "step": 18571 + }, + { + "epoch": 2.43, + "learning_rate": 1.631188995150687e-05, + "loss": 0.0675, + "step": 18572 + }, + { + "epoch": 2.43, + "learning_rate": 1.631147860709466e-05, + "loss": 0.0546, + "step": 18573 + }, + { + "epoch": 2.43, + "learning_rate": 1.6311067244931757e-05, + "loss": 0.065, + "step": 18574 + }, + { + "epoch": 2.43, + "learning_rate": 1.6310655865019312e-05, + "loss": 0.0569, + "step": 18575 + }, + { + "epoch": 2.43, + "learning_rate": 1.631024446735848e-05, + "loss": 0.0586, + "step": 18576 + }, + { + "epoch": 2.43, + "learning_rate": 1.6309833051950422e-05, + "loss": 0.0668, + "step": 18577 + }, + { + "epoch": 2.43, + "learning_rate": 1.6309421618796295e-05, + "loss": 0.0752, + "step": 18578 + }, + { + "epoch": 2.43, + "learning_rate": 1.6309010167897258e-05, + "loss": 0.0642, + "step": 18579 + }, + { + "epoch": 2.43, + "learning_rate": 1.6308598699254466e-05, + "loss": 0.0586, + "step": 18580 + }, + { + "epoch": 2.43, + "learning_rate": 1.630818721286907e-05, + "loss": 0.0754, + "step": 18581 + }, + { + "epoch": 2.43, + "learning_rate": 1.630777570874224e-05, + "loss": 0.0626, + "step": 18582 + }, + { + "epoch": 2.43, + "learning_rate": 1.6307364186875127e-05, + "loss": 0.066, + "step": 18583 + }, + { + "epoch": 2.43, + "learning_rate": 1.6306952647268884e-05, + "loss": 0.0574, + "step": 18584 + }, + { + "epoch": 2.43, + "learning_rate": 1.6306541089924676e-05, + "loss": 0.0654, + "step": 18585 + }, + { + "epoch": 2.43, + "learning_rate": 1.6306129514843653e-05, + "loss": 0.061, + "step": 18586 + }, + { + "epoch": 2.43, + "learning_rate": 1.630571792202698e-05, + "loss": 0.0672, + "step": 18587 + }, + { + "epoch": 2.43, + "learning_rate": 1.630530631147581e-05, + "loss": 0.0707, + "step": 18588 + }, + { + "epoch": 2.44, + "learning_rate": 1.63048946831913e-05, + "loss": 0.0571, + "step": 18589 + }, + { + "epoch": 2.44, + "learning_rate": 1.6304483037174616e-05, + "loss": 0.0629, + "step": 18590 + }, + { + "epoch": 2.44, + "learning_rate": 1.6304071373426902e-05, + "loss": 0.0655, + "step": 18591 + }, + { + "epoch": 2.44, + "learning_rate": 1.6303659691949325e-05, + "loss": 0.0604, + "step": 18592 + }, + { + "epoch": 2.44, + "learning_rate": 1.630324799274304e-05, + "loss": 0.0581, + "step": 18593 + }, + { + "epoch": 2.44, + "learning_rate": 1.630283627580921e-05, + "loss": 0.0666, + "step": 18594 + }, + { + "epoch": 2.44, + "learning_rate": 1.630242454114898e-05, + "loss": 0.0561, + "step": 18595 + }, + { + "epoch": 2.44, + "learning_rate": 1.6302012788763525e-05, + "loss": 0.0518, + "step": 18596 + }, + { + "epoch": 2.44, + "learning_rate": 1.6301601018653992e-05, + "loss": 0.0575, + "step": 18597 + }, + { + "epoch": 2.44, + "learning_rate": 1.6301189230821542e-05, + "loss": 0.0577, + "step": 18598 + }, + { + "epoch": 2.44, + "learning_rate": 1.630077742526733e-05, + "loss": 0.0531, + "step": 18599 + }, + { + "epoch": 2.44, + "learning_rate": 1.630036560199252e-05, + "loss": 0.0556, + "step": 18600 + }, + { + "epoch": 2.44, + "learning_rate": 1.6299953760998267e-05, + "loss": 0.0653, + "step": 18601 + }, + { + "epoch": 2.44, + "learning_rate": 1.629954190228573e-05, + "loss": 0.0571, + "step": 18602 + }, + { + "epoch": 2.44, + "learning_rate": 1.6299130025856067e-05, + "loss": 0.0614, + "step": 18603 + }, + { + "epoch": 2.44, + "learning_rate": 1.6298718131710435e-05, + "loss": 0.0513, + "step": 18604 + }, + { + "epoch": 2.44, + "learning_rate": 1.6298306219849993e-05, + "loss": 0.065, + "step": 18605 + }, + { + "epoch": 2.44, + "learning_rate": 1.6297894290275903e-05, + "loss": 0.0683, + "step": 18606 + }, + { + "epoch": 2.44, + "learning_rate": 1.6297482342989317e-05, + "loss": 0.0642, + "step": 18607 + }, + { + "epoch": 2.44, + "learning_rate": 1.62970703779914e-05, + "loss": 0.0593, + "step": 18608 + }, + { + "epoch": 2.44, + "learning_rate": 1.6296658395283308e-05, + "loss": 0.057, + "step": 18609 + }, + { + "epoch": 2.44, + "learning_rate": 1.6296246394866198e-05, + "loss": 0.0675, + "step": 18610 + }, + { + "epoch": 2.44, + "learning_rate": 1.629583437674123e-05, + "loss": 0.0582, + "step": 18611 + }, + { + "epoch": 2.44, + "learning_rate": 1.6295422340909565e-05, + "loss": 0.0619, + "step": 18612 + }, + { + "epoch": 2.44, + "learning_rate": 1.6295010287372356e-05, + "loss": 0.0547, + "step": 18613 + }, + { + "epoch": 2.44, + "learning_rate": 1.629459821613077e-05, + "loss": 0.0554, + "step": 18614 + }, + { + "epoch": 2.44, + "learning_rate": 1.6294186127185964e-05, + "loss": 0.0614, + "step": 18615 + }, + { + "epoch": 2.44, + "learning_rate": 1.6293774020539087e-05, + "loss": 0.061, + "step": 18616 + }, + { + "epoch": 2.44, + "learning_rate": 1.629336189619131e-05, + "loss": 0.0526, + "step": 18617 + }, + { + "epoch": 2.44, + "learning_rate": 1.629294975414379e-05, + "loss": 0.0531, + "step": 18618 + }, + { + "epoch": 2.44, + "learning_rate": 1.6292537594397678e-05, + "loss": 0.0624, + "step": 18619 + }, + { + "epoch": 2.44, + "learning_rate": 1.6292125416954145e-05, + "loss": 0.0575, + "step": 18620 + }, + { + "epoch": 2.44, + "learning_rate": 1.629171322181434e-05, + "loss": 0.0565, + "step": 18621 + }, + { + "epoch": 2.44, + "learning_rate": 1.629130100897943e-05, + "loss": 0.0645, + "step": 18622 + }, + { + "epoch": 2.44, + "learning_rate": 1.629088877845057e-05, + "loss": 0.0571, + "step": 18623 + }, + { + "epoch": 2.44, + "learning_rate": 1.629047653022892e-05, + "loss": 0.0658, + "step": 18624 + }, + { + "epoch": 2.44, + "learning_rate": 1.629006426431564e-05, + "loss": 0.0671, + "step": 18625 + }, + { + "epoch": 2.44, + "learning_rate": 1.628965198071189e-05, + "loss": 0.0693, + "step": 18626 + }, + { + "epoch": 2.44, + "learning_rate": 1.628923967941883e-05, + "loss": 0.057, + "step": 18627 + }, + { + "epoch": 2.44, + "learning_rate": 1.628882736043761e-05, + "loss": 0.0682, + "step": 18628 + }, + { + "epoch": 2.44, + "learning_rate": 1.6288415023769407e-05, + "loss": 0.0625, + "step": 18629 + }, + { + "epoch": 2.44, + "learning_rate": 1.628800266941537e-05, + "loss": 0.0511, + "step": 18630 + }, + { + "epoch": 2.44, + "learning_rate": 1.6287590297376658e-05, + "loss": 0.0769, + "step": 18631 + }, + { + "epoch": 2.44, + "learning_rate": 1.6287177907654434e-05, + "loss": 0.0522, + "step": 18632 + }, + { + "epoch": 2.44, + "learning_rate": 1.6286765500249855e-05, + "loss": 0.0556, + "step": 18633 + }, + { + "epoch": 2.44, + "learning_rate": 1.6286353075164086e-05, + "loss": 0.0554, + "step": 18634 + }, + { + "epoch": 2.44, + "learning_rate": 1.6285940632398284e-05, + "loss": 0.0583, + "step": 18635 + }, + { + "epoch": 2.44, + "learning_rate": 1.6285528171953606e-05, + "loss": 0.0543, + "step": 18636 + }, + { + "epoch": 2.44, + "learning_rate": 1.6285115693831215e-05, + "loss": 0.0668, + "step": 18637 + }, + { + "epoch": 2.44, + "learning_rate": 1.6284703198032273e-05, + "loss": 0.0599, + "step": 18638 + }, + { + "epoch": 2.44, + "learning_rate": 1.6284290684557935e-05, + "loss": 0.0556, + "step": 18639 + }, + { + "epoch": 2.44, + "learning_rate": 1.6283878153409368e-05, + "loss": 0.0642, + "step": 18640 + }, + { + "epoch": 2.44, + "learning_rate": 1.6283465604587725e-05, + "loss": 0.0685, + "step": 18641 + }, + { + "epoch": 2.44, + "learning_rate": 1.628305303809417e-05, + "loss": 0.0634, + "step": 18642 + }, + { + "epoch": 2.44, + "learning_rate": 1.6282640453929864e-05, + "loss": 0.0572, + "step": 18643 + }, + { + "epoch": 2.44, + "learning_rate": 1.6282227852095967e-05, + "loss": 0.0634, + "step": 18644 + }, + { + "epoch": 2.44, + "learning_rate": 1.6281815232593636e-05, + "loss": 0.0606, + "step": 18645 + }, + { + "epoch": 2.44, + "learning_rate": 1.6281402595424035e-05, + "loss": 0.0553, + "step": 18646 + }, + { + "epoch": 2.44, + "learning_rate": 1.6280989940588328e-05, + "loss": 0.0672, + "step": 18647 + }, + { + "epoch": 2.44, + "learning_rate": 1.6280577268087668e-05, + "loss": 0.0651, + "step": 18648 + }, + { + "epoch": 2.44, + "learning_rate": 1.6280164577923216e-05, + "loss": 0.057, + "step": 18649 + }, + { + "epoch": 2.44, + "learning_rate": 1.627975187009614e-05, + "loss": 0.0599, + "step": 18650 + }, + { + "epoch": 2.44, + "learning_rate": 1.6279339144607594e-05, + "loss": 0.0729, + "step": 18651 + }, + { + "epoch": 2.44, + "learning_rate": 1.627892640145874e-05, + "loss": 0.0568, + "step": 18652 + }, + { + "epoch": 2.44, + "learning_rate": 1.6278513640650743e-05, + "loss": 0.0673, + "step": 18653 + }, + { + "epoch": 2.44, + "learning_rate": 1.627810086218476e-05, + "loss": 0.0638, + "step": 18654 + }, + { + "epoch": 2.44, + "learning_rate": 1.627768806606195e-05, + "loss": 0.0627, + "step": 18655 + }, + { + "epoch": 2.44, + "learning_rate": 1.627727525228348e-05, + "loss": 0.0572, + "step": 18656 + }, + { + "epoch": 2.44, + "learning_rate": 1.6276862420850505e-05, + "loss": 0.0628, + "step": 18657 + }, + { + "epoch": 2.44, + "learning_rate": 1.6276449571764193e-05, + "loss": 0.062, + "step": 18658 + }, + { + "epoch": 2.44, + "learning_rate": 1.6276036705025695e-05, + "loss": 0.0698, + "step": 18659 + }, + { + "epoch": 2.44, + "learning_rate": 1.627562382063618e-05, + "loss": 0.0678, + "step": 18660 + }, + { + "epoch": 2.44, + "learning_rate": 1.627521091859681e-05, + "loss": 0.0607, + "step": 18661 + }, + { + "epoch": 2.44, + "learning_rate": 1.627479799890874e-05, + "loss": 0.0644, + "step": 18662 + }, + { + "epoch": 2.44, + "learning_rate": 1.6274385061573136e-05, + "loss": 0.0614, + "step": 18663 + }, + { + "epoch": 2.44, + "learning_rate": 1.627397210659116e-05, + "loss": 0.0713, + "step": 18664 + }, + { + "epoch": 2.44, + "learning_rate": 1.627355913396397e-05, + "loss": 0.0648, + "step": 18665 + }, + { + "epoch": 2.45, + "learning_rate": 1.6273146143692727e-05, + "loss": 0.0607, + "step": 18666 + }, + { + "epoch": 2.45, + "learning_rate": 1.6272733135778597e-05, + "loss": 0.0664, + "step": 18667 + }, + { + "epoch": 2.45, + "learning_rate": 1.6272320110222735e-05, + "loss": 0.0574, + "step": 18668 + }, + { + "epoch": 2.45, + "learning_rate": 1.6271907067026315e-05, + "loss": 0.0537, + "step": 18669 + }, + { + "epoch": 2.45, + "learning_rate": 1.6271494006190485e-05, + "loss": 0.0652, + "step": 18670 + }, + { + "epoch": 2.45, + "learning_rate": 1.627108092771641e-05, + "loss": 0.059, + "step": 18671 + }, + { + "epoch": 2.45, + "learning_rate": 1.6270667831605258e-05, + "loss": 0.0636, + "step": 18672 + }, + { + "epoch": 2.45, + "learning_rate": 1.6270254717858187e-05, + "loss": 0.063, + "step": 18673 + }, + { + "epoch": 2.45, + "learning_rate": 1.6269841586476353e-05, + "loss": 0.0643, + "step": 18674 + }, + { + "epoch": 2.45, + "learning_rate": 1.6269428437460927e-05, + "loss": 0.0653, + "step": 18675 + }, + { + "epoch": 2.45, + "learning_rate": 1.626901527081307e-05, + "loss": 0.0588, + "step": 18676 + }, + { + "epoch": 2.45, + "learning_rate": 1.6268602086533937e-05, + "loss": 0.056, + "step": 18677 + }, + { + "epoch": 2.45, + "learning_rate": 1.6268188884624695e-05, + "loss": 0.0616, + "step": 18678 + }, + { + "epoch": 2.45, + "learning_rate": 1.6267775665086508e-05, + "loss": 0.0651, + "step": 18679 + }, + { + "epoch": 2.45, + "learning_rate": 1.6267362427920536e-05, + "loss": 0.0522, + "step": 18680 + }, + { + "epoch": 2.45, + "learning_rate": 1.626694917312794e-05, + "loss": 0.0685, + "step": 18681 + }, + { + "epoch": 2.45, + "learning_rate": 1.6266535900709882e-05, + "loss": 0.0531, + "step": 18682 + }, + { + "epoch": 2.45, + "learning_rate": 1.6266122610667526e-05, + "loss": 0.0674, + "step": 18683 + }, + { + "epoch": 2.45, + "learning_rate": 1.6265709303002032e-05, + "loss": 0.062, + "step": 18684 + }, + { + "epoch": 2.45, + "learning_rate": 1.626529597771457e-05, + "loss": 0.0603, + "step": 18685 + }, + { + "epoch": 2.45, + "learning_rate": 1.6264882634806293e-05, + "loss": 0.0592, + "step": 18686 + }, + { + "epoch": 2.45, + "learning_rate": 1.6264469274278364e-05, + "loss": 0.0771, + "step": 18687 + }, + { + "epoch": 2.45, + "learning_rate": 1.6264055896131953e-05, + "loss": 0.0646, + "step": 18688 + }, + { + "epoch": 2.45, + "learning_rate": 1.6263642500368216e-05, + "loss": 0.0645, + "step": 18689 + }, + { + "epoch": 2.45, + "learning_rate": 1.626322908698832e-05, + "loss": 0.075, + "step": 18690 + }, + { + "epoch": 2.45, + "learning_rate": 1.6262815655993428e-05, + "loss": 0.0515, + "step": 18691 + }, + { + "epoch": 2.45, + "learning_rate": 1.6262402207384698e-05, + "loss": 0.0614, + "step": 18692 + }, + { + "epoch": 2.45, + "learning_rate": 1.6261988741163294e-05, + "loss": 0.0639, + "step": 18693 + }, + { + "epoch": 2.45, + "learning_rate": 1.626157525733038e-05, + "loss": 0.0647, + "step": 18694 + }, + { + "epoch": 2.45, + "learning_rate": 1.626116175588712e-05, + "loss": 0.0662, + "step": 18695 + }, + { + "epoch": 2.45, + "learning_rate": 1.6260748236834677e-05, + "loss": 0.0605, + "step": 18696 + }, + { + "epoch": 2.45, + "learning_rate": 1.6260334700174213e-05, + "loss": 0.0572, + "step": 18697 + }, + { + "epoch": 2.45, + "learning_rate": 1.6259921145906888e-05, + "loss": 0.0595, + "step": 18698 + }, + { + "epoch": 2.45, + "learning_rate": 1.625950757403387e-05, + "loss": 0.0725, + "step": 18699 + }, + { + "epoch": 2.45, + "learning_rate": 1.625909398455632e-05, + "loss": 0.0549, + "step": 18700 + }, + { + "epoch": 2.45, + "learning_rate": 1.6258680377475404e-05, + "loss": 0.0597, + "step": 18701 + }, + { + "epoch": 2.45, + "learning_rate": 1.625826675279228e-05, + "loss": 0.0536, + "step": 18702 + }, + { + "epoch": 2.45, + "learning_rate": 1.6257853110508117e-05, + "loss": 0.07, + "step": 18703 + }, + { + "epoch": 2.45, + "learning_rate": 1.625743945062407e-05, + "loss": 0.0711, + "step": 18704 + }, + { + "epoch": 2.45, + "learning_rate": 1.6257025773141312e-05, + "loss": 0.0594, + "step": 18705 + }, + { + "epoch": 2.45, + "learning_rate": 1.6256612078061e-05, + "loss": 0.0743, + "step": 18706 + }, + { + "epoch": 2.45, + "learning_rate": 1.62561983653843e-05, + "loss": 0.0629, + "step": 18707 + }, + { + "epoch": 2.45, + "learning_rate": 1.625578463511238e-05, + "loss": 0.0585, + "step": 18708 + }, + { + "epoch": 2.45, + "learning_rate": 1.6255370887246394e-05, + "loss": 0.0667, + "step": 18709 + }, + { + "epoch": 2.45, + "learning_rate": 1.625495712178751e-05, + "loss": 0.0615, + "step": 18710 + }, + { + "epoch": 2.45, + "learning_rate": 1.625454333873689e-05, + "loss": 0.0606, + "step": 18711 + }, + { + "epoch": 2.45, + "learning_rate": 1.6254129538095708e-05, + "loss": 0.0626, + "step": 18712 + }, + { + "epoch": 2.45, + "learning_rate": 1.6253715719865114e-05, + "loss": 0.0621, + "step": 18713 + }, + { + "epoch": 2.45, + "learning_rate": 1.625330188404628e-05, + "loss": 0.0615, + "step": 18714 + }, + { + "epoch": 2.45, + "learning_rate": 1.6252888030640365e-05, + "loss": 0.0578, + "step": 18715 + }, + { + "epoch": 2.45, + "learning_rate": 1.6252474159648536e-05, + "loss": 0.0666, + "step": 18716 + }, + { + "epoch": 2.45, + "learning_rate": 1.625206027107196e-05, + "loss": 0.0638, + "step": 18717 + }, + { + "epoch": 2.45, + "learning_rate": 1.6251646364911793e-05, + "loss": 0.0578, + "step": 18718 + }, + { + "epoch": 2.45, + "learning_rate": 1.6251232441169208e-05, + "loss": 0.0559, + "step": 18719 + }, + { + "epoch": 2.45, + "learning_rate": 1.625081849984536e-05, + "loss": 0.0651, + "step": 18720 + }, + { + "epoch": 2.45, + "learning_rate": 1.6250404540941417e-05, + "loss": 0.0598, + "step": 18721 + }, + { + "epoch": 2.45, + "learning_rate": 1.624999056445855e-05, + "loss": 0.0621, + "step": 18722 + }, + { + "epoch": 2.45, + "learning_rate": 1.6249576570397915e-05, + "loss": 0.0619, + "step": 18723 + }, + { + "epoch": 2.45, + "learning_rate": 1.6249162558760676e-05, + "loss": 0.0586, + "step": 18724 + }, + { + "epoch": 2.45, + "learning_rate": 1.6248748529548e-05, + "loss": 0.0699, + "step": 18725 + }, + { + "epoch": 2.45, + "learning_rate": 1.6248334482761056e-05, + "loss": 0.0581, + "step": 18726 + }, + { + "epoch": 2.45, + "learning_rate": 1.6247920418401e-05, + "loss": 0.0645, + "step": 18727 + }, + { + "epoch": 2.45, + "learning_rate": 1.6247506336469e-05, + "loss": 0.0674, + "step": 18728 + }, + { + "epoch": 2.45, + "learning_rate": 1.6247092236966226e-05, + "loss": 0.0577, + "step": 18729 + }, + { + "epoch": 2.45, + "learning_rate": 1.6246678119893835e-05, + "loss": 0.0612, + "step": 18730 + }, + { + "epoch": 2.45, + "learning_rate": 1.6246263985252993e-05, + "loss": 0.0628, + "step": 18731 + }, + { + "epoch": 2.45, + "learning_rate": 1.6245849833044866e-05, + "loss": 0.0688, + "step": 18732 + }, + { + "epoch": 2.45, + "learning_rate": 1.624543566327062e-05, + "loss": 0.0516, + "step": 18733 + }, + { + "epoch": 2.45, + "learning_rate": 1.624502147593142e-05, + "loss": 0.0699, + "step": 18734 + }, + { + "epoch": 2.45, + "learning_rate": 1.6244607271028428e-05, + "loss": 0.0709, + "step": 18735 + }, + { + "epoch": 2.45, + "learning_rate": 1.624419304856281e-05, + "loss": 0.0659, + "step": 18736 + }, + { + "epoch": 2.45, + "learning_rate": 1.6243778808535735e-05, + "loss": 0.0631, + "step": 18737 + }, + { + "epoch": 2.45, + "learning_rate": 1.624336455094836e-05, + "loss": 0.0643, + "step": 18738 + }, + { + "epoch": 2.45, + "learning_rate": 1.624295027580186e-05, + "loss": 0.0631, + "step": 18739 + }, + { + "epoch": 2.45, + "learning_rate": 1.6242535983097394e-05, + "loss": 0.0564, + "step": 18740 + }, + { + "epoch": 2.45, + "learning_rate": 1.6242121672836125e-05, + "loss": 0.0604, + "step": 18741 + }, + { + "epoch": 2.46, + "learning_rate": 1.6241707345019223e-05, + "loss": 0.0619, + "step": 18742 + }, + { + "epoch": 2.46, + "learning_rate": 1.6241292999647852e-05, + "loss": 0.0565, + "step": 18743 + }, + { + "epoch": 2.46, + "learning_rate": 1.6240878636723173e-05, + "loss": 0.0593, + "step": 18744 + }, + { + "epoch": 2.46, + "learning_rate": 1.624046425624636e-05, + "loss": 0.0644, + "step": 18745 + }, + { + "epoch": 2.46, + "learning_rate": 1.624004985821857e-05, + "loss": 0.072, + "step": 18746 + }, + { + "epoch": 2.46, + "learning_rate": 1.6239635442640972e-05, + "loss": 0.0568, + "step": 18747 + }, + { + "epoch": 2.46, + "learning_rate": 1.6239221009514734e-05, + "loss": 0.0523, + "step": 18748 + }, + { + "epoch": 2.46, + "learning_rate": 1.6238806558841016e-05, + "loss": 0.0697, + "step": 18749 + }, + { + "epoch": 2.46, + "learning_rate": 1.623839209062099e-05, + "loss": 0.0668, + "step": 18750 + }, + { + "epoch": 2.46, + "learning_rate": 1.623797760485582e-05, + "loss": 0.0742, + "step": 18751 + }, + { + "epoch": 2.46, + "learning_rate": 1.6237563101546668e-05, + "loss": 0.0568, + "step": 18752 + }, + { + "epoch": 2.46, + "learning_rate": 1.6237148580694698e-05, + "loss": 0.0542, + "step": 18753 + }, + { + "epoch": 2.46, + "learning_rate": 1.6236734042301082e-05, + "loss": 0.0645, + "step": 18754 + }, + { + "epoch": 2.46, + "learning_rate": 1.6236319486366987e-05, + "loss": 0.0636, + "step": 18755 + }, + { + "epoch": 2.46, + "learning_rate": 1.6235904912893573e-05, + "loss": 0.061, + "step": 18756 + }, + { + "epoch": 2.46, + "learning_rate": 1.6235490321882013e-05, + "loss": 0.0592, + "step": 18757 + }, + { + "epoch": 2.46, + "learning_rate": 1.623507571333346e-05, + "loss": 0.0679, + "step": 18758 + }, + { + "epoch": 2.46, + "learning_rate": 1.6234661087249095e-05, + "loss": 0.0634, + "step": 18759 + }, + { + "epoch": 2.46, + "learning_rate": 1.6234246443630074e-05, + "loss": 0.0692, + "step": 18760 + }, + { + "epoch": 2.46, + "learning_rate": 1.6233831782477573e-05, + "loss": 0.0614, + "step": 18761 + }, + { + "epoch": 2.46, + "learning_rate": 1.6233417103792747e-05, + "loss": 0.0707, + "step": 18762 + }, + { + "epoch": 2.46, + "learning_rate": 1.6233002407576764e-05, + "loss": 0.0594, + "step": 18763 + }, + { + "epoch": 2.46, + "learning_rate": 1.6232587693830798e-05, + "loss": 0.0715, + "step": 18764 + }, + { + "epoch": 2.46, + "learning_rate": 1.6232172962556014e-05, + "loss": 0.0597, + "step": 18765 + }, + { + "epoch": 2.46, + "learning_rate": 1.623175821375357e-05, + "loss": 0.0581, + "step": 18766 + }, + { + "epoch": 2.46, + "learning_rate": 1.6231343447424642e-05, + "loss": 0.0649, + "step": 18767 + }, + { + "epoch": 2.46, + "learning_rate": 1.623092866357039e-05, + "loss": 0.0585, + "step": 18768 + }, + { + "epoch": 2.46, + "learning_rate": 1.6230513862191983e-05, + "loss": 0.0695, + "step": 18769 + }, + { + "epoch": 2.46, + "learning_rate": 1.6230099043290586e-05, + "loss": 0.0748, + "step": 18770 + }, + { + "epoch": 2.46, + "learning_rate": 1.622968420686737e-05, + "loss": 0.0614, + "step": 18771 + }, + { + "epoch": 2.46, + "learning_rate": 1.62292693529235e-05, + "loss": 0.0524, + "step": 18772 + }, + { + "epoch": 2.46, + "learning_rate": 1.6228854481460138e-05, + "loss": 0.0592, + "step": 18773 + }, + { + "epoch": 2.46, + "learning_rate": 1.6228439592478458e-05, + "loss": 0.0571, + "step": 18774 + }, + { + "epoch": 2.46, + "learning_rate": 1.622802468597962e-05, + "loss": 0.0631, + "step": 18775 + }, + { + "epoch": 2.46, + "learning_rate": 1.6227609761964796e-05, + "loss": 0.0686, + "step": 18776 + }, + { + "epoch": 2.46, + "learning_rate": 1.6227194820435154e-05, + "loss": 0.0605, + "step": 18777 + }, + { + "epoch": 2.46, + "learning_rate": 1.6226779861391854e-05, + "loss": 0.0734, + "step": 18778 + }, + { + "epoch": 2.46, + "learning_rate": 1.622636488483607e-05, + "loss": 0.0627, + "step": 18779 + }, + { + "epoch": 2.46, + "learning_rate": 1.6225949890768962e-05, + "loss": 0.067, + "step": 18780 + }, + { + "epoch": 2.46, + "learning_rate": 1.6225534879191704e-05, + "loss": 0.0668, + "step": 18781 + }, + { + "epoch": 2.46, + "learning_rate": 1.622511985010546e-05, + "loss": 0.0609, + "step": 18782 + }, + { + "epoch": 2.46, + "learning_rate": 1.6224704803511403e-05, + "loss": 0.065, + "step": 18783 + }, + { + "epoch": 2.46, + "learning_rate": 1.622428973941069e-05, + "loss": 0.07, + "step": 18784 + }, + { + "epoch": 2.46, + "learning_rate": 1.6223874657804494e-05, + "loss": 0.0632, + "step": 18785 + }, + { + "epoch": 2.46, + "learning_rate": 1.622345955869398e-05, + "loss": 0.0535, + "step": 18786 + }, + { + "epoch": 2.46, + "learning_rate": 1.6223044442080323e-05, + "loss": 0.0517, + "step": 18787 + }, + { + "epoch": 2.46, + "learning_rate": 1.622262930796468e-05, + "loss": 0.0512, + "step": 18788 + }, + { + "epoch": 2.46, + "learning_rate": 1.622221415634823e-05, + "loss": 0.0602, + "step": 18789 + }, + { + "epoch": 2.46, + "learning_rate": 1.6221798987232125e-05, + "loss": 0.0593, + "step": 18790 + }, + { + "epoch": 2.46, + "learning_rate": 1.6221383800617548e-05, + "loss": 0.0677, + "step": 18791 + }, + { + "epoch": 2.46, + "learning_rate": 1.622096859650566e-05, + "loss": 0.0701, + "step": 18792 + }, + { + "epoch": 2.46, + "learning_rate": 1.6220553374897626e-05, + "loss": 0.0635, + "step": 18793 + }, + { + "epoch": 2.46, + "learning_rate": 1.622013813579462e-05, + "loss": 0.0642, + "step": 18794 + }, + { + "epoch": 2.46, + "learning_rate": 1.6219722879197804e-05, + "loss": 0.07, + "step": 18795 + }, + { + "epoch": 2.46, + "learning_rate": 1.621930760510835e-05, + "loss": 0.0597, + "step": 18796 + }, + { + "epoch": 2.46, + "learning_rate": 1.6218892313527422e-05, + "loss": 0.0733, + "step": 18797 + }, + { + "epoch": 2.46, + "learning_rate": 1.6218477004456194e-05, + "loss": 0.0463, + "step": 18798 + }, + { + "epoch": 2.46, + "learning_rate": 1.6218061677895833e-05, + "loss": 0.057, + "step": 18799 + }, + { + "epoch": 2.46, + "learning_rate": 1.62176463338475e-05, + "loss": 0.0638, + "step": 18800 + }, + { + "epoch": 2.46, + "learning_rate": 1.621723097231237e-05, + "loss": 0.0615, + "step": 18801 + }, + { + "epoch": 2.46, + "learning_rate": 1.6216815593291605e-05, + "loss": 0.0594, + "step": 18802 + }, + { + "epoch": 2.46, + "learning_rate": 1.621640019678638e-05, + "loss": 0.0581, + "step": 18803 + }, + { + "epoch": 2.46, + "learning_rate": 1.6215984782797863e-05, + "loss": 0.0684, + "step": 18804 + }, + { + "epoch": 2.46, + "learning_rate": 1.6215569351327216e-05, + "loss": 0.0612, + "step": 18805 + }, + { + "epoch": 2.46, + "learning_rate": 1.6215153902375612e-05, + "loss": 0.0621, + "step": 18806 + }, + { + "epoch": 2.46, + "learning_rate": 1.621473843594422e-05, + "loss": 0.0611, + "step": 18807 + }, + { + "epoch": 2.46, + "learning_rate": 1.6214322952034206e-05, + "loss": 0.0558, + "step": 18808 + }, + { + "epoch": 2.46, + "learning_rate": 1.621390745064674e-05, + "loss": 0.0613, + "step": 18809 + }, + { + "epoch": 2.46, + "learning_rate": 1.621349193178299e-05, + "loss": 0.0609, + "step": 18810 + }, + { + "epoch": 2.46, + "learning_rate": 1.6213076395444125e-05, + "loss": 0.0725, + "step": 18811 + }, + { + "epoch": 2.46, + "learning_rate": 1.6212660841631315e-05, + "loss": 0.0582, + "step": 18812 + }, + { + "epoch": 2.46, + "learning_rate": 1.6212245270345724e-05, + "loss": 0.0629, + "step": 18813 + }, + { + "epoch": 2.46, + "learning_rate": 1.6211829681588526e-05, + "loss": 0.0557, + "step": 18814 + }, + { + "epoch": 2.46, + "learning_rate": 1.6211414075360888e-05, + "loss": 0.0729, + "step": 18815 + }, + { + "epoch": 2.46, + "learning_rate": 1.621099845166398e-05, + "loss": 0.0639, + "step": 18816 + }, + { + "epoch": 2.46, + "learning_rate": 1.6210582810498967e-05, + "loss": 0.0587, + "step": 18817 + }, + { + "epoch": 2.47, + "learning_rate": 1.6210167151867023e-05, + "loss": 0.0616, + "step": 18818 + }, + { + "epoch": 2.47, + "learning_rate": 1.620975147576931e-05, + "loss": 0.0653, + "step": 18819 + }, + { + "epoch": 2.47, + "learning_rate": 1.620933578220701e-05, + "loss": 0.057, + "step": 18820 + }, + { + "epoch": 2.47, + "learning_rate": 1.6208920071181276e-05, + "loss": 0.0607, + "step": 18821 + }, + { + "epoch": 2.47, + "learning_rate": 1.6208504342693284e-05, + "loss": 0.0566, + "step": 18822 + }, + { + "epoch": 2.47, + "learning_rate": 1.620808859674421e-05, + "loss": 0.0713, + "step": 18823 + }, + { + "epoch": 2.47, + "learning_rate": 1.6207672833335216e-05, + "loss": 0.0691, + "step": 18824 + }, + { + "epoch": 2.47, + "learning_rate": 1.6207257052467473e-05, + "loss": 0.0598, + "step": 18825 + }, + { + "epoch": 2.47, + "learning_rate": 1.6206841254142147e-05, + "loss": 0.0599, + "step": 18826 + }, + { + "epoch": 2.47, + "learning_rate": 1.6206425438360414e-05, + "loss": 0.068, + "step": 18827 + }, + { + "epoch": 2.47, + "learning_rate": 1.6206009605123436e-05, + "loss": 0.0582, + "step": 18828 + }, + { + "epoch": 2.47, + "learning_rate": 1.620559375443239e-05, + "loss": 0.0545, + "step": 18829 + }, + { + "epoch": 2.47, + "learning_rate": 1.620517788628844e-05, + "loss": 0.0593, + "step": 18830 + }, + { + "epoch": 2.47, + "learning_rate": 1.6204762000692758e-05, + "loss": 0.058, + "step": 18831 + }, + { + "epoch": 2.47, + "learning_rate": 1.6204346097646513e-05, + "loss": 0.0583, + "step": 18832 + }, + { + "epoch": 2.47, + "learning_rate": 1.6203930177150877e-05, + "loss": 0.0659, + "step": 18833 + }, + { + "epoch": 2.47, + "learning_rate": 1.6203514239207017e-05, + "loss": 0.0595, + "step": 18834 + }, + { + "epoch": 2.47, + "learning_rate": 1.6203098283816104e-05, + "loss": 0.0586, + "step": 18835 + }, + { + "epoch": 2.47, + "learning_rate": 1.6202682310979304e-05, + "loss": 0.059, + "step": 18836 + }, + { + "epoch": 2.47, + "learning_rate": 1.6202266320697796e-05, + "loss": 0.0674, + "step": 18837 + }, + { + "epoch": 2.47, + "learning_rate": 1.620185031297274e-05, + "loss": 0.0601, + "step": 18838 + }, + { + "epoch": 2.47, + "learning_rate": 1.6201434287805312e-05, + "loss": 0.0553, + "step": 18839 + }, + { + "epoch": 2.47, + "learning_rate": 1.6201018245196677e-05, + "loss": 0.0566, + "step": 18840 + }, + { + "epoch": 2.47, + "learning_rate": 1.620060218514801e-05, + "loss": 0.0613, + "step": 18841 + }, + { + "epoch": 2.47, + "learning_rate": 1.6200186107660482e-05, + "loss": 0.0656, + "step": 18842 + }, + { + "epoch": 2.47, + "learning_rate": 1.619977001273526e-05, + "loss": 0.0597, + "step": 18843 + }, + { + "epoch": 2.47, + "learning_rate": 1.6199353900373516e-05, + "loss": 0.067, + "step": 18844 + }, + { + "epoch": 2.47, + "learning_rate": 1.6198937770576416e-05, + "loss": 0.0602, + "step": 18845 + }, + { + "epoch": 2.47, + "learning_rate": 1.6198521623345137e-05, + "loss": 0.0661, + "step": 18846 + }, + { + "epoch": 2.47, + "learning_rate": 1.6198105458680843e-05, + "loss": 0.0638, + "step": 18847 + }, + { + "epoch": 2.47, + "learning_rate": 1.619768927658471e-05, + "loss": 0.0626, + "step": 18848 + }, + { + "epoch": 2.47, + "learning_rate": 1.6197273077057905e-05, + "loss": 0.0606, + "step": 18849 + }, + { + "epoch": 2.47, + "learning_rate": 1.61968568601016e-05, + "loss": 0.07, + "step": 18850 + }, + { + "epoch": 2.47, + "learning_rate": 1.6196440625716963e-05, + "loss": 0.0685, + "step": 18851 + }, + { + "epoch": 2.47, + "learning_rate": 1.619602437390517e-05, + "loss": 0.0724, + "step": 18852 + }, + { + "epoch": 2.47, + "learning_rate": 1.6195608104667383e-05, + "loss": 0.0632, + "step": 18853 + }, + { + "epoch": 2.47, + "learning_rate": 1.619519181800478e-05, + "loss": 0.0598, + "step": 18854 + }, + { + "epoch": 2.47, + "learning_rate": 1.6194775513918536e-05, + "loss": 0.0576, + "step": 18855 + }, + { + "epoch": 2.47, + "learning_rate": 1.619435919240981e-05, + "loss": 0.0548, + "step": 18856 + }, + { + "epoch": 2.47, + "learning_rate": 1.6193942853479782e-05, + "loss": 0.0661, + "step": 18857 + }, + { + "epoch": 2.47, + "learning_rate": 1.6193526497129615e-05, + "loss": 0.0722, + "step": 18858 + }, + { + "epoch": 2.47, + "learning_rate": 1.6193110123360486e-05, + "loss": 0.0541, + "step": 18859 + }, + { + "epoch": 2.47, + "learning_rate": 1.6192693732173568e-05, + "loss": 0.0579, + "step": 18860 + }, + { + "epoch": 2.47, + "learning_rate": 1.6192277323570026e-05, + "loss": 0.0638, + "step": 18861 + }, + { + "epoch": 2.47, + "learning_rate": 1.6191860897551033e-05, + "loss": 0.0633, + "step": 18862 + }, + { + "epoch": 2.47, + "learning_rate": 1.619144445411776e-05, + "loss": 0.0564, + "step": 18863 + }, + { + "epoch": 2.47, + "learning_rate": 1.619102799327138e-05, + "loss": 0.0603, + "step": 18864 + }, + { + "epoch": 2.47, + "learning_rate": 1.6190611515013068e-05, + "loss": 0.0584, + "step": 18865 + }, + { + "epoch": 2.47, + "learning_rate": 1.619019501934399e-05, + "loss": 0.0653, + "step": 18866 + }, + { + "epoch": 2.47, + "learning_rate": 1.6189778506265313e-05, + "loss": 0.0699, + "step": 18867 + }, + { + "epoch": 2.47, + "learning_rate": 1.6189361975778216e-05, + "loss": 0.0633, + "step": 18868 + }, + { + "epoch": 2.47, + "learning_rate": 1.618894542788387e-05, + "loss": 0.0624, + "step": 18869 + }, + { + "epoch": 2.47, + "learning_rate": 1.618852886258344e-05, + "loss": 0.0636, + "step": 18870 + }, + { + "epoch": 2.47, + "learning_rate": 1.6188112279878105e-05, + "loss": 0.0697, + "step": 18871 + }, + { + "epoch": 2.47, + "learning_rate": 1.6187695679769037e-05, + "loss": 0.0661, + "step": 18872 + }, + { + "epoch": 2.47, + "learning_rate": 1.61872790622574e-05, + "loss": 0.0612, + "step": 18873 + }, + { + "epoch": 2.47, + "learning_rate": 1.618686242734437e-05, + "loss": 0.0738, + "step": 18874 + }, + { + "epoch": 2.47, + "learning_rate": 1.618644577503112e-05, + "loss": 0.066, + "step": 18875 + }, + { + "epoch": 2.47, + "learning_rate": 1.618602910531882e-05, + "loss": 0.0658, + "step": 18876 + }, + { + "epoch": 2.47, + "learning_rate": 1.6185612418208646e-05, + "loss": 0.0588, + "step": 18877 + }, + { + "epoch": 2.47, + "learning_rate": 1.6185195713701764e-05, + "loss": 0.0644, + "step": 18878 + }, + { + "epoch": 2.47, + "learning_rate": 1.6184778991799346e-05, + "loss": 0.0708, + "step": 18879 + }, + { + "epoch": 2.47, + "learning_rate": 1.618436225250257e-05, + "loss": 0.0587, + "step": 18880 + }, + { + "epoch": 2.47, + "learning_rate": 1.6183945495812605e-05, + "loss": 0.0582, + "step": 18881 + }, + { + "epoch": 2.47, + "learning_rate": 1.618352872173062e-05, + "loss": 0.0725, + "step": 18882 + }, + { + "epoch": 2.47, + "learning_rate": 1.618311193025779e-05, + "loss": 0.0574, + "step": 18883 + }, + { + "epoch": 2.47, + "learning_rate": 1.6182695121395288e-05, + "loss": 0.0647, + "step": 18884 + }, + { + "epoch": 2.47, + "learning_rate": 1.6182278295144286e-05, + "loss": 0.0513, + "step": 18885 + }, + { + "epoch": 2.47, + "learning_rate": 1.6181861451505954e-05, + "loss": 0.0547, + "step": 18886 + }, + { + "epoch": 2.47, + "learning_rate": 1.6181444590481464e-05, + "loss": 0.0567, + "step": 18887 + }, + { + "epoch": 2.47, + "learning_rate": 1.6181027712071996e-05, + "loss": 0.0646, + "step": 18888 + }, + { + "epoch": 2.47, + "learning_rate": 1.6180610816278712e-05, + "loss": 0.0599, + "step": 18889 + }, + { + "epoch": 2.47, + "learning_rate": 1.6180193903102787e-05, + "loss": 0.0536, + "step": 18890 + }, + { + "epoch": 2.47, + "learning_rate": 1.6179776972545403e-05, + "loss": 0.058, + "step": 18891 + }, + { + "epoch": 2.47, + "learning_rate": 1.617936002460772e-05, + "loss": 0.0712, + "step": 18892 + }, + { + "epoch": 2.47, + "learning_rate": 1.6178943059290917e-05, + "loss": 0.0575, + "step": 18893 + }, + { + "epoch": 2.47, + "learning_rate": 1.6178526076596165e-05, + "loss": 0.0606, + "step": 18894 + }, + { + "epoch": 2.48, + "learning_rate": 1.617810907652464e-05, + "loss": 0.0594, + "step": 18895 + }, + { + "epoch": 2.48, + "learning_rate": 1.617769205907751e-05, + "loss": 0.0585, + "step": 18896 + }, + { + "epoch": 2.48, + "learning_rate": 1.617727502425595e-05, + "loss": 0.0683, + "step": 18897 + }, + { + "epoch": 2.48, + "learning_rate": 1.6176857972061136e-05, + "loss": 0.0568, + "step": 18898 + }, + { + "epoch": 2.48, + "learning_rate": 1.6176440902494238e-05, + "loss": 0.0636, + "step": 18899 + }, + { + "epoch": 2.48, + "learning_rate": 1.6176023815556426e-05, + "loss": 0.0632, + "step": 18900 + }, + { + "epoch": 2.48, + "learning_rate": 1.6175606711248874e-05, + "loss": 0.0617, + "step": 18901 + }, + { + "epoch": 2.48, + "learning_rate": 1.617518958957276e-05, + "loss": 0.0518, + "step": 18902 + }, + { + "epoch": 2.48, + "learning_rate": 1.6174772450529255e-05, + "loss": 0.0611, + "step": 18903 + }, + { + "epoch": 2.48, + "learning_rate": 1.6174355294119532e-05, + "loss": 0.0631, + "step": 18904 + }, + { + "epoch": 2.48, + "learning_rate": 1.6173938120344762e-05, + "loss": 0.0528, + "step": 18905 + }, + { + "epoch": 2.48, + "learning_rate": 1.6173520929206116e-05, + "loss": 0.0611, + "step": 18906 + }, + { + "epoch": 2.48, + "learning_rate": 1.6173103720704774e-05, + "loss": 0.0691, + "step": 18907 + }, + { + "epoch": 2.48, + "learning_rate": 1.6172686494841906e-05, + "loss": 0.0652, + "step": 18908 + }, + { + "epoch": 2.48, + "learning_rate": 1.6172269251618692e-05, + "loss": 0.07, + "step": 18909 + }, + { + "epoch": 2.48, + "learning_rate": 1.6171851991036293e-05, + "loss": 0.0596, + "step": 18910 + }, + { + "epoch": 2.48, + "learning_rate": 1.6171434713095888e-05, + "loss": 0.0587, + "step": 18911 + }, + { + "epoch": 2.48, + "learning_rate": 1.6171017417798657e-05, + "loss": 0.0571, + "step": 18912 + }, + { + "epoch": 2.48, + "learning_rate": 1.6170600105145765e-05, + "loss": 0.0735, + "step": 18913 + }, + { + "epoch": 2.48, + "learning_rate": 1.6170182775138387e-05, + "loss": 0.0597, + "step": 18914 + }, + { + "epoch": 2.48, + "learning_rate": 1.6169765427777703e-05, + "loss": 0.0764, + "step": 18915 + }, + { + "epoch": 2.48, + "learning_rate": 1.616934806306488e-05, + "loss": 0.0627, + "step": 18916 + }, + { + "epoch": 2.48, + "learning_rate": 1.616893068100109e-05, + "loss": 0.0544, + "step": 18917 + }, + { + "epoch": 2.48, + "learning_rate": 1.6168513281587516e-05, + "loss": 0.0636, + "step": 18918 + }, + { + "epoch": 2.48, + "learning_rate": 1.6168095864825323e-05, + "loss": 0.0634, + "step": 18919 + }, + { + "epoch": 2.48, + "learning_rate": 1.6167678430715696e-05, + "loss": 0.0657, + "step": 18920 + }, + { + "epoch": 2.48, + "learning_rate": 1.6167260979259795e-05, + "loss": 0.0715, + "step": 18921 + }, + { + "epoch": 2.48, + "learning_rate": 1.6166843510458802e-05, + "loss": 0.0721, + "step": 18922 + }, + { + "epoch": 2.48, + "learning_rate": 1.6166426024313894e-05, + "loss": 0.0609, + "step": 18923 + }, + { + "epoch": 2.48, + "learning_rate": 1.6166008520826237e-05, + "loss": 0.0624, + "step": 18924 + }, + { + "epoch": 2.48, + "learning_rate": 1.616559099999701e-05, + "loss": 0.0651, + "step": 18925 + }, + { + "epoch": 2.48, + "learning_rate": 1.616517346182739e-05, + "loss": 0.07, + "step": 18926 + }, + { + "epoch": 2.48, + "learning_rate": 1.6164755906318547e-05, + "loss": 0.0676, + "step": 18927 + }, + { + "epoch": 2.48, + "learning_rate": 1.616433833347165e-05, + "loss": 0.0519, + "step": 18928 + }, + { + "epoch": 2.48, + "learning_rate": 1.6163920743287885e-05, + "loss": 0.064, + "step": 18929 + }, + { + "epoch": 2.48, + "learning_rate": 1.616350313576842e-05, + "loss": 0.062, + "step": 18930 + }, + { + "epoch": 2.48, + "learning_rate": 1.6163085510914434e-05, + "loss": 0.0654, + "step": 18931 + }, + { + "epoch": 2.48, + "learning_rate": 1.6162667868727094e-05, + "loss": 0.0698, + "step": 18932 + }, + { + "epoch": 2.48, + "learning_rate": 1.6162250209207583e-05, + "loss": 0.0569, + "step": 18933 + }, + { + "epoch": 2.48, + "learning_rate": 1.6161832532357066e-05, + "loss": 0.0538, + "step": 18934 + }, + { + "epoch": 2.48, + "learning_rate": 1.6161414838176726e-05, + "loss": 0.0645, + "step": 18935 + }, + { + "epoch": 2.48, + "learning_rate": 1.6160997126667735e-05, + "loss": 0.0578, + "step": 18936 + }, + { + "epoch": 2.48, + "learning_rate": 1.616057939783127e-05, + "loss": 0.0552, + "step": 18937 + }, + { + "epoch": 2.48, + "learning_rate": 1.61601616516685e-05, + "loss": 0.0601, + "step": 18938 + }, + { + "epoch": 2.48, + "learning_rate": 1.61597438881806e-05, + "loss": 0.0574, + "step": 18939 + }, + { + "epoch": 2.48, + "learning_rate": 1.6159326107368755e-05, + "loss": 0.0618, + "step": 18940 + }, + { + "epoch": 2.48, + "learning_rate": 1.6158908309234132e-05, + "loss": 0.0593, + "step": 18941 + }, + { + "epoch": 2.48, + "learning_rate": 1.6158490493777908e-05, + "loss": 0.0671, + "step": 18942 + }, + { + "epoch": 2.48, + "learning_rate": 1.6158072661001253e-05, + "loss": 0.0582, + "step": 18943 + }, + { + "epoch": 2.48, + "learning_rate": 1.615765481090535e-05, + "loss": 0.0538, + "step": 18944 + }, + { + "epoch": 2.48, + "learning_rate": 1.615723694349137e-05, + "loss": 0.0534, + "step": 18945 + }, + { + "epoch": 2.48, + "learning_rate": 1.6156819058760488e-05, + "loss": 0.0676, + "step": 18946 + }, + { + "epoch": 2.48, + "learning_rate": 1.6156401156713885e-05, + "loss": 0.0635, + "step": 18947 + }, + { + "epoch": 2.48, + "learning_rate": 1.615598323735273e-05, + "loss": 0.0645, + "step": 18948 + }, + { + "epoch": 2.48, + "learning_rate": 1.6155565300678196e-05, + "loss": 0.0617, + "step": 18949 + }, + { + "epoch": 2.48, + "learning_rate": 1.6155147346691462e-05, + "loss": 0.0561, + "step": 18950 + }, + { + "epoch": 2.48, + "learning_rate": 1.615472937539371e-05, + "loss": 0.0646, + "step": 18951 + }, + { + "epoch": 2.48, + "learning_rate": 1.6154311386786105e-05, + "loss": 0.0642, + "step": 18952 + }, + { + "epoch": 2.48, + "learning_rate": 1.6153893380869824e-05, + "loss": 0.0678, + "step": 18953 + }, + { + "epoch": 2.48, + "learning_rate": 1.615347535764605e-05, + "loss": 0.0566, + "step": 18954 + }, + { + "epoch": 2.48, + "learning_rate": 1.6153057317115954e-05, + "loss": 0.0532, + "step": 18955 + }, + { + "epoch": 2.48, + "learning_rate": 1.615263925928071e-05, + "loss": 0.0661, + "step": 18956 + }, + { + "epoch": 2.48, + "learning_rate": 1.61522211841415e-05, + "loss": 0.0582, + "step": 18957 + }, + { + "epoch": 2.48, + "learning_rate": 1.615180309169949e-05, + "loss": 0.0619, + "step": 18958 + }, + { + "epoch": 2.48, + "learning_rate": 1.6151384981955865e-05, + "loss": 0.0671, + "step": 18959 + }, + { + "epoch": 2.48, + "learning_rate": 1.61509668549118e-05, + "loss": 0.0533, + "step": 18960 + }, + { + "epoch": 2.48, + "learning_rate": 1.615054871056846e-05, + "loss": 0.0596, + "step": 18961 + }, + { + "epoch": 2.48, + "learning_rate": 1.6150130548927035e-05, + "loss": 0.0749, + "step": 18962 + }, + { + "epoch": 2.48, + "learning_rate": 1.6149712369988696e-05, + "loss": 0.0661, + "step": 18963 + }, + { + "epoch": 2.48, + "learning_rate": 1.6149294173754615e-05, + "loss": 0.06, + "step": 18964 + }, + { + "epoch": 2.48, + "learning_rate": 1.6148875960225977e-05, + "loss": 0.0715, + "step": 18965 + }, + { + "epoch": 2.48, + "learning_rate": 1.6148457729403946e-05, + "loss": 0.0613, + "step": 18966 + }, + { + "epoch": 2.48, + "learning_rate": 1.6148039481289706e-05, + "loss": 0.0624, + "step": 18967 + }, + { + "epoch": 2.48, + "learning_rate": 1.6147621215884438e-05, + "loss": 0.0652, + "step": 18968 + }, + { + "epoch": 2.48, + "learning_rate": 1.614720293318931e-05, + "loss": 0.0638, + "step": 18969 + }, + { + "epoch": 2.48, + "learning_rate": 1.6146784633205503e-05, + "loss": 0.0597, + "step": 18970 + }, + { + "epoch": 2.49, + "learning_rate": 1.6146366315934184e-05, + "loss": 0.0604, + "step": 18971 + }, + { + "epoch": 2.49, + "learning_rate": 1.6145947981376543e-05, + "loss": 0.0579, + "step": 18972 + }, + { + "epoch": 2.49, + "learning_rate": 1.614552962953375e-05, + "loss": 0.0608, + "step": 18973 + }, + { + "epoch": 2.49, + "learning_rate": 1.6145111260406984e-05, + "loss": 0.058, + "step": 18974 + }, + { + "epoch": 2.49, + "learning_rate": 1.6144692873997417e-05, + "loss": 0.0622, + "step": 18975 + }, + { + "epoch": 2.49, + "learning_rate": 1.614427447030623e-05, + "loss": 0.0597, + "step": 18976 + }, + { + "epoch": 2.49, + "learning_rate": 1.6143856049334593e-05, + "loss": 0.0587, + "step": 18977 + }, + { + "epoch": 2.49, + "learning_rate": 1.6143437611083693e-05, + "loss": 0.0698, + "step": 18978 + }, + { + "epoch": 2.49, + "learning_rate": 1.6143019155554704e-05, + "loss": 0.049, + "step": 18979 + }, + { + "epoch": 2.49, + "learning_rate": 1.6142600682748797e-05, + "loss": 0.0637, + "step": 18980 + }, + { + "epoch": 2.49, + "learning_rate": 1.6142182192667154e-05, + "loss": 0.0497, + "step": 18981 + }, + { + "epoch": 2.49, + "learning_rate": 1.614176368531095e-05, + "loss": 0.0514, + "step": 18982 + }, + { + "epoch": 2.49, + "learning_rate": 1.6141345160681365e-05, + "loss": 0.0627, + "step": 18983 + }, + { + "epoch": 2.49, + "learning_rate": 1.6140926618779568e-05, + "loss": 0.056, + "step": 18984 + }, + { + "epoch": 2.49, + "learning_rate": 1.6140508059606746e-05, + "loss": 0.0605, + "step": 18985 + }, + { + "epoch": 2.49, + "learning_rate": 1.6140089483164074e-05, + "loss": 0.0653, + "step": 18986 + }, + { + "epoch": 2.49, + "learning_rate": 1.6139670889452725e-05, + "loss": 0.0599, + "step": 18987 + }, + { + "epoch": 2.49, + "learning_rate": 1.6139252278473877e-05, + "loss": 0.0496, + "step": 18988 + }, + { + "epoch": 2.49, + "learning_rate": 1.613883365022871e-05, + "loss": 0.065, + "step": 18989 + }, + { + "epoch": 2.49, + "learning_rate": 1.6138415004718403e-05, + "loss": 0.0555, + "step": 18990 + }, + { + "epoch": 2.49, + "learning_rate": 1.6137996341944127e-05, + "loss": 0.0549, + "step": 18991 + }, + { + "epoch": 2.49, + "learning_rate": 1.6137577661907066e-05, + "loss": 0.0621, + "step": 18992 + }, + { + "epoch": 2.49, + "learning_rate": 1.6137158964608395e-05, + "loss": 0.0629, + "step": 18993 + }, + { + "epoch": 2.49, + "learning_rate": 1.613674025004929e-05, + "loss": 0.0696, + "step": 18994 + }, + { + "epoch": 2.49, + "learning_rate": 1.613632151823093e-05, + "loss": 0.0692, + "step": 18995 + }, + { + "epoch": 2.49, + "learning_rate": 1.6135902769154493e-05, + "loss": 0.0591, + "step": 18996 + }, + { + "epoch": 2.49, + "learning_rate": 1.6135484002821153e-05, + "loss": 0.0623, + "step": 18997 + }, + { + "epoch": 2.49, + "learning_rate": 1.6135065219232095e-05, + "loss": 0.0638, + "step": 18998 + }, + { + "epoch": 2.49, + "learning_rate": 1.6134646418388488e-05, + "loss": 0.0597, + "step": 18999 + }, + { + "epoch": 2.49, + "learning_rate": 1.613422760029152e-05, + "loss": 0.0684, + "step": 19000 + }, + { + "epoch": 2.49, + "learning_rate": 1.613380876494236e-05, + "loss": 0.0674, + "step": 19001 + }, + { + "epoch": 2.49, + "learning_rate": 1.613338991234219e-05, + "loss": 0.0657, + "step": 19002 + }, + { + "epoch": 2.49, + "learning_rate": 1.613297104249219e-05, + "loss": 0.0584, + "step": 19003 + }, + { + "epoch": 2.49, + "learning_rate": 1.613255215539353e-05, + "loss": 0.0571, + "step": 19004 + }, + { + "epoch": 2.49, + "learning_rate": 1.61321332510474e-05, + "loss": 0.0618, + "step": 19005 + }, + { + "epoch": 2.49, + "learning_rate": 1.6131714329454967e-05, + "loss": 0.0601, + "step": 19006 + }, + { + "epoch": 2.49, + "learning_rate": 1.6131295390617418e-05, + "loss": 0.07, + "step": 19007 + }, + { + "epoch": 2.49, + "learning_rate": 1.6130876434535928e-05, + "loss": 0.0622, + "step": 19008 + }, + { + "epoch": 2.49, + "learning_rate": 1.6130457461211672e-05, + "loss": 0.055, + "step": 19009 + }, + { + "epoch": 2.49, + "learning_rate": 1.613003847064583e-05, + "loss": 0.0589, + "step": 19010 + }, + { + "epoch": 2.49, + "learning_rate": 1.612961946283958e-05, + "loss": 0.0558, + "step": 19011 + }, + { + "epoch": 2.49, + "learning_rate": 1.6129200437794106e-05, + "loss": 0.0636, + "step": 19012 + }, + { + "epoch": 2.49, + "learning_rate": 1.612878139551058e-05, + "loss": 0.0588, + "step": 19013 + }, + { + "epoch": 2.49, + "learning_rate": 1.6128362335990183e-05, + "loss": 0.0515, + "step": 19014 + }, + { + "epoch": 2.49, + "learning_rate": 1.612794325923409e-05, + "loss": 0.0609, + "step": 19015 + }, + { + "epoch": 2.49, + "learning_rate": 1.612752416524349e-05, + "loss": 0.0602, + "step": 19016 + }, + { + "epoch": 2.49, + "learning_rate": 1.612710505401955e-05, + "loss": 0.0665, + "step": 19017 + }, + { + "epoch": 2.49, + "learning_rate": 1.6126685925563454e-05, + "loss": 0.0628, + "step": 19018 + }, + { + "epoch": 2.49, + "learning_rate": 1.612626677987638e-05, + "loss": 0.0575, + "step": 19019 + }, + { + "epoch": 2.49, + "learning_rate": 1.6125847616959506e-05, + "loss": 0.0671, + "step": 19020 + }, + { + "epoch": 2.49, + "learning_rate": 1.612542843681401e-05, + "loss": 0.0757, + "step": 19021 + }, + { + "epoch": 2.49, + "learning_rate": 1.6125009239441073e-05, + "loss": 0.0606, + "step": 19022 + }, + { + "epoch": 2.49, + "learning_rate": 1.612459002484188e-05, + "loss": 0.0834, + "step": 19023 + }, + { + "epoch": 2.49, + "learning_rate": 1.6124170793017597e-05, + "loss": 0.064, + "step": 19024 + }, + { + "epoch": 2.49, + "learning_rate": 1.6123751543969416e-05, + "loss": 0.0564, + "step": 19025 + }, + { + "epoch": 2.49, + "learning_rate": 1.61233322776985e-05, + "loss": 0.064, + "step": 19026 + }, + { + "epoch": 2.49, + "learning_rate": 1.6122912994206047e-05, + "loss": 0.0623, + "step": 19027 + }, + { + "epoch": 2.49, + "learning_rate": 1.6122493693493227e-05, + "loss": 0.0672, + "step": 19028 + }, + { + "epoch": 2.49, + "learning_rate": 1.6122074375561217e-05, + "loss": 0.0575, + "step": 19029 + }, + { + "epoch": 2.49, + "learning_rate": 1.61216550404112e-05, + "loss": 0.0641, + "step": 19030 + }, + { + "epoch": 2.49, + "learning_rate": 1.612123568804435e-05, + "loss": 0.058, + "step": 19031 + }, + { + "epoch": 2.49, + "learning_rate": 1.6120816318461857e-05, + "loss": 0.0583, + "step": 19032 + }, + { + "epoch": 2.49, + "learning_rate": 1.612039693166489e-05, + "loss": 0.0647, + "step": 19033 + }, + { + "epoch": 2.49, + "learning_rate": 1.6119977527654633e-05, + "loss": 0.0578, + "step": 19034 + }, + { + "epoch": 2.49, + "learning_rate": 1.6119558106432264e-05, + "loss": 0.0717, + "step": 19035 + }, + { + "epoch": 2.49, + "learning_rate": 1.6119138667998968e-05, + "loss": 0.0641, + "step": 19036 + }, + { + "epoch": 2.49, + "learning_rate": 1.6118719212355914e-05, + "loss": 0.0602, + "step": 19037 + }, + { + "epoch": 2.49, + "learning_rate": 1.6118299739504295e-05, + "loss": 0.0593, + "step": 19038 + }, + { + "epoch": 2.49, + "learning_rate": 1.611788024944528e-05, + "loss": 0.068, + "step": 19039 + }, + { + "epoch": 2.49, + "learning_rate": 1.6117460742180053e-05, + "loss": 0.0747, + "step": 19040 + }, + { + "epoch": 2.49, + "learning_rate": 1.6117041217709792e-05, + "loss": 0.0622, + "step": 19041 + }, + { + "epoch": 2.49, + "learning_rate": 1.611662167603568e-05, + "loss": 0.0509, + "step": 19042 + }, + { + "epoch": 2.49, + "learning_rate": 1.6116202117158896e-05, + "loss": 0.054, + "step": 19043 + }, + { + "epoch": 2.49, + "learning_rate": 1.6115782541080615e-05, + "loss": 0.0651, + "step": 19044 + }, + { + "epoch": 2.49, + "learning_rate": 1.6115362947802025e-05, + "loss": 0.057, + "step": 19045 + }, + { + "epoch": 2.49, + "learning_rate": 1.6114943337324303e-05, + "loss": 0.0655, + "step": 19046 + }, + { + "epoch": 2.5, + "learning_rate": 1.6114523709648627e-05, + "loss": 0.0536, + "step": 19047 + }, + { + "epoch": 2.5, + "learning_rate": 1.611410406477618e-05, + "loss": 0.0629, + "step": 19048 + }, + { + "epoch": 2.5, + "learning_rate": 1.611368440270814e-05, + "loss": 0.0542, + "step": 19049 + }, + { + "epoch": 2.5, + "learning_rate": 1.611326472344569e-05, + "loss": 0.0599, + "step": 19050 + }, + { + "epoch": 2.5, + "learning_rate": 1.6112845026990006e-05, + "loss": 0.0671, + "step": 19051 + }, + { + "epoch": 2.5, + "learning_rate": 1.611242531334227e-05, + "loss": 0.0598, + "step": 19052 + }, + { + "epoch": 2.5, + "learning_rate": 1.6112005582503668e-05, + "loss": 0.0604, + "step": 19053 + }, + { + "epoch": 2.5, + "learning_rate": 1.611158583447537e-05, + "loss": 0.0611, + "step": 19054 + }, + { + "epoch": 2.5, + "learning_rate": 1.6111166069258567e-05, + "loss": 0.0568, + "step": 19055 + }, + { + "epoch": 2.5, + "learning_rate": 1.6110746286854434e-05, + "loss": 0.063, + "step": 19056 + }, + { + "epoch": 2.5, + "learning_rate": 1.611032648726415e-05, + "loss": 0.0585, + "step": 19057 + }, + { + "epoch": 2.5, + "learning_rate": 1.6109906670488902e-05, + "loss": 0.0546, + "step": 19058 + }, + { + "epoch": 2.5, + "learning_rate": 1.610948683652986e-05, + "loss": 0.0594, + "step": 19059 + }, + { + "epoch": 2.5, + "learning_rate": 1.610906698538822e-05, + "loss": 0.0642, + "step": 19060 + }, + { + "epoch": 2.5, + "learning_rate": 1.610864711706515e-05, + "loss": 0.0622, + "step": 19061 + }, + { + "epoch": 2.5, + "learning_rate": 1.6108227231561837e-05, + "loss": 0.0579, + "step": 19062 + }, + { + "epoch": 2.5, + "learning_rate": 1.6107807328879455e-05, + "loss": 0.0559, + "step": 19063 + }, + { + "epoch": 2.5, + "learning_rate": 1.6107387409019195e-05, + "loss": 0.0594, + "step": 19064 + }, + { + "epoch": 2.5, + "learning_rate": 1.610696747198223e-05, + "loss": 0.0503, + "step": 19065 + }, + { + "epoch": 2.5, + "learning_rate": 1.6106547517769747e-05, + "loss": 0.0626, + "step": 19066 + }, + { + "epoch": 2.5, + "learning_rate": 1.6106127546382925e-05, + "loss": 0.0687, + "step": 19067 + }, + { + "epoch": 2.5, + "learning_rate": 1.6105707557822946e-05, + "loss": 0.0657, + "step": 19068 + }, + { + "epoch": 2.5, + "learning_rate": 1.6105287552090984e-05, + "loss": 0.0694, + "step": 19069 + }, + { + "epoch": 2.5, + "learning_rate": 1.6104867529188227e-05, + "loss": 0.0619, + "step": 19070 + }, + { + "epoch": 2.5, + "learning_rate": 1.6104447489115858e-05, + "loss": 0.0577, + "step": 19071 + }, + { + "epoch": 2.5, + "learning_rate": 1.6104027431875057e-05, + "loss": 0.0548, + "step": 19072 + }, + { + "epoch": 2.5, + "learning_rate": 1.6103607357467e-05, + "loss": 0.0734, + "step": 19073 + }, + { + "epoch": 2.5, + "learning_rate": 1.6103187265892874e-05, + "loss": 0.0668, + "step": 19074 + }, + { + "epoch": 2.5, + "learning_rate": 1.610276715715386e-05, + "loss": 0.0582, + "step": 19075 + }, + { + "epoch": 2.5, + "learning_rate": 1.6102347031251137e-05, + "loss": 0.0589, + "step": 19076 + }, + { + "epoch": 2.5, + "learning_rate": 1.6101926888185887e-05, + "loss": 0.0551, + "step": 19077 + }, + { + "epoch": 2.5, + "learning_rate": 1.6101506727959293e-05, + "loss": 0.0649, + "step": 19078 + }, + { + "epoch": 2.5, + "learning_rate": 1.6101086550572536e-05, + "loss": 0.0622, + "step": 19079 + }, + { + "epoch": 2.5, + "learning_rate": 1.61006663560268e-05, + "loss": 0.063, + "step": 19080 + }, + { + "epoch": 2.5, + "learning_rate": 1.6100246144323262e-05, + "loss": 0.0714, + "step": 19081 + }, + { + "epoch": 2.5, + "learning_rate": 1.609982591546311e-05, + "loss": 0.0714, + "step": 19082 + }, + { + "epoch": 2.5, + "learning_rate": 1.609940566944752e-05, + "loss": 0.0599, + "step": 19083 + }, + { + "epoch": 2.5, + "learning_rate": 1.6098985406277678e-05, + "loss": 0.0667, + "step": 19084 + }, + { + "epoch": 2.5, + "learning_rate": 1.609856512595476e-05, + "loss": 0.0506, + "step": 19085 + }, + { + "epoch": 2.5, + "learning_rate": 1.6098144828479956e-05, + "loss": 0.0488, + "step": 19086 + }, + { + "epoch": 2.5, + "learning_rate": 1.6097724513854444e-05, + "loss": 0.0657, + "step": 19087 + }, + { + "epoch": 2.5, + "learning_rate": 1.609730418207941e-05, + "loss": 0.0574, + "step": 19088 + }, + { + "epoch": 2.5, + "learning_rate": 1.609688383315603e-05, + "loss": 0.0523, + "step": 19089 + }, + { + "epoch": 2.5, + "learning_rate": 1.6096463467085483e-05, + "loss": 0.0593, + "step": 19090 + }, + { + "epoch": 2.5, + "learning_rate": 1.6096043083868965e-05, + "loss": 0.0539, + "step": 19091 + }, + { + "epoch": 2.5, + "learning_rate": 1.6095622683507645e-05, + "loss": 0.0639, + "step": 19092 + }, + { + "epoch": 2.5, + "learning_rate": 1.6095202266002715e-05, + "loss": 0.0682, + "step": 19093 + }, + { + "epoch": 2.5, + "learning_rate": 1.6094781831355352e-05, + "loss": 0.0654, + "step": 19094 + }, + { + "epoch": 2.5, + "learning_rate": 1.609436137956674e-05, + "loss": 0.066, + "step": 19095 + }, + { + "epoch": 2.5, + "learning_rate": 1.609394091063806e-05, + "loss": 0.0622, + "step": 19096 + }, + { + "epoch": 2.5, + "learning_rate": 1.6093520424570493e-05, + "loss": 0.0568, + "step": 19097 + }, + { + "epoch": 2.5, + "learning_rate": 1.609309992136523e-05, + "loss": 0.0688, + "step": 19098 + }, + { + "epoch": 2.5, + "learning_rate": 1.6092679401023444e-05, + "loss": 0.0693, + "step": 19099 + }, + { + "epoch": 2.5, + "learning_rate": 1.6092258863546322e-05, + "loss": 0.0703, + "step": 19100 + }, + { + "epoch": 2.5, + "learning_rate": 1.6091838308935052e-05, + "loss": 0.0746, + "step": 19101 + }, + { + "epoch": 2.5, + "learning_rate": 1.6091417737190802e-05, + "loss": 0.0644, + "step": 19102 + }, + { + "epoch": 2.5, + "learning_rate": 1.609099714831477e-05, + "loss": 0.0629, + "step": 19103 + }, + { + "epoch": 2.5, + "learning_rate": 1.609057654230813e-05, + "loss": 0.0597, + "step": 19104 + }, + { + "epoch": 2.5, + "learning_rate": 1.609015591917207e-05, + "loss": 0.067, + "step": 19105 + }, + { + "epoch": 2.5, + "learning_rate": 1.6089735278907772e-05, + "loss": 0.0735, + "step": 19106 + }, + { + "epoch": 2.5, + "learning_rate": 1.6089314621516415e-05, + "loss": 0.0561, + "step": 19107 + }, + { + "epoch": 2.5, + "learning_rate": 1.6088893946999188e-05, + "loss": 0.0551, + "step": 19108 + }, + { + "epoch": 2.5, + "learning_rate": 1.6088473255357265e-05, + "loss": 0.0621, + "step": 19109 + }, + { + "epoch": 2.5, + "learning_rate": 1.6088052546591843e-05, + "loss": 0.0641, + "step": 19110 + }, + { + "epoch": 2.5, + "learning_rate": 1.6087631820704093e-05, + "loss": 0.0616, + "step": 19111 + }, + { + "epoch": 2.5, + "learning_rate": 1.6087211077695202e-05, + "loss": 0.0575, + "step": 19112 + }, + { + "epoch": 2.5, + "learning_rate": 1.6086790317566353e-05, + "loss": 0.0701, + "step": 19113 + }, + { + "epoch": 2.5, + "learning_rate": 1.6086369540318733e-05, + "loss": 0.0541, + "step": 19114 + }, + { + "epoch": 2.5, + "learning_rate": 1.6085948745953523e-05, + "loss": 0.0671, + "step": 19115 + }, + { + "epoch": 2.5, + "learning_rate": 1.608552793447191e-05, + "loss": 0.0675, + "step": 19116 + }, + { + "epoch": 2.5, + "learning_rate": 1.6085107105875065e-05, + "loss": 0.0592, + "step": 19117 + }, + { + "epoch": 2.5, + "learning_rate": 1.6084686260164185e-05, + "loss": 0.067, + "step": 19118 + }, + { + "epoch": 2.5, + "learning_rate": 1.6084265397340447e-05, + "loss": 0.0599, + "step": 19119 + }, + { + "epoch": 2.5, + "learning_rate": 1.608384451740504e-05, + "loss": 0.0575, + "step": 19120 + }, + { + "epoch": 2.5, + "learning_rate": 1.6083423620359143e-05, + "loss": 0.0551, + "step": 19121 + }, + { + "epoch": 2.5, + "learning_rate": 1.6083002706203942e-05, + "loss": 0.0579, + "step": 19122 + }, + { + "epoch": 2.5, + "learning_rate": 1.6082581774940617e-05, + "loss": 0.0693, + "step": 19123 + }, + { + "epoch": 2.51, + "learning_rate": 1.6082160826570356e-05, + "loss": 0.0651, + "step": 19124 + }, + { + "epoch": 2.51, + "learning_rate": 1.608173986109434e-05, + "loss": 0.0551, + "step": 19125 + }, + { + "epoch": 2.51, + "learning_rate": 1.6081318878513757e-05, + "loss": 0.058, + "step": 19126 + }, + { + "epoch": 2.51, + "learning_rate": 1.6080897878829787e-05, + "loss": 0.057, + "step": 19127 + }, + { + "epoch": 2.51, + "learning_rate": 1.608047686204362e-05, + "loss": 0.0566, + "step": 19128 + }, + { + "epoch": 2.51, + "learning_rate": 1.6080055828156427e-05, + "loss": 0.0578, + "step": 19129 + }, + { + "epoch": 2.51, + "learning_rate": 1.6079634777169406e-05, + "loss": 0.0644, + "step": 19130 + }, + { + "epoch": 2.51, + "learning_rate": 1.6079213709083737e-05, + "loss": 0.0587, + "step": 19131 + }, + { + "epoch": 2.51, + "learning_rate": 1.60787926239006e-05, + "loss": 0.0598, + "step": 19132 + }, + { + "epoch": 2.51, + "learning_rate": 1.6078371521621184e-05, + "loss": 0.0747, + "step": 19133 + }, + { + "epoch": 2.51, + "learning_rate": 1.6077950402246674e-05, + "loss": 0.051, + "step": 19134 + }, + { + "epoch": 2.51, + "learning_rate": 1.607752926577825e-05, + "loss": 0.065, + "step": 19135 + }, + { + "epoch": 2.51, + "learning_rate": 1.6077108112217097e-05, + "loss": 0.0677, + "step": 19136 + }, + { + "epoch": 2.51, + "learning_rate": 1.6076686941564402e-05, + "loss": 0.0579, + "step": 19137 + }, + { + "epoch": 2.51, + "learning_rate": 1.607626575382135e-05, + "loss": 0.0651, + "step": 19138 + }, + { + "epoch": 2.51, + "learning_rate": 1.6075844548989124e-05, + "loss": 0.0543, + "step": 19139 + }, + { + "epoch": 2.51, + "learning_rate": 1.607542332706891e-05, + "loss": 0.0631, + "step": 19140 + }, + { + "epoch": 2.51, + "learning_rate": 1.607500208806189e-05, + "loss": 0.055, + "step": 19141 + }, + { + "epoch": 2.51, + "learning_rate": 1.6074580831969248e-05, + "loss": 0.0666, + "step": 19142 + }, + { + "epoch": 2.51, + "learning_rate": 1.6074159558792173e-05, + "loss": 0.0507, + "step": 19143 + }, + { + "epoch": 2.51, + "learning_rate": 1.6073738268531845e-05, + "loss": 0.0571, + "step": 19144 + }, + { + "epoch": 2.51, + "learning_rate": 1.6073316961189456e-05, + "loss": 0.0627, + "step": 19145 + }, + { + "epoch": 2.51, + "learning_rate": 1.6072895636766183e-05, + "loss": 0.0619, + "step": 19146 + }, + { + "epoch": 2.51, + "learning_rate": 1.6072474295263215e-05, + "loss": 0.0611, + "step": 19147 + }, + { + "epoch": 2.51, + "learning_rate": 1.607205293668174e-05, + "loss": 0.0684, + "step": 19148 + }, + { + "epoch": 2.51, + "learning_rate": 1.6071631561022937e-05, + "loss": 0.0638, + "step": 19149 + }, + { + "epoch": 2.51, + "learning_rate": 1.607121016828799e-05, + "loss": 0.0682, + "step": 19150 + }, + { + "epoch": 2.51, + "learning_rate": 1.6070788758478095e-05, + "loss": 0.0577, + "step": 19151 + }, + { + "epoch": 2.51, + "learning_rate": 1.6070367331594423e-05, + "loss": 0.0558, + "step": 19152 + }, + { + "epoch": 2.51, + "learning_rate": 1.606994588763817e-05, + "loss": 0.0567, + "step": 19153 + }, + { + "epoch": 2.51, + "learning_rate": 1.6069524426610517e-05, + "loss": 0.0568, + "step": 19154 + }, + { + "epoch": 2.51, + "learning_rate": 1.6069102948512648e-05, + "loss": 0.0576, + "step": 19155 + }, + { + "epoch": 2.51, + "learning_rate": 1.606868145334575e-05, + "loss": 0.0634, + "step": 19156 + }, + { + "epoch": 2.51, + "learning_rate": 1.606825994111101e-05, + "loss": 0.0595, + "step": 19157 + }, + { + "epoch": 2.51, + "learning_rate": 1.6067838411809614e-05, + "loss": 0.0639, + "step": 19158 + }, + { + "epoch": 2.51, + "learning_rate": 1.6067416865442743e-05, + "loss": 0.0649, + "step": 19159 + }, + { + "epoch": 2.51, + "learning_rate": 1.6066995302011586e-05, + "loss": 0.0572, + "step": 19160 + }, + { + "epoch": 2.51, + "learning_rate": 1.6066573721517327e-05, + "loss": 0.0628, + "step": 19161 + }, + { + "epoch": 2.51, + "learning_rate": 1.606615212396115e-05, + "loss": 0.0596, + "step": 19162 + }, + { + "epoch": 2.51, + "learning_rate": 1.606573050934425e-05, + "loss": 0.0604, + "step": 19163 + }, + { + "epoch": 2.51, + "learning_rate": 1.6065308877667803e-05, + "loss": 0.0585, + "step": 19164 + }, + { + "epoch": 2.51, + "learning_rate": 1.6064887228932996e-05, + "loss": 0.061, + "step": 19165 + }, + { + "epoch": 2.51, + "learning_rate": 1.606446556314102e-05, + "loss": 0.061, + "step": 19166 + }, + { + "epoch": 2.51, + "learning_rate": 1.6064043880293053e-05, + "loss": 0.0572, + "step": 19167 + }, + { + "epoch": 2.51, + "learning_rate": 1.606362218039029e-05, + "loss": 0.0654, + "step": 19168 + }, + { + "epoch": 2.51, + "learning_rate": 1.6063200463433906e-05, + "loss": 0.0665, + "step": 19169 + }, + { + "epoch": 2.51, + "learning_rate": 1.60627787294251e-05, + "loss": 0.0518, + "step": 19170 + }, + { + "epoch": 2.51, + "learning_rate": 1.606235697836505e-05, + "loss": 0.0636, + "step": 19171 + }, + { + "epoch": 2.51, + "learning_rate": 1.6061935210254945e-05, + "loss": 0.0576, + "step": 19172 + }, + { + "epoch": 2.51, + "learning_rate": 1.6061513425095964e-05, + "loss": 0.0539, + "step": 19173 + }, + { + "epoch": 2.51, + "learning_rate": 1.6061091622889303e-05, + "loss": 0.0729, + "step": 19174 + }, + { + "epoch": 2.51, + "learning_rate": 1.6060669803636148e-05, + "loss": 0.0657, + "step": 19175 + }, + { + "epoch": 2.51, + "learning_rate": 1.606024796733768e-05, + "loss": 0.0601, + "step": 19176 + }, + { + "epoch": 2.51, + "learning_rate": 1.6059826113995086e-05, + "loss": 0.0593, + "step": 19177 + }, + { + "epoch": 2.51, + "learning_rate": 1.6059404243609554e-05, + "loss": 0.0582, + "step": 19178 + }, + { + "epoch": 2.51, + "learning_rate": 1.605898235618227e-05, + "loss": 0.0545, + "step": 19179 + }, + { + "epoch": 2.51, + "learning_rate": 1.6058560451714424e-05, + "loss": 0.0681, + "step": 19180 + }, + { + "epoch": 2.51, + "learning_rate": 1.6058138530207194e-05, + "loss": 0.0647, + "step": 19181 + }, + { + "epoch": 2.51, + "learning_rate": 1.6057716591661778e-05, + "loss": 0.0591, + "step": 19182 + }, + { + "epoch": 2.51, + "learning_rate": 1.605729463607935e-05, + "loss": 0.0593, + "step": 19183 + }, + { + "epoch": 2.51, + "learning_rate": 1.6056872663461106e-05, + "loss": 0.0571, + "step": 19184 + }, + { + "epoch": 2.51, + "learning_rate": 1.6056450673808233e-05, + "loss": 0.0622, + "step": 19185 + }, + { + "epoch": 2.51, + "learning_rate": 1.6056028667121913e-05, + "loss": 0.0612, + "step": 19186 + }, + { + "epoch": 2.51, + "learning_rate": 1.6055606643403336e-05, + "loss": 0.0527, + "step": 19187 + }, + { + "epoch": 2.51, + "learning_rate": 1.6055184602653687e-05, + "loss": 0.07, + "step": 19188 + }, + { + "epoch": 2.51, + "learning_rate": 1.605476254487415e-05, + "loss": 0.0564, + "step": 19189 + }, + { + "epoch": 2.51, + "learning_rate": 1.605434047006592e-05, + "loss": 0.0743, + "step": 19190 + }, + { + "epoch": 2.51, + "learning_rate": 1.605391837823018e-05, + "loss": 0.0636, + "step": 19191 + }, + { + "epoch": 2.51, + "learning_rate": 1.6053496269368116e-05, + "loss": 0.076, + "step": 19192 + }, + { + "epoch": 2.51, + "learning_rate": 1.6053074143480916e-05, + "loss": 0.0609, + "step": 19193 + }, + { + "epoch": 2.51, + "learning_rate": 1.6052652000569767e-05, + "loss": 0.0495, + "step": 19194 + }, + { + "epoch": 2.51, + "learning_rate": 1.6052229840635858e-05, + "loss": 0.0577, + "step": 19195 + }, + { + "epoch": 2.51, + "learning_rate": 1.605180766368037e-05, + "loss": 0.0545, + "step": 19196 + }, + { + "epoch": 2.51, + "learning_rate": 1.6051385469704502e-05, + "loss": 0.0707, + "step": 19197 + }, + { + "epoch": 2.51, + "learning_rate": 1.6050963258709434e-05, + "loss": 0.0626, + "step": 19198 + }, + { + "epoch": 2.51, + "learning_rate": 1.6050541030696352e-05, + "loss": 0.0671, + "step": 19199 + }, + { + "epoch": 2.52, + "learning_rate": 1.6050118785666444e-05, + "loss": 0.0557, + "step": 19200 + }, + { + "epoch": 2.52, + "learning_rate": 1.60496965236209e-05, + "loss": 0.0627, + "step": 19201 + }, + { + "epoch": 2.52, + "learning_rate": 1.6049274244560907e-05, + "loss": 0.0697, + "step": 19202 + }, + { + "epoch": 2.52, + "learning_rate": 1.604885194848765e-05, + "loss": 0.0579, + "step": 19203 + }, + { + "epoch": 2.52, + "learning_rate": 1.6048429635402324e-05, + "loss": 0.0595, + "step": 19204 + }, + { + "epoch": 2.52, + "learning_rate": 1.604800730530611e-05, + "loss": 0.0456, + "step": 19205 + }, + { + "epoch": 2.52, + "learning_rate": 1.6047584958200196e-05, + "loss": 0.0597, + "step": 19206 + }, + { + "epoch": 2.52, + "learning_rate": 1.6047162594085772e-05, + "loss": 0.0613, + "step": 19207 + }, + { + "epoch": 2.52, + "learning_rate": 1.6046740212964025e-05, + "loss": 0.0569, + "step": 19208 + }, + { + "epoch": 2.52, + "learning_rate": 1.6046317814836143e-05, + "loss": 0.0543, + "step": 19209 + }, + { + "epoch": 2.52, + "learning_rate": 1.6045895399703316e-05, + "loss": 0.0595, + "step": 19210 + }, + { + "epoch": 2.52, + "learning_rate": 1.604547296756673e-05, + "loss": 0.063, + "step": 19211 + }, + { + "epoch": 2.52, + "learning_rate": 1.604505051842757e-05, + "loss": 0.0628, + "step": 19212 + }, + { + "epoch": 2.52, + "learning_rate": 1.6044628052287027e-05, + "loss": 0.0564, + "step": 19213 + }, + { + "epoch": 2.52, + "learning_rate": 1.6044205569146294e-05, + "loss": 0.0511, + "step": 19214 + }, + { + "epoch": 2.52, + "learning_rate": 1.6043783069006552e-05, + "loss": 0.0491, + "step": 19215 + }, + { + "epoch": 2.52, + "learning_rate": 1.6043360551868992e-05, + "loss": 0.0627, + "step": 19216 + }, + { + "epoch": 2.52, + "learning_rate": 1.6042938017734802e-05, + "loss": 0.0556, + "step": 19217 + }, + { + "epoch": 2.52, + "learning_rate": 1.604251546660517e-05, + "loss": 0.056, + "step": 19218 + }, + { + "epoch": 2.52, + "learning_rate": 1.604209289848129e-05, + "loss": 0.0609, + "step": 19219 + }, + { + "epoch": 2.52, + "learning_rate": 1.6041670313364342e-05, + "loss": 0.0699, + "step": 19220 + }, + { + "epoch": 2.52, + "learning_rate": 1.6041247711255517e-05, + "loss": 0.0691, + "step": 19221 + }, + { + "epoch": 2.52, + "learning_rate": 1.6040825092156005e-05, + "loss": 0.0652, + "step": 19222 + }, + { + "epoch": 2.52, + "learning_rate": 1.6040402456066996e-05, + "loss": 0.0621, + "step": 19223 + }, + { + "epoch": 2.52, + "learning_rate": 1.6039979802989675e-05, + "loss": 0.0672, + "step": 19224 + }, + { + "epoch": 2.52, + "learning_rate": 1.6039557132925233e-05, + "loss": 0.075, + "step": 19225 + }, + { + "epoch": 2.52, + "learning_rate": 1.6039134445874857e-05, + "loss": 0.0593, + "step": 19226 + }, + { + "epoch": 2.52, + "learning_rate": 1.603871174183974e-05, + "loss": 0.0697, + "step": 19227 + }, + { + "epoch": 2.52, + "learning_rate": 1.6038289020821064e-05, + "loss": 0.0694, + "step": 19228 + }, + { + "epoch": 2.52, + "learning_rate": 1.6037866282820023e-05, + "loss": 0.0497, + "step": 19229 + }, + { + "epoch": 2.52, + "learning_rate": 1.6037443527837805e-05, + "loss": 0.0608, + "step": 19230 + }, + { + "epoch": 2.52, + "learning_rate": 1.6037020755875597e-05, + "loss": 0.0607, + "step": 19231 + }, + { + "epoch": 2.52, + "learning_rate": 1.6036597966934594e-05, + "loss": 0.0629, + "step": 19232 + }, + { + "epoch": 2.52, + "learning_rate": 1.6036175161015975e-05, + "loss": 0.0625, + "step": 19233 + }, + { + "epoch": 2.52, + "learning_rate": 1.603575233812094e-05, + "loss": 0.0538, + "step": 19234 + }, + { + "epoch": 2.52, + "learning_rate": 1.603532949825067e-05, + "loss": 0.0679, + "step": 19235 + }, + { + "epoch": 2.52, + "learning_rate": 1.603490664140636e-05, + "loss": 0.0493, + "step": 19236 + }, + { + "epoch": 2.52, + "learning_rate": 1.6034483767589192e-05, + "loss": 0.0501, + "step": 19237 + }, + { + "epoch": 2.52, + "learning_rate": 1.6034060876800365e-05, + "loss": 0.0656, + "step": 19238 + }, + { + "epoch": 2.52, + "learning_rate": 1.6033637969041057e-05, + "loss": 0.0628, + "step": 19239 + }, + { + "epoch": 2.52, + "learning_rate": 1.6033215044312466e-05, + "loss": 0.0516, + "step": 19240 + }, + { + "epoch": 2.52, + "learning_rate": 1.6032792102615783e-05, + "loss": 0.0617, + "step": 19241 + }, + { + "epoch": 2.52, + "learning_rate": 1.603236914395219e-05, + "loss": 0.0654, + "step": 19242 + }, + { + "epoch": 2.52, + "learning_rate": 1.603194616832288e-05, + "loss": 0.077, + "step": 19243 + }, + { + "epoch": 2.52, + "learning_rate": 1.603152317572904e-05, + "loss": 0.0596, + "step": 19244 + }, + { + "epoch": 2.52, + "learning_rate": 1.6031100166171865e-05, + "loss": 0.0698, + "step": 19245 + }, + { + "epoch": 2.52, + "learning_rate": 1.6030677139652544e-05, + "loss": 0.0567, + "step": 19246 + }, + { + "epoch": 2.52, + "learning_rate": 1.6030254096172262e-05, + "loss": 0.0602, + "step": 19247 + }, + { + "epoch": 2.52, + "learning_rate": 1.6029831035732212e-05, + "loss": 0.0668, + "step": 19248 + }, + { + "epoch": 2.52, + "learning_rate": 1.6029407958333583e-05, + "loss": 0.06, + "step": 19249 + }, + { + "epoch": 2.52, + "learning_rate": 1.6028984863977565e-05, + "loss": 0.0572, + "step": 19250 + }, + { + "epoch": 2.52, + "learning_rate": 1.602856175266535e-05, + "loss": 0.0628, + "step": 19251 + }, + { + "epoch": 2.52, + "learning_rate": 1.6028138624398127e-05, + "loss": 0.0639, + "step": 19252 + }, + { + "epoch": 2.52, + "learning_rate": 1.6027715479177082e-05, + "loss": 0.0669, + "step": 19253 + }, + { + "epoch": 2.52, + "learning_rate": 1.6027292317003408e-05, + "loss": 0.0608, + "step": 19254 + }, + { + "epoch": 2.52, + "learning_rate": 1.6026869137878298e-05, + "loss": 0.0533, + "step": 19255 + }, + { + "epoch": 2.52, + "learning_rate": 1.6026445941802937e-05, + "loss": 0.0612, + "step": 19256 + }, + { + "epoch": 2.52, + "learning_rate": 1.602602272877852e-05, + "loss": 0.0651, + "step": 19257 + }, + { + "epoch": 2.52, + "learning_rate": 1.6025599498806236e-05, + "loss": 0.0608, + "step": 19258 + }, + { + "epoch": 2.52, + "learning_rate": 1.6025176251887268e-05, + "loss": 0.0542, + "step": 19259 + }, + { + "epoch": 2.52, + "learning_rate": 1.602475298802282e-05, + "loss": 0.0596, + "step": 19260 + }, + { + "epoch": 2.52, + "learning_rate": 1.602432970721407e-05, + "loss": 0.0571, + "step": 19261 + }, + { + "epoch": 2.52, + "learning_rate": 1.6023906409462216e-05, + "loss": 0.0629, + "step": 19262 + }, + { + "epoch": 2.52, + "learning_rate": 1.6023483094768445e-05, + "loss": 0.0704, + "step": 19263 + }, + { + "epoch": 2.52, + "learning_rate": 1.602305976313395e-05, + "loss": 0.0538, + "step": 19264 + }, + { + "epoch": 2.52, + "learning_rate": 1.6022636414559917e-05, + "loss": 0.0559, + "step": 19265 + }, + { + "epoch": 2.52, + "learning_rate": 1.602221304904754e-05, + "loss": 0.0719, + "step": 19266 + }, + { + "epoch": 2.52, + "learning_rate": 1.6021789666598015e-05, + "loss": 0.0544, + "step": 19267 + }, + { + "epoch": 2.52, + "learning_rate": 1.6021366267212524e-05, + "loss": 0.063, + "step": 19268 + }, + { + "epoch": 2.52, + "learning_rate": 1.602094285089226e-05, + "loss": 0.0718, + "step": 19269 + }, + { + "epoch": 2.52, + "learning_rate": 1.6020519417638416e-05, + "loss": 0.0524, + "step": 19270 + }, + { + "epoch": 2.52, + "learning_rate": 1.602009596745218e-05, + "loss": 0.0635, + "step": 19271 + }, + { + "epoch": 2.52, + "learning_rate": 1.6019672500334747e-05, + "loss": 0.069, + "step": 19272 + }, + { + "epoch": 2.52, + "learning_rate": 1.6019249016287303e-05, + "loss": 0.0738, + "step": 19273 + }, + { + "epoch": 2.52, + "learning_rate": 1.6018825515311045e-05, + "loss": 0.0572, + "step": 19274 + }, + { + "epoch": 2.52, + "learning_rate": 1.601840199740716e-05, + "loss": 0.0661, + "step": 19275 + }, + { + "epoch": 2.53, + "learning_rate": 1.601797846257684e-05, + "loss": 0.0519, + "step": 19276 + }, + { + "epoch": 2.53, + "learning_rate": 1.6017554910821272e-05, + "loss": 0.0611, + "step": 19277 + }, + { + "epoch": 2.53, + "learning_rate": 1.6017131342141653e-05, + "loss": 0.0589, + "step": 19278 + }, + { + "epoch": 2.53, + "learning_rate": 1.6016707756539177e-05, + "loss": 0.0661, + "step": 19279 + }, + { + "epoch": 2.53, + "learning_rate": 1.6016284154015026e-05, + "loss": 0.0707, + "step": 19280 + }, + { + "epoch": 2.53, + "learning_rate": 1.60158605345704e-05, + "loss": 0.0566, + "step": 19281 + }, + { + "epoch": 2.53, + "learning_rate": 1.6015436898206484e-05, + "loss": 0.0572, + "step": 19282 + }, + { + "epoch": 2.53, + "learning_rate": 1.601501324492447e-05, + "loss": 0.0541, + "step": 19283 + }, + { + "epoch": 2.53, + "learning_rate": 1.6014589574725557e-05, + "loss": 0.0669, + "step": 19284 + }, + { + "epoch": 2.53, + "learning_rate": 1.601416588761093e-05, + "loss": 0.0619, + "step": 19285 + }, + { + "epoch": 2.53, + "learning_rate": 1.6013742183581778e-05, + "loss": 0.0613, + "step": 19286 + }, + { + "epoch": 2.53, + "learning_rate": 1.6013318462639296e-05, + "loss": 0.0631, + "step": 19287 + }, + { + "epoch": 2.53, + "learning_rate": 1.601289472478468e-05, + "loss": 0.0682, + "step": 19288 + }, + { + "epoch": 2.53, + "learning_rate": 1.6012470970019117e-05, + "loss": 0.0603, + "step": 19289 + }, + { + "epoch": 2.53, + "learning_rate": 1.60120471983438e-05, + "loss": 0.066, + "step": 19290 + }, + { + "epoch": 2.53, + "learning_rate": 1.6011623409759917e-05, + "loss": 0.0572, + "step": 19291 + }, + { + "epoch": 2.53, + "learning_rate": 1.6011199604268667e-05, + "loss": 0.0697, + "step": 19292 + }, + { + "epoch": 2.53, + "learning_rate": 1.6010775781871235e-05, + "loss": 0.0635, + "step": 19293 + }, + { + "epoch": 2.53, + "learning_rate": 1.601035194256882e-05, + "loss": 0.0664, + "step": 19294 + }, + { + "epoch": 2.53, + "learning_rate": 1.6009928086362606e-05, + "loss": 0.0582, + "step": 19295 + }, + { + "epoch": 2.53, + "learning_rate": 1.6009504213253793e-05, + "loss": 0.0611, + "step": 19296 + }, + { + "epoch": 2.53, + "learning_rate": 1.6009080323243566e-05, + "loss": 0.0599, + "step": 19297 + }, + { + "epoch": 2.53, + "learning_rate": 1.600865641633312e-05, + "loss": 0.0602, + "step": 19298 + }, + { + "epoch": 2.53, + "learning_rate": 1.6008232492523653e-05, + "loss": 0.0658, + "step": 19299 + }, + { + "epoch": 2.53, + "learning_rate": 1.600780855181635e-05, + "loss": 0.0649, + "step": 19300 + }, + { + "epoch": 2.53, + "learning_rate": 1.6007384594212405e-05, + "loss": 0.0621, + "step": 19301 + }, + { + "epoch": 2.53, + "learning_rate": 1.600696061971301e-05, + "loss": 0.0631, + "step": 19302 + }, + { + "epoch": 2.53, + "learning_rate": 1.600653662831936e-05, + "loss": 0.0512, + "step": 19303 + }, + { + "epoch": 2.53, + "learning_rate": 1.6006112620032643e-05, + "loss": 0.0465, + "step": 19304 + }, + { + "epoch": 2.53, + "learning_rate": 1.6005688594854053e-05, + "loss": 0.063, + "step": 19305 + }, + { + "epoch": 2.53, + "learning_rate": 1.6005264552784786e-05, + "loss": 0.0707, + "step": 19306 + }, + { + "epoch": 2.53, + "learning_rate": 1.600484049382603e-05, + "loss": 0.0616, + "step": 19307 + }, + { + "epoch": 2.53, + "learning_rate": 1.6004416417978985e-05, + "loss": 0.0597, + "step": 19308 + }, + { + "epoch": 2.53, + "learning_rate": 1.600399232524484e-05, + "loss": 0.0562, + "step": 19309 + }, + { + "epoch": 2.53, + "learning_rate": 1.600356821562478e-05, + "loss": 0.0604, + "step": 19310 + }, + { + "epoch": 2.53, + "learning_rate": 1.6003144089120005e-05, + "loss": 0.0619, + "step": 19311 + }, + { + "epoch": 2.53, + "learning_rate": 1.6002719945731707e-05, + "loss": 0.0641, + "step": 19312 + }, + { + "epoch": 2.53, + "learning_rate": 1.6002295785461083e-05, + "loss": 0.0558, + "step": 19313 + }, + { + "epoch": 2.53, + "learning_rate": 1.6001871608309316e-05, + "loss": 0.0589, + "step": 19314 + }, + { + "epoch": 2.53, + "learning_rate": 1.6001447414277605e-05, + "loss": 0.0624, + "step": 19315 + }, + { + "epoch": 2.53, + "learning_rate": 1.6001023203367147e-05, + "loss": 0.0621, + "step": 19316 + }, + { + "epoch": 2.53, + "learning_rate": 1.600059897557913e-05, + "loss": 0.058, + "step": 19317 + }, + { + "epoch": 2.53, + "learning_rate": 1.6000174730914747e-05, + "loss": 0.0618, + "step": 19318 + }, + { + "epoch": 2.53, + "learning_rate": 1.5999750469375192e-05, + "loss": 0.0553, + "step": 19319 + }, + { + "epoch": 2.53, + "learning_rate": 1.5999326190961656e-05, + "loss": 0.0715, + "step": 19320 + }, + { + "epoch": 2.53, + "learning_rate": 1.5998901895675335e-05, + "loss": 0.0652, + "step": 19321 + }, + { + "epoch": 2.53, + "learning_rate": 1.5998477583517426e-05, + "loss": 0.059, + "step": 19322 + }, + { + "epoch": 2.53, + "learning_rate": 1.5998053254489114e-05, + "loss": 0.0626, + "step": 19323 + }, + { + "epoch": 2.53, + "learning_rate": 1.59976289085916e-05, + "loss": 0.0715, + "step": 19324 + }, + { + "epoch": 2.53, + "learning_rate": 1.5997204545826074e-05, + "loss": 0.0666, + "step": 19325 + }, + { + "epoch": 2.53, + "learning_rate": 1.5996780166193727e-05, + "loss": 0.0649, + "step": 19326 + }, + { + "epoch": 2.53, + "learning_rate": 1.5996355769695754e-05, + "loss": 0.0658, + "step": 19327 + }, + { + "epoch": 2.53, + "learning_rate": 1.5995931356333353e-05, + "loss": 0.0587, + "step": 19328 + }, + { + "epoch": 2.53, + "learning_rate": 1.5995506926107714e-05, + "loss": 0.0538, + "step": 19329 + }, + { + "epoch": 2.53, + "learning_rate": 1.599508247902003e-05, + "loss": 0.0572, + "step": 19330 + }, + { + "epoch": 2.53, + "learning_rate": 1.5994658015071497e-05, + "loss": 0.0566, + "step": 19331 + }, + { + "epoch": 2.53, + "learning_rate": 1.5994233534263308e-05, + "loss": 0.0689, + "step": 19332 + }, + { + "epoch": 2.53, + "learning_rate": 1.5993809036596655e-05, + "loss": 0.06, + "step": 19333 + }, + { + "epoch": 2.53, + "learning_rate": 1.5993384522072736e-05, + "loss": 0.0545, + "step": 19334 + }, + { + "epoch": 2.53, + "learning_rate": 1.599295999069274e-05, + "loss": 0.0708, + "step": 19335 + }, + { + "epoch": 2.53, + "learning_rate": 1.599253544245786e-05, + "loss": 0.0721, + "step": 19336 + }, + { + "epoch": 2.53, + "learning_rate": 1.5992110877369296e-05, + "loss": 0.057, + "step": 19337 + }, + { + "epoch": 2.53, + "learning_rate": 1.599168629542824e-05, + "loss": 0.0574, + "step": 19338 + }, + { + "epoch": 2.53, + "learning_rate": 1.599126169663588e-05, + "loss": 0.0592, + "step": 19339 + }, + { + "epoch": 2.53, + "learning_rate": 1.5990837080993423e-05, + "loss": 0.0631, + "step": 19340 + }, + { + "epoch": 2.53, + "learning_rate": 1.5990412448502053e-05, + "loss": 0.0669, + "step": 19341 + }, + { + "epoch": 2.53, + "learning_rate": 1.598998779916297e-05, + "loss": 0.0527, + "step": 19342 + }, + { + "epoch": 2.53, + "learning_rate": 1.5989563132977358e-05, + "loss": 0.0644, + "step": 19343 + }, + { + "epoch": 2.53, + "learning_rate": 1.5989138449946425e-05, + "loss": 0.0591, + "step": 19344 + }, + { + "epoch": 2.53, + "learning_rate": 1.5988713750071355e-05, + "loss": 0.0672, + "step": 19345 + }, + { + "epoch": 2.53, + "learning_rate": 1.598828903335335e-05, + "loss": 0.0684, + "step": 19346 + }, + { + "epoch": 2.53, + "learning_rate": 1.5987864299793598e-05, + "loss": 0.0593, + "step": 19347 + }, + { + "epoch": 2.53, + "learning_rate": 1.5987439549393297e-05, + "loss": 0.0544, + "step": 19348 + }, + { + "epoch": 2.53, + "learning_rate": 1.5987014782153645e-05, + "loss": 0.0565, + "step": 19349 + }, + { + "epoch": 2.53, + "learning_rate": 1.598658999807583e-05, + "loss": 0.0529, + "step": 19350 + }, + { + "epoch": 2.53, + "learning_rate": 1.5986165197161047e-05, + "loss": 0.0564, + "step": 19351 + }, + { + "epoch": 2.53, + "learning_rate": 1.5985740379410497e-05, + "loss": 0.0604, + "step": 19352 + }, + { + "epoch": 2.54, + "learning_rate": 1.598531554482537e-05, + "loss": 0.0583, + "step": 19353 + }, + { + "epoch": 2.54, + "learning_rate": 1.598489069340686e-05, + "loss": 0.0581, + "step": 19354 + }, + { + "epoch": 2.54, + "learning_rate": 1.5984465825156163e-05, + "loss": 0.0697, + "step": 19355 + }, + { + "epoch": 2.54, + "learning_rate": 1.598404094007448e-05, + "loss": 0.0538, + "step": 19356 + }, + { + "epoch": 2.54, + "learning_rate": 1.5983616038162996e-05, + "loss": 0.053, + "step": 19357 + }, + { + "epoch": 2.54, + "learning_rate": 1.5983191119422913e-05, + "loss": 0.0636, + "step": 19358 + }, + { + "epoch": 2.54, + "learning_rate": 1.598276618385542e-05, + "loss": 0.0595, + "step": 19359 + }, + { + "epoch": 2.54, + "learning_rate": 1.5982341231461717e-05, + "loss": 0.0756, + "step": 19360 + }, + { + "epoch": 2.54, + "learning_rate": 1.5981916262242997e-05, + "loss": 0.0506, + "step": 19361 + }, + { + "epoch": 2.54, + "learning_rate": 1.598149127620046e-05, + "loss": 0.0575, + "step": 19362 + }, + { + "epoch": 2.54, + "learning_rate": 1.598106627333529e-05, + "loss": 0.062, + "step": 19363 + }, + { + "epoch": 2.54, + "learning_rate": 1.5980641253648698e-05, + "loss": 0.0589, + "step": 19364 + }, + { + "epoch": 2.54, + "learning_rate": 1.5980216217141867e-05, + "loss": 0.0594, + "step": 19365 + }, + { + "epoch": 2.54, + "learning_rate": 1.5979791163815997e-05, + "loss": 0.0585, + "step": 19366 + }, + { + "epoch": 2.54, + "learning_rate": 1.5979366093672285e-05, + "loss": 0.0704, + "step": 19367 + }, + { + "epoch": 2.54, + "learning_rate": 1.5978941006711922e-05, + "loss": 0.0618, + "step": 19368 + }, + { + "epoch": 2.54, + "learning_rate": 1.5978515902936106e-05, + "loss": 0.0598, + "step": 19369 + }, + { + "epoch": 2.54, + "learning_rate": 1.597809078234603e-05, + "loss": 0.0592, + "step": 19370 + }, + { + "epoch": 2.54, + "learning_rate": 1.5977665644942896e-05, + "loss": 0.0622, + "step": 19371 + }, + { + "epoch": 2.54, + "learning_rate": 1.5977240490727894e-05, + "loss": 0.0546, + "step": 19372 + }, + { + "epoch": 2.54, + "learning_rate": 1.597681531970222e-05, + "loss": 0.0552, + "step": 19373 + }, + { + "epoch": 2.54, + "learning_rate": 1.5976390131867073e-05, + "loss": 0.0704, + "step": 19374 + }, + { + "epoch": 2.54, + "learning_rate": 1.5975964927223646e-05, + "loss": 0.0623, + "step": 19375 + }, + { + "epoch": 2.54, + "learning_rate": 1.5975539705773137e-05, + "loss": 0.0609, + "step": 19376 + }, + { + "epoch": 2.54, + "learning_rate": 1.597511446751674e-05, + "loss": 0.0555, + "step": 19377 + }, + { + "epoch": 2.54, + "learning_rate": 1.597468921245565e-05, + "loss": 0.0529, + "step": 19378 + }, + { + "epoch": 2.54, + "learning_rate": 1.597426394059107e-05, + "loss": 0.0726, + "step": 19379 + }, + { + "epoch": 2.54, + "learning_rate": 1.5973838651924187e-05, + "loss": 0.0627, + "step": 19380 + }, + { + "epoch": 2.54, + "learning_rate": 1.59734133464562e-05, + "loss": 0.0627, + "step": 19381 + }, + { + "epoch": 2.54, + "learning_rate": 1.5972988024188308e-05, + "loss": 0.064, + "step": 19382 + }, + { + "epoch": 2.54, + "learning_rate": 1.5972562685121704e-05, + "loss": 0.0591, + "step": 19383 + }, + { + "epoch": 2.54, + "learning_rate": 1.5972137329257587e-05, + "loss": 0.0577, + "step": 19384 + }, + { + "epoch": 2.54, + "learning_rate": 1.5971711956597152e-05, + "loss": 0.0614, + "step": 19385 + }, + { + "epoch": 2.54, + "learning_rate": 1.597128656714159e-05, + "loss": 0.0509, + "step": 19386 + }, + { + "epoch": 2.54, + "learning_rate": 1.597086116089211e-05, + "loss": 0.052, + "step": 19387 + }, + { + "epoch": 2.54, + "learning_rate": 1.5970435737849897e-05, + "loss": 0.064, + "step": 19388 + }, + { + "epoch": 2.54, + "learning_rate": 1.597001029801615e-05, + "loss": 0.0641, + "step": 19389 + }, + { + "epoch": 2.54, + "learning_rate": 1.596958484139207e-05, + "loss": 0.0617, + "step": 19390 + }, + { + "epoch": 2.54, + "learning_rate": 1.596915936797885e-05, + "loss": 0.0592, + "step": 19391 + }, + { + "epoch": 2.54, + "learning_rate": 1.5968733877777687e-05, + "loss": 0.0497, + "step": 19392 + }, + { + "epoch": 2.54, + "learning_rate": 1.5968308370789777e-05, + "loss": 0.0703, + "step": 19393 + }, + { + "epoch": 2.54, + "learning_rate": 1.596788284701632e-05, + "loss": 0.0612, + "step": 19394 + }, + { + "epoch": 2.54, + "learning_rate": 1.596745730645851e-05, + "loss": 0.0553, + "step": 19395 + }, + { + "epoch": 2.54, + "learning_rate": 1.596703174911754e-05, + "loss": 0.0617, + "step": 19396 + }, + { + "epoch": 2.54, + "learning_rate": 1.5966606174994615e-05, + "loss": 0.0571, + "step": 19397 + }, + { + "epoch": 2.54, + "learning_rate": 1.5966180584090927e-05, + "loss": 0.0516, + "step": 19398 + }, + { + "epoch": 2.54, + "learning_rate": 1.5965754976407672e-05, + "loss": 0.0571, + "step": 19399 + }, + { + "epoch": 2.54, + "learning_rate": 1.596532935194605e-05, + "loss": 0.0551, + "step": 19400 + }, + { + "epoch": 2.54, + "learning_rate": 1.5964903710707255e-05, + "loss": 0.0648, + "step": 19401 + }, + { + "epoch": 2.54, + "learning_rate": 1.596447805269249e-05, + "loss": 0.05, + "step": 19402 + }, + { + "epoch": 2.54, + "learning_rate": 1.5964052377902947e-05, + "loss": 0.0657, + "step": 19403 + }, + { + "epoch": 2.54, + "learning_rate": 1.596362668633982e-05, + "loss": 0.046, + "step": 19404 + }, + { + "epoch": 2.54, + "learning_rate": 1.5963200978004315e-05, + "loss": 0.0505, + "step": 19405 + }, + { + "epoch": 2.54, + "learning_rate": 1.5962775252897625e-05, + "loss": 0.0566, + "step": 19406 + }, + { + "epoch": 2.54, + "learning_rate": 1.5962349511020946e-05, + "loss": 0.0634, + "step": 19407 + }, + { + "epoch": 2.54, + "learning_rate": 1.5961923752375474e-05, + "loss": 0.0492, + "step": 19408 + }, + { + "epoch": 2.54, + "learning_rate": 1.5961497976962412e-05, + "loss": 0.0615, + "step": 19409 + }, + { + "epoch": 2.54, + "learning_rate": 1.5961072184782953e-05, + "loss": 0.062, + "step": 19410 + }, + { + "epoch": 2.54, + "learning_rate": 1.5960646375838297e-05, + "loss": 0.0595, + "step": 19411 + }, + { + "epoch": 2.54, + "learning_rate": 1.5960220550129642e-05, + "loss": 0.0622, + "step": 19412 + }, + { + "epoch": 2.54, + "learning_rate": 1.595979470765818e-05, + "loss": 0.0629, + "step": 19413 + }, + { + "epoch": 2.54, + "learning_rate": 1.5959368848425115e-05, + "loss": 0.0643, + "step": 19414 + }, + { + "epoch": 2.54, + "learning_rate": 1.5958942972431644e-05, + "loss": 0.0567, + "step": 19415 + }, + { + "epoch": 2.54, + "learning_rate": 1.5958517079678963e-05, + "loss": 0.0619, + "step": 19416 + }, + { + "epoch": 2.54, + "learning_rate": 1.5958091170168266e-05, + "loss": 0.0664, + "step": 19417 + }, + { + "epoch": 2.54, + "learning_rate": 1.5957665243900758e-05, + "loss": 0.0561, + "step": 19418 + }, + { + "epoch": 2.54, + "learning_rate": 1.5957239300877633e-05, + "loss": 0.0697, + "step": 19419 + }, + { + "epoch": 2.54, + "learning_rate": 1.595681334110009e-05, + "loss": 0.0627, + "step": 19420 + }, + { + "epoch": 2.54, + "learning_rate": 1.5956387364569326e-05, + "loss": 0.0696, + "step": 19421 + }, + { + "epoch": 2.54, + "learning_rate": 1.595596137128654e-05, + "loss": 0.0656, + "step": 19422 + }, + { + "epoch": 2.54, + "learning_rate": 1.595553536125293e-05, + "loss": 0.0591, + "step": 19423 + }, + { + "epoch": 2.54, + "learning_rate": 1.5955109334469696e-05, + "loss": 0.0558, + "step": 19424 + }, + { + "epoch": 2.54, + "learning_rate": 1.5954683290938028e-05, + "loss": 0.0667, + "step": 19425 + }, + { + "epoch": 2.54, + "learning_rate": 1.5954257230659136e-05, + "loss": 0.0481, + "step": 19426 + }, + { + "epoch": 2.54, + "learning_rate": 1.595383115363421e-05, + "loss": 0.0718, + "step": 19427 + }, + { + "epoch": 2.54, + "learning_rate": 1.5953405059864456e-05, + "loss": 0.0681, + "step": 19428 + }, + { + "epoch": 2.55, + "learning_rate": 1.5952978949351062e-05, + "loss": 0.0512, + "step": 19429 + }, + { + "epoch": 2.55, + "learning_rate": 1.5952552822095232e-05, + "loss": 0.0693, + "step": 19430 + }, + { + "epoch": 2.55, + "learning_rate": 1.5952126678098166e-05, + "loss": 0.0579, + "step": 19431 + }, + { + "epoch": 2.55, + "learning_rate": 1.595170051736106e-05, + "loss": 0.0552, + "step": 19432 + }, + { + "epoch": 2.55, + "learning_rate": 1.5951274339885116e-05, + "loss": 0.0658, + "step": 19433 + }, + { + "epoch": 2.55, + "learning_rate": 1.5950848145671525e-05, + "loss": 0.067, + "step": 19434 + }, + { + "epoch": 2.55, + "learning_rate": 1.595042193472149e-05, + "loss": 0.0731, + "step": 19435 + }, + { + "epoch": 2.55, + "learning_rate": 1.5949995707036214e-05, + "loss": 0.0575, + "step": 19436 + }, + { + "epoch": 2.55, + "learning_rate": 1.5949569462616893e-05, + "loss": 0.0669, + "step": 19437 + }, + { + "epoch": 2.55, + "learning_rate": 1.5949143201464724e-05, + "loss": 0.0605, + "step": 19438 + }, + { + "epoch": 2.55, + "learning_rate": 1.5948716923580906e-05, + "loss": 0.0651, + "step": 19439 + }, + { + "epoch": 2.55, + "learning_rate": 1.5948290628966633e-05, + "loss": 0.0595, + "step": 19440 + }, + { + "epoch": 2.55, + "learning_rate": 1.5947864317623113e-05, + "loss": 0.0641, + "step": 19441 + }, + { + "epoch": 2.55, + "learning_rate": 1.5947437989551543e-05, + "loss": 0.0592, + "step": 19442 + }, + { + "epoch": 2.55, + "learning_rate": 1.594701164475312e-05, + "loss": 0.0617, + "step": 19443 + }, + { + "epoch": 2.55, + "learning_rate": 1.5946585283229044e-05, + "loss": 0.0532, + "step": 19444 + }, + { + "epoch": 2.55, + "learning_rate": 1.5946158904980513e-05, + "loss": 0.075, + "step": 19445 + }, + { + "epoch": 2.55, + "learning_rate": 1.5945732510008725e-05, + "loss": 0.0516, + "step": 19446 + }, + { + "epoch": 2.55, + "learning_rate": 1.5945306098314885e-05, + "loss": 0.0642, + "step": 19447 + }, + { + "epoch": 2.55, + "learning_rate": 1.5944879669900184e-05, + "loss": 0.0533, + "step": 19448 + }, + { + "epoch": 2.55, + "learning_rate": 1.5944453224765825e-05, + "loss": 0.0627, + "step": 19449 + }, + { + "epoch": 2.55, + "learning_rate": 1.594402676291301e-05, + "loss": 0.0627, + "step": 19450 + }, + { + "epoch": 2.55, + "learning_rate": 1.5943600284342937e-05, + "loss": 0.0579, + "step": 19451 + }, + { + "epoch": 2.55, + "learning_rate": 1.5943173789056804e-05, + "loss": 0.0597, + "step": 19452 + }, + { + "epoch": 2.55, + "learning_rate": 1.594274727705581e-05, + "loss": 0.0703, + "step": 19453 + }, + { + "epoch": 2.55, + "learning_rate": 1.5942320748341157e-05, + "loss": 0.0685, + "step": 19454 + }, + { + "epoch": 2.55, + "learning_rate": 1.5941894202914045e-05, + "loss": 0.0481, + "step": 19455 + }, + { + "epoch": 2.55, + "learning_rate": 1.594146764077567e-05, + "loss": 0.0672, + "step": 19456 + }, + { + "epoch": 2.55, + "learning_rate": 1.594104106192723e-05, + "loss": 0.069, + "step": 19457 + }, + { + "epoch": 2.55, + "learning_rate": 1.5940614466369933e-05, + "loss": 0.0666, + "step": 19458 + }, + { + "epoch": 2.55, + "learning_rate": 1.5940187854104972e-05, + "loss": 0.0655, + "step": 19459 + }, + { + "epoch": 2.55, + "learning_rate": 1.5939761225133554e-05, + "loss": 0.0603, + "step": 19460 + }, + { + "epoch": 2.55, + "learning_rate": 1.593933457945687e-05, + "loss": 0.0677, + "step": 19461 + }, + { + "epoch": 2.55, + "learning_rate": 1.5938907917076124e-05, + "loss": 0.0691, + "step": 19462 + }, + { + "epoch": 2.55, + "learning_rate": 1.5938481237992512e-05, + "loss": 0.0659, + "step": 19463 + }, + { + "epoch": 2.55, + "learning_rate": 1.5938054542207243e-05, + "loss": 0.0627, + "step": 19464 + }, + { + "epoch": 2.55, + "learning_rate": 1.593762782972151e-05, + "loss": 0.0646, + "step": 19465 + }, + { + "epoch": 2.55, + "learning_rate": 1.5937201100536516e-05, + "loss": 0.0517, + "step": 19466 + }, + { + "epoch": 2.55, + "learning_rate": 1.5936774354653463e-05, + "loss": 0.065, + "step": 19467 + }, + { + "epoch": 2.55, + "learning_rate": 1.5936347592073544e-05, + "loss": 0.0524, + "step": 19468 + }, + { + "epoch": 2.55, + "learning_rate": 1.593592081279796e-05, + "loss": 0.0641, + "step": 19469 + }, + { + "epoch": 2.55, + "learning_rate": 1.593549401682792e-05, + "loss": 0.0556, + "step": 19470 + }, + { + "epoch": 2.55, + "learning_rate": 1.593506720416462e-05, + "loss": 0.0585, + "step": 19471 + }, + { + "epoch": 2.55, + "learning_rate": 1.5934640374809254e-05, + "loss": 0.0665, + "step": 19472 + }, + { + "epoch": 2.55, + "learning_rate": 1.5934213528763038e-05, + "loss": 0.0656, + "step": 19473 + }, + { + "epoch": 2.55, + "learning_rate": 1.593378666602715e-05, + "loss": 0.0522, + "step": 19474 + }, + { + "epoch": 2.55, + "learning_rate": 1.593335978660281e-05, + "loss": 0.0622, + "step": 19475 + }, + { + "epoch": 2.55, + "learning_rate": 1.593293289049121e-05, + "loss": 0.0638, + "step": 19476 + }, + { + "epoch": 2.55, + "learning_rate": 1.5932505977693557e-05, + "loss": 0.059, + "step": 19477 + }, + { + "epoch": 2.55, + "learning_rate": 1.593207904821104e-05, + "loss": 0.0661, + "step": 19478 + }, + { + "epoch": 2.55, + "learning_rate": 1.593165210204487e-05, + "loss": 0.0608, + "step": 19479 + }, + { + "epoch": 2.55, + "learning_rate": 1.5931225139196244e-05, + "loss": 0.0629, + "step": 19480 + }, + { + "epoch": 2.55, + "learning_rate": 1.5930798159666362e-05, + "loss": 0.0585, + "step": 19481 + }, + { + "epoch": 2.55, + "learning_rate": 1.5930371163456427e-05, + "loss": 0.0559, + "step": 19482 + }, + { + "epoch": 2.55, + "learning_rate": 1.592994415056764e-05, + "loss": 0.068, + "step": 19483 + }, + { + "epoch": 2.55, + "learning_rate": 1.5929517121001198e-05, + "loss": 0.0676, + "step": 19484 + }, + { + "epoch": 2.55, + "learning_rate": 1.5929090074758307e-05, + "loss": 0.0721, + "step": 19485 + }, + { + "epoch": 2.55, + "learning_rate": 1.5928663011840163e-05, + "loss": 0.0664, + "step": 19486 + }, + { + "epoch": 2.55, + "learning_rate": 1.5928235932247975e-05, + "loss": 0.0621, + "step": 19487 + }, + { + "epoch": 2.55, + "learning_rate": 1.5927808835982937e-05, + "loss": 0.0623, + "step": 19488 + }, + { + "epoch": 2.55, + "learning_rate": 1.592738172304625e-05, + "loss": 0.0648, + "step": 19489 + }, + { + "epoch": 2.55, + "learning_rate": 1.592695459343912e-05, + "loss": 0.0672, + "step": 19490 + }, + { + "epoch": 2.55, + "learning_rate": 1.592652744716274e-05, + "loss": 0.0591, + "step": 19491 + }, + { + "epoch": 2.55, + "learning_rate": 1.5926100284218327e-05, + "loss": 0.0714, + "step": 19492 + }, + { + "epoch": 2.55, + "learning_rate": 1.592567310460707e-05, + "loss": 0.0602, + "step": 19493 + }, + { + "epoch": 2.55, + "learning_rate": 1.5925245908330168e-05, + "loss": 0.062, + "step": 19494 + }, + { + "epoch": 2.55, + "learning_rate": 1.592481869538883e-05, + "loss": 0.0632, + "step": 19495 + }, + { + "epoch": 2.55, + "learning_rate": 1.5924391465784256e-05, + "loss": 0.0552, + "step": 19496 + }, + { + "epoch": 2.55, + "learning_rate": 1.592396421951765e-05, + "loss": 0.0612, + "step": 19497 + }, + { + "epoch": 2.55, + "learning_rate": 1.59235369565902e-05, + "loss": 0.064, + "step": 19498 + }, + { + "epoch": 2.55, + "learning_rate": 1.5923109677003126e-05, + "loss": 0.0573, + "step": 19499 + }, + { + "epoch": 2.55, + "learning_rate": 1.592268238075762e-05, + "loss": 0.0675, + "step": 19500 + }, + { + "epoch": 2.55, + "learning_rate": 1.5922255067854883e-05, + "loss": 0.0626, + "step": 19501 + }, + { + "epoch": 2.55, + "learning_rate": 1.5921827738296117e-05, + "loss": 0.0594, + "step": 19502 + }, + { + "epoch": 2.55, + "learning_rate": 1.5921400392082532e-05, + "loss": 0.0671, + "step": 19503 + }, + { + "epoch": 2.55, + "learning_rate": 1.5920973029215323e-05, + "loss": 0.0667, + "step": 19504 + }, + { + "epoch": 2.56, + "learning_rate": 1.5920545649695688e-05, + "loss": 0.0571, + "step": 19505 + }, + { + "epoch": 2.56, + "learning_rate": 1.5920118253524838e-05, + "loss": 0.0579, + "step": 19506 + }, + { + "epoch": 2.56, + "learning_rate": 1.5919690840703967e-05, + "loss": 0.0686, + "step": 19507 + }, + { + "epoch": 2.56, + "learning_rate": 1.5919263411234283e-05, + "loss": 0.0587, + "step": 19508 + }, + { + "epoch": 2.56, + "learning_rate": 1.5918835965116987e-05, + "loss": 0.0662, + "step": 19509 + }, + { + "epoch": 2.56, + "learning_rate": 1.5918408502353274e-05, + "loss": 0.0586, + "step": 19510 + }, + { + "epoch": 2.56, + "learning_rate": 1.591798102294436e-05, + "loss": 0.0668, + "step": 19511 + }, + { + "epoch": 2.56, + "learning_rate": 1.5917553526891434e-05, + "loss": 0.0609, + "step": 19512 + }, + { + "epoch": 2.56, + "learning_rate": 1.5917126014195708e-05, + "loss": 0.064, + "step": 19513 + }, + { + "epoch": 2.56, + "learning_rate": 1.5916698484858377e-05, + "loss": 0.0674, + "step": 19514 + }, + { + "epoch": 2.56, + "learning_rate": 1.591627093888065e-05, + "loss": 0.0612, + "step": 19515 + }, + { + "epoch": 2.56, + "learning_rate": 1.5915843376263723e-05, + "loss": 0.0611, + "step": 19516 + }, + { + "epoch": 2.56, + "learning_rate": 1.5915415797008802e-05, + "loss": 0.0615, + "step": 19517 + }, + { + "epoch": 2.56, + "learning_rate": 1.591498820111709e-05, + "loss": 0.0496, + "step": 19518 + }, + { + "epoch": 2.56, + "learning_rate": 1.5914560588589787e-05, + "loss": 0.0651, + "step": 19519 + }, + { + "epoch": 2.56, + "learning_rate": 1.59141329594281e-05, + "loss": 0.0639, + "step": 19520 + }, + { + "epoch": 2.56, + "learning_rate": 1.5913705313633227e-05, + "loss": 0.0539, + "step": 19521 + }, + { + "epoch": 2.56, + "learning_rate": 1.5913277651206373e-05, + "loss": 0.0639, + "step": 19522 + }, + { + "epoch": 2.56, + "learning_rate": 1.591284997214874e-05, + "loss": 0.0517, + "step": 19523 + }, + { + "epoch": 2.56, + "learning_rate": 1.5912422276461532e-05, + "loss": 0.058, + "step": 19524 + }, + { + "epoch": 2.56, + "learning_rate": 1.5911994564145952e-05, + "loss": 0.0655, + "step": 19525 + }, + { + "epoch": 2.56, + "learning_rate": 1.5911566835203203e-05, + "loss": 0.0605, + "step": 19526 + }, + { + "epoch": 2.56, + "learning_rate": 1.5911139089634484e-05, + "loss": 0.065, + "step": 19527 + }, + { + "epoch": 2.56, + "learning_rate": 1.5910711327441003e-05, + "loss": 0.0585, + "step": 19528 + }, + { + "epoch": 2.56, + "learning_rate": 1.591028354862396e-05, + "loss": 0.0514, + "step": 19529 + }, + { + "epoch": 2.56, + "learning_rate": 1.590985575318456e-05, + "loss": 0.0574, + "step": 19530 + }, + { + "epoch": 2.56, + "learning_rate": 1.5909427941124004e-05, + "loss": 0.0454, + "step": 19531 + }, + { + "epoch": 2.56, + "learning_rate": 1.59090001124435e-05, + "loss": 0.0634, + "step": 19532 + }, + { + "epoch": 2.56, + "learning_rate": 1.5908572267144243e-05, + "loss": 0.0624, + "step": 19533 + }, + { + "epoch": 2.56, + "learning_rate": 1.5908144405227445e-05, + "loss": 0.058, + "step": 19534 + }, + { + "epoch": 2.56, + "learning_rate": 1.5907716526694306e-05, + "loss": 0.0528, + "step": 19535 + }, + { + "epoch": 2.56, + "learning_rate": 1.5907288631546027e-05, + "loss": 0.0639, + "step": 19536 + }, + { + "epoch": 2.56, + "learning_rate": 1.5906860719783813e-05, + "loss": 0.0704, + "step": 19537 + }, + { + "epoch": 2.56, + "learning_rate": 1.590643279140887e-05, + "loss": 0.0508, + "step": 19538 + }, + { + "epoch": 2.56, + "learning_rate": 1.5906004846422397e-05, + "loss": 0.0568, + "step": 19539 + }, + { + "epoch": 2.56, + "learning_rate": 1.59055768848256e-05, + "loss": 0.0592, + "step": 19540 + }, + { + "epoch": 2.56, + "learning_rate": 1.5905148906619685e-05, + "loss": 0.0611, + "step": 19541 + }, + { + "epoch": 2.56, + "learning_rate": 1.590472091180585e-05, + "loss": 0.0517, + "step": 19542 + }, + { + "epoch": 2.56, + "learning_rate": 1.5904292900385304e-05, + "loss": 0.0523, + "step": 19543 + }, + { + "epoch": 2.56, + "learning_rate": 1.5903864872359246e-05, + "loss": 0.0592, + "step": 19544 + }, + { + "epoch": 2.56, + "learning_rate": 1.5903436827728883e-05, + "loss": 0.0604, + "step": 19545 + }, + { + "epoch": 2.56, + "learning_rate": 1.5903008766495424e-05, + "loss": 0.0596, + "step": 19546 + }, + { + "epoch": 2.56, + "learning_rate": 1.590258068866006e-05, + "loss": 0.0469, + "step": 19547 + }, + { + "epoch": 2.56, + "learning_rate": 1.5902152594224004e-05, + "loss": 0.0483, + "step": 19548 + }, + { + "epoch": 2.56, + "learning_rate": 1.590172448318846e-05, + "loss": 0.0592, + "step": 19549 + }, + { + "epoch": 2.56, + "learning_rate": 1.5901296355554628e-05, + "loss": 0.0573, + "step": 19550 + }, + { + "epoch": 2.56, + "learning_rate": 1.590086821132372e-05, + "loss": 0.0507, + "step": 19551 + }, + { + "epoch": 2.56, + "learning_rate": 1.5900440050496927e-05, + "loss": 0.0522, + "step": 19552 + }, + { + "epoch": 2.56, + "learning_rate": 1.5900011873075463e-05, + "loss": 0.064, + "step": 19553 + }, + { + "epoch": 2.56, + "learning_rate": 1.589958367906053e-05, + "loss": 0.0595, + "step": 19554 + }, + { + "epoch": 2.56, + "learning_rate": 1.5899155468453335e-05, + "loss": 0.0654, + "step": 19555 + }, + { + "epoch": 2.56, + "learning_rate": 1.5898727241255073e-05, + "loss": 0.0627, + "step": 19556 + }, + { + "epoch": 2.56, + "learning_rate": 1.5898298997466962e-05, + "loss": 0.0629, + "step": 19557 + }, + { + "epoch": 2.56, + "learning_rate": 1.5897870737090192e-05, + "loss": 0.052, + "step": 19558 + }, + { + "epoch": 2.56, + "learning_rate": 1.589744246012598e-05, + "loss": 0.0525, + "step": 19559 + }, + { + "epoch": 2.56, + "learning_rate": 1.5897014166575523e-05, + "loss": 0.0614, + "step": 19560 + }, + { + "epoch": 2.56, + "learning_rate": 1.589658585644003e-05, + "loss": 0.0622, + "step": 19561 + }, + { + "epoch": 2.56, + "learning_rate": 1.5896157529720698e-05, + "loss": 0.0597, + "step": 19562 + }, + { + "epoch": 2.56, + "learning_rate": 1.589572918641874e-05, + "loss": 0.0613, + "step": 19563 + }, + { + "epoch": 2.56, + "learning_rate": 1.589530082653536e-05, + "loss": 0.0751, + "step": 19564 + }, + { + "epoch": 2.56, + "learning_rate": 1.5894872450071762e-05, + "loss": 0.0713, + "step": 19565 + }, + { + "epoch": 2.56, + "learning_rate": 1.5894444057029145e-05, + "loss": 0.0491, + "step": 19566 + }, + { + "epoch": 2.56, + "learning_rate": 1.5894015647408717e-05, + "loss": 0.0595, + "step": 19567 + }, + { + "epoch": 2.56, + "learning_rate": 1.5893587221211684e-05, + "loss": 0.0549, + "step": 19568 + }, + { + "epoch": 2.56, + "learning_rate": 1.5893158778439254e-05, + "loss": 0.0674, + "step": 19569 + }, + { + "epoch": 2.56, + "learning_rate": 1.5892730319092625e-05, + "loss": 0.0567, + "step": 19570 + }, + { + "epoch": 2.56, + "learning_rate": 1.5892301843173008e-05, + "loss": 0.0563, + "step": 19571 + }, + { + "epoch": 2.56, + "learning_rate": 1.5891873350681607e-05, + "loss": 0.0564, + "step": 19572 + }, + { + "epoch": 2.56, + "learning_rate": 1.5891444841619623e-05, + "loss": 0.0669, + "step": 19573 + }, + { + "epoch": 2.56, + "learning_rate": 1.5891016315988265e-05, + "loss": 0.06, + "step": 19574 + }, + { + "epoch": 2.56, + "learning_rate": 1.589058777378874e-05, + "loss": 0.0636, + "step": 19575 + }, + { + "epoch": 2.56, + "learning_rate": 1.5890159215022248e-05, + "loss": 0.0573, + "step": 19576 + }, + { + "epoch": 2.56, + "learning_rate": 1.588973063969e-05, + "loss": 0.0602, + "step": 19577 + }, + { + "epoch": 2.56, + "learning_rate": 1.5889302047793193e-05, + "loss": 0.0627, + "step": 19578 + }, + { + "epoch": 2.56, + "learning_rate": 1.5888873439333038e-05, + "loss": 0.0624, + "step": 19579 + }, + { + "epoch": 2.56, + "learning_rate": 1.5888444814310745e-05, + "loss": 0.0706, + "step": 19580 + }, + { + "epoch": 2.56, + "learning_rate": 1.5888016172727512e-05, + "loss": 0.0701, + "step": 19581 + }, + { + "epoch": 2.57, + "learning_rate": 1.5887587514584545e-05, + "loss": 0.0569, + "step": 19582 + }, + { + "epoch": 2.57, + "learning_rate": 1.5887158839883055e-05, + "loss": 0.0548, + "step": 19583 + }, + { + "epoch": 2.57, + "learning_rate": 1.5886730148624243e-05, + "loss": 0.0563, + "step": 19584 + }, + { + "epoch": 2.57, + "learning_rate": 1.5886301440809315e-05, + "loss": 0.0609, + "step": 19585 + }, + { + "epoch": 2.57, + "learning_rate": 1.5885872716439477e-05, + "loss": 0.0656, + "step": 19586 + }, + { + "epoch": 2.57, + "learning_rate": 1.588544397551594e-05, + "loss": 0.0632, + "step": 19587 + }, + { + "epoch": 2.57, + "learning_rate": 1.58850152180399e-05, + "loss": 0.0618, + "step": 19588 + }, + { + "epoch": 2.57, + "learning_rate": 1.5884586444012572e-05, + "loss": 0.0594, + "step": 19589 + }, + { + "epoch": 2.57, + "learning_rate": 1.5884157653435156e-05, + "loss": 0.0595, + "step": 19590 + }, + { + "epoch": 2.57, + "learning_rate": 1.588372884630886e-05, + "loss": 0.0491, + "step": 19591 + }, + { + "epoch": 2.57, + "learning_rate": 1.588330002263489e-05, + "loss": 0.0538, + "step": 19592 + }, + { + "epoch": 2.57, + "learning_rate": 1.588287118241445e-05, + "loss": 0.05, + "step": 19593 + }, + { + "epoch": 2.57, + "learning_rate": 1.5882442325648753e-05, + "loss": 0.0665, + "step": 19594 + }, + { + "epoch": 2.57, + "learning_rate": 1.5882013452338996e-05, + "loss": 0.0507, + "step": 19595 + }, + { + "epoch": 2.57, + "learning_rate": 1.588158456248639e-05, + "loss": 0.0595, + "step": 19596 + }, + { + "epoch": 2.57, + "learning_rate": 1.5881155656092144e-05, + "loss": 0.0681, + "step": 19597 + }, + { + "epoch": 2.57, + "learning_rate": 1.5880726733157458e-05, + "loss": 0.0555, + "step": 19598 + }, + { + "epoch": 2.57, + "learning_rate": 1.5880297793683542e-05, + "loss": 0.0629, + "step": 19599 + }, + { + "epoch": 2.57, + "learning_rate": 1.58798688376716e-05, + "loss": 0.0597, + "step": 19600 + }, + { + "epoch": 2.57, + "learning_rate": 1.587943986512284e-05, + "loss": 0.0495, + "step": 19601 + }, + { + "epoch": 2.57, + "learning_rate": 1.587901087603847e-05, + "loss": 0.0649, + "step": 19602 + }, + { + "epoch": 2.57, + "learning_rate": 1.5878581870419695e-05, + "loss": 0.0552, + "step": 19603 + }, + { + "epoch": 2.57, + "learning_rate": 1.5878152848267723e-05, + "loss": 0.0527, + "step": 19604 + }, + { + "epoch": 2.57, + "learning_rate": 1.5877723809583752e-05, + "loss": 0.0465, + "step": 19605 + }, + { + "epoch": 2.57, + "learning_rate": 1.5877294754369005e-05, + "loss": 0.0573, + "step": 19606 + }, + { + "epoch": 2.57, + "learning_rate": 1.5876865682624677e-05, + "loss": 0.065, + "step": 19607 + }, + { + "epoch": 2.57, + "learning_rate": 1.5876436594351975e-05, + "loss": 0.0631, + "step": 19608 + }, + { + "epoch": 2.57, + "learning_rate": 1.587600748955211e-05, + "loss": 0.0606, + "step": 19609 + }, + { + "epoch": 2.57, + "learning_rate": 1.5875578368226284e-05, + "loss": 0.0535, + "step": 19610 + }, + { + "epoch": 2.57, + "learning_rate": 1.587514923037571e-05, + "loss": 0.0461, + "step": 19611 + }, + { + "epoch": 2.57, + "learning_rate": 1.587472007600159e-05, + "loss": 0.0568, + "step": 19612 + }, + { + "epoch": 2.57, + "learning_rate": 1.587429090510513e-05, + "loss": 0.058, + "step": 19613 + }, + { + "epoch": 2.57, + "learning_rate": 1.5873861717687544e-05, + "loss": 0.0561, + "step": 19614 + }, + { + "epoch": 2.57, + "learning_rate": 1.5873432513750033e-05, + "loss": 0.055, + "step": 19615 + }, + { + "epoch": 2.57, + "learning_rate": 1.5873003293293808e-05, + "loss": 0.0645, + "step": 19616 + }, + { + "epoch": 2.57, + "learning_rate": 1.5872574056320068e-05, + "loss": 0.0606, + "step": 19617 + }, + { + "epoch": 2.57, + "learning_rate": 1.5872144802830032e-05, + "loss": 0.051, + "step": 19618 + }, + { + "epoch": 2.57, + "learning_rate": 1.58717155328249e-05, + "loss": 0.0713, + "step": 19619 + }, + { + "epoch": 2.57, + "learning_rate": 1.587128624630588e-05, + "loss": 0.0524, + "step": 19620 + }, + { + "epoch": 2.57, + "learning_rate": 1.587085694327418e-05, + "loss": 0.0671, + "step": 19621 + }, + { + "epoch": 2.57, + "learning_rate": 1.5870427623731007e-05, + "loss": 0.0621, + "step": 19622 + }, + { + "epoch": 2.57, + "learning_rate": 1.586999828767757e-05, + "loss": 0.0616, + "step": 19623 + }, + { + "epoch": 2.57, + "learning_rate": 1.5869568935115077e-05, + "loss": 0.064, + "step": 19624 + }, + { + "epoch": 2.57, + "learning_rate": 1.5869139566044733e-05, + "loss": 0.0508, + "step": 19625 + }, + { + "epoch": 2.57, + "learning_rate": 1.5868710180467742e-05, + "loss": 0.0572, + "step": 19626 + }, + { + "epoch": 2.57, + "learning_rate": 1.5868280778385322e-05, + "loss": 0.0534, + "step": 19627 + }, + { + "epoch": 2.57, + "learning_rate": 1.5867851359798675e-05, + "loss": 0.0615, + "step": 19628 + }, + { + "epoch": 2.57, + "learning_rate": 1.5867421924709003e-05, + "loss": 0.0683, + "step": 19629 + }, + { + "epoch": 2.57, + "learning_rate": 1.5866992473117524e-05, + "loss": 0.0627, + "step": 19630 + }, + { + "epoch": 2.57, + "learning_rate": 1.586656300502544e-05, + "loss": 0.0583, + "step": 19631 + }, + { + "epoch": 2.57, + "learning_rate": 1.5866133520433957e-05, + "loss": 0.0529, + "step": 19632 + }, + { + "epoch": 2.57, + "learning_rate": 1.586570401934429e-05, + "loss": 0.0599, + "step": 19633 + }, + { + "epoch": 2.57, + "learning_rate": 1.5865274501757643e-05, + "loss": 0.0625, + "step": 19634 + }, + { + "epoch": 2.57, + "learning_rate": 1.5864844967675218e-05, + "loss": 0.05, + "step": 19635 + }, + { + "epoch": 2.57, + "learning_rate": 1.5864415417098235e-05, + "loss": 0.0545, + "step": 19636 + }, + { + "epoch": 2.57, + "learning_rate": 1.5863985850027895e-05, + "loss": 0.057, + "step": 19637 + }, + { + "epoch": 2.57, + "learning_rate": 1.5863556266465403e-05, + "loss": 0.0693, + "step": 19638 + }, + { + "epoch": 2.57, + "learning_rate": 1.5863126666411975e-05, + "loss": 0.0592, + "step": 19639 + }, + { + "epoch": 2.57, + "learning_rate": 1.5862697049868816e-05, + "loss": 0.0585, + "step": 19640 + }, + { + "epoch": 2.57, + "learning_rate": 1.5862267416837133e-05, + "loss": 0.0565, + "step": 19641 + }, + { + "epoch": 2.57, + "learning_rate": 1.5861837767318134e-05, + "loss": 0.0613, + "step": 19642 + }, + { + "epoch": 2.57, + "learning_rate": 1.586140810131303e-05, + "loss": 0.0564, + "step": 19643 + }, + { + "epoch": 2.57, + "learning_rate": 1.5860978418823026e-05, + "loss": 0.0665, + "step": 19644 + }, + { + "epoch": 2.57, + "learning_rate": 1.586054871984933e-05, + "loss": 0.0532, + "step": 19645 + }, + { + "epoch": 2.57, + "learning_rate": 1.5860119004393156e-05, + "loss": 0.0559, + "step": 19646 + }, + { + "epoch": 2.57, + "learning_rate": 1.5859689272455713e-05, + "loss": 0.0629, + "step": 19647 + }, + { + "epoch": 2.57, + "learning_rate": 1.58592595240382e-05, + "loss": 0.0638, + "step": 19648 + }, + { + "epoch": 2.57, + "learning_rate": 1.5858829759141835e-05, + "loss": 0.0547, + "step": 19649 + }, + { + "epoch": 2.57, + "learning_rate": 1.5858399977767822e-05, + "loss": 0.0568, + "step": 19650 + }, + { + "epoch": 2.57, + "learning_rate": 1.585797017991737e-05, + "loss": 0.0602, + "step": 19651 + }, + { + "epoch": 2.57, + "learning_rate": 1.5857540365591693e-05, + "loss": 0.0613, + "step": 19652 + }, + { + "epoch": 2.57, + "learning_rate": 1.585711053479199e-05, + "loss": 0.0549, + "step": 19653 + }, + { + "epoch": 2.57, + "learning_rate": 1.5856680687519478e-05, + "loss": 0.0588, + "step": 19654 + }, + { + "epoch": 2.57, + "learning_rate": 1.5856250823775363e-05, + "loss": 0.057, + "step": 19655 + }, + { + "epoch": 2.57, + "learning_rate": 1.585582094356086e-05, + "loss": 0.0505, + "step": 19656 + }, + { + "epoch": 2.57, + "learning_rate": 1.5855391046877164e-05, + "loss": 0.0658, + "step": 19657 + }, + { + "epoch": 2.58, + "learning_rate": 1.5854961133725496e-05, + "loss": 0.0542, + "step": 19658 + }, + { + "epoch": 2.58, + "learning_rate": 1.5854531204107063e-05, + "loss": 0.0622, + "step": 19659 + }, + { + "epoch": 2.58, + "learning_rate": 1.585410125802307e-05, + "loss": 0.0626, + "step": 19660 + }, + { + "epoch": 2.58, + "learning_rate": 1.585367129547473e-05, + "loss": 0.0682, + "step": 19661 + }, + { + "epoch": 2.58, + "learning_rate": 1.5853241316463254e-05, + "loss": 0.0629, + "step": 19662 + }, + { + "epoch": 2.58, + "learning_rate": 1.5852811320989846e-05, + "loss": 0.0656, + "step": 19663 + }, + { + "epoch": 2.58, + "learning_rate": 1.5852381309055718e-05, + "loss": 0.0528, + "step": 19664 + }, + { + "epoch": 2.58, + "learning_rate": 1.5851951280662076e-05, + "loss": 0.0627, + "step": 19665 + }, + { + "epoch": 2.58, + "learning_rate": 1.5851521235810137e-05, + "loss": 0.0608, + "step": 19666 + }, + { + "epoch": 2.58, + "learning_rate": 1.5851091174501107e-05, + "loss": 0.0678, + "step": 19667 + }, + { + "epoch": 2.58, + "learning_rate": 1.585066109673619e-05, + "loss": 0.0623, + "step": 19668 + }, + { + "epoch": 2.58, + "learning_rate": 1.5850231002516604e-05, + "loss": 0.062, + "step": 19669 + }, + { + "epoch": 2.58, + "learning_rate": 1.5849800891843554e-05, + "loss": 0.0598, + "step": 19670 + }, + { + "epoch": 2.58, + "learning_rate": 1.584937076471825e-05, + "loss": 0.057, + "step": 19671 + }, + { + "epoch": 2.58, + "learning_rate": 1.58489406211419e-05, + "loss": 0.0649, + "step": 19672 + }, + { + "epoch": 2.58, + "learning_rate": 1.584851046111572e-05, + "loss": 0.0614, + "step": 19673 + }, + { + "epoch": 2.58, + "learning_rate": 1.5848080284640916e-05, + "loss": 0.056, + "step": 19674 + }, + { + "epoch": 2.58, + "learning_rate": 1.5847650091718697e-05, + "loss": 0.0579, + "step": 19675 + }, + { + "epoch": 2.58, + "learning_rate": 1.584721988235027e-05, + "loss": 0.0657, + "step": 19676 + }, + { + "epoch": 2.58, + "learning_rate": 1.584678965653685e-05, + "loss": 0.0634, + "step": 19677 + }, + { + "epoch": 2.58, + "learning_rate": 1.5846359414279646e-05, + "loss": 0.0531, + "step": 19678 + }, + { + "epoch": 2.58, + "learning_rate": 1.5845929155579867e-05, + "loss": 0.0547, + "step": 19679 + }, + { + "epoch": 2.58, + "learning_rate": 1.5845498880438724e-05, + "loss": 0.0659, + "step": 19680 + }, + { + "epoch": 2.58, + "learning_rate": 1.5845068588857428e-05, + "loss": 0.0617, + "step": 19681 + }, + { + "epoch": 2.58, + "learning_rate": 1.5844638280837187e-05, + "loss": 0.0583, + "step": 19682 + }, + { + "epoch": 2.58, + "learning_rate": 1.584420795637921e-05, + "loss": 0.0621, + "step": 19683 + }, + { + "epoch": 2.58, + "learning_rate": 1.5843777615484713e-05, + "loss": 0.0635, + "step": 19684 + }, + { + "epoch": 2.58, + "learning_rate": 1.5843347258154903e-05, + "loss": 0.0622, + "step": 19685 + }, + { + "epoch": 2.58, + "learning_rate": 1.5842916884390987e-05, + "loss": 0.0566, + "step": 19686 + }, + { + "epoch": 2.58, + "learning_rate": 1.5842486494194178e-05, + "loss": 0.0558, + "step": 19687 + }, + { + "epoch": 2.58, + "learning_rate": 1.584205608756569e-05, + "loss": 0.0654, + "step": 19688 + }, + { + "epoch": 2.58, + "learning_rate": 1.5841625664506728e-05, + "loss": 0.0644, + "step": 19689 + }, + { + "epoch": 2.58, + "learning_rate": 1.5841195225018507e-05, + "loss": 0.0568, + "step": 19690 + }, + { + "epoch": 2.58, + "learning_rate": 1.5840764769102234e-05, + "loss": 0.0592, + "step": 19691 + }, + { + "epoch": 2.58, + "learning_rate": 1.5840334296759116e-05, + "loss": 0.0558, + "step": 19692 + }, + { + "epoch": 2.58, + "learning_rate": 1.583990380799038e-05, + "loss": 0.0537, + "step": 19693 + }, + { + "epoch": 2.58, + "learning_rate": 1.583947330279722e-05, + "loss": 0.063, + "step": 19694 + }, + { + "epoch": 2.58, + "learning_rate": 1.5839042781180845e-05, + "loss": 0.0544, + "step": 19695 + }, + { + "epoch": 2.58, + "learning_rate": 1.5838612243142485e-05, + "loss": 0.0521, + "step": 19696 + }, + { + "epoch": 2.58, + "learning_rate": 1.583818168868333e-05, + "loss": 0.0651, + "step": 19697 + }, + { + "epoch": 2.58, + "learning_rate": 1.5837751117804606e-05, + "loss": 0.06, + "step": 19698 + }, + { + "epoch": 2.58, + "learning_rate": 1.5837320530507516e-05, + "loss": 0.0557, + "step": 19699 + }, + { + "epoch": 2.58, + "learning_rate": 1.583688992679327e-05, + "loss": 0.0544, + "step": 19700 + }, + { + "epoch": 2.58, + "learning_rate": 1.5836459306663086e-05, + "loss": 0.0698, + "step": 19701 + }, + { + "epoch": 2.58, + "learning_rate": 1.583602867011817e-05, + "loss": 0.062, + "step": 19702 + }, + { + "epoch": 2.58, + "learning_rate": 1.583559801715973e-05, + "loss": 0.0573, + "step": 19703 + }, + { + "epoch": 2.58, + "learning_rate": 1.5835167347788987e-05, + "loss": 0.0542, + "step": 19704 + }, + { + "epoch": 2.58, + "learning_rate": 1.5834736662007143e-05, + "loss": 0.0707, + "step": 19705 + }, + { + "epoch": 2.58, + "learning_rate": 1.5834305959815413e-05, + "loss": 0.0707, + "step": 19706 + }, + { + "epoch": 2.58, + "learning_rate": 1.5833875241215013e-05, + "loss": 0.0656, + "step": 19707 + }, + { + "epoch": 2.58, + "learning_rate": 1.5833444506207144e-05, + "loss": 0.0558, + "step": 19708 + }, + { + "epoch": 2.58, + "learning_rate": 1.5833013754793026e-05, + "loss": 0.0707, + "step": 19709 + }, + { + "epoch": 2.58, + "learning_rate": 1.5832582986973868e-05, + "loss": 0.066, + "step": 19710 + }, + { + "epoch": 2.58, + "learning_rate": 1.5832152202750878e-05, + "loss": 0.0524, + "step": 19711 + }, + { + "epoch": 2.58, + "learning_rate": 1.5831721402125274e-05, + "loss": 0.0657, + "step": 19712 + }, + { + "epoch": 2.58, + "learning_rate": 1.583129058509826e-05, + "loss": 0.0565, + "step": 19713 + }, + { + "epoch": 2.58, + "learning_rate": 1.5830859751671056e-05, + "loss": 0.0593, + "step": 19714 + }, + { + "epoch": 2.58, + "learning_rate": 1.5830428901844866e-05, + "loss": 0.0599, + "step": 19715 + }, + { + "epoch": 2.58, + "learning_rate": 1.5829998035620907e-05, + "loss": 0.0588, + "step": 19716 + }, + { + "epoch": 2.58, + "learning_rate": 1.5829567153000386e-05, + "loss": 0.0625, + "step": 19717 + }, + { + "epoch": 2.58, + "learning_rate": 1.5829136253984522e-05, + "loss": 0.0669, + "step": 19718 + }, + { + "epoch": 2.58, + "learning_rate": 1.582870533857452e-05, + "loss": 0.0548, + "step": 19719 + }, + { + "epoch": 2.58, + "learning_rate": 1.58282744067716e-05, + "loss": 0.0661, + "step": 19720 + }, + { + "epoch": 2.58, + "learning_rate": 1.5827843458576962e-05, + "loss": 0.0478, + "step": 19721 + }, + { + "epoch": 2.58, + "learning_rate": 1.5827412493991824e-05, + "loss": 0.0624, + "step": 19722 + }, + { + "epoch": 2.58, + "learning_rate": 1.5826981513017404e-05, + "loss": 0.0715, + "step": 19723 + }, + { + "epoch": 2.58, + "learning_rate": 1.5826550515654907e-05, + "loss": 0.0585, + "step": 19724 + }, + { + "epoch": 2.58, + "learning_rate": 1.5826119501905544e-05, + "loss": 0.0704, + "step": 19725 + }, + { + "epoch": 2.58, + "learning_rate": 1.582568847177053e-05, + "loss": 0.062, + "step": 19726 + }, + { + "epoch": 2.58, + "learning_rate": 1.582525742525108e-05, + "loss": 0.0643, + "step": 19727 + }, + { + "epoch": 2.58, + "learning_rate": 1.5824826362348405e-05, + "loss": 0.0505, + "step": 19728 + }, + { + "epoch": 2.58, + "learning_rate": 1.5824395283063717e-05, + "loss": 0.0622, + "step": 19729 + }, + { + "epoch": 2.58, + "learning_rate": 1.582396418739822e-05, + "loss": 0.0579, + "step": 19730 + }, + { + "epoch": 2.58, + "learning_rate": 1.582353307535314e-05, + "loss": 0.0597, + "step": 19731 + }, + { + "epoch": 2.58, + "learning_rate": 1.5823101946929686e-05, + "loss": 0.0609, + "step": 19732 + }, + { + "epoch": 2.58, + "learning_rate": 1.5822670802129062e-05, + "loss": 0.0535, + "step": 19733 + }, + { + "epoch": 2.59, + "learning_rate": 1.582223964095249e-05, + "loss": 0.0633, + "step": 19734 + }, + { + "epoch": 2.59, + "learning_rate": 1.5821808463401178e-05, + "loss": 0.0597, + "step": 19735 + }, + { + "epoch": 2.59, + "learning_rate": 1.582137726947634e-05, + "loss": 0.0699, + "step": 19736 + }, + { + "epoch": 2.59, + "learning_rate": 1.582094605917919e-05, + "loss": 0.0539, + "step": 19737 + }, + { + "epoch": 2.59, + "learning_rate": 1.5820514832510934e-05, + "loss": 0.0625, + "step": 19738 + }, + { + "epoch": 2.59, + "learning_rate": 1.582008358947279e-05, + "loss": 0.0615, + "step": 19739 + }, + { + "epoch": 2.59, + "learning_rate": 1.5819652330065977e-05, + "loss": 0.0579, + "step": 19740 + }, + { + "epoch": 2.59, + "learning_rate": 1.5819221054291698e-05, + "loss": 0.0627, + "step": 19741 + }, + { + "epoch": 2.59, + "learning_rate": 1.581878976215117e-05, + "loss": 0.0663, + "step": 19742 + }, + { + "epoch": 2.59, + "learning_rate": 1.5818358453645604e-05, + "loss": 0.0633, + "step": 19743 + }, + { + "epoch": 2.59, + "learning_rate": 1.5817927128776222e-05, + "loss": 0.0656, + "step": 19744 + }, + { + "epoch": 2.59, + "learning_rate": 1.5817495787544223e-05, + "loss": 0.0571, + "step": 19745 + }, + { + "epoch": 2.59, + "learning_rate": 1.5817064429950827e-05, + "loss": 0.0563, + "step": 19746 + }, + { + "epoch": 2.59, + "learning_rate": 1.5816633055997247e-05, + "loss": 0.0529, + "step": 19747 + }, + { + "epoch": 2.59, + "learning_rate": 1.58162016656847e-05, + "loss": 0.068, + "step": 19748 + }, + { + "epoch": 2.59, + "learning_rate": 1.581577025901439e-05, + "loss": 0.0517, + "step": 19749 + }, + { + "epoch": 2.59, + "learning_rate": 1.581533883598754e-05, + "loss": 0.0548, + "step": 19750 + }, + { + "epoch": 2.59, + "learning_rate": 1.5814907396605362e-05, + "loss": 0.0627, + "step": 19751 + }, + { + "epoch": 2.59, + "learning_rate": 1.5814475940869057e-05, + "loss": 0.0562, + "step": 19752 + }, + { + "epoch": 2.59, + "learning_rate": 1.5814044468779856e-05, + "loss": 0.0616, + "step": 19753 + }, + { + "epoch": 2.59, + "learning_rate": 1.5813612980338962e-05, + "loss": 0.0606, + "step": 19754 + }, + { + "epoch": 2.59, + "learning_rate": 1.581318147554759e-05, + "loss": 0.0603, + "step": 19755 + }, + { + "epoch": 2.59, + "learning_rate": 1.581274995440696e-05, + "loss": 0.0638, + "step": 19756 + }, + { + "epoch": 2.59, + "learning_rate": 1.5812318416918277e-05, + "loss": 0.0529, + "step": 19757 + }, + { + "epoch": 2.59, + "learning_rate": 1.5811886863082756e-05, + "loss": 0.0555, + "step": 19758 + }, + { + "epoch": 2.59, + "learning_rate": 1.5811455292901614e-05, + "loss": 0.0647, + "step": 19759 + }, + { + "epoch": 2.59, + "learning_rate": 1.5811023706376067e-05, + "loss": 0.0562, + "step": 19760 + }, + { + "epoch": 2.59, + "learning_rate": 1.581059210350732e-05, + "loss": 0.0519, + "step": 19761 + }, + { + "epoch": 2.59, + "learning_rate": 1.5810160484296594e-05, + "loss": 0.0574, + "step": 19762 + }, + { + "epoch": 2.59, + "learning_rate": 1.5809728848745098e-05, + "loss": 0.0596, + "step": 19763 + }, + { + "epoch": 2.59, + "learning_rate": 1.5809297196854055e-05, + "loss": 0.0575, + "step": 19764 + }, + { + "epoch": 2.59, + "learning_rate": 1.5808865528624668e-05, + "loss": 0.0505, + "step": 19765 + }, + { + "epoch": 2.59, + "learning_rate": 1.580843384405816e-05, + "loss": 0.0575, + "step": 19766 + }, + { + "epoch": 2.59, + "learning_rate": 1.5808002143155738e-05, + "loss": 0.0588, + "step": 19767 + }, + { + "epoch": 2.59, + "learning_rate": 1.580757042591862e-05, + "loss": 0.0498, + "step": 19768 + }, + { + "epoch": 2.59, + "learning_rate": 1.580713869234802e-05, + "loss": 0.0561, + "step": 19769 + }, + { + "epoch": 2.59, + "learning_rate": 1.5806706942445153e-05, + "loss": 0.053, + "step": 19770 + }, + { + "epoch": 2.59, + "learning_rate": 1.580627517621123e-05, + "loss": 0.0624, + "step": 19771 + }, + { + "epoch": 2.59, + "learning_rate": 1.5805843393647467e-05, + "loss": 0.0605, + "step": 19772 + }, + { + "epoch": 2.59, + "learning_rate": 1.5805411594755083e-05, + "loss": 0.0585, + "step": 19773 + }, + { + "epoch": 2.59, + "learning_rate": 1.580497977953528e-05, + "loss": 0.0593, + "step": 19774 + }, + { + "epoch": 2.59, + "learning_rate": 1.5804547947989286e-05, + "loss": 0.0611, + "step": 19775 + }, + { + "epoch": 2.59, + "learning_rate": 1.5804116100118312e-05, + "loss": 0.0585, + "step": 19776 + }, + { + "epoch": 2.59, + "learning_rate": 1.5803684235923567e-05, + "loss": 0.06, + "step": 19777 + }, + { + "epoch": 2.59, + "learning_rate": 1.580325235540627e-05, + "loss": 0.0469, + "step": 19778 + }, + { + "epoch": 2.59, + "learning_rate": 1.5802820458567634e-05, + "loss": 0.0556, + "step": 19779 + }, + { + "epoch": 2.59, + "learning_rate": 1.5802388545408875e-05, + "loss": 0.0784, + "step": 19780 + }, + { + "epoch": 2.59, + "learning_rate": 1.5801956615931207e-05, + "loss": 0.0673, + "step": 19781 + }, + { + "epoch": 2.59, + "learning_rate": 1.5801524670135842e-05, + "loss": 0.0596, + "step": 19782 + }, + { + "epoch": 2.59, + "learning_rate": 1.5801092708024004e-05, + "loss": 0.0496, + "step": 19783 + }, + { + "epoch": 2.59, + "learning_rate": 1.58006607295969e-05, + "loss": 0.055, + "step": 19784 + }, + { + "epoch": 2.59, + "learning_rate": 1.5800228734855744e-05, + "loss": 0.0542, + "step": 19785 + }, + { + "epoch": 2.59, + "learning_rate": 1.5799796723801756e-05, + "loss": 0.0555, + "step": 19786 + }, + { + "epoch": 2.59, + "learning_rate": 1.5799364696436145e-05, + "loss": 0.059, + "step": 19787 + }, + { + "epoch": 2.59, + "learning_rate": 1.5798932652760137e-05, + "loss": 0.0589, + "step": 19788 + }, + { + "epoch": 2.59, + "learning_rate": 1.5798500592774932e-05, + "loss": 0.0563, + "step": 19789 + }, + { + "epoch": 2.59, + "learning_rate": 1.5798068516481754e-05, + "loss": 0.0492, + "step": 19790 + }, + { + "epoch": 2.59, + "learning_rate": 1.579763642388182e-05, + "loss": 0.0644, + "step": 19791 + }, + { + "epoch": 2.59, + "learning_rate": 1.5797204314976343e-05, + "loss": 0.0672, + "step": 19792 + }, + { + "epoch": 2.59, + "learning_rate": 1.5796772189766535e-05, + "loss": 0.0615, + "step": 19793 + }, + { + "epoch": 2.59, + "learning_rate": 1.5796340048253616e-05, + "loss": 0.0532, + "step": 19794 + }, + { + "epoch": 2.59, + "learning_rate": 1.5795907890438796e-05, + "loss": 0.0568, + "step": 19795 + }, + { + "epoch": 2.59, + "learning_rate": 1.5795475716323295e-05, + "loss": 0.0654, + "step": 19796 + }, + { + "epoch": 2.59, + "learning_rate": 1.5795043525908328e-05, + "loss": 0.0596, + "step": 19797 + }, + { + "epoch": 2.59, + "learning_rate": 1.579461131919511e-05, + "loss": 0.0534, + "step": 19798 + }, + { + "epoch": 2.59, + "learning_rate": 1.5794179096184855e-05, + "loss": 0.0569, + "step": 19799 + }, + { + "epoch": 2.59, + "learning_rate": 1.5793746856878784e-05, + "loss": 0.0645, + "step": 19800 + }, + { + "epoch": 2.59, + "learning_rate": 1.57933146012781e-05, + "loss": 0.0605, + "step": 19801 + }, + { + "epoch": 2.59, + "learning_rate": 1.5792882329384035e-05, + "loss": 0.0595, + "step": 19802 + }, + { + "epoch": 2.59, + "learning_rate": 1.5792450041197794e-05, + "loss": 0.0515, + "step": 19803 + }, + { + "epoch": 2.59, + "learning_rate": 1.5792017736720594e-05, + "loss": 0.0544, + "step": 19804 + }, + { + "epoch": 2.59, + "learning_rate": 1.5791585415953657e-05, + "loss": 0.058, + "step": 19805 + }, + { + "epoch": 2.59, + "learning_rate": 1.579115307889819e-05, + "loss": 0.0594, + "step": 19806 + }, + { + "epoch": 2.59, + "learning_rate": 1.579072072555541e-05, + "loss": 0.061, + "step": 19807 + }, + { + "epoch": 2.59, + "learning_rate": 1.5790288355926545e-05, + "loss": 0.0653, + "step": 19808 + }, + { + "epoch": 2.59, + "learning_rate": 1.57898559700128e-05, + "loss": 0.061, + "step": 19809 + }, + { + "epoch": 2.59, + "learning_rate": 1.578942356781539e-05, + "loss": 0.0629, + "step": 19810 + }, + { + "epoch": 2.6, + "learning_rate": 1.5788991149335544e-05, + "loss": 0.0563, + "step": 19811 + }, + { + "epoch": 2.6, + "learning_rate": 1.578855871457446e-05, + "loss": 0.0556, + "step": 19812 + }, + { + "epoch": 2.6, + "learning_rate": 1.578812626353336e-05, + "loss": 0.0527, + "step": 19813 + }, + { + "epoch": 2.6, + "learning_rate": 1.578769379621347e-05, + "loss": 0.0616, + "step": 19814 + }, + { + "epoch": 2.6, + "learning_rate": 1.5787261312615995e-05, + "loss": 0.0566, + "step": 19815 + }, + { + "epoch": 2.6, + "learning_rate": 1.578682881274216e-05, + "loss": 0.0543, + "step": 19816 + }, + { + "epoch": 2.6, + "learning_rate": 1.5786396296593175e-05, + "loss": 0.057, + "step": 19817 + }, + { + "epoch": 2.6, + "learning_rate": 1.5785963764170258e-05, + "loss": 0.0663, + "step": 19818 + }, + { + "epoch": 2.6, + "learning_rate": 1.5785531215474626e-05, + "loss": 0.0626, + "step": 19819 + }, + { + "epoch": 2.6, + "learning_rate": 1.5785098650507497e-05, + "loss": 0.0625, + "step": 19820 + }, + { + "epoch": 2.6, + "learning_rate": 1.5784666069270085e-05, + "loss": 0.0574, + "step": 19821 + }, + { + "epoch": 2.6, + "learning_rate": 1.5784233471763607e-05, + "loss": 0.0665, + "step": 19822 + }, + { + "epoch": 2.6, + "learning_rate": 1.578380085798928e-05, + "loss": 0.0696, + "step": 19823 + }, + { + "epoch": 2.6, + "learning_rate": 1.5783368227948324e-05, + "loss": 0.0511, + "step": 19824 + }, + { + "epoch": 2.6, + "learning_rate": 1.578293558164195e-05, + "loss": 0.0639, + "step": 19825 + }, + { + "epoch": 2.6, + "learning_rate": 1.5782502919071377e-05, + "loss": 0.0572, + "step": 19826 + }, + { + "epoch": 2.6, + "learning_rate": 1.5782070240237826e-05, + "loss": 0.0571, + "step": 19827 + }, + { + "epoch": 2.6, + "learning_rate": 1.5781637545142506e-05, + "loss": 0.0617, + "step": 19828 + }, + { + "epoch": 2.6, + "learning_rate": 1.5781204833786644e-05, + "loss": 0.0565, + "step": 19829 + }, + { + "epoch": 2.6, + "learning_rate": 1.5780772106171446e-05, + "loss": 0.0553, + "step": 19830 + }, + { + "epoch": 2.6, + "learning_rate": 1.578033936229814e-05, + "loss": 0.059, + "step": 19831 + }, + { + "epoch": 2.6, + "learning_rate": 1.577990660216793e-05, + "loss": 0.0525, + "step": 19832 + }, + { + "epoch": 2.6, + "learning_rate": 1.577947382578205e-05, + "loss": 0.0597, + "step": 19833 + }, + { + "epoch": 2.6, + "learning_rate": 1.5779041033141697e-05, + "loss": 0.0654, + "step": 19834 + }, + { + "epoch": 2.6, + "learning_rate": 1.5778608224248104e-05, + "loss": 0.0488, + "step": 19835 + }, + { + "epoch": 2.6, + "learning_rate": 1.5778175399102483e-05, + "loss": 0.0558, + "step": 19836 + }, + { + "epoch": 2.6, + "learning_rate": 1.5777742557706053e-05, + "loss": 0.0644, + "step": 19837 + }, + { + "epoch": 2.6, + "learning_rate": 1.577730970006003e-05, + "loss": 0.0604, + "step": 19838 + }, + { + "epoch": 2.6, + "learning_rate": 1.5776876826165626e-05, + "loss": 0.0563, + "step": 19839 + }, + { + "epoch": 2.6, + "learning_rate": 1.5776443936024066e-05, + "loss": 0.0535, + "step": 19840 + }, + { + "epoch": 2.6, + "learning_rate": 1.5776011029636568e-05, + "loss": 0.0575, + "step": 19841 + }, + { + "epoch": 2.6, + "learning_rate": 1.5775578107004346e-05, + "loss": 0.0613, + "step": 19842 + }, + { + "epoch": 2.6, + "learning_rate": 1.5775145168128614e-05, + "loss": 0.0607, + "step": 19843 + }, + { + "epoch": 2.6, + "learning_rate": 1.5774712213010598e-05, + "loss": 0.0545, + "step": 19844 + }, + { + "epoch": 2.6, + "learning_rate": 1.577427924165151e-05, + "loss": 0.0681, + "step": 19845 + }, + { + "epoch": 2.6, + "learning_rate": 1.5773846254052568e-05, + "loss": 0.0713, + "step": 19846 + }, + { + "epoch": 2.6, + "learning_rate": 1.5773413250214992e-05, + "loss": 0.0671, + "step": 19847 + }, + { + "epoch": 2.6, + "learning_rate": 1.5772980230139995e-05, + "loss": 0.0643, + "step": 19848 + }, + { + "epoch": 2.6, + "learning_rate": 1.5772547193828803e-05, + "loss": 0.0537, + "step": 19849 + }, + { + "epoch": 2.6, + "learning_rate": 1.5772114141282632e-05, + "loss": 0.0558, + "step": 19850 + }, + { + "epoch": 2.6, + "learning_rate": 1.577168107250269e-05, + "loss": 0.0558, + "step": 19851 + }, + { + "epoch": 2.6, + "learning_rate": 1.5771247987490208e-05, + "loss": 0.0557, + "step": 19852 + }, + { + "epoch": 2.6, + "learning_rate": 1.5770814886246395e-05, + "loss": 0.066, + "step": 19853 + }, + { + "epoch": 2.6, + "learning_rate": 1.5770381768772475e-05, + "loss": 0.0508, + "step": 19854 + }, + { + "epoch": 2.6, + "learning_rate": 1.576994863506966e-05, + "loss": 0.0559, + "step": 19855 + }, + { + "epoch": 2.6, + "learning_rate": 1.5769515485139174e-05, + "loss": 0.0472, + "step": 19856 + }, + { + "epoch": 2.6, + "learning_rate": 1.5769082318982235e-05, + "loss": 0.0618, + "step": 19857 + }, + { + "epoch": 2.6, + "learning_rate": 1.5768649136600056e-05, + "loss": 0.0522, + "step": 19858 + }, + { + "epoch": 2.6, + "learning_rate": 1.576821593799386e-05, + "loss": 0.0528, + "step": 19859 + }, + { + "epoch": 2.6, + "learning_rate": 1.5767782723164867e-05, + "loss": 0.0499, + "step": 19860 + }, + { + "epoch": 2.6, + "learning_rate": 1.5767349492114285e-05, + "loss": 0.0467, + "step": 19861 + }, + { + "epoch": 2.6, + "learning_rate": 1.5766916244843344e-05, + "loss": 0.0598, + "step": 19862 + }, + { + "epoch": 2.6, + "learning_rate": 1.576648298135326e-05, + "loss": 0.0593, + "step": 19863 + }, + { + "epoch": 2.6, + "learning_rate": 1.5766049701645246e-05, + "loss": 0.0562, + "step": 19864 + }, + { + "epoch": 2.6, + "learning_rate": 1.576561640572053e-05, + "loss": 0.0523, + "step": 19865 + }, + { + "epoch": 2.6, + "learning_rate": 1.5765183093580318e-05, + "loss": 0.0403, + "step": 19866 + }, + { + "epoch": 2.6, + "learning_rate": 1.5764749765225837e-05, + "loss": 0.0731, + "step": 19867 + }, + { + "epoch": 2.6, + "learning_rate": 1.5764316420658305e-05, + "loss": 0.065, + "step": 19868 + }, + { + "epoch": 2.6, + "learning_rate": 1.5763883059878938e-05, + "loss": 0.0552, + "step": 19869 + }, + { + "epoch": 2.6, + "learning_rate": 1.5763449682888963e-05, + "loss": 0.0575, + "step": 19870 + }, + { + "epoch": 2.6, + "learning_rate": 1.5763016289689586e-05, + "loss": 0.0629, + "step": 19871 + }, + { + "epoch": 2.6, + "learning_rate": 1.5762582880282038e-05, + "loss": 0.051, + "step": 19872 + }, + { + "epoch": 2.6, + "learning_rate": 1.5762149454667527e-05, + "loss": 0.053, + "step": 19873 + }, + { + "epoch": 2.6, + "learning_rate": 1.5761716012847283e-05, + "loss": 0.0618, + "step": 19874 + }, + { + "epoch": 2.6, + "learning_rate": 1.5761282554822517e-05, + "loss": 0.064, + "step": 19875 + }, + { + "epoch": 2.6, + "learning_rate": 1.5760849080594454e-05, + "loss": 0.0635, + "step": 19876 + }, + { + "epoch": 2.6, + "learning_rate": 1.5760415590164307e-05, + "loss": 0.0651, + "step": 19877 + }, + { + "epoch": 2.6, + "learning_rate": 1.5759982083533294e-05, + "loss": 0.0668, + "step": 19878 + }, + { + "epoch": 2.6, + "learning_rate": 1.5759548560702643e-05, + "loss": 0.0569, + "step": 19879 + }, + { + "epoch": 2.6, + "learning_rate": 1.5759115021673568e-05, + "loss": 0.0657, + "step": 19880 + }, + { + "epoch": 2.6, + "learning_rate": 1.575868146644729e-05, + "loss": 0.0685, + "step": 19881 + }, + { + "epoch": 2.6, + "learning_rate": 1.575824789502502e-05, + "loss": 0.067, + "step": 19882 + }, + { + "epoch": 2.6, + "learning_rate": 1.5757814307407987e-05, + "loss": 0.0634, + "step": 19883 + }, + { + "epoch": 2.6, + "learning_rate": 1.5757380703597412e-05, + "loss": 0.0571, + "step": 19884 + }, + { + "epoch": 2.6, + "learning_rate": 1.575694708359451e-05, + "loss": 0.052, + "step": 19885 + }, + { + "epoch": 2.6, + "learning_rate": 1.57565134474005e-05, + "loss": 0.058, + "step": 19886 + }, + { + "epoch": 2.61, + "learning_rate": 1.5756079795016602e-05, + "loss": 0.0527, + "step": 19887 + }, + { + "epoch": 2.61, + "learning_rate": 1.5755646126444033e-05, + "loss": 0.0523, + "step": 19888 + }, + { + "epoch": 2.61, + "learning_rate": 1.575521244168402e-05, + "loss": 0.0484, + "step": 19889 + }, + { + "epoch": 2.61, + "learning_rate": 1.5754778740737777e-05, + "loss": 0.0512, + "step": 19890 + }, + { + "epoch": 2.61, + "learning_rate": 1.5754345023606526e-05, + "loss": 0.0611, + "step": 19891 + }, + { + "epoch": 2.61, + "learning_rate": 1.5753911290291485e-05, + "loss": 0.0623, + "step": 19892 + }, + { + "epoch": 2.61, + "learning_rate": 1.5753477540793875e-05, + "loss": 0.0617, + "step": 19893 + }, + { + "epoch": 2.61, + "learning_rate": 1.5753043775114914e-05, + "loss": 0.0525, + "step": 19894 + }, + { + "epoch": 2.61, + "learning_rate": 1.575260999325583e-05, + "loss": 0.0674, + "step": 19895 + }, + { + "epoch": 2.61, + "learning_rate": 1.575217619521783e-05, + "loss": 0.07, + "step": 19896 + }, + { + "epoch": 2.61, + "learning_rate": 1.5751742381002144e-05, + "loss": 0.064, + "step": 19897 + }, + { + "epoch": 2.61, + "learning_rate": 1.575130855060999e-05, + "loss": 0.0601, + "step": 19898 + }, + { + "epoch": 2.61, + "learning_rate": 1.5750874704042582e-05, + "loss": 0.0576, + "step": 19899 + }, + { + "epoch": 2.61, + "learning_rate": 1.575044084130115e-05, + "loss": 0.0575, + "step": 19900 + }, + { + "epoch": 2.61, + "learning_rate": 1.5750006962386908e-05, + "loss": 0.0553, + "step": 19901 + }, + { + "epoch": 2.61, + "learning_rate": 1.574957306730108e-05, + "loss": 0.059, + "step": 19902 + }, + { + "epoch": 2.61, + "learning_rate": 1.574913915604488e-05, + "loss": 0.0673, + "step": 19903 + }, + { + "epoch": 2.61, + "learning_rate": 1.5748705228619533e-05, + "loss": 0.0607, + "step": 19904 + }, + { + "epoch": 2.61, + "learning_rate": 1.574827128502626e-05, + "loss": 0.0576, + "step": 19905 + }, + { + "epoch": 2.61, + "learning_rate": 1.574783732526628e-05, + "loss": 0.0511, + "step": 19906 + }, + { + "epoch": 2.61, + "learning_rate": 1.5747403349340812e-05, + "loss": 0.0546, + "step": 19907 + }, + { + "epoch": 2.61, + "learning_rate": 1.574696935725108e-05, + "loss": 0.0621, + "step": 19908 + }, + { + "epoch": 2.61, + "learning_rate": 1.57465353489983e-05, + "loss": 0.0607, + "step": 19909 + }, + { + "epoch": 2.61, + "learning_rate": 1.5746101324583697e-05, + "loss": 0.0517, + "step": 19910 + }, + { + "epoch": 2.61, + "learning_rate": 1.574566728400849e-05, + "loss": 0.06, + "step": 19911 + }, + { + "epoch": 2.61, + "learning_rate": 1.5745233227273898e-05, + "loss": 0.0713, + "step": 19912 + }, + { + "epoch": 2.61, + "learning_rate": 1.574479915438115e-05, + "loss": 0.0569, + "step": 19913 + }, + { + "epoch": 2.61, + "learning_rate": 1.5744365065331454e-05, + "loss": 0.0647, + "step": 19914 + }, + { + "epoch": 2.61, + "learning_rate": 1.574393096012604e-05, + "loss": 0.059, + "step": 19915 + }, + { + "epoch": 2.61, + "learning_rate": 1.5743496838766124e-05, + "loss": 0.062, + "step": 19916 + }, + { + "epoch": 2.61, + "learning_rate": 1.574306270125293e-05, + "loss": 0.0573, + "step": 19917 + }, + { + "epoch": 2.61, + "learning_rate": 1.574262854758768e-05, + "loss": 0.0563, + "step": 19918 + }, + { + "epoch": 2.61, + "learning_rate": 1.574219437777159e-05, + "loss": 0.062, + "step": 19919 + }, + { + "epoch": 2.61, + "learning_rate": 1.5741760191805884e-05, + "loss": 0.0529, + "step": 19920 + }, + { + "epoch": 2.61, + "learning_rate": 1.5741325989691784e-05, + "loss": 0.0598, + "step": 19921 + }, + { + "epoch": 2.61, + "learning_rate": 1.5740891771430512e-05, + "loss": 0.0629, + "step": 19922 + }, + { + "epoch": 2.61, + "learning_rate": 1.5740457537023285e-05, + "loss": 0.0659, + "step": 19923 + }, + { + "epoch": 2.61, + "learning_rate": 1.5740023286471328e-05, + "loss": 0.0511, + "step": 19924 + }, + { + "epoch": 2.61, + "learning_rate": 1.5739589019775862e-05, + "loss": 0.0673, + "step": 19925 + }, + { + "epoch": 2.61, + "learning_rate": 1.573915473693811e-05, + "loss": 0.0589, + "step": 19926 + }, + { + "epoch": 2.61, + "learning_rate": 1.5738720437959284e-05, + "loss": 0.0712, + "step": 19927 + }, + { + "epoch": 2.61, + "learning_rate": 1.5738286122840618e-05, + "loss": 0.0592, + "step": 19928 + }, + { + "epoch": 2.61, + "learning_rate": 1.5737851791583328e-05, + "loss": 0.0607, + "step": 19929 + }, + { + "epoch": 2.61, + "learning_rate": 1.573741744418863e-05, + "loss": 0.0624, + "step": 19930 + }, + { + "epoch": 2.61, + "learning_rate": 1.5736983080657756e-05, + "loss": 0.0667, + "step": 19931 + }, + { + "epoch": 2.61, + "learning_rate": 1.5736548700991923e-05, + "loss": 0.0621, + "step": 19932 + }, + { + "epoch": 2.61, + "learning_rate": 1.573611430519235e-05, + "loss": 0.0534, + "step": 19933 + }, + { + "epoch": 2.61, + "learning_rate": 1.573567989326026e-05, + "loss": 0.0572, + "step": 19934 + }, + { + "epoch": 2.61, + "learning_rate": 1.573524546519688e-05, + "loss": 0.0575, + "step": 19935 + }, + { + "epoch": 2.61, + "learning_rate": 1.5734811021003423e-05, + "loss": 0.0535, + "step": 19936 + }, + { + "epoch": 2.61, + "learning_rate": 1.5734376560681117e-05, + "loss": 0.0593, + "step": 19937 + }, + { + "epoch": 2.61, + "learning_rate": 1.573394208423118e-05, + "loss": 0.0636, + "step": 19938 + }, + { + "epoch": 2.61, + "learning_rate": 1.5733507591654838e-05, + "loss": 0.0522, + "step": 19939 + }, + { + "epoch": 2.61, + "learning_rate": 1.5733073082953314e-05, + "loss": 0.0599, + "step": 19940 + }, + { + "epoch": 2.61, + "learning_rate": 1.5732638558127823e-05, + "loss": 0.0647, + "step": 19941 + }, + { + "epoch": 2.61, + "learning_rate": 1.5732204017179596e-05, + "loss": 0.0578, + "step": 19942 + }, + { + "epoch": 2.61, + "learning_rate": 1.5731769460109848e-05, + "loss": 0.0534, + "step": 19943 + }, + { + "epoch": 2.61, + "learning_rate": 1.5731334886919803e-05, + "loss": 0.0576, + "step": 19944 + }, + { + "epoch": 2.61, + "learning_rate": 1.5730900297610686e-05, + "loss": 0.0642, + "step": 19945 + }, + { + "epoch": 2.61, + "learning_rate": 1.5730465692183716e-05, + "loss": 0.0552, + "step": 19946 + }, + { + "epoch": 2.61, + "learning_rate": 1.5730031070640113e-05, + "loss": 0.0603, + "step": 19947 + }, + { + "epoch": 2.61, + "learning_rate": 1.5729596432981106e-05, + "loss": 0.0671, + "step": 19948 + }, + { + "epoch": 2.61, + "learning_rate": 1.5729161779207914e-05, + "loss": 0.0519, + "step": 19949 + }, + { + "epoch": 2.61, + "learning_rate": 1.572872710932176e-05, + "loss": 0.0656, + "step": 19950 + }, + { + "epoch": 2.61, + "learning_rate": 1.5728292423323866e-05, + "loss": 0.0516, + "step": 19951 + }, + { + "epoch": 2.61, + "learning_rate": 1.572785772121545e-05, + "loss": 0.055, + "step": 19952 + }, + { + "epoch": 2.61, + "learning_rate": 1.5727423002997745e-05, + "loss": 0.0482, + "step": 19953 + }, + { + "epoch": 2.61, + "learning_rate": 1.5726988268671967e-05, + "loss": 0.0597, + "step": 19954 + }, + { + "epoch": 2.61, + "learning_rate": 1.572655351823934e-05, + "loss": 0.0685, + "step": 19955 + }, + { + "epoch": 2.61, + "learning_rate": 1.5726118751701084e-05, + "loss": 0.0618, + "step": 19956 + }, + { + "epoch": 2.61, + "learning_rate": 1.5725683969058426e-05, + "loss": 0.0604, + "step": 19957 + }, + { + "epoch": 2.61, + "learning_rate": 1.5725249170312583e-05, + "loss": 0.0639, + "step": 19958 + }, + { + "epoch": 2.61, + "learning_rate": 1.5724814355464783e-05, + "loss": 0.0492, + "step": 19959 + }, + { + "epoch": 2.61, + "learning_rate": 1.572437952451625e-05, + "loss": 0.0483, + "step": 19960 + }, + { + "epoch": 2.61, + "learning_rate": 1.57239446774682e-05, + "loss": 0.0519, + "step": 19961 + }, + { + "epoch": 2.61, + "learning_rate": 1.5723509814321864e-05, + "loss": 0.0549, + "step": 19962 + }, + { + "epoch": 2.62, + "learning_rate": 1.5723074935078462e-05, + "loss": 0.0596, + "step": 19963 + }, + { + "epoch": 2.62, + "learning_rate": 1.5722640039739215e-05, + "loss": 0.0602, + "step": 19964 + }, + { + "epoch": 2.62, + "learning_rate": 1.572220512830535e-05, + "loss": 0.0594, + "step": 19965 + }, + { + "epoch": 2.62, + "learning_rate": 1.5721770200778085e-05, + "loss": 0.0639, + "step": 19966 + }, + { + "epoch": 2.62, + "learning_rate": 1.5721335257158648e-05, + "loss": 0.0476, + "step": 19967 + }, + { + "epoch": 2.62, + "learning_rate": 1.5720900297448258e-05, + "loss": 0.0549, + "step": 19968 + }, + { + "epoch": 2.62, + "learning_rate": 1.572046532164814e-05, + "loss": 0.0567, + "step": 19969 + }, + { + "epoch": 2.62, + "learning_rate": 1.572003032975952e-05, + "loss": 0.0722, + "step": 19970 + }, + { + "epoch": 2.62, + "learning_rate": 1.571959532178362e-05, + "loss": 0.0537, + "step": 19971 + }, + { + "epoch": 2.62, + "learning_rate": 1.571916029772166e-05, + "loss": 0.0546, + "step": 19972 + }, + { + "epoch": 2.62, + "learning_rate": 1.571872525757487e-05, + "loss": 0.0792, + "step": 19973 + }, + { + "epoch": 2.62, + "learning_rate": 1.5718290201344466e-05, + "loss": 0.0527, + "step": 19974 + }, + { + "epoch": 2.62, + "learning_rate": 1.5717855129031677e-05, + "loss": 0.0488, + "step": 19975 + }, + { + "epoch": 2.62, + "learning_rate": 1.5717420040637725e-05, + "loss": 0.0673, + "step": 19976 + }, + { + "epoch": 2.62, + "learning_rate": 1.5716984936163834e-05, + "loss": 0.059, + "step": 19977 + }, + { + "epoch": 2.62, + "learning_rate": 1.5716549815611225e-05, + "loss": 0.0595, + "step": 19978 + }, + { + "epoch": 2.62, + "learning_rate": 1.5716114678981128e-05, + "loss": 0.0527, + "step": 19979 + }, + { + "epoch": 2.62, + "learning_rate": 1.5715679526274762e-05, + "loss": 0.0464, + "step": 19980 + }, + { + "epoch": 2.62, + "learning_rate": 1.5715244357493347e-05, + "loss": 0.0586, + "step": 19981 + }, + { + "epoch": 2.62, + "learning_rate": 1.5714809172638114e-05, + "loss": 0.0578, + "step": 19982 + }, + { + "epoch": 2.62, + "learning_rate": 1.5714373971710287e-05, + "loss": 0.0544, + "step": 19983 + }, + { + "epoch": 2.62, + "learning_rate": 1.5713938754711085e-05, + "loss": 0.0499, + "step": 19984 + }, + { + "epoch": 2.62, + "learning_rate": 1.5713503521641737e-05, + "loss": 0.0609, + "step": 19985 + }, + { + "epoch": 2.62, + "learning_rate": 1.5713068272503465e-05, + "loss": 0.0524, + "step": 19986 + }, + { + "epoch": 2.62, + "learning_rate": 1.571263300729749e-05, + "loss": 0.0575, + "step": 19987 + }, + { + "epoch": 2.62, + "learning_rate": 1.5712197726025038e-05, + "loss": 0.0728, + "step": 19988 + }, + { + "epoch": 2.62, + "learning_rate": 1.5711762428687336e-05, + "loss": 0.0587, + "step": 19989 + }, + { + "epoch": 2.62, + "learning_rate": 1.5711327115285606e-05, + "loss": 0.0562, + "step": 19990 + }, + { + "epoch": 2.62, + "learning_rate": 1.5710891785821076e-05, + "loss": 0.0645, + "step": 19991 + }, + { + "epoch": 2.62, + "learning_rate": 1.5710456440294957e-05, + "loss": 0.0598, + "step": 19992 + }, + { + "epoch": 2.62, + "learning_rate": 1.571002107870849e-05, + "loss": 0.0666, + "step": 19993 + }, + { + "epoch": 2.62, + "learning_rate": 1.5709585701062897e-05, + "loss": 0.0537, + "step": 19994 + }, + { + "epoch": 2.62, + "learning_rate": 1.570915030735939e-05, + "loss": 0.0616, + "step": 19995 + }, + { + "epoch": 2.62, + "learning_rate": 1.570871489759921e-05, + "loss": 0.0526, + "step": 19996 + }, + { + "epoch": 2.62, + "learning_rate": 1.5708279471783568e-05, + "loss": 0.0678, + "step": 19997 + }, + { + "epoch": 2.62, + "learning_rate": 1.5707844029913696e-05, + "loss": 0.0574, + "step": 19998 + }, + { + "epoch": 2.62, + "learning_rate": 1.5707408571990816e-05, + "loss": 0.0642, + "step": 19999 + }, + { + "epoch": 2.62, + "learning_rate": 1.5706973098016152e-05, + "loss": 0.0563, + "step": 20000 + }, + { + "epoch": 2.62, + "learning_rate": 1.5706537607990933e-05, + "loss": 0.0641, + "step": 20001 + }, + { + "epoch": 2.62, + "learning_rate": 1.5706102101916377e-05, + "loss": 0.062, + "step": 20002 + }, + { + "epoch": 2.62, + "learning_rate": 1.5705666579793715e-05, + "loss": 0.0495, + "step": 20003 + }, + { + "epoch": 2.62, + "learning_rate": 1.5705231041624173e-05, + "loss": 0.0578, + "step": 20004 + }, + { + "epoch": 2.62, + "learning_rate": 1.5704795487408967e-05, + "loss": 0.067, + "step": 20005 + }, + { + "epoch": 2.62, + "learning_rate": 1.570435991714933e-05, + "loss": 0.0596, + "step": 20006 + }, + { + "epoch": 2.62, + "learning_rate": 1.5703924330846485e-05, + "loss": 0.0616, + "step": 20007 + }, + { + "epoch": 2.62, + "learning_rate": 1.5703488728501657e-05, + "loss": 0.0549, + "step": 20008 + }, + { + "epoch": 2.62, + "learning_rate": 1.5703053110116066e-05, + "loss": 0.0672, + "step": 20009 + }, + { + "epoch": 2.62, + "learning_rate": 1.5702617475690946e-05, + "loss": 0.0734, + "step": 20010 + }, + { + "epoch": 2.62, + "learning_rate": 1.570218182522752e-05, + "loss": 0.0587, + "step": 20011 + }, + { + "epoch": 2.62, + "learning_rate": 1.5701746158727007e-05, + "loss": 0.0507, + "step": 20012 + }, + { + "epoch": 2.62, + "learning_rate": 1.5701310476190636e-05, + "loss": 0.0584, + "step": 20013 + }, + { + "epoch": 2.62, + "learning_rate": 1.5700874777619634e-05, + "loss": 0.0591, + "step": 20014 + }, + { + "epoch": 2.62, + "learning_rate": 1.570043906301523e-05, + "loss": 0.0664, + "step": 20015 + }, + { + "epoch": 2.62, + "learning_rate": 1.570000333237864e-05, + "loss": 0.0538, + "step": 20016 + }, + { + "epoch": 2.62, + "learning_rate": 1.5699567585711094e-05, + "loss": 0.0633, + "step": 20017 + }, + { + "epoch": 2.62, + "learning_rate": 1.569913182301382e-05, + "loss": 0.0599, + "step": 20018 + }, + { + "epoch": 2.62, + "learning_rate": 1.5698696044288038e-05, + "loss": 0.0658, + "step": 20019 + }, + { + "epoch": 2.62, + "learning_rate": 1.5698260249534977e-05, + "loss": 0.058, + "step": 20020 + }, + { + "epoch": 2.62, + "learning_rate": 1.5697824438755867e-05, + "loss": 0.0597, + "step": 20021 + }, + { + "epoch": 2.62, + "learning_rate": 1.5697388611951926e-05, + "loss": 0.0552, + "step": 20022 + }, + { + "epoch": 2.62, + "learning_rate": 1.5696952769124382e-05, + "loss": 0.0582, + "step": 20023 + }, + { + "epoch": 2.62, + "learning_rate": 1.5696516910274465e-05, + "loss": 0.057, + "step": 20024 + }, + { + "epoch": 2.62, + "learning_rate": 1.5696081035403394e-05, + "loss": 0.0594, + "step": 20025 + }, + { + "epoch": 2.62, + "learning_rate": 1.5695645144512402e-05, + "loss": 0.066, + "step": 20026 + }, + { + "epoch": 2.62, + "learning_rate": 1.569520923760271e-05, + "loss": 0.062, + "step": 20027 + }, + { + "epoch": 2.62, + "learning_rate": 1.5694773314675546e-05, + "loss": 0.0665, + "step": 20028 + }, + { + "epoch": 2.62, + "learning_rate": 1.5694337375732134e-05, + "loss": 0.0621, + "step": 20029 + }, + { + "epoch": 2.62, + "learning_rate": 1.5693901420773698e-05, + "loss": 0.056, + "step": 20030 + }, + { + "epoch": 2.62, + "learning_rate": 1.5693465449801474e-05, + "loss": 0.0572, + "step": 20031 + }, + { + "epoch": 2.62, + "learning_rate": 1.5693029462816678e-05, + "loss": 0.0639, + "step": 20032 + }, + { + "epoch": 2.62, + "learning_rate": 1.569259345982054e-05, + "loss": 0.061, + "step": 20033 + }, + { + "epoch": 2.62, + "learning_rate": 1.5692157440814287e-05, + "loss": 0.0625, + "step": 20034 + }, + { + "epoch": 2.62, + "learning_rate": 1.569172140579914e-05, + "loss": 0.0588, + "step": 20035 + }, + { + "epoch": 2.62, + "learning_rate": 1.5691285354776337e-05, + "loss": 0.0552, + "step": 20036 + }, + { + "epoch": 2.62, + "learning_rate": 1.5690849287747094e-05, + "loss": 0.0513, + "step": 20037 + }, + { + "epoch": 2.62, + "learning_rate": 1.5690413204712637e-05, + "loss": 0.0623, + "step": 20038 + }, + { + "epoch": 2.62, + "learning_rate": 1.56899771056742e-05, + "loss": 0.073, + "step": 20039 + }, + { + "epoch": 2.63, + "learning_rate": 1.5689540990633003e-05, + "loss": 0.0602, + "step": 20040 + }, + { + "epoch": 2.63, + "learning_rate": 1.5689104859590275e-05, + "loss": 0.0534, + "step": 20041 + }, + { + "epoch": 2.63, + "learning_rate": 1.5688668712547243e-05, + "loss": 0.0531, + "step": 20042 + }, + { + "epoch": 2.63, + "learning_rate": 1.5688232549505134e-05, + "loss": 0.0694, + "step": 20043 + }, + { + "epoch": 2.63, + "learning_rate": 1.5687796370465175e-05, + "loss": 0.0552, + "step": 20044 + }, + { + "epoch": 2.63, + "learning_rate": 1.5687360175428593e-05, + "loss": 0.0574, + "step": 20045 + }, + { + "epoch": 2.63, + "learning_rate": 1.5686923964396606e-05, + "loss": 0.062, + "step": 20046 + }, + { + "epoch": 2.63, + "learning_rate": 1.5686487737370452e-05, + "loss": 0.0572, + "step": 20047 + }, + { + "epoch": 2.63, + "learning_rate": 1.5686051494351357e-05, + "loss": 0.0602, + "step": 20048 + }, + { + "epoch": 2.63, + "learning_rate": 1.5685615235340544e-05, + "loss": 0.0697, + "step": 20049 + }, + { + "epoch": 2.63, + "learning_rate": 1.568517896033924e-05, + "loss": 0.0598, + "step": 20050 + }, + { + "epoch": 2.63, + "learning_rate": 1.5684742669348675e-05, + "loss": 0.0565, + "step": 20051 + }, + { + "epoch": 2.63, + "learning_rate": 1.568430636237007e-05, + "loss": 0.0727, + "step": 20052 + }, + { + "epoch": 2.63, + "learning_rate": 1.5683870039404656e-05, + "loss": 0.0565, + "step": 20053 + }, + { + "epoch": 2.63, + "learning_rate": 1.5683433700453662e-05, + "loss": 0.0591, + "step": 20054 + }, + { + "epoch": 2.63, + "learning_rate": 1.5682997345518316e-05, + "loss": 0.0648, + "step": 20055 + }, + { + "epoch": 2.63, + "learning_rate": 1.5682560974599838e-05, + "loss": 0.0651, + "step": 20056 + }, + { + "epoch": 2.63, + "learning_rate": 1.568212458769946e-05, + "loss": 0.0661, + "step": 20057 + }, + { + "epoch": 2.63, + "learning_rate": 1.568168818481841e-05, + "loss": 0.0664, + "step": 20058 + }, + { + "epoch": 2.63, + "learning_rate": 1.568125176595792e-05, + "loss": 0.0539, + "step": 20059 + }, + { + "epoch": 2.63, + "learning_rate": 1.5680815331119204e-05, + "loss": 0.0613, + "step": 20060 + }, + { + "epoch": 2.63, + "learning_rate": 1.5680378880303503e-05, + "loss": 0.0636, + "step": 20061 + }, + { + "epoch": 2.63, + "learning_rate": 1.5679942413512036e-05, + "loss": 0.0649, + "step": 20062 + }, + { + "epoch": 2.63, + "learning_rate": 1.5679505930746035e-05, + "loss": 0.054, + "step": 20063 + }, + { + "epoch": 2.63, + "learning_rate": 1.5679069432006728e-05, + "loss": 0.0527, + "step": 20064 + }, + { + "epoch": 2.63, + "learning_rate": 1.5678632917295337e-05, + "loss": 0.0655, + "step": 20065 + }, + { + "epoch": 2.63, + "learning_rate": 1.5678196386613092e-05, + "loss": 0.0675, + "step": 20066 + }, + { + "epoch": 2.63, + "learning_rate": 1.5677759839961227e-05, + "loss": 0.0631, + "step": 20067 + }, + { + "epoch": 2.63, + "learning_rate": 1.5677323277340958e-05, + "loss": 0.0552, + "step": 20068 + }, + { + "epoch": 2.63, + "learning_rate": 1.5676886698753525e-05, + "loss": 0.0588, + "step": 20069 + }, + { + "epoch": 2.63, + "learning_rate": 1.5676450104200148e-05, + "loss": 0.0594, + "step": 20070 + }, + { + "epoch": 2.63, + "learning_rate": 1.567601349368206e-05, + "loss": 0.0628, + "step": 20071 + }, + { + "epoch": 2.63, + "learning_rate": 1.5675576867200484e-05, + "loss": 0.0603, + "step": 20072 + }, + { + "epoch": 2.63, + "learning_rate": 1.5675140224756653e-05, + "loss": 0.0598, + "step": 20073 + }, + { + "epoch": 2.63, + "learning_rate": 1.567470356635179e-05, + "loss": 0.0615, + "step": 20074 + }, + { + "epoch": 2.63, + "learning_rate": 1.5674266891987127e-05, + "loss": 0.0551, + "step": 20075 + }, + { + "epoch": 2.63, + "learning_rate": 1.5673830201663885e-05, + "loss": 0.0507, + "step": 20076 + }, + { + "epoch": 2.63, + "learning_rate": 1.5673393495383305e-05, + "loss": 0.0498, + "step": 20077 + }, + { + "epoch": 2.63, + "learning_rate": 1.5672956773146606e-05, + "loss": 0.0745, + "step": 20078 + }, + { + "epoch": 2.63, + "learning_rate": 1.5672520034955018e-05, + "loss": 0.0675, + "step": 20079 + }, + { + "epoch": 2.63, + "learning_rate": 1.5672083280809767e-05, + "loss": 0.0575, + "step": 20080 + }, + { + "epoch": 2.63, + "learning_rate": 1.5671646510712085e-05, + "loss": 0.0549, + "step": 20081 + }, + { + "epoch": 2.63, + "learning_rate": 1.56712097246632e-05, + "loss": 0.0601, + "step": 20082 + }, + { + "epoch": 2.63, + "learning_rate": 1.5670772922664342e-05, + "loss": 0.0626, + "step": 20083 + }, + { + "epoch": 2.63, + "learning_rate": 1.567033610471673e-05, + "loss": 0.0584, + "step": 20084 + }, + { + "epoch": 2.63, + "learning_rate": 1.5669899270821605e-05, + "loss": 0.0544, + "step": 20085 + }, + { + "epoch": 2.63, + "learning_rate": 1.5669462420980192e-05, + "loss": 0.0674, + "step": 20086 + }, + { + "epoch": 2.63, + "learning_rate": 1.5669025555193712e-05, + "loss": 0.064, + "step": 20087 + }, + { + "epoch": 2.63, + "learning_rate": 1.5668588673463406e-05, + "loss": 0.0562, + "step": 20088 + }, + { + "epoch": 2.63, + "learning_rate": 1.5668151775790493e-05, + "loss": 0.0683, + "step": 20089 + }, + { + "epoch": 2.63, + "learning_rate": 1.5667714862176203e-05, + "loss": 0.0755, + "step": 20090 + }, + { + "epoch": 2.63, + "learning_rate": 1.5667277932621773e-05, + "loss": 0.0667, + "step": 20091 + }, + { + "epoch": 2.63, + "learning_rate": 1.566684098712842e-05, + "loss": 0.061, + "step": 20092 + }, + { + "epoch": 2.63, + "learning_rate": 1.5666404025697378e-05, + "loss": 0.0714, + "step": 20093 + }, + { + "epoch": 2.63, + "learning_rate": 1.566596704832988e-05, + "loss": 0.0434, + "step": 20094 + }, + { + "epoch": 2.63, + "learning_rate": 1.5665530055027153e-05, + "loss": 0.0744, + "step": 20095 + }, + { + "epoch": 2.63, + "learning_rate": 1.5665093045790416e-05, + "loss": 0.0518, + "step": 20096 + }, + { + "epoch": 2.63, + "learning_rate": 1.5664656020620914e-05, + "loss": 0.0583, + "step": 20097 + }, + { + "epoch": 2.63, + "learning_rate": 1.566421897951987e-05, + "loss": 0.0719, + "step": 20098 + }, + { + "epoch": 2.63, + "learning_rate": 1.5663781922488504e-05, + "loss": 0.0616, + "step": 20099 + }, + { + "epoch": 2.63, + "learning_rate": 1.5663344849528056e-05, + "loss": 0.0613, + "step": 20100 + }, + { + "epoch": 2.63, + "learning_rate": 1.5662907760639755e-05, + "loss": 0.0557, + "step": 20101 + }, + { + "epoch": 2.63, + "learning_rate": 1.5662470655824823e-05, + "loss": 0.0592, + "step": 20102 + }, + { + "epoch": 2.63, + "learning_rate": 1.5662033535084496e-05, + "loss": 0.0628, + "step": 20103 + }, + { + "epoch": 2.63, + "learning_rate": 1.5661596398420004e-05, + "loss": 0.0576, + "step": 20104 + }, + { + "epoch": 2.63, + "learning_rate": 1.5661159245832573e-05, + "loss": 0.0531, + "step": 20105 + }, + { + "epoch": 2.63, + "learning_rate": 1.5660722077323433e-05, + "loss": 0.0596, + "step": 20106 + }, + { + "epoch": 2.63, + "learning_rate": 1.5660284892893807e-05, + "loss": 0.0662, + "step": 20107 + }, + { + "epoch": 2.63, + "learning_rate": 1.565984769254494e-05, + "loss": 0.061, + "step": 20108 + }, + { + "epoch": 2.63, + "learning_rate": 1.5659410476278046e-05, + "loss": 0.0614, + "step": 20109 + }, + { + "epoch": 2.63, + "learning_rate": 1.5658973244094367e-05, + "loss": 0.0526, + "step": 20110 + }, + { + "epoch": 2.63, + "learning_rate": 1.5658535995995124e-05, + "loss": 0.0624, + "step": 20111 + }, + { + "epoch": 2.63, + "learning_rate": 1.565809873198155e-05, + "loss": 0.0528, + "step": 20112 + }, + { + "epoch": 2.63, + "learning_rate": 1.5657661452054875e-05, + "loss": 0.0492, + "step": 20113 + }, + { + "epoch": 2.63, + "learning_rate": 1.5657224156216327e-05, + "loss": 0.0527, + "step": 20114 + }, + { + "epoch": 2.63, + "learning_rate": 1.565678684446714e-05, + "loss": 0.0593, + "step": 20115 + }, + { + "epoch": 2.64, + "learning_rate": 1.565634951680854e-05, + "loss": 0.042, + "step": 20116 + }, + { + "epoch": 2.64, + "learning_rate": 1.565591217324176e-05, + "loss": 0.0654, + "step": 20117 + }, + { + "epoch": 2.64, + "learning_rate": 1.5655474813768027e-05, + "loss": 0.0457, + "step": 20118 + }, + { + "epoch": 2.64, + "learning_rate": 1.565503743838857e-05, + "loss": 0.0639, + "step": 20119 + }, + { + "epoch": 2.64, + "learning_rate": 1.565460004710462e-05, + "loss": 0.0568, + "step": 20120 + }, + { + "epoch": 2.64, + "learning_rate": 1.5654162639917413e-05, + "loss": 0.0558, + "step": 20121 + }, + { + "epoch": 2.64, + "learning_rate": 1.5653725216828173e-05, + "loss": 0.0552, + "step": 20122 + }, + { + "epoch": 2.64, + "learning_rate": 1.5653287777838134e-05, + "loss": 0.0534, + "step": 20123 + }, + { + "epoch": 2.64, + "learning_rate": 1.565285032294852e-05, + "loss": 0.0576, + "step": 20124 + }, + { + "epoch": 2.64, + "learning_rate": 1.565241285216057e-05, + "loss": 0.0595, + "step": 20125 + }, + { + "epoch": 2.64, + "learning_rate": 1.5651975365475506e-05, + "loss": 0.0546, + "step": 20126 + }, + { + "epoch": 2.64, + "learning_rate": 1.5651537862894566e-05, + "loss": 0.0603, + "step": 20127 + }, + { + "epoch": 2.64, + "learning_rate": 1.5651100344418975e-05, + "loss": 0.058, + "step": 20128 + }, + { + "epoch": 2.64, + "learning_rate": 1.5650662810049963e-05, + "loss": 0.0626, + "step": 20129 + }, + { + "epoch": 2.64, + "learning_rate": 1.5650225259788765e-05, + "loss": 0.0632, + "step": 20130 + }, + { + "epoch": 2.64, + "learning_rate": 1.564978769363661e-05, + "loss": 0.0559, + "step": 20131 + }, + { + "epoch": 2.64, + "learning_rate": 1.5649350111594727e-05, + "loss": 0.0495, + "step": 20132 + }, + { + "epoch": 2.64, + "learning_rate": 1.564891251366435e-05, + "loss": 0.0657, + "step": 20133 + }, + { + "epoch": 2.64, + "learning_rate": 1.56484748998467e-05, + "loss": 0.0606, + "step": 20134 + }, + { + "epoch": 2.64, + "learning_rate": 1.5648037270143023e-05, + "loss": 0.0563, + "step": 20135 + }, + { + "epoch": 2.64, + "learning_rate": 1.564759962455454e-05, + "loss": 0.063, + "step": 20136 + }, + { + "epoch": 2.64, + "learning_rate": 1.5647161963082483e-05, + "loss": 0.0629, + "step": 20137 + }, + { + "epoch": 2.64, + "learning_rate": 1.5646724285728082e-05, + "loss": 0.0471, + "step": 20138 + }, + { + "epoch": 2.64, + "learning_rate": 1.5646286592492572e-05, + "loss": 0.0662, + "step": 20139 + }, + { + "epoch": 2.64, + "learning_rate": 1.5645848883377183e-05, + "loss": 0.0716, + "step": 20140 + }, + { + "epoch": 2.64, + "learning_rate": 1.5645411158383138e-05, + "loss": 0.0589, + "step": 20141 + }, + { + "epoch": 2.64, + "learning_rate": 1.564497341751168e-05, + "loss": 0.0678, + "step": 20142 + }, + { + "epoch": 2.64, + "learning_rate": 1.5644535660764034e-05, + "loss": 0.0597, + "step": 20143 + }, + { + "epoch": 2.64, + "learning_rate": 1.5644097888141435e-05, + "loss": 0.0694, + "step": 20144 + }, + { + "epoch": 2.64, + "learning_rate": 1.564366009964511e-05, + "loss": 0.0587, + "step": 20145 + }, + { + "epoch": 2.64, + "learning_rate": 1.5643222295276288e-05, + "loss": 0.0523, + "step": 20146 + }, + { + "epoch": 2.64, + "learning_rate": 1.5642784475036207e-05, + "loss": 0.059, + "step": 20147 + }, + { + "epoch": 2.64, + "learning_rate": 1.5642346638926093e-05, + "loss": 0.0575, + "step": 20148 + }, + { + "epoch": 2.64, + "learning_rate": 1.564190878694718e-05, + "loss": 0.0599, + "step": 20149 + }, + { + "epoch": 2.64, + "learning_rate": 1.56414709191007e-05, + "loss": 0.0626, + "step": 20150 + }, + { + "epoch": 2.64, + "learning_rate": 1.5641033035387883e-05, + "loss": 0.0607, + "step": 20151 + }, + { + "epoch": 2.64, + "learning_rate": 1.564059513580996e-05, + "loss": 0.0523, + "step": 20152 + }, + { + "epoch": 2.64, + "learning_rate": 1.5640157220368166e-05, + "loss": 0.061, + "step": 20153 + }, + { + "epoch": 2.64, + "learning_rate": 1.563971928906373e-05, + "loss": 0.0576, + "step": 20154 + }, + { + "epoch": 2.64, + "learning_rate": 1.563928134189788e-05, + "loss": 0.0661, + "step": 20155 + }, + { + "epoch": 2.64, + "learning_rate": 1.563884337887185e-05, + "loss": 0.0533, + "step": 20156 + }, + { + "epoch": 2.64, + "learning_rate": 1.5638405399986884e-05, + "loss": 0.0655, + "step": 20157 + }, + { + "epoch": 2.64, + "learning_rate": 1.5637967405244196e-05, + "loss": 0.0557, + "step": 20158 + }, + { + "epoch": 2.64, + "learning_rate": 1.5637529394645024e-05, + "loss": 0.0658, + "step": 20159 + }, + { + "epoch": 2.64, + "learning_rate": 1.5637091368190604e-05, + "loss": 0.0602, + "step": 20160 + }, + { + "epoch": 2.64, + "learning_rate": 1.563665332588216e-05, + "loss": 0.0623, + "step": 20161 + }, + { + "epoch": 2.64, + "learning_rate": 1.563621526772093e-05, + "loss": 0.0585, + "step": 20162 + }, + { + "epoch": 2.64, + "learning_rate": 1.5635777193708152e-05, + "loss": 0.0565, + "step": 20163 + }, + { + "epoch": 2.64, + "learning_rate": 1.5635339103845043e-05, + "loss": 0.0607, + "step": 20164 + }, + { + "epoch": 2.64, + "learning_rate": 1.5634900998132842e-05, + "loss": 0.0585, + "step": 20165 + }, + { + "epoch": 2.64, + "learning_rate": 1.5634462876572788e-05, + "loss": 0.0583, + "step": 20166 + }, + { + "epoch": 2.64, + "learning_rate": 1.5634024739166102e-05, + "loss": 0.0632, + "step": 20167 + }, + { + "epoch": 2.64, + "learning_rate": 1.5633586585914025e-05, + "loss": 0.0542, + "step": 20168 + }, + { + "epoch": 2.64, + "learning_rate": 1.5633148416817784e-05, + "loss": 0.0549, + "step": 20169 + }, + { + "epoch": 2.64, + "learning_rate": 1.5632710231878614e-05, + "loss": 0.0575, + "step": 20170 + }, + { + "epoch": 2.64, + "learning_rate": 1.5632272031097747e-05, + "loss": 0.0564, + "step": 20171 + }, + { + "epoch": 2.64, + "learning_rate": 1.563183381447641e-05, + "loss": 0.0601, + "step": 20172 + }, + { + "epoch": 2.64, + "learning_rate": 1.5631395582015847e-05, + "loss": 0.0634, + "step": 20173 + }, + { + "epoch": 2.64, + "learning_rate": 1.5630957333717282e-05, + "loss": 0.0651, + "step": 20174 + }, + { + "epoch": 2.64, + "learning_rate": 1.5630519069581945e-05, + "loss": 0.0562, + "step": 20175 + }, + { + "epoch": 2.64, + "learning_rate": 1.5630080789611076e-05, + "loss": 0.0499, + "step": 20176 + }, + { + "epoch": 2.64, + "learning_rate": 1.5629642493805905e-05, + "loss": 0.0689, + "step": 20177 + }, + { + "epoch": 2.64, + "learning_rate": 1.5629204182167663e-05, + "loss": 0.0708, + "step": 20178 + }, + { + "epoch": 2.64, + "learning_rate": 1.562876585469759e-05, + "loss": 0.0583, + "step": 20179 + }, + { + "epoch": 2.64, + "learning_rate": 1.5628327511396902e-05, + "loss": 0.0482, + "step": 20180 + }, + { + "epoch": 2.64, + "learning_rate": 1.562788915226685e-05, + "loss": 0.0615, + "step": 20181 + }, + { + "epoch": 2.64, + "learning_rate": 1.562745077730866e-05, + "loss": 0.0707, + "step": 20182 + }, + { + "epoch": 2.64, + "learning_rate": 1.562701238652356e-05, + "loss": 0.0665, + "step": 20183 + }, + { + "epoch": 2.64, + "learning_rate": 1.5626573979912793e-05, + "loss": 0.0626, + "step": 20184 + }, + { + "epoch": 2.64, + "learning_rate": 1.562613555747758e-05, + "loss": 0.0601, + "step": 20185 + }, + { + "epoch": 2.64, + "learning_rate": 1.5625697119219166e-05, + "loss": 0.0609, + "step": 20186 + }, + { + "epoch": 2.64, + "learning_rate": 1.5625258665138774e-05, + "loss": 0.0529, + "step": 20187 + }, + { + "epoch": 2.64, + "learning_rate": 1.5624820195237643e-05, + "loss": 0.0581, + "step": 20188 + }, + { + "epoch": 2.64, + "learning_rate": 1.5624381709517003e-05, + "loss": 0.0606, + "step": 20189 + }, + { + "epoch": 2.64, + "learning_rate": 1.562394320797809e-05, + "loss": 0.0539, + "step": 20190 + }, + { + "epoch": 2.64, + "learning_rate": 1.562350469062214e-05, + "loss": 0.054, + "step": 20191 + }, + { + "epoch": 2.65, + "learning_rate": 1.562306615745038e-05, + "loss": 0.0554, + "step": 20192 + }, + { + "epoch": 2.65, + "learning_rate": 1.5622627608464046e-05, + "loss": 0.0566, + "step": 20193 + }, + { + "epoch": 2.65, + "learning_rate": 1.562218904366437e-05, + "loss": 0.0527, + "step": 20194 + }, + { + "epoch": 2.65, + "learning_rate": 1.5621750463052587e-05, + "loss": 0.0632, + "step": 20195 + }, + { + "epoch": 2.65, + "learning_rate": 1.562131186662993e-05, + "loss": 0.0531, + "step": 20196 + }, + { + "epoch": 2.65, + "learning_rate": 1.5620873254397636e-05, + "loss": 0.0475, + "step": 20197 + }, + { + "epoch": 2.65, + "learning_rate": 1.5620434626356934e-05, + "loss": 0.0668, + "step": 20198 + }, + { + "epoch": 2.65, + "learning_rate": 1.5619995982509056e-05, + "loss": 0.0574, + "step": 20199 + }, + { + "epoch": 2.65, + "learning_rate": 1.561955732285524e-05, + "loss": 0.0548, + "step": 20200 + }, + { + "epoch": 2.65, + "learning_rate": 1.5619118647396718e-05, + "loss": 0.056, + "step": 20201 + }, + { + "epoch": 2.65, + "learning_rate": 1.5618679956134727e-05, + "loss": 0.0639, + "step": 20202 + }, + { + "epoch": 2.65, + "learning_rate": 1.5618241249070493e-05, + "loss": 0.0703, + "step": 20203 + }, + { + "epoch": 2.65, + "learning_rate": 1.5617802526205256e-05, + "loss": 0.0549, + "step": 20204 + }, + { + "epoch": 2.65, + "learning_rate": 1.561736378754025e-05, + "loss": 0.0596, + "step": 20205 + }, + { + "epoch": 2.65, + "learning_rate": 1.561692503307671e-05, + "loss": 0.0634, + "step": 20206 + }, + { + "epoch": 2.65, + "learning_rate": 1.5616486262815862e-05, + "loss": 0.0615, + "step": 20207 + }, + { + "epoch": 2.65, + "learning_rate": 1.5616047476758948e-05, + "loss": 0.0657, + "step": 20208 + }, + { + "epoch": 2.65, + "learning_rate": 1.56156086749072e-05, + "loss": 0.0671, + "step": 20209 + }, + { + "epoch": 2.65, + "learning_rate": 1.5615169857261854e-05, + "loss": 0.052, + "step": 20210 + }, + { + "epoch": 2.65, + "learning_rate": 1.5614731023824136e-05, + "loss": 0.0503, + "step": 20211 + }, + { + "epoch": 2.65, + "learning_rate": 1.5614292174595292e-05, + "loss": 0.0576, + "step": 20212 + }, + { + "epoch": 2.65, + "learning_rate": 1.5613853309576547e-05, + "loss": 0.0549, + "step": 20213 + }, + { + "epoch": 2.65, + "learning_rate": 1.561341442876914e-05, + "loss": 0.0643, + "step": 20214 + }, + { + "epoch": 2.65, + "learning_rate": 1.56129755321743e-05, + "loss": 0.0639, + "step": 20215 + }, + { + "epoch": 2.65, + "learning_rate": 1.5612536619793268e-05, + "loss": 0.0573, + "step": 20216 + }, + { + "epoch": 2.65, + "learning_rate": 1.5612097691627275e-05, + "loss": 0.0602, + "step": 20217 + }, + { + "epoch": 2.65, + "learning_rate": 1.5611658747677555e-05, + "loss": 0.0662, + "step": 20218 + }, + { + "epoch": 2.65, + "learning_rate": 1.5611219787945348e-05, + "loss": 0.062, + "step": 20219 + }, + { + "epoch": 2.65, + "learning_rate": 1.561078081243188e-05, + "loss": 0.0577, + "step": 20220 + }, + { + "epoch": 2.65, + "learning_rate": 1.5610341821138392e-05, + "loss": 0.0661, + "step": 20221 + }, + { + "epoch": 2.65, + "learning_rate": 1.5609902814066117e-05, + "loss": 0.0535, + "step": 20222 + }, + { + "epoch": 2.65, + "learning_rate": 1.5609463791216287e-05, + "loss": 0.0508, + "step": 20223 + }, + { + "epoch": 2.65, + "learning_rate": 1.560902475259014e-05, + "loss": 0.0656, + "step": 20224 + }, + { + "epoch": 2.65, + "learning_rate": 1.560858569818891e-05, + "loss": 0.0646, + "step": 20225 + }, + { + "epoch": 2.65, + "learning_rate": 1.560814662801383e-05, + "loss": 0.0575, + "step": 20226 + }, + { + "epoch": 2.65, + "learning_rate": 1.5607707542066136e-05, + "loss": 0.0582, + "step": 20227 + }, + { + "epoch": 2.65, + "learning_rate": 1.5607268440347064e-05, + "loss": 0.0553, + "step": 20228 + }, + { + "epoch": 2.65, + "learning_rate": 1.5606829322857847e-05, + "loss": 0.0644, + "step": 20229 + }, + { + "epoch": 2.65, + "learning_rate": 1.5606390189599722e-05, + "loss": 0.0517, + "step": 20230 + }, + { + "epoch": 2.65, + "learning_rate": 1.5605951040573924e-05, + "loss": 0.0483, + "step": 20231 + }, + { + "epoch": 2.65, + "learning_rate": 1.5605511875781684e-05, + "loss": 0.0614, + "step": 20232 + }, + { + "epoch": 2.65, + "learning_rate": 1.5605072695224244e-05, + "loss": 0.0552, + "step": 20233 + }, + { + "epoch": 2.65, + "learning_rate": 1.5604633498902836e-05, + "loss": 0.0617, + "step": 20234 + }, + { + "epoch": 2.65, + "learning_rate": 1.5604194286818694e-05, + "loss": 0.0645, + "step": 20235 + }, + { + "epoch": 2.65, + "learning_rate": 1.560375505897305e-05, + "loss": 0.0589, + "step": 20236 + }, + { + "epoch": 2.65, + "learning_rate": 1.560331581536715e-05, + "loss": 0.0547, + "step": 20237 + }, + { + "epoch": 2.65, + "learning_rate": 1.5602876556002215e-05, + "loss": 0.056, + "step": 20238 + }, + { + "epoch": 2.65, + "learning_rate": 1.5602437280879496e-05, + "loss": 0.0679, + "step": 20239 + }, + { + "epoch": 2.65, + "learning_rate": 1.5601997990000216e-05, + "loss": 0.0574, + "step": 20240 + }, + { + "epoch": 2.65, + "learning_rate": 1.5601558683365614e-05, + "loss": 0.0614, + "step": 20241 + }, + { + "epoch": 2.65, + "learning_rate": 1.5601119360976932e-05, + "loss": 0.0665, + "step": 20242 + }, + { + "epoch": 2.65, + "learning_rate": 1.560068002283539e-05, + "loss": 0.0667, + "step": 20243 + }, + { + "epoch": 2.65, + "learning_rate": 1.5600240668942244e-05, + "loss": 0.0645, + "step": 20244 + }, + { + "epoch": 2.65, + "learning_rate": 1.5599801299298715e-05, + "loss": 0.0636, + "step": 20245 + }, + { + "epoch": 2.65, + "learning_rate": 1.5599361913906044e-05, + "loss": 0.055, + "step": 20246 + }, + { + "epoch": 2.65, + "learning_rate": 1.5598922512765463e-05, + "loss": 0.0566, + "step": 20247 + }, + { + "epoch": 2.65, + "learning_rate": 1.5598483095878214e-05, + "loss": 0.0567, + "step": 20248 + }, + { + "epoch": 2.65, + "learning_rate": 1.5598043663245527e-05, + "loss": 0.0619, + "step": 20249 + }, + { + "epoch": 2.65, + "learning_rate": 1.5597604214868642e-05, + "loss": 0.0496, + "step": 20250 + }, + { + "epoch": 2.65, + "learning_rate": 1.5597164750748793e-05, + "loss": 0.0539, + "step": 20251 + }, + { + "epoch": 2.65, + "learning_rate": 1.5596725270887216e-05, + "loss": 0.0629, + "step": 20252 + }, + { + "epoch": 2.65, + "learning_rate": 1.559628577528515e-05, + "loss": 0.0524, + "step": 20253 + }, + { + "epoch": 2.65, + "learning_rate": 1.559584626394382e-05, + "loss": 0.0585, + "step": 20254 + }, + { + "epoch": 2.65, + "learning_rate": 1.5595406736864477e-05, + "loss": 0.0487, + "step": 20255 + }, + { + "epoch": 2.65, + "learning_rate": 1.5594967194048352e-05, + "loss": 0.064, + "step": 20256 + }, + { + "epoch": 2.65, + "learning_rate": 1.5594527635496675e-05, + "loss": 0.0562, + "step": 20257 + }, + { + "epoch": 2.65, + "learning_rate": 1.559408806121069e-05, + "loss": 0.0569, + "step": 20258 + }, + { + "epoch": 2.65, + "learning_rate": 1.559364847119163e-05, + "loss": 0.0636, + "step": 20259 + }, + { + "epoch": 2.65, + "learning_rate": 1.559320886544073e-05, + "loss": 0.0669, + "step": 20260 + }, + { + "epoch": 2.65, + "learning_rate": 1.5592769243959227e-05, + "loss": 0.0605, + "step": 20261 + }, + { + "epoch": 2.65, + "learning_rate": 1.559232960674836e-05, + "loss": 0.0613, + "step": 20262 + }, + { + "epoch": 2.65, + "learning_rate": 1.5591889953809367e-05, + "loss": 0.0678, + "step": 20263 + }, + { + "epoch": 2.65, + "learning_rate": 1.559145028514348e-05, + "loss": 0.054, + "step": 20264 + }, + { + "epoch": 2.65, + "learning_rate": 1.559101060075193e-05, + "loss": 0.0643, + "step": 20265 + }, + { + "epoch": 2.65, + "learning_rate": 1.5590570900635968e-05, + "loss": 0.0641, + "step": 20266 + }, + { + "epoch": 2.65, + "learning_rate": 1.559013118479682e-05, + "loss": 0.0558, + "step": 20267 + }, + { + "epoch": 2.65, + "learning_rate": 1.5589691453235725e-05, + "loss": 0.0511, + "step": 20268 + }, + { + "epoch": 2.66, + "learning_rate": 1.558925170595392e-05, + "loss": 0.06, + "step": 20269 + }, + { + "epoch": 2.66, + "learning_rate": 1.5588811942952648e-05, + "loss": 0.0657, + "step": 20270 + }, + { + "epoch": 2.66, + "learning_rate": 1.5588372164233134e-05, + "loss": 0.0519, + "step": 20271 + }, + { + "epoch": 2.66, + "learning_rate": 1.5587932369796622e-05, + "loss": 0.0582, + "step": 20272 + }, + { + "epoch": 2.66, + "learning_rate": 1.558749255964435e-05, + "loss": 0.0633, + "step": 20273 + }, + { + "epoch": 2.66, + "learning_rate": 1.558705273377755e-05, + "loss": 0.069, + "step": 20274 + }, + { + "epoch": 2.66, + "learning_rate": 1.5586612892197463e-05, + "loss": 0.0635, + "step": 20275 + }, + { + "epoch": 2.66, + "learning_rate": 1.5586173034905323e-05, + "loss": 0.0624, + "step": 20276 + }, + { + "epoch": 2.66, + "learning_rate": 1.558573316190237e-05, + "loss": 0.0565, + "step": 20277 + }, + { + "epoch": 2.66, + "learning_rate": 1.5585293273189842e-05, + "loss": 0.061, + "step": 20278 + }, + { + "epoch": 2.66, + "learning_rate": 1.5584853368768972e-05, + "loss": 0.0649, + "step": 20279 + }, + { + "epoch": 2.66, + "learning_rate": 1.5584413448641004e-05, + "loss": 0.0488, + "step": 20280 + }, + { + "epoch": 2.66, + "learning_rate": 1.5583973512807165e-05, + "loss": 0.0619, + "step": 20281 + }, + { + "epoch": 2.66, + "learning_rate": 1.5583533561268698e-05, + "loss": 0.0711, + "step": 20282 + }, + { + "epoch": 2.66, + "learning_rate": 1.5583093594026843e-05, + "loss": 0.0674, + "step": 20283 + }, + { + "epoch": 2.66, + "learning_rate": 1.5582653611082833e-05, + "loss": 0.0632, + "step": 20284 + }, + { + "epoch": 2.66, + "learning_rate": 1.5582213612437906e-05, + "loss": 0.0593, + "step": 20285 + }, + { + "epoch": 2.66, + "learning_rate": 1.55817735980933e-05, + "loss": 0.056, + "step": 20286 + }, + { + "epoch": 2.66, + "learning_rate": 1.5581333568050255e-05, + "loss": 0.0599, + "step": 20287 + }, + { + "epoch": 2.66, + "learning_rate": 1.558089352231001e-05, + "loss": 0.0566, + "step": 20288 + }, + { + "epoch": 2.66, + "learning_rate": 1.5580453460873794e-05, + "loss": 0.0569, + "step": 20289 + }, + { + "epoch": 2.66, + "learning_rate": 1.558001338374285e-05, + "loss": 0.0668, + "step": 20290 + }, + { + "epoch": 2.66, + "learning_rate": 1.5579573290918418e-05, + "loss": 0.0549, + "step": 20291 + }, + { + "epoch": 2.66, + "learning_rate": 1.557913318240173e-05, + "loss": 0.0558, + "step": 20292 + }, + { + "epoch": 2.66, + "learning_rate": 1.557869305819403e-05, + "loss": 0.064, + "step": 20293 + }, + { + "epoch": 2.66, + "learning_rate": 1.557825291829655e-05, + "loss": 0.0568, + "step": 20294 + }, + { + "epoch": 2.66, + "learning_rate": 1.5577812762710534e-05, + "loss": 0.053, + "step": 20295 + }, + { + "epoch": 2.66, + "learning_rate": 1.5577372591437216e-05, + "loss": 0.0655, + "step": 20296 + }, + { + "epoch": 2.66, + "learning_rate": 1.5576932404477833e-05, + "loss": 0.0677, + "step": 20297 + }, + { + "epoch": 2.66, + "learning_rate": 1.5576492201833623e-05, + "loss": 0.0651, + "step": 20298 + }, + { + "epoch": 2.66, + "learning_rate": 1.557605198350583e-05, + "loss": 0.0532, + "step": 20299 + }, + { + "epoch": 2.66, + "learning_rate": 1.5575611749495684e-05, + "loss": 0.0644, + "step": 20300 + }, + { + "epoch": 2.66, + "learning_rate": 1.5575171499804432e-05, + "loss": 0.0618, + "step": 20301 + }, + { + "epoch": 2.66, + "learning_rate": 1.5574731234433302e-05, + "loss": 0.0604, + "step": 20302 + }, + { + "epoch": 2.66, + "learning_rate": 1.5574290953383538e-05, + "loss": 0.0647, + "step": 20303 + }, + { + "epoch": 2.66, + "learning_rate": 1.557385065665638e-05, + "loss": 0.0606, + "step": 20304 + }, + { + "epoch": 2.66, + "learning_rate": 1.5573410344253065e-05, + "loss": 0.057, + "step": 20305 + }, + { + "epoch": 2.66, + "learning_rate": 1.5572970016174825e-05, + "loss": 0.0592, + "step": 20306 + }, + { + "epoch": 2.66, + "learning_rate": 1.5572529672422905e-05, + "loss": 0.0533, + "step": 20307 + }, + { + "epoch": 2.66, + "learning_rate": 1.5572089312998545e-05, + "loss": 0.0546, + "step": 20308 + }, + { + "epoch": 2.66, + "learning_rate": 1.5571648937902977e-05, + "loss": 0.0567, + "step": 20309 + }, + { + "epoch": 2.66, + "learning_rate": 1.5571208547137445e-05, + "loss": 0.0546, + "step": 20310 + }, + { + "epoch": 2.66, + "learning_rate": 1.5570768140703187e-05, + "loss": 0.0607, + "step": 20311 + }, + { + "epoch": 2.66, + "learning_rate": 1.557032771860144e-05, + "loss": 0.0558, + "step": 20312 + }, + { + "epoch": 2.66, + "learning_rate": 1.556988728083344e-05, + "loss": 0.0599, + "step": 20313 + }, + { + "epoch": 2.66, + "learning_rate": 1.556944682740043e-05, + "loss": 0.054, + "step": 20314 + }, + { + "epoch": 2.66, + "learning_rate": 1.5569006358303647e-05, + "loss": 0.0614, + "step": 20315 + }, + { + "epoch": 2.66, + "learning_rate": 1.556856587354433e-05, + "loss": 0.0558, + "step": 20316 + }, + { + "epoch": 2.66, + "learning_rate": 1.5568125373123718e-05, + "loss": 0.0632, + "step": 20317 + }, + { + "epoch": 2.66, + "learning_rate": 1.5567684857043053e-05, + "loss": 0.0527, + "step": 20318 + }, + { + "epoch": 2.66, + "learning_rate": 1.556724432530357e-05, + "loss": 0.062, + "step": 20319 + }, + { + "epoch": 2.66, + "learning_rate": 1.5566803777906507e-05, + "loss": 0.0571, + "step": 20320 + }, + { + "epoch": 2.66, + "learning_rate": 1.5566363214853105e-05, + "loss": 0.0654, + "step": 20321 + }, + { + "epoch": 2.66, + "learning_rate": 1.5565922636144603e-05, + "loss": 0.0684, + "step": 20322 + }, + { + "epoch": 2.66, + "learning_rate": 1.556548204178224e-05, + "loss": 0.0666, + "step": 20323 + }, + { + "epoch": 2.66, + "learning_rate": 1.5565041431767255e-05, + "loss": 0.0591, + "step": 20324 + }, + { + "epoch": 2.66, + "learning_rate": 1.5564600806100886e-05, + "loss": 0.0647, + "step": 20325 + }, + { + "epoch": 2.66, + "learning_rate": 1.5564160164784378e-05, + "loss": 0.0649, + "step": 20326 + }, + { + "epoch": 2.66, + "learning_rate": 1.5563719507818963e-05, + "loss": 0.0553, + "step": 20327 + }, + { + "epoch": 2.66, + "learning_rate": 1.5563278835205884e-05, + "loss": 0.0656, + "step": 20328 + }, + { + "epoch": 2.66, + "learning_rate": 1.5562838146946376e-05, + "loss": 0.0586, + "step": 20329 + }, + { + "epoch": 2.66, + "learning_rate": 1.5562397443041685e-05, + "loss": 0.0628, + "step": 20330 + }, + { + "epoch": 2.66, + "learning_rate": 1.5561956723493047e-05, + "loss": 0.0546, + "step": 20331 + }, + { + "epoch": 2.66, + "learning_rate": 1.5561515988301703e-05, + "loss": 0.064, + "step": 20332 + }, + { + "epoch": 2.66, + "learning_rate": 1.556107523746889e-05, + "loss": 0.0592, + "step": 20333 + }, + { + "epoch": 2.66, + "learning_rate": 1.556063447099585e-05, + "loss": 0.0588, + "step": 20334 + }, + { + "epoch": 2.66, + "learning_rate": 1.5560193688883823e-05, + "loss": 0.0596, + "step": 20335 + }, + { + "epoch": 2.66, + "learning_rate": 1.5559752891134045e-05, + "loss": 0.0546, + "step": 20336 + }, + { + "epoch": 2.66, + "learning_rate": 1.5559312077747753e-05, + "loss": 0.057, + "step": 20337 + }, + { + "epoch": 2.66, + "learning_rate": 1.5558871248726202e-05, + "loss": 0.0646, + "step": 20338 + }, + { + "epoch": 2.66, + "learning_rate": 1.555843040407062e-05, + "loss": 0.0591, + "step": 20339 + }, + { + "epoch": 2.66, + "learning_rate": 1.5557989543782242e-05, + "loss": 0.0566, + "step": 20340 + }, + { + "epoch": 2.66, + "learning_rate": 1.555754866786232e-05, + "loss": 0.0622, + "step": 20341 + }, + { + "epoch": 2.66, + "learning_rate": 1.5557107776312082e-05, + "loss": 0.0548, + "step": 20342 + }, + { + "epoch": 2.66, + "learning_rate": 1.555666686913278e-05, + "loss": 0.0644, + "step": 20343 + }, + { + "epoch": 2.66, + "learning_rate": 1.5556225946325646e-05, + "loss": 0.0623, + "step": 20344 + }, + { + "epoch": 2.67, + "learning_rate": 1.5555785007891925e-05, + "loss": 0.0518, + "step": 20345 + }, + { + "epoch": 2.67, + "learning_rate": 1.5555344053832854e-05, + "loss": 0.0529, + "step": 20346 + }, + { + "epoch": 2.67, + "learning_rate": 1.555490308414967e-05, + "loss": 0.0634, + "step": 20347 + }, + { + "epoch": 2.67, + "learning_rate": 1.555446209884362e-05, + "loss": 0.056, + "step": 20348 + }, + { + "epoch": 2.67, + "learning_rate": 1.5554021097915943e-05, + "loss": 0.058, + "step": 20349 + }, + { + "epoch": 2.67, + "learning_rate": 1.5553580081367874e-05, + "loss": 0.0514, + "step": 20350 + }, + { + "epoch": 2.67, + "learning_rate": 1.555313904920066e-05, + "loss": 0.0546, + "step": 20351 + }, + { + "epoch": 2.67, + "learning_rate": 1.5552698001415533e-05, + "loss": 0.0727, + "step": 20352 + }, + { + "epoch": 2.67, + "learning_rate": 1.5552256938013742e-05, + "loss": 0.058, + "step": 20353 + }, + { + "epoch": 2.67, + "learning_rate": 1.5551815858996522e-05, + "loss": 0.0548, + "step": 20354 + }, + { + "epoch": 2.67, + "learning_rate": 1.555137476436512e-05, + "loss": 0.0531, + "step": 20355 + }, + { + "epoch": 2.67, + "learning_rate": 1.5550933654120767e-05, + "loss": 0.0713, + "step": 20356 + }, + { + "epoch": 2.67, + "learning_rate": 1.5550492528264713e-05, + "loss": 0.0574, + "step": 20357 + }, + { + "epoch": 2.67, + "learning_rate": 1.555005138679819e-05, + "loss": 0.0524, + "step": 20358 + }, + { + "epoch": 2.67, + "learning_rate": 1.5549610229722445e-05, + "loss": 0.055, + "step": 20359 + }, + { + "epoch": 2.67, + "learning_rate": 1.5549169057038716e-05, + "loss": 0.0611, + "step": 20360 + }, + { + "epoch": 2.67, + "learning_rate": 1.554872786874825e-05, + "loss": 0.052, + "step": 20361 + }, + { + "epoch": 2.67, + "learning_rate": 1.5548286664852278e-05, + "loss": 0.0545, + "step": 20362 + }, + { + "epoch": 2.67, + "learning_rate": 1.554784544535204e-05, + "loss": 0.0585, + "step": 20363 + }, + { + "epoch": 2.67, + "learning_rate": 1.5547404210248787e-05, + "loss": 0.0579, + "step": 20364 + }, + { + "epoch": 2.67, + "learning_rate": 1.5546962959543756e-05, + "loss": 0.052, + "step": 20365 + }, + { + "epoch": 2.67, + "learning_rate": 1.554652169323819e-05, + "loss": 0.061, + "step": 20366 + }, + { + "epoch": 2.67, + "learning_rate": 1.5546080411333323e-05, + "loss": 0.0532, + "step": 20367 + }, + { + "epoch": 2.67, + "learning_rate": 1.5545639113830402e-05, + "loss": 0.0528, + "step": 20368 + }, + { + "epoch": 2.67, + "learning_rate": 1.5545197800730664e-05, + "loss": 0.0584, + "step": 20369 + }, + { + "epoch": 2.67, + "learning_rate": 1.554475647203535e-05, + "loss": 0.0668, + "step": 20370 + }, + { + "epoch": 2.67, + "learning_rate": 1.554431512774571e-05, + "loss": 0.0585, + "step": 20371 + }, + { + "epoch": 2.67, + "learning_rate": 1.5543873767862975e-05, + "loss": 0.0629, + "step": 20372 + }, + { + "epoch": 2.67, + "learning_rate": 1.5543432392388394e-05, + "loss": 0.0565, + "step": 20373 + }, + { + "epoch": 2.67, + "learning_rate": 1.55429910013232e-05, + "loss": 0.0476, + "step": 20374 + }, + { + "epoch": 2.67, + "learning_rate": 1.5542549594668642e-05, + "loss": 0.0662, + "step": 20375 + }, + { + "epoch": 2.67, + "learning_rate": 1.5542108172425957e-05, + "loss": 0.0486, + "step": 20376 + }, + { + "epoch": 2.67, + "learning_rate": 1.554166673459639e-05, + "loss": 0.0617, + "step": 20377 + }, + { + "epoch": 2.67, + "learning_rate": 1.5541225281181177e-05, + "loss": 0.0595, + "step": 20378 + }, + { + "epoch": 2.67, + "learning_rate": 1.5540783812181562e-05, + "loss": 0.0595, + "step": 20379 + }, + { + "epoch": 2.67, + "learning_rate": 1.554034232759879e-05, + "loss": 0.0583, + "step": 20380 + }, + { + "epoch": 2.67, + "learning_rate": 1.5539900827434102e-05, + "loss": 0.0621, + "step": 20381 + }, + { + "epoch": 2.67, + "learning_rate": 1.5539459311688738e-05, + "loss": 0.0606, + "step": 20382 + }, + { + "epoch": 2.67, + "learning_rate": 1.5539017780363937e-05, + "loss": 0.0555, + "step": 20383 + }, + { + "epoch": 2.67, + "learning_rate": 1.5538576233460947e-05, + "loss": 0.0635, + "step": 20384 + }, + { + "epoch": 2.67, + "learning_rate": 1.5538134670981e-05, + "loss": 0.0525, + "step": 20385 + }, + { + "epoch": 2.67, + "learning_rate": 1.553769309292535e-05, + "loss": 0.0567, + "step": 20386 + }, + { + "epoch": 2.67, + "learning_rate": 1.5537251499295233e-05, + "loss": 0.056, + "step": 20387 + }, + { + "epoch": 2.67, + "learning_rate": 1.553680989009189e-05, + "loss": 0.0669, + "step": 20388 + }, + { + "epoch": 2.67, + "learning_rate": 1.5536368265316562e-05, + "loss": 0.0559, + "step": 20389 + }, + { + "epoch": 2.67, + "learning_rate": 1.5535926624970493e-05, + "loss": 0.0683, + "step": 20390 + }, + { + "epoch": 2.67, + "learning_rate": 1.5535484969054925e-05, + "loss": 0.0521, + "step": 20391 + }, + { + "epoch": 2.67, + "learning_rate": 1.5535043297571105e-05, + "loss": 0.0561, + "step": 20392 + }, + { + "epoch": 2.67, + "learning_rate": 1.553460161052027e-05, + "loss": 0.0551, + "step": 20393 + }, + { + "epoch": 2.67, + "learning_rate": 1.5534159907903658e-05, + "loss": 0.0536, + "step": 20394 + }, + { + "epoch": 2.67, + "learning_rate": 1.5533718189722518e-05, + "loss": 0.0538, + "step": 20395 + }, + { + "epoch": 2.67, + "learning_rate": 1.5533276455978087e-05, + "loss": 0.0538, + "step": 20396 + }, + { + "epoch": 2.67, + "learning_rate": 1.5532834706671613e-05, + "loss": 0.0669, + "step": 20397 + }, + { + "epoch": 2.67, + "learning_rate": 1.553239294180434e-05, + "loss": 0.072, + "step": 20398 + }, + { + "epoch": 2.67, + "learning_rate": 1.5531951161377502e-05, + "loss": 0.0561, + "step": 20399 + }, + { + "epoch": 2.67, + "learning_rate": 1.5531509365392348e-05, + "loss": 0.0562, + "step": 20400 + }, + { + "epoch": 2.67, + "learning_rate": 1.5531067553850115e-05, + "loss": 0.0547, + "step": 20401 + }, + { + "epoch": 2.67, + "learning_rate": 1.5530625726752055e-05, + "loss": 0.0586, + "step": 20402 + }, + { + "epoch": 2.67, + "learning_rate": 1.55301838840994e-05, + "loss": 0.0543, + "step": 20403 + }, + { + "epoch": 2.67, + "learning_rate": 1.55297420258934e-05, + "loss": 0.0535, + "step": 20404 + }, + { + "epoch": 2.67, + "learning_rate": 1.5529300152135293e-05, + "loss": 0.0561, + "step": 20405 + }, + { + "epoch": 2.67, + "learning_rate": 1.5528858262826326e-05, + "loss": 0.0624, + "step": 20406 + }, + { + "epoch": 2.67, + "learning_rate": 1.5528416357967737e-05, + "loss": 0.0659, + "step": 20407 + }, + { + "epoch": 2.67, + "learning_rate": 1.5527974437560775e-05, + "loss": 0.0582, + "step": 20408 + }, + { + "epoch": 2.67, + "learning_rate": 1.5527532501606675e-05, + "loss": 0.0521, + "step": 20409 + }, + { + "epoch": 2.67, + "learning_rate": 1.5527090550106688e-05, + "loss": 0.0645, + "step": 20410 + }, + { + "epoch": 2.67, + "learning_rate": 1.552664858306205e-05, + "loss": 0.0567, + "step": 20411 + }, + { + "epoch": 2.67, + "learning_rate": 1.552620660047401e-05, + "loss": 0.0636, + "step": 20412 + }, + { + "epoch": 2.67, + "learning_rate": 1.5525764602343802e-05, + "loss": 0.0556, + "step": 20413 + }, + { + "epoch": 2.67, + "learning_rate": 1.552532258867268e-05, + "loss": 0.0581, + "step": 20414 + }, + { + "epoch": 2.67, + "learning_rate": 1.5524880559461887e-05, + "loss": 0.0612, + "step": 20415 + }, + { + "epoch": 2.67, + "learning_rate": 1.5524438514712653e-05, + "loss": 0.0549, + "step": 20416 + }, + { + "epoch": 2.67, + "learning_rate": 1.5523996454426235e-05, + "loss": 0.0589, + "step": 20417 + }, + { + "epoch": 2.67, + "learning_rate": 1.5523554378603872e-05, + "loss": 0.0625, + "step": 20418 + }, + { + "epoch": 2.67, + "learning_rate": 1.5523112287246802e-05, + "loss": 0.0535, + "step": 20419 + }, + { + "epoch": 2.67, + "learning_rate": 1.5522670180356277e-05, + "loss": 0.0603, + "step": 20420 + }, + { + "epoch": 2.68, + "learning_rate": 1.5522228057933535e-05, + "loss": 0.0508, + "step": 20421 + }, + { + "epoch": 2.68, + "learning_rate": 1.5521785919979818e-05, + "loss": 0.0717, + "step": 20422 + }, + { + "epoch": 2.68, + "learning_rate": 1.5521343766496375e-05, + "loss": 0.0605, + "step": 20423 + }, + { + "epoch": 2.68, + "learning_rate": 1.5520901597484446e-05, + "loss": 0.0603, + "step": 20424 + }, + { + "epoch": 2.68, + "learning_rate": 1.5520459412945275e-05, + "loss": 0.0587, + "step": 20425 + }, + { + "epoch": 2.68, + "learning_rate": 1.552001721288011e-05, + "loss": 0.0488, + "step": 20426 + }, + { + "epoch": 2.68, + "learning_rate": 1.5519574997290187e-05, + "loss": 0.0561, + "step": 20427 + }, + { + "epoch": 2.68, + "learning_rate": 1.551913276617675e-05, + "loss": 0.0773, + "step": 20428 + }, + { + "epoch": 2.68, + "learning_rate": 1.551869051954105e-05, + "loss": 0.06, + "step": 20429 + }, + { + "epoch": 2.68, + "learning_rate": 1.5518248257384325e-05, + "loss": 0.0561, + "step": 20430 + }, + { + "epoch": 2.68, + "learning_rate": 1.5517805979707824e-05, + "loss": 0.0626, + "step": 20431 + }, + { + "epoch": 2.68, + "learning_rate": 1.5517363686512784e-05, + "loss": 0.0636, + "step": 20432 + }, + { + "epoch": 2.68, + "learning_rate": 1.5516921377800453e-05, + "loss": 0.0581, + "step": 20433 + }, + { + "epoch": 2.68, + "learning_rate": 1.5516479053572075e-05, + "loss": 0.0541, + "step": 20434 + }, + { + "epoch": 2.68, + "learning_rate": 1.5516036713828892e-05, + "loss": 0.0552, + "step": 20435 + }, + { + "epoch": 2.68, + "learning_rate": 1.551559435857215e-05, + "loss": 0.0485, + "step": 20436 + }, + { + "epoch": 2.68, + "learning_rate": 1.5515151987803094e-05, + "loss": 0.0591, + "step": 20437 + }, + { + "epoch": 2.68, + "learning_rate": 1.5514709601522965e-05, + "loss": 0.0491, + "step": 20438 + }, + { + "epoch": 2.68, + "learning_rate": 1.5514267199733013e-05, + "loss": 0.0482, + "step": 20439 + }, + { + "epoch": 2.68, + "learning_rate": 1.5513824782434474e-05, + "loss": 0.0542, + "step": 20440 + }, + { + "epoch": 2.68, + "learning_rate": 1.5513382349628597e-05, + "loss": 0.0678, + "step": 20441 + }, + { + "epoch": 2.68, + "learning_rate": 1.5512939901316628e-05, + "loss": 0.05, + "step": 20442 + }, + { + "epoch": 2.68, + "learning_rate": 1.5512497437499808e-05, + "loss": 0.0605, + "step": 20443 + }, + { + "epoch": 2.68, + "learning_rate": 1.5512054958179378e-05, + "loss": 0.0598, + "step": 20444 + }, + { + "epoch": 2.68, + "learning_rate": 1.5511612463356592e-05, + "loss": 0.0525, + "step": 20445 + }, + { + "epoch": 2.68, + "learning_rate": 1.5511169953032687e-05, + "loss": 0.0577, + "step": 20446 + }, + { + "epoch": 2.68, + "learning_rate": 1.5510727427208915e-05, + "loss": 0.0536, + "step": 20447 + }, + { + "epoch": 2.68, + "learning_rate": 1.551028488588651e-05, + "loss": 0.0632, + "step": 20448 + }, + { + "epoch": 2.68, + "learning_rate": 1.5509842329066723e-05, + "loss": 0.0523, + "step": 20449 + }, + { + "epoch": 2.68, + "learning_rate": 1.55093997567508e-05, + "loss": 0.0527, + "step": 20450 + }, + { + "epoch": 2.68, + "learning_rate": 1.5508957168939982e-05, + "loss": 0.0506, + "step": 20451 + }, + { + "epoch": 2.68, + "learning_rate": 1.5508514565635518e-05, + "loss": 0.0568, + "step": 20452 + }, + { + "epoch": 2.68, + "learning_rate": 1.550807194683865e-05, + "loss": 0.0572, + "step": 20453 + }, + { + "epoch": 2.68, + "learning_rate": 1.5507629312550617e-05, + "loss": 0.0688, + "step": 20454 + }, + { + "epoch": 2.68, + "learning_rate": 1.5507186662772674e-05, + "loss": 0.0588, + "step": 20455 + }, + { + "epoch": 2.68, + "learning_rate": 1.5506743997506063e-05, + "loss": 0.0647, + "step": 20456 + }, + { + "epoch": 2.68, + "learning_rate": 1.5506301316752027e-05, + "loss": 0.0553, + "step": 20457 + }, + { + "epoch": 2.68, + "learning_rate": 1.5505858620511812e-05, + "loss": 0.0478, + "step": 20458 + }, + { + "epoch": 2.68, + "learning_rate": 1.550541590878666e-05, + "loss": 0.0536, + "step": 20459 + }, + { + "epoch": 2.68, + "learning_rate": 1.550497318157782e-05, + "loss": 0.0613, + "step": 20460 + }, + { + "epoch": 2.68, + "learning_rate": 1.550453043888654e-05, + "loss": 0.051, + "step": 20461 + }, + { + "epoch": 2.68, + "learning_rate": 1.5504087680714058e-05, + "loss": 0.0532, + "step": 20462 + }, + { + "epoch": 2.68, + "learning_rate": 1.5503644907061622e-05, + "loss": 0.0592, + "step": 20463 + }, + { + "epoch": 2.68, + "learning_rate": 1.550320211793048e-05, + "loss": 0.0564, + "step": 20464 + }, + { + "epoch": 2.68, + "learning_rate": 1.5502759313321873e-05, + "loss": 0.0475, + "step": 20465 + }, + { + "epoch": 2.68, + "learning_rate": 1.550231649323705e-05, + "loss": 0.0621, + "step": 20466 + }, + { + "epoch": 2.68, + "learning_rate": 1.5501873657677252e-05, + "loss": 0.0584, + "step": 20467 + }, + { + "epoch": 2.68, + "learning_rate": 1.550143080664373e-05, + "loss": 0.0642, + "step": 20468 + }, + { + "epoch": 2.68, + "learning_rate": 1.5500987940137725e-05, + "loss": 0.0552, + "step": 20469 + }, + { + "epoch": 2.68, + "learning_rate": 1.550054505816049e-05, + "loss": 0.0635, + "step": 20470 + }, + { + "epoch": 2.68, + "learning_rate": 1.5500102160713256e-05, + "loss": 0.0595, + "step": 20471 + }, + { + "epoch": 2.68, + "learning_rate": 1.5499659247797284e-05, + "loss": 0.0566, + "step": 20472 + }, + { + "epoch": 2.68, + "learning_rate": 1.549921631941381e-05, + "loss": 0.0603, + "step": 20473 + }, + { + "epoch": 2.68, + "learning_rate": 1.5498773375564086e-05, + "loss": 0.069, + "step": 20474 + }, + { + "epoch": 2.68, + "learning_rate": 1.5498330416249354e-05, + "loss": 0.0611, + "step": 20475 + }, + { + "epoch": 2.68, + "learning_rate": 1.549788744147086e-05, + "loss": 0.055, + "step": 20476 + }, + { + "epoch": 2.68, + "learning_rate": 1.549744445122985e-05, + "loss": 0.0526, + "step": 20477 + }, + { + "epoch": 2.68, + "learning_rate": 1.5497001445527567e-05, + "loss": 0.0599, + "step": 20478 + }, + { + "epoch": 2.68, + "learning_rate": 1.5496558424365263e-05, + "loss": 0.0547, + "step": 20479 + }, + { + "epoch": 2.68, + "learning_rate": 1.5496115387744182e-05, + "loss": 0.0515, + "step": 20480 + }, + { + "epoch": 2.68, + "learning_rate": 1.549567233566557e-05, + "loss": 0.0627, + "step": 20481 + }, + { + "epoch": 2.68, + "learning_rate": 1.5495229268130668e-05, + "loss": 0.0525, + "step": 20482 + }, + { + "epoch": 2.68, + "learning_rate": 1.5494786185140732e-05, + "loss": 0.0555, + "step": 20483 + }, + { + "epoch": 2.68, + "learning_rate": 1.5494343086697e-05, + "loss": 0.0651, + "step": 20484 + }, + { + "epoch": 2.68, + "learning_rate": 1.549389997280072e-05, + "loss": 0.0596, + "step": 20485 + }, + { + "epoch": 2.68, + "learning_rate": 1.5493456843453138e-05, + "loss": 0.0596, + "step": 20486 + }, + { + "epoch": 2.68, + "learning_rate": 1.54930136986555e-05, + "loss": 0.0609, + "step": 20487 + }, + { + "epoch": 2.68, + "learning_rate": 1.549257053840906e-05, + "loss": 0.0592, + "step": 20488 + }, + { + "epoch": 2.68, + "learning_rate": 1.549212736271505e-05, + "loss": 0.0534, + "step": 20489 + }, + { + "epoch": 2.68, + "learning_rate": 1.549168417157473e-05, + "loss": 0.0503, + "step": 20490 + }, + { + "epoch": 2.68, + "learning_rate": 1.5491240964989337e-05, + "loss": 0.0556, + "step": 20491 + }, + { + "epoch": 2.68, + "learning_rate": 1.5490797742960126e-05, + "loss": 0.0685, + "step": 20492 + }, + { + "epoch": 2.68, + "learning_rate": 1.5490354505488335e-05, + "loss": 0.0512, + "step": 20493 + }, + { + "epoch": 2.68, + "learning_rate": 1.5489911252575215e-05, + "loss": 0.0584, + "step": 20494 + }, + { + "epoch": 2.68, + "learning_rate": 1.548946798422201e-05, + "loss": 0.054, + "step": 20495 + }, + { + "epoch": 2.68, + "learning_rate": 1.548902470042997e-05, + "loss": 0.0581, + "step": 20496 + }, + { + "epoch": 2.68, + "learning_rate": 1.5488581401200344e-05, + "loss": 0.0631, + "step": 20497 + }, + { + "epoch": 2.69, + "learning_rate": 1.5488138086534372e-05, + "loss": 0.058, + "step": 20498 + }, + { + "epoch": 2.69, + "learning_rate": 1.54876947564333e-05, + "loss": 0.0532, + "step": 20499 + }, + { + "epoch": 2.69, + "learning_rate": 1.5487251410898382e-05, + "loss": 0.0562, + "step": 20500 + }, + { + "epoch": 2.69, + "learning_rate": 1.5486808049930865e-05, + "loss": 0.0539, + "step": 20501 + }, + { + "epoch": 2.69, + "learning_rate": 1.548636467353199e-05, + "loss": 0.0604, + "step": 20502 + }, + { + "epoch": 2.69, + "learning_rate": 1.5485921281703003e-05, + "loss": 0.0627, + "step": 20503 + }, + { + "epoch": 2.69, + "learning_rate": 1.548547787444516e-05, + "loss": 0.0577, + "step": 20504 + }, + { + "epoch": 2.69, + "learning_rate": 1.54850344517597e-05, + "loss": 0.0585, + "step": 20505 + }, + { + "epoch": 2.69, + "learning_rate": 1.5484591013647875e-05, + "loss": 0.064, + "step": 20506 + }, + { + "epoch": 2.69, + "learning_rate": 1.5484147560110926e-05, + "loss": 0.0591, + "step": 20507 + }, + { + "epoch": 2.69, + "learning_rate": 1.5483704091150106e-05, + "loss": 0.0515, + "step": 20508 + }, + { + "epoch": 2.69, + "learning_rate": 1.548326060676666e-05, + "loss": 0.0529, + "step": 20509 + }, + { + "epoch": 2.69, + "learning_rate": 1.5482817106961837e-05, + "loss": 0.0509, + "step": 20510 + }, + { + "epoch": 2.69, + "learning_rate": 1.548237359173688e-05, + "loss": 0.0533, + "step": 20511 + }, + { + "epoch": 2.69, + "learning_rate": 1.548193006109304e-05, + "loss": 0.0522, + "step": 20512 + }, + { + "epoch": 2.69, + "learning_rate": 1.5481486515031566e-05, + "loss": 0.0643, + "step": 20513 + }, + { + "epoch": 2.69, + "learning_rate": 1.5481042953553705e-05, + "loss": 0.0578, + "step": 20514 + }, + { + "epoch": 2.69, + "learning_rate": 1.5480599376660697e-05, + "loss": 0.0582, + "step": 20515 + }, + { + "epoch": 2.69, + "learning_rate": 1.5480155784353795e-05, + "loss": 0.0574, + "step": 20516 + }, + { + "epoch": 2.69, + "learning_rate": 1.547971217663425e-05, + "loss": 0.0698, + "step": 20517 + }, + { + "epoch": 2.69, + "learning_rate": 1.5479268553503307e-05, + "loss": 0.0626, + "step": 20518 + }, + { + "epoch": 2.69, + "learning_rate": 1.5478824914962212e-05, + "loss": 0.0556, + "step": 20519 + }, + { + "epoch": 2.69, + "learning_rate": 1.5478381261012212e-05, + "loss": 0.0572, + "step": 20520 + }, + { + "epoch": 2.69, + "learning_rate": 1.547793759165456e-05, + "loss": 0.0544, + "step": 20521 + }, + { + "epoch": 2.69, + "learning_rate": 1.5477493906890495e-05, + "loss": 0.057, + "step": 20522 + }, + { + "epoch": 2.69, + "learning_rate": 1.5477050206721275e-05, + "loss": 0.0609, + "step": 20523 + }, + { + "epoch": 2.69, + "learning_rate": 1.5476606491148138e-05, + "loss": 0.0529, + "step": 20524 + }, + { + "epoch": 2.69, + "learning_rate": 1.547616276017234e-05, + "loss": 0.062, + "step": 20525 + }, + { + "epoch": 2.69, + "learning_rate": 1.5475719013795124e-05, + "loss": 0.0618, + "step": 20526 + }, + { + "epoch": 2.69, + "learning_rate": 1.5475275252017742e-05, + "loss": 0.0642, + "step": 20527 + }, + { + "epoch": 2.69, + "learning_rate": 1.547483147484144e-05, + "loss": 0.0483, + "step": 20528 + }, + { + "epoch": 2.69, + "learning_rate": 1.5474387682267465e-05, + "loss": 0.0469, + "step": 20529 + }, + { + "epoch": 2.69, + "learning_rate": 1.547394387429707e-05, + "loss": 0.0605, + "step": 20530 + }, + { + "epoch": 2.69, + "learning_rate": 1.5473500050931495e-05, + "loss": 0.0588, + "step": 20531 + }, + { + "epoch": 2.69, + "learning_rate": 1.5473056212171994e-05, + "loss": 0.0559, + "step": 20532 + }, + { + "epoch": 2.69, + "learning_rate": 1.5472612358019808e-05, + "loss": 0.0582, + "step": 20533 + }, + { + "epoch": 2.69, + "learning_rate": 1.54721684884762e-05, + "loss": 0.0643, + "step": 20534 + }, + { + "epoch": 2.69, + "learning_rate": 1.5471724603542405e-05, + "loss": 0.0562, + "step": 20535 + }, + { + "epoch": 2.69, + "learning_rate": 1.547128070321968e-05, + "loss": 0.0627, + "step": 20536 + }, + { + "epoch": 2.69, + "learning_rate": 1.5470836787509265e-05, + "loss": 0.0487, + "step": 20537 + }, + { + "epoch": 2.69, + "learning_rate": 1.5470392856412412e-05, + "loss": 0.0538, + "step": 20538 + }, + { + "epoch": 2.69, + "learning_rate": 1.5469948909930372e-05, + "loss": 0.0648, + "step": 20539 + }, + { + "epoch": 2.69, + "learning_rate": 1.5469504948064394e-05, + "loss": 0.0488, + "step": 20540 + }, + { + "epoch": 2.69, + "learning_rate": 1.5469060970815724e-05, + "loss": 0.0487, + "step": 20541 + }, + { + "epoch": 2.69, + "learning_rate": 1.546861697818561e-05, + "loss": 0.0524, + "step": 20542 + }, + { + "epoch": 2.69, + "learning_rate": 1.54681729701753e-05, + "loss": 0.0457, + "step": 20543 + }, + { + "epoch": 2.69, + "learning_rate": 1.546772894678605e-05, + "loss": 0.0629, + "step": 20544 + }, + { + "epoch": 2.69, + "learning_rate": 1.54672849080191e-05, + "loss": 0.0539, + "step": 20545 + }, + { + "epoch": 2.69, + "learning_rate": 1.54668408538757e-05, + "loss": 0.0563, + "step": 20546 + }, + { + "epoch": 2.69, + "learning_rate": 1.5466396784357107e-05, + "loss": 0.0662, + "step": 20547 + }, + { + "epoch": 2.69, + "learning_rate": 1.546595269946456e-05, + "loss": 0.051, + "step": 20548 + }, + { + "epoch": 2.69, + "learning_rate": 1.546550859919931e-05, + "loss": 0.0538, + "step": 20549 + }, + { + "epoch": 2.69, + "learning_rate": 1.5465064483562616e-05, + "loss": 0.0528, + "step": 20550 + }, + { + "epoch": 2.69, + "learning_rate": 1.5464620352555715e-05, + "loss": 0.0605, + "step": 20551 + }, + { + "epoch": 2.69, + "learning_rate": 1.546417620617986e-05, + "loss": 0.0553, + "step": 20552 + }, + { + "epoch": 2.69, + "learning_rate": 1.54637320444363e-05, + "loss": 0.0573, + "step": 20553 + }, + { + "epoch": 2.69, + "learning_rate": 1.5463287867326284e-05, + "loss": 0.0552, + "step": 20554 + }, + { + "epoch": 2.69, + "learning_rate": 1.5462843674851066e-05, + "loss": 0.0505, + "step": 20555 + }, + { + "epoch": 2.69, + "learning_rate": 1.5462399467011887e-05, + "loss": 0.0584, + "step": 20556 + }, + { + "epoch": 2.69, + "learning_rate": 1.546195524381e-05, + "loss": 0.0596, + "step": 20557 + }, + { + "epoch": 2.69, + "learning_rate": 1.5461511005246655e-05, + "loss": 0.0599, + "step": 20558 + }, + { + "epoch": 2.69, + "learning_rate": 1.5461066751323104e-05, + "loss": 0.059, + "step": 20559 + }, + { + "epoch": 2.69, + "learning_rate": 1.546062248204059e-05, + "loss": 0.059, + "step": 20560 + }, + { + "epoch": 2.69, + "learning_rate": 1.546017819740037e-05, + "loss": 0.0583, + "step": 20561 + }, + { + "epoch": 2.69, + "learning_rate": 1.5459733897403687e-05, + "loss": 0.0632, + "step": 20562 + }, + { + "epoch": 2.69, + "learning_rate": 1.5459289582051794e-05, + "loss": 0.0598, + "step": 20563 + }, + { + "epoch": 2.69, + "learning_rate": 1.5458845251345936e-05, + "loss": 0.0655, + "step": 20564 + }, + { + "epoch": 2.69, + "learning_rate": 1.545840090528737e-05, + "loss": 0.0515, + "step": 20565 + }, + { + "epoch": 2.69, + "learning_rate": 1.5457956543877345e-05, + "loss": 0.0683, + "step": 20566 + }, + { + "epoch": 2.69, + "learning_rate": 1.5457512167117102e-05, + "loss": 0.0639, + "step": 20567 + }, + { + "epoch": 2.69, + "learning_rate": 1.54570677750079e-05, + "loss": 0.0601, + "step": 20568 + }, + { + "epoch": 2.69, + "learning_rate": 1.5456623367550987e-05, + "loss": 0.0532, + "step": 20569 + }, + { + "epoch": 2.69, + "learning_rate": 1.545617894474761e-05, + "loss": 0.0585, + "step": 20570 + }, + { + "epoch": 2.69, + "learning_rate": 1.5455734506599016e-05, + "loss": 0.0611, + "step": 20571 + }, + { + "epoch": 2.69, + "learning_rate": 1.5455290053106466e-05, + "loss": 0.0598, + "step": 20572 + }, + { + "epoch": 2.69, + "learning_rate": 1.5454845584271197e-05, + "loss": 0.0645, + "step": 20573 + }, + { + "epoch": 2.7, + "learning_rate": 1.545440110009447e-05, + "loss": 0.0523, + "step": 20574 + }, + { + "epoch": 2.7, + "learning_rate": 1.5453956600577528e-05, + "loss": 0.0693, + "step": 20575 + }, + { + "epoch": 2.7, + "learning_rate": 1.5453512085721625e-05, + "loss": 0.061, + "step": 20576 + }, + { + "epoch": 2.7, + "learning_rate": 1.545306755552801e-05, + "loss": 0.0668, + "step": 20577 + }, + { + "epoch": 2.7, + "learning_rate": 1.5452623009997928e-05, + "loss": 0.0467, + "step": 20578 + }, + { + "epoch": 2.7, + "learning_rate": 1.545217844913264e-05, + "loss": 0.0633, + "step": 20579 + }, + { + "epoch": 2.7, + "learning_rate": 1.545173387293339e-05, + "loss": 0.0672, + "step": 20580 + }, + { + "epoch": 2.7, + "learning_rate": 1.5451289281401427e-05, + "loss": 0.0623, + "step": 20581 + }, + { + "epoch": 2.7, + "learning_rate": 1.5450844674538003e-05, + "loss": 0.0626, + "step": 20582 + }, + { + "epoch": 2.7, + "learning_rate": 1.5450400052344366e-05, + "loss": 0.0569, + "step": 20583 + }, + { + "epoch": 2.7, + "learning_rate": 1.5449955414821775e-05, + "loss": 0.0526, + "step": 20584 + }, + { + "epoch": 2.7, + "learning_rate": 1.5449510761971472e-05, + "loss": 0.0543, + "step": 20585 + }, + { + "epoch": 2.7, + "learning_rate": 1.544906609379471e-05, + "loss": 0.0502, + "step": 20586 + }, + { + "epoch": 2.7, + "learning_rate": 1.5448621410292738e-05, + "loss": 0.0623, + "step": 20587 + }, + { + "epoch": 2.7, + "learning_rate": 1.544817671146681e-05, + "loss": 0.0587, + "step": 20588 + }, + { + "epoch": 2.7, + "learning_rate": 1.5447731997318177e-05, + "loss": 0.0696, + "step": 20589 + }, + { + "epoch": 2.7, + "learning_rate": 1.5447287267848086e-05, + "loss": 0.0638, + "step": 20590 + }, + { + "epoch": 2.7, + "learning_rate": 1.5446842523057787e-05, + "loss": 0.0541, + "step": 20591 + }, + { + "epoch": 2.7, + "learning_rate": 1.544639776294854e-05, + "loss": 0.0701, + "step": 20592 + }, + { + "epoch": 2.7, + "learning_rate": 1.5445952987521584e-05, + "loss": 0.0599, + "step": 20593 + }, + { + "epoch": 2.7, + "learning_rate": 1.5445508196778176e-05, + "loss": 0.0585, + "step": 20594 + }, + { + "epoch": 2.7, + "learning_rate": 1.544506339071957e-05, + "loss": 0.0664, + "step": 20595 + }, + { + "epoch": 2.7, + "learning_rate": 1.5444618569347006e-05, + "loss": 0.0539, + "step": 20596 + }, + { + "epoch": 2.7, + "learning_rate": 1.544417373266175e-05, + "loss": 0.0597, + "step": 20597 + }, + { + "epoch": 2.7, + "learning_rate": 1.544372888066504e-05, + "loss": 0.0646, + "step": 20598 + }, + { + "epoch": 2.7, + "learning_rate": 1.5443284013358134e-05, + "loss": 0.0493, + "step": 20599 + }, + { + "epoch": 2.7, + "learning_rate": 1.544283913074228e-05, + "loss": 0.0512, + "step": 20600 + }, + { + "epoch": 2.7, + "learning_rate": 1.5442394232818735e-05, + "loss": 0.0599, + "step": 20601 + }, + { + "epoch": 2.7, + "learning_rate": 1.5441949319588742e-05, + "loss": 0.0597, + "step": 20602 + }, + { + "epoch": 2.7, + "learning_rate": 1.544150439105356e-05, + "loss": 0.0553, + "step": 20603 + }, + { + "epoch": 2.7, + "learning_rate": 1.5441059447214434e-05, + "loss": 0.0515, + "step": 20604 + }, + { + "epoch": 2.7, + "learning_rate": 1.544061448807262e-05, + "loss": 0.0531, + "step": 20605 + }, + { + "epoch": 2.7, + "learning_rate": 1.5440169513629365e-05, + "loss": 0.0441, + "step": 20606 + }, + { + "epoch": 2.7, + "learning_rate": 1.543972452388592e-05, + "loss": 0.0527, + "step": 20607 + }, + { + "epoch": 2.7, + "learning_rate": 1.5439279518843546e-05, + "loss": 0.0607, + "step": 20608 + }, + { + "epoch": 2.7, + "learning_rate": 1.5438834498503486e-05, + "loss": 0.0486, + "step": 20609 + }, + { + "epoch": 2.7, + "learning_rate": 1.5438389462866993e-05, + "loss": 0.0594, + "step": 20610 + }, + { + "epoch": 2.7, + "learning_rate": 1.5437944411935318e-05, + "loss": 0.0509, + "step": 20611 + }, + { + "epoch": 2.7, + "learning_rate": 1.5437499345709718e-05, + "loss": 0.0648, + "step": 20612 + }, + { + "epoch": 2.7, + "learning_rate": 1.543705426419144e-05, + "loss": 0.0578, + "step": 20613 + }, + { + "epoch": 2.7, + "learning_rate": 1.5436609167381732e-05, + "loss": 0.0637, + "step": 20614 + }, + { + "epoch": 2.7, + "learning_rate": 1.543616405528185e-05, + "loss": 0.0567, + "step": 20615 + }, + { + "epoch": 2.7, + "learning_rate": 1.5435718927893048e-05, + "loss": 0.0469, + "step": 20616 + }, + { + "epoch": 2.7, + "learning_rate": 1.5435273785216578e-05, + "loss": 0.0534, + "step": 20617 + }, + { + "epoch": 2.7, + "learning_rate": 1.543482862725369e-05, + "loss": 0.0568, + "step": 20618 + }, + { + "epoch": 2.7, + "learning_rate": 1.5434383454005634e-05, + "loss": 0.0477, + "step": 20619 + }, + { + "epoch": 2.7, + "learning_rate": 1.5433938265473663e-05, + "loss": 0.0544, + "step": 20620 + }, + { + "epoch": 2.7, + "learning_rate": 1.5433493061659033e-05, + "loss": 0.0606, + "step": 20621 + }, + { + "epoch": 2.7, + "learning_rate": 1.5433047842562993e-05, + "loss": 0.0636, + "step": 20622 + }, + { + "epoch": 2.7, + "learning_rate": 1.5432602608186794e-05, + "loss": 0.0737, + "step": 20623 + }, + { + "epoch": 2.7, + "learning_rate": 1.5432157358531686e-05, + "loss": 0.0566, + "step": 20624 + }, + { + "epoch": 2.7, + "learning_rate": 1.543171209359893e-05, + "loss": 0.06, + "step": 20625 + }, + { + "epoch": 2.7, + "learning_rate": 1.543126681338977e-05, + "loss": 0.0584, + "step": 20626 + }, + { + "epoch": 2.7, + "learning_rate": 1.5430821517905464e-05, + "loss": 0.0544, + "step": 20627 + }, + { + "epoch": 2.7, + "learning_rate": 1.543037620714726e-05, + "loss": 0.0577, + "step": 20628 + }, + { + "epoch": 2.7, + "learning_rate": 1.5429930881116413e-05, + "loss": 0.0569, + "step": 20629 + }, + { + "epoch": 2.7, + "learning_rate": 1.5429485539814173e-05, + "loss": 0.0569, + "step": 20630 + }, + { + "epoch": 2.7, + "learning_rate": 1.5429040183241792e-05, + "loss": 0.0582, + "step": 20631 + }, + { + "epoch": 2.7, + "learning_rate": 1.5428594811400528e-05, + "loss": 0.0634, + "step": 20632 + }, + { + "epoch": 2.7, + "learning_rate": 1.542814942429163e-05, + "loss": 0.0575, + "step": 20633 + }, + { + "epoch": 2.7, + "learning_rate": 1.5427704021916353e-05, + "loss": 0.0487, + "step": 20634 + }, + { + "epoch": 2.7, + "learning_rate": 1.5427258604275943e-05, + "loss": 0.0544, + "step": 20635 + }, + { + "epoch": 2.7, + "learning_rate": 1.5426813171371657e-05, + "loss": 0.0475, + "step": 20636 + }, + { + "epoch": 2.7, + "learning_rate": 1.5426367723204752e-05, + "loss": 0.0743, + "step": 20637 + }, + { + "epoch": 2.7, + "learning_rate": 1.5425922259776475e-05, + "loss": 0.0642, + "step": 20638 + }, + { + "epoch": 2.7, + "learning_rate": 1.542547678108808e-05, + "loss": 0.0524, + "step": 20639 + }, + { + "epoch": 2.7, + "learning_rate": 1.5425031287140822e-05, + "loss": 0.0592, + "step": 20640 + }, + { + "epoch": 2.7, + "learning_rate": 1.542458577793595e-05, + "loss": 0.0554, + "step": 20641 + }, + { + "epoch": 2.7, + "learning_rate": 1.542414025347472e-05, + "loss": 0.057, + "step": 20642 + }, + { + "epoch": 2.7, + "learning_rate": 1.5423694713758384e-05, + "loss": 0.0609, + "step": 20643 + }, + { + "epoch": 2.7, + "learning_rate": 1.5423249158788197e-05, + "loss": 0.0684, + "step": 20644 + }, + { + "epoch": 2.7, + "learning_rate": 1.542280358856541e-05, + "loss": 0.0619, + "step": 20645 + }, + { + "epoch": 2.7, + "learning_rate": 1.542235800309128e-05, + "loss": 0.0525, + "step": 20646 + }, + { + "epoch": 2.7, + "learning_rate": 1.542191240236705e-05, + "loss": 0.0501, + "step": 20647 + }, + { + "epoch": 2.7, + "learning_rate": 1.5421466786393982e-05, + "loss": 0.0537, + "step": 20648 + }, + { + "epoch": 2.7, + "learning_rate": 1.542102115517333e-05, + "loss": 0.0592, + "step": 20649 + }, + { + "epoch": 2.71, + "learning_rate": 1.5420575508706342e-05, + "loss": 0.0564, + "step": 20650 + }, + { + "epoch": 2.71, + "learning_rate": 1.542012984699428e-05, + "loss": 0.0653, + "step": 20651 + }, + { + "epoch": 2.71, + "learning_rate": 1.5419684170038385e-05, + "loss": 0.069, + "step": 20652 + }, + { + "epoch": 2.71, + "learning_rate": 1.541923847783992e-05, + "loss": 0.0511, + "step": 20653 + }, + { + "epoch": 2.71, + "learning_rate": 1.5418792770400132e-05, + "loss": 0.046, + "step": 20654 + }, + { + "epoch": 2.71, + "learning_rate": 1.541834704772028e-05, + "loss": 0.0537, + "step": 20655 + }, + { + "epoch": 2.71, + "learning_rate": 1.5417901309801616e-05, + "loss": 0.0527, + "step": 20656 + }, + { + "epoch": 2.71, + "learning_rate": 1.5417455556645394e-05, + "loss": 0.0469, + "step": 20657 + }, + { + "epoch": 2.71, + "learning_rate": 1.5417009788252865e-05, + "loss": 0.0601, + "step": 20658 + }, + { + "epoch": 2.71, + "learning_rate": 1.5416564004625287e-05, + "loss": 0.0618, + "step": 20659 + }, + { + "epoch": 2.71, + "learning_rate": 1.541611820576391e-05, + "loss": 0.0623, + "step": 20660 + }, + { + "epoch": 2.71, + "learning_rate": 1.5415672391669985e-05, + "loss": 0.0593, + "step": 20661 + }, + { + "epoch": 2.71, + "learning_rate": 1.5415226562344776e-05, + "loss": 0.0511, + "step": 20662 + }, + { + "epoch": 2.71, + "learning_rate": 1.541478071778953e-05, + "loss": 0.0559, + "step": 20663 + }, + { + "epoch": 2.71, + "learning_rate": 1.5414334858005496e-05, + "loss": 0.0527, + "step": 20664 + }, + { + "epoch": 2.71, + "learning_rate": 1.5413888982993938e-05, + "loss": 0.0595, + "step": 20665 + }, + { + "epoch": 2.71, + "learning_rate": 1.5413443092756103e-05, + "loss": 0.0538, + "step": 20666 + }, + { + "epoch": 2.71, + "learning_rate": 1.541299718729325e-05, + "loss": 0.0555, + "step": 20667 + }, + { + "epoch": 2.71, + "learning_rate": 1.541255126660663e-05, + "loss": 0.0616, + "step": 20668 + }, + { + "epoch": 2.71, + "learning_rate": 1.5412105330697497e-05, + "loss": 0.0565, + "step": 20669 + }, + { + "epoch": 2.71, + "learning_rate": 1.5411659379567107e-05, + "loss": 0.0601, + "step": 20670 + }, + { + "epoch": 2.71, + "learning_rate": 1.541121341321671e-05, + "loss": 0.0578, + "step": 20671 + }, + { + "epoch": 2.71, + "learning_rate": 1.5410767431647568e-05, + "loss": 0.0603, + "step": 20672 + }, + { + "epoch": 2.71, + "learning_rate": 1.541032143486093e-05, + "loss": 0.0625, + "step": 20673 + }, + { + "epoch": 2.71, + "learning_rate": 1.5409875422858046e-05, + "loss": 0.0549, + "step": 20674 + }, + { + "epoch": 2.71, + "learning_rate": 1.540942939564018e-05, + "loss": 0.0656, + "step": 20675 + }, + { + "epoch": 2.71, + "learning_rate": 1.5408983353208583e-05, + "loss": 0.054, + "step": 20676 + }, + { + "epoch": 2.71, + "learning_rate": 1.5408537295564504e-05, + "loss": 0.065, + "step": 20677 + }, + { + "epoch": 2.71, + "learning_rate": 1.5408091222709205e-05, + "loss": 0.058, + "step": 20678 + }, + { + "epoch": 2.71, + "learning_rate": 1.5407645134643938e-05, + "loss": 0.0566, + "step": 20679 + }, + { + "epoch": 2.71, + "learning_rate": 1.540719903136995e-05, + "loss": 0.0622, + "step": 20680 + }, + { + "epoch": 2.71, + "learning_rate": 1.540675291288851e-05, + "loss": 0.0612, + "step": 20681 + }, + { + "epoch": 2.71, + "learning_rate": 1.5406306779200867e-05, + "loss": 0.0591, + "step": 20682 + }, + { + "epoch": 2.71, + "learning_rate": 1.540586063030827e-05, + "loss": 0.0582, + "step": 20683 + }, + { + "epoch": 2.71, + "learning_rate": 1.540541446621198e-05, + "loss": 0.0512, + "step": 20684 + }, + { + "epoch": 2.71, + "learning_rate": 1.5404968286913246e-05, + "loss": 0.0476, + "step": 20685 + }, + { + "epoch": 2.71, + "learning_rate": 1.540452209241333e-05, + "loss": 0.0632, + "step": 20686 + }, + { + "epoch": 2.71, + "learning_rate": 1.5404075882713484e-05, + "loss": 0.0549, + "step": 20687 + }, + { + "epoch": 2.71, + "learning_rate": 1.5403629657814964e-05, + "loss": 0.0623, + "step": 20688 + }, + { + "epoch": 2.71, + "learning_rate": 1.5403183417719016e-05, + "loss": 0.0505, + "step": 20689 + }, + { + "epoch": 2.71, + "learning_rate": 1.5402737162426908e-05, + "loss": 0.0566, + "step": 20690 + }, + { + "epoch": 2.71, + "learning_rate": 1.540229089193989e-05, + "loss": 0.0561, + "step": 20691 + }, + { + "epoch": 2.71, + "learning_rate": 1.540184460625921e-05, + "loss": 0.0604, + "step": 20692 + }, + { + "epoch": 2.71, + "learning_rate": 1.5401398305386137e-05, + "loss": 0.057, + "step": 20693 + }, + { + "epoch": 2.71, + "learning_rate": 1.5400951989321916e-05, + "loss": 0.0517, + "step": 20694 + }, + { + "epoch": 2.71, + "learning_rate": 1.5400505658067804e-05, + "loss": 0.0567, + "step": 20695 + }, + { + "epoch": 2.71, + "learning_rate": 1.5400059311625057e-05, + "loss": 0.0558, + "step": 20696 + }, + { + "epoch": 2.71, + "learning_rate": 1.5399612949994935e-05, + "loss": 0.0563, + "step": 20697 + }, + { + "epoch": 2.71, + "learning_rate": 1.5399166573178685e-05, + "loss": 0.0575, + "step": 20698 + }, + { + "epoch": 2.71, + "learning_rate": 1.5398720181177573e-05, + "loss": 0.0579, + "step": 20699 + }, + { + "epoch": 2.71, + "learning_rate": 1.539827377399284e-05, + "loss": 0.0585, + "step": 20700 + }, + { + "epoch": 2.71, + "learning_rate": 1.5397827351625752e-05, + "loss": 0.0603, + "step": 20701 + }, + { + "epoch": 2.71, + "learning_rate": 1.5397380914077565e-05, + "loss": 0.064, + "step": 20702 + }, + { + "epoch": 2.71, + "learning_rate": 1.539693446134953e-05, + "loss": 0.0425, + "step": 20703 + }, + { + "epoch": 2.71, + "learning_rate": 1.5396487993442902e-05, + "loss": 0.0559, + "step": 20704 + }, + { + "epoch": 2.71, + "learning_rate": 1.539604151035894e-05, + "loss": 0.054, + "step": 20705 + }, + { + "epoch": 2.71, + "learning_rate": 1.53955950120989e-05, + "loss": 0.069, + "step": 20706 + }, + { + "epoch": 2.71, + "learning_rate": 1.5395148498664033e-05, + "loss": 0.0585, + "step": 20707 + }, + { + "epoch": 2.71, + "learning_rate": 1.5394701970055602e-05, + "loss": 0.0578, + "step": 20708 + }, + { + "epoch": 2.71, + "learning_rate": 1.5394255426274856e-05, + "loss": 0.059, + "step": 20709 + }, + { + "epoch": 2.71, + "learning_rate": 1.5393808867323056e-05, + "loss": 0.0556, + "step": 20710 + }, + { + "epoch": 2.71, + "learning_rate": 1.5393362293201458e-05, + "loss": 0.0609, + "step": 20711 + }, + { + "epoch": 2.71, + "learning_rate": 1.5392915703911312e-05, + "loss": 0.0556, + "step": 20712 + }, + { + "epoch": 2.71, + "learning_rate": 1.5392469099453876e-05, + "loss": 0.0574, + "step": 20713 + }, + { + "epoch": 2.71, + "learning_rate": 1.5392022479830413e-05, + "loss": 0.054, + "step": 20714 + }, + { + "epoch": 2.71, + "learning_rate": 1.539157584504217e-05, + "loss": 0.0582, + "step": 20715 + }, + { + "epoch": 2.71, + "learning_rate": 1.539112919509041e-05, + "loss": 0.0594, + "step": 20716 + }, + { + "epoch": 2.71, + "learning_rate": 1.5390682529976385e-05, + "loss": 0.0658, + "step": 20717 + }, + { + "epoch": 2.71, + "learning_rate": 1.5390235849701348e-05, + "loss": 0.0498, + "step": 20718 + }, + { + "epoch": 2.71, + "learning_rate": 1.5389789154266566e-05, + "loss": 0.0636, + "step": 20719 + }, + { + "epoch": 2.71, + "learning_rate": 1.538934244367329e-05, + "loss": 0.0557, + "step": 20720 + }, + { + "epoch": 2.71, + "learning_rate": 1.538889571792277e-05, + "loss": 0.0635, + "step": 20721 + }, + { + "epoch": 2.71, + "learning_rate": 1.538844897701627e-05, + "loss": 0.0593, + "step": 20722 + }, + { + "epoch": 2.71, + "learning_rate": 1.5388002220955044e-05, + "loss": 0.0599, + "step": 20723 + }, + { + "epoch": 2.71, + "learning_rate": 1.538755544974035e-05, + "loss": 0.0569, + "step": 20724 + }, + { + "epoch": 2.71, + "learning_rate": 1.5387108663373442e-05, + "loss": 0.0548, + "step": 20725 + }, + { + "epoch": 2.71, + "learning_rate": 1.5386661861855576e-05, + "loss": 0.0634, + "step": 20726 + }, + { + "epoch": 2.72, + "learning_rate": 1.5386215045188015e-05, + "loss": 0.0555, + "step": 20727 + }, + { + "epoch": 2.72, + "learning_rate": 1.5385768213372004e-05, + "loss": 0.0522, + "step": 20728 + }, + { + "epoch": 2.72, + "learning_rate": 1.5385321366408815e-05, + "loss": 0.0492, + "step": 20729 + }, + { + "epoch": 2.72, + "learning_rate": 1.538487450429969e-05, + "loss": 0.0584, + "step": 20730 + }, + { + "epoch": 2.72, + "learning_rate": 1.5384427627045896e-05, + "loss": 0.0582, + "step": 20731 + }, + { + "epoch": 2.72, + "learning_rate": 1.5383980734648686e-05, + "loss": 0.0617, + "step": 20732 + }, + { + "epoch": 2.72, + "learning_rate": 1.5383533827109316e-05, + "loss": 0.0509, + "step": 20733 + }, + { + "epoch": 2.72, + "learning_rate": 1.5383086904429042e-05, + "loss": 0.0532, + "step": 20734 + }, + { + "epoch": 2.72, + "learning_rate": 1.5382639966609125e-05, + "loss": 0.0675, + "step": 20735 + }, + { + "epoch": 2.72, + "learning_rate": 1.5382193013650818e-05, + "loss": 0.0681, + "step": 20736 + }, + { + "epoch": 2.72, + "learning_rate": 1.538174604555538e-05, + "loss": 0.0592, + "step": 20737 + }, + { + "epoch": 2.72, + "learning_rate": 1.538129906232407e-05, + "loss": 0.0497, + "step": 20738 + }, + { + "epoch": 2.72, + "learning_rate": 1.5380852063958142e-05, + "loss": 0.0518, + "step": 20739 + }, + { + "epoch": 2.72, + "learning_rate": 1.538040505045885e-05, + "loss": 0.063, + "step": 20740 + }, + { + "epoch": 2.72, + "learning_rate": 1.537995802182746e-05, + "loss": 0.0558, + "step": 20741 + }, + { + "epoch": 2.72, + "learning_rate": 1.5379510978065225e-05, + "loss": 0.0597, + "step": 20742 + }, + { + "epoch": 2.72, + "learning_rate": 1.53790639191734e-05, + "loss": 0.0606, + "step": 20743 + }, + { + "epoch": 2.72, + "learning_rate": 1.5378616845153245e-05, + "loss": 0.0638, + "step": 20744 + }, + { + "epoch": 2.72, + "learning_rate": 1.5378169756006016e-05, + "loss": 0.047, + "step": 20745 + }, + { + "epoch": 2.72, + "learning_rate": 1.537772265173297e-05, + "loss": 0.06, + "step": 20746 + }, + { + "epoch": 2.72, + "learning_rate": 1.5377275532335366e-05, + "loss": 0.0482, + "step": 20747 + }, + { + "epoch": 2.72, + "learning_rate": 1.5376828397814463e-05, + "loss": 0.0531, + "step": 20748 + }, + { + "epoch": 2.72, + "learning_rate": 1.5376381248171514e-05, + "loss": 0.0613, + "step": 20749 + }, + { + "epoch": 2.72, + "learning_rate": 1.537593408340778e-05, + "loss": 0.0583, + "step": 20750 + }, + { + "epoch": 2.72, + "learning_rate": 1.5375486903524516e-05, + "loss": 0.0644, + "step": 20751 + }, + { + "epoch": 2.72, + "learning_rate": 1.5375039708522987e-05, + "loss": 0.0505, + "step": 20752 + }, + { + "epoch": 2.72, + "learning_rate": 1.5374592498404437e-05, + "loss": 0.0543, + "step": 20753 + }, + { + "epoch": 2.72, + "learning_rate": 1.5374145273170137e-05, + "loss": 0.0726, + "step": 20754 + }, + { + "epoch": 2.72, + "learning_rate": 1.5373698032821337e-05, + "loss": 0.0487, + "step": 20755 + }, + { + "epoch": 2.72, + "learning_rate": 1.53732507773593e-05, + "loss": 0.0563, + "step": 20756 + }, + { + "epoch": 2.72, + "learning_rate": 1.5372803506785282e-05, + "loss": 0.0572, + "step": 20757 + }, + { + "epoch": 2.72, + "learning_rate": 1.5372356221100538e-05, + "loss": 0.0592, + "step": 20758 + }, + { + "epoch": 2.72, + "learning_rate": 1.5371908920306328e-05, + "loss": 0.0683, + "step": 20759 + }, + { + "epoch": 2.72, + "learning_rate": 1.5371461604403914e-05, + "loss": 0.0699, + "step": 20760 + }, + { + "epoch": 2.72, + "learning_rate": 1.5371014273394545e-05, + "loss": 0.0615, + "step": 20761 + }, + { + "epoch": 2.72, + "learning_rate": 1.5370566927279485e-05, + "loss": 0.0608, + "step": 20762 + }, + { + "epoch": 2.72, + "learning_rate": 1.5370119566059995e-05, + "loss": 0.0498, + "step": 20763 + }, + { + "epoch": 2.72, + "learning_rate": 1.5369672189737328e-05, + "loss": 0.0586, + "step": 20764 + }, + { + "epoch": 2.72, + "learning_rate": 1.5369224798312745e-05, + "loss": 0.0537, + "step": 20765 + }, + { + "epoch": 2.72, + "learning_rate": 1.53687773917875e-05, + "loss": 0.0607, + "step": 20766 + }, + { + "epoch": 2.72, + "learning_rate": 1.5368329970162853e-05, + "loss": 0.0598, + "step": 20767 + }, + { + "epoch": 2.72, + "learning_rate": 1.536788253344007e-05, + "loss": 0.0595, + "step": 20768 + }, + { + "epoch": 2.72, + "learning_rate": 1.53674350816204e-05, + "loss": 0.0566, + "step": 20769 + }, + { + "epoch": 2.72, + "learning_rate": 1.5366987614705103e-05, + "loss": 0.0554, + "step": 20770 + }, + { + "epoch": 2.72, + "learning_rate": 1.5366540132695445e-05, + "loss": 0.0574, + "step": 20771 + }, + { + "epoch": 2.72, + "learning_rate": 1.5366092635592673e-05, + "loss": 0.0524, + "step": 20772 + }, + { + "epoch": 2.72, + "learning_rate": 1.536564512339805e-05, + "loss": 0.0517, + "step": 20773 + }, + { + "epoch": 2.72, + "learning_rate": 1.5365197596112842e-05, + "loss": 0.0532, + "step": 20774 + }, + { + "epoch": 2.72, + "learning_rate": 1.5364750053738297e-05, + "loss": 0.0506, + "step": 20775 + }, + { + "epoch": 2.72, + "learning_rate": 1.5364302496275682e-05, + "loss": 0.0593, + "step": 20776 + }, + { + "epoch": 2.72, + "learning_rate": 1.5363854923726252e-05, + "loss": 0.0653, + "step": 20777 + }, + { + "epoch": 2.72, + "learning_rate": 1.5363407336091257e-05, + "loss": 0.0654, + "step": 20778 + }, + { + "epoch": 2.72, + "learning_rate": 1.5362959733371974e-05, + "loss": 0.0611, + "step": 20779 + }, + { + "epoch": 2.72, + "learning_rate": 1.5362512115569646e-05, + "loss": 0.0508, + "step": 20780 + }, + { + "epoch": 2.72, + "learning_rate": 1.5362064482685545e-05, + "loss": 0.0591, + "step": 20781 + }, + { + "epoch": 2.72, + "learning_rate": 1.536161683472092e-05, + "loss": 0.0631, + "step": 20782 + }, + { + "epoch": 2.72, + "learning_rate": 1.5361169171677034e-05, + "loss": 0.0591, + "step": 20783 + }, + { + "epoch": 2.72, + "learning_rate": 1.536072149355514e-05, + "loss": 0.0533, + "step": 20784 + }, + { + "epoch": 2.72, + "learning_rate": 1.5360273800356508e-05, + "loss": 0.0637, + "step": 20785 + }, + { + "epoch": 2.72, + "learning_rate": 1.5359826092082388e-05, + "loss": 0.0566, + "step": 20786 + }, + { + "epoch": 2.72, + "learning_rate": 1.5359378368734042e-05, + "loss": 0.0573, + "step": 20787 + }, + { + "epoch": 2.72, + "learning_rate": 1.5358930630312736e-05, + "loss": 0.067, + "step": 20788 + }, + { + "epoch": 2.72, + "learning_rate": 1.5358482876819718e-05, + "loss": 0.0443, + "step": 20789 + }, + { + "epoch": 2.72, + "learning_rate": 1.5358035108256252e-05, + "loss": 0.0531, + "step": 20790 + }, + { + "epoch": 2.72, + "learning_rate": 1.5357587324623596e-05, + "loss": 0.0641, + "step": 20791 + }, + { + "epoch": 2.72, + "learning_rate": 1.5357139525923014e-05, + "loss": 0.0618, + "step": 20792 + }, + { + "epoch": 2.72, + "learning_rate": 1.5356691712155765e-05, + "loss": 0.0562, + "step": 20793 + }, + { + "epoch": 2.72, + "learning_rate": 1.5356243883323103e-05, + "loss": 0.0659, + "step": 20794 + }, + { + "epoch": 2.72, + "learning_rate": 1.535579603942629e-05, + "loss": 0.063, + "step": 20795 + }, + { + "epoch": 2.72, + "learning_rate": 1.5355348180466583e-05, + "loss": 0.0516, + "step": 20796 + }, + { + "epoch": 2.72, + "learning_rate": 1.535490030644525e-05, + "loss": 0.0533, + "step": 20797 + }, + { + "epoch": 2.72, + "learning_rate": 1.5354452417363544e-05, + "loss": 0.053, + "step": 20798 + }, + { + "epoch": 2.72, + "learning_rate": 1.535400451322272e-05, + "loss": 0.0609, + "step": 20799 + }, + { + "epoch": 2.72, + "learning_rate": 1.5353556594024048e-05, + "loss": 0.0548, + "step": 20800 + }, + { + "epoch": 2.72, + "learning_rate": 1.535310865976878e-05, + "loss": 0.0592, + "step": 20801 + }, + { + "epoch": 2.72, + "learning_rate": 1.5352660710458185e-05, + "loss": 0.0517, + "step": 20802 + }, + { + "epoch": 2.73, + "learning_rate": 1.5352212746093513e-05, + "loss": 0.0598, + "step": 20803 + }, + { + "epoch": 2.73, + "learning_rate": 1.535176476667603e-05, + "loss": 0.0716, + "step": 20804 + }, + { + "epoch": 2.73, + "learning_rate": 1.535131677220699e-05, + "loss": 0.0461, + "step": 20805 + }, + { + "epoch": 2.73, + "learning_rate": 1.5350868762687658e-05, + "loss": 0.0593, + "step": 20806 + }, + { + "epoch": 2.73, + "learning_rate": 1.5350420738119295e-05, + "loss": 0.0584, + "step": 20807 + }, + { + "epoch": 2.73, + "learning_rate": 1.5349972698503156e-05, + "loss": 0.0699, + "step": 20808 + }, + { + "epoch": 2.73, + "learning_rate": 1.5349524643840502e-05, + "loss": 0.0519, + "step": 20809 + }, + { + "epoch": 2.73, + "learning_rate": 1.53490765741326e-05, + "loss": 0.0604, + "step": 20810 + }, + { + "epoch": 2.73, + "learning_rate": 1.53486284893807e-05, + "loss": 0.0511, + "step": 20811 + }, + { + "epoch": 2.73, + "learning_rate": 1.5348180389586067e-05, + "loss": 0.0637, + "step": 20812 + }, + { + "epoch": 2.73, + "learning_rate": 1.5347732274749966e-05, + "loss": 0.056, + "step": 20813 + }, + { + "epoch": 2.73, + "learning_rate": 1.534728414487365e-05, + "loss": 0.054, + "step": 20814 + }, + { + "epoch": 2.73, + "learning_rate": 1.5346835999958382e-05, + "loss": 0.0667, + "step": 20815 + }, + { + "epoch": 2.73, + "learning_rate": 1.5346387840005424e-05, + "loss": 0.0562, + "step": 20816 + }, + { + "epoch": 2.73, + "learning_rate": 1.5345939665016034e-05, + "loss": 0.048, + "step": 20817 + }, + { + "epoch": 2.73, + "learning_rate": 1.534549147499147e-05, + "loss": 0.057, + "step": 20818 + }, + { + "epoch": 2.73, + "learning_rate": 1.5345043269933002e-05, + "loss": 0.0615, + "step": 20819 + }, + { + "epoch": 2.73, + "learning_rate": 1.534459504984188e-05, + "loss": 0.0461, + "step": 20820 + }, + { + "epoch": 2.73, + "learning_rate": 1.5344146814719372e-05, + "loss": 0.0571, + "step": 20821 + }, + { + "epoch": 2.73, + "learning_rate": 1.5343698564566732e-05, + "loss": 0.0628, + "step": 20822 + }, + { + "epoch": 2.73, + "learning_rate": 1.5343250299385228e-05, + "loss": 0.0532, + "step": 20823 + }, + { + "epoch": 2.73, + "learning_rate": 1.5342802019176115e-05, + "loss": 0.0531, + "step": 20824 + }, + { + "epoch": 2.73, + "learning_rate": 1.5342353723940658e-05, + "loss": 0.0666, + "step": 20825 + }, + { + "epoch": 2.73, + "learning_rate": 1.5341905413680114e-05, + "loss": 0.0552, + "step": 20826 + }, + { + "epoch": 2.73, + "learning_rate": 1.5341457088395746e-05, + "loss": 0.0555, + "step": 20827 + }, + { + "epoch": 2.73, + "learning_rate": 1.5341008748088812e-05, + "loss": 0.0601, + "step": 20828 + }, + { + "epoch": 2.73, + "learning_rate": 1.534056039276058e-05, + "loss": 0.0594, + "step": 20829 + }, + { + "epoch": 2.73, + "learning_rate": 1.53401120224123e-05, + "loss": 0.0583, + "step": 20830 + }, + { + "epoch": 2.73, + "learning_rate": 1.5339663637045247e-05, + "loss": 0.0517, + "step": 20831 + }, + { + "epoch": 2.73, + "learning_rate": 1.5339215236660668e-05, + "loss": 0.0627, + "step": 20832 + }, + { + "epoch": 2.73, + "learning_rate": 1.5338766821259835e-05, + "loss": 0.0538, + "step": 20833 + }, + { + "epoch": 2.73, + "learning_rate": 1.5338318390844e-05, + "loss": 0.057, + "step": 20834 + }, + { + "epoch": 2.73, + "learning_rate": 1.533786994541443e-05, + "loss": 0.0635, + "step": 20835 + }, + { + "epoch": 2.73, + "learning_rate": 1.533742148497239e-05, + "loss": 0.0595, + "step": 20836 + }, + { + "epoch": 2.73, + "learning_rate": 1.533697300951913e-05, + "loss": 0.0615, + "step": 20837 + }, + { + "epoch": 2.73, + "learning_rate": 1.5336524519055922e-05, + "loss": 0.0587, + "step": 20838 + }, + { + "epoch": 2.73, + "learning_rate": 1.533607601358402e-05, + "loss": 0.0616, + "step": 20839 + }, + { + "epoch": 2.73, + "learning_rate": 1.5335627493104692e-05, + "loss": 0.0651, + "step": 20840 + }, + { + "epoch": 2.73, + "learning_rate": 1.5335178957619192e-05, + "loss": 0.0553, + "step": 20841 + }, + { + "epoch": 2.73, + "learning_rate": 1.5334730407128788e-05, + "loss": 0.063, + "step": 20842 + }, + { + "epoch": 2.73, + "learning_rate": 1.5334281841634736e-05, + "loss": 0.0655, + "step": 20843 + }, + { + "epoch": 2.73, + "learning_rate": 1.5333833261138302e-05, + "loss": 0.0622, + "step": 20844 + }, + { + "epoch": 2.73, + "learning_rate": 1.5333384665640748e-05, + "loss": 0.0486, + "step": 20845 + }, + { + "epoch": 2.73, + "learning_rate": 1.533293605514333e-05, + "loss": 0.0602, + "step": 20846 + }, + { + "epoch": 2.73, + "learning_rate": 1.5332487429647315e-05, + "loss": 0.0526, + "step": 20847 + }, + { + "epoch": 2.73, + "learning_rate": 1.5332038789153965e-05, + "loss": 0.0577, + "step": 20848 + }, + { + "epoch": 2.73, + "learning_rate": 1.5331590133664533e-05, + "loss": 0.0533, + "step": 20849 + }, + { + "epoch": 2.73, + "learning_rate": 1.5331141463180298e-05, + "loss": 0.0518, + "step": 20850 + }, + { + "epoch": 2.73, + "learning_rate": 1.53306927777025e-05, + "loss": 0.0523, + "step": 20851 + }, + { + "epoch": 2.73, + "learning_rate": 1.533024407723242e-05, + "loss": 0.0488, + "step": 20852 + }, + { + "epoch": 2.73, + "learning_rate": 1.5329795361771312e-05, + "loss": 0.0569, + "step": 20853 + }, + { + "epoch": 2.73, + "learning_rate": 1.5329346631320437e-05, + "loss": 0.0553, + "step": 20854 + }, + { + "epoch": 2.73, + "learning_rate": 1.532889788588106e-05, + "loss": 0.044, + "step": 20855 + }, + { + "epoch": 2.73, + "learning_rate": 1.5328449125454437e-05, + "loss": 0.0617, + "step": 20856 + }, + { + "epoch": 2.73, + "learning_rate": 1.532800035004184e-05, + "loss": 0.0618, + "step": 20857 + }, + { + "epoch": 2.73, + "learning_rate": 1.5327551559644526e-05, + "loss": 0.0557, + "step": 20858 + }, + { + "epoch": 2.73, + "learning_rate": 1.5327102754263754e-05, + "loss": 0.0477, + "step": 20859 + }, + { + "epoch": 2.73, + "learning_rate": 1.5326653933900785e-05, + "loss": 0.0569, + "step": 20860 + }, + { + "epoch": 2.73, + "learning_rate": 1.5326205098556894e-05, + "loss": 0.0563, + "step": 20861 + }, + { + "epoch": 2.73, + "learning_rate": 1.532575624823333e-05, + "loss": 0.0539, + "step": 20862 + }, + { + "epoch": 2.73, + "learning_rate": 1.532530738293136e-05, + "loss": 0.0627, + "step": 20863 + }, + { + "epoch": 2.73, + "learning_rate": 1.532485850265225e-05, + "loss": 0.052, + "step": 20864 + }, + { + "epoch": 2.73, + "learning_rate": 1.5324409607397257e-05, + "loss": 0.0467, + "step": 20865 + }, + { + "epoch": 2.73, + "learning_rate": 1.5323960697167643e-05, + "loss": 0.0557, + "step": 20866 + }, + { + "epoch": 2.73, + "learning_rate": 1.5323511771964676e-05, + "loss": 0.053, + "step": 20867 + }, + { + "epoch": 2.73, + "learning_rate": 1.5323062831789615e-05, + "loss": 0.0447, + "step": 20868 + }, + { + "epoch": 2.73, + "learning_rate": 1.5322613876643725e-05, + "loss": 0.0647, + "step": 20869 + }, + { + "epoch": 2.73, + "learning_rate": 1.5322164906528262e-05, + "loss": 0.0637, + "step": 20870 + }, + { + "epoch": 2.73, + "learning_rate": 1.5321715921444497e-05, + "loss": 0.0551, + "step": 20871 + }, + { + "epoch": 2.73, + "learning_rate": 1.532126692139369e-05, + "loss": 0.0639, + "step": 20872 + }, + { + "epoch": 2.73, + "learning_rate": 1.5320817906377103e-05, + "loss": 0.061, + "step": 20873 + }, + { + "epoch": 2.73, + "learning_rate": 1.5320368876395997e-05, + "loss": 0.0587, + "step": 20874 + }, + { + "epoch": 2.73, + "learning_rate": 1.5319919831451638e-05, + "loss": 0.0558, + "step": 20875 + }, + { + "epoch": 2.73, + "learning_rate": 1.531947077154529e-05, + "loss": 0.0642, + "step": 20876 + }, + { + "epoch": 2.73, + "learning_rate": 1.531902169667821e-05, + "loss": 0.0596, + "step": 20877 + }, + { + "epoch": 2.73, + "learning_rate": 1.5318572606851665e-05, + "loss": 0.0575, + "step": 20878 + }, + { + "epoch": 2.74, + "learning_rate": 1.531812350206692e-05, + "loss": 0.0499, + "step": 20879 + }, + { + "epoch": 2.74, + "learning_rate": 1.5317674382325237e-05, + "loss": 0.0579, + "step": 20880 + }, + { + "epoch": 2.74, + "learning_rate": 1.5317225247627874e-05, + "loss": 0.0578, + "step": 20881 + }, + { + "epoch": 2.74, + "learning_rate": 1.53167760979761e-05, + "loss": 0.0556, + "step": 20882 + }, + { + "epoch": 2.74, + "learning_rate": 1.5316326933371174e-05, + "loss": 0.0557, + "step": 20883 + }, + { + "epoch": 2.74, + "learning_rate": 1.5315877753814367e-05, + "loss": 0.0572, + "step": 20884 + }, + { + "epoch": 2.74, + "learning_rate": 1.5315428559306933e-05, + "loss": 0.0575, + "step": 20885 + }, + { + "epoch": 2.74, + "learning_rate": 1.531497934985014e-05, + "loss": 0.0466, + "step": 20886 + }, + { + "epoch": 2.74, + "learning_rate": 1.531453012544525e-05, + "loss": 0.0519, + "step": 20887 + }, + { + "epoch": 2.74, + "learning_rate": 1.5314080886093528e-05, + "loss": 0.0523, + "step": 20888 + }, + { + "epoch": 2.74, + "learning_rate": 1.5313631631796234e-05, + "loss": 0.0604, + "step": 20889 + }, + { + "epoch": 2.74, + "learning_rate": 1.531318236255464e-05, + "loss": 0.0603, + "step": 20890 + }, + { + "epoch": 2.74, + "learning_rate": 1.531273307837e-05, + "loss": 0.0588, + "step": 20891 + }, + { + "epoch": 2.74, + "learning_rate": 1.531228377924358e-05, + "loss": 0.0428, + "step": 20892 + }, + { + "epoch": 2.74, + "learning_rate": 1.5311834465176645e-05, + "loss": 0.0574, + "step": 20893 + }, + { + "epoch": 2.74, + "learning_rate": 1.531138513617046e-05, + "loss": 0.0568, + "step": 20894 + }, + { + "epoch": 2.74, + "learning_rate": 1.5310935792226285e-05, + "loss": 0.0561, + "step": 20895 + }, + { + "epoch": 2.74, + "learning_rate": 1.531048643334539e-05, + "loss": 0.0591, + "step": 20896 + }, + { + "epoch": 2.74, + "learning_rate": 1.531003705952903e-05, + "loss": 0.0608, + "step": 20897 + }, + { + "epoch": 2.74, + "learning_rate": 1.5309587670778477e-05, + "loss": 0.063, + "step": 20898 + }, + { + "epoch": 2.74, + "learning_rate": 1.5309138267094987e-05, + "loss": 0.0527, + "step": 20899 + }, + { + "epoch": 2.74, + "learning_rate": 1.5308688848479834e-05, + "loss": 0.055, + "step": 20900 + }, + { + "epoch": 2.74, + "learning_rate": 1.530823941493427e-05, + "loss": 0.0604, + "step": 20901 + }, + { + "epoch": 2.74, + "learning_rate": 1.530778996645957e-05, + "loss": 0.0448, + "step": 20902 + }, + { + "epoch": 2.74, + "learning_rate": 1.5307340503056995e-05, + "loss": 0.054, + "step": 20903 + }, + { + "epoch": 2.74, + "learning_rate": 1.5306891024727805e-05, + "loss": 0.0582, + "step": 20904 + }, + { + "epoch": 2.74, + "learning_rate": 1.5306441531473267e-05, + "loss": 0.0591, + "step": 20905 + }, + { + "epoch": 2.74, + "learning_rate": 1.5305992023294643e-05, + "loss": 0.0514, + "step": 20906 + }, + { + "epoch": 2.74, + "learning_rate": 1.53055425001932e-05, + "loss": 0.0603, + "step": 20907 + }, + { + "epoch": 2.74, + "learning_rate": 1.5305092962170202e-05, + "loss": 0.0582, + "step": 20908 + }, + { + "epoch": 2.74, + "learning_rate": 1.530464340922691e-05, + "loss": 0.0564, + "step": 20909 + }, + { + "epoch": 2.74, + "learning_rate": 1.5304193841364595e-05, + "loss": 0.054, + "step": 20910 + }, + { + "epoch": 2.74, + "learning_rate": 1.5303744258584513e-05, + "loss": 0.0514, + "step": 20911 + }, + { + "epoch": 2.74, + "learning_rate": 1.5303294660887936e-05, + "loss": 0.0587, + "step": 20912 + }, + { + "epoch": 2.74, + "learning_rate": 1.5302845048276126e-05, + "loss": 0.0536, + "step": 20913 + }, + { + "epoch": 2.74, + "learning_rate": 1.5302395420750343e-05, + "loss": 0.0561, + "step": 20914 + }, + { + "epoch": 2.74, + "learning_rate": 1.5301945778311856e-05, + "loss": 0.0571, + "step": 20915 + }, + { + "epoch": 2.74, + "learning_rate": 1.530149612096193e-05, + "loss": 0.0616, + "step": 20916 + }, + { + "epoch": 2.74, + "learning_rate": 1.530104644870183e-05, + "loss": 0.0564, + "step": 20917 + }, + { + "epoch": 2.74, + "learning_rate": 1.5300596761532814e-05, + "loss": 0.0496, + "step": 20918 + }, + { + "epoch": 2.74, + "learning_rate": 1.5300147059456158e-05, + "loss": 0.0471, + "step": 20919 + }, + { + "epoch": 2.74, + "learning_rate": 1.5299697342473117e-05, + "loss": 0.0546, + "step": 20920 + }, + { + "epoch": 2.74, + "learning_rate": 1.529924761058496e-05, + "loss": 0.0626, + "step": 20921 + }, + { + "epoch": 2.74, + "learning_rate": 1.5298797863792953e-05, + "loss": 0.0646, + "step": 20922 + }, + { + "epoch": 2.74, + "learning_rate": 1.5298348102098357e-05, + "loss": 0.0612, + "step": 20923 + }, + { + "epoch": 2.74, + "learning_rate": 1.529789832550244e-05, + "loss": 0.0668, + "step": 20924 + }, + { + "epoch": 2.74, + "learning_rate": 1.529744853400647e-05, + "loss": 0.0614, + "step": 20925 + }, + { + "epoch": 2.74, + "learning_rate": 1.52969987276117e-05, + "loss": 0.0458, + "step": 20926 + }, + { + "epoch": 2.74, + "learning_rate": 1.529654890631941e-05, + "loss": 0.0576, + "step": 20927 + }, + { + "epoch": 2.74, + "learning_rate": 1.5296099070130853e-05, + "loss": 0.0671, + "step": 20928 + }, + { + "epoch": 2.74, + "learning_rate": 1.5295649219047305e-05, + "loss": 0.0646, + "step": 20929 + }, + { + "epoch": 2.74, + "learning_rate": 1.5295199353070024e-05, + "loss": 0.066, + "step": 20930 + }, + { + "epoch": 2.74, + "learning_rate": 1.529474947220027e-05, + "loss": 0.0616, + "step": 20931 + }, + { + "epoch": 2.74, + "learning_rate": 1.529429957643932e-05, + "loss": 0.0649, + "step": 20932 + }, + { + "epoch": 2.74, + "learning_rate": 1.5293849665788435e-05, + "loss": 0.053, + "step": 20933 + }, + { + "epoch": 2.74, + "learning_rate": 1.529339974024888e-05, + "loss": 0.0505, + "step": 20934 + }, + { + "epoch": 2.74, + "learning_rate": 1.529294979982192e-05, + "loss": 0.0579, + "step": 20935 + }, + { + "epoch": 2.74, + "learning_rate": 1.5292499844508823e-05, + "loss": 0.0618, + "step": 20936 + }, + { + "epoch": 2.74, + "learning_rate": 1.5292049874310846e-05, + "loss": 0.064, + "step": 20937 + }, + { + "epoch": 2.74, + "learning_rate": 1.5291599889229267e-05, + "loss": 0.0509, + "step": 20938 + }, + { + "epoch": 2.74, + "learning_rate": 1.5291149889265343e-05, + "loss": 0.0506, + "step": 20939 + }, + { + "epoch": 2.74, + "learning_rate": 1.5290699874420338e-05, + "loss": 0.0595, + "step": 20940 + }, + { + "epoch": 2.74, + "learning_rate": 1.5290249844695525e-05, + "loss": 0.052, + "step": 20941 + }, + { + "epoch": 2.74, + "learning_rate": 1.5289799800092165e-05, + "loss": 0.0563, + "step": 20942 + }, + { + "epoch": 2.74, + "learning_rate": 1.5289349740611525e-05, + "loss": 0.0544, + "step": 20943 + }, + { + "epoch": 2.74, + "learning_rate": 1.5288899666254872e-05, + "loss": 0.0704, + "step": 20944 + }, + { + "epoch": 2.74, + "learning_rate": 1.5288449577023467e-05, + "loss": 0.0575, + "step": 20945 + }, + { + "epoch": 2.74, + "learning_rate": 1.5287999472918583e-05, + "loss": 0.0589, + "step": 20946 + }, + { + "epoch": 2.74, + "learning_rate": 1.528754935394148e-05, + "loss": 0.0563, + "step": 20947 + }, + { + "epoch": 2.74, + "learning_rate": 1.5287099220093426e-05, + "loss": 0.0585, + "step": 20948 + }, + { + "epoch": 2.74, + "learning_rate": 1.5286649071375686e-05, + "loss": 0.0632, + "step": 20949 + }, + { + "epoch": 2.74, + "learning_rate": 1.5286198907789527e-05, + "loss": 0.0566, + "step": 20950 + }, + { + "epoch": 2.74, + "learning_rate": 1.5285748729336218e-05, + "loss": 0.0572, + "step": 20951 + }, + { + "epoch": 2.74, + "learning_rate": 1.528529853601702e-05, + "loss": 0.0692, + "step": 20952 + }, + { + "epoch": 2.74, + "learning_rate": 1.52848483278332e-05, + "loss": 0.0579, + "step": 20953 + }, + { + "epoch": 2.74, + "learning_rate": 1.5284398104786026e-05, + "loss": 0.0527, + "step": 20954 + }, + { + "epoch": 2.74, + "learning_rate": 1.5283947866876767e-05, + "loss": 0.0485, + "step": 20955 + }, + { + "epoch": 2.75, + "learning_rate": 1.5283497614106685e-05, + "loss": 0.0618, + "step": 20956 + }, + { + "epoch": 2.75, + "learning_rate": 1.5283047346477046e-05, + "loss": 0.0666, + "step": 20957 + }, + { + "epoch": 2.75, + "learning_rate": 1.5282597063989114e-05, + "loss": 0.0625, + "step": 20958 + }, + { + "epoch": 2.75, + "learning_rate": 1.5282146766644167e-05, + "loss": 0.057, + "step": 20959 + }, + { + "epoch": 2.75, + "learning_rate": 1.5281696454443458e-05, + "loss": 0.0514, + "step": 20960 + }, + { + "epoch": 2.75, + "learning_rate": 1.528124612738826e-05, + "loss": 0.0537, + "step": 20961 + }, + { + "epoch": 2.75, + "learning_rate": 1.528079578547984e-05, + "loss": 0.0573, + "step": 20962 + }, + { + "epoch": 2.75, + "learning_rate": 1.528034542871946e-05, + "loss": 0.0578, + "step": 20963 + }, + { + "epoch": 2.75, + "learning_rate": 1.5279895057108393e-05, + "loss": 0.056, + "step": 20964 + }, + { + "epoch": 2.75, + "learning_rate": 1.52794446706479e-05, + "loss": 0.062, + "step": 20965 + }, + { + "epoch": 2.75, + "learning_rate": 1.5278994269339253e-05, + "loss": 0.055, + "step": 20966 + }, + { + "epoch": 2.75, + "learning_rate": 1.527854385318371e-05, + "loss": 0.0526, + "step": 20967 + }, + { + "epoch": 2.75, + "learning_rate": 1.5278093422182552e-05, + "loss": 0.0539, + "step": 20968 + }, + { + "epoch": 2.75, + "learning_rate": 1.5277642976337032e-05, + "loss": 0.0472, + "step": 20969 + }, + { + "epoch": 2.75, + "learning_rate": 1.527719251564842e-05, + "loss": 0.0545, + "step": 20970 + }, + { + "epoch": 2.75, + "learning_rate": 1.5276742040117988e-05, + "loss": 0.0564, + "step": 20971 + }, + { + "epoch": 2.75, + "learning_rate": 1.5276291549747002e-05, + "loss": 0.0653, + "step": 20972 + }, + { + "epoch": 2.75, + "learning_rate": 1.5275841044536727e-05, + "loss": 0.0471, + "step": 20973 + }, + { + "epoch": 2.75, + "learning_rate": 1.5275390524488427e-05, + "loss": 0.0482, + "step": 20974 + }, + { + "epoch": 2.75, + "learning_rate": 1.527493998960337e-05, + "loss": 0.0561, + "step": 20975 + }, + { + "epoch": 2.75, + "learning_rate": 1.5274489439882828e-05, + "loss": 0.0626, + "step": 20976 + }, + { + "epoch": 2.75, + "learning_rate": 1.5274038875328066e-05, + "loss": 0.0589, + "step": 20977 + }, + { + "epoch": 2.75, + "learning_rate": 1.5273588295940348e-05, + "loss": 0.0573, + "step": 20978 + }, + { + "epoch": 2.75, + "learning_rate": 1.527313770172095e-05, + "loss": 0.0508, + "step": 20979 + }, + { + "epoch": 2.75, + "learning_rate": 1.5272687092671125e-05, + "loss": 0.0555, + "step": 20980 + }, + { + "epoch": 2.75, + "learning_rate": 1.5272236468792154e-05, + "loss": 0.0493, + "step": 20981 + }, + { + "epoch": 2.75, + "learning_rate": 1.5271785830085296e-05, + "loss": 0.0664, + "step": 20982 + }, + { + "epoch": 2.75, + "learning_rate": 1.5271335176551824e-05, + "loss": 0.0604, + "step": 20983 + }, + { + "epoch": 2.75, + "learning_rate": 1.5270884508192998e-05, + "loss": 0.0578, + "step": 20984 + }, + { + "epoch": 2.75, + "learning_rate": 1.5270433825010096e-05, + "loss": 0.0553, + "step": 20985 + }, + { + "epoch": 2.75, + "learning_rate": 1.5269983127004372e-05, + "loss": 0.055, + "step": 20986 + }, + { + "epoch": 2.75, + "learning_rate": 1.5269532414177106e-05, + "loss": 0.0574, + "step": 20987 + }, + { + "epoch": 2.75, + "learning_rate": 1.526908168652956e-05, + "loss": 0.067, + "step": 20988 + }, + { + "epoch": 2.75, + "learning_rate": 1.5268630944063004e-05, + "loss": 0.06, + "step": 20989 + }, + { + "epoch": 2.75, + "learning_rate": 1.52681801867787e-05, + "loss": 0.0679, + "step": 20990 + }, + { + "epoch": 2.75, + "learning_rate": 1.5267729414677924e-05, + "loss": 0.0546, + "step": 20991 + }, + { + "epoch": 2.75, + "learning_rate": 1.5267278627761933e-05, + "loss": 0.0526, + "step": 20992 + }, + { + "epoch": 2.75, + "learning_rate": 1.526682782603201e-05, + "loss": 0.0437, + "step": 20993 + }, + { + "epoch": 2.75, + "learning_rate": 1.526637700948941e-05, + "loss": 0.0578, + "step": 20994 + }, + { + "epoch": 2.75, + "learning_rate": 1.5265926178135406e-05, + "loss": 0.0617, + "step": 20995 + }, + { + "epoch": 2.75, + "learning_rate": 1.5265475331971263e-05, + "loss": 0.0616, + "step": 20996 + }, + { + "epoch": 2.75, + "learning_rate": 1.526502447099825e-05, + "loss": 0.0626, + "step": 20997 + }, + { + "epoch": 2.75, + "learning_rate": 1.526457359521764e-05, + "loss": 0.0564, + "step": 20998 + }, + { + "epoch": 2.75, + "learning_rate": 1.52641227046307e-05, + "loss": 0.0647, + "step": 20999 + }, + { + "epoch": 2.75, + "learning_rate": 1.5263671799238688e-05, + "loss": 0.0603, + "step": 21000 + }, + { + "epoch": 2.75, + "learning_rate": 1.526322087904288e-05, + "loss": 0.0652, + "step": 21001 + }, + { + "epoch": 2.75, + "learning_rate": 1.5262769944044546e-05, + "loss": 0.0553, + "step": 21002 + }, + { + "epoch": 2.75, + "learning_rate": 1.526231899424495e-05, + "loss": 0.049, + "step": 21003 + }, + { + "epoch": 2.75, + "learning_rate": 1.5261868029645363e-05, + "loss": 0.0581, + "step": 21004 + }, + { + "epoch": 2.75, + "learning_rate": 1.5261417050247054e-05, + "loss": 0.0498, + "step": 21005 + }, + { + "epoch": 2.75, + "learning_rate": 1.5260966056051293e-05, + "loss": 0.045, + "step": 21006 + }, + { + "epoch": 2.75, + "learning_rate": 1.526051504705934e-05, + "loss": 0.0723, + "step": 21007 + }, + { + "epoch": 2.75, + "learning_rate": 1.5260064023272465e-05, + "loss": 0.0642, + "step": 21008 + }, + { + "epoch": 2.75, + "learning_rate": 1.5259612984691944e-05, + "loss": 0.0629, + "step": 21009 + }, + { + "epoch": 2.75, + "learning_rate": 1.5259161931319046e-05, + "loss": 0.0634, + "step": 21010 + }, + { + "epoch": 2.75, + "learning_rate": 1.525871086315503e-05, + "loss": 0.0518, + "step": 21011 + }, + { + "epoch": 2.75, + "learning_rate": 1.5258259780201172e-05, + "loss": 0.0534, + "step": 21012 + }, + { + "epoch": 2.75, + "learning_rate": 1.5257808682458732e-05, + "loss": 0.0651, + "step": 21013 + }, + { + "epoch": 2.75, + "learning_rate": 1.5257357569928991e-05, + "loss": 0.0582, + "step": 21014 + }, + { + "epoch": 2.75, + "learning_rate": 1.5256906442613214e-05, + "loss": 0.0577, + "step": 21015 + }, + { + "epoch": 2.75, + "learning_rate": 1.5256455300512665e-05, + "loss": 0.0566, + "step": 21016 + }, + { + "epoch": 2.75, + "learning_rate": 1.5256004143628614e-05, + "loss": 0.0568, + "step": 21017 + }, + { + "epoch": 2.75, + "learning_rate": 1.5255552971962333e-05, + "loss": 0.0533, + "step": 21018 + }, + { + "epoch": 2.75, + "learning_rate": 1.5255101785515086e-05, + "loss": 0.0578, + "step": 21019 + }, + { + "epoch": 2.75, + "learning_rate": 1.5254650584288148e-05, + "loss": 0.0634, + "step": 21020 + }, + { + "epoch": 2.75, + "learning_rate": 1.5254199368282784e-05, + "loss": 0.0543, + "step": 21021 + }, + { + "epoch": 2.75, + "learning_rate": 1.5253748137500263e-05, + "loss": 0.0526, + "step": 21022 + }, + { + "epoch": 2.75, + "learning_rate": 1.525329689194186e-05, + "loss": 0.0606, + "step": 21023 + }, + { + "epoch": 2.75, + "learning_rate": 1.5252845631608835e-05, + "loss": 0.0539, + "step": 21024 + }, + { + "epoch": 2.75, + "learning_rate": 1.5252394356502463e-05, + "loss": 0.0482, + "step": 21025 + }, + { + "epoch": 2.75, + "learning_rate": 1.5251943066624008e-05, + "loss": 0.0528, + "step": 21026 + }, + { + "epoch": 2.75, + "learning_rate": 1.5251491761974747e-05, + "loss": 0.0482, + "step": 21027 + }, + { + "epoch": 2.75, + "learning_rate": 1.5251040442555946e-05, + "loss": 0.0606, + "step": 21028 + }, + { + "epoch": 2.75, + "learning_rate": 1.525058910836887e-05, + "loss": 0.0679, + "step": 21029 + }, + { + "epoch": 2.75, + "learning_rate": 1.5250137759414791e-05, + "loss": 0.048, + "step": 21030 + }, + { + "epoch": 2.75, + "learning_rate": 1.5249686395694983e-05, + "loss": 0.0607, + "step": 21031 + }, + { + "epoch": 2.76, + "learning_rate": 1.524923501721071e-05, + "loss": 0.0679, + "step": 21032 + }, + { + "epoch": 2.76, + "learning_rate": 1.5248783623963243e-05, + "loss": 0.059, + "step": 21033 + }, + { + "epoch": 2.76, + "learning_rate": 1.5248332215953852e-05, + "loss": 0.0585, + "step": 21034 + }, + { + "epoch": 2.76, + "learning_rate": 1.5247880793183804e-05, + "loss": 0.0618, + "step": 21035 + }, + { + "epoch": 2.76, + "learning_rate": 1.5247429355654373e-05, + "loss": 0.0598, + "step": 21036 + }, + { + "epoch": 2.76, + "learning_rate": 1.5246977903366827e-05, + "loss": 0.0573, + "step": 21037 + }, + { + "epoch": 2.76, + "learning_rate": 1.5246526436322433e-05, + "loss": 0.0571, + "step": 21038 + }, + { + "epoch": 2.76, + "learning_rate": 1.5246074954522465e-05, + "loss": 0.0598, + "step": 21039 + }, + { + "epoch": 2.76, + "learning_rate": 1.5245623457968187e-05, + "loss": 0.058, + "step": 21040 + }, + { + "epoch": 2.76, + "learning_rate": 1.5245171946660874e-05, + "loss": 0.0523, + "step": 21041 + }, + { + "epoch": 2.76, + "learning_rate": 1.5244720420601796e-05, + "loss": 0.0618, + "step": 21042 + }, + { + "epoch": 2.76, + "learning_rate": 1.5244268879792219e-05, + "loss": 0.0534, + "step": 21043 + }, + { + "epoch": 2.76, + "learning_rate": 1.5243817324233417e-05, + "loss": 0.0503, + "step": 21044 + }, + { + "epoch": 2.76, + "learning_rate": 1.5243365753926655e-05, + "loss": 0.0607, + "step": 21045 + }, + { + "epoch": 2.76, + "learning_rate": 1.5242914168873207e-05, + "loss": 0.0604, + "step": 21046 + }, + { + "epoch": 2.76, + "learning_rate": 1.5242462569074343e-05, + "loss": 0.0559, + "step": 21047 + }, + { + "epoch": 2.76, + "learning_rate": 1.5242010954531332e-05, + "loss": 0.0556, + "step": 21048 + }, + { + "epoch": 2.76, + "learning_rate": 1.5241559325245445e-05, + "loss": 0.0597, + "step": 21049 + }, + { + "epoch": 2.76, + "learning_rate": 1.5241107681217953e-05, + "loss": 0.065, + "step": 21050 + }, + { + "epoch": 2.76, + "learning_rate": 1.524065602245012e-05, + "loss": 0.0482, + "step": 21051 + }, + { + "epoch": 2.76, + "learning_rate": 1.5240204348943227e-05, + "loss": 0.0599, + "step": 21052 + }, + { + "epoch": 2.76, + "learning_rate": 1.5239752660698533e-05, + "loss": 0.0582, + "step": 21053 + }, + { + "epoch": 2.76, + "learning_rate": 1.5239300957717316e-05, + "loss": 0.057, + "step": 21054 + }, + { + "epoch": 2.76, + "learning_rate": 1.5238849240000844e-05, + "loss": 0.0618, + "step": 21055 + }, + { + "epoch": 2.76, + "learning_rate": 1.5238397507550388e-05, + "loss": 0.0611, + "step": 21056 + }, + { + "epoch": 2.76, + "learning_rate": 1.5237945760367216e-05, + "loss": 0.063, + "step": 21057 + }, + { + "epoch": 2.76, + "learning_rate": 1.5237493998452601e-05, + "loss": 0.0673, + "step": 21058 + }, + { + "epoch": 2.76, + "learning_rate": 1.5237042221807813e-05, + "loss": 0.0478, + "step": 21059 + }, + { + "epoch": 2.76, + "learning_rate": 1.5236590430434126e-05, + "loss": 0.0611, + "step": 21060 + }, + { + "epoch": 2.76, + "learning_rate": 1.5236138624332805e-05, + "loss": 0.0644, + "step": 21061 + }, + { + "epoch": 2.76, + "learning_rate": 1.523568680350512e-05, + "loss": 0.0595, + "step": 21062 + }, + { + "epoch": 2.76, + "learning_rate": 1.5235234967952348e-05, + "loss": 0.061, + "step": 21063 + }, + { + "epoch": 2.76, + "learning_rate": 1.5234783117675759e-05, + "loss": 0.0627, + "step": 21064 + }, + { + "epoch": 2.76, + "learning_rate": 1.5234331252676615e-05, + "loss": 0.0591, + "step": 21065 + }, + { + "epoch": 2.76, + "learning_rate": 1.5233879372956195e-05, + "loss": 0.0557, + "step": 21066 + }, + { + "epoch": 2.76, + "learning_rate": 1.5233427478515772e-05, + "loss": 0.0564, + "step": 21067 + }, + { + "epoch": 2.76, + "learning_rate": 1.5232975569356607e-05, + "loss": 0.0659, + "step": 21068 + }, + { + "epoch": 2.76, + "learning_rate": 1.5232523645479982e-05, + "loss": 0.0514, + "step": 21069 + }, + { + "epoch": 2.76, + "learning_rate": 1.5232071706887158e-05, + "loss": 0.0584, + "step": 21070 + }, + { + "epoch": 2.76, + "learning_rate": 1.5231619753579416e-05, + "loss": 0.0561, + "step": 21071 + }, + { + "epoch": 2.76, + "learning_rate": 1.523116778555802e-05, + "loss": 0.0554, + "step": 21072 + }, + { + "epoch": 2.76, + "learning_rate": 1.5230715802824244e-05, + "loss": 0.0559, + "step": 21073 + }, + { + "epoch": 2.76, + "learning_rate": 1.5230263805379358e-05, + "loss": 0.065, + "step": 21074 + }, + { + "epoch": 2.76, + "learning_rate": 1.5229811793224633e-05, + "loss": 0.0559, + "step": 21075 + }, + { + "epoch": 2.76, + "learning_rate": 1.5229359766361338e-05, + "loss": 0.0503, + "step": 21076 + }, + { + "epoch": 2.76, + "learning_rate": 1.5228907724790753e-05, + "loss": 0.0525, + "step": 21077 + }, + { + "epoch": 2.76, + "learning_rate": 1.522845566851414e-05, + "loss": 0.0611, + "step": 21078 + }, + { + "epoch": 2.76, + "learning_rate": 1.5228003597532773e-05, + "loss": 0.0626, + "step": 21079 + }, + { + "epoch": 2.76, + "learning_rate": 1.5227551511847925e-05, + "loss": 0.0557, + "step": 21080 + }, + { + "epoch": 2.76, + "learning_rate": 1.5227099411460868e-05, + "loss": 0.0521, + "step": 21081 + }, + { + "epoch": 2.76, + "learning_rate": 1.5226647296372872e-05, + "loss": 0.0489, + "step": 21082 + }, + { + "epoch": 2.76, + "learning_rate": 1.5226195166585207e-05, + "loss": 0.0522, + "step": 21083 + }, + { + "epoch": 2.76, + "learning_rate": 1.5225743022099147e-05, + "loss": 0.059, + "step": 21084 + }, + { + "epoch": 2.76, + "learning_rate": 1.5225290862915966e-05, + "loss": 0.0585, + "step": 21085 + }, + { + "epoch": 2.76, + "learning_rate": 1.522483868903693e-05, + "loss": 0.0555, + "step": 21086 + }, + { + "epoch": 2.76, + "learning_rate": 1.5224386500463313e-05, + "loss": 0.0556, + "step": 21087 + }, + { + "epoch": 2.76, + "learning_rate": 1.522393429719639e-05, + "loss": 0.056, + "step": 21088 + }, + { + "epoch": 2.76, + "learning_rate": 1.5223482079237428e-05, + "loss": 0.0479, + "step": 21089 + }, + { + "epoch": 2.76, + "learning_rate": 1.52230298465877e-05, + "loss": 0.0508, + "step": 21090 + }, + { + "epoch": 2.76, + "learning_rate": 1.5222577599248477e-05, + "loss": 0.0708, + "step": 21091 + }, + { + "epoch": 2.76, + "learning_rate": 1.522212533722104e-05, + "loss": 0.0521, + "step": 21092 + }, + { + "epoch": 2.76, + "learning_rate": 1.5221673060506648e-05, + "loss": 0.0531, + "step": 21093 + }, + { + "epoch": 2.76, + "learning_rate": 1.5221220769106579e-05, + "loss": 0.0539, + "step": 21094 + }, + { + "epoch": 2.76, + "learning_rate": 1.5220768463022105e-05, + "loss": 0.0546, + "step": 21095 + }, + { + "epoch": 2.76, + "learning_rate": 1.5220316142254499e-05, + "loss": 0.045, + "step": 21096 + }, + { + "epoch": 2.76, + "learning_rate": 1.5219863806805029e-05, + "loss": 0.0656, + "step": 21097 + }, + { + "epoch": 2.76, + "learning_rate": 1.5219411456674973e-05, + "loss": 0.0582, + "step": 21098 + }, + { + "epoch": 2.76, + "learning_rate": 1.5218959091865598e-05, + "loss": 0.0535, + "step": 21099 + }, + { + "epoch": 2.76, + "learning_rate": 1.521850671237818e-05, + "loss": 0.0574, + "step": 21100 + }, + { + "epoch": 2.76, + "learning_rate": 1.5218054318213988e-05, + "loss": 0.0636, + "step": 21101 + }, + { + "epoch": 2.76, + "learning_rate": 1.5217601909374297e-05, + "loss": 0.0492, + "step": 21102 + }, + { + "epoch": 2.76, + "learning_rate": 1.5217149485860382e-05, + "loss": 0.0497, + "step": 21103 + }, + { + "epoch": 2.76, + "learning_rate": 1.5216697047673506e-05, + "loss": 0.0598, + "step": 21104 + }, + { + "epoch": 2.76, + "learning_rate": 1.521624459481495e-05, + "loss": 0.0535, + "step": 21105 + }, + { + "epoch": 2.76, + "learning_rate": 1.5215792127285985e-05, + "loss": 0.0578, + "step": 21106 + }, + { + "epoch": 2.76, + "learning_rate": 1.5215339645087878e-05, + "loss": 0.0491, + "step": 21107 + }, + { + "epoch": 2.76, + "learning_rate": 1.5214887148221911e-05, + "loss": 0.0624, + "step": 21108 + }, + { + "epoch": 2.77, + "learning_rate": 1.521443463668935e-05, + "loss": 0.0609, + "step": 21109 + }, + { + "epoch": 2.77, + "learning_rate": 1.5213982110491468e-05, + "loss": 0.0481, + "step": 21110 + }, + { + "epoch": 2.77, + "learning_rate": 1.5213529569629542e-05, + "loss": 0.0451, + "step": 21111 + }, + { + "epoch": 2.77, + "learning_rate": 1.5213077014104837e-05, + "loss": 0.0589, + "step": 21112 + }, + { + "epoch": 2.77, + "learning_rate": 1.5212624443918636e-05, + "loss": 0.0531, + "step": 21113 + }, + { + "epoch": 2.77, + "learning_rate": 1.5212171859072203e-05, + "loss": 0.0632, + "step": 21114 + }, + { + "epoch": 2.77, + "learning_rate": 1.5211719259566815e-05, + "loss": 0.0543, + "step": 21115 + }, + { + "epoch": 2.77, + "learning_rate": 1.5211266645403744e-05, + "loss": 0.0606, + "step": 21116 + }, + { + "epoch": 2.77, + "learning_rate": 1.5210814016584263e-05, + "loss": 0.0515, + "step": 21117 + }, + { + "epoch": 2.77, + "learning_rate": 1.5210361373109646e-05, + "loss": 0.0505, + "step": 21118 + }, + { + "epoch": 2.77, + "learning_rate": 1.5209908714981163e-05, + "loss": 0.0484, + "step": 21119 + }, + { + "epoch": 2.77, + "learning_rate": 1.520945604220009e-05, + "loss": 0.0527, + "step": 21120 + }, + { + "epoch": 2.77, + "learning_rate": 1.5209003354767701e-05, + "loss": 0.0589, + "step": 21121 + }, + { + "epoch": 2.77, + "learning_rate": 1.5208550652685263e-05, + "loss": 0.0548, + "step": 21122 + }, + { + "epoch": 2.77, + "learning_rate": 1.5208097935954059e-05, + "loss": 0.0602, + "step": 21123 + }, + { + "epoch": 2.77, + "learning_rate": 1.5207645204575356e-05, + "loss": 0.0533, + "step": 21124 + }, + { + "epoch": 2.77, + "learning_rate": 1.5207192458550428e-05, + "loss": 0.0631, + "step": 21125 + }, + { + "epoch": 2.77, + "learning_rate": 1.5206739697880547e-05, + "loss": 0.047, + "step": 21126 + }, + { + "epoch": 2.77, + "learning_rate": 1.5206286922566985e-05, + "loss": 0.0587, + "step": 21127 + }, + { + "epoch": 2.77, + "learning_rate": 1.5205834132611022e-05, + "loss": 0.0582, + "step": 21128 + }, + { + "epoch": 2.77, + "learning_rate": 1.5205381328013929e-05, + "loss": 0.0548, + "step": 21129 + }, + { + "epoch": 2.77, + "learning_rate": 1.5204928508776978e-05, + "loss": 0.0607, + "step": 21130 + }, + { + "epoch": 2.77, + "learning_rate": 1.5204475674901437e-05, + "loss": 0.059, + "step": 21131 + }, + { + "epoch": 2.77, + "learning_rate": 1.5204022826388593e-05, + "loss": 0.0545, + "step": 21132 + }, + { + "epoch": 2.77, + "learning_rate": 1.5203569963239706e-05, + "loss": 0.052, + "step": 21133 + }, + { + "epoch": 2.77, + "learning_rate": 1.5203117085456058e-05, + "loss": 0.0672, + "step": 21134 + }, + { + "epoch": 2.77, + "learning_rate": 1.5202664193038921e-05, + "loss": 0.0527, + "step": 21135 + }, + { + "epoch": 2.77, + "learning_rate": 1.520221128598957e-05, + "loss": 0.063, + "step": 21136 + }, + { + "epoch": 2.77, + "learning_rate": 1.5201758364309273e-05, + "loss": 0.0609, + "step": 21137 + }, + { + "epoch": 2.77, + "learning_rate": 1.5201305427999309e-05, + "loss": 0.06, + "step": 21138 + }, + { + "epoch": 2.77, + "learning_rate": 1.5200852477060947e-05, + "loss": 0.0595, + "step": 21139 + }, + { + "epoch": 2.77, + "learning_rate": 1.520039951149547e-05, + "loss": 0.0562, + "step": 21140 + }, + { + "epoch": 2.77, + "learning_rate": 1.5199946531304144e-05, + "loss": 0.0631, + "step": 21141 + }, + { + "epoch": 2.77, + "learning_rate": 1.5199493536488246e-05, + "loss": 0.0579, + "step": 21142 + }, + { + "epoch": 2.77, + "learning_rate": 1.5199040527049049e-05, + "loss": 0.0637, + "step": 21143 + }, + { + "epoch": 2.77, + "learning_rate": 1.5198587502987825e-05, + "loss": 0.0582, + "step": 21144 + }, + { + "epoch": 2.77, + "learning_rate": 1.5198134464305854e-05, + "loss": 0.0653, + "step": 21145 + }, + { + "epoch": 2.77, + "learning_rate": 1.5197681411004406e-05, + "loss": 0.0521, + "step": 21146 + }, + { + "epoch": 2.77, + "learning_rate": 1.5197228343084755e-05, + "loss": 0.0675, + "step": 21147 + }, + { + "epoch": 2.77, + "learning_rate": 1.5196775260548173e-05, + "loss": 0.058, + "step": 21148 + }, + { + "epoch": 2.77, + "learning_rate": 1.519632216339594e-05, + "loss": 0.0563, + "step": 21149 + }, + { + "epoch": 2.77, + "learning_rate": 1.5195869051629331e-05, + "loss": 0.0592, + "step": 21150 + }, + { + "epoch": 2.77, + "learning_rate": 1.5195415925249612e-05, + "loss": 0.0502, + "step": 21151 + }, + { + "epoch": 2.77, + "learning_rate": 1.5194962784258065e-05, + "loss": 0.0529, + "step": 21152 + }, + { + "epoch": 2.77, + "learning_rate": 1.5194509628655962e-05, + "loss": 0.0668, + "step": 21153 + }, + { + "epoch": 2.77, + "learning_rate": 1.5194056458444576e-05, + "loss": 0.0558, + "step": 21154 + }, + { + "epoch": 2.77, + "learning_rate": 1.5193603273625184e-05, + "loss": 0.0592, + "step": 21155 + }, + { + "epoch": 2.77, + "learning_rate": 1.5193150074199057e-05, + "loss": 0.0636, + "step": 21156 + }, + { + "epoch": 2.77, + "learning_rate": 1.5192696860167476e-05, + "loss": 0.0505, + "step": 21157 + }, + { + "epoch": 2.77, + "learning_rate": 1.5192243631531707e-05, + "loss": 0.0603, + "step": 21158 + }, + { + "epoch": 2.77, + "learning_rate": 1.5191790388293032e-05, + "loss": 0.0502, + "step": 21159 + }, + { + "epoch": 2.77, + "learning_rate": 1.519133713045272e-05, + "loss": 0.0469, + "step": 21160 + }, + { + "epoch": 2.77, + "learning_rate": 1.5190883858012053e-05, + "loss": 0.0545, + "step": 21161 + }, + { + "epoch": 2.77, + "learning_rate": 1.51904305709723e-05, + "loss": 0.0538, + "step": 21162 + }, + { + "epoch": 2.77, + "learning_rate": 1.5189977269334736e-05, + "loss": 0.0509, + "step": 21163 + }, + { + "epoch": 2.77, + "learning_rate": 1.518952395310064e-05, + "loss": 0.0584, + "step": 21164 + }, + { + "epoch": 2.77, + "learning_rate": 1.5189070622271286e-05, + "loss": 0.0574, + "step": 21165 + }, + { + "epoch": 2.77, + "learning_rate": 1.5188617276847942e-05, + "loss": 0.0529, + "step": 21166 + }, + { + "epoch": 2.77, + "learning_rate": 1.5188163916831888e-05, + "loss": 0.0708, + "step": 21167 + }, + { + "epoch": 2.77, + "learning_rate": 1.5187710542224403e-05, + "loss": 0.0596, + "step": 21168 + }, + { + "epoch": 2.77, + "learning_rate": 1.5187257153026758e-05, + "loss": 0.0468, + "step": 21169 + }, + { + "epoch": 2.77, + "learning_rate": 1.5186803749240227e-05, + "loss": 0.0551, + "step": 21170 + }, + { + "epoch": 2.77, + "learning_rate": 1.5186350330866086e-05, + "loss": 0.0585, + "step": 21171 + }, + { + "epoch": 2.77, + "learning_rate": 1.5185896897905613e-05, + "loss": 0.0488, + "step": 21172 + }, + { + "epoch": 2.77, + "learning_rate": 1.5185443450360077e-05, + "loss": 0.0571, + "step": 21173 + }, + { + "epoch": 2.77, + "learning_rate": 1.5184989988230762e-05, + "loss": 0.056, + "step": 21174 + }, + { + "epoch": 2.77, + "learning_rate": 1.518453651151894e-05, + "loss": 0.0499, + "step": 21175 + }, + { + "epoch": 2.77, + "learning_rate": 1.5184083020225882e-05, + "loss": 0.058, + "step": 21176 + }, + { + "epoch": 2.77, + "learning_rate": 1.5183629514352864e-05, + "loss": 0.0672, + "step": 21177 + }, + { + "epoch": 2.77, + "learning_rate": 1.518317599390117e-05, + "loss": 0.0623, + "step": 21178 + }, + { + "epoch": 2.77, + "learning_rate": 1.5182722458872064e-05, + "loss": 0.0616, + "step": 21179 + }, + { + "epoch": 2.77, + "learning_rate": 1.518226890926683e-05, + "loss": 0.0529, + "step": 21180 + }, + { + "epoch": 2.77, + "learning_rate": 1.518181534508674e-05, + "loss": 0.0545, + "step": 21181 + }, + { + "epoch": 2.77, + "learning_rate": 1.518136176633307e-05, + "loss": 0.0578, + "step": 21182 + }, + { + "epoch": 2.77, + "learning_rate": 1.5180908173007099e-05, + "loss": 0.0558, + "step": 21183 + }, + { + "epoch": 2.77, + "learning_rate": 1.5180454565110096e-05, + "loss": 0.0626, + "step": 21184 + }, + { + "epoch": 2.78, + "learning_rate": 1.518000094264334e-05, + "loss": 0.0536, + "step": 21185 + }, + { + "epoch": 2.78, + "learning_rate": 1.5179547305608111e-05, + "loss": 0.063, + "step": 21186 + }, + { + "epoch": 2.78, + "learning_rate": 1.517909365400568e-05, + "loss": 0.0572, + "step": 21187 + }, + { + "epoch": 2.78, + "learning_rate": 1.517863998783732e-05, + "loss": 0.0558, + "step": 21188 + }, + { + "epoch": 2.78, + "learning_rate": 1.5178186307104314e-05, + "loss": 0.0618, + "step": 21189 + }, + { + "epoch": 2.78, + "learning_rate": 1.5177732611807937e-05, + "loss": 0.0617, + "step": 21190 + }, + { + "epoch": 2.78, + "learning_rate": 1.5177278901949457e-05, + "loss": 0.0622, + "step": 21191 + }, + { + "epoch": 2.78, + "learning_rate": 1.5176825177530159e-05, + "loss": 0.0515, + "step": 21192 + }, + { + "epoch": 2.78, + "learning_rate": 1.5176371438551317e-05, + "loss": 0.0506, + "step": 21193 + }, + { + "epoch": 2.78, + "learning_rate": 1.5175917685014203e-05, + "loss": 0.0484, + "step": 21194 + }, + { + "epoch": 2.78, + "learning_rate": 1.5175463916920099e-05, + "loss": 0.0578, + "step": 21195 + }, + { + "epoch": 2.78, + "learning_rate": 1.5175010134270279e-05, + "loss": 0.0528, + "step": 21196 + }, + { + "epoch": 2.78, + "learning_rate": 1.5174556337066016e-05, + "loss": 0.0628, + "step": 21197 + }, + { + "epoch": 2.78, + "learning_rate": 1.517410252530859e-05, + "loss": 0.0534, + "step": 21198 + }, + { + "epoch": 2.78, + "learning_rate": 1.5173648698999275e-05, + "loss": 0.0607, + "step": 21199 + }, + { + "epoch": 2.78, + "learning_rate": 1.517319485813935e-05, + "loss": 0.0561, + "step": 21200 + }, + { + "epoch": 2.78, + "learning_rate": 1.5172741002730089e-05, + "loss": 0.0544, + "step": 21201 + }, + { + "epoch": 2.78, + "learning_rate": 1.517228713277277e-05, + "loss": 0.0397, + "step": 21202 + }, + { + "epoch": 2.78, + "learning_rate": 1.517183324826867e-05, + "loss": 0.0609, + "step": 21203 + }, + { + "epoch": 2.78, + "learning_rate": 1.5171379349219063e-05, + "loss": 0.0549, + "step": 21204 + }, + { + "epoch": 2.78, + "learning_rate": 1.5170925435625225e-05, + "loss": 0.0532, + "step": 21205 + }, + { + "epoch": 2.78, + "learning_rate": 1.5170471507488436e-05, + "loss": 0.0628, + "step": 21206 + }, + { + "epoch": 2.78, + "learning_rate": 1.5170017564809974e-05, + "loss": 0.0572, + "step": 21207 + }, + { + "epoch": 2.78, + "learning_rate": 1.516956360759111e-05, + "loss": 0.0626, + "step": 21208 + }, + { + "epoch": 2.78, + "learning_rate": 1.5169109635833123e-05, + "loss": 0.0568, + "step": 21209 + }, + { + "epoch": 2.78, + "learning_rate": 1.5168655649537295e-05, + "loss": 0.0545, + "step": 21210 + }, + { + "epoch": 2.78, + "learning_rate": 1.5168201648704895e-05, + "loss": 0.0594, + "step": 21211 + }, + { + "epoch": 2.78, + "learning_rate": 1.5167747633337201e-05, + "loss": 0.0476, + "step": 21212 + }, + { + "epoch": 2.78, + "learning_rate": 1.5167293603435493e-05, + "loss": 0.0525, + "step": 21213 + }, + { + "epoch": 2.78, + "learning_rate": 1.516683955900105e-05, + "loss": 0.0611, + "step": 21214 + }, + { + "epoch": 2.78, + "learning_rate": 1.5166385500035143e-05, + "loss": 0.0573, + "step": 21215 + }, + { + "epoch": 2.78, + "learning_rate": 1.5165931426539053e-05, + "loss": 0.0554, + "step": 21216 + }, + { + "epoch": 2.78, + "learning_rate": 1.5165477338514057e-05, + "loss": 0.0557, + "step": 21217 + }, + { + "epoch": 2.78, + "learning_rate": 1.5165023235961428e-05, + "loss": 0.0685, + "step": 21218 + }, + { + "epoch": 2.78, + "learning_rate": 1.5164569118882449e-05, + "loss": 0.0571, + "step": 21219 + }, + { + "epoch": 2.78, + "learning_rate": 1.5164114987278394e-05, + "loss": 0.0572, + "step": 21220 + }, + { + "epoch": 2.78, + "learning_rate": 1.5163660841150538e-05, + "loss": 0.0551, + "step": 21221 + }, + { + "epoch": 2.78, + "learning_rate": 1.5163206680500165e-05, + "loss": 0.0602, + "step": 21222 + }, + { + "epoch": 2.78, + "learning_rate": 1.5162752505328544e-05, + "loss": 0.0527, + "step": 21223 + }, + { + "epoch": 2.78, + "learning_rate": 1.5162298315636962e-05, + "loss": 0.0529, + "step": 21224 + }, + { + "epoch": 2.78, + "learning_rate": 1.5161844111426684e-05, + "loss": 0.0428, + "step": 21225 + }, + { + "epoch": 2.78, + "learning_rate": 1.5161389892698999e-05, + "loss": 0.0586, + "step": 21226 + }, + { + "epoch": 2.78, + "learning_rate": 1.5160935659455178e-05, + "loss": 0.0568, + "step": 21227 + }, + { + "epoch": 2.78, + "learning_rate": 1.5160481411696505e-05, + "loss": 0.0554, + "step": 21228 + }, + { + "epoch": 2.78, + "learning_rate": 1.5160027149424246e-05, + "loss": 0.0528, + "step": 21229 + }, + { + "epoch": 2.78, + "learning_rate": 1.515957287263969e-05, + "loss": 0.0592, + "step": 21230 + }, + { + "epoch": 2.78, + "learning_rate": 1.5159118581344104e-05, + "loss": 0.0655, + "step": 21231 + }, + { + "epoch": 2.78, + "learning_rate": 1.5158664275538777e-05, + "loss": 0.0685, + "step": 21232 + }, + { + "epoch": 2.78, + "learning_rate": 1.515820995522498e-05, + "loss": 0.0621, + "step": 21233 + }, + { + "epoch": 2.78, + "learning_rate": 1.5157755620403993e-05, + "loss": 0.0601, + "step": 21234 + }, + { + "epoch": 2.78, + "learning_rate": 1.5157301271077093e-05, + "loss": 0.0515, + "step": 21235 + }, + { + "epoch": 2.78, + "learning_rate": 1.5156846907245558e-05, + "loss": 0.0681, + "step": 21236 + }, + { + "epoch": 2.78, + "learning_rate": 1.5156392528910663e-05, + "loss": 0.0529, + "step": 21237 + }, + { + "epoch": 2.78, + "learning_rate": 1.515593813607369e-05, + "loss": 0.0536, + "step": 21238 + }, + { + "epoch": 2.78, + "learning_rate": 1.515548372873592e-05, + "loss": 0.052, + "step": 21239 + }, + { + "epoch": 2.78, + "learning_rate": 1.5155029306898621e-05, + "loss": 0.0478, + "step": 21240 + }, + { + "epoch": 2.78, + "learning_rate": 1.515457487056308e-05, + "loss": 0.0512, + "step": 21241 + }, + { + "epoch": 2.78, + "learning_rate": 1.5154120419730568e-05, + "loss": 0.0456, + "step": 21242 + }, + { + "epoch": 2.78, + "learning_rate": 1.515366595440237e-05, + "loss": 0.0533, + "step": 21243 + }, + { + "epoch": 2.78, + "learning_rate": 1.515321147457976e-05, + "loss": 0.0502, + "step": 21244 + }, + { + "epoch": 2.78, + "learning_rate": 1.5152756980264016e-05, + "loss": 0.0646, + "step": 21245 + }, + { + "epoch": 2.78, + "learning_rate": 1.5152302471456419e-05, + "loss": 0.058, + "step": 21246 + }, + { + "epoch": 2.78, + "learning_rate": 1.5151847948158247e-05, + "loss": 0.0643, + "step": 21247 + }, + { + "epoch": 2.78, + "learning_rate": 1.515139341037077e-05, + "loss": 0.0503, + "step": 21248 + }, + { + "epoch": 2.78, + "learning_rate": 1.5150938858095284e-05, + "loss": 0.0676, + "step": 21249 + }, + { + "epoch": 2.78, + "learning_rate": 1.5150484291333048e-05, + "loss": 0.0571, + "step": 21250 + }, + { + "epoch": 2.78, + "learning_rate": 1.5150029710085355e-05, + "loss": 0.0568, + "step": 21251 + }, + { + "epoch": 2.78, + "learning_rate": 1.5149575114353474e-05, + "loss": 0.0582, + "step": 21252 + }, + { + "epoch": 2.78, + "learning_rate": 1.5149120504138686e-05, + "loss": 0.0537, + "step": 21253 + }, + { + "epoch": 2.78, + "learning_rate": 1.5148665879442273e-05, + "loss": 0.0579, + "step": 21254 + }, + { + "epoch": 2.78, + "learning_rate": 1.5148211240265514e-05, + "loss": 0.0489, + "step": 21255 + }, + { + "epoch": 2.78, + "learning_rate": 1.5147756586609683e-05, + "loss": 0.0563, + "step": 21256 + }, + { + "epoch": 2.78, + "learning_rate": 1.5147301918476059e-05, + "loss": 0.0557, + "step": 21257 + }, + { + "epoch": 2.78, + "learning_rate": 1.5146847235865925e-05, + "loss": 0.0588, + "step": 21258 + }, + { + "epoch": 2.78, + "learning_rate": 1.5146392538780554e-05, + "loss": 0.0636, + "step": 21259 + }, + { + "epoch": 2.78, + "learning_rate": 1.5145937827221229e-05, + "loss": 0.0532, + "step": 21260 + }, + { + "epoch": 2.79, + "learning_rate": 1.5145483101189232e-05, + "loss": 0.0561, + "step": 21261 + }, + { + "epoch": 2.79, + "learning_rate": 1.5145028360685834e-05, + "loss": 0.0654, + "step": 21262 + }, + { + "epoch": 2.79, + "learning_rate": 1.5144573605712319e-05, + "loss": 0.0585, + "step": 21263 + }, + { + "epoch": 2.79, + "learning_rate": 1.5144118836269961e-05, + "loss": 0.0581, + "step": 21264 + }, + { + "epoch": 2.79, + "learning_rate": 1.5143664052360048e-05, + "loss": 0.0573, + "step": 21265 + }, + { + "epoch": 2.79, + "learning_rate": 1.5143209253983853e-05, + "loss": 0.0558, + "step": 21266 + }, + { + "epoch": 2.79, + "learning_rate": 1.514275444114265e-05, + "loss": 0.0601, + "step": 21267 + }, + { + "epoch": 2.79, + "learning_rate": 1.5142299613837733e-05, + "loss": 0.048, + "step": 21268 + }, + { + "epoch": 2.79, + "learning_rate": 1.5141844772070367e-05, + "loss": 0.0594, + "step": 21269 + }, + { + "epoch": 2.79, + "learning_rate": 1.5141389915841835e-05, + "loss": 0.0566, + "step": 21270 + }, + { + "epoch": 2.79, + "learning_rate": 1.514093504515342e-05, + "loss": 0.0713, + "step": 21271 + }, + { + "epoch": 2.79, + "learning_rate": 1.51404801600064e-05, + "loss": 0.0592, + "step": 21272 + }, + { + "epoch": 2.79, + "learning_rate": 1.5140025260402052e-05, + "loss": 0.0605, + "step": 21273 + }, + { + "epoch": 2.79, + "learning_rate": 1.5139570346341655e-05, + "loss": 0.0559, + "step": 21274 + }, + { + "epoch": 2.79, + "learning_rate": 1.513911541782649e-05, + "loss": 0.0428, + "step": 21275 + }, + { + "epoch": 2.79, + "learning_rate": 1.5138660474857841e-05, + "loss": 0.0665, + "step": 21276 + }, + { + "epoch": 2.79, + "learning_rate": 1.513820551743698e-05, + "loss": 0.0494, + "step": 21277 + }, + { + "epoch": 2.79, + "learning_rate": 1.513775054556519e-05, + "loss": 0.0586, + "step": 21278 + }, + { + "epoch": 2.79, + "learning_rate": 1.5137295559243752e-05, + "loss": 0.0564, + "step": 21279 + }, + { + "epoch": 2.79, + "learning_rate": 1.513684055847394e-05, + "loss": 0.0666, + "step": 21280 + }, + { + "epoch": 2.79, + "learning_rate": 1.5136385543257039e-05, + "loss": 0.0588, + "step": 21281 + }, + { + "epoch": 2.79, + "learning_rate": 1.5135930513594325e-05, + "loss": 0.0616, + "step": 21282 + }, + { + "epoch": 2.79, + "learning_rate": 1.5135475469487087e-05, + "loss": 0.0558, + "step": 21283 + }, + { + "epoch": 2.79, + "learning_rate": 1.5135020410936592e-05, + "loss": 0.0622, + "step": 21284 + }, + { + "epoch": 2.79, + "learning_rate": 1.513456533794413e-05, + "loss": 0.0492, + "step": 21285 + }, + { + "epoch": 2.79, + "learning_rate": 1.5134110250510969e-05, + "loss": 0.0534, + "step": 21286 + }, + { + "epoch": 2.79, + "learning_rate": 1.5133655148638403e-05, + "loss": 0.061, + "step": 21287 + }, + { + "epoch": 2.79, + "learning_rate": 1.51332000323277e-05, + "loss": 0.0566, + "step": 21288 + }, + { + "epoch": 2.79, + "learning_rate": 1.5132744901580152e-05, + "loss": 0.0531, + "step": 21289 + }, + { + "epoch": 2.79, + "learning_rate": 1.5132289756397028e-05, + "loss": 0.062, + "step": 21290 + }, + { + "epoch": 2.79, + "learning_rate": 1.513183459677961e-05, + "loss": 0.0519, + "step": 21291 + }, + { + "epoch": 2.79, + "learning_rate": 1.5131379422729185e-05, + "loss": 0.0581, + "step": 21292 + }, + { + "epoch": 2.79, + "learning_rate": 1.5130924234247026e-05, + "loss": 0.0612, + "step": 21293 + }, + { + "epoch": 2.79, + "learning_rate": 1.5130469031334416e-05, + "loss": 0.0641, + "step": 21294 + }, + { + "epoch": 2.79, + "learning_rate": 1.5130013813992635e-05, + "loss": 0.0674, + "step": 21295 + }, + { + "epoch": 2.79, + "learning_rate": 1.5129558582222967e-05, + "loss": 0.0516, + "step": 21296 + }, + { + "epoch": 2.79, + "learning_rate": 1.5129103336026685e-05, + "loss": 0.0575, + "step": 21297 + }, + { + "epoch": 2.79, + "learning_rate": 1.5128648075405073e-05, + "loss": 0.0549, + "step": 21298 + }, + { + "epoch": 2.79, + "learning_rate": 1.5128192800359411e-05, + "loss": 0.0524, + "step": 21299 + }, + { + "epoch": 2.79, + "learning_rate": 1.5127737510890984e-05, + "loss": 0.0487, + "step": 21300 + }, + { + "epoch": 2.79, + "learning_rate": 1.5127282207001063e-05, + "loss": 0.0562, + "step": 21301 + }, + { + "epoch": 2.79, + "learning_rate": 1.5126826888690937e-05, + "loss": 0.0535, + "step": 21302 + }, + { + "epoch": 2.79, + "learning_rate": 1.5126371555961883e-05, + "loss": 0.0606, + "step": 21303 + }, + { + "epoch": 2.79, + "learning_rate": 1.5125916208815181e-05, + "loss": 0.058, + "step": 21304 + }, + { + "epoch": 2.79, + "learning_rate": 1.5125460847252114e-05, + "loss": 0.0629, + "step": 21305 + }, + { + "epoch": 2.79, + "learning_rate": 1.5125005471273962e-05, + "loss": 0.0547, + "step": 21306 + }, + { + "epoch": 2.79, + "learning_rate": 1.5124550080882001e-05, + "loss": 0.0589, + "step": 21307 + }, + { + "epoch": 2.79, + "learning_rate": 1.5124094676077519e-05, + "loss": 0.0638, + "step": 21308 + }, + { + "epoch": 2.79, + "learning_rate": 1.5123639256861793e-05, + "loss": 0.0606, + "step": 21309 + }, + { + "epoch": 2.79, + "learning_rate": 1.5123183823236105e-05, + "loss": 0.057, + "step": 21310 + }, + { + "epoch": 2.79, + "learning_rate": 1.5122728375201735e-05, + "loss": 0.0586, + "step": 21311 + }, + { + "epoch": 2.79, + "learning_rate": 1.5122272912759966e-05, + "loss": 0.0519, + "step": 21312 + }, + { + "epoch": 2.79, + "learning_rate": 1.5121817435912072e-05, + "loss": 0.0565, + "step": 21313 + }, + { + "epoch": 2.79, + "learning_rate": 1.5121361944659343e-05, + "loss": 0.0603, + "step": 21314 + }, + { + "epoch": 2.79, + "learning_rate": 1.5120906439003056e-05, + "loss": 0.0515, + "step": 21315 + }, + { + "epoch": 2.79, + "learning_rate": 1.5120450918944491e-05, + "loss": 0.0491, + "step": 21316 + }, + { + "epoch": 2.79, + "learning_rate": 1.5119995384484933e-05, + "loss": 0.0614, + "step": 21317 + }, + { + "epoch": 2.79, + "learning_rate": 1.5119539835625658e-05, + "loss": 0.0608, + "step": 21318 + }, + { + "epoch": 2.79, + "learning_rate": 1.511908427236795e-05, + "loss": 0.0532, + "step": 21319 + }, + { + "epoch": 2.79, + "learning_rate": 1.511862869471309e-05, + "loss": 0.0577, + "step": 21320 + }, + { + "epoch": 2.79, + "learning_rate": 1.5118173102662362e-05, + "loss": 0.0484, + "step": 21321 + }, + { + "epoch": 2.79, + "learning_rate": 1.5117717496217043e-05, + "loss": 0.0526, + "step": 21322 + }, + { + "epoch": 2.79, + "learning_rate": 1.5117261875378417e-05, + "loss": 0.0565, + "step": 21323 + }, + { + "epoch": 2.79, + "learning_rate": 1.5116806240147761e-05, + "loss": 0.0526, + "step": 21324 + }, + { + "epoch": 2.79, + "learning_rate": 1.5116350590526364e-05, + "loss": 0.0645, + "step": 21325 + }, + { + "epoch": 2.79, + "learning_rate": 1.5115894926515503e-05, + "loss": 0.0599, + "step": 21326 + }, + { + "epoch": 2.79, + "learning_rate": 1.5115439248116461e-05, + "loss": 0.0536, + "step": 21327 + }, + { + "epoch": 2.79, + "learning_rate": 1.5114983555330516e-05, + "loss": 0.0497, + "step": 21328 + }, + { + "epoch": 2.79, + "learning_rate": 1.5114527848158953e-05, + "loss": 0.0667, + "step": 21329 + }, + { + "epoch": 2.79, + "learning_rate": 1.511407212660305e-05, + "loss": 0.0503, + "step": 21330 + }, + { + "epoch": 2.79, + "learning_rate": 1.5113616390664097e-05, + "loss": 0.0678, + "step": 21331 + }, + { + "epoch": 2.79, + "learning_rate": 1.5113160640343364e-05, + "loss": 0.0559, + "step": 21332 + }, + { + "epoch": 2.79, + "learning_rate": 1.5112704875642145e-05, + "loss": 0.0633, + "step": 21333 + }, + { + "epoch": 2.79, + "learning_rate": 1.5112249096561713e-05, + "loss": 0.0627, + "step": 21334 + }, + { + "epoch": 2.79, + "learning_rate": 1.511179330310335e-05, + "loss": 0.0608, + "step": 21335 + }, + { + "epoch": 2.79, + "learning_rate": 1.5111337495268343e-05, + "loss": 0.0617, + "step": 21336 + }, + { + "epoch": 2.79, + "learning_rate": 1.5110881673057973e-05, + "loss": 0.0549, + "step": 21337 + }, + { + "epoch": 2.8, + "learning_rate": 1.511042583647352e-05, + "loss": 0.0588, + "step": 21338 + }, + { + "epoch": 2.8, + "learning_rate": 1.5109969985516264e-05, + "loss": 0.0519, + "step": 21339 + }, + { + "epoch": 2.8, + "learning_rate": 1.5109514120187494e-05, + "loss": 0.0466, + "step": 21340 + }, + { + "epoch": 2.8, + "learning_rate": 1.5109058240488484e-05, + "loss": 0.0533, + "step": 21341 + }, + { + "epoch": 2.8, + "learning_rate": 1.5108602346420517e-05, + "loss": 0.0599, + "step": 21342 + }, + { + "epoch": 2.8, + "learning_rate": 1.5108146437984883e-05, + "loss": 0.0618, + "step": 21343 + }, + { + "epoch": 2.8, + "learning_rate": 1.510769051518286e-05, + "loss": 0.0513, + "step": 21344 + }, + { + "epoch": 2.8, + "learning_rate": 1.5107234578015727e-05, + "loss": 0.0609, + "step": 21345 + }, + { + "epoch": 2.8, + "learning_rate": 1.5106778626484766e-05, + "loss": 0.0518, + "step": 21346 + }, + { + "epoch": 2.8, + "learning_rate": 1.5106322660591266e-05, + "loss": 0.0475, + "step": 21347 + }, + { + "epoch": 2.8, + "learning_rate": 1.5105866680336505e-05, + "loss": 0.055, + "step": 21348 + }, + { + "epoch": 2.8, + "learning_rate": 1.5105410685721763e-05, + "loss": 0.0484, + "step": 21349 + }, + { + "epoch": 2.8, + "learning_rate": 1.5104954676748331e-05, + "loss": 0.0499, + "step": 21350 + }, + { + "epoch": 2.8, + "learning_rate": 1.510449865341748e-05, + "loss": 0.0563, + "step": 21351 + }, + { + "epoch": 2.8, + "learning_rate": 1.5104042615730504e-05, + "loss": 0.0577, + "step": 21352 + }, + { + "epoch": 2.8, + "learning_rate": 1.5103586563688675e-05, + "loss": 0.0532, + "step": 21353 + }, + { + "epoch": 2.8, + "learning_rate": 1.5103130497293284e-05, + "loss": 0.059, + "step": 21354 + }, + { + "epoch": 2.8, + "learning_rate": 1.5102674416545607e-05, + "loss": 0.0555, + "step": 21355 + }, + { + "epoch": 2.8, + "learning_rate": 1.5102218321446933e-05, + "loss": 0.0556, + "step": 21356 + }, + { + "epoch": 2.8, + "learning_rate": 1.5101762211998536e-05, + "loss": 0.0517, + "step": 21357 + }, + { + "epoch": 2.8, + "learning_rate": 1.510130608820171e-05, + "loss": 0.0553, + "step": 21358 + }, + { + "epoch": 2.8, + "learning_rate": 1.510084995005773e-05, + "loss": 0.0672, + "step": 21359 + }, + { + "epoch": 2.8, + "learning_rate": 1.5100393797567883e-05, + "loss": 0.0542, + "step": 21360 + }, + { + "epoch": 2.8, + "learning_rate": 1.509993763073345e-05, + "loss": 0.053, + "step": 21361 + }, + { + "epoch": 2.8, + "learning_rate": 1.5099481449555713e-05, + "loss": 0.0615, + "step": 21362 + }, + { + "epoch": 2.8, + "learning_rate": 1.5099025254035956e-05, + "loss": 0.066, + "step": 21363 + }, + { + "epoch": 2.8, + "learning_rate": 1.509856904417546e-05, + "loss": 0.0627, + "step": 21364 + }, + { + "epoch": 2.8, + "learning_rate": 1.5098112819975515e-05, + "loss": 0.054, + "step": 21365 + }, + { + "epoch": 2.8, + "learning_rate": 1.5097656581437394e-05, + "loss": 0.0593, + "step": 21366 + }, + { + "epoch": 2.8, + "learning_rate": 1.509720032856239e-05, + "loss": 0.0621, + "step": 21367 + }, + { + "epoch": 2.8, + "learning_rate": 1.5096744061351776e-05, + "loss": 0.0602, + "step": 21368 + }, + { + "epoch": 2.8, + "learning_rate": 1.5096287779806847e-05, + "loss": 0.0522, + "step": 21369 + }, + { + "epoch": 2.8, + "learning_rate": 1.5095831483928876e-05, + "loss": 0.0642, + "step": 21370 + }, + { + "epoch": 2.8, + "learning_rate": 1.509537517371915e-05, + "loss": 0.0636, + "step": 21371 + }, + { + "epoch": 2.8, + "learning_rate": 1.5094918849178957e-05, + "loss": 0.0646, + "step": 21372 + }, + { + "epoch": 2.8, + "learning_rate": 1.5094462510309572e-05, + "loss": 0.0546, + "step": 21373 + }, + { + "epoch": 2.8, + "learning_rate": 1.5094006157112282e-05, + "loss": 0.0511, + "step": 21374 + }, + { + "epoch": 2.8, + "learning_rate": 1.5093549789588373e-05, + "loss": 0.0493, + "step": 21375 + }, + { + "epoch": 2.8, + "learning_rate": 1.5093093407739125e-05, + "loss": 0.0683, + "step": 21376 + }, + { + "epoch": 2.8, + "learning_rate": 1.5092637011565825e-05, + "loss": 0.0539, + "step": 21377 + }, + { + "epoch": 2.8, + "learning_rate": 1.5092180601069753e-05, + "loss": 0.0556, + "step": 21378 + }, + { + "epoch": 2.8, + "learning_rate": 1.5091724176252193e-05, + "loss": 0.0547, + "step": 21379 + }, + { + "epoch": 2.8, + "learning_rate": 1.5091267737114432e-05, + "loss": 0.0549, + "step": 21380 + }, + { + "epoch": 2.8, + "learning_rate": 1.5090811283657754e-05, + "loss": 0.0524, + "step": 21381 + }, + { + "epoch": 2.8, + "learning_rate": 1.5090354815883437e-05, + "loss": 0.0552, + "step": 21382 + }, + { + "epoch": 2.8, + "learning_rate": 1.508989833379277e-05, + "loss": 0.0557, + "step": 21383 + }, + { + "epoch": 2.8, + "learning_rate": 1.5089441837387031e-05, + "loss": 0.0545, + "step": 21384 + }, + { + "epoch": 2.8, + "learning_rate": 1.508898532666751e-05, + "loss": 0.0497, + "step": 21385 + }, + { + "epoch": 2.8, + "learning_rate": 1.5088528801635491e-05, + "loss": 0.0521, + "step": 21386 + }, + { + "epoch": 2.8, + "learning_rate": 1.5088072262292253e-05, + "loss": 0.0535, + "step": 21387 + }, + { + "epoch": 2.8, + "learning_rate": 1.5087615708639086e-05, + "loss": 0.0557, + "step": 21388 + }, + { + "epoch": 2.8, + "learning_rate": 1.5087159140677268e-05, + "loss": 0.0497, + "step": 21389 + }, + { + "epoch": 2.8, + "learning_rate": 1.5086702558408086e-05, + "loss": 0.0597, + "step": 21390 + }, + { + "epoch": 2.8, + "learning_rate": 1.5086245961832825e-05, + "loss": 0.0585, + "step": 21391 + }, + { + "epoch": 2.8, + "learning_rate": 1.5085789350952769e-05, + "loss": 0.055, + "step": 21392 + }, + { + "epoch": 2.8, + "learning_rate": 1.5085332725769197e-05, + "loss": 0.0673, + "step": 21393 + }, + { + "epoch": 2.8, + "learning_rate": 1.5084876086283402e-05, + "loss": 0.0661, + "step": 21394 + }, + { + "epoch": 2.8, + "learning_rate": 1.508441943249666e-05, + "loss": 0.0552, + "step": 21395 + }, + { + "epoch": 2.8, + "learning_rate": 1.5083962764410263e-05, + "loss": 0.0604, + "step": 21396 + }, + { + "epoch": 2.8, + "learning_rate": 1.508350608202549e-05, + "loss": 0.0551, + "step": 21397 + }, + { + "epoch": 2.8, + "learning_rate": 1.5083049385343626e-05, + "loss": 0.0639, + "step": 21398 + }, + { + "epoch": 2.8, + "learning_rate": 1.5082592674365957e-05, + "loss": 0.0616, + "step": 21399 + }, + { + "epoch": 2.8, + "learning_rate": 1.5082135949093766e-05, + "loss": 0.0561, + "step": 21400 + }, + { + "epoch": 2.8, + "learning_rate": 1.5081679209528337e-05, + "loss": 0.0464, + "step": 21401 + }, + { + "epoch": 2.8, + "learning_rate": 1.5081222455670958e-05, + "loss": 0.0611, + "step": 21402 + }, + { + "epoch": 2.8, + "learning_rate": 1.5080765687522912e-05, + "loss": 0.0635, + "step": 21403 + }, + { + "epoch": 2.8, + "learning_rate": 1.508030890508548e-05, + "loss": 0.0619, + "step": 21404 + }, + { + "epoch": 2.8, + "learning_rate": 1.5079852108359956e-05, + "loss": 0.061, + "step": 21405 + }, + { + "epoch": 2.8, + "learning_rate": 1.5079395297347614e-05, + "loss": 0.0549, + "step": 21406 + }, + { + "epoch": 2.8, + "learning_rate": 1.5078938472049742e-05, + "loss": 0.0569, + "step": 21407 + }, + { + "epoch": 2.8, + "learning_rate": 1.5078481632467626e-05, + "loss": 0.0605, + "step": 21408 + }, + { + "epoch": 2.8, + "learning_rate": 1.5078024778602554e-05, + "loss": 0.0562, + "step": 21409 + }, + { + "epoch": 2.8, + "learning_rate": 1.5077567910455805e-05, + "loss": 0.0578, + "step": 21410 + }, + { + "epoch": 2.8, + "learning_rate": 1.5077111028028671e-05, + "loss": 0.0543, + "step": 21411 + }, + { + "epoch": 2.8, + "learning_rate": 1.5076654131322426e-05, + "loss": 0.0585, + "step": 21412 + }, + { + "epoch": 2.8, + "learning_rate": 1.5076197220338367e-05, + "loss": 0.046, + "step": 21413 + }, + { + "epoch": 2.81, + "learning_rate": 1.5075740295077773e-05, + "loss": 0.0539, + "step": 21414 + }, + { + "epoch": 2.81, + "learning_rate": 1.5075283355541927e-05, + "loss": 0.065, + "step": 21415 + }, + { + "epoch": 2.81, + "learning_rate": 1.5074826401732119e-05, + "loss": 0.0473, + "step": 21416 + }, + { + "epoch": 2.81, + "learning_rate": 1.507436943364963e-05, + "loss": 0.0575, + "step": 21417 + }, + { + "epoch": 2.81, + "learning_rate": 1.5073912451295747e-05, + "loss": 0.0603, + "step": 21418 + }, + { + "epoch": 2.81, + "learning_rate": 1.5073455454671759e-05, + "loss": 0.0596, + "step": 21419 + }, + { + "epoch": 2.81, + "learning_rate": 1.5072998443778946e-05, + "loss": 0.0622, + "step": 21420 + }, + { + "epoch": 2.81, + "learning_rate": 1.5072541418618596e-05, + "loss": 0.0582, + "step": 21421 + }, + { + "epoch": 2.81, + "learning_rate": 1.5072084379191989e-05, + "loss": 0.0633, + "step": 21422 + }, + { + "epoch": 2.81, + "learning_rate": 1.5071627325500418e-05, + "loss": 0.0584, + "step": 21423 + }, + { + "epoch": 2.81, + "learning_rate": 1.5071170257545165e-05, + "loss": 0.052, + "step": 21424 + }, + { + "epoch": 2.81, + "learning_rate": 1.5070713175327516e-05, + "loss": 0.0566, + "step": 21425 + }, + { + "epoch": 2.81, + "learning_rate": 1.5070256078848757e-05, + "loss": 0.0476, + "step": 21426 + }, + { + "epoch": 2.81, + "learning_rate": 1.506979896811017e-05, + "loss": 0.056, + "step": 21427 + }, + { + "epoch": 2.81, + "learning_rate": 1.5069341843113045e-05, + "loss": 0.0556, + "step": 21428 + }, + { + "epoch": 2.81, + "learning_rate": 1.5068884703858667e-05, + "loss": 0.0533, + "step": 21429 + }, + { + "epoch": 2.81, + "learning_rate": 1.506842755034832e-05, + "loss": 0.0535, + "step": 21430 + }, + { + "epoch": 2.81, + "learning_rate": 1.506797038258329e-05, + "loss": 0.0647, + "step": 21431 + }, + { + "epoch": 2.81, + "learning_rate": 1.5067513200564866e-05, + "loss": 0.0525, + "step": 21432 + }, + { + "epoch": 2.81, + "learning_rate": 1.5067056004294325e-05, + "loss": 0.0544, + "step": 21433 + }, + { + "epoch": 2.81, + "learning_rate": 1.5066598793772964e-05, + "loss": 0.0509, + "step": 21434 + }, + { + "epoch": 2.81, + "learning_rate": 1.5066141569002062e-05, + "loss": 0.0565, + "step": 21435 + }, + { + "epoch": 2.81, + "learning_rate": 1.5065684329982908e-05, + "loss": 0.0526, + "step": 21436 + }, + { + "epoch": 2.81, + "learning_rate": 1.5065227076716784e-05, + "loss": 0.061, + "step": 21437 + }, + { + "epoch": 2.81, + "learning_rate": 1.506476980920498e-05, + "loss": 0.0561, + "step": 21438 + }, + { + "epoch": 2.81, + "learning_rate": 1.5064312527448779e-05, + "loss": 0.0595, + "step": 21439 + }, + { + "epoch": 2.81, + "learning_rate": 1.5063855231449471e-05, + "loss": 0.0606, + "step": 21440 + }, + { + "epoch": 2.81, + "learning_rate": 1.5063397921208338e-05, + "loss": 0.0525, + "step": 21441 + }, + { + "epoch": 2.81, + "learning_rate": 1.506294059672667e-05, + "loss": 0.0524, + "step": 21442 + }, + { + "epoch": 2.81, + "learning_rate": 1.5062483258005748e-05, + "loss": 0.0609, + "step": 21443 + }, + { + "epoch": 2.81, + "learning_rate": 1.5062025905046867e-05, + "loss": 0.0541, + "step": 21444 + }, + { + "epoch": 2.81, + "learning_rate": 1.50615685378513e-05, + "loss": 0.0569, + "step": 21445 + }, + { + "epoch": 2.81, + "learning_rate": 1.506111115642035e-05, + "loss": 0.0551, + "step": 21446 + }, + { + "epoch": 2.81, + "learning_rate": 1.5060653760755285e-05, + "loss": 0.063, + "step": 21447 + }, + { + "epoch": 2.81, + "learning_rate": 1.5060196350857407e-05, + "loss": 0.0632, + "step": 21448 + }, + { + "epoch": 2.81, + "learning_rate": 1.5059738926727999e-05, + "loss": 0.0564, + "step": 21449 + }, + { + "epoch": 2.81, + "learning_rate": 1.5059281488368336e-05, + "loss": 0.0493, + "step": 21450 + }, + { + "epoch": 2.81, + "learning_rate": 1.5058824035779718e-05, + "loss": 0.0576, + "step": 21451 + }, + { + "epoch": 2.81, + "learning_rate": 1.5058366568963427e-05, + "loss": 0.0556, + "step": 21452 + }, + { + "epoch": 2.81, + "learning_rate": 1.5057909087920751e-05, + "loss": 0.0567, + "step": 21453 + }, + { + "epoch": 2.81, + "learning_rate": 1.5057451592652972e-05, + "loss": 0.0463, + "step": 21454 + }, + { + "epoch": 2.81, + "learning_rate": 1.5056994083161378e-05, + "loss": 0.0514, + "step": 21455 + }, + { + "epoch": 2.81, + "learning_rate": 1.505653655944726e-05, + "loss": 0.0548, + "step": 21456 + }, + { + "epoch": 2.81, + "learning_rate": 1.5056079021511905e-05, + "loss": 0.0483, + "step": 21457 + }, + { + "epoch": 2.81, + "learning_rate": 1.5055621469356592e-05, + "loss": 0.0541, + "step": 21458 + }, + { + "epoch": 2.81, + "learning_rate": 1.5055163902982618e-05, + "loss": 0.0516, + "step": 21459 + }, + { + "epoch": 2.81, + "learning_rate": 1.505470632239126e-05, + "loss": 0.051, + "step": 21460 + }, + { + "epoch": 2.81, + "learning_rate": 1.505424872758381e-05, + "loss": 0.0539, + "step": 21461 + }, + { + "epoch": 2.81, + "learning_rate": 1.5053791118561557e-05, + "loss": 0.0485, + "step": 21462 + }, + { + "epoch": 2.81, + "learning_rate": 1.5053333495325786e-05, + "loss": 0.0524, + "step": 21463 + }, + { + "epoch": 2.81, + "learning_rate": 1.5052875857877782e-05, + "loss": 0.0588, + "step": 21464 + }, + { + "epoch": 2.81, + "learning_rate": 1.5052418206218832e-05, + "loss": 0.0611, + "step": 21465 + }, + { + "epoch": 2.81, + "learning_rate": 1.5051960540350228e-05, + "loss": 0.0689, + "step": 21466 + }, + { + "epoch": 2.81, + "learning_rate": 1.5051502860273255e-05, + "loss": 0.059, + "step": 21467 + }, + { + "epoch": 2.81, + "learning_rate": 1.5051045165989196e-05, + "loss": 0.0447, + "step": 21468 + }, + { + "epoch": 2.81, + "learning_rate": 1.5050587457499341e-05, + "loss": 0.064, + "step": 21469 + }, + { + "epoch": 2.81, + "learning_rate": 1.5050129734804981e-05, + "loss": 0.0627, + "step": 21470 + }, + { + "epoch": 2.81, + "learning_rate": 1.5049671997907398e-05, + "loss": 0.0542, + "step": 21471 + }, + { + "epoch": 2.81, + "learning_rate": 1.5049214246807882e-05, + "loss": 0.0556, + "step": 21472 + }, + { + "epoch": 2.81, + "learning_rate": 1.5048756481507717e-05, + "loss": 0.0577, + "step": 21473 + }, + { + "epoch": 2.81, + "learning_rate": 1.5048298702008199e-05, + "loss": 0.0611, + "step": 21474 + }, + { + "epoch": 2.81, + "learning_rate": 1.5047840908310605e-05, + "loss": 0.0584, + "step": 21475 + }, + { + "epoch": 2.81, + "learning_rate": 1.5047383100416231e-05, + "loss": 0.05, + "step": 21476 + }, + { + "epoch": 2.81, + "learning_rate": 1.5046925278326356e-05, + "loss": 0.05, + "step": 21477 + }, + { + "epoch": 2.81, + "learning_rate": 1.5046467442042276e-05, + "loss": 0.0571, + "step": 21478 + }, + { + "epoch": 2.81, + "learning_rate": 1.5046009591565275e-05, + "loss": 0.0515, + "step": 21479 + }, + { + "epoch": 2.81, + "learning_rate": 1.5045551726896641e-05, + "loss": 0.0565, + "step": 21480 + }, + { + "epoch": 2.81, + "learning_rate": 1.5045093848037659e-05, + "loss": 0.0549, + "step": 21481 + }, + { + "epoch": 2.81, + "learning_rate": 1.5044635954989621e-05, + "loss": 0.0491, + "step": 21482 + }, + { + "epoch": 2.81, + "learning_rate": 1.5044178047753812e-05, + "loss": 0.053, + "step": 21483 + }, + { + "epoch": 2.81, + "learning_rate": 1.5043720126331523e-05, + "loss": 0.0622, + "step": 21484 + }, + { + "epoch": 2.81, + "learning_rate": 1.5043262190724037e-05, + "loss": 0.0527, + "step": 21485 + }, + { + "epoch": 2.81, + "learning_rate": 1.5042804240932647e-05, + "loss": 0.0516, + "step": 21486 + }, + { + "epoch": 2.81, + "learning_rate": 1.5042346276958638e-05, + "loss": 0.0628, + "step": 21487 + }, + { + "epoch": 2.81, + "learning_rate": 1.5041888298803296e-05, + "loss": 0.0609, + "step": 21488 + }, + { + "epoch": 2.81, + "learning_rate": 1.5041430306467915e-05, + "loss": 0.0632, + "step": 21489 + }, + { + "epoch": 2.82, + "learning_rate": 1.5040972299953778e-05, + "loss": 0.0507, + "step": 21490 + }, + { + "epoch": 2.82, + "learning_rate": 1.5040514279262178e-05, + "loss": 0.0567, + "step": 21491 + }, + { + "epoch": 2.82, + "learning_rate": 1.5040056244394394e-05, + "loss": 0.0613, + "step": 21492 + }, + { + "epoch": 2.82, + "learning_rate": 1.5039598195351724e-05, + "loss": 0.0474, + "step": 21493 + }, + { + "epoch": 2.82, + "learning_rate": 1.503914013213545e-05, + "loss": 0.0646, + "step": 21494 + }, + { + "epoch": 2.82, + "learning_rate": 1.5038682054746866e-05, + "loss": 0.064, + "step": 21495 + }, + { + "epoch": 2.82, + "learning_rate": 1.5038223963187255e-05, + "loss": 0.0469, + "step": 21496 + }, + { + "epoch": 2.82, + "learning_rate": 1.5037765857457911e-05, + "loss": 0.0506, + "step": 21497 + }, + { + "epoch": 2.82, + "learning_rate": 1.5037307737560114e-05, + "loss": 0.0547, + "step": 21498 + }, + { + "epoch": 2.82, + "learning_rate": 1.5036849603495161e-05, + "loss": 0.0565, + "step": 21499 + }, + { + "epoch": 2.82, + "learning_rate": 1.5036391455264335e-05, + "loss": 0.0574, + "step": 21500 + }, + { + "epoch": 2.82, + "learning_rate": 1.5035933292868926e-05, + "loss": 0.0596, + "step": 21501 + }, + { + "epoch": 2.82, + "learning_rate": 1.5035475116310223e-05, + "loss": 0.0506, + "step": 21502 + }, + { + "epoch": 2.82, + "learning_rate": 1.5035016925589517e-05, + "loss": 0.0595, + "step": 21503 + }, + { + "epoch": 2.82, + "learning_rate": 1.5034558720708088e-05, + "loss": 0.0466, + "step": 21504 + }, + { + "epoch": 2.82, + "learning_rate": 1.5034100501667236e-05, + "loss": 0.0545, + "step": 21505 + }, + { + "epoch": 2.82, + "learning_rate": 1.503364226846824e-05, + "loss": 0.0561, + "step": 21506 + }, + { + "epoch": 2.82, + "learning_rate": 1.5033184021112398e-05, + "loss": 0.0476, + "step": 21507 + }, + { + "epoch": 2.82, + "learning_rate": 1.5032725759600993e-05, + "loss": 0.0468, + "step": 21508 + }, + { + "epoch": 2.82, + "learning_rate": 1.503226748393531e-05, + "loss": 0.0495, + "step": 21509 + }, + { + "epoch": 2.82, + "learning_rate": 1.5031809194116647e-05, + "loss": 0.0605, + "step": 21510 + }, + { + "epoch": 2.82, + "learning_rate": 1.5031350890146285e-05, + "loss": 0.0585, + "step": 21511 + }, + { + "epoch": 2.82, + "learning_rate": 1.5030892572025524e-05, + "loss": 0.0584, + "step": 21512 + }, + { + "epoch": 2.82, + "learning_rate": 1.5030434239755641e-05, + "loss": 0.0562, + "step": 21513 + }, + { + "epoch": 2.82, + "learning_rate": 1.5029975893337926e-05, + "loss": 0.0567, + "step": 21514 + }, + { + "epoch": 2.82, + "learning_rate": 1.5029517532773677e-05, + "loss": 0.0493, + "step": 21515 + }, + { + "epoch": 2.82, + "learning_rate": 1.5029059158064177e-05, + "loss": 0.0655, + "step": 21516 + }, + { + "epoch": 2.82, + "learning_rate": 1.5028600769210713e-05, + "loss": 0.0546, + "step": 21517 + }, + { + "epoch": 2.82, + "learning_rate": 1.502814236621458e-05, + "loss": 0.0467, + "step": 21518 + }, + { + "epoch": 2.82, + "learning_rate": 1.5027683949077063e-05, + "loss": 0.0609, + "step": 21519 + }, + { + "epoch": 2.82, + "learning_rate": 1.5027225517799453e-05, + "loss": 0.062, + "step": 21520 + }, + { + "epoch": 2.82, + "learning_rate": 1.5026767072383037e-05, + "loss": 0.0449, + "step": 21521 + }, + { + "epoch": 2.82, + "learning_rate": 1.502630861282911e-05, + "loss": 0.0474, + "step": 21522 + }, + { + "epoch": 2.82, + "learning_rate": 1.5025850139138956e-05, + "loss": 0.0521, + "step": 21523 + }, + { + "epoch": 2.82, + "learning_rate": 1.5025391651313868e-05, + "loss": 0.0617, + "step": 21524 + }, + { + "epoch": 2.82, + "learning_rate": 1.502493314935513e-05, + "loss": 0.054, + "step": 21525 + }, + { + "epoch": 2.82, + "learning_rate": 1.5024474633264036e-05, + "loss": 0.0475, + "step": 21526 + }, + { + "epoch": 2.82, + "learning_rate": 1.5024016103041876e-05, + "loss": 0.0573, + "step": 21527 + }, + { + "epoch": 2.82, + "learning_rate": 1.5023557558689939e-05, + "loss": 0.0547, + "step": 21528 + }, + { + "epoch": 2.82, + "learning_rate": 1.5023099000209511e-05, + "loss": 0.0611, + "step": 21529 + }, + { + "epoch": 2.82, + "learning_rate": 1.5022640427601887e-05, + "loss": 0.0546, + "step": 21530 + }, + { + "epoch": 2.82, + "learning_rate": 1.5022181840868353e-05, + "loss": 0.0608, + "step": 21531 + }, + { + "epoch": 2.82, + "learning_rate": 1.50217232400102e-05, + "loss": 0.0629, + "step": 21532 + }, + { + "epoch": 2.82, + "learning_rate": 1.5021264625028718e-05, + "loss": 0.0519, + "step": 21533 + }, + { + "epoch": 2.82, + "learning_rate": 1.5020805995925196e-05, + "loss": 0.0558, + "step": 21534 + }, + { + "epoch": 2.82, + "learning_rate": 1.5020347352700926e-05, + "loss": 0.0564, + "step": 21535 + }, + { + "epoch": 2.82, + "learning_rate": 1.5019888695357197e-05, + "loss": 0.0482, + "step": 21536 + }, + { + "epoch": 2.82, + "learning_rate": 1.5019430023895297e-05, + "loss": 0.0555, + "step": 21537 + }, + { + "epoch": 2.82, + "learning_rate": 1.5018971338316516e-05, + "loss": 0.0585, + "step": 21538 + }, + { + "epoch": 2.82, + "learning_rate": 1.5018512638622147e-05, + "loss": 0.0624, + "step": 21539 + }, + { + "epoch": 2.82, + "learning_rate": 1.5018053924813478e-05, + "loss": 0.0537, + "step": 21540 + }, + { + "epoch": 2.82, + "learning_rate": 1.50175951968918e-05, + "loss": 0.0553, + "step": 21541 + }, + { + "epoch": 2.82, + "learning_rate": 1.5017136454858406e-05, + "loss": 0.0477, + "step": 21542 + }, + { + "epoch": 2.82, + "learning_rate": 1.5016677698714576e-05, + "loss": 0.0584, + "step": 21543 + }, + { + "epoch": 2.82, + "learning_rate": 1.5016218928461612e-05, + "loss": 0.0612, + "step": 21544 + }, + { + "epoch": 2.82, + "learning_rate": 1.50157601441008e-05, + "loss": 0.07, + "step": 21545 + }, + { + "epoch": 2.82, + "learning_rate": 1.5015301345633427e-05, + "loss": 0.053, + "step": 21546 + }, + { + "epoch": 2.82, + "learning_rate": 1.5014842533060787e-05, + "loss": 0.056, + "step": 21547 + }, + { + "epoch": 2.82, + "learning_rate": 1.5014383706384168e-05, + "loss": 0.0517, + "step": 21548 + }, + { + "epoch": 2.82, + "learning_rate": 1.5013924865604862e-05, + "loss": 0.0525, + "step": 21549 + }, + { + "epoch": 2.82, + "learning_rate": 1.5013466010724162e-05, + "loss": 0.0497, + "step": 21550 + }, + { + "epoch": 2.82, + "learning_rate": 1.5013007141743352e-05, + "loss": 0.0571, + "step": 21551 + }, + { + "epoch": 2.82, + "learning_rate": 1.501254825866373e-05, + "loss": 0.0632, + "step": 21552 + }, + { + "epoch": 2.82, + "learning_rate": 1.5012089361486582e-05, + "loss": 0.0537, + "step": 21553 + }, + { + "epoch": 2.82, + "learning_rate": 1.5011630450213198e-05, + "loss": 0.059, + "step": 21554 + }, + { + "epoch": 2.82, + "learning_rate": 1.5011171524844872e-05, + "loss": 0.0653, + "step": 21555 + }, + { + "epoch": 2.82, + "learning_rate": 1.5010712585382892e-05, + "loss": 0.0613, + "step": 21556 + }, + { + "epoch": 2.82, + "learning_rate": 1.5010253631828549e-05, + "loss": 0.0608, + "step": 21557 + }, + { + "epoch": 2.82, + "learning_rate": 1.5009794664183139e-05, + "loss": 0.0514, + "step": 21558 + }, + { + "epoch": 2.82, + "learning_rate": 1.5009335682447942e-05, + "loss": 0.0569, + "step": 21559 + }, + { + "epoch": 2.82, + "learning_rate": 1.5008876686624259e-05, + "loss": 0.0594, + "step": 21560 + }, + { + "epoch": 2.82, + "learning_rate": 1.5008417676713377e-05, + "loss": 0.0566, + "step": 21561 + }, + { + "epoch": 2.82, + "learning_rate": 1.5007958652716585e-05, + "loss": 0.0513, + "step": 21562 + }, + { + "epoch": 2.82, + "learning_rate": 1.5007499614635178e-05, + "loss": 0.0509, + "step": 21563 + }, + { + "epoch": 2.82, + "learning_rate": 1.5007040562470443e-05, + "loss": 0.0499, + "step": 21564 + }, + { + "epoch": 2.82, + "learning_rate": 1.500658149622367e-05, + "loss": 0.0509, + "step": 21565 + }, + { + "epoch": 2.82, + "learning_rate": 1.5006122415896162e-05, + "loss": 0.066, + "step": 21566 + }, + { + "epoch": 2.83, + "learning_rate": 1.5005663321489197e-05, + "loss": 0.0563, + "step": 21567 + }, + { + "epoch": 2.83, + "learning_rate": 1.5005204213004069e-05, + "loss": 0.056, + "step": 21568 + }, + { + "epoch": 2.83, + "learning_rate": 1.500474509044207e-05, + "loss": 0.0544, + "step": 21569 + }, + { + "epoch": 2.83, + "learning_rate": 1.5004285953804493e-05, + "loss": 0.0497, + "step": 21570 + }, + { + "epoch": 2.83, + "learning_rate": 1.500382680309263e-05, + "loss": 0.0608, + "step": 21571 + }, + { + "epoch": 2.83, + "learning_rate": 1.500336763830777e-05, + "loss": 0.0571, + "step": 21572 + }, + { + "epoch": 2.83, + "learning_rate": 1.5002908459451205e-05, + "loss": 0.05, + "step": 21573 + }, + { + "epoch": 2.83, + "learning_rate": 1.5002449266524224e-05, + "loss": 0.0553, + "step": 21574 + }, + { + "epoch": 2.83, + "learning_rate": 1.5001990059528124e-05, + "loss": 0.0485, + "step": 21575 + }, + { + "epoch": 2.83, + "learning_rate": 1.5001530838464188e-05, + "loss": 0.0598, + "step": 21576 + }, + { + "epoch": 2.83, + "learning_rate": 1.5001071603333718e-05, + "loss": 0.0587, + "step": 21577 + }, + { + "epoch": 2.83, + "learning_rate": 1.5000612354138002e-05, + "loss": 0.052, + "step": 21578 + }, + { + "epoch": 2.83, + "learning_rate": 1.5000153090878326e-05, + "loss": 0.0544, + "step": 21579 + }, + { + "epoch": 2.83, + "learning_rate": 1.4999693813555988e-05, + "loss": 0.0472, + "step": 21580 + }, + { + "epoch": 2.83, + "learning_rate": 1.4999234522172276e-05, + "loss": 0.0546, + "step": 21581 + }, + { + "epoch": 2.83, + "learning_rate": 1.4998775216728483e-05, + "loss": 0.058, + "step": 21582 + }, + { + "epoch": 2.83, + "learning_rate": 1.4998315897225903e-05, + "loss": 0.0595, + "step": 21583 + }, + { + "epoch": 2.83, + "learning_rate": 1.4997856563665822e-05, + "loss": 0.0549, + "step": 21584 + }, + { + "epoch": 2.83, + "learning_rate": 1.4997397216049541e-05, + "loss": 0.0466, + "step": 21585 + }, + { + "epoch": 2.83, + "learning_rate": 1.4996937854378339e-05, + "loss": 0.0558, + "step": 21586 + }, + { + "epoch": 2.83, + "learning_rate": 1.4996478478653522e-05, + "loss": 0.0573, + "step": 21587 + }, + { + "epoch": 2.83, + "learning_rate": 1.4996019088876372e-05, + "loss": 0.0563, + "step": 21588 + }, + { + "epoch": 2.83, + "learning_rate": 1.4995559685048186e-05, + "loss": 0.0666, + "step": 21589 + }, + { + "epoch": 2.83, + "learning_rate": 1.4995100267170254e-05, + "loss": 0.0595, + "step": 21590 + }, + { + "epoch": 2.83, + "learning_rate": 1.4994640835243868e-05, + "loss": 0.059, + "step": 21591 + }, + { + "epoch": 2.83, + "learning_rate": 1.499418138927032e-05, + "loss": 0.0625, + "step": 21592 + }, + { + "epoch": 2.83, + "learning_rate": 1.4993721929250904e-05, + "loss": 0.0495, + "step": 21593 + }, + { + "epoch": 2.83, + "learning_rate": 1.4993262455186912e-05, + "loss": 0.0581, + "step": 21594 + }, + { + "epoch": 2.83, + "learning_rate": 1.4992802967079634e-05, + "loss": 0.057, + "step": 21595 + }, + { + "epoch": 2.83, + "learning_rate": 1.4992343464930364e-05, + "loss": 0.0648, + "step": 21596 + }, + { + "epoch": 2.83, + "learning_rate": 1.4991883948740391e-05, + "loss": 0.0547, + "step": 21597 + }, + { + "epoch": 2.83, + "learning_rate": 1.4991424418511013e-05, + "loss": 0.0568, + "step": 21598 + }, + { + "epoch": 2.83, + "learning_rate": 1.4990964874243519e-05, + "loss": 0.0606, + "step": 21599 + }, + { + "epoch": 2.83, + "learning_rate": 1.4990505315939203e-05, + "loss": 0.0763, + "step": 21600 + }, + { + "epoch": 2.83, + "learning_rate": 1.4990045743599356e-05, + "loss": 0.0589, + "step": 21601 + }, + { + "epoch": 2.83, + "learning_rate": 1.4989586157225271e-05, + "loss": 0.0537, + "step": 21602 + }, + { + "epoch": 2.83, + "learning_rate": 1.498912655681824e-05, + "loss": 0.0601, + "step": 21603 + }, + { + "epoch": 2.83, + "learning_rate": 1.4988666942379558e-05, + "loss": 0.0462, + "step": 21604 + }, + { + "epoch": 2.83, + "learning_rate": 1.4988207313910513e-05, + "loss": 0.0617, + "step": 21605 + }, + { + "epoch": 2.83, + "learning_rate": 1.4987747671412403e-05, + "loss": 0.0529, + "step": 21606 + }, + { + "epoch": 2.83, + "learning_rate": 1.4987288014886518e-05, + "loss": 0.0597, + "step": 21607 + }, + { + "epoch": 2.83, + "learning_rate": 1.498682834433415e-05, + "loss": 0.0607, + "step": 21608 + }, + { + "epoch": 2.83, + "learning_rate": 1.4986368659756592e-05, + "loss": 0.0561, + "step": 21609 + }, + { + "epoch": 2.83, + "learning_rate": 1.498590896115514e-05, + "loss": 0.0601, + "step": 21610 + }, + { + "epoch": 2.83, + "learning_rate": 1.4985449248531083e-05, + "loss": 0.0567, + "step": 21611 + }, + { + "epoch": 2.83, + "learning_rate": 1.4984989521885716e-05, + "loss": 0.0594, + "step": 21612 + }, + { + "epoch": 2.83, + "learning_rate": 1.4984529781220332e-05, + "loss": 0.0476, + "step": 21613 + }, + { + "epoch": 2.83, + "learning_rate": 1.4984070026536223e-05, + "loss": 0.0595, + "step": 21614 + }, + { + "epoch": 2.83, + "learning_rate": 1.4983610257834682e-05, + "loss": 0.0657, + "step": 21615 + }, + { + "epoch": 2.83, + "learning_rate": 1.4983150475117003e-05, + "loss": 0.0508, + "step": 21616 + }, + { + "epoch": 2.83, + "learning_rate": 1.498269067838448e-05, + "loss": 0.0632, + "step": 21617 + }, + { + "epoch": 2.83, + "learning_rate": 1.4982230867638403e-05, + "loss": 0.059, + "step": 21618 + }, + { + "epoch": 2.83, + "learning_rate": 1.4981771042880067e-05, + "loss": 0.0588, + "step": 21619 + }, + { + "epoch": 2.83, + "learning_rate": 1.4981311204110764e-05, + "loss": 0.0544, + "step": 21620 + }, + { + "epoch": 2.83, + "learning_rate": 1.498085135133179e-05, + "loss": 0.0531, + "step": 21621 + }, + { + "epoch": 2.83, + "learning_rate": 1.4980391484544439e-05, + "loss": 0.0577, + "step": 21622 + }, + { + "epoch": 2.83, + "learning_rate": 1.4979931603750003e-05, + "loss": 0.0617, + "step": 21623 + }, + { + "epoch": 2.83, + "learning_rate": 1.4979471708949768e-05, + "loss": 0.07, + "step": 21624 + }, + { + "epoch": 2.83, + "learning_rate": 1.4979011800145037e-05, + "loss": 0.0605, + "step": 21625 + }, + { + "epoch": 2.83, + "learning_rate": 1.4978551877337101e-05, + "loss": 0.0517, + "step": 21626 + }, + { + "epoch": 2.83, + "learning_rate": 1.4978091940527257e-05, + "loss": 0.0613, + "step": 21627 + }, + { + "epoch": 2.83, + "learning_rate": 1.497763198971679e-05, + "loss": 0.0574, + "step": 21628 + }, + { + "epoch": 2.83, + "learning_rate": 1.4977172024906998e-05, + "loss": 0.0574, + "step": 21629 + }, + { + "epoch": 2.83, + "learning_rate": 1.4976712046099175e-05, + "loss": 0.0496, + "step": 21630 + }, + { + "epoch": 2.83, + "learning_rate": 1.4976252053294615e-05, + "loss": 0.0491, + "step": 21631 + }, + { + "epoch": 2.83, + "learning_rate": 1.497579204649461e-05, + "loss": 0.0585, + "step": 21632 + }, + { + "epoch": 2.83, + "learning_rate": 1.4975332025700458e-05, + "loss": 0.0611, + "step": 21633 + }, + { + "epoch": 2.83, + "learning_rate": 1.4974871990913447e-05, + "loss": 0.0588, + "step": 21634 + }, + { + "epoch": 2.83, + "learning_rate": 1.4974411942134874e-05, + "loss": 0.0532, + "step": 21635 + }, + { + "epoch": 2.83, + "learning_rate": 1.4973951879366031e-05, + "loss": 0.0545, + "step": 21636 + }, + { + "epoch": 2.83, + "learning_rate": 1.4973491802608212e-05, + "loss": 0.0616, + "step": 21637 + }, + { + "epoch": 2.83, + "learning_rate": 1.4973031711862718e-05, + "loss": 0.0551, + "step": 21638 + }, + { + "epoch": 2.83, + "learning_rate": 1.4972571607130833e-05, + "loss": 0.0512, + "step": 21639 + }, + { + "epoch": 2.83, + "learning_rate": 1.4972111488413858e-05, + "loss": 0.0557, + "step": 21640 + }, + { + "epoch": 2.83, + "learning_rate": 1.4971651355713079e-05, + "loss": 0.0565, + "step": 21641 + }, + { + "epoch": 2.83, + "learning_rate": 1.49711912090298e-05, + "loss": 0.0534, + "step": 21642 + }, + { + "epoch": 2.84, + "learning_rate": 1.497073104836531e-05, + "loss": 0.0644, + "step": 21643 + }, + { + "epoch": 2.84, + "learning_rate": 1.4970270873720903e-05, + "loss": 0.0481, + "step": 21644 + }, + { + "epoch": 2.84, + "learning_rate": 1.4969810685097872e-05, + "loss": 0.0653, + "step": 21645 + }, + { + "epoch": 2.84, + "learning_rate": 1.4969350482497515e-05, + "loss": 0.061, + "step": 21646 + }, + { + "epoch": 2.84, + "learning_rate": 1.4968890265921123e-05, + "loss": 0.0516, + "step": 21647 + }, + { + "epoch": 2.84, + "learning_rate": 1.4968430035369994e-05, + "loss": 0.0645, + "step": 21648 + }, + { + "epoch": 2.84, + "learning_rate": 1.4967969790845418e-05, + "loss": 0.0589, + "step": 21649 + }, + { + "epoch": 2.84, + "learning_rate": 1.4967509532348694e-05, + "loss": 0.0538, + "step": 21650 + }, + { + "epoch": 2.84, + "learning_rate": 1.4967049259881113e-05, + "loss": 0.0595, + "step": 21651 + }, + { + "epoch": 2.84, + "learning_rate": 1.4966588973443967e-05, + "loss": 0.0458, + "step": 21652 + }, + { + "epoch": 2.84, + "learning_rate": 1.4966128673038559e-05, + "loss": 0.0534, + "step": 21653 + }, + { + "epoch": 2.84, + "learning_rate": 1.4965668358666175e-05, + "loss": 0.0633, + "step": 21654 + }, + { + "epoch": 2.84, + "learning_rate": 1.4965208030328113e-05, + "loss": 0.0608, + "step": 21655 + }, + { + "epoch": 2.84, + "learning_rate": 1.4964747688025672e-05, + "loss": 0.0551, + "step": 21656 + }, + { + "epoch": 2.84, + "learning_rate": 1.4964287331760137e-05, + "loss": 0.0439, + "step": 21657 + }, + { + "epoch": 2.84, + "learning_rate": 1.4963826961532813e-05, + "loss": 0.0507, + "step": 21658 + }, + { + "epoch": 2.84, + "learning_rate": 1.4963366577344989e-05, + "loss": 0.0533, + "step": 21659 + }, + { + "epoch": 2.84, + "learning_rate": 1.4962906179197958e-05, + "loss": 0.0551, + "step": 21660 + }, + { + "epoch": 2.84, + "learning_rate": 1.4962445767093021e-05, + "loss": 0.0526, + "step": 21661 + }, + { + "epoch": 2.84, + "learning_rate": 1.4961985341031469e-05, + "loss": 0.0539, + "step": 21662 + }, + { + "epoch": 2.84, + "learning_rate": 1.4961524901014595e-05, + "loss": 0.0555, + "step": 21663 + }, + { + "epoch": 2.84, + "learning_rate": 1.4961064447043696e-05, + "loss": 0.0483, + "step": 21664 + }, + { + "epoch": 2.84, + "learning_rate": 1.4960603979120072e-05, + "loss": 0.0501, + "step": 21665 + }, + { + "epoch": 2.84, + "learning_rate": 1.496014349724501e-05, + "loss": 0.052, + "step": 21666 + }, + { + "epoch": 2.84, + "learning_rate": 1.4959683001419812e-05, + "loss": 0.0565, + "step": 21667 + }, + { + "epoch": 2.84, + "learning_rate": 1.4959222491645764e-05, + "loss": 0.0469, + "step": 21668 + }, + { + "epoch": 2.84, + "learning_rate": 1.4958761967924171e-05, + "loss": 0.0674, + "step": 21669 + }, + { + "epoch": 2.84, + "learning_rate": 1.4958301430256321e-05, + "loss": 0.0673, + "step": 21670 + }, + { + "epoch": 2.84, + "learning_rate": 1.4957840878643518e-05, + "loss": 0.0487, + "step": 21671 + }, + { + "epoch": 2.84, + "learning_rate": 1.4957380313087047e-05, + "loss": 0.0568, + "step": 21672 + }, + { + "epoch": 2.84, + "learning_rate": 1.495691973358821e-05, + "loss": 0.0575, + "step": 21673 + }, + { + "epoch": 2.84, + "learning_rate": 1.4956459140148295e-05, + "loss": 0.057, + "step": 21674 + }, + { + "epoch": 2.84, + "learning_rate": 1.495599853276861e-05, + "loss": 0.0577, + "step": 21675 + }, + { + "epoch": 2.84, + "learning_rate": 1.4955537911450437e-05, + "loss": 0.0533, + "step": 21676 + }, + { + "epoch": 2.84, + "learning_rate": 1.495507727619508e-05, + "loss": 0.0504, + "step": 21677 + }, + { + "epoch": 2.84, + "learning_rate": 1.4954616627003833e-05, + "loss": 0.0556, + "step": 21678 + }, + { + "epoch": 2.84, + "learning_rate": 1.495415596387799e-05, + "loss": 0.0628, + "step": 21679 + }, + { + "epoch": 2.84, + "learning_rate": 1.4953695286818848e-05, + "loss": 0.0601, + "step": 21680 + }, + { + "epoch": 2.84, + "learning_rate": 1.4953234595827699e-05, + "loss": 0.0602, + "step": 21681 + }, + { + "epoch": 2.84, + "learning_rate": 1.4952773890905845e-05, + "loss": 0.0443, + "step": 21682 + }, + { + "epoch": 2.84, + "learning_rate": 1.4952313172054575e-05, + "loss": 0.0548, + "step": 21683 + }, + { + "epoch": 2.84, + "learning_rate": 1.495185243927519e-05, + "loss": 0.0584, + "step": 21684 + }, + { + "epoch": 2.84, + "learning_rate": 1.4951391692568982e-05, + "loss": 0.0549, + "step": 21685 + }, + { + "epoch": 2.84, + "learning_rate": 1.4950930931937251e-05, + "loss": 0.0467, + "step": 21686 + }, + { + "epoch": 2.84, + "learning_rate": 1.4950470157381288e-05, + "loss": 0.0541, + "step": 21687 + }, + { + "epoch": 2.84, + "learning_rate": 1.4950009368902393e-05, + "loss": 0.0618, + "step": 21688 + }, + { + "epoch": 2.84, + "learning_rate": 1.4949548566501859e-05, + "loss": 0.0606, + "step": 21689 + }, + { + "epoch": 2.84, + "learning_rate": 1.4949087750180984e-05, + "loss": 0.0562, + "step": 21690 + }, + { + "epoch": 2.84, + "learning_rate": 1.4948626919941063e-05, + "loss": 0.0483, + "step": 21691 + }, + { + "epoch": 2.84, + "learning_rate": 1.4948166075783393e-05, + "loss": 0.0433, + "step": 21692 + }, + { + "epoch": 2.84, + "learning_rate": 1.4947705217709268e-05, + "loss": 0.0534, + "step": 21693 + }, + { + "epoch": 2.84, + "learning_rate": 1.4947244345719987e-05, + "loss": 0.0629, + "step": 21694 + }, + { + "epoch": 2.84, + "learning_rate": 1.4946783459816842e-05, + "loss": 0.0535, + "step": 21695 + }, + { + "epoch": 2.84, + "learning_rate": 1.4946322560001137e-05, + "loss": 0.0422, + "step": 21696 + }, + { + "epoch": 2.84, + "learning_rate": 1.4945861646274159e-05, + "loss": 0.0571, + "step": 21697 + }, + { + "epoch": 2.84, + "learning_rate": 1.494540071863721e-05, + "loss": 0.0537, + "step": 21698 + }, + { + "epoch": 2.84, + "learning_rate": 1.4944939777091585e-05, + "loss": 0.0451, + "step": 21699 + }, + { + "epoch": 2.84, + "learning_rate": 1.4944478821638576e-05, + "loss": 0.0455, + "step": 21700 + }, + { + "epoch": 2.84, + "learning_rate": 1.4944017852279489e-05, + "loss": 0.0556, + "step": 21701 + }, + { + "epoch": 2.84, + "learning_rate": 1.4943556869015614e-05, + "loss": 0.0488, + "step": 21702 + }, + { + "epoch": 2.84, + "learning_rate": 1.4943095871848249e-05, + "loss": 0.0571, + "step": 21703 + }, + { + "epoch": 2.84, + "learning_rate": 1.4942634860778689e-05, + "loss": 0.0626, + "step": 21704 + }, + { + "epoch": 2.84, + "learning_rate": 1.4942173835808231e-05, + "loss": 0.0634, + "step": 21705 + }, + { + "epoch": 2.84, + "learning_rate": 1.4941712796938173e-05, + "loss": 0.056, + "step": 21706 + }, + { + "epoch": 2.84, + "learning_rate": 1.4941251744169811e-05, + "loss": 0.0535, + "step": 21707 + }, + { + "epoch": 2.84, + "learning_rate": 1.4940790677504441e-05, + "loss": 0.0521, + "step": 21708 + }, + { + "epoch": 2.84, + "learning_rate": 1.4940329596943364e-05, + "loss": 0.0497, + "step": 21709 + }, + { + "epoch": 2.84, + "learning_rate": 1.4939868502487869e-05, + "loss": 0.0574, + "step": 21710 + }, + { + "epoch": 2.84, + "learning_rate": 1.4939407394139257e-05, + "loss": 0.0649, + "step": 21711 + }, + { + "epoch": 2.84, + "learning_rate": 1.4938946271898824e-05, + "loss": 0.0597, + "step": 21712 + }, + { + "epoch": 2.84, + "learning_rate": 1.493848513576787e-05, + "loss": 0.0625, + "step": 21713 + }, + { + "epoch": 2.84, + "learning_rate": 1.4938023985747689e-05, + "loss": 0.0579, + "step": 21714 + }, + { + "epoch": 2.84, + "learning_rate": 1.493756282183958e-05, + "loss": 0.0623, + "step": 21715 + }, + { + "epoch": 2.84, + "learning_rate": 1.4937101644044836e-05, + "loss": 0.0532, + "step": 21716 + }, + { + "epoch": 2.84, + "learning_rate": 1.4936640452364757e-05, + "loss": 0.059, + "step": 21717 + }, + { + "epoch": 2.84, + "learning_rate": 1.4936179246800639e-05, + "loss": 0.0571, + "step": 21718 + }, + { + "epoch": 2.85, + "learning_rate": 1.4935718027353783e-05, + "loss": 0.0552, + "step": 21719 + }, + { + "epoch": 2.85, + "learning_rate": 1.4935256794025482e-05, + "loss": 0.0522, + "step": 21720 + }, + { + "epoch": 2.85, + "learning_rate": 1.4934795546817031e-05, + "loss": 0.0557, + "step": 21721 + }, + { + "epoch": 2.85, + "learning_rate": 1.4934334285729732e-05, + "loss": 0.0519, + "step": 21722 + }, + { + "epoch": 2.85, + "learning_rate": 1.4933873010764881e-05, + "loss": 0.0558, + "step": 21723 + }, + { + "epoch": 2.85, + "learning_rate": 1.4933411721923774e-05, + "loss": 0.0558, + "step": 21724 + }, + { + "epoch": 2.85, + "learning_rate": 1.4932950419207708e-05, + "loss": 0.0553, + "step": 21725 + }, + { + "epoch": 2.85, + "learning_rate": 1.4932489102617987e-05, + "loss": 0.0558, + "step": 21726 + }, + { + "epoch": 2.85, + "learning_rate": 1.4932027772155896e-05, + "loss": 0.0611, + "step": 21727 + }, + { + "epoch": 2.85, + "learning_rate": 1.4931566427822743e-05, + "loss": 0.0417, + "step": 21728 + }, + { + "epoch": 2.85, + "learning_rate": 1.4931105069619822e-05, + "loss": 0.0599, + "step": 21729 + }, + { + "epoch": 2.85, + "learning_rate": 1.4930643697548432e-05, + "loss": 0.0579, + "step": 21730 + }, + { + "epoch": 2.85, + "learning_rate": 1.4930182311609866e-05, + "loss": 0.0549, + "step": 21731 + }, + { + "epoch": 2.85, + "learning_rate": 1.492972091180543e-05, + "loss": 0.0563, + "step": 21732 + }, + { + "epoch": 2.85, + "learning_rate": 1.4929259498136413e-05, + "loss": 0.0489, + "step": 21733 + }, + { + "epoch": 2.85, + "learning_rate": 1.4928798070604114e-05, + "loss": 0.0567, + "step": 21734 + }, + { + "epoch": 2.85, + "learning_rate": 1.4928336629209837e-05, + "loss": 0.0559, + "step": 21735 + }, + { + "epoch": 2.85, + "learning_rate": 1.4927875173954873e-05, + "loss": 0.0538, + "step": 21736 + }, + { + "epoch": 2.85, + "learning_rate": 1.4927413704840523e-05, + "loss": 0.0561, + "step": 21737 + }, + { + "epoch": 2.85, + "learning_rate": 1.4926952221868087e-05, + "loss": 0.0534, + "step": 21738 + }, + { + "epoch": 2.85, + "learning_rate": 1.4926490725038855e-05, + "loss": 0.0502, + "step": 21739 + }, + { + "epoch": 2.85, + "learning_rate": 1.4926029214354134e-05, + "loss": 0.0479, + "step": 21740 + }, + { + "epoch": 2.85, + "learning_rate": 1.4925567689815217e-05, + "loss": 0.0522, + "step": 21741 + }, + { + "epoch": 2.85, + "learning_rate": 1.4925106151423403e-05, + "loss": 0.055, + "step": 21742 + }, + { + "epoch": 2.85, + "learning_rate": 1.4924644599179992e-05, + "loss": 0.0547, + "step": 21743 + }, + { + "epoch": 2.85, + "learning_rate": 1.492418303308628e-05, + "loss": 0.0661, + "step": 21744 + }, + { + "epoch": 2.85, + "learning_rate": 1.4923721453143563e-05, + "loss": 0.0486, + "step": 21745 + }, + { + "epoch": 2.85, + "learning_rate": 1.492325985935314e-05, + "loss": 0.0642, + "step": 21746 + }, + { + "epoch": 2.85, + "learning_rate": 1.4922798251716316e-05, + "loss": 0.0577, + "step": 21747 + }, + { + "epoch": 2.85, + "learning_rate": 1.4922336630234381e-05, + "loss": 0.0523, + "step": 21748 + }, + { + "epoch": 2.85, + "learning_rate": 1.4921874994908641e-05, + "loss": 0.0567, + "step": 21749 + }, + { + "epoch": 2.85, + "learning_rate": 1.4921413345740383e-05, + "loss": 0.0502, + "step": 21750 + }, + { + "epoch": 2.85, + "learning_rate": 1.4920951682730915e-05, + "loss": 0.0475, + "step": 21751 + }, + { + "epoch": 2.85, + "learning_rate": 1.4920490005881534e-05, + "loss": 0.0552, + "step": 21752 + }, + { + "epoch": 2.85, + "learning_rate": 1.4920028315193536e-05, + "loss": 0.0616, + "step": 21753 + }, + { + "epoch": 2.85, + "learning_rate": 1.491956661066822e-05, + "loss": 0.0554, + "step": 21754 + }, + { + "epoch": 2.85, + "learning_rate": 1.4919104892306884e-05, + "loss": 0.0398, + "step": 21755 + }, + { + "epoch": 2.85, + "learning_rate": 1.4918643160110825e-05, + "loss": 0.0572, + "step": 21756 + }, + { + "epoch": 2.85, + "learning_rate": 1.491818141408135e-05, + "loss": 0.0529, + "step": 21757 + }, + { + "epoch": 2.85, + "learning_rate": 1.4917719654219749e-05, + "loss": 0.0539, + "step": 21758 + }, + { + "epoch": 2.85, + "learning_rate": 1.4917257880527323e-05, + "loss": 0.0587, + "step": 21759 + }, + { + "epoch": 2.85, + "learning_rate": 1.4916796093005373e-05, + "loss": 0.0573, + "step": 21760 + }, + { + "epoch": 2.85, + "learning_rate": 1.4916334291655192e-05, + "loss": 0.0543, + "step": 21761 + }, + { + "epoch": 2.85, + "learning_rate": 1.4915872476478083e-05, + "loss": 0.0613, + "step": 21762 + }, + { + "epoch": 2.85, + "learning_rate": 1.4915410647475348e-05, + "loss": 0.0618, + "step": 21763 + }, + { + "epoch": 2.85, + "learning_rate": 1.4914948804648282e-05, + "loss": 0.0579, + "step": 21764 + }, + { + "epoch": 2.85, + "learning_rate": 1.4914486947998181e-05, + "loss": 0.0572, + "step": 21765 + }, + { + "epoch": 2.85, + "learning_rate": 1.4914025077526349e-05, + "loss": 0.0465, + "step": 21766 + }, + { + "epoch": 2.85, + "learning_rate": 1.4913563193234081e-05, + "loss": 0.0596, + "step": 21767 + }, + { + "epoch": 2.85, + "learning_rate": 1.491310129512268e-05, + "loss": 0.0546, + "step": 21768 + }, + { + "epoch": 2.85, + "learning_rate": 1.4912639383193445e-05, + "loss": 0.0619, + "step": 21769 + }, + { + "epoch": 2.85, + "learning_rate": 1.4912177457447671e-05, + "loss": 0.0532, + "step": 21770 + }, + { + "epoch": 2.85, + "learning_rate": 1.4911715517886662e-05, + "loss": 0.0587, + "step": 21771 + }, + { + "epoch": 2.85, + "learning_rate": 1.4911253564511712e-05, + "loss": 0.0508, + "step": 21772 + }, + { + "epoch": 2.85, + "learning_rate": 1.4910791597324124e-05, + "loss": 0.0559, + "step": 21773 + }, + { + "epoch": 2.85, + "learning_rate": 1.4910329616325197e-05, + "loss": 0.0567, + "step": 21774 + }, + { + "epoch": 2.85, + "learning_rate": 1.4909867621516228e-05, + "loss": 0.0633, + "step": 21775 + }, + { + "epoch": 2.85, + "learning_rate": 1.4909405612898518e-05, + "loss": 0.0561, + "step": 21776 + }, + { + "epoch": 2.85, + "learning_rate": 1.4908943590473364e-05, + "loss": 0.0601, + "step": 21777 + }, + { + "epoch": 2.85, + "learning_rate": 1.4908481554242071e-05, + "loss": 0.0575, + "step": 21778 + }, + { + "epoch": 2.85, + "learning_rate": 1.4908019504205936e-05, + "loss": 0.0506, + "step": 21779 + }, + { + "epoch": 2.85, + "learning_rate": 1.4907557440366256e-05, + "loss": 0.0665, + "step": 21780 + }, + { + "epoch": 2.85, + "learning_rate": 1.490709536272433e-05, + "loss": 0.0461, + "step": 21781 + }, + { + "epoch": 2.85, + "learning_rate": 1.490663327128146e-05, + "loss": 0.0568, + "step": 21782 + }, + { + "epoch": 2.85, + "learning_rate": 1.4906171166038944e-05, + "loss": 0.0468, + "step": 21783 + }, + { + "epoch": 2.85, + "learning_rate": 1.4905709046998089e-05, + "loss": 0.0497, + "step": 21784 + }, + { + "epoch": 2.85, + "learning_rate": 1.4905246914160185e-05, + "loss": 0.0615, + "step": 21785 + }, + { + "epoch": 2.85, + "learning_rate": 1.4904784767526535e-05, + "loss": 0.0483, + "step": 21786 + }, + { + "epoch": 2.85, + "learning_rate": 1.4904322607098438e-05, + "loss": 0.0536, + "step": 21787 + }, + { + "epoch": 2.85, + "learning_rate": 1.4903860432877198e-05, + "loss": 0.0651, + "step": 21788 + }, + { + "epoch": 2.85, + "learning_rate": 1.4903398244864106e-05, + "loss": 0.0625, + "step": 21789 + }, + { + "epoch": 2.85, + "learning_rate": 1.4902936043060473e-05, + "loss": 0.0592, + "step": 21790 + }, + { + "epoch": 2.85, + "learning_rate": 1.4902473827467592e-05, + "loss": 0.0525, + "step": 21791 + }, + { + "epoch": 2.85, + "learning_rate": 1.4902011598086763e-05, + "loss": 0.0502, + "step": 21792 + }, + { + "epoch": 2.85, + "learning_rate": 1.4901549354919289e-05, + "loss": 0.0549, + "step": 21793 + }, + { + "epoch": 2.85, + "learning_rate": 1.4901087097966467e-05, + "loss": 0.049, + "step": 21794 + }, + { + "epoch": 2.85, + "learning_rate": 1.49006248272296e-05, + "loss": 0.0571, + "step": 21795 + }, + { + "epoch": 2.86, + "learning_rate": 1.4900162542709985e-05, + "loss": 0.0535, + "step": 21796 + }, + { + "epoch": 2.86, + "learning_rate": 1.4899700244408928e-05, + "loss": 0.06, + "step": 21797 + }, + { + "epoch": 2.86, + "learning_rate": 1.489923793232772e-05, + "loss": 0.0477, + "step": 21798 + }, + { + "epoch": 2.86, + "learning_rate": 1.4898775606467667e-05, + "loss": 0.061, + "step": 21799 + }, + { + "epoch": 2.86, + "learning_rate": 1.4898313266830071e-05, + "loss": 0.0614, + "step": 21800 + }, + { + "epoch": 2.86, + "learning_rate": 1.4897850913416227e-05, + "loss": 0.052, + "step": 21801 + }, + { + "epoch": 2.86, + "learning_rate": 1.489738854622744e-05, + "loss": 0.0546, + "step": 21802 + }, + { + "epoch": 2.86, + "learning_rate": 1.4896926165265009e-05, + "loss": 0.0579, + "step": 21803 + }, + { + "epoch": 2.86, + "learning_rate": 1.4896463770530233e-05, + "loss": 0.0584, + "step": 21804 + }, + { + "epoch": 2.86, + "learning_rate": 1.4896001362024413e-05, + "loss": 0.054, + "step": 21805 + }, + { + "epoch": 2.86, + "learning_rate": 1.489553893974885e-05, + "loss": 0.063, + "step": 21806 + }, + { + "epoch": 2.86, + "learning_rate": 1.4895076503704844e-05, + "loss": 0.0568, + "step": 21807 + }, + { + "epoch": 2.86, + "learning_rate": 1.4894614053893697e-05, + "loss": 0.0541, + "step": 21808 + }, + { + "epoch": 2.86, + "learning_rate": 1.4894151590316709e-05, + "loss": 0.0603, + "step": 21809 + }, + { + "epoch": 2.86, + "learning_rate": 1.4893689112975179e-05, + "loss": 0.0521, + "step": 21810 + }, + { + "epoch": 2.86, + "learning_rate": 1.4893226621870408e-05, + "loss": 0.0568, + "step": 21811 + }, + { + "epoch": 2.86, + "learning_rate": 1.4892764117003703e-05, + "loss": 0.0519, + "step": 21812 + }, + { + "epoch": 2.86, + "learning_rate": 1.4892301598376353e-05, + "loss": 0.0604, + "step": 21813 + }, + { + "epoch": 2.86, + "learning_rate": 1.489183906598967e-05, + "loss": 0.0538, + "step": 21814 + }, + { + "epoch": 2.86, + "learning_rate": 1.4891376519844944e-05, + "loss": 0.047, + "step": 21815 + }, + { + "epoch": 2.86, + "learning_rate": 1.489091395994349e-05, + "loss": 0.055, + "step": 21816 + }, + { + "epoch": 2.86, + "learning_rate": 1.4890451386286596e-05, + "loss": 0.0448, + "step": 21817 + }, + { + "epoch": 2.86, + "learning_rate": 1.488998879887557e-05, + "loss": 0.051, + "step": 21818 + }, + { + "epoch": 2.86, + "learning_rate": 1.488952619771171e-05, + "loss": 0.055, + "step": 21819 + }, + { + "epoch": 2.86, + "learning_rate": 1.488906358279632e-05, + "loss": 0.0598, + "step": 21820 + }, + { + "epoch": 2.86, + "learning_rate": 1.4888600954130694e-05, + "loss": 0.0586, + "step": 21821 + }, + { + "epoch": 2.86, + "learning_rate": 1.4888138311716142e-05, + "loss": 0.0592, + "step": 21822 + }, + { + "epoch": 2.86, + "learning_rate": 1.488767565555396e-05, + "loss": 0.0448, + "step": 21823 + }, + { + "epoch": 2.86, + "learning_rate": 1.4887212985645452e-05, + "loss": 0.0509, + "step": 21824 + }, + { + "epoch": 2.86, + "learning_rate": 1.4886750301991914e-05, + "loss": 0.0481, + "step": 21825 + }, + { + "epoch": 2.86, + "learning_rate": 1.4886287604594657e-05, + "loss": 0.052, + "step": 21826 + }, + { + "epoch": 2.86, + "learning_rate": 1.488582489345497e-05, + "loss": 0.051, + "step": 21827 + }, + { + "epoch": 2.86, + "learning_rate": 1.4885362168574165e-05, + "loss": 0.0492, + "step": 21828 + }, + { + "epoch": 2.86, + "learning_rate": 1.4884899429953539e-05, + "loss": 0.0594, + "step": 21829 + }, + { + "epoch": 2.86, + "learning_rate": 1.4884436677594392e-05, + "loss": 0.048, + "step": 21830 + }, + { + "epoch": 2.86, + "learning_rate": 1.4883973911498026e-05, + "loss": 0.0508, + "step": 21831 + }, + { + "epoch": 2.86, + "learning_rate": 1.4883511131665741e-05, + "loss": 0.049, + "step": 21832 + }, + { + "epoch": 2.86, + "learning_rate": 1.4883048338098846e-05, + "loss": 0.0634, + "step": 21833 + }, + { + "epoch": 2.86, + "learning_rate": 1.4882585530798637e-05, + "loss": 0.0597, + "step": 21834 + }, + { + "epoch": 2.86, + "learning_rate": 1.4882122709766418e-05, + "loss": 0.0552, + "step": 21835 + }, + { + "epoch": 2.86, + "learning_rate": 1.4881659875003484e-05, + "loss": 0.0569, + "step": 21836 + }, + { + "epoch": 2.86, + "learning_rate": 1.4881197026511145e-05, + "loss": 0.0503, + "step": 21837 + }, + { + "epoch": 2.86, + "learning_rate": 1.4880734164290696e-05, + "loss": 0.0558, + "step": 21838 + }, + { + "epoch": 2.86, + "learning_rate": 1.4880271288343447e-05, + "loss": 0.0562, + "step": 21839 + }, + { + "epoch": 2.86, + "learning_rate": 1.4879808398670691e-05, + "loss": 0.0592, + "step": 21840 + }, + { + "epoch": 2.86, + "learning_rate": 1.4879345495273735e-05, + "loss": 0.0485, + "step": 21841 + }, + { + "epoch": 2.86, + "learning_rate": 1.487888257815388e-05, + "loss": 0.0549, + "step": 21842 + }, + { + "epoch": 2.86, + "learning_rate": 1.4878419647312426e-05, + "loss": 0.061, + "step": 21843 + }, + { + "epoch": 2.86, + "learning_rate": 1.4877956702750678e-05, + "loss": 0.0512, + "step": 21844 + }, + { + "epoch": 2.86, + "learning_rate": 1.4877493744469938e-05, + "loss": 0.0473, + "step": 21845 + }, + { + "epoch": 2.86, + "learning_rate": 1.4877030772471504e-05, + "loss": 0.0529, + "step": 21846 + }, + { + "epoch": 2.86, + "learning_rate": 1.4876567786756679e-05, + "loss": 0.0488, + "step": 21847 + }, + { + "epoch": 2.86, + "learning_rate": 1.4876104787326768e-05, + "loss": 0.0573, + "step": 21848 + }, + { + "epoch": 2.86, + "learning_rate": 1.4875641774183076e-05, + "loss": 0.0506, + "step": 21849 + }, + { + "epoch": 2.86, + "learning_rate": 1.4875178747326898e-05, + "loss": 0.0487, + "step": 21850 + }, + { + "epoch": 2.86, + "learning_rate": 1.4874715706759539e-05, + "loss": 0.0619, + "step": 21851 + }, + { + "epoch": 2.86, + "learning_rate": 1.4874252652482305e-05, + "loss": 0.0501, + "step": 21852 + }, + { + "epoch": 2.86, + "learning_rate": 1.487378958449649e-05, + "loss": 0.0537, + "step": 21853 + }, + { + "epoch": 2.86, + "learning_rate": 1.4873326502803402e-05, + "loss": 0.0523, + "step": 21854 + }, + { + "epoch": 2.86, + "learning_rate": 1.4872863407404345e-05, + "loss": 0.0601, + "step": 21855 + }, + { + "epoch": 2.86, + "learning_rate": 1.487240029830062e-05, + "loss": 0.0556, + "step": 21856 + }, + { + "epoch": 2.86, + "learning_rate": 1.4871937175493528e-05, + "loss": 0.06, + "step": 21857 + }, + { + "epoch": 2.86, + "learning_rate": 1.4871474038984373e-05, + "loss": 0.0599, + "step": 21858 + }, + { + "epoch": 2.86, + "learning_rate": 1.487101088877445e-05, + "loss": 0.0611, + "step": 21859 + }, + { + "epoch": 2.86, + "learning_rate": 1.4870547724865075e-05, + "loss": 0.0538, + "step": 21860 + }, + { + "epoch": 2.86, + "learning_rate": 1.4870084547257541e-05, + "loss": 0.0531, + "step": 21861 + }, + { + "epoch": 2.86, + "learning_rate": 1.4869621355953156e-05, + "loss": 0.0662, + "step": 21862 + }, + { + "epoch": 2.86, + "learning_rate": 1.486915815095322e-05, + "loss": 0.0485, + "step": 21863 + }, + { + "epoch": 2.86, + "learning_rate": 1.4868694932259037e-05, + "loss": 0.058, + "step": 21864 + }, + { + "epoch": 2.86, + "learning_rate": 1.4868231699871905e-05, + "loss": 0.0632, + "step": 21865 + }, + { + "epoch": 2.86, + "learning_rate": 1.4867768453793132e-05, + "loss": 0.0478, + "step": 21866 + }, + { + "epoch": 2.86, + "learning_rate": 1.486730519402402e-05, + "loss": 0.0552, + "step": 21867 + }, + { + "epoch": 2.86, + "learning_rate": 1.4866841920565872e-05, + "loss": 0.0549, + "step": 21868 + }, + { + "epoch": 2.86, + "learning_rate": 1.4866378633419988e-05, + "loss": 0.0535, + "step": 21869 + }, + { + "epoch": 2.86, + "learning_rate": 1.4865915332587675e-05, + "loss": 0.0589, + "step": 21870 + }, + { + "epoch": 2.86, + "learning_rate": 1.4865452018070232e-05, + "loss": 0.0604, + "step": 21871 + }, + { + "epoch": 2.87, + "learning_rate": 1.4864988689868967e-05, + "loss": 0.0608, + "step": 21872 + }, + { + "epoch": 2.87, + "learning_rate": 1.4864525347985182e-05, + "loss": 0.0514, + "step": 21873 + }, + { + "epoch": 2.87, + "learning_rate": 1.4864061992420173e-05, + "loss": 0.0576, + "step": 21874 + }, + { + "epoch": 2.87, + "learning_rate": 1.4863598623175255e-05, + "loss": 0.0513, + "step": 21875 + }, + { + "epoch": 2.87, + "learning_rate": 1.4863135240251718e-05, + "loss": 0.0592, + "step": 21876 + }, + { + "epoch": 2.87, + "learning_rate": 1.4862671843650878e-05, + "loss": 0.0612, + "step": 21877 + }, + { + "epoch": 2.87, + "learning_rate": 1.486220843337403e-05, + "loss": 0.0526, + "step": 21878 + }, + { + "epoch": 2.87, + "learning_rate": 1.4861745009422481e-05, + "loss": 0.0421, + "step": 21879 + }, + { + "epoch": 2.87, + "learning_rate": 1.4861281571797532e-05, + "loss": 0.0522, + "step": 21880 + }, + { + "epoch": 2.87, + "learning_rate": 1.4860818120500487e-05, + "loss": 0.0606, + "step": 21881 + }, + { + "epoch": 2.87, + "learning_rate": 1.4860354655532649e-05, + "loss": 0.0572, + "step": 21882 + }, + { + "epoch": 2.87, + "learning_rate": 1.4859891176895325e-05, + "loss": 0.0609, + "step": 21883 + }, + { + "epoch": 2.87, + "learning_rate": 1.4859427684589813e-05, + "loss": 0.0454, + "step": 21884 + }, + { + "epoch": 2.87, + "learning_rate": 1.4858964178617423e-05, + "loss": 0.0525, + "step": 21885 + }, + { + "epoch": 2.87, + "learning_rate": 1.4858500658979451e-05, + "loss": 0.0546, + "step": 21886 + }, + { + "epoch": 2.87, + "learning_rate": 1.4858037125677208e-05, + "loss": 0.0574, + "step": 21887 + }, + { + "epoch": 2.87, + "learning_rate": 1.4857573578711993e-05, + "loss": 0.0543, + "step": 21888 + }, + { + "epoch": 2.87, + "learning_rate": 1.485711001808511e-05, + "loss": 0.0576, + "step": 21889 + }, + { + "epoch": 2.87, + "learning_rate": 1.4856646443797865e-05, + "loss": 0.0555, + "step": 21890 + }, + { + "epoch": 2.87, + "learning_rate": 1.4856182855851559e-05, + "loss": 0.057, + "step": 21891 + }, + { + "epoch": 2.87, + "learning_rate": 1.4855719254247498e-05, + "loss": 0.0479, + "step": 21892 + }, + { + "epoch": 2.87, + "learning_rate": 1.4855255638986985e-05, + "loss": 0.0547, + "step": 21893 + }, + { + "epoch": 2.87, + "learning_rate": 1.4854792010071325e-05, + "loss": 0.0603, + "step": 21894 + }, + { + "epoch": 2.87, + "learning_rate": 1.485432836750182e-05, + "loss": 0.0549, + "step": 21895 + }, + { + "epoch": 2.87, + "learning_rate": 1.4853864711279778e-05, + "loss": 0.0507, + "step": 21896 + }, + { + "epoch": 2.87, + "learning_rate": 1.4853401041406494e-05, + "loss": 0.0511, + "step": 21897 + }, + { + "epoch": 2.87, + "learning_rate": 1.4852937357883286e-05, + "loss": 0.0486, + "step": 21898 + }, + { + "epoch": 2.87, + "learning_rate": 1.4852473660711444e-05, + "loss": 0.0555, + "step": 21899 + }, + { + "epoch": 2.87, + "learning_rate": 1.485200994989228e-05, + "loss": 0.0529, + "step": 21900 + }, + { + "epoch": 2.87, + "learning_rate": 1.4851546225427096e-05, + "loss": 0.0475, + "step": 21901 + }, + { + "epoch": 2.87, + "learning_rate": 1.4851082487317199e-05, + "loss": 0.0559, + "step": 21902 + }, + { + "epoch": 2.87, + "learning_rate": 1.4850618735563886e-05, + "loss": 0.0553, + "step": 21903 + }, + { + "epoch": 2.87, + "learning_rate": 1.4850154970168471e-05, + "loss": 0.0537, + "step": 21904 + }, + { + "epoch": 2.87, + "learning_rate": 1.4849691191132251e-05, + "loss": 0.0549, + "step": 21905 + }, + { + "epoch": 2.87, + "learning_rate": 1.4849227398456534e-05, + "loss": 0.0436, + "step": 21906 + }, + { + "epoch": 2.87, + "learning_rate": 1.4848763592142624e-05, + "loss": 0.0493, + "step": 21907 + }, + { + "epoch": 2.87, + "learning_rate": 1.484829977219182e-05, + "loss": 0.0511, + "step": 21908 + }, + { + "epoch": 2.87, + "learning_rate": 1.4847835938605433e-05, + "loss": 0.0574, + "step": 21909 + }, + { + "epoch": 2.87, + "learning_rate": 1.4847372091384768e-05, + "loss": 0.048, + "step": 21910 + }, + { + "epoch": 2.87, + "learning_rate": 1.4846908230531125e-05, + "loss": 0.0646, + "step": 21911 + }, + { + "epoch": 2.87, + "learning_rate": 1.4846444356045813e-05, + "loss": 0.0572, + "step": 21912 + }, + { + "epoch": 2.87, + "learning_rate": 1.4845980467930133e-05, + "loss": 0.0616, + "step": 21913 + }, + { + "epoch": 2.87, + "learning_rate": 1.484551656618539e-05, + "loss": 0.0501, + "step": 21914 + }, + { + "epoch": 2.87, + "learning_rate": 1.484505265081289e-05, + "loss": 0.0555, + "step": 21915 + }, + { + "epoch": 2.87, + "learning_rate": 1.4844588721813938e-05, + "loss": 0.0531, + "step": 21916 + }, + { + "epoch": 2.87, + "learning_rate": 1.484412477918984e-05, + "loss": 0.0618, + "step": 21917 + }, + { + "epoch": 2.87, + "learning_rate": 1.4843660822941896e-05, + "loss": 0.0567, + "step": 21918 + }, + { + "epoch": 2.87, + "learning_rate": 1.4843196853071416e-05, + "loss": 0.0518, + "step": 21919 + }, + { + "epoch": 2.87, + "learning_rate": 1.4842732869579699e-05, + "loss": 0.0546, + "step": 21920 + }, + { + "epoch": 2.87, + "learning_rate": 1.484226887246806e-05, + "loss": 0.0438, + "step": 21921 + }, + { + "epoch": 2.87, + "learning_rate": 1.4841804861737792e-05, + "loss": 0.0598, + "step": 21922 + }, + { + "epoch": 2.87, + "learning_rate": 1.484134083739021e-05, + "loss": 0.0523, + "step": 21923 + }, + { + "epoch": 2.87, + "learning_rate": 1.484087679942661e-05, + "loss": 0.0506, + "step": 21924 + }, + { + "epoch": 2.87, + "learning_rate": 1.4840412747848305e-05, + "loss": 0.0601, + "step": 21925 + }, + { + "epoch": 2.87, + "learning_rate": 1.4839948682656595e-05, + "loss": 0.0503, + "step": 21926 + }, + { + "epoch": 2.87, + "learning_rate": 1.4839484603852792e-05, + "loss": 0.0584, + "step": 21927 + }, + { + "epoch": 2.87, + "learning_rate": 1.4839020511438191e-05, + "loss": 0.0533, + "step": 21928 + }, + { + "epoch": 2.87, + "learning_rate": 1.4838556405414104e-05, + "loss": 0.0527, + "step": 21929 + }, + { + "epoch": 2.87, + "learning_rate": 1.4838092285781833e-05, + "loss": 0.0571, + "step": 21930 + }, + { + "epoch": 2.87, + "learning_rate": 1.4837628152542689e-05, + "loss": 0.0494, + "step": 21931 + }, + { + "epoch": 2.87, + "learning_rate": 1.4837164005697971e-05, + "loss": 0.0562, + "step": 21932 + }, + { + "epoch": 2.87, + "learning_rate": 1.4836699845248984e-05, + "loss": 0.0467, + "step": 21933 + }, + { + "epoch": 2.87, + "learning_rate": 1.4836235671197043e-05, + "loss": 0.0576, + "step": 21934 + }, + { + "epoch": 2.87, + "learning_rate": 1.4835771483543444e-05, + "loss": 0.0421, + "step": 21935 + }, + { + "epoch": 2.87, + "learning_rate": 1.4835307282289492e-05, + "loss": 0.061, + "step": 21936 + }, + { + "epoch": 2.87, + "learning_rate": 1.48348430674365e-05, + "loss": 0.0494, + "step": 21937 + }, + { + "epoch": 2.87, + "learning_rate": 1.4834378838985768e-05, + "loss": 0.0519, + "step": 21938 + }, + { + "epoch": 2.87, + "learning_rate": 1.4833914596938601e-05, + "loss": 0.0541, + "step": 21939 + }, + { + "epoch": 2.87, + "learning_rate": 1.483345034129631e-05, + "loss": 0.0543, + "step": 21940 + }, + { + "epoch": 2.87, + "learning_rate": 1.4832986072060193e-05, + "loss": 0.0465, + "step": 21941 + }, + { + "epoch": 2.87, + "learning_rate": 1.4832521789231564e-05, + "loss": 0.061, + "step": 21942 + }, + { + "epoch": 2.87, + "learning_rate": 1.4832057492811723e-05, + "loss": 0.0639, + "step": 21943 + }, + { + "epoch": 2.87, + "learning_rate": 1.483159318280198e-05, + "loss": 0.0585, + "step": 21944 + }, + { + "epoch": 2.87, + "learning_rate": 1.4831128859203634e-05, + "loss": 0.0532, + "step": 21945 + }, + { + "epoch": 2.87, + "learning_rate": 1.4830664522017996e-05, + "loss": 0.0547, + "step": 21946 + }, + { + "epoch": 2.87, + "learning_rate": 1.4830200171246375e-05, + "loss": 0.0515, + "step": 21947 + }, + { + "epoch": 2.88, + "learning_rate": 1.4829735806890071e-05, + "loss": 0.0605, + "step": 21948 + }, + { + "epoch": 2.88, + "learning_rate": 1.4829271428950391e-05, + "loss": 0.0596, + "step": 21949 + }, + { + "epoch": 2.88, + "learning_rate": 1.4828807037428648e-05, + "loss": 0.0535, + "step": 21950 + }, + { + "epoch": 2.88, + "learning_rate": 1.4828342632326136e-05, + "loss": 0.0511, + "step": 21951 + }, + { + "epoch": 2.88, + "learning_rate": 1.482787821364417e-05, + "loss": 0.06, + "step": 21952 + }, + { + "epoch": 2.88, + "learning_rate": 1.4827413781384052e-05, + "loss": 0.0509, + "step": 21953 + }, + { + "epoch": 2.88, + "learning_rate": 1.4826949335547092e-05, + "loss": 0.0554, + "step": 21954 + }, + { + "epoch": 2.88, + "learning_rate": 1.4826484876134593e-05, + "loss": 0.0657, + "step": 21955 + }, + { + "epoch": 2.88, + "learning_rate": 1.4826020403147862e-05, + "loss": 0.0537, + "step": 21956 + }, + { + "epoch": 2.88, + "learning_rate": 1.4825555916588206e-05, + "loss": 0.0591, + "step": 21957 + }, + { + "epoch": 2.88, + "learning_rate": 1.4825091416456928e-05, + "loss": 0.0582, + "step": 21958 + }, + { + "epoch": 2.88, + "learning_rate": 1.4824626902755343e-05, + "loss": 0.0633, + "step": 21959 + }, + { + "epoch": 2.88, + "learning_rate": 1.482416237548475e-05, + "loss": 0.0535, + "step": 21960 + }, + { + "epoch": 2.88, + "learning_rate": 1.4823697834646456e-05, + "loss": 0.0524, + "step": 21961 + }, + { + "epoch": 2.88, + "learning_rate": 1.482323328024177e-05, + "loss": 0.0562, + "step": 21962 + }, + { + "epoch": 2.88, + "learning_rate": 1.4822768712271995e-05, + "loss": 0.0553, + "step": 21963 + }, + { + "epoch": 2.88, + "learning_rate": 1.482230413073844e-05, + "loss": 0.0496, + "step": 21964 + }, + { + "epoch": 2.88, + "learning_rate": 1.4821839535642415e-05, + "loss": 0.055, + "step": 21965 + }, + { + "epoch": 2.88, + "learning_rate": 1.482137492698522e-05, + "loss": 0.0574, + "step": 21966 + }, + { + "epoch": 2.88, + "learning_rate": 1.4820910304768167e-05, + "loss": 0.0561, + "step": 21967 + }, + { + "epoch": 2.88, + "learning_rate": 1.4820445668992555e-05, + "loss": 0.0529, + "step": 21968 + }, + { + "epoch": 2.88, + "learning_rate": 1.4819981019659701e-05, + "loss": 0.0571, + "step": 21969 + }, + { + "epoch": 2.88, + "learning_rate": 1.4819516356770907e-05, + "loss": 0.0543, + "step": 21970 + }, + { + "epoch": 2.88, + "learning_rate": 1.4819051680327481e-05, + "loss": 0.054, + "step": 21971 + }, + { + "epoch": 2.88, + "learning_rate": 1.4818586990330726e-05, + "loss": 0.0606, + "step": 21972 + }, + { + "epoch": 2.88, + "learning_rate": 1.4818122286781952e-05, + "loss": 0.0563, + "step": 21973 + }, + { + "epoch": 2.88, + "learning_rate": 1.4817657569682467e-05, + "loss": 0.0591, + "step": 21974 + }, + { + "epoch": 2.88, + "learning_rate": 1.4817192839033576e-05, + "loss": 0.0535, + "step": 21975 + }, + { + "epoch": 2.88, + "learning_rate": 1.4816728094836585e-05, + "loss": 0.0515, + "step": 21976 + }, + { + "epoch": 2.88, + "learning_rate": 1.4816263337092804e-05, + "loss": 0.0535, + "step": 21977 + }, + { + "epoch": 2.88, + "learning_rate": 1.481579856580354e-05, + "loss": 0.0516, + "step": 21978 + }, + { + "epoch": 2.88, + "learning_rate": 1.4815333780970095e-05, + "loss": 0.0619, + "step": 21979 + }, + { + "epoch": 2.88, + "learning_rate": 1.4814868982593786e-05, + "loss": 0.0477, + "step": 21980 + }, + { + "epoch": 2.88, + "learning_rate": 1.4814404170675912e-05, + "loss": 0.0549, + "step": 21981 + }, + { + "epoch": 2.88, + "learning_rate": 1.4813939345217782e-05, + "loss": 0.0639, + "step": 21982 + }, + { + "epoch": 2.88, + "learning_rate": 1.4813474506220705e-05, + "loss": 0.0544, + "step": 21983 + }, + { + "epoch": 2.88, + "learning_rate": 1.4813009653685988e-05, + "loss": 0.0519, + "step": 21984 + }, + { + "epoch": 2.88, + "learning_rate": 1.4812544787614932e-05, + "loss": 0.0608, + "step": 21985 + }, + { + "epoch": 2.88, + "learning_rate": 1.4812079908008856e-05, + "loss": 0.0653, + "step": 21986 + }, + { + "epoch": 2.88, + "learning_rate": 1.481161501486906e-05, + "loss": 0.0514, + "step": 21987 + }, + { + "epoch": 2.88, + "learning_rate": 1.4811150108196854e-05, + "loss": 0.0515, + "step": 21988 + }, + { + "epoch": 2.88, + "learning_rate": 1.4810685187993543e-05, + "loss": 0.0443, + "step": 21989 + }, + { + "epoch": 2.88, + "learning_rate": 1.4810220254260436e-05, + "loss": 0.0573, + "step": 21990 + }, + { + "epoch": 2.88, + "learning_rate": 1.480975530699884e-05, + "loss": 0.0564, + "step": 21991 + }, + { + "epoch": 2.88, + "learning_rate": 1.4809290346210065e-05, + "loss": 0.0541, + "step": 21992 + }, + { + "epoch": 2.88, + "learning_rate": 1.4808825371895417e-05, + "loss": 0.0569, + "step": 21993 + }, + { + "epoch": 2.88, + "learning_rate": 1.4808360384056201e-05, + "loss": 0.0596, + "step": 21994 + }, + { + "epoch": 2.88, + "learning_rate": 1.4807895382693733e-05, + "loss": 0.0487, + "step": 21995 + }, + { + "epoch": 2.88, + "learning_rate": 1.4807430367809312e-05, + "loss": 0.056, + "step": 21996 + }, + { + "epoch": 2.88, + "learning_rate": 1.4806965339404249e-05, + "loss": 0.053, + "step": 21997 + }, + { + "epoch": 2.88, + "learning_rate": 1.4806500297479851e-05, + "loss": 0.0524, + "step": 21998 + }, + { + "epoch": 2.88, + "learning_rate": 1.480603524203743e-05, + "loss": 0.0545, + "step": 21999 + }, + { + "epoch": 2.88, + "learning_rate": 1.4805570173078288e-05, + "loss": 0.0628, + "step": 22000 + }, + { + "epoch": 2.88, + "learning_rate": 1.4805105090603735e-05, + "loss": 0.048, + "step": 22001 + }, + { + "epoch": 2.88, + "learning_rate": 1.480463999461508e-05, + "loss": 0.0503, + "step": 22002 + }, + { + "epoch": 2.88, + "learning_rate": 1.4804174885113633e-05, + "loss": 0.0646, + "step": 22003 + }, + { + "epoch": 2.88, + "learning_rate": 1.48037097621007e-05, + "loss": 0.0579, + "step": 22004 + }, + { + "epoch": 2.88, + "learning_rate": 1.4803244625577589e-05, + "loss": 0.0584, + "step": 22005 + }, + { + "epoch": 2.88, + "learning_rate": 1.4802779475545607e-05, + "loss": 0.0581, + "step": 22006 + }, + { + "epoch": 2.88, + "learning_rate": 1.4802314312006064e-05, + "loss": 0.0551, + "step": 22007 + }, + { + "epoch": 2.88, + "learning_rate": 1.4801849134960267e-05, + "loss": 0.0579, + "step": 22008 + }, + { + "epoch": 2.88, + "learning_rate": 1.4801383944409526e-05, + "loss": 0.0547, + "step": 22009 + }, + { + "epoch": 2.88, + "learning_rate": 1.480091874035515e-05, + "loss": 0.0536, + "step": 22010 + }, + { + "epoch": 2.88, + "learning_rate": 1.4800453522798444e-05, + "loss": 0.0472, + "step": 22011 + }, + { + "epoch": 2.88, + "learning_rate": 1.4799988291740715e-05, + "loss": 0.0588, + "step": 22012 + }, + { + "epoch": 2.88, + "learning_rate": 1.479952304718328e-05, + "loss": 0.0486, + "step": 22013 + }, + { + "epoch": 2.88, + "learning_rate": 1.4799057789127438e-05, + "loss": 0.0508, + "step": 22014 + }, + { + "epoch": 2.88, + "learning_rate": 1.4798592517574504e-05, + "loss": 0.0559, + "step": 22015 + }, + { + "epoch": 2.88, + "learning_rate": 1.4798127232525783e-05, + "loss": 0.0529, + "step": 22016 + }, + { + "epoch": 2.88, + "learning_rate": 1.4797661933982586e-05, + "loss": 0.0604, + "step": 22017 + }, + { + "epoch": 2.88, + "learning_rate": 1.4797196621946217e-05, + "loss": 0.0577, + "step": 22018 + }, + { + "epoch": 2.88, + "learning_rate": 1.479673129641799e-05, + "loss": 0.0565, + "step": 22019 + }, + { + "epoch": 2.88, + "learning_rate": 1.4796265957399214e-05, + "loss": 0.0504, + "step": 22020 + }, + { + "epoch": 2.88, + "learning_rate": 1.4795800604891191e-05, + "loss": 0.0514, + "step": 22021 + }, + { + "epoch": 2.88, + "learning_rate": 1.4795335238895239e-05, + "loss": 0.0558, + "step": 22022 + }, + { + "epoch": 2.88, + "learning_rate": 1.4794869859412657e-05, + "loss": 0.0562, + "step": 22023 + }, + { + "epoch": 2.88, + "learning_rate": 1.4794404466444764e-05, + "loss": 0.0539, + "step": 22024 + }, + { + "epoch": 2.89, + "learning_rate": 1.479393905999286e-05, + "loss": 0.0535, + "step": 22025 + }, + { + "epoch": 2.89, + "learning_rate": 1.4793473640058257e-05, + "loss": 0.05, + "step": 22026 + }, + { + "epoch": 2.89, + "learning_rate": 1.4793008206642267e-05, + "loss": 0.0547, + "step": 22027 + }, + { + "epoch": 2.89, + "learning_rate": 1.4792542759746196e-05, + "loss": 0.0539, + "step": 22028 + }, + { + "epoch": 2.89, + "learning_rate": 1.4792077299371353e-05, + "loss": 0.0626, + "step": 22029 + }, + { + "epoch": 2.89, + "learning_rate": 1.479161182551905e-05, + "loss": 0.0578, + "step": 22030 + }, + { + "epoch": 2.89, + "learning_rate": 1.4791146338190592e-05, + "loss": 0.0576, + "step": 22031 + }, + { + "epoch": 2.89, + "learning_rate": 1.4790680837387293e-05, + "loss": 0.0534, + "step": 22032 + }, + { + "epoch": 2.89, + "learning_rate": 1.4790215323110456e-05, + "loss": 0.0623, + "step": 22033 + }, + { + "epoch": 2.89, + "learning_rate": 1.4789749795361392e-05, + "loss": 0.0563, + "step": 22034 + }, + { + "epoch": 2.89, + "learning_rate": 1.4789284254141415e-05, + "loss": 0.0512, + "step": 22035 + }, + { + "epoch": 2.89, + "learning_rate": 1.478881869945183e-05, + "loss": 0.0493, + "step": 22036 + }, + { + "epoch": 2.89, + "learning_rate": 1.4788353131293946e-05, + "loss": 0.0465, + "step": 22037 + }, + { + "epoch": 2.89, + "learning_rate": 1.4787887549669076e-05, + "loss": 0.0505, + "step": 22038 + }, + { + "epoch": 2.89, + "learning_rate": 1.4787421954578528e-05, + "loss": 0.0552, + "step": 22039 + }, + { + "epoch": 2.89, + "learning_rate": 1.4786956346023608e-05, + "loss": 0.0605, + "step": 22040 + }, + { + "epoch": 2.89, + "learning_rate": 1.4786490724005629e-05, + "loss": 0.0583, + "step": 22041 + }, + { + "epoch": 2.89, + "learning_rate": 1.4786025088525899e-05, + "loss": 0.0618, + "step": 22042 + }, + { + "epoch": 2.89, + "learning_rate": 1.4785559439585733e-05, + "loss": 0.0447, + "step": 22043 + }, + { + "epoch": 2.89, + "learning_rate": 1.478509377718643e-05, + "loss": 0.0616, + "step": 22044 + }, + { + "epoch": 2.89, + "learning_rate": 1.4784628101329309e-05, + "loss": 0.0587, + "step": 22045 + }, + { + "epoch": 2.89, + "learning_rate": 1.4784162412015675e-05, + "loss": 0.0419, + "step": 22046 + }, + { + "epoch": 2.89, + "learning_rate": 1.478369670924684e-05, + "loss": 0.0524, + "step": 22047 + }, + { + "epoch": 2.89, + "learning_rate": 1.4783230993024112e-05, + "loss": 0.0496, + "step": 22048 + }, + { + "epoch": 2.89, + "learning_rate": 1.4782765263348802e-05, + "loss": 0.0538, + "step": 22049 + }, + { + "epoch": 2.89, + "learning_rate": 1.4782299520222216e-05, + "loss": 0.0518, + "step": 22050 + }, + { + "epoch": 2.89, + "learning_rate": 1.4781833763645672e-05, + "loss": 0.0567, + "step": 22051 + }, + { + "epoch": 2.89, + "learning_rate": 1.4781367993620473e-05, + "loss": 0.0558, + "step": 22052 + }, + { + "epoch": 2.89, + "learning_rate": 1.4780902210147932e-05, + "loss": 0.0481, + "step": 22053 + }, + { + "epoch": 2.89, + "learning_rate": 1.4780436413229357e-05, + "loss": 0.0622, + "step": 22054 + }, + { + "epoch": 2.89, + "learning_rate": 1.4779970602866063e-05, + "loss": 0.0625, + "step": 22055 + }, + { + "epoch": 2.89, + "learning_rate": 1.4779504779059348e-05, + "loss": 0.0598, + "step": 22056 + }, + { + "epoch": 2.89, + "learning_rate": 1.4779038941810538e-05, + "loss": 0.046, + "step": 22057 + }, + { + "epoch": 2.89, + "learning_rate": 1.4778573091120932e-05, + "loss": 0.0505, + "step": 22058 + }, + { + "epoch": 2.89, + "learning_rate": 1.4778107226991844e-05, + "loss": 0.0467, + "step": 22059 + }, + { + "epoch": 2.89, + "learning_rate": 1.4777641349424587e-05, + "loss": 0.05, + "step": 22060 + }, + { + "epoch": 2.89, + "learning_rate": 1.4777175458420465e-05, + "loss": 0.0583, + "step": 22061 + }, + { + "epoch": 2.89, + "learning_rate": 1.4776709553980791e-05, + "loss": 0.059, + "step": 22062 + }, + { + "epoch": 2.89, + "learning_rate": 1.4776243636106875e-05, + "loss": 0.0545, + "step": 22063 + }, + { + "epoch": 2.89, + "learning_rate": 1.4775777704800032e-05, + "loss": 0.062, + "step": 22064 + }, + { + "epoch": 2.89, + "learning_rate": 1.4775311760061567e-05, + "loss": 0.0521, + "step": 22065 + }, + { + "epoch": 2.89, + "learning_rate": 1.4774845801892792e-05, + "loss": 0.0584, + "step": 22066 + }, + { + "epoch": 2.89, + "learning_rate": 1.4774379830295013e-05, + "loss": 0.0539, + "step": 22067 + }, + { + "epoch": 2.89, + "learning_rate": 1.4773913845269552e-05, + "loss": 0.0614, + "step": 22068 + }, + { + "epoch": 2.89, + "learning_rate": 1.4773447846817708e-05, + "loss": 0.0585, + "step": 22069 + }, + { + "epoch": 2.89, + "learning_rate": 1.4772981834940798e-05, + "loss": 0.0545, + "step": 22070 + }, + { + "epoch": 2.89, + "learning_rate": 1.477251580964013e-05, + "loss": 0.0585, + "step": 22071 + }, + { + "epoch": 2.89, + "learning_rate": 1.4772049770917014e-05, + "loss": 0.0553, + "step": 22072 + }, + { + "epoch": 2.89, + "learning_rate": 1.4771583718772763e-05, + "loss": 0.0499, + "step": 22073 + }, + { + "epoch": 2.89, + "learning_rate": 1.4771117653208691e-05, + "loss": 0.0527, + "step": 22074 + }, + { + "epoch": 2.89, + "learning_rate": 1.47706515742261e-05, + "loss": 0.0681, + "step": 22075 + }, + { + "epoch": 2.89, + "learning_rate": 1.4770185481826308e-05, + "loss": 0.0459, + "step": 22076 + }, + { + "epoch": 2.89, + "learning_rate": 1.4769719376010621e-05, + "loss": 0.0618, + "step": 22077 + }, + { + "epoch": 2.89, + "learning_rate": 1.4769253256780353e-05, + "loss": 0.051, + "step": 22078 + }, + { + "epoch": 2.89, + "learning_rate": 1.4768787124136816e-05, + "loss": 0.0536, + "step": 22079 + }, + { + "epoch": 2.89, + "learning_rate": 1.4768320978081318e-05, + "loss": 0.0582, + "step": 22080 + }, + { + "epoch": 2.89, + "learning_rate": 1.476785481861517e-05, + "loss": 0.0502, + "step": 22081 + }, + { + "epoch": 2.89, + "learning_rate": 1.4767388645739686e-05, + "loss": 0.0574, + "step": 22082 + }, + { + "epoch": 2.89, + "learning_rate": 1.4766922459456172e-05, + "loss": 0.0554, + "step": 22083 + }, + { + "epoch": 2.89, + "learning_rate": 1.4766456259765946e-05, + "loss": 0.0522, + "step": 22084 + }, + { + "epoch": 2.89, + "learning_rate": 1.4765990046670316e-05, + "loss": 0.0517, + "step": 22085 + }, + { + "epoch": 2.89, + "learning_rate": 1.476552382017059e-05, + "loss": 0.0485, + "step": 22086 + }, + { + "epoch": 2.89, + "learning_rate": 1.4765057580268085e-05, + "loss": 0.0578, + "step": 22087 + }, + { + "epoch": 2.89, + "learning_rate": 1.4764591326964106e-05, + "loss": 0.0497, + "step": 22088 + }, + { + "epoch": 2.89, + "learning_rate": 1.476412506025997e-05, + "loss": 0.0557, + "step": 22089 + }, + { + "epoch": 2.89, + "learning_rate": 1.4763658780156985e-05, + "loss": 0.0499, + "step": 22090 + }, + { + "epoch": 2.89, + "learning_rate": 1.4763192486656465e-05, + "loss": 0.0538, + "step": 22091 + }, + { + "epoch": 2.89, + "learning_rate": 1.4762726179759719e-05, + "loss": 0.054, + "step": 22092 + }, + { + "epoch": 2.89, + "learning_rate": 1.476225985946806e-05, + "loss": 0.0666, + "step": 22093 + }, + { + "epoch": 2.89, + "learning_rate": 1.4761793525782796e-05, + "loss": 0.0501, + "step": 22094 + }, + { + "epoch": 2.89, + "learning_rate": 1.4761327178705242e-05, + "loss": 0.0525, + "step": 22095 + }, + { + "epoch": 2.89, + "learning_rate": 1.4760860818236709e-05, + "loss": 0.056, + "step": 22096 + }, + { + "epoch": 2.89, + "learning_rate": 1.4760394444378513e-05, + "loss": 0.055, + "step": 22097 + }, + { + "epoch": 2.89, + "learning_rate": 1.4759928057131953e-05, + "loss": 0.0509, + "step": 22098 + }, + { + "epoch": 2.89, + "learning_rate": 1.4759461656498352e-05, + "loss": 0.0584, + "step": 22099 + }, + { + "epoch": 2.89, + "learning_rate": 1.4758995242479022e-05, + "loss": 0.0509, + "step": 22100 + }, + { + "epoch": 2.9, + "learning_rate": 1.4758528815075268e-05, + "loss": 0.0559, + "step": 22101 + }, + { + "epoch": 2.9, + "learning_rate": 1.4758062374288407e-05, + "loss": 0.0517, + "step": 22102 + }, + { + "epoch": 2.9, + "learning_rate": 1.4757595920119748e-05, + "loss": 0.0492, + "step": 22103 + }, + { + "epoch": 2.9, + "learning_rate": 1.4757129452570604e-05, + "loss": 0.0578, + "step": 22104 + }, + { + "epoch": 2.9, + "learning_rate": 1.4756662971642287e-05, + "loss": 0.0511, + "step": 22105 + }, + { + "epoch": 2.9, + "learning_rate": 1.4756196477336107e-05, + "loss": 0.0562, + "step": 22106 + }, + { + "epoch": 2.9, + "learning_rate": 1.4755729969653378e-05, + "loss": 0.0606, + "step": 22107 + }, + { + "epoch": 2.9, + "learning_rate": 1.4755263448595415e-05, + "loss": 0.049, + "step": 22108 + }, + { + "epoch": 2.9, + "learning_rate": 1.4754796914163522e-05, + "loss": 0.0566, + "step": 22109 + }, + { + "epoch": 2.9, + "learning_rate": 1.475433036635902e-05, + "loss": 0.0584, + "step": 22110 + }, + { + "epoch": 2.9, + "learning_rate": 1.4753863805183213e-05, + "loss": 0.0472, + "step": 22111 + }, + { + "epoch": 2.9, + "learning_rate": 1.4753397230637422e-05, + "loss": 0.0578, + "step": 22112 + }, + { + "epoch": 2.9, + "learning_rate": 1.4752930642722952e-05, + "loss": 0.0543, + "step": 22113 + }, + { + "epoch": 2.9, + "learning_rate": 1.4752464041441119e-05, + "loss": 0.052, + "step": 22114 + }, + { + "epoch": 2.9, + "learning_rate": 1.4751997426793231e-05, + "loss": 0.0473, + "step": 22115 + }, + { + "epoch": 2.9, + "learning_rate": 1.4751530798780604e-05, + "loss": 0.0533, + "step": 22116 + }, + { + "epoch": 2.9, + "learning_rate": 1.4751064157404551e-05, + "loss": 0.0487, + "step": 22117 + }, + { + "epoch": 2.9, + "learning_rate": 1.4750597502666384e-05, + "loss": 0.0534, + "step": 22118 + }, + { + "epoch": 2.9, + "learning_rate": 1.4750130834567412e-05, + "loss": 0.0584, + "step": 22119 + }, + { + "epoch": 2.9, + "learning_rate": 1.474966415310895e-05, + "loss": 0.0599, + "step": 22120 + }, + { + "epoch": 2.9, + "learning_rate": 1.4749197458292312e-05, + "loss": 0.0426, + "step": 22121 + }, + { + "epoch": 2.9, + "learning_rate": 1.4748730750118809e-05, + "loss": 0.0534, + "step": 22122 + }, + { + "epoch": 2.9, + "learning_rate": 1.4748264028589755e-05, + "loss": 0.0513, + "step": 22123 + }, + { + "epoch": 2.9, + "learning_rate": 1.4747797293706458e-05, + "loss": 0.0561, + "step": 22124 + }, + { + "epoch": 2.9, + "learning_rate": 1.4747330545470238e-05, + "loss": 0.0472, + "step": 22125 + }, + { + "epoch": 2.9, + "learning_rate": 1.4746863783882399e-05, + "loss": 0.0538, + "step": 22126 + }, + { + "epoch": 2.9, + "learning_rate": 1.474639700894426e-05, + "loss": 0.0516, + "step": 22127 + }, + { + "epoch": 2.9, + "learning_rate": 1.4745930220657134e-05, + "loss": 0.0547, + "step": 22128 + }, + { + "epoch": 2.9, + "learning_rate": 1.474546341902233e-05, + "loss": 0.0581, + "step": 22129 + }, + { + "epoch": 2.9, + "learning_rate": 1.4744996604041164e-05, + "loss": 0.0513, + "step": 22130 + }, + { + "epoch": 2.9, + "learning_rate": 1.4744529775714948e-05, + "loss": 0.0648, + "step": 22131 + }, + { + "epoch": 2.9, + "learning_rate": 1.4744062934044992e-05, + "loss": 0.0586, + "step": 22132 + }, + { + "epoch": 2.9, + "learning_rate": 1.4743596079032615e-05, + "loss": 0.0439, + "step": 22133 + }, + { + "epoch": 2.9, + "learning_rate": 1.4743129210679126e-05, + "loss": 0.0615, + "step": 22134 + }, + { + "epoch": 2.9, + "learning_rate": 1.4742662328985838e-05, + "loss": 0.0608, + "step": 22135 + }, + { + "epoch": 2.9, + "learning_rate": 1.4742195433954065e-05, + "loss": 0.055, + "step": 22136 + }, + { + "epoch": 2.9, + "learning_rate": 1.4741728525585121e-05, + "loss": 0.062, + "step": 22137 + }, + { + "epoch": 2.9, + "learning_rate": 1.4741261603880314e-05, + "loss": 0.0488, + "step": 22138 + }, + { + "epoch": 2.9, + "learning_rate": 1.4740794668840966e-05, + "loss": 0.056, + "step": 22139 + }, + { + "epoch": 2.9, + "learning_rate": 1.4740327720468382e-05, + "loss": 0.0455, + "step": 22140 + }, + { + "epoch": 2.9, + "learning_rate": 1.4739860758763883e-05, + "loss": 0.0635, + "step": 22141 + }, + { + "epoch": 2.9, + "learning_rate": 1.4739393783728773e-05, + "loss": 0.0536, + "step": 22142 + }, + { + "epoch": 2.9, + "learning_rate": 1.4738926795364373e-05, + "loss": 0.0588, + "step": 22143 + }, + { + "epoch": 2.9, + "learning_rate": 1.4738459793671992e-05, + "loss": 0.0632, + "step": 22144 + }, + { + "epoch": 2.9, + "learning_rate": 1.4737992778652945e-05, + "loss": 0.0544, + "step": 22145 + }, + { + "epoch": 2.9, + "learning_rate": 1.4737525750308548e-05, + "loss": 0.054, + "step": 22146 + }, + { + "epoch": 2.9, + "learning_rate": 1.473705870864011e-05, + "loss": 0.064, + "step": 22147 + }, + { + "epoch": 2.9, + "learning_rate": 1.4736591653648948e-05, + "loss": 0.055, + "step": 22148 + }, + { + "epoch": 2.9, + "learning_rate": 1.4736124585336374e-05, + "loss": 0.0517, + "step": 22149 + }, + { + "epoch": 2.9, + "learning_rate": 1.4735657503703701e-05, + "loss": 0.0501, + "step": 22150 + }, + { + "epoch": 2.9, + "learning_rate": 1.4735190408752247e-05, + "loss": 0.0517, + "step": 22151 + }, + { + "epoch": 2.9, + "learning_rate": 1.473472330048332e-05, + "loss": 0.0495, + "step": 22152 + }, + { + "epoch": 2.9, + "learning_rate": 1.4734256178898234e-05, + "loss": 0.0504, + "step": 22153 + }, + { + "epoch": 2.9, + "learning_rate": 1.4733789043998306e-05, + "loss": 0.0494, + "step": 22154 + }, + { + "epoch": 2.9, + "learning_rate": 1.4733321895784848e-05, + "loss": 0.0534, + "step": 22155 + }, + { + "epoch": 2.9, + "learning_rate": 1.4732854734259174e-05, + "loss": 0.0514, + "step": 22156 + }, + { + "epoch": 2.9, + "learning_rate": 1.47323875594226e-05, + "loss": 0.0555, + "step": 22157 + }, + { + "epoch": 2.9, + "learning_rate": 1.4731920371276438e-05, + "loss": 0.0607, + "step": 22158 + }, + { + "epoch": 2.9, + "learning_rate": 1.4731453169821999e-05, + "loss": 0.0598, + "step": 22159 + }, + { + "epoch": 2.9, + "learning_rate": 1.4730985955060601e-05, + "loss": 0.0636, + "step": 22160 + }, + { + "epoch": 2.9, + "learning_rate": 1.4730518726993557e-05, + "loss": 0.0566, + "step": 22161 + }, + { + "epoch": 2.9, + "learning_rate": 1.4730051485622184e-05, + "loss": 0.0525, + "step": 22162 + }, + { + "epoch": 2.9, + "learning_rate": 1.4729584230947789e-05, + "loss": 0.0505, + "step": 22163 + }, + { + "epoch": 2.9, + "learning_rate": 1.4729116962971693e-05, + "loss": 0.0561, + "step": 22164 + }, + { + "epoch": 2.9, + "learning_rate": 1.4728649681695205e-05, + "loss": 0.0508, + "step": 22165 + }, + { + "epoch": 2.9, + "learning_rate": 1.4728182387119642e-05, + "loss": 0.0617, + "step": 22166 + }, + { + "epoch": 2.9, + "learning_rate": 1.4727715079246318e-05, + "loss": 0.0526, + "step": 22167 + }, + { + "epoch": 2.9, + "learning_rate": 1.4727247758076548e-05, + "loss": 0.0531, + "step": 22168 + }, + { + "epoch": 2.9, + "learning_rate": 1.4726780423611646e-05, + "loss": 0.0619, + "step": 22169 + }, + { + "epoch": 2.9, + "learning_rate": 1.4726313075852922e-05, + "loss": 0.0524, + "step": 22170 + }, + { + "epoch": 2.9, + "learning_rate": 1.4725845714801699e-05, + "loss": 0.0471, + "step": 22171 + }, + { + "epoch": 2.9, + "learning_rate": 1.4725378340459284e-05, + "loss": 0.0554, + "step": 22172 + }, + { + "epoch": 2.9, + "learning_rate": 1.4724910952826996e-05, + "loss": 0.0613, + "step": 22173 + }, + { + "epoch": 2.9, + "learning_rate": 1.4724443551906145e-05, + "loss": 0.0644, + "step": 22174 + }, + { + "epoch": 2.9, + "learning_rate": 1.4723976137698051e-05, + "loss": 0.0566, + "step": 22175 + }, + { + "epoch": 2.9, + "learning_rate": 1.4723508710204022e-05, + "loss": 0.0542, + "step": 22176 + }, + { + "epoch": 2.91, + "learning_rate": 1.472304126942538e-05, + "loss": 0.05, + "step": 22177 + }, + { + "epoch": 2.91, + "learning_rate": 1.4722573815363432e-05, + "loss": 0.054, + "step": 22178 + }, + { + "epoch": 2.91, + "learning_rate": 1.4722106348019501e-05, + "loss": 0.0634, + "step": 22179 + }, + { + "epoch": 2.91, + "learning_rate": 1.4721638867394895e-05, + "loss": 0.0643, + "step": 22180 + }, + { + "epoch": 2.91, + "learning_rate": 1.472117137349093e-05, + "loss": 0.053, + "step": 22181 + }, + { + "epoch": 2.91, + "learning_rate": 1.4720703866308923e-05, + "loss": 0.0501, + "step": 22182 + }, + { + "epoch": 2.91, + "learning_rate": 1.472023634585019e-05, + "loss": 0.0601, + "step": 22183 + }, + { + "epoch": 2.91, + "learning_rate": 1.471976881211604e-05, + "loss": 0.0538, + "step": 22184 + }, + { + "epoch": 2.91, + "learning_rate": 1.4719301265107793e-05, + "loss": 0.0495, + "step": 22185 + }, + { + "epoch": 2.91, + "learning_rate": 1.4718833704826765e-05, + "loss": 0.0569, + "step": 22186 + }, + { + "epoch": 2.91, + "learning_rate": 1.4718366131274264e-05, + "loss": 0.058, + "step": 22187 + }, + { + "epoch": 2.91, + "learning_rate": 1.4717898544451613e-05, + "loss": 0.0429, + "step": 22188 + }, + { + "epoch": 2.91, + "learning_rate": 1.471743094436012e-05, + "loss": 0.0666, + "step": 22189 + }, + { + "epoch": 2.91, + "learning_rate": 1.4716963331001108e-05, + "loss": 0.0627, + "step": 22190 + }, + { + "epoch": 2.91, + "learning_rate": 1.4716495704375885e-05, + "loss": 0.0591, + "step": 22191 + }, + { + "epoch": 2.91, + "learning_rate": 1.471602806448577e-05, + "loss": 0.0583, + "step": 22192 + }, + { + "epoch": 2.91, + "learning_rate": 1.4715560411332077e-05, + "loss": 0.0522, + "step": 22193 + }, + { + "epoch": 2.91, + "learning_rate": 1.4715092744916124e-05, + "loss": 0.049, + "step": 22194 + }, + { + "epoch": 2.91, + "learning_rate": 1.471462506523922e-05, + "loss": 0.0482, + "step": 22195 + }, + { + "epoch": 2.91, + "learning_rate": 1.4714157372302686e-05, + "loss": 0.057, + "step": 22196 + }, + { + "epoch": 2.91, + "learning_rate": 1.4713689666107836e-05, + "loss": 0.0545, + "step": 22197 + }, + { + "epoch": 2.91, + "learning_rate": 1.4713221946655983e-05, + "loss": 0.0508, + "step": 22198 + }, + { + "epoch": 2.91, + "learning_rate": 1.4712754213948444e-05, + "loss": 0.0512, + "step": 22199 + }, + { + "epoch": 2.91, + "learning_rate": 1.4712286467986537e-05, + "loss": 0.0578, + "step": 22200 + }, + { + "epoch": 2.91, + "learning_rate": 1.4711818708771572e-05, + "loss": 0.0515, + "step": 22201 + }, + { + "epoch": 2.91, + "learning_rate": 1.4711350936304871e-05, + "loss": 0.0421, + "step": 22202 + }, + { + "epoch": 2.91, + "learning_rate": 1.4710883150587742e-05, + "loss": 0.0546, + "step": 22203 + }, + { + "epoch": 2.91, + "learning_rate": 1.471041535162151e-05, + "loss": 0.0517, + "step": 22204 + }, + { + "epoch": 2.91, + "learning_rate": 1.4709947539407482e-05, + "loss": 0.0607, + "step": 22205 + }, + { + "epoch": 2.91, + "learning_rate": 1.4709479713946981e-05, + "loss": 0.0421, + "step": 22206 + }, + { + "epoch": 2.91, + "learning_rate": 1.4709011875241319e-05, + "loss": 0.0571, + "step": 22207 + }, + { + "epoch": 2.91, + "learning_rate": 1.4708544023291809e-05, + "loss": 0.0603, + "step": 22208 + }, + { + "epoch": 2.91, + "learning_rate": 1.4708076158099769e-05, + "loss": 0.0533, + "step": 22209 + }, + { + "epoch": 2.91, + "learning_rate": 1.4707608279666516e-05, + "loss": 0.0601, + "step": 22210 + }, + { + "epoch": 2.91, + "learning_rate": 1.4707140387993368e-05, + "loss": 0.0533, + "step": 22211 + }, + { + "epoch": 2.91, + "learning_rate": 1.4706672483081636e-05, + "loss": 0.0598, + "step": 22212 + }, + { + "epoch": 2.91, + "learning_rate": 1.4706204564932642e-05, + "loss": 0.0498, + "step": 22213 + }, + { + "epoch": 2.91, + "learning_rate": 1.4705736633547694e-05, + "loss": 0.0497, + "step": 22214 + }, + { + "epoch": 2.91, + "learning_rate": 1.4705268688928116e-05, + "loss": 0.0546, + "step": 22215 + }, + { + "epoch": 2.91, + "learning_rate": 1.4704800731075218e-05, + "loss": 0.059, + "step": 22216 + }, + { + "epoch": 2.91, + "learning_rate": 1.4704332759990319e-05, + "loss": 0.0522, + "step": 22217 + }, + { + "epoch": 2.91, + "learning_rate": 1.4703864775674734e-05, + "loss": 0.0552, + "step": 22218 + }, + { + "epoch": 2.91, + "learning_rate": 1.470339677812978e-05, + "loss": 0.0519, + "step": 22219 + }, + { + "epoch": 2.91, + "learning_rate": 1.4702928767356772e-05, + "loss": 0.0424, + "step": 22220 + }, + { + "epoch": 2.91, + "learning_rate": 1.470246074335703e-05, + "loss": 0.0573, + "step": 22221 + }, + { + "epoch": 2.91, + "learning_rate": 1.4701992706131867e-05, + "loss": 0.0615, + "step": 22222 + }, + { + "epoch": 2.91, + "learning_rate": 1.47015246556826e-05, + "loss": 0.0652, + "step": 22223 + }, + { + "epoch": 2.91, + "learning_rate": 1.4701056592010543e-05, + "loss": 0.0594, + "step": 22224 + }, + { + "epoch": 2.91, + "learning_rate": 1.4700588515117015e-05, + "loss": 0.056, + "step": 22225 + }, + { + "epoch": 2.91, + "learning_rate": 1.4700120425003333e-05, + "loss": 0.0525, + "step": 22226 + }, + { + "epoch": 2.91, + "learning_rate": 1.4699652321670814e-05, + "loss": 0.0613, + "step": 22227 + }, + { + "epoch": 2.91, + "learning_rate": 1.469918420512077e-05, + "loss": 0.0571, + "step": 22228 + }, + { + "epoch": 2.91, + "learning_rate": 1.4698716075354522e-05, + "loss": 0.0538, + "step": 22229 + }, + { + "epoch": 2.91, + "learning_rate": 1.4698247932373386e-05, + "loss": 0.0525, + "step": 22230 + }, + { + "epoch": 2.91, + "learning_rate": 1.469777977617868e-05, + "loss": 0.052, + "step": 22231 + }, + { + "epoch": 2.91, + "learning_rate": 1.4697311606771715e-05, + "loss": 0.0563, + "step": 22232 + }, + { + "epoch": 2.91, + "learning_rate": 1.469684342415381e-05, + "loss": 0.0607, + "step": 22233 + }, + { + "epoch": 2.91, + "learning_rate": 1.469637522832629e-05, + "loss": 0.0496, + "step": 22234 + }, + { + "epoch": 2.91, + "learning_rate": 1.4695907019290459e-05, + "loss": 0.0546, + "step": 22235 + }, + { + "epoch": 2.91, + "learning_rate": 1.4695438797047641e-05, + "loss": 0.0574, + "step": 22236 + }, + { + "epoch": 2.91, + "learning_rate": 1.4694970561599148e-05, + "loss": 0.0515, + "step": 22237 + }, + { + "epoch": 2.91, + "learning_rate": 1.4694502312946306e-05, + "loss": 0.0528, + "step": 22238 + }, + { + "epoch": 2.91, + "learning_rate": 1.4694034051090422e-05, + "loss": 0.0475, + "step": 22239 + }, + { + "epoch": 2.91, + "learning_rate": 1.4693565776032821e-05, + "loss": 0.0503, + "step": 22240 + }, + { + "epoch": 2.91, + "learning_rate": 1.4693097487774812e-05, + "loss": 0.0509, + "step": 22241 + }, + { + "epoch": 2.91, + "learning_rate": 1.4692629186317719e-05, + "loss": 0.0583, + "step": 22242 + }, + { + "epoch": 2.91, + "learning_rate": 1.4692160871662856e-05, + "loss": 0.0599, + "step": 22243 + }, + { + "epoch": 2.91, + "learning_rate": 1.4691692543811542e-05, + "loss": 0.0557, + "step": 22244 + }, + { + "epoch": 2.91, + "learning_rate": 1.4691224202765091e-05, + "loss": 0.0527, + "step": 22245 + }, + { + "epoch": 2.91, + "learning_rate": 1.469075584852482e-05, + "loss": 0.0515, + "step": 22246 + }, + { + "epoch": 2.91, + "learning_rate": 1.469028748109205e-05, + "loss": 0.0458, + "step": 22247 + }, + { + "epoch": 2.91, + "learning_rate": 1.4689819100468097e-05, + "loss": 0.0512, + "step": 22248 + }, + { + "epoch": 2.91, + "learning_rate": 1.4689350706654275e-05, + "loss": 0.0548, + "step": 22249 + }, + { + "epoch": 2.91, + "learning_rate": 1.4688882299651903e-05, + "loss": 0.056, + "step": 22250 + }, + { + "epoch": 2.91, + "learning_rate": 1.4688413879462304e-05, + "loss": 0.0533, + "step": 22251 + }, + { + "epoch": 2.91, + "learning_rate": 1.4687945446086786e-05, + "loss": 0.0546, + "step": 22252 + }, + { + "epoch": 2.91, + "learning_rate": 1.4687476999526673e-05, + "loss": 0.0565, + "step": 22253 + }, + { + "epoch": 2.92, + "learning_rate": 1.4687008539783281e-05, + "loss": 0.0525, + "step": 22254 + }, + { + "epoch": 2.92, + "learning_rate": 1.4686540066857929e-05, + "loss": 0.047, + "step": 22255 + }, + { + "epoch": 2.92, + "learning_rate": 1.4686071580751929e-05, + "loss": 0.0513, + "step": 22256 + }, + { + "epoch": 2.92, + "learning_rate": 1.4685603081466604e-05, + "loss": 0.049, + "step": 22257 + }, + { + "epoch": 2.92, + "learning_rate": 1.4685134569003268e-05, + "loss": 0.0552, + "step": 22258 + }, + { + "epoch": 2.92, + "learning_rate": 1.4684666043363242e-05, + "loss": 0.0439, + "step": 22259 + }, + { + "epoch": 2.92, + "learning_rate": 1.468419750454784e-05, + "loss": 0.0508, + "step": 22260 + }, + { + "epoch": 2.92, + "learning_rate": 1.4683728952558387e-05, + "loss": 0.0612, + "step": 22261 + }, + { + "epoch": 2.92, + "learning_rate": 1.4683260387396191e-05, + "loss": 0.0617, + "step": 22262 + }, + { + "epoch": 2.92, + "learning_rate": 1.4682791809062578e-05, + "loss": 0.0536, + "step": 22263 + }, + { + "epoch": 2.92, + "learning_rate": 1.468232321755886e-05, + "loss": 0.0551, + "step": 22264 + }, + { + "epoch": 2.92, + "learning_rate": 1.4681854612886359e-05, + "loss": 0.0603, + "step": 22265 + }, + { + "epoch": 2.92, + "learning_rate": 1.4681385995046392e-05, + "loss": 0.0565, + "step": 22266 + }, + { + "epoch": 2.92, + "learning_rate": 1.4680917364040275e-05, + "loss": 0.0505, + "step": 22267 + }, + { + "epoch": 2.92, + "learning_rate": 1.4680448719869325e-05, + "loss": 0.0462, + "step": 22268 + }, + { + "epoch": 2.92, + "learning_rate": 1.4679980062534864e-05, + "loss": 0.0528, + "step": 22269 + }, + { + "epoch": 2.92, + "learning_rate": 1.4679511392038208e-05, + "loss": 0.0498, + "step": 22270 + }, + { + "epoch": 2.92, + "learning_rate": 1.4679042708380675e-05, + "loss": 0.0581, + "step": 22271 + }, + { + "epoch": 2.92, + "learning_rate": 1.4678574011563585e-05, + "loss": 0.0528, + "step": 22272 + }, + { + "epoch": 2.92, + "learning_rate": 1.4678105301588254e-05, + "loss": 0.0502, + "step": 22273 + }, + { + "epoch": 2.92, + "learning_rate": 1.4677636578456001e-05, + "loss": 0.061, + "step": 22274 + }, + { + "epoch": 2.92, + "learning_rate": 1.4677167842168143e-05, + "loss": 0.0538, + "step": 22275 + }, + { + "epoch": 2.92, + "learning_rate": 1.4676699092726001e-05, + "loss": 0.0531, + "step": 22276 + }, + { + "epoch": 2.92, + "learning_rate": 1.4676230330130892e-05, + "loss": 0.0469, + "step": 22277 + }, + { + "epoch": 2.92, + "learning_rate": 1.4675761554384134e-05, + "loss": 0.0572, + "step": 22278 + }, + { + "epoch": 2.92, + "learning_rate": 1.4675292765487044e-05, + "loss": 0.0541, + "step": 22279 + }, + { + "epoch": 2.92, + "learning_rate": 1.4674823963440944e-05, + "loss": 0.0572, + "step": 22280 + }, + { + "epoch": 2.92, + "learning_rate": 1.467435514824715e-05, + "loss": 0.0571, + "step": 22281 + }, + { + "epoch": 2.92, + "learning_rate": 1.4673886319906982e-05, + "loss": 0.0474, + "step": 22282 + }, + { + "epoch": 2.92, + "learning_rate": 1.4673417478421756e-05, + "loss": 0.0654, + "step": 22283 + }, + { + "epoch": 2.92, + "learning_rate": 1.4672948623792795e-05, + "loss": 0.0526, + "step": 22284 + }, + { + "epoch": 2.92, + "learning_rate": 1.467247975602141e-05, + "loss": 0.0642, + "step": 22285 + }, + { + "epoch": 2.92, + "learning_rate": 1.4672010875108928e-05, + "loss": 0.0578, + "step": 22286 + }, + { + "epoch": 2.92, + "learning_rate": 1.4671541981056662e-05, + "loss": 0.0544, + "step": 22287 + }, + { + "epoch": 2.92, + "learning_rate": 1.4671073073865935e-05, + "loss": 0.0516, + "step": 22288 + }, + { + "epoch": 2.92, + "learning_rate": 1.4670604153538063e-05, + "loss": 0.0496, + "step": 22289 + }, + { + "epoch": 2.92, + "learning_rate": 1.4670135220074367e-05, + "loss": 0.0552, + "step": 22290 + }, + { + "epoch": 2.92, + "learning_rate": 1.4669666273476162e-05, + "loss": 0.0555, + "step": 22291 + }, + { + "epoch": 2.92, + "learning_rate": 1.4669197313744771e-05, + "loss": 0.0532, + "step": 22292 + }, + { + "epoch": 2.92, + "learning_rate": 1.466872834088151e-05, + "loss": 0.0493, + "step": 22293 + }, + { + "epoch": 2.92, + "learning_rate": 1.46682593548877e-05, + "loss": 0.0569, + "step": 22294 + }, + { + "epoch": 2.92, + "learning_rate": 1.466779035576466e-05, + "loss": 0.0596, + "step": 22295 + }, + { + "epoch": 2.92, + "learning_rate": 1.4667321343513705e-05, + "loss": 0.0487, + "step": 22296 + }, + { + "epoch": 2.92, + "learning_rate": 1.466685231813616e-05, + "loss": 0.0526, + "step": 22297 + }, + { + "epoch": 2.92, + "learning_rate": 1.466638327963334e-05, + "loss": 0.056, + "step": 22298 + }, + { + "epoch": 2.92, + "learning_rate": 1.4665914228006568e-05, + "loss": 0.0588, + "step": 22299 + }, + { + "epoch": 2.92, + "learning_rate": 1.4665445163257159e-05, + "loss": 0.0544, + "step": 22300 + }, + { + "epoch": 2.92, + "learning_rate": 1.4664976085386434e-05, + "loss": 0.0479, + "step": 22301 + }, + { + "epoch": 2.92, + "learning_rate": 1.466450699439571e-05, + "loss": 0.056, + "step": 22302 + }, + { + "epoch": 2.92, + "learning_rate": 1.466403789028631e-05, + "loss": 0.0609, + "step": 22303 + }, + { + "epoch": 2.92, + "learning_rate": 1.4663568773059553e-05, + "loss": 0.0609, + "step": 22304 + }, + { + "epoch": 2.92, + "learning_rate": 1.4663099642716758e-05, + "loss": 0.0535, + "step": 22305 + }, + { + "epoch": 2.92, + "learning_rate": 1.4662630499259244e-05, + "loss": 0.0551, + "step": 22306 + }, + { + "epoch": 2.92, + "learning_rate": 1.4662161342688326e-05, + "loss": 0.0586, + "step": 22307 + }, + { + "epoch": 2.92, + "learning_rate": 1.4661692173005328e-05, + "loss": 0.0459, + "step": 22308 + }, + { + "epoch": 2.92, + "learning_rate": 1.4661222990211573e-05, + "loss": 0.0552, + "step": 22309 + }, + { + "epoch": 2.92, + "learning_rate": 1.4660753794308374e-05, + "loss": 0.0595, + "step": 22310 + }, + { + "epoch": 2.92, + "learning_rate": 1.4660284585297051e-05, + "loss": 0.058, + "step": 22311 + }, + { + "epoch": 2.92, + "learning_rate": 1.4659815363178928e-05, + "loss": 0.0561, + "step": 22312 + }, + { + "epoch": 2.92, + "learning_rate": 1.4659346127955326e-05, + "loss": 0.0631, + "step": 22313 + }, + { + "epoch": 2.92, + "learning_rate": 1.4658876879627557e-05, + "loss": 0.0589, + "step": 22314 + }, + { + "epoch": 2.92, + "learning_rate": 1.4658407618196943e-05, + "loss": 0.0522, + "step": 22315 + }, + { + "epoch": 2.92, + "learning_rate": 1.465793834366481e-05, + "loss": 0.0533, + "step": 22316 + }, + { + "epoch": 2.92, + "learning_rate": 1.4657469056032471e-05, + "loss": 0.0513, + "step": 22317 + }, + { + "epoch": 2.92, + "learning_rate": 1.4656999755301249e-05, + "loss": 0.0474, + "step": 22318 + }, + { + "epoch": 2.92, + "learning_rate": 1.4656530441472462e-05, + "loss": 0.0535, + "step": 22319 + }, + { + "epoch": 2.92, + "learning_rate": 1.4656061114547435e-05, + "loss": 0.0484, + "step": 22320 + }, + { + "epoch": 2.92, + "learning_rate": 1.465559177452748e-05, + "loss": 0.0546, + "step": 22321 + }, + { + "epoch": 2.92, + "learning_rate": 1.4655122421413923e-05, + "loss": 0.0555, + "step": 22322 + }, + { + "epoch": 2.92, + "learning_rate": 1.465465305520808e-05, + "loss": 0.0474, + "step": 22323 + }, + { + "epoch": 2.92, + "learning_rate": 1.4654183675911275e-05, + "loss": 0.0517, + "step": 22324 + }, + { + "epoch": 2.92, + "learning_rate": 1.4653714283524828e-05, + "loss": 0.0541, + "step": 22325 + }, + { + "epoch": 2.92, + "learning_rate": 1.4653244878050055e-05, + "loss": 0.0512, + "step": 22326 + }, + { + "epoch": 2.92, + "learning_rate": 1.465277545948828e-05, + "loss": 0.0521, + "step": 22327 + }, + { + "epoch": 2.92, + "learning_rate": 1.4652306027840821e-05, + "loss": 0.0568, + "step": 22328 + }, + { + "epoch": 2.92, + "learning_rate": 1.4651836583108997e-05, + "loss": 0.0489, + "step": 22329 + }, + { + "epoch": 2.93, + "learning_rate": 1.4651367125294131e-05, + "loss": 0.0554, + "step": 22330 + }, + { + "epoch": 2.93, + "learning_rate": 1.4650897654397544e-05, + "loss": 0.0468, + "step": 22331 + }, + { + "epoch": 2.93, + "learning_rate": 1.4650428170420557e-05, + "loss": 0.0569, + "step": 22332 + }, + { + "epoch": 2.93, + "learning_rate": 1.4649958673364487e-05, + "loss": 0.0543, + "step": 22333 + }, + { + "epoch": 2.93, + "learning_rate": 1.4649489163230653e-05, + "loss": 0.0559, + "step": 22334 + }, + { + "epoch": 2.93, + "learning_rate": 1.464901964002038e-05, + "loss": 0.0593, + "step": 22335 + }, + { + "epoch": 2.93, + "learning_rate": 1.4648550103734987e-05, + "loss": 0.0597, + "step": 22336 + }, + { + "epoch": 2.93, + "learning_rate": 1.4648080554375796e-05, + "loss": 0.048, + "step": 22337 + }, + { + "epoch": 2.93, + "learning_rate": 1.4647610991944124e-05, + "loss": 0.041, + "step": 22338 + }, + { + "epoch": 2.93, + "learning_rate": 1.4647141416441295e-05, + "loss": 0.0565, + "step": 22339 + }, + { + "epoch": 2.93, + "learning_rate": 1.4646671827868626e-05, + "loss": 0.062, + "step": 22340 + }, + { + "epoch": 2.93, + "learning_rate": 1.4646202226227441e-05, + "loss": 0.0552, + "step": 22341 + }, + { + "epoch": 2.93, + "learning_rate": 1.4645732611519058e-05, + "loss": 0.058, + "step": 22342 + }, + { + "epoch": 2.93, + "learning_rate": 1.4645262983744807e-05, + "loss": 0.0512, + "step": 22343 + }, + { + "epoch": 2.93, + "learning_rate": 1.4644793342905991e-05, + "loss": 0.0462, + "step": 22344 + }, + { + "epoch": 2.93, + "learning_rate": 1.4644323689003948e-05, + "loss": 0.0508, + "step": 22345 + }, + { + "epoch": 2.93, + "learning_rate": 1.4643854022039987e-05, + "loss": 0.0651, + "step": 22346 + }, + { + "epoch": 2.93, + "learning_rate": 1.4643384342015438e-05, + "loss": 0.0495, + "step": 22347 + }, + { + "epoch": 2.93, + "learning_rate": 1.4642914648931616e-05, + "loss": 0.0497, + "step": 22348 + }, + { + "epoch": 2.93, + "learning_rate": 1.4642444942789847e-05, + "loss": 0.0495, + "step": 22349 + }, + { + "epoch": 2.93, + "learning_rate": 1.4641975223591443e-05, + "loss": 0.0544, + "step": 22350 + }, + { + "epoch": 2.93, + "learning_rate": 1.4641505491337735e-05, + "loss": 0.0545, + "step": 22351 + }, + { + "epoch": 2.93, + "learning_rate": 1.4641035746030038e-05, + "loss": 0.0517, + "step": 22352 + }, + { + "epoch": 2.93, + "learning_rate": 1.4640565987669677e-05, + "loss": 0.0558, + "step": 22353 + }, + { + "epoch": 2.93, + "learning_rate": 1.464009621625797e-05, + "loss": 0.0493, + "step": 22354 + }, + { + "epoch": 2.93, + "learning_rate": 1.4639626431796238e-05, + "loss": 0.061, + "step": 22355 + }, + { + "epoch": 2.93, + "learning_rate": 1.4639156634285808e-05, + "loss": 0.054, + "step": 22356 + }, + { + "epoch": 2.93, + "learning_rate": 1.4638686823727994e-05, + "loss": 0.0583, + "step": 22357 + }, + { + "epoch": 2.93, + "learning_rate": 1.463821700012412e-05, + "loss": 0.0475, + "step": 22358 + }, + { + "epoch": 2.93, + "learning_rate": 1.4637747163475508e-05, + "loss": 0.0493, + "step": 22359 + }, + { + "epoch": 2.93, + "learning_rate": 1.4637277313783482e-05, + "loss": 0.0618, + "step": 22360 + }, + { + "epoch": 2.93, + "learning_rate": 1.4636807451049355e-05, + "loss": 0.0566, + "step": 22361 + }, + { + "epoch": 2.93, + "learning_rate": 1.4636337575274458e-05, + "loss": 0.0565, + "step": 22362 + }, + { + "epoch": 2.93, + "learning_rate": 1.4635867686460108e-05, + "loss": 0.0456, + "step": 22363 + }, + { + "epoch": 2.93, + "learning_rate": 1.4635397784607628e-05, + "loss": 0.0466, + "step": 22364 + }, + { + "epoch": 2.93, + "learning_rate": 1.4634927869718336e-05, + "loss": 0.069, + "step": 22365 + }, + { + "epoch": 2.93, + "learning_rate": 1.4634457941793558e-05, + "loss": 0.0603, + "step": 22366 + }, + { + "epoch": 2.93, + "learning_rate": 1.4633988000834612e-05, + "loss": 0.0577, + "step": 22367 + }, + { + "epoch": 2.93, + "learning_rate": 1.4633518046842823e-05, + "loss": 0.0472, + "step": 22368 + }, + { + "epoch": 2.93, + "learning_rate": 1.463304807981951e-05, + "loss": 0.0552, + "step": 22369 + }, + { + "epoch": 2.93, + "learning_rate": 1.4632578099765998e-05, + "loss": 0.049, + "step": 22370 + }, + { + "epoch": 2.93, + "learning_rate": 1.4632108106683607e-05, + "loss": 0.0538, + "step": 22371 + }, + { + "epoch": 2.93, + "learning_rate": 1.4631638100573656e-05, + "loss": 0.0454, + "step": 22372 + }, + { + "epoch": 2.93, + "learning_rate": 1.463116808143747e-05, + "loss": 0.063, + "step": 22373 + }, + { + "epoch": 2.93, + "learning_rate": 1.4630698049276372e-05, + "loss": 0.0541, + "step": 22374 + }, + { + "epoch": 2.93, + "learning_rate": 1.463022800409168e-05, + "loss": 0.0567, + "step": 22375 + }, + { + "epoch": 2.93, + "learning_rate": 1.4629757945884722e-05, + "loss": 0.0516, + "step": 22376 + }, + { + "epoch": 2.93, + "learning_rate": 1.4629287874656814e-05, + "loss": 0.0515, + "step": 22377 + }, + { + "epoch": 2.93, + "learning_rate": 1.462881779040928e-05, + "loss": 0.0439, + "step": 22378 + }, + { + "epoch": 2.93, + "learning_rate": 1.4628347693143443e-05, + "loss": 0.0597, + "step": 22379 + }, + { + "epoch": 2.93, + "learning_rate": 1.4627877582860623e-05, + "loss": 0.0563, + "step": 22380 + }, + { + "epoch": 2.93, + "learning_rate": 1.4627407459562148e-05, + "loss": 0.064, + "step": 22381 + }, + { + "epoch": 2.93, + "learning_rate": 1.4626937323249333e-05, + "loss": 0.0506, + "step": 22382 + }, + { + "epoch": 2.93, + "learning_rate": 1.4626467173923502e-05, + "loss": 0.0546, + "step": 22383 + }, + { + "epoch": 2.93, + "learning_rate": 1.462599701158598e-05, + "loss": 0.0661, + "step": 22384 + }, + { + "epoch": 2.93, + "learning_rate": 1.4625526836238089e-05, + "loss": 0.0596, + "step": 22385 + }, + { + "epoch": 2.93, + "learning_rate": 1.4625056647881145e-05, + "loss": 0.0527, + "step": 22386 + }, + { + "epoch": 2.93, + "learning_rate": 1.4624586446516485e-05, + "loss": 0.0545, + "step": 22387 + }, + { + "epoch": 2.93, + "learning_rate": 1.4624116232145413e-05, + "loss": 0.0599, + "step": 22388 + }, + { + "epoch": 2.93, + "learning_rate": 1.4623646004769262e-05, + "loss": 0.0515, + "step": 22389 + }, + { + "epoch": 2.93, + "learning_rate": 1.4623175764389355e-05, + "loss": 0.0584, + "step": 22390 + }, + { + "epoch": 2.93, + "learning_rate": 1.4622705511007013e-05, + "loss": 0.0616, + "step": 22391 + }, + { + "epoch": 2.93, + "learning_rate": 1.4622235244623557e-05, + "loss": 0.0601, + "step": 22392 + }, + { + "epoch": 2.93, + "learning_rate": 1.4621764965240311e-05, + "loss": 0.0571, + "step": 22393 + }, + { + "epoch": 2.93, + "learning_rate": 1.4621294672858594e-05, + "loss": 0.0574, + "step": 22394 + }, + { + "epoch": 2.93, + "learning_rate": 1.4620824367479736e-05, + "loss": 0.0525, + "step": 22395 + }, + { + "epoch": 2.93, + "learning_rate": 1.4620354049105052e-05, + "loss": 0.0588, + "step": 22396 + }, + { + "epoch": 2.93, + "learning_rate": 1.4619883717735872e-05, + "loss": 0.0532, + "step": 22397 + }, + { + "epoch": 2.93, + "learning_rate": 1.4619413373373512e-05, + "loss": 0.0493, + "step": 22398 + }, + { + "epoch": 2.93, + "learning_rate": 1.4618943016019298e-05, + "loss": 0.058, + "step": 22399 + }, + { + "epoch": 2.93, + "learning_rate": 1.4618472645674553e-05, + "loss": 0.0656, + "step": 22400 + }, + { + "epoch": 2.93, + "learning_rate": 1.46180022623406e-05, + "loss": 0.0498, + "step": 22401 + }, + { + "epoch": 2.93, + "learning_rate": 1.4617531866018764e-05, + "loss": 0.0611, + "step": 22402 + }, + { + "epoch": 2.93, + "learning_rate": 1.4617061456710363e-05, + "loss": 0.0579, + "step": 22403 + }, + { + "epoch": 2.93, + "learning_rate": 1.4616591034416723e-05, + "loss": 0.0562, + "step": 22404 + }, + { + "epoch": 2.93, + "learning_rate": 1.4616120599139164e-05, + "loss": 0.064, + "step": 22405 + }, + { + "epoch": 2.94, + "learning_rate": 1.4615650150879016e-05, + "loss": 0.0482, + "step": 22406 + }, + { + "epoch": 2.94, + "learning_rate": 1.4615179689637594e-05, + "loss": 0.0533, + "step": 22407 + }, + { + "epoch": 2.94, + "learning_rate": 1.461470921541623e-05, + "loss": 0.056, + "step": 22408 + }, + { + "epoch": 2.94, + "learning_rate": 1.4614238728216236e-05, + "loss": 0.0466, + "step": 22409 + }, + { + "epoch": 2.94, + "learning_rate": 1.4613768228038945e-05, + "loss": 0.051, + "step": 22410 + }, + { + "epoch": 2.94, + "learning_rate": 1.4613297714885675e-05, + "loss": 0.0597, + "step": 22411 + }, + { + "epoch": 2.94, + "learning_rate": 1.4612827188757752e-05, + "loss": 0.0529, + "step": 22412 + }, + { + "epoch": 2.94, + "learning_rate": 1.4612356649656499e-05, + "loss": 0.0673, + "step": 22413 + }, + { + "epoch": 2.94, + "learning_rate": 1.4611886097583239e-05, + "loss": 0.0522, + "step": 22414 + }, + { + "epoch": 2.94, + "learning_rate": 1.461141553253929e-05, + "loss": 0.0598, + "step": 22415 + }, + { + "epoch": 2.94, + "learning_rate": 1.4610944954525983e-05, + "loss": 0.0525, + "step": 22416 + }, + { + "epoch": 2.94, + "learning_rate": 1.461047436354464e-05, + "loss": 0.0585, + "step": 22417 + }, + { + "epoch": 2.94, + "learning_rate": 1.4610003759596584e-05, + "loss": 0.0576, + "step": 22418 + }, + { + "epoch": 2.94, + "learning_rate": 1.4609533142683137e-05, + "loss": 0.0529, + "step": 22419 + }, + { + "epoch": 2.94, + "learning_rate": 1.4609062512805623e-05, + "loss": 0.06, + "step": 22420 + }, + { + "epoch": 2.94, + "learning_rate": 1.4608591869965365e-05, + "loss": 0.061, + "step": 22421 + }, + { + "epoch": 2.94, + "learning_rate": 1.460812121416369e-05, + "loss": 0.0507, + "step": 22422 + }, + { + "epoch": 2.94, + "learning_rate": 1.4607650545401918e-05, + "loss": 0.0503, + "step": 22423 + }, + { + "epoch": 2.94, + "learning_rate": 1.4607179863681375e-05, + "loss": 0.0633, + "step": 22424 + }, + { + "epoch": 2.94, + "learning_rate": 1.4606709169003386e-05, + "loss": 0.052, + "step": 22425 + }, + { + "epoch": 2.94, + "learning_rate": 1.4606238461369268e-05, + "loss": 0.0527, + "step": 22426 + }, + { + "epoch": 2.94, + "learning_rate": 1.4605767740780351e-05, + "loss": 0.0543, + "step": 22427 + }, + { + "epoch": 2.94, + "learning_rate": 1.460529700723796e-05, + "loss": 0.0598, + "step": 22428 + }, + { + "epoch": 2.94, + "learning_rate": 1.4604826260743417e-05, + "loss": 0.053, + "step": 22429 + }, + { + "epoch": 2.94, + "learning_rate": 1.4604355501298043e-05, + "loss": 0.0633, + "step": 22430 + }, + { + "epoch": 2.94, + "learning_rate": 1.4603884728903166e-05, + "loss": 0.0528, + "step": 22431 + }, + { + "epoch": 2.94, + "learning_rate": 1.4603413943560106e-05, + "loss": 0.0539, + "step": 22432 + }, + { + "epoch": 2.94, + "learning_rate": 1.460294314527019e-05, + "loss": 0.0581, + "step": 22433 + }, + { + "epoch": 2.94, + "learning_rate": 1.4602472334034744e-05, + "loss": 0.053, + "step": 22434 + }, + { + "epoch": 2.94, + "learning_rate": 1.4602001509855085e-05, + "loss": 0.0617, + "step": 22435 + }, + { + "epoch": 2.94, + "learning_rate": 1.4601530672732546e-05, + "loss": 0.0563, + "step": 22436 + }, + { + "epoch": 2.94, + "learning_rate": 1.4601059822668447e-05, + "loss": 0.0586, + "step": 22437 + }, + { + "epoch": 2.94, + "learning_rate": 1.4600588959664108e-05, + "loss": 0.0448, + "step": 22438 + }, + { + "epoch": 2.94, + "learning_rate": 1.4600118083720862e-05, + "loss": 0.0565, + "step": 22439 + }, + { + "epoch": 2.94, + "learning_rate": 1.4599647194840028e-05, + "loss": 0.0523, + "step": 22440 + }, + { + "epoch": 2.94, + "learning_rate": 1.459917629302293e-05, + "loss": 0.0637, + "step": 22441 + }, + { + "epoch": 2.94, + "learning_rate": 1.4598705378270896e-05, + "loss": 0.0566, + "step": 22442 + }, + { + "epoch": 2.94, + "learning_rate": 1.4598234450585244e-05, + "loss": 0.0585, + "step": 22443 + }, + { + "epoch": 2.94, + "learning_rate": 1.4597763509967304e-05, + "loss": 0.0485, + "step": 22444 + }, + { + "epoch": 2.94, + "learning_rate": 1.4597292556418401e-05, + "loss": 0.0499, + "step": 22445 + }, + { + "epoch": 2.94, + "learning_rate": 1.4596821589939855e-05, + "loss": 0.0567, + "step": 22446 + }, + { + "epoch": 2.94, + "learning_rate": 1.4596350610532995e-05, + "loss": 0.0529, + "step": 22447 + }, + { + "epoch": 2.94, + "learning_rate": 1.4595879618199144e-05, + "loss": 0.0561, + "step": 22448 + }, + { + "epoch": 2.94, + "learning_rate": 1.4595408612939623e-05, + "loss": 0.0589, + "step": 22449 + }, + { + "epoch": 2.94, + "learning_rate": 1.4594937594755766e-05, + "loss": 0.0542, + "step": 22450 + }, + { + "epoch": 2.94, + "learning_rate": 1.4594466563648885e-05, + "loss": 0.0619, + "step": 22451 + }, + { + "epoch": 2.94, + "learning_rate": 1.4593995519620316e-05, + "loss": 0.0579, + "step": 22452 + }, + { + "epoch": 2.94, + "learning_rate": 1.4593524462671377e-05, + "loss": 0.0443, + "step": 22453 + }, + { + "epoch": 2.94, + "learning_rate": 1.4593053392803395e-05, + "loss": 0.0567, + "step": 22454 + }, + { + "epoch": 2.94, + "learning_rate": 1.4592582310017695e-05, + "loss": 0.0497, + "step": 22455 + }, + { + "epoch": 2.94, + "learning_rate": 1.4592111214315603e-05, + "loss": 0.0497, + "step": 22456 + }, + { + "epoch": 2.94, + "learning_rate": 1.4591640105698443e-05, + "loss": 0.0451, + "step": 22457 + }, + { + "epoch": 2.94, + "learning_rate": 1.459116898416754e-05, + "loss": 0.0356, + "step": 22458 + }, + { + "epoch": 2.94, + "learning_rate": 1.4590697849724217e-05, + "loss": 0.0626, + "step": 22459 + }, + { + "epoch": 2.94, + "learning_rate": 1.45902267023698e-05, + "loss": 0.0583, + "step": 22460 + }, + { + "epoch": 2.94, + "learning_rate": 1.4589755542105617e-05, + "loss": 0.0457, + "step": 22461 + }, + { + "epoch": 2.94, + "learning_rate": 1.458928436893299e-05, + "loss": 0.0612, + "step": 22462 + }, + { + "epoch": 2.94, + "learning_rate": 1.4588813182853248e-05, + "loss": 0.0552, + "step": 22463 + }, + { + "epoch": 2.94, + "learning_rate": 1.4588341983867709e-05, + "loss": 0.0577, + "step": 22464 + }, + { + "epoch": 2.94, + "learning_rate": 1.4587870771977703e-05, + "loss": 0.06, + "step": 22465 + }, + { + "epoch": 2.94, + "learning_rate": 1.4587399547184555e-05, + "loss": 0.0628, + "step": 22466 + }, + { + "epoch": 2.94, + "learning_rate": 1.4586928309489592e-05, + "loss": 0.0469, + "step": 22467 + }, + { + "epoch": 2.94, + "learning_rate": 1.4586457058894137e-05, + "loss": 0.0607, + "step": 22468 + }, + { + "epoch": 2.94, + "learning_rate": 1.4585985795399516e-05, + "loss": 0.0537, + "step": 22469 + }, + { + "epoch": 2.94, + "learning_rate": 1.4585514519007054e-05, + "loss": 0.0488, + "step": 22470 + }, + { + "epoch": 2.94, + "learning_rate": 1.4585043229718074e-05, + "loss": 0.054, + "step": 22471 + }, + { + "epoch": 2.94, + "learning_rate": 1.4584571927533906e-05, + "loss": 0.0576, + "step": 22472 + }, + { + "epoch": 2.94, + "learning_rate": 1.4584100612455873e-05, + "loss": 0.0482, + "step": 22473 + }, + { + "epoch": 2.94, + "learning_rate": 1.4583629284485304e-05, + "loss": 0.0531, + "step": 22474 + }, + { + "epoch": 2.94, + "learning_rate": 1.4583157943623518e-05, + "loss": 0.062, + "step": 22475 + }, + { + "epoch": 2.94, + "learning_rate": 1.4582686589871844e-05, + "loss": 0.0535, + "step": 22476 + }, + { + "epoch": 2.94, + "learning_rate": 1.4582215223231613e-05, + "loss": 0.0487, + "step": 22477 + }, + { + "epoch": 2.94, + "learning_rate": 1.458174384370414e-05, + "loss": 0.0574, + "step": 22478 + }, + { + "epoch": 2.94, + "learning_rate": 1.458127245129076e-05, + "loss": 0.0511, + "step": 22479 + }, + { + "epoch": 2.94, + "learning_rate": 1.4580801045992793e-05, + "loss": 0.0522, + "step": 22480 + }, + { + "epoch": 2.94, + "learning_rate": 1.4580329627811569e-05, + "loss": 0.0602, + "step": 22481 + }, + { + "epoch": 2.94, + "learning_rate": 1.457985819674841e-05, + "loss": 0.0511, + "step": 22482 + }, + { + "epoch": 2.95, + "learning_rate": 1.4579386752804646e-05, + "loss": 0.0556, + "step": 22483 + }, + { + "epoch": 2.95, + "learning_rate": 1.4578915295981599e-05, + "loss": 0.0565, + "step": 22484 + }, + { + "epoch": 2.95, + "learning_rate": 1.4578443826280598e-05, + "loss": 0.061, + "step": 22485 + }, + { + "epoch": 2.95, + "learning_rate": 1.4577972343702966e-05, + "loss": 0.0578, + "step": 22486 + }, + { + "epoch": 2.95, + "learning_rate": 1.4577500848250031e-05, + "loss": 0.0533, + "step": 22487 + }, + { + "epoch": 2.95, + "learning_rate": 1.457702933992312e-05, + "loss": 0.0534, + "step": 22488 + }, + { + "epoch": 2.95, + "learning_rate": 1.4576557818723554e-05, + "loss": 0.0621, + "step": 22489 + }, + { + "epoch": 2.95, + "learning_rate": 1.4576086284652667e-05, + "loss": 0.0572, + "step": 22490 + }, + { + "epoch": 2.95, + "learning_rate": 1.4575614737711778e-05, + "loss": 0.0532, + "step": 22491 + }, + { + "epoch": 2.95, + "learning_rate": 1.4575143177902218e-05, + "loss": 0.06, + "step": 22492 + }, + { + "epoch": 2.95, + "learning_rate": 1.457467160522531e-05, + "loss": 0.0508, + "step": 22493 + }, + { + "epoch": 2.95, + "learning_rate": 1.4574200019682384e-05, + "loss": 0.0592, + "step": 22494 + }, + { + "epoch": 2.95, + "learning_rate": 1.4573728421274763e-05, + "loss": 0.0541, + "step": 22495 + }, + { + "epoch": 2.95, + "learning_rate": 1.4573256810003775e-05, + "loss": 0.0519, + "step": 22496 + }, + { + "epoch": 2.95, + "learning_rate": 1.4572785185870744e-05, + "loss": 0.0646, + "step": 22497 + }, + { + "epoch": 2.95, + "learning_rate": 1.4572313548876999e-05, + "loss": 0.0561, + "step": 22498 + }, + { + "epoch": 2.95, + "learning_rate": 1.4571841899023864e-05, + "loss": 0.0484, + "step": 22499 + }, + { + "epoch": 2.95, + "learning_rate": 1.457137023631267e-05, + "loss": 0.0584, + "step": 22500 + }, + { + "epoch": 2.95, + "learning_rate": 1.457089856074474e-05, + "loss": 0.0519, + "step": 22501 + }, + { + "epoch": 2.95, + "learning_rate": 1.4570426872321399e-05, + "loss": 0.0571, + "step": 22502 + }, + { + "epoch": 2.95, + "learning_rate": 1.4569955171043977e-05, + "loss": 0.0665, + "step": 22503 + }, + { + "epoch": 2.95, + "learning_rate": 1.4569483456913802e-05, + "loss": 0.0465, + "step": 22504 + }, + { + "epoch": 2.95, + "learning_rate": 1.4569011729932195e-05, + "loss": 0.0457, + "step": 22505 + }, + { + "epoch": 2.95, + "learning_rate": 1.4568539990100488e-05, + "loss": 0.0616, + "step": 22506 + }, + { + "epoch": 2.95, + "learning_rate": 1.4568068237420004e-05, + "loss": 0.0573, + "step": 22507 + }, + { + "epoch": 2.95, + "learning_rate": 1.4567596471892072e-05, + "loss": 0.0643, + "step": 22508 + }, + { + "epoch": 2.95, + "learning_rate": 1.4567124693518016e-05, + "loss": 0.0515, + "step": 22509 + }, + { + "epoch": 2.95, + "learning_rate": 1.4566652902299168e-05, + "loss": 0.0485, + "step": 22510 + }, + { + "epoch": 2.95, + "learning_rate": 1.4566181098236851e-05, + "loss": 0.0511, + "step": 22511 + }, + { + "epoch": 2.95, + "learning_rate": 1.4565709281332391e-05, + "loss": 0.0566, + "step": 22512 + }, + { + "epoch": 2.95, + "learning_rate": 1.456523745158712e-05, + "loss": 0.0617, + "step": 22513 + }, + { + "epoch": 2.95, + "learning_rate": 1.4564765609002358e-05, + "loss": 0.0531, + "step": 22514 + }, + { + "epoch": 2.95, + "learning_rate": 1.456429375357944e-05, + "loss": 0.0533, + "step": 22515 + }, + { + "epoch": 2.95, + "learning_rate": 1.4563821885319687e-05, + "loss": 0.0559, + "step": 22516 + }, + { + "epoch": 2.95, + "learning_rate": 1.4563350004224428e-05, + "loss": 0.0481, + "step": 22517 + }, + { + "epoch": 2.95, + "learning_rate": 1.456287811029499e-05, + "loss": 0.0556, + "step": 22518 + }, + { + "epoch": 2.95, + "learning_rate": 1.4562406203532703e-05, + "loss": 0.0502, + "step": 22519 + }, + { + "epoch": 2.95, + "learning_rate": 1.4561934283938886e-05, + "loss": 0.0591, + "step": 22520 + }, + { + "epoch": 2.95, + "learning_rate": 1.4561462351514874e-05, + "loss": 0.0523, + "step": 22521 + }, + { + "epoch": 2.95, + "learning_rate": 1.4560990406261995e-05, + "loss": 0.049, + "step": 22522 + }, + { + "epoch": 2.95, + "learning_rate": 1.4560518448181571e-05, + "loss": 0.0548, + "step": 22523 + }, + { + "epoch": 2.95, + "learning_rate": 1.4560046477274931e-05, + "loss": 0.0546, + "step": 22524 + }, + { + "epoch": 2.95, + "learning_rate": 1.4559574493543402e-05, + "loss": 0.0563, + "step": 22525 + }, + { + "epoch": 2.95, + "learning_rate": 1.4559102496988316e-05, + "loss": 0.0542, + "step": 22526 + }, + { + "epoch": 2.95, + "learning_rate": 1.4558630487610994e-05, + "loss": 0.0435, + "step": 22527 + }, + { + "epoch": 2.95, + "learning_rate": 1.4558158465412769e-05, + "loss": 0.0454, + "step": 22528 + }, + { + "epoch": 2.95, + "learning_rate": 1.4557686430394964e-05, + "loss": 0.0574, + "step": 22529 + }, + { + "epoch": 2.95, + "learning_rate": 1.4557214382558912e-05, + "loss": 0.0557, + "step": 22530 + }, + { + "epoch": 2.95, + "learning_rate": 1.4556742321905933e-05, + "loss": 0.0517, + "step": 22531 + }, + { + "epoch": 2.95, + "learning_rate": 1.4556270248437362e-05, + "loss": 0.0558, + "step": 22532 + }, + { + "epoch": 2.95, + "learning_rate": 1.455579816215452e-05, + "loss": 0.0587, + "step": 22533 + }, + { + "epoch": 2.95, + "learning_rate": 1.4555326063058744e-05, + "loss": 0.0477, + "step": 22534 + }, + { + "epoch": 2.95, + "learning_rate": 1.455485395115135e-05, + "loss": 0.0532, + "step": 22535 + }, + { + "epoch": 2.95, + "learning_rate": 1.4554381826433674e-05, + "loss": 0.0627, + "step": 22536 + }, + { + "epoch": 2.95, + "learning_rate": 1.4553909688907042e-05, + "loss": 0.0527, + "step": 22537 + }, + { + "epoch": 2.95, + "learning_rate": 1.4553437538572784e-05, + "loss": 0.0507, + "step": 22538 + }, + { + "epoch": 2.95, + "learning_rate": 1.4552965375432223e-05, + "loss": 0.0589, + "step": 22539 + }, + { + "epoch": 2.95, + "learning_rate": 1.4552493199486689e-05, + "loss": 0.0543, + "step": 22540 + }, + { + "epoch": 2.95, + "learning_rate": 1.4552021010737508e-05, + "loss": 0.0624, + "step": 22541 + }, + { + "epoch": 2.95, + "learning_rate": 1.4551548809186015e-05, + "loss": 0.0596, + "step": 22542 + }, + { + "epoch": 2.95, + "learning_rate": 1.455107659483353e-05, + "loss": 0.0475, + "step": 22543 + }, + { + "epoch": 2.95, + "learning_rate": 1.4550604367681387e-05, + "loss": 0.0542, + "step": 22544 + }, + { + "epoch": 2.95, + "learning_rate": 1.4550132127730909e-05, + "loss": 0.0514, + "step": 22545 + }, + { + "epoch": 2.95, + "learning_rate": 1.4549659874983429e-05, + "loss": 0.0539, + "step": 22546 + }, + { + "epoch": 2.95, + "learning_rate": 1.4549187609440272e-05, + "loss": 0.0562, + "step": 22547 + }, + { + "epoch": 2.95, + "learning_rate": 1.4548715331102768e-05, + "loss": 0.0515, + "step": 22548 + }, + { + "epoch": 2.95, + "learning_rate": 1.4548243039972244e-05, + "loss": 0.05, + "step": 22549 + }, + { + "epoch": 2.95, + "learning_rate": 1.4547770736050026e-05, + "loss": 0.0547, + "step": 22550 + }, + { + "epoch": 2.95, + "learning_rate": 1.4547298419337449e-05, + "loss": 0.0532, + "step": 22551 + }, + { + "epoch": 2.95, + "learning_rate": 1.4546826089835832e-05, + "loss": 0.0501, + "step": 22552 + }, + { + "epoch": 2.95, + "learning_rate": 1.4546353747546513e-05, + "loss": 0.0506, + "step": 22553 + }, + { + "epoch": 2.95, + "learning_rate": 1.4545881392470816e-05, + "loss": 0.0556, + "step": 22554 + }, + { + "epoch": 2.95, + "learning_rate": 1.4545409024610069e-05, + "loss": 0.0531, + "step": 22555 + }, + { + "epoch": 2.95, + "learning_rate": 1.4544936643965601e-05, + "loss": 0.0537, + "step": 22556 + }, + { + "epoch": 2.95, + "learning_rate": 1.4544464250538743e-05, + "loss": 0.0469, + "step": 22557 + }, + { + "epoch": 2.95, + "learning_rate": 1.4543991844330816e-05, + "loss": 0.0534, + "step": 22558 + }, + { + "epoch": 2.96, + "learning_rate": 1.4543519425343157e-05, + "loss": 0.0622, + "step": 22559 + }, + { + "epoch": 2.96, + "learning_rate": 1.4543046993577093e-05, + "loss": 0.0598, + "step": 22560 + }, + { + "epoch": 2.96, + "learning_rate": 1.4542574549033952e-05, + "loss": 0.0579, + "step": 22561 + }, + { + "epoch": 2.96, + "learning_rate": 1.4542102091715059e-05, + "loss": 0.0519, + "step": 22562 + }, + { + "epoch": 2.96, + "learning_rate": 1.4541629621621746e-05, + "loss": 0.0612, + "step": 22563 + }, + { + "epoch": 2.96, + "learning_rate": 1.4541157138755342e-05, + "loss": 0.0614, + "step": 22564 + }, + { + "epoch": 2.96, + "learning_rate": 1.4540684643117176e-05, + "loss": 0.0584, + "step": 22565 + }, + { + "epoch": 2.96, + "learning_rate": 1.4540212134708576e-05, + "loss": 0.06, + "step": 22566 + }, + { + "epoch": 2.96, + "learning_rate": 1.453973961353087e-05, + "loss": 0.0448, + "step": 22567 + }, + { + "epoch": 2.96, + "learning_rate": 1.453926707958539e-05, + "loss": 0.0514, + "step": 22568 + }, + { + "epoch": 2.96, + "learning_rate": 1.4538794532873461e-05, + "loss": 0.0626, + "step": 22569 + }, + { + "epoch": 2.96, + "learning_rate": 1.4538321973396414e-05, + "loss": 0.0516, + "step": 22570 + }, + { + "epoch": 2.96, + "learning_rate": 1.4537849401155582e-05, + "loss": 0.0541, + "step": 22571 + }, + { + "epoch": 2.96, + "learning_rate": 1.4537376816152286e-05, + "loss": 0.0461, + "step": 22572 + }, + { + "epoch": 2.96, + "learning_rate": 1.4536904218387861e-05, + "loss": 0.0643, + "step": 22573 + }, + { + "epoch": 2.96, + "learning_rate": 1.4536431607863635e-05, + "loss": 0.0581, + "step": 22574 + }, + { + "epoch": 2.96, + "learning_rate": 1.4535958984580933e-05, + "loss": 0.0513, + "step": 22575 + }, + { + "epoch": 2.96, + "learning_rate": 1.4535486348541095e-05, + "loss": 0.0569, + "step": 22576 + }, + { + "epoch": 2.96, + "learning_rate": 1.4535013699745437e-05, + "loss": 0.059, + "step": 22577 + }, + { + "epoch": 2.96, + "learning_rate": 1.4534541038195297e-05, + "loss": 0.0513, + "step": 22578 + }, + { + "epoch": 2.96, + "learning_rate": 1.4534068363892003e-05, + "loss": 0.0624, + "step": 22579 + }, + { + "epoch": 2.96, + "learning_rate": 1.453359567683688e-05, + "loss": 0.0569, + "step": 22580 + }, + { + "epoch": 2.96, + "learning_rate": 1.4533122977031263e-05, + "loss": 0.0525, + "step": 22581 + }, + { + "epoch": 2.96, + "learning_rate": 1.4532650264476477e-05, + "loss": 0.0514, + "step": 22582 + }, + { + "epoch": 2.96, + "learning_rate": 1.4532177539173856e-05, + "loss": 0.052, + "step": 22583 + }, + { + "epoch": 2.96, + "learning_rate": 1.4531704801124727e-05, + "loss": 0.0513, + "step": 22584 + }, + { + "epoch": 2.96, + "learning_rate": 1.4531232050330415e-05, + "loss": 0.0545, + "step": 22585 + }, + { + "epoch": 2.96, + "learning_rate": 1.453075928679226e-05, + "loss": 0.0532, + "step": 22586 + }, + { + "epoch": 2.96, + "learning_rate": 1.4530286510511583e-05, + "loss": 0.0515, + "step": 22587 + }, + { + "epoch": 2.96, + "learning_rate": 1.4529813721489717e-05, + "loss": 0.0583, + "step": 22588 + }, + { + "epoch": 2.96, + "learning_rate": 1.4529340919727992e-05, + "loss": 0.0497, + "step": 22589 + }, + { + "epoch": 2.96, + "learning_rate": 1.4528868105227733e-05, + "loss": 0.0541, + "step": 22590 + }, + { + "epoch": 2.96, + "learning_rate": 1.4528395277990277e-05, + "loss": 0.0469, + "step": 22591 + }, + { + "epoch": 2.96, + "learning_rate": 1.452792243801695e-05, + "loss": 0.0527, + "step": 22592 + }, + { + "epoch": 2.96, + "learning_rate": 1.4527449585309084e-05, + "loss": 0.0615, + "step": 22593 + }, + { + "epoch": 2.96, + "learning_rate": 1.4526976719868002e-05, + "loss": 0.0555, + "step": 22594 + }, + { + "epoch": 2.96, + "learning_rate": 1.4526503841695046e-05, + "loss": 0.0452, + "step": 22595 + }, + { + "epoch": 2.96, + "learning_rate": 1.4526030950791532e-05, + "loss": 0.0594, + "step": 22596 + }, + { + "epoch": 2.96, + "learning_rate": 1.45255580471588e-05, + "loss": 0.0465, + "step": 22597 + }, + { + "epoch": 2.96, + "learning_rate": 1.4525085130798177e-05, + "loss": 0.062, + "step": 22598 + }, + { + "epoch": 2.96, + "learning_rate": 1.4524612201710997e-05, + "loss": 0.0508, + "step": 22599 + }, + { + "epoch": 2.96, + "learning_rate": 1.4524139259898582e-05, + "loss": 0.0507, + "step": 22600 + }, + { + "epoch": 2.96, + "learning_rate": 1.4523666305362267e-05, + "loss": 0.0455, + "step": 22601 + }, + { + "epoch": 2.96, + "learning_rate": 1.4523193338103378e-05, + "loss": 0.0613, + "step": 22602 + }, + { + "epoch": 2.96, + "learning_rate": 1.4522720358123254e-05, + "loss": 0.0543, + "step": 22603 + }, + { + "epoch": 2.96, + "learning_rate": 1.4522247365423217e-05, + "loss": 0.0592, + "step": 22604 + }, + { + "epoch": 2.96, + "learning_rate": 1.4521774360004601e-05, + "loss": 0.052, + "step": 22605 + }, + { + "epoch": 2.96, + "learning_rate": 1.4521301341868734e-05, + "loss": 0.0619, + "step": 22606 + }, + { + "epoch": 2.96, + "learning_rate": 1.452082831101695e-05, + "loss": 0.0579, + "step": 22607 + }, + { + "epoch": 2.96, + "learning_rate": 1.4520355267450574e-05, + "loss": 0.0658, + "step": 22608 + }, + { + "epoch": 2.96, + "learning_rate": 1.4519882211170944e-05, + "loss": 0.051, + "step": 22609 + }, + { + "epoch": 2.96, + "learning_rate": 1.4519409142179381e-05, + "loss": 0.0554, + "step": 22610 + }, + { + "epoch": 2.96, + "learning_rate": 1.4518936060477225e-05, + "loss": 0.0539, + "step": 22611 + }, + { + "epoch": 2.96, + "learning_rate": 1.4518462966065799e-05, + "loss": 0.0523, + "step": 22612 + }, + { + "epoch": 2.96, + "learning_rate": 1.4517989858946436e-05, + "loss": 0.0612, + "step": 22613 + }, + { + "epoch": 2.96, + "learning_rate": 1.451751673912047e-05, + "loss": 0.0594, + "step": 22614 + }, + { + "epoch": 2.96, + "learning_rate": 1.4517043606589228e-05, + "loss": 0.0608, + "step": 22615 + }, + { + "epoch": 2.96, + "learning_rate": 1.451657046135404e-05, + "loss": 0.0514, + "step": 22616 + }, + { + "epoch": 2.96, + "learning_rate": 1.451609730341624e-05, + "loss": 0.0515, + "step": 22617 + }, + { + "epoch": 2.96, + "learning_rate": 1.4515624132777152e-05, + "loss": 0.0473, + "step": 22618 + }, + { + "epoch": 2.96, + "learning_rate": 1.4515150949438117e-05, + "loss": 0.056, + "step": 22619 + }, + { + "epoch": 2.96, + "learning_rate": 1.4514677753400459e-05, + "loss": 0.046, + "step": 22620 + }, + { + "epoch": 2.96, + "learning_rate": 1.4514204544665509e-05, + "loss": 0.0497, + "step": 22621 + }, + { + "epoch": 2.96, + "learning_rate": 1.4513731323234601e-05, + "loss": 0.0522, + "step": 22622 + }, + { + "epoch": 2.96, + "learning_rate": 1.4513258089109061e-05, + "loss": 0.0479, + "step": 22623 + }, + { + "epoch": 2.96, + "learning_rate": 1.4512784842290226e-05, + "loss": 0.0498, + "step": 22624 + }, + { + "epoch": 2.96, + "learning_rate": 1.4512311582779423e-05, + "loss": 0.0531, + "step": 22625 + }, + { + "epoch": 2.96, + "learning_rate": 1.4511838310577986e-05, + "loss": 0.0586, + "step": 22626 + }, + { + "epoch": 2.96, + "learning_rate": 1.451136502568724e-05, + "loss": 0.0505, + "step": 22627 + }, + { + "epoch": 2.96, + "learning_rate": 1.4510891728108521e-05, + "loss": 0.0595, + "step": 22628 + }, + { + "epoch": 2.96, + "learning_rate": 1.4510418417843159e-05, + "loss": 0.0525, + "step": 22629 + }, + { + "epoch": 2.96, + "learning_rate": 1.4509945094892487e-05, + "loss": 0.0466, + "step": 22630 + }, + { + "epoch": 2.96, + "learning_rate": 1.4509471759257836e-05, + "loss": 0.0633, + "step": 22631 + }, + { + "epoch": 2.96, + "learning_rate": 1.4508998410940534e-05, + "loss": 0.0499, + "step": 22632 + }, + { + "epoch": 2.96, + "learning_rate": 1.4508525049941915e-05, + "loss": 0.0584, + "step": 22633 + }, + { + "epoch": 2.96, + "learning_rate": 1.450805167626331e-05, + "loss": 0.0627, + "step": 22634 + }, + { + "epoch": 2.97, + "learning_rate": 1.4507578289906045e-05, + "loss": 0.045, + "step": 22635 + }, + { + "epoch": 2.97, + "learning_rate": 1.4507104890871459e-05, + "loss": 0.051, + "step": 22636 + }, + { + "epoch": 2.97, + "learning_rate": 1.4506631479160883e-05, + "loss": 0.0536, + "step": 22637 + }, + { + "epoch": 2.97, + "learning_rate": 1.4506158054775642e-05, + "loss": 0.0509, + "step": 22638 + }, + { + "epoch": 2.97, + "learning_rate": 1.4505684617717075e-05, + "loss": 0.0582, + "step": 22639 + }, + { + "epoch": 2.97, + "learning_rate": 1.4505211167986506e-05, + "loss": 0.0509, + "step": 22640 + }, + { + "epoch": 2.97, + "learning_rate": 1.4504737705585275e-05, + "loss": 0.0471, + "step": 22641 + }, + { + "epoch": 2.97, + "learning_rate": 1.4504264230514707e-05, + "loss": 0.0565, + "step": 22642 + }, + { + "epoch": 2.97, + "learning_rate": 1.4503790742776137e-05, + "loss": 0.0516, + "step": 22643 + }, + { + "epoch": 2.97, + "learning_rate": 1.450331724237089e-05, + "loss": 0.0422, + "step": 22644 + }, + { + "epoch": 2.97, + "learning_rate": 1.450284372930031e-05, + "loss": 0.0581, + "step": 22645 + }, + { + "epoch": 2.97, + "learning_rate": 1.4502370203565717e-05, + "loss": 0.0537, + "step": 22646 + }, + { + "epoch": 2.97, + "learning_rate": 1.450189666516845e-05, + "loss": 0.0485, + "step": 22647 + }, + { + "epoch": 2.97, + "learning_rate": 1.4501423114109839e-05, + "loss": 0.0393, + "step": 22648 + }, + { + "epoch": 2.97, + "learning_rate": 1.4500949550391212e-05, + "loss": 0.0361, + "step": 22649 + }, + { + "epoch": 2.97, + "learning_rate": 1.4500475974013907e-05, + "loss": 0.0415, + "step": 22650 + }, + { + "epoch": 2.97, + "learning_rate": 1.450000238497925e-05, + "loss": 0.0464, + "step": 22651 + }, + { + "epoch": 2.97, + "learning_rate": 1.4499528783288579e-05, + "loss": 0.0538, + "step": 22652 + }, + { + "epoch": 2.97, + "learning_rate": 1.4499055168943222e-05, + "loss": 0.0546, + "step": 22653 + }, + { + "epoch": 2.97, + "learning_rate": 1.4498581541944512e-05, + "loss": 0.0556, + "step": 22654 + }, + { + "epoch": 2.97, + "learning_rate": 1.4498107902293778e-05, + "loss": 0.0586, + "step": 22655 + }, + { + "epoch": 2.97, + "learning_rate": 1.4497634249992358e-05, + "loss": 0.0563, + "step": 22656 + }, + { + "epoch": 2.97, + "learning_rate": 1.449716058504158e-05, + "loss": 0.0498, + "step": 22657 + }, + { + "epoch": 2.97, + "learning_rate": 1.4496686907442781e-05, + "loss": 0.0489, + "step": 22658 + }, + { + "epoch": 2.97, + "learning_rate": 1.4496213217197285e-05, + "loss": 0.0551, + "step": 22659 + }, + { + "epoch": 2.97, + "learning_rate": 1.4495739514306433e-05, + "loss": 0.0577, + "step": 22660 + }, + { + "epoch": 2.97, + "learning_rate": 1.449526579877155e-05, + "loss": 0.0572, + "step": 22661 + }, + { + "epoch": 2.97, + "learning_rate": 1.4494792070593972e-05, + "loss": 0.0496, + "step": 22662 + }, + { + "epoch": 2.97, + "learning_rate": 1.449431832977503e-05, + "loss": 0.0531, + "step": 22663 + }, + { + "epoch": 2.97, + "learning_rate": 1.449384457631606e-05, + "loss": 0.052, + "step": 22664 + }, + { + "epoch": 2.97, + "learning_rate": 1.4493370810218389e-05, + "loss": 0.0474, + "step": 22665 + }, + { + "epoch": 2.97, + "learning_rate": 1.449289703148335e-05, + "loss": 0.0458, + "step": 22666 + }, + { + "epoch": 2.97, + "learning_rate": 1.4492423240112278e-05, + "loss": 0.049, + "step": 22667 + }, + { + "epoch": 2.97, + "learning_rate": 1.4491949436106509e-05, + "loss": 0.0635, + "step": 22668 + }, + { + "epoch": 2.97, + "learning_rate": 1.4491475619467368e-05, + "loss": 0.0567, + "step": 22669 + }, + { + "epoch": 2.97, + "learning_rate": 1.4491001790196193e-05, + "loss": 0.0538, + "step": 22670 + }, + { + "epoch": 2.97, + "learning_rate": 1.4490527948294313e-05, + "loss": 0.0573, + "step": 22671 + }, + { + "epoch": 2.97, + "learning_rate": 1.4490054093763062e-05, + "loss": 0.0514, + "step": 22672 + }, + { + "epoch": 2.97, + "learning_rate": 1.4489580226603773e-05, + "loss": 0.0577, + "step": 22673 + }, + { + "epoch": 2.97, + "learning_rate": 1.448910634681778e-05, + "loss": 0.0499, + "step": 22674 + }, + { + "epoch": 2.97, + "learning_rate": 1.4488632454406413e-05, + "loss": 0.0471, + "step": 22675 + }, + { + "epoch": 2.97, + "learning_rate": 1.4488158549371005e-05, + "loss": 0.0595, + "step": 22676 + }, + { + "epoch": 2.97, + "learning_rate": 1.4487684631712894e-05, + "loss": 0.0482, + "step": 22677 + }, + { + "epoch": 2.97, + "learning_rate": 1.4487210701433404e-05, + "loss": 0.0552, + "step": 22678 + }, + { + "epoch": 2.97, + "learning_rate": 1.448673675853388e-05, + "loss": 0.0501, + "step": 22679 + }, + { + "epoch": 2.97, + "learning_rate": 1.448626280301564e-05, + "loss": 0.0561, + "step": 22680 + }, + { + "epoch": 2.97, + "learning_rate": 1.4485788834880032e-05, + "loss": 0.0641, + "step": 22681 + }, + { + "epoch": 2.97, + "learning_rate": 1.4485314854128376e-05, + "loss": 0.0459, + "step": 22682 + }, + { + "epoch": 2.97, + "learning_rate": 1.4484840860762013e-05, + "loss": 0.0604, + "step": 22683 + }, + { + "epoch": 2.97, + "learning_rate": 1.4484366854782272e-05, + "loss": 0.0573, + "step": 22684 + }, + { + "epoch": 2.97, + "learning_rate": 1.4483892836190494e-05, + "loss": 0.0457, + "step": 22685 + }, + { + "epoch": 2.97, + "learning_rate": 1.4483418804988e-05, + "loss": 0.047, + "step": 22686 + }, + { + "epoch": 2.97, + "learning_rate": 1.4482944761176133e-05, + "loss": 0.0489, + "step": 22687 + }, + { + "epoch": 2.97, + "learning_rate": 1.4482470704756222e-05, + "loss": 0.0565, + "step": 22688 + }, + { + "epoch": 2.97, + "learning_rate": 1.4481996635729598e-05, + "loss": 0.0538, + "step": 22689 + }, + { + "epoch": 2.97, + "learning_rate": 1.44815225540976e-05, + "loss": 0.0619, + "step": 22690 + }, + { + "epoch": 2.97, + "learning_rate": 1.4481048459861557e-05, + "loss": 0.0539, + "step": 22691 + }, + { + "epoch": 2.97, + "learning_rate": 1.4480574353022804e-05, + "loss": 0.0555, + "step": 22692 + }, + { + "epoch": 2.97, + "learning_rate": 1.4480100233582675e-05, + "loss": 0.0668, + "step": 22693 + }, + { + "epoch": 2.97, + "learning_rate": 1.4479626101542502e-05, + "loss": 0.0463, + "step": 22694 + }, + { + "epoch": 2.97, + "learning_rate": 1.4479151956903622e-05, + "loss": 0.0543, + "step": 22695 + }, + { + "epoch": 2.97, + "learning_rate": 1.447867779966736e-05, + "loss": 0.0542, + "step": 22696 + }, + { + "epoch": 2.97, + "learning_rate": 1.447820362983506e-05, + "loss": 0.0623, + "step": 22697 + }, + { + "epoch": 2.97, + "learning_rate": 1.4477729447408049e-05, + "loss": 0.0545, + "step": 22698 + }, + { + "epoch": 2.97, + "learning_rate": 1.4477255252387661e-05, + "loss": 0.0592, + "step": 22699 + }, + { + "epoch": 2.97, + "learning_rate": 1.4476781044775232e-05, + "loss": 0.0563, + "step": 22700 + }, + { + "epoch": 2.97, + "learning_rate": 1.4476306824572095e-05, + "loss": 0.0489, + "step": 22701 + }, + { + "epoch": 2.97, + "learning_rate": 1.4475832591779586e-05, + "loss": 0.051, + "step": 22702 + }, + { + "epoch": 2.97, + "learning_rate": 1.4475358346399034e-05, + "loss": 0.0575, + "step": 22703 + }, + { + "epoch": 2.97, + "learning_rate": 1.4474884088431775e-05, + "loss": 0.0596, + "step": 22704 + }, + { + "epoch": 2.97, + "learning_rate": 1.447440981787914e-05, + "loss": 0.0522, + "step": 22705 + }, + { + "epoch": 2.97, + "learning_rate": 1.447393553474247e-05, + "loss": 0.0488, + "step": 22706 + }, + { + "epoch": 2.97, + "learning_rate": 1.4473461239023092e-05, + "loss": 0.06, + "step": 22707 + }, + { + "epoch": 2.97, + "learning_rate": 1.4472986930722344e-05, + "loss": 0.0562, + "step": 22708 + }, + { + "epoch": 2.97, + "learning_rate": 1.4472512609841559e-05, + "loss": 0.0435, + "step": 22709 + }, + { + "epoch": 2.97, + "learning_rate": 1.447203827638207e-05, + "loss": 0.0597, + "step": 22710 + }, + { + "epoch": 2.97, + "learning_rate": 1.4471563930345206e-05, + "loss": 0.0475, + "step": 22711 + }, + { + "epoch": 2.98, + "learning_rate": 1.4471089571732316e-05, + "loss": 0.0659, + "step": 22712 + }, + { + "epoch": 2.98, + "learning_rate": 1.4470615200544718e-05, + "loss": 0.0553, + "step": 22713 + }, + { + "epoch": 2.98, + "learning_rate": 1.4470140816783756e-05, + "loss": 0.0529, + "step": 22714 + }, + { + "epoch": 2.98, + "learning_rate": 1.4469666420450758e-05, + "loss": 0.0588, + "step": 22715 + }, + { + "epoch": 2.98, + "learning_rate": 1.4469192011547065e-05, + "loss": 0.0588, + "step": 22716 + }, + { + "epoch": 2.98, + "learning_rate": 1.4468717590074004e-05, + "loss": 0.0498, + "step": 22717 + }, + { + "epoch": 2.98, + "learning_rate": 1.4468243156032912e-05, + "loss": 0.0479, + "step": 22718 + }, + { + "epoch": 2.98, + "learning_rate": 1.4467768709425128e-05, + "loss": 0.0522, + "step": 22719 + }, + { + "epoch": 2.98, + "learning_rate": 1.4467294250251983e-05, + "loss": 0.0623, + "step": 22720 + }, + { + "epoch": 2.98, + "learning_rate": 1.4466819778514806e-05, + "loss": 0.0524, + "step": 22721 + }, + { + "epoch": 2.98, + "learning_rate": 1.4466345294214938e-05, + "loss": 0.058, + "step": 22722 + }, + { + "epoch": 2.98, + "learning_rate": 1.4465870797353714e-05, + "loss": 0.0568, + "step": 22723 + }, + { + "epoch": 2.98, + "learning_rate": 1.4465396287932465e-05, + "loss": 0.053, + "step": 22724 + }, + { + "epoch": 2.98, + "learning_rate": 1.4464921765952526e-05, + "loss": 0.0469, + "step": 22725 + }, + { + "epoch": 2.98, + "learning_rate": 1.4464447231415232e-05, + "loss": 0.0507, + "step": 22726 + }, + { + "epoch": 2.98, + "learning_rate": 1.4463972684321919e-05, + "loss": 0.0438, + "step": 22727 + }, + { + "epoch": 2.98, + "learning_rate": 1.446349812467392e-05, + "loss": 0.0467, + "step": 22728 + }, + { + "epoch": 2.98, + "learning_rate": 1.4463023552472573e-05, + "loss": 0.0489, + "step": 22729 + }, + { + "epoch": 2.98, + "learning_rate": 1.4462548967719205e-05, + "loss": 0.0564, + "step": 22730 + }, + { + "epoch": 2.98, + "learning_rate": 1.446207437041516e-05, + "loss": 0.0478, + "step": 22731 + }, + { + "epoch": 2.98, + "learning_rate": 1.4461599760561767e-05, + "loss": 0.0478, + "step": 22732 + }, + { + "epoch": 2.98, + "learning_rate": 1.4461125138160358e-05, + "loss": 0.0487, + "step": 22733 + }, + { + "epoch": 2.98, + "learning_rate": 1.4460650503212277e-05, + "loss": 0.0469, + "step": 22734 + }, + { + "epoch": 2.98, + "learning_rate": 1.4460175855718854e-05, + "loss": 0.0484, + "step": 22735 + }, + { + "epoch": 2.98, + "learning_rate": 1.4459701195681422e-05, + "loss": 0.0448, + "step": 22736 + }, + { + "epoch": 2.98, + "learning_rate": 1.4459226523101317e-05, + "loss": 0.052, + "step": 22737 + }, + { + "epoch": 2.98, + "learning_rate": 1.4458751837979877e-05, + "loss": 0.0478, + "step": 22738 + }, + { + "epoch": 2.98, + "learning_rate": 1.4458277140318435e-05, + "loss": 0.0443, + "step": 22739 + }, + { + "epoch": 2.98, + "learning_rate": 1.4457802430118326e-05, + "loss": 0.0518, + "step": 22740 + }, + { + "epoch": 2.98, + "learning_rate": 1.4457327707380884e-05, + "loss": 0.043, + "step": 22741 + }, + { + "epoch": 2.98, + "learning_rate": 1.4456852972107448e-05, + "loss": 0.0479, + "step": 22742 + }, + { + "epoch": 2.98, + "learning_rate": 1.4456378224299345e-05, + "loss": 0.0537, + "step": 22743 + }, + { + "epoch": 2.98, + "learning_rate": 1.445590346395792e-05, + "loss": 0.0539, + "step": 22744 + }, + { + "epoch": 2.98, + "learning_rate": 1.44554286910845e-05, + "loss": 0.0544, + "step": 22745 + }, + { + "epoch": 2.98, + "learning_rate": 1.4454953905680429e-05, + "loss": 0.0586, + "step": 22746 + }, + { + "epoch": 2.98, + "learning_rate": 1.4454479107747036e-05, + "loss": 0.0661, + "step": 22747 + }, + { + "epoch": 2.98, + "learning_rate": 1.4454004297285656e-05, + "loss": 0.0515, + "step": 22748 + }, + { + "epoch": 2.98, + "learning_rate": 1.4453529474297624e-05, + "loss": 0.0486, + "step": 22749 + }, + { + "epoch": 2.98, + "learning_rate": 1.4453054638784281e-05, + "loss": 0.0521, + "step": 22750 + }, + { + "epoch": 2.98, + "learning_rate": 1.445257979074696e-05, + "loss": 0.0589, + "step": 22751 + }, + { + "epoch": 2.98, + "learning_rate": 1.4452104930186993e-05, + "loss": 0.064, + "step": 22752 + }, + { + "epoch": 2.98, + "learning_rate": 1.4451630057105719e-05, + "loss": 0.0516, + "step": 22753 + }, + { + "epoch": 2.98, + "learning_rate": 1.445115517150447e-05, + "loss": 0.0449, + "step": 22754 + }, + { + "epoch": 2.98, + "learning_rate": 1.4450680273384587e-05, + "loss": 0.0534, + "step": 22755 + }, + { + "epoch": 2.98, + "learning_rate": 1.4450205362747402e-05, + "loss": 0.0528, + "step": 22756 + }, + { + "epoch": 2.98, + "learning_rate": 1.4449730439594252e-05, + "loss": 0.0471, + "step": 22757 + }, + { + "epoch": 2.98, + "learning_rate": 1.4449255503926471e-05, + "loss": 0.0549, + "step": 22758 + }, + { + "epoch": 2.98, + "learning_rate": 1.4448780555745396e-05, + "loss": 0.0534, + "step": 22759 + }, + { + "epoch": 2.98, + "learning_rate": 1.4448305595052364e-05, + "loss": 0.06, + "step": 22760 + }, + { + "epoch": 2.98, + "learning_rate": 1.4447830621848706e-05, + "loss": 0.0587, + "step": 22761 + }, + { + "epoch": 2.98, + "learning_rate": 1.4447355636135765e-05, + "loss": 0.0499, + "step": 22762 + }, + { + "epoch": 2.98, + "learning_rate": 1.444688063791487e-05, + "loss": 0.0485, + "step": 22763 + }, + { + "epoch": 2.98, + "learning_rate": 1.4446405627187363e-05, + "loss": 0.0537, + "step": 22764 + }, + { + "epoch": 2.98, + "learning_rate": 1.4445930603954575e-05, + "loss": 0.0565, + "step": 22765 + }, + { + "epoch": 2.98, + "learning_rate": 1.4445455568217844e-05, + "loss": 0.0566, + "step": 22766 + }, + { + "epoch": 2.98, + "learning_rate": 1.4444980519978508e-05, + "loss": 0.0502, + "step": 22767 + }, + { + "epoch": 2.98, + "learning_rate": 1.4444505459237898e-05, + "loss": 0.0484, + "step": 22768 + }, + { + "epoch": 2.98, + "learning_rate": 1.4444030385997356e-05, + "loss": 0.0577, + "step": 22769 + }, + { + "epoch": 2.98, + "learning_rate": 1.4443555300258215e-05, + "loss": 0.0633, + "step": 22770 + }, + { + "epoch": 2.98, + "learning_rate": 1.444308020202181e-05, + "loss": 0.0506, + "step": 22771 + }, + { + "epoch": 2.98, + "learning_rate": 1.4442605091289477e-05, + "loss": 0.0609, + "step": 22772 + }, + { + "epoch": 2.98, + "learning_rate": 1.4442129968062557e-05, + "loss": 0.0503, + "step": 22773 + }, + { + "epoch": 2.98, + "learning_rate": 1.444165483234238e-05, + "loss": 0.0585, + "step": 22774 + }, + { + "epoch": 2.98, + "learning_rate": 1.444117968413029e-05, + "loss": 0.0511, + "step": 22775 + }, + { + "epoch": 2.98, + "learning_rate": 1.4440704523427611e-05, + "loss": 0.0554, + "step": 22776 + }, + { + "epoch": 2.98, + "learning_rate": 1.4440229350235693e-05, + "loss": 0.0499, + "step": 22777 + }, + { + "epoch": 2.98, + "learning_rate": 1.4439754164555863e-05, + "loss": 0.0607, + "step": 22778 + }, + { + "epoch": 2.98, + "learning_rate": 1.4439278966389463e-05, + "loss": 0.0542, + "step": 22779 + }, + { + "epoch": 2.98, + "learning_rate": 1.4438803755737828e-05, + "loss": 0.0565, + "step": 22780 + }, + { + "epoch": 2.98, + "learning_rate": 1.4438328532602293e-05, + "loss": 0.0619, + "step": 22781 + }, + { + "epoch": 2.98, + "learning_rate": 1.4437853296984194e-05, + "loss": 0.0559, + "step": 22782 + }, + { + "epoch": 2.98, + "learning_rate": 1.443737804888487e-05, + "loss": 0.0577, + "step": 22783 + }, + { + "epoch": 2.98, + "learning_rate": 1.4436902788305657e-05, + "loss": 0.0513, + "step": 22784 + }, + { + "epoch": 2.98, + "learning_rate": 1.443642751524789e-05, + "loss": 0.0582, + "step": 22785 + }, + { + "epoch": 2.98, + "learning_rate": 1.443595222971291e-05, + "loss": 0.0549, + "step": 22786 + }, + { + "epoch": 2.98, + "learning_rate": 1.4435476931702046e-05, + "loss": 0.0529, + "step": 22787 + }, + { + "epoch": 2.99, + "learning_rate": 1.4435001621216643e-05, + "loss": 0.057, + "step": 22788 + }, + { + "epoch": 2.99, + "learning_rate": 1.443452629825803e-05, + "loss": 0.0516, + "step": 22789 + }, + { + "epoch": 2.99, + "learning_rate": 1.4434050962827555e-05, + "loss": 0.0577, + "step": 22790 + }, + { + "epoch": 2.99, + "learning_rate": 1.4433575614926543e-05, + "loss": 0.0496, + "step": 22791 + }, + { + "epoch": 2.99, + "learning_rate": 1.4433100254556336e-05, + "loss": 0.0558, + "step": 22792 + }, + { + "epoch": 2.99, + "learning_rate": 1.443262488171827e-05, + "loss": 0.0449, + "step": 22793 + }, + { + "epoch": 2.99, + "learning_rate": 1.4432149496413684e-05, + "loss": 0.0543, + "step": 22794 + }, + { + "epoch": 2.99, + "learning_rate": 1.443167409864391e-05, + "loss": 0.0602, + "step": 22795 + }, + { + "epoch": 2.99, + "learning_rate": 1.4431198688410295e-05, + "loss": 0.0615, + "step": 22796 + }, + { + "epoch": 2.99, + "learning_rate": 1.4430723265714166e-05, + "loss": 0.0561, + "step": 22797 + }, + { + "epoch": 2.99, + "learning_rate": 1.4430247830556862e-05, + "loss": 0.0466, + "step": 22798 + }, + { + "epoch": 2.99, + "learning_rate": 1.4429772382939723e-05, + "loss": 0.0524, + "step": 22799 + }, + { + "epoch": 2.99, + "learning_rate": 1.4429296922864087e-05, + "loss": 0.0479, + "step": 22800 + }, + { + "epoch": 2.99, + "learning_rate": 1.4428821450331288e-05, + "loss": 0.0632, + "step": 22801 + }, + { + "epoch": 2.99, + "learning_rate": 1.4428345965342666e-05, + "loss": 0.0553, + "step": 22802 + }, + { + "epoch": 2.99, + "learning_rate": 1.4427870467899556e-05, + "loss": 0.0445, + "step": 22803 + }, + { + "epoch": 2.99, + "learning_rate": 1.4427394958003293e-05, + "loss": 0.0444, + "step": 22804 + }, + { + "epoch": 2.99, + "learning_rate": 1.442691943565522e-05, + "loss": 0.0444, + "step": 22805 + }, + { + "epoch": 2.99, + "learning_rate": 1.4426443900856672e-05, + "loss": 0.0518, + "step": 22806 + }, + { + "epoch": 2.99, + "learning_rate": 1.4425968353608987e-05, + "loss": 0.0456, + "step": 22807 + }, + { + "epoch": 2.99, + "learning_rate": 1.4425492793913501e-05, + "loss": 0.0445, + "step": 22808 + }, + { + "epoch": 2.99, + "learning_rate": 1.4425017221771552e-05, + "loss": 0.0605, + "step": 22809 + }, + { + "epoch": 2.99, + "learning_rate": 1.4424541637184478e-05, + "loss": 0.044, + "step": 22810 + }, + { + "epoch": 2.99, + "learning_rate": 1.4424066040153616e-05, + "loss": 0.0458, + "step": 22811 + }, + { + "epoch": 2.99, + "learning_rate": 1.4423590430680303e-05, + "loss": 0.0487, + "step": 22812 + }, + { + "epoch": 2.99, + "learning_rate": 1.4423114808765878e-05, + "loss": 0.0475, + "step": 22813 + }, + { + "epoch": 2.99, + "learning_rate": 1.4422639174411676e-05, + "loss": 0.0448, + "step": 22814 + }, + { + "epoch": 2.99, + "learning_rate": 1.4422163527619041e-05, + "loss": 0.0525, + "step": 22815 + }, + { + "epoch": 2.99, + "learning_rate": 1.4421687868389305e-05, + "loss": 0.0676, + "step": 22816 + }, + { + "epoch": 2.99, + "learning_rate": 1.4421212196723808e-05, + "loss": 0.0659, + "step": 22817 + }, + { + "epoch": 2.99, + "learning_rate": 1.4420736512623888e-05, + "loss": 0.0491, + "step": 22818 + }, + { + "epoch": 2.99, + "learning_rate": 1.4420260816090878e-05, + "loss": 0.0516, + "step": 22819 + }, + { + "epoch": 2.99, + "learning_rate": 1.4419785107126123e-05, + "loss": 0.059, + "step": 22820 + }, + { + "epoch": 2.99, + "learning_rate": 1.4419309385730958e-05, + "loss": 0.0483, + "step": 22821 + }, + { + "epoch": 2.99, + "learning_rate": 1.441883365190672e-05, + "loss": 0.0563, + "step": 22822 + }, + { + "epoch": 2.99, + "learning_rate": 1.4418357905654745e-05, + "loss": 0.0553, + "step": 22823 + }, + { + "epoch": 2.99, + "learning_rate": 1.4417882146976378e-05, + "loss": 0.0482, + "step": 22824 + }, + { + "epoch": 2.99, + "learning_rate": 1.4417406375872949e-05, + "loss": 0.0566, + "step": 22825 + }, + { + "epoch": 2.99, + "learning_rate": 1.4416930592345801e-05, + "loss": 0.0519, + "step": 22826 + }, + { + "epoch": 2.99, + "learning_rate": 1.4416454796396272e-05, + "loss": 0.0597, + "step": 22827 + }, + { + "epoch": 2.99, + "learning_rate": 1.44159789880257e-05, + "loss": 0.0466, + "step": 22828 + }, + { + "epoch": 2.99, + "learning_rate": 1.4415503167235421e-05, + "loss": 0.0506, + "step": 22829 + }, + { + "epoch": 2.99, + "learning_rate": 1.4415027334026774e-05, + "loss": 0.056, + "step": 22830 + }, + { + "epoch": 2.99, + "learning_rate": 1.4414551488401097e-05, + "loss": 0.0489, + "step": 22831 + }, + { + "epoch": 2.99, + "learning_rate": 1.441407563035973e-05, + "loss": 0.0661, + "step": 22832 + }, + { + "epoch": 2.99, + "learning_rate": 1.4413599759904009e-05, + "loss": 0.0541, + "step": 22833 + }, + { + "epoch": 2.99, + "learning_rate": 1.4413123877035278e-05, + "loss": 0.0508, + "step": 22834 + }, + { + "epoch": 2.99, + "learning_rate": 1.4412647981754866e-05, + "loss": 0.0478, + "step": 22835 + }, + { + "epoch": 2.99, + "learning_rate": 1.441217207406412e-05, + "loss": 0.0549, + "step": 22836 + }, + { + "epoch": 2.99, + "learning_rate": 1.4411696153964371e-05, + "loss": 0.0515, + "step": 22837 + }, + { + "epoch": 2.99, + "learning_rate": 1.4411220221456967e-05, + "loss": 0.0447, + "step": 22838 + }, + { + "epoch": 2.99, + "learning_rate": 1.4410744276543238e-05, + "loss": 0.0516, + "step": 22839 + }, + { + "epoch": 2.99, + "learning_rate": 1.4410268319224528e-05, + "loss": 0.0464, + "step": 22840 + }, + { + "epoch": 2.99, + "learning_rate": 1.440979234950217e-05, + "loss": 0.0441, + "step": 22841 + }, + { + "epoch": 2.99, + "learning_rate": 1.4409316367377507e-05, + "loss": 0.0512, + "step": 22842 + }, + { + "epoch": 2.99, + "learning_rate": 1.4408840372851875e-05, + "loss": 0.0564, + "step": 22843 + }, + { + "epoch": 2.99, + "learning_rate": 1.4408364365926614e-05, + "loss": 0.0538, + "step": 22844 + }, + { + "epoch": 2.99, + "learning_rate": 1.4407888346603067e-05, + "loss": 0.0524, + "step": 22845 + }, + { + "epoch": 2.99, + "learning_rate": 1.4407412314882567e-05, + "loss": 0.0533, + "step": 22846 + }, + { + "epoch": 2.99, + "learning_rate": 1.4406936270766451e-05, + "loss": 0.0515, + "step": 22847 + }, + { + "epoch": 2.99, + "learning_rate": 1.4406460214256063e-05, + "loss": 0.0411, + "step": 22848 + }, + { + "epoch": 2.99, + "learning_rate": 1.4405984145352745e-05, + "loss": 0.0479, + "step": 22849 + }, + { + "epoch": 2.99, + "learning_rate": 1.4405508064057827e-05, + "loss": 0.052, + "step": 22850 + }, + { + "epoch": 2.99, + "learning_rate": 1.4405031970372655e-05, + "loss": 0.0505, + "step": 22851 + }, + { + "epoch": 2.99, + "learning_rate": 1.4404555864298563e-05, + "loss": 0.0596, + "step": 22852 + }, + { + "epoch": 2.99, + "learning_rate": 1.4404079745836892e-05, + "loss": 0.0664, + "step": 22853 + }, + { + "epoch": 2.99, + "learning_rate": 1.4403603614988982e-05, + "loss": 0.05, + "step": 22854 + }, + { + "epoch": 2.99, + "learning_rate": 1.440312747175617e-05, + "loss": 0.0551, + "step": 22855 + }, + { + "epoch": 2.99, + "learning_rate": 1.44026513161398e-05, + "loss": 0.0598, + "step": 22856 + }, + { + "epoch": 2.99, + "learning_rate": 1.4402175148141205e-05, + "loss": 0.0563, + "step": 22857 + }, + { + "epoch": 2.99, + "learning_rate": 1.4401698967761726e-05, + "loss": 0.0524, + "step": 22858 + }, + { + "epoch": 2.99, + "learning_rate": 1.4401222775002704e-05, + "loss": 0.0543, + "step": 22859 + }, + { + "epoch": 2.99, + "learning_rate": 1.4400746569865476e-05, + "loss": 0.0545, + "step": 22860 + }, + { + "epoch": 2.99, + "learning_rate": 1.4400270352351386e-05, + "loss": 0.0575, + "step": 22861 + }, + { + "epoch": 2.99, + "learning_rate": 1.4399794122461768e-05, + "loss": 0.0508, + "step": 22862 + }, + { + "epoch": 2.99, + "learning_rate": 1.4399317880197963e-05, + "loss": 0.0641, + "step": 22863 + }, + { + "epoch": 3.0, + "learning_rate": 1.439884162556131e-05, + "loss": 0.0523, + "step": 22864 + }, + { + "epoch": 3.0, + "learning_rate": 1.4398365358553153e-05, + "loss": 0.0496, + "step": 22865 + }, + { + "epoch": 3.0, + "learning_rate": 1.4397889079174822e-05, + "loss": 0.0626, + "step": 22866 + }, + { + "epoch": 3.0, + "learning_rate": 1.4397412787427666e-05, + "loss": 0.0519, + "step": 22867 + }, + { + "epoch": 3.0, + "learning_rate": 1.439693648331302e-05, + "loss": 0.0566, + "step": 22868 + }, + { + "epoch": 3.0, + "learning_rate": 1.4396460166832222e-05, + "loss": 0.0563, + "step": 22869 + }, + { + "epoch": 3.0, + "learning_rate": 1.4395983837986618e-05, + "loss": 0.044, + "step": 22870 + }, + { + "epoch": 3.0, + "learning_rate": 1.4395507496777541e-05, + "loss": 0.0597, + "step": 22871 + }, + { + "epoch": 3.0, + "learning_rate": 1.4395031143206336e-05, + "loss": 0.0559, + "step": 22872 + }, + { + "epoch": 3.0, + "learning_rate": 1.4394554777274334e-05, + "loss": 0.0588, + "step": 22873 + }, + { + "epoch": 3.0, + "learning_rate": 1.4394078398982884e-05, + "loss": 0.0573, + "step": 22874 + }, + { + "epoch": 3.0, + "learning_rate": 1.4393602008333325e-05, + "loss": 0.0531, + "step": 22875 + }, + { + "epoch": 3.0, + "learning_rate": 1.4393125605326991e-05, + "loss": 0.0529, + "step": 22876 + }, + { + "epoch": 3.0, + "learning_rate": 1.4392649189965227e-05, + "loss": 0.0427, + "step": 22877 + }, + { + "epoch": 3.0, + "learning_rate": 1.4392172762249372e-05, + "loss": 0.0496, + "step": 22878 + }, + { + "epoch": 3.0, + "learning_rate": 1.4391696322180764e-05, + "loss": 0.0496, + "step": 22879 + }, + { + "epoch": 3.0, + "learning_rate": 1.4391219869760743e-05, + "loss": 0.0537, + "step": 22880 + }, + { + "epoch": 3.0, + "learning_rate": 1.4390743404990649e-05, + "loss": 0.0491, + "step": 22881 + }, + { + "epoch": 3.0, + "learning_rate": 1.4390266927871826e-05, + "loss": 0.0465, + "step": 22882 + }, + { + "epoch": 3.0, + "learning_rate": 1.4389790438405609e-05, + "loss": 0.0592, + "step": 22883 + }, + { + "epoch": 3.0, + "learning_rate": 1.4389313936593339e-05, + "loss": 0.0517, + "step": 22884 + }, + { + "epoch": 3.0, + "learning_rate": 1.4388837422436359e-05, + "loss": 0.0594, + "step": 22885 + }, + { + "epoch": 3.0, + "learning_rate": 1.4388360895936008e-05, + "loss": 0.0626, + "step": 22886 + }, + { + "epoch": 3.0, + "learning_rate": 1.4387884357093627e-05, + "loss": 0.0506, + "step": 22887 + }, + { + "epoch": 3.0, + "learning_rate": 1.4387407805910551e-05, + "loss": 0.0531, + "step": 22888 + }, + { + "epoch": 3.0, + "learning_rate": 1.438693124238813e-05, + "loss": 0.0571, + "step": 22889 + }, + { + "epoch": 3.0, + "learning_rate": 1.4386454666527691e-05, + "loss": 0.0535, + "step": 22890 + }, + { + "epoch": 3.0, + "learning_rate": 1.4385978078330585e-05, + "loss": 0.048, + "step": 22891 + }, + { + "epoch": 3.0, + "learning_rate": 1.4385501477798149e-05, + "loss": 0.0485, + "step": 22892 + }, + { + "epoch": 3.0, + "learning_rate": 1.4385024864931727e-05, + "loss": 0.0449, + "step": 22893 + }, + { + "epoch": 3.0, + "learning_rate": 1.4384548239732652e-05, + "loss": 0.0661, + "step": 22894 + }, + { + "epoch": 3.0, + "learning_rate": 1.4384071602202272e-05, + "loss": 0.05, + "step": 22895 + }, + { + "epoch": 3.0, + "learning_rate": 1.4383594952341919e-05, + "loss": 0.0475, + "step": 22896 + }, + { + "epoch": 3.0, + "learning_rate": 1.4383118290152944e-05, + "loss": 0.0491, + "step": 22897 + }, + { + "epoch": 3.0, + "learning_rate": 1.4382641615636679e-05, + "loss": 0.0474, + "step": 22898 + }, + { + "epoch": 3.0, + "learning_rate": 1.438216492879447e-05, + "loss": 0.053, + "step": 22899 + }, + { + "epoch": 3.0, + "learning_rate": 1.4381688229627654e-05, + "loss": 0.0504, + "step": 22900 + }, + { + "epoch": 3.0, + "learning_rate": 1.4381211518137574e-05, + "loss": 0.0508, + "step": 22901 + }, + { + "epoch": 3.0, + "learning_rate": 1.4380734794325569e-05, + "loss": 0.0523, + "step": 22902 + }, + { + "epoch": 3.0, + "learning_rate": 1.4380258058192982e-05, + "loss": 0.0441, + "step": 22903 + }, + { + "epoch": 3.0, + "learning_rate": 1.437978130974115e-05, + "loss": 0.0399, + "step": 22904 + }, + { + "epoch": 3.0, + "learning_rate": 1.4379304548971419e-05, + "loss": 0.0419, + "step": 22905 + }, + { + "epoch": 3.0, + "learning_rate": 1.4378827775885126e-05, + "loss": 0.0467, + "step": 22906 + }, + { + "epoch": 3.0, + "learning_rate": 1.4378350990483614e-05, + "loss": 0.048, + "step": 22907 + }, + { + "epoch": 3.0, + "learning_rate": 1.437787419276822e-05, + "loss": 0.057, + "step": 22908 + }, + { + "epoch": 3.0, + "learning_rate": 1.437739738274029e-05, + "loss": 0.0428, + "step": 22909 + }, + { + "epoch": 3.0, + "learning_rate": 1.4376920560401165e-05, + "loss": 0.0523, + "step": 22910 + }, + { + "epoch": 3.0, + "learning_rate": 1.437644372575218e-05, + "loss": 0.0424, + "step": 22911 + }, + { + "epoch": 3.0, + "learning_rate": 1.4375966878794685e-05, + "loss": 0.0403, + "step": 22912 + }, + { + "epoch": 3.0, + "learning_rate": 1.4375490019530012e-05, + "loss": 0.041, + "step": 22913 + }, + { + "epoch": 3.0, + "learning_rate": 1.437501314795951e-05, + "loss": 0.0443, + "step": 22914 + }, + { + "epoch": 3.0, + "learning_rate": 1.4374536264084511e-05, + "loss": 0.0494, + "step": 22915 + }, + { + "epoch": 3.0, + "learning_rate": 1.4374059367906368e-05, + "loss": 0.0416, + "step": 22916 + }, + { + "epoch": 3.0, + "learning_rate": 1.4373582459426412e-05, + "loss": 0.039, + "step": 22917 + }, + { + "epoch": 3.0, + "learning_rate": 1.4373105538645988e-05, + "loss": 0.0381, + "step": 22918 + }, + { + "epoch": 3.0, + "learning_rate": 1.437262860556644e-05, + "loss": 0.0482, + "step": 22919 + }, + { + "epoch": 3.0, + "learning_rate": 1.4372151660189104e-05, + "loss": 0.0428, + "step": 22920 + }, + { + "epoch": 3.0, + "learning_rate": 1.4371674702515325e-05, + "loss": 0.0434, + "step": 22921 + }, + { + "epoch": 3.0, + "learning_rate": 1.4371197732546447e-05, + "loss": 0.0448, + "step": 22922 + }, + { + "epoch": 3.0, + "learning_rate": 1.4370720750283805e-05, + "loss": 0.0404, + "step": 22923 + }, + { + "epoch": 3.0, + "learning_rate": 1.4370243755728743e-05, + "loss": 0.0441, + "step": 22924 + }, + { + "epoch": 3.0, + "learning_rate": 1.4369766748882606e-05, + "loss": 0.048, + "step": 22925 + }, + { + "epoch": 3.0, + "learning_rate": 1.436928972974673e-05, + "loss": 0.048, + "step": 22926 + }, + { + "epoch": 3.0, + "learning_rate": 1.4368812698322461e-05, + "loss": 0.0418, + "step": 22927 + }, + { + "epoch": 3.0, + "learning_rate": 1.4368335654611138e-05, + "loss": 0.0456, + "step": 22928 + }, + { + "epoch": 3.0, + "learning_rate": 1.4367858598614103e-05, + "loss": 0.0457, + "step": 22929 + }, + { + "epoch": 3.0, + "learning_rate": 1.43673815303327e-05, + "loss": 0.04, + "step": 22930 + }, + { + "epoch": 3.0, + "learning_rate": 1.4366904449768272e-05, + "loss": 0.0401, + "step": 22931 + }, + { + "epoch": 3.0, + "learning_rate": 1.4366427356922153e-05, + "loss": 0.0415, + "step": 22932 + }, + { + "epoch": 3.0, + "learning_rate": 1.4365950251795691e-05, + "loss": 0.0439, + "step": 22933 + }, + { + "epoch": 3.0, + "learning_rate": 1.4365473134390225e-05, + "loss": 0.0377, + "step": 22934 + }, + { + "epoch": 3.0, + "learning_rate": 1.4364996004707099e-05, + "loss": 0.0363, + "step": 22935 + }, + { + "epoch": 3.0, + "learning_rate": 1.4364518862747653e-05, + "loss": 0.0492, + "step": 22936 + }, + { + "epoch": 3.0, + "learning_rate": 1.4364041708513235e-05, + "loss": 0.0394, + "step": 22937 + }, + { + "epoch": 3.0, + "learning_rate": 1.4363564542005177e-05, + "loss": 0.0456, + "step": 22938 + }, + { + "epoch": 3.0, + "learning_rate": 1.436308736322483e-05, + "loss": 0.0482, + "step": 22939 + }, + { + "epoch": 3.0, + "learning_rate": 1.4362610172173529e-05, + "loss": 0.0493, + "step": 22940 + }, + { + "epoch": 3.01, + "learning_rate": 1.436213296885262e-05, + "loss": 0.044, + "step": 22941 + }, + { + "epoch": 3.01, + "learning_rate": 1.4361655753263444e-05, + "loss": 0.0328, + "step": 22942 + }, + { + "epoch": 3.01, + "learning_rate": 1.4361178525407346e-05, + "loss": 0.0398, + "step": 22943 + }, + { + "epoch": 3.01, + "learning_rate": 1.4360701285285663e-05, + "loss": 0.0482, + "step": 22944 + }, + { + "epoch": 3.01, + "learning_rate": 1.436022403289974e-05, + "loss": 0.0428, + "step": 22945 + }, + { + "epoch": 3.01, + "learning_rate": 1.435974676825092e-05, + "loss": 0.0443, + "step": 22946 + }, + { + "epoch": 3.01, + "learning_rate": 1.4359269491340543e-05, + "loss": 0.0458, + "step": 22947 + }, + { + "epoch": 3.01, + "learning_rate": 1.4358792202169956e-05, + "loss": 0.0429, + "step": 22948 + }, + { + "epoch": 3.01, + "learning_rate": 1.4358314900740494e-05, + "loss": 0.037, + "step": 22949 + }, + { + "epoch": 3.01, + "learning_rate": 1.4357837587053507e-05, + "loss": 0.0467, + "step": 22950 + }, + { + "epoch": 3.01, + "learning_rate": 1.4357360261110332e-05, + "loss": 0.0436, + "step": 22951 + }, + { + "epoch": 3.01, + "learning_rate": 1.4356882922912312e-05, + "loss": 0.0448, + "step": 22952 + }, + { + "epoch": 3.01, + "learning_rate": 1.4356405572460792e-05, + "loss": 0.0514, + "step": 22953 + }, + { + "epoch": 3.01, + "learning_rate": 1.4355928209757116e-05, + "loss": 0.0439, + "step": 22954 + }, + { + "epoch": 3.01, + "learning_rate": 1.435545083480262e-05, + "loss": 0.048, + "step": 22955 + }, + { + "epoch": 3.01, + "learning_rate": 1.435497344759865e-05, + "loss": 0.0487, + "step": 22956 + }, + { + "epoch": 3.01, + "learning_rate": 1.435449604814655e-05, + "loss": 0.0363, + "step": 22957 + }, + { + "epoch": 3.01, + "learning_rate": 1.4354018636447666e-05, + "loss": 0.0483, + "step": 22958 + }, + { + "epoch": 3.01, + "learning_rate": 1.4353541212503332e-05, + "loss": 0.0359, + "step": 22959 + }, + { + "epoch": 3.01, + "learning_rate": 1.4353063776314899e-05, + "loss": 0.0412, + "step": 22960 + }, + { + "epoch": 3.01, + "learning_rate": 1.4352586327883702e-05, + "loss": 0.0499, + "step": 22961 + }, + { + "epoch": 3.01, + "learning_rate": 1.4352108867211088e-05, + "loss": 0.0412, + "step": 22962 + }, + { + "epoch": 3.01, + "learning_rate": 1.43516313942984e-05, + "loss": 0.0427, + "step": 22963 + }, + { + "epoch": 3.01, + "learning_rate": 1.4351153909146983e-05, + "loss": 0.0487, + "step": 22964 + }, + { + "epoch": 3.01, + "learning_rate": 1.4350676411758175e-05, + "loss": 0.044, + "step": 22965 + }, + { + "epoch": 3.01, + "learning_rate": 1.4350198902133323e-05, + "loss": 0.0382, + "step": 22966 + }, + { + "epoch": 3.01, + "learning_rate": 1.4349721380273765e-05, + "loss": 0.0466, + "step": 22967 + }, + { + "epoch": 3.01, + "learning_rate": 1.4349243846180851e-05, + "loss": 0.0421, + "step": 22968 + }, + { + "epoch": 3.01, + "learning_rate": 1.4348766299855919e-05, + "loss": 0.0452, + "step": 22969 + }, + { + "epoch": 3.01, + "learning_rate": 1.4348288741300311e-05, + "loss": 0.044, + "step": 22970 + }, + { + "epoch": 3.01, + "learning_rate": 1.4347811170515376e-05, + "loss": 0.0425, + "step": 22971 + }, + { + "epoch": 3.01, + "learning_rate": 1.4347333587502453e-05, + "loss": 0.0381, + "step": 22972 + }, + { + "epoch": 3.01, + "learning_rate": 1.4346855992262882e-05, + "loss": 0.0505, + "step": 22973 + }, + { + "epoch": 3.01, + "learning_rate": 1.4346378384798012e-05, + "loss": 0.0363, + "step": 22974 + }, + { + "epoch": 3.01, + "learning_rate": 1.4345900765109187e-05, + "loss": 0.0396, + "step": 22975 + }, + { + "epoch": 3.01, + "learning_rate": 1.4345423133197745e-05, + "loss": 0.046, + "step": 22976 + }, + { + "epoch": 3.01, + "learning_rate": 1.4344945489065033e-05, + "loss": 0.0413, + "step": 22977 + }, + { + "epoch": 3.01, + "learning_rate": 1.434446783271239e-05, + "loss": 0.0523, + "step": 22978 + }, + { + "epoch": 3.01, + "learning_rate": 1.4343990164141168e-05, + "loss": 0.0506, + "step": 22979 + }, + { + "epoch": 3.01, + "learning_rate": 1.4343512483352702e-05, + "loss": 0.0494, + "step": 22980 + }, + { + "epoch": 3.01, + "learning_rate": 1.434303479034834e-05, + "loss": 0.0367, + "step": 22981 + }, + { + "epoch": 3.01, + "learning_rate": 1.434255708512942e-05, + "loss": 0.0348, + "step": 22982 + }, + { + "epoch": 3.01, + "learning_rate": 1.4342079367697294e-05, + "loss": 0.0466, + "step": 22983 + }, + { + "epoch": 3.01, + "learning_rate": 1.4341601638053298e-05, + "loss": 0.0365, + "step": 22984 + }, + { + "epoch": 3.01, + "learning_rate": 1.434112389619878e-05, + "loss": 0.0398, + "step": 22985 + }, + { + "epoch": 3.01, + "learning_rate": 1.4340646142135082e-05, + "loss": 0.0433, + "step": 22986 + }, + { + "epoch": 3.01, + "learning_rate": 1.4340168375863548e-05, + "loss": 0.0378, + "step": 22987 + }, + { + "epoch": 3.01, + "learning_rate": 1.4339690597385521e-05, + "loss": 0.0443, + "step": 22988 + }, + { + "epoch": 3.01, + "learning_rate": 1.4339212806702345e-05, + "loss": 0.0436, + "step": 22989 + }, + { + "epoch": 3.01, + "learning_rate": 1.4338735003815365e-05, + "loss": 0.039, + "step": 22990 + }, + { + "epoch": 3.01, + "learning_rate": 1.4338257188725924e-05, + "loss": 0.0422, + "step": 22991 + }, + { + "epoch": 3.01, + "learning_rate": 1.4337779361435365e-05, + "loss": 0.0505, + "step": 22992 + }, + { + "epoch": 3.01, + "learning_rate": 1.4337301521945031e-05, + "loss": 0.0478, + "step": 22993 + }, + { + "epoch": 3.01, + "learning_rate": 1.4336823670256271e-05, + "loss": 0.0488, + "step": 22994 + }, + { + "epoch": 3.01, + "learning_rate": 1.4336345806370423e-05, + "loss": 0.0399, + "step": 22995 + }, + { + "epoch": 3.01, + "learning_rate": 1.4335867930288833e-05, + "loss": 0.0536, + "step": 22996 + }, + { + "epoch": 3.01, + "learning_rate": 1.4335390042012847e-05, + "loss": 0.0392, + "step": 22997 + }, + { + "epoch": 3.01, + "learning_rate": 1.4334912141543807e-05, + "loss": 0.0487, + "step": 22998 + }, + { + "epoch": 3.01, + "learning_rate": 1.4334434228883057e-05, + "loss": 0.0446, + "step": 22999 + }, + { + "epoch": 3.01, + "learning_rate": 1.4333956304031943e-05, + "loss": 0.0438, + "step": 23000 + }, + { + "epoch": 3.01, + "learning_rate": 1.4333478366991804e-05, + "loss": 0.0416, + "step": 23001 + }, + { + "epoch": 3.01, + "learning_rate": 1.4333000417763991e-05, + "loss": 0.0404, + "step": 23002 + }, + { + "epoch": 3.01, + "learning_rate": 1.4332522456349845e-05, + "loss": 0.0454, + "step": 23003 + }, + { + "epoch": 3.01, + "learning_rate": 1.4332044482750709e-05, + "loss": 0.0314, + "step": 23004 + }, + { + "epoch": 3.01, + "learning_rate": 1.4331566496967929e-05, + "loss": 0.0518, + "step": 23005 + }, + { + "epoch": 3.01, + "learning_rate": 1.4331088499002848e-05, + "loss": 0.0535, + "step": 23006 + }, + { + "epoch": 3.01, + "learning_rate": 1.4330610488856812e-05, + "loss": 0.0412, + "step": 23007 + }, + { + "epoch": 3.01, + "learning_rate": 1.4330132466531166e-05, + "loss": 0.039, + "step": 23008 + }, + { + "epoch": 3.01, + "learning_rate": 1.432965443202725e-05, + "loss": 0.0527, + "step": 23009 + }, + { + "epoch": 3.01, + "learning_rate": 1.4329176385346415e-05, + "loss": 0.0454, + "step": 23010 + }, + { + "epoch": 3.01, + "learning_rate": 1.432869832649e-05, + "loss": 0.0429, + "step": 23011 + }, + { + "epoch": 3.01, + "learning_rate": 1.4328220255459353e-05, + "loss": 0.0445, + "step": 23012 + }, + { + "epoch": 3.01, + "learning_rate": 1.4327742172255813e-05, + "loss": 0.0377, + "step": 23013 + }, + { + "epoch": 3.01, + "learning_rate": 1.4327264076880734e-05, + "loss": 0.04, + "step": 23014 + }, + { + "epoch": 3.01, + "learning_rate": 1.4326785969335452e-05, + "loss": 0.0469, + "step": 23015 + }, + { + "epoch": 3.01, + "learning_rate": 1.4326307849621317e-05, + "loss": 0.042, + "step": 23016 + }, + { + "epoch": 3.02, + "learning_rate": 1.4325829717739668e-05, + "loss": 0.0438, + "step": 23017 + }, + { + "epoch": 3.02, + "learning_rate": 1.4325351573691855e-05, + "loss": 0.0419, + "step": 23018 + }, + { + "epoch": 3.02, + "learning_rate": 1.4324873417479224e-05, + "loss": 0.0391, + "step": 23019 + }, + { + "epoch": 3.02, + "learning_rate": 1.4324395249103114e-05, + "loss": 0.0458, + "step": 23020 + }, + { + "epoch": 3.02, + "learning_rate": 1.4323917068564874e-05, + "loss": 0.0433, + "step": 23021 + }, + { + "epoch": 3.02, + "learning_rate": 1.4323438875865843e-05, + "loss": 0.0418, + "step": 23022 + }, + { + "epoch": 3.02, + "learning_rate": 1.4322960671007375e-05, + "loss": 0.0471, + "step": 23023 + }, + { + "epoch": 3.02, + "learning_rate": 1.432248245399081e-05, + "loss": 0.044, + "step": 23024 + }, + { + "epoch": 3.02, + "learning_rate": 1.4322004224817491e-05, + "loss": 0.042, + "step": 23025 + }, + { + "epoch": 3.02, + "learning_rate": 1.432152598348877e-05, + "loss": 0.0395, + "step": 23026 + }, + { + "epoch": 3.02, + "learning_rate": 1.4321047730005982e-05, + "loss": 0.0412, + "step": 23027 + }, + { + "epoch": 3.02, + "learning_rate": 1.4320569464370477e-05, + "loss": 0.0443, + "step": 23028 + }, + { + "epoch": 3.02, + "learning_rate": 1.4320091186583605e-05, + "loss": 0.0442, + "step": 23029 + }, + { + "epoch": 3.02, + "learning_rate": 1.4319612896646705e-05, + "loss": 0.0435, + "step": 23030 + }, + { + "epoch": 3.02, + "learning_rate": 1.4319134594561124e-05, + "loss": 0.0437, + "step": 23031 + }, + { + "epoch": 3.02, + "learning_rate": 1.4318656280328207e-05, + "loss": 0.0444, + "step": 23032 + }, + { + "epoch": 3.02, + "learning_rate": 1.4318177953949296e-05, + "loss": 0.0473, + "step": 23033 + }, + { + "epoch": 3.02, + "learning_rate": 1.4317699615425743e-05, + "loss": 0.0487, + "step": 23034 + }, + { + "epoch": 3.02, + "learning_rate": 1.4317221264758887e-05, + "loss": 0.0389, + "step": 23035 + }, + { + "epoch": 3.02, + "learning_rate": 1.4316742901950079e-05, + "loss": 0.0492, + "step": 23036 + }, + { + "epoch": 3.02, + "learning_rate": 1.4316264527000661e-05, + "loss": 0.0422, + "step": 23037 + }, + { + "epoch": 3.02, + "learning_rate": 1.4315786139911975e-05, + "loss": 0.0454, + "step": 23038 + }, + { + "epoch": 3.02, + "learning_rate": 1.4315307740685373e-05, + "loss": 0.0427, + "step": 23039 + }, + { + "epoch": 3.02, + "learning_rate": 1.43148293293222e-05, + "loss": 0.0392, + "step": 23040 + }, + { + "epoch": 3.02, + "learning_rate": 1.4314350905823798e-05, + "loss": 0.0367, + "step": 23041 + }, + { + "epoch": 3.02, + "learning_rate": 1.4313872470191515e-05, + "loss": 0.0402, + "step": 23042 + }, + { + "epoch": 3.02, + "learning_rate": 1.4313394022426693e-05, + "loss": 0.0411, + "step": 23043 + }, + { + "epoch": 3.02, + "learning_rate": 1.431291556253068e-05, + "loss": 0.0459, + "step": 23044 + }, + { + "epoch": 3.02, + "learning_rate": 1.431243709050482e-05, + "loss": 0.0495, + "step": 23045 + }, + { + "epoch": 3.02, + "learning_rate": 1.4311958606350464e-05, + "loss": 0.0445, + "step": 23046 + }, + { + "epoch": 3.02, + "learning_rate": 1.4311480110068954e-05, + "loss": 0.0512, + "step": 23047 + }, + { + "epoch": 3.02, + "learning_rate": 1.4311001601661634e-05, + "loss": 0.0385, + "step": 23048 + }, + { + "epoch": 3.02, + "learning_rate": 1.431052308112985e-05, + "loss": 0.0429, + "step": 23049 + }, + { + "epoch": 3.02, + "learning_rate": 1.4310044548474952e-05, + "loss": 0.043, + "step": 23050 + }, + { + "epoch": 3.02, + "learning_rate": 1.4309566003698283e-05, + "loss": 0.0395, + "step": 23051 + }, + { + "epoch": 3.02, + "learning_rate": 1.430908744680119e-05, + "loss": 0.0501, + "step": 23052 + }, + { + "epoch": 3.02, + "learning_rate": 1.4308608877785018e-05, + "loss": 0.0385, + "step": 23053 + }, + { + "epoch": 3.02, + "learning_rate": 1.4308130296651111e-05, + "loss": 0.0359, + "step": 23054 + }, + { + "epoch": 3.02, + "learning_rate": 1.4307651703400816e-05, + "loss": 0.0471, + "step": 23055 + }, + { + "epoch": 3.02, + "learning_rate": 1.4307173098035484e-05, + "loss": 0.0442, + "step": 23056 + }, + { + "epoch": 3.02, + "learning_rate": 1.4306694480556453e-05, + "loss": 0.0365, + "step": 23057 + }, + { + "epoch": 3.02, + "learning_rate": 1.4306215850965075e-05, + "loss": 0.0386, + "step": 23058 + }, + { + "epoch": 3.02, + "learning_rate": 1.4305737209262695e-05, + "loss": 0.0465, + "step": 23059 + }, + { + "epoch": 3.02, + "learning_rate": 1.4305258555450654e-05, + "loss": 0.0438, + "step": 23060 + }, + { + "epoch": 3.02, + "learning_rate": 1.430477988953031e-05, + "loss": 0.0358, + "step": 23061 + }, + { + "epoch": 3.02, + "learning_rate": 1.4304301211502996e-05, + "loss": 0.0452, + "step": 23062 + }, + { + "epoch": 3.02, + "learning_rate": 1.4303822521370067e-05, + "loss": 0.0444, + "step": 23063 + }, + { + "epoch": 3.02, + "learning_rate": 1.4303343819132864e-05, + "loss": 0.0435, + "step": 23064 + }, + { + "epoch": 3.02, + "learning_rate": 1.430286510479274e-05, + "loss": 0.0354, + "step": 23065 + }, + { + "epoch": 3.02, + "learning_rate": 1.4302386378351032e-05, + "loss": 0.0408, + "step": 23066 + }, + { + "epoch": 3.02, + "learning_rate": 1.4301907639809093e-05, + "loss": 0.0453, + "step": 23067 + }, + { + "epoch": 3.02, + "learning_rate": 1.4301428889168269e-05, + "loss": 0.0515, + "step": 23068 + }, + { + "epoch": 3.02, + "learning_rate": 1.4300950126429904e-05, + "loss": 0.041, + "step": 23069 + }, + { + "epoch": 3.02, + "learning_rate": 1.4300471351595349e-05, + "loss": 0.0431, + "step": 23070 + }, + { + "epoch": 3.02, + "learning_rate": 1.4299992564665942e-05, + "loss": 0.0393, + "step": 23071 + }, + { + "epoch": 3.02, + "learning_rate": 1.4299513765643039e-05, + "loss": 0.0415, + "step": 23072 + }, + { + "epoch": 3.02, + "learning_rate": 1.4299034954527982e-05, + "loss": 0.0432, + "step": 23073 + }, + { + "epoch": 3.02, + "learning_rate": 1.4298556131322116e-05, + "loss": 0.0326, + "step": 23074 + }, + { + "epoch": 3.02, + "learning_rate": 1.4298077296026792e-05, + "loss": 0.0395, + "step": 23075 + }, + { + "epoch": 3.02, + "learning_rate": 1.4297598448643353e-05, + "loss": 0.0387, + "step": 23076 + }, + { + "epoch": 3.02, + "learning_rate": 1.4297119589173148e-05, + "loss": 0.0427, + "step": 23077 + }, + { + "epoch": 3.02, + "learning_rate": 1.4296640717617523e-05, + "loss": 0.0451, + "step": 23078 + }, + { + "epoch": 3.02, + "learning_rate": 1.4296161833977822e-05, + "loss": 0.0453, + "step": 23079 + }, + { + "epoch": 3.02, + "learning_rate": 1.4295682938255399e-05, + "loss": 0.0399, + "step": 23080 + }, + { + "epoch": 3.02, + "learning_rate": 1.4295204030451594e-05, + "loss": 0.0422, + "step": 23081 + }, + { + "epoch": 3.02, + "learning_rate": 1.4294725110567758e-05, + "loss": 0.0417, + "step": 23082 + }, + { + "epoch": 3.02, + "learning_rate": 1.4294246178605234e-05, + "loss": 0.04, + "step": 23083 + }, + { + "epoch": 3.02, + "learning_rate": 1.4293767234565375e-05, + "loss": 0.05, + "step": 23084 + }, + { + "epoch": 3.02, + "learning_rate": 1.4293288278449521e-05, + "loss": 0.0502, + "step": 23085 + }, + { + "epoch": 3.02, + "learning_rate": 1.4292809310259023e-05, + "loss": 0.0431, + "step": 23086 + }, + { + "epoch": 3.02, + "learning_rate": 1.4292330329995228e-05, + "loss": 0.0365, + "step": 23087 + }, + { + "epoch": 3.02, + "learning_rate": 1.4291851337659482e-05, + "loss": 0.0451, + "step": 23088 + }, + { + "epoch": 3.02, + "learning_rate": 1.4291372333253133e-05, + "loss": 0.0359, + "step": 23089 + }, + { + "epoch": 3.02, + "learning_rate": 1.429089331677753e-05, + "loss": 0.0438, + "step": 23090 + }, + { + "epoch": 3.02, + "learning_rate": 1.4290414288234015e-05, + "loss": 0.0375, + "step": 23091 + }, + { + "epoch": 3.02, + "learning_rate": 1.428993524762394e-05, + "loss": 0.0484, + "step": 23092 + }, + { + "epoch": 3.03, + "learning_rate": 1.4289456194948648e-05, + "loss": 0.0428, + "step": 23093 + }, + { + "epoch": 3.03, + "learning_rate": 1.4288977130209491e-05, + "loss": 0.0425, + "step": 23094 + }, + { + "epoch": 3.03, + "learning_rate": 1.4288498053407813e-05, + "loss": 0.0321, + "step": 23095 + }, + { + "epoch": 3.03, + "learning_rate": 1.4288018964544962e-05, + "loss": 0.0461, + "step": 23096 + }, + { + "epoch": 3.03, + "learning_rate": 1.428753986362229e-05, + "loss": 0.0405, + "step": 23097 + }, + { + "epoch": 3.03, + "learning_rate": 1.4287060750641138e-05, + "loss": 0.0457, + "step": 23098 + }, + { + "epoch": 3.03, + "learning_rate": 1.4286581625602856e-05, + "loss": 0.0434, + "step": 23099 + }, + { + "epoch": 3.03, + "learning_rate": 1.428610248850879e-05, + "loss": 0.044, + "step": 23100 + }, + { + "epoch": 3.03, + "learning_rate": 1.428562333936029e-05, + "loss": 0.0355, + "step": 23101 + }, + { + "epoch": 3.03, + "learning_rate": 1.4285144178158703e-05, + "loss": 0.0504, + "step": 23102 + }, + { + "epoch": 3.03, + "learning_rate": 1.4284665004905378e-05, + "loss": 0.0423, + "step": 23103 + }, + { + "epoch": 3.03, + "learning_rate": 1.4284185819601657e-05, + "loss": 0.046, + "step": 23104 + }, + { + "epoch": 3.03, + "learning_rate": 1.4283706622248894e-05, + "loss": 0.0415, + "step": 23105 + }, + { + "epoch": 3.03, + "learning_rate": 1.4283227412848435e-05, + "loss": 0.0385, + "step": 23106 + }, + { + "epoch": 3.03, + "learning_rate": 1.4282748191401625e-05, + "loss": 0.0505, + "step": 23107 + }, + { + "epoch": 3.03, + "learning_rate": 1.4282268957909815e-05, + "loss": 0.0488, + "step": 23108 + }, + { + "epoch": 3.03, + "learning_rate": 1.428178971237435e-05, + "loss": 0.0352, + "step": 23109 + }, + { + "epoch": 3.03, + "learning_rate": 1.428131045479658e-05, + "loss": 0.0432, + "step": 23110 + }, + { + "epoch": 3.03, + "learning_rate": 1.4280831185177855e-05, + "loss": 0.0472, + "step": 23111 + }, + { + "epoch": 3.03, + "learning_rate": 1.4280351903519515e-05, + "loss": 0.0427, + "step": 23112 + }, + { + "epoch": 3.03, + "learning_rate": 1.427987260982292e-05, + "loss": 0.0386, + "step": 23113 + }, + { + "epoch": 3.03, + "learning_rate": 1.4279393304089405e-05, + "loss": 0.0404, + "step": 23114 + }, + { + "epoch": 3.03, + "learning_rate": 1.4278913986320328e-05, + "loss": 0.0413, + "step": 23115 + }, + { + "epoch": 3.03, + "learning_rate": 1.4278434656517031e-05, + "loss": 0.0473, + "step": 23116 + }, + { + "epoch": 3.03, + "learning_rate": 1.4277955314680868e-05, + "loss": 0.0419, + "step": 23117 + }, + { + "epoch": 3.03, + "learning_rate": 1.4277475960813179e-05, + "loss": 0.0431, + "step": 23118 + }, + { + "epoch": 3.03, + "learning_rate": 1.4276996594915318e-05, + "loss": 0.0448, + "step": 23119 + }, + { + "epoch": 3.03, + "learning_rate": 1.4276517216988633e-05, + "loss": 0.0441, + "step": 23120 + }, + { + "epoch": 3.03, + "learning_rate": 1.427603782703447e-05, + "loss": 0.0312, + "step": 23121 + }, + { + "epoch": 3.03, + "learning_rate": 1.4275558425054178e-05, + "loss": 0.0501, + "step": 23122 + }, + { + "epoch": 3.03, + "learning_rate": 1.4275079011049107e-05, + "loss": 0.0465, + "step": 23123 + }, + { + "epoch": 3.03, + "learning_rate": 1.4274599585020606e-05, + "loss": 0.0431, + "step": 23124 + }, + { + "epoch": 3.03, + "learning_rate": 1.4274120146970019e-05, + "loss": 0.0478, + "step": 23125 + }, + { + "epoch": 3.03, + "learning_rate": 1.4273640696898695e-05, + "loss": 0.0428, + "step": 23126 + }, + { + "epoch": 3.03, + "learning_rate": 1.4273161234807984e-05, + "loss": 0.0375, + "step": 23127 + }, + { + "epoch": 3.03, + "learning_rate": 1.4272681760699239e-05, + "loss": 0.0443, + "step": 23128 + }, + { + "epoch": 3.03, + "learning_rate": 1.42722022745738e-05, + "loss": 0.0498, + "step": 23129 + }, + { + "epoch": 3.03, + "learning_rate": 1.4271722776433022e-05, + "loss": 0.0324, + "step": 23130 + }, + { + "epoch": 3.03, + "learning_rate": 1.4271243266278245e-05, + "loss": 0.0423, + "step": 23131 + }, + { + "epoch": 3.03, + "learning_rate": 1.4270763744110832e-05, + "loss": 0.0421, + "step": 23132 + }, + { + "epoch": 3.03, + "learning_rate": 1.427028420993212e-05, + "loss": 0.0458, + "step": 23133 + }, + { + "epoch": 3.03, + "learning_rate": 1.4269804663743463e-05, + "loss": 0.0463, + "step": 23134 + }, + { + "epoch": 3.03, + "learning_rate": 1.4269325105546203e-05, + "loss": 0.0382, + "step": 23135 + }, + { + "epoch": 3.03, + "learning_rate": 1.4268845535341698e-05, + "loss": 0.039, + "step": 23136 + }, + { + "epoch": 3.03, + "learning_rate": 1.426836595313129e-05, + "loss": 0.0453, + "step": 23137 + }, + { + "epoch": 3.03, + "learning_rate": 1.4267886358916331e-05, + "loss": 0.0363, + "step": 23138 + }, + { + "epoch": 3.03, + "learning_rate": 1.4267406752698168e-05, + "loss": 0.0454, + "step": 23139 + }, + { + "epoch": 3.03, + "learning_rate": 1.4266927134478152e-05, + "loss": 0.0512, + "step": 23140 + }, + { + "epoch": 3.03, + "learning_rate": 1.4266447504257631e-05, + "loss": 0.0388, + "step": 23141 + }, + { + "epoch": 3.03, + "learning_rate": 1.4265967862037951e-05, + "loss": 0.0412, + "step": 23142 + }, + { + "epoch": 3.03, + "learning_rate": 1.4265488207820467e-05, + "loss": 0.0437, + "step": 23143 + }, + { + "epoch": 3.03, + "learning_rate": 1.426500854160652e-05, + "loss": 0.0465, + "step": 23144 + }, + { + "epoch": 3.03, + "learning_rate": 1.4264528863397468e-05, + "loss": 0.0443, + "step": 23145 + }, + { + "epoch": 3.03, + "learning_rate": 1.4264049173194655e-05, + "loss": 0.0441, + "step": 23146 + }, + { + "epoch": 3.03, + "learning_rate": 1.4263569470999428e-05, + "loss": 0.0454, + "step": 23147 + }, + { + "epoch": 3.03, + "learning_rate": 1.426308975681314e-05, + "loss": 0.0458, + "step": 23148 + }, + { + "epoch": 3.03, + "learning_rate": 1.4262610030637141e-05, + "loss": 0.0414, + "step": 23149 + }, + { + "epoch": 3.03, + "learning_rate": 1.4262130292472778e-05, + "loss": 0.0402, + "step": 23150 + }, + { + "epoch": 3.03, + "learning_rate": 1.4261650542321399e-05, + "loss": 0.0405, + "step": 23151 + }, + { + "epoch": 3.03, + "learning_rate": 1.4261170780184356e-05, + "loss": 0.0384, + "step": 23152 + }, + { + "epoch": 3.03, + "learning_rate": 1.4260691006062997e-05, + "loss": 0.0331, + "step": 23153 + }, + { + "epoch": 3.03, + "learning_rate": 1.4260211219958669e-05, + "loss": 0.0532, + "step": 23154 + }, + { + "epoch": 3.03, + "learning_rate": 1.4259731421872726e-05, + "loss": 0.0462, + "step": 23155 + }, + { + "epoch": 3.03, + "learning_rate": 1.4259251611806514e-05, + "loss": 0.0443, + "step": 23156 + }, + { + "epoch": 3.03, + "learning_rate": 1.4258771789761388e-05, + "loss": 0.0374, + "step": 23157 + }, + { + "epoch": 3.03, + "learning_rate": 1.4258291955738688e-05, + "loss": 0.0495, + "step": 23158 + }, + { + "epoch": 3.03, + "learning_rate": 1.4257812109739773e-05, + "loss": 0.028, + "step": 23159 + }, + { + "epoch": 3.03, + "learning_rate": 1.4257332251765982e-05, + "loss": 0.0434, + "step": 23160 + }, + { + "epoch": 3.03, + "learning_rate": 1.4256852381818676e-05, + "loss": 0.0456, + "step": 23161 + }, + { + "epoch": 3.03, + "learning_rate": 1.4256372499899197e-05, + "loss": 0.0385, + "step": 23162 + }, + { + "epoch": 3.03, + "learning_rate": 1.4255892606008899e-05, + "loss": 0.048, + "step": 23163 + }, + { + "epoch": 3.03, + "learning_rate": 1.4255412700149127e-05, + "loss": 0.0386, + "step": 23164 + }, + { + "epoch": 3.03, + "learning_rate": 1.4254932782321237e-05, + "loss": 0.0382, + "step": 23165 + }, + { + "epoch": 3.03, + "learning_rate": 1.4254452852526572e-05, + "loss": 0.0398, + "step": 23166 + }, + { + "epoch": 3.03, + "learning_rate": 1.4253972910766486e-05, + "loss": 0.0395, + "step": 23167 + }, + { + "epoch": 3.03, + "learning_rate": 1.425349295704233e-05, + "loss": 0.0387, + "step": 23168 + }, + { + "epoch": 3.03, + "learning_rate": 1.4253012991355445e-05, + "loss": 0.0442, + "step": 23169 + }, + { + "epoch": 3.04, + "learning_rate": 1.4252533013707193e-05, + "loss": 0.0429, + "step": 23170 + }, + { + "epoch": 3.04, + "learning_rate": 1.4252053024098918e-05, + "loss": 0.0412, + "step": 23171 + }, + { + "epoch": 3.04, + "learning_rate": 1.4251573022531968e-05, + "loss": 0.0344, + "step": 23172 + }, + { + "epoch": 3.04, + "learning_rate": 1.4251093009007696e-05, + "loss": 0.044, + "step": 23173 + }, + { + "epoch": 3.04, + "learning_rate": 1.4250612983527455e-05, + "loss": 0.0393, + "step": 23174 + }, + { + "epoch": 3.04, + "learning_rate": 1.4250132946092585e-05, + "loss": 0.0479, + "step": 23175 + }, + { + "epoch": 3.04, + "learning_rate": 1.4249652896704447e-05, + "loss": 0.0489, + "step": 23176 + }, + { + "epoch": 3.04, + "learning_rate": 1.4249172835364385e-05, + "loss": 0.0449, + "step": 23177 + }, + { + "epoch": 3.04, + "learning_rate": 1.4248692762073753e-05, + "loss": 0.047, + "step": 23178 + }, + { + "epoch": 3.04, + "learning_rate": 1.4248212676833894e-05, + "loss": 0.0508, + "step": 23179 + }, + { + "epoch": 3.04, + "learning_rate": 1.4247732579646167e-05, + "loss": 0.0431, + "step": 23180 + }, + { + "epoch": 3.04, + "learning_rate": 1.4247252470511917e-05, + "loss": 0.0404, + "step": 23181 + }, + { + "epoch": 3.04, + "learning_rate": 1.4246772349432496e-05, + "loss": 0.0477, + "step": 23182 + }, + { + "epoch": 3.04, + "learning_rate": 1.4246292216409255e-05, + "loss": 0.0413, + "step": 23183 + }, + { + "epoch": 3.04, + "learning_rate": 1.424581207144354e-05, + "loss": 0.0364, + "step": 23184 + }, + { + "epoch": 3.04, + "learning_rate": 1.4245331914536708e-05, + "loss": 0.0372, + "step": 23185 + }, + { + "epoch": 3.04, + "learning_rate": 1.4244851745690103e-05, + "loss": 0.0434, + "step": 23186 + }, + { + "epoch": 3.04, + "learning_rate": 1.424437156490508e-05, + "loss": 0.0396, + "step": 23187 + }, + { + "epoch": 3.04, + "learning_rate": 1.4243891372182987e-05, + "loss": 0.041, + "step": 23188 + }, + { + "epoch": 3.04, + "learning_rate": 1.424341116752518e-05, + "loss": 0.0384, + "step": 23189 + }, + { + "epoch": 3.04, + "learning_rate": 1.4242930950933001e-05, + "loss": 0.052, + "step": 23190 + }, + { + "epoch": 3.04, + "learning_rate": 1.4242450722407805e-05, + "loss": 0.0334, + "step": 23191 + }, + { + "epoch": 3.04, + "learning_rate": 1.4241970481950942e-05, + "loss": 0.0326, + "step": 23192 + }, + { + "epoch": 3.04, + "learning_rate": 1.4241490229563764e-05, + "loss": 0.0385, + "step": 23193 + }, + { + "epoch": 3.04, + "learning_rate": 1.4241009965247621e-05, + "loss": 0.0417, + "step": 23194 + }, + { + "epoch": 3.04, + "learning_rate": 1.4240529689003863e-05, + "loss": 0.0391, + "step": 23195 + }, + { + "epoch": 3.04, + "learning_rate": 1.424004940083384e-05, + "loss": 0.046, + "step": 23196 + }, + { + "epoch": 3.04, + "learning_rate": 1.4239569100738904e-05, + "loss": 0.0462, + "step": 23197 + }, + { + "epoch": 3.04, + "learning_rate": 1.4239088788720405e-05, + "loss": 0.0427, + "step": 23198 + }, + { + "epoch": 3.04, + "learning_rate": 1.4238608464779697e-05, + "loss": 0.0438, + "step": 23199 + }, + { + "epoch": 3.04, + "learning_rate": 1.4238128128918127e-05, + "loss": 0.045, + "step": 23200 + }, + { + "epoch": 3.04, + "learning_rate": 1.4237647781137048e-05, + "loss": 0.0476, + "step": 23201 + }, + { + "epoch": 3.04, + "learning_rate": 1.4237167421437807e-05, + "loss": 0.0499, + "step": 23202 + }, + { + "epoch": 3.04, + "learning_rate": 1.4236687049821763e-05, + "loss": 0.0459, + "step": 23203 + }, + { + "epoch": 3.04, + "learning_rate": 1.4236206666290259e-05, + "loss": 0.0421, + "step": 23204 + }, + { + "epoch": 3.04, + "learning_rate": 1.4235726270844649e-05, + "loss": 0.0405, + "step": 23205 + }, + { + "epoch": 3.04, + "learning_rate": 1.4235245863486288e-05, + "loss": 0.045, + "step": 23206 + }, + { + "epoch": 3.04, + "learning_rate": 1.4234765444216523e-05, + "loss": 0.0439, + "step": 23207 + }, + { + "epoch": 3.04, + "learning_rate": 1.4234285013036702e-05, + "loss": 0.0367, + "step": 23208 + }, + { + "epoch": 3.04, + "learning_rate": 1.4233804569948182e-05, + "loss": 0.0412, + "step": 23209 + }, + { + "epoch": 3.04, + "learning_rate": 1.4233324114952314e-05, + "loss": 0.0424, + "step": 23210 + }, + { + "epoch": 3.04, + "learning_rate": 1.4232843648050445e-05, + "loss": 0.0469, + "step": 23211 + }, + { + "epoch": 3.04, + "learning_rate": 1.423236316924393e-05, + "loss": 0.0398, + "step": 23212 + }, + { + "epoch": 3.04, + "learning_rate": 1.4231882678534117e-05, + "loss": 0.0353, + "step": 23213 + }, + { + "epoch": 3.04, + "learning_rate": 1.4231402175922363e-05, + "loss": 0.0421, + "step": 23214 + }, + { + "epoch": 3.04, + "learning_rate": 1.423092166141001e-05, + "loss": 0.0473, + "step": 23215 + }, + { + "epoch": 3.04, + "learning_rate": 1.4230441134998421e-05, + "loss": 0.0449, + "step": 23216 + }, + { + "epoch": 3.04, + "learning_rate": 1.422996059668894e-05, + "loss": 0.0494, + "step": 23217 + }, + { + "epoch": 3.04, + "learning_rate": 1.4229480046482917e-05, + "loss": 0.0373, + "step": 23218 + }, + { + "epoch": 3.04, + "learning_rate": 1.422899948438171e-05, + "loss": 0.0382, + "step": 23219 + }, + { + "epoch": 3.04, + "learning_rate": 1.4228518910386667e-05, + "loss": 0.0408, + "step": 23220 + }, + { + "epoch": 3.04, + "learning_rate": 1.4228038324499136e-05, + "loss": 0.0412, + "step": 23221 + }, + { + "epoch": 3.04, + "learning_rate": 1.4227557726720475e-05, + "loss": 0.0386, + "step": 23222 + }, + { + "epoch": 3.04, + "learning_rate": 1.4227077117052035e-05, + "loss": 0.037, + "step": 23223 + }, + { + "epoch": 3.04, + "learning_rate": 1.4226596495495161e-05, + "loss": 0.0466, + "step": 23224 + }, + { + "epoch": 3.04, + "learning_rate": 1.4226115862051213e-05, + "loss": 0.0391, + "step": 23225 + }, + { + "epoch": 3.04, + "learning_rate": 1.4225635216721539e-05, + "loss": 0.0409, + "step": 23226 + }, + { + "epoch": 3.04, + "learning_rate": 1.422515455950749e-05, + "loss": 0.0391, + "step": 23227 + }, + { + "epoch": 3.04, + "learning_rate": 1.4224673890410418e-05, + "loss": 0.0409, + "step": 23228 + }, + { + "epoch": 3.04, + "learning_rate": 1.4224193209431679e-05, + "loss": 0.0385, + "step": 23229 + }, + { + "epoch": 3.04, + "learning_rate": 1.4223712516572616e-05, + "loss": 0.0402, + "step": 23230 + }, + { + "epoch": 3.04, + "learning_rate": 1.4223231811834592e-05, + "loss": 0.0451, + "step": 23231 + }, + { + "epoch": 3.04, + "learning_rate": 1.4222751095218951e-05, + "loss": 0.0421, + "step": 23232 + }, + { + "epoch": 3.04, + "learning_rate": 1.4222270366727049e-05, + "loss": 0.0406, + "step": 23233 + }, + { + "epoch": 3.04, + "learning_rate": 1.4221789626360234e-05, + "loss": 0.0412, + "step": 23234 + }, + { + "epoch": 3.04, + "learning_rate": 1.422130887411986e-05, + "loss": 0.0445, + "step": 23235 + }, + { + "epoch": 3.04, + "learning_rate": 1.422082811000728e-05, + "loss": 0.0393, + "step": 23236 + }, + { + "epoch": 3.04, + "learning_rate": 1.422034733402385e-05, + "loss": 0.0464, + "step": 23237 + }, + { + "epoch": 3.04, + "learning_rate": 1.4219866546170915e-05, + "loss": 0.0475, + "step": 23238 + }, + { + "epoch": 3.04, + "learning_rate": 1.4219385746449832e-05, + "loss": 0.0384, + "step": 23239 + }, + { + "epoch": 3.04, + "learning_rate": 1.4218904934861946e-05, + "loss": 0.0423, + "step": 23240 + }, + { + "epoch": 3.04, + "learning_rate": 1.4218424111408617e-05, + "loss": 0.0468, + "step": 23241 + }, + { + "epoch": 3.04, + "learning_rate": 1.4217943276091198e-05, + "loss": 0.0528, + "step": 23242 + }, + { + "epoch": 3.04, + "learning_rate": 1.4217462428911035e-05, + "loss": 0.0362, + "step": 23243 + }, + { + "epoch": 3.04, + "learning_rate": 1.4216981569869484e-05, + "loss": 0.0459, + "step": 23244 + }, + { + "epoch": 3.04, + "learning_rate": 1.4216500698967897e-05, + "loss": 0.0465, + "step": 23245 + }, + { + "epoch": 3.05, + "learning_rate": 1.4216019816207625e-05, + "loss": 0.0412, + "step": 23246 + }, + { + "epoch": 3.05, + "learning_rate": 1.4215538921590023e-05, + "loss": 0.0407, + "step": 23247 + }, + { + "epoch": 3.05, + "learning_rate": 1.4215058015116444e-05, + "loss": 0.0484, + "step": 23248 + }, + { + "epoch": 3.05, + "learning_rate": 1.4214577096788235e-05, + "loss": 0.0436, + "step": 23249 + }, + { + "epoch": 3.05, + "learning_rate": 1.4214096166606757e-05, + "loss": 0.0416, + "step": 23250 + }, + { + "epoch": 3.05, + "learning_rate": 1.4213615224573354e-05, + "loss": 0.0436, + "step": 23251 + }, + { + "epoch": 3.05, + "learning_rate": 1.4213134270689386e-05, + "loss": 0.0418, + "step": 23252 + }, + { + "epoch": 3.05, + "learning_rate": 1.4212653304956198e-05, + "loss": 0.0398, + "step": 23253 + }, + { + "epoch": 3.05, + "learning_rate": 1.4212172327375152e-05, + "loss": 0.0447, + "step": 23254 + }, + { + "epoch": 3.05, + "learning_rate": 1.4211691337947591e-05, + "loss": 0.0508, + "step": 23255 + }, + { + "epoch": 3.05, + "learning_rate": 1.4211210336674876e-05, + "loss": 0.039, + "step": 23256 + }, + { + "epoch": 3.05, + "learning_rate": 1.4210729323558351e-05, + "loss": 0.0408, + "step": 23257 + }, + { + "epoch": 3.05, + "learning_rate": 1.4210248298599381e-05, + "loss": 0.0445, + "step": 23258 + }, + { + "epoch": 3.05, + "learning_rate": 1.4209767261799306e-05, + "loss": 0.0366, + "step": 23259 + }, + { + "epoch": 3.05, + "learning_rate": 1.420928621315949e-05, + "loss": 0.0437, + "step": 23260 + }, + { + "epoch": 3.05, + "learning_rate": 1.4208805152681277e-05, + "loss": 0.0447, + "step": 23261 + }, + { + "epoch": 3.05, + "learning_rate": 1.4208324080366023e-05, + "loss": 0.0427, + "step": 23262 + }, + { + "epoch": 3.05, + "learning_rate": 1.4207842996215081e-05, + "loss": 0.0499, + "step": 23263 + }, + { + "epoch": 3.05, + "learning_rate": 1.420736190022981e-05, + "loss": 0.0449, + "step": 23264 + }, + { + "epoch": 3.05, + "learning_rate": 1.4206880792411553e-05, + "loss": 0.053, + "step": 23265 + }, + { + "epoch": 3.05, + "learning_rate": 1.4206399672761669e-05, + "loss": 0.0376, + "step": 23266 + }, + { + "epoch": 3.05, + "learning_rate": 1.4205918541281512e-05, + "loss": 0.0453, + "step": 23267 + }, + { + "epoch": 3.05, + "learning_rate": 1.420543739797243e-05, + "loss": 0.0443, + "step": 23268 + }, + { + "epoch": 3.05, + "learning_rate": 1.420495624283578e-05, + "loss": 0.0403, + "step": 23269 + }, + { + "epoch": 3.05, + "learning_rate": 1.4204475075872914e-05, + "loss": 0.0427, + "step": 23270 + }, + { + "epoch": 3.05, + "learning_rate": 1.4203993897085189e-05, + "loss": 0.0429, + "step": 23271 + }, + { + "epoch": 3.05, + "learning_rate": 1.4203512706473952e-05, + "loss": 0.0422, + "step": 23272 + }, + { + "epoch": 3.05, + "learning_rate": 1.4203031504040561e-05, + "loss": 0.0382, + "step": 23273 + }, + { + "epoch": 3.05, + "learning_rate": 1.4202550289786365e-05, + "loss": 0.0529, + "step": 23274 + }, + { + "epoch": 3.05, + "learning_rate": 1.4202069063712722e-05, + "loss": 0.0422, + "step": 23275 + }, + { + "epoch": 3.05, + "learning_rate": 1.4201587825820983e-05, + "loss": 0.047, + "step": 23276 + }, + { + "epoch": 3.05, + "learning_rate": 1.4201106576112507e-05, + "loss": 0.0381, + "step": 23277 + }, + { + "epoch": 3.05, + "learning_rate": 1.4200625314588634e-05, + "loss": 0.0419, + "step": 23278 + }, + { + "epoch": 3.05, + "learning_rate": 1.4200144041250732e-05, + "loss": 0.0391, + "step": 23279 + }, + { + "epoch": 3.05, + "learning_rate": 1.4199662756100145e-05, + "loss": 0.0409, + "step": 23280 + }, + { + "epoch": 3.05, + "learning_rate": 1.4199181459138233e-05, + "loss": 0.0446, + "step": 23281 + }, + { + "epoch": 3.05, + "learning_rate": 1.4198700150366344e-05, + "loss": 0.0471, + "step": 23282 + }, + { + "epoch": 3.05, + "learning_rate": 1.4198218829785838e-05, + "loss": 0.0399, + "step": 23283 + }, + { + "epoch": 3.05, + "learning_rate": 1.4197737497398061e-05, + "loss": 0.044, + "step": 23284 + }, + { + "epoch": 3.05, + "learning_rate": 1.4197256153204376e-05, + "loss": 0.0519, + "step": 23285 + }, + { + "epoch": 3.05, + "learning_rate": 1.4196774797206126e-05, + "loss": 0.0505, + "step": 23286 + }, + { + "epoch": 3.05, + "learning_rate": 1.4196293429404673e-05, + "loss": 0.0363, + "step": 23287 + }, + { + "epoch": 3.05, + "learning_rate": 1.419581204980137e-05, + "loss": 0.0441, + "step": 23288 + }, + { + "epoch": 3.05, + "learning_rate": 1.4195330658397563e-05, + "loss": 0.0463, + "step": 23289 + }, + { + "epoch": 3.05, + "learning_rate": 1.4194849255194619e-05, + "loss": 0.0385, + "step": 23290 + }, + { + "epoch": 3.05, + "learning_rate": 1.4194367840193878e-05, + "loss": 0.0385, + "step": 23291 + }, + { + "epoch": 3.05, + "learning_rate": 1.4193886413396707e-05, + "loss": 0.0497, + "step": 23292 + }, + { + "epoch": 3.05, + "learning_rate": 1.419340497480445e-05, + "loss": 0.0353, + "step": 23293 + }, + { + "epoch": 3.05, + "learning_rate": 1.419292352441847e-05, + "loss": 0.0374, + "step": 23294 + }, + { + "epoch": 3.05, + "learning_rate": 1.4192442062240108e-05, + "loss": 0.0395, + "step": 23295 + }, + { + "epoch": 3.05, + "learning_rate": 1.4191960588270732e-05, + "loss": 0.0354, + "step": 23296 + }, + { + "epoch": 3.05, + "learning_rate": 1.4191479102511687e-05, + "loss": 0.0392, + "step": 23297 + }, + { + "epoch": 3.05, + "learning_rate": 1.4190997604964335e-05, + "loss": 0.0437, + "step": 23298 + }, + { + "epoch": 3.05, + "learning_rate": 1.4190516095630021e-05, + "loss": 0.0447, + "step": 23299 + }, + { + "epoch": 3.05, + "learning_rate": 1.4190034574510103e-05, + "loss": 0.0461, + "step": 23300 + }, + { + "epoch": 3.05, + "learning_rate": 1.4189553041605937e-05, + "loss": 0.0381, + "step": 23301 + }, + { + "epoch": 3.05, + "learning_rate": 1.4189071496918878e-05, + "loss": 0.0426, + "step": 23302 + }, + { + "epoch": 3.05, + "learning_rate": 1.4188589940450275e-05, + "loss": 0.046, + "step": 23303 + }, + { + "epoch": 3.05, + "learning_rate": 1.418810837220149e-05, + "loss": 0.0372, + "step": 23304 + }, + { + "epoch": 3.05, + "learning_rate": 1.418762679217387e-05, + "loss": 0.0423, + "step": 23305 + }, + { + "epoch": 3.05, + "learning_rate": 1.4187145200368772e-05, + "loss": 0.0456, + "step": 23306 + }, + { + "epoch": 3.05, + "learning_rate": 1.4186663596787554e-05, + "loss": 0.0439, + "step": 23307 + }, + { + "epoch": 3.05, + "learning_rate": 1.4186181981431568e-05, + "loss": 0.0429, + "step": 23308 + }, + { + "epoch": 3.05, + "learning_rate": 1.4185700354302166e-05, + "loss": 0.0391, + "step": 23309 + }, + { + "epoch": 3.05, + "learning_rate": 1.4185218715400704e-05, + "loss": 0.0456, + "step": 23310 + }, + { + "epoch": 3.05, + "learning_rate": 1.4184737064728538e-05, + "loss": 0.0439, + "step": 23311 + }, + { + "epoch": 3.05, + "learning_rate": 1.4184255402287022e-05, + "loss": 0.0456, + "step": 23312 + }, + { + "epoch": 3.05, + "learning_rate": 1.4183773728077513e-05, + "loss": 0.0443, + "step": 23313 + }, + { + "epoch": 3.05, + "learning_rate": 1.418329204210136e-05, + "loss": 0.0375, + "step": 23314 + }, + { + "epoch": 3.05, + "learning_rate": 1.4182810344359923e-05, + "loss": 0.0477, + "step": 23315 + }, + { + "epoch": 3.05, + "learning_rate": 1.4182328634854551e-05, + "loss": 0.0413, + "step": 23316 + }, + { + "epoch": 3.05, + "learning_rate": 1.4181846913586606e-05, + "loss": 0.0466, + "step": 23317 + }, + { + "epoch": 3.05, + "learning_rate": 1.4181365180557438e-05, + "loss": 0.0397, + "step": 23318 + }, + { + "epoch": 3.05, + "learning_rate": 1.4180883435768404e-05, + "loss": 0.0488, + "step": 23319 + }, + { + "epoch": 3.05, + "learning_rate": 1.4180401679220857e-05, + "loss": 0.0497, + "step": 23320 + }, + { + "epoch": 3.05, + "learning_rate": 1.4179919910916155e-05, + "loss": 0.0413, + "step": 23321 + }, + { + "epoch": 3.06, + "learning_rate": 1.4179438130855646e-05, + "loss": 0.0435, + "step": 23322 + }, + { + "epoch": 3.06, + "learning_rate": 1.4178956339040694e-05, + "loss": 0.0451, + "step": 23323 + }, + { + "epoch": 3.06, + "learning_rate": 1.4178474535472649e-05, + "loss": 0.0339, + "step": 23324 + }, + { + "epoch": 3.06, + "learning_rate": 1.4177992720152869e-05, + "loss": 0.04, + "step": 23325 + }, + { + "epoch": 3.06, + "learning_rate": 1.4177510893082702e-05, + "loss": 0.0463, + "step": 23326 + }, + { + "epoch": 3.06, + "learning_rate": 1.4177029054263512e-05, + "loss": 0.0425, + "step": 23327 + }, + { + "epoch": 3.06, + "learning_rate": 1.4176547203696647e-05, + "loss": 0.0458, + "step": 23328 + }, + { + "epoch": 3.06, + "learning_rate": 1.4176065341383469e-05, + "loss": 0.0455, + "step": 23329 + }, + { + "epoch": 3.06, + "learning_rate": 1.4175583467325326e-05, + "loss": 0.0414, + "step": 23330 + }, + { + "epoch": 3.06, + "learning_rate": 1.417510158152358e-05, + "loss": 0.0423, + "step": 23331 + }, + { + "epoch": 3.06, + "learning_rate": 1.4174619683979583e-05, + "loss": 0.0479, + "step": 23332 + }, + { + "epoch": 3.06, + "learning_rate": 1.4174137774694687e-05, + "loss": 0.0431, + "step": 23333 + }, + { + "epoch": 3.06, + "learning_rate": 1.4173655853670255e-05, + "loss": 0.0464, + "step": 23334 + }, + { + "epoch": 3.06, + "learning_rate": 1.4173173920907635e-05, + "loss": 0.0389, + "step": 23335 + }, + { + "epoch": 3.06, + "learning_rate": 1.4172691976408188e-05, + "loss": 0.0426, + "step": 23336 + }, + { + "epoch": 3.06, + "learning_rate": 1.4172210020173266e-05, + "loss": 0.0335, + "step": 23337 + }, + { + "epoch": 3.06, + "learning_rate": 1.4171728052204223e-05, + "loss": 0.0452, + "step": 23338 + }, + { + "epoch": 3.06, + "learning_rate": 1.417124607250242e-05, + "loss": 0.0534, + "step": 23339 + }, + { + "epoch": 3.06, + "learning_rate": 1.417076408106921e-05, + "loss": 0.0361, + "step": 23340 + }, + { + "epoch": 3.06, + "learning_rate": 1.4170282077905948e-05, + "loss": 0.0452, + "step": 23341 + }, + { + "epoch": 3.06, + "learning_rate": 1.416980006301399e-05, + "loss": 0.0427, + "step": 23342 + }, + { + "epoch": 3.06, + "learning_rate": 1.4169318036394688e-05, + "loss": 0.0345, + "step": 23343 + }, + { + "epoch": 3.06, + "learning_rate": 1.4168835998049404e-05, + "loss": 0.0425, + "step": 23344 + }, + { + "epoch": 3.06, + "learning_rate": 1.4168353947979489e-05, + "loss": 0.0429, + "step": 23345 + }, + { + "epoch": 3.06, + "learning_rate": 1.4167871886186304e-05, + "loss": 0.0421, + "step": 23346 + }, + { + "epoch": 3.06, + "learning_rate": 1.4167389812671197e-05, + "loss": 0.0384, + "step": 23347 + }, + { + "epoch": 3.06, + "learning_rate": 1.4166907727435533e-05, + "loss": 0.0425, + "step": 23348 + }, + { + "epoch": 3.06, + "learning_rate": 1.4166425630480655e-05, + "loss": 0.0417, + "step": 23349 + }, + { + "epoch": 3.06, + "learning_rate": 1.4165943521807934e-05, + "loss": 0.0462, + "step": 23350 + }, + { + "epoch": 3.06, + "learning_rate": 1.4165461401418714e-05, + "loss": 0.046, + "step": 23351 + }, + { + "epoch": 3.06, + "learning_rate": 1.4164979269314358e-05, + "loss": 0.0391, + "step": 23352 + }, + { + "epoch": 3.06, + "learning_rate": 1.4164497125496222e-05, + "loss": 0.0488, + "step": 23353 + }, + { + "epoch": 3.06, + "learning_rate": 1.4164014969965656e-05, + "loss": 0.0508, + "step": 23354 + }, + { + "epoch": 3.06, + "learning_rate": 1.416353280272402e-05, + "loss": 0.0397, + "step": 23355 + }, + { + "epoch": 3.06, + "learning_rate": 1.4163050623772669e-05, + "loss": 0.043, + "step": 23356 + }, + { + "epoch": 3.06, + "learning_rate": 1.4162568433112964e-05, + "loss": 0.0411, + "step": 23357 + }, + { + "epoch": 3.06, + "learning_rate": 1.4162086230746255e-05, + "loss": 0.0372, + "step": 23358 + }, + { + "epoch": 3.06, + "learning_rate": 1.41616040166739e-05, + "loss": 0.038, + "step": 23359 + }, + { + "epoch": 3.06, + "learning_rate": 1.4161121790897251e-05, + "loss": 0.0362, + "step": 23360 + }, + { + "epoch": 3.06, + "learning_rate": 1.4160639553417675e-05, + "loss": 0.0407, + "step": 23361 + }, + { + "epoch": 3.06, + "learning_rate": 1.4160157304236518e-05, + "loss": 0.0453, + "step": 23362 + }, + { + "epoch": 3.06, + "learning_rate": 1.4159675043355142e-05, + "loss": 0.0465, + "step": 23363 + }, + { + "epoch": 3.06, + "learning_rate": 1.4159192770774901e-05, + "loss": 0.0465, + "step": 23364 + }, + { + "epoch": 3.06, + "learning_rate": 1.4158710486497152e-05, + "loss": 0.0456, + "step": 23365 + }, + { + "epoch": 3.06, + "learning_rate": 1.4158228190523249e-05, + "loss": 0.0399, + "step": 23366 + }, + { + "epoch": 3.06, + "learning_rate": 1.4157745882854555e-05, + "loss": 0.0385, + "step": 23367 + }, + { + "epoch": 3.06, + "learning_rate": 1.4157263563492417e-05, + "loss": 0.0414, + "step": 23368 + }, + { + "epoch": 3.06, + "learning_rate": 1.41567812324382e-05, + "loss": 0.0479, + "step": 23369 + }, + { + "epoch": 3.06, + "learning_rate": 1.4156298889693257e-05, + "loss": 0.036, + "step": 23370 + }, + { + "epoch": 3.06, + "learning_rate": 1.4155816535258943e-05, + "loss": 0.0473, + "step": 23371 + }, + { + "epoch": 3.06, + "learning_rate": 1.4155334169136616e-05, + "loss": 0.0443, + "step": 23372 + }, + { + "epoch": 3.06, + "learning_rate": 1.4154851791327636e-05, + "loss": 0.0395, + "step": 23373 + }, + { + "epoch": 3.06, + "learning_rate": 1.4154369401833355e-05, + "loss": 0.0508, + "step": 23374 + }, + { + "epoch": 3.06, + "learning_rate": 1.4153887000655129e-05, + "loss": 0.0448, + "step": 23375 + }, + { + "epoch": 3.06, + "learning_rate": 1.4153404587794321e-05, + "loss": 0.04, + "step": 23376 + }, + { + "epoch": 3.06, + "learning_rate": 1.415292216325228e-05, + "loss": 0.0471, + "step": 23377 + }, + { + "epoch": 3.06, + "learning_rate": 1.415243972703037e-05, + "loss": 0.0429, + "step": 23378 + }, + { + "epoch": 3.06, + "learning_rate": 1.4151957279129942e-05, + "loss": 0.0413, + "step": 23379 + }, + { + "epoch": 3.06, + "learning_rate": 1.4151474819552358e-05, + "loss": 0.0471, + "step": 23380 + }, + { + "epoch": 3.06, + "learning_rate": 1.4150992348298969e-05, + "loss": 0.0464, + "step": 23381 + }, + { + "epoch": 3.06, + "learning_rate": 1.4150509865371135e-05, + "loss": 0.04, + "step": 23382 + }, + { + "epoch": 3.06, + "learning_rate": 1.4150027370770215e-05, + "loss": 0.0481, + "step": 23383 + }, + { + "epoch": 3.06, + "learning_rate": 1.4149544864497565e-05, + "loss": 0.0527, + "step": 23384 + }, + { + "epoch": 3.06, + "learning_rate": 1.4149062346554539e-05, + "loss": 0.0504, + "step": 23385 + }, + { + "epoch": 3.06, + "learning_rate": 1.4148579816942496e-05, + "loss": 0.0401, + "step": 23386 + }, + { + "epoch": 3.06, + "learning_rate": 1.4148097275662793e-05, + "loss": 0.0374, + "step": 23387 + }, + { + "epoch": 3.06, + "learning_rate": 1.4147614722716788e-05, + "loss": 0.0361, + "step": 23388 + }, + { + "epoch": 3.06, + "learning_rate": 1.4147132158105836e-05, + "loss": 0.0369, + "step": 23389 + }, + { + "epoch": 3.06, + "learning_rate": 1.41466495818313e-05, + "loss": 0.0428, + "step": 23390 + }, + { + "epoch": 3.06, + "learning_rate": 1.4146166993894527e-05, + "loss": 0.0494, + "step": 23391 + }, + { + "epoch": 3.06, + "learning_rate": 1.4145684394296883e-05, + "loss": 0.0419, + "step": 23392 + }, + { + "epoch": 3.06, + "learning_rate": 1.4145201783039724e-05, + "loss": 0.042, + "step": 23393 + }, + { + "epoch": 3.06, + "learning_rate": 1.4144719160124405e-05, + "loss": 0.0424, + "step": 23394 + }, + { + "epoch": 3.06, + "learning_rate": 1.4144236525552284e-05, + "loss": 0.0401, + "step": 23395 + }, + { + "epoch": 3.06, + "learning_rate": 1.4143753879324718e-05, + "loss": 0.0412, + "step": 23396 + }, + { + "epoch": 3.06, + "learning_rate": 1.4143271221443064e-05, + "loss": 0.0402, + "step": 23397 + }, + { + "epoch": 3.06, + "learning_rate": 1.414278855190868e-05, + "loss": 0.0406, + "step": 23398 + }, + { + "epoch": 3.07, + "learning_rate": 1.4142305870722926e-05, + "loss": 0.0403, + "step": 23399 + }, + { + "epoch": 3.07, + "learning_rate": 1.4141823177887155e-05, + "loss": 0.0455, + "step": 23400 + }, + { + "epoch": 3.07, + "learning_rate": 1.4141340473402729e-05, + "loss": 0.0387, + "step": 23401 + }, + { + "epoch": 3.07, + "learning_rate": 1.4140857757271002e-05, + "loss": 0.034, + "step": 23402 + }, + { + "epoch": 3.07, + "learning_rate": 1.4140375029493334e-05, + "loss": 0.049, + "step": 23403 + }, + { + "epoch": 3.07, + "learning_rate": 1.413989229007108e-05, + "loss": 0.0455, + "step": 23404 + }, + { + "epoch": 3.07, + "learning_rate": 1.4139409539005601e-05, + "loss": 0.044, + "step": 23405 + }, + { + "epoch": 3.07, + "learning_rate": 1.4138926776298255e-05, + "loss": 0.0421, + "step": 23406 + }, + { + "epoch": 3.07, + "learning_rate": 1.4138444001950395e-05, + "loss": 0.0407, + "step": 23407 + }, + { + "epoch": 3.07, + "learning_rate": 1.413796121596338e-05, + "loss": 0.0404, + "step": 23408 + }, + { + "epoch": 3.07, + "learning_rate": 1.4137478418338573e-05, + "loss": 0.043, + "step": 23409 + }, + { + "epoch": 3.07, + "learning_rate": 1.4136995609077326e-05, + "loss": 0.0409, + "step": 23410 + }, + { + "epoch": 3.07, + "learning_rate": 1.4136512788181002e-05, + "loss": 0.0391, + "step": 23411 + }, + { + "epoch": 3.07, + "learning_rate": 1.4136029955650954e-05, + "loss": 0.0479, + "step": 23412 + }, + { + "epoch": 3.07, + "learning_rate": 1.413554711148854e-05, + "loss": 0.0473, + "step": 23413 + }, + { + "epoch": 3.07, + "learning_rate": 1.4135064255695126e-05, + "loss": 0.0466, + "step": 23414 + }, + { + "epoch": 3.07, + "learning_rate": 1.4134581388272058e-05, + "loss": 0.0507, + "step": 23415 + }, + { + "epoch": 3.07, + "learning_rate": 1.4134098509220702e-05, + "loss": 0.0487, + "step": 23416 + }, + { + "epoch": 3.07, + "learning_rate": 1.4133615618542414e-05, + "loss": 0.0479, + "step": 23417 + }, + { + "epoch": 3.07, + "learning_rate": 1.4133132716238554e-05, + "loss": 0.0473, + "step": 23418 + }, + { + "epoch": 3.07, + "learning_rate": 1.4132649802310477e-05, + "loss": 0.0345, + "step": 23419 + }, + { + "epoch": 3.07, + "learning_rate": 1.4132166876759542e-05, + "loss": 0.0378, + "step": 23420 + }, + { + "epoch": 3.07, + "learning_rate": 1.4131683939587107e-05, + "loss": 0.0319, + "step": 23421 + }, + { + "epoch": 3.07, + "learning_rate": 1.4131200990794533e-05, + "loss": 0.0416, + "step": 23422 + }, + { + "epoch": 3.07, + "learning_rate": 1.4130718030383174e-05, + "loss": 0.0429, + "step": 23423 + }, + { + "epoch": 3.07, + "learning_rate": 1.4130235058354393e-05, + "loss": 0.044, + "step": 23424 + }, + { + "epoch": 3.07, + "learning_rate": 1.4129752074709544e-05, + "loss": 0.0442, + "step": 23425 + }, + { + "epoch": 3.07, + "learning_rate": 1.4129269079449987e-05, + "loss": 0.0444, + "step": 23426 + }, + { + "epoch": 3.07, + "learning_rate": 1.4128786072577082e-05, + "loss": 0.0413, + "step": 23427 + }, + { + "epoch": 3.07, + "learning_rate": 1.4128303054092188e-05, + "loss": 0.0435, + "step": 23428 + }, + { + "epoch": 3.07, + "learning_rate": 1.4127820023996657e-05, + "loss": 0.04, + "step": 23429 + }, + { + "epoch": 3.07, + "learning_rate": 1.4127336982291855e-05, + "loss": 0.0498, + "step": 23430 + }, + { + "epoch": 3.07, + "learning_rate": 1.4126853928979134e-05, + "loss": 0.0458, + "step": 23431 + }, + { + "epoch": 3.07, + "learning_rate": 1.412637086405986e-05, + "loss": 0.0398, + "step": 23432 + }, + { + "epoch": 3.07, + "learning_rate": 1.4125887787535385e-05, + "loss": 0.0422, + "step": 23433 + }, + { + "epoch": 3.07, + "learning_rate": 1.4125404699407074e-05, + "loss": 0.047, + "step": 23434 + }, + { + "epoch": 3.07, + "learning_rate": 1.412492159967628e-05, + "loss": 0.0497, + "step": 23435 + }, + { + "epoch": 3.07, + "learning_rate": 1.4124438488344363e-05, + "loss": 0.0501, + "step": 23436 + }, + { + "epoch": 3.07, + "learning_rate": 1.412395536541268e-05, + "loss": 0.0452, + "step": 23437 + }, + { + "epoch": 3.07, + "learning_rate": 1.4123472230882595e-05, + "loss": 0.0403, + "step": 23438 + }, + { + "epoch": 3.07, + "learning_rate": 1.4122989084755465e-05, + "loss": 0.041, + "step": 23439 + }, + { + "epoch": 3.07, + "learning_rate": 1.4122505927032645e-05, + "loss": 0.0444, + "step": 23440 + }, + { + "epoch": 3.07, + "learning_rate": 1.4122022757715501e-05, + "loss": 0.0332, + "step": 23441 + }, + { + "epoch": 3.07, + "learning_rate": 1.4121539576805382e-05, + "loss": 0.0362, + "step": 23442 + }, + { + "epoch": 3.07, + "learning_rate": 1.4121056384303655e-05, + "loss": 0.034, + "step": 23443 + }, + { + "epoch": 3.07, + "learning_rate": 1.4120573180211677e-05, + "loss": 0.0398, + "step": 23444 + }, + { + "epoch": 3.07, + "learning_rate": 1.4120089964530807e-05, + "loss": 0.0401, + "step": 23445 + }, + { + "epoch": 3.07, + "learning_rate": 1.4119606737262401e-05, + "loss": 0.0374, + "step": 23446 + }, + { + "epoch": 3.07, + "learning_rate": 1.4119123498407823e-05, + "loss": 0.0413, + "step": 23447 + }, + { + "epoch": 3.07, + "learning_rate": 1.4118640247968424e-05, + "loss": 0.0436, + "step": 23448 + }, + { + "epoch": 3.07, + "learning_rate": 1.4118156985945578e-05, + "loss": 0.0393, + "step": 23449 + }, + { + "epoch": 3.07, + "learning_rate": 1.4117673712340627e-05, + "loss": 0.0477, + "step": 23450 + }, + { + "epoch": 3.07, + "learning_rate": 1.4117190427154941e-05, + "loss": 0.0436, + "step": 23451 + }, + { + "epoch": 3.07, + "learning_rate": 1.4116707130389876e-05, + "loss": 0.0368, + "step": 23452 + }, + { + "epoch": 3.07, + "learning_rate": 1.411622382204679e-05, + "loss": 0.0485, + "step": 23453 + }, + { + "epoch": 3.07, + "learning_rate": 1.4115740502127044e-05, + "loss": 0.0428, + "step": 23454 + }, + { + "epoch": 3.07, + "learning_rate": 1.4115257170632e-05, + "loss": 0.0423, + "step": 23455 + }, + { + "epoch": 3.07, + "learning_rate": 1.4114773827563011e-05, + "loss": 0.0325, + "step": 23456 + }, + { + "epoch": 3.07, + "learning_rate": 1.4114290472921441e-05, + "loss": 0.0385, + "step": 23457 + }, + { + "epoch": 3.07, + "learning_rate": 1.4113807106708648e-05, + "loss": 0.0359, + "step": 23458 + }, + { + "epoch": 3.07, + "learning_rate": 1.4113323728925992e-05, + "loss": 0.0396, + "step": 23459 + }, + { + "epoch": 3.07, + "learning_rate": 1.4112840339574832e-05, + "loss": 0.0376, + "step": 23460 + }, + { + "epoch": 3.07, + "learning_rate": 1.4112356938656527e-05, + "loss": 0.0467, + "step": 23461 + }, + { + "epoch": 3.07, + "learning_rate": 1.4111873526172438e-05, + "loss": 0.0327, + "step": 23462 + }, + { + "epoch": 3.07, + "learning_rate": 1.4111390102123923e-05, + "loss": 0.0534, + "step": 23463 + }, + { + "epoch": 3.07, + "learning_rate": 1.411090666651234e-05, + "loss": 0.0385, + "step": 23464 + }, + { + "epoch": 3.07, + "learning_rate": 1.4110423219339054e-05, + "loss": 0.0404, + "step": 23465 + }, + { + "epoch": 3.07, + "learning_rate": 1.4109939760605422e-05, + "loss": 0.0489, + "step": 23466 + }, + { + "epoch": 3.07, + "learning_rate": 1.4109456290312801e-05, + "loss": 0.0507, + "step": 23467 + }, + { + "epoch": 3.07, + "learning_rate": 1.4108972808462555e-05, + "loss": 0.0373, + "step": 23468 + }, + { + "epoch": 3.07, + "learning_rate": 1.410848931505604e-05, + "loss": 0.0444, + "step": 23469 + }, + { + "epoch": 3.07, + "learning_rate": 1.4108005810094618e-05, + "loss": 0.0493, + "step": 23470 + }, + { + "epoch": 3.07, + "learning_rate": 1.410752229357965e-05, + "loss": 0.0601, + "step": 23471 + }, + { + "epoch": 3.07, + "learning_rate": 1.4107038765512494e-05, + "loss": 0.0395, + "step": 23472 + }, + { + "epoch": 3.07, + "learning_rate": 1.4106555225894509e-05, + "loss": 0.0467, + "step": 23473 + }, + { + "epoch": 3.07, + "learning_rate": 1.4106071674727058e-05, + "loss": 0.0479, + "step": 23474 + }, + { + "epoch": 3.08, + "learning_rate": 1.4105588112011494e-05, + "loss": 0.0356, + "step": 23475 + }, + { + "epoch": 3.08, + "learning_rate": 1.4105104537749188e-05, + "loss": 0.0445, + "step": 23476 + }, + { + "epoch": 3.08, + "learning_rate": 1.410462095194149e-05, + "loss": 0.0413, + "step": 23477 + }, + { + "epoch": 3.08, + "learning_rate": 1.4104137354589766e-05, + "loss": 0.0431, + "step": 23478 + }, + { + "epoch": 3.08, + "learning_rate": 1.4103653745695375e-05, + "loss": 0.0523, + "step": 23479 + }, + { + "epoch": 3.08, + "learning_rate": 1.4103170125259675e-05, + "loss": 0.0432, + "step": 23480 + }, + { + "epoch": 3.08, + "learning_rate": 1.4102686493284028e-05, + "loss": 0.0555, + "step": 23481 + }, + { + "epoch": 3.08, + "learning_rate": 1.4102202849769793e-05, + "loss": 0.0408, + "step": 23482 + }, + { + "epoch": 3.08, + "learning_rate": 1.4101719194718332e-05, + "loss": 0.0461, + "step": 23483 + }, + { + "epoch": 3.08, + "learning_rate": 1.4101235528131005e-05, + "loss": 0.0487, + "step": 23484 + }, + { + "epoch": 3.08, + "learning_rate": 1.4100751850009172e-05, + "loss": 0.0457, + "step": 23485 + }, + { + "epoch": 3.08, + "learning_rate": 1.410026816035419e-05, + "loss": 0.0448, + "step": 23486 + }, + { + "epoch": 3.08, + "learning_rate": 1.4099784459167423e-05, + "loss": 0.0401, + "step": 23487 + }, + { + "epoch": 3.08, + "learning_rate": 1.409930074645023e-05, + "loss": 0.0449, + "step": 23488 + }, + { + "epoch": 3.08, + "learning_rate": 1.4098817022203974e-05, + "loss": 0.045, + "step": 23489 + }, + { + "epoch": 3.08, + "learning_rate": 1.4098333286430012e-05, + "loss": 0.0373, + "step": 23490 + }, + { + "epoch": 3.08, + "learning_rate": 1.4097849539129706e-05, + "loss": 0.0441, + "step": 23491 + }, + { + "epoch": 3.08, + "learning_rate": 1.4097365780304416e-05, + "loss": 0.0481, + "step": 23492 + }, + { + "epoch": 3.08, + "learning_rate": 1.4096882009955506e-05, + "loss": 0.0461, + "step": 23493 + }, + { + "epoch": 3.08, + "learning_rate": 1.409639822808433e-05, + "loss": 0.0546, + "step": 23494 + }, + { + "epoch": 3.08, + "learning_rate": 1.4095914434692253e-05, + "loss": 0.0435, + "step": 23495 + }, + { + "epoch": 3.08, + "learning_rate": 1.4095430629780636e-05, + "loss": 0.046, + "step": 23496 + }, + { + "epoch": 3.08, + "learning_rate": 1.4094946813350837e-05, + "loss": 0.0459, + "step": 23497 + }, + { + "epoch": 3.08, + "learning_rate": 1.4094462985404215e-05, + "loss": 0.044, + "step": 23498 + }, + { + "epoch": 3.08, + "learning_rate": 1.409397914594214e-05, + "loss": 0.0383, + "step": 23499 + }, + { + "epoch": 3.08, + "learning_rate": 1.4093495294965964e-05, + "loss": 0.0444, + "step": 23500 + }, + { + "epoch": 3.08, + "learning_rate": 1.4093011432477051e-05, + "loss": 0.0455, + "step": 23501 + }, + { + "epoch": 3.08, + "learning_rate": 1.4092527558476759e-05, + "loss": 0.0375, + "step": 23502 + }, + { + "epoch": 3.08, + "learning_rate": 1.4092043672966451e-05, + "loss": 0.0449, + "step": 23503 + }, + { + "epoch": 3.08, + "learning_rate": 1.4091559775947495e-05, + "loss": 0.0453, + "step": 23504 + }, + { + "epoch": 3.08, + "learning_rate": 1.4091075867421238e-05, + "loss": 0.037, + "step": 23505 + }, + { + "epoch": 3.08, + "learning_rate": 1.4090591947389052e-05, + "loss": 0.0384, + "step": 23506 + }, + { + "epoch": 3.08, + "learning_rate": 1.409010801585229e-05, + "loss": 0.0411, + "step": 23507 + }, + { + "epoch": 3.08, + "learning_rate": 1.4089624072812318e-05, + "loss": 0.0398, + "step": 23508 + }, + { + "epoch": 3.08, + "learning_rate": 1.4089140118270499e-05, + "loss": 0.0357, + "step": 23509 + }, + { + "epoch": 3.08, + "learning_rate": 1.4088656152228191e-05, + "loss": 0.0394, + "step": 23510 + }, + { + "epoch": 3.08, + "learning_rate": 1.4088172174686752e-05, + "loss": 0.0543, + "step": 23511 + }, + { + "epoch": 3.08, + "learning_rate": 1.408768818564755e-05, + "loss": 0.0396, + "step": 23512 + }, + { + "epoch": 3.08, + "learning_rate": 1.4087204185111938e-05, + "loss": 0.035, + "step": 23513 + }, + { + "epoch": 3.08, + "learning_rate": 1.4086720173081286e-05, + "loss": 0.0425, + "step": 23514 + }, + { + "epoch": 3.08, + "learning_rate": 1.408623614955695e-05, + "loss": 0.0463, + "step": 23515 + }, + { + "epoch": 3.08, + "learning_rate": 1.4085752114540291e-05, + "loss": 0.0371, + "step": 23516 + }, + { + "epoch": 3.08, + "learning_rate": 1.4085268068032673e-05, + "loss": 0.0424, + "step": 23517 + }, + { + "epoch": 3.08, + "learning_rate": 1.4084784010035457e-05, + "loss": 0.0435, + "step": 23518 + }, + { + "epoch": 3.08, + "learning_rate": 1.4084299940550002e-05, + "loss": 0.0473, + "step": 23519 + }, + { + "epoch": 3.08, + "learning_rate": 1.4083815859577671e-05, + "loss": 0.0458, + "step": 23520 + }, + { + "epoch": 3.08, + "learning_rate": 1.4083331767119828e-05, + "loss": 0.0449, + "step": 23521 + }, + { + "epoch": 3.08, + "learning_rate": 1.4082847663177828e-05, + "loss": 0.0414, + "step": 23522 + }, + { + "epoch": 3.08, + "learning_rate": 1.408236354775304e-05, + "loss": 0.0511, + "step": 23523 + }, + { + "epoch": 3.08, + "learning_rate": 1.4081879420846818e-05, + "loss": 0.0374, + "step": 23524 + }, + { + "epoch": 3.08, + "learning_rate": 1.4081395282460531e-05, + "loss": 0.0405, + "step": 23525 + }, + { + "epoch": 3.08, + "learning_rate": 1.4080911132595533e-05, + "loss": 0.0533, + "step": 23526 + }, + { + "epoch": 3.08, + "learning_rate": 1.4080426971253193e-05, + "loss": 0.0416, + "step": 23527 + }, + { + "epoch": 3.08, + "learning_rate": 1.4079942798434871e-05, + "loss": 0.0433, + "step": 23528 + }, + { + "epoch": 3.08, + "learning_rate": 1.4079458614141926e-05, + "loss": 0.0388, + "step": 23529 + }, + { + "epoch": 3.08, + "learning_rate": 1.4078974418375718e-05, + "loss": 0.0392, + "step": 23530 + }, + { + "epoch": 3.08, + "learning_rate": 1.4078490211137614e-05, + "loss": 0.0373, + "step": 23531 + }, + { + "epoch": 3.08, + "learning_rate": 1.4078005992428972e-05, + "loss": 0.0393, + "step": 23532 + }, + { + "epoch": 3.08, + "learning_rate": 1.4077521762251159e-05, + "loss": 0.0335, + "step": 23533 + }, + { + "epoch": 3.08, + "learning_rate": 1.407703752060553e-05, + "loss": 0.0369, + "step": 23534 + }, + { + "epoch": 3.08, + "learning_rate": 1.4076553267493448e-05, + "loss": 0.0419, + "step": 23535 + }, + { + "epoch": 3.08, + "learning_rate": 1.407606900291628e-05, + "loss": 0.0492, + "step": 23536 + }, + { + "epoch": 3.08, + "learning_rate": 1.4075584726875387e-05, + "loss": 0.0357, + "step": 23537 + }, + { + "epoch": 3.08, + "learning_rate": 1.4075100439372124e-05, + "loss": 0.04, + "step": 23538 + }, + { + "epoch": 3.08, + "learning_rate": 1.4074616140407864e-05, + "loss": 0.0476, + "step": 23539 + }, + { + "epoch": 3.08, + "learning_rate": 1.4074131829983957e-05, + "loss": 0.0476, + "step": 23540 + }, + { + "epoch": 3.08, + "learning_rate": 1.4073647508101775e-05, + "loss": 0.0427, + "step": 23541 + }, + { + "epoch": 3.08, + "learning_rate": 1.4073163174762673e-05, + "loss": 0.0419, + "step": 23542 + }, + { + "epoch": 3.08, + "learning_rate": 1.407267882996802e-05, + "loss": 0.0406, + "step": 23543 + }, + { + "epoch": 3.08, + "learning_rate": 1.4072194473719174e-05, + "loss": 0.0486, + "step": 23544 + }, + { + "epoch": 3.08, + "learning_rate": 1.4071710106017498e-05, + "loss": 0.0403, + "step": 23545 + }, + { + "epoch": 3.08, + "learning_rate": 1.4071225726864353e-05, + "loss": 0.044, + "step": 23546 + }, + { + "epoch": 3.08, + "learning_rate": 1.4070741336261102e-05, + "loss": 0.0462, + "step": 23547 + }, + { + "epoch": 3.08, + "learning_rate": 1.407025693420911e-05, + "loss": 0.039, + "step": 23548 + }, + { + "epoch": 3.08, + "learning_rate": 1.4069772520709735e-05, + "loss": 0.0401, + "step": 23549 + }, + { + "epoch": 3.08, + "learning_rate": 1.4069288095764344e-05, + "loss": 0.0414, + "step": 23550 + }, + { + "epoch": 3.09, + "learning_rate": 1.4068803659374293e-05, + "loss": 0.0442, + "step": 23551 + }, + { + "epoch": 3.09, + "learning_rate": 1.4068319211540953e-05, + "loss": 0.0422, + "step": 23552 + }, + { + "epoch": 3.09, + "learning_rate": 1.4067834752265679e-05, + "loss": 0.0431, + "step": 23553 + }, + { + "epoch": 3.09, + "learning_rate": 1.4067350281549837e-05, + "loss": 0.0379, + "step": 23554 + }, + { + "epoch": 3.09, + "learning_rate": 1.406686579939479e-05, + "loss": 0.0429, + "step": 23555 + }, + { + "epoch": 3.09, + "learning_rate": 1.4066381305801899e-05, + "loss": 0.0445, + "step": 23556 + }, + { + "epoch": 3.09, + "learning_rate": 1.4065896800772523e-05, + "loss": 0.0489, + "step": 23557 + }, + { + "epoch": 3.09, + "learning_rate": 1.4065412284308034e-05, + "loss": 0.0418, + "step": 23558 + }, + { + "epoch": 3.09, + "learning_rate": 1.4064927756409787e-05, + "loss": 0.0425, + "step": 23559 + }, + { + "epoch": 3.09, + "learning_rate": 1.4064443217079148e-05, + "loss": 0.0407, + "step": 23560 + }, + { + "epoch": 3.09, + "learning_rate": 1.4063958666317478e-05, + "loss": 0.0416, + "step": 23561 + }, + { + "epoch": 3.09, + "learning_rate": 1.4063474104126142e-05, + "loss": 0.0424, + "step": 23562 + }, + { + "epoch": 3.09, + "learning_rate": 1.4062989530506502e-05, + "loss": 0.0417, + "step": 23563 + }, + { + "epoch": 3.09, + "learning_rate": 1.4062504945459918e-05, + "loss": 0.0438, + "step": 23564 + }, + { + "epoch": 3.09, + "learning_rate": 1.4062020348987757e-05, + "loss": 0.0442, + "step": 23565 + }, + { + "epoch": 3.09, + "learning_rate": 1.4061535741091379e-05, + "loss": 0.0441, + "step": 23566 + }, + { + "epoch": 3.09, + "learning_rate": 1.406105112177215e-05, + "loss": 0.0395, + "step": 23567 + }, + { + "epoch": 3.09, + "learning_rate": 1.4060566491031426e-05, + "loss": 0.0487, + "step": 23568 + }, + { + "epoch": 3.09, + "learning_rate": 1.4060081848870581e-05, + "loss": 0.0378, + "step": 23569 + }, + { + "epoch": 3.09, + "learning_rate": 1.4059597195290968e-05, + "loss": 0.0392, + "step": 23570 + }, + { + "epoch": 3.09, + "learning_rate": 1.4059112530293957e-05, + "loss": 0.0395, + "step": 23571 + }, + { + "epoch": 3.09, + "learning_rate": 1.4058627853880905e-05, + "loss": 0.0433, + "step": 23572 + }, + { + "epoch": 3.09, + "learning_rate": 1.405814316605318e-05, + "loss": 0.0417, + "step": 23573 + }, + { + "epoch": 3.09, + "learning_rate": 1.4057658466812144e-05, + "loss": 0.0422, + "step": 23574 + }, + { + "epoch": 3.09, + "learning_rate": 1.4057173756159159e-05, + "loss": 0.0498, + "step": 23575 + }, + { + "epoch": 3.09, + "learning_rate": 1.405668903409559e-05, + "loss": 0.04, + "step": 23576 + }, + { + "epoch": 3.09, + "learning_rate": 1.40562043006228e-05, + "loss": 0.0455, + "step": 23577 + }, + { + "epoch": 3.09, + "learning_rate": 1.4055719555742149e-05, + "loss": 0.0483, + "step": 23578 + }, + { + "epoch": 3.09, + "learning_rate": 1.4055234799455002e-05, + "loss": 0.0496, + "step": 23579 + }, + { + "epoch": 3.09, + "learning_rate": 1.4054750031762722e-05, + "loss": 0.0445, + "step": 23580 + }, + { + "epoch": 3.09, + "learning_rate": 1.4054265252666678e-05, + "loss": 0.0377, + "step": 23581 + }, + { + "epoch": 3.09, + "learning_rate": 1.4053780462168228e-05, + "loss": 0.0416, + "step": 23582 + }, + { + "epoch": 3.09, + "learning_rate": 1.4053295660268733e-05, + "loss": 0.0444, + "step": 23583 + }, + { + "epoch": 3.09, + "learning_rate": 1.4052810846969563e-05, + "loss": 0.0491, + "step": 23584 + }, + { + "epoch": 3.09, + "learning_rate": 1.4052326022272076e-05, + "loss": 0.0374, + "step": 23585 + }, + { + "epoch": 3.09, + "learning_rate": 1.4051841186177639e-05, + "loss": 0.0445, + "step": 23586 + }, + { + "epoch": 3.09, + "learning_rate": 1.4051356338687613e-05, + "loss": 0.0458, + "step": 23587 + }, + { + "epoch": 3.09, + "learning_rate": 1.4050871479803367e-05, + "loss": 0.0469, + "step": 23588 + }, + { + "epoch": 3.09, + "learning_rate": 1.405038660952626e-05, + "loss": 0.0393, + "step": 23589 + }, + { + "epoch": 3.09, + "learning_rate": 1.4049901727857651e-05, + "loss": 0.0417, + "step": 23590 + }, + { + "epoch": 3.09, + "learning_rate": 1.4049416834798912e-05, + "loss": 0.0429, + "step": 23591 + }, + { + "epoch": 3.09, + "learning_rate": 1.4048931930351406e-05, + "loss": 0.0447, + "step": 23592 + }, + { + "epoch": 3.09, + "learning_rate": 1.4048447014516492e-05, + "loss": 0.0437, + "step": 23593 + }, + { + "epoch": 3.09, + "learning_rate": 1.4047962087295538e-05, + "loss": 0.0438, + "step": 23594 + }, + { + "epoch": 3.09, + "learning_rate": 1.4047477148689903e-05, + "loss": 0.0464, + "step": 23595 + }, + { + "epoch": 3.09, + "learning_rate": 1.4046992198700959e-05, + "loss": 0.0424, + "step": 23596 + }, + { + "epoch": 3.09, + "learning_rate": 1.4046507237330062e-05, + "loss": 0.0477, + "step": 23597 + }, + { + "epoch": 3.09, + "learning_rate": 1.404602226457858e-05, + "loss": 0.032, + "step": 23598 + }, + { + "epoch": 3.09, + "learning_rate": 1.4045537280447876e-05, + "loss": 0.0416, + "step": 23599 + }, + { + "epoch": 3.09, + "learning_rate": 1.4045052284939313e-05, + "loss": 0.0408, + "step": 23600 + }, + { + "epoch": 3.09, + "learning_rate": 1.4044567278054254e-05, + "loss": 0.0406, + "step": 23601 + }, + { + "epoch": 3.09, + "learning_rate": 1.4044082259794069e-05, + "loss": 0.0432, + "step": 23602 + }, + { + "epoch": 3.09, + "learning_rate": 1.4043597230160115e-05, + "loss": 0.0461, + "step": 23603 + }, + { + "epoch": 3.09, + "learning_rate": 1.404311218915376e-05, + "loss": 0.0439, + "step": 23604 + }, + { + "epoch": 3.09, + "learning_rate": 1.4042627136776369e-05, + "loss": 0.0466, + "step": 23605 + }, + { + "epoch": 3.09, + "learning_rate": 1.40421420730293e-05, + "loss": 0.0488, + "step": 23606 + }, + { + "epoch": 3.09, + "learning_rate": 1.4041656997913924e-05, + "loss": 0.0418, + "step": 23607 + }, + { + "epoch": 3.09, + "learning_rate": 1.4041171911431603e-05, + "loss": 0.0429, + "step": 23608 + }, + { + "epoch": 3.09, + "learning_rate": 1.4040686813583704e-05, + "loss": 0.0482, + "step": 23609 + }, + { + "epoch": 3.09, + "learning_rate": 1.4040201704371585e-05, + "loss": 0.04, + "step": 23610 + }, + { + "epoch": 3.09, + "learning_rate": 1.4039716583796614e-05, + "loss": 0.0464, + "step": 23611 + }, + { + "epoch": 3.09, + "learning_rate": 1.4039231451860155e-05, + "loss": 0.0401, + "step": 23612 + }, + { + "epoch": 3.09, + "learning_rate": 1.4038746308563575e-05, + "loss": 0.0406, + "step": 23613 + }, + { + "epoch": 3.09, + "learning_rate": 1.4038261153908235e-05, + "loss": 0.0479, + "step": 23614 + }, + { + "epoch": 3.09, + "learning_rate": 1.4037775987895503e-05, + "loss": 0.0384, + "step": 23615 + }, + { + "epoch": 3.09, + "learning_rate": 1.4037290810526738e-05, + "loss": 0.038, + "step": 23616 + }, + { + "epoch": 3.09, + "learning_rate": 1.4036805621803307e-05, + "loss": 0.0453, + "step": 23617 + }, + { + "epoch": 3.09, + "learning_rate": 1.4036320421726574e-05, + "loss": 0.0434, + "step": 23618 + }, + { + "epoch": 3.09, + "learning_rate": 1.403583521029791e-05, + "loss": 0.0448, + "step": 23619 + }, + { + "epoch": 3.09, + "learning_rate": 1.4035349987518672e-05, + "loss": 0.0432, + "step": 23620 + }, + { + "epoch": 3.09, + "learning_rate": 1.4034864753390224e-05, + "loss": 0.0399, + "step": 23621 + }, + { + "epoch": 3.09, + "learning_rate": 1.4034379507913935e-05, + "loss": 0.0487, + "step": 23622 + }, + { + "epoch": 3.09, + "learning_rate": 1.4033894251091172e-05, + "loss": 0.0429, + "step": 23623 + }, + { + "epoch": 3.09, + "learning_rate": 1.4033408982923294e-05, + "loss": 0.0448, + "step": 23624 + }, + { + "epoch": 3.09, + "learning_rate": 1.4032923703411667e-05, + "loss": 0.0367, + "step": 23625 + }, + { + "epoch": 3.09, + "learning_rate": 1.4032438412557658e-05, + "loss": 0.0472, + "step": 23626 + }, + { + "epoch": 3.09, + "learning_rate": 1.403195311036263e-05, + "loss": 0.0398, + "step": 23627 + }, + { + "epoch": 3.1, + "learning_rate": 1.4031467796827948e-05, + "loss": 0.0463, + "step": 23628 + }, + { + "epoch": 3.1, + "learning_rate": 1.4030982471954978e-05, + "loss": 0.0414, + "step": 23629 + }, + { + "epoch": 3.1, + "learning_rate": 1.4030497135745088e-05, + "loss": 0.0432, + "step": 23630 + }, + { + "epoch": 3.1, + "learning_rate": 1.4030011788199636e-05, + "loss": 0.0432, + "step": 23631 + }, + { + "epoch": 3.1, + "learning_rate": 1.4029526429319992e-05, + "loss": 0.0392, + "step": 23632 + }, + { + "epoch": 3.1, + "learning_rate": 1.4029041059107515e-05, + "loss": 0.0424, + "step": 23633 + }, + { + "epoch": 3.1, + "learning_rate": 1.4028555677563581e-05, + "loss": 0.05, + "step": 23634 + }, + { + "epoch": 3.1, + "learning_rate": 1.4028070284689544e-05, + "loss": 0.0494, + "step": 23635 + }, + { + "epoch": 3.1, + "learning_rate": 1.4027584880486778e-05, + "loss": 0.0423, + "step": 23636 + }, + { + "epoch": 3.1, + "learning_rate": 1.402709946495664e-05, + "loss": 0.045, + "step": 23637 + }, + { + "epoch": 3.1, + "learning_rate": 1.4026614038100501e-05, + "loss": 0.0414, + "step": 23638 + }, + { + "epoch": 3.1, + "learning_rate": 1.4026128599919724e-05, + "loss": 0.0341, + "step": 23639 + }, + { + "epoch": 3.1, + "learning_rate": 1.4025643150415675e-05, + "loss": 0.0381, + "step": 23640 + }, + { + "epoch": 3.1, + "learning_rate": 1.4025157689589719e-05, + "loss": 0.0352, + "step": 23641 + }, + { + "epoch": 3.1, + "learning_rate": 1.4024672217443222e-05, + "loss": 0.0431, + "step": 23642 + }, + { + "epoch": 3.1, + "learning_rate": 1.4024186733977548e-05, + "loss": 0.0414, + "step": 23643 + }, + { + "epoch": 3.1, + "learning_rate": 1.4023701239194063e-05, + "loss": 0.0436, + "step": 23644 + }, + { + "epoch": 3.1, + "learning_rate": 1.4023215733094132e-05, + "loss": 0.0491, + "step": 23645 + }, + { + "epoch": 3.1, + "learning_rate": 1.4022730215679121e-05, + "loss": 0.0416, + "step": 23646 + }, + { + "epoch": 3.1, + "learning_rate": 1.4022244686950396e-05, + "loss": 0.0354, + "step": 23647 + }, + { + "epoch": 3.1, + "learning_rate": 1.4021759146909321e-05, + "loss": 0.0378, + "step": 23648 + }, + { + "epoch": 3.1, + "learning_rate": 1.4021273595557268e-05, + "loss": 0.0432, + "step": 23649 + }, + { + "epoch": 3.1, + "learning_rate": 1.4020788032895592e-05, + "loss": 0.0438, + "step": 23650 + }, + { + "epoch": 3.1, + "learning_rate": 1.4020302458925664e-05, + "loss": 0.0363, + "step": 23651 + }, + { + "epoch": 3.1, + "learning_rate": 1.4019816873648848e-05, + "loss": 0.0418, + "step": 23652 + }, + { + "epoch": 3.1, + "learning_rate": 1.4019331277066515e-05, + "loss": 0.0386, + "step": 23653 + }, + { + "epoch": 3.1, + "learning_rate": 1.4018845669180023e-05, + "loss": 0.0511, + "step": 23654 + }, + { + "epoch": 3.1, + "learning_rate": 1.4018360049990745e-05, + "loss": 0.045, + "step": 23655 + }, + { + "epoch": 3.1, + "learning_rate": 1.4017874419500042e-05, + "loss": 0.0527, + "step": 23656 + }, + { + "epoch": 3.1, + "learning_rate": 1.4017388777709282e-05, + "loss": 0.0435, + "step": 23657 + }, + { + "epoch": 3.1, + "learning_rate": 1.4016903124619827e-05, + "loss": 0.0461, + "step": 23658 + }, + { + "epoch": 3.1, + "learning_rate": 1.4016417460233051e-05, + "loss": 0.046, + "step": 23659 + }, + { + "epoch": 3.1, + "learning_rate": 1.4015931784550309e-05, + "loss": 0.0512, + "step": 23660 + }, + { + "epoch": 3.1, + "learning_rate": 1.4015446097572975e-05, + "loss": 0.0341, + "step": 23661 + }, + { + "epoch": 3.1, + "learning_rate": 1.4014960399302413e-05, + "loss": 0.0469, + "step": 23662 + }, + { + "epoch": 3.1, + "learning_rate": 1.4014474689739991e-05, + "loss": 0.0399, + "step": 23663 + }, + { + "epoch": 3.1, + "learning_rate": 1.4013988968887069e-05, + "loss": 0.0424, + "step": 23664 + }, + { + "epoch": 3.1, + "learning_rate": 1.4013503236745021e-05, + "loss": 0.0487, + "step": 23665 + }, + { + "epoch": 3.1, + "learning_rate": 1.4013017493315205e-05, + "loss": 0.0391, + "step": 23666 + }, + { + "epoch": 3.1, + "learning_rate": 1.4012531738598991e-05, + "loss": 0.0405, + "step": 23667 + }, + { + "epoch": 3.1, + "learning_rate": 1.4012045972597747e-05, + "loss": 0.0442, + "step": 23668 + }, + { + "epoch": 3.1, + "learning_rate": 1.4011560195312837e-05, + "loss": 0.0469, + "step": 23669 + }, + { + "epoch": 3.1, + "learning_rate": 1.4011074406745629e-05, + "loss": 0.0439, + "step": 23670 + }, + { + "epoch": 3.1, + "learning_rate": 1.4010588606897484e-05, + "loss": 0.0456, + "step": 23671 + }, + { + "epoch": 3.1, + "learning_rate": 1.4010102795769775e-05, + "loss": 0.047, + "step": 23672 + }, + { + "epoch": 3.1, + "learning_rate": 1.4009616973363863e-05, + "loss": 0.0432, + "step": 23673 + }, + { + "epoch": 3.1, + "learning_rate": 1.4009131139681121e-05, + "loss": 0.0389, + "step": 23674 + }, + { + "epoch": 3.1, + "learning_rate": 1.4008645294722906e-05, + "loss": 0.046, + "step": 23675 + }, + { + "epoch": 3.1, + "learning_rate": 1.4008159438490597e-05, + "loss": 0.0354, + "step": 23676 + }, + { + "epoch": 3.1, + "learning_rate": 1.4007673570985544e-05, + "loss": 0.0455, + "step": 23677 + }, + { + "epoch": 3.1, + "learning_rate": 1.4007187692209128e-05, + "loss": 0.0417, + "step": 23678 + }, + { + "epoch": 3.1, + "learning_rate": 1.4006701802162707e-05, + "loss": 0.0462, + "step": 23679 + }, + { + "epoch": 3.1, + "learning_rate": 1.4006215900847655e-05, + "loss": 0.0416, + "step": 23680 + }, + { + "epoch": 3.1, + "learning_rate": 1.400572998826533e-05, + "loss": 0.0459, + "step": 23681 + }, + { + "epoch": 3.1, + "learning_rate": 1.4005244064417104e-05, + "loss": 0.0422, + "step": 23682 + }, + { + "epoch": 3.1, + "learning_rate": 1.4004758129304343e-05, + "loss": 0.0433, + "step": 23683 + }, + { + "epoch": 3.1, + "learning_rate": 1.4004272182928414e-05, + "loss": 0.0448, + "step": 23684 + }, + { + "epoch": 3.1, + "learning_rate": 1.400378622529068e-05, + "loss": 0.0421, + "step": 23685 + }, + { + "epoch": 3.1, + "learning_rate": 1.4003300256392513e-05, + "loss": 0.0388, + "step": 23686 + }, + { + "epoch": 3.1, + "learning_rate": 1.4002814276235275e-05, + "loss": 0.0388, + "step": 23687 + }, + { + "epoch": 3.1, + "learning_rate": 1.4002328284820337e-05, + "loss": 0.0395, + "step": 23688 + }, + { + "epoch": 3.1, + "learning_rate": 1.400184228214906e-05, + "loss": 0.0377, + "step": 23689 + }, + { + "epoch": 3.1, + "learning_rate": 1.4001356268222821e-05, + "loss": 0.0469, + "step": 23690 + }, + { + "epoch": 3.1, + "learning_rate": 1.4000870243042977e-05, + "loss": 0.0366, + "step": 23691 + }, + { + "epoch": 3.1, + "learning_rate": 1.4000384206610897e-05, + "loss": 0.0457, + "step": 23692 + }, + { + "epoch": 3.1, + "learning_rate": 1.3999898158927949e-05, + "loss": 0.0427, + "step": 23693 + }, + { + "epoch": 3.1, + "learning_rate": 1.3999412099995504e-05, + "loss": 0.0432, + "step": 23694 + }, + { + "epoch": 3.1, + "learning_rate": 1.3998926029814925e-05, + "loss": 0.0482, + "step": 23695 + }, + { + "epoch": 3.1, + "learning_rate": 1.399843994838758e-05, + "loss": 0.0449, + "step": 23696 + }, + { + "epoch": 3.1, + "learning_rate": 1.3997953855714834e-05, + "loss": 0.0442, + "step": 23697 + }, + { + "epoch": 3.1, + "learning_rate": 1.3997467751798056e-05, + "loss": 0.0335, + "step": 23698 + }, + { + "epoch": 3.1, + "learning_rate": 1.3996981636638611e-05, + "loss": 0.0381, + "step": 23699 + }, + { + "epoch": 3.1, + "learning_rate": 1.3996495510237869e-05, + "loss": 0.0356, + "step": 23700 + }, + { + "epoch": 3.1, + "learning_rate": 1.3996009372597197e-05, + "loss": 0.0328, + "step": 23701 + }, + { + "epoch": 3.1, + "learning_rate": 1.3995523223717959e-05, + "loss": 0.0398, + "step": 23702 + }, + { + "epoch": 3.1, + "learning_rate": 1.399503706360153e-05, + "loss": 0.0442, + "step": 23703 + }, + { + "epoch": 3.11, + "learning_rate": 1.3994550892249265e-05, + "loss": 0.0445, + "step": 23704 + }, + { + "epoch": 3.11, + "learning_rate": 1.3994064709662545e-05, + "loss": 0.0393, + "step": 23705 + }, + { + "epoch": 3.11, + "learning_rate": 1.3993578515842727e-05, + "loss": 0.0419, + "step": 23706 + }, + { + "epoch": 3.11, + "learning_rate": 1.3993092310791186e-05, + "loss": 0.0413, + "step": 23707 + }, + { + "epoch": 3.11, + "learning_rate": 1.399260609450928e-05, + "loss": 0.0427, + "step": 23708 + }, + { + "epoch": 3.11, + "learning_rate": 1.3992119866998386e-05, + "loss": 0.0449, + "step": 23709 + }, + { + "epoch": 3.11, + "learning_rate": 1.3991633628259863e-05, + "loss": 0.0407, + "step": 23710 + }, + { + "epoch": 3.11, + "learning_rate": 1.399114737829509e-05, + "loss": 0.0366, + "step": 23711 + }, + { + "epoch": 3.11, + "learning_rate": 1.3990661117105422e-05, + "loss": 0.0452, + "step": 23712 + }, + { + "epoch": 3.11, + "learning_rate": 1.3990174844692232e-05, + "loss": 0.0418, + "step": 23713 + }, + { + "epoch": 3.11, + "learning_rate": 1.398968856105689e-05, + "loss": 0.0418, + "step": 23714 + }, + { + "epoch": 3.11, + "learning_rate": 1.3989202266200761e-05, + "loss": 0.046, + "step": 23715 + }, + { + "epoch": 3.11, + "learning_rate": 1.3988715960125212e-05, + "loss": 0.0374, + "step": 23716 + }, + { + "epoch": 3.11, + "learning_rate": 1.3988229642831612e-05, + "loss": 0.0514, + "step": 23717 + }, + { + "epoch": 3.11, + "learning_rate": 1.3987743314321332e-05, + "loss": 0.0413, + "step": 23718 + }, + { + "epoch": 3.11, + "learning_rate": 1.3987256974595732e-05, + "loss": 0.0444, + "step": 23719 + }, + { + "epoch": 3.11, + "learning_rate": 1.3986770623656187e-05, + "loss": 0.0489, + "step": 23720 + }, + { + "epoch": 3.11, + "learning_rate": 1.3986284261504058e-05, + "loss": 0.0424, + "step": 23721 + }, + { + "epoch": 3.11, + "learning_rate": 1.398579788814072e-05, + "loss": 0.0406, + "step": 23722 + }, + { + "epoch": 3.11, + "learning_rate": 1.3985311503567536e-05, + "loss": 0.0441, + "step": 23723 + }, + { + "epoch": 3.11, + "learning_rate": 1.398482510778588e-05, + "loss": 0.0489, + "step": 23724 + }, + { + "epoch": 3.11, + "learning_rate": 1.3984338700797112e-05, + "loss": 0.04, + "step": 23725 + }, + { + "epoch": 3.11, + "learning_rate": 1.39838522826026e-05, + "loss": 0.0323, + "step": 23726 + }, + { + "epoch": 3.11, + "learning_rate": 1.398336585320372e-05, + "loss": 0.0356, + "step": 23727 + }, + { + "epoch": 3.11, + "learning_rate": 1.3982879412601835e-05, + "loss": 0.0427, + "step": 23728 + }, + { + "epoch": 3.11, + "learning_rate": 1.3982392960798315e-05, + "loss": 0.0468, + "step": 23729 + }, + { + "epoch": 3.11, + "learning_rate": 1.3981906497794525e-05, + "loss": 0.0493, + "step": 23730 + }, + { + "epoch": 3.11, + "learning_rate": 1.3981420023591835e-05, + "loss": 0.0391, + "step": 23731 + }, + { + "epoch": 3.11, + "learning_rate": 1.3980933538191615e-05, + "loss": 0.0436, + "step": 23732 + }, + { + "epoch": 3.11, + "learning_rate": 1.398044704159523e-05, + "loss": 0.0387, + "step": 23733 + }, + { + "epoch": 3.11, + "learning_rate": 1.3979960533804048e-05, + "loss": 0.0372, + "step": 23734 + }, + { + "epoch": 3.11, + "learning_rate": 1.3979474014819442e-05, + "loss": 0.0463, + "step": 23735 + }, + { + "epoch": 3.11, + "learning_rate": 1.3978987484642776e-05, + "loss": 0.0427, + "step": 23736 + }, + { + "epoch": 3.11, + "learning_rate": 1.397850094327542e-05, + "loss": 0.0345, + "step": 23737 + }, + { + "epoch": 3.11, + "learning_rate": 1.3978014390718742e-05, + "loss": 0.0376, + "step": 23738 + }, + { + "epoch": 3.11, + "learning_rate": 1.3977527826974111e-05, + "loss": 0.0514, + "step": 23739 + }, + { + "epoch": 3.11, + "learning_rate": 1.3977041252042894e-05, + "loss": 0.0401, + "step": 23740 + }, + { + "epoch": 3.11, + "learning_rate": 1.3976554665926461e-05, + "loss": 0.0382, + "step": 23741 + }, + { + "epoch": 3.11, + "learning_rate": 1.3976068068626176e-05, + "loss": 0.04, + "step": 23742 + }, + { + "epoch": 3.11, + "learning_rate": 1.3975581460143415e-05, + "loss": 0.0388, + "step": 23743 + }, + { + "epoch": 3.11, + "learning_rate": 1.3975094840479541e-05, + "loss": 0.0384, + "step": 23744 + }, + { + "epoch": 3.11, + "learning_rate": 1.3974608209635927e-05, + "loss": 0.038, + "step": 23745 + }, + { + "epoch": 3.11, + "learning_rate": 1.3974121567613939e-05, + "loss": 0.048, + "step": 23746 + }, + { + "epoch": 3.11, + "learning_rate": 1.3973634914414944e-05, + "loss": 0.0392, + "step": 23747 + }, + { + "epoch": 3.11, + "learning_rate": 1.397314825004031e-05, + "loss": 0.0381, + "step": 23748 + }, + { + "epoch": 3.11, + "learning_rate": 1.3972661574491414e-05, + "loss": 0.0413, + "step": 23749 + }, + { + "epoch": 3.11, + "learning_rate": 1.3972174887769613e-05, + "loss": 0.0398, + "step": 23750 + }, + { + "epoch": 3.11, + "learning_rate": 1.3971688189876287e-05, + "loss": 0.0474, + "step": 23751 + }, + { + "epoch": 3.11, + "learning_rate": 1.3971201480812795e-05, + "loss": 0.037, + "step": 23752 + }, + { + "epoch": 3.11, + "learning_rate": 1.3970714760580512e-05, + "loss": 0.0445, + "step": 23753 + }, + { + "epoch": 3.11, + "learning_rate": 1.3970228029180806e-05, + "loss": 0.0433, + "step": 23754 + }, + { + "epoch": 3.11, + "learning_rate": 1.3969741286615044e-05, + "loss": 0.0361, + "step": 23755 + }, + { + "epoch": 3.11, + "learning_rate": 1.3969254532884598e-05, + "loss": 0.0454, + "step": 23756 + }, + { + "epoch": 3.11, + "learning_rate": 1.3968767767990832e-05, + "loss": 0.0455, + "step": 23757 + }, + { + "epoch": 3.11, + "learning_rate": 1.396828099193512e-05, + "loss": 0.0412, + "step": 23758 + }, + { + "epoch": 3.11, + "learning_rate": 1.3967794204718828e-05, + "loss": 0.0475, + "step": 23759 + }, + { + "epoch": 3.11, + "learning_rate": 1.3967307406343326e-05, + "loss": 0.0392, + "step": 23760 + }, + { + "epoch": 3.11, + "learning_rate": 1.3966820596809982e-05, + "loss": 0.0448, + "step": 23761 + }, + { + "epoch": 3.11, + "learning_rate": 1.3966333776120171e-05, + "loss": 0.0419, + "step": 23762 + }, + { + "epoch": 3.11, + "learning_rate": 1.3965846944275254e-05, + "loss": 0.048, + "step": 23763 + }, + { + "epoch": 3.11, + "learning_rate": 1.3965360101276603e-05, + "loss": 0.0421, + "step": 23764 + }, + { + "epoch": 3.11, + "learning_rate": 1.396487324712559e-05, + "loss": 0.0456, + "step": 23765 + }, + { + "epoch": 3.11, + "learning_rate": 1.3964386381823582e-05, + "loss": 0.046, + "step": 23766 + }, + { + "epoch": 3.11, + "learning_rate": 1.3963899505371948e-05, + "loss": 0.035, + "step": 23767 + }, + { + "epoch": 3.11, + "learning_rate": 1.3963412617772059e-05, + "loss": 0.0396, + "step": 23768 + }, + { + "epoch": 3.11, + "learning_rate": 1.396292571902528e-05, + "loss": 0.0458, + "step": 23769 + }, + { + "epoch": 3.11, + "learning_rate": 1.3962438809132982e-05, + "loss": 0.0436, + "step": 23770 + }, + { + "epoch": 3.11, + "learning_rate": 1.396195188809654e-05, + "loss": 0.0351, + "step": 23771 + }, + { + "epoch": 3.11, + "learning_rate": 1.3961464955917322e-05, + "loss": 0.0346, + "step": 23772 + }, + { + "epoch": 3.11, + "learning_rate": 1.3960978012596688e-05, + "loss": 0.0453, + "step": 23773 + }, + { + "epoch": 3.11, + "learning_rate": 1.3960491058136017e-05, + "loss": 0.0461, + "step": 23774 + }, + { + "epoch": 3.11, + "learning_rate": 1.3960004092536673e-05, + "loss": 0.0416, + "step": 23775 + }, + { + "epoch": 3.11, + "learning_rate": 1.3959517115800035e-05, + "loss": 0.0402, + "step": 23776 + }, + { + "epoch": 3.11, + "learning_rate": 1.3959030127927462e-05, + "loss": 0.045, + "step": 23777 + }, + { + "epoch": 3.11, + "learning_rate": 1.3958543128920328e-05, + "loss": 0.0339, + "step": 23778 + }, + { + "epoch": 3.11, + "learning_rate": 1.3958056118780002e-05, + "loss": 0.0414, + "step": 23779 + }, + { + "epoch": 3.12, + "learning_rate": 1.3957569097507855e-05, + "loss": 0.0384, + "step": 23780 + }, + { + "epoch": 3.12, + "learning_rate": 1.3957082065105253e-05, + "loss": 0.0426, + "step": 23781 + }, + { + "epoch": 3.12, + "learning_rate": 1.395659502157357e-05, + "loss": 0.041, + "step": 23782 + }, + { + "epoch": 3.12, + "learning_rate": 1.3956107966914173e-05, + "loss": 0.0382, + "step": 23783 + }, + { + "epoch": 3.12, + "learning_rate": 1.3955620901128436e-05, + "loss": 0.0458, + "step": 23784 + }, + { + "epoch": 3.12, + "learning_rate": 1.3955133824217725e-05, + "loss": 0.0415, + "step": 23785 + }, + { + "epoch": 3.12, + "learning_rate": 1.3954646736183408e-05, + "loss": 0.038, + "step": 23786 + }, + { + "epoch": 3.12, + "learning_rate": 1.3954159637026861e-05, + "loss": 0.0363, + "step": 23787 + }, + { + "epoch": 3.12, + "learning_rate": 1.395367252674945e-05, + "loss": 0.0411, + "step": 23788 + }, + { + "epoch": 3.12, + "learning_rate": 1.3953185405352544e-05, + "loss": 0.0515, + "step": 23789 + }, + { + "epoch": 3.12, + "learning_rate": 1.3952698272837517e-05, + "loss": 0.0413, + "step": 23790 + }, + { + "epoch": 3.12, + "learning_rate": 1.3952211129205735e-05, + "loss": 0.0456, + "step": 23791 + }, + { + "epoch": 3.12, + "learning_rate": 1.3951723974458568e-05, + "loss": 0.0424, + "step": 23792 + }, + { + "epoch": 3.12, + "learning_rate": 1.395123680859739e-05, + "loss": 0.0431, + "step": 23793 + }, + { + "epoch": 3.12, + "learning_rate": 1.3950749631623567e-05, + "loss": 0.0341, + "step": 23794 + }, + { + "epoch": 3.12, + "learning_rate": 1.3950262443538471e-05, + "loss": 0.0415, + "step": 23795 + }, + { + "epoch": 3.12, + "learning_rate": 1.3949775244343475e-05, + "loss": 0.0506, + "step": 23796 + }, + { + "epoch": 3.12, + "learning_rate": 1.3949288034039948e-05, + "loss": 0.0444, + "step": 23797 + }, + { + "epoch": 3.12, + "learning_rate": 1.3948800812629253e-05, + "loss": 0.0485, + "step": 23798 + }, + { + "epoch": 3.12, + "learning_rate": 1.394831358011277e-05, + "loss": 0.0408, + "step": 23799 + }, + { + "epoch": 3.12, + "learning_rate": 1.3947826336491863e-05, + "loss": 0.0406, + "step": 23800 + }, + { + "epoch": 3.12, + "learning_rate": 1.3947339081767905e-05, + "loss": 0.0341, + "step": 23801 + }, + { + "epoch": 3.12, + "learning_rate": 1.3946851815942269e-05, + "loss": 0.0378, + "step": 23802 + }, + { + "epoch": 3.12, + "learning_rate": 1.3946364539016317e-05, + "loss": 0.0413, + "step": 23803 + }, + { + "epoch": 3.12, + "learning_rate": 1.3945877250991431e-05, + "loss": 0.0503, + "step": 23804 + }, + { + "epoch": 3.12, + "learning_rate": 1.394538995186897e-05, + "loss": 0.0349, + "step": 23805 + }, + { + "epoch": 3.12, + "learning_rate": 1.3944902641650315e-05, + "loss": 0.0396, + "step": 23806 + }, + { + "epoch": 3.12, + "learning_rate": 1.3944415320336828e-05, + "loss": 0.0336, + "step": 23807 + }, + { + "epoch": 3.12, + "learning_rate": 1.3943927987929884e-05, + "loss": 0.0372, + "step": 23808 + }, + { + "epoch": 3.12, + "learning_rate": 1.394344064443085e-05, + "loss": 0.0455, + "step": 23809 + }, + { + "epoch": 3.12, + "learning_rate": 1.3942953289841103e-05, + "loss": 0.0403, + "step": 23810 + }, + { + "epoch": 3.12, + "learning_rate": 1.3942465924162009e-05, + "loss": 0.042, + "step": 23811 + }, + { + "epoch": 3.12, + "learning_rate": 1.3941978547394942e-05, + "loss": 0.0427, + "step": 23812 + }, + { + "epoch": 3.12, + "learning_rate": 1.3941491159541263e-05, + "loss": 0.0441, + "step": 23813 + }, + { + "epoch": 3.12, + "learning_rate": 1.3941003760602355e-05, + "loss": 0.0398, + "step": 23814 + }, + { + "epoch": 3.12, + "learning_rate": 1.3940516350579583e-05, + "loss": 0.0442, + "step": 23815 + }, + { + "epoch": 3.12, + "learning_rate": 1.394002892947432e-05, + "loss": 0.0422, + "step": 23816 + }, + { + "epoch": 3.12, + "learning_rate": 1.3939541497287934e-05, + "loss": 0.0467, + "step": 23817 + }, + { + "epoch": 3.12, + "learning_rate": 1.3939054054021797e-05, + "loss": 0.0433, + "step": 23818 + }, + { + "epoch": 3.12, + "learning_rate": 1.393856659967728e-05, + "loss": 0.0474, + "step": 23819 + }, + { + "epoch": 3.12, + "learning_rate": 1.3938079134255753e-05, + "loss": 0.0357, + "step": 23820 + }, + { + "epoch": 3.12, + "learning_rate": 1.393759165775859e-05, + "loss": 0.0448, + "step": 23821 + }, + { + "epoch": 3.12, + "learning_rate": 1.3937104170187161e-05, + "loss": 0.0419, + "step": 23822 + }, + { + "epoch": 3.12, + "learning_rate": 1.3936616671542835e-05, + "loss": 0.0375, + "step": 23823 + }, + { + "epoch": 3.12, + "learning_rate": 1.3936129161826984e-05, + "loss": 0.0396, + "step": 23824 + }, + { + "epoch": 3.12, + "learning_rate": 1.393564164104098e-05, + "loss": 0.04, + "step": 23825 + }, + { + "epoch": 3.12, + "learning_rate": 1.3935154109186193e-05, + "loss": 0.0424, + "step": 23826 + }, + { + "epoch": 3.12, + "learning_rate": 1.3934666566263996e-05, + "loss": 0.0455, + "step": 23827 + }, + { + "epoch": 3.12, + "learning_rate": 1.3934179012275758e-05, + "loss": 0.0385, + "step": 23828 + }, + { + "epoch": 3.12, + "learning_rate": 1.3933691447222853e-05, + "loss": 0.0386, + "step": 23829 + }, + { + "epoch": 3.12, + "learning_rate": 1.3933203871106644e-05, + "loss": 0.0432, + "step": 23830 + }, + { + "epoch": 3.12, + "learning_rate": 1.3932716283928514e-05, + "loss": 0.043, + "step": 23831 + }, + { + "epoch": 3.12, + "learning_rate": 1.3932228685689829e-05, + "loss": 0.0363, + "step": 23832 + }, + { + "epoch": 3.12, + "learning_rate": 1.3931741076391958e-05, + "loss": 0.0341, + "step": 23833 + }, + { + "epoch": 3.12, + "learning_rate": 1.3931253456036276e-05, + "loss": 0.043, + "step": 23834 + }, + { + "epoch": 3.12, + "learning_rate": 1.393076582462415e-05, + "loss": 0.0407, + "step": 23835 + }, + { + "epoch": 3.12, + "learning_rate": 1.3930278182156957e-05, + "loss": 0.0392, + "step": 23836 + }, + { + "epoch": 3.12, + "learning_rate": 1.3929790528636067e-05, + "loss": 0.0403, + "step": 23837 + }, + { + "epoch": 3.12, + "learning_rate": 1.3929302864062847e-05, + "loss": 0.0448, + "step": 23838 + }, + { + "epoch": 3.12, + "learning_rate": 1.3928815188438675e-05, + "loss": 0.038, + "step": 23839 + }, + { + "epoch": 3.12, + "learning_rate": 1.392832750176492e-05, + "loss": 0.0404, + "step": 23840 + }, + { + "epoch": 3.12, + "learning_rate": 1.3927839804042951e-05, + "loss": 0.0506, + "step": 23841 + }, + { + "epoch": 3.12, + "learning_rate": 1.3927352095274142e-05, + "loss": 0.0398, + "step": 23842 + }, + { + "epoch": 3.12, + "learning_rate": 1.3926864375459863e-05, + "loss": 0.0402, + "step": 23843 + }, + { + "epoch": 3.12, + "learning_rate": 1.3926376644601492e-05, + "loss": 0.0381, + "step": 23844 + }, + { + "epoch": 3.12, + "learning_rate": 1.392588890270039e-05, + "loss": 0.0486, + "step": 23845 + }, + { + "epoch": 3.12, + "learning_rate": 1.3925401149757937e-05, + "loss": 0.0442, + "step": 23846 + }, + { + "epoch": 3.12, + "learning_rate": 1.3924913385775503e-05, + "loss": 0.0411, + "step": 23847 + }, + { + "epoch": 3.12, + "learning_rate": 1.392442561075446e-05, + "loss": 0.0486, + "step": 23848 + }, + { + "epoch": 3.12, + "learning_rate": 1.3923937824696177e-05, + "loss": 0.037, + "step": 23849 + }, + { + "epoch": 3.12, + "learning_rate": 1.3923450027602029e-05, + "loss": 0.046, + "step": 23850 + }, + { + "epoch": 3.12, + "learning_rate": 1.3922962219473384e-05, + "loss": 0.0399, + "step": 23851 + }, + { + "epoch": 3.12, + "learning_rate": 1.3922474400311618e-05, + "loss": 0.0385, + "step": 23852 + }, + { + "epoch": 3.12, + "learning_rate": 1.3921986570118102e-05, + "loss": 0.0437, + "step": 23853 + }, + { + "epoch": 3.12, + "learning_rate": 1.3921498728894211e-05, + "loss": 0.041, + "step": 23854 + }, + { + "epoch": 3.12, + "learning_rate": 1.392101087664131e-05, + "loss": 0.0366, + "step": 23855 + }, + { + "epoch": 3.12, + "learning_rate": 1.3920523013360775e-05, + "loss": 0.0408, + "step": 23856 + }, + { + "epoch": 3.13, + "learning_rate": 1.3920035139053978e-05, + "loss": 0.045, + "step": 23857 + }, + { + "epoch": 3.13, + "learning_rate": 1.3919547253722292e-05, + "loss": 0.0364, + "step": 23858 + }, + { + "epoch": 3.13, + "learning_rate": 1.3919059357367086e-05, + "loss": 0.0413, + "step": 23859 + }, + { + "epoch": 3.13, + "learning_rate": 1.3918571449989734e-05, + "loss": 0.042, + "step": 23860 + }, + { + "epoch": 3.13, + "learning_rate": 1.3918083531591613e-05, + "loss": 0.0495, + "step": 23861 + }, + { + "epoch": 3.13, + "learning_rate": 1.3917595602174086e-05, + "loss": 0.0375, + "step": 23862 + }, + { + "epoch": 3.13, + "learning_rate": 1.3917107661738532e-05, + "loss": 0.042, + "step": 23863 + }, + { + "epoch": 3.13, + "learning_rate": 1.3916619710286319e-05, + "loss": 0.0413, + "step": 23864 + }, + { + "epoch": 3.13, + "learning_rate": 1.3916131747818825e-05, + "loss": 0.0443, + "step": 23865 + }, + { + "epoch": 3.13, + "learning_rate": 1.3915643774337417e-05, + "loss": 0.0452, + "step": 23866 + }, + { + "epoch": 3.13, + "learning_rate": 1.391515578984347e-05, + "loss": 0.0374, + "step": 23867 + }, + { + "epoch": 3.13, + "learning_rate": 1.3914667794338354e-05, + "loss": 0.0381, + "step": 23868 + }, + { + "epoch": 3.13, + "learning_rate": 1.3914179787823447e-05, + "loss": 0.0378, + "step": 23869 + }, + { + "epoch": 3.13, + "learning_rate": 1.3913691770300115e-05, + "loss": 0.0515, + "step": 23870 + }, + { + "epoch": 3.13, + "learning_rate": 1.3913203741769736e-05, + "loss": 0.0347, + "step": 23871 + }, + { + "epoch": 3.13, + "learning_rate": 1.3912715702233677e-05, + "loss": 0.0425, + "step": 23872 + }, + { + "epoch": 3.13, + "learning_rate": 1.3912227651693315e-05, + "loss": 0.038, + "step": 23873 + }, + { + "epoch": 3.13, + "learning_rate": 1.3911739590150019e-05, + "loss": 0.0479, + "step": 23874 + }, + { + "epoch": 3.13, + "learning_rate": 1.3911251517605167e-05, + "loss": 0.0467, + "step": 23875 + }, + { + "epoch": 3.13, + "learning_rate": 1.3910763434060126e-05, + "loss": 0.0363, + "step": 23876 + }, + { + "epoch": 3.13, + "learning_rate": 1.3910275339516273e-05, + "loss": 0.0346, + "step": 23877 + }, + { + "epoch": 3.13, + "learning_rate": 1.3909787233974976e-05, + "loss": 0.0469, + "step": 23878 + }, + { + "epoch": 3.13, + "learning_rate": 1.3909299117437613e-05, + "loss": 0.0467, + "step": 23879 + }, + { + "epoch": 3.13, + "learning_rate": 1.3908810989905552e-05, + "loss": 0.0423, + "step": 23880 + }, + { + "epoch": 3.13, + "learning_rate": 1.3908322851380172e-05, + "loss": 0.0404, + "step": 23881 + }, + { + "epoch": 3.13, + "learning_rate": 1.3907834701862839e-05, + "loss": 0.0392, + "step": 23882 + }, + { + "epoch": 3.13, + "learning_rate": 1.390734654135493e-05, + "loss": 0.0359, + "step": 23883 + }, + { + "epoch": 3.13, + "learning_rate": 1.3906858369857816e-05, + "loss": 0.0395, + "step": 23884 + }, + { + "epoch": 3.13, + "learning_rate": 1.3906370187372872e-05, + "loss": 0.042, + "step": 23885 + }, + { + "epoch": 3.13, + "learning_rate": 1.390588199390147e-05, + "loss": 0.0392, + "step": 23886 + }, + { + "epoch": 3.13, + "learning_rate": 1.3905393789444983e-05, + "loss": 0.0418, + "step": 23887 + }, + { + "epoch": 3.13, + "learning_rate": 1.3904905574004785e-05, + "loss": 0.039, + "step": 23888 + }, + { + "epoch": 3.13, + "learning_rate": 1.3904417347582247e-05, + "loss": 0.0436, + "step": 23889 + }, + { + "epoch": 3.13, + "learning_rate": 1.3903929110178739e-05, + "loss": 0.0392, + "step": 23890 + }, + { + "epoch": 3.13, + "learning_rate": 1.3903440861795643e-05, + "loss": 0.0316, + "step": 23891 + }, + { + "epoch": 3.13, + "learning_rate": 1.3902952602434328e-05, + "loss": 0.0393, + "step": 23892 + }, + { + "epoch": 3.13, + "learning_rate": 1.3902464332096165e-05, + "loss": 0.0519, + "step": 23893 + }, + { + "epoch": 3.13, + "learning_rate": 1.3901976050782532e-05, + "loss": 0.053, + "step": 23894 + }, + { + "epoch": 3.13, + "learning_rate": 1.3901487758494792e-05, + "loss": 0.0433, + "step": 23895 + }, + { + "epoch": 3.13, + "learning_rate": 1.3900999455234329e-05, + "loss": 0.0467, + "step": 23896 + }, + { + "epoch": 3.13, + "learning_rate": 1.3900511141002512e-05, + "loss": 0.0421, + "step": 23897 + }, + { + "epoch": 3.13, + "learning_rate": 1.390002281580072e-05, + "loss": 0.0432, + "step": 23898 + }, + { + "epoch": 3.13, + "learning_rate": 1.3899534479630317e-05, + "loss": 0.0471, + "step": 23899 + }, + { + "epoch": 3.13, + "learning_rate": 1.3899046132492681e-05, + "loss": 0.0429, + "step": 23900 + }, + { + "epoch": 3.13, + "learning_rate": 1.3898557774389186e-05, + "loss": 0.0395, + "step": 23901 + }, + { + "epoch": 3.13, + "learning_rate": 1.3898069405321207e-05, + "loss": 0.0461, + "step": 23902 + }, + { + "epoch": 3.13, + "learning_rate": 1.3897581025290112e-05, + "loss": 0.0365, + "step": 23903 + }, + { + "epoch": 3.13, + "learning_rate": 1.389709263429728e-05, + "loss": 0.0337, + "step": 23904 + }, + { + "epoch": 3.13, + "learning_rate": 1.3896604232344083e-05, + "loss": 0.0403, + "step": 23905 + }, + { + "epoch": 3.13, + "learning_rate": 1.3896115819431892e-05, + "loss": 0.045, + "step": 23906 + }, + { + "epoch": 3.13, + "learning_rate": 1.3895627395562084e-05, + "loss": 0.0543, + "step": 23907 + }, + { + "epoch": 3.13, + "learning_rate": 1.389513896073603e-05, + "loss": 0.0426, + "step": 23908 + }, + { + "epoch": 3.13, + "learning_rate": 1.3894650514955108e-05, + "loss": 0.0611, + "step": 23909 + }, + { + "epoch": 3.13, + "learning_rate": 1.3894162058220685e-05, + "loss": 0.04, + "step": 23910 + }, + { + "epoch": 3.13, + "learning_rate": 1.3893673590534144e-05, + "loss": 0.0414, + "step": 23911 + }, + { + "epoch": 3.13, + "learning_rate": 1.3893185111896847e-05, + "loss": 0.0464, + "step": 23912 + }, + { + "epoch": 3.13, + "learning_rate": 1.3892696622310177e-05, + "loss": 0.0406, + "step": 23913 + }, + { + "epoch": 3.13, + "learning_rate": 1.3892208121775505e-05, + "loss": 0.0455, + "step": 23914 + }, + { + "epoch": 3.13, + "learning_rate": 1.3891719610294208e-05, + "loss": 0.0421, + "step": 23915 + }, + { + "epoch": 3.13, + "learning_rate": 1.3891231087867654e-05, + "loss": 0.0467, + "step": 23916 + }, + { + "epoch": 3.13, + "learning_rate": 1.389074255449722e-05, + "loss": 0.0476, + "step": 23917 + }, + { + "epoch": 3.13, + "learning_rate": 1.389025401018428e-05, + "loss": 0.0421, + "step": 23918 + }, + { + "epoch": 3.13, + "learning_rate": 1.3889765454930206e-05, + "loss": 0.0441, + "step": 23919 + }, + { + "epoch": 3.13, + "learning_rate": 1.3889276888736376e-05, + "loss": 0.0396, + "step": 23920 + }, + { + "epoch": 3.13, + "learning_rate": 1.388878831160416e-05, + "loss": 0.0381, + "step": 23921 + }, + { + "epoch": 3.13, + "learning_rate": 1.3888299723534936e-05, + "loss": 0.0526, + "step": 23922 + }, + { + "epoch": 3.13, + "learning_rate": 1.3887811124530075e-05, + "loss": 0.0366, + "step": 23923 + }, + { + "epoch": 3.13, + "learning_rate": 1.3887322514590952e-05, + "loss": 0.0392, + "step": 23924 + }, + { + "epoch": 3.13, + "learning_rate": 1.3886833893718943e-05, + "loss": 0.0416, + "step": 23925 + }, + { + "epoch": 3.13, + "learning_rate": 1.3886345261915419e-05, + "loss": 0.0432, + "step": 23926 + }, + { + "epoch": 3.13, + "learning_rate": 1.3885856619181757e-05, + "loss": 0.04, + "step": 23927 + }, + { + "epoch": 3.13, + "learning_rate": 1.3885367965519327e-05, + "loss": 0.0396, + "step": 23928 + }, + { + "epoch": 3.13, + "learning_rate": 1.388487930092951e-05, + "loss": 0.0449, + "step": 23929 + }, + { + "epoch": 3.13, + "learning_rate": 1.3884390625413678e-05, + "loss": 0.0397, + "step": 23930 + }, + { + "epoch": 3.13, + "learning_rate": 1.38839019389732e-05, + "loss": 0.0383, + "step": 23931 + }, + { + "epoch": 3.13, + "learning_rate": 1.3883413241609459e-05, + "loss": 0.0418, + "step": 23932 + }, + { + "epoch": 3.14, + "learning_rate": 1.388292453332382e-05, + "loss": 0.0414, + "step": 23933 + }, + { + "epoch": 3.14, + "learning_rate": 1.3882435814117667e-05, + "loss": 0.043, + "step": 23934 + }, + { + "epoch": 3.14, + "learning_rate": 1.3881947083992368e-05, + "loss": 0.0458, + "step": 23935 + }, + { + "epoch": 3.14, + "learning_rate": 1.3881458342949301e-05, + "loss": 0.0447, + "step": 23936 + }, + { + "epoch": 3.14, + "learning_rate": 1.3880969590989839e-05, + "loss": 0.0454, + "step": 23937 + }, + { + "epoch": 3.14, + "learning_rate": 1.3880480828115355e-05, + "loss": 0.0452, + "step": 23938 + }, + { + "epoch": 3.14, + "learning_rate": 1.3879992054327225e-05, + "loss": 0.0411, + "step": 23939 + }, + { + "epoch": 3.14, + "learning_rate": 1.3879503269626826e-05, + "loss": 0.041, + "step": 23940 + }, + { + "epoch": 3.14, + "learning_rate": 1.3879014474015528e-05, + "loss": 0.039, + "step": 23941 + }, + { + "epoch": 3.14, + "learning_rate": 1.3878525667494713e-05, + "loss": 0.0405, + "step": 23942 + }, + { + "epoch": 3.14, + "learning_rate": 1.3878036850065749e-05, + "loss": 0.0447, + "step": 23943 + }, + { + "epoch": 3.14, + "learning_rate": 1.387754802173001e-05, + "loss": 0.0406, + "step": 23944 + }, + { + "epoch": 3.14, + "learning_rate": 1.3877059182488875e-05, + "loss": 0.0427, + "step": 23945 + }, + { + "epoch": 3.14, + "learning_rate": 1.3876570332343717e-05, + "loss": 0.0427, + "step": 23946 + }, + { + "epoch": 3.14, + "learning_rate": 1.3876081471295915e-05, + "loss": 0.0453, + "step": 23947 + }, + { + "epoch": 3.14, + "learning_rate": 1.3875592599346837e-05, + "loss": 0.0436, + "step": 23948 + }, + { + "epoch": 3.14, + "learning_rate": 1.3875103716497864e-05, + "loss": 0.0354, + "step": 23949 + }, + { + "epoch": 3.14, + "learning_rate": 1.3874614822750363e-05, + "loss": 0.0486, + "step": 23950 + }, + { + "epoch": 3.14, + "learning_rate": 1.3874125918105718e-05, + "loss": 0.0466, + "step": 23951 + }, + { + "epoch": 3.14, + "learning_rate": 1.3873637002565299e-05, + "loss": 0.0436, + "step": 23952 + }, + { + "epoch": 3.14, + "learning_rate": 1.3873148076130485e-05, + "loss": 0.0396, + "step": 23953 + }, + { + "epoch": 3.14, + "learning_rate": 1.3872659138802647e-05, + "loss": 0.0409, + "step": 23954 + }, + { + "epoch": 3.14, + "learning_rate": 1.3872170190583159e-05, + "loss": 0.0444, + "step": 23955 + }, + { + "epoch": 3.14, + "learning_rate": 1.3871681231473399e-05, + "loss": 0.0415, + "step": 23956 + }, + { + "epoch": 3.14, + "learning_rate": 1.3871192261474744e-05, + "loss": 0.0475, + "step": 23957 + }, + { + "epoch": 3.14, + "learning_rate": 1.3870703280588567e-05, + "loss": 0.0366, + "step": 23958 + }, + { + "epoch": 3.14, + "learning_rate": 1.3870214288816242e-05, + "loss": 0.038, + "step": 23959 + }, + { + "epoch": 3.14, + "learning_rate": 1.3869725286159143e-05, + "loss": 0.0378, + "step": 23960 + }, + { + "epoch": 3.14, + "learning_rate": 1.3869236272618651e-05, + "loss": 0.0407, + "step": 23961 + }, + { + "epoch": 3.14, + "learning_rate": 1.3868747248196136e-05, + "loss": 0.0354, + "step": 23962 + }, + { + "epoch": 3.14, + "learning_rate": 1.3868258212892977e-05, + "loss": 0.0411, + "step": 23963 + }, + { + "epoch": 3.14, + "learning_rate": 1.3867769166710548e-05, + "loss": 0.0451, + "step": 23964 + }, + { + "epoch": 3.14, + "learning_rate": 1.3867280109650221e-05, + "loss": 0.0483, + "step": 23965 + }, + { + "epoch": 3.14, + "learning_rate": 1.3866791041713377e-05, + "loss": 0.0423, + "step": 23966 + }, + { + "epoch": 3.14, + "learning_rate": 1.3866301962901392e-05, + "loss": 0.0369, + "step": 23967 + }, + { + "epoch": 3.14, + "learning_rate": 1.3865812873215633e-05, + "loss": 0.0378, + "step": 23968 + }, + { + "epoch": 3.14, + "learning_rate": 1.3865323772657483e-05, + "loss": 0.0414, + "step": 23969 + }, + { + "epoch": 3.14, + "learning_rate": 1.386483466122832e-05, + "loss": 0.0322, + "step": 23970 + }, + { + "epoch": 3.14, + "learning_rate": 1.386434553892951e-05, + "loss": 0.0378, + "step": 23971 + }, + { + "epoch": 3.14, + "learning_rate": 1.3863856405762435e-05, + "loss": 0.0399, + "step": 23972 + }, + { + "epoch": 3.14, + "learning_rate": 1.3863367261728471e-05, + "loss": 0.0394, + "step": 23973 + }, + { + "epoch": 3.14, + "learning_rate": 1.3862878106828994e-05, + "loss": 0.0447, + "step": 23974 + }, + { + "epoch": 3.14, + "learning_rate": 1.3862388941065375e-05, + "loss": 0.0491, + "step": 23975 + }, + { + "epoch": 3.14, + "learning_rate": 1.3861899764438996e-05, + "loss": 0.0517, + "step": 23976 + }, + { + "epoch": 3.14, + "learning_rate": 1.3861410576951225e-05, + "loss": 0.0438, + "step": 23977 + }, + { + "epoch": 3.14, + "learning_rate": 1.3860921378603445e-05, + "loss": 0.0407, + "step": 23978 + }, + { + "epoch": 3.14, + "learning_rate": 1.3860432169397026e-05, + "loss": 0.0441, + "step": 23979 + }, + { + "epoch": 3.14, + "learning_rate": 1.3859942949333354e-05, + "loss": 0.0444, + "step": 23980 + }, + { + "epoch": 3.14, + "learning_rate": 1.3859453718413791e-05, + "loss": 0.0415, + "step": 23981 + }, + { + "epoch": 3.14, + "learning_rate": 1.3858964476639724e-05, + "loss": 0.0421, + "step": 23982 + }, + { + "epoch": 3.14, + "learning_rate": 1.3858475224012523e-05, + "loss": 0.046, + "step": 23983 + }, + { + "epoch": 3.14, + "learning_rate": 1.3857985960533569e-05, + "loss": 0.0493, + "step": 23984 + }, + { + "epoch": 3.14, + "learning_rate": 1.385749668620423e-05, + "loss": 0.0447, + "step": 23985 + }, + { + "epoch": 3.14, + "learning_rate": 1.3857007401025891e-05, + "loss": 0.0417, + "step": 23986 + }, + { + "epoch": 3.14, + "learning_rate": 1.3856518104999923e-05, + "loss": 0.0369, + "step": 23987 + }, + { + "epoch": 3.14, + "learning_rate": 1.3856028798127703e-05, + "loss": 0.0475, + "step": 23988 + }, + { + "epoch": 3.14, + "learning_rate": 1.3855539480410607e-05, + "loss": 0.0436, + "step": 23989 + }, + { + "epoch": 3.14, + "learning_rate": 1.385505015185001e-05, + "loss": 0.0408, + "step": 23990 + }, + { + "epoch": 3.14, + "learning_rate": 1.3854560812447293e-05, + "loss": 0.0416, + "step": 23991 + }, + { + "epoch": 3.14, + "learning_rate": 1.3854071462203827e-05, + "loss": 0.0378, + "step": 23992 + }, + { + "epoch": 3.14, + "learning_rate": 1.3853582101120992e-05, + "loss": 0.0367, + "step": 23993 + }, + { + "epoch": 3.14, + "learning_rate": 1.3853092729200159e-05, + "loss": 0.0389, + "step": 23994 + }, + { + "epoch": 3.14, + "learning_rate": 1.3852603346442713e-05, + "loss": 0.0457, + "step": 23995 + }, + { + "epoch": 3.14, + "learning_rate": 1.3852113952850022e-05, + "loss": 0.0416, + "step": 23996 + }, + { + "epoch": 3.14, + "learning_rate": 1.3851624548423466e-05, + "loss": 0.0417, + "step": 23997 + }, + { + "epoch": 3.14, + "learning_rate": 1.3851135133164419e-05, + "loss": 0.0365, + "step": 23998 + }, + { + "epoch": 3.14, + "learning_rate": 1.3850645707074262e-05, + "loss": 0.0455, + "step": 23999 + }, + { + "epoch": 3.14, + "learning_rate": 1.3850156270154368e-05, + "loss": 0.0417, + "step": 24000 + }, + { + "epoch": 3.14, + "learning_rate": 1.3849666822406116e-05, + "loss": 0.0371, + "step": 24001 + }, + { + "epoch": 3.14, + "learning_rate": 1.3849177363830881e-05, + "loss": 0.0333, + "step": 24002 + }, + { + "epoch": 3.14, + "learning_rate": 1.3848687894430039e-05, + "loss": 0.0524, + "step": 24003 + }, + { + "epoch": 3.14, + "learning_rate": 1.3848198414204966e-05, + "loss": 0.0371, + "step": 24004 + }, + { + "epoch": 3.14, + "learning_rate": 1.3847708923157041e-05, + "loss": 0.0435, + "step": 24005 + }, + { + "epoch": 3.14, + "learning_rate": 1.384721942128764e-05, + "loss": 0.0402, + "step": 24006 + }, + { + "epoch": 3.14, + "learning_rate": 1.384672990859814e-05, + "loss": 0.0384, + "step": 24007 + }, + { + "epoch": 3.14, + "learning_rate": 1.3846240385089917e-05, + "loss": 0.0429, + "step": 24008 + }, + { + "epoch": 3.15, + "learning_rate": 1.3845750850764346e-05, + "loss": 0.0373, + "step": 24009 + }, + { + "epoch": 3.15, + "learning_rate": 1.3845261305622804e-05, + "loss": 0.0492, + "step": 24010 + }, + { + "epoch": 3.15, + "learning_rate": 1.3844771749666672e-05, + "loss": 0.0375, + "step": 24011 + }, + { + "epoch": 3.15, + "learning_rate": 1.3844282182897324e-05, + "loss": 0.039, + "step": 24012 + }, + { + "epoch": 3.15, + "learning_rate": 1.3843792605316137e-05, + "loss": 0.0398, + "step": 24013 + }, + { + "epoch": 3.15, + "learning_rate": 1.3843303016924489e-05, + "loss": 0.0405, + "step": 24014 + }, + { + "epoch": 3.15, + "learning_rate": 1.3842813417723752e-05, + "loss": 0.0424, + "step": 24015 + }, + { + "epoch": 3.15, + "learning_rate": 1.3842323807715312e-05, + "loss": 0.0421, + "step": 24016 + }, + { + "epoch": 3.15, + "learning_rate": 1.3841834186900539e-05, + "loss": 0.0442, + "step": 24017 + }, + { + "epoch": 3.15, + "learning_rate": 1.3841344555280812e-05, + "loss": 0.0518, + "step": 24018 + }, + { + "epoch": 3.15, + "learning_rate": 1.3840854912857506e-05, + "loss": 0.0424, + "step": 24019 + }, + { + "epoch": 3.15, + "learning_rate": 1.3840365259632003e-05, + "loss": 0.0492, + "step": 24020 + }, + { + "epoch": 3.15, + "learning_rate": 1.3839875595605672e-05, + "loss": 0.04, + "step": 24021 + }, + { + "epoch": 3.15, + "learning_rate": 1.3839385920779902e-05, + "loss": 0.0504, + "step": 24022 + }, + { + "epoch": 3.15, + "learning_rate": 1.3838896235156058e-05, + "loss": 0.0499, + "step": 24023 + }, + { + "epoch": 3.15, + "learning_rate": 1.3838406538735529e-05, + "loss": 0.0379, + "step": 24024 + }, + { + "epoch": 3.15, + "learning_rate": 1.3837916831519682e-05, + "loss": 0.0368, + "step": 24025 + }, + { + "epoch": 3.15, + "learning_rate": 1.3837427113509899e-05, + "loss": 0.0408, + "step": 24026 + }, + { + "epoch": 3.15, + "learning_rate": 1.3836937384707554e-05, + "loss": 0.0436, + "step": 24027 + }, + { + "epoch": 3.15, + "learning_rate": 1.383644764511403e-05, + "loss": 0.0438, + "step": 24028 + }, + { + "epoch": 3.15, + "learning_rate": 1.3835957894730702e-05, + "loss": 0.0428, + "step": 24029 + }, + { + "epoch": 3.15, + "learning_rate": 1.3835468133558945e-05, + "loss": 0.036, + "step": 24030 + }, + { + "epoch": 3.15, + "learning_rate": 1.383497836160014e-05, + "loss": 0.0345, + "step": 24031 + }, + { + "epoch": 3.15, + "learning_rate": 1.3834488578855658e-05, + "loss": 0.0365, + "step": 24032 + }, + { + "epoch": 3.15, + "learning_rate": 1.3833998785326885e-05, + "loss": 0.0394, + "step": 24033 + }, + { + "epoch": 3.15, + "learning_rate": 1.3833508981015193e-05, + "loss": 0.0478, + "step": 24034 + }, + { + "epoch": 3.15, + "learning_rate": 1.3833019165921964e-05, + "loss": 0.044, + "step": 24035 + }, + { + "epoch": 3.15, + "learning_rate": 1.3832529340048569e-05, + "loss": 0.041, + "step": 24036 + }, + { + "epoch": 3.15, + "learning_rate": 1.383203950339639e-05, + "loss": 0.034, + "step": 24037 + }, + { + "epoch": 3.15, + "learning_rate": 1.3831549655966806e-05, + "loss": 0.0439, + "step": 24038 + }, + { + "epoch": 3.15, + "learning_rate": 1.383105979776119e-05, + "loss": 0.0416, + "step": 24039 + }, + { + "epoch": 3.15, + "learning_rate": 1.3830569928780927e-05, + "loss": 0.0407, + "step": 24040 + }, + { + "epoch": 3.15, + "learning_rate": 1.3830080049027385e-05, + "loss": 0.0453, + "step": 24041 + }, + { + "epoch": 3.15, + "learning_rate": 1.3829590158501948e-05, + "loss": 0.0478, + "step": 24042 + }, + { + "epoch": 3.15, + "learning_rate": 1.3829100257205994e-05, + "loss": 0.0342, + "step": 24043 + }, + { + "epoch": 3.15, + "learning_rate": 1.3828610345140899e-05, + "loss": 0.0367, + "step": 24044 + }, + { + "epoch": 3.15, + "learning_rate": 1.3828120422308043e-05, + "loss": 0.0416, + "step": 24045 + }, + { + "epoch": 3.15, + "learning_rate": 1.3827630488708798e-05, + "loss": 0.0393, + "step": 24046 + }, + { + "epoch": 3.15, + "learning_rate": 1.3827140544344549e-05, + "loss": 0.049, + "step": 24047 + }, + { + "epoch": 3.15, + "learning_rate": 1.3826650589216668e-05, + "loss": 0.0485, + "step": 24048 + }, + { + "epoch": 3.15, + "learning_rate": 1.3826160623326539e-05, + "loss": 0.0461, + "step": 24049 + }, + { + "epoch": 3.15, + "learning_rate": 1.3825670646675537e-05, + "loss": 0.037, + "step": 24050 + }, + { + "epoch": 3.15, + "learning_rate": 1.3825180659265038e-05, + "loss": 0.0411, + "step": 24051 + }, + { + "epoch": 3.15, + "learning_rate": 1.3824690661096424e-05, + "loss": 0.0411, + "step": 24052 + }, + { + "epoch": 3.15, + "learning_rate": 1.3824200652171072e-05, + "loss": 0.0403, + "step": 24053 + }, + { + "epoch": 3.15, + "learning_rate": 1.3823710632490357e-05, + "loss": 0.0408, + "step": 24054 + }, + { + "epoch": 3.15, + "learning_rate": 1.382322060205566e-05, + "loss": 0.0456, + "step": 24055 + }, + { + "epoch": 3.15, + "learning_rate": 1.3822730560868361e-05, + "loss": 0.0426, + "step": 24056 + }, + { + "epoch": 3.15, + "learning_rate": 1.3822240508929832e-05, + "loss": 0.0473, + "step": 24057 + }, + { + "epoch": 3.15, + "learning_rate": 1.382175044624146e-05, + "loss": 0.0419, + "step": 24058 + }, + { + "epoch": 3.15, + "learning_rate": 1.3821260372804613e-05, + "loss": 0.0413, + "step": 24059 + }, + { + "epoch": 3.15, + "learning_rate": 1.3820770288620677e-05, + "loss": 0.0337, + "step": 24060 + }, + { + "epoch": 3.15, + "learning_rate": 1.382028019369103e-05, + "loss": 0.0445, + "step": 24061 + }, + { + "epoch": 3.15, + "learning_rate": 1.3819790088017047e-05, + "loss": 0.0412, + "step": 24062 + }, + { + "epoch": 3.15, + "learning_rate": 1.3819299971600107e-05, + "loss": 0.0487, + "step": 24063 + }, + { + "epoch": 3.15, + "learning_rate": 1.3818809844441589e-05, + "loss": 0.0411, + "step": 24064 + }, + { + "epoch": 3.15, + "learning_rate": 1.3818319706542871e-05, + "loss": 0.0441, + "step": 24065 + }, + { + "epoch": 3.15, + "learning_rate": 1.3817829557905333e-05, + "loss": 0.0426, + "step": 24066 + }, + { + "epoch": 3.15, + "learning_rate": 1.3817339398530354e-05, + "loss": 0.0423, + "step": 24067 + }, + { + "epoch": 3.15, + "learning_rate": 1.381684922841931e-05, + "loss": 0.0448, + "step": 24068 + }, + { + "epoch": 3.15, + "learning_rate": 1.3816359047573582e-05, + "loss": 0.0355, + "step": 24069 + }, + { + "epoch": 3.15, + "learning_rate": 1.3815868855994547e-05, + "loss": 0.0353, + "step": 24070 + }, + { + "epoch": 3.15, + "learning_rate": 1.3815378653683582e-05, + "loss": 0.0418, + "step": 24071 + }, + { + "epoch": 3.15, + "learning_rate": 1.381488844064207e-05, + "loss": 0.0411, + "step": 24072 + }, + { + "epoch": 3.15, + "learning_rate": 1.3814398216871387e-05, + "loss": 0.0429, + "step": 24073 + }, + { + "epoch": 3.15, + "learning_rate": 1.3813907982372911e-05, + "loss": 0.0455, + "step": 24074 + }, + { + "epoch": 3.15, + "learning_rate": 1.3813417737148022e-05, + "loss": 0.0397, + "step": 24075 + }, + { + "epoch": 3.15, + "learning_rate": 1.38129274811981e-05, + "loss": 0.038, + "step": 24076 + }, + { + "epoch": 3.15, + "learning_rate": 1.3812437214524524e-05, + "loss": 0.049, + "step": 24077 + }, + { + "epoch": 3.15, + "learning_rate": 1.3811946937128669e-05, + "loss": 0.0427, + "step": 24078 + }, + { + "epoch": 3.15, + "learning_rate": 1.3811456649011916e-05, + "loss": 0.0409, + "step": 24079 + }, + { + "epoch": 3.15, + "learning_rate": 1.3810966350175646e-05, + "loss": 0.0501, + "step": 24080 + }, + { + "epoch": 3.15, + "learning_rate": 1.3810476040621232e-05, + "loss": 0.0507, + "step": 24081 + }, + { + "epoch": 3.15, + "learning_rate": 1.3809985720350062e-05, + "loss": 0.0335, + "step": 24082 + }, + { + "epoch": 3.15, + "learning_rate": 1.3809495389363507e-05, + "loss": 0.038, + "step": 24083 + }, + { + "epoch": 3.15, + "learning_rate": 1.3809005047662952e-05, + "loss": 0.0444, + "step": 24084 + }, + { + "epoch": 3.15, + "learning_rate": 1.3808514695249771e-05, + "loss": 0.0441, + "step": 24085 + }, + { + "epoch": 3.16, + "learning_rate": 1.3808024332125343e-05, + "loss": 0.0408, + "step": 24086 + }, + { + "epoch": 3.16, + "learning_rate": 1.3807533958291053e-05, + "loss": 0.044, + "step": 24087 + }, + { + "epoch": 3.16, + "learning_rate": 1.3807043573748273e-05, + "loss": 0.0392, + "step": 24088 + }, + { + "epoch": 3.16, + "learning_rate": 1.380655317849839e-05, + "loss": 0.0444, + "step": 24089 + }, + { + "epoch": 3.16, + "learning_rate": 1.3806062772542777e-05, + "loss": 0.0433, + "step": 24090 + }, + { + "epoch": 3.16, + "learning_rate": 1.3805572355882814e-05, + "loss": 0.0369, + "step": 24091 + }, + { + "epoch": 3.16, + "learning_rate": 1.380508192851988e-05, + "loss": 0.0443, + "step": 24092 + }, + { + "epoch": 3.16, + "learning_rate": 1.3804591490455362e-05, + "loss": 0.043, + "step": 24093 + }, + { + "epoch": 3.16, + "learning_rate": 1.3804101041690628e-05, + "loss": 0.0411, + "step": 24094 + }, + { + "epoch": 3.16, + "learning_rate": 1.3803610582227063e-05, + "loss": 0.0438, + "step": 24095 + }, + { + "epoch": 3.16, + "learning_rate": 1.3803120112066048e-05, + "loss": 0.0443, + "step": 24096 + }, + { + "epoch": 3.16, + "learning_rate": 1.3802629631208957e-05, + "loss": 0.041, + "step": 24097 + }, + { + "epoch": 3.16, + "learning_rate": 1.3802139139657174e-05, + "loss": 0.0432, + "step": 24098 + }, + { + "epoch": 3.16, + "learning_rate": 1.3801648637412078e-05, + "loss": 0.0379, + "step": 24099 + }, + { + "epoch": 3.16, + "learning_rate": 1.3801158124475047e-05, + "loss": 0.0363, + "step": 24100 + }, + { + "epoch": 3.16, + "learning_rate": 1.380066760084746e-05, + "loss": 0.0386, + "step": 24101 + }, + { + "epoch": 3.16, + "learning_rate": 1.3800177066530702e-05, + "loss": 0.0394, + "step": 24102 + }, + { + "epoch": 3.16, + "learning_rate": 1.379968652152614e-05, + "loss": 0.051, + "step": 24103 + }, + { + "epoch": 3.16, + "learning_rate": 1.379919596583517e-05, + "loss": 0.0458, + "step": 24104 + }, + { + "epoch": 3.16, + "learning_rate": 1.379870539945916e-05, + "loss": 0.0368, + "step": 24105 + }, + { + "epoch": 3.16, + "learning_rate": 1.3798214822399496e-05, + "loss": 0.0437, + "step": 24106 + }, + { + "epoch": 3.16, + "learning_rate": 1.3797724234657555e-05, + "loss": 0.0445, + "step": 24107 + }, + { + "epoch": 3.16, + "learning_rate": 1.3797233636234713e-05, + "loss": 0.0364, + "step": 24108 + }, + { + "epoch": 3.16, + "learning_rate": 1.3796743027132355e-05, + "loss": 0.0435, + "step": 24109 + }, + { + "epoch": 3.16, + "learning_rate": 1.3796252407351862e-05, + "loss": 0.047, + "step": 24110 + }, + { + "epoch": 3.16, + "learning_rate": 1.379576177689461e-05, + "loss": 0.0505, + "step": 24111 + }, + { + "epoch": 3.16, + "learning_rate": 1.3795271135761979e-05, + "loss": 0.0433, + "step": 24112 + }, + { + "epoch": 3.16, + "learning_rate": 1.3794780483955352e-05, + "loss": 0.0435, + "step": 24113 + }, + { + "epoch": 3.16, + "learning_rate": 1.3794289821476106e-05, + "loss": 0.0427, + "step": 24114 + }, + { + "epoch": 3.16, + "learning_rate": 1.3793799148325619e-05, + "loss": 0.0472, + "step": 24115 + }, + { + "epoch": 3.16, + "learning_rate": 1.3793308464505278e-05, + "loss": 0.0488, + "step": 24116 + }, + { + "epoch": 3.16, + "learning_rate": 1.3792817770016457e-05, + "loss": 0.0462, + "step": 24117 + }, + { + "epoch": 3.16, + "learning_rate": 1.3792327064860537e-05, + "loss": 0.0349, + "step": 24118 + }, + { + "epoch": 3.16, + "learning_rate": 1.3791836349038904e-05, + "loss": 0.0479, + "step": 24119 + }, + { + "epoch": 3.16, + "learning_rate": 1.3791345622552927e-05, + "loss": 0.0359, + "step": 24120 + }, + { + "epoch": 3.16, + "learning_rate": 1.3790854885403997e-05, + "loss": 0.0442, + "step": 24121 + }, + { + "epoch": 3.16, + "learning_rate": 1.3790364137593488e-05, + "loss": 0.0388, + "step": 24122 + }, + { + "epoch": 3.16, + "learning_rate": 1.3789873379122782e-05, + "loss": 0.0399, + "step": 24123 + }, + { + "epoch": 3.16, + "learning_rate": 1.3789382609993258e-05, + "loss": 0.0441, + "step": 24124 + }, + { + "epoch": 3.16, + "learning_rate": 1.3788891830206297e-05, + "loss": 0.0409, + "step": 24125 + }, + { + "epoch": 3.16, + "learning_rate": 1.378840103976328e-05, + "loss": 0.0395, + "step": 24126 + }, + { + "epoch": 3.16, + "learning_rate": 1.3787910238665586e-05, + "loss": 0.0421, + "step": 24127 + }, + { + "epoch": 3.16, + "learning_rate": 1.3787419426914599e-05, + "loss": 0.0419, + "step": 24128 + }, + { + "epoch": 3.16, + "learning_rate": 1.3786928604511695e-05, + "loss": 0.0413, + "step": 24129 + }, + { + "epoch": 3.16, + "learning_rate": 1.3786437771458252e-05, + "loss": 0.0445, + "step": 24130 + }, + { + "epoch": 3.16, + "learning_rate": 1.378594692775566e-05, + "loss": 0.0409, + "step": 24131 + }, + { + "epoch": 3.16, + "learning_rate": 1.378545607340529e-05, + "loss": 0.0429, + "step": 24132 + }, + { + "epoch": 3.16, + "learning_rate": 1.3784965208408532e-05, + "loss": 0.0376, + "step": 24133 + }, + { + "epoch": 3.16, + "learning_rate": 1.3784474332766755e-05, + "loss": 0.0388, + "step": 24134 + }, + { + "epoch": 3.16, + "learning_rate": 1.3783983446481348e-05, + "loss": 0.038, + "step": 24135 + }, + { + "epoch": 3.16, + "learning_rate": 1.3783492549553688e-05, + "loss": 0.0419, + "step": 24136 + }, + { + "epoch": 3.16, + "learning_rate": 1.3783001641985157e-05, + "loss": 0.0458, + "step": 24137 + }, + { + "epoch": 3.16, + "learning_rate": 1.3782510723777136e-05, + "loss": 0.035, + "step": 24138 + }, + { + "epoch": 3.16, + "learning_rate": 1.3782019794931004e-05, + "loss": 0.0428, + "step": 24139 + }, + { + "epoch": 3.16, + "learning_rate": 1.3781528855448145e-05, + "loss": 0.0417, + "step": 24140 + }, + { + "epoch": 3.16, + "learning_rate": 1.3781037905329932e-05, + "loss": 0.0426, + "step": 24141 + }, + { + "epoch": 3.16, + "learning_rate": 1.378054694457776e-05, + "loss": 0.0429, + "step": 24142 + }, + { + "epoch": 3.16, + "learning_rate": 1.3780055973192995e-05, + "loss": 0.0442, + "step": 24143 + }, + { + "epoch": 3.16, + "learning_rate": 1.3779564991177026e-05, + "loss": 0.048, + "step": 24144 + }, + { + "epoch": 3.16, + "learning_rate": 1.377907399853123e-05, + "loss": 0.0385, + "step": 24145 + }, + { + "epoch": 3.16, + "learning_rate": 1.3778582995256989e-05, + "loss": 0.0383, + "step": 24146 + }, + { + "epoch": 3.16, + "learning_rate": 1.3778091981355686e-05, + "loss": 0.0408, + "step": 24147 + }, + { + "epoch": 3.16, + "learning_rate": 1.37776009568287e-05, + "loss": 0.0478, + "step": 24148 + }, + { + "epoch": 3.16, + "learning_rate": 1.3777109921677414e-05, + "loss": 0.0467, + "step": 24149 + }, + { + "epoch": 3.16, + "learning_rate": 1.3776618875903207e-05, + "loss": 0.0369, + "step": 24150 + }, + { + "epoch": 3.16, + "learning_rate": 1.377612781950746e-05, + "loss": 0.0471, + "step": 24151 + }, + { + "epoch": 3.16, + "learning_rate": 1.3775636752491555e-05, + "loss": 0.0495, + "step": 24152 + }, + { + "epoch": 3.16, + "learning_rate": 1.3775145674856871e-05, + "loss": 0.0349, + "step": 24153 + }, + { + "epoch": 3.16, + "learning_rate": 1.3774654586604795e-05, + "loss": 0.0407, + "step": 24154 + }, + { + "epoch": 3.16, + "learning_rate": 1.3774163487736699e-05, + "loss": 0.0404, + "step": 24155 + }, + { + "epoch": 3.16, + "learning_rate": 1.377367237825397e-05, + "loss": 0.0477, + "step": 24156 + }, + { + "epoch": 3.16, + "learning_rate": 1.377318125815799e-05, + "loss": 0.0333, + "step": 24157 + }, + { + "epoch": 3.16, + "learning_rate": 1.377269012745014e-05, + "loss": 0.0505, + "step": 24158 + }, + { + "epoch": 3.16, + "learning_rate": 1.3772198986131798e-05, + "loss": 0.054, + "step": 24159 + }, + { + "epoch": 3.16, + "learning_rate": 1.3771707834204347e-05, + "loss": 0.033, + "step": 24160 + }, + { + "epoch": 3.16, + "learning_rate": 1.377121667166917e-05, + "loss": 0.0376, + "step": 24161 + }, + { + "epoch": 3.17, + "learning_rate": 1.3770725498527643e-05, + "loss": 0.0388, + "step": 24162 + }, + { + "epoch": 3.17, + "learning_rate": 1.3770234314781154e-05, + "loss": 0.0422, + "step": 24163 + }, + { + "epoch": 3.17, + "learning_rate": 1.3769743120431082e-05, + "loss": 0.048, + "step": 24164 + }, + { + "epoch": 3.17, + "learning_rate": 1.3769251915478811e-05, + "loss": 0.0441, + "step": 24165 + }, + { + "epoch": 3.17, + "learning_rate": 1.3768760699925716e-05, + "loss": 0.0435, + "step": 24166 + }, + { + "epoch": 3.17, + "learning_rate": 1.3768269473773183e-05, + "loss": 0.0399, + "step": 24167 + }, + { + "epoch": 3.17, + "learning_rate": 1.376777823702259e-05, + "loss": 0.0427, + "step": 24168 + }, + { + "epoch": 3.17, + "learning_rate": 1.3767286989675328e-05, + "loss": 0.0346, + "step": 24169 + }, + { + "epoch": 3.17, + "learning_rate": 1.3766795731732767e-05, + "loss": 0.0379, + "step": 24170 + }, + { + "epoch": 3.17, + "learning_rate": 1.3766304463196298e-05, + "loss": 0.0478, + "step": 24171 + }, + { + "epoch": 3.17, + "learning_rate": 1.3765813184067293e-05, + "loss": 0.036, + "step": 24172 + }, + { + "epoch": 3.17, + "learning_rate": 1.3765321894347143e-05, + "loss": 0.0387, + "step": 24173 + }, + { + "epoch": 3.17, + "learning_rate": 1.3764830594037223e-05, + "loss": 0.0425, + "step": 24174 + }, + { + "epoch": 3.17, + "learning_rate": 1.376433928313892e-05, + "loss": 0.0433, + "step": 24175 + }, + { + "epoch": 3.17, + "learning_rate": 1.3763847961653609e-05, + "loss": 0.041, + "step": 24176 + }, + { + "epoch": 3.17, + "learning_rate": 1.3763356629582679e-05, + "loss": 0.0432, + "step": 24177 + }, + { + "epoch": 3.17, + "learning_rate": 1.376286528692751e-05, + "loss": 0.0369, + "step": 24178 + }, + { + "epoch": 3.17, + "learning_rate": 1.376237393368948e-05, + "loss": 0.0427, + "step": 24179 + }, + { + "epoch": 3.17, + "learning_rate": 1.3761882569869972e-05, + "loss": 0.042, + "step": 24180 + }, + { + "epoch": 3.17, + "learning_rate": 1.3761391195470375e-05, + "loss": 0.0393, + "step": 24181 + }, + { + "epoch": 3.17, + "learning_rate": 1.3760899810492064e-05, + "loss": 0.0363, + "step": 24182 + }, + { + "epoch": 3.17, + "learning_rate": 1.3760408414936421e-05, + "loss": 0.0434, + "step": 24183 + }, + { + "epoch": 3.17, + "learning_rate": 1.3759917008804831e-05, + "loss": 0.0403, + "step": 24184 + }, + { + "epoch": 3.17, + "learning_rate": 1.3759425592098672e-05, + "loss": 0.0399, + "step": 24185 + }, + { + "epoch": 3.17, + "learning_rate": 1.3758934164819331e-05, + "loss": 0.0342, + "step": 24186 + }, + { + "epoch": 3.17, + "learning_rate": 1.3758442726968187e-05, + "loss": 0.0456, + "step": 24187 + }, + { + "epoch": 3.17, + "learning_rate": 1.3757951278546624e-05, + "loss": 0.039, + "step": 24188 + }, + { + "epoch": 3.17, + "learning_rate": 1.3757459819556023e-05, + "loss": 0.0497, + "step": 24189 + }, + { + "epoch": 3.17, + "learning_rate": 1.3756968349997766e-05, + "loss": 0.039, + "step": 24190 + }, + { + "epoch": 3.17, + "learning_rate": 1.3756476869873234e-05, + "loss": 0.0402, + "step": 24191 + }, + { + "epoch": 3.17, + "learning_rate": 1.3755985379183813e-05, + "loss": 0.0466, + "step": 24192 + }, + { + "epoch": 3.17, + "learning_rate": 1.3755493877930883e-05, + "loss": 0.0348, + "step": 24193 + }, + { + "epoch": 3.17, + "learning_rate": 1.3755002366115826e-05, + "loss": 0.0415, + "step": 24194 + }, + { + "epoch": 3.17, + "learning_rate": 1.3754510843740026e-05, + "loss": 0.0447, + "step": 24195 + }, + { + "epoch": 3.17, + "learning_rate": 1.3754019310804863e-05, + "loss": 0.032, + "step": 24196 + }, + { + "epoch": 3.17, + "learning_rate": 1.375352776731172e-05, + "loss": 0.035, + "step": 24197 + }, + { + "epoch": 3.17, + "learning_rate": 1.3753036213261982e-05, + "loss": 0.0482, + "step": 24198 + }, + { + "epoch": 3.17, + "learning_rate": 1.3752544648657024e-05, + "loss": 0.041, + "step": 24199 + }, + { + "epoch": 3.17, + "learning_rate": 1.375205307349824e-05, + "loss": 0.0497, + "step": 24200 + }, + { + "epoch": 3.17, + "learning_rate": 1.3751561487787002e-05, + "loss": 0.039, + "step": 24201 + }, + { + "epoch": 3.17, + "learning_rate": 1.37510698915247e-05, + "loss": 0.0363, + "step": 24202 + }, + { + "epoch": 3.17, + "learning_rate": 1.3750578284712715e-05, + "loss": 0.0424, + "step": 24203 + }, + { + "epoch": 3.17, + "learning_rate": 1.3750086667352427e-05, + "loss": 0.0399, + "step": 24204 + }, + { + "epoch": 3.17, + "learning_rate": 1.374959503944522e-05, + "loss": 0.0472, + "step": 24205 + }, + { + "epoch": 3.17, + "learning_rate": 1.3749103400992473e-05, + "loss": 0.0431, + "step": 24206 + }, + { + "epoch": 3.17, + "learning_rate": 1.3748611751995577e-05, + "loss": 0.038, + "step": 24207 + }, + { + "epoch": 3.17, + "learning_rate": 1.374812009245591e-05, + "loss": 0.0437, + "step": 24208 + }, + { + "epoch": 3.17, + "learning_rate": 1.3747628422374855e-05, + "loss": 0.0421, + "step": 24209 + }, + { + "epoch": 3.17, + "learning_rate": 1.3747136741753792e-05, + "loss": 0.0368, + "step": 24210 + }, + { + "epoch": 3.17, + "learning_rate": 1.374664505059411e-05, + "loss": 0.0425, + "step": 24211 + }, + { + "epoch": 3.17, + "learning_rate": 1.3746153348897183e-05, + "loss": 0.0442, + "step": 24212 + }, + { + "epoch": 3.17, + "learning_rate": 1.3745661636664403e-05, + "loss": 0.0396, + "step": 24213 + }, + { + "epoch": 3.17, + "learning_rate": 1.3745169913897147e-05, + "loss": 0.0429, + "step": 24214 + }, + { + "epoch": 3.17, + "learning_rate": 1.3744678180596802e-05, + "loss": 0.0417, + "step": 24215 + }, + { + "epoch": 3.17, + "learning_rate": 1.3744186436764748e-05, + "loss": 0.0444, + "step": 24216 + }, + { + "epoch": 3.17, + "learning_rate": 1.374369468240237e-05, + "loss": 0.0409, + "step": 24217 + }, + { + "epoch": 3.17, + "learning_rate": 1.374320291751105e-05, + "loss": 0.0416, + "step": 24218 + }, + { + "epoch": 3.17, + "learning_rate": 1.374271114209217e-05, + "loss": 0.0422, + "step": 24219 + }, + { + "epoch": 3.17, + "learning_rate": 1.3742219356147116e-05, + "loss": 0.0414, + "step": 24220 + }, + { + "epoch": 3.17, + "learning_rate": 1.3741727559677267e-05, + "loss": 0.0459, + "step": 24221 + }, + { + "epoch": 3.17, + "learning_rate": 1.374123575268401e-05, + "loss": 0.0416, + "step": 24222 + }, + { + "epoch": 3.17, + "learning_rate": 1.3740743935168728e-05, + "loss": 0.0459, + "step": 24223 + }, + { + "epoch": 3.17, + "learning_rate": 1.3740252107132798e-05, + "loss": 0.0412, + "step": 24224 + }, + { + "epoch": 3.17, + "learning_rate": 1.3739760268577613e-05, + "loss": 0.0388, + "step": 24225 + }, + { + "epoch": 3.17, + "learning_rate": 1.373926841950455e-05, + "loss": 0.0462, + "step": 24226 + }, + { + "epoch": 3.17, + "learning_rate": 1.3738776559914993e-05, + "loss": 0.0389, + "step": 24227 + }, + { + "epoch": 3.17, + "learning_rate": 1.3738284689810327e-05, + "loss": 0.047, + "step": 24228 + }, + { + "epoch": 3.17, + "learning_rate": 1.3737792809191932e-05, + "loss": 0.0386, + "step": 24229 + }, + { + "epoch": 3.17, + "learning_rate": 1.3737300918061198e-05, + "loss": 0.0422, + "step": 24230 + }, + { + "epoch": 3.17, + "learning_rate": 1.3736809016419502e-05, + "loss": 0.0448, + "step": 24231 + }, + { + "epoch": 3.17, + "learning_rate": 1.3736317104268229e-05, + "loss": 0.0391, + "step": 24232 + }, + { + "epoch": 3.17, + "learning_rate": 1.3735825181608765e-05, + "loss": 0.0439, + "step": 24233 + }, + { + "epoch": 3.17, + "learning_rate": 1.3735333248442489e-05, + "loss": 0.0422, + "step": 24234 + }, + { + "epoch": 3.17, + "learning_rate": 1.3734841304770787e-05, + "loss": 0.051, + "step": 24235 + }, + { + "epoch": 3.17, + "learning_rate": 1.3734349350595047e-05, + "loss": 0.0375, + "step": 24236 + }, + { + "epoch": 3.17, + "learning_rate": 1.3733857385916642e-05, + "loss": 0.0374, + "step": 24237 + }, + { + "epoch": 3.18, + "learning_rate": 1.3733365410736965e-05, + "loss": 0.0428, + "step": 24238 + }, + { + "epoch": 3.18, + "learning_rate": 1.3732873425057397e-05, + "loss": 0.0434, + "step": 24239 + }, + { + "epoch": 3.18, + "learning_rate": 1.3732381428879323e-05, + "loss": 0.0445, + "step": 24240 + }, + { + "epoch": 3.18, + "learning_rate": 1.3731889422204123e-05, + "loss": 0.0409, + "step": 24241 + }, + { + "epoch": 3.18, + "learning_rate": 1.3731397405033182e-05, + "loss": 0.044, + "step": 24242 + }, + { + "epoch": 3.18, + "learning_rate": 1.3730905377367887e-05, + "loss": 0.0444, + "step": 24243 + }, + { + "epoch": 3.18, + "learning_rate": 1.3730413339209618e-05, + "loss": 0.0355, + "step": 24244 + }, + { + "epoch": 3.18, + "learning_rate": 1.3729921290559758e-05, + "loss": 0.0385, + "step": 24245 + }, + { + "epoch": 3.18, + "learning_rate": 1.3729429231419695e-05, + "loss": 0.0433, + "step": 24246 + }, + { + "epoch": 3.18, + "learning_rate": 1.3728937161790812e-05, + "loss": 0.0476, + "step": 24247 + }, + { + "epoch": 3.18, + "learning_rate": 1.3728445081674488e-05, + "loss": 0.0378, + "step": 24248 + }, + { + "epoch": 3.18, + "learning_rate": 1.3727952991072115e-05, + "loss": 0.0454, + "step": 24249 + }, + { + "epoch": 3.18, + "learning_rate": 1.372746088998507e-05, + "loss": 0.0308, + "step": 24250 + }, + { + "epoch": 3.18, + "learning_rate": 1.3726968778414742e-05, + "loss": 0.0442, + "step": 24251 + }, + { + "epoch": 3.18, + "learning_rate": 1.3726476656362511e-05, + "loss": 0.0408, + "step": 24252 + }, + { + "epoch": 3.18, + "learning_rate": 1.3725984523829766e-05, + "loss": 0.0369, + "step": 24253 + }, + { + "epoch": 3.18, + "learning_rate": 1.3725492380817884e-05, + "loss": 0.0422, + "step": 24254 + }, + { + "epoch": 3.18, + "learning_rate": 1.3725000227328256e-05, + "loss": 0.046, + "step": 24255 + }, + { + "epoch": 3.18, + "learning_rate": 1.3724508063362258e-05, + "loss": 0.0454, + "step": 24256 + }, + { + "epoch": 3.18, + "learning_rate": 1.3724015888921285e-05, + "loss": 0.0392, + "step": 24257 + }, + { + "epoch": 3.18, + "learning_rate": 1.3723523704006714e-05, + "loss": 0.0372, + "step": 24258 + }, + { + "epoch": 3.18, + "learning_rate": 1.3723031508619933e-05, + "loss": 0.0366, + "step": 24259 + }, + { + "epoch": 3.18, + "learning_rate": 1.3722539302762321e-05, + "loss": 0.0398, + "step": 24260 + }, + { + "epoch": 3.18, + "learning_rate": 1.3722047086435266e-05, + "loss": 0.0422, + "step": 24261 + }, + { + "epoch": 3.18, + "learning_rate": 1.3721554859640152e-05, + "loss": 0.0341, + "step": 24262 + }, + { + "epoch": 3.18, + "learning_rate": 1.372106262237836e-05, + "loss": 0.0378, + "step": 24263 + }, + { + "epoch": 3.18, + "learning_rate": 1.3720570374651284e-05, + "loss": 0.0369, + "step": 24264 + }, + { + "epoch": 3.18, + "learning_rate": 1.3720078116460295e-05, + "loss": 0.0385, + "step": 24265 + }, + { + "epoch": 3.18, + "learning_rate": 1.3719585847806789e-05, + "loss": 0.0385, + "step": 24266 + }, + { + "epoch": 3.18, + "learning_rate": 1.3719093568692145e-05, + "loss": 0.0452, + "step": 24267 + }, + { + "epoch": 3.18, + "learning_rate": 1.371860127911775e-05, + "loss": 0.0453, + "step": 24268 + }, + { + "epoch": 3.18, + "learning_rate": 1.3718108979084982e-05, + "loss": 0.0389, + "step": 24269 + }, + { + "epoch": 3.18, + "learning_rate": 1.3717616668595238e-05, + "loss": 0.0496, + "step": 24270 + }, + { + "epoch": 3.18, + "learning_rate": 1.3717124347649891e-05, + "loss": 0.0411, + "step": 24271 + }, + { + "epoch": 3.18, + "learning_rate": 1.3716632016250328e-05, + "loss": 0.0336, + "step": 24272 + }, + { + "epoch": 3.18, + "learning_rate": 1.3716139674397936e-05, + "loss": 0.0362, + "step": 24273 + }, + { + "epoch": 3.18, + "learning_rate": 1.3715647322094102e-05, + "loss": 0.0388, + "step": 24274 + }, + { + "epoch": 3.18, + "learning_rate": 1.3715154959340205e-05, + "loss": 0.044, + "step": 24275 + }, + { + "epoch": 3.18, + "learning_rate": 1.3714662586137635e-05, + "loss": 0.0342, + "step": 24276 + }, + { + "epoch": 3.18, + "learning_rate": 1.3714170202487769e-05, + "loss": 0.0449, + "step": 24277 + }, + { + "epoch": 3.18, + "learning_rate": 1.3713677808392003e-05, + "loss": 0.0415, + "step": 24278 + }, + { + "epoch": 3.18, + "learning_rate": 1.3713185403851713e-05, + "loss": 0.0377, + "step": 24279 + }, + { + "epoch": 3.18, + "learning_rate": 1.3712692988868289e-05, + "loss": 0.0398, + "step": 24280 + }, + { + "epoch": 3.18, + "learning_rate": 1.3712200563443112e-05, + "loss": 0.0395, + "step": 24281 + }, + { + "epoch": 3.18, + "learning_rate": 1.3711708127577568e-05, + "loss": 0.0481, + "step": 24282 + }, + { + "epoch": 3.18, + "learning_rate": 1.3711215681273042e-05, + "loss": 0.0436, + "step": 24283 + }, + { + "epoch": 3.18, + "learning_rate": 1.3710723224530921e-05, + "loss": 0.0365, + "step": 24284 + }, + { + "epoch": 3.18, + "learning_rate": 1.371023075735259e-05, + "loss": 0.037, + "step": 24285 + }, + { + "epoch": 3.18, + "learning_rate": 1.3709738279739428e-05, + "loss": 0.0429, + "step": 24286 + }, + { + "epoch": 3.18, + "learning_rate": 1.370924579169283e-05, + "loss": 0.0366, + "step": 24287 + }, + { + "epoch": 3.18, + "learning_rate": 1.3708753293214171e-05, + "loss": 0.0409, + "step": 24288 + }, + { + "epoch": 3.18, + "learning_rate": 1.3708260784304846e-05, + "loss": 0.0419, + "step": 24289 + }, + { + "epoch": 3.18, + "learning_rate": 1.370776826496623e-05, + "loss": 0.0387, + "step": 24290 + }, + { + "epoch": 3.18, + "learning_rate": 1.3707275735199716e-05, + "loss": 0.04, + "step": 24291 + }, + { + "epoch": 3.18, + "learning_rate": 1.3706783195006686e-05, + "loss": 0.0372, + "step": 24292 + }, + { + "epoch": 3.18, + "learning_rate": 1.3706290644388526e-05, + "loss": 0.0423, + "step": 24293 + }, + { + "epoch": 3.18, + "learning_rate": 1.3705798083346617e-05, + "loss": 0.0391, + "step": 24294 + }, + { + "epoch": 3.18, + "learning_rate": 1.3705305511882353e-05, + "loss": 0.0388, + "step": 24295 + }, + { + "epoch": 3.18, + "learning_rate": 1.3704812929997113e-05, + "loss": 0.0356, + "step": 24296 + }, + { + "epoch": 3.18, + "learning_rate": 1.3704320337692287e-05, + "loss": 0.0441, + "step": 24297 + }, + { + "epoch": 3.18, + "learning_rate": 1.3703827734969253e-05, + "loss": 0.0441, + "step": 24298 + }, + { + "epoch": 3.18, + "learning_rate": 1.3703335121829402e-05, + "loss": 0.0439, + "step": 24299 + }, + { + "epoch": 3.18, + "learning_rate": 1.3702842498274118e-05, + "loss": 0.0377, + "step": 24300 + }, + { + "epoch": 3.18, + "learning_rate": 1.3702349864304788e-05, + "loss": 0.0431, + "step": 24301 + }, + { + "epoch": 3.18, + "learning_rate": 1.3701857219922797e-05, + "loss": 0.0435, + "step": 24302 + }, + { + "epoch": 3.18, + "learning_rate": 1.3701364565129524e-05, + "loss": 0.0484, + "step": 24303 + }, + { + "epoch": 3.18, + "learning_rate": 1.3700871899926367e-05, + "loss": 0.0373, + "step": 24304 + }, + { + "epoch": 3.18, + "learning_rate": 1.3700379224314702e-05, + "loss": 0.0411, + "step": 24305 + }, + { + "epoch": 3.18, + "learning_rate": 1.3699886538295917e-05, + "loss": 0.0412, + "step": 24306 + }, + { + "epoch": 3.18, + "learning_rate": 1.3699393841871398e-05, + "loss": 0.0405, + "step": 24307 + }, + { + "epoch": 3.18, + "learning_rate": 1.3698901135042532e-05, + "loss": 0.0369, + "step": 24308 + }, + { + "epoch": 3.18, + "learning_rate": 1.3698408417810703e-05, + "loss": 0.0319, + "step": 24309 + }, + { + "epoch": 3.18, + "learning_rate": 1.3697915690177296e-05, + "loss": 0.0371, + "step": 24310 + }, + { + "epoch": 3.18, + "learning_rate": 1.3697422952143699e-05, + "loss": 0.0375, + "step": 24311 + }, + { + "epoch": 3.18, + "learning_rate": 1.3696930203711298e-05, + "loss": 0.0418, + "step": 24312 + }, + { + "epoch": 3.18, + "learning_rate": 1.3696437444881476e-05, + "loss": 0.0429, + "step": 24313 + }, + { + "epoch": 3.18, + "learning_rate": 1.3695944675655623e-05, + "loss": 0.0422, + "step": 24314 + }, + { + "epoch": 3.19, + "learning_rate": 1.3695451896035122e-05, + "loss": 0.0443, + "step": 24315 + }, + { + "epoch": 3.19, + "learning_rate": 1.3694959106021357e-05, + "loss": 0.0381, + "step": 24316 + }, + { + "epoch": 3.19, + "learning_rate": 1.3694466305615718e-05, + "loss": 0.0474, + "step": 24317 + }, + { + "epoch": 3.19, + "learning_rate": 1.3693973494819588e-05, + "loss": 0.0384, + "step": 24318 + }, + { + "epoch": 3.19, + "learning_rate": 1.3693480673634355e-05, + "loss": 0.0476, + "step": 24319 + }, + { + "epoch": 3.19, + "learning_rate": 1.3692987842061405e-05, + "loss": 0.0466, + "step": 24320 + }, + { + "epoch": 3.19, + "learning_rate": 1.369249500010212e-05, + "loss": 0.0405, + "step": 24321 + }, + { + "epoch": 3.19, + "learning_rate": 1.3692002147757895e-05, + "loss": 0.0445, + "step": 24322 + }, + { + "epoch": 3.19, + "learning_rate": 1.3691509285030104e-05, + "loss": 0.0382, + "step": 24323 + }, + { + "epoch": 3.19, + "learning_rate": 1.3691016411920146e-05, + "loss": 0.0384, + "step": 24324 + }, + { + "epoch": 3.19, + "learning_rate": 1.3690523528429399e-05, + "loss": 0.0417, + "step": 24325 + }, + { + "epoch": 3.19, + "learning_rate": 1.3690030634559249e-05, + "loss": 0.0419, + "step": 24326 + }, + { + "epoch": 3.19, + "learning_rate": 1.3689537730311085e-05, + "loss": 0.0399, + "step": 24327 + }, + { + "epoch": 3.19, + "learning_rate": 1.3689044815686292e-05, + "loss": 0.0455, + "step": 24328 + }, + { + "epoch": 3.19, + "learning_rate": 1.3688551890686261e-05, + "loss": 0.0457, + "step": 24329 + }, + { + "epoch": 3.19, + "learning_rate": 1.3688058955312371e-05, + "loss": 0.0474, + "step": 24330 + }, + { + "epoch": 3.19, + "learning_rate": 1.3687566009566014e-05, + "loss": 0.042, + "step": 24331 + }, + { + "epoch": 3.19, + "learning_rate": 1.3687073053448568e-05, + "loss": 0.0455, + "step": 24332 + }, + { + "epoch": 3.19, + "learning_rate": 1.368658008696143e-05, + "loss": 0.0475, + "step": 24333 + }, + { + "epoch": 3.19, + "learning_rate": 1.3686087110105981e-05, + "loss": 0.039, + "step": 24334 + }, + { + "epoch": 3.19, + "learning_rate": 1.368559412288361e-05, + "loss": 0.0433, + "step": 24335 + }, + { + "epoch": 3.19, + "learning_rate": 1.3685101125295699e-05, + "loss": 0.0486, + "step": 24336 + }, + { + "epoch": 3.19, + "learning_rate": 1.3684608117343641e-05, + "loss": 0.0412, + "step": 24337 + }, + { + "epoch": 3.19, + "learning_rate": 1.3684115099028813e-05, + "loss": 0.0416, + "step": 24338 + }, + { + "epoch": 3.19, + "learning_rate": 1.3683622070352614e-05, + "loss": 0.0422, + "step": 24339 + }, + { + "epoch": 3.19, + "learning_rate": 1.368312903131642e-05, + "loss": 0.0361, + "step": 24340 + }, + { + "epoch": 3.19, + "learning_rate": 1.3682635981921624e-05, + "loss": 0.0505, + "step": 24341 + }, + { + "epoch": 3.19, + "learning_rate": 1.368214292216961e-05, + "loss": 0.0482, + "step": 24342 + }, + { + "epoch": 3.19, + "learning_rate": 1.3681649852061762e-05, + "loss": 0.0396, + "step": 24343 + }, + { + "epoch": 3.19, + "learning_rate": 1.3681156771599473e-05, + "loss": 0.0438, + "step": 24344 + }, + { + "epoch": 3.19, + "learning_rate": 1.3680663680784127e-05, + "loss": 0.0439, + "step": 24345 + }, + { + "epoch": 3.19, + "learning_rate": 1.3680170579617108e-05, + "loss": 0.0413, + "step": 24346 + }, + { + "epoch": 3.19, + "learning_rate": 1.3679677468099806e-05, + "loss": 0.0459, + "step": 24347 + }, + { + "epoch": 3.19, + "learning_rate": 1.3679184346233607e-05, + "loss": 0.0314, + "step": 24348 + }, + { + "epoch": 3.19, + "learning_rate": 1.3678691214019902e-05, + "loss": 0.04, + "step": 24349 + }, + { + "epoch": 3.19, + "learning_rate": 1.3678198071460069e-05, + "loss": 0.0354, + "step": 24350 + }, + { + "epoch": 3.19, + "learning_rate": 1.3677704918555498e-05, + "loss": 0.0502, + "step": 24351 + }, + { + "epoch": 3.19, + "learning_rate": 1.3677211755307585e-05, + "loss": 0.0312, + "step": 24352 + }, + { + "epoch": 3.19, + "learning_rate": 1.3676718581717703e-05, + "loss": 0.041, + "step": 24353 + }, + { + "epoch": 3.19, + "learning_rate": 1.3676225397787248e-05, + "loss": 0.0339, + "step": 24354 + }, + { + "epoch": 3.19, + "learning_rate": 1.3675732203517607e-05, + "loss": 0.0401, + "step": 24355 + }, + { + "epoch": 3.19, + "learning_rate": 1.3675238998910162e-05, + "loss": 0.05, + "step": 24356 + }, + { + "epoch": 3.19, + "learning_rate": 1.3674745783966306e-05, + "loss": 0.044, + "step": 24357 + }, + { + "epoch": 3.19, + "learning_rate": 1.3674252558687422e-05, + "loss": 0.0439, + "step": 24358 + }, + { + "epoch": 3.19, + "learning_rate": 1.3673759323074896e-05, + "loss": 0.0398, + "step": 24359 + }, + { + "epoch": 3.19, + "learning_rate": 1.367326607713012e-05, + "loss": 0.031, + "step": 24360 + }, + { + "epoch": 3.19, + "learning_rate": 1.3672772820854477e-05, + "loss": 0.0415, + "step": 24361 + }, + { + "epoch": 3.19, + "learning_rate": 1.3672279554249359e-05, + "loss": 0.0446, + "step": 24362 + }, + { + "epoch": 3.19, + "learning_rate": 1.3671786277316148e-05, + "loss": 0.0399, + "step": 24363 + }, + { + "epoch": 3.19, + "learning_rate": 1.3671292990056233e-05, + "loss": 0.0386, + "step": 24364 + }, + { + "epoch": 3.19, + "learning_rate": 1.3670799692471002e-05, + "loss": 0.044, + "step": 24365 + }, + { + "epoch": 3.19, + "learning_rate": 1.3670306384561845e-05, + "loss": 0.0439, + "step": 24366 + }, + { + "epoch": 3.19, + "learning_rate": 1.3669813066330144e-05, + "loss": 0.0351, + "step": 24367 + }, + { + "epoch": 3.19, + "learning_rate": 1.3669319737777291e-05, + "loss": 0.0428, + "step": 24368 + }, + { + "epoch": 3.19, + "learning_rate": 1.3668826398904672e-05, + "loss": 0.0419, + "step": 24369 + }, + { + "epoch": 3.19, + "learning_rate": 1.3668333049713671e-05, + "loss": 0.0432, + "step": 24370 + }, + { + "epoch": 3.19, + "learning_rate": 1.3667839690205681e-05, + "loss": 0.0351, + "step": 24371 + }, + { + "epoch": 3.19, + "learning_rate": 1.3667346320382087e-05, + "loss": 0.0479, + "step": 24372 + }, + { + "epoch": 3.19, + "learning_rate": 1.3666852940244278e-05, + "loss": 0.0419, + "step": 24373 + }, + { + "epoch": 3.19, + "learning_rate": 1.3666359549793638e-05, + "loss": 0.0425, + "step": 24374 + }, + { + "epoch": 3.19, + "learning_rate": 1.3665866149031559e-05, + "loss": 0.0405, + "step": 24375 + }, + { + "epoch": 3.19, + "learning_rate": 1.3665372737959423e-05, + "loss": 0.0326, + "step": 24376 + }, + { + "epoch": 3.19, + "learning_rate": 1.3664879316578628e-05, + "loss": 0.0432, + "step": 24377 + }, + { + "epoch": 3.19, + "learning_rate": 1.366438588489055e-05, + "loss": 0.0417, + "step": 24378 + }, + { + "epoch": 3.19, + "learning_rate": 1.3663892442896582e-05, + "loss": 0.0427, + "step": 24379 + }, + { + "epoch": 3.19, + "learning_rate": 1.3663398990598113e-05, + "loss": 0.0424, + "step": 24380 + }, + { + "epoch": 3.19, + "learning_rate": 1.366290552799653e-05, + "loss": 0.0391, + "step": 24381 + }, + { + "epoch": 3.19, + "learning_rate": 1.3662412055093218e-05, + "loss": 0.0454, + "step": 24382 + }, + { + "epoch": 3.19, + "learning_rate": 1.366191857188957e-05, + "loss": 0.0506, + "step": 24383 + }, + { + "epoch": 3.19, + "learning_rate": 1.366142507838697e-05, + "loss": 0.0395, + "step": 24384 + }, + { + "epoch": 3.19, + "learning_rate": 1.3660931574586807e-05, + "loss": 0.046, + "step": 24385 + }, + { + "epoch": 3.19, + "learning_rate": 1.3660438060490468e-05, + "loss": 0.0505, + "step": 24386 + }, + { + "epoch": 3.19, + "learning_rate": 1.3659944536099341e-05, + "loss": 0.0456, + "step": 24387 + }, + { + "epoch": 3.19, + "learning_rate": 1.3659451001414816e-05, + "loss": 0.0444, + "step": 24388 + }, + { + "epoch": 3.19, + "learning_rate": 1.365895745643828e-05, + "loss": 0.0462, + "step": 24389 + }, + { + "epoch": 3.19, + "learning_rate": 1.3658463901171121e-05, + "loss": 0.0358, + "step": 24390 + }, + { + "epoch": 3.2, + "learning_rate": 1.3657970335614728e-05, + "loss": 0.0454, + "step": 24391 + }, + { + "epoch": 3.2, + "learning_rate": 1.3657476759770487e-05, + "loss": 0.0424, + "step": 24392 + }, + { + "epoch": 3.2, + "learning_rate": 1.3656983173639786e-05, + "loss": 0.0429, + "step": 24393 + }, + { + "epoch": 3.2, + "learning_rate": 1.365648957722402e-05, + "loss": 0.0434, + "step": 24394 + }, + { + "epoch": 3.2, + "learning_rate": 1.3655995970524565e-05, + "loss": 0.0416, + "step": 24395 + }, + { + "epoch": 3.2, + "learning_rate": 1.3655502353542821e-05, + "loss": 0.0418, + "step": 24396 + }, + { + "epoch": 3.2, + "learning_rate": 1.3655008726280168e-05, + "loss": 0.0433, + "step": 24397 + }, + { + "epoch": 3.2, + "learning_rate": 1.3654515088738001e-05, + "loss": 0.0404, + "step": 24398 + }, + { + "epoch": 3.2, + "learning_rate": 1.3654021440917702e-05, + "loss": 0.0353, + "step": 24399 + }, + { + "epoch": 3.2, + "learning_rate": 1.3653527782820663e-05, + "loss": 0.0351, + "step": 24400 + }, + { + "epoch": 3.2, + "learning_rate": 1.3653034114448272e-05, + "loss": 0.036, + "step": 24401 + }, + { + "epoch": 3.2, + "learning_rate": 1.365254043580192e-05, + "loss": 0.0485, + "step": 24402 + }, + { + "epoch": 3.2, + "learning_rate": 1.3652046746882985e-05, + "loss": 0.0447, + "step": 24403 + }, + { + "epoch": 3.2, + "learning_rate": 1.3651553047692869e-05, + "loss": 0.0392, + "step": 24404 + }, + { + "epoch": 3.2, + "learning_rate": 1.3651059338232955e-05, + "loss": 0.0447, + "step": 24405 + }, + { + "epoch": 3.2, + "learning_rate": 1.3650565618504629e-05, + "loss": 0.0445, + "step": 24406 + }, + { + "epoch": 3.2, + "learning_rate": 1.3650071888509282e-05, + "loss": 0.0387, + "step": 24407 + }, + { + "epoch": 3.2, + "learning_rate": 1.36495781482483e-05, + "loss": 0.0372, + "step": 24408 + }, + { + "epoch": 3.2, + "learning_rate": 1.3649084397723076e-05, + "loss": 0.0379, + "step": 24409 + }, + { + "epoch": 3.2, + "learning_rate": 1.3648590636934998e-05, + "loss": 0.044, + "step": 24410 + }, + { + "epoch": 3.2, + "learning_rate": 1.3648096865885449e-05, + "loss": 0.0394, + "step": 24411 + }, + { + "epoch": 3.2, + "learning_rate": 1.3647603084575824e-05, + "loss": 0.0417, + "step": 24412 + }, + { + "epoch": 3.2, + "learning_rate": 1.3647109293007512e-05, + "loss": 0.044, + "step": 24413 + }, + { + "epoch": 3.2, + "learning_rate": 1.3646615491181894e-05, + "loss": 0.0386, + "step": 24414 + }, + { + "epoch": 3.2, + "learning_rate": 1.3646121679100368e-05, + "loss": 0.0405, + "step": 24415 + }, + { + "epoch": 3.2, + "learning_rate": 1.3645627856764318e-05, + "loss": 0.0437, + "step": 24416 + }, + { + "epoch": 3.2, + "learning_rate": 1.3645134024175136e-05, + "loss": 0.0448, + "step": 24417 + }, + { + "epoch": 3.2, + "learning_rate": 1.3644640181334205e-05, + "loss": 0.045, + "step": 24418 + }, + { + "epoch": 3.2, + "learning_rate": 1.364414632824292e-05, + "loss": 0.0375, + "step": 24419 + }, + { + "epoch": 3.2, + "learning_rate": 1.3643652464902666e-05, + "loss": 0.0435, + "step": 24420 + }, + { + "epoch": 3.2, + "learning_rate": 1.3643158591314836e-05, + "loss": 0.0437, + "step": 24421 + }, + { + "epoch": 3.2, + "learning_rate": 1.3642664707480812e-05, + "loss": 0.0485, + "step": 24422 + }, + { + "epoch": 3.2, + "learning_rate": 1.3642170813401993e-05, + "loss": 0.0445, + "step": 24423 + }, + { + "epoch": 3.2, + "learning_rate": 1.364167690907976e-05, + "loss": 0.0445, + "step": 24424 + }, + { + "epoch": 3.2, + "learning_rate": 1.3641182994515503e-05, + "loss": 0.0417, + "step": 24425 + }, + { + "epoch": 3.2, + "learning_rate": 1.3640689069710613e-05, + "loss": 0.033, + "step": 24426 + }, + { + "epoch": 3.2, + "learning_rate": 1.3640195134666483e-05, + "loss": 0.0406, + "step": 24427 + }, + { + "epoch": 3.2, + "learning_rate": 1.3639701189384494e-05, + "loss": 0.0415, + "step": 24428 + }, + { + "epoch": 3.2, + "learning_rate": 1.3639207233866038e-05, + "loss": 0.0332, + "step": 24429 + }, + { + "epoch": 3.2, + "learning_rate": 1.3638713268112507e-05, + "loss": 0.0428, + "step": 24430 + }, + { + "epoch": 3.2, + "learning_rate": 1.3638219292125292e-05, + "loss": 0.0427, + "step": 24431 + }, + { + "epoch": 3.2, + "learning_rate": 1.3637725305905776e-05, + "loss": 0.0389, + "step": 24432 + }, + { + "epoch": 3.2, + "learning_rate": 1.363723130945535e-05, + "loss": 0.0459, + "step": 24433 + }, + { + "epoch": 3.2, + "learning_rate": 1.3636737302775407e-05, + "loss": 0.0457, + "step": 24434 + }, + { + "epoch": 3.2, + "learning_rate": 1.3636243285867334e-05, + "loss": 0.0391, + "step": 24435 + }, + { + "epoch": 3.2, + "learning_rate": 1.3635749258732519e-05, + "loss": 0.0526, + "step": 24436 + }, + { + "epoch": 3.2, + "learning_rate": 1.3635255221372352e-05, + "loss": 0.0458, + "step": 24437 + }, + { + "epoch": 3.2, + "learning_rate": 1.3634761173788227e-05, + "loss": 0.042, + "step": 24438 + }, + { + "epoch": 3.2, + "learning_rate": 1.3634267115981525e-05, + "loss": 0.0369, + "step": 24439 + }, + { + "epoch": 3.2, + "learning_rate": 1.3633773047953646e-05, + "loss": 0.0448, + "step": 24440 + }, + { + "epoch": 3.2, + "learning_rate": 1.3633278969705968e-05, + "loss": 0.0412, + "step": 24441 + }, + { + "epoch": 3.2, + "learning_rate": 1.363278488123989e-05, + "loss": 0.0457, + "step": 24442 + }, + { + "epoch": 3.2, + "learning_rate": 1.3632290782556797e-05, + "loss": 0.0374, + "step": 24443 + }, + { + "epoch": 3.2, + "learning_rate": 1.3631796673658083e-05, + "loss": 0.0475, + "step": 24444 + }, + { + "epoch": 3.2, + "learning_rate": 1.3631302554545127e-05, + "loss": 0.0402, + "step": 24445 + }, + { + "epoch": 3.2, + "learning_rate": 1.3630808425219334e-05, + "loss": 0.0354, + "step": 24446 + }, + { + "epoch": 3.2, + "learning_rate": 1.3630314285682076e-05, + "loss": 0.0449, + "step": 24447 + }, + { + "epoch": 3.2, + "learning_rate": 1.3629820135934759e-05, + "loss": 0.0413, + "step": 24448 + }, + { + "epoch": 3.2, + "learning_rate": 1.3629325975978767e-05, + "loss": 0.0469, + "step": 24449 + }, + { + "epoch": 3.2, + "learning_rate": 1.3628831805815488e-05, + "loss": 0.0444, + "step": 24450 + }, + { + "epoch": 3.2, + "learning_rate": 1.362833762544631e-05, + "loss": 0.0479, + "step": 24451 + }, + { + "epoch": 3.2, + "learning_rate": 1.3627843434872627e-05, + "loss": 0.0388, + "step": 24452 + }, + { + "epoch": 3.2, + "learning_rate": 1.3627349234095829e-05, + "loss": 0.0434, + "step": 24453 + }, + { + "epoch": 3.2, + "learning_rate": 1.36268550231173e-05, + "loss": 0.0392, + "step": 24454 + }, + { + "epoch": 3.2, + "learning_rate": 1.362636080193844e-05, + "loss": 0.0471, + "step": 24455 + }, + { + "epoch": 3.2, + "learning_rate": 1.362586657056063e-05, + "loss": 0.0492, + "step": 24456 + }, + { + "epoch": 3.2, + "learning_rate": 1.3625372328985267e-05, + "loss": 0.0451, + "step": 24457 + }, + { + "epoch": 3.2, + "learning_rate": 1.3624878077213732e-05, + "loss": 0.0418, + "step": 24458 + }, + { + "epoch": 3.2, + "learning_rate": 1.3624383815247426e-05, + "loss": 0.0468, + "step": 24459 + }, + { + "epoch": 3.2, + "learning_rate": 1.3623889543087729e-05, + "loss": 0.0396, + "step": 24460 + }, + { + "epoch": 3.2, + "learning_rate": 1.3623395260736039e-05, + "loss": 0.0433, + "step": 24461 + }, + { + "epoch": 3.2, + "learning_rate": 1.3622900968193742e-05, + "loss": 0.0402, + "step": 24462 + }, + { + "epoch": 3.2, + "learning_rate": 1.3622406665462226e-05, + "loss": 0.0398, + "step": 24463 + }, + { + "epoch": 3.2, + "learning_rate": 1.3621912352542887e-05, + "loss": 0.0393, + "step": 24464 + }, + { + "epoch": 3.2, + "learning_rate": 1.3621418029437113e-05, + "loss": 0.0417, + "step": 24465 + }, + { + "epoch": 3.2, + "learning_rate": 1.3620923696146291e-05, + "loss": 0.038, + "step": 24466 + }, + { + "epoch": 3.21, + "learning_rate": 1.3620429352671817e-05, + "loss": 0.0399, + "step": 24467 + }, + { + "epoch": 3.21, + "learning_rate": 1.3619934999015074e-05, + "loss": 0.0408, + "step": 24468 + }, + { + "epoch": 3.21, + "learning_rate": 1.361944063517746e-05, + "loss": 0.0365, + "step": 24469 + }, + { + "epoch": 3.21, + "learning_rate": 1.3618946261160361e-05, + "loss": 0.0487, + "step": 24470 + }, + { + "epoch": 3.21, + "learning_rate": 1.361845187696517e-05, + "loss": 0.0449, + "step": 24471 + }, + { + "epoch": 3.21, + "learning_rate": 1.3617957482593274e-05, + "loss": 0.0414, + "step": 24472 + }, + { + "epoch": 3.21, + "learning_rate": 1.3617463078046065e-05, + "loss": 0.0387, + "step": 24473 + }, + { + "epoch": 3.21, + "learning_rate": 1.3616968663324934e-05, + "loss": 0.041, + "step": 24474 + }, + { + "epoch": 3.21, + "learning_rate": 1.3616474238431274e-05, + "loss": 0.0401, + "step": 24475 + }, + { + "epoch": 3.21, + "learning_rate": 1.3615979803366468e-05, + "loss": 0.0357, + "step": 24476 + }, + { + "epoch": 3.21, + "learning_rate": 1.3615485358131914e-05, + "loss": 0.0409, + "step": 24477 + }, + { + "epoch": 3.21, + "learning_rate": 1.3614990902729001e-05, + "loss": 0.0367, + "step": 24478 + }, + { + "epoch": 3.21, + "learning_rate": 1.3614496437159115e-05, + "loss": 0.0459, + "step": 24479 + }, + { + "epoch": 3.21, + "learning_rate": 1.3614001961423652e-05, + "loss": 0.0417, + "step": 24480 + }, + { + "epoch": 3.21, + "learning_rate": 1.3613507475524004e-05, + "loss": 0.0419, + "step": 24481 + }, + { + "epoch": 3.21, + "learning_rate": 1.3613012979461556e-05, + "loss": 0.0469, + "step": 24482 + }, + { + "epoch": 3.21, + "learning_rate": 1.3612518473237702e-05, + "loss": 0.0394, + "step": 24483 + }, + { + "epoch": 3.21, + "learning_rate": 1.3612023956853832e-05, + "loss": 0.0484, + "step": 24484 + }, + { + "epoch": 3.21, + "learning_rate": 1.3611529430311334e-05, + "loss": 0.0443, + "step": 24485 + }, + { + "epoch": 3.21, + "learning_rate": 1.3611034893611606e-05, + "loss": 0.0402, + "step": 24486 + }, + { + "epoch": 3.21, + "learning_rate": 1.3610540346756031e-05, + "loss": 0.0481, + "step": 24487 + }, + { + "epoch": 3.21, + "learning_rate": 1.361004578974601e-05, + "loss": 0.0377, + "step": 24488 + }, + { + "epoch": 3.21, + "learning_rate": 1.3609551222582922e-05, + "loss": 0.0448, + "step": 24489 + }, + { + "epoch": 3.21, + "learning_rate": 1.3609056645268164e-05, + "loss": 0.0336, + "step": 24490 + }, + { + "epoch": 3.21, + "learning_rate": 1.3608562057803126e-05, + "loss": 0.0453, + "step": 24491 + }, + { + "epoch": 3.21, + "learning_rate": 1.3608067460189201e-05, + "loss": 0.041, + "step": 24492 + }, + { + "epoch": 3.21, + "learning_rate": 1.3607572852427778e-05, + "loss": 0.0405, + "step": 24493 + }, + { + "epoch": 3.21, + "learning_rate": 1.3607078234520245e-05, + "loss": 0.0396, + "step": 24494 + }, + { + "epoch": 3.21, + "learning_rate": 1.3606583606468002e-05, + "loss": 0.0413, + "step": 24495 + }, + { + "epoch": 3.21, + "learning_rate": 1.3606088968272433e-05, + "loss": 0.0378, + "step": 24496 + }, + { + "epoch": 3.21, + "learning_rate": 1.360559431993493e-05, + "loss": 0.0423, + "step": 24497 + }, + { + "epoch": 3.21, + "learning_rate": 1.3605099661456884e-05, + "loss": 0.039, + "step": 24498 + }, + { + "epoch": 3.21, + "learning_rate": 1.3604604992839689e-05, + "loss": 0.036, + "step": 24499 + }, + { + "epoch": 3.21, + "learning_rate": 1.3604110314084733e-05, + "loss": 0.0488, + "step": 24500 + }, + { + "epoch": 3.21, + "learning_rate": 1.360361562519341e-05, + "loss": 0.0423, + "step": 24501 + }, + { + "epoch": 3.21, + "learning_rate": 1.3603120926167109e-05, + "loss": 0.0394, + "step": 24502 + }, + { + "epoch": 3.21, + "learning_rate": 1.3602626217007224e-05, + "loss": 0.044, + "step": 24503 + }, + { + "epoch": 3.21, + "learning_rate": 1.3602131497715143e-05, + "loss": 0.0357, + "step": 24504 + }, + { + "epoch": 3.21, + "learning_rate": 1.3601636768292262e-05, + "loss": 0.046, + "step": 24505 + }, + { + "epoch": 3.21, + "learning_rate": 1.3601142028739964e-05, + "loss": 0.0461, + "step": 24506 + }, + { + "epoch": 3.21, + "learning_rate": 1.360064727905965e-05, + "loss": 0.0394, + "step": 24507 + }, + { + "epoch": 3.21, + "learning_rate": 1.3600152519252702e-05, + "loss": 0.0422, + "step": 24508 + }, + { + "epoch": 3.21, + "learning_rate": 1.3599657749320523e-05, + "loss": 0.0392, + "step": 24509 + }, + { + "epoch": 3.21, + "learning_rate": 1.3599162969264494e-05, + "loss": 0.0425, + "step": 24510 + }, + { + "epoch": 3.21, + "learning_rate": 1.3598668179086016e-05, + "loss": 0.0473, + "step": 24511 + }, + { + "epoch": 3.21, + "learning_rate": 1.359817337878647e-05, + "loss": 0.035, + "step": 24512 + }, + { + "epoch": 3.21, + "learning_rate": 1.3597678568367256e-05, + "loss": 0.0462, + "step": 24513 + }, + { + "epoch": 3.21, + "learning_rate": 1.359718374782976e-05, + "loss": 0.0416, + "step": 24514 + }, + { + "epoch": 3.21, + "learning_rate": 1.359668891717538e-05, + "loss": 0.042, + "step": 24515 + }, + { + "epoch": 3.21, + "learning_rate": 1.3596194076405503e-05, + "loss": 0.0387, + "step": 24516 + }, + { + "epoch": 3.21, + "learning_rate": 1.3595699225521518e-05, + "loss": 0.0467, + "step": 24517 + }, + { + "epoch": 3.21, + "learning_rate": 1.3595204364524822e-05, + "loss": 0.037, + "step": 24518 + }, + { + "epoch": 3.21, + "learning_rate": 1.3594709493416807e-05, + "loss": 0.0383, + "step": 24519 + }, + { + "epoch": 3.21, + "learning_rate": 1.3594214612198862e-05, + "loss": 0.0443, + "step": 24520 + }, + { + "epoch": 3.21, + "learning_rate": 1.359371972087238e-05, + "loss": 0.0392, + "step": 24521 + }, + { + "epoch": 3.21, + "learning_rate": 1.3593224819438753e-05, + "loss": 0.0428, + "step": 24522 + }, + { + "epoch": 3.21, + "learning_rate": 1.3592729907899372e-05, + "loss": 0.0426, + "step": 24523 + }, + { + "epoch": 3.21, + "learning_rate": 1.3592234986255627e-05, + "loss": 0.0346, + "step": 24524 + }, + { + "epoch": 3.21, + "learning_rate": 1.3591740054508917e-05, + "loss": 0.0432, + "step": 24525 + }, + { + "epoch": 3.21, + "learning_rate": 1.359124511266063e-05, + "loss": 0.0442, + "step": 24526 + }, + { + "epoch": 3.21, + "learning_rate": 1.3590750160712152e-05, + "loss": 0.0471, + "step": 24527 + }, + { + "epoch": 3.21, + "learning_rate": 1.3590255198664886e-05, + "loss": 0.0386, + "step": 24528 + }, + { + "epoch": 3.21, + "learning_rate": 1.3589760226520212e-05, + "loss": 0.0395, + "step": 24529 + }, + { + "epoch": 3.21, + "learning_rate": 1.3589265244279533e-05, + "loss": 0.0431, + "step": 24530 + }, + { + "epoch": 3.21, + "learning_rate": 1.3588770251944237e-05, + "loss": 0.0531, + "step": 24531 + }, + { + "epoch": 3.21, + "learning_rate": 1.3588275249515715e-05, + "loss": 0.0366, + "step": 24532 + }, + { + "epoch": 3.21, + "learning_rate": 1.3587780236995357e-05, + "loss": 0.0367, + "step": 24533 + }, + { + "epoch": 3.21, + "learning_rate": 1.3587285214384563e-05, + "loss": 0.0384, + "step": 24534 + }, + { + "epoch": 3.21, + "learning_rate": 1.3586790181684714e-05, + "loss": 0.0455, + "step": 24535 + }, + { + "epoch": 3.21, + "learning_rate": 1.3586295138897217e-05, + "loss": 0.0458, + "step": 24536 + }, + { + "epoch": 3.21, + "learning_rate": 1.3585800086023451e-05, + "loss": 0.0411, + "step": 24537 + }, + { + "epoch": 3.21, + "learning_rate": 1.3585305023064813e-05, + "loss": 0.0438, + "step": 24538 + }, + { + "epoch": 3.21, + "learning_rate": 1.3584809950022693e-05, + "loss": 0.0437, + "step": 24539 + }, + { + "epoch": 3.21, + "learning_rate": 1.3584314866898492e-05, + "loss": 0.0349, + "step": 24540 + }, + { + "epoch": 3.21, + "learning_rate": 1.3583819773693591e-05, + "loss": 0.0381, + "step": 24541 + }, + { + "epoch": 3.21, + "learning_rate": 1.358332467040939e-05, + "loss": 0.0388, + "step": 24542 + }, + { + "epoch": 3.21, + "learning_rate": 1.3582829557047282e-05, + "loss": 0.0439, + "step": 24543 + }, + { + "epoch": 3.22, + "learning_rate": 1.3582334433608653e-05, + "loss": 0.0385, + "step": 24544 + }, + { + "epoch": 3.22, + "learning_rate": 1.3581839300094899e-05, + "loss": 0.0429, + "step": 24545 + }, + { + "epoch": 3.22, + "learning_rate": 1.3581344156507415e-05, + "loss": 0.0417, + "step": 24546 + }, + { + "epoch": 3.22, + "learning_rate": 1.3580849002847592e-05, + "loss": 0.042, + "step": 24547 + }, + { + "epoch": 3.22, + "learning_rate": 1.3580353839116817e-05, + "loss": 0.0429, + "step": 24548 + }, + { + "epoch": 3.22, + "learning_rate": 1.3579858665316494e-05, + "loss": 0.045, + "step": 24549 + }, + { + "epoch": 3.22, + "learning_rate": 1.3579363481448002e-05, + "loss": 0.0476, + "step": 24550 + }, + { + "epoch": 3.22, + "learning_rate": 1.3578868287512746e-05, + "loss": 0.0433, + "step": 24551 + }, + { + "epoch": 3.22, + "learning_rate": 1.3578373083512114e-05, + "loss": 0.0408, + "step": 24552 + }, + { + "epoch": 3.22, + "learning_rate": 1.3577877869447495e-05, + "loss": 0.0455, + "step": 24553 + }, + { + "epoch": 3.22, + "learning_rate": 1.3577382645320288e-05, + "loss": 0.0447, + "step": 24554 + }, + { + "epoch": 3.22, + "learning_rate": 1.3576887411131878e-05, + "loss": 0.046, + "step": 24555 + }, + { + "epoch": 3.22, + "learning_rate": 1.3576392166883667e-05, + "loss": 0.0421, + "step": 24556 + }, + { + "epoch": 3.22, + "learning_rate": 1.3575896912577042e-05, + "loss": 0.0415, + "step": 24557 + }, + { + "epoch": 3.22, + "learning_rate": 1.3575401648213398e-05, + "loss": 0.045, + "step": 24558 + }, + { + "epoch": 3.22, + "learning_rate": 1.3574906373794126e-05, + "loss": 0.0407, + "step": 24559 + }, + { + "epoch": 3.22, + "learning_rate": 1.3574411089320622e-05, + "loss": 0.0422, + "step": 24560 + }, + { + "epoch": 3.22, + "learning_rate": 1.3573915794794277e-05, + "loss": 0.0462, + "step": 24561 + }, + { + "epoch": 3.22, + "learning_rate": 1.3573420490216482e-05, + "loss": 0.0405, + "step": 24562 + }, + { + "epoch": 3.22, + "learning_rate": 1.3572925175588634e-05, + "loss": 0.0407, + "step": 24563 + }, + { + "epoch": 3.22, + "learning_rate": 1.3572429850912126e-05, + "loss": 0.0425, + "step": 24564 + }, + { + "epoch": 3.22, + "learning_rate": 1.3571934516188345e-05, + "loss": 0.0408, + "step": 24565 + }, + { + "epoch": 3.22, + "learning_rate": 1.3571439171418693e-05, + "loss": 0.0395, + "step": 24566 + }, + { + "epoch": 3.22, + "learning_rate": 1.3570943816604554e-05, + "loss": 0.0414, + "step": 24567 + }, + { + "epoch": 3.22, + "learning_rate": 1.3570448451747327e-05, + "loss": 0.0438, + "step": 24568 + }, + { + "epoch": 3.22, + "learning_rate": 1.3569953076848403e-05, + "loss": 0.0427, + "step": 24569 + }, + { + "epoch": 3.22, + "learning_rate": 1.356945769190918e-05, + "loss": 0.0427, + "step": 24570 + }, + { + "epoch": 3.22, + "learning_rate": 1.3568962296931045e-05, + "loss": 0.0475, + "step": 24571 + }, + { + "epoch": 3.22, + "learning_rate": 1.3568466891915393e-05, + "loss": 0.0416, + "step": 24572 + }, + { + "epoch": 3.22, + "learning_rate": 1.3567971476863617e-05, + "loss": 0.0398, + "step": 24573 + }, + { + "epoch": 3.22, + "learning_rate": 1.3567476051777116e-05, + "loss": 0.0395, + "step": 24574 + }, + { + "epoch": 3.22, + "learning_rate": 1.3566980616657275e-05, + "loss": 0.0395, + "step": 24575 + }, + { + "epoch": 3.22, + "learning_rate": 1.356648517150549e-05, + "loss": 0.0441, + "step": 24576 + }, + { + "epoch": 3.22, + "learning_rate": 1.3565989716323158e-05, + "loss": 0.037, + "step": 24577 + }, + { + "epoch": 3.22, + "learning_rate": 1.3565494251111666e-05, + "loss": 0.0433, + "step": 24578 + }, + { + "epoch": 3.22, + "learning_rate": 1.3564998775872413e-05, + "loss": 0.0424, + "step": 24579 + }, + { + "epoch": 3.22, + "learning_rate": 1.3564503290606793e-05, + "loss": 0.0364, + "step": 24580 + }, + { + "epoch": 3.22, + "learning_rate": 1.3564007795316197e-05, + "loss": 0.0453, + "step": 24581 + }, + { + "epoch": 3.22, + "learning_rate": 1.3563512290002017e-05, + "loss": 0.0498, + "step": 24582 + }, + { + "epoch": 3.22, + "learning_rate": 1.356301677466565e-05, + "loss": 0.0482, + "step": 24583 + }, + { + "epoch": 3.22, + "learning_rate": 1.3562521249308485e-05, + "loss": 0.0341, + "step": 24584 + }, + { + "epoch": 3.22, + "learning_rate": 1.3562025713931921e-05, + "loss": 0.0407, + "step": 24585 + }, + { + "epoch": 3.22, + "learning_rate": 1.3561530168537348e-05, + "loss": 0.0372, + "step": 24586 + }, + { + "epoch": 3.22, + "learning_rate": 1.3561034613126165e-05, + "loss": 0.0343, + "step": 24587 + }, + { + "epoch": 3.22, + "learning_rate": 1.356053904769976e-05, + "loss": 0.0448, + "step": 24588 + }, + { + "epoch": 3.22, + "learning_rate": 1.3560043472259525e-05, + "loss": 0.043, + "step": 24589 + }, + { + "epoch": 3.22, + "learning_rate": 1.3559547886806861e-05, + "loss": 0.0444, + "step": 24590 + }, + { + "epoch": 3.22, + "learning_rate": 1.3559052291343158e-05, + "loss": 0.0415, + "step": 24591 + }, + { + "epoch": 3.22, + "learning_rate": 1.3558556685869808e-05, + "loss": 0.0378, + "step": 24592 + }, + { + "epoch": 3.22, + "learning_rate": 1.3558061070388209e-05, + "loss": 0.0485, + "step": 24593 + }, + { + "epoch": 3.22, + "learning_rate": 1.3557565444899749e-05, + "loss": 0.0371, + "step": 24594 + }, + { + "epoch": 3.22, + "learning_rate": 1.3557069809405832e-05, + "loss": 0.0402, + "step": 24595 + }, + { + "epoch": 3.22, + "learning_rate": 1.355657416390784e-05, + "loss": 0.0503, + "step": 24596 + }, + { + "epoch": 3.22, + "learning_rate": 1.3556078508407175e-05, + "loss": 0.0438, + "step": 24597 + }, + { + "epoch": 3.22, + "learning_rate": 1.3555582842905229e-05, + "loss": 0.0527, + "step": 24598 + }, + { + "epoch": 3.22, + "learning_rate": 1.3555087167403393e-05, + "loss": 0.0322, + "step": 24599 + }, + { + "epoch": 3.22, + "learning_rate": 1.3554591481903066e-05, + "loss": 0.0404, + "step": 24600 + }, + { + "epoch": 3.22, + "learning_rate": 1.3554095786405638e-05, + "loss": 0.038, + "step": 24601 + }, + { + "epoch": 3.22, + "learning_rate": 1.3553600080912506e-05, + "loss": 0.0373, + "step": 24602 + }, + { + "epoch": 3.22, + "learning_rate": 1.3553104365425064e-05, + "loss": 0.0465, + "step": 24603 + }, + { + "epoch": 3.22, + "learning_rate": 1.3552608639944704e-05, + "loss": 0.0401, + "step": 24604 + }, + { + "epoch": 3.22, + "learning_rate": 1.3552112904472818e-05, + "loss": 0.0434, + "step": 24605 + }, + { + "epoch": 3.22, + "learning_rate": 1.355161715901081e-05, + "loss": 0.0382, + "step": 24606 + }, + { + "epoch": 3.22, + "learning_rate": 1.3551121403560063e-05, + "loss": 0.0379, + "step": 24607 + }, + { + "epoch": 3.22, + "learning_rate": 1.3550625638121982e-05, + "loss": 0.0428, + "step": 24608 + }, + { + "epoch": 3.22, + "learning_rate": 1.355012986269795e-05, + "loss": 0.0459, + "step": 24609 + }, + { + "epoch": 3.22, + "learning_rate": 1.3549634077289366e-05, + "loss": 0.0342, + "step": 24610 + }, + { + "epoch": 3.22, + "learning_rate": 1.354913828189763e-05, + "loss": 0.035, + "step": 24611 + }, + { + "epoch": 3.22, + "learning_rate": 1.354864247652413e-05, + "loss": 0.0404, + "step": 24612 + }, + { + "epoch": 3.22, + "learning_rate": 1.3548146661170258e-05, + "loss": 0.0412, + "step": 24613 + }, + { + "epoch": 3.22, + "learning_rate": 1.3547650835837419e-05, + "loss": 0.0447, + "step": 24614 + }, + { + "epoch": 3.22, + "learning_rate": 1.3547155000526997e-05, + "loss": 0.0449, + "step": 24615 + }, + { + "epoch": 3.22, + "learning_rate": 1.3546659155240388e-05, + "loss": 0.0395, + "step": 24616 + }, + { + "epoch": 3.22, + "learning_rate": 1.3546163299978993e-05, + "loss": 0.039, + "step": 24617 + }, + { + "epoch": 3.22, + "learning_rate": 1.3545667434744203e-05, + "loss": 0.0439, + "step": 24618 + }, + { + "epoch": 3.22, + "learning_rate": 1.3545171559537411e-05, + "loss": 0.0402, + "step": 24619 + }, + { + "epoch": 3.23, + "learning_rate": 1.354467567436001e-05, + "loss": 0.0443, + "step": 24620 + }, + { + "epoch": 3.23, + "learning_rate": 1.3544179779213402e-05, + "loss": 0.0411, + "step": 24621 + }, + { + "epoch": 3.23, + "learning_rate": 1.3543683874098976e-05, + "loss": 0.0371, + "step": 24622 + }, + { + "epoch": 3.23, + "learning_rate": 1.3543187959018127e-05, + "loss": 0.0438, + "step": 24623 + }, + { + "epoch": 3.23, + "learning_rate": 1.3542692033972248e-05, + "loss": 0.04, + "step": 24624 + }, + { + "epoch": 3.23, + "learning_rate": 1.3542196098962742e-05, + "loss": 0.0419, + "step": 24625 + }, + { + "epoch": 3.23, + "learning_rate": 1.3541700153990992e-05, + "loss": 0.0374, + "step": 24626 + }, + { + "epoch": 3.23, + "learning_rate": 1.3541204199058402e-05, + "loss": 0.0447, + "step": 24627 + }, + { + "epoch": 3.23, + "learning_rate": 1.3540708234166364e-05, + "loss": 0.0354, + "step": 24628 + }, + { + "epoch": 3.23, + "learning_rate": 1.3540212259316273e-05, + "loss": 0.0455, + "step": 24629 + }, + { + "epoch": 3.23, + "learning_rate": 1.3539716274509523e-05, + "loss": 0.0411, + "step": 24630 + }, + { + "epoch": 3.23, + "learning_rate": 1.3539220279747509e-05, + "loss": 0.0495, + "step": 24631 + }, + { + "epoch": 3.23, + "learning_rate": 1.3538724275031625e-05, + "loss": 0.048, + "step": 24632 + }, + { + "epoch": 3.23, + "learning_rate": 1.3538228260363271e-05, + "loss": 0.0471, + "step": 24633 + }, + { + "epoch": 3.23, + "learning_rate": 1.3537732235743835e-05, + "loss": 0.0492, + "step": 24634 + }, + { + "epoch": 3.23, + "learning_rate": 1.353723620117472e-05, + "loss": 0.0359, + "step": 24635 + }, + { + "epoch": 3.23, + "learning_rate": 1.353674015665731e-05, + "loss": 0.0349, + "step": 24636 + }, + { + "epoch": 3.23, + "learning_rate": 1.3536244102193012e-05, + "loss": 0.0435, + "step": 24637 + }, + { + "epoch": 3.23, + "learning_rate": 1.3535748037783214e-05, + "loss": 0.0417, + "step": 24638 + }, + { + "epoch": 3.23, + "learning_rate": 1.3535251963429313e-05, + "loss": 0.0406, + "step": 24639 + }, + { + "epoch": 3.23, + "learning_rate": 1.3534755879132704e-05, + "loss": 0.0429, + "step": 24640 + }, + { + "epoch": 3.23, + "learning_rate": 1.3534259784894783e-05, + "loss": 0.0389, + "step": 24641 + }, + { + "epoch": 3.23, + "learning_rate": 1.3533763680716945e-05, + "loss": 0.0354, + "step": 24642 + }, + { + "epoch": 3.23, + "learning_rate": 1.3533267566600584e-05, + "loss": 0.0457, + "step": 24643 + }, + { + "epoch": 3.23, + "learning_rate": 1.3532771442547092e-05, + "loss": 0.0477, + "step": 24644 + }, + { + "epoch": 3.23, + "learning_rate": 1.3532275308557874e-05, + "loss": 0.0398, + "step": 24645 + }, + { + "epoch": 3.23, + "learning_rate": 1.3531779164634318e-05, + "loss": 0.0395, + "step": 24646 + }, + { + "epoch": 3.23, + "learning_rate": 1.3531283010777822e-05, + "loss": 0.042, + "step": 24647 + }, + { + "epoch": 3.23, + "learning_rate": 1.353078684698978e-05, + "loss": 0.0384, + "step": 24648 + }, + { + "epoch": 3.23, + "learning_rate": 1.3530290673271586e-05, + "loss": 0.0379, + "step": 24649 + }, + { + "epoch": 3.23, + "learning_rate": 1.3529794489624641e-05, + "loss": 0.0389, + "step": 24650 + }, + { + "epoch": 3.23, + "learning_rate": 1.3529298296050335e-05, + "loss": 0.0442, + "step": 24651 + }, + { + "epoch": 3.23, + "learning_rate": 1.3528802092550066e-05, + "loss": 0.0403, + "step": 24652 + }, + { + "epoch": 3.23, + "learning_rate": 1.3528305879125229e-05, + "loss": 0.0417, + "step": 24653 + }, + { + "epoch": 3.23, + "learning_rate": 1.352780965577722e-05, + "loss": 0.0356, + "step": 24654 + }, + { + "epoch": 3.23, + "learning_rate": 1.3527313422507433e-05, + "loss": 0.0428, + "step": 24655 + }, + { + "epoch": 3.23, + "learning_rate": 1.3526817179317267e-05, + "loss": 0.0359, + "step": 24656 + }, + { + "epoch": 3.23, + "learning_rate": 1.3526320926208115e-05, + "loss": 0.0456, + "step": 24657 + }, + { + "epoch": 3.23, + "learning_rate": 1.352582466318137e-05, + "loss": 0.0438, + "step": 24658 + }, + { + "epoch": 3.23, + "learning_rate": 1.3525328390238435e-05, + "loss": 0.0412, + "step": 24659 + }, + { + "epoch": 3.23, + "learning_rate": 1.3524832107380698e-05, + "loss": 0.042, + "step": 24660 + }, + { + "epoch": 3.23, + "learning_rate": 1.3524335814609559e-05, + "loss": 0.0397, + "step": 24661 + }, + { + "epoch": 3.23, + "learning_rate": 1.3523839511926418e-05, + "loss": 0.0438, + "step": 24662 + }, + { + "epoch": 3.23, + "learning_rate": 1.352334319933266e-05, + "loss": 0.044, + "step": 24663 + }, + { + "epoch": 3.23, + "learning_rate": 1.3522846876829689e-05, + "loss": 0.0352, + "step": 24664 + }, + { + "epoch": 3.23, + "learning_rate": 1.3522350544418898e-05, + "loss": 0.0389, + "step": 24665 + }, + { + "epoch": 3.23, + "learning_rate": 1.3521854202101688e-05, + "loss": 0.0454, + "step": 24666 + }, + { + "epoch": 3.23, + "learning_rate": 1.3521357849879447e-05, + "loss": 0.0384, + "step": 24667 + }, + { + "epoch": 3.23, + "learning_rate": 1.3520861487753573e-05, + "loss": 0.0384, + "step": 24668 + }, + { + "epoch": 3.23, + "learning_rate": 1.3520365115725466e-05, + "loss": 0.0434, + "step": 24669 + }, + { + "epoch": 3.23, + "learning_rate": 1.3519868733796518e-05, + "loss": 0.0478, + "step": 24670 + }, + { + "epoch": 3.23, + "learning_rate": 1.3519372341968129e-05, + "loss": 0.0456, + "step": 24671 + }, + { + "epoch": 3.23, + "learning_rate": 1.351887594024169e-05, + "loss": 0.0377, + "step": 24672 + }, + { + "epoch": 3.23, + "learning_rate": 1.3518379528618604e-05, + "loss": 0.0345, + "step": 24673 + }, + { + "epoch": 3.23, + "learning_rate": 1.3517883107100259e-05, + "loss": 0.0421, + "step": 24674 + }, + { + "epoch": 3.23, + "learning_rate": 1.3517386675688057e-05, + "loss": 0.0431, + "step": 24675 + }, + { + "epoch": 3.23, + "learning_rate": 1.3516890234383388e-05, + "loss": 0.0425, + "step": 24676 + }, + { + "epoch": 3.23, + "learning_rate": 1.3516393783187658e-05, + "loss": 0.0477, + "step": 24677 + }, + { + "epoch": 3.23, + "learning_rate": 1.3515897322102256e-05, + "loss": 0.0494, + "step": 24678 + }, + { + "epoch": 3.23, + "learning_rate": 1.3515400851128581e-05, + "loss": 0.0368, + "step": 24679 + }, + { + "epoch": 3.23, + "learning_rate": 1.3514904370268024e-05, + "loss": 0.0385, + "step": 24680 + }, + { + "epoch": 3.23, + "learning_rate": 1.3514407879521988e-05, + "loss": 0.033, + "step": 24681 + }, + { + "epoch": 3.23, + "learning_rate": 1.3513911378891869e-05, + "loss": 0.0416, + "step": 24682 + }, + { + "epoch": 3.23, + "learning_rate": 1.3513414868379062e-05, + "loss": 0.038, + "step": 24683 + }, + { + "epoch": 3.23, + "learning_rate": 1.351291834798496e-05, + "loss": 0.0423, + "step": 24684 + }, + { + "epoch": 3.23, + "learning_rate": 1.3512421817710964e-05, + "loss": 0.0386, + "step": 24685 + }, + { + "epoch": 3.23, + "learning_rate": 1.3511925277558469e-05, + "loss": 0.0411, + "step": 24686 + }, + { + "epoch": 3.23, + "learning_rate": 1.3511428727528868e-05, + "loss": 0.0375, + "step": 24687 + }, + { + "epoch": 3.23, + "learning_rate": 1.3510932167623562e-05, + "loss": 0.031, + "step": 24688 + }, + { + "epoch": 3.23, + "learning_rate": 1.3510435597843948e-05, + "loss": 0.0492, + "step": 24689 + }, + { + "epoch": 3.23, + "learning_rate": 1.3509939018191421e-05, + "loss": 0.0443, + "step": 24690 + }, + { + "epoch": 3.23, + "learning_rate": 1.3509442428667374e-05, + "loss": 0.0411, + "step": 24691 + }, + { + "epoch": 3.23, + "learning_rate": 1.350894582927321e-05, + "loss": 0.0399, + "step": 24692 + }, + { + "epoch": 3.23, + "learning_rate": 1.350844922001032e-05, + "loss": 0.0368, + "step": 24693 + }, + { + "epoch": 3.23, + "learning_rate": 1.3507952600880108e-05, + "loss": 0.0408, + "step": 24694 + }, + { + "epoch": 3.23, + "learning_rate": 1.3507455971883963e-05, + "loss": 0.0415, + "step": 24695 + }, + { + "epoch": 3.24, + "learning_rate": 1.3506959333023286e-05, + "loss": 0.0403, + "step": 24696 + }, + { + "epoch": 3.24, + "learning_rate": 1.3506462684299471e-05, + "loss": 0.0421, + "step": 24697 + }, + { + "epoch": 3.24, + "learning_rate": 1.3505966025713917e-05, + "loss": 0.0406, + "step": 24698 + }, + { + "epoch": 3.24, + "learning_rate": 1.350546935726802e-05, + "loss": 0.045, + "step": 24699 + }, + { + "epoch": 3.24, + "learning_rate": 1.350497267896318e-05, + "loss": 0.0415, + "step": 24700 + }, + { + "epoch": 3.24, + "learning_rate": 1.3504475990800787e-05, + "loss": 0.0449, + "step": 24701 + }, + { + "epoch": 3.24, + "learning_rate": 1.3503979292782245e-05, + "loss": 0.0459, + "step": 24702 + }, + { + "epoch": 3.24, + "learning_rate": 1.3503482584908945e-05, + "loss": 0.037, + "step": 24703 + }, + { + "epoch": 3.24, + "learning_rate": 1.3502985867182287e-05, + "loss": 0.0456, + "step": 24704 + }, + { + "epoch": 3.24, + "learning_rate": 1.3502489139603672e-05, + "loss": 0.0337, + "step": 24705 + }, + { + "epoch": 3.24, + "learning_rate": 1.3501992402174486e-05, + "loss": 0.0319, + "step": 24706 + }, + { + "epoch": 3.24, + "learning_rate": 1.350149565489614e-05, + "loss": 0.0357, + "step": 24707 + }, + { + "epoch": 3.24, + "learning_rate": 1.3500998897770022e-05, + "loss": 0.0415, + "step": 24708 + }, + { + "epoch": 3.24, + "learning_rate": 1.3500502130797526e-05, + "loss": 0.0371, + "step": 24709 + }, + { + "epoch": 3.24, + "learning_rate": 1.350000535398006e-05, + "loss": 0.0416, + "step": 24710 + }, + { + "epoch": 3.24, + "learning_rate": 1.3499508567319012e-05, + "loss": 0.0447, + "step": 24711 + }, + { + "epoch": 3.24, + "learning_rate": 1.3499011770815783e-05, + "loss": 0.0417, + "step": 24712 + }, + { + "epoch": 3.24, + "learning_rate": 1.3498514964471775e-05, + "loss": 0.0487, + "step": 24713 + }, + { + "epoch": 3.24, + "learning_rate": 1.3498018148288372e-05, + "loss": 0.037, + "step": 24714 + }, + { + "epoch": 3.24, + "learning_rate": 1.3497521322266981e-05, + "loss": 0.0355, + "step": 24715 + }, + { + "epoch": 3.24, + "learning_rate": 1.3497024486409e-05, + "loss": 0.0356, + "step": 24716 + }, + { + "epoch": 3.24, + "learning_rate": 1.3496527640715824e-05, + "loss": 0.0433, + "step": 24717 + }, + { + "epoch": 3.24, + "learning_rate": 1.3496030785188849e-05, + "loss": 0.047, + "step": 24718 + }, + { + "epoch": 3.24, + "learning_rate": 1.3495533919829476e-05, + "loss": 0.0375, + "step": 24719 + }, + { + "epoch": 3.24, + "learning_rate": 1.3495037044639094e-05, + "loss": 0.0451, + "step": 24720 + }, + { + "epoch": 3.24, + "learning_rate": 1.3494540159619112e-05, + "loss": 0.0394, + "step": 24721 + }, + { + "epoch": 3.24, + "learning_rate": 1.349404326477092e-05, + "loss": 0.0335, + "step": 24722 + }, + { + "epoch": 3.24, + "learning_rate": 1.349354636009592e-05, + "loss": 0.0381, + "step": 24723 + }, + { + "epoch": 3.24, + "learning_rate": 1.3493049445595502e-05, + "loss": 0.0467, + "step": 24724 + }, + { + "epoch": 3.24, + "learning_rate": 1.3492552521271071e-05, + "loss": 0.0492, + "step": 24725 + }, + { + "epoch": 3.24, + "learning_rate": 1.3492055587124022e-05, + "loss": 0.037, + "step": 24726 + }, + { + "epoch": 3.24, + "learning_rate": 1.3491558643155754e-05, + "loss": 0.0454, + "step": 24727 + }, + { + "epoch": 3.24, + "learning_rate": 1.3491061689367661e-05, + "loss": 0.0394, + "step": 24728 + }, + { + "epoch": 3.24, + "learning_rate": 1.3490564725761144e-05, + "loss": 0.0466, + "step": 24729 + }, + { + "epoch": 3.24, + "learning_rate": 1.3490067752337599e-05, + "loss": 0.0494, + "step": 24730 + }, + { + "epoch": 3.24, + "learning_rate": 1.3489570769098423e-05, + "loss": 0.0366, + "step": 24731 + }, + { + "epoch": 3.24, + "learning_rate": 1.3489073776045019e-05, + "loss": 0.0419, + "step": 24732 + }, + { + "epoch": 3.24, + "learning_rate": 1.3488576773178777e-05, + "loss": 0.0427, + "step": 24733 + }, + { + "epoch": 3.24, + "learning_rate": 1.3488079760501103e-05, + "loss": 0.036, + "step": 24734 + }, + { + "epoch": 3.24, + "learning_rate": 1.3487582738013387e-05, + "loss": 0.0521, + "step": 24735 + }, + { + "epoch": 3.24, + "learning_rate": 1.3487085705717029e-05, + "loss": 0.0386, + "step": 24736 + }, + { + "epoch": 3.24, + "learning_rate": 1.3486588663613429e-05, + "loss": 0.0454, + "step": 24737 + }, + { + "epoch": 3.24, + "learning_rate": 1.3486091611703986e-05, + "loss": 0.039, + "step": 24738 + }, + { + "epoch": 3.24, + "learning_rate": 1.3485594549990096e-05, + "loss": 0.0385, + "step": 24739 + }, + { + "epoch": 3.24, + "learning_rate": 1.3485097478473156e-05, + "loss": 0.0381, + "step": 24740 + }, + { + "epoch": 3.24, + "learning_rate": 1.3484600397154563e-05, + "loss": 0.042, + "step": 24741 + }, + { + "epoch": 3.24, + "learning_rate": 1.3484103306035721e-05, + "loss": 0.0428, + "step": 24742 + }, + { + "epoch": 3.24, + "learning_rate": 1.348360620511802e-05, + "loss": 0.0444, + "step": 24743 + }, + { + "epoch": 3.24, + "learning_rate": 1.3483109094402867e-05, + "loss": 0.0411, + "step": 24744 + }, + { + "epoch": 3.24, + "learning_rate": 1.348261197389165e-05, + "loss": 0.0402, + "step": 24745 + }, + { + "epoch": 3.24, + "learning_rate": 1.3482114843585775e-05, + "loss": 0.0361, + "step": 24746 + }, + { + "epoch": 3.24, + "learning_rate": 1.3481617703486636e-05, + "loss": 0.0306, + "step": 24747 + }, + { + "epoch": 3.24, + "learning_rate": 1.3481120553595635e-05, + "loss": 0.0348, + "step": 24748 + }, + { + "epoch": 3.24, + "learning_rate": 1.3480623393914165e-05, + "loss": 0.0428, + "step": 24749 + }, + { + "epoch": 3.24, + "learning_rate": 1.3480126224443625e-05, + "loss": 0.0422, + "step": 24750 + }, + { + "epoch": 3.24, + "learning_rate": 1.3479629045185423e-05, + "loss": 0.0418, + "step": 24751 + }, + { + "epoch": 3.24, + "learning_rate": 1.3479131856140943e-05, + "loss": 0.0378, + "step": 24752 + }, + { + "epoch": 3.24, + "learning_rate": 1.3478634657311594e-05, + "loss": 0.0454, + "step": 24753 + }, + { + "epoch": 3.24, + "learning_rate": 1.3478137448698765e-05, + "loss": 0.0409, + "step": 24754 + }, + { + "epoch": 3.24, + "learning_rate": 1.3477640230303866e-05, + "loss": 0.0435, + "step": 24755 + }, + { + "epoch": 3.24, + "learning_rate": 1.3477143002128285e-05, + "loss": 0.0404, + "step": 24756 + }, + { + "epoch": 3.24, + "learning_rate": 1.3476645764173429e-05, + "loss": 0.0387, + "step": 24757 + }, + { + "epoch": 3.24, + "learning_rate": 1.3476148516440683e-05, + "loss": 0.0364, + "step": 24758 + }, + { + "epoch": 3.24, + "learning_rate": 1.3475651258931461e-05, + "loss": 0.0417, + "step": 24759 + }, + { + "epoch": 3.24, + "learning_rate": 1.3475153991647154e-05, + "loss": 0.0356, + "step": 24760 + }, + { + "epoch": 3.24, + "learning_rate": 1.3474656714589162e-05, + "loss": 0.0457, + "step": 24761 + }, + { + "epoch": 3.24, + "learning_rate": 1.3474159427758884e-05, + "loss": 0.0436, + "step": 24762 + }, + { + "epoch": 3.24, + "learning_rate": 1.3473662131157714e-05, + "loss": 0.0409, + "step": 24763 + }, + { + "epoch": 3.24, + "learning_rate": 1.3473164824787055e-05, + "loss": 0.0381, + "step": 24764 + }, + { + "epoch": 3.24, + "learning_rate": 1.3472667508648306e-05, + "loss": 0.045, + "step": 24765 + }, + { + "epoch": 3.24, + "learning_rate": 1.3472170182742868e-05, + "loss": 0.0426, + "step": 24766 + }, + { + "epoch": 3.24, + "learning_rate": 1.3471672847072133e-05, + "loss": 0.0388, + "step": 24767 + }, + { + "epoch": 3.24, + "learning_rate": 1.3471175501637501e-05, + "loss": 0.0395, + "step": 24768 + }, + { + "epoch": 3.24, + "learning_rate": 1.3470678146440375e-05, + "loss": 0.0393, + "step": 24769 + }, + { + "epoch": 3.24, + "learning_rate": 1.3470180781482149e-05, + "loss": 0.0526, + "step": 24770 + }, + { + "epoch": 3.24, + "learning_rate": 1.3469683406764226e-05, + "loss": 0.0346, + "step": 24771 + }, + { + "epoch": 3.24, + "learning_rate": 1.3469186022288008e-05, + "loss": 0.0434, + "step": 24772 + }, + { + "epoch": 3.25, + "learning_rate": 1.3468688628054882e-05, + "loss": 0.036, + "step": 24773 + }, + { + "epoch": 3.25, + "learning_rate": 1.3468191224066255e-05, + "loss": 0.0429, + "step": 24774 + }, + { + "epoch": 3.25, + "learning_rate": 1.3467693810323528e-05, + "loss": 0.0448, + "step": 24775 + }, + { + "epoch": 3.25, + "learning_rate": 1.3467196386828096e-05, + "loss": 0.042, + "step": 24776 + }, + { + "epoch": 3.25, + "learning_rate": 1.3466698953581358e-05, + "loss": 0.0381, + "step": 24777 + }, + { + "epoch": 3.25, + "learning_rate": 1.3466201510584716e-05, + "loss": 0.0374, + "step": 24778 + }, + { + "epoch": 3.25, + "learning_rate": 1.3465704057839563e-05, + "loss": 0.0403, + "step": 24779 + }, + { + "epoch": 3.25, + "learning_rate": 1.3465206595347304e-05, + "loss": 0.044, + "step": 24780 + }, + { + "epoch": 3.25, + "learning_rate": 1.3464709123109337e-05, + "loss": 0.0341, + "step": 24781 + }, + { + "epoch": 3.25, + "learning_rate": 1.3464211641127058e-05, + "loss": 0.0351, + "step": 24782 + }, + { + "epoch": 3.25, + "learning_rate": 1.346371414940187e-05, + "loss": 0.0363, + "step": 24783 + }, + { + "epoch": 3.25, + "learning_rate": 1.3463216647935168e-05, + "loss": 0.0414, + "step": 24784 + }, + { + "epoch": 3.25, + "learning_rate": 1.3462719136728354e-05, + "loss": 0.044, + "step": 24785 + }, + { + "epoch": 3.25, + "learning_rate": 1.346222161578283e-05, + "loss": 0.0337, + "step": 24786 + }, + { + "epoch": 3.25, + "learning_rate": 1.346172408509999e-05, + "loss": 0.0449, + "step": 24787 + }, + { + "epoch": 3.25, + "learning_rate": 1.3461226544681237e-05, + "loss": 0.0356, + "step": 24788 + }, + { + "epoch": 3.25, + "learning_rate": 1.3460728994527965e-05, + "loss": 0.0376, + "step": 24789 + }, + { + "epoch": 3.25, + "learning_rate": 1.3460231434641577e-05, + "loss": 0.0398, + "step": 24790 + }, + { + "epoch": 3.25, + "learning_rate": 1.3459733865023474e-05, + "loss": 0.0404, + "step": 24791 + }, + { + "epoch": 3.25, + "learning_rate": 1.3459236285675057e-05, + "loss": 0.0449, + "step": 24792 + }, + { + "epoch": 3.25, + "learning_rate": 1.345873869659772e-05, + "loss": 0.0388, + "step": 24793 + }, + { + "epoch": 3.25, + "learning_rate": 1.345824109779286e-05, + "loss": 0.0376, + "step": 24794 + }, + { + "epoch": 3.25, + "learning_rate": 1.345774348926189e-05, + "loss": 0.0412, + "step": 24795 + }, + { + "epoch": 3.25, + "learning_rate": 1.345724587100619e-05, + "loss": 0.0394, + "step": 24796 + }, + { + "epoch": 3.25, + "learning_rate": 1.3456748243027177e-05, + "loss": 0.0355, + "step": 24797 + }, + { + "epoch": 3.25, + "learning_rate": 1.3456250605326245e-05, + "loss": 0.0332, + "step": 24798 + }, + { + "epoch": 3.25, + "learning_rate": 1.3455752957904789e-05, + "loss": 0.0393, + "step": 24799 + }, + { + "epoch": 3.25, + "learning_rate": 1.3455255300764212e-05, + "loss": 0.0398, + "step": 24800 + }, + { + "epoch": 3.25, + "learning_rate": 1.3454757633905913e-05, + "loss": 0.0437, + "step": 24801 + }, + { + "epoch": 3.25, + "learning_rate": 1.3454259957331293e-05, + "loss": 0.0448, + "step": 24802 + }, + { + "epoch": 3.25, + "learning_rate": 1.345376227104175e-05, + "loss": 0.0415, + "step": 24803 + }, + { + "epoch": 3.25, + "learning_rate": 1.3453264575038688e-05, + "loss": 0.0381, + "step": 24804 + }, + { + "epoch": 3.25, + "learning_rate": 1.34527668693235e-05, + "loss": 0.0404, + "step": 24805 + }, + { + "epoch": 3.25, + "learning_rate": 1.345226915389759e-05, + "loss": 0.04, + "step": 24806 + }, + { + "epoch": 3.25, + "learning_rate": 1.3451771428762355e-05, + "loss": 0.0466, + "step": 24807 + }, + { + "epoch": 3.25, + "learning_rate": 1.3451273693919198e-05, + "loss": 0.0389, + "step": 24808 + }, + { + "epoch": 3.25, + "learning_rate": 1.345077594936952e-05, + "loss": 0.0457, + "step": 24809 + }, + { + "epoch": 3.25, + "learning_rate": 1.3450278195114714e-05, + "loss": 0.0398, + "step": 24810 + }, + { + "epoch": 3.25, + "learning_rate": 1.3449780431156185e-05, + "loss": 0.0445, + "step": 24811 + }, + { + "epoch": 3.25, + "learning_rate": 1.3449282657495335e-05, + "loss": 0.0416, + "step": 24812 + }, + { + "epoch": 3.25, + "learning_rate": 1.344878487413356e-05, + "loss": 0.0395, + "step": 24813 + }, + { + "epoch": 3.25, + "learning_rate": 1.344828708107226e-05, + "loss": 0.0409, + "step": 24814 + }, + { + "epoch": 3.25, + "learning_rate": 1.3447789278312838e-05, + "loss": 0.0393, + "step": 24815 + }, + { + "epoch": 3.25, + "learning_rate": 1.3447291465856694e-05, + "loss": 0.0419, + "step": 24816 + }, + { + "epoch": 3.25, + "learning_rate": 1.3446793643705223e-05, + "loss": 0.0472, + "step": 24817 + }, + { + "epoch": 3.25, + "learning_rate": 1.3446295811859828e-05, + "loss": 0.0346, + "step": 24818 + }, + { + "epoch": 3.25, + "learning_rate": 1.344579797032191e-05, + "loss": 0.0412, + "step": 24819 + }, + { + "epoch": 3.25, + "learning_rate": 1.3445300119092873e-05, + "loss": 0.0485, + "step": 24820 + }, + { + "epoch": 3.25, + "learning_rate": 1.3444802258174105e-05, + "loss": 0.0425, + "step": 24821 + }, + { + "epoch": 3.25, + "learning_rate": 1.3444304387567022e-05, + "loss": 0.0379, + "step": 24822 + }, + { + "epoch": 3.25, + "learning_rate": 1.3443806507273009e-05, + "loss": 0.0456, + "step": 24823 + }, + { + "epoch": 3.25, + "learning_rate": 1.344330861729348e-05, + "loss": 0.041, + "step": 24824 + }, + { + "epoch": 3.25, + "learning_rate": 1.3442810717629827e-05, + "loss": 0.038, + "step": 24825 + }, + { + "epoch": 3.25, + "learning_rate": 1.3442312808283453e-05, + "loss": 0.0429, + "step": 24826 + }, + { + "epoch": 3.25, + "learning_rate": 1.3441814889255755e-05, + "loss": 0.0355, + "step": 24827 + }, + { + "epoch": 3.25, + "learning_rate": 1.3441316960548139e-05, + "loss": 0.0446, + "step": 24828 + }, + { + "epoch": 3.25, + "learning_rate": 1.3440819022161997e-05, + "loss": 0.0441, + "step": 24829 + }, + { + "epoch": 3.25, + "learning_rate": 1.3440321074098739e-05, + "loss": 0.0415, + "step": 24830 + }, + { + "epoch": 3.25, + "learning_rate": 1.343982311635976e-05, + "loss": 0.0388, + "step": 24831 + }, + { + "epoch": 3.25, + "learning_rate": 1.3439325148946463e-05, + "loss": 0.0427, + "step": 24832 + }, + { + "epoch": 3.25, + "learning_rate": 1.3438827171860245e-05, + "loss": 0.0391, + "step": 24833 + }, + { + "epoch": 3.25, + "learning_rate": 1.343832918510251e-05, + "loss": 0.0381, + "step": 24834 + }, + { + "epoch": 3.25, + "learning_rate": 1.3437831188674655e-05, + "loss": 0.0382, + "step": 24835 + }, + { + "epoch": 3.25, + "learning_rate": 1.3437333182578085e-05, + "loss": 0.043, + "step": 24836 + }, + { + "epoch": 3.25, + "learning_rate": 1.34368351668142e-05, + "loss": 0.0362, + "step": 24837 + }, + { + "epoch": 3.25, + "learning_rate": 1.3436337141384397e-05, + "loss": 0.0376, + "step": 24838 + }, + { + "epoch": 3.25, + "learning_rate": 1.3435839106290081e-05, + "loss": 0.0449, + "step": 24839 + }, + { + "epoch": 3.25, + "learning_rate": 1.3435341061532645e-05, + "loss": 0.038, + "step": 24840 + }, + { + "epoch": 3.25, + "learning_rate": 1.3434843007113501e-05, + "loss": 0.0459, + "step": 24841 + }, + { + "epoch": 3.25, + "learning_rate": 1.3434344943034042e-05, + "loss": 0.0406, + "step": 24842 + }, + { + "epoch": 3.25, + "learning_rate": 1.3433846869295673e-05, + "loss": 0.0422, + "step": 24843 + }, + { + "epoch": 3.25, + "learning_rate": 1.343334878589979e-05, + "loss": 0.0421, + "step": 24844 + }, + { + "epoch": 3.25, + "learning_rate": 1.3432850692847794e-05, + "loss": 0.04, + "step": 24845 + }, + { + "epoch": 3.25, + "learning_rate": 1.343235259014109e-05, + "loss": 0.0396, + "step": 24846 + }, + { + "epoch": 3.25, + "learning_rate": 1.343185447778108e-05, + "loss": 0.0344, + "step": 24847 + }, + { + "epoch": 3.25, + "learning_rate": 1.3431356355769158e-05, + "loss": 0.0394, + "step": 24848 + }, + { + "epoch": 3.26, + "learning_rate": 1.3430858224106733e-05, + "loss": 0.0449, + "step": 24849 + }, + { + "epoch": 3.26, + "learning_rate": 1.34303600827952e-05, + "loss": 0.0383, + "step": 24850 + }, + { + "epoch": 3.26, + "learning_rate": 1.3429861931835961e-05, + "loss": 0.039, + "step": 24851 + }, + { + "epoch": 3.26, + "learning_rate": 1.342936377123042e-05, + "loss": 0.0357, + "step": 24852 + }, + { + "epoch": 3.26, + "learning_rate": 1.3428865600979975e-05, + "loss": 0.0412, + "step": 24853 + }, + { + "epoch": 3.26, + "learning_rate": 1.3428367421086028e-05, + "loss": 0.043, + "step": 24854 + }, + { + "epoch": 3.26, + "learning_rate": 1.3427869231549979e-05, + "loss": 0.048, + "step": 24855 + }, + { + "epoch": 3.26, + "learning_rate": 1.342737103237323e-05, + "loss": 0.046, + "step": 24856 + }, + { + "epoch": 3.26, + "learning_rate": 1.3426872823557187e-05, + "loss": 0.0433, + "step": 24857 + }, + { + "epoch": 3.26, + "learning_rate": 1.3426374605103242e-05, + "loss": 0.0519, + "step": 24858 + }, + { + "epoch": 3.26, + "learning_rate": 1.3425876377012804e-05, + "loss": 0.0379, + "step": 24859 + }, + { + "epoch": 3.26, + "learning_rate": 1.3425378139287269e-05, + "loss": 0.0387, + "step": 24860 + }, + { + "epoch": 3.26, + "learning_rate": 1.3424879891928038e-05, + "loss": 0.0389, + "step": 24861 + }, + { + "epoch": 3.26, + "learning_rate": 1.342438163493652e-05, + "loss": 0.0381, + "step": 24862 + }, + { + "epoch": 3.26, + "learning_rate": 1.3423883368314107e-05, + "loss": 0.0415, + "step": 24863 + }, + { + "epoch": 3.26, + "learning_rate": 1.3423385092062208e-05, + "loss": 0.0493, + "step": 24864 + }, + { + "epoch": 3.26, + "learning_rate": 1.3422886806182218e-05, + "loss": 0.0469, + "step": 24865 + }, + { + "epoch": 3.26, + "learning_rate": 1.3422388510675543e-05, + "loss": 0.0402, + "step": 24866 + }, + { + "epoch": 3.26, + "learning_rate": 1.342189020554358e-05, + "loss": 0.041, + "step": 24867 + }, + { + "epoch": 3.26, + "learning_rate": 1.3421391890787736e-05, + "loss": 0.0443, + "step": 24868 + }, + { + "epoch": 3.26, + "learning_rate": 1.3420893566409407e-05, + "loss": 0.0449, + "step": 24869 + }, + { + "epoch": 3.26, + "learning_rate": 1.342039523241e-05, + "loss": 0.04, + "step": 24870 + }, + { + "epoch": 3.26, + "learning_rate": 1.3419896888790912e-05, + "loss": 0.0397, + "step": 24871 + }, + { + "epoch": 3.26, + "learning_rate": 1.3419398535553545e-05, + "loss": 0.0422, + "step": 24872 + }, + { + "epoch": 3.26, + "learning_rate": 1.34189001726993e-05, + "loss": 0.0445, + "step": 24873 + }, + { + "epoch": 3.26, + "learning_rate": 1.3418401800229586e-05, + "loss": 0.0432, + "step": 24874 + }, + { + "epoch": 3.26, + "learning_rate": 1.3417903418145794e-05, + "loss": 0.0394, + "step": 24875 + }, + { + "epoch": 3.26, + "learning_rate": 1.3417405026449332e-05, + "loss": 0.0394, + "step": 24876 + }, + { + "epoch": 3.26, + "learning_rate": 1.3416906625141603e-05, + "loss": 0.0419, + "step": 24877 + }, + { + "epoch": 3.26, + "learning_rate": 1.3416408214224003e-05, + "loss": 0.04, + "step": 24878 + }, + { + "epoch": 3.26, + "learning_rate": 1.3415909793697937e-05, + "loss": 0.0467, + "step": 24879 + }, + { + "epoch": 3.26, + "learning_rate": 1.3415411363564807e-05, + "loss": 0.0432, + "step": 24880 + }, + { + "epoch": 3.26, + "learning_rate": 1.3414912923826015e-05, + "loss": 0.0475, + "step": 24881 + }, + { + "epoch": 3.26, + "learning_rate": 1.3414414474482963e-05, + "loss": 0.0358, + "step": 24882 + }, + { + "epoch": 3.26, + "learning_rate": 1.341391601553705e-05, + "loss": 0.0396, + "step": 24883 + }, + { + "epoch": 3.26, + "learning_rate": 1.341341754698968e-05, + "loss": 0.0443, + "step": 24884 + }, + { + "epoch": 3.26, + "learning_rate": 1.3412919068842259e-05, + "loss": 0.0396, + "step": 24885 + }, + { + "epoch": 3.26, + "learning_rate": 1.341242058109618e-05, + "loss": 0.0363, + "step": 24886 + }, + { + "epoch": 3.26, + "learning_rate": 1.3411922083752852e-05, + "loss": 0.0397, + "step": 24887 + }, + { + "epoch": 3.26, + "learning_rate": 1.3411423576813675e-05, + "loss": 0.043, + "step": 24888 + }, + { + "epoch": 3.26, + "learning_rate": 1.341092506028005e-05, + "loss": 0.043, + "step": 24889 + }, + { + "epoch": 3.26, + "learning_rate": 1.341042653415338e-05, + "loss": 0.043, + "step": 24890 + }, + { + "epoch": 3.26, + "learning_rate": 1.3409927998435068e-05, + "loss": 0.0536, + "step": 24891 + }, + { + "epoch": 3.26, + "learning_rate": 1.3409429453126514e-05, + "loss": 0.043, + "step": 24892 + }, + { + "epoch": 3.26, + "learning_rate": 1.3408930898229125e-05, + "loss": 0.0367, + "step": 24893 + }, + { + "epoch": 3.26, + "learning_rate": 1.3408432333744293e-05, + "loss": 0.0392, + "step": 24894 + }, + { + "epoch": 3.26, + "learning_rate": 1.3407933759673431e-05, + "loss": 0.034, + "step": 24895 + }, + { + "epoch": 3.26, + "learning_rate": 1.3407435176017933e-05, + "loss": 0.0404, + "step": 24896 + }, + { + "epoch": 3.26, + "learning_rate": 1.3406936582779207e-05, + "loss": 0.0462, + "step": 24897 + }, + { + "epoch": 3.26, + "learning_rate": 1.3406437979958655e-05, + "loss": 0.042, + "step": 24898 + }, + { + "epoch": 3.26, + "learning_rate": 1.3405939367557677e-05, + "loss": 0.0415, + "step": 24899 + }, + { + "epoch": 3.26, + "learning_rate": 1.3405440745577673e-05, + "loss": 0.0353, + "step": 24900 + }, + { + "epoch": 3.26, + "learning_rate": 1.340494211402005e-05, + "loss": 0.0446, + "step": 24901 + }, + { + "epoch": 3.26, + "learning_rate": 1.3404443472886213e-05, + "loss": 0.0396, + "step": 24902 + }, + { + "epoch": 3.26, + "learning_rate": 1.3403944822177554e-05, + "loss": 0.037, + "step": 24903 + }, + { + "epoch": 3.26, + "learning_rate": 1.3403446161895486e-05, + "loss": 0.0362, + "step": 24904 + }, + { + "epoch": 3.26, + "learning_rate": 1.34029474920414e-05, + "loss": 0.0386, + "step": 24905 + }, + { + "epoch": 3.26, + "learning_rate": 1.3402448812616713e-05, + "loss": 0.046, + "step": 24906 + }, + { + "epoch": 3.26, + "learning_rate": 1.3401950123622815e-05, + "loss": 0.0465, + "step": 24907 + }, + { + "epoch": 3.26, + "learning_rate": 1.3401451425061116e-05, + "loss": 0.0407, + "step": 24908 + }, + { + "epoch": 3.26, + "learning_rate": 1.3400952716933016e-05, + "loss": 0.0442, + "step": 24909 + }, + { + "epoch": 3.26, + "learning_rate": 1.3400453999239917e-05, + "loss": 0.0396, + "step": 24910 + }, + { + "epoch": 3.26, + "learning_rate": 1.339995527198322e-05, + "loss": 0.0433, + "step": 24911 + }, + { + "epoch": 3.26, + "learning_rate": 1.3399456535164333e-05, + "loss": 0.037, + "step": 24912 + }, + { + "epoch": 3.26, + "learning_rate": 1.3398957788784654e-05, + "loss": 0.041, + "step": 24913 + }, + { + "epoch": 3.26, + "learning_rate": 1.339845903284559e-05, + "loss": 0.0409, + "step": 24914 + }, + { + "epoch": 3.26, + "learning_rate": 1.3397960267348538e-05, + "loss": 0.0423, + "step": 24915 + }, + { + "epoch": 3.26, + "learning_rate": 1.3397461492294904e-05, + "loss": 0.0467, + "step": 24916 + }, + { + "epoch": 3.26, + "learning_rate": 1.339696270768609e-05, + "loss": 0.0395, + "step": 24917 + }, + { + "epoch": 3.26, + "learning_rate": 1.33964639135235e-05, + "loss": 0.0388, + "step": 24918 + }, + { + "epoch": 3.26, + "learning_rate": 1.3395965109808538e-05, + "loss": 0.0411, + "step": 24919 + }, + { + "epoch": 3.26, + "learning_rate": 1.33954662965426e-05, + "loss": 0.0366, + "step": 24920 + }, + { + "epoch": 3.26, + "learning_rate": 1.33949674737271e-05, + "loss": 0.0401, + "step": 24921 + }, + { + "epoch": 3.26, + "learning_rate": 1.339446864136343e-05, + "loss": 0.0436, + "step": 24922 + }, + { + "epoch": 3.26, + "learning_rate": 1.3393969799452999e-05, + "loss": 0.048, + "step": 24923 + }, + { + "epoch": 3.26, + "learning_rate": 1.3393470947997209e-05, + "loss": 0.04, + "step": 24924 + }, + { + "epoch": 3.26, + "learning_rate": 1.3392972086997465e-05, + "loss": 0.052, + "step": 24925 + }, + { + "epoch": 3.27, + "learning_rate": 1.3392473216455164e-05, + "loss": 0.036, + "step": 24926 + }, + { + "epoch": 3.27, + "learning_rate": 1.3391974336371714e-05, + "loss": 0.0329, + "step": 24927 + }, + { + "epoch": 3.27, + "learning_rate": 1.3391475446748516e-05, + "loss": 0.0358, + "step": 24928 + }, + { + "epoch": 3.27, + "learning_rate": 1.3390976547586979e-05, + "loss": 0.0368, + "step": 24929 + }, + { + "epoch": 3.27, + "learning_rate": 1.3390477638888496e-05, + "loss": 0.0409, + "step": 24930 + }, + { + "epoch": 3.27, + "learning_rate": 1.3389978720654475e-05, + "loss": 0.046, + "step": 24931 + }, + { + "epoch": 3.27, + "learning_rate": 1.3389479792886319e-05, + "loss": 0.0404, + "step": 24932 + }, + { + "epoch": 3.27, + "learning_rate": 1.3388980855585434e-05, + "loss": 0.0375, + "step": 24933 + }, + { + "epoch": 3.27, + "learning_rate": 1.338848190875322e-05, + "loss": 0.0396, + "step": 24934 + }, + { + "epoch": 3.27, + "learning_rate": 1.3387982952391083e-05, + "loss": 0.0397, + "step": 24935 + }, + { + "epoch": 3.27, + "learning_rate": 1.3387483986500422e-05, + "loss": 0.0457, + "step": 24936 + }, + { + "epoch": 3.27, + "learning_rate": 1.3386985011082643e-05, + "loss": 0.0424, + "step": 24937 + }, + { + "epoch": 3.27, + "learning_rate": 1.3386486026139147e-05, + "loss": 0.0362, + "step": 24938 + }, + { + "epoch": 3.27, + "learning_rate": 1.3385987031671344e-05, + "loss": 0.0409, + "step": 24939 + }, + { + "epoch": 3.27, + "learning_rate": 1.3385488027680632e-05, + "loss": 0.0311, + "step": 24940 + }, + { + "epoch": 3.27, + "learning_rate": 1.338498901416841e-05, + "loss": 0.0479, + "step": 24941 + }, + { + "epoch": 3.27, + "learning_rate": 1.3384489991136091e-05, + "loss": 0.0417, + "step": 24942 + }, + { + "epoch": 3.27, + "learning_rate": 1.3383990958585074e-05, + "loss": 0.0407, + "step": 24943 + }, + { + "epoch": 3.27, + "learning_rate": 1.3383491916516763e-05, + "loss": 0.0359, + "step": 24944 + }, + { + "epoch": 3.27, + "learning_rate": 1.3382992864932558e-05, + "loss": 0.0414, + "step": 24945 + }, + { + "epoch": 3.27, + "learning_rate": 1.3382493803833869e-05, + "loss": 0.0434, + "step": 24946 + }, + { + "epoch": 3.27, + "learning_rate": 1.3381994733222094e-05, + "loss": 0.0477, + "step": 24947 + }, + { + "epoch": 3.27, + "learning_rate": 1.3381495653098642e-05, + "loss": 0.0444, + "step": 24948 + }, + { + "epoch": 3.27, + "learning_rate": 1.3380996563464909e-05, + "loss": 0.0318, + "step": 24949 + }, + { + "epoch": 3.27, + "learning_rate": 1.3380497464322306e-05, + "loss": 0.035, + "step": 24950 + }, + { + "epoch": 3.27, + "learning_rate": 1.3379998355672235e-05, + "loss": 0.0399, + "step": 24951 + }, + { + "epoch": 3.27, + "learning_rate": 1.3379499237516099e-05, + "loss": 0.0411, + "step": 24952 + }, + { + "epoch": 3.27, + "learning_rate": 1.3379000109855298e-05, + "loss": 0.0412, + "step": 24953 + }, + { + "epoch": 3.27, + "learning_rate": 1.3378500972691241e-05, + "loss": 0.0445, + "step": 24954 + }, + { + "epoch": 3.27, + "learning_rate": 1.3378001826025328e-05, + "loss": 0.0395, + "step": 24955 + }, + { + "epoch": 3.27, + "learning_rate": 1.3377502669858968e-05, + "loss": 0.0445, + "step": 24956 + }, + { + "epoch": 3.27, + "learning_rate": 1.3377003504193562e-05, + "loss": 0.0431, + "step": 24957 + }, + { + "epoch": 3.27, + "learning_rate": 1.3376504329030513e-05, + "loss": 0.0405, + "step": 24958 + }, + { + "epoch": 3.27, + "learning_rate": 1.3376005144371222e-05, + "loss": 0.0393, + "step": 24959 + }, + { + "epoch": 3.27, + "learning_rate": 1.3375505950217098e-05, + "loss": 0.0416, + "step": 24960 + }, + { + "epoch": 3.27, + "learning_rate": 1.3375006746569544e-05, + "loss": 0.043, + "step": 24961 + }, + { + "epoch": 3.27, + "learning_rate": 1.337450753342996e-05, + "loss": 0.0417, + "step": 24962 + }, + { + "epoch": 3.27, + "learning_rate": 1.3374008310799758e-05, + "loss": 0.0385, + "step": 24963 + }, + { + "epoch": 3.27, + "learning_rate": 1.3373509078680336e-05, + "loss": 0.0464, + "step": 24964 + }, + { + "epoch": 3.27, + "learning_rate": 1.3373009837073099e-05, + "loss": 0.0463, + "step": 24965 + }, + { + "epoch": 3.27, + "learning_rate": 1.337251058597945e-05, + "loss": 0.0455, + "step": 24966 + }, + { + "epoch": 3.27, + "learning_rate": 1.3372011325400797e-05, + "loss": 0.0388, + "step": 24967 + }, + { + "epoch": 3.27, + "learning_rate": 1.337151205533854e-05, + "loss": 0.0394, + "step": 24968 + }, + { + "epoch": 3.27, + "learning_rate": 1.3371012775794086e-05, + "loss": 0.0365, + "step": 24969 + }, + { + "epoch": 3.27, + "learning_rate": 1.3370513486768838e-05, + "loss": 0.0385, + "step": 24970 + }, + { + "epoch": 3.27, + "learning_rate": 1.3370014188264198e-05, + "loss": 0.0453, + "step": 24971 + }, + { + "epoch": 3.27, + "learning_rate": 1.3369514880281575e-05, + "loss": 0.0364, + "step": 24972 + }, + { + "epoch": 3.27, + "learning_rate": 1.3369015562822371e-05, + "loss": 0.0382, + "step": 24973 + }, + { + "epoch": 3.27, + "learning_rate": 1.336851623588799e-05, + "loss": 0.0493, + "step": 24974 + }, + { + "epoch": 3.27, + "learning_rate": 1.3368016899479837e-05, + "loss": 0.0413, + "step": 24975 + }, + { + "epoch": 3.27, + "learning_rate": 1.3367517553599311e-05, + "loss": 0.0353, + "step": 24976 + }, + { + "epoch": 3.27, + "learning_rate": 1.3367018198247828e-05, + "loss": 0.0358, + "step": 24977 + }, + { + "epoch": 3.27, + "learning_rate": 1.336651883342678e-05, + "loss": 0.0408, + "step": 24978 + }, + { + "epoch": 3.27, + "learning_rate": 1.3366019459137583e-05, + "loss": 0.0393, + "step": 24979 + }, + { + "epoch": 3.27, + "learning_rate": 1.336552007538163e-05, + "loss": 0.0356, + "step": 24980 + }, + { + "epoch": 3.27, + "learning_rate": 1.3365020682160332e-05, + "loss": 0.0442, + "step": 24981 + }, + { + "epoch": 3.27, + "learning_rate": 1.3364521279475093e-05, + "loss": 0.041, + "step": 24982 + }, + { + "epoch": 3.27, + "learning_rate": 1.336402186732732e-05, + "loss": 0.0444, + "step": 24983 + }, + { + "epoch": 3.27, + "learning_rate": 1.336352244571841e-05, + "loss": 0.0357, + "step": 24984 + }, + { + "epoch": 3.27, + "learning_rate": 1.3363023014649776e-05, + "loss": 0.0311, + "step": 24985 + }, + { + "epoch": 3.27, + "learning_rate": 1.3362523574122819e-05, + "loss": 0.0436, + "step": 24986 + }, + { + "epoch": 3.27, + "learning_rate": 1.3362024124138939e-05, + "loss": 0.0482, + "step": 24987 + }, + { + "epoch": 3.27, + "learning_rate": 1.336152466469955e-05, + "loss": 0.0449, + "step": 24988 + }, + { + "epoch": 3.27, + "learning_rate": 1.3361025195806048e-05, + "loss": 0.0376, + "step": 24989 + }, + { + "epoch": 3.27, + "learning_rate": 1.3360525717459844e-05, + "loss": 0.0483, + "step": 24990 + }, + { + "epoch": 3.27, + "learning_rate": 1.3360026229662342e-05, + "loss": 0.0459, + "step": 24991 + }, + { + "epoch": 3.27, + "learning_rate": 1.3359526732414942e-05, + "loss": 0.0357, + "step": 24992 + }, + { + "epoch": 3.27, + "learning_rate": 1.3359027225719052e-05, + "loss": 0.0436, + "step": 24993 + }, + { + "epoch": 3.27, + "learning_rate": 1.3358527709576081e-05, + "loss": 0.0487, + "step": 24994 + }, + { + "epoch": 3.27, + "learning_rate": 1.3358028183987425e-05, + "loss": 0.036, + "step": 24995 + }, + { + "epoch": 3.27, + "learning_rate": 1.3357528648954496e-05, + "loss": 0.0385, + "step": 24996 + }, + { + "epoch": 3.27, + "learning_rate": 1.3357029104478695e-05, + "loss": 0.0467, + "step": 24997 + }, + { + "epoch": 3.27, + "learning_rate": 1.3356529550561428e-05, + "loss": 0.0421, + "step": 24998 + }, + { + "epoch": 3.27, + "learning_rate": 1.3356029987204103e-05, + "loss": 0.0414, + "step": 24999 + }, + { + "epoch": 3.27, + "learning_rate": 1.3355530414408122e-05, + "loss": 0.0421, + "step": 25000 + }, + { + "epoch": 3.27, + "learning_rate": 1.3355030832174887e-05, + "loss": 0.042, + "step": 25001 + }, + { + "epoch": 3.28, + "learning_rate": 1.3354531240505809e-05, + "loss": 0.0453, + "step": 25002 + }, + { + "epoch": 3.28, + "learning_rate": 1.335403163940229e-05, + "loss": 0.0369, + "step": 25003 + }, + { + "epoch": 3.28, + "learning_rate": 1.3353532028865737e-05, + "loss": 0.04, + "step": 25004 + }, + { + "epoch": 3.28, + "learning_rate": 1.3353032408897552e-05, + "loss": 0.0348, + "step": 25005 + }, + { + "epoch": 3.28, + "learning_rate": 1.3352532779499141e-05, + "loss": 0.0434, + "step": 25006 + }, + { + "epoch": 3.28, + "learning_rate": 1.3352033140671914e-05, + "loss": 0.0381, + "step": 25007 + }, + { + "epoch": 3.28, + "learning_rate": 1.335153349241727e-05, + "loss": 0.0415, + "step": 25008 + }, + { + "epoch": 3.28, + "learning_rate": 1.3351033834736616e-05, + "loss": 0.0358, + "step": 25009 + }, + { + "epoch": 3.28, + "learning_rate": 1.3350534167631357e-05, + "loss": 0.0399, + "step": 25010 + }, + { + "epoch": 3.28, + "learning_rate": 1.3350034491102902e-05, + "loss": 0.0437, + "step": 25011 + }, + { + "epoch": 3.28, + "learning_rate": 1.3349534805152651e-05, + "loss": 0.0472, + "step": 25012 + }, + { + "epoch": 3.28, + "learning_rate": 1.3349035109782014e-05, + "loss": 0.0341, + "step": 25013 + }, + { + "epoch": 3.28, + "learning_rate": 1.3348535404992388e-05, + "loss": 0.0314, + "step": 25014 + }, + { + "epoch": 3.28, + "learning_rate": 1.3348035690785193e-05, + "loss": 0.0421, + "step": 25015 + }, + { + "epoch": 3.28, + "learning_rate": 1.3347535967161819e-05, + "loss": 0.0425, + "step": 25016 + }, + { + "epoch": 3.28, + "learning_rate": 1.3347036234123684e-05, + "loss": 0.0473, + "step": 25017 + }, + { + "epoch": 3.28, + "learning_rate": 1.3346536491672184e-05, + "loss": 0.0424, + "step": 25018 + }, + { + "epoch": 3.28, + "learning_rate": 1.3346036739808733e-05, + "loss": 0.0422, + "step": 25019 + }, + { + "epoch": 3.28, + "learning_rate": 1.3345536978534726e-05, + "loss": 0.0447, + "step": 25020 + }, + { + "epoch": 3.28, + "learning_rate": 1.3345037207851577e-05, + "loss": 0.0426, + "step": 25021 + }, + { + "epoch": 3.28, + "learning_rate": 1.3344537427760687e-05, + "loss": 0.0415, + "step": 25022 + }, + { + "epoch": 3.28, + "learning_rate": 1.3344037638263465e-05, + "loss": 0.045, + "step": 25023 + }, + { + "epoch": 3.28, + "learning_rate": 1.3343537839361319e-05, + "loss": 0.0423, + "step": 25024 + }, + { + "epoch": 3.28, + "learning_rate": 1.3343038031055646e-05, + "loss": 0.0409, + "step": 25025 + }, + { + "epoch": 3.28, + "learning_rate": 1.3342538213347855e-05, + "loss": 0.0472, + "step": 25026 + }, + { + "epoch": 3.28, + "learning_rate": 1.3342038386239358e-05, + "loss": 0.0353, + "step": 25027 + }, + { + "epoch": 3.28, + "learning_rate": 1.3341538549731554e-05, + "loss": 0.0435, + "step": 25028 + }, + { + "epoch": 3.28, + "learning_rate": 1.334103870382585e-05, + "loss": 0.046, + "step": 25029 + }, + { + "epoch": 3.28, + "learning_rate": 1.3340538848523657e-05, + "loss": 0.0415, + "step": 25030 + }, + { + "epoch": 3.28, + "learning_rate": 1.334003898382637e-05, + "loss": 0.0422, + "step": 25031 + }, + { + "epoch": 3.28, + "learning_rate": 1.3339539109735404e-05, + "loss": 0.0383, + "step": 25032 + }, + { + "epoch": 3.28, + "learning_rate": 1.3339039226252164e-05, + "loss": 0.0402, + "step": 25033 + }, + { + "epoch": 3.28, + "learning_rate": 1.3338539333378053e-05, + "loss": 0.0411, + "step": 25034 + }, + { + "epoch": 3.28, + "learning_rate": 1.3338039431114478e-05, + "loss": 0.0356, + "step": 25035 + }, + { + "epoch": 3.28, + "learning_rate": 1.3337539519462844e-05, + "loss": 0.0471, + "step": 25036 + }, + { + "epoch": 3.28, + "learning_rate": 1.3337039598424555e-05, + "loss": 0.0433, + "step": 25037 + }, + { + "epoch": 3.28, + "learning_rate": 1.3336539668001025e-05, + "loss": 0.0401, + "step": 25038 + }, + { + "epoch": 3.28, + "learning_rate": 1.3336039728193653e-05, + "loss": 0.0461, + "step": 25039 + }, + { + "epoch": 3.28, + "learning_rate": 1.3335539779003845e-05, + "loss": 0.0436, + "step": 25040 + }, + { + "epoch": 3.28, + "learning_rate": 1.3335039820433011e-05, + "loss": 0.035, + "step": 25041 + }, + { + "epoch": 3.28, + "learning_rate": 1.3334539852482555e-05, + "loss": 0.0383, + "step": 25042 + }, + { + "epoch": 3.28, + "learning_rate": 1.3334039875153882e-05, + "loss": 0.0403, + "step": 25043 + }, + { + "epoch": 3.28, + "learning_rate": 1.3333539888448402e-05, + "loss": 0.043, + "step": 25044 + }, + { + "epoch": 3.28, + "learning_rate": 1.3333039892367517e-05, + "loss": 0.0387, + "step": 25045 + }, + { + "epoch": 3.28, + "learning_rate": 1.3332539886912632e-05, + "loss": 0.0432, + "step": 25046 + }, + { + "epoch": 3.28, + "learning_rate": 1.3332039872085158e-05, + "loss": 0.0406, + "step": 25047 + }, + { + "epoch": 3.28, + "learning_rate": 1.3331539847886502e-05, + "loss": 0.0387, + "step": 25048 + }, + { + "epoch": 3.28, + "learning_rate": 1.3331039814318065e-05, + "loss": 0.04, + "step": 25049 + }, + { + "epoch": 3.28, + "learning_rate": 1.3330539771381257e-05, + "loss": 0.0467, + "step": 25050 + }, + { + "epoch": 3.28, + "learning_rate": 1.3330039719077483e-05, + "loss": 0.0406, + "step": 25051 + }, + { + "epoch": 3.28, + "learning_rate": 1.3329539657408149e-05, + "loss": 0.0467, + "step": 25052 + }, + { + "epoch": 3.28, + "learning_rate": 1.3329039586374664e-05, + "loss": 0.0408, + "step": 25053 + }, + { + "epoch": 3.28, + "learning_rate": 1.3328539505978429e-05, + "loss": 0.0422, + "step": 25054 + }, + { + "epoch": 3.28, + "learning_rate": 1.3328039416220857e-05, + "loss": 0.0452, + "step": 25055 + }, + { + "epoch": 3.28, + "learning_rate": 1.3327539317103349e-05, + "loss": 0.0363, + "step": 25056 + }, + { + "epoch": 3.28, + "learning_rate": 1.3327039208627316e-05, + "loss": 0.0386, + "step": 25057 + }, + { + "epoch": 3.28, + "learning_rate": 1.332653909079416e-05, + "loss": 0.0444, + "step": 25058 + }, + { + "epoch": 3.28, + "learning_rate": 1.3326038963605292e-05, + "loss": 0.0404, + "step": 25059 + }, + { + "epoch": 3.28, + "learning_rate": 1.3325538827062116e-05, + "loss": 0.0342, + "step": 25060 + }, + { + "epoch": 3.28, + "learning_rate": 1.332503868116604e-05, + "loss": 0.0356, + "step": 25061 + }, + { + "epoch": 3.28, + "learning_rate": 1.3324538525918469e-05, + "loss": 0.0388, + "step": 25062 + }, + { + "epoch": 3.28, + "learning_rate": 1.3324038361320808e-05, + "loss": 0.0388, + "step": 25063 + }, + { + "epoch": 3.28, + "learning_rate": 1.3323538187374467e-05, + "loss": 0.0446, + "step": 25064 + }, + { + "epoch": 3.28, + "learning_rate": 1.3323038004080857e-05, + "loss": 0.0436, + "step": 25065 + }, + { + "epoch": 3.28, + "learning_rate": 1.3322537811441374e-05, + "loss": 0.0425, + "step": 25066 + }, + { + "epoch": 3.28, + "learning_rate": 1.3322037609457433e-05, + "loss": 0.0447, + "step": 25067 + }, + { + "epoch": 3.28, + "learning_rate": 1.3321537398130438e-05, + "loss": 0.0375, + "step": 25068 + }, + { + "epoch": 3.28, + "learning_rate": 1.3321037177461794e-05, + "loss": 0.0398, + "step": 25069 + }, + { + "epoch": 3.28, + "learning_rate": 1.3320536947452912e-05, + "loss": 0.0404, + "step": 25070 + }, + { + "epoch": 3.28, + "learning_rate": 1.3320036708105193e-05, + "loss": 0.0415, + "step": 25071 + }, + { + "epoch": 3.28, + "learning_rate": 1.3319536459420052e-05, + "loss": 0.0359, + "step": 25072 + }, + { + "epoch": 3.28, + "learning_rate": 1.331903620139889e-05, + "loss": 0.0453, + "step": 25073 + }, + { + "epoch": 3.28, + "learning_rate": 1.3318535934043113e-05, + "loss": 0.0314, + "step": 25074 + }, + { + "epoch": 3.28, + "learning_rate": 1.3318035657354133e-05, + "loss": 0.0376, + "step": 25075 + }, + { + "epoch": 3.28, + "learning_rate": 1.3317535371333357e-05, + "loss": 0.046, + "step": 25076 + }, + { + "epoch": 3.28, + "learning_rate": 1.3317035075982183e-05, + "loss": 0.0421, + "step": 25077 + }, + { + "epoch": 3.29, + "learning_rate": 1.331653477130203e-05, + "loss": 0.0443, + "step": 25078 + }, + { + "epoch": 3.29, + "learning_rate": 1.3316034457294298e-05, + "loss": 0.0431, + "step": 25079 + }, + { + "epoch": 3.29, + "learning_rate": 1.3315534133960393e-05, + "loss": 0.0387, + "step": 25080 + }, + { + "epoch": 3.29, + "learning_rate": 1.3315033801301727e-05, + "loss": 0.0424, + "step": 25081 + }, + { + "epoch": 3.29, + "learning_rate": 1.3314533459319707e-05, + "loss": 0.0413, + "step": 25082 + }, + { + "epoch": 3.29, + "learning_rate": 1.3314033108015735e-05, + "loss": 0.0413, + "step": 25083 + }, + { + "epoch": 3.29, + "learning_rate": 1.3313532747391225e-05, + "loss": 0.0491, + "step": 25084 + }, + { + "epoch": 3.29, + "learning_rate": 1.3313032377447575e-05, + "loss": 0.0366, + "step": 25085 + }, + { + "epoch": 3.29, + "learning_rate": 1.3312531998186202e-05, + "loss": 0.0353, + "step": 25086 + }, + { + "epoch": 3.29, + "learning_rate": 1.3312031609608506e-05, + "loss": 0.0399, + "step": 25087 + }, + { + "epoch": 3.29, + "learning_rate": 1.33115312117159e-05, + "loss": 0.0446, + "step": 25088 + }, + { + "epoch": 3.29, + "learning_rate": 1.331103080450979e-05, + "loss": 0.0392, + "step": 25089 + }, + { + "epoch": 3.29, + "learning_rate": 1.3310530387991581e-05, + "loss": 0.0393, + "step": 25090 + }, + { + "epoch": 3.29, + "learning_rate": 1.331002996216268e-05, + "loss": 0.0395, + "step": 25091 + }, + { + "epoch": 3.29, + "learning_rate": 1.3309529527024496e-05, + "loss": 0.0385, + "step": 25092 + }, + { + "epoch": 3.29, + "learning_rate": 1.330902908257844e-05, + "loss": 0.0413, + "step": 25093 + }, + { + "epoch": 3.29, + "learning_rate": 1.3308528628825911e-05, + "loss": 0.0462, + "step": 25094 + }, + { + "epoch": 3.29, + "learning_rate": 1.3308028165768327e-05, + "loss": 0.0359, + "step": 25095 + }, + { + "epoch": 3.29, + "learning_rate": 1.3307527693407083e-05, + "loss": 0.0412, + "step": 25096 + }, + { + "epoch": 3.29, + "learning_rate": 1.3307027211743599e-05, + "loss": 0.0405, + "step": 25097 + }, + { + "epoch": 3.29, + "learning_rate": 1.3306526720779276e-05, + "loss": 0.0445, + "step": 25098 + }, + { + "epoch": 3.29, + "learning_rate": 1.3306026220515523e-05, + "loss": 0.0474, + "step": 25099 + }, + { + "epoch": 3.29, + "learning_rate": 1.3305525710953747e-05, + "loss": 0.0423, + "step": 25100 + }, + { + "epoch": 3.29, + "learning_rate": 1.3305025192095355e-05, + "loss": 0.0431, + "step": 25101 + }, + { + "epoch": 3.29, + "learning_rate": 1.3304524663941753e-05, + "loss": 0.048, + "step": 25102 + }, + { + "epoch": 3.29, + "learning_rate": 1.3304024126494358e-05, + "loss": 0.0389, + "step": 25103 + }, + { + "epoch": 3.29, + "learning_rate": 1.3303523579754566e-05, + "loss": 0.039, + "step": 25104 + }, + { + "epoch": 3.29, + "learning_rate": 1.3303023023723793e-05, + "loss": 0.0358, + "step": 25105 + }, + { + "epoch": 3.29, + "learning_rate": 1.330252245840344e-05, + "loss": 0.0493, + "step": 25106 + }, + { + "epoch": 3.29, + "learning_rate": 1.330202188379492e-05, + "loss": 0.0454, + "step": 25107 + }, + { + "epoch": 3.29, + "learning_rate": 1.330152129989964e-05, + "loss": 0.0365, + "step": 25108 + }, + { + "epoch": 3.29, + "learning_rate": 1.3301020706719007e-05, + "loss": 0.0448, + "step": 25109 + }, + { + "epoch": 3.29, + "learning_rate": 1.330052010425443e-05, + "loss": 0.0385, + "step": 25110 + }, + { + "epoch": 3.29, + "learning_rate": 1.3300019492507314e-05, + "loss": 0.0372, + "step": 25111 + }, + { + "epoch": 3.29, + "learning_rate": 1.3299518871479072e-05, + "loss": 0.0381, + "step": 25112 + }, + { + "epoch": 3.29, + "learning_rate": 1.3299018241171102e-05, + "loss": 0.0402, + "step": 25113 + }, + { + "epoch": 3.29, + "learning_rate": 1.3298517601584827e-05, + "loss": 0.0367, + "step": 25114 + }, + { + "epoch": 3.29, + "learning_rate": 1.3298016952721642e-05, + "loss": 0.0398, + "step": 25115 + }, + { + "epoch": 3.29, + "learning_rate": 1.3297516294582961e-05, + "loss": 0.0389, + "step": 25116 + }, + { + "epoch": 3.29, + "learning_rate": 1.3297015627170193e-05, + "loss": 0.0393, + "step": 25117 + }, + { + "epoch": 3.29, + "learning_rate": 1.3296514950484742e-05, + "loss": 0.0394, + "step": 25118 + }, + { + "epoch": 3.29, + "learning_rate": 1.329601426452802e-05, + "loss": 0.0379, + "step": 25119 + }, + { + "epoch": 3.29, + "learning_rate": 1.3295513569301432e-05, + "loss": 0.042, + "step": 25120 + }, + { + "epoch": 3.29, + "learning_rate": 1.3295012864806387e-05, + "loss": 0.0412, + "step": 25121 + }, + { + "epoch": 3.29, + "learning_rate": 1.3294512151044297e-05, + "loss": 0.0433, + "step": 25122 + }, + { + "epoch": 3.29, + "learning_rate": 1.3294011428016562e-05, + "loss": 0.042, + "step": 25123 + }, + { + "epoch": 3.29, + "learning_rate": 1.3293510695724599e-05, + "loss": 0.0345, + "step": 25124 + }, + { + "epoch": 3.29, + "learning_rate": 1.329300995416981e-05, + "loss": 0.0475, + "step": 25125 + }, + { + "epoch": 3.29, + "learning_rate": 1.329250920335361e-05, + "loss": 0.0369, + "step": 25126 + }, + { + "epoch": 3.29, + "learning_rate": 1.32920084432774e-05, + "loss": 0.039, + "step": 25127 + }, + { + "epoch": 3.29, + "learning_rate": 1.3291507673942592e-05, + "loss": 0.0375, + "step": 25128 + }, + { + "epoch": 3.29, + "learning_rate": 1.3291006895350593e-05, + "loss": 0.0455, + "step": 25129 + }, + { + "epoch": 3.29, + "learning_rate": 1.3290506107502813e-05, + "loss": 0.0422, + "step": 25130 + }, + { + "epoch": 3.29, + "learning_rate": 1.329000531040066e-05, + "loss": 0.039, + "step": 25131 + }, + { + "epoch": 3.29, + "learning_rate": 1.3289504504045542e-05, + "loss": 0.0375, + "step": 25132 + }, + { + "epoch": 3.29, + "learning_rate": 1.3289003688438871e-05, + "loss": 0.0374, + "step": 25133 + }, + { + "epoch": 3.29, + "learning_rate": 1.3288502863582049e-05, + "loss": 0.0371, + "step": 25134 + }, + { + "epoch": 3.29, + "learning_rate": 1.3288002029476488e-05, + "loss": 0.041, + "step": 25135 + }, + { + "epoch": 3.29, + "learning_rate": 1.3287501186123595e-05, + "loss": 0.0386, + "step": 25136 + }, + { + "epoch": 3.29, + "learning_rate": 1.3287000333524783e-05, + "loss": 0.0345, + "step": 25137 + }, + { + "epoch": 3.29, + "learning_rate": 1.3286499471681455e-05, + "loss": 0.0369, + "step": 25138 + }, + { + "epoch": 3.29, + "learning_rate": 1.3285998600595024e-05, + "loss": 0.034, + "step": 25139 + }, + { + "epoch": 3.29, + "learning_rate": 1.3285497720266896e-05, + "loss": 0.034, + "step": 25140 + }, + { + "epoch": 3.29, + "learning_rate": 1.3284996830698482e-05, + "loss": 0.0367, + "step": 25141 + }, + { + "epoch": 3.29, + "learning_rate": 1.3284495931891189e-05, + "loss": 0.0459, + "step": 25142 + }, + { + "epoch": 3.29, + "learning_rate": 1.3283995023846425e-05, + "loss": 0.0384, + "step": 25143 + }, + { + "epoch": 3.29, + "learning_rate": 1.3283494106565602e-05, + "loss": 0.0394, + "step": 25144 + }, + { + "epoch": 3.29, + "learning_rate": 1.3282993180050123e-05, + "loss": 0.0354, + "step": 25145 + }, + { + "epoch": 3.29, + "learning_rate": 1.3282492244301404e-05, + "loss": 0.0337, + "step": 25146 + }, + { + "epoch": 3.29, + "learning_rate": 1.3281991299320848e-05, + "loss": 0.0385, + "step": 25147 + }, + { + "epoch": 3.29, + "learning_rate": 1.3281490345109868e-05, + "loss": 0.0355, + "step": 25148 + }, + { + "epoch": 3.29, + "learning_rate": 1.3280989381669872e-05, + "loss": 0.0403, + "step": 25149 + }, + { + "epoch": 3.29, + "learning_rate": 1.3280488409002266e-05, + "loss": 0.0486, + "step": 25150 + }, + { + "epoch": 3.29, + "learning_rate": 1.327998742710846e-05, + "loss": 0.0389, + "step": 25151 + }, + { + "epoch": 3.29, + "learning_rate": 1.3279486435989866e-05, + "loss": 0.0404, + "step": 25152 + }, + { + "epoch": 3.29, + "learning_rate": 1.3278985435647891e-05, + "loss": 0.0396, + "step": 25153 + }, + { + "epoch": 3.29, + "learning_rate": 1.3278484426083944e-05, + "loss": 0.0424, + "step": 25154 + }, + { + "epoch": 3.3, + "learning_rate": 1.3277983407299432e-05, + "loss": 0.0345, + "step": 25155 + }, + { + "epoch": 3.3, + "learning_rate": 1.3277482379295767e-05, + "loss": 0.0346, + "step": 25156 + }, + { + "epoch": 3.3, + "learning_rate": 1.327698134207436e-05, + "loss": 0.0483, + "step": 25157 + }, + { + "epoch": 3.3, + "learning_rate": 1.3276480295636617e-05, + "loss": 0.0367, + "step": 25158 + }, + { + "epoch": 3.3, + "learning_rate": 1.3275979239983947e-05, + "loss": 0.0409, + "step": 25159 + }, + { + "epoch": 3.3, + "learning_rate": 1.327547817511776e-05, + "loss": 0.04, + "step": 25160 + }, + { + "epoch": 3.3, + "learning_rate": 1.3274977101039463e-05, + "loss": 0.0538, + "step": 25161 + }, + { + "epoch": 3.3, + "learning_rate": 1.3274476017750469e-05, + "loss": 0.0436, + "step": 25162 + }, + { + "epoch": 3.3, + "learning_rate": 1.3273974925252184e-05, + "loss": 0.033, + "step": 25163 + }, + { + "epoch": 3.3, + "learning_rate": 1.3273473823546022e-05, + "loss": 0.0414, + "step": 25164 + }, + { + "epoch": 3.3, + "learning_rate": 1.3272972712633387e-05, + "loss": 0.041, + "step": 25165 + }, + { + "epoch": 3.3, + "learning_rate": 1.3272471592515692e-05, + "loss": 0.0447, + "step": 25166 + }, + { + "epoch": 3.3, + "learning_rate": 1.3271970463194342e-05, + "loss": 0.0383, + "step": 25167 + }, + { + "epoch": 3.3, + "learning_rate": 1.3271469324670751e-05, + "loss": 0.0422, + "step": 25168 + }, + { + "epoch": 3.3, + "learning_rate": 1.3270968176946326e-05, + "loss": 0.0394, + "step": 25169 + }, + { + "epoch": 3.3, + "learning_rate": 1.3270467020022482e-05, + "loss": 0.0382, + "step": 25170 + }, + { + "epoch": 3.3, + "learning_rate": 1.3269965853900617e-05, + "loss": 0.0365, + "step": 25171 + }, + { + "epoch": 3.3, + "learning_rate": 1.3269464678582152e-05, + "loss": 0.0424, + "step": 25172 + }, + { + "epoch": 3.3, + "learning_rate": 1.3268963494068487e-05, + "loss": 0.0352, + "step": 25173 + }, + { + "epoch": 3.3, + "learning_rate": 1.3268462300361042e-05, + "loss": 0.0389, + "step": 25174 + }, + { + "epoch": 3.3, + "learning_rate": 1.3267961097461214e-05, + "loss": 0.0371, + "step": 25175 + }, + { + "epoch": 3.3, + "learning_rate": 1.3267459885370423e-05, + "loss": 0.0535, + "step": 25176 + }, + { + "epoch": 3.3, + "learning_rate": 1.3266958664090077e-05, + "loss": 0.0485, + "step": 25177 + }, + { + "epoch": 3.3, + "learning_rate": 1.326645743362158e-05, + "loss": 0.0392, + "step": 25178 + }, + { + "epoch": 3.3, + "learning_rate": 1.326595619396635e-05, + "loss": 0.0341, + "step": 25179 + }, + { + "epoch": 3.3, + "learning_rate": 1.326545494512579e-05, + "loss": 0.0378, + "step": 25180 + }, + { + "epoch": 3.3, + "learning_rate": 1.3264953687101311e-05, + "loss": 0.0453, + "step": 25181 + }, + { + "epoch": 3.3, + "learning_rate": 1.3264452419894325e-05, + "loss": 0.0413, + "step": 25182 + }, + { + "epoch": 3.3, + "learning_rate": 1.326395114350624e-05, + "loss": 0.0419, + "step": 25183 + }, + { + "epoch": 3.3, + "learning_rate": 1.3263449857938464e-05, + "loss": 0.0424, + "step": 25184 + }, + { + "epoch": 3.3, + "learning_rate": 1.3262948563192411e-05, + "loss": 0.0471, + "step": 25185 + }, + { + "epoch": 3.3, + "learning_rate": 1.326244725926949e-05, + "loss": 0.0365, + "step": 25186 + }, + { + "epoch": 3.3, + "learning_rate": 1.3261945946171109e-05, + "loss": 0.0396, + "step": 25187 + }, + { + "epoch": 3.3, + "learning_rate": 1.3261444623898675e-05, + "loss": 0.0378, + "step": 25188 + }, + { + "epoch": 3.3, + "learning_rate": 1.3260943292453605e-05, + "loss": 0.0427, + "step": 25189 + }, + { + "epoch": 3.3, + "learning_rate": 1.3260441951837306e-05, + "loss": 0.036, + "step": 25190 + }, + { + "epoch": 3.3, + "learning_rate": 1.325994060205119e-05, + "loss": 0.0425, + "step": 25191 + }, + { + "epoch": 3.3, + "learning_rate": 1.3259439243096659e-05, + "loss": 0.0438, + "step": 25192 + }, + { + "epoch": 3.3, + "learning_rate": 1.3258937874975133e-05, + "loss": 0.0456, + "step": 25193 + }, + { + "epoch": 3.3, + "learning_rate": 1.3258436497688015e-05, + "loss": 0.0465, + "step": 25194 + }, + { + "epoch": 3.3, + "learning_rate": 1.325793511123672e-05, + "loss": 0.0401, + "step": 25195 + }, + { + "epoch": 3.3, + "learning_rate": 1.3257433715622655e-05, + "loss": 0.0392, + "step": 25196 + }, + { + "epoch": 3.3, + "learning_rate": 1.325693231084723e-05, + "loss": 0.0393, + "step": 25197 + }, + { + "epoch": 3.3, + "learning_rate": 1.325643089691186e-05, + "loss": 0.0354, + "step": 25198 + }, + { + "epoch": 3.3, + "learning_rate": 1.3255929473817948e-05, + "loss": 0.0347, + "step": 25199 + }, + { + "epoch": 3.3, + "learning_rate": 1.3255428041566909e-05, + "loss": 0.0381, + "step": 25200 + }, + { + "epoch": 3.3, + "learning_rate": 1.325492660016015e-05, + "loss": 0.0394, + "step": 25201 + }, + { + "epoch": 3.3, + "learning_rate": 1.3254425149599088e-05, + "loss": 0.0441, + "step": 25202 + }, + { + "epoch": 3.3, + "learning_rate": 1.3253923689885125e-05, + "loss": 0.0392, + "step": 25203 + }, + { + "epoch": 3.3, + "learning_rate": 1.3253422221019676e-05, + "loss": 0.0498, + "step": 25204 + }, + { + "epoch": 3.3, + "learning_rate": 1.3252920743004148e-05, + "loss": 0.0339, + "step": 25205 + }, + { + "epoch": 3.3, + "learning_rate": 1.3252419255839956e-05, + "loss": 0.0436, + "step": 25206 + }, + { + "epoch": 3.3, + "learning_rate": 1.3251917759528507e-05, + "loss": 0.0437, + "step": 25207 + }, + { + "epoch": 3.3, + "learning_rate": 1.3251416254071214e-05, + "loss": 0.042, + "step": 25208 + }, + { + "epoch": 3.3, + "learning_rate": 1.3250914739469486e-05, + "loss": 0.0413, + "step": 25209 + }, + { + "epoch": 3.3, + "learning_rate": 1.3250413215724733e-05, + "loss": 0.0415, + "step": 25210 + }, + { + "epoch": 3.3, + "learning_rate": 1.3249911682838363e-05, + "loss": 0.0418, + "step": 25211 + }, + { + "epoch": 3.3, + "learning_rate": 1.3249410140811793e-05, + "loss": 0.0447, + "step": 25212 + }, + { + "epoch": 3.3, + "learning_rate": 1.324890858964643e-05, + "loss": 0.0444, + "step": 25213 + }, + { + "epoch": 3.3, + "learning_rate": 1.3248407029343682e-05, + "loss": 0.042, + "step": 25214 + }, + { + "epoch": 3.3, + "learning_rate": 1.3247905459904964e-05, + "loss": 0.0421, + "step": 25215 + }, + { + "epoch": 3.3, + "learning_rate": 1.3247403881331684e-05, + "loss": 0.0371, + "step": 25216 + }, + { + "epoch": 3.3, + "learning_rate": 1.3246902293625254e-05, + "loss": 0.0357, + "step": 25217 + }, + { + "epoch": 3.3, + "learning_rate": 1.3246400696787082e-05, + "loss": 0.0458, + "step": 25218 + }, + { + "epoch": 3.3, + "learning_rate": 1.3245899090818587e-05, + "loss": 0.0386, + "step": 25219 + }, + { + "epoch": 3.3, + "learning_rate": 1.324539747572117e-05, + "loss": 0.0451, + "step": 25220 + }, + { + "epoch": 3.3, + "learning_rate": 1.3244895851496244e-05, + "loss": 0.0378, + "step": 25221 + }, + { + "epoch": 3.3, + "learning_rate": 1.324439421814522e-05, + "loss": 0.0414, + "step": 25222 + }, + { + "epoch": 3.3, + "learning_rate": 1.3243892575669515e-05, + "loss": 0.036, + "step": 25223 + }, + { + "epoch": 3.3, + "learning_rate": 1.3243390924070531e-05, + "loss": 0.041, + "step": 25224 + }, + { + "epoch": 3.3, + "learning_rate": 1.3242889263349685e-05, + "loss": 0.0429, + "step": 25225 + }, + { + "epoch": 3.3, + "learning_rate": 1.3242387593508386e-05, + "loss": 0.0378, + "step": 25226 + }, + { + "epoch": 3.3, + "learning_rate": 1.3241885914548043e-05, + "loss": 0.0333, + "step": 25227 + }, + { + "epoch": 3.3, + "learning_rate": 1.3241384226470067e-05, + "loss": 0.0434, + "step": 25228 + }, + { + "epoch": 3.3, + "learning_rate": 1.3240882529275874e-05, + "loss": 0.0375, + "step": 25229 + }, + { + "epoch": 3.3, + "learning_rate": 1.3240380822966869e-05, + "loss": 0.0407, + "step": 25230 + }, + { + "epoch": 3.31, + "learning_rate": 1.323987910754447e-05, + "loss": 0.0449, + "step": 25231 + }, + { + "epoch": 3.31, + "learning_rate": 1.3239377383010077e-05, + "loss": 0.0412, + "step": 25232 + }, + { + "epoch": 3.31, + "learning_rate": 1.323887564936511e-05, + "loss": 0.0339, + "step": 25233 + }, + { + "epoch": 3.31, + "learning_rate": 1.3238373906610978e-05, + "loss": 0.0423, + "step": 25234 + }, + { + "epoch": 3.31, + "learning_rate": 1.3237872154749093e-05, + "loss": 0.0372, + "step": 25235 + }, + { + "epoch": 3.31, + "learning_rate": 1.3237370393780865e-05, + "loss": 0.0394, + "step": 25236 + }, + { + "epoch": 3.31, + "learning_rate": 1.32368686237077e-05, + "loss": 0.04, + "step": 25237 + }, + { + "epoch": 3.31, + "learning_rate": 1.3236366844531018e-05, + "loss": 0.0438, + "step": 25238 + }, + { + "epoch": 3.31, + "learning_rate": 1.3235865056252228e-05, + "loss": 0.0442, + "step": 25239 + }, + { + "epoch": 3.31, + "learning_rate": 1.323536325887274e-05, + "loss": 0.0384, + "step": 25240 + }, + { + "epoch": 3.31, + "learning_rate": 1.3234861452393964e-05, + "loss": 0.0383, + "step": 25241 + }, + { + "epoch": 3.31, + "learning_rate": 1.3234359636817313e-05, + "loss": 0.0483, + "step": 25242 + }, + { + "epoch": 3.31, + "learning_rate": 1.3233857812144197e-05, + "loss": 0.0353, + "step": 25243 + }, + { + "epoch": 3.31, + "learning_rate": 1.3233355978376027e-05, + "loss": 0.0325, + "step": 25244 + }, + { + "epoch": 3.31, + "learning_rate": 1.3232854135514217e-05, + "loss": 0.0327, + "step": 25245 + }, + { + "epoch": 3.31, + "learning_rate": 1.3232352283560179e-05, + "loss": 0.0358, + "step": 25246 + }, + { + "epoch": 3.31, + "learning_rate": 1.323185042251532e-05, + "loss": 0.0434, + "step": 25247 + }, + { + "epoch": 3.31, + "learning_rate": 1.3231348552381054e-05, + "loss": 0.045, + "step": 25248 + }, + { + "epoch": 3.31, + "learning_rate": 1.323084667315879e-05, + "loss": 0.0339, + "step": 25249 + }, + { + "epoch": 3.31, + "learning_rate": 1.3230344784849946e-05, + "loss": 0.0344, + "step": 25250 + }, + { + "epoch": 3.31, + "learning_rate": 1.3229842887455926e-05, + "loss": 0.0428, + "step": 25251 + }, + { + "epoch": 3.31, + "learning_rate": 1.322934098097815e-05, + "loss": 0.0365, + "step": 25252 + }, + { + "epoch": 3.31, + "learning_rate": 1.3228839065418018e-05, + "loss": 0.0436, + "step": 25253 + }, + { + "epoch": 3.31, + "learning_rate": 1.322833714077695e-05, + "loss": 0.0347, + "step": 25254 + }, + { + "epoch": 3.31, + "learning_rate": 1.3227835207056357e-05, + "loss": 0.0465, + "step": 25255 + }, + { + "epoch": 3.31, + "learning_rate": 1.3227333264257649e-05, + "loss": 0.0381, + "step": 25256 + }, + { + "epoch": 3.31, + "learning_rate": 1.3226831312382237e-05, + "loss": 0.0383, + "step": 25257 + }, + { + "epoch": 3.31, + "learning_rate": 1.3226329351431536e-05, + "loss": 0.0421, + "step": 25258 + }, + { + "epoch": 3.31, + "learning_rate": 1.3225827381406954e-05, + "loss": 0.0374, + "step": 25259 + }, + { + "epoch": 3.31, + "learning_rate": 1.3225325402309905e-05, + "loss": 0.0384, + "step": 25260 + }, + { + "epoch": 3.31, + "learning_rate": 1.3224823414141798e-05, + "loss": 0.0363, + "step": 25261 + }, + { + "epoch": 3.31, + "learning_rate": 1.3224321416904049e-05, + "loss": 0.0465, + "step": 25262 + }, + { + "epoch": 3.31, + "learning_rate": 1.3223819410598066e-05, + "loss": 0.0457, + "step": 25263 + }, + { + "epoch": 3.31, + "learning_rate": 1.3223317395225266e-05, + "loss": 0.0444, + "step": 25264 + }, + { + "epoch": 3.31, + "learning_rate": 1.3222815370787054e-05, + "loss": 0.0435, + "step": 25265 + }, + { + "epoch": 3.31, + "learning_rate": 1.3222313337284844e-05, + "loss": 0.0433, + "step": 25266 + }, + { + "epoch": 3.31, + "learning_rate": 1.3221811294720055e-05, + "loss": 0.0433, + "step": 25267 + }, + { + "epoch": 3.31, + "learning_rate": 1.322130924309409e-05, + "loss": 0.0411, + "step": 25268 + }, + { + "epoch": 3.31, + "learning_rate": 1.3220807182408364e-05, + "loss": 0.04, + "step": 25269 + }, + { + "epoch": 3.31, + "learning_rate": 1.3220305112664291e-05, + "loss": 0.0432, + "step": 25270 + }, + { + "epoch": 3.31, + "learning_rate": 1.321980303386328e-05, + "loss": 0.0423, + "step": 25271 + }, + { + "epoch": 3.31, + "learning_rate": 1.3219300946006742e-05, + "loss": 0.0397, + "step": 25272 + }, + { + "epoch": 3.31, + "learning_rate": 1.3218798849096097e-05, + "loss": 0.0378, + "step": 25273 + }, + { + "epoch": 3.31, + "learning_rate": 1.3218296743132748e-05, + "loss": 0.0347, + "step": 25274 + }, + { + "epoch": 3.31, + "learning_rate": 1.3217794628118113e-05, + "loss": 0.0372, + "step": 25275 + }, + { + "epoch": 3.31, + "learning_rate": 1.3217292504053596e-05, + "loss": 0.0432, + "step": 25276 + }, + { + "epoch": 3.31, + "learning_rate": 1.3216790370940621e-05, + "loss": 0.0373, + "step": 25277 + }, + { + "epoch": 3.31, + "learning_rate": 1.3216288228780593e-05, + "loss": 0.0396, + "step": 25278 + }, + { + "epoch": 3.31, + "learning_rate": 1.3215786077574926e-05, + "loss": 0.0393, + "step": 25279 + }, + { + "epoch": 3.31, + "learning_rate": 1.3215283917325032e-05, + "loss": 0.0439, + "step": 25280 + }, + { + "epoch": 3.31, + "learning_rate": 1.3214781748032323e-05, + "loss": 0.0391, + "step": 25281 + }, + { + "epoch": 3.31, + "learning_rate": 1.3214279569698212e-05, + "loss": 0.0402, + "step": 25282 + }, + { + "epoch": 3.31, + "learning_rate": 1.3213777382324109e-05, + "loss": 0.036, + "step": 25283 + }, + { + "epoch": 3.31, + "learning_rate": 1.3213275185911432e-05, + "loss": 0.0358, + "step": 25284 + }, + { + "epoch": 3.31, + "learning_rate": 1.3212772980461585e-05, + "loss": 0.0406, + "step": 25285 + }, + { + "epoch": 3.31, + "learning_rate": 1.3212270765975989e-05, + "loss": 0.0428, + "step": 25286 + }, + { + "epoch": 3.31, + "learning_rate": 1.321176854245605e-05, + "loss": 0.042, + "step": 25287 + }, + { + "epoch": 3.31, + "learning_rate": 1.3211266309903184e-05, + "loss": 0.0453, + "step": 25288 + }, + { + "epoch": 3.31, + "learning_rate": 1.32107640683188e-05, + "loss": 0.0449, + "step": 25289 + }, + { + "epoch": 3.31, + "learning_rate": 1.3210261817704317e-05, + "loss": 0.0376, + "step": 25290 + }, + { + "epoch": 3.31, + "learning_rate": 1.3209759558061142e-05, + "loss": 0.0377, + "step": 25291 + }, + { + "epoch": 3.31, + "learning_rate": 1.3209257289390692e-05, + "loss": 0.0347, + "step": 25292 + }, + { + "epoch": 3.31, + "learning_rate": 1.320875501169437e-05, + "loss": 0.0459, + "step": 25293 + }, + { + "epoch": 3.31, + "learning_rate": 1.3208252724973601e-05, + "loss": 0.0406, + "step": 25294 + }, + { + "epoch": 3.31, + "learning_rate": 1.320775042922979e-05, + "loss": 0.0411, + "step": 25295 + }, + { + "epoch": 3.31, + "learning_rate": 1.3207248124464354e-05, + "loss": 0.0343, + "step": 25296 + }, + { + "epoch": 3.31, + "learning_rate": 1.3206745810678701e-05, + "loss": 0.0352, + "step": 25297 + }, + { + "epoch": 3.31, + "learning_rate": 1.3206243487874244e-05, + "loss": 0.0324, + "step": 25298 + }, + { + "epoch": 3.31, + "learning_rate": 1.32057411560524e-05, + "loss": 0.049, + "step": 25299 + }, + { + "epoch": 3.31, + "learning_rate": 1.3205238815214582e-05, + "loss": 0.0399, + "step": 25300 + }, + { + "epoch": 3.31, + "learning_rate": 1.32047364653622e-05, + "loss": 0.032, + "step": 25301 + }, + { + "epoch": 3.31, + "learning_rate": 1.3204234106496666e-05, + "loss": 0.0371, + "step": 25302 + }, + { + "epoch": 3.31, + "learning_rate": 1.3203731738619395e-05, + "loss": 0.0358, + "step": 25303 + }, + { + "epoch": 3.31, + "learning_rate": 1.3203229361731795e-05, + "loss": 0.0426, + "step": 25304 + }, + { + "epoch": 3.31, + "learning_rate": 1.3202726975835289e-05, + "loss": 0.0486, + "step": 25305 + }, + { + "epoch": 3.31, + "learning_rate": 1.320222458093128e-05, + "loss": 0.035, + "step": 25306 + }, + { + "epoch": 3.32, + "learning_rate": 1.3201722177021188e-05, + "loss": 0.0315, + "step": 25307 + }, + { + "epoch": 3.32, + "learning_rate": 1.3201219764106421e-05, + "loss": 0.0447, + "step": 25308 + }, + { + "epoch": 3.32, + "learning_rate": 1.3200717342188394e-05, + "loss": 0.0355, + "step": 25309 + }, + { + "epoch": 3.32, + "learning_rate": 1.3200214911268519e-05, + "loss": 0.0445, + "step": 25310 + }, + { + "epoch": 3.32, + "learning_rate": 1.3199712471348212e-05, + "loss": 0.0446, + "step": 25311 + }, + { + "epoch": 3.32, + "learning_rate": 1.3199210022428883e-05, + "loss": 0.043, + "step": 25312 + }, + { + "epoch": 3.32, + "learning_rate": 1.3198707564511947e-05, + "loss": 0.0458, + "step": 25313 + }, + { + "epoch": 3.32, + "learning_rate": 1.3198205097598814e-05, + "loss": 0.0415, + "step": 25314 + }, + { + "epoch": 3.32, + "learning_rate": 1.3197702621690902e-05, + "loss": 0.0459, + "step": 25315 + }, + { + "epoch": 3.32, + "learning_rate": 1.319720013678962e-05, + "loss": 0.0418, + "step": 25316 + }, + { + "epoch": 3.32, + "learning_rate": 1.3196697642896386e-05, + "loss": 0.0474, + "step": 25317 + }, + { + "epoch": 3.32, + "learning_rate": 1.3196195140012607e-05, + "loss": 0.0458, + "step": 25318 + }, + { + "epoch": 3.32, + "learning_rate": 1.3195692628139697e-05, + "loss": 0.0387, + "step": 25319 + }, + { + "epoch": 3.32, + "learning_rate": 1.3195190107279077e-05, + "loss": 0.0353, + "step": 25320 + }, + { + "epoch": 3.32, + "learning_rate": 1.3194687577432152e-05, + "loss": 0.0328, + "step": 25321 + }, + { + "epoch": 3.32, + "learning_rate": 1.319418503860034e-05, + "loss": 0.0392, + "step": 25322 + }, + { + "epoch": 3.32, + "learning_rate": 1.3193682490785052e-05, + "loss": 0.0448, + "step": 25323 + }, + { + "epoch": 3.32, + "learning_rate": 1.3193179933987703e-05, + "loss": 0.0404, + "step": 25324 + }, + { + "epoch": 3.32, + "learning_rate": 1.3192677368209703e-05, + "loss": 0.0471, + "step": 25325 + }, + { + "epoch": 3.32, + "learning_rate": 1.3192174793452468e-05, + "loss": 0.0398, + "step": 25326 + }, + { + "epoch": 3.32, + "learning_rate": 1.3191672209717412e-05, + "loss": 0.0426, + "step": 25327 + }, + { + "epoch": 3.32, + "learning_rate": 1.319116961700595e-05, + "loss": 0.0423, + "step": 25328 + }, + { + "epoch": 3.32, + "learning_rate": 1.319066701531949e-05, + "loss": 0.0437, + "step": 25329 + }, + { + "epoch": 3.32, + "learning_rate": 1.3190164404659453e-05, + "loss": 0.0422, + "step": 25330 + }, + { + "epoch": 3.32, + "learning_rate": 1.3189661785027244e-05, + "loss": 0.0438, + "step": 25331 + }, + { + "epoch": 3.32, + "learning_rate": 1.3189159156424285e-05, + "loss": 0.0467, + "step": 25332 + }, + { + "epoch": 3.32, + "learning_rate": 1.3188656518851983e-05, + "loss": 0.0433, + "step": 25333 + }, + { + "epoch": 3.32, + "learning_rate": 1.3188153872311758e-05, + "loss": 0.0373, + "step": 25334 + }, + { + "epoch": 3.32, + "learning_rate": 1.3187651216805017e-05, + "loss": 0.0386, + "step": 25335 + }, + { + "epoch": 3.32, + "learning_rate": 1.3187148552333176e-05, + "loss": 0.0416, + "step": 25336 + }, + { + "epoch": 3.32, + "learning_rate": 1.318664587889765e-05, + "loss": 0.0438, + "step": 25337 + }, + { + "epoch": 3.32, + "learning_rate": 1.3186143196499855e-05, + "loss": 0.0345, + "step": 25338 + }, + { + "epoch": 3.32, + "learning_rate": 1.3185640505141199e-05, + "loss": 0.0422, + "step": 25339 + }, + { + "epoch": 3.32, + "learning_rate": 1.3185137804823098e-05, + "loss": 0.0401, + "step": 25340 + }, + { + "epoch": 3.32, + "learning_rate": 1.318463509554697e-05, + "loss": 0.0402, + "step": 25341 + }, + { + "epoch": 3.32, + "learning_rate": 1.3184132377314224e-05, + "loss": 0.0422, + "step": 25342 + }, + { + "epoch": 3.32, + "learning_rate": 1.3183629650126272e-05, + "loss": 0.0421, + "step": 25343 + }, + { + "epoch": 3.32, + "learning_rate": 1.3183126913984534e-05, + "loss": 0.0416, + "step": 25344 + }, + { + "epoch": 3.32, + "learning_rate": 1.3182624168890423e-05, + "loss": 0.0329, + "step": 25345 + }, + { + "epoch": 3.32, + "learning_rate": 1.3182121414845348e-05, + "loss": 0.0461, + "step": 25346 + }, + { + "epoch": 3.32, + "learning_rate": 1.3181618651850725e-05, + "loss": 0.039, + "step": 25347 + }, + { + "epoch": 3.32, + "learning_rate": 1.318111587990797e-05, + "loss": 0.0373, + "step": 25348 + }, + { + "epoch": 3.32, + "learning_rate": 1.3180613099018498e-05, + "loss": 0.0437, + "step": 25349 + }, + { + "epoch": 3.32, + "learning_rate": 1.3180110309183721e-05, + "loss": 0.047, + "step": 25350 + }, + { + "epoch": 3.32, + "learning_rate": 1.3179607510405053e-05, + "loss": 0.0372, + "step": 25351 + }, + { + "epoch": 3.32, + "learning_rate": 1.3179104702683905e-05, + "loss": 0.035, + "step": 25352 + }, + { + "epoch": 3.32, + "learning_rate": 1.3178601886021697e-05, + "loss": 0.0413, + "step": 25353 + }, + { + "epoch": 3.32, + "learning_rate": 1.3178099060419838e-05, + "loss": 0.0397, + "step": 25354 + }, + { + "epoch": 3.32, + "learning_rate": 1.317759622587975e-05, + "loss": 0.0385, + "step": 25355 + }, + { + "epoch": 3.32, + "learning_rate": 1.3177093382402837e-05, + "loss": 0.0395, + "step": 25356 + }, + { + "epoch": 3.32, + "learning_rate": 1.317659052999052e-05, + "loss": 0.0403, + "step": 25357 + }, + { + "epoch": 3.32, + "learning_rate": 1.3176087668644209e-05, + "loss": 0.0411, + "step": 25358 + }, + { + "epoch": 3.32, + "learning_rate": 1.3175584798365323e-05, + "loss": 0.0374, + "step": 25359 + }, + { + "epoch": 3.32, + "learning_rate": 1.3175081919155272e-05, + "loss": 0.034, + "step": 25360 + }, + { + "epoch": 3.32, + "learning_rate": 1.3174579031015475e-05, + "loss": 0.0417, + "step": 25361 + }, + { + "epoch": 3.32, + "learning_rate": 1.3174076133947341e-05, + "loss": 0.0406, + "step": 25362 + }, + { + "epoch": 3.32, + "learning_rate": 1.3173573227952287e-05, + "loss": 0.0432, + "step": 25363 + }, + { + "epoch": 3.32, + "learning_rate": 1.3173070313031728e-05, + "loss": 0.0423, + "step": 25364 + }, + { + "epoch": 3.32, + "learning_rate": 1.3172567389187077e-05, + "loss": 0.039, + "step": 25365 + }, + { + "epoch": 3.32, + "learning_rate": 1.317206445641975e-05, + "loss": 0.036, + "step": 25366 + }, + { + "epoch": 3.32, + "learning_rate": 1.317156151473116e-05, + "loss": 0.0378, + "step": 25367 + }, + { + "epoch": 3.32, + "learning_rate": 1.3171058564122725e-05, + "loss": 0.0419, + "step": 25368 + }, + { + "epoch": 3.32, + "learning_rate": 1.3170555604595854e-05, + "loss": 0.0398, + "step": 25369 + }, + { + "epoch": 3.32, + "learning_rate": 1.3170052636151964e-05, + "loss": 0.0378, + "step": 25370 + }, + { + "epoch": 3.32, + "learning_rate": 1.316954965879247e-05, + "loss": 0.0379, + "step": 25371 + }, + { + "epoch": 3.32, + "learning_rate": 1.3169046672518788e-05, + "loss": 0.0415, + "step": 25372 + }, + { + "epoch": 3.32, + "learning_rate": 1.3168543677332331e-05, + "loss": 0.0392, + "step": 25373 + }, + { + "epoch": 3.32, + "learning_rate": 1.3168040673234512e-05, + "loss": 0.0442, + "step": 25374 + }, + { + "epoch": 3.32, + "learning_rate": 1.3167537660226746e-05, + "loss": 0.0406, + "step": 25375 + }, + { + "epoch": 3.32, + "learning_rate": 1.3167034638310452e-05, + "loss": 0.0461, + "step": 25376 + }, + { + "epoch": 3.32, + "learning_rate": 1.316653160748704e-05, + "loss": 0.0317, + "step": 25377 + }, + { + "epoch": 3.32, + "learning_rate": 1.316602856775793e-05, + "loss": 0.0434, + "step": 25378 + }, + { + "epoch": 3.32, + "learning_rate": 1.316552551912453e-05, + "loss": 0.043, + "step": 25379 + }, + { + "epoch": 3.32, + "learning_rate": 1.3165022461588255e-05, + "loss": 0.0481, + "step": 25380 + }, + { + "epoch": 3.32, + "learning_rate": 1.3164519395150525e-05, + "loss": 0.0512, + "step": 25381 + }, + { + "epoch": 3.32, + "learning_rate": 1.3164016319812756e-05, + "loss": 0.0438, + "step": 25382 + }, + { + "epoch": 3.32, + "learning_rate": 1.3163513235576357e-05, + "loss": 0.0366, + "step": 25383 + }, + { + "epoch": 3.33, + "learning_rate": 1.3163010142442746e-05, + "loss": 0.037, + "step": 25384 + }, + { + "epoch": 3.33, + "learning_rate": 1.316250704041334e-05, + "loss": 0.0386, + "step": 25385 + }, + { + "epoch": 3.33, + "learning_rate": 1.3162003929489547e-05, + "loss": 0.0409, + "step": 25386 + }, + { + "epoch": 3.33, + "learning_rate": 1.3161500809672786e-05, + "loss": 0.0444, + "step": 25387 + }, + { + "epoch": 3.33, + "learning_rate": 1.3160997680964474e-05, + "loss": 0.0398, + "step": 25388 + }, + { + "epoch": 3.33, + "learning_rate": 1.3160494543366027e-05, + "loss": 0.0427, + "step": 25389 + }, + { + "epoch": 3.33, + "learning_rate": 1.3159991396878856e-05, + "loss": 0.0299, + "step": 25390 + }, + { + "epoch": 3.33, + "learning_rate": 1.3159488241504374e-05, + "loss": 0.0355, + "step": 25391 + }, + { + "epoch": 3.33, + "learning_rate": 1.3158985077244002e-05, + "loss": 0.0433, + "step": 25392 + }, + { + "epoch": 3.33, + "learning_rate": 1.3158481904099154e-05, + "loss": 0.0381, + "step": 25393 + }, + { + "epoch": 3.33, + "learning_rate": 1.3157978722071244e-05, + "loss": 0.038, + "step": 25394 + }, + { + "epoch": 3.33, + "learning_rate": 1.3157475531161685e-05, + "loss": 0.0453, + "step": 25395 + }, + { + "epoch": 3.33, + "learning_rate": 1.3156972331371894e-05, + "loss": 0.0408, + "step": 25396 + }, + { + "epoch": 3.33, + "learning_rate": 1.3156469122703289e-05, + "loss": 0.0396, + "step": 25397 + }, + { + "epoch": 3.33, + "learning_rate": 1.3155965905157278e-05, + "loss": 0.0352, + "step": 25398 + }, + { + "epoch": 3.33, + "learning_rate": 1.3155462678735287e-05, + "loss": 0.0346, + "step": 25399 + }, + { + "epoch": 3.33, + "learning_rate": 1.3154959443438719e-05, + "loss": 0.0348, + "step": 25400 + }, + { + "epoch": 3.33, + "learning_rate": 1.3154456199269e-05, + "loss": 0.0438, + "step": 25401 + }, + { + "epoch": 3.33, + "learning_rate": 1.315395294622754e-05, + "loss": 0.0358, + "step": 25402 + }, + { + "epoch": 3.33, + "learning_rate": 1.3153449684315756e-05, + "loss": 0.04, + "step": 25403 + }, + { + "epoch": 3.33, + "learning_rate": 1.3152946413535061e-05, + "loss": 0.0311, + "step": 25404 + }, + { + "epoch": 3.33, + "learning_rate": 1.3152443133886872e-05, + "loss": 0.037, + "step": 25405 + }, + { + "epoch": 3.33, + "learning_rate": 1.3151939845372607e-05, + "loss": 0.0364, + "step": 25406 + }, + { + "epoch": 3.33, + "learning_rate": 1.3151436547993677e-05, + "loss": 0.0431, + "step": 25407 + }, + { + "epoch": 3.33, + "learning_rate": 1.31509332417515e-05, + "loss": 0.0416, + "step": 25408 + }, + { + "epoch": 3.33, + "learning_rate": 1.315042992664749e-05, + "loss": 0.045, + "step": 25409 + }, + { + "epoch": 3.33, + "learning_rate": 1.3149926602683065e-05, + "loss": 0.0369, + "step": 25410 + }, + { + "epoch": 3.33, + "learning_rate": 1.314942326985964e-05, + "loss": 0.0407, + "step": 25411 + }, + { + "epoch": 3.33, + "learning_rate": 1.314891992817863e-05, + "loss": 0.0355, + "step": 25412 + }, + { + "epoch": 3.33, + "learning_rate": 1.3148416577641445e-05, + "loss": 0.0444, + "step": 25413 + }, + { + "epoch": 3.33, + "learning_rate": 1.3147913218249513e-05, + "loss": 0.0393, + "step": 25414 + }, + { + "epoch": 3.33, + "learning_rate": 1.3147409850004237e-05, + "loss": 0.0421, + "step": 25415 + }, + { + "epoch": 3.33, + "learning_rate": 1.3146906472907044e-05, + "loss": 0.0466, + "step": 25416 + }, + { + "epoch": 3.33, + "learning_rate": 1.3146403086959341e-05, + "loss": 0.0417, + "step": 25417 + }, + { + "epoch": 3.33, + "learning_rate": 1.3145899692162547e-05, + "loss": 0.034, + "step": 25418 + }, + { + "epoch": 3.33, + "learning_rate": 1.3145396288518075e-05, + "loss": 0.0375, + "step": 25419 + }, + { + "epoch": 3.33, + "learning_rate": 1.3144892876027349e-05, + "loss": 0.0405, + "step": 25420 + }, + { + "epoch": 3.33, + "learning_rate": 1.3144389454691774e-05, + "loss": 0.0435, + "step": 25421 + }, + { + "epoch": 3.33, + "learning_rate": 1.3143886024512777e-05, + "loss": 0.0351, + "step": 25422 + }, + { + "epoch": 3.33, + "learning_rate": 1.3143382585491765e-05, + "loss": 0.0412, + "step": 25423 + }, + { + "epoch": 3.33, + "learning_rate": 1.3142879137630154e-05, + "loss": 0.0426, + "step": 25424 + }, + { + "epoch": 3.33, + "learning_rate": 1.3142375680929364e-05, + "loss": 0.0393, + "step": 25425 + }, + { + "epoch": 3.33, + "learning_rate": 1.3141872215390814e-05, + "loss": 0.0384, + "step": 25426 + }, + { + "epoch": 3.33, + "learning_rate": 1.3141368741015912e-05, + "loss": 0.0478, + "step": 25427 + }, + { + "epoch": 3.33, + "learning_rate": 1.3140865257806078e-05, + "loss": 0.0354, + "step": 25428 + }, + { + "epoch": 3.33, + "learning_rate": 1.3140361765762727e-05, + "loss": 0.0426, + "step": 25429 + }, + { + "epoch": 3.33, + "learning_rate": 1.3139858264887275e-05, + "loss": 0.0434, + "step": 25430 + }, + { + "epoch": 3.33, + "learning_rate": 1.3139354755181144e-05, + "loss": 0.0408, + "step": 25431 + }, + { + "epoch": 3.33, + "learning_rate": 1.313885123664574e-05, + "loss": 0.051, + "step": 25432 + }, + { + "epoch": 3.33, + "learning_rate": 1.3138347709282489e-05, + "loss": 0.0385, + "step": 25433 + }, + { + "epoch": 3.33, + "learning_rate": 1.3137844173092797e-05, + "loss": 0.0356, + "step": 25434 + }, + { + "epoch": 3.33, + "learning_rate": 1.3137340628078086e-05, + "loss": 0.0373, + "step": 25435 + }, + { + "epoch": 3.33, + "learning_rate": 1.3136837074239772e-05, + "loss": 0.0401, + "step": 25436 + }, + { + "epoch": 3.33, + "learning_rate": 1.3136333511579274e-05, + "loss": 0.038, + "step": 25437 + }, + { + "epoch": 3.33, + "learning_rate": 1.3135829940098002e-05, + "loss": 0.0431, + "step": 25438 + }, + { + "epoch": 3.33, + "learning_rate": 1.3135326359797377e-05, + "loss": 0.0347, + "step": 25439 + }, + { + "epoch": 3.33, + "learning_rate": 1.313482277067881e-05, + "loss": 0.0437, + "step": 25440 + }, + { + "epoch": 3.33, + "learning_rate": 1.3134319172743726e-05, + "loss": 0.0435, + "step": 25441 + }, + { + "epoch": 3.33, + "learning_rate": 1.3133815565993531e-05, + "loss": 0.0435, + "step": 25442 + }, + { + "epoch": 3.33, + "learning_rate": 1.3133311950429653e-05, + "loss": 0.0385, + "step": 25443 + }, + { + "epoch": 3.33, + "learning_rate": 1.3132808326053496e-05, + "loss": 0.0424, + "step": 25444 + }, + { + "epoch": 3.33, + "learning_rate": 1.3132304692866485e-05, + "loss": 0.039, + "step": 25445 + }, + { + "epoch": 3.33, + "learning_rate": 1.3131801050870032e-05, + "loss": 0.045, + "step": 25446 + }, + { + "epoch": 3.33, + "learning_rate": 1.313129740006556e-05, + "loss": 0.0331, + "step": 25447 + }, + { + "epoch": 3.33, + "learning_rate": 1.3130793740454475e-05, + "loss": 0.0492, + "step": 25448 + }, + { + "epoch": 3.33, + "learning_rate": 1.31302900720382e-05, + "loss": 0.0341, + "step": 25449 + }, + { + "epoch": 3.33, + "learning_rate": 1.3129786394818156e-05, + "loss": 0.0478, + "step": 25450 + }, + { + "epoch": 3.33, + "learning_rate": 1.3129282708795751e-05, + "loss": 0.0395, + "step": 25451 + }, + { + "epoch": 3.33, + "learning_rate": 1.3128779013972404e-05, + "loss": 0.044, + "step": 25452 + }, + { + "epoch": 3.33, + "learning_rate": 1.3128275310349534e-05, + "loss": 0.0354, + "step": 25453 + }, + { + "epoch": 3.33, + "learning_rate": 1.3127771597928557e-05, + "loss": 0.048, + "step": 25454 + }, + { + "epoch": 3.33, + "learning_rate": 1.3127267876710888e-05, + "loss": 0.0424, + "step": 25455 + }, + { + "epoch": 3.33, + "learning_rate": 1.3126764146697944e-05, + "loss": 0.0461, + "step": 25456 + }, + { + "epoch": 3.33, + "learning_rate": 1.3126260407891142e-05, + "loss": 0.0285, + "step": 25457 + }, + { + "epoch": 3.33, + "learning_rate": 1.3125756660291902e-05, + "loss": 0.0401, + "step": 25458 + }, + { + "epoch": 3.33, + "learning_rate": 1.3125252903901635e-05, + "loss": 0.0412, + "step": 25459 + }, + { + "epoch": 3.34, + "learning_rate": 1.3124749138721764e-05, + "loss": 0.0384, + "step": 25460 + }, + { + "epoch": 3.34, + "learning_rate": 1.31242453647537e-05, + "loss": 0.0432, + "step": 25461 + }, + { + "epoch": 3.34, + "learning_rate": 1.312374158199886e-05, + "loss": 0.0365, + "step": 25462 + }, + { + "epoch": 3.34, + "learning_rate": 1.3123237790458668e-05, + "loss": 0.0421, + "step": 25463 + }, + { + "epoch": 3.34, + "learning_rate": 1.3122733990134533e-05, + "loss": 0.0374, + "step": 25464 + }, + { + "epoch": 3.34, + "learning_rate": 1.3122230181027875e-05, + "loss": 0.0424, + "step": 25465 + }, + { + "epoch": 3.34, + "learning_rate": 1.3121726363140112e-05, + "loss": 0.0352, + "step": 25466 + }, + { + "epoch": 3.34, + "learning_rate": 1.3121222536472658e-05, + "loss": 0.0391, + "step": 25467 + }, + { + "epoch": 3.34, + "learning_rate": 1.3120718701026936e-05, + "loss": 0.0414, + "step": 25468 + }, + { + "epoch": 3.34, + "learning_rate": 1.3120214856804353e-05, + "loss": 0.0385, + "step": 25469 + }, + { + "epoch": 3.34, + "learning_rate": 1.3119711003806335e-05, + "loss": 0.0366, + "step": 25470 + }, + { + "epoch": 3.34, + "learning_rate": 1.3119207142034298e-05, + "loss": 0.0353, + "step": 25471 + }, + { + "epoch": 3.34, + "learning_rate": 1.3118703271489652e-05, + "loss": 0.0473, + "step": 25472 + }, + { + "epoch": 3.34, + "learning_rate": 1.3118199392173823e-05, + "loss": 0.0373, + "step": 25473 + }, + { + "epoch": 3.34, + "learning_rate": 1.3117695504088222e-05, + "loss": 0.0405, + "step": 25474 + }, + { + "epoch": 3.34, + "learning_rate": 1.3117191607234272e-05, + "loss": 0.0389, + "step": 25475 + }, + { + "epoch": 3.34, + "learning_rate": 1.3116687701613384e-05, + "loss": 0.0373, + "step": 25476 + }, + { + "epoch": 3.34, + "learning_rate": 1.311618378722698e-05, + "loss": 0.0394, + "step": 25477 + }, + { + "epoch": 3.34, + "learning_rate": 1.311567986407647e-05, + "loss": 0.0451, + "step": 25478 + }, + { + "epoch": 3.34, + "learning_rate": 1.311517593216328e-05, + "loss": 0.0346, + "step": 25479 + }, + { + "epoch": 3.34, + "learning_rate": 1.3114671991488823e-05, + "loss": 0.0472, + "step": 25480 + }, + { + "epoch": 3.34, + "learning_rate": 1.3114168042054518e-05, + "loss": 0.0434, + "step": 25481 + }, + { + "epoch": 3.34, + "learning_rate": 1.3113664083861779e-05, + "loss": 0.0401, + "step": 25482 + }, + { + "epoch": 3.34, + "learning_rate": 1.311316011691203e-05, + "loss": 0.0391, + "step": 25483 + }, + { + "epoch": 3.34, + "learning_rate": 1.3112656141206676e-05, + "loss": 0.0491, + "step": 25484 + }, + { + "epoch": 3.34, + "learning_rate": 1.3112152156747148e-05, + "loss": 0.0465, + "step": 25485 + }, + { + "epoch": 3.34, + "learning_rate": 1.3111648163534855e-05, + "loss": 0.0458, + "step": 25486 + }, + { + "epoch": 3.34, + "learning_rate": 1.3111144161571224e-05, + "loss": 0.0367, + "step": 25487 + }, + { + "epoch": 3.34, + "learning_rate": 1.311064015085766e-05, + "loss": 0.0343, + "step": 25488 + }, + { + "epoch": 3.34, + "learning_rate": 1.3110136131395585e-05, + "loss": 0.0404, + "step": 25489 + }, + { + "epoch": 3.34, + "learning_rate": 1.3109632103186417e-05, + "loss": 0.0434, + "step": 25490 + }, + { + "epoch": 3.34, + "learning_rate": 1.3109128066231581e-05, + "loss": 0.0501, + "step": 25491 + }, + { + "epoch": 3.34, + "learning_rate": 1.3108624020532483e-05, + "loss": 0.0478, + "step": 25492 + }, + { + "epoch": 3.34, + "learning_rate": 1.3108119966090548e-05, + "loss": 0.0464, + "step": 25493 + }, + { + "epoch": 3.34, + "learning_rate": 1.310761590290719e-05, + "loss": 0.0439, + "step": 25494 + }, + { + "epoch": 3.34, + "learning_rate": 1.3107111830983823e-05, + "loss": 0.037, + "step": 25495 + }, + { + "epoch": 3.34, + "learning_rate": 1.3106607750321876e-05, + "loss": 0.0316, + "step": 25496 + }, + { + "epoch": 3.34, + "learning_rate": 1.3106103660922758e-05, + "loss": 0.0368, + "step": 25497 + }, + { + "epoch": 3.34, + "learning_rate": 1.3105599562787887e-05, + "loss": 0.0316, + "step": 25498 + }, + { + "epoch": 3.34, + "learning_rate": 1.3105095455918686e-05, + "loss": 0.0356, + "step": 25499 + }, + { + "epoch": 3.34, + "learning_rate": 1.3104591340316566e-05, + "loss": 0.0382, + "step": 25500 + }, + { + "epoch": 3.34, + "learning_rate": 1.310408721598295e-05, + "loss": 0.0413, + "step": 25501 + }, + { + "epoch": 3.34, + "learning_rate": 1.3103583082919255e-05, + "loss": 0.0376, + "step": 25502 + }, + { + "epoch": 3.34, + "learning_rate": 1.3103078941126896e-05, + "loss": 0.0436, + "step": 25503 + }, + { + "epoch": 3.34, + "learning_rate": 1.3102574790607295e-05, + "loss": 0.0427, + "step": 25504 + }, + { + "epoch": 3.34, + "learning_rate": 1.3102070631361865e-05, + "loss": 0.0415, + "step": 25505 + }, + { + "epoch": 3.34, + "learning_rate": 1.3101566463392029e-05, + "loss": 0.0443, + "step": 25506 + }, + { + "epoch": 3.34, + "learning_rate": 1.31010622866992e-05, + "loss": 0.0414, + "step": 25507 + }, + { + "epoch": 3.34, + "learning_rate": 1.3100558101284801e-05, + "loss": 0.0401, + "step": 25508 + }, + { + "epoch": 3.34, + "learning_rate": 1.3100053907150245e-05, + "loss": 0.0476, + "step": 25509 + }, + { + "epoch": 3.34, + "learning_rate": 1.3099549704296952e-05, + "loss": 0.0454, + "step": 25510 + }, + { + "epoch": 3.34, + "learning_rate": 1.309904549272634e-05, + "loss": 0.0418, + "step": 25511 + }, + { + "epoch": 3.34, + "learning_rate": 1.3098541272439834e-05, + "loss": 0.0431, + "step": 25512 + }, + { + "epoch": 3.34, + "learning_rate": 1.309803704343884e-05, + "loss": 0.0361, + "step": 25513 + }, + { + "epoch": 3.34, + "learning_rate": 1.3097532805724782e-05, + "loss": 0.0408, + "step": 25514 + }, + { + "epoch": 3.34, + "learning_rate": 1.309702855929908e-05, + "loss": 0.0331, + "step": 25515 + }, + { + "epoch": 3.34, + "learning_rate": 1.309652430416315e-05, + "loss": 0.0297, + "step": 25516 + }, + { + "epoch": 3.34, + "learning_rate": 1.3096020040318408e-05, + "loss": 0.0439, + "step": 25517 + }, + { + "epoch": 3.34, + "learning_rate": 1.3095515767766275e-05, + "loss": 0.0428, + "step": 25518 + }, + { + "epoch": 3.34, + "learning_rate": 1.3095011486508172e-05, + "loss": 0.0413, + "step": 25519 + }, + { + "epoch": 3.34, + "learning_rate": 1.3094507196545512e-05, + "loss": 0.0371, + "step": 25520 + }, + { + "epoch": 3.34, + "learning_rate": 1.3094002897879716e-05, + "loss": 0.0336, + "step": 25521 + }, + { + "epoch": 3.34, + "learning_rate": 1.3093498590512198e-05, + "loss": 0.0385, + "step": 25522 + }, + { + "epoch": 3.34, + "learning_rate": 1.3092994274444384e-05, + "loss": 0.0405, + "step": 25523 + }, + { + "epoch": 3.34, + "learning_rate": 1.3092489949677684e-05, + "loss": 0.0421, + "step": 25524 + }, + { + "epoch": 3.34, + "learning_rate": 1.3091985616213527e-05, + "loss": 0.0427, + "step": 25525 + }, + { + "epoch": 3.34, + "learning_rate": 1.3091481274053322e-05, + "loss": 0.0497, + "step": 25526 + }, + { + "epoch": 3.34, + "learning_rate": 1.3090976923198489e-05, + "loss": 0.0363, + "step": 25527 + }, + { + "epoch": 3.34, + "learning_rate": 1.309047256365045e-05, + "loss": 0.0462, + "step": 25528 + }, + { + "epoch": 3.34, + "learning_rate": 1.308996819541062e-05, + "loss": 0.0425, + "step": 25529 + }, + { + "epoch": 3.34, + "learning_rate": 1.308946381848042e-05, + "loss": 0.0423, + "step": 25530 + }, + { + "epoch": 3.34, + "learning_rate": 1.3088959432861267e-05, + "loss": 0.0406, + "step": 25531 + }, + { + "epoch": 3.34, + "learning_rate": 1.3088455038554582e-05, + "loss": 0.0425, + "step": 25532 + }, + { + "epoch": 3.34, + "learning_rate": 1.3087950635561779e-05, + "loss": 0.0337, + "step": 25533 + }, + { + "epoch": 3.34, + "learning_rate": 1.308744622388428e-05, + "loss": 0.0406, + "step": 25534 + }, + { + "epoch": 3.34, + "learning_rate": 1.3086941803523503e-05, + "loss": 0.0361, + "step": 25535 + }, + { + "epoch": 3.35, + "learning_rate": 1.3086437374480867e-05, + "loss": 0.0317, + "step": 25536 + }, + { + "epoch": 3.35, + "learning_rate": 1.308593293675779e-05, + "loss": 0.0459, + "step": 25537 + }, + { + "epoch": 3.35, + "learning_rate": 1.3085428490355691e-05, + "loss": 0.0412, + "step": 25538 + }, + { + "epoch": 3.35, + "learning_rate": 1.3084924035275988e-05, + "loss": 0.0446, + "step": 25539 + }, + { + "epoch": 3.35, + "learning_rate": 1.3084419571520102e-05, + "loss": 0.0311, + "step": 25540 + }, + { + "epoch": 3.35, + "learning_rate": 1.3083915099089448e-05, + "loss": 0.0354, + "step": 25541 + }, + { + "epoch": 3.35, + "learning_rate": 1.308341061798545e-05, + "loss": 0.0392, + "step": 25542 + }, + { + "epoch": 3.35, + "learning_rate": 1.3082906128209522e-05, + "loss": 0.0311, + "step": 25543 + }, + { + "epoch": 3.35, + "learning_rate": 1.3082401629763084e-05, + "loss": 0.0409, + "step": 25544 + }, + { + "epoch": 3.35, + "learning_rate": 1.3081897122647558e-05, + "loss": 0.0447, + "step": 25545 + }, + { + "epoch": 3.35, + "learning_rate": 1.3081392606864358e-05, + "loss": 0.0373, + "step": 25546 + }, + { + "epoch": 3.35, + "learning_rate": 1.3080888082414906e-05, + "loss": 0.0407, + "step": 25547 + }, + { + "epoch": 3.35, + "learning_rate": 1.3080383549300622e-05, + "loss": 0.0478, + "step": 25548 + }, + { + "epoch": 3.35, + "learning_rate": 1.307987900752292e-05, + "loss": 0.0397, + "step": 25549 + }, + { + "epoch": 3.35, + "learning_rate": 1.3079374457083225e-05, + "loss": 0.0427, + "step": 25550 + }, + { + "epoch": 3.35, + "learning_rate": 1.3078869897982953e-05, + "loss": 0.046, + "step": 25551 + }, + { + "epoch": 3.35, + "learning_rate": 1.3078365330223524e-05, + "loss": 0.0361, + "step": 25552 + }, + { + "epoch": 3.35, + "learning_rate": 1.3077860753806356e-05, + "loss": 0.041, + "step": 25553 + }, + { + "epoch": 3.35, + "learning_rate": 1.3077356168732867e-05, + "loss": 0.0387, + "step": 25554 + }, + { + "epoch": 3.35, + "learning_rate": 1.3076851575004477e-05, + "loss": 0.0341, + "step": 25555 + }, + { + "epoch": 3.35, + "learning_rate": 1.3076346972622607e-05, + "loss": 0.0401, + "step": 25556 + }, + { + "epoch": 3.35, + "learning_rate": 1.3075842361588676e-05, + "loss": 0.0407, + "step": 25557 + }, + { + "epoch": 3.35, + "learning_rate": 1.3075337741904101e-05, + "loss": 0.0366, + "step": 25558 + }, + { + "epoch": 3.35, + "learning_rate": 1.3074833113570304e-05, + "loss": 0.046, + "step": 25559 + }, + { + "epoch": 3.35, + "learning_rate": 1.30743284765887e-05, + "loss": 0.0432, + "step": 25560 + }, + { + "epoch": 3.35, + "learning_rate": 1.3073823830960715e-05, + "loss": 0.0414, + "step": 25561 + }, + { + "epoch": 3.35, + "learning_rate": 1.307331917668776e-05, + "loss": 0.0441, + "step": 25562 + }, + { + "epoch": 3.35, + "learning_rate": 1.307281451377126e-05, + "loss": 0.0328, + "step": 25563 + }, + { + "epoch": 3.35, + "learning_rate": 1.3072309842212633e-05, + "loss": 0.0361, + "step": 25564 + }, + { + "epoch": 3.35, + "learning_rate": 1.30718051620133e-05, + "loss": 0.0381, + "step": 25565 + }, + { + "epoch": 3.35, + "learning_rate": 1.3071300473174673e-05, + "loss": 0.0426, + "step": 25566 + }, + { + "epoch": 3.35, + "learning_rate": 1.3070795775698182e-05, + "loss": 0.0333, + "step": 25567 + }, + { + "epoch": 3.35, + "learning_rate": 1.3070291069585238e-05, + "loss": 0.0401, + "step": 25568 + }, + { + "epoch": 3.35, + "learning_rate": 1.3069786354837268e-05, + "loss": 0.0369, + "step": 25569 + }, + { + "epoch": 3.35, + "learning_rate": 1.3069281631455686e-05, + "loss": 0.0388, + "step": 25570 + }, + { + "epoch": 3.35, + "learning_rate": 1.306877689944191e-05, + "loss": 0.0397, + "step": 25571 + }, + { + "epoch": 3.35, + "learning_rate": 1.3068272158797366e-05, + "loss": 0.0402, + "step": 25572 + }, + { + "epoch": 3.35, + "learning_rate": 1.306776740952347e-05, + "loss": 0.0367, + "step": 25573 + }, + { + "epoch": 3.35, + "learning_rate": 1.3067262651621637e-05, + "loss": 0.0401, + "step": 25574 + }, + { + "epoch": 3.35, + "learning_rate": 1.3066757885093293e-05, + "loss": 0.034, + "step": 25575 + }, + { + "epoch": 3.35, + "learning_rate": 1.3066253109939859e-05, + "loss": 0.0365, + "step": 25576 + }, + { + "epoch": 3.35, + "learning_rate": 1.3065748326162751e-05, + "loss": 0.0416, + "step": 25577 + }, + { + "epoch": 3.35, + "learning_rate": 1.3065243533763386e-05, + "loss": 0.0342, + "step": 25578 + }, + { + "epoch": 3.35, + "learning_rate": 1.3064738732743187e-05, + "loss": 0.0405, + "step": 25579 + }, + { + "epoch": 3.35, + "learning_rate": 1.3064233923103575e-05, + "loss": 0.0419, + "step": 25580 + }, + { + "epoch": 3.35, + "learning_rate": 1.3063729104845968e-05, + "loss": 0.0438, + "step": 25581 + }, + { + "epoch": 3.35, + "learning_rate": 1.3063224277971784e-05, + "loss": 0.0419, + "step": 25582 + }, + { + "epoch": 3.35, + "learning_rate": 1.3062719442482447e-05, + "loss": 0.0402, + "step": 25583 + }, + { + "epoch": 3.35, + "learning_rate": 1.3062214598379375e-05, + "loss": 0.0432, + "step": 25584 + }, + { + "epoch": 3.35, + "learning_rate": 1.3061709745663986e-05, + "loss": 0.0448, + "step": 25585 + }, + { + "epoch": 3.35, + "learning_rate": 1.3061204884337705e-05, + "loss": 0.0368, + "step": 25586 + }, + { + "epoch": 3.35, + "learning_rate": 1.3060700014401946e-05, + "loss": 0.0362, + "step": 25587 + }, + { + "epoch": 3.35, + "learning_rate": 1.3060195135858129e-05, + "loss": 0.0447, + "step": 25588 + }, + { + "epoch": 3.35, + "learning_rate": 1.3059690248707676e-05, + "loss": 0.0323, + "step": 25589 + }, + { + "epoch": 3.35, + "learning_rate": 1.3059185352952013e-05, + "loss": 0.0407, + "step": 25590 + }, + { + "epoch": 3.35, + "learning_rate": 1.3058680448592548e-05, + "loss": 0.0405, + "step": 25591 + }, + { + "epoch": 3.35, + "learning_rate": 1.3058175535630713e-05, + "loss": 0.0353, + "step": 25592 + }, + { + "epoch": 3.35, + "learning_rate": 1.3057670614067914e-05, + "loss": 0.0396, + "step": 25593 + }, + { + "epoch": 3.35, + "learning_rate": 1.3057165683905586e-05, + "loss": 0.0406, + "step": 25594 + }, + { + "epoch": 3.35, + "learning_rate": 1.305666074514514e-05, + "loss": 0.0359, + "step": 25595 + }, + { + "epoch": 3.35, + "learning_rate": 1.3056155797787997e-05, + "loss": 0.035, + "step": 25596 + }, + { + "epoch": 3.35, + "learning_rate": 1.3055650841835581e-05, + "loss": 0.0376, + "step": 25597 + }, + { + "epoch": 3.35, + "learning_rate": 1.3055145877289308e-05, + "loss": 0.0382, + "step": 25598 + }, + { + "epoch": 3.35, + "learning_rate": 1.3054640904150601e-05, + "loss": 0.0398, + "step": 25599 + }, + { + "epoch": 3.35, + "learning_rate": 1.3054135922420879e-05, + "loss": 0.0427, + "step": 25600 + }, + { + "epoch": 3.35, + "learning_rate": 1.3053630932101562e-05, + "loss": 0.0385, + "step": 25601 + }, + { + "epoch": 3.35, + "learning_rate": 1.3053125933194071e-05, + "loss": 0.0391, + "step": 25602 + }, + { + "epoch": 3.35, + "learning_rate": 1.3052620925699829e-05, + "loss": 0.0424, + "step": 25603 + }, + { + "epoch": 3.35, + "learning_rate": 1.3052115909620245e-05, + "loss": 0.041, + "step": 25604 + }, + { + "epoch": 3.35, + "learning_rate": 1.3051610884956753e-05, + "loss": 0.0366, + "step": 25605 + }, + { + "epoch": 3.35, + "learning_rate": 1.3051105851710768e-05, + "loss": 0.0354, + "step": 25606 + }, + { + "epoch": 3.35, + "learning_rate": 1.3050600809883712e-05, + "loss": 0.0355, + "step": 25607 + }, + { + "epoch": 3.35, + "learning_rate": 1.3050095759477e-05, + "loss": 0.0402, + "step": 25608 + }, + { + "epoch": 3.35, + "learning_rate": 1.3049590700492057e-05, + "loss": 0.0377, + "step": 25609 + }, + { + "epoch": 3.35, + "learning_rate": 1.3049085632930304e-05, + "loss": 0.0405, + "step": 25610 + }, + { + "epoch": 3.35, + "learning_rate": 1.3048580556793158e-05, + "loss": 0.0356, + "step": 25611 + }, + { + "epoch": 3.35, + "learning_rate": 1.3048075472082042e-05, + "loss": 0.0448, + "step": 25612 + }, + { + "epoch": 3.36, + "learning_rate": 1.3047570378798377e-05, + "loss": 0.0353, + "step": 25613 + }, + { + "epoch": 3.36, + "learning_rate": 1.3047065276943582e-05, + "loss": 0.0394, + "step": 25614 + }, + { + "epoch": 3.36, + "learning_rate": 1.3046560166519078e-05, + "loss": 0.043, + "step": 25615 + }, + { + "epoch": 3.36, + "learning_rate": 1.3046055047526284e-05, + "loss": 0.0462, + "step": 25616 + }, + { + "epoch": 3.36, + "learning_rate": 1.3045549919966627e-05, + "loss": 0.0451, + "step": 25617 + }, + { + "epoch": 3.36, + "learning_rate": 1.3045044783841521e-05, + "loss": 0.0407, + "step": 25618 + }, + { + "epoch": 3.36, + "learning_rate": 1.3044539639152387e-05, + "loss": 0.0447, + "step": 25619 + }, + { + "epoch": 3.36, + "learning_rate": 1.3044034485900647e-05, + "loss": 0.0463, + "step": 25620 + }, + { + "epoch": 3.36, + "learning_rate": 1.3043529324087725e-05, + "loss": 0.0367, + "step": 25621 + }, + { + "epoch": 3.36, + "learning_rate": 1.3043024153715037e-05, + "loss": 0.0414, + "step": 25622 + }, + { + "epoch": 3.36, + "learning_rate": 1.3042518974784007e-05, + "loss": 0.0401, + "step": 25623 + }, + { + "epoch": 3.36, + "learning_rate": 1.3042013787296056e-05, + "loss": 0.0358, + "step": 25624 + }, + { + "epoch": 3.36, + "learning_rate": 1.3041508591252598e-05, + "loss": 0.0519, + "step": 25625 + }, + { + "epoch": 3.36, + "learning_rate": 1.3041003386655063e-05, + "loss": 0.0344, + "step": 25626 + }, + { + "epoch": 3.36, + "learning_rate": 1.3040498173504865e-05, + "loss": 0.0392, + "step": 25627 + }, + { + "epoch": 3.36, + "learning_rate": 1.303999295180343e-05, + "loss": 0.0363, + "step": 25628 + }, + { + "epoch": 3.36, + "learning_rate": 1.3039487721552177e-05, + "loss": 0.0449, + "step": 25629 + }, + { + "epoch": 3.36, + "learning_rate": 1.3038982482752528e-05, + "loss": 0.0414, + "step": 25630 + }, + { + "epoch": 3.36, + "learning_rate": 1.3038477235405897e-05, + "loss": 0.046, + "step": 25631 + }, + { + "epoch": 3.36, + "learning_rate": 1.3037971979513717e-05, + "loss": 0.0388, + "step": 25632 + }, + { + "epoch": 3.36, + "learning_rate": 1.3037466715077398e-05, + "loss": 0.0391, + "step": 25633 + }, + { + "epoch": 3.36, + "learning_rate": 1.303696144209837e-05, + "loss": 0.045, + "step": 25634 + }, + { + "epoch": 3.36, + "learning_rate": 1.3036456160578046e-05, + "loss": 0.0336, + "step": 25635 + }, + { + "epoch": 3.36, + "learning_rate": 1.303595087051785e-05, + "loss": 0.0404, + "step": 25636 + }, + { + "epoch": 3.36, + "learning_rate": 1.3035445571919206e-05, + "loss": 0.0404, + "step": 25637 + }, + { + "epoch": 3.36, + "learning_rate": 1.3034940264783533e-05, + "loss": 0.0504, + "step": 25638 + }, + { + "epoch": 3.36, + "learning_rate": 1.3034434949112252e-05, + "loss": 0.0399, + "step": 25639 + }, + { + "epoch": 3.36, + "learning_rate": 1.3033929624906785e-05, + "loss": 0.0477, + "step": 25640 + }, + { + "epoch": 3.36, + "learning_rate": 1.3033424292168552e-05, + "loss": 0.042, + "step": 25641 + }, + { + "epoch": 3.36, + "learning_rate": 1.3032918950898974e-05, + "loss": 0.04, + "step": 25642 + }, + { + "epoch": 3.36, + "learning_rate": 1.3032413601099472e-05, + "loss": 0.0483, + "step": 25643 + }, + { + "epoch": 3.36, + "learning_rate": 1.3031908242771469e-05, + "loss": 0.0308, + "step": 25644 + }, + { + "epoch": 3.36, + "learning_rate": 1.3031402875916385e-05, + "loss": 0.0458, + "step": 25645 + }, + { + "epoch": 3.36, + "learning_rate": 1.3030897500535644e-05, + "loss": 0.0469, + "step": 25646 + }, + { + "epoch": 3.36, + "learning_rate": 1.3030392116630666e-05, + "loss": 0.0345, + "step": 25647 + }, + { + "epoch": 3.36, + "learning_rate": 1.3029886724202867e-05, + "loss": 0.0399, + "step": 25648 + }, + { + "epoch": 3.36, + "learning_rate": 1.3029381323253676e-05, + "loss": 0.0398, + "step": 25649 + }, + { + "epoch": 3.36, + "learning_rate": 1.3028875913784512e-05, + "loss": 0.0398, + "step": 25650 + }, + { + "epoch": 3.36, + "learning_rate": 1.3028370495796794e-05, + "loss": 0.0452, + "step": 25651 + }, + { + "epoch": 3.36, + "learning_rate": 1.3027865069291946e-05, + "loss": 0.0363, + "step": 25652 + }, + { + "epoch": 3.36, + "learning_rate": 1.3027359634271388e-05, + "loss": 0.0391, + "step": 25653 + }, + { + "epoch": 3.36, + "learning_rate": 1.3026854190736542e-05, + "loss": 0.0445, + "step": 25654 + }, + { + "epoch": 3.36, + "learning_rate": 1.3026348738688831e-05, + "loss": 0.0402, + "step": 25655 + }, + { + "epoch": 3.36, + "learning_rate": 1.3025843278129674e-05, + "loss": 0.0382, + "step": 25656 + }, + { + "epoch": 3.36, + "learning_rate": 1.3025337809060496e-05, + "loss": 0.0405, + "step": 25657 + }, + { + "epoch": 3.36, + "learning_rate": 1.3024832331482715e-05, + "loss": 0.0382, + "step": 25658 + }, + { + "epoch": 3.36, + "learning_rate": 1.3024326845397756e-05, + "loss": 0.0392, + "step": 25659 + }, + { + "epoch": 3.36, + "learning_rate": 1.3023821350807038e-05, + "loss": 0.0427, + "step": 25660 + }, + { + "epoch": 3.36, + "learning_rate": 1.3023315847711982e-05, + "loss": 0.0337, + "step": 25661 + }, + { + "epoch": 3.36, + "learning_rate": 1.3022810336114011e-05, + "loss": 0.0431, + "step": 25662 + }, + { + "epoch": 3.36, + "learning_rate": 1.3022304816014548e-05, + "loss": 0.0455, + "step": 25663 + }, + { + "epoch": 3.36, + "learning_rate": 1.3021799287415013e-05, + "loss": 0.0414, + "step": 25664 + }, + { + "epoch": 3.36, + "learning_rate": 1.302129375031683e-05, + "loss": 0.0403, + "step": 25665 + }, + { + "epoch": 3.36, + "learning_rate": 1.302078820472142e-05, + "loss": 0.038, + "step": 25666 + }, + { + "epoch": 3.36, + "learning_rate": 1.30202826506302e-05, + "loss": 0.0432, + "step": 25667 + }, + { + "epoch": 3.36, + "learning_rate": 1.30197770880446e-05, + "loss": 0.0445, + "step": 25668 + }, + { + "epoch": 3.36, + "learning_rate": 1.3019271516966034e-05, + "loss": 0.0479, + "step": 25669 + }, + { + "epoch": 3.36, + "learning_rate": 1.3018765937395929e-05, + "loss": 0.0377, + "step": 25670 + }, + { + "epoch": 3.36, + "learning_rate": 1.3018260349335708e-05, + "loss": 0.0407, + "step": 25671 + }, + { + "epoch": 3.36, + "learning_rate": 1.3017754752786788e-05, + "loss": 0.0479, + "step": 25672 + }, + { + "epoch": 3.36, + "learning_rate": 1.3017249147750595e-05, + "loss": 0.0421, + "step": 25673 + }, + { + "epoch": 3.36, + "learning_rate": 1.3016743534228548e-05, + "loss": 0.0415, + "step": 25674 + }, + { + "epoch": 3.36, + "learning_rate": 1.3016237912222069e-05, + "loss": 0.0397, + "step": 25675 + }, + { + "epoch": 3.36, + "learning_rate": 1.3015732281732585e-05, + "loss": 0.0343, + "step": 25676 + }, + { + "epoch": 3.36, + "learning_rate": 1.3015226642761514e-05, + "loss": 0.0429, + "step": 25677 + }, + { + "epoch": 3.36, + "learning_rate": 1.301472099531028e-05, + "loss": 0.045, + "step": 25678 + }, + { + "epoch": 3.36, + "learning_rate": 1.3014215339380299e-05, + "loss": 0.0475, + "step": 25679 + }, + { + "epoch": 3.36, + "learning_rate": 1.3013709674973e-05, + "loss": 0.039, + "step": 25680 + }, + { + "epoch": 3.36, + "learning_rate": 1.3013204002089803e-05, + "loss": 0.0393, + "step": 25681 + }, + { + "epoch": 3.36, + "learning_rate": 1.3012698320732128e-05, + "loss": 0.0368, + "step": 25682 + }, + { + "epoch": 3.36, + "learning_rate": 1.3012192630901405e-05, + "loss": 0.0413, + "step": 25683 + }, + { + "epoch": 3.36, + "learning_rate": 1.3011686932599046e-05, + "loss": 0.0358, + "step": 25684 + }, + { + "epoch": 3.36, + "learning_rate": 1.3011181225826481e-05, + "loss": 0.0401, + "step": 25685 + }, + { + "epoch": 3.36, + "learning_rate": 1.3010675510585125e-05, + "loss": 0.0339, + "step": 25686 + }, + { + "epoch": 3.36, + "learning_rate": 1.301016978687641e-05, + "loss": 0.0408, + "step": 25687 + }, + { + "epoch": 3.36, + "learning_rate": 1.3009664054701747e-05, + "loss": 0.0366, + "step": 25688 + }, + { + "epoch": 3.37, + "learning_rate": 1.3009158314062569e-05, + "loss": 0.0334, + "step": 25689 + }, + { + "epoch": 3.37, + "learning_rate": 1.3008652564960291e-05, + "loss": 0.0323, + "step": 25690 + }, + { + "epoch": 3.37, + "learning_rate": 1.3008146807396338e-05, + "loss": 0.0381, + "step": 25691 + }, + { + "epoch": 3.37, + "learning_rate": 1.3007641041372133e-05, + "loss": 0.0487, + "step": 25692 + }, + { + "epoch": 3.37, + "learning_rate": 1.3007135266889098e-05, + "loss": 0.0335, + "step": 25693 + }, + { + "epoch": 3.37, + "learning_rate": 1.3006629483948654e-05, + "loss": 0.0412, + "step": 25694 + }, + { + "epoch": 3.37, + "learning_rate": 1.3006123692552226e-05, + "loss": 0.0314, + "step": 25695 + }, + { + "epoch": 3.37, + "learning_rate": 1.3005617892701235e-05, + "loss": 0.0398, + "step": 25696 + }, + { + "epoch": 3.37, + "learning_rate": 1.3005112084397102e-05, + "loss": 0.037, + "step": 25697 + }, + { + "epoch": 3.37, + "learning_rate": 1.3004606267641251e-05, + "loss": 0.0347, + "step": 25698 + }, + { + "epoch": 3.37, + "learning_rate": 1.3004100442435108e-05, + "loss": 0.0471, + "step": 25699 + }, + { + "epoch": 3.37, + "learning_rate": 1.300359460878009e-05, + "loss": 0.0419, + "step": 25700 + }, + { + "epoch": 3.37, + "learning_rate": 1.3003088766677623e-05, + "loss": 0.0383, + "step": 25701 + }, + { + "epoch": 3.37, + "learning_rate": 1.300258291612913e-05, + "loss": 0.0455, + "step": 25702 + }, + { + "epoch": 3.37, + "learning_rate": 1.300207705713603e-05, + "loss": 0.0393, + "step": 25703 + }, + { + "epoch": 3.37, + "learning_rate": 1.3001571189699749e-05, + "loss": 0.0459, + "step": 25704 + }, + { + "epoch": 3.37, + "learning_rate": 1.3001065313821708e-05, + "loss": 0.0348, + "step": 25705 + }, + { + "epoch": 3.37, + "learning_rate": 1.3000559429503335e-05, + "loss": 0.0402, + "step": 25706 + }, + { + "epoch": 3.37, + "learning_rate": 1.3000053536746045e-05, + "loss": 0.0421, + "step": 25707 + }, + { + "epoch": 3.37, + "learning_rate": 1.2999547635551263e-05, + "loss": 0.0423, + "step": 25708 + }, + { + "epoch": 3.37, + "learning_rate": 1.2999041725920413e-05, + "loss": 0.0459, + "step": 25709 + }, + { + "epoch": 3.37, + "learning_rate": 1.2998535807854922e-05, + "loss": 0.0443, + "step": 25710 + }, + { + "epoch": 3.37, + "learning_rate": 1.2998029881356204e-05, + "loss": 0.0406, + "step": 25711 + }, + { + "epoch": 3.37, + "learning_rate": 1.2997523946425692e-05, + "loss": 0.0381, + "step": 25712 + }, + { + "epoch": 3.37, + "learning_rate": 1.2997018003064796e-05, + "loss": 0.0465, + "step": 25713 + }, + { + "epoch": 3.37, + "learning_rate": 1.2996512051274953e-05, + "loss": 0.0409, + "step": 25714 + }, + { + "epoch": 3.37, + "learning_rate": 1.2996006091057575e-05, + "loss": 0.0515, + "step": 25715 + }, + { + "epoch": 3.37, + "learning_rate": 1.2995500122414094e-05, + "loss": 0.0451, + "step": 25716 + }, + { + "epoch": 3.37, + "learning_rate": 1.2994994145345924e-05, + "loss": 0.0397, + "step": 25717 + }, + { + "epoch": 3.37, + "learning_rate": 1.2994488159854497e-05, + "loss": 0.0458, + "step": 25718 + }, + { + "epoch": 3.37, + "learning_rate": 1.2993982165941226e-05, + "loss": 0.0369, + "step": 25719 + }, + { + "epoch": 3.37, + "learning_rate": 1.2993476163607544e-05, + "loss": 0.0402, + "step": 25720 + }, + { + "epoch": 3.37, + "learning_rate": 1.2992970152854866e-05, + "loss": 0.0397, + "step": 25721 + }, + { + "epoch": 3.37, + "learning_rate": 1.2992464133684621e-05, + "loss": 0.0404, + "step": 25722 + }, + { + "epoch": 3.37, + "learning_rate": 1.2991958106098233e-05, + "loss": 0.0391, + "step": 25723 + }, + { + "epoch": 3.37, + "learning_rate": 1.2991452070097117e-05, + "loss": 0.0437, + "step": 25724 + }, + { + "epoch": 3.37, + "learning_rate": 1.2990946025682702e-05, + "loss": 0.0421, + "step": 25725 + }, + { + "epoch": 3.37, + "learning_rate": 1.2990439972856412e-05, + "loss": 0.0388, + "step": 25726 + }, + { + "epoch": 3.37, + "learning_rate": 1.2989933911619671e-05, + "loss": 0.0485, + "step": 25727 + }, + { + "epoch": 3.37, + "learning_rate": 1.2989427841973896e-05, + "loss": 0.0363, + "step": 25728 + }, + { + "epoch": 3.37, + "learning_rate": 1.2988921763920517e-05, + "loss": 0.0383, + "step": 25729 + }, + { + "epoch": 3.37, + "learning_rate": 1.2988415677460953e-05, + "loss": 0.0417, + "step": 25730 + }, + { + "epoch": 3.37, + "learning_rate": 1.2987909582596631e-05, + "loss": 0.0506, + "step": 25731 + }, + { + "epoch": 3.37, + "learning_rate": 1.2987403479328972e-05, + "loss": 0.0335, + "step": 25732 + }, + { + "epoch": 3.37, + "learning_rate": 1.2986897367659402e-05, + "loss": 0.0338, + "step": 25733 + }, + { + "epoch": 3.37, + "learning_rate": 1.2986391247589338e-05, + "loss": 0.0397, + "step": 25734 + }, + { + "epoch": 3.37, + "learning_rate": 1.2985885119120208e-05, + "loss": 0.0509, + "step": 25735 + }, + { + "epoch": 3.37, + "learning_rate": 1.2985378982253435e-05, + "loss": 0.0427, + "step": 25736 + }, + { + "epoch": 3.37, + "learning_rate": 1.2984872836990445e-05, + "loss": 0.0458, + "step": 25737 + }, + { + "epoch": 3.37, + "learning_rate": 1.2984366683332659e-05, + "loss": 0.0386, + "step": 25738 + }, + { + "epoch": 3.37, + "learning_rate": 1.2983860521281501e-05, + "loss": 0.0402, + "step": 25739 + }, + { + "epoch": 3.37, + "learning_rate": 1.2983354350838391e-05, + "loss": 0.0427, + "step": 25740 + }, + { + "epoch": 3.37, + "learning_rate": 1.2982848172004758e-05, + "loss": 0.0397, + "step": 25741 + }, + { + "epoch": 3.37, + "learning_rate": 1.2982341984782022e-05, + "loss": 0.0376, + "step": 25742 + }, + { + "epoch": 3.37, + "learning_rate": 1.298183578917161e-05, + "loss": 0.0435, + "step": 25743 + }, + { + "epoch": 3.37, + "learning_rate": 1.2981329585174943e-05, + "loss": 0.0327, + "step": 25744 + }, + { + "epoch": 3.37, + "learning_rate": 1.2980823372793443e-05, + "loss": 0.0381, + "step": 25745 + }, + { + "epoch": 3.37, + "learning_rate": 1.298031715202854e-05, + "loss": 0.0305, + "step": 25746 + }, + { + "epoch": 3.37, + "learning_rate": 1.2979810922881649e-05, + "loss": 0.0391, + "step": 25747 + }, + { + "epoch": 3.37, + "learning_rate": 1.2979304685354204e-05, + "loss": 0.0356, + "step": 25748 + }, + { + "epoch": 3.37, + "learning_rate": 1.2978798439447618e-05, + "loss": 0.0332, + "step": 25749 + }, + { + "epoch": 3.37, + "learning_rate": 1.2978292185163324e-05, + "loss": 0.0349, + "step": 25750 + }, + { + "epoch": 3.37, + "learning_rate": 1.2977785922502736e-05, + "loss": 0.0366, + "step": 25751 + }, + { + "epoch": 3.37, + "learning_rate": 1.297727965146729e-05, + "loss": 0.0409, + "step": 25752 + }, + { + "epoch": 3.37, + "learning_rate": 1.29767733720584e-05, + "loss": 0.0364, + "step": 25753 + }, + { + "epoch": 3.37, + "learning_rate": 1.2976267084277496e-05, + "loss": 0.0433, + "step": 25754 + }, + { + "epoch": 3.37, + "learning_rate": 1.2975760788126e-05, + "loss": 0.0381, + "step": 25755 + }, + { + "epoch": 3.37, + "learning_rate": 1.2975254483605333e-05, + "loss": 0.0422, + "step": 25756 + }, + { + "epoch": 3.37, + "learning_rate": 1.297474817071692e-05, + "loss": 0.0394, + "step": 25757 + }, + { + "epoch": 3.37, + "learning_rate": 1.2974241849462188e-05, + "loss": 0.0429, + "step": 25758 + }, + { + "epoch": 3.37, + "learning_rate": 1.2973735519842558e-05, + "loss": 0.04, + "step": 25759 + }, + { + "epoch": 3.37, + "learning_rate": 1.2973229181859459e-05, + "loss": 0.035, + "step": 25760 + }, + { + "epoch": 3.37, + "learning_rate": 1.297272283551431e-05, + "loss": 0.0509, + "step": 25761 + }, + { + "epoch": 3.37, + "learning_rate": 1.297221648080853e-05, + "loss": 0.0377, + "step": 25762 + }, + { + "epoch": 3.37, + "learning_rate": 1.2971710117743555e-05, + "loss": 0.0395, + "step": 25763 + }, + { + "epoch": 3.37, + "learning_rate": 1.2971203746320805e-05, + "loss": 0.047, + "step": 25764 + }, + { + "epoch": 3.38, + "learning_rate": 1.2970697366541698e-05, + "loss": 0.0378, + "step": 25765 + }, + { + "epoch": 3.38, + "learning_rate": 1.2970190978407663e-05, + "loss": 0.0336, + "step": 25766 + }, + { + "epoch": 3.38, + "learning_rate": 1.296968458192013e-05, + "loss": 0.0418, + "step": 25767 + }, + { + "epoch": 3.38, + "learning_rate": 1.2969178177080512e-05, + "loss": 0.0353, + "step": 25768 + }, + { + "epoch": 3.38, + "learning_rate": 1.2968671763890242e-05, + "loss": 0.045, + "step": 25769 + }, + { + "epoch": 3.38, + "learning_rate": 1.2968165342350736e-05, + "loss": 0.0373, + "step": 25770 + }, + { + "epoch": 3.38, + "learning_rate": 1.2967658912463428e-05, + "loss": 0.0445, + "step": 25771 + }, + { + "epoch": 3.38, + "learning_rate": 1.2967152474229734e-05, + "loss": 0.0414, + "step": 25772 + }, + { + "epoch": 3.38, + "learning_rate": 1.2966646027651082e-05, + "loss": 0.0421, + "step": 25773 + }, + { + "epoch": 3.38, + "learning_rate": 1.2966139572728897e-05, + "loss": 0.0388, + "step": 25774 + }, + { + "epoch": 3.38, + "learning_rate": 1.2965633109464604e-05, + "loss": 0.0367, + "step": 25775 + }, + { + "epoch": 3.38, + "learning_rate": 1.2965126637859621e-05, + "loss": 0.039, + "step": 25776 + }, + { + "epoch": 3.38, + "learning_rate": 1.2964620157915384e-05, + "loss": 0.0445, + "step": 25777 + }, + { + "epoch": 3.38, + "learning_rate": 1.2964113669633307e-05, + "loss": 0.0445, + "step": 25778 + }, + { + "epoch": 3.38, + "learning_rate": 1.2963607173014816e-05, + "loss": 0.041, + "step": 25779 + }, + { + "epoch": 3.38, + "learning_rate": 1.296310066806134e-05, + "loss": 0.0427, + "step": 25780 + }, + { + "epoch": 3.38, + "learning_rate": 1.2962594154774302e-05, + "loss": 0.0417, + "step": 25781 + }, + { + "epoch": 3.38, + "learning_rate": 1.2962087633155124e-05, + "loss": 0.0476, + "step": 25782 + }, + { + "epoch": 3.38, + "learning_rate": 1.2961581103205235e-05, + "loss": 0.0387, + "step": 25783 + }, + { + "epoch": 3.38, + "learning_rate": 1.2961074564926053e-05, + "loss": 0.0378, + "step": 25784 + }, + { + "epoch": 3.38, + "learning_rate": 1.2960568018319009e-05, + "loss": 0.0423, + "step": 25785 + }, + { + "epoch": 3.38, + "learning_rate": 1.2960061463385523e-05, + "loss": 0.0402, + "step": 25786 + }, + { + "epoch": 3.38, + "learning_rate": 1.2959554900127023e-05, + "loss": 0.0333, + "step": 25787 + }, + { + "epoch": 3.38, + "learning_rate": 1.2959048328544934e-05, + "loss": 0.0389, + "step": 25788 + }, + { + "epoch": 3.38, + "learning_rate": 1.2958541748640675e-05, + "loss": 0.0336, + "step": 25789 + }, + { + "epoch": 3.38, + "learning_rate": 1.2958035160415678e-05, + "loss": 0.033, + "step": 25790 + }, + { + "epoch": 3.38, + "learning_rate": 1.2957528563871364e-05, + "loss": 0.0409, + "step": 25791 + }, + { + "epoch": 3.38, + "learning_rate": 1.295702195900916e-05, + "loss": 0.0493, + "step": 25792 + }, + { + "epoch": 3.38, + "learning_rate": 1.295651534583049e-05, + "loss": 0.048, + "step": 25793 + }, + { + "epoch": 3.38, + "learning_rate": 1.2956008724336775e-05, + "loss": 0.0335, + "step": 25794 + }, + { + "epoch": 3.38, + "learning_rate": 1.2955502094529441e-05, + "loss": 0.0406, + "step": 25795 + }, + { + "epoch": 3.38, + "learning_rate": 1.295499545640992e-05, + "loss": 0.0385, + "step": 25796 + }, + { + "epoch": 3.38, + "learning_rate": 1.2954488809979628e-05, + "loss": 0.0347, + "step": 25797 + }, + { + "epoch": 3.38, + "learning_rate": 1.2953982155239996e-05, + "loss": 0.0425, + "step": 25798 + }, + { + "epoch": 3.38, + "learning_rate": 1.2953475492192444e-05, + "loss": 0.043, + "step": 25799 + }, + { + "epoch": 3.38, + "learning_rate": 1.2952968820838401e-05, + "loss": 0.0397, + "step": 25800 + }, + { + "epoch": 3.38, + "learning_rate": 1.2952462141179289e-05, + "loss": 0.0411, + "step": 25801 + }, + { + "epoch": 3.38, + "learning_rate": 1.2951955453216538e-05, + "loss": 0.0395, + "step": 25802 + }, + { + "epoch": 3.38, + "learning_rate": 1.2951448756951565e-05, + "loss": 0.037, + "step": 25803 + }, + { + "epoch": 3.38, + "learning_rate": 1.2950942052385804e-05, + "loss": 0.038, + "step": 25804 + }, + { + "epoch": 3.38, + "learning_rate": 1.2950435339520673e-05, + "loss": 0.0299, + "step": 25805 + }, + { + "epoch": 3.38, + "learning_rate": 1.2949928618357598e-05, + "loss": 0.0457, + "step": 25806 + }, + { + "epoch": 3.38, + "learning_rate": 1.2949421888898009e-05, + "loss": 0.0317, + "step": 25807 + }, + { + "epoch": 3.38, + "learning_rate": 1.294891515114333e-05, + "loss": 0.0351, + "step": 25808 + }, + { + "epoch": 3.38, + "learning_rate": 1.2948408405094979e-05, + "loss": 0.0335, + "step": 25809 + }, + { + "epoch": 3.38, + "learning_rate": 1.2947901650754386e-05, + "loss": 0.039, + "step": 25810 + }, + { + "epoch": 3.38, + "learning_rate": 1.2947394888122979e-05, + "loss": 0.0446, + "step": 25811 + }, + { + "epoch": 3.38, + "learning_rate": 1.294688811720218e-05, + "loss": 0.0444, + "step": 25812 + }, + { + "epoch": 3.38, + "learning_rate": 1.2946381337993419e-05, + "loss": 0.0445, + "step": 25813 + }, + { + "epoch": 3.38, + "learning_rate": 1.2945874550498112e-05, + "loss": 0.0378, + "step": 25814 + }, + { + "epoch": 3.38, + "learning_rate": 1.2945367754717696e-05, + "loss": 0.0401, + "step": 25815 + }, + { + "epoch": 3.38, + "learning_rate": 1.2944860950653584e-05, + "loss": 0.0376, + "step": 25816 + }, + { + "epoch": 3.38, + "learning_rate": 1.2944354138307209e-05, + "loss": 0.046, + "step": 25817 + }, + { + "epoch": 3.38, + "learning_rate": 1.2943847317679995e-05, + "loss": 0.0414, + "step": 25818 + }, + { + "epoch": 3.38, + "learning_rate": 1.2943340488773369e-05, + "loss": 0.032, + "step": 25819 + }, + { + "epoch": 3.38, + "learning_rate": 1.294283365158875e-05, + "loss": 0.0349, + "step": 25820 + }, + { + "epoch": 3.38, + "learning_rate": 1.2942326806127574e-05, + "loss": 0.0351, + "step": 25821 + }, + { + "epoch": 3.38, + "learning_rate": 1.2941819952391255e-05, + "loss": 0.0311, + "step": 25822 + }, + { + "epoch": 3.38, + "learning_rate": 1.2941313090381229e-05, + "loss": 0.0446, + "step": 25823 + }, + { + "epoch": 3.38, + "learning_rate": 1.2940806220098914e-05, + "loss": 0.0397, + "step": 25824 + }, + { + "epoch": 3.38, + "learning_rate": 1.2940299341545739e-05, + "loss": 0.0453, + "step": 25825 + }, + { + "epoch": 3.38, + "learning_rate": 1.2939792454723128e-05, + "loss": 0.0381, + "step": 25826 + }, + { + "epoch": 3.38, + "learning_rate": 1.2939285559632506e-05, + "loss": 0.0448, + "step": 25827 + }, + { + "epoch": 3.38, + "learning_rate": 1.2938778656275302e-05, + "loss": 0.0396, + "step": 25828 + }, + { + "epoch": 3.38, + "learning_rate": 1.2938271744652939e-05, + "loss": 0.0418, + "step": 25829 + }, + { + "epoch": 3.38, + "learning_rate": 1.2937764824766843e-05, + "loss": 0.0463, + "step": 25830 + }, + { + "epoch": 3.38, + "learning_rate": 1.2937257896618439e-05, + "loss": 0.0378, + "step": 25831 + }, + { + "epoch": 3.38, + "learning_rate": 1.2936750960209157e-05, + "loss": 0.0387, + "step": 25832 + }, + { + "epoch": 3.38, + "learning_rate": 1.2936244015540416e-05, + "loss": 0.0403, + "step": 25833 + }, + { + "epoch": 3.38, + "learning_rate": 1.2935737062613645e-05, + "loss": 0.0384, + "step": 25834 + }, + { + "epoch": 3.38, + "learning_rate": 1.2935230101430269e-05, + "loss": 0.043, + "step": 25835 + }, + { + "epoch": 3.38, + "learning_rate": 1.2934723131991717e-05, + "loss": 0.0345, + "step": 25836 + }, + { + "epoch": 3.38, + "learning_rate": 1.2934216154299412e-05, + "loss": 0.0369, + "step": 25837 + }, + { + "epoch": 3.38, + "learning_rate": 1.2933709168354782e-05, + "loss": 0.045, + "step": 25838 + }, + { + "epoch": 3.38, + "learning_rate": 1.2933202174159244e-05, + "loss": 0.0423, + "step": 25839 + }, + { + "epoch": 3.38, + "learning_rate": 1.293269517171424e-05, + "loss": 0.0417, + "step": 25840 + }, + { + "epoch": 3.38, + "learning_rate": 1.2932188161021182e-05, + "loss": 0.0403, + "step": 25841 + }, + { + "epoch": 3.39, + "learning_rate": 1.2931681142081505e-05, + "loss": 0.0456, + "step": 25842 + }, + { + "epoch": 3.39, + "learning_rate": 1.2931174114896627e-05, + "loss": 0.035, + "step": 25843 + }, + { + "epoch": 3.39, + "learning_rate": 1.2930667079467978e-05, + "loss": 0.0401, + "step": 25844 + }, + { + "epoch": 3.39, + "learning_rate": 1.2930160035796985e-05, + "loss": 0.0343, + "step": 25845 + }, + { + "epoch": 3.39, + "learning_rate": 1.2929652983885075e-05, + "loss": 0.0475, + "step": 25846 + }, + { + "epoch": 3.39, + "learning_rate": 1.2929145923733669e-05, + "loss": 0.0428, + "step": 25847 + }, + { + "epoch": 3.39, + "learning_rate": 1.2928638855344196e-05, + "loss": 0.034, + "step": 25848 + }, + { + "epoch": 3.39, + "learning_rate": 1.2928131778718086e-05, + "loss": 0.0319, + "step": 25849 + }, + { + "epoch": 3.39, + "learning_rate": 1.2927624693856758e-05, + "loss": 0.0351, + "step": 25850 + }, + { + "epoch": 3.39, + "learning_rate": 1.292711760076164e-05, + "loss": 0.0443, + "step": 25851 + }, + { + "epoch": 3.39, + "learning_rate": 1.292661049943416e-05, + "loss": 0.0391, + "step": 25852 + }, + { + "epoch": 3.39, + "learning_rate": 1.2926103389875748e-05, + "loss": 0.0421, + "step": 25853 + }, + { + "epoch": 3.39, + "learning_rate": 1.292559627208782e-05, + "loss": 0.0378, + "step": 25854 + }, + { + "epoch": 3.39, + "learning_rate": 1.2925089146071814e-05, + "loss": 0.0431, + "step": 25855 + }, + { + "epoch": 3.39, + "learning_rate": 1.2924582011829148e-05, + "loss": 0.039, + "step": 25856 + }, + { + "epoch": 3.39, + "learning_rate": 1.2924074869361252e-05, + "loss": 0.0382, + "step": 25857 + }, + { + "epoch": 3.39, + "learning_rate": 1.292356771866955e-05, + "loss": 0.0372, + "step": 25858 + }, + { + "epoch": 3.39, + "learning_rate": 1.2923060559755471e-05, + "loss": 0.045, + "step": 25859 + }, + { + "epoch": 3.39, + "learning_rate": 1.2922553392620436e-05, + "loss": 0.0412, + "step": 25860 + }, + { + "epoch": 3.39, + "learning_rate": 1.292204621726588e-05, + "loss": 0.0384, + "step": 25861 + }, + { + "epoch": 3.39, + "learning_rate": 1.292153903369322e-05, + "loss": 0.0413, + "step": 25862 + }, + { + "epoch": 3.39, + "learning_rate": 1.2921031841903892e-05, + "loss": 0.0406, + "step": 25863 + }, + { + "epoch": 3.39, + "learning_rate": 1.2920524641899314e-05, + "loss": 0.0385, + "step": 25864 + }, + { + "epoch": 3.39, + "learning_rate": 1.292001743368092e-05, + "loss": 0.0374, + "step": 25865 + }, + { + "epoch": 3.39, + "learning_rate": 1.2919510217250127e-05, + "loss": 0.0367, + "step": 25866 + }, + { + "epoch": 3.39, + "learning_rate": 1.291900299260837e-05, + "loss": 0.0417, + "step": 25867 + }, + { + "epoch": 3.39, + "learning_rate": 1.2918495759757071e-05, + "loss": 0.0364, + "step": 25868 + }, + { + "epoch": 3.39, + "learning_rate": 1.291798851869766e-05, + "loss": 0.0381, + "step": 25869 + }, + { + "epoch": 3.39, + "learning_rate": 1.291748126943156e-05, + "loss": 0.0374, + "step": 25870 + }, + { + "epoch": 3.39, + "learning_rate": 1.2916974011960199e-05, + "loss": 0.0426, + "step": 25871 + }, + { + "epoch": 3.39, + "learning_rate": 1.2916466746285006e-05, + "loss": 0.0418, + "step": 25872 + }, + { + "epoch": 3.39, + "learning_rate": 1.2915959472407403e-05, + "loss": 0.0336, + "step": 25873 + }, + { + "epoch": 3.39, + "learning_rate": 1.2915452190328824e-05, + "loss": 0.0421, + "step": 25874 + }, + { + "epoch": 3.39, + "learning_rate": 1.2914944900050685e-05, + "loss": 0.0364, + "step": 25875 + }, + { + "epoch": 3.39, + "learning_rate": 1.2914437601574423e-05, + "loss": 0.0392, + "step": 25876 + }, + { + "epoch": 3.39, + "learning_rate": 1.2913930294901458e-05, + "loss": 0.0374, + "step": 25877 + }, + { + "epoch": 3.39, + "learning_rate": 1.2913422980033221e-05, + "loss": 0.0405, + "step": 25878 + }, + { + "epoch": 3.39, + "learning_rate": 1.2912915656971135e-05, + "loss": 0.0407, + "step": 25879 + }, + { + "epoch": 3.39, + "learning_rate": 1.2912408325716632e-05, + "loss": 0.0482, + "step": 25880 + }, + { + "epoch": 3.39, + "learning_rate": 1.2911900986271133e-05, + "loss": 0.0342, + "step": 25881 + }, + { + "epoch": 3.39, + "learning_rate": 1.2911393638636066e-05, + "loss": 0.0376, + "step": 25882 + }, + { + "epoch": 3.39, + "learning_rate": 1.2910886282812861e-05, + "loss": 0.0437, + "step": 25883 + }, + { + "epoch": 3.39, + "learning_rate": 1.2910378918802946e-05, + "loss": 0.0471, + "step": 25884 + }, + { + "epoch": 3.39, + "learning_rate": 1.2909871546607744e-05, + "loss": 0.033, + "step": 25885 + }, + { + "epoch": 3.39, + "learning_rate": 1.2909364166228685e-05, + "loss": 0.0349, + "step": 25886 + }, + { + "epoch": 3.39, + "learning_rate": 1.2908856777667189e-05, + "loss": 0.0386, + "step": 25887 + }, + { + "epoch": 3.39, + "learning_rate": 1.290834938092469e-05, + "loss": 0.0334, + "step": 25888 + }, + { + "epoch": 3.39, + "learning_rate": 1.2907841976002615e-05, + "loss": 0.0372, + "step": 25889 + }, + { + "epoch": 3.39, + "learning_rate": 1.290733456290239e-05, + "loss": 0.041, + "step": 25890 + }, + { + "epoch": 3.39, + "learning_rate": 1.2906827141625437e-05, + "loss": 0.0451, + "step": 25891 + }, + { + "epoch": 3.39, + "learning_rate": 1.290631971217319e-05, + "loss": 0.0475, + "step": 25892 + }, + { + "epoch": 3.39, + "learning_rate": 1.2905812274547074e-05, + "loss": 0.0362, + "step": 25893 + }, + { + "epoch": 3.39, + "learning_rate": 1.2905304828748517e-05, + "loss": 0.0442, + "step": 25894 + }, + { + "epoch": 3.39, + "learning_rate": 1.2904797374778944e-05, + "loss": 0.0415, + "step": 25895 + }, + { + "epoch": 3.39, + "learning_rate": 1.290428991263978e-05, + "loss": 0.042, + "step": 25896 + }, + { + "epoch": 3.39, + "learning_rate": 1.2903782442332463e-05, + "loss": 0.0428, + "step": 25897 + }, + { + "epoch": 3.39, + "learning_rate": 1.2903274963858406e-05, + "loss": 0.0491, + "step": 25898 + }, + { + "epoch": 3.39, + "learning_rate": 1.2902767477219043e-05, + "loss": 0.0357, + "step": 25899 + }, + { + "epoch": 3.39, + "learning_rate": 1.2902259982415802e-05, + "loss": 0.0473, + "step": 25900 + }, + { + "epoch": 3.39, + "learning_rate": 1.2901752479450109e-05, + "loss": 0.0329, + "step": 25901 + }, + { + "epoch": 3.39, + "learning_rate": 1.2901244968323393e-05, + "loss": 0.0432, + "step": 25902 + }, + { + "epoch": 3.39, + "learning_rate": 1.290073744903708e-05, + "loss": 0.0392, + "step": 25903 + }, + { + "epoch": 3.39, + "learning_rate": 1.2900229921592594e-05, + "loss": 0.0434, + "step": 25904 + }, + { + "epoch": 3.39, + "learning_rate": 1.2899722385991369e-05, + "loss": 0.0435, + "step": 25905 + }, + { + "epoch": 3.39, + "learning_rate": 1.289921484223483e-05, + "loss": 0.0359, + "step": 25906 + }, + { + "epoch": 3.39, + "learning_rate": 1.2898707290324403e-05, + "loss": 0.0404, + "step": 25907 + }, + { + "epoch": 3.39, + "learning_rate": 1.2898199730261514e-05, + "loss": 0.0397, + "step": 25908 + }, + { + "epoch": 3.39, + "learning_rate": 1.2897692162047598e-05, + "loss": 0.0333, + "step": 25909 + }, + { + "epoch": 3.39, + "learning_rate": 1.2897184585684067e-05, + "loss": 0.0434, + "step": 25910 + }, + { + "epoch": 3.39, + "learning_rate": 1.2896677001172368e-05, + "loss": 0.0377, + "step": 25911 + }, + { + "epoch": 3.39, + "learning_rate": 1.2896169408513915e-05, + "loss": 0.0364, + "step": 25912 + }, + { + "epoch": 3.39, + "learning_rate": 1.289566180771014e-05, + "loss": 0.0411, + "step": 25913 + }, + { + "epoch": 3.39, + "learning_rate": 1.2895154198762472e-05, + "loss": 0.0372, + "step": 25914 + }, + { + "epoch": 3.39, + "learning_rate": 1.2894646581672337e-05, + "loss": 0.0425, + "step": 25915 + }, + { + "epoch": 3.39, + "learning_rate": 1.289413895644116e-05, + "loss": 0.0425, + "step": 25916 + }, + { + "epoch": 3.39, + "learning_rate": 1.2893631323070373e-05, + "loss": 0.033, + "step": 25917 + }, + { + "epoch": 3.4, + "learning_rate": 1.2893123681561403e-05, + "loss": 0.0432, + "step": 25918 + }, + { + "epoch": 3.4, + "learning_rate": 1.2892616031915674e-05, + "loss": 0.0376, + "step": 25919 + }, + { + "epoch": 3.4, + "learning_rate": 1.289210837413462e-05, + "loss": 0.0352, + "step": 25920 + }, + { + "epoch": 3.4, + "learning_rate": 1.289160070821966e-05, + "loss": 0.0464, + "step": 25921 + }, + { + "epoch": 3.4, + "learning_rate": 1.2891093034172233e-05, + "loss": 0.0428, + "step": 25922 + }, + { + "epoch": 3.4, + "learning_rate": 1.289058535199376e-05, + "loss": 0.0433, + "step": 25923 + }, + { + "epoch": 3.4, + "learning_rate": 1.2890077661685668e-05, + "loss": 0.0424, + "step": 25924 + }, + { + "epoch": 3.4, + "learning_rate": 1.2889569963249384e-05, + "loss": 0.0349, + "step": 25925 + }, + { + "epoch": 3.4, + "learning_rate": 1.288906225668634e-05, + "loss": 0.0377, + "step": 25926 + }, + { + "epoch": 3.4, + "learning_rate": 1.2888554541997965e-05, + "loss": 0.039, + "step": 25927 + }, + { + "epoch": 3.4, + "learning_rate": 1.2888046819185682e-05, + "loss": 0.0419, + "step": 25928 + }, + { + "epoch": 3.4, + "learning_rate": 1.288753908825092e-05, + "loss": 0.0409, + "step": 25929 + }, + { + "epoch": 3.4, + "learning_rate": 1.2887031349195111e-05, + "loss": 0.0411, + "step": 25930 + }, + { + "epoch": 3.4, + "learning_rate": 1.2886523602019676e-05, + "loss": 0.0383, + "step": 25931 + }, + { + "epoch": 3.4, + "learning_rate": 1.288601584672605e-05, + "loss": 0.0319, + "step": 25932 + }, + { + "epoch": 3.4, + "learning_rate": 1.2885508083315657e-05, + "loss": 0.0402, + "step": 25933 + }, + { + "epoch": 3.4, + "learning_rate": 1.2885000311789929e-05, + "loss": 0.0357, + "step": 25934 + }, + { + "epoch": 3.4, + "learning_rate": 1.2884492532150287e-05, + "loss": 0.0423, + "step": 25935 + }, + { + "epoch": 3.4, + "learning_rate": 1.2883984744398165e-05, + "loss": 0.0382, + "step": 25936 + }, + { + "epoch": 3.4, + "learning_rate": 1.288347694853499e-05, + "loss": 0.0396, + "step": 25937 + }, + { + "epoch": 3.4, + "learning_rate": 1.288296914456219e-05, + "loss": 0.0394, + "step": 25938 + }, + { + "epoch": 3.4, + "learning_rate": 1.2882461332481193e-05, + "loss": 0.047, + "step": 25939 + }, + { + "epoch": 3.4, + "learning_rate": 1.2881953512293425e-05, + "loss": 0.0427, + "step": 25940 + }, + { + "epoch": 3.4, + "learning_rate": 1.288144568400032e-05, + "loss": 0.0428, + "step": 25941 + }, + { + "epoch": 3.4, + "learning_rate": 1.2880937847603297e-05, + "loss": 0.0482, + "step": 25942 + }, + { + "epoch": 3.4, + "learning_rate": 1.2880430003103795e-05, + "loss": 0.0349, + "step": 25943 + }, + { + "epoch": 3.4, + "learning_rate": 1.2879922150503232e-05, + "loss": 0.0377, + "step": 25944 + }, + { + "epoch": 3.4, + "learning_rate": 1.2879414289803047e-05, + "loss": 0.0404, + "step": 25945 + }, + { + "epoch": 3.4, + "learning_rate": 1.2878906421004658e-05, + "loss": 0.0412, + "step": 25946 + }, + { + "epoch": 3.4, + "learning_rate": 1.2878398544109501e-05, + "loss": 0.0418, + "step": 25947 + }, + { + "epoch": 3.4, + "learning_rate": 1.2877890659118996e-05, + "loss": 0.0389, + "step": 25948 + }, + { + "epoch": 3.4, + "learning_rate": 1.2877382766034585e-05, + "loss": 0.0408, + "step": 25949 + }, + { + "epoch": 3.4, + "learning_rate": 1.2876874864857682e-05, + "loss": 0.0394, + "step": 25950 + }, + { + "epoch": 3.4, + "learning_rate": 1.2876366955589726e-05, + "loss": 0.0411, + "step": 25951 + }, + { + "epoch": 3.4, + "learning_rate": 1.2875859038232138e-05, + "loss": 0.0439, + "step": 25952 + }, + { + "epoch": 3.4, + "learning_rate": 1.287535111278635e-05, + "loss": 0.036, + "step": 25953 + }, + { + "epoch": 3.4, + "learning_rate": 1.2874843179253791e-05, + "loss": 0.0407, + "step": 25954 + }, + { + "epoch": 3.4, + "learning_rate": 1.2874335237635888e-05, + "loss": 0.0424, + "step": 25955 + }, + { + "epoch": 3.4, + "learning_rate": 1.287382728793407e-05, + "loss": 0.0426, + "step": 25956 + }, + { + "epoch": 3.4, + "learning_rate": 1.2873319330149764e-05, + "loss": 0.0346, + "step": 25957 + }, + { + "epoch": 3.4, + "learning_rate": 1.2872811364284405e-05, + "loss": 0.0438, + "step": 25958 + }, + { + "epoch": 3.4, + "learning_rate": 1.2872303390339416e-05, + "loss": 0.0417, + "step": 25959 + }, + { + "epoch": 3.4, + "learning_rate": 1.2871795408316224e-05, + "loss": 0.0368, + "step": 25960 + }, + { + "epoch": 3.4, + "learning_rate": 1.2871287418216261e-05, + "loss": 0.041, + "step": 25961 + }, + { + "epoch": 3.4, + "learning_rate": 1.2870779420040958e-05, + "loss": 0.0358, + "step": 25962 + }, + { + "epoch": 3.4, + "learning_rate": 1.2870271413791736e-05, + "loss": 0.0391, + "step": 25963 + }, + { + "epoch": 3.4, + "learning_rate": 1.286976339947003e-05, + "loss": 0.032, + "step": 25964 + }, + { + "epoch": 3.4, + "learning_rate": 1.2869255377077268e-05, + "loss": 0.0416, + "step": 25965 + }, + { + "epoch": 3.4, + "learning_rate": 1.286874734661488e-05, + "loss": 0.0432, + "step": 25966 + }, + { + "epoch": 3.4, + "learning_rate": 1.2868239308084292e-05, + "loss": 0.034, + "step": 25967 + }, + { + "epoch": 3.4, + "learning_rate": 1.2867731261486933e-05, + "loss": 0.0448, + "step": 25968 + }, + { + "epoch": 3.4, + "learning_rate": 1.2867223206824232e-05, + "loss": 0.0382, + "step": 25969 + }, + { + "epoch": 3.4, + "learning_rate": 1.2866715144097617e-05, + "loss": 0.0462, + "step": 25970 + }, + { + "epoch": 3.4, + "learning_rate": 1.286620707330852e-05, + "loss": 0.0397, + "step": 25971 + }, + { + "epoch": 3.4, + "learning_rate": 1.2865698994458372e-05, + "loss": 0.0404, + "step": 25972 + }, + { + "epoch": 3.4, + "learning_rate": 1.286519090754859e-05, + "loss": 0.0473, + "step": 25973 + }, + { + "epoch": 3.4, + "learning_rate": 1.286468281258062e-05, + "loss": 0.0408, + "step": 25974 + }, + { + "epoch": 3.4, + "learning_rate": 1.2864174709555875e-05, + "loss": 0.0326, + "step": 25975 + }, + { + "epoch": 3.4, + "learning_rate": 1.2863666598475795e-05, + "loss": 0.0402, + "step": 25976 + }, + { + "epoch": 3.4, + "learning_rate": 1.2863158479341802e-05, + "loss": 0.0325, + "step": 25977 + }, + { + "epoch": 3.4, + "learning_rate": 1.286265035215533e-05, + "loss": 0.0422, + "step": 25978 + }, + { + "epoch": 3.4, + "learning_rate": 1.2862142216917808e-05, + "loss": 0.0381, + "step": 25979 + }, + { + "epoch": 3.4, + "learning_rate": 1.2861634073630662e-05, + "loss": 0.0359, + "step": 25980 + }, + { + "epoch": 3.4, + "learning_rate": 1.2861125922295322e-05, + "loss": 0.0454, + "step": 25981 + }, + { + "epoch": 3.4, + "learning_rate": 1.2860617762913218e-05, + "loss": 0.0409, + "step": 25982 + }, + { + "epoch": 3.4, + "learning_rate": 1.2860109595485782e-05, + "loss": 0.0446, + "step": 25983 + }, + { + "epoch": 3.4, + "learning_rate": 1.2859601420014435e-05, + "loss": 0.0469, + "step": 25984 + }, + { + "epoch": 3.4, + "learning_rate": 1.2859093236500614e-05, + "loss": 0.038, + "step": 25985 + }, + { + "epoch": 3.4, + "learning_rate": 1.2858585044945743e-05, + "loss": 0.0324, + "step": 25986 + }, + { + "epoch": 3.4, + "learning_rate": 1.2858076845351257e-05, + "loss": 0.0361, + "step": 25987 + }, + { + "epoch": 3.4, + "learning_rate": 1.2857568637718583e-05, + "loss": 0.0423, + "step": 25988 + }, + { + "epoch": 3.4, + "learning_rate": 1.2857060422049148e-05, + "loss": 0.0387, + "step": 25989 + }, + { + "epoch": 3.4, + "learning_rate": 1.2856552198344381e-05, + "loss": 0.0439, + "step": 25990 + }, + { + "epoch": 3.4, + "learning_rate": 1.2856043966605713e-05, + "loss": 0.0395, + "step": 25991 + }, + { + "epoch": 3.4, + "learning_rate": 1.2855535726834574e-05, + "loss": 0.0372, + "step": 25992 + }, + { + "epoch": 3.4, + "learning_rate": 1.2855027479032396e-05, + "loss": 0.0372, + "step": 25993 + }, + { + "epoch": 3.41, + "learning_rate": 1.28545192232006e-05, + "loss": 0.0394, + "step": 25994 + }, + { + "epoch": 3.41, + "learning_rate": 1.2854010959340629e-05, + "loss": 0.0382, + "step": 25995 + }, + { + "epoch": 3.41, + "learning_rate": 1.2853502687453895e-05, + "loss": 0.0367, + "step": 25996 + }, + { + "epoch": 3.41, + "learning_rate": 1.2852994407541841e-05, + "loss": 0.038, + "step": 25997 + }, + { + "epoch": 3.41, + "learning_rate": 1.2852486119605891e-05, + "loss": 0.0431, + "step": 25998 + }, + { + "epoch": 3.41, + "learning_rate": 1.2851977823647478e-05, + "loss": 0.0423, + "step": 25999 + }, + { + "epoch": 3.41, + "learning_rate": 1.2851469519668026e-05, + "loss": 0.0377, + "step": 26000 + }, + { + "epoch": 3.41, + "learning_rate": 1.2850961207668968e-05, + "loss": 0.0416, + "step": 26001 + }, + { + "epoch": 3.41, + "learning_rate": 1.2850452887651735e-05, + "loss": 0.0407, + "step": 26002 + }, + { + "epoch": 3.41, + "learning_rate": 1.2849944559617754e-05, + "loss": 0.0421, + "step": 26003 + }, + { + "epoch": 3.41, + "learning_rate": 1.2849436223568457e-05, + "loss": 0.0405, + "step": 26004 + }, + { + "epoch": 3.41, + "learning_rate": 1.284892787950527e-05, + "loss": 0.039, + "step": 26005 + }, + { + "epoch": 3.41, + "learning_rate": 1.284841952742963e-05, + "loss": 0.0376, + "step": 26006 + }, + { + "epoch": 3.41, + "learning_rate": 1.2847911167342958e-05, + "loss": 0.0375, + "step": 26007 + }, + { + "epoch": 3.41, + "learning_rate": 1.2847402799246686e-05, + "loss": 0.0435, + "step": 26008 + }, + { + "epoch": 3.41, + "learning_rate": 1.2846894423142247e-05, + "loss": 0.0435, + "step": 26009 + }, + { + "epoch": 3.41, + "learning_rate": 1.2846386039031071e-05, + "loss": 0.0315, + "step": 26010 + }, + { + "epoch": 3.41, + "learning_rate": 1.2845877646914584e-05, + "loss": 0.0423, + "step": 26011 + }, + { + "epoch": 3.41, + "learning_rate": 1.2845369246794217e-05, + "loss": 0.05, + "step": 26012 + }, + { + "epoch": 3.41, + "learning_rate": 1.28448608386714e-05, + "loss": 0.0428, + "step": 26013 + }, + { + "epoch": 3.41, + "learning_rate": 1.2844352422547565e-05, + "loss": 0.0393, + "step": 26014 + }, + { + "epoch": 3.41, + "learning_rate": 1.2843843998424142e-05, + "loss": 0.0371, + "step": 26015 + }, + { + "epoch": 3.41, + "learning_rate": 1.2843335566302558e-05, + "loss": 0.0336, + "step": 26016 + }, + { + "epoch": 3.41, + "learning_rate": 1.284282712618424e-05, + "loss": 0.0409, + "step": 26017 + }, + { + "epoch": 3.41, + "learning_rate": 1.2842318678070627e-05, + "loss": 0.0431, + "step": 26018 + }, + { + "epoch": 3.41, + "learning_rate": 1.2841810221963143e-05, + "loss": 0.041, + "step": 26019 + }, + { + "epoch": 3.41, + "learning_rate": 1.2841301757863219e-05, + "loss": 0.0367, + "step": 26020 + }, + { + "epoch": 3.41, + "learning_rate": 1.2840793285772284e-05, + "loss": 0.0405, + "step": 26021 + }, + { + "epoch": 3.41, + "learning_rate": 1.2840284805691769e-05, + "loss": 0.0363, + "step": 26022 + }, + { + "epoch": 3.41, + "learning_rate": 1.2839776317623108e-05, + "loss": 0.0429, + "step": 26023 + }, + { + "epoch": 3.41, + "learning_rate": 1.2839267821567722e-05, + "loss": 0.0304, + "step": 26024 + }, + { + "epoch": 3.41, + "learning_rate": 1.283875931752705e-05, + "loss": 0.0434, + "step": 26025 + }, + { + "epoch": 3.41, + "learning_rate": 1.2838250805502517e-05, + "loss": 0.0397, + "step": 26026 + }, + { + "epoch": 3.41, + "learning_rate": 1.2837742285495556e-05, + "loss": 0.0325, + "step": 26027 + }, + { + "epoch": 3.41, + "learning_rate": 1.2837233757507594e-05, + "loss": 0.0482, + "step": 26028 + }, + { + "epoch": 3.41, + "learning_rate": 1.2836725221540065e-05, + "loss": 0.0412, + "step": 26029 + }, + { + "epoch": 3.41, + "learning_rate": 1.2836216677594395e-05, + "loss": 0.0457, + "step": 26030 + }, + { + "epoch": 3.41, + "learning_rate": 1.283570812567202e-05, + "loss": 0.0411, + "step": 26031 + }, + { + "epoch": 3.41, + "learning_rate": 1.2835199565774366e-05, + "loss": 0.0353, + "step": 26032 + }, + { + "epoch": 3.41, + "learning_rate": 1.2834690997902863e-05, + "loss": 0.0433, + "step": 26033 + }, + { + "epoch": 3.41, + "learning_rate": 1.2834182422058942e-05, + "loss": 0.0509, + "step": 26034 + }, + { + "epoch": 3.41, + "learning_rate": 1.2833673838244037e-05, + "loss": 0.0463, + "step": 26035 + }, + { + "epoch": 3.41, + "learning_rate": 1.2833165246459571e-05, + "loss": 0.0367, + "step": 26036 + }, + { + "epoch": 3.41, + "learning_rate": 1.2832656646706982e-05, + "loss": 0.0355, + "step": 26037 + }, + { + "epoch": 3.41, + "learning_rate": 1.2832148038987695e-05, + "loss": 0.0463, + "step": 26038 + }, + { + "epoch": 3.41, + "learning_rate": 1.2831639423303141e-05, + "loss": 0.043, + "step": 26039 + }, + { + "epoch": 3.41, + "learning_rate": 1.2831130799654756e-05, + "loss": 0.0359, + "step": 26040 + }, + { + "epoch": 3.41, + "learning_rate": 1.2830622168043965e-05, + "loss": 0.0332, + "step": 26041 + }, + { + "epoch": 3.41, + "learning_rate": 1.2830113528472199e-05, + "loss": 0.0415, + "step": 26042 + }, + { + "epoch": 3.41, + "learning_rate": 1.2829604880940888e-05, + "loss": 0.0388, + "step": 26043 + }, + { + "epoch": 3.41, + "learning_rate": 1.2829096225451468e-05, + "loss": 0.041, + "step": 26044 + }, + { + "epoch": 3.41, + "learning_rate": 1.2828587562005362e-05, + "loss": 0.0384, + "step": 26045 + }, + { + "epoch": 3.41, + "learning_rate": 1.2828078890604005e-05, + "loss": 0.0396, + "step": 26046 + }, + { + "epoch": 3.41, + "learning_rate": 1.2827570211248828e-05, + "loss": 0.0424, + "step": 26047 + }, + { + "epoch": 3.41, + "learning_rate": 1.282706152394126e-05, + "loss": 0.031, + "step": 26048 + }, + { + "epoch": 3.41, + "learning_rate": 1.282655282868273e-05, + "loss": 0.0368, + "step": 26049 + }, + { + "epoch": 3.41, + "learning_rate": 1.2826044125474674e-05, + "loss": 0.0399, + "step": 26050 + }, + { + "epoch": 3.41, + "learning_rate": 1.2825535414318518e-05, + "loss": 0.0471, + "step": 26051 + }, + { + "epoch": 3.41, + "learning_rate": 1.2825026695215693e-05, + "loss": 0.0347, + "step": 26052 + }, + { + "epoch": 3.41, + "learning_rate": 1.282451796816763e-05, + "loss": 0.0405, + "step": 26053 + }, + { + "epoch": 3.41, + "learning_rate": 1.2824009233175766e-05, + "loss": 0.0411, + "step": 26054 + }, + { + "epoch": 3.41, + "learning_rate": 1.2823500490241521e-05, + "loss": 0.0369, + "step": 26055 + }, + { + "epoch": 3.41, + "learning_rate": 1.2822991739366333e-05, + "loss": 0.0441, + "step": 26056 + }, + { + "epoch": 3.41, + "learning_rate": 1.2822482980551629e-05, + "loss": 0.0422, + "step": 26057 + }, + { + "epoch": 3.41, + "learning_rate": 1.2821974213798845e-05, + "loss": 0.0397, + "step": 26058 + }, + { + "epoch": 3.41, + "learning_rate": 1.2821465439109407e-05, + "loss": 0.0368, + "step": 26059 + }, + { + "epoch": 3.41, + "learning_rate": 1.282095665648475e-05, + "loss": 0.0334, + "step": 26060 + }, + { + "epoch": 3.41, + "learning_rate": 1.28204478659263e-05, + "loss": 0.0482, + "step": 26061 + }, + { + "epoch": 3.41, + "learning_rate": 1.2819939067435492e-05, + "loss": 0.041, + "step": 26062 + }, + { + "epoch": 3.41, + "learning_rate": 1.2819430261013755e-05, + "loss": 0.0444, + "step": 26063 + }, + { + "epoch": 3.41, + "learning_rate": 1.2818921446662517e-05, + "loss": 0.0414, + "step": 26064 + }, + { + "epoch": 3.41, + "learning_rate": 1.2818412624383219e-05, + "loss": 0.0426, + "step": 26065 + }, + { + "epoch": 3.41, + "learning_rate": 1.2817903794177283e-05, + "loss": 0.0385, + "step": 26066 + }, + { + "epoch": 3.41, + "learning_rate": 1.2817394956046142e-05, + "loss": 0.0375, + "step": 26067 + }, + { + "epoch": 3.41, + "learning_rate": 1.2816886109991227e-05, + "loss": 0.0389, + "step": 26068 + }, + { + "epoch": 3.41, + "learning_rate": 1.2816377256013972e-05, + "loss": 0.0402, + "step": 26069 + }, + { + "epoch": 3.41, + "learning_rate": 1.2815868394115805e-05, + "loss": 0.0368, + "step": 26070 + }, + { + "epoch": 3.42, + "learning_rate": 1.2815359524298158e-05, + "loss": 0.0474, + "step": 26071 + }, + { + "epoch": 3.42, + "learning_rate": 1.2814850646562463e-05, + "loss": 0.0354, + "step": 26072 + }, + { + "epoch": 3.42, + "learning_rate": 1.2814341760910148e-05, + "loss": 0.0406, + "step": 26073 + }, + { + "epoch": 3.42, + "learning_rate": 1.2813832867342648e-05, + "loss": 0.0404, + "step": 26074 + }, + { + "epoch": 3.42, + "learning_rate": 1.2813323965861395e-05, + "loss": 0.0321, + "step": 26075 + }, + { + "epoch": 3.42, + "learning_rate": 1.2812815056467815e-05, + "loss": 0.0474, + "step": 26076 + }, + { + "epoch": 3.42, + "learning_rate": 1.2812306139163347e-05, + "loss": 0.0416, + "step": 26077 + }, + { + "epoch": 3.42, + "learning_rate": 1.2811797213949413e-05, + "loss": 0.0424, + "step": 26078 + }, + { + "epoch": 3.42, + "learning_rate": 1.2811288280827449e-05, + "loss": 0.0425, + "step": 26079 + }, + { + "epoch": 3.42, + "learning_rate": 1.2810779339798888e-05, + "loss": 0.0411, + "step": 26080 + }, + { + "epoch": 3.42, + "learning_rate": 1.2810270390865162e-05, + "loss": 0.0383, + "step": 26081 + }, + { + "epoch": 3.42, + "learning_rate": 1.2809761434027697e-05, + "loss": 0.0328, + "step": 26082 + }, + { + "epoch": 3.42, + "learning_rate": 1.280925246928793e-05, + "loss": 0.0414, + "step": 26083 + }, + { + "epoch": 3.42, + "learning_rate": 1.2808743496647287e-05, + "loss": 0.0374, + "step": 26084 + }, + { + "epoch": 3.42, + "learning_rate": 1.2808234516107205e-05, + "loss": 0.0409, + "step": 26085 + }, + { + "epoch": 3.42, + "learning_rate": 1.2807725527669113e-05, + "loss": 0.0406, + "step": 26086 + }, + { + "epoch": 3.42, + "learning_rate": 1.280721653133444e-05, + "loss": 0.0441, + "step": 26087 + }, + { + "epoch": 3.42, + "learning_rate": 1.2806707527104623e-05, + "loss": 0.0337, + "step": 26088 + }, + { + "epoch": 3.42, + "learning_rate": 1.2806198514981088e-05, + "loss": 0.0428, + "step": 26089 + }, + { + "epoch": 3.42, + "learning_rate": 1.280568949496527e-05, + "loss": 0.0318, + "step": 26090 + }, + { + "epoch": 3.42, + "learning_rate": 1.28051804670586e-05, + "loss": 0.0399, + "step": 26091 + }, + { + "epoch": 3.42, + "learning_rate": 1.2804671431262512e-05, + "loss": 0.0356, + "step": 26092 + }, + { + "epoch": 3.42, + "learning_rate": 1.2804162387578432e-05, + "loss": 0.0372, + "step": 26093 + }, + { + "epoch": 3.42, + "learning_rate": 1.2803653336007794e-05, + "loss": 0.0407, + "step": 26094 + }, + { + "epoch": 3.42, + "learning_rate": 1.2803144276552029e-05, + "loss": 0.0301, + "step": 26095 + }, + { + "epoch": 3.42, + "learning_rate": 1.2802635209212575e-05, + "loss": 0.0361, + "step": 26096 + }, + { + "epoch": 3.42, + "learning_rate": 1.2802126133990855e-05, + "loss": 0.0338, + "step": 26097 + }, + { + "epoch": 3.42, + "learning_rate": 1.2801617050888306e-05, + "loss": 0.0361, + "step": 26098 + }, + { + "epoch": 3.42, + "learning_rate": 1.2801107959906358e-05, + "loss": 0.0346, + "step": 26099 + }, + { + "epoch": 3.42, + "learning_rate": 1.2800598861046442e-05, + "loss": 0.0428, + "step": 26100 + }, + { + "epoch": 3.42, + "learning_rate": 1.280008975430999e-05, + "loss": 0.0366, + "step": 26101 + }, + { + "epoch": 3.42, + "learning_rate": 1.2799580639698439e-05, + "loss": 0.0336, + "step": 26102 + }, + { + "epoch": 3.42, + "learning_rate": 1.279907151721321e-05, + "loss": 0.0469, + "step": 26103 + }, + { + "epoch": 3.42, + "learning_rate": 1.2798562386855747e-05, + "loss": 0.0386, + "step": 26104 + }, + { + "epoch": 3.42, + "learning_rate": 1.2798053248627475e-05, + "loss": 0.0406, + "step": 26105 + }, + { + "epoch": 3.42, + "learning_rate": 1.2797544102529826e-05, + "loss": 0.0395, + "step": 26106 + }, + { + "epoch": 3.42, + "learning_rate": 1.2797034948564233e-05, + "loss": 0.0381, + "step": 26107 + }, + { + "epoch": 3.42, + "learning_rate": 1.279652578673213e-05, + "loss": 0.0407, + "step": 26108 + }, + { + "epoch": 3.42, + "learning_rate": 1.2796016617034948e-05, + "loss": 0.0383, + "step": 26109 + }, + { + "epoch": 3.42, + "learning_rate": 1.2795507439474116e-05, + "loss": 0.041, + "step": 26110 + }, + { + "epoch": 3.42, + "learning_rate": 1.279499825405107e-05, + "loss": 0.039, + "step": 26111 + }, + { + "epoch": 3.42, + "learning_rate": 1.2794489060767234e-05, + "loss": 0.0395, + "step": 26112 + }, + { + "epoch": 3.42, + "learning_rate": 1.2793979859624052e-05, + "loss": 0.0365, + "step": 26113 + }, + { + "epoch": 3.42, + "learning_rate": 1.279347065062295e-05, + "loss": 0.0369, + "step": 26114 + }, + { + "epoch": 3.42, + "learning_rate": 1.2792961433765362e-05, + "loss": 0.0416, + "step": 26115 + }, + { + "epoch": 3.42, + "learning_rate": 1.2792452209052717e-05, + "loss": 0.0419, + "step": 26116 + }, + { + "epoch": 3.42, + "learning_rate": 1.2791942976486446e-05, + "loss": 0.0437, + "step": 26117 + }, + { + "epoch": 3.42, + "learning_rate": 1.2791433736067988e-05, + "loss": 0.044, + "step": 26118 + }, + { + "epoch": 3.42, + "learning_rate": 1.279092448779877e-05, + "loss": 0.0371, + "step": 26119 + }, + { + "epoch": 3.42, + "learning_rate": 1.2790415231680226e-05, + "loss": 0.0442, + "step": 26120 + }, + { + "epoch": 3.42, + "learning_rate": 1.278990596771379e-05, + "loss": 0.0404, + "step": 26121 + }, + { + "epoch": 3.42, + "learning_rate": 1.2789396695900886e-05, + "loss": 0.0398, + "step": 26122 + }, + { + "epoch": 3.42, + "learning_rate": 1.2788887416242956e-05, + "loss": 0.0396, + "step": 26123 + }, + { + "epoch": 3.42, + "learning_rate": 1.2788378128741428e-05, + "loss": 0.0387, + "step": 26124 + }, + { + "epoch": 3.42, + "learning_rate": 1.2787868833397737e-05, + "loss": 0.0412, + "step": 26125 + }, + { + "epoch": 3.42, + "learning_rate": 1.2787359530213311e-05, + "loss": 0.0353, + "step": 26126 + }, + { + "epoch": 3.42, + "learning_rate": 1.2786850219189587e-05, + "loss": 0.0375, + "step": 26127 + }, + { + "epoch": 3.42, + "learning_rate": 1.2786340900327992e-05, + "loss": 0.0421, + "step": 26128 + }, + { + "epoch": 3.42, + "learning_rate": 1.278583157362996e-05, + "loss": 0.0447, + "step": 26129 + }, + { + "epoch": 3.42, + "learning_rate": 1.2785322239096931e-05, + "loss": 0.0361, + "step": 26130 + }, + { + "epoch": 3.42, + "learning_rate": 1.278481289673033e-05, + "loss": 0.0418, + "step": 26131 + }, + { + "epoch": 3.42, + "learning_rate": 1.278430354653159e-05, + "loss": 0.0373, + "step": 26132 + }, + { + "epoch": 3.42, + "learning_rate": 1.278379418850214e-05, + "loss": 0.045, + "step": 26133 + }, + { + "epoch": 3.42, + "learning_rate": 1.2783284822643424e-05, + "loss": 0.0368, + "step": 26134 + }, + { + "epoch": 3.42, + "learning_rate": 1.2782775448956865e-05, + "loss": 0.0421, + "step": 26135 + }, + { + "epoch": 3.42, + "learning_rate": 1.27822660674439e-05, + "loss": 0.0326, + "step": 26136 + }, + { + "epoch": 3.42, + "learning_rate": 1.278175667810596e-05, + "loss": 0.0379, + "step": 26137 + }, + { + "epoch": 3.42, + "learning_rate": 1.2781247280944475e-05, + "loss": 0.0359, + "step": 26138 + }, + { + "epoch": 3.42, + "learning_rate": 1.2780737875960878e-05, + "loss": 0.0397, + "step": 26139 + }, + { + "epoch": 3.42, + "learning_rate": 1.2780228463156608e-05, + "loss": 0.0427, + "step": 26140 + }, + { + "epoch": 3.42, + "learning_rate": 1.2779719042533091e-05, + "loss": 0.0283, + "step": 26141 + }, + { + "epoch": 3.42, + "learning_rate": 1.2779209614091766e-05, + "loss": 0.042, + "step": 26142 + }, + { + "epoch": 3.42, + "learning_rate": 1.2778700177834058e-05, + "loss": 0.0405, + "step": 26143 + }, + { + "epoch": 3.42, + "learning_rate": 1.2778190733761405e-05, + "loss": 0.044, + "step": 26144 + }, + { + "epoch": 3.42, + "learning_rate": 1.2777681281875239e-05, + "loss": 0.0395, + "step": 26145 + }, + { + "epoch": 3.42, + "learning_rate": 1.2777171822176992e-05, + "loss": 0.0351, + "step": 26146 + }, + { + "epoch": 3.43, + "learning_rate": 1.2776662354668096e-05, + "loss": 0.047, + "step": 26147 + }, + { + "epoch": 3.43, + "learning_rate": 1.2776152879349985e-05, + "loss": 0.0361, + "step": 26148 + }, + { + "epoch": 3.43, + "learning_rate": 1.2775643396224096e-05, + "loss": 0.037, + "step": 26149 + }, + { + "epoch": 3.43, + "learning_rate": 1.2775133905291854e-05, + "loss": 0.0403, + "step": 26150 + }, + { + "epoch": 3.43, + "learning_rate": 1.2774624406554695e-05, + "loss": 0.0404, + "step": 26151 + }, + { + "epoch": 3.43, + "learning_rate": 1.2774114900014052e-05, + "loss": 0.0416, + "step": 26152 + }, + { + "epoch": 3.43, + "learning_rate": 1.2773605385671365e-05, + "loss": 0.036, + "step": 26153 + }, + { + "epoch": 3.43, + "learning_rate": 1.2773095863528053e-05, + "loss": 0.0408, + "step": 26154 + }, + { + "epoch": 3.43, + "learning_rate": 1.277258633358556e-05, + "loss": 0.0456, + "step": 26155 + }, + { + "epoch": 3.43, + "learning_rate": 1.2772076795845313e-05, + "loss": 0.0412, + "step": 26156 + }, + { + "epoch": 3.43, + "learning_rate": 1.2771567250308752e-05, + "loss": 0.0415, + "step": 26157 + }, + { + "epoch": 3.43, + "learning_rate": 1.2771057696977303e-05, + "loss": 0.0426, + "step": 26158 + }, + { + "epoch": 3.43, + "learning_rate": 1.2770548135852403e-05, + "loss": 0.0331, + "step": 26159 + }, + { + "epoch": 3.43, + "learning_rate": 1.2770038566935483e-05, + "loss": 0.0369, + "step": 26160 + }, + { + "epoch": 3.43, + "learning_rate": 1.2769528990227977e-05, + "loss": 0.0378, + "step": 26161 + }, + { + "epoch": 3.43, + "learning_rate": 1.2769019405731317e-05, + "loss": 0.0368, + "step": 26162 + }, + { + "epoch": 3.43, + "learning_rate": 1.2768509813446942e-05, + "loss": 0.0381, + "step": 26163 + }, + { + "epoch": 3.43, + "learning_rate": 1.2768000213376277e-05, + "loss": 0.0395, + "step": 26164 + }, + { + "epoch": 3.43, + "learning_rate": 1.2767490605520757e-05, + "loss": 0.0395, + "step": 26165 + }, + { + "epoch": 3.43, + "learning_rate": 1.2766980989881818e-05, + "loss": 0.0418, + "step": 26166 + }, + { + "epoch": 3.43, + "learning_rate": 1.2766471366460898e-05, + "loss": 0.0504, + "step": 26167 + }, + { + "epoch": 3.43, + "learning_rate": 1.2765961735259417e-05, + "loss": 0.042, + "step": 26168 + }, + { + "epoch": 3.43, + "learning_rate": 1.276545209627882e-05, + "loss": 0.0406, + "step": 26169 + }, + { + "epoch": 3.43, + "learning_rate": 1.2764942449520536e-05, + "loss": 0.0439, + "step": 26170 + }, + { + "epoch": 3.43, + "learning_rate": 1.2764432794986e-05, + "loss": 0.041, + "step": 26171 + }, + { + "epoch": 3.43, + "learning_rate": 1.2763923132676638e-05, + "loss": 0.0382, + "step": 26172 + }, + { + "epoch": 3.43, + "learning_rate": 1.2763413462593894e-05, + "loss": 0.0397, + "step": 26173 + }, + { + "epoch": 3.43, + "learning_rate": 1.2762903784739198e-05, + "loss": 0.0459, + "step": 26174 + }, + { + "epoch": 3.43, + "learning_rate": 1.276239409911398e-05, + "loss": 0.0345, + "step": 26175 + }, + { + "epoch": 3.43, + "learning_rate": 1.2761884405719678e-05, + "loss": 0.0436, + "step": 26176 + }, + { + "epoch": 3.43, + "learning_rate": 1.2761374704557718e-05, + "loss": 0.0392, + "step": 26177 + }, + { + "epoch": 3.43, + "learning_rate": 1.2760864995629543e-05, + "loss": 0.0346, + "step": 26178 + }, + { + "epoch": 3.43, + "learning_rate": 1.276035527893658e-05, + "loss": 0.0491, + "step": 26179 + }, + { + "epoch": 3.43, + "learning_rate": 1.2759845554480269e-05, + "loss": 0.042, + "step": 26180 + }, + { + "epoch": 3.43, + "learning_rate": 1.2759335822262035e-05, + "loss": 0.0376, + "step": 26181 + }, + { + "epoch": 3.43, + "learning_rate": 1.275882608228332e-05, + "loss": 0.0368, + "step": 26182 + }, + { + "epoch": 3.43, + "learning_rate": 1.2758316334545546e-05, + "loss": 0.0359, + "step": 26183 + }, + { + "epoch": 3.43, + "learning_rate": 1.275780657905016e-05, + "loss": 0.0386, + "step": 26184 + }, + { + "epoch": 3.43, + "learning_rate": 1.275729681579859e-05, + "loss": 0.0336, + "step": 26185 + }, + { + "epoch": 3.43, + "learning_rate": 1.2756787044792268e-05, + "loss": 0.0408, + "step": 26186 + }, + { + "epoch": 3.43, + "learning_rate": 1.275627726603263e-05, + "loss": 0.0377, + "step": 26187 + }, + { + "epoch": 3.43, + "learning_rate": 1.2755767479521108e-05, + "loss": 0.0394, + "step": 26188 + }, + { + "epoch": 3.43, + "learning_rate": 1.2755257685259136e-05, + "loss": 0.0451, + "step": 26189 + }, + { + "epoch": 3.43, + "learning_rate": 1.2754747883248151e-05, + "loss": 0.0448, + "step": 26190 + }, + { + "epoch": 3.43, + "learning_rate": 1.2754238073489584e-05, + "loss": 0.0375, + "step": 26191 + }, + { + "epoch": 3.43, + "learning_rate": 1.2753728255984867e-05, + "loss": 0.0474, + "step": 26192 + }, + { + "epoch": 3.43, + "learning_rate": 1.2753218430735437e-05, + "loss": 0.0371, + "step": 26193 + }, + { + "epoch": 3.43, + "learning_rate": 1.2752708597742727e-05, + "loss": 0.0386, + "step": 26194 + }, + { + "epoch": 3.43, + "learning_rate": 1.2752198757008172e-05, + "loss": 0.0375, + "step": 26195 + }, + { + "epoch": 3.43, + "learning_rate": 1.2751688908533202e-05, + "loss": 0.0408, + "step": 26196 + }, + { + "epoch": 3.43, + "learning_rate": 1.2751179052319255e-05, + "loss": 0.0379, + "step": 26197 + }, + { + "epoch": 3.43, + "learning_rate": 1.2750669188367764e-05, + "loss": 0.0447, + "step": 26198 + }, + { + "epoch": 3.43, + "learning_rate": 1.275015931668016e-05, + "loss": 0.0376, + "step": 26199 + }, + { + "epoch": 3.43, + "learning_rate": 1.274964943725788e-05, + "loss": 0.0445, + "step": 26200 + }, + { + "epoch": 3.43, + "learning_rate": 1.274913955010236e-05, + "loss": 0.0442, + "step": 26201 + }, + { + "epoch": 3.43, + "learning_rate": 1.274862965521503e-05, + "loss": 0.0435, + "step": 26202 + }, + { + "epoch": 3.43, + "learning_rate": 1.2748119752597325e-05, + "loss": 0.0437, + "step": 26203 + }, + { + "epoch": 3.43, + "learning_rate": 1.2747609842250678e-05, + "loss": 0.0439, + "step": 26204 + }, + { + "epoch": 3.43, + "learning_rate": 1.2747099924176526e-05, + "loss": 0.0371, + "step": 26205 + }, + { + "epoch": 3.43, + "learning_rate": 1.2746589998376302e-05, + "loss": 0.0466, + "step": 26206 + }, + { + "epoch": 3.43, + "learning_rate": 1.2746080064851443e-05, + "loss": 0.0372, + "step": 26207 + }, + { + "epoch": 3.43, + "learning_rate": 1.2745570123603374e-05, + "loss": 0.0404, + "step": 26208 + }, + { + "epoch": 3.43, + "learning_rate": 1.2745060174633538e-05, + "loss": 0.0378, + "step": 26209 + }, + { + "epoch": 3.43, + "learning_rate": 1.2744550217943368e-05, + "loss": 0.0411, + "step": 26210 + }, + { + "epoch": 3.43, + "learning_rate": 1.2744040253534296e-05, + "loss": 0.0385, + "step": 26211 + }, + { + "epoch": 3.43, + "learning_rate": 1.2743530281407757e-05, + "loss": 0.0347, + "step": 26212 + }, + { + "epoch": 3.43, + "learning_rate": 1.2743020301565182e-05, + "loss": 0.0404, + "step": 26213 + }, + { + "epoch": 3.43, + "learning_rate": 1.2742510314008012e-05, + "loss": 0.044, + "step": 26214 + }, + { + "epoch": 3.43, + "learning_rate": 1.2742000318737676e-05, + "loss": 0.0407, + "step": 26215 + }, + { + "epoch": 3.43, + "learning_rate": 1.2741490315755611e-05, + "loss": 0.0376, + "step": 26216 + }, + { + "epoch": 3.43, + "learning_rate": 1.274098030506325e-05, + "loss": 0.0427, + "step": 26217 + }, + { + "epoch": 3.43, + "learning_rate": 1.274047028666203e-05, + "loss": 0.0452, + "step": 26218 + }, + { + "epoch": 3.43, + "learning_rate": 1.273996026055338e-05, + "loss": 0.0371, + "step": 26219 + }, + { + "epoch": 3.43, + "learning_rate": 1.2739450226738742e-05, + "loss": 0.0425, + "step": 26220 + }, + { + "epoch": 3.43, + "learning_rate": 1.273894018521954e-05, + "loss": 0.0451, + "step": 26221 + }, + { + "epoch": 3.43, + "learning_rate": 1.2738430135997217e-05, + "loss": 0.04, + "step": 26222 + }, + { + "epoch": 3.44, + "learning_rate": 1.2737920079073208e-05, + "loss": 0.0428, + "step": 26223 + }, + { + "epoch": 3.44, + "learning_rate": 1.2737410014448944e-05, + "loss": 0.0416, + "step": 26224 + }, + { + "epoch": 3.44, + "learning_rate": 1.2736899942125857e-05, + "loss": 0.0329, + "step": 26225 + }, + { + "epoch": 3.44, + "learning_rate": 1.2736389862105386e-05, + "loss": 0.0417, + "step": 26226 + }, + { + "epoch": 3.44, + "learning_rate": 1.2735879774388961e-05, + "loss": 0.0455, + "step": 26227 + }, + { + "epoch": 3.44, + "learning_rate": 1.2735369678978026e-05, + "loss": 0.0415, + "step": 26228 + }, + { + "epoch": 3.44, + "learning_rate": 1.2734859575874006e-05, + "loss": 0.0464, + "step": 26229 + }, + { + "epoch": 3.44, + "learning_rate": 1.2734349465078339e-05, + "loss": 0.0431, + "step": 26230 + }, + { + "epoch": 3.44, + "learning_rate": 1.2733839346592463e-05, + "loss": 0.038, + "step": 26231 + }, + { + "epoch": 3.44, + "learning_rate": 1.2733329220417807e-05, + "loss": 0.0421, + "step": 26232 + }, + { + "epoch": 3.44, + "learning_rate": 1.2732819086555805e-05, + "loss": 0.0372, + "step": 26233 + }, + { + "epoch": 3.44, + "learning_rate": 1.2732308945007898e-05, + "loss": 0.0433, + "step": 26234 + }, + { + "epoch": 3.44, + "learning_rate": 1.2731798795775521e-05, + "loss": 0.0377, + "step": 26235 + }, + { + "epoch": 3.44, + "learning_rate": 1.27312886388601e-05, + "loss": 0.0366, + "step": 26236 + }, + { + "epoch": 3.44, + "learning_rate": 1.2730778474263076e-05, + "loss": 0.0415, + "step": 26237 + }, + { + "epoch": 3.44, + "learning_rate": 1.2730268301985884e-05, + "loss": 0.0462, + "step": 26238 + }, + { + "epoch": 3.44, + "learning_rate": 1.2729758122029958e-05, + "loss": 0.0388, + "step": 26239 + }, + { + "epoch": 3.44, + "learning_rate": 1.2729247934396734e-05, + "loss": 0.0412, + "step": 26240 + }, + { + "epoch": 3.44, + "learning_rate": 1.2728737739087645e-05, + "loss": 0.0452, + "step": 26241 + }, + { + "epoch": 3.44, + "learning_rate": 1.2728227536104125e-05, + "loss": 0.0346, + "step": 26242 + }, + { + "epoch": 3.44, + "learning_rate": 1.272771732544761e-05, + "loss": 0.0395, + "step": 26243 + }, + { + "epoch": 3.44, + "learning_rate": 1.2727207107119535e-05, + "loss": 0.0411, + "step": 26244 + }, + { + "epoch": 3.44, + "learning_rate": 1.2726696881121336e-05, + "loss": 0.0365, + "step": 26245 + }, + { + "epoch": 3.44, + "learning_rate": 1.2726186647454448e-05, + "loss": 0.039, + "step": 26246 + }, + { + "epoch": 3.44, + "learning_rate": 1.2725676406120306e-05, + "loss": 0.0409, + "step": 26247 + }, + { + "epoch": 3.44, + "learning_rate": 1.2725166157120341e-05, + "loss": 0.0359, + "step": 26248 + }, + { + "epoch": 3.44, + "learning_rate": 1.2724655900455995e-05, + "loss": 0.0354, + "step": 26249 + }, + { + "epoch": 3.44, + "learning_rate": 1.2724145636128698e-05, + "loss": 0.0428, + "step": 26250 + }, + { + "epoch": 3.44, + "learning_rate": 1.2723635364139886e-05, + "loss": 0.0369, + "step": 26251 + }, + { + "epoch": 3.44, + "learning_rate": 1.2723125084490995e-05, + "loss": 0.0331, + "step": 26252 + }, + { + "epoch": 3.44, + "learning_rate": 1.272261479718346e-05, + "loss": 0.0335, + "step": 26253 + }, + { + "epoch": 3.44, + "learning_rate": 1.2722104502218713e-05, + "loss": 0.0341, + "step": 26254 + }, + { + "epoch": 3.44, + "learning_rate": 1.2721594199598193e-05, + "loss": 0.0398, + "step": 26255 + }, + { + "epoch": 3.44, + "learning_rate": 1.2721083889323338e-05, + "loss": 0.041, + "step": 26256 + }, + { + "epoch": 3.44, + "learning_rate": 1.2720573571395576e-05, + "loss": 0.0416, + "step": 26257 + }, + { + "epoch": 3.44, + "learning_rate": 1.2720063245816348e-05, + "loss": 0.0386, + "step": 26258 + }, + { + "epoch": 3.44, + "learning_rate": 1.2719552912587083e-05, + "loss": 0.0439, + "step": 26259 + }, + { + "epoch": 3.44, + "learning_rate": 1.2719042571709226e-05, + "loss": 0.0352, + "step": 26260 + }, + { + "epoch": 3.44, + "learning_rate": 1.2718532223184204e-05, + "loss": 0.0397, + "step": 26261 + }, + { + "epoch": 3.44, + "learning_rate": 1.2718021867013455e-05, + "loss": 0.0395, + "step": 26262 + }, + { + "epoch": 3.44, + "learning_rate": 1.2717511503198411e-05, + "loss": 0.0346, + "step": 26263 + }, + { + "epoch": 3.44, + "learning_rate": 1.2717001131740515e-05, + "loss": 0.0347, + "step": 26264 + }, + { + "epoch": 3.44, + "learning_rate": 1.2716490752641195e-05, + "loss": 0.0426, + "step": 26265 + }, + { + "epoch": 3.44, + "learning_rate": 1.2715980365901892e-05, + "loss": 0.0404, + "step": 26266 + }, + { + "epoch": 3.44, + "learning_rate": 1.2715469971524036e-05, + "loss": 0.0442, + "step": 26267 + }, + { + "epoch": 3.44, + "learning_rate": 1.2714959569509069e-05, + "loss": 0.0436, + "step": 26268 + }, + { + "epoch": 3.44, + "learning_rate": 1.271444915985842e-05, + "loss": 0.0362, + "step": 26269 + }, + { + "epoch": 3.44, + "learning_rate": 1.2713938742573528e-05, + "loss": 0.0433, + "step": 26270 + }, + { + "epoch": 3.44, + "learning_rate": 1.2713428317655827e-05, + "loss": 0.0373, + "step": 26271 + }, + { + "epoch": 3.44, + "learning_rate": 1.2712917885106754e-05, + "loss": 0.0358, + "step": 26272 + }, + { + "epoch": 3.44, + "learning_rate": 1.2712407444927745e-05, + "loss": 0.0421, + "step": 26273 + }, + { + "epoch": 3.44, + "learning_rate": 1.2711896997120231e-05, + "loss": 0.0447, + "step": 26274 + }, + { + "epoch": 3.44, + "learning_rate": 1.2711386541685656e-05, + "loss": 0.0446, + "step": 26275 + }, + { + "epoch": 3.44, + "learning_rate": 1.2710876078625449e-05, + "loss": 0.0428, + "step": 26276 + }, + { + "epoch": 3.44, + "learning_rate": 1.2710365607941046e-05, + "loss": 0.0404, + "step": 26277 + }, + { + "epoch": 3.44, + "learning_rate": 1.2709855129633883e-05, + "loss": 0.0395, + "step": 26278 + }, + { + "epoch": 3.44, + "learning_rate": 1.2709344643705401e-05, + "loss": 0.0435, + "step": 26279 + }, + { + "epoch": 3.44, + "learning_rate": 1.2708834150157028e-05, + "loss": 0.0424, + "step": 26280 + }, + { + "epoch": 3.44, + "learning_rate": 1.2708323648990205e-05, + "loss": 0.0416, + "step": 26281 + }, + { + "epoch": 3.44, + "learning_rate": 1.2707813140206365e-05, + "loss": 0.0484, + "step": 26282 + }, + { + "epoch": 3.44, + "learning_rate": 1.2707302623806945e-05, + "loss": 0.0331, + "step": 26283 + }, + { + "epoch": 3.44, + "learning_rate": 1.2706792099793381e-05, + "loss": 0.0407, + "step": 26284 + }, + { + "epoch": 3.44, + "learning_rate": 1.270628156816711e-05, + "loss": 0.04, + "step": 26285 + }, + { + "epoch": 3.44, + "learning_rate": 1.2705771028929562e-05, + "loss": 0.0382, + "step": 26286 + }, + { + "epoch": 3.44, + "learning_rate": 1.270526048208218e-05, + "loss": 0.0377, + "step": 26287 + }, + { + "epoch": 3.44, + "learning_rate": 1.2704749927626394e-05, + "loss": 0.0361, + "step": 26288 + }, + { + "epoch": 3.44, + "learning_rate": 1.2704239365563646e-05, + "loss": 0.0385, + "step": 26289 + }, + { + "epoch": 3.44, + "learning_rate": 1.270372879589537e-05, + "loss": 0.0315, + "step": 26290 + }, + { + "epoch": 3.44, + "learning_rate": 1.2703218218623e-05, + "loss": 0.0425, + "step": 26291 + }, + { + "epoch": 3.44, + "learning_rate": 1.2702707633747968e-05, + "loss": 0.0462, + "step": 26292 + }, + { + "epoch": 3.44, + "learning_rate": 1.2702197041271718e-05, + "loss": 0.0347, + "step": 26293 + }, + { + "epoch": 3.44, + "learning_rate": 1.2701686441195683e-05, + "loss": 0.0393, + "step": 26294 + }, + { + "epoch": 3.44, + "learning_rate": 1.2701175833521297e-05, + "loss": 0.0368, + "step": 26295 + }, + { + "epoch": 3.44, + "learning_rate": 1.2700665218250002e-05, + "loss": 0.0381, + "step": 26296 + }, + { + "epoch": 3.44, + "learning_rate": 1.2700154595383225e-05, + "loss": 0.04, + "step": 26297 + }, + { + "epoch": 3.44, + "learning_rate": 1.269964396492241e-05, + "loss": 0.0352, + "step": 26298 + }, + { + "epoch": 3.44, + "learning_rate": 1.2699133326868988e-05, + "loss": 0.0397, + "step": 26299 + }, + { + "epoch": 3.45, + "learning_rate": 1.2698622681224399e-05, + "loss": 0.0359, + "step": 26300 + }, + { + "epoch": 3.45, + "learning_rate": 1.2698112027990077e-05, + "loss": 0.0392, + "step": 26301 + }, + { + "epoch": 3.45, + "learning_rate": 1.269760136716746e-05, + "loss": 0.0407, + "step": 26302 + }, + { + "epoch": 3.45, + "learning_rate": 1.2697090698757978e-05, + "loss": 0.0335, + "step": 26303 + }, + { + "epoch": 3.45, + "learning_rate": 1.2696580022763079e-05, + "loss": 0.0389, + "step": 26304 + }, + { + "epoch": 3.45, + "learning_rate": 1.2696069339184187e-05, + "loss": 0.0404, + "step": 26305 + }, + { + "epoch": 3.45, + "learning_rate": 1.2695558648022747e-05, + "loss": 0.033, + "step": 26306 + }, + { + "epoch": 3.45, + "learning_rate": 1.269504794928019e-05, + "loss": 0.0395, + "step": 26307 + }, + { + "epoch": 3.45, + "learning_rate": 1.2694537242957953e-05, + "loss": 0.0489, + "step": 26308 + }, + { + "epoch": 3.45, + "learning_rate": 1.2694026529057475e-05, + "loss": 0.047, + "step": 26309 + }, + { + "epoch": 3.45, + "learning_rate": 1.2693515807580193e-05, + "loss": 0.0318, + "step": 26310 + }, + { + "epoch": 3.45, + "learning_rate": 1.2693005078527536e-05, + "loss": 0.0389, + "step": 26311 + }, + { + "epoch": 3.45, + "learning_rate": 1.2692494341900952e-05, + "loss": 0.0382, + "step": 26312 + }, + { + "epoch": 3.45, + "learning_rate": 1.2691983597701865e-05, + "loss": 0.0438, + "step": 26313 + }, + { + "epoch": 3.45, + "learning_rate": 1.269147284593172e-05, + "loss": 0.0455, + "step": 26314 + }, + { + "epoch": 3.45, + "learning_rate": 1.2690962086591952e-05, + "loss": 0.0423, + "step": 26315 + }, + { + "epoch": 3.45, + "learning_rate": 1.2690451319683996e-05, + "loss": 0.041, + "step": 26316 + }, + { + "epoch": 3.45, + "learning_rate": 1.2689940545209287e-05, + "loss": 0.0393, + "step": 26317 + }, + { + "epoch": 3.45, + "learning_rate": 1.2689429763169266e-05, + "loss": 0.0357, + "step": 26318 + }, + { + "epoch": 3.45, + "learning_rate": 1.2688918973565362e-05, + "loss": 0.039, + "step": 26319 + }, + { + "epoch": 3.45, + "learning_rate": 1.2688408176399021e-05, + "loss": 0.0429, + "step": 26320 + }, + { + "epoch": 3.45, + "learning_rate": 1.2687897371671675e-05, + "loss": 0.0376, + "step": 26321 + }, + { + "epoch": 3.45, + "learning_rate": 1.268738655938476e-05, + "loss": 0.0334, + "step": 26322 + }, + { + "epoch": 3.45, + "learning_rate": 1.2686875739539716e-05, + "loss": 0.0383, + "step": 26323 + }, + { + "epoch": 3.45, + "learning_rate": 1.2686364912137972e-05, + "loss": 0.0387, + "step": 26324 + }, + { + "epoch": 3.45, + "learning_rate": 1.2685854077180974e-05, + "loss": 0.0394, + "step": 26325 + }, + { + "epoch": 3.45, + "learning_rate": 1.2685343234670154e-05, + "loss": 0.0352, + "step": 26326 + }, + { + "epoch": 3.45, + "learning_rate": 1.2684832384606948e-05, + "loss": 0.039, + "step": 26327 + }, + { + "epoch": 3.45, + "learning_rate": 1.2684321526992794e-05, + "loss": 0.0443, + "step": 26328 + }, + { + "epoch": 3.45, + "learning_rate": 1.268381066182913e-05, + "loss": 0.0451, + "step": 26329 + }, + { + "epoch": 3.45, + "learning_rate": 1.2683299789117386e-05, + "loss": 0.0354, + "step": 26330 + }, + { + "epoch": 3.45, + "learning_rate": 1.2682788908859012e-05, + "loss": 0.0405, + "step": 26331 + }, + { + "epoch": 3.45, + "learning_rate": 1.2682278021055432e-05, + "loss": 0.0337, + "step": 26332 + }, + { + "epoch": 3.45, + "learning_rate": 1.2681767125708091e-05, + "loss": 0.0369, + "step": 26333 + }, + { + "epoch": 3.45, + "learning_rate": 1.2681256222818422e-05, + "loss": 0.0379, + "step": 26334 + }, + { + "epoch": 3.45, + "learning_rate": 1.2680745312387859e-05, + "loss": 0.0408, + "step": 26335 + }, + { + "epoch": 3.45, + "learning_rate": 1.2680234394417845e-05, + "loss": 0.0341, + "step": 26336 + }, + { + "epoch": 3.45, + "learning_rate": 1.2679723468909817e-05, + "loss": 0.0327, + "step": 26337 + }, + { + "epoch": 3.45, + "learning_rate": 1.2679212535865206e-05, + "loss": 0.0404, + "step": 26338 + }, + { + "epoch": 3.45, + "learning_rate": 1.2678701595285454e-05, + "loss": 0.0397, + "step": 26339 + }, + { + "epoch": 3.45, + "learning_rate": 1.2678190647171997e-05, + "loss": 0.0412, + "step": 26340 + }, + { + "epoch": 3.45, + "learning_rate": 1.267767969152627e-05, + "loss": 0.0361, + "step": 26341 + }, + { + "epoch": 3.45, + "learning_rate": 1.2677168728349712e-05, + "loss": 0.0453, + "step": 26342 + }, + { + "epoch": 3.45, + "learning_rate": 1.2676657757643757e-05, + "loss": 0.046, + "step": 26343 + }, + { + "epoch": 3.45, + "learning_rate": 1.267614677940985e-05, + "loss": 0.0413, + "step": 26344 + }, + { + "epoch": 3.45, + "learning_rate": 1.267563579364942e-05, + "loss": 0.046, + "step": 26345 + }, + { + "epoch": 3.45, + "learning_rate": 1.2675124800363904e-05, + "loss": 0.0455, + "step": 26346 + }, + { + "epoch": 3.45, + "learning_rate": 1.2674613799554744e-05, + "loss": 0.0358, + "step": 26347 + }, + { + "epoch": 3.45, + "learning_rate": 1.2674102791223378e-05, + "loss": 0.0339, + "step": 26348 + }, + { + "epoch": 3.45, + "learning_rate": 1.2673591775371236e-05, + "loss": 0.0393, + "step": 26349 + }, + { + "epoch": 3.45, + "learning_rate": 1.2673080751999765e-05, + "loss": 0.0323, + "step": 26350 + }, + { + "epoch": 3.45, + "learning_rate": 1.2672569721110393e-05, + "loss": 0.0367, + "step": 26351 + }, + { + "epoch": 3.45, + "learning_rate": 1.267205868270456e-05, + "loss": 0.0337, + "step": 26352 + }, + { + "epoch": 3.45, + "learning_rate": 1.2671547636783706e-05, + "loss": 0.0385, + "step": 26353 + }, + { + "epoch": 3.45, + "learning_rate": 1.2671036583349268e-05, + "loss": 0.0346, + "step": 26354 + }, + { + "epoch": 3.45, + "learning_rate": 1.267052552240268e-05, + "loss": 0.0408, + "step": 26355 + }, + { + "epoch": 3.45, + "learning_rate": 1.2670014453945379e-05, + "loss": 0.0367, + "step": 26356 + }, + { + "epoch": 3.45, + "learning_rate": 1.2669503377978804e-05, + "loss": 0.0406, + "step": 26357 + }, + { + "epoch": 3.45, + "learning_rate": 1.26689922945044e-05, + "loss": 0.0371, + "step": 26358 + }, + { + "epoch": 3.45, + "learning_rate": 1.266848120352359e-05, + "loss": 0.0457, + "step": 26359 + }, + { + "epoch": 3.45, + "learning_rate": 1.266797010503782e-05, + "loss": 0.0334, + "step": 26360 + }, + { + "epoch": 3.45, + "learning_rate": 1.266745899904853e-05, + "loss": 0.0391, + "step": 26361 + }, + { + "epoch": 3.45, + "learning_rate": 1.2666947885557152e-05, + "loss": 0.0373, + "step": 26362 + }, + { + "epoch": 3.45, + "learning_rate": 1.2666436764565125e-05, + "loss": 0.0367, + "step": 26363 + }, + { + "epoch": 3.45, + "learning_rate": 1.2665925636073883e-05, + "loss": 0.0395, + "step": 26364 + }, + { + "epoch": 3.45, + "learning_rate": 1.2665414500084871e-05, + "loss": 0.0383, + "step": 26365 + }, + { + "epoch": 3.45, + "learning_rate": 1.2664903356599524e-05, + "loss": 0.0342, + "step": 26366 + }, + { + "epoch": 3.45, + "learning_rate": 1.2664392205619276e-05, + "loss": 0.0387, + "step": 26367 + }, + { + "epoch": 3.45, + "learning_rate": 1.2663881047145563e-05, + "loss": 0.0403, + "step": 26368 + }, + { + "epoch": 3.45, + "learning_rate": 1.2663369881179832e-05, + "loss": 0.0311, + "step": 26369 + }, + { + "epoch": 3.45, + "learning_rate": 1.2662858707723512e-05, + "loss": 0.0422, + "step": 26370 + }, + { + "epoch": 3.45, + "learning_rate": 1.2662347526778048e-05, + "loss": 0.0471, + "step": 26371 + }, + { + "epoch": 3.45, + "learning_rate": 1.2661836338344871e-05, + "loss": 0.0408, + "step": 26372 + }, + { + "epoch": 3.45, + "learning_rate": 1.266132514242542e-05, + "loss": 0.0415, + "step": 26373 + }, + { + "epoch": 3.45, + "learning_rate": 1.266081393902113e-05, + "loss": 0.0438, + "step": 26374 + }, + { + "epoch": 3.45, + "learning_rate": 1.2660302728133449e-05, + "loss": 0.0377, + "step": 26375 + }, + { + "epoch": 3.46, + "learning_rate": 1.2659791509763805e-05, + "loss": 0.0362, + "step": 26376 + }, + { + "epoch": 3.46, + "learning_rate": 1.2659280283913643e-05, + "loss": 0.0337, + "step": 26377 + }, + { + "epoch": 3.46, + "learning_rate": 1.2658769050584393e-05, + "loss": 0.0399, + "step": 26378 + }, + { + "epoch": 3.46, + "learning_rate": 1.2658257809777497e-05, + "loss": 0.0365, + "step": 26379 + }, + { + "epoch": 3.46, + "learning_rate": 1.265774656149439e-05, + "loss": 0.0369, + "step": 26380 + }, + { + "epoch": 3.46, + "learning_rate": 1.2657235305736517e-05, + "loss": 0.0468, + "step": 26381 + }, + { + "epoch": 3.46, + "learning_rate": 1.2656724042505309e-05, + "loss": 0.0458, + "step": 26382 + }, + { + "epoch": 3.46, + "learning_rate": 1.2656212771802205e-05, + "loss": 0.0433, + "step": 26383 + }, + { + "epoch": 3.46, + "learning_rate": 1.2655701493628648e-05, + "loss": 0.0355, + "step": 26384 + }, + { + "epoch": 3.46, + "learning_rate": 1.2655190207986068e-05, + "loss": 0.0424, + "step": 26385 + }, + { + "epoch": 3.46, + "learning_rate": 1.2654678914875908e-05, + "loss": 0.044, + "step": 26386 + }, + { + "epoch": 3.46, + "learning_rate": 1.2654167614299606e-05, + "loss": 0.0473, + "step": 26387 + }, + { + "epoch": 3.46, + "learning_rate": 1.26536563062586e-05, + "loss": 0.0414, + "step": 26388 + }, + { + "epoch": 3.46, + "learning_rate": 1.2653144990754324e-05, + "loss": 0.0367, + "step": 26389 + }, + { + "epoch": 3.46, + "learning_rate": 1.265263366778822e-05, + "loss": 0.041, + "step": 26390 + }, + { + "epoch": 3.46, + "learning_rate": 1.2652122337361722e-05, + "loss": 0.0366, + "step": 26391 + }, + { + "epoch": 3.46, + "learning_rate": 1.2651610999476279e-05, + "loss": 0.0407, + "step": 26392 + }, + { + "epoch": 3.46, + "learning_rate": 1.2651099654133315e-05, + "loss": 0.0372, + "step": 26393 + }, + { + "epoch": 3.46, + "learning_rate": 1.2650588301334276e-05, + "loss": 0.0395, + "step": 26394 + }, + { + "epoch": 3.46, + "learning_rate": 1.2650076941080598e-05, + "loss": 0.0438, + "step": 26395 + }, + { + "epoch": 3.46, + "learning_rate": 1.2649565573373721e-05, + "loss": 0.0331, + "step": 26396 + }, + { + "epoch": 3.46, + "learning_rate": 1.2649054198215081e-05, + "loss": 0.0387, + "step": 26397 + }, + { + "epoch": 3.46, + "learning_rate": 1.264854281560612e-05, + "loss": 0.04, + "step": 26398 + }, + { + "epoch": 3.46, + "learning_rate": 1.2648031425548268e-05, + "loss": 0.0355, + "step": 26399 + }, + { + "epoch": 3.46, + "learning_rate": 1.2647520028042972e-05, + "loss": 0.0409, + "step": 26400 + }, + { + "epoch": 3.46, + "learning_rate": 1.2647008623091666e-05, + "loss": 0.0347, + "step": 26401 + }, + { + "epoch": 3.46, + "learning_rate": 1.2646497210695792e-05, + "loss": 0.0339, + "step": 26402 + }, + { + "epoch": 3.46, + "learning_rate": 1.2645985790856781e-05, + "loss": 0.0333, + "step": 26403 + }, + { + "epoch": 3.46, + "learning_rate": 1.2645474363576079e-05, + "loss": 0.0367, + "step": 26404 + }, + { + "epoch": 3.46, + "learning_rate": 1.264496292885512e-05, + "loss": 0.0345, + "step": 26405 + }, + { + "epoch": 3.46, + "learning_rate": 1.2644451486695341e-05, + "loss": 0.0394, + "step": 26406 + }, + { + "epoch": 3.46, + "learning_rate": 1.2643940037098189e-05, + "loss": 0.0411, + "step": 26407 + }, + { + "epoch": 3.46, + "learning_rate": 1.2643428580065091e-05, + "loss": 0.0452, + "step": 26408 + }, + { + "epoch": 3.46, + "learning_rate": 1.2642917115597495e-05, + "loss": 0.0343, + "step": 26409 + }, + { + "epoch": 3.46, + "learning_rate": 1.2642405643696833e-05, + "loss": 0.0389, + "step": 26410 + }, + { + "epoch": 3.46, + "learning_rate": 1.2641894164364548e-05, + "loss": 0.0281, + "step": 26411 + }, + { + "epoch": 3.46, + "learning_rate": 1.2641382677602071e-05, + "loss": 0.0428, + "step": 26412 + }, + { + "epoch": 3.46, + "learning_rate": 1.2640871183410851e-05, + "loss": 0.0416, + "step": 26413 + }, + { + "epoch": 3.46, + "learning_rate": 1.264035968179232e-05, + "loss": 0.0424, + "step": 26414 + }, + { + "epoch": 3.46, + "learning_rate": 1.2639848172747921e-05, + "loss": 0.0412, + "step": 26415 + }, + { + "epoch": 3.46, + "learning_rate": 1.2639336656279086e-05, + "loss": 0.0397, + "step": 26416 + }, + { + "epoch": 3.46, + "learning_rate": 1.2638825132387255e-05, + "loss": 0.0411, + "step": 26417 + }, + { + "epoch": 3.46, + "learning_rate": 1.2638313601073873e-05, + "loss": 0.0387, + "step": 26418 + }, + { + "epoch": 3.46, + "learning_rate": 1.2637802062340375e-05, + "loss": 0.0422, + "step": 26419 + }, + { + "epoch": 3.46, + "learning_rate": 1.2637290516188196e-05, + "loss": 0.0435, + "step": 26420 + }, + { + "epoch": 3.46, + "learning_rate": 1.263677896261878e-05, + "loss": 0.0449, + "step": 26421 + }, + { + "epoch": 3.46, + "learning_rate": 1.2636267401633565e-05, + "loss": 0.038, + "step": 26422 + }, + { + "epoch": 3.46, + "learning_rate": 1.2635755833233985e-05, + "loss": 0.0396, + "step": 26423 + }, + { + "epoch": 3.46, + "learning_rate": 1.2635244257421482e-05, + "loss": 0.0479, + "step": 26424 + }, + { + "epoch": 3.46, + "learning_rate": 1.2634732674197494e-05, + "loss": 0.0378, + "step": 26425 + }, + { + "epoch": 3.46, + "learning_rate": 1.2634221083563465e-05, + "loss": 0.0365, + "step": 26426 + }, + { + "epoch": 3.46, + "learning_rate": 1.2633709485520827e-05, + "loss": 0.0381, + "step": 26427 + }, + { + "epoch": 3.46, + "learning_rate": 1.263319788007102e-05, + "loss": 0.0461, + "step": 26428 + }, + { + "epoch": 3.46, + "learning_rate": 1.2632686267215486e-05, + "loss": 0.0391, + "step": 26429 + }, + { + "epoch": 3.46, + "learning_rate": 1.2632174646955662e-05, + "loss": 0.0392, + "step": 26430 + }, + { + "epoch": 3.46, + "learning_rate": 1.2631663019292986e-05, + "loss": 0.0417, + "step": 26431 + }, + { + "epoch": 3.46, + "learning_rate": 1.2631151384228901e-05, + "loss": 0.0391, + "step": 26432 + }, + { + "epoch": 3.46, + "learning_rate": 1.2630639741764838e-05, + "loss": 0.0341, + "step": 26433 + }, + { + "epoch": 3.46, + "learning_rate": 1.2630128091902246e-05, + "loss": 0.0342, + "step": 26434 + }, + { + "epoch": 3.46, + "learning_rate": 1.2629616434642555e-05, + "loss": 0.0357, + "step": 26435 + }, + { + "epoch": 3.46, + "learning_rate": 1.2629104769987208e-05, + "loss": 0.037, + "step": 26436 + }, + { + "epoch": 3.46, + "learning_rate": 1.2628593097937646e-05, + "loss": 0.0331, + "step": 26437 + }, + { + "epoch": 3.46, + "learning_rate": 1.2628081418495306e-05, + "loss": 0.0305, + "step": 26438 + }, + { + "epoch": 3.46, + "learning_rate": 1.2627569731661625e-05, + "loss": 0.0393, + "step": 26439 + }, + { + "epoch": 3.46, + "learning_rate": 1.2627058037438045e-05, + "loss": 0.0415, + "step": 26440 + }, + { + "epoch": 3.46, + "learning_rate": 1.2626546335826001e-05, + "loss": 0.0452, + "step": 26441 + }, + { + "epoch": 3.46, + "learning_rate": 1.262603462682694e-05, + "loss": 0.0316, + "step": 26442 + }, + { + "epoch": 3.46, + "learning_rate": 1.2625522910442296e-05, + "loss": 0.0421, + "step": 26443 + }, + { + "epoch": 3.46, + "learning_rate": 1.2625011186673505e-05, + "loss": 0.0384, + "step": 26444 + }, + { + "epoch": 3.46, + "learning_rate": 1.2624499455522013e-05, + "loss": 0.0403, + "step": 26445 + }, + { + "epoch": 3.46, + "learning_rate": 1.2623987716989254e-05, + "loss": 0.0388, + "step": 26446 + }, + { + "epoch": 3.46, + "learning_rate": 1.262347597107667e-05, + "loss": 0.0419, + "step": 26447 + }, + { + "epoch": 3.46, + "learning_rate": 1.26229642177857e-05, + "loss": 0.0351, + "step": 26448 + }, + { + "epoch": 3.46, + "learning_rate": 1.2622452457117786e-05, + "loss": 0.0414, + "step": 26449 + }, + { + "epoch": 3.46, + "learning_rate": 1.2621940689074357e-05, + "loss": 0.0383, + "step": 26450 + }, + { + "epoch": 3.46, + "learning_rate": 1.2621428913656867e-05, + "loss": 0.0374, + "step": 26451 + }, + { + "epoch": 3.47, + "learning_rate": 1.262091713086674e-05, + "loss": 0.0406, + "step": 26452 + }, + { + "epoch": 3.47, + "learning_rate": 1.2620405340705431e-05, + "loss": 0.0373, + "step": 26453 + }, + { + "epoch": 3.47, + "learning_rate": 1.2619893543174367e-05, + "loss": 0.0348, + "step": 26454 + }, + { + "epoch": 3.47, + "learning_rate": 1.2619381738274992e-05, + "loss": 0.0375, + "step": 26455 + }, + { + "epoch": 3.47, + "learning_rate": 1.2618869926008745e-05, + "loss": 0.0354, + "step": 26456 + }, + { + "epoch": 3.47, + "learning_rate": 1.2618358106377068e-05, + "loss": 0.0481, + "step": 26457 + }, + { + "epoch": 3.47, + "learning_rate": 1.2617846279381399e-05, + "loss": 0.0416, + "step": 26458 + }, + { + "epoch": 3.47, + "learning_rate": 1.2617334445023175e-05, + "loss": 0.0359, + "step": 26459 + }, + { + "epoch": 3.47, + "learning_rate": 1.2616822603303836e-05, + "loss": 0.0287, + "step": 26460 + }, + { + "epoch": 3.47, + "learning_rate": 1.2616310754224823e-05, + "loss": 0.0339, + "step": 26461 + }, + { + "epoch": 3.47, + "learning_rate": 1.2615798897787576e-05, + "loss": 0.0402, + "step": 26462 + }, + { + "epoch": 3.47, + "learning_rate": 1.2615287033993535e-05, + "loss": 0.0378, + "step": 26463 + }, + { + "epoch": 3.47, + "learning_rate": 1.2614775162844137e-05, + "loss": 0.0402, + "step": 26464 + }, + { + "epoch": 3.47, + "learning_rate": 1.2614263284340825e-05, + "loss": 0.0442, + "step": 26465 + }, + { + "epoch": 3.47, + "learning_rate": 1.2613751398485033e-05, + "loss": 0.0398, + "step": 26466 + }, + { + "epoch": 3.47, + "learning_rate": 1.2613239505278209e-05, + "loss": 0.0359, + "step": 26467 + }, + { + "epoch": 3.47, + "learning_rate": 1.2612727604721786e-05, + "loss": 0.0438, + "step": 26468 + }, + { + "epoch": 3.47, + "learning_rate": 1.2612215696817204e-05, + "loss": 0.0415, + "step": 26469 + }, + { + "epoch": 3.47, + "learning_rate": 1.2611703781565908e-05, + "loss": 0.0424, + "step": 26470 + }, + { + "epoch": 3.47, + "learning_rate": 1.2611191858969331e-05, + "loss": 0.0422, + "step": 26471 + }, + { + "epoch": 3.47, + "learning_rate": 1.2610679929028917e-05, + "loss": 0.0499, + "step": 26472 + }, + { + "epoch": 3.47, + "learning_rate": 1.2610167991746103e-05, + "loss": 0.0402, + "step": 26473 + }, + { + "epoch": 3.47, + "learning_rate": 1.2609656047122334e-05, + "loss": 0.034, + "step": 26474 + }, + { + "epoch": 3.47, + "learning_rate": 1.2609144095159045e-05, + "loss": 0.0423, + "step": 26475 + }, + { + "epoch": 3.47, + "learning_rate": 1.2608632135857679e-05, + "loss": 0.0391, + "step": 26476 + }, + { + "epoch": 3.47, + "learning_rate": 1.2608120169219668e-05, + "loss": 0.0388, + "step": 26477 + }, + { + "epoch": 3.47, + "learning_rate": 1.2607608195246464e-05, + "loss": 0.0426, + "step": 26478 + }, + { + "epoch": 3.47, + "learning_rate": 1.2607096213939498e-05, + "loss": 0.0419, + "step": 26479 + }, + { + "epoch": 3.47, + "learning_rate": 1.2606584225300215e-05, + "loss": 0.0415, + "step": 26480 + }, + { + "epoch": 3.47, + "learning_rate": 1.260607222933005e-05, + "loss": 0.0439, + "step": 26481 + }, + { + "epoch": 3.47, + "learning_rate": 1.2605560226030446e-05, + "loss": 0.0371, + "step": 26482 + }, + { + "epoch": 3.47, + "learning_rate": 1.2605048215402843e-05, + "loss": 0.0383, + "step": 26483 + }, + { + "epoch": 3.47, + "learning_rate": 1.2604536197448685e-05, + "loss": 0.0347, + "step": 26484 + }, + { + "epoch": 3.47, + "learning_rate": 1.2604024172169403e-05, + "loss": 0.0369, + "step": 26485 + }, + { + "epoch": 3.47, + "learning_rate": 1.260351213956644e-05, + "loss": 0.0385, + "step": 26486 + }, + { + "epoch": 3.47, + "learning_rate": 1.2603000099641242e-05, + "loss": 0.0418, + "step": 26487 + }, + { + "epoch": 3.47, + "learning_rate": 1.2602488052395243e-05, + "loss": 0.0414, + "step": 26488 + }, + { + "epoch": 3.47, + "learning_rate": 1.2601975997829885e-05, + "loss": 0.0414, + "step": 26489 + }, + { + "epoch": 3.47, + "learning_rate": 1.2601463935946607e-05, + "loss": 0.0434, + "step": 26490 + }, + { + "epoch": 3.47, + "learning_rate": 1.2600951866746854e-05, + "loss": 0.044, + "step": 26491 + }, + { + "epoch": 3.47, + "learning_rate": 1.2600439790232057e-05, + "loss": 0.0343, + "step": 26492 + }, + { + "epoch": 3.47, + "learning_rate": 1.2599927706403665e-05, + "loss": 0.0372, + "step": 26493 + }, + { + "epoch": 3.47, + "learning_rate": 1.2599415615263113e-05, + "loss": 0.0315, + "step": 26494 + }, + { + "epoch": 3.47, + "learning_rate": 1.2598903516811844e-05, + "loss": 0.0351, + "step": 26495 + }, + { + "epoch": 3.47, + "learning_rate": 1.2598391411051298e-05, + "loss": 0.0374, + "step": 26496 + }, + { + "epoch": 3.47, + "learning_rate": 1.2597879297982916e-05, + "loss": 0.039, + "step": 26497 + }, + { + "epoch": 3.47, + "learning_rate": 1.2597367177608134e-05, + "loss": 0.0358, + "step": 26498 + }, + { + "epoch": 3.47, + "learning_rate": 1.2596855049928394e-05, + "loss": 0.0405, + "step": 26499 + }, + { + "epoch": 3.47, + "learning_rate": 1.2596342914945137e-05, + "loss": 0.0282, + "step": 26500 + }, + { + "epoch": 3.47, + "learning_rate": 1.259583077265981e-05, + "loss": 0.0348, + "step": 26501 + }, + { + "epoch": 3.47, + "learning_rate": 1.2595318623073842e-05, + "loss": 0.0367, + "step": 26502 + }, + { + "epoch": 3.47, + "learning_rate": 1.259480646618868e-05, + "loss": 0.0448, + "step": 26503 + }, + { + "epoch": 3.47, + "learning_rate": 1.2594294302005763e-05, + "loss": 0.0338, + "step": 26504 + }, + { + "epoch": 3.47, + "learning_rate": 1.259378213052653e-05, + "loss": 0.0405, + "step": 26505 + }, + { + "epoch": 3.47, + "learning_rate": 1.2593269951752422e-05, + "loss": 0.0496, + "step": 26506 + }, + { + "epoch": 3.47, + "learning_rate": 1.2592757765684884e-05, + "loss": 0.0442, + "step": 26507 + }, + { + "epoch": 3.47, + "learning_rate": 1.2592245572325352e-05, + "loss": 0.0442, + "step": 26508 + }, + { + "epoch": 3.47, + "learning_rate": 1.2591733371675266e-05, + "loss": 0.0376, + "step": 26509 + }, + { + "epoch": 3.47, + "learning_rate": 1.2591221163736068e-05, + "loss": 0.0437, + "step": 26510 + }, + { + "epoch": 3.47, + "learning_rate": 1.2590708948509198e-05, + "loss": 0.0409, + "step": 26511 + }, + { + "epoch": 3.47, + "learning_rate": 1.25901967259961e-05, + "loss": 0.0348, + "step": 26512 + }, + { + "epoch": 3.47, + "learning_rate": 1.258968449619821e-05, + "loss": 0.0403, + "step": 26513 + }, + { + "epoch": 3.47, + "learning_rate": 1.258917225911697e-05, + "loss": 0.0449, + "step": 26514 + }, + { + "epoch": 3.47, + "learning_rate": 1.2588660014753819e-05, + "loss": 0.0365, + "step": 26515 + }, + { + "epoch": 3.47, + "learning_rate": 1.2588147763110201e-05, + "loss": 0.0367, + "step": 26516 + }, + { + "epoch": 3.47, + "learning_rate": 1.2587635504187556e-05, + "loss": 0.0385, + "step": 26517 + }, + { + "epoch": 3.47, + "learning_rate": 1.2587123237987327e-05, + "loss": 0.0364, + "step": 26518 + }, + { + "epoch": 3.47, + "learning_rate": 1.2586610964510945e-05, + "loss": 0.0468, + "step": 26519 + }, + { + "epoch": 3.47, + "learning_rate": 1.2586098683759865e-05, + "loss": 0.0405, + "step": 26520 + }, + { + "epoch": 3.47, + "learning_rate": 1.2585586395735511e-05, + "loss": 0.0372, + "step": 26521 + }, + { + "epoch": 3.47, + "learning_rate": 1.258507410043934e-05, + "loss": 0.0371, + "step": 26522 + }, + { + "epoch": 3.47, + "learning_rate": 1.2584561797872783e-05, + "loss": 0.0418, + "step": 26523 + }, + { + "epoch": 3.47, + "learning_rate": 1.2584049488037286e-05, + "loss": 0.0415, + "step": 26524 + }, + { + "epoch": 3.47, + "learning_rate": 1.2583537170934283e-05, + "loss": 0.0379, + "step": 26525 + }, + { + "epoch": 3.47, + "learning_rate": 1.2583024846565221e-05, + "loss": 0.0412, + "step": 26526 + }, + { + "epoch": 3.47, + "learning_rate": 1.2582512514931542e-05, + "loss": 0.0382, + "step": 26527 + }, + { + "epoch": 3.47, + "learning_rate": 1.2582000176034684e-05, + "loss": 0.0414, + "step": 26528 + }, + { + "epoch": 3.48, + "learning_rate": 1.2581487829876084e-05, + "loss": 0.0424, + "step": 26529 + }, + { + "epoch": 3.48, + "learning_rate": 1.2580975476457188e-05, + "loss": 0.0354, + "step": 26530 + }, + { + "epoch": 3.48, + "learning_rate": 1.2580463115779439e-05, + "loss": 0.0403, + "step": 26531 + }, + { + "epoch": 3.48, + "learning_rate": 1.2579950747844269e-05, + "loss": 0.0388, + "step": 26532 + }, + { + "epoch": 3.48, + "learning_rate": 1.2579438372653132e-05, + "loss": 0.0411, + "step": 26533 + }, + { + "epoch": 3.48, + "learning_rate": 1.2578925990207457e-05, + "loss": 0.0359, + "step": 26534 + }, + { + "epoch": 3.48, + "learning_rate": 1.2578413600508694e-05, + "loss": 0.0395, + "step": 26535 + }, + { + "epoch": 3.48, + "learning_rate": 1.2577901203558278e-05, + "loss": 0.0403, + "step": 26536 + }, + { + "epoch": 3.48, + "learning_rate": 1.257738879935765e-05, + "loss": 0.0256, + "step": 26537 + }, + { + "epoch": 3.48, + "learning_rate": 1.2576876387908255e-05, + "loss": 0.0432, + "step": 26538 + }, + { + "epoch": 3.48, + "learning_rate": 1.2576363969211533e-05, + "loss": 0.0315, + "step": 26539 + }, + { + "epoch": 3.48, + "learning_rate": 1.2575851543268925e-05, + "loss": 0.045, + "step": 26540 + }, + { + "epoch": 3.48, + "learning_rate": 1.2575339110081872e-05, + "loss": 0.0359, + "step": 26541 + }, + { + "epoch": 3.48, + "learning_rate": 1.2574826669651813e-05, + "loss": 0.0487, + "step": 26542 + }, + { + "epoch": 3.48, + "learning_rate": 1.2574314221980192e-05, + "loss": 0.0481, + "step": 26543 + }, + { + "epoch": 3.48, + "learning_rate": 1.2573801767068447e-05, + "loss": 0.0322, + "step": 26544 + }, + { + "epoch": 3.48, + "learning_rate": 1.2573289304918026e-05, + "loss": 0.0305, + "step": 26545 + }, + { + "epoch": 3.48, + "learning_rate": 1.2572776835530362e-05, + "loss": 0.0388, + "step": 26546 + }, + { + "epoch": 3.48, + "learning_rate": 1.2572264358906902e-05, + "loss": 0.0402, + "step": 26547 + }, + { + "epoch": 3.48, + "learning_rate": 1.2571751875049085e-05, + "loss": 0.035, + "step": 26548 + }, + { + "epoch": 3.48, + "learning_rate": 1.2571239383958354e-05, + "loss": 0.0363, + "step": 26549 + }, + { + "epoch": 3.48, + "learning_rate": 1.2570726885636148e-05, + "loss": 0.0404, + "step": 26550 + }, + { + "epoch": 3.48, + "learning_rate": 1.257021438008391e-05, + "loss": 0.0336, + "step": 26551 + }, + { + "epoch": 3.48, + "learning_rate": 1.2569701867303081e-05, + "loss": 0.0422, + "step": 26552 + }, + { + "epoch": 3.48, + "learning_rate": 1.25691893472951e-05, + "loss": 0.0379, + "step": 26553 + }, + { + "epoch": 3.48, + "learning_rate": 1.2568676820061414e-05, + "loss": 0.0405, + "step": 26554 + }, + { + "epoch": 3.48, + "learning_rate": 1.256816428560346e-05, + "loss": 0.0451, + "step": 26555 + }, + { + "epoch": 3.48, + "learning_rate": 1.2567651743922682e-05, + "loss": 0.0411, + "step": 26556 + }, + { + "epoch": 3.48, + "learning_rate": 1.2567139195020519e-05, + "loss": 0.0417, + "step": 26557 + }, + { + "epoch": 3.48, + "learning_rate": 1.2566626638898414e-05, + "loss": 0.0405, + "step": 26558 + }, + { + "epoch": 3.48, + "learning_rate": 1.2566114075557807e-05, + "loss": 0.0416, + "step": 26559 + }, + { + "epoch": 3.48, + "learning_rate": 1.2565601505000142e-05, + "loss": 0.0442, + "step": 26560 + }, + { + "epoch": 3.48, + "learning_rate": 1.2565088927226859e-05, + "loss": 0.0421, + "step": 26561 + }, + { + "epoch": 3.48, + "learning_rate": 1.2564576342239401e-05, + "loss": 0.0496, + "step": 26562 + }, + { + "epoch": 3.48, + "learning_rate": 1.2564063750039205e-05, + "loss": 0.0341, + "step": 26563 + }, + { + "epoch": 3.48, + "learning_rate": 1.2563551150627724e-05, + "loss": 0.0345, + "step": 26564 + }, + { + "epoch": 3.48, + "learning_rate": 1.2563038544006382e-05, + "loss": 0.0452, + "step": 26565 + }, + { + "epoch": 3.48, + "learning_rate": 1.256252593017664e-05, + "loss": 0.0391, + "step": 26566 + }, + { + "epoch": 3.48, + "learning_rate": 1.2562013309139922e-05, + "loss": 0.0441, + "step": 26567 + }, + { + "epoch": 3.48, + "learning_rate": 1.2561500680897685e-05, + "loss": 0.0361, + "step": 26568 + }, + { + "epoch": 3.48, + "learning_rate": 1.2560988045451359e-05, + "loss": 0.0302, + "step": 26569 + }, + { + "epoch": 3.48, + "learning_rate": 1.2560475402802392e-05, + "loss": 0.0376, + "step": 26570 + }, + { + "epoch": 3.48, + "learning_rate": 1.2559962752952225e-05, + "loss": 0.0379, + "step": 26571 + }, + { + "epoch": 3.48, + "learning_rate": 1.2559450095902297e-05, + "loss": 0.0362, + "step": 26572 + }, + { + "epoch": 3.48, + "learning_rate": 1.2558937431654055e-05, + "loss": 0.0361, + "step": 26573 + }, + { + "epoch": 3.48, + "learning_rate": 1.2558424760208933e-05, + "loss": 0.036, + "step": 26574 + }, + { + "epoch": 3.48, + "learning_rate": 1.2557912081568383e-05, + "loss": 0.036, + "step": 26575 + }, + { + "epoch": 3.48, + "learning_rate": 1.2557399395733839e-05, + "loss": 0.04, + "step": 26576 + }, + { + "epoch": 3.48, + "learning_rate": 1.2556886702706746e-05, + "loss": 0.0358, + "step": 26577 + }, + { + "epoch": 3.48, + "learning_rate": 1.2556374002488543e-05, + "loss": 0.0384, + "step": 26578 + }, + { + "epoch": 3.48, + "learning_rate": 1.2555861295080678e-05, + "loss": 0.0434, + "step": 26579 + }, + { + "epoch": 3.48, + "learning_rate": 1.2555348580484585e-05, + "loss": 0.0381, + "step": 26580 + }, + { + "epoch": 3.48, + "learning_rate": 1.2554835858701713e-05, + "loss": 0.0392, + "step": 26581 + }, + { + "epoch": 3.48, + "learning_rate": 1.2554323129733501e-05, + "loss": 0.0459, + "step": 26582 + }, + { + "epoch": 3.48, + "learning_rate": 1.2553810393581391e-05, + "loss": 0.0344, + "step": 26583 + }, + { + "epoch": 3.48, + "learning_rate": 1.2553297650246825e-05, + "loss": 0.0335, + "step": 26584 + }, + { + "epoch": 3.48, + "learning_rate": 1.2552784899731247e-05, + "loss": 0.0314, + "step": 26585 + }, + { + "epoch": 3.48, + "learning_rate": 1.2552272142036092e-05, + "loss": 0.0368, + "step": 26586 + }, + { + "epoch": 3.48, + "learning_rate": 1.2551759377162816e-05, + "loss": 0.0442, + "step": 26587 + }, + { + "epoch": 3.48, + "learning_rate": 1.2551246605112845e-05, + "loss": 0.05, + "step": 26588 + }, + { + "epoch": 3.48, + "learning_rate": 1.2550733825887634e-05, + "loss": 0.0337, + "step": 26589 + }, + { + "epoch": 3.48, + "learning_rate": 1.2550221039488616e-05, + "loss": 0.0382, + "step": 26590 + }, + { + "epoch": 3.48, + "learning_rate": 1.254970824591724e-05, + "loss": 0.0469, + "step": 26591 + }, + { + "epoch": 3.48, + "learning_rate": 1.2549195445174945e-05, + "loss": 0.035, + "step": 26592 + }, + { + "epoch": 3.48, + "learning_rate": 1.2548682637263173e-05, + "loss": 0.0331, + "step": 26593 + }, + { + "epoch": 3.48, + "learning_rate": 1.2548169822183365e-05, + "loss": 0.0408, + "step": 26594 + }, + { + "epoch": 3.48, + "learning_rate": 1.2547656999936968e-05, + "loss": 0.0353, + "step": 26595 + }, + { + "epoch": 3.48, + "learning_rate": 1.2547144170525421e-05, + "loss": 0.0368, + "step": 26596 + }, + { + "epoch": 3.48, + "learning_rate": 1.2546631333950164e-05, + "loss": 0.0383, + "step": 26597 + }, + { + "epoch": 3.48, + "learning_rate": 1.2546118490212645e-05, + "loss": 0.0387, + "step": 26598 + }, + { + "epoch": 3.48, + "learning_rate": 1.25456056393143e-05, + "loss": 0.0425, + "step": 26599 + }, + { + "epoch": 3.48, + "learning_rate": 1.2545092781256578e-05, + "loss": 0.0388, + "step": 26600 + }, + { + "epoch": 3.48, + "learning_rate": 1.2544579916040917e-05, + "loss": 0.0417, + "step": 26601 + }, + { + "epoch": 3.48, + "learning_rate": 1.2544067043668762e-05, + "loss": 0.0431, + "step": 26602 + }, + { + "epoch": 3.48, + "learning_rate": 1.254355416414155e-05, + "loss": 0.0387, + "step": 26603 + }, + { + "epoch": 3.48, + "learning_rate": 1.254304127746073e-05, + "loss": 0.0463, + "step": 26604 + }, + { + "epoch": 3.49, + "learning_rate": 1.2542528383627741e-05, + "loss": 0.0357, + "step": 26605 + }, + { + "epoch": 3.49, + "learning_rate": 1.254201548264403e-05, + "loss": 0.0397, + "step": 26606 + }, + { + "epoch": 3.49, + "learning_rate": 1.254150257451103e-05, + "loss": 0.0375, + "step": 26607 + }, + { + "epoch": 3.49, + "learning_rate": 1.2540989659230192e-05, + "loss": 0.0373, + "step": 26608 + }, + { + "epoch": 3.49, + "learning_rate": 1.2540476736802955e-05, + "loss": 0.0403, + "step": 26609 + }, + { + "epoch": 3.49, + "learning_rate": 1.2539963807230766e-05, + "loss": 0.0372, + "step": 26610 + }, + { + "epoch": 3.49, + "learning_rate": 1.253945087051506e-05, + "loss": 0.0407, + "step": 26611 + }, + { + "epoch": 3.49, + "learning_rate": 1.2538937926657284e-05, + "loss": 0.0456, + "step": 26612 + }, + { + "epoch": 3.49, + "learning_rate": 1.2538424975658882e-05, + "loss": 0.0445, + "step": 26613 + }, + { + "epoch": 3.49, + "learning_rate": 1.2537912017521293e-05, + "loss": 0.0373, + "step": 26614 + }, + { + "epoch": 3.49, + "learning_rate": 1.2537399052245965e-05, + "loss": 0.0355, + "step": 26615 + }, + { + "epoch": 3.49, + "learning_rate": 1.2536886079834334e-05, + "loss": 0.0406, + "step": 26616 + }, + { + "epoch": 3.49, + "learning_rate": 1.2536373100287849e-05, + "loss": 0.0442, + "step": 26617 + }, + { + "epoch": 3.49, + "learning_rate": 1.2535860113607947e-05, + "loss": 0.0413, + "step": 26618 + }, + { + "epoch": 3.49, + "learning_rate": 1.2535347119796073e-05, + "loss": 0.0384, + "step": 26619 + }, + { + "epoch": 3.49, + "learning_rate": 1.2534834118853671e-05, + "loss": 0.0425, + "step": 26620 + }, + { + "epoch": 3.49, + "learning_rate": 1.2534321110782186e-05, + "loss": 0.0409, + "step": 26621 + }, + { + "epoch": 3.49, + "learning_rate": 1.2533808095583054e-05, + "loss": 0.0411, + "step": 26622 + }, + { + "epoch": 3.49, + "learning_rate": 1.2533295073257725e-05, + "loss": 0.037, + "step": 26623 + }, + { + "epoch": 3.49, + "learning_rate": 1.2532782043807636e-05, + "loss": 0.0329, + "step": 26624 + }, + { + "epoch": 3.49, + "learning_rate": 1.2532269007234233e-05, + "loss": 0.0398, + "step": 26625 + }, + { + "epoch": 3.49, + "learning_rate": 1.2531755963538956e-05, + "loss": 0.0415, + "step": 26626 + }, + { + "epoch": 3.49, + "learning_rate": 1.2531242912723255e-05, + "loss": 0.0392, + "step": 26627 + }, + { + "epoch": 3.49, + "learning_rate": 1.2530729854788563e-05, + "loss": 0.0371, + "step": 26628 + }, + { + "epoch": 3.49, + "learning_rate": 1.2530216789736333e-05, + "loss": 0.0423, + "step": 26629 + }, + { + "epoch": 3.49, + "learning_rate": 1.2529703717568001e-05, + "loss": 0.0404, + "step": 26630 + }, + { + "epoch": 3.49, + "learning_rate": 1.2529190638285011e-05, + "loss": 0.0385, + "step": 26631 + }, + { + "epoch": 3.49, + "learning_rate": 1.2528677551888806e-05, + "loss": 0.049, + "step": 26632 + }, + { + "epoch": 3.49, + "learning_rate": 1.2528164458380834e-05, + "loss": 0.0359, + "step": 26633 + }, + { + "epoch": 3.49, + "learning_rate": 1.2527651357762532e-05, + "loss": 0.0327, + "step": 26634 + }, + { + "epoch": 3.49, + "learning_rate": 1.2527138250035344e-05, + "loss": 0.0344, + "step": 26635 + }, + { + "epoch": 3.49, + "learning_rate": 1.2526625135200714e-05, + "loss": 0.0472, + "step": 26636 + }, + { + "epoch": 3.49, + "learning_rate": 1.2526112013260087e-05, + "loss": 0.0401, + "step": 26637 + }, + { + "epoch": 3.49, + "learning_rate": 1.2525598884214904e-05, + "loss": 0.0398, + "step": 26638 + }, + { + "epoch": 3.49, + "learning_rate": 1.2525085748066609e-05, + "loss": 0.0347, + "step": 26639 + }, + { + "epoch": 3.49, + "learning_rate": 1.2524572604816648e-05, + "loss": 0.037, + "step": 26640 + }, + { + "epoch": 3.49, + "learning_rate": 1.2524059454466453e-05, + "loss": 0.0304, + "step": 26641 + }, + { + "epoch": 3.49, + "learning_rate": 1.2523546297017482e-05, + "loss": 0.033, + "step": 26642 + }, + { + "epoch": 3.49, + "learning_rate": 1.2523033132471168e-05, + "loss": 0.0461, + "step": 26643 + }, + { + "epoch": 3.49, + "learning_rate": 1.252251996082896e-05, + "loss": 0.0411, + "step": 26644 + }, + { + "epoch": 3.49, + "learning_rate": 1.2522006782092298e-05, + "loss": 0.0374, + "step": 26645 + }, + { + "epoch": 3.49, + "learning_rate": 1.2521493596262625e-05, + "loss": 0.0374, + "step": 26646 + }, + { + "epoch": 3.49, + "learning_rate": 1.2520980403341385e-05, + "loss": 0.0393, + "step": 26647 + }, + { + "epoch": 3.49, + "learning_rate": 1.2520467203330026e-05, + "loss": 0.0342, + "step": 26648 + }, + { + "epoch": 3.49, + "learning_rate": 1.2519953996229983e-05, + "loss": 0.0397, + "step": 26649 + }, + { + "epoch": 3.49, + "learning_rate": 1.2519440782042706e-05, + "loss": 0.0413, + "step": 26650 + }, + { + "epoch": 3.49, + "learning_rate": 1.2518927560769637e-05, + "loss": 0.0367, + "step": 26651 + }, + { + "epoch": 3.49, + "learning_rate": 1.2518414332412214e-05, + "loss": 0.0411, + "step": 26652 + }, + { + "epoch": 3.49, + "learning_rate": 1.2517901096971888e-05, + "loss": 0.037, + "step": 26653 + }, + { + "epoch": 3.49, + "learning_rate": 1.25173878544501e-05, + "loss": 0.0356, + "step": 26654 + }, + { + "epoch": 3.49, + "learning_rate": 1.251687460484829e-05, + "loss": 0.0387, + "step": 26655 + }, + { + "epoch": 3.49, + "learning_rate": 1.2516361348167905e-05, + "loss": 0.0405, + "step": 26656 + }, + { + "epoch": 3.49, + "learning_rate": 1.2515848084410387e-05, + "loss": 0.0383, + "step": 26657 + }, + { + "epoch": 3.49, + "learning_rate": 1.2515334813577185e-05, + "loss": 0.0461, + "step": 26658 + }, + { + "epoch": 3.49, + "learning_rate": 1.2514821535669733e-05, + "loss": 0.0427, + "step": 26659 + }, + { + "epoch": 3.49, + "learning_rate": 1.251430825068948e-05, + "loss": 0.0402, + "step": 26660 + }, + { + "epoch": 3.49, + "learning_rate": 1.251379495863787e-05, + "loss": 0.0357, + "step": 26661 + }, + { + "epoch": 3.49, + "learning_rate": 1.2513281659516343e-05, + "loss": 0.0516, + "step": 26662 + }, + { + "epoch": 3.49, + "learning_rate": 1.2512768353326347e-05, + "loss": 0.0352, + "step": 26663 + }, + { + "epoch": 3.49, + "learning_rate": 1.2512255040069326e-05, + "loss": 0.0403, + "step": 26664 + }, + { + "epoch": 3.49, + "learning_rate": 1.2511741719746719e-05, + "loss": 0.0285, + "step": 26665 + }, + { + "epoch": 3.49, + "learning_rate": 1.2511228392359972e-05, + "loss": 0.0404, + "step": 26666 + }, + { + "epoch": 3.49, + "learning_rate": 1.2510715057910533e-05, + "loss": 0.0376, + "step": 26667 + }, + { + "epoch": 3.49, + "learning_rate": 1.2510201716399835e-05, + "loss": 0.034, + "step": 26668 + }, + { + "epoch": 3.49, + "learning_rate": 1.2509688367829331e-05, + "loss": 0.0369, + "step": 26669 + }, + { + "epoch": 3.49, + "learning_rate": 1.2509175012200463e-05, + "loss": 0.0456, + "step": 26670 + }, + { + "epoch": 3.49, + "learning_rate": 1.2508661649514675e-05, + "loss": 0.0354, + "step": 26671 + }, + { + "epoch": 3.49, + "learning_rate": 1.2508148279773406e-05, + "loss": 0.0405, + "step": 26672 + }, + { + "epoch": 3.49, + "learning_rate": 1.2507634902978108e-05, + "loss": 0.0377, + "step": 26673 + }, + { + "epoch": 3.49, + "learning_rate": 1.2507121519130217e-05, + "loss": 0.0433, + "step": 26674 + }, + { + "epoch": 3.49, + "learning_rate": 1.2506608128231181e-05, + "loss": 0.0425, + "step": 26675 + }, + { + "epoch": 3.49, + "learning_rate": 1.2506094730282445e-05, + "loss": 0.0344, + "step": 26676 + }, + { + "epoch": 3.49, + "learning_rate": 1.250558132528545e-05, + "loss": 0.0384, + "step": 26677 + }, + { + "epoch": 3.49, + "learning_rate": 1.250506791324164e-05, + "loss": 0.0427, + "step": 26678 + }, + { + "epoch": 3.49, + "learning_rate": 1.2504554494152462e-05, + "loss": 0.0389, + "step": 26679 + }, + { + "epoch": 3.49, + "learning_rate": 1.2504041068019354e-05, + "loss": 0.0424, + "step": 26680 + }, + { + "epoch": 3.5, + "learning_rate": 1.2503527634843766e-05, + "loss": 0.0397, + "step": 26681 + }, + { + "epoch": 3.5, + "learning_rate": 1.2503014194627142e-05, + "loss": 0.0428, + "step": 26682 + }, + { + "epoch": 3.5, + "learning_rate": 1.250250074737092e-05, + "loss": 0.043, + "step": 26683 + }, + { + "epoch": 3.5, + "learning_rate": 1.2501987293076552e-05, + "loss": 0.0374, + "step": 26684 + }, + { + "epoch": 3.5, + "learning_rate": 1.2501473831745471e-05, + "loss": 0.0445, + "step": 26685 + }, + { + "epoch": 3.5, + "learning_rate": 1.2500960363379135e-05, + "loss": 0.0424, + "step": 26686 + }, + { + "epoch": 3.5, + "learning_rate": 1.2500446887978978e-05, + "loss": 0.0404, + "step": 26687 + }, + { + "epoch": 3.5, + "learning_rate": 1.249993340554645e-05, + "loss": 0.0371, + "step": 26688 + }, + { + "epoch": 3.5, + "learning_rate": 1.2499419916082991e-05, + "loss": 0.0384, + "step": 26689 + }, + { + "epoch": 3.5, + "learning_rate": 1.2498906419590045e-05, + "loss": 0.0412, + "step": 26690 + }, + { + "epoch": 3.5, + "learning_rate": 1.249839291606906e-05, + "loss": 0.0361, + "step": 26691 + }, + { + "epoch": 3.5, + "learning_rate": 1.2497879405521477e-05, + "loss": 0.0324, + "step": 26692 + }, + { + "epoch": 3.5, + "learning_rate": 1.2497365887948741e-05, + "loss": 0.0381, + "step": 26693 + }, + { + "epoch": 3.5, + "learning_rate": 1.2496852363352299e-05, + "loss": 0.0432, + "step": 26694 + }, + { + "epoch": 3.5, + "learning_rate": 1.2496338831733587e-05, + "loss": 0.0388, + "step": 26695 + }, + { + "epoch": 3.5, + "learning_rate": 1.2495825293094059e-05, + "loss": 0.0413, + "step": 26696 + }, + { + "epoch": 3.5, + "learning_rate": 1.2495311747435155e-05, + "loss": 0.0448, + "step": 26697 + }, + { + "epoch": 3.5, + "learning_rate": 1.2494798194758317e-05, + "loss": 0.0323, + "step": 26698 + }, + { + "epoch": 3.5, + "learning_rate": 1.2494284635064996e-05, + "loss": 0.0332, + "step": 26699 + }, + { + "epoch": 3.5, + "learning_rate": 1.249377106835663e-05, + "loss": 0.0331, + "step": 26700 + }, + { + "epoch": 3.5, + "learning_rate": 1.2493257494634664e-05, + "loss": 0.0289, + "step": 26701 + }, + { + "epoch": 3.5, + "learning_rate": 1.2492743913900546e-05, + "loss": 0.039, + "step": 26702 + }, + { + "epoch": 3.5, + "learning_rate": 1.249223032615572e-05, + "loss": 0.0345, + "step": 26703 + }, + { + "epoch": 3.5, + "learning_rate": 1.2491716731401628e-05, + "loss": 0.0305, + "step": 26704 + }, + { + "epoch": 3.5, + "learning_rate": 1.2491203129639717e-05, + "loss": 0.0353, + "step": 26705 + }, + { + "epoch": 3.5, + "learning_rate": 1.2490689520871426e-05, + "loss": 0.0394, + "step": 26706 + }, + { + "epoch": 3.5, + "learning_rate": 1.2490175905098205e-05, + "loss": 0.0342, + "step": 26707 + }, + { + "epoch": 3.5, + "learning_rate": 1.2489662282321495e-05, + "loss": 0.0358, + "step": 26708 + }, + { + "epoch": 3.5, + "learning_rate": 1.2489148652542746e-05, + "loss": 0.0443, + "step": 26709 + }, + { + "epoch": 3.5, + "learning_rate": 1.2488635015763397e-05, + "loss": 0.0464, + "step": 26710 + }, + { + "epoch": 3.5, + "learning_rate": 1.2488121371984897e-05, + "loss": 0.0385, + "step": 26711 + }, + { + "epoch": 3.5, + "learning_rate": 1.2487607721208683e-05, + "loss": 0.0359, + "step": 26712 + }, + { + "epoch": 3.5, + "learning_rate": 1.248709406343621e-05, + "loss": 0.039, + "step": 26713 + }, + { + "epoch": 3.5, + "learning_rate": 1.2486580398668915e-05, + "loss": 0.0374, + "step": 26714 + }, + { + "epoch": 3.5, + "learning_rate": 1.2486066726908245e-05, + "loss": 0.0348, + "step": 26715 + }, + { + "epoch": 3.5, + "learning_rate": 1.2485553048155645e-05, + "loss": 0.0364, + "step": 26716 + }, + { + "epoch": 3.5, + "learning_rate": 1.248503936241256e-05, + "loss": 0.0393, + "step": 26717 + }, + { + "epoch": 3.5, + "learning_rate": 1.2484525669680432e-05, + "loss": 0.0403, + "step": 26718 + }, + { + "epoch": 3.5, + "learning_rate": 1.2484011969960712e-05, + "loss": 0.0383, + "step": 26719 + }, + { + "epoch": 3.5, + "learning_rate": 1.2483498263254836e-05, + "loss": 0.0444, + "step": 26720 + }, + { + "epoch": 3.5, + "learning_rate": 1.2482984549564256e-05, + "loss": 0.0433, + "step": 26721 + }, + { + "epoch": 3.5, + "learning_rate": 1.2482470828890415e-05, + "loss": 0.0374, + "step": 26722 + }, + { + "epoch": 3.5, + "learning_rate": 1.2481957101234751e-05, + "loss": 0.0385, + "step": 26723 + }, + { + "epoch": 3.5, + "learning_rate": 1.2481443366598722e-05, + "loss": 0.0313, + "step": 26724 + }, + { + "epoch": 3.5, + "learning_rate": 1.2480929624983764e-05, + "loss": 0.0441, + "step": 26725 + }, + { + "epoch": 3.5, + "learning_rate": 1.2480415876391324e-05, + "loss": 0.0367, + "step": 26726 + }, + { + "epoch": 3.5, + "learning_rate": 1.2479902120822844e-05, + "loss": 0.0362, + "step": 26727 + }, + { + "epoch": 3.5, + "learning_rate": 1.2479388358279771e-05, + "loss": 0.0475, + "step": 26728 + }, + { + "epoch": 3.5, + "learning_rate": 1.2478874588763553e-05, + "loss": 0.0445, + "step": 26729 + }, + { + "epoch": 3.5, + "learning_rate": 1.247836081227563e-05, + "loss": 0.0428, + "step": 26730 + }, + { + "epoch": 3.5, + "learning_rate": 1.2477847028817451e-05, + "loss": 0.0344, + "step": 26731 + }, + { + "epoch": 3.5, + "learning_rate": 1.2477333238390459e-05, + "loss": 0.0351, + "step": 26732 + }, + { + "epoch": 3.5, + "learning_rate": 1.2476819440996097e-05, + "loss": 0.0312, + "step": 26733 + }, + { + "epoch": 3.5, + "learning_rate": 1.2476305636635814e-05, + "loss": 0.0328, + "step": 26734 + }, + { + "epoch": 3.5, + "learning_rate": 1.2475791825311052e-05, + "loss": 0.0363, + "step": 26735 + }, + { + "epoch": 3.5, + "learning_rate": 1.2475278007023262e-05, + "loss": 0.0392, + "step": 26736 + }, + { + "epoch": 3.5, + "learning_rate": 1.2474764181773878e-05, + "loss": 0.038, + "step": 26737 + }, + { + "epoch": 3.5, + "learning_rate": 1.2474250349564355e-05, + "loss": 0.0445, + "step": 26738 + }, + { + "epoch": 3.5, + "learning_rate": 1.2473736510396132e-05, + "loss": 0.0349, + "step": 26739 + }, + { + "epoch": 3.5, + "learning_rate": 1.247322266427066e-05, + "loss": 0.0388, + "step": 26740 + }, + { + "epoch": 3.5, + "learning_rate": 1.247270881118938e-05, + "loss": 0.032, + "step": 26741 + }, + { + "epoch": 3.5, + "learning_rate": 1.2472194951153738e-05, + "loss": 0.0419, + "step": 26742 + }, + { + "epoch": 3.5, + "learning_rate": 1.2471681084165179e-05, + "loss": 0.032, + "step": 26743 + }, + { + "epoch": 3.5, + "learning_rate": 1.2471167210225148e-05, + "loss": 0.0399, + "step": 26744 + }, + { + "epoch": 3.5, + "learning_rate": 1.2470653329335092e-05, + "loss": 0.0404, + "step": 26745 + }, + { + "epoch": 3.5, + "learning_rate": 1.2470139441496453e-05, + "loss": 0.038, + "step": 26746 + }, + { + "epoch": 3.5, + "learning_rate": 1.2469625546710682e-05, + "loss": 0.0385, + "step": 26747 + }, + { + "epoch": 3.5, + "learning_rate": 1.246911164497922e-05, + "loss": 0.0423, + "step": 26748 + }, + { + "epoch": 3.5, + "learning_rate": 1.2468597736303511e-05, + "loss": 0.0398, + "step": 26749 + }, + { + "epoch": 3.5, + "learning_rate": 1.2468083820685002e-05, + "loss": 0.0335, + "step": 26750 + }, + { + "epoch": 3.5, + "learning_rate": 1.246756989812514e-05, + "loss": 0.0362, + "step": 26751 + }, + { + "epoch": 3.5, + "learning_rate": 1.2467055968625369e-05, + "loss": 0.0389, + "step": 26752 + }, + { + "epoch": 3.5, + "learning_rate": 1.2466542032187135e-05, + "loss": 0.0352, + "step": 26753 + }, + { + "epoch": 3.5, + "learning_rate": 1.2466028088811881e-05, + "loss": 0.0404, + "step": 26754 + }, + { + "epoch": 3.5, + "learning_rate": 1.2465514138501059e-05, + "loss": 0.0273, + "step": 26755 + }, + { + "epoch": 3.5, + "learning_rate": 1.2465000181256101e-05, + "loss": 0.0362, + "step": 26756 + }, + { + "epoch": 3.5, + "learning_rate": 1.2464486217078468e-05, + "loss": 0.0375, + "step": 26757 + }, + { + "epoch": 3.51, + "learning_rate": 1.2463972245969597e-05, + "loss": 0.0371, + "step": 26758 + }, + { + "epoch": 3.51, + "learning_rate": 1.2463458267930936e-05, + "loss": 0.0373, + "step": 26759 + }, + { + "epoch": 3.51, + "learning_rate": 1.2462944282963929e-05, + "loss": 0.0418, + "step": 26760 + }, + { + "epoch": 3.51, + "learning_rate": 1.2462430291070023e-05, + "loss": 0.04, + "step": 26761 + }, + { + "epoch": 3.51, + "learning_rate": 1.2461916292250662e-05, + "loss": 0.0342, + "step": 26762 + }, + { + "epoch": 3.51, + "learning_rate": 1.2461402286507293e-05, + "loss": 0.0393, + "step": 26763 + }, + { + "epoch": 3.51, + "learning_rate": 1.2460888273841363e-05, + "loss": 0.0333, + "step": 26764 + }, + { + "epoch": 3.51, + "learning_rate": 1.2460374254254312e-05, + "loss": 0.0374, + "step": 26765 + }, + { + "epoch": 3.51, + "learning_rate": 1.2459860227747593e-05, + "loss": 0.0431, + "step": 26766 + }, + { + "epoch": 3.51, + "learning_rate": 1.2459346194322645e-05, + "loss": 0.0375, + "step": 26767 + }, + { + "epoch": 3.51, + "learning_rate": 1.245883215398092e-05, + "loss": 0.0417, + "step": 26768 + }, + { + "epoch": 3.51, + "learning_rate": 1.2458318106723859e-05, + "loss": 0.0406, + "step": 26769 + }, + { + "epoch": 3.51, + "learning_rate": 1.2457804052552911e-05, + "loss": 0.0352, + "step": 26770 + }, + { + "epoch": 3.51, + "learning_rate": 1.2457289991469517e-05, + "loss": 0.0407, + "step": 26771 + }, + { + "epoch": 3.51, + "learning_rate": 1.2456775923475127e-05, + "loss": 0.0342, + "step": 26772 + }, + { + "epoch": 3.51, + "learning_rate": 1.2456261848571182e-05, + "loss": 0.037, + "step": 26773 + }, + { + "epoch": 3.51, + "learning_rate": 1.2455747766759139e-05, + "loss": 0.0352, + "step": 26774 + }, + { + "epoch": 3.51, + "learning_rate": 1.245523367804043e-05, + "loss": 0.0409, + "step": 26775 + }, + { + "epoch": 3.51, + "learning_rate": 1.245471958241651e-05, + "loss": 0.0321, + "step": 26776 + }, + { + "epoch": 3.51, + "learning_rate": 1.245420547988882e-05, + "loss": 0.0348, + "step": 26777 + }, + { + "epoch": 3.51, + "learning_rate": 1.2453691370458809e-05, + "loss": 0.0392, + "step": 26778 + }, + { + "epoch": 3.51, + "learning_rate": 1.2453177254127921e-05, + "loss": 0.0452, + "step": 26779 + }, + { + "epoch": 3.51, + "learning_rate": 1.2452663130897604e-05, + "loss": 0.0374, + "step": 26780 + }, + { + "epoch": 3.51, + "learning_rate": 1.24521490007693e-05, + "loss": 0.0332, + "step": 26781 + }, + { + "epoch": 3.51, + "learning_rate": 1.245163486374446e-05, + "loss": 0.0437, + "step": 26782 + }, + { + "epoch": 3.51, + "learning_rate": 1.2451120719824526e-05, + "loss": 0.0368, + "step": 26783 + }, + { + "epoch": 3.51, + "learning_rate": 1.2450606569010947e-05, + "loss": 0.034, + "step": 26784 + }, + { + "epoch": 3.51, + "learning_rate": 1.2450092411305165e-05, + "loss": 0.0406, + "step": 26785 + }, + { + "epoch": 3.51, + "learning_rate": 1.244957824670863e-05, + "loss": 0.0465, + "step": 26786 + }, + { + "epoch": 3.51, + "learning_rate": 1.244906407522279e-05, + "loss": 0.0326, + "step": 26787 + }, + { + "epoch": 3.51, + "learning_rate": 1.2448549896849081e-05, + "loss": 0.0517, + "step": 26788 + }, + { + "epoch": 3.51, + "learning_rate": 1.244803571158896e-05, + "loss": 0.0422, + "step": 26789 + }, + { + "epoch": 3.51, + "learning_rate": 1.2447521519443868e-05, + "loss": 0.0376, + "step": 26790 + }, + { + "epoch": 3.51, + "learning_rate": 1.2447007320415254e-05, + "loss": 0.0381, + "step": 26791 + }, + { + "epoch": 3.51, + "learning_rate": 1.2446493114504561e-05, + "loss": 0.0427, + "step": 26792 + }, + { + "epoch": 3.51, + "learning_rate": 1.2445978901713236e-05, + "loss": 0.0359, + "step": 26793 + }, + { + "epoch": 3.51, + "learning_rate": 1.2445464682042723e-05, + "loss": 0.0386, + "step": 26794 + }, + { + "epoch": 3.51, + "learning_rate": 1.2444950455494475e-05, + "loss": 0.0444, + "step": 26795 + }, + { + "epoch": 3.51, + "learning_rate": 1.244443622206993e-05, + "loss": 0.0473, + "step": 26796 + }, + { + "epoch": 3.51, + "learning_rate": 1.2443921981770543e-05, + "loss": 0.0383, + "step": 26797 + }, + { + "epoch": 3.51, + "learning_rate": 1.2443407734597752e-05, + "loss": 0.0408, + "step": 26798 + }, + { + "epoch": 3.51, + "learning_rate": 1.2442893480553009e-05, + "loss": 0.0425, + "step": 26799 + }, + { + "epoch": 3.51, + "learning_rate": 1.2442379219637756e-05, + "loss": 0.04, + "step": 26800 + }, + { + "epoch": 3.51, + "learning_rate": 1.2441864951853444e-05, + "loss": 0.0421, + "step": 26801 + }, + { + "epoch": 3.51, + "learning_rate": 1.2441350677201516e-05, + "loss": 0.0428, + "step": 26802 + }, + { + "epoch": 3.51, + "learning_rate": 1.2440836395683418e-05, + "loss": 0.0395, + "step": 26803 + }, + { + "epoch": 3.51, + "learning_rate": 1.24403221073006e-05, + "loss": 0.0393, + "step": 26804 + }, + { + "epoch": 3.51, + "learning_rate": 1.2439807812054503e-05, + "loss": 0.0445, + "step": 26805 + }, + { + "epoch": 3.51, + "learning_rate": 1.243929350994658e-05, + "loss": 0.0476, + "step": 26806 + }, + { + "epoch": 3.51, + "learning_rate": 1.243877920097827e-05, + "loss": 0.0335, + "step": 26807 + }, + { + "epoch": 3.51, + "learning_rate": 1.2438264885151026e-05, + "loss": 0.0365, + "step": 26808 + }, + { + "epoch": 3.51, + "learning_rate": 1.243775056246629e-05, + "loss": 0.0369, + "step": 26809 + }, + { + "epoch": 3.51, + "learning_rate": 1.243723623292551e-05, + "loss": 0.0381, + "step": 26810 + }, + { + "epoch": 3.51, + "learning_rate": 1.2436721896530136e-05, + "loss": 0.0392, + "step": 26811 + }, + { + "epoch": 3.51, + "learning_rate": 1.2436207553281612e-05, + "loss": 0.0404, + "step": 26812 + }, + { + "epoch": 3.51, + "learning_rate": 1.243569320318138e-05, + "loss": 0.0373, + "step": 26813 + }, + { + "epoch": 3.51, + "learning_rate": 1.2435178846230892e-05, + "loss": 0.0448, + "step": 26814 + }, + { + "epoch": 3.51, + "learning_rate": 1.2434664482431596e-05, + "loss": 0.034, + "step": 26815 + }, + { + "epoch": 3.51, + "learning_rate": 1.243415011178493e-05, + "loss": 0.04, + "step": 26816 + }, + { + "epoch": 3.51, + "learning_rate": 1.243363573429235e-05, + "loss": 0.0315, + "step": 26817 + }, + { + "epoch": 3.51, + "learning_rate": 1.2433121349955301e-05, + "loss": 0.0416, + "step": 26818 + }, + { + "epoch": 3.51, + "learning_rate": 1.2432606958775226e-05, + "loss": 0.0391, + "step": 26819 + }, + { + "epoch": 3.51, + "learning_rate": 1.2432092560753572e-05, + "loss": 0.0381, + "step": 26820 + }, + { + "epoch": 3.51, + "learning_rate": 1.2431578155891789e-05, + "loss": 0.0443, + "step": 26821 + }, + { + "epoch": 3.51, + "learning_rate": 1.243106374419132e-05, + "loss": 0.0489, + "step": 26822 + }, + { + "epoch": 3.51, + "learning_rate": 1.2430549325653615e-05, + "loss": 0.0427, + "step": 26823 + }, + { + "epoch": 3.51, + "learning_rate": 1.243003490028012e-05, + "loss": 0.0441, + "step": 26824 + }, + { + "epoch": 3.51, + "learning_rate": 1.242952046807228e-05, + "loss": 0.0374, + "step": 26825 + }, + { + "epoch": 3.51, + "learning_rate": 1.2429006029031545e-05, + "loss": 0.0401, + "step": 26826 + }, + { + "epoch": 3.51, + "learning_rate": 1.2428491583159357e-05, + "loss": 0.0337, + "step": 26827 + }, + { + "epoch": 3.51, + "learning_rate": 1.2427977130457168e-05, + "loss": 0.0408, + "step": 26828 + }, + { + "epoch": 3.51, + "learning_rate": 1.2427462670926422e-05, + "loss": 0.0442, + "step": 26829 + }, + { + "epoch": 3.51, + "learning_rate": 1.2426948204568564e-05, + "loss": 0.0364, + "step": 26830 + }, + { + "epoch": 3.51, + "learning_rate": 1.242643373138505e-05, + "loss": 0.0324, + "step": 26831 + }, + { + "epoch": 3.51, + "learning_rate": 1.2425919251377315e-05, + "loss": 0.0358, + "step": 26832 + }, + { + "epoch": 3.51, + "learning_rate": 1.2425404764546811e-05, + "loss": 0.0366, + "step": 26833 + }, + { + "epoch": 3.52, + "learning_rate": 1.2424890270894987e-05, + "loss": 0.0396, + "step": 26834 + }, + { + "epoch": 3.52, + "learning_rate": 1.2424375770423289e-05, + "loss": 0.0421, + "step": 26835 + }, + { + "epoch": 3.52, + "learning_rate": 1.2423861263133162e-05, + "loss": 0.046, + "step": 26836 + }, + { + "epoch": 3.52, + "learning_rate": 1.2423346749026057e-05, + "loss": 0.0481, + "step": 26837 + }, + { + "epoch": 3.52, + "learning_rate": 1.2422832228103411e-05, + "loss": 0.0386, + "step": 26838 + }, + { + "epoch": 3.52, + "learning_rate": 1.2422317700366686e-05, + "loss": 0.037, + "step": 26839 + }, + { + "epoch": 3.52, + "learning_rate": 1.242180316581732e-05, + "loss": 0.0422, + "step": 26840 + }, + { + "epoch": 3.52, + "learning_rate": 1.242128862445676e-05, + "loss": 0.0347, + "step": 26841 + }, + { + "epoch": 3.52, + "learning_rate": 1.2420774076286454e-05, + "loss": 0.0355, + "step": 26842 + }, + { + "epoch": 3.52, + "learning_rate": 1.2420259521307852e-05, + "loss": 0.0393, + "step": 26843 + }, + { + "epoch": 3.52, + "learning_rate": 1.2419744959522395e-05, + "loss": 0.0392, + "step": 26844 + }, + { + "epoch": 3.52, + "learning_rate": 1.241923039093154e-05, + "loss": 0.0365, + "step": 26845 + }, + { + "epoch": 3.52, + "learning_rate": 1.2418715815536725e-05, + "loss": 0.0337, + "step": 26846 + }, + { + "epoch": 3.52, + "learning_rate": 1.24182012333394e-05, + "loss": 0.0346, + "step": 26847 + }, + { + "epoch": 3.52, + "learning_rate": 1.2417686644341012e-05, + "loss": 0.0404, + "step": 26848 + }, + { + "epoch": 3.52, + "learning_rate": 1.2417172048543011e-05, + "loss": 0.048, + "step": 26849 + }, + { + "epoch": 3.52, + "learning_rate": 1.2416657445946843e-05, + "loss": 0.0411, + "step": 26850 + }, + { + "epoch": 3.52, + "learning_rate": 1.2416142836553953e-05, + "loss": 0.0329, + "step": 26851 + }, + { + "epoch": 3.52, + "learning_rate": 1.2415628220365793e-05, + "loss": 0.0397, + "step": 26852 + }, + { + "epoch": 3.52, + "learning_rate": 1.2415113597383803e-05, + "loss": 0.0333, + "step": 26853 + }, + { + "epoch": 3.52, + "learning_rate": 1.2414598967609437e-05, + "loss": 0.0464, + "step": 26854 + }, + { + "epoch": 3.52, + "learning_rate": 1.2414084331044138e-05, + "loss": 0.038, + "step": 26855 + }, + { + "epoch": 3.52, + "learning_rate": 1.2413569687689358e-05, + "loss": 0.0442, + "step": 26856 + }, + { + "epoch": 3.52, + "learning_rate": 1.241305503754654e-05, + "loss": 0.0352, + "step": 26857 + }, + { + "epoch": 3.52, + "learning_rate": 1.2412540380617134e-05, + "loss": 0.0412, + "step": 26858 + }, + { + "epoch": 3.52, + "learning_rate": 1.2412025716902585e-05, + "loss": 0.0368, + "step": 26859 + }, + { + "epoch": 3.52, + "learning_rate": 1.2411511046404346e-05, + "loss": 0.0408, + "step": 26860 + }, + { + "epoch": 3.52, + "learning_rate": 1.2410996369123858e-05, + "loss": 0.0399, + "step": 26861 + }, + { + "epoch": 3.52, + "learning_rate": 1.2410481685062572e-05, + "loss": 0.0385, + "step": 26862 + }, + { + "epoch": 3.52, + "learning_rate": 1.2409966994221934e-05, + "loss": 0.0341, + "step": 26863 + }, + { + "epoch": 3.52, + "learning_rate": 1.240945229660339e-05, + "loss": 0.0325, + "step": 26864 + }, + { + "epoch": 3.52, + "learning_rate": 1.2408937592208393e-05, + "loss": 0.0443, + "step": 26865 + }, + { + "epoch": 3.52, + "learning_rate": 1.2408422881038387e-05, + "loss": 0.039, + "step": 26866 + }, + { + "epoch": 3.52, + "learning_rate": 1.2407908163094819e-05, + "loss": 0.0411, + "step": 26867 + }, + { + "epoch": 3.52, + "learning_rate": 1.2407393438379135e-05, + "loss": 0.0372, + "step": 26868 + }, + { + "epoch": 3.52, + "learning_rate": 1.240687870689279e-05, + "loss": 0.0366, + "step": 26869 + }, + { + "epoch": 3.52, + "learning_rate": 1.2406363968637225e-05, + "loss": 0.0376, + "step": 26870 + }, + { + "epoch": 3.52, + "learning_rate": 1.2405849223613889e-05, + "loss": 0.0366, + "step": 26871 + }, + { + "epoch": 3.52, + "learning_rate": 1.2405334471824229e-05, + "loss": 0.0404, + "step": 26872 + }, + { + "epoch": 3.52, + "learning_rate": 1.2404819713269696e-05, + "loss": 0.0381, + "step": 26873 + }, + { + "epoch": 3.52, + "learning_rate": 1.2404304947951736e-05, + "loss": 0.0365, + "step": 26874 + }, + { + "epoch": 3.52, + "learning_rate": 1.2403790175871794e-05, + "loss": 0.0427, + "step": 26875 + }, + { + "epoch": 3.52, + "learning_rate": 1.240327539703132e-05, + "loss": 0.0351, + "step": 26876 + }, + { + "epoch": 3.52, + "learning_rate": 1.2402760611431763e-05, + "loss": 0.0378, + "step": 26877 + }, + { + "epoch": 3.52, + "learning_rate": 1.240224581907457e-05, + "loss": 0.0287, + "step": 26878 + }, + { + "epoch": 3.52, + "learning_rate": 1.2401731019961192e-05, + "loss": 0.0368, + "step": 26879 + }, + { + "epoch": 3.52, + "learning_rate": 1.2401216214093067e-05, + "loss": 0.0401, + "step": 26880 + }, + { + "epoch": 3.52, + "learning_rate": 1.2400701401471653e-05, + "loss": 0.0351, + "step": 26881 + }, + { + "epoch": 3.52, + "learning_rate": 1.2400186582098392e-05, + "loss": 0.0346, + "step": 26882 + }, + { + "epoch": 3.52, + "learning_rate": 1.2399671755974735e-05, + "loss": 0.0331, + "step": 26883 + }, + { + "epoch": 3.52, + "learning_rate": 1.2399156923102131e-05, + "loss": 0.0387, + "step": 26884 + }, + { + "epoch": 3.52, + "learning_rate": 1.2398642083482026e-05, + "loss": 0.0326, + "step": 26885 + }, + { + "epoch": 3.52, + "learning_rate": 1.2398127237115865e-05, + "loss": 0.0362, + "step": 26886 + }, + { + "epoch": 3.52, + "learning_rate": 1.23976123840051e-05, + "loss": 0.0377, + "step": 26887 + }, + { + "epoch": 3.52, + "learning_rate": 1.2397097524151176e-05, + "loss": 0.0369, + "step": 26888 + }, + { + "epoch": 3.52, + "learning_rate": 1.2396582657555544e-05, + "loss": 0.0418, + "step": 26889 + }, + { + "epoch": 3.52, + "learning_rate": 1.2396067784219655e-05, + "loss": 0.0413, + "step": 26890 + }, + { + "epoch": 3.52, + "learning_rate": 1.2395552904144947e-05, + "loss": 0.0329, + "step": 26891 + }, + { + "epoch": 3.52, + "learning_rate": 1.2395038017332877e-05, + "loss": 0.0355, + "step": 26892 + }, + { + "epoch": 3.52, + "learning_rate": 1.239452312378489e-05, + "loss": 0.0389, + "step": 26893 + }, + { + "epoch": 3.52, + "learning_rate": 1.2394008223502434e-05, + "loss": 0.0415, + "step": 26894 + }, + { + "epoch": 3.52, + "learning_rate": 1.2393493316486957e-05, + "loss": 0.0435, + "step": 26895 + }, + { + "epoch": 3.52, + "learning_rate": 1.2392978402739909e-05, + "loss": 0.0359, + "step": 26896 + }, + { + "epoch": 3.52, + "learning_rate": 1.2392463482262736e-05, + "loss": 0.0387, + "step": 26897 + }, + { + "epoch": 3.52, + "learning_rate": 1.2391948555056886e-05, + "loss": 0.0295, + "step": 26898 + }, + { + "epoch": 3.52, + "learning_rate": 1.2391433621123805e-05, + "loss": 0.0393, + "step": 26899 + }, + { + "epoch": 3.52, + "learning_rate": 1.239091868046495e-05, + "loss": 0.0362, + "step": 26900 + }, + { + "epoch": 3.52, + "learning_rate": 1.2390403733081761e-05, + "loss": 0.0437, + "step": 26901 + }, + { + "epoch": 3.52, + "learning_rate": 1.2389888778975692e-05, + "loss": 0.0409, + "step": 26902 + }, + { + "epoch": 3.52, + "learning_rate": 1.238937381814818e-05, + "loss": 0.0358, + "step": 26903 + }, + { + "epoch": 3.52, + "learning_rate": 1.2388858850600688e-05, + "loss": 0.0358, + "step": 26904 + }, + { + "epoch": 3.52, + "learning_rate": 1.2388343876334657e-05, + "loss": 0.0435, + "step": 26905 + }, + { + "epoch": 3.52, + "learning_rate": 1.2387828895351537e-05, + "loss": 0.0448, + "step": 26906 + }, + { + "epoch": 3.52, + "learning_rate": 1.238731390765277e-05, + "loss": 0.0377, + "step": 26907 + }, + { + "epoch": 3.52, + "learning_rate": 1.2386798913239815e-05, + "loss": 0.039, + "step": 26908 + }, + { + "epoch": 3.52, + "learning_rate": 1.2386283912114113e-05, + "loss": 0.0385, + "step": 26909 + }, + { + "epoch": 3.53, + "learning_rate": 1.2385768904277118e-05, + "loss": 0.0357, + "step": 26910 + }, + { + "epoch": 3.53, + "learning_rate": 1.2385253889730272e-05, + "loss": 0.0381, + "step": 26911 + }, + { + "epoch": 3.53, + "learning_rate": 1.2384738868475023e-05, + "loss": 0.0334, + "step": 26912 + }, + { + "epoch": 3.53, + "learning_rate": 1.2384223840512829e-05, + "loss": 0.0424, + "step": 26913 + }, + { + "epoch": 3.53, + "learning_rate": 1.2383708805845126e-05, + "loss": 0.0386, + "step": 26914 + }, + { + "epoch": 3.53, + "learning_rate": 1.2383193764473375e-05, + "loss": 0.0346, + "step": 26915 + }, + { + "epoch": 3.53, + "learning_rate": 1.2382678716399017e-05, + "loss": 0.0342, + "step": 26916 + }, + { + "epoch": 3.53, + "learning_rate": 1.2382163661623504e-05, + "loss": 0.0398, + "step": 26917 + }, + { + "epoch": 3.53, + "learning_rate": 1.2381648600148278e-05, + "loss": 0.0369, + "step": 26918 + }, + { + "epoch": 3.53, + "learning_rate": 1.2381133531974791e-05, + "loss": 0.0346, + "step": 26919 + }, + { + "epoch": 3.53, + "learning_rate": 1.2380618457104497e-05, + "loss": 0.0345, + "step": 26920 + }, + { + "epoch": 3.53, + "learning_rate": 1.238010337553884e-05, + "loss": 0.0378, + "step": 26921 + }, + { + "epoch": 3.53, + "learning_rate": 1.2379588287279267e-05, + "loss": 0.0419, + "step": 26922 + }, + { + "epoch": 3.53, + "learning_rate": 1.2379073192327234e-05, + "loss": 0.0371, + "step": 26923 + }, + { + "epoch": 3.53, + "learning_rate": 1.2378558090684178e-05, + "loss": 0.0334, + "step": 26924 + }, + { + "epoch": 3.53, + "learning_rate": 1.2378042982351555e-05, + "loss": 0.0359, + "step": 26925 + }, + { + "epoch": 3.53, + "learning_rate": 1.2377527867330815e-05, + "loss": 0.0386, + "step": 26926 + }, + { + "epoch": 3.53, + "learning_rate": 1.2377012745623403e-05, + "loss": 0.0455, + "step": 26927 + }, + { + "epoch": 3.53, + "learning_rate": 1.237649761723077e-05, + "loss": 0.0334, + "step": 26928 + }, + { + "epoch": 3.53, + "learning_rate": 1.2375982482154363e-05, + "loss": 0.0399, + "step": 26929 + }, + { + "epoch": 3.53, + "learning_rate": 1.237546734039563e-05, + "loss": 0.0397, + "step": 26930 + }, + { + "epoch": 3.53, + "learning_rate": 1.2374952191956027e-05, + "loss": 0.038, + "step": 26931 + }, + { + "epoch": 3.53, + "learning_rate": 1.2374437036836993e-05, + "loss": 0.0305, + "step": 26932 + }, + { + "epoch": 3.53, + "learning_rate": 1.2373921875039983e-05, + "loss": 0.0378, + "step": 26933 + }, + { + "epoch": 3.53, + "learning_rate": 1.2373406706566443e-05, + "loss": 0.033, + "step": 26934 + }, + { + "epoch": 3.53, + "learning_rate": 1.2372891531417826e-05, + "loss": 0.0388, + "step": 26935 + }, + { + "epoch": 3.53, + "learning_rate": 1.2372376349595574e-05, + "loss": 0.0407, + "step": 26936 + }, + { + "epoch": 3.53, + "learning_rate": 1.2371861161101143e-05, + "loss": 0.0378, + "step": 26937 + }, + { + "epoch": 3.53, + "learning_rate": 1.2371345965935978e-05, + "loss": 0.0351, + "step": 26938 + }, + { + "epoch": 3.53, + "learning_rate": 1.2370830764101527e-05, + "loss": 0.0379, + "step": 26939 + }, + { + "epoch": 3.53, + "learning_rate": 1.2370315555599244e-05, + "loss": 0.0376, + "step": 26940 + }, + { + "epoch": 3.53, + "learning_rate": 1.236980034043057e-05, + "loss": 0.0383, + "step": 26941 + }, + { + "epoch": 3.53, + "learning_rate": 1.2369285118596963e-05, + "loss": 0.0456, + "step": 26942 + }, + { + "epoch": 3.53, + "learning_rate": 1.2368769890099866e-05, + "loss": 0.0385, + "step": 26943 + }, + { + "epoch": 3.53, + "learning_rate": 1.236825465494073e-05, + "loss": 0.045, + "step": 26944 + }, + { + "epoch": 3.53, + "learning_rate": 1.2367739413121003e-05, + "loss": 0.0357, + "step": 26945 + }, + { + "epoch": 3.53, + "learning_rate": 1.2367224164642137e-05, + "loss": 0.0391, + "step": 26946 + }, + { + "epoch": 3.53, + "learning_rate": 1.2366708909505576e-05, + "loss": 0.0406, + "step": 26947 + }, + { + "epoch": 3.53, + "learning_rate": 1.2366193647712777e-05, + "loss": 0.0436, + "step": 26948 + }, + { + "epoch": 3.53, + "learning_rate": 1.236567837926518e-05, + "loss": 0.0346, + "step": 26949 + }, + { + "epoch": 3.53, + "learning_rate": 1.2365163104164241e-05, + "loss": 0.0408, + "step": 26950 + }, + { + "epoch": 3.53, + "learning_rate": 1.2364647822411406e-05, + "loss": 0.0435, + "step": 26951 + }, + { + "epoch": 3.53, + "learning_rate": 1.2364132534008123e-05, + "loss": 0.0378, + "step": 26952 + }, + { + "epoch": 3.53, + "learning_rate": 1.2363617238955846e-05, + "loss": 0.0331, + "step": 26953 + }, + { + "epoch": 3.53, + "learning_rate": 1.2363101937256018e-05, + "loss": 0.0394, + "step": 26954 + }, + { + "epoch": 3.53, + "learning_rate": 1.2362586628910096e-05, + "loss": 0.0418, + "step": 26955 + }, + { + "epoch": 3.53, + "learning_rate": 1.2362071313919523e-05, + "loss": 0.039, + "step": 26956 + }, + { + "epoch": 3.53, + "learning_rate": 1.2361555992285752e-05, + "loss": 0.0373, + "step": 26957 + }, + { + "epoch": 3.53, + "learning_rate": 1.2361040664010227e-05, + "loss": 0.0331, + "step": 26958 + }, + { + "epoch": 3.53, + "learning_rate": 1.2360525329094404e-05, + "loss": 0.0355, + "step": 26959 + }, + { + "epoch": 3.53, + "learning_rate": 1.2360009987539728e-05, + "loss": 0.0478, + "step": 26960 + }, + { + "epoch": 3.53, + "learning_rate": 1.2359494639347649e-05, + "loss": 0.0473, + "step": 26961 + }, + { + "epoch": 3.53, + "learning_rate": 1.2358979284519616e-05, + "loss": 0.0384, + "step": 26962 + }, + { + "epoch": 3.53, + "learning_rate": 1.235846392305708e-05, + "loss": 0.0393, + "step": 26963 + }, + { + "epoch": 3.53, + "learning_rate": 1.235794855496149e-05, + "loss": 0.0329, + "step": 26964 + }, + { + "epoch": 3.53, + "learning_rate": 1.2357433180234298e-05, + "loss": 0.0431, + "step": 26965 + }, + { + "epoch": 3.53, + "learning_rate": 1.2356917798876949e-05, + "loss": 0.0432, + "step": 26966 + }, + { + "epoch": 3.53, + "learning_rate": 1.2356402410890893e-05, + "loss": 0.0309, + "step": 26967 + }, + { + "epoch": 3.53, + "learning_rate": 1.2355887016277582e-05, + "loss": 0.0364, + "step": 26968 + }, + { + "epoch": 3.53, + "learning_rate": 1.2355371615038462e-05, + "loss": 0.0443, + "step": 26969 + }, + { + "epoch": 3.53, + "learning_rate": 1.2354856207174986e-05, + "loss": 0.0473, + "step": 26970 + }, + { + "epoch": 3.53, + "learning_rate": 1.2354340792688603e-05, + "loss": 0.041, + "step": 26971 + }, + { + "epoch": 3.53, + "learning_rate": 1.2353825371580764e-05, + "loss": 0.0363, + "step": 26972 + }, + { + "epoch": 3.53, + "learning_rate": 1.2353309943852911e-05, + "loss": 0.0403, + "step": 26973 + }, + { + "epoch": 3.53, + "learning_rate": 1.2352794509506503e-05, + "loss": 0.0407, + "step": 26974 + }, + { + "epoch": 3.53, + "learning_rate": 1.2352279068542985e-05, + "loss": 0.033, + "step": 26975 + }, + { + "epoch": 3.53, + "learning_rate": 1.2351763620963809e-05, + "loss": 0.0358, + "step": 26976 + }, + { + "epoch": 3.53, + "learning_rate": 1.235124816677042e-05, + "loss": 0.0294, + "step": 26977 + }, + { + "epoch": 3.53, + "learning_rate": 1.2350732705964273e-05, + "loss": 0.0398, + "step": 26978 + }, + { + "epoch": 3.53, + "learning_rate": 1.2350217238546813e-05, + "loss": 0.0309, + "step": 26979 + }, + { + "epoch": 3.53, + "learning_rate": 1.2349701764519494e-05, + "loss": 0.0295, + "step": 26980 + }, + { + "epoch": 3.53, + "learning_rate": 1.2349186283883763e-05, + "loss": 0.0383, + "step": 26981 + }, + { + "epoch": 3.53, + "learning_rate": 1.2348670796641074e-05, + "loss": 0.0439, + "step": 26982 + }, + { + "epoch": 3.53, + "learning_rate": 1.234815530279287e-05, + "loss": 0.0457, + "step": 26983 + }, + { + "epoch": 3.53, + "learning_rate": 1.2347639802340606e-05, + "loss": 0.038, + "step": 26984 + }, + { + "epoch": 3.53, + "learning_rate": 1.2347124295285726e-05, + "loss": 0.0384, + "step": 26985 + }, + { + "epoch": 3.53, + "learning_rate": 1.234660878162969e-05, + "loss": 0.039, + "step": 26986 + }, + { + "epoch": 3.54, + "learning_rate": 1.2346093261373937e-05, + "loss": 0.0417, + "step": 26987 + }, + { + "epoch": 3.54, + "learning_rate": 1.2345577734519924e-05, + "loss": 0.0388, + "step": 26988 + }, + { + "epoch": 3.54, + "learning_rate": 1.2345062201069099e-05, + "loss": 0.0386, + "step": 26989 + }, + { + "epoch": 3.54, + "learning_rate": 1.2344546661022909e-05, + "loss": 0.0426, + "step": 26990 + }, + { + "epoch": 3.54, + "learning_rate": 1.2344031114382806e-05, + "loss": 0.0373, + "step": 26991 + }, + { + "epoch": 3.54, + "learning_rate": 1.2343515561150246e-05, + "loss": 0.0352, + "step": 26992 + }, + { + "epoch": 3.54, + "learning_rate": 1.2343000001326669e-05, + "loss": 0.035, + "step": 26993 + }, + { + "epoch": 3.54, + "learning_rate": 1.2342484434913526e-05, + "loss": 0.0368, + "step": 26994 + }, + { + "epoch": 3.54, + "learning_rate": 1.2341968861912278e-05, + "loss": 0.0419, + "step": 26995 + }, + { + "epoch": 3.54, + "learning_rate": 1.2341453282324362e-05, + "loss": 0.039, + "step": 26996 + }, + { + "epoch": 3.54, + "learning_rate": 1.2340937696151235e-05, + "loss": 0.0345, + "step": 26997 + }, + { + "epoch": 3.54, + "learning_rate": 1.2340422103394342e-05, + "loss": 0.0323, + "step": 26998 + }, + { + "epoch": 3.54, + "learning_rate": 1.2339906504055141e-05, + "loss": 0.0396, + "step": 26999 + }, + { + "epoch": 3.54, + "learning_rate": 1.2339390898135075e-05, + "loss": 0.0426, + "step": 27000 + }, + { + "epoch": 3.54, + "learning_rate": 1.2338875285635598e-05, + "loss": 0.041, + "step": 27001 + }, + { + "epoch": 3.54, + "learning_rate": 1.2338359666558158e-05, + "loss": 0.0441, + "step": 27002 + }, + { + "epoch": 3.54, + "learning_rate": 1.2337844040904208e-05, + "loss": 0.0385, + "step": 27003 + }, + { + "epoch": 3.54, + "learning_rate": 1.2337328408675194e-05, + "loss": 0.0405, + "step": 27004 + }, + { + "epoch": 3.54, + "learning_rate": 1.233681276987257e-05, + "loss": 0.0354, + "step": 27005 + }, + { + "epoch": 3.54, + "learning_rate": 1.2336297124497783e-05, + "loss": 0.0353, + "step": 27006 + }, + { + "epoch": 3.54, + "learning_rate": 1.2335781472552284e-05, + "loss": 0.0407, + "step": 27007 + }, + { + "epoch": 3.54, + "learning_rate": 1.2335265814037525e-05, + "loss": 0.0383, + "step": 27008 + }, + { + "epoch": 3.54, + "learning_rate": 1.2334750148954958e-05, + "loss": 0.0393, + "step": 27009 + }, + { + "epoch": 3.54, + "learning_rate": 1.2334234477306026e-05, + "loss": 0.0399, + "step": 27010 + }, + { + "epoch": 3.54, + "learning_rate": 1.2333718799092186e-05, + "loss": 0.037, + "step": 27011 + }, + { + "epoch": 3.54, + "learning_rate": 1.2333203114314884e-05, + "loss": 0.0432, + "step": 27012 + }, + { + "epoch": 3.54, + "learning_rate": 1.2332687422975576e-05, + "loss": 0.0459, + "step": 27013 + }, + { + "epoch": 3.54, + "learning_rate": 1.2332171725075708e-05, + "loss": 0.0396, + "step": 27014 + }, + { + "epoch": 3.54, + "learning_rate": 1.2331656020616728e-05, + "loss": 0.0379, + "step": 27015 + }, + { + "epoch": 3.54, + "learning_rate": 1.2331140309600096e-05, + "loss": 0.0424, + "step": 27016 + }, + { + "epoch": 3.54, + "learning_rate": 1.2330624592027252e-05, + "loss": 0.0392, + "step": 27017 + }, + { + "epoch": 3.54, + "learning_rate": 1.2330108867899649e-05, + "loss": 0.0363, + "step": 27018 + }, + { + "epoch": 3.54, + "learning_rate": 1.2329593137218738e-05, + "loss": 0.04, + "step": 27019 + }, + { + "epoch": 3.54, + "learning_rate": 1.2329077399985975e-05, + "loss": 0.0347, + "step": 27020 + }, + { + "epoch": 3.54, + "learning_rate": 1.2328561656202805e-05, + "loss": 0.0354, + "step": 27021 + }, + { + "epoch": 3.54, + "learning_rate": 1.2328045905870677e-05, + "loss": 0.0339, + "step": 27022 + }, + { + "epoch": 3.54, + "learning_rate": 1.2327530148991043e-05, + "loss": 0.0324, + "step": 27023 + }, + { + "epoch": 3.54, + "learning_rate": 1.232701438556536e-05, + "loss": 0.0385, + "step": 27024 + }, + { + "epoch": 3.54, + "learning_rate": 1.2326498615595066e-05, + "loss": 0.0409, + "step": 27025 + }, + { + "epoch": 3.54, + "learning_rate": 1.2325982839081625e-05, + "loss": 0.0409, + "step": 27026 + }, + { + "epoch": 3.54, + "learning_rate": 1.2325467056026476e-05, + "loss": 0.0436, + "step": 27027 + }, + { + "epoch": 3.54, + "learning_rate": 1.2324951266431079e-05, + "loss": 0.0378, + "step": 27028 + }, + { + "epoch": 3.54, + "learning_rate": 1.2324435470296874e-05, + "loss": 0.0406, + "step": 27029 + }, + { + "epoch": 3.54, + "learning_rate": 1.2323919667625324e-05, + "loss": 0.0402, + "step": 27030 + }, + { + "epoch": 3.54, + "learning_rate": 1.232340385841787e-05, + "loss": 0.0395, + "step": 27031 + }, + { + "epoch": 3.54, + "learning_rate": 1.2322888042675972e-05, + "loss": 0.0422, + "step": 27032 + }, + { + "epoch": 3.54, + "learning_rate": 1.2322372220401069e-05, + "loss": 0.0427, + "step": 27033 + }, + { + "epoch": 3.54, + "learning_rate": 1.232185639159462e-05, + "loss": 0.044, + "step": 27034 + }, + { + "epoch": 3.54, + "learning_rate": 1.2321340556258073e-05, + "loss": 0.0434, + "step": 27035 + }, + { + "epoch": 3.54, + "learning_rate": 1.2320824714392881e-05, + "loss": 0.043, + "step": 27036 + }, + { + "epoch": 3.54, + "learning_rate": 1.2320308866000491e-05, + "loss": 0.0453, + "step": 27037 + }, + { + "epoch": 3.54, + "learning_rate": 1.2319793011082357e-05, + "loss": 0.033, + "step": 27038 + }, + { + "epoch": 3.54, + "learning_rate": 1.231927714963993e-05, + "loss": 0.0423, + "step": 27039 + }, + { + "epoch": 3.54, + "learning_rate": 1.2318761281674657e-05, + "loss": 0.0348, + "step": 27040 + }, + { + "epoch": 3.54, + "learning_rate": 1.2318245407187992e-05, + "loss": 0.0401, + "step": 27041 + }, + { + "epoch": 3.54, + "learning_rate": 1.2317729526181385e-05, + "loss": 0.0357, + "step": 27042 + }, + { + "epoch": 3.54, + "learning_rate": 1.231721363865629e-05, + "loss": 0.0484, + "step": 27043 + }, + { + "epoch": 3.54, + "learning_rate": 1.2316697744614152e-05, + "loss": 0.0547, + "step": 27044 + }, + { + "epoch": 3.54, + "learning_rate": 1.2316181844056425e-05, + "loss": 0.043, + "step": 27045 + }, + { + "epoch": 3.54, + "learning_rate": 1.231566593698456e-05, + "loss": 0.0358, + "step": 27046 + }, + { + "epoch": 3.54, + "learning_rate": 1.231515002340001e-05, + "loss": 0.0398, + "step": 27047 + }, + { + "epoch": 3.54, + "learning_rate": 1.2314634103304222e-05, + "loss": 0.0351, + "step": 27048 + }, + { + "epoch": 3.54, + "learning_rate": 1.231411817669865e-05, + "loss": 0.04, + "step": 27049 + }, + { + "epoch": 3.54, + "learning_rate": 1.2313602243584741e-05, + "loss": 0.0445, + "step": 27050 + }, + { + "epoch": 3.54, + "learning_rate": 1.2313086303963952e-05, + "loss": 0.0351, + "step": 27051 + }, + { + "epoch": 3.54, + "learning_rate": 1.231257035783773e-05, + "loss": 0.0444, + "step": 27052 + }, + { + "epoch": 3.54, + "learning_rate": 1.2312054405207529e-05, + "loss": 0.0424, + "step": 27053 + }, + { + "epoch": 3.54, + "learning_rate": 1.2311538446074795e-05, + "loss": 0.0482, + "step": 27054 + }, + { + "epoch": 3.54, + "learning_rate": 1.2311022480440982e-05, + "loss": 0.0397, + "step": 27055 + }, + { + "epoch": 3.54, + "learning_rate": 1.2310506508307543e-05, + "loss": 0.0335, + "step": 27056 + }, + { + "epoch": 3.54, + "learning_rate": 1.2309990529675928e-05, + "loss": 0.0433, + "step": 27057 + }, + { + "epoch": 3.54, + "learning_rate": 1.2309474544547585e-05, + "loss": 0.0403, + "step": 27058 + }, + { + "epoch": 3.54, + "learning_rate": 1.230895855292397e-05, + "loss": 0.0375, + "step": 27059 + }, + { + "epoch": 3.54, + "learning_rate": 1.2308442554806535e-05, + "loss": 0.0388, + "step": 27060 + }, + { + "epoch": 3.54, + "learning_rate": 1.2307926550196723e-05, + "loss": 0.0356, + "step": 27061 + }, + { + "epoch": 3.54, + "learning_rate": 1.2307410539095992e-05, + "loss": 0.0401, + "step": 27062 + }, + { + "epoch": 3.55, + "learning_rate": 1.2306894521505793e-05, + "loss": 0.0439, + "step": 27063 + }, + { + "epoch": 3.55, + "learning_rate": 1.2306378497427575e-05, + "loss": 0.0377, + "step": 27064 + }, + { + "epoch": 3.55, + "learning_rate": 1.230586246686279e-05, + "loss": 0.0314, + "step": 27065 + }, + { + "epoch": 3.55, + "learning_rate": 1.2305346429812893e-05, + "loss": 0.0383, + "step": 27066 + }, + { + "epoch": 3.55, + "learning_rate": 1.2304830386279325e-05, + "loss": 0.0381, + "step": 27067 + }, + { + "epoch": 3.55, + "learning_rate": 1.230431433626355e-05, + "loss": 0.0319, + "step": 27068 + }, + { + "epoch": 3.55, + "learning_rate": 1.230379827976701e-05, + "loss": 0.041, + "step": 27069 + }, + { + "epoch": 3.55, + "learning_rate": 1.2303282216791163e-05, + "loss": 0.041, + "step": 27070 + }, + { + "epoch": 3.55, + "learning_rate": 1.2302766147337457e-05, + "loss": 0.0393, + "step": 27071 + }, + { + "epoch": 3.55, + "learning_rate": 1.2302250071407343e-05, + "loss": 0.046, + "step": 27072 + }, + { + "epoch": 3.55, + "learning_rate": 1.230173398900227e-05, + "loss": 0.045, + "step": 27073 + }, + { + "epoch": 3.55, + "learning_rate": 1.2301217900123698e-05, + "loss": 0.0393, + "step": 27074 + }, + { + "epoch": 3.55, + "learning_rate": 1.2300701804773073e-05, + "loss": 0.0376, + "step": 27075 + }, + { + "epoch": 3.55, + "learning_rate": 1.2300185702951845e-05, + "loss": 0.0379, + "step": 27076 + }, + { + "epoch": 3.55, + "learning_rate": 1.2299669594661468e-05, + "loss": 0.0384, + "step": 27077 + }, + { + "epoch": 3.55, + "learning_rate": 1.2299153479903391e-05, + "loss": 0.038, + "step": 27078 + }, + { + "epoch": 3.55, + "learning_rate": 1.2298637358679069e-05, + "loss": 0.0394, + "step": 27079 + }, + { + "epoch": 3.55, + "learning_rate": 1.229812123098995e-05, + "loss": 0.0341, + "step": 27080 + }, + { + "epoch": 3.55, + "learning_rate": 1.2297605096837488e-05, + "loss": 0.0312, + "step": 27081 + }, + { + "epoch": 3.55, + "learning_rate": 1.2297088956223136e-05, + "loss": 0.0376, + "step": 27082 + }, + { + "epoch": 3.55, + "learning_rate": 1.2296572809148342e-05, + "loss": 0.0362, + "step": 27083 + }, + { + "epoch": 3.55, + "learning_rate": 1.2296056655614557e-05, + "loss": 0.0405, + "step": 27084 + }, + { + "epoch": 3.55, + "learning_rate": 1.2295540495623239e-05, + "loss": 0.0394, + "step": 27085 + }, + { + "epoch": 3.55, + "learning_rate": 1.2295024329175834e-05, + "loss": 0.0411, + "step": 27086 + }, + { + "epoch": 3.55, + "learning_rate": 1.2294508156273797e-05, + "loss": 0.0363, + "step": 27087 + }, + { + "epoch": 3.55, + "learning_rate": 1.2293991976918575e-05, + "loss": 0.0404, + "step": 27088 + }, + { + "epoch": 3.55, + "learning_rate": 1.2293475791111624e-05, + "loss": 0.0365, + "step": 27089 + }, + { + "epoch": 3.55, + "learning_rate": 1.2292959598854395e-05, + "loss": 0.0427, + "step": 27090 + }, + { + "epoch": 3.55, + "learning_rate": 1.229244340014834e-05, + "loss": 0.0343, + "step": 27091 + }, + { + "epoch": 3.55, + "learning_rate": 1.2291927194994908e-05, + "loss": 0.0438, + "step": 27092 + }, + { + "epoch": 3.55, + "learning_rate": 1.2291410983395555e-05, + "loss": 0.0434, + "step": 27093 + }, + { + "epoch": 3.55, + "learning_rate": 1.2290894765351727e-05, + "loss": 0.0304, + "step": 27094 + }, + { + "epoch": 3.55, + "learning_rate": 1.2290378540864882e-05, + "loss": 0.035, + "step": 27095 + }, + { + "epoch": 3.55, + "learning_rate": 1.2289862309936469e-05, + "loss": 0.0457, + "step": 27096 + }, + { + "epoch": 3.55, + "learning_rate": 1.228934607256794e-05, + "loss": 0.0378, + "step": 27097 + }, + { + "epoch": 3.55, + "learning_rate": 1.2288829828760748e-05, + "loss": 0.0419, + "step": 27098 + }, + { + "epoch": 3.55, + "learning_rate": 1.2288313578516342e-05, + "loss": 0.0377, + "step": 27099 + }, + { + "epoch": 3.55, + "learning_rate": 1.2287797321836177e-05, + "loss": 0.0455, + "step": 27100 + }, + { + "epoch": 3.55, + "learning_rate": 1.2287281058721706e-05, + "loss": 0.0342, + "step": 27101 + }, + { + "epoch": 3.55, + "learning_rate": 1.2286764789174375e-05, + "loss": 0.043, + "step": 27102 + }, + { + "epoch": 3.55, + "learning_rate": 1.2286248513195642e-05, + "loss": 0.0424, + "step": 27103 + }, + { + "epoch": 3.55, + "learning_rate": 1.2285732230786958e-05, + "loss": 0.0362, + "step": 27104 + }, + { + "epoch": 3.55, + "learning_rate": 1.2285215941949769e-05, + "loss": 0.049, + "step": 27105 + }, + { + "epoch": 3.55, + "learning_rate": 1.2284699646685537e-05, + "loss": 0.039, + "step": 27106 + }, + { + "epoch": 3.55, + "learning_rate": 1.2284183344995706e-05, + "loss": 0.0342, + "step": 27107 + }, + { + "epoch": 3.55, + "learning_rate": 1.2283667036881734e-05, + "loss": 0.0367, + "step": 27108 + }, + { + "epoch": 3.55, + "learning_rate": 1.2283150722345067e-05, + "loss": 0.043, + "step": 27109 + }, + { + "epoch": 3.55, + "learning_rate": 1.2282634401387163e-05, + "loss": 0.0383, + "step": 27110 + }, + { + "epoch": 3.55, + "learning_rate": 1.2282118074009468e-05, + "loss": 0.0445, + "step": 27111 + }, + { + "epoch": 3.55, + "learning_rate": 1.228160174021344e-05, + "loss": 0.0326, + "step": 27112 + }, + { + "epoch": 3.55, + "learning_rate": 1.2281085400000527e-05, + "loss": 0.0351, + "step": 27113 + }, + { + "epoch": 3.55, + "learning_rate": 1.2280569053372184e-05, + "loss": 0.0459, + "step": 27114 + }, + { + "epoch": 3.55, + "learning_rate": 1.2280052700329863e-05, + "loss": 0.0395, + "step": 27115 + }, + { + "epoch": 3.55, + "learning_rate": 1.2279536340875011e-05, + "loss": 0.0408, + "step": 27116 + }, + { + "epoch": 3.55, + "learning_rate": 1.2279019975009088e-05, + "loss": 0.0386, + "step": 27117 + }, + { + "epoch": 3.55, + "learning_rate": 1.2278503602733542e-05, + "loss": 0.0373, + "step": 27118 + }, + { + "epoch": 3.55, + "learning_rate": 1.2277987224049826e-05, + "loss": 0.0339, + "step": 27119 + }, + { + "epoch": 3.55, + "learning_rate": 1.227747083895939e-05, + "loss": 0.0314, + "step": 27120 + }, + { + "epoch": 3.55, + "learning_rate": 1.2276954447463691e-05, + "loss": 0.0308, + "step": 27121 + }, + { + "epoch": 3.55, + "learning_rate": 1.2276438049564181e-05, + "loss": 0.0363, + "step": 27122 + }, + { + "epoch": 3.55, + "learning_rate": 1.2275921645262306e-05, + "loss": 0.0429, + "step": 27123 + }, + { + "epoch": 3.55, + "learning_rate": 1.2275405234559523e-05, + "loss": 0.0459, + "step": 27124 + }, + { + "epoch": 3.55, + "learning_rate": 1.2274888817457286e-05, + "loss": 0.0426, + "step": 27125 + }, + { + "epoch": 3.55, + "learning_rate": 1.2274372393957045e-05, + "loss": 0.0329, + "step": 27126 + }, + { + "epoch": 3.55, + "learning_rate": 1.227385596406025e-05, + "loss": 0.0405, + "step": 27127 + }, + { + "epoch": 3.55, + "learning_rate": 1.227333952776836e-05, + "loss": 0.0453, + "step": 27128 + }, + { + "epoch": 3.55, + "learning_rate": 1.2272823085082823e-05, + "loss": 0.0379, + "step": 27129 + }, + { + "epoch": 3.55, + "learning_rate": 1.227230663600509e-05, + "loss": 0.0311, + "step": 27130 + }, + { + "epoch": 3.55, + "learning_rate": 1.2271790180536619e-05, + "loss": 0.0306, + "step": 27131 + }, + { + "epoch": 3.55, + "learning_rate": 1.2271273718678853e-05, + "loss": 0.0322, + "step": 27132 + }, + { + "epoch": 3.55, + "learning_rate": 1.2270757250433257e-05, + "loss": 0.0407, + "step": 27133 + }, + { + "epoch": 3.55, + "learning_rate": 1.2270240775801274e-05, + "loss": 0.0408, + "step": 27134 + }, + { + "epoch": 3.55, + "learning_rate": 1.2269724294784362e-05, + "loss": 0.0327, + "step": 27135 + }, + { + "epoch": 3.55, + "learning_rate": 1.2269207807383969e-05, + "loss": 0.0316, + "step": 27136 + }, + { + "epoch": 3.55, + "learning_rate": 1.226869131360155e-05, + "loss": 0.0399, + "step": 27137 + }, + { + "epoch": 3.55, + "learning_rate": 1.2268174813438556e-05, + "loss": 0.0425, + "step": 27138 + }, + { + "epoch": 3.56, + "learning_rate": 1.2267658306896447e-05, + "loss": 0.0366, + "step": 27139 + }, + { + "epoch": 3.56, + "learning_rate": 1.2267141793976664e-05, + "loss": 0.0349, + "step": 27140 + }, + { + "epoch": 3.56, + "learning_rate": 1.2266625274680666e-05, + "loss": 0.0302, + "step": 27141 + }, + { + "epoch": 3.56, + "learning_rate": 1.2266108749009908e-05, + "loss": 0.0454, + "step": 27142 + }, + { + "epoch": 3.56, + "learning_rate": 1.2265592216965837e-05, + "loss": 0.0435, + "step": 27143 + }, + { + "epoch": 3.56, + "learning_rate": 1.2265075678549912e-05, + "loss": 0.034, + "step": 27144 + }, + { + "epoch": 3.56, + "learning_rate": 1.2264559133763577e-05, + "loss": 0.0338, + "step": 27145 + }, + { + "epoch": 3.56, + "learning_rate": 1.2264042582608295e-05, + "loss": 0.0379, + "step": 27146 + }, + { + "epoch": 3.56, + "learning_rate": 1.2263526025085512e-05, + "loss": 0.0392, + "step": 27147 + }, + { + "epoch": 3.56, + "learning_rate": 1.2263009461196684e-05, + "loss": 0.038, + "step": 27148 + }, + { + "epoch": 3.56, + "learning_rate": 1.2262492890943259e-05, + "loss": 0.0448, + "step": 27149 + }, + { + "epoch": 3.56, + "learning_rate": 1.2261976314326696e-05, + "loss": 0.0291, + "step": 27150 + }, + { + "epoch": 3.56, + "learning_rate": 1.2261459731348444e-05, + "loss": 0.0382, + "step": 27151 + }, + { + "epoch": 3.56, + "learning_rate": 1.2260943142009958e-05, + "loss": 0.042, + "step": 27152 + }, + { + "epoch": 3.56, + "learning_rate": 1.226042654631269e-05, + "loss": 0.0407, + "step": 27153 + }, + { + "epoch": 3.56, + "learning_rate": 1.225990994425809e-05, + "loss": 0.0478, + "step": 27154 + }, + { + "epoch": 3.56, + "learning_rate": 1.2259393335847614e-05, + "loss": 0.0331, + "step": 27155 + }, + { + "epoch": 3.56, + "learning_rate": 1.2258876721082717e-05, + "loss": 0.0443, + "step": 27156 + }, + { + "epoch": 3.56, + "learning_rate": 1.2258360099964848e-05, + "loss": 0.0433, + "step": 27157 + }, + { + "epoch": 3.56, + "learning_rate": 1.2257843472495463e-05, + "loss": 0.0436, + "step": 27158 + }, + { + "epoch": 3.56, + "learning_rate": 1.2257326838676013e-05, + "loss": 0.0352, + "step": 27159 + }, + { + "epoch": 3.56, + "learning_rate": 1.2256810198507951e-05, + "loss": 0.0391, + "step": 27160 + }, + { + "epoch": 3.56, + "learning_rate": 1.2256293551992728e-05, + "loss": 0.0358, + "step": 27161 + }, + { + "epoch": 3.56, + "learning_rate": 1.2255776899131804e-05, + "loss": 0.0326, + "step": 27162 + }, + { + "epoch": 3.56, + "learning_rate": 1.2255260239926626e-05, + "loss": 0.0412, + "step": 27163 + }, + { + "epoch": 3.56, + "learning_rate": 1.2254743574378646e-05, + "loss": 0.0345, + "step": 27164 + }, + { + "epoch": 3.56, + "learning_rate": 1.225422690248932e-05, + "loss": 0.0337, + "step": 27165 + }, + { + "epoch": 3.56, + "learning_rate": 1.2253710224260104e-05, + "loss": 0.0356, + "step": 27166 + }, + { + "epoch": 3.56, + "learning_rate": 1.2253193539692449e-05, + "loss": 0.0362, + "step": 27167 + }, + { + "epoch": 3.56, + "learning_rate": 1.2252676848787803e-05, + "loss": 0.0394, + "step": 27168 + }, + { + "epoch": 3.56, + "learning_rate": 1.2252160151547628e-05, + "loss": 0.045, + "step": 27169 + }, + { + "epoch": 3.56, + "learning_rate": 1.225164344797337e-05, + "loss": 0.0372, + "step": 27170 + }, + { + "epoch": 3.56, + "learning_rate": 1.2251126738066483e-05, + "loss": 0.0448, + "step": 27171 + }, + { + "epoch": 3.56, + "learning_rate": 1.2250610021828422e-05, + "loss": 0.0368, + "step": 27172 + }, + { + "epoch": 3.56, + "learning_rate": 1.2250093299260644e-05, + "loss": 0.0429, + "step": 27173 + }, + { + "epoch": 3.56, + "learning_rate": 1.2249576570364593e-05, + "loss": 0.0375, + "step": 27174 + }, + { + "epoch": 3.56, + "learning_rate": 1.2249059835141735e-05, + "loss": 0.0329, + "step": 27175 + }, + { + "epoch": 3.56, + "learning_rate": 1.2248543093593508e-05, + "loss": 0.0404, + "step": 27176 + }, + { + "epoch": 3.56, + "learning_rate": 1.224802634572138e-05, + "loss": 0.0355, + "step": 27177 + }, + { + "epoch": 3.56, + "learning_rate": 1.2247509591526794e-05, + "loss": 0.0341, + "step": 27178 + }, + { + "epoch": 3.56, + "learning_rate": 1.2246992831011209e-05, + "loss": 0.0397, + "step": 27179 + }, + { + "epoch": 3.56, + "learning_rate": 1.2246476064176074e-05, + "loss": 0.0338, + "step": 27180 + }, + { + "epoch": 3.56, + "learning_rate": 1.2245959291022844e-05, + "loss": 0.0379, + "step": 27181 + }, + { + "epoch": 3.56, + "learning_rate": 1.2245442511552975e-05, + "loss": 0.0381, + "step": 27182 + }, + { + "epoch": 3.56, + "learning_rate": 1.224492572576792e-05, + "loss": 0.0429, + "step": 27183 + }, + { + "epoch": 3.56, + "learning_rate": 1.2244408933669128e-05, + "loss": 0.0381, + "step": 27184 + }, + { + "epoch": 3.56, + "learning_rate": 1.2243892135258056e-05, + "loss": 0.0409, + "step": 27185 + }, + { + "epoch": 3.56, + "learning_rate": 1.2243375330536158e-05, + "loss": 0.036, + "step": 27186 + }, + { + "epoch": 3.56, + "learning_rate": 1.2242858519504886e-05, + "loss": 0.041, + "step": 27187 + }, + { + "epoch": 3.56, + "learning_rate": 1.2242341702165692e-05, + "loss": 0.0401, + "step": 27188 + }, + { + "epoch": 3.56, + "learning_rate": 1.2241824878520033e-05, + "loss": 0.0377, + "step": 27189 + }, + { + "epoch": 3.56, + "learning_rate": 1.2241308048569363e-05, + "loss": 0.036, + "step": 27190 + }, + { + "epoch": 3.56, + "learning_rate": 1.224079121231513e-05, + "loss": 0.037, + "step": 27191 + }, + { + "epoch": 3.56, + "learning_rate": 1.2240274369758792e-05, + "loss": 0.04, + "step": 27192 + }, + { + "epoch": 3.56, + "learning_rate": 1.22397575209018e-05, + "loss": 0.0427, + "step": 27193 + }, + { + "epoch": 3.56, + "learning_rate": 1.2239240665745614e-05, + "loss": 0.0431, + "step": 27194 + }, + { + "epoch": 3.56, + "learning_rate": 1.2238723804291678e-05, + "loss": 0.0397, + "step": 27195 + }, + { + "epoch": 3.56, + "learning_rate": 1.2238206936541457e-05, + "loss": 0.0389, + "step": 27196 + }, + { + "epoch": 3.56, + "learning_rate": 1.2237690062496392e-05, + "loss": 0.0418, + "step": 27197 + }, + { + "epoch": 3.56, + "learning_rate": 1.2237173182157944e-05, + "loss": 0.0392, + "step": 27198 + }, + { + "epoch": 3.56, + "learning_rate": 1.2236656295527566e-05, + "loss": 0.0367, + "step": 27199 + }, + { + "epoch": 3.56, + "learning_rate": 1.2236139402606713e-05, + "loss": 0.0339, + "step": 27200 + }, + { + "epoch": 3.56, + "learning_rate": 1.2235622503396836e-05, + "loss": 0.0459, + "step": 27201 + }, + { + "epoch": 3.56, + "learning_rate": 1.2235105597899391e-05, + "loss": 0.0401, + "step": 27202 + }, + { + "epoch": 3.56, + "learning_rate": 1.2234588686115825e-05, + "loss": 0.0463, + "step": 27203 + }, + { + "epoch": 3.56, + "learning_rate": 1.2234071768047602e-05, + "loss": 0.0406, + "step": 27204 + }, + { + "epoch": 3.56, + "learning_rate": 1.2233554843696171e-05, + "loss": 0.0475, + "step": 27205 + }, + { + "epoch": 3.56, + "learning_rate": 1.2233037913062985e-05, + "loss": 0.0424, + "step": 27206 + }, + { + "epoch": 3.56, + "learning_rate": 1.22325209761495e-05, + "loss": 0.0431, + "step": 27207 + }, + { + "epoch": 3.56, + "learning_rate": 1.2232004032957168e-05, + "loss": 0.0417, + "step": 27208 + }, + { + "epoch": 3.56, + "learning_rate": 1.223148708348744e-05, + "loss": 0.0354, + "step": 27209 + }, + { + "epoch": 3.56, + "learning_rate": 1.2230970127741777e-05, + "loss": 0.0347, + "step": 27210 + }, + { + "epoch": 3.56, + "learning_rate": 1.2230453165721633e-05, + "loss": 0.0334, + "step": 27211 + }, + { + "epoch": 3.56, + "learning_rate": 1.2229936197428452e-05, + "loss": 0.037, + "step": 27212 + }, + { + "epoch": 3.56, + "learning_rate": 1.2229419222863696e-05, + "loss": 0.0405, + "step": 27213 + }, + { + "epoch": 3.56, + "learning_rate": 1.2228902242028817e-05, + "loss": 0.0407, + "step": 27214 + }, + { + "epoch": 3.56, + "learning_rate": 1.222838525492527e-05, + "loss": 0.0313, + "step": 27215 + }, + { + "epoch": 3.57, + "learning_rate": 1.2227868261554507e-05, + "loss": 0.0347, + "step": 27216 + }, + { + "epoch": 3.57, + "learning_rate": 1.2227351261917987e-05, + "loss": 0.0449, + "step": 27217 + }, + { + "epoch": 3.57, + "learning_rate": 1.2226834256017154e-05, + "loss": 0.0399, + "step": 27218 + }, + { + "epoch": 3.57, + "learning_rate": 1.2226317243853476e-05, + "loss": 0.037, + "step": 27219 + }, + { + "epoch": 3.57, + "learning_rate": 1.222580022542839e-05, + "loss": 0.037, + "step": 27220 + }, + { + "epoch": 3.57, + "learning_rate": 1.2225283200743366e-05, + "loss": 0.0358, + "step": 27221 + }, + { + "epoch": 3.57, + "learning_rate": 1.2224766169799849e-05, + "loss": 0.0445, + "step": 27222 + }, + { + "epoch": 3.57, + "learning_rate": 1.2224249132599298e-05, + "loss": 0.0349, + "step": 27223 + }, + { + "epoch": 3.57, + "learning_rate": 1.2223732089143164e-05, + "loss": 0.0338, + "step": 27224 + }, + { + "epoch": 3.57, + "learning_rate": 1.2223215039432898e-05, + "loss": 0.0341, + "step": 27225 + }, + { + "epoch": 3.57, + "learning_rate": 1.2222697983469963e-05, + "loss": 0.0343, + "step": 27226 + }, + { + "epoch": 3.57, + "learning_rate": 1.2222180921255807e-05, + "loss": 0.0435, + "step": 27227 + }, + { + "epoch": 3.57, + "learning_rate": 1.2221663852791883e-05, + "loss": 0.0444, + "step": 27228 + }, + { + "epoch": 3.57, + "learning_rate": 1.2221146778079649e-05, + "loss": 0.0401, + "step": 27229 + }, + { + "epoch": 3.57, + "learning_rate": 1.222062969712056e-05, + "loss": 0.0365, + "step": 27230 + }, + { + "epoch": 3.57, + "learning_rate": 1.2220112609916063e-05, + "loss": 0.0384, + "step": 27231 + }, + { + "epoch": 3.57, + "learning_rate": 1.2219595516467623e-05, + "loss": 0.0394, + "step": 27232 + }, + { + "epoch": 3.57, + "learning_rate": 1.2219078416776688e-05, + "loss": 0.0412, + "step": 27233 + }, + { + "epoch": 3.57, + "learning_rate": 1.2218561310844714e-05, + "loss": 0.0475, + "step": 27234 + }, + { + "epoch": 3.57, + "learning_rate": 1.221804419867315e-05, + "loss": 0.045, + "step": 27235 + }, + { + "epoch": 3.57, + "learning_rate": 1.2217527080263458e-05, + "loss": 0.0309, + "step": 27236 + }, + { + "epoch": 3.57, + "learning_rate": 1.2217009955617087e-05, + "loss": 0.0383, + "step": 27237 + }, + { + "epoch": 3.57, + "learning_rate": 1.2216492824735498e-05, + "loss": 0.0432, + "step": 27238 + }, + { + "epoch": 3.57, + "learning_rate": 1.2215975687620138e-05, + "loss": 0.0346, + "step": 27239 + }, + { + "epoch": 3.57, + "learning_rate": 1.2215458544272466e-05, + "loss": 0.0399, + "step": 27240 + }, + { + "epoch": 3.57, + "learning_rate": 1.2214941394693934e-05, + "loss": 0.0408, + "step": 27241 + }, + { + "epoch": 3.57, + "learning_rate": 1.2214424238885995e-05, + "loss": 0.0348, + "step": 27242 + }, + { + "epoch": 3.57, + "learning_rate": 1.2213907076850106e-05, + "loss": 0.0342, + "step": 27243 + }, + { + "epoch": 3.57, + "learning_rate": 1.2213389908587725e-05, + "loss": 0.049, + "step": 27244 + }, + { + "epoch": 3.57, + "learning_rate": 1.2212872734100299e-05, + "loss": 0.0341, + "step": 27245 + }, + { + "epoch": 3.57, + "learning_rate": 1.2212355553389288e-05, + "loss": 0.0397, + "step": 27246 + }, + { + "epoch": 3.57, + "learning_rate": 1.2211838366456144e-05, + "loss": 0.0394, + "step": 27247 + }, + { + "epoch": 3.57, + "learning_rate": 1.2211321173302325e-05, + "loss": 0.0403, + "step": 27248 + }, + { + "epoch": 3.57, + "learning_rate": 1.2210803973929279e-05, + "loss": 0.0329, + "step": 27249 + }, + { + "epoch": 3.57, + "learning_rate": 1.2210286768338466e-05, + "loss": 0.0365, + "step": 27250 + }, + { + "epoch": 3.57, + "learning_rate": 1.2209769556531341e-05, + "loss": 0.0416, + "step": 27251 + }, + { + "epoch": 3.57, + "learning_rate": 1.2209252338509354e-05, + "loss": 0.0297, + "step": 27252 + }, + { + "epoch": 3.57, + "learning_rate": 1.2208735114273963e-05, + "loss": 0.0382, + "step": 27253 + }, + { + "epoch": 3.57, + "learning_rate": 1.2208217883826623e-05, + "loss": 0.0406, + "step": 27254 + }, + { + "epoch": 3.57, + "learning_rate": 1.220770064716879e-05, + "loss": 0.0427, + "step": 27255 + }, + { + "epoch": 3.57, + "learning_rate": 1.2207183404301911e-05, + "loss": 0.0376, + "step": 27256 + }, + { + "epoch": 3.57, + "learning_rate": 1.2206666155227454e-05, + "loss": 0.0397, + "step": 27257 + }, + { + "epoch": 3.57, + "learning_rate": 1.2206148899946856e-05, + "loss": 0.0411, + "step": 27258 + }, + { + "epoch": 3.57, + "learning_rate": 1.2205631638461591e-05, + "loss": 0.0347, + "step": 27259 + }, + { + "epoch": 3.57, + "learning_rate": 1.2205114370773098e-05, + "loss": 0.035, + "step": 27260 + }, + { + "epoch": 3.57, + "learning_rate": 1.2204597096882844e-05, + "loss": 0.0363, + "step": 27261 + }, + { + "epoch": 3.57, + "learning_rate": 1.2204079816792273e-05, + "loss": 0.0309, + "step": 27262 + }, + { + "epoch": 3.57, + "learning_rate": 1.2203562530502846e-05, + "loss": 0.0435, + "step": 27263 + }, + { + "epoch": 3.57, + "learning_rate": 1.2203045238016017e-05, + "loss": 0.0349, + "step": 27264 + }, + { + "epoch": 3.57, + "learning_rate": 1.2202527939333244e-05, + "loss": 0.0374, + "step": 27265 + }, + { + "epoch": 3.57, + "learning_rate": 1.2202010634455974e-05, + "loss": 0.0298, + "step": 27266 + }, + { + "epoch": 3.57, + "learning_rate": 1.2201493323385669e-05, + "loss": 0.0351, + "step": 27267 + }, + { + "epoch": 3.57, + "learning_rate": 1.2200976006123782e-05, + "loss": 0.0257, + "step": 27268 + }, + { + "epoch": 3.57, + "learning_rate": 1.2200458682671763e-05, + "loss": 0.0349, + "step": 27269 + }, + { + "epoch": 3.57, + "learning_rate": 1.2199941353031073e-05, + "loss": 0.0389, + "step": 27270 + }, + { + "epoch": 3.57, + "learning_rate": 1.2199424017203166e-05, + "loss": 0.0413, + "step": 27271 + }, + { + "epoch": 3.57, + "learning_rate": 1.2198906675189497e-05, + "loss": 0.0413, + "step": 27272 + }, + { + "epoch": 3.57, + "learning_rate": 1.2198389326991518e-05, + "loss": 0.0392, + "step": 27273 + }, + { + "epoch": 3.57, + "learning_rate": 1.2197871972610687e-05, + "loss": 0.0339, + "step": 27274 + }, + { + "epoch": 3.57, + "learning_rate": 1.2197354612048456e-05, + "loss": 0.0394, + "step": 27275 + }, + { + "epoch": 3.57, + "learning_rate": 1.2196837245306286e-05, + "loss": 0.0415, + "step": 27276 + }, + { + "epoch": 3.57, + "learning_rate": 1.2196319872385625e-05, + "loss": 0.0277, + "step": 27277 + }, + { + "epoch": 3.57, + "learning_rate": 1.2195802493287934e-05, + "loss": 0.04, + "step": 27278 + }, + { + "epoch": 3.57, + "learning_rate": 1.2195285108014665e-05, + "loss": 0.0363, + "step": 27279 + }, + { + "epoch": 3.57, + "learning_rate": 1.2194767716567271e-05, + "loss": 0.0383, + "step": 27280 + }, + { + "epoch": 3.57, + "learning_rate": 1.2194250318947211e-05, + "loss": 0.0368, + "step": 27281 + }, + { + "epoch": 3.57, + "learning_rate": 1.219373291515594e-05, + "loss": 0.0376, + "step": 27282 + }, + { + "epoch": 3.57, + "learning_rate": 1.219321550519491e-05, + "loss": 0.0377, + "step": 27283 + }, + { + "epoch": 3.57, + "learning_rate": 1.2192698089065583e-05, + "loss": 0.0369, + "step": 27284 + }, + { + "epoch": 3.57, + "learning_rate": 1.2192180666769401e-05, + "loss": 0.0413, + "step": 27285 + }, + { + "epoch": 3.57, + "learning_rate": 1.2191663238307837e-05, + "loss": 0.0319, + "step": 27286 + }, + { + "epoch": 3.57, + "learning_rate": 1.2191145803682329e-05, + "loss": 0.0375, + "step": 27287 + }, + { + "epoch": 3.57, + "learning_rate": 1.2190628362894346e-05, + "loss": 0.0344, + "step": 27288 + }, + { + "epoch": 3.57, + "learning_rate": 1.2190110915945335e-05, + "loss": 0.0376, + "step": 27289 + }, + { + "epoch": 3.57, + "learning_rate": 1.2189593462836751e-05, + "loss": 0.0377, + "step": 27290 + }, + { + "epoch": 3.57, + "learning_rate": 1.2189076003570053e-05, + "loss": 0.0366, + "step": 27291 + }, + { + "epoch": 3.58, + "learning_rate": 1.21885585381467e-05, + "loss": 0.0365, + "step": 27292 + }, + { + "epoch": 3.58, + "learning_rate": 1.2188041066568138e-05, + "loss": 0.0385, + "step": 27293 + }, + { + "epoch": 3.58, + "learning_rate": 1.2187523588835829e-05, + "loss": 0.0379, + "step": 27294 + }, + { + "epoch": 3.58, + "learning_rate": 1.2187006104951226e-05, + "loss": 0.0358, + "step": 27295 + }, + { + "epoch": 3.58, + "learning_rate": 1.2186488614915782e-05, + "loss": 0.0362, + "step": 27296 + }, + { + "epoch": 3.58, + "learning_rate": 1.2185971118730958e-05, + "loss": 0.0486, + "step": 27297 + }, + { + "epoch": 3.58, + "learning_rate": 1.2185453616398207e-05, + "loss": 0.0363, + "step": 27298 + }, + { + "epoch": 3.58, + "learning_rate": 1.2184936107918985e-05, + "loss": 0.0383, + "step": 27299 + }, + { + "epoch": 3.58, + "learning_rate": 1.2184418593294745e-05, + "loss": 0.0332, + "step": 27300 + }, + { + "epoch": 3.58, + "learning_rate": 1.2183901072526945e-05, + "loss": 0.0463, + "step": 27301 + }, + { + "epoch": 3.58, + "learning_rate": 1.2183383545617036e-05, + "loss": 0.0313, + "step": 27302 + }, + { + "epoch": 3.58, + "learning_rate": 1.2182866012566481e-05, + "loss": 0.0314, + "step": 27303 + }, + { + "epoch": 3.58, + "learning_rate": 1.218234847337673e-05, + "loss": 0.0354, + "step": 27304 + }, + { + "epoch": 3.58, + "learning_rate": 1.2181830928049241e-05, + "loss": 0.0387, + "step": 27305 + }, + { + "epoch": 3.58, + "learning_rate": 1.2181313376585468e-05, + "loss": 0.0366, + "step": 27306 + }, + { + "epoch": 3.58, + "learning_rate": 1.2180795818986866e-05, + "loss": 0.0354, + "step": 27307 + }, + { + "epoch": 3.58, + "learning_rate": 1.2180278255254894e-05, + "loss": 0.0368, + "step": 27308 + }, + { + "epoch": 3.58, + "learning_rate": 1.2179760685391006e-05, + "loss": 0.0364, + "step": 27309 + }, + { + "epoch": 3.58, + "learning_rate": 1.2179243109396654e-05, + "loss": 0.0506, + "step": 27310 + }, + { + "epoch": 3.58, + "learning_rate": 1.2178725527273298e-05, + "loss": 0.0369, + "step": 27311 + }, + { + "epoch": 3.58, + "learning_rate": 1.2178207939022393e-05, + "loss": 0.0484, + "step": 27312 + }, + { + "epoch": 3.58, + "learning_rate": 1.2177690344645397e-05, + "loss": 0.0316, + "step": 27313 + }, + { + "epoch": 3.58, + "learning_rate": 1.217717274414376e-05, + "loss": 0.0398, + "step": 27314 + }, + { + "epoch": 3.58, + "learning_rate": 1.217665513751894e-05, + "loss": 0.0396, + "step": 27315 + }, + { + "epoch": 3.58, + "learning_rate": 1.2176137524772396e-05, + "loss": 0.0391, + "step": 27316 + }, + { + "epoch": 3.58, + "learning_rate": 1.2175619905905579e-05, + "loss": 0.0371, + "step": 27317 + }, + { + "epoch": 3.58, + "learning_rate": 1.2175102280919946e-05, + "loss": 0.0349, + "step": 27318 + }, + { + "epoch": 3.58, + "learning_rate": 1.2174584649816955e-05, + "loss": 0.0373, + "step": 27319 + }, + { + "epoch": 3.58, + "learning_rate": 1.217406701259806e-05, + "loss": 0.04, + "step": 27320 + }, + { + "epoch": 3.58, + "learning_rate": 1.2173549369264718e-05, + "loss": 0.0371, + "step": 27321 + }, + { + "epoch": 3.58, + "learning_rate": 1.2173031719818385e-05, + "loss": 0.0474, + "step": 27322 + }, + { + "epoch": 3.58, + "learning_rate": 1.2172514064260513e-05, + "loss": 0.0423, + "step": 27323 + }, + { + "epoch": 3.58, + "learning_rate": 1.2171996402592563e-05, + "loss": 0.035, + "step": 27324 + }, + { + "epoch": 3.58, + "learning_rate": 1.217147873481599e-05, + "loss": 0.041, + "step": 27325 + }, + { + "epoch": 3.58, + "learning_rate": 1.2170961060932248e-05, + "loss": 0.0366, + "step": 27326 + }, + { + "epoch": 3.58, + "learning_rate": 1.217044338094279e-05, + "loss": 0.0385, + "step": 27327 + }, + { + "epoch": 3.58, + "learning_rate": 1.216992569484908e-05, + "loss": 0.0383, + "step": 27328 + }, + { + "epoch": 3.58, + "learning_rate": 1.2169408002652566e-05, + "loss": 0.033, + "step": 27329 + }, + { + "epoch": 3.58, + "learning_rate": 1.216889030435471e-05, + "loss": 0.0295, + "step": 27330 + }, + { + "epoch": 3.58, + "learning_rate": 1.2168372599956966e-05, + "loss": 0.037, + "step": 27331 + }, + { + "epoch": 3.58, + "learning_rate": 1.2167854889460787e-05, + "loss": 0.0414, + "step": 27332 + }, + { + "epoch": 3.58, + "learning_rate": 1.2167337172867635e-05, + "loss": 0.0426, + "step": 27333 + }, + { + "epoch": 3.58, + "learning_rate": 1.2166819450178959e-05, + "loss": 0.0354, + "step": 27334 + }, + { + "epoch": 3.58, + "learning_rate": 1.2166301721396221e-05, + "loss": 0.0344, + "step": 27335 + }, + { + "epoch": 3.58, + "learning_rate": 1.216578398652087e-05, + "loss": 0.0408, + "step": 27336 + }, + { + "epoch": 3.58, + "learning_rate": 1.2165266245554374e-05, + "loss": 0.0402, + "step": 27337 + }, + { + "epoch": 3.58, + "learning_rate": 1.216474849849818e-05, + "loss": 0.0364, + "step": 27338 + }, + { + "epoch": 3.58, + "learning_rate": 1.2164230745353747e-05, + "loss": 0.0377, + "step": 27339 + }, + { + "epoch": 3.58, + "learning_rate": 1.2163712986122525e-05, + "loss": 0.0319, + "step": 27340 + }, + { + "epoch": 3.58, + "learning_rate": 1.2163195220805979e-05, + "loss": 0.0354, + "step": 27341 + }, + { + "epoch": 3.58, + "learning_rate": 1.2162677449405561e-05, + "loss": 0.0357, + "step": 27342 + }, + { + "epoch": 3.58, + "learning_rate": 1.216215967192273e-05, + "loss": 0.0391, + "step": 27343 + }, + { + "epoch": 3.58, + "learning_rate": 1.2161641888358939e-05, + "loss": 0.0328, + "step": 27344 + }, + { + "epoch": 3.58, + "learning_rate": 1.2161124098715644e-05, + "loss": 0.0321, + "step": 27345 + }, + { + "epoch": 3.58, + "learning_rate": 1.2160606302994302e-05, + "loss": 0.0369, + "step": 27346 + }, + { + "epoch": 3.58, + "learning_rate": 1.2160088501196372e-05, + "loss": 0.0373, + "step": 27347 + }, + { + "epoch": 3.58, + "learning_rate": 1.2159570693323309e-05, + "loss": 0.0422, + "step": 27348 + }, + { + "epoch": 3.58, + "learning_rate": 1.2159052879376567e-05, + "loss": 0.0416, + "step": 27349 + }, + { + "epoch": 3.58, + "learning_rate": 1.2158535059357602e-05, + "loss": 0.0409, + "step": 27350 + }, + { + "epoch": 3.58, + "learning_rate": 1.2158017233267874e-05, + "loss": 0.0396, + "step": 27351 + }, + { + "epoch": 3.58, + "learning_rate": 1.2157499401108836e-05, + "loss": 0.0361, + "step": 27352 + }, + { + "epoch": 3.58, + "learning_rate": 1.2156981562881947e-05, + "loss": 0.0406, + "step": 27353 + }, + { + "epoch": 3.58, + "learning_rate": 1.2156463718588663e-05, + "loss": 0.0356, + "step": 27354 + }, + { + "epoch": 3.58, + "learning_rate": 1.2155945868230436e-05, + "loss": 0.0367, + "step": 27355 + }, + { + "epoch": 3.58, + "learning_rate": 1.2155428011808729e-05, + "loss": 0.0328, + "step": 27356 + }, + { + "epoch": 3.58, + "learning_rate": 1.2154910149324994e-05, + "loss": 0.0447, + "step": 27357 + }, + { + "epoch": 3.58, + "learning_rate": 1.2154392280780692e-05, + "loss": 0.0349, + "step": 27358 + }, + { + "epoch": 3.58, + "learning_rate": 1.2153874406177275e-05, + "loss": 0.0428, + "step": 27359 + }, + { + "epoch": 3.58, + "learning_rate": 1.21533565255162e-05, + "loss": 0.0291, + "step": 27360 + }, + { + "epoch": 3.58, + "learning_rate": 1.2152838638798925e-05, + "loss": 0.0397, + "step": 27361 + }, + { + "epoch": 3.58, + "learning_rate": 1.2152320746026906e-05, + "loss": 0.036, + "step": 27362 + }, + { + "epoch": 3.58, + "learning_rate": 1.21518028472016e-05, + "loss": 0.0415, + "step": 27363 + }, + { + "epoch": 3.58, + "learning_rate": 1.2151284942324462e-05, + "loss": 0.0384, + "step": 27364 + }, + { + "epoch": 3.58, + "learning_rate": 1.215076703139695e-05, + "loss": 0.043, + "step": 27365 + }, + { + "epoch": 3.58, + "learning_rate": 1.2150249114420522e-05, + "loss": 0.0397, + "step": 27366 + }, + { + "epoch": 3.58, + "learning_rate": 1.2149731191396628e-05, + "loss": 0.0392, + "step": 27367 + }, + { + "epoch": 3.59, + "learning_rate": 1.2149213262326735e-05, + "loss": 0.0398, + "step": 27368 + }, + { + "epoch": 3.59, + "learning_rate": 1.2148695327212293e-05, + "loss": 0.0358, + "step": 27369 + }, + { + "epoch": 3.59, + "learning_rate": 1.2148177386054758e-05, + "loss": 0.0285, + "step": 27370 + }, + { + "epoch": 3.59, + "learning_rate": 1.214765943885559e-05, + "loss": 0.0377, + "step": 27371 + }, + { + "epoch": 3.59, + "learning_rate": 1.2147141485616243e-05, + "loss": 0.0403, + "step": 27372 + }, + { + "epoch": 3.59, + "learning_rate": 1.2146623526338175e-05, + "loss": 0.0444, + "step": 27373 + }, + { + "epoch": 3.59, + "learning_rate": 1.2146105561022843e-05, + "loss": 0.0373, + "step": 27374 + }, + { + "epoch": 3.59, + "learning_rate": 1.2145587589671704e-05, + "loss": 0.0435, + "step": 27375 + }, + { + "epoch": 3.59, + "learning_rate": 1.2145069612286213e-05, + "loss": 0.041, + "step": 27376 + }, + { + "epoch": 3.59, + "learning_rate": 1.214455162886783e-05, + "loss": 0.038, + "step": 27377 + }, + { + "epoch": 3.59, + "learning_rate": 1.2144033639418007e-05, + "loss": 0.0355, + "step": 27378 + }, + { + "epoch": 3.59, + "learning_rate": 1.2143515643938203e-05, + "loss": 0.0391, + "step": 27379 + }, + { + "epoch": 3.59, + "learning_rate": 1.2142997642429876e-05, + "loss": 0.0384, + "step": 27380 + }, + { + "epoch": 3.59, + "learning_rate": 1.2142479634894486e-05, + "loss": 0.0409, + "step": 27381 + }, + { + "epoch": 3.59, + "learning_rate": 1.2141961621333485e-05, + "loss": 0.0412, + "step": 27382 + }, + { + "epoch": 3.59, + "learning_rate": 1.2141443601748328e-05, + "loss": 0.041, + "step": 27383 + }, + { + "epoch": 3.59, + "learning_rate": 1.2140925576140477e-05, + "loss": 0.0384, + "step": 27384 + }, + { + "epoch": 3.59, + "learning_rate": 1.2140407544511389e-05, + "loss": 0.0327, + "step": 27385 + }, + { + "epoch": 3.59, + "learning_rate": 1.2139889506862512e-05, + "loss": 0.0366, + "step": 27386 + }, + { + "epoch": 3.59, + "learning_rate": 1.2139371463195317e-05, + "loss": 0.0394, + "step": 27387 + }, + { + "epoch": 3.59, + "learning_rate": 1.2138853413511249e-05, + "loss": 0.0317, + "step": 27388 + }, + { + "epoch": 3.59, + "learning_rate": 1.213833535781177e-05, + "loss": 0.0354, + "step": 27389 + }, + { + "epoch": 3.59, + "learning_rate": 1.2137817296098338e-05, + "loss": 0.035, + "step": 27390 + }, + { + "epoch": 3.59, + "learning_rate": 1.213729922837241e-05, + "loss": 0.0304, + "step": 27391 + }, + { + "epoch": 3.59, + "learning_rate": 1.213678115463544e-05, + "loss": 0.0369, + "step": 27392 + }, + { + "epoch": 3.59, + "learning_rate": 1.2136263074888888e-05, + "loss": 0.0357, + "step": 27393 + }, + { + "epoch": 3.59, + "learning_rate": 1.2135744989134207e-05, + "loss": 0.037, + "step": 27394 + }, + { + "epoch": 3.59, + "learning_rate": 1.213522689737286e-05, + "loss": 0.0378, + "step": 27395 + }, + { + "epoch": 3.59, + "learning_rate": 1.2134708799606301e-05, + "loss": 0.0375, + "step": 27396 + }, + { + "epoch": 3.59, + "learning_rate": 1.2134190695835986e-05, + "loss": 0.0361, + "step": 27397 + }, + { + "epoch": 3.59, + "learning_rate": 1.2133672586063374e-05, + "loss": 0.04, + "step": 27398 + }, + { + "epoch": 3.59, + "learning_rate": 1.2133154470289921e-05, + "loss": 0.0418, + "step": 27399 + }, + { + "epoch": 3.59, + "learning_rate": 1.2132636348517081e-05, + "loss": 0.0408, + "step": 27400 + }, + { + "epoch": 3.59, + "learning_rate": 1.2132118220746321e-05, + "loss": 0.034, + "step": 27401 + }, + { + "epoch": 3.59, + "learning_rate": 1.213160008697909e-05, + "loss": 0.0342, + "step": 27402 + }, + { + "epoch": 3.59, + "learning_rate": 1.2131081947216844e-05, + "loss": 0.0404, + "step": 27403 + }, + { + "epoch": 3.59, + "learning_rate": 1.2130563801461051e-05, + "loss": 0.0386, + "step": 27404 + }, + { + "epoch": 3.59, + "learning_rate": 1.2130045649713152e-05, + "loss": 0.0348, + "step": 27405 + }, + { + "epoch": 3.59, + "learning_rate": 1.212952749197462e-05, + "loss": 0.0443, + "step": 27406 + }, + { + "epoch": 3.59, + "learning_rate": 1.2129009328246902e-05, + "loss": 0.0369, + "step": 27407 + }, + { + "epoch": 3.59, + "learning_rate": 1.2128491158531458e-05, + "loss": 0.0379, + "step": 27408 + }, + { + "epoch": 3.59, + "learning_rate": 1.2127972982829747e-05, + "loss": 0.0351, + "step": 27409 + }, + { + "epoch": 3.59, + "learning_rate": 1.2127454801143228e-05, + "loss": 0.0436, + "step": 27410 + }, + { + "epoch": 3.59, + "learning_rate": 1.212693661347335e-05, + "loss": 0.0321, + "step": 27411 + }, + { + "epoch": 3.59, + "learning_rate": 1.212641841982158e-05, + "loss": 0.0355, + "step": 27412 + }, + { + "epoch": 3.59, + "learning_rate": 1.2125900220189371e-05, + "loss": 0.0424, + "step": 27413 + }, + { + "epoch": 3.59, + "learning_rate": 1.2125382014578179e-05, + "loss": 0.0438, + "step": 27414 + }, + { + "epoch": 3.59, + "learning_rate": 1.2124863802989465e-05, + "loss": 0.039, + "step": 27415 + }, + { + "epoch": 3.59, + "learning_rate": 1.2124345585424683e-05, + "loss": 0.0423, + "step": 27416 + }, + { + "epoch": 3.59, + "learning_rate": 1.2123827361885294e-05, + "loss": 0.0413, + "step": 27417 + }, + { + "epoch": 3.59, + "learning_rate": 1.2123309132372756e-05, + "loss": 0.0452, + "step": 27418 + }, + { + "epoch": 3.59, + "learning_rate": 1.2122790896888518e-05, + "loss": 0.0463, + "step": 27419 + }, + { + "epoch": 3.59, + "learning_rate": 1.2122272655434048e-05, + "loss": 0.0407, + "step": 27420 + }, + { + "epoch": 3.59, + "learning_rate": 1.21217544080108e-05, + "loss": 0.0421, + "step": 27421 + }, + { + "epoch": 3.59, + "learning_rate": 1.2121236154620227e-05, + "loss": 0.0332, + "step": 27422 + }, + { + "epoch": 3.59, + "learning_rate": 1.2120717895263792e-05, + "loss": 0.0463, + "step": 27423 + }, + { + "epoch": 3.59, + "learning_rate": 1.212019962994295e-05, + "loss": 0.0327, + "step": 27424 + }, + { + "epoch": 3.59, + "learning_rate": 1.2119681358659164e-05, + "loss": 0.0392, + "step": 27425 + }, + { + "epoch": 3.59, + "learning_rate": 1.211916308141388e-05, + "loss": 0.0353, + "step": 27426 + }, + { + "epoch": 3.59, + "learning_rate": 1.2118644798208567e-05, + "loss": 0.0373, + "step": 27427 + }, + { + "epoch": 3.59, + "learning_rate": 1.2118126509044676e-05, + "loss": 0.0425, + "step": 27428 + }, + { + "epoch": 3.59, + "learning_rate": 1.211760821392367e-05, + "loss": 0.0391, + "step": 27429 + }, + { + "epoch": 3.59, + "learning_rate": 1.2117089912847002e-05, + "loss": 0.0406, + "step": 27430 + }, + { + "epoch": 3.59, + "learning_rate": 1.2116571605816132e-05, + "loss": 0.0395, + "step": 27431 + }, + { + "epoch": 3.59, + "learning_rate": 1.2116053292832518e-05, + "loss": 0.0378, + "step": 27432 + }, + { + "epoch": 3.59, + "learning_rate": 1.2115534973897614e-05, + "loss": 0.0348, + "step": 27433 + }, + { + "epoch": 3.59, + "learning_rate": 1.211501664901288e-05, + "loss": 0.0417, + "step": 27434 + }, + { + "epoch": 3.59, + "learning_rate": 1.211449831817978e-05, + "loss": 0.0374, + "step": 27435 + }, + { + "epoch": 3.59, + "learning_rate": 1.211397998139976e-05, + "loss": 0.0438, + "step": 27436 + }, + { + "epoch": 3.59, + "learning_rate": 1.2113461638674286e-05, + "loss": 0.0319, + "step": 27437 + }, + { + "epoch": 3.59, + "learning_rate": 1.2112943290004813e-05, + "loss": 0.0388, + "step": 27438 + }, + { + "epoch": 3.59, + "learning_rate": 1.2112424935392802e-05, + "loss": 0.044, + "step": 27439 + }, + { + "epoch": 3.59, + "learning_rate": 1.2111906574839707e-05, + "loss": 0.0405, + "step": 27440 + }, + { + "epoch": 3.59, + "learning_rate": 1.2111388208346986e-05, + "loss": 0.0425, + "step": 27441 + }, + { + "epoch": 3.59, + "learning_rate": 1.2110869835916099e-05, + "loss": 0.0354, + "step": 27442 + }, + { + "epoch": 3.59, + "learning_rate": 1.2110351457548504e-05, + "loss": 0.0319, + "step": 27443 + }, + { + "epoch": 3.59, + "learning_rate": 1.2109833073245655e-05, + "loss": 0.0388, + "step": 27444 + }, + { + "epoch": 3.6, + "learning_rate": 1.2109314683009014e-05, + "loss": 0.0413, + "step": 27445 + }, + { + "epoch": 3.6, + "learning_rate": 1.2108796286840041e-05, + "loss": 0.0393, + "step": 27446 + }, + { + "epoch": 3.6, + "learning_rate": 1.2108277884740187e-05, + "loss": 0.0364, + "step": 27447 + }, + { + "epoch": 3.6, + "learning_rate": 1.210775947671092e-05, + "loss": 0.0454, + "step": 27448 + }, + { + "epoch": 3.6, + "learning_rate": 1.2107241062753681e-05, + "loss": 0.0405, + "step": 27449 + }, + { + "epoch": 3.6, + "learning_rate": 1.2106722642869945e-05, + "loss": 0.038, + "step": 27450 + }, + { + "epoch": 3.6, + "learning_rate": 1.2106204217061163e-05, + "loss": 0.0346, + "step": 27451 + }, + { + "epoch": 3.6, + "learning_rate": 1.2105685785328795e-05, + "loss": 0.0376, + "step": 27452 + }, + { + "epoch": 3.6, + "learning_rate": 1.2105167347674297e-05, + "loss": 0.039, + "step": 27453 + }, + { + "epoch": 3.6, + "learning_rate": 1.2104648904099128e-05, + "loss": 0.0477, + "step": 27454 + }, + { + "epoch": 3.6, + "learning_rate": 1.2104130454604746e-05, + "loss": 0.0349, + "step": 27455 + }, + { + "epoch": 3.6, + "learning_rate": 1.210361199919261e-05, + "loss": 0.0381, + "step": 27456 + }, + { + "epoch": 3.6, + "learning_rate": 1.2103093537864177e-05, + "loss": 0.0368, + "step": 27457 + }, + { + "epoch": 3.6, + "learning_rate": 1.2102575070620903e-05, + "loss": 0.0368, + "step": 27458 + }, + { + "epoch": 3.6, + "learning_rate": 1.2102056597464251e-05, + "loss": 0.0327, + "step": 27459 + }, + { + "epoch": 3.6, + "learning_rate": 1.2101538118395677e-05, + "loss": 0.0336, + "step": 27460 + }, + { + "epoch": 3.6, + "learning_rate": 1.2101019633416637e-05, + "loss": 0.0418, + "step": 27461 + }, + { + "epoch": 3.6, + "learning_rate": 1.210050114252859e-05, + "loss": 0.0376, + "step": 27462 + }, + { + "epoch": 3.6, + "learning_rate": 1.2099982645733e-05, + "loss": 0.0382, + "step": 27463 + }, + { + "epoch": 3.6, + "learning_rate": 1.2099464143031318e-05, + "loss": 0.0334, + "step": 27464 + }, + { + "epoch": 3.6, + "learning_rate": 1.2098945634425003e-05, + "loss": 0.0311, + "step": 27465 + }, + { + "epoch": 3.6, + "learning_rate": 1.2098427119915516e-05, + "loss": 0.0383, + "step": 27466 + }, + { + "epoch": 3.6, + "learning_rate": 1.209790859950432e-05, + "loss": 0.0353, + "step": 27467 + }, + { + "epoch": 3.6, + "learning_rate": 1.2097390073192862e-05, + "loss": 0.0335, + "step": 27468 + }, + { + "epoch": 3.6, + "learning_rate": 1.2096871540982608e-05, + "loss": 0.0465, + "step": 27469 + }, + { + "epoch": 3.6, + "learning_rate": 1.2096353002875014e-05, + "loss": 0.0395, + "step": 27470 + }, + { + "epoch": 3.6, + "learning_rate": 1.2095834458871537e-05, + "loss": 0.0393, + "step": 27471 + }, + { + "epoch": 3.6, + "learning_rate": 1.2095315908973641e-05, + "loss": 0.0365, + "step": 27472 + }, + { + "epoch": 3.6, + "learning_rate": 1.209479735318278e-05, + "loss": 0.038, + "step": 27473 + }, + { + "epoch": 3.6, + "learning_rate": 1.209427879150041e-05, + "loss": 0.0365, + "step": 27474 + }, + { + "epoch": 3.6, + "learning_rate": 1.2093760223927998e-05, + "loss": 0.0383, + "step": 27475 + }, + { + "epoch": 3.6, + "learning_rate": 1.2093241650466989e-05, + "loss": 0.0443, + "step": 27476 + }, + { + "epoch": 3.6, + "learning_rate": 1.2092723071118855e-05, + "loss": 0.0464, + "step": 27477 + }, + { + "epoch": 3.6, + "learning_rate": 1.2092204485885046e-05, + "loss": 0.041, + "step": 27478 + }, + { + "epoch": 3.6, + "learning_rate": 1.2091685894767027e-05, + "loss": 0.032, + "step": 27479 + }, + { + "epoch": 3.6, + "learning_rate": 1.2091167297766249e-05, + "loss": 0.0329, + "step": 27480 + }, + { + "epoch": 3.6, + "learning_rate": 1.2090648694884175e-05, + "loss": 0.0394, + "step": 27481 + }, + { + "epoch": 3.6, + "learning_rate": 1.2090130086122263e-05, + "loss": 0.044, + "step": 27482 + }, + { + "epoch": 3.6, + "learning_rate": 1.2089611471481973e-05, + "loss": 0.0517, + "step": 27483 + }, + { + "epoch": 3.6, + "learning_rate": 1.2089092850964761e-05, + "loss": 0.0415, + "step": 27484 + }, + { + "epoch": 3.6, + "learning_rate": 1.2088574224572085e-05, + "loss": 0.0378, + "step": 27485 + }, + { + "epoch": 3.6, + "learning_rate": 1.208805559230541e-05, + "loss": 0.0439, + "step": 27486 + }, + { + "epoch": 3.6, + "learning_rate": 1.2087536954166184e-05, + "loss": 0.0385, + "step": 27487 + }, + { + "epoch": 3.6, + "learning_rate": 1.2087018310155873e-05, + "loss": 0.04, + "step": 27488 + }, + { + "epoch": 3.6, + "learning_rate": 1.2086499660275937e-05, + "loss": 0.0368, + "step": 27489 + }, + { + "epoch": 3.6, + "learning_rate": 1.2085981004527832e-05, + "loss": 0.0387, + "step": 27490 + }, + { + "epoch": 3.6, + "learning_rate": 1.2085462342913012e-05, + "loss": 0.0353, + "step": 27491 + }, + { + "epoch": 3.6, + "learning_rate": 1.2084943675432945e-05, + "loss": 0.039, + "step": 27492 + }, + { + "epoch": 3.6, + "learning_rate": 1.208442500208908e-05, + "loss": 0.0322, + "step": 27493 + }, + { + "epoch": 3.6, + "learning_rate": 1.2083906322882886e-05, + "loss": 0.0382, + "step": 27494 + }, + { + "epoch": 3.6, + "learning_rate": 1.208338763781581e-05, + "loss": 0.0315, + "step": 27495 + }, + { + "epoch": 3.6, + "learning_rate": 1.2082868946889326e-05, + "loss": 0.0328, + "step": 27496 + }, + { + "epoch": 3.6, + "learning_rate": 1.2082350250104877e-05, + "loss": 0.0373, + "step": 27497 + }, + { + "epoch": 3.6, + "learning_rate": 1.2081831547463931e-05, + "loss": 0.0429, + "step": 27498 + }, + { + "epoch": 3.6, + "learning_rate": 1.2081312838967942e-05, + "loss": 0.0405, + "step": 27499 + }, + { + "epoch": 3.6, + "learning_rate": 1.2080794124618375e-05, + "loss": 0.0361, + "step": 27500 + }, + { + "epoch": 3.6, + "learning_rate": 1.2080275404416685e-05, + "loss": 0.0321, + "step": 27501 + }, + { + "epoch": 3.6, + "learning_rate": 1.2079756678364329e-05, + "loss": 0.0396, + "step": 27502 + }, + { + "epoch": 3.6, + "learning_rate": 1.2079237946462771e-05, + "loss": 0.0351, + "step": 27503 + }, + { + "epoch": 3.6, + "learning_rate": 1.2078719208713467e-05, + "loss": 0.0374, + "step": 27504 + }, + { + "epoch": 3.6, + "learning_rate": 1.2078200465117872e-05, + "loss": 0.0435, + "step": 27505 + }, + { + "epoch": 3.6, + "learning_rate": 1.2077681715677452e-05, + "loss": 0.043, + "step": 27506 + }, + { + "epoch": 3.6, + "learning_rate": 1.2077162960393665e-05, + "loss": 0.0401, + "step": 27507 + }, + { + "epoch": 3.6, + "learning_rate": 1.2076644199267964e-05, + "loss": 0.0352, + "step": 27508 + }, + { + "epoch": 3.6, + "learning_rate": 1.207612543230181e-05, + "loss": 0.0301, + "step": 27509 + }, + { + "epoch": 3.6, + "learning_rate": 1.2075606659496668e-05, + "loss": 0.0337, + "step": 27510 + }, + { + "epoch": 3.6, + "learning_rate": 1.2075087880853992e-05, + "loss": 0.0414, + "step": 27511 + }, + { + "epoch": 3.6, + "learning_rate": 1.207456909637524e-05, + "loss": 0.0307, + "step": 27512 + }, + { + "epoch": 3.6, + "learning_rate": 1.2074050306061878e-05, + "loss": 0.0346, + "step": 27513 + }, + { + "epoch": 3.6, + "learning_rate": 1.2073531509915352e-05, + "loss": 0.0405, + "step": 27514 + }, + { + "epoch": 3.6, + "learning_rate": 1.2073012707937134e-05, + "loss": 0.0451, + "step": 27515 + }, + { + "epoch": 3.6, + "learning_rate": 1.2072493900128677e-05, + "loss": 0.038, + "step": 27516 + }, + { + "epoch": 3.6, + "learning_rate": 1.2071975086491443e-05, + "loss": 0.0422, + "step": 27517 + }, + { + "epoch": 3.6, + "learning_rate": 1.2071456267026888e-05, + "loss": 0.038, + "step": 27518 + }, + { + "epoch": 3.6, + "learning_rate": 1.2070937441736474e-05, + "loss": 0.0435, + "step": 27519 + }, + { + "epoch": 3.6, + "learning_rate": 1.2070418610621655e-05, + "loss": 0.0337, + "step": 27520 + }, + { + "epoch": 3.61, + "learning_rate": 1.2069899773683897e-05, + "loss": 0.0394, + "step": 27521 + }, + { + "epoch": 3.61, + "learning_rate": 1.2069380930924654e-05, + "loss": 0.0438, + "step": 27522 + }, + { + "epoch": 3.61, + "learning_rate": 1.2068862082345389e-05, + "loss": 0.0381, + "step": 27523 + }, + { + "epoch": 3.61, + "learning_rate": 1.2068343227947561e-05, + "loss": 0.0389, + "step": 27524 + }, + { + "epoch": 3.61, + "learning_rate": 1.2067824367732626e-05, + "loss": 0.0332, + "step": 27525 + }, + { + "epoch": 3.61, + "learning_rate": 1.2067305501702042e-05, + "loss": 0.0348, + "step": 27526 + }, + { + "epoch": 3.61, + "learning_rate": 1.2066786629857275e-05, + "loss": 0.0431, + "step": 27527 + }, + { + "epoch": 3.61, + "learning_rate": 1.206626775219978e-05, + "loss": 0.0356, + "step": 27528 + }, + { + "epoch": 3.61, + "learning_rate": 1.2065748868731018e-05, + "loss": 0.037, + "step": 27529 + }, + { + "epoch": 3.61, + "learning_rate": 1.2065229979452449e-05, + "loss": 0.0377, + "step": 27530 + }, + { + "epoch": 3.61, + "learning_rate": 1.2064711084365523e-05, + "loss": 0.031, + "step": 27531 + }, + { + "epoch": 3.61, + "learning_rate": 1.2064192183471714e-05, + "loss": 0.0354, + "step": 27532 + }, + { + "epoch": 3.61, + "learning_rate": 1.2063673276772475e-05, + "loss": 0.0343, + "step": 27533 + }, + { + "epoch": 3.61, + "learning_rate": 1.2063154364269262e-05, + "loss": 0.0456, + "step": 27534 + }, + { + "epoch": 3.61, + "learning_rate": 1.2062635445963537e-05, + "loss": 0.0403, + "step": 27535 + }, + { + "epoch": 3.61, + "learning_rate": 1.206211652185676e-05, + "loss": 0.0397, + "step": 27536 + }, + { + "epoch": 3.61, + "learning_rate": 1.206159759195039e-05, + "loss": 0.0346, + "step": 27537 + }, + { + "epoch": 3.61, + "learning_rate": 1.2061078656245888e-05, + "loss": 0.0369, + "step": 27538 + }, + { + "epoch": 3.61, + "learning_rate": 1.2060559714744711e-05, + "loss": 0.0315, + "step": 27539 + }, + { + "epoch": 3.61, + "learning_rate": 1.206004076744832e-05, + "loss": 0.0393, + "step": 27540 + }, + { + "epoch": 3.61, + "learning_rate": 1.2059521814358173e-05, + "loss": 0.0328, + "step": 27541 + }, + { + "epoch": 3.61, + "learning_rate": 1.205900285547573e-05, + "loss": 0.0333, + "step": 27542 + }, + { + "epoch": 3.61, + "learning_rate": 1.205848389080245e-05, + "loss": 0.0387, + "step": 27543 + }, + { + "epoch": 3.61, + "learning_rate": 1.20579649203398e-05, + "loss": 0.0377, + "step": 27544 + }, + { + "epoch": 3.61, + "learning_rate": 1.2057445944089227e-05, + "loss": 0.0288, + "step": 27545 + }, + { + "epoch": 3.61, + "learning_rate": 1.2056926962052197e-05, + "loss": 0.0356, + "step": 27546 + }, + { + "epoch": 3.61, + "learning_rate": 1.2056407974230173e-05, + "loss": 0.0363, + "step": 27547 + }, + { + "epoch": 3.61, + "learning_rate": 1.2055888980624608e-05, + "loss": 0.0437, + "step": 27548 + }, + { + "epoch": 3.61, + "learning_rate": 1.2055369981236968e-05, + "loss": 0.0427, + "step": 27549 + }, + { + "epoch": 3.61, + "learning_rate": 1.2054850976068707e-05, + "loss": 0.048, + "step": 27550 + }, + { + "epoch": 3.61, + "learning_rate": 1.2054331965121289e-05, + "loss": 0.0386, + "step": 27551 + }, + { + "epoch": 3.61, + "learning_rate": 1.2053812948396171e-05, + "loss": 0.037, + "step": 27552 + }, + { + "epoch": 3.61, + "learning_rate": 1.2053293925894811e-05, + "loss": 0.0336, + "step": 27553 + }, + { + "epoch": 3.61, + "learning_rate": 1.2052774897618673e-05, + "loss": 0.0384, + "step": 27554 + }, + { + "epoch": 3.61, + "learning_rate": 1.2052255863569216e-05, + "loss": 0.0424, + "step": 27555 + }, + { + "epoch": 3.61, + "learning_rate": 1.2051736823747897e-05, + "loss": 0.0359, + "step": 27556 + }, + { + "epoch": 3.61, + "learning_rate": 1.2051217778156179e-05, + "loss": 0.0383, + "step": 27557 + }, + { + "epoch": 3.61, + "learning_rate": 1.2050698726795518e-05, + "loss": 0.0377, + "step": 27558 + }, + { + "epoch": 3.61, + "learning_rate": 1.2050179669667378e-05, + "loss": 0.0368, + "step": 27559 + }, + { + "epoch": 3.61, + "learning_rate": 1.2049660606773215e-05, + "loss": 0.0302, + "step": 27560 + }, + { + "epoch": 3.61, + "learning_rate": 1.2049141538114495e-05, + "loss": 0.0378, + "step": 27561 + }, + { + "epoch": 3.61, + "learning_rate": 1.204862246369267e-05, + "loss": 0.0379, + "step": 27562 + }, + { + "epoch": 3.61, + "learning_rate": 1.2048103383509202e-05, + "loss": 0.0383, + "step": 27563 + }, + { + "epoch": 3.61, + "learning_rate": 1.2047584297565553e-05, + "loss": 0.039, + "step": 27564 + }, + { + "epoch": 3.61, + "learning_rate": 1.2047065205863183e-05, + "loss": 0.0354, + "step": 27565 + }, + { + "epoch": 3.61, + "learning_rate": 1.2046546108403554e-05, + "loss": 0.0386, + "step": 27566 + }, + { + "epoch": 3.61, + "learning_rate": 1.204602700518812e-05, + "loss": 0.0403, + "step": 27567 + }, + { + "epoch": 3.61, + "learning_rate": 1.2045507896218344e-05, + "loss": 0.0382, + "step": 27568 + }, + { + "epoch": 3.61, + "learning_rate": 1.2044988781495688e-05, + "loss": 0.0305, + "step": 27569 + }, + { + "epoch": 3.61, + "learning_rate": 1.2044469661021607e-05, + "loss": 0.0349, + "step": 27570 + }, + { + "epoch": 3.61, + "learning_rate": 1.2043950534797563e-05, + "loss": 0.031, + "step": 27571 + }, + { + "epoch": 3.61, + "learning_rate": 1.2043431402825023e-05, + "loss": 0.045, + "step": 27572 + }, + { + "epoch": 3.61, + "learning_rate": 1.2042912265105438e-05, + "loss": 0.0394, + "step": 27573 + }, + { + "epoch": 3.61, + "learning_rate": 1.204239312164027e-05, + "loss": 0.037, + "step": 27574 + }, + { + "epoch": 3.61, + "learning_rate": 1.2041873972430978e-05, + "loss": 0.0403, + "step": 27575 + }, + { + "epoch": 3.61, + "learning_rate": 1.204135481747903e-05, + "loss": 0.0385, + "step": 27576 + }, + { + "epoch": 3.61, + "learning_rate": 1.2040835656785874e-05, + "loss": 0.0418, + "step": 27577 + }, + { + "epoch": 3.61, + "learning_rate": 1.2040316490352982e-05, + "loss": 0.0448, + "step": 27578 + }, + { + "epoch": 3.61, + "learning_rate": 1.2039797318181806e-05, + "loss": 0.0388, + "step": 27579 + }, + { + "epoch": 3.61, + "learning_rate": 1.2039278140273808e-05, + "loss": 0.0418, + "step": 27580 + }, + { + "epoch": 3.61, + "learning_rate": 1.2038758956630448e-05, + "loss": 0.031, + "step": 27581 + }, + { + "epoch": 3.61, + "learning_rate": 1.203823976725319e-05, + "loss": 0.0358, + "step": 27582 + }, + { + "epoch": 3.61, + "learning_rate": 1.2037720572143488e-05, + "loss": 0.0372, + "step": 27583 + }, + { + "epoch": 3.61, + "learning_rate": 1.203720137130281e-05, + "loss": 0.0319, + "step": 27584 + }, + { + "epoch": 3.61, + "learning_rate": 1.2036682164732604e-05, + "loss": 0.0376, + "step": 27585 + }, + { + "epoch": 3.61, + "learning_rate": 1.2036162952434345e-05, + "loss": 0.041, + "step": 27586 + }, + { + "epoch": 3.61, + "learning_rate": 1.2035643734409483e-05, + "loss": 0.0349, + "step": 27587 + }, + { + "epoch": 3.61, + "learning_rate": 1.2035124510659482e-05, + "loss": 0.0444, + "step": 27588 + }, + { + "epoch": 3.61, + "learning_rate": 1.2034605281185803e-05, + "loss": 0.036, + "step": 27589 + }, + { + "epoch": 3.61, + "learning_rate": 1.2034086045989903e-05, + "loss": 0.0355, + "step": 27590 + }, + { + "epoch": 3.61, + "learning_rate": 1.2033566805073244e-05, + "loss": 0.041, + "step": 27591 + }, + { + "epoch": 3.61, + "learning_rate": 1.2033047558437287e-05, + "loss": 0.0326, + "step": 27592 + }, + { + "epoch": 3.61, + "learning_rate": 1.2032528306083496e-05, + "loss": 0.0338, + "step": 27593 + }, + { + "epoch": 3.61, + "learning_rate": 1.2032009048013323e-05, + "loss": 0.0455, + "step": 27594 + }, + { + "epoch": 3.61, + "learning_rate": 1.2031489784228234e-05, + "loss": 0.0344, + "step": 27595 + }, + { + "epoch": 3.61, + "learning_rate": 1.2030970514729686e-05, + "loss": 0.0383, + "step": 27596 + }, + { + "epoch": 3.62, + "learning_rate": 1.2030451239519146e-05, + "loss": 0.044, + "step": 27597 + }, + { + "epoch": 3.62, + "learning_rate": 1.2029931958598067e-05, + "loss": 0.0409, + "step": 27598 + }, + { + "epoch": 3.62, + "learning_rate": 1.2029412671967915e-05, + "loss": 0.0342, + "step": 27599 + }, + { + "epoch": 3.62, + "learning_rate": 1.2028893379630145e-05, + "loss": 0.0331, + "step": 27600 + }, + { + "epoch": 3.62, + "learning_rate": 1.2028374081586222e-05, + "loss": 0.0473, + "step": 27601 + }, + { + "epoch": 3.62, + "learning_rate": 1.2027854777837602e-05, + "loss": 0.0371, + "step": 27602 + }, + { + "epoch": 3.62, + "learning_rate": 1.2027335468385752e-05, + "loss": 0.0388, + "step": 27603 + }, + { + "epoch": 3.62, + "learning_rate": 1.2026816153232126e-05, + "loss": 0.0367, + "step": 27604 + }, + { + "epoch": 3.62, + "learning_rate": 1.2026296832378192e-05, + "loss": 0.0395, + "step": 27605 + }, + { + "epoch": 3.62, + "learning_rate": 1.2025777505825402e-05, + "loss": 0.0395, + "step": 27606 + }, + { + "epoch": 3.62, + "learning_rate": 1.2025258173575223e-05, + "loss": 0.0325, + "step": 27607 + }, + { + "epoch": 3.62, + "learning_rate": 1.202473883562911e-05, + "loss": 0.0465, + "step": 27608 + }, + { + "epoch": 3.62, + "learning_rate": 1.2024219491988531e-05, + "loss": 0.0324, + "step": 27609 + }, + { + "epoch": 3.62, + "learning_rate": 1.2023700142654939e-05, + "loss": 0.0348, + "step": 27610 + }, + { + "epoch": 3.62, + "learning_rate": 1.20231807876298e-05, + "loss": 0.0438, + "step": 27611 + }, + { + "epoch": 3.62, + "learning_rate": 1.2022661426914574e-05, + "loss": 0.0388, + "step": 27612 + }, + { + "epoch": 3.62, + "learning_rate": 1.2022142060510715e-05, + "loss": 0.0321, + "step": 27613 + }, + { + "epoch": 3.62, + "learning_rate": 1.2021622688419696e-05, + "loss": 0.0483, + "step": 27614 + }, + { + "epoch": 3.62, + "learning_rate": 1.2021103310642968e-05, + "loss": 0.0444, + "step": 27615 + }, + { + "epoch": 3.62, + "learning_rate": 1.2020583927181995e-05, + "loss": 0.0413, + "step": 27616 + }, + { + "epoch": 3.62, + "learning_rate": 1.2020064538038237e-05, + "loss": 0.0431, + "step": 27617 + }, + { + "epoch": 3.62, + "learning_rate": 1.2019545143213153e-05, + "loss": 0.0307, + "step": 27618 + }, + { + "epoch": 3.62, + "learning_rate": 1.2019025742708209e-05, + "loss": 0.0329, + "step": 27619 + }, + { + "epoch": 3.62, + "learning_rate": 1.201850633652486e-05, + "loss": 0.0357, + "step": 27620 + }, + { + "epoch": 3.62, + "learning_rate": 1.2017986924664571e-05, + "loss": 0.041, + "step": 27621 + }, + { + "epoch": 3.62, + "learning_rate": 1.2017467507128801e-05, + "loss": 0.0374, + "step": 27622 + }, + { + "epoch": 3.62, + "learning_rate": 1.2016948083919011e-05, + "loss": 0.0359, + "step": 27623 + }, + { + "epoch": 3.62, + "learning_rate": 1.2016428655036662e-05, + "loss": 0.0416, + "step": 27624 + }, + { + "epoch": 3.62, + "learning_rate": 1.2015909220483212e-05, + "loss": 0.0418, + "step": 27625 + }, + { + "epoch": 3.62, + "learning_rate": 1.2015389780260132e-05, + "loss": 0.0374, + "step": 27626 + }, + { + "epoch": 3.62, + "learning_rate": 1.2014870334368869e-05, + "loss": 0.0402, + "step": 27627 + }, + { + "epoch": 3.62, + "learning_rate": 1.2014350882810892e-05, + "loss": 0.0415, + "step": 27628 + }, + { + "epoch": 3.62, + "learning_rate": 1.201383142558766e-05, + "loss": 0.0339, + "step": 27629 + }, + { + "epoch": 3.62, + "learning_rate": 1.2013311962700635e-05, + "loss": 0.0375, + "step": 27630 + }, + { + "epoch": 3.62, + "learning_rate": 1.2012792494151277e-05, + "loss": 0.0346, + "step": 27631 + }, + { + "epoch": 3.62, + "learning_rate": 1.2012273019941047e-05, + "loss": 0.0425, + "step": 27632 + }, + { + "epoch": 3.62, + "learning_rate": 1.2011753540071411e-05, + "loss": 0.038, + "step": 27633 + }, + { + "epoch": 3.62, + "learning_rate": 1.2011234054543822e-05, + "loss": 0.0389, + "step": 27634 + }, + { + "epoch": 3.62, + "learning_rate": 1.2010714563359743e-05, + "loss": 0.033, + "step": 27635 + }, + { + "epoch": 3.62, + "learning_rate": 1.2010195066520635e-05, + "loss": 0.038, + "step": 27636 + }, + { + "epoch": 3.62, + "learning_rate": 1.2009675564027963e-05, + "loss": 0.037, + "step": 27637 + }, + { + "epoch": 3.62, + "learning_rate": 1.2009156055883185e-05, + "loss": 0.0415, + "step": 27638 + }, + { + "epoch": 3.62, + "learning_rate": 1.2008636542087763e-05, + "loss": 0.0388, + "step": 27639 + }, + { + "epoch": 3.62, + "learning_rate": 1.2008117022643155e-05, + "loss": 0.0422, + "step": 27640 + }, + { + "epoch": 3.62, + "learning_rate": 1.200759749755083e-05, + "loss": 0.0386, + "step": 27641 + }, + { + "epoch": 3.62, + "learning_rate": 1.200707796681224e-05, + "loss": 0.0381, + "step": 27642 + }, + { + "epoch": 3.62, + "learning_rate": 1.2006558430428852e-05, + "loss": 0.04, + "step": 27643 + }, + { + "epoch": 3.62, + "learning_rate": 1.2006038888402125e-05, + "loss": 0.0383, + "step": 27644 + }, + { + "epoch": 3.62, + "learning_rate": 1.2005519340733519e-05, + "loss": 0.0394, + "step": 27645 + }, + { + "epoch": 3.62, + "learning_rate": 1.2004999787424496e-05, + "loss": 0.0376, + "step": 27646 + }, + { + "epoch": 3.62, + "learning_rate": 1.2004480228476521e-05, + "loss": 0.0348, + "step": 27647 + }, + { + "epoch": 3.62, + "learning_rate": 1.200396066389105e-05, + "loss": 0.0344, + "step": 27648 + }, + { + "epoch": 3.62, + "learning_rate": 1.2003441093669545e-05, + "loss": 0.0311, + "step": 27649 + }, + { + "epoch": 3.62, + "learning_rate": 1.2002921517813473e-05, + "loss": 0.0389, + "step": 27650 + }, + { + "epoch": 3.62, + "learning_rate": 1.2002401936324289e-05, + "loss": 0.036, + "step": 27651 + }, + { + "epoch": 3.62, + "learning_rate": 1.2001882349203453e-05, + "loss": 0.0374, + "step": 27652 + }, + { + "epoch": 3.62, + "learning_rate": 1.2001362756452432e-05, + "loss": 0.036, + "step": 27653 + }, + { + "epoch": 3.62, + "learning_rate": 1.2000843158072685e-05, + "loss": 0.0433, + "step": 27654 + }, + { + "epoch": 3.62, + "learning_rate": 1.2000323554065674e-05, + "loss": 0.0402, + "step": 27655 + }, + { + "epoch": 3.62, + "learning_rate": 1.1999803944432857e-05, + "loss": 0.0429, + "step": 27656 + }, + { + "epoch": 3.62, + "learning_rate": 1.19992843291757e-05, + "loss": 0.0377, + "step": 27657 + }, + { + "epoch": 3.62, + "learning_rate": 1.1998764708295663e-05, + "loss": 0.0386, + "step": 27658 + }, + { + "epoch": 3.62, + "learning_rate": 1.1998245081794204e-05, + "loss": 0.037, + "step": 27659 + }, + { + "epoch": 3.62, + "learning_rate": 1.199772544967279e-05, + "loss": 0.0397, + "step": 27660 + }, + { + "epoch": 3.62, + "learning_rate": 1.1997205811932879e-05, + "loss": 0.0365, + "step": 27661 + }, + { + "epoch": 3.62, + "learning_rate": 1.1996686168575931e-05, + "loss": 0.0345, + "step": 27662 + }, + { + "epoch": 3.62, + "learning_rate": 1.1996166519603409e-05, + "loss": 0.0372, + "step": 27663 + }, + { + "epoch": 3.62, + "learning_rate": 1.199564686501678e-05, + "loss": 0.0393, + "step": 27664 + }, + { + "epoch": 3.62, + "learning_rate": 1.1995127204817497e-05, + "loss": 0.0454, + "step": 27665 + }, + { + "epoch": 3.62, + "learning_rate": 1.1994607539007027e-05, + "loss": 0.0376, + "step": 27666 + }, + { + "epoch": 3.62, + "learning_rate": 1.1994087867586824e-05, + "loss": 0.0463, + "step": 27667 + }, + { + "epoch": 3.62, + "learning_rate": 1.199356819055836e-05, + "loss": 0.0402, + "step": 27668 + }, + { + "epoch": 3.62, + "learning_rate": 1.1993048507923088e-05, + "loss": 0.0335, + "step": 27669 + }, + { + "epoch": 3.62, + "learning_rate": 1.1992528819682478e-05, + "loss": 0.0386, + "step": 27670 + }, + { + "epoch": 3.62, + "learning_rate": 1.1992009125837982e-05, + "loss": 0.0369, + "step": 27671 + }, + { + "epoch": 3.62, + "learning_rate": 1.199148942639107e-05, + "loss": 0.0363, + "step": 27672 + }, + { + "epoch": 3.62, + "learning_rate": 1.1990969721343198e-05, + "loss": 0.0436, + "step": 27673 + }, + { + "epoch": 3.63, + "learning_rate": 1.1990450010695829e-05, + "loss": 0.0348, + "step": 27674 + }, + { + "epoch": 3.63, + "learning_rate": 1.1989930294450429e-05, + "loss": 0.04, + "step": 27675 + }, + { + "epoch": 3.63, + "learning_rate": 1.1989410572608453e-05, + "loss": 0.0357, + "step": 27676 + }, + { + "epoch": 3.63, + "learning_rate": 1.1988890845171368e-05, + "loss": 0.0311, + "step": 27677 + }, + { + "epoch": 3.63, + "learning_rate": 1.198837111214063e-05, + "loss": 0.0412, + "step": 27678 + }, + { + "epoch": 3.63, + "learning_rate": 1.1987851373517708e-05, + "loss": 0.0364, + "step": 27679 + }, + { + "epoch": 3.63, + "learning_rate": 1.1987331629304056e-05, + "loss": 0.0396, + "step": 27680 + }, + { + "epoch": 3.63, + "learning_rate": 1.1986811879501143e-05, + "loss": 0.0412, + "step": 27681 + }, + { + "epoch": 3.63, + "learning_rate": 1.1986292124110426e-05, + "loss": 0.0437, + "step": 27682 + }, + { + "epoch": 3.63, + "learning_rate": 1.1985772363133368e-05, + "loss": 0.0435, + "step": 27683 + }, + { + "epoch": 3.63, + "learning_rate": 1.1985252596571428e-05, + "loss": 0.0331, + "step": 27684 + }, + { + "epoch": 3.63, + "learning_rate": 1.1984732824426075e-05, + "loss": 0.0308, + "step": 27685 + }, + { + "epoch": 3.63, + "learning_rate": 1.1984213046698765e-05, + "loss": 0.0387, + "step": 27686 + }, + { + "epoch": 3.63, + "learning_rate": 1.1983693263390966e-05, + "loss": 0.0439, + "step": 27687 + }, + { + "epoch": 3.63, + "learning_rate": 1.1983173474504129e-05, + "loss": 0.0414, + "step": 27688 + }, + { + "epoch": 3.63, + "learning_rate": 1.1982653680039723e-05, + "loss": 0.0474, + "step": 27689 + }, + { + "epoch": 3.63, + "learning_rate": 1.1982133879999212e-05, + "loss": 0.0385, + "step": 27690 + }, + { + "epoch": 3.63, + "learning_rate": 1.1981614074384054e-05, + "loss": 0.0411, + "step": 27691 + }, + { + "epoch": 3.63, + "learning_rate": 1.1981094263195712e-05, + "loss": 0.0352, + "step": 27692 + }, + { + "epoch": 3.63, + "learning_rate": 1.1980574446435646e-05, + "loss": 0.04, + "step": 27693 + }, + { + "epoch": 3.63, + "learning_rate": 1.1980054624105322e-05, + "loss": 0.04, + "step": 27694 + }, + { + "epoch": 3.63, + "learning_rate": 1.1979534796206197e-05, + "loss": 0.0374, + "step": 27695 + }, + { + "epoch": 3.63, + "learning_rate": 1.1979014962739738e-05, + "loss": 0.0417, + "step": 27696 + }, + { + "epoch": 3.63, + "learning_rate": 1.1978495123707405e-05, + "loss": 0.0373, + "step": 27697 + }, + { + "epoch": 3.63, + "learning_rate": 1.197797527911066e-05, + "loss": 0.0344, + "step": 27698 + }, + { + "epoch": 3.63, + "learning_rate": 1.1977455428950966e-05, + "loss": 0.042, + "step": 27699 + }, + { + "epoch": 3.63, + "learning_rate": 1.197693557322978e-05, + "loss": 0.0437, + "step": 27700 + }, + { + "epoch": 3.63, + "learning_rate": 1.197641571194857e-05, + "loss": 0.0405, + "step": 27701 + }, + { + "epoch": 3.63, + "learning_rate": 1.1975895845108798e-05, + "loss": 0.0427, + "step": 27702 + }, + { + "epoch": 3.63, + "learning_rate": 1.1975375972711922e-05, + "loss": 0.0378, + "step": 27703 + }, + { + "epoch": 3.63, + "learning_rate": 1.1974856094759408e-05, + "loss": 0.0436, + "step": 27704 + }, + { + "epoch": 3.63, + "learning_rate": 1.1974336211252716e-05, + "loss": 0.0362, + "step": 27705 + }, + { + "epoch": 3.63, + "learning_rate": 1.1973816322193306e-05, + "loss": 0.0379, + "step": 27706 + }, + { + "epoch": 3.63, + "learning_rate": 1.1973296427582644e-05, + "loss": 0.0381, + "step": 27707 + }, + { + "epoch": 3.63, + "learning_rate": 1.1972776527422193e-05, + "loss": 0.0414, + "step": 27708 + }, + { + "epoch": 3.63, + "learning_rate": 1.197225662171341e-05, + "loss": 0.0391, + "step": 27709 + }, + { + "epoch": 3.63, + "learning_rate": 1.1971736710457764e-05, + "loss": 0.0399, + "step": 27710 + }, + { + "epoch": 3.63, + "learning_rate": 1.1971216793656707e-05, + "loss": 0.0356, + "step": 27711 + }, + { + "epoch": 3.63, + "learning_rate": 1.1970696871311716e-05, + "loss": 0.0417, + "step": 27712 + }, + { + "epoch": 3.63, + "learning_rate": 1.1970176943424238e-05, + "loss": 0.0383, + "step": 27713 + }, + { + "epoch": 3.63, + "learning_rate": 1.1969657009995745e-05, + "loss": 0.0411, + "step": 27714 + }, + { + "epoch": 3.63, + "learning_rate": 1.1969137071027699e-05, + "loss": 0.0403, + "step": 27715 + }, + { + "epoch": 3.63, + "learning_rate": 1.1968617126521556e-05, + "loss": 0.041, + "step": 27716 + }, + { + "epoch": 3.63, + "learning_rate": 1.1968097176478784e-05, + "loss": 0.0361, + "step": 27717 + }, + { + "epoch": 3.63, + "learning_rate": 1.1967577220900842e-05, + "loss": 0.0366, + "step": 27718 + }, + { + "epoch": 3.63, + "learning_rate": 1.1967057259789196e-05, + "loss": 0.0393, + "step": 27719 + }, + { + "epoch": 3.63, + "learning_rate": 1.1966537293145305e-05, + "loss": 0.0391, + "step": 27720 + }, + { + "epoch": 3.63, + "learning_rate": 1.1966017320970635e-05, + "loss": 0.0462, + "step": 27721 + }, + { + "epoch": 3.63, + "learning_rate": 1.196549734326664e-05, + "loss": 0.0364, + "step": 27722 + }, + { + "epoch": 3.63, + "learning_rate": 1.1964977360034795e-05, + "loss": 0.0456, + "step": 27723 + }, + { + "epoch": 3.63, + "learning_rate": 1.1964457371276553e-05, + "loss": 0.0409, + "step": 27724 + }, + { + "epoch": 3.63, + "learning_rate": 1.1963937376993379e-05, + "loss": 0.0348, + "step": 27725 + }, + { + "epoch": 3.63, + "learning_rate": 1.1963417377186737e-05, + "loss": 0.0391, + "step": 27726 + }, + { + "epoch": 3.63, + "learning_rate": 1.1962897371858086e-05, + "loss": 0.0349, + "step": 27727 + }, + { + "epoch": 3.63, + "learning_rate": 1.1962377361008893e-05, + "loss": 0.04, + "step": 27728 + }, + { + "epoch": 3.63, + "learning_rate": 1.1961857344640617e-05, + "loss": 0.0444, + "step": 27729 + }, + { + "epoch": 3.63, + "learning_rate": 1.196133732275472e-05, + "loss": 0.0351, + "step": 27730 + }, + { + "epoch": 3.63, + "learning_rate": 1.1960817295352671e-05, + "loss": 0.0429, + "step": 27731 + }, + { + "epoch": 3.63, + "learning_rate": 1.1960297262435923e-05, + "loss": 0.0365, + "step": 27732 + }, + { + "epoch": 3.63, + "learning_rate": 1.1959777224005946e-05, + "loss": 0.0383, + "step": 27733 + }, + { + "epoch": 3.63, + "learning_rate": 1.1959257180064197e-05, + "loss": 0.0412, + "step": 27734 + }, + { + "epoch": 3.63, + "learning_rate": 1.1958737130612146e-05, + "loss": 0.0409, + "step": 27735 + }, + { + "epoch": 3.63, + "learning_rate": 1.1958217075651248e-05, + "loss": 0.0354, + "step": 27736 + }, + { + "epoch": 3.63, + "learning_rate": 1.1957697015182968e-05, + "loss": 0.0371, + "step": 27737 + }, + { + "epoch": 3.63, + "learning_rate": 1.195717694920877e-05, + "loss": 0.0441, + "step": 27738 + }, + { + "epoch": 3.63, + "learning_rate": 1.1956656877730117e-05, + "loss": 0.0363, + "step": 27739 + }, + { + "epoch": 3.63, + "learning_rate": 1.1956136800748472e-05, + "loss": 0.0363, + "step": 27740 + }, + { + "epoch": 3.63, + "learning_rate": 1.1955616718265294e-05, + "loss": 0.0378, + "step": 27741 + }, + { + "epoch": 3.63, + "learning_rate": 1.195509663028205e-05, + "loss": 0.0392, + "step": 27742 + }, + { + "epoch": 3.63, + "learning_rate": 1.19545765368002e-05, + "loss": 0.0397, + "step": 27743 + }, + { + "epoch": 3.63, + "learning_rate": 1.1954056437821206e-05, + "loss": 0.0383, + "step": 27744 + }, + { + "epoch": 3.63, + "learning_rate": 1.1953536333346533e-05, + "loss": 0.035, + "step": 27745 + }, + { + "epoch": 3.63, + "learning_rate": 1.1953016223377648e-05, + "loss": 0.0381, + "step": 27746 + }, + { + "epoch": 3.63, + "learning_rate": 1.1952496107916002e-05, + "loss": 0.0352, + "step": 27747 + }, + { + "epoch": 3.63, + "learning_rate": 1.1951975986963071e-05, + "loss": 0.0375, + "step": 27748 + }, + { + "epoch": 3.63, + "learning_rate": 1.1951455860520305e-05, + "loss": 0.0364, + "step": 27749 + }, + { + "epoch": 3.64, + "learning_rate": 1.1950935728589178e-05, + "loss": 0.0411, + "step": 27750 + }, + { + "epoch": 3.64, + "learning_rate": 1.1950415591171146e-05, + "loss": 0.0368, + "step": 27751 + }, + { + "epoch": 3.64, + "learning_rate": 1.1949895448267674e-05, + "loss": 0.0411, + "step": 27752 + }, + { + "epoch": 3.64, + "learning_rate": 1.1949375299880225e-05, + "loss": 0.046, + "step": 27753 + }, + { + "epoch": 3.64, + "learning_rate": 1.1948855146010262e-05, + "loss": 0.0405, + "step": 27754 + }, + { + "epoch": 3.64, + "learning_rate": 1.1948334986659248e-05, + "loss": 0.0377, + "step": 27755 + }, + { + "epoch": 3.64, + "learning_rate": 1.1947814821828648e-05, + "loss": 0.0365, + "step": 27756 + }, + { + "epoch": 3.64, + "learning_rate": 1.1947294651519918e-05, + "loss": 0.0459, + "step": 27757 + }, + { + "epoch": 3.64, + "learning_rate": 1.1946774475734526e-05, + "loss": 0.0359, + "step": 27758 + }, + { + "epoch": 3.64, + "learning_rate": 1.1946254294473939e-05, + "loss": 0.0324, + "step": 27759 + }, + { + "epoch": 3.64, + "learning_rate": 1.194573410773961e-05, + "loss": 0.0305, + "step": 27760 + }, + { + "epoch": 3.64, + "learning_rate": 1.194521391553301e-05, + "loss": 0.0368, + "step": 27761 + }, + { + "epoch": 3.64, + "learning_rate": 1.19446937178556e-05, + "loss": 0.0369, + "step": 27762 + }, + { + "epoch": 3.64, + "learning_rate": 1.1944173514708843e-05, + "loss": 0.034, + "step": 27763 + }, + { + "epoch": 3.64, + "learning_rate": 1.1943653306094199e-05, + "loss": 0.0442, + "step": 27764 + }, + { + "epoch": 3.64, + "learning_rate": 1.1943133092013138e-05, + "loss": 0.0347, + "step": 27765 + }, + { + "epoch": 3.64, + "learning_rate": 1.1942612872467111e-05, + "loss": 0.0367, + "step": 27766 + }, + { + "epoch": 3.64, + "learning_rate": 1.1942092647457596e-05, + "loss": 0.0339, + "step": 27767 + }, + { + "epoch": 3.64, + "learning_rate": 1.1941572416986045e-05, + "loss": 0.033, + "step": 27768 + }, + { + "epoch": 3.64, + "learning_rate": 1.1941052181053926e-05, + "loss": 0.0369, + "step": 27769 + }, + { + "epoch": 3.64, + "learning_rate": 1.1940531939662704e-05, + "loss": 0.0352, + "step": 27770 + }, + { + "epoch": 3.64, + "learning_rate": 1.1940011692813834e-05, + "loss": 0.0407, + "step": 27771 + }, + { + "epoch": 3.64, + "learning_rate": 1.1939491440508787e-05, + "loss": 0.041, + "step": 27772 + }, + { + "epoch": 3.64, + "learning_rate": 1.1938971182749025e-05, + "loss": 0.0427, + "step": 27773 + }, + { + "epoch": 3.64, + "learning_rate": 1.1938450919536007e-05, + "loss": 0.0401, + "step": 27774 + }, + { + "epoch": 3.64, + "learning_rate": 1.19379306508712e-05, + "loss": 0.0396, + "step": 27775 + }, + { + "epoch": 3.64, + "learning_rate": 1.1937410376756065e-05, + "loss": 0.0275, + "step": 27776 + }, + { + "epoch": 3.64, + "learning_rate": 1.1936890097192069e-05, + "loss": 0.0278, + "step": 27777 + }, + { + "epoch": 3.64, + "learning_rate": 1.1936369812180672e-05, + "loss": 0.0391, + "step": 27778 + }, + { + "epoch": 3.64, + "learning_rate": 1.1935849521723338e-05, + "loss": 0.0412, + "step": 27779 + }, + { + "epoch": 3.64, + "learning_rate": 1.193532922582153e-05, + "loss": 0.0353, + "step": 27780 + }, + { + "epoch": 3.64, + "learning_rate": 1.1934808924476712e-05, + "loss": 0.041, + "step": 27781 + }, + { + "epoch": 3.64, + "learning_rate": 1.1934288617690346e-05, + "loss": 0.0478, + "step": 27782 + }, + { + "epoch": 3.64, + "learning_rate": 1.1933768305463898e-05, + "loss": 0.0342, + "step": 27783 + }, + { + "epoch": 3.64, + "learning_rate": 1.1933247987798828e-05, + "loss": 0.0414, + "step": 27784 + }, + { + "epoch": 3.64, + "learning_rate": 1.1932727664696602e-05, + "loss": 0.0414, + "step": 27785 + }, + { + "epoch": 3.64, + "learning_rate": 1.1932207336158683e-05, + "loss": 0.0392, + "step": 27786 + }, + { + "epoch": 3.64, + "learning_rate": 1.193168700218653e-05, + "loss": 0.0396, + "step": 27787 + }, + { + "epoch": 3.64, + "learning_rate": 1.1931166662781615e-05, + "loss": 0.0373, + "step": 27788 + }, + { + "epoch": 3.64, + "learning_rate": 1.1930646317945394e-05, + "loss": 0.0389, + "step": 27789 + }, + { + "epoch": 3.64, + "learning_rate": 1.1930125967679334e-05, + "loss": 0.0387, + "step": 27790 + }, + { + "epoch": 3.64, + "learning_rate": 1.1929605611984896e-05, + "loss": 0.0322, + "step": 27791 + }, + { + "epoch": 3.64, + "learning_rate": 1.1929085250863549e-05, + "loss": 0.0377, + "step": 27792 + }, + { + "epoch": 3.64, + "learning_rate": 1.1928564884316747e-05, + "loss": 0.0441, + "step": 27793 + }, + { + "epoch": 3.64, + "learning_rate": 1.1928044512345963e-05, + "loss": 0.0444, + "step": 27794 + }, + { + "epoch": 3.64, + "learning_rate": 1.1927524134952653e-05, + "loss": 0.0402, + "step": 27795 + }, + { + "epoch": 3.64, + "learning_rate": 1.1927003752138287e-05, + "loss": 0.0411, + "step": 27796 + }, + { + "epoch": 3.64, + "learning_rate": 1.1926483363904325e-05, + "loss": 0.0387, + "step": 27797 + }, + { + "epoch": 3.64, + "learning_rate": 1.192596297025223e-05, + "loss": 0.0342, + "step": 27798 + }, + { + "epoch": 3.64, + "learning_rate": 1.1925442571183465e-05, + "loss": 0.0412, + "step": 27799 + }, + { + "epoch": 3.64, + "learning_rate": 1.1924922166699498e-05, + "loss": 0.0374, + "step": 27800 + }, + { + "epoch": 3.64, + "learning_rate": 1.1924401756801792e-05, + "loss": 0.0387, + "step": 27801 + }, + { + "epoch": 3.64, + "learning_rate": 1.1923881341491805e-05, + "loss": 0.0449, + "step": 27802 + }, + { + "epoch": 3.64, + "learning_rate": 1.1923360920771005e-05, + "loss": 0.0411, + "step": 27803 + }, + { + "epoch": 3.64, + "learning_rate": 1.1922840494640853e-05, + "loss": 0.0409, + "step": 27804 + }, + { + "epoch": 3.64, + "learning_rate": 1.1922320063102819e-05, + "loss": 0.0389, + "step": 27805 + }, + { + "epoch": 3.64, + "learning_rate": 1.192179962615836e-05, + "loss": 0.0355, + "step": 27806 + }, + { + "epoch": 3.64, + "learning_rate": 1.1921279183808943e-05, + "loss": 0.0352, + "step": 27807 + }, + { + "epoch": 3.64, + "learning_rate": 1.1920758736056028e-05, + "loss": 0.0414, + "step": 27808 + }, + { + "epoch": 3.64, + "learning_rate": 1.1920238282901082e-05, + "loss": 0.0358, + "step": 27809 + }, + { + "epoch": 3.64, + "learning_rate": 1.1919717824345566e-05, + "loss": 0.047, + "step": 27810 + }, + { + "epoch": 3.64, + "learning_rate": 1.1919197360390953e-05, + "loss": 0.0388, + "step": 27811 + }, + { + "epoch": 3.64, + "learning_rate": 1.1918676891038693e-05, + "loss": 0.0507, + "step": 27812 + }, + { + "epoch": 3.64, + "learning_rate": 1.1918156416290261e-05, + "loss": 0.0408, + "step": 27813 + }, + { + "epoch": 3.64, + "learning_rate": 1.1917635936147112e-05, + "loss": 0.0363, + "step": 27814 + }, + { + "epoch": 3.64, + "learning_rate": 1.1917115450610717e-05, + "loss": 0.0387, + "step": 27815 + }, + { + "epoch": 3.64, + "learning_rate": 1.1916594959682534e-05, + "loss": 0.041, + "step": 27816 + }, + { + "epoch": 3.64, + "learning_rate": 1.1916074463364034e-05, + "loss": 0.0363, + "step": 27817 + }, + { + "epoch": 3.64, + "learning_rate": 1.1915553961656673e-05, + "loss": 0.0333, + "step": 27818 + }, + { + "epoch": 3.64, + "learning_rate": 1.1915033454561919e-05, + "loss": 0.0341, + "step": 27819 + }, + { + "epoch": 3.64, + "learning_rate": 1.1914512942081235e-05, + "loss": 0.0393, + "step": 27820 + }, + { + "epoch": 3.64, + "learning_rate": 1.1913992424216087e-05, + "loss": 0.0397, + "step": 27821 + }, + { + "epoch": 3.64, + "learning_rate": 1.1913471900967937e-05, + "loss": 0.0382, + "step": 27822 + }, + { + "epoch": 3.64, + "learning_rate": 1.1912951372338249e-05, + "loss": 0.0329, + "step": 27823 + }, + { + "epoch": 3.64, + "learning_rate": 1.1912430838328488e-05, + "loss": 0.0424, + "step": 27824 + }, + { + "epoch": 3.64, + "learning_rate": 1.1911910298940115e-05, + "loss": 0.042, + "step": 27825 + }, + { + "epoch": 3.65, + "learning_rate": 1.1911389754174596e-05, + "loss": 0.0339, + "step": 27826 + }, + { + "epoch": 3.65, + "learning_rate": 1.1910869204033395e-05, + "loss": 0.0364, + "step": 27827 + }, + { + "epoch": 3.65, + "learning_rate": 1.191034864851798e-05, + "loss": 0.0392, + "step": 27828 + }, + { + "epoch": 3.65, + "learning_rate": 1.1909828087629806e-05, + "loss": 0.0353, + "step": 27829 + }, + { + "epoch": 3.65, + "learning_rate": 1.1909307521370348e-05, + "loss": 0.0397, + "step": 27830 + }, + { + "epoch": 3.65, + "learning_rate": 1.1908786949741057e-05, + "loss": 0.0362, + "step": 27831 + }, + { + "epoch": 3.65, + "learning_rate": 1.190826637274341e-05, + "loss": 0.0313, + "step": 27832 + }, + { + "epoch": 3.65, + "learning_rate": 1.1907745790378863e-05, + "loss": 0.0327, + "step": 27833 + }, + { + "epoch": 3.65, + "learning_rate": 1.1907225202648887e-05, + "loss": 0.036, + "step": 27834 + }, + { + "epoch": 3.65, + "learning_rate": 1.1906704609554936e-05, + "loss": 0.0389, + "step": 27835 + }, + { + "epoch": 3.65, + "learning_rate": 1.1906184011098482e-05, + "loss": 0.0396, + "step": 27836 + }, + { + "epoch": 3.65, + "learning_rate": 1.1905663407280986e-05, + "loss": 0.0394, + "step": 27837 + }, + { + "epoch": 3.65, + "learning_rate": 1.1905142798103915e-05, + "loss": 0.0339, + "step": 27838 + }, + { + "epoch": 3.65, + "learning_rate": 1.1904622183568731e-05, + "loss": 0.0396, + "step": 27839 + }, + { + "epoch": 3.65, + "learning_rate": 1.1904101563676895e-05, + "loss": 0.0392, + "step": 27840 + }, + { + "epoch": 3.65, + "learning_rate": 1.190358093842988e-05, + "loss": 0.0359, + "step": 27841 + }, + { + "epoch": 3.65, + "learning_rate": 1.1903060307829142e-05, + "loss": 0.0431, + "step": 27842 + }, + { + "epoch": 3.65, + "learning_rate": 1.1902539671876148e-05, + "loss": 0.0348, + "step": 27843 + }, + { + "epoch": 3.65, + "learning_rate": 1.1902019030572363e-05, + "loss": 0.0371, + "step": 27844 + }, + { + "epoch": 3.65, + "learning_rate": 1.1901498383919253e-05, + "loss": 0.0325, + "step": 27845 + }, + { + "epoch": 3.65, + "learning_rate": 1.1900977731918277e-05, + "loss": 0.0416, + "step": 27846 + }, + { + "epoch": 3.65, + "learning_rate": 1.1900457074570902e-05, + "loss": 0.0354, + "step": 27847 + }, + { + "epoch": 3.65, + "learning_rate": 1.1899936411878594e-05, + "loss": 0.0409, + "step": 27848 + }, + { + "epoch": 3.65, + "learning_rate": 1.1899415743842817e-05, + "loss": 0.0287, + "step": 27849 + }, + { + "epoch": 3.65, + "learning_rate": 1.1898895070465034e-05, + "loss": 0.0389, + "step": 27850 + }, + { + "epoch": 3.65, + "learning_rate": 1.189837439174671e-05, + "loss": 0.0415, + "step": 27851 + }, + { + "epoch": 3.65, + "learning_rate": 1.1897853707689309e-05, + "loss": 0.0298, + "step": 27852 + }, + { + "epoch": 3.65, + "learning_rate": 1.1897333018294293e-05, + "loss": 0.0367, + "step": 27853 + }, + { + "epoch": 3.65, + "learning_rate": 1.1896812323563129e-05, + "loss": 0.0442, + "step": 27854 + }, + { + "epoch": 3.65, + "learning_rate": 1.1896291623497285e-05, + "loss": 0.0388, + "step": 27855 + }, + { + "epoch": 3.65, + "learning_rate": 1.1895770918098219e-05, + "loss": 0.0382, + "step": 27856 + }, + { + "epoch": 3.65, + "learning_rate": 1.1895250207367402e-05, + "loss": 0.0286, + "step": 27857 + }, + { + "epoch": 3.65, + "learning_rate": 1.189472949130629e-05, + "loss": 0.0354, + "step": 27858 + }, + { + "epoch": 3.65, + "learning_rate": 1.1894208769916356e-05, + "loss": 0.0314, + "step": 27859 + }, + { + "epoch": 3.65, + "learning_rate": 1.1893688043199056e-05, + "loss": 0.0368, + "step": 27860 + }, + { + "epoch": 3.65, + "learning_rate": 1.1893167311155864e-05, + "loss": 0.0429, + "step": 27861 + }, + { + "epoch": 3.65, + "learning_rate": 1.1892646573788238e-05, + "loss": 0.0316, + "step": 27862 + }, + { + "epoch": 3.65, + "learning_rate": 1.1892125831097643e-05, + "loss": 0.0359, + "step": 27863 + }, + { + "epoch": 3.65, + "learning_rate": 1.1891605083085547e-05, + "loss": 0.0432, + "step": 27864 + }, + { + "epoch": 3.65, + "learning_rate": 1.1891084329753408e-05, + "loss": 0.03, + "step": 27865 + }, + { + "epoch": 3.65, + "learning_rate": 1.1890563571102703e-05, + "loss": 0.0334, + "step": 27866 + }, + { + "epoch": 3.65, + "learning_rate": 1.1890042807134883e-05, + "loss": 0.0349, + "step": 27867 + }, + { + "epoch": 3.65, + "learning_rate": 1.188952203785142e-05, + "loss": 0.0363, + "step": 27868 + }, + { + "epoch": 3.65, + "learning_rate": 1.1889001263253773e-05, + "loss": 0.0306, + "step": 27869 + }, + { + "epoch": 3.65, + "learning_rate": 1.1888480483343417e-05, + "loss": 0.0409, + "step": 27870 + }, + { + "epoch": 3.65, + "learning_rate": 1.1887959698121806e-05, + "loss": 0.0348, + "step": 27871 + }, + { + "epoch": 3.65, + "learning_rate": 1.1887438907590409e-05, + "loss": 0.0338, + "step": 27872 + }, + { + "epoch": 3.65, + "learning_rate": 1.1886918111750692e-05, + "loss": 0.0394, + "step": 27873 + }, + { + "epoch": 3.65, + "learning_rate": 1.188639731060412e-05, + "loss": 0.0346, + "step": 27874 + }, + { + "epoch": 3.65, + "learning_rate": 1.188587650415215e-05, + "loss": 0.0279, + "step": 27875 + }, + { + "epoch": 3.65, + "learning_rate": 1.188535569239626e-05, + "loss": 0.0382, + "step": 27876 + }, + { + "epoch": 3.65, + "learning_rate": 1.18848348753379e-05, + "loss": 0.04, + "step": 27877 + }, + { + "epoch": 3.65, + "learning_rate": 1.1884314052978549e-05, + "loss": 0.0455, + "step": 27878 + }, + { + "epoch": 3.65, + "learning_rate": 1.1883793225319662e-05, + "loss": 0.036, + "step": 27879 + }, + { + "epoch": 3.65, + "learning_rate": 1.1883272392362705e-05, + "loss": 0.0299, + "step": 27880 + }, + { + "epoch": 3.65, + "learning_rate": 1.1882751554109149e-05, + "loss": 0.0328, + "step": 27881 + }, + { + "epoch": 3.65, + "learning_rate": 1.1882230710560454e-05, + "loss": 0.0408, + "step": 27882 + }, + { + "epoch": 3.65, + "learning_rate": 1.1881709861718083e-05, + "loss": 0.0359, + "step": 27883 + }, + { + "epoch": 3.65, + "learning_rate": 1.1881189007583502e-05, + "loss": 0.036, + "step": 27884 + }, + { + "epoch": 3.65, + "learning_rate": 1.188066814815818e-05, + "loss": 0.0397, + "step": 27885 + }, + { + "epoch": 3.65, + "learning_rate": 1.1880147283443579e-05, + "loss": 0.0361, + "step": 27886 + }, + { + "epoch": 3.65, + "learning_rate": 1.1879626413441162e-05, + "loss": 0.0362, + "step": 27887 + }, + { + "epoch": 3.65, + "learning_rate": 1.1879105538152396e-05, + "loss": 0.0438, + "step": 27888 + }, + { + "epoch": 3.65, + "learning_rate": 1.187858465757875e-05, + "loss": 0.0396, + "step": 27889 + }, + { + "epoch": 3.65, + "learning_rate": 1.187806377172168e-05, + "loss": 0.0294, + "step": 27890 + }, + { + "epoch": 3.65, + "learning_rate": 1.1877542880582656e-05, + "loss": 0.0353, + "step": 27891 + }, + { + "epoch": 3.65, + "learning_rate": 1.1877021984163142e-05, + "loss": 0.0362, + "step": 27892 + }, + { + "epoch": 3.65, + "learning_rate": 1.1876501082464607e-05, + "loss": 0.0395, + "step": 27893 + }, + { + "epoch": 3.65, + "learning_rate": 1.1875980175488511e-05, + "loss": 0.0343, + "step": 27894 + }, + { + "epoch": 3.65, + "learning_rate": 1.1875459263236322e-05, + "loss": 0.0382, + "step": 27895 + }, + { + "epoch": 3.65, + "learning_rate": 1.1874938345709502e-05, + "loss": 0.0356, + "step": 27896 + }, + { + "epoch": 3.65, + "learning_rate": 1.1874417422909516e-05, + "loss": 0.0365, + "step": 27897 + }, + { + "epoch": 3.65, + "learning_rate": 1.1873896494837833e-05, + "loss": 0.0365, + "step": 27898 + }, + { + "epoch": 3.65, + "learning_rate": 1.1873375561495917e-05, + "loss": 0.0383, + "step": 27899 + }, + { + "epoch": 3.65, + "learning_rate": 1.1872854622885228e-05, + "loss": 0.0388, + "step": 27900 + }, + { + "epoch": 3.65, + "learning_rate": 1.1872333679007242e-05, + "loss": 0.0361, + "step": 27901 + }, + { + "epoch": 3.65, + "learning_rate": 1.187181272986341e-05, + "loss": 0.0411, + "step": 27902 + }, + { + "epoch": 3.66, + "learning_rate": 1.1871291775455207e-05, + "loss": 0.0403, + "step": 27903 + }, + { + "epoch": 3.66, + "learning_rate": 1.1870770815784095e-05, + "loss": 0.037, + "step": 27904 + }, + { + "epoch": 3.66, + "learning_rate": 1.187024985085154e-05, + "loss": 0.0364, + "step": 27905 + }, + { + "epoch": 3.66, + "learning_rate": 1.1869728880659011e-05, + "loss": 0.0399, + "step": 27906 + }, + { + "epoch": 3.66, + "learning_rate": 1.1869207905207965e-05, + "loss": 0.0407, + "step": 27907 + }, + { + "epoch": 3.66, + "learning_rate": 1.186868692449987e-05, + "loss": 0.0325, + "step": 27908 + }, + { + "epoch": 3.66, + "learning_rate": 1.1868165938536194e-05, + "loss": 0.0373, + "step": 27909 + }, + { + "epoch": 3.66, + "learning_rate": 1.1867644947318402e-05, + "loss": 0.0432, + "step": 27910 + }, + { + "epoch": 3.66, + "learning_rate": 1.1867123950847956e-05, + "loss": 0.0374, + "step": 27911 + }, + { + "epoch": 3.66, + "learning_rate": 1.1866602949126326e-05, + "loss": 0.0467, + "step": 27912 + }, + { + "epoch": 3.66, + "learning_rate": 1.1866081942154972e-05, + "loss": 0.0342, + "step": 27913 + }, + { + "epoch": 3.66, + "learning_rate": 1.1865560929935364e-05, + "loss": 0.0374, + "step": 27914 + }, + { + "epoch": 3.66, + "learning_rate": 1.1865039912468964e-05, + "loss": 0.0376, + "step": 27915 + }, + { + "epoch": 3.66, + "learning_rate": 1.1864518889757242e-05, + "loss": 0.0382, + "step": 27916 + }, + { + "epoch": 3.66, + "learning_rate": 1.1863997861801655e-05, + "loss": 0.0357, + "step": 27917 + }, + { + "epoch": 3.66, + "learning_rate": 1.1863476828603676e-05, + "loss": 0.039, + "step": 27918 + }, + { + "epoch": 3.66, + "learning_rate": 1.1862955790164765e-05, + "loss": 0.0361, + "step": 27919 + }, + { + "epoch": 3.66, + "learning_rate": 1.1862434746486392e-05, + "loss": 0.0329, + "step": 27920 + }, + { + "epoch": 3.66, + "learning_rate": 1.186191369757002e-05, + "loss": 0.0336, + "step": 27921 + }, + { + "epoch": 3.66, + "learning_rate": 1.1861392643417116e-05, + "loss": 0.0376, + "step": 27922 + }, + { + "epoch": 3.66, + "learning_rate": 1.1860871584029144e-05, + "loss": 0.0422, + "step": 27923 + }, + { + "epoch": 3.66, + "learning_rate": 1.1860350519407568e-05, + "loss": 0.0355, + "step": 27924 + }, + { + "epoch": 3.66, + "learning_rate": 1.1859829449553857e-05, + "loss": 0.039, + "step": 27925 + }, + { + "epoch": 3.66, + "learning_rate": 1.1859308374469473e-05, + "loss": 0.0399, + "step": 27926 + }, + { + "epoch": 3.66, + "learning_rate": 1.1858787294155884e-05, + "loss": 0.033, + "step": 27927 + }, + { + "epoch": 3.66, + "learning_rate": 1.1858266208614555e-05, + "loss": 0.0406, + "step": 27928 + }, + { + "epoch": 3.66, + "learning_rate": 1.185774511784695e-05, + "loss": 0.0355, + "step": 27929 + }, + { + "epoch": 3.66, + "learning_rate": 1.1857224021854539e-05, + "loss": 0.0285, + "step": 27930 + }, + { + "epoch": 3.66, + "learning_rate": 1.1856702920638782e-05, + "loss": 0.0287, + "step": 27931 + }, + { + "epoch": 3.66, + "learning_rate": 1.1856181814201149e-05, + "loss": 0.0304, + "step": 27932 + }, + { + "epoch": 3.66, + "learning_rate": 1.1855660702543102e-05, + "loss": 0.0391, + "step": 27933 + }, + { + "epoch": 3.66, + "learning_rate": 1.1855139585666106e-05, + "loss": 0.035, + "step": 27934 + }, + { + "epoch": 3.66, + "learning_rate": 1.1854618463571633e-05, + "loss": 0.0333, + "step": 27935 + }, + { + "epoch": 3.66, + "learning_rate": 1.1854097336261139e-05, + "loss": 0.0426, + "step": 27936 + }, + { + "epoch": 3.66, + "learning_rate": 1.1853576203736102e-05, + "loss": 0.0418, + "step": 27937 + }, + { + "epoch": 3.66, + "learning_rate": 1.1853055065997973e-05, + "loss": 0.0347, + "step": 27938 + }, + { + "epoch": 3.66, + "learning_rate": 1.1852533923048233e-05, + "loss": 0.0396, + "step": 27939 + }, + { + "epoch": 3.66, + "learning_rate": 1.1852012774888331e-05, + "loss": 0.0373, + "step": 27940 + }, + { + "epoch": 3.66, + "learning_rate": 1.1851491621519748e-05, + "loss": 0.0323, + "step": 27941 + }, + { + "epoch": 3.66, + "learning_rate": 1.1850970462943943e-05, + "loss": 0.0312, + "step": 27942 + }, + { + "epoch": 3.66, + "learning_rate": 1.185044929916238e-05, + "loss": 0.0386, + "step": 27943 + }, + { + "epoch": 3.66, + "learning_rate": 1.184992813017653e-05, + "loss": 0.0309, + "step": 27944 + }, + { + "epoch": 3.66, + "learning_rate": 1.1849406955987853e-05, + "loss": 0.0436, + "step": 27945 + }, + { + "epoch": 3.66, + "learning_rate": 1.1848885776597816e-05, + "loss": 0.0471, + "step": 27946 + }, + { + "epoch": 3.66, + "learning_rate": 1.1848364592007891e-05, + "loss": 0.0368, + "step": 27947 + }, + { + "epoch": 3.66, + "learning_rate": 1.1847843402219534e-05, + "loss": 0.0337, + "step": 27948 + }, + { + "epoch": 3.66, + "learning_rate": 1.184732220723422e-05, + "loss": 0.036, + "step": 27949 + }, + { + "epoch": 3.66, + "learning_rate": 1.184680100705341e-05, + "loss": 0.0364, + "step": 27950 + }, + { + "epoch": 3.66, + "learning_rate": 1.1846279801678568e-05, + "loss": 0.0429, + "step": 27951 + }, + { + "epoch": 3.66, + "learning_rate": 1.1845758591111162e-05, + "loss": 0.0326, + "step": 27952 + }, + { + "epoch": 3.66, + "learning_rate": 1.184523737535266e-05, + "loss": 0.0401, + "step": 27953 + }, + { + "epoch": 3.66, + "learning_rate": 1.1844716154404527e-05, + "loss": 0.0429, + "step": 27954 + }, + { + "epoch": 3.66, + "learning_rate": 1.1844194928268226e-05, + "loss": 0.0361, + "step": 27955 + }, + { + "epoch": 3.66, + "learning_rate": 1.184367369694523e-05, + "loss": 0.0374, + "step": 27956 + }, + { + "epoch": 3.66, + "learning_rate": 1.1843152460436992e-05, + "loss": 0.0359, + "step": 27957 + }, + { + "epoch": 3.66, + "learning_rate": 1.1842631218744994e-05, + "loss": 0.0389, + "step": 27958 + }, + { + "epoch": 3.66, + "learning_rate": 1.1842109971870688e-05, + "loss": 0.0335, + "step": 27959 + }, + { + "epoch": 3.66, + "learning_rate": 1.1841588719815548e-05, + "loss": 0.0352, + "step": 27960 + }, + { + "epoch": 3.66, + "learning_rate": 1.1841067462581037e-05, + "loss": 0.0366, + "step": 27961 + }, + { + "epoch": 3.66, + "learning_rate": 1.184054620016862e-05, + "loss": 0.039, + "step": 27962 + }, + { + "epoch": 3.66, + "learning_rate": 1.1840024932579767e-05, + "loss": 0.0396, + "step": 27963 + }, + { + "epoch": 3.66, + "learning_rate": 1.1839503659815946e-05, + "loss": 0.0335, + "step": 27964 + }, + { + "epoch": 3.66, + "learning_rate": 1.1838982381878613e-05, + "loss": 0.0358, + "step": 27965 + }, + { + "epoch": 3.66, + "learning_rate": 1.1838461098769241e-05, + "loss": 0.0323, + "step": 27966 + }, + { + "epoch": 3.66, + "learning_rate": 1.1837939810489299e-05, + "loss": 0.0427, + "step": 27967 + }, + { + "epoch": 3.66, + "learning_rate": 1.1837418517040245e-05, + "loss": 0.036, + "step": 27968 + }, + { + "epoch": 3.66, + "learning_rate": 1.183689721842355e-05, + "loss": 0.0377, + "step": 27969 + }, + { + "epoch": 3.66, + "learning_rate": 1.1836375914640677e-05, + "loss": 0.0334, + "step": 27970 + }, + { + "epoch": 3.66, + "learning_rate": 1.1835854605693098e-05, + "loss": 0.0353, + "step": 27971 + }, + { + "epoch": 3.66, + "learning_rate": 1.1835333291582276e-05, + "loss": 0.0309, + "step": 27972 + }, + { + "epoch": 3.66, + "learning_rate": 1.1834811972309675e-05, + "loss": 0.0308, + "step": 27973 + }, + { + "epoch": 3.66, + "learning_rate": 1.1834290647876762e-05, + "loss": 0.0407, + "step": 27974 + }, + { + "epoch": 3.66, + "learning_rate": 1.1833769318285007e-05, + "loss": 0.0389, + "step": 27975 + }, + { + "epoch": 3.66, + "learning_rate": 1.1833247983535872e-05, + "loss": 0.035, + "step": 27976 + }, + { + "epoch": 3.66, + "learning_rate": 1.1832726643630828e-05, + "loss": 0.0367, + "step": 27977 + }, + { + "epoch": 3.66, + "learning_rate": 1.1832205298571332e-05, + "loss": 0.0322, + "step": 27978 + }, + { + "epoch": 3.67, + "learning_rate": 1.1831683948358861e-05, + "loss": 0.0332, + "step": 27979 + }, + { + "epoch": 3.67, + "learning_rate": 1.1831162592994873e-05, + "loss": 0.0316, + "step": 27980 + }, + { + "epoch": 3.67, + "learning_rate": 1.1830641232480842e-05, + "loss": 0.0387, + "step": 27981 + }, + { + "epoch": 3.67, + "learning_rate": 1.1830119866818227e-05, + "loss": 0.0333, + "step": 27982 + }, + { + "epoch": 3.67, + "learning_rate": 1.1829598496008498e-05, + "loss": 0.0364, + "step": 27983 + }, + { + "epoch": 3.67, + "learning_rate": 1.1829077120053119e-05, + "loss": 0.0341, + "step": 27984 + }, + { + "epoch": 3.67, + "learning_rate": 1.182855573895356e-05, + "loss": 0.0339, + "step": 27985 + }, + { + "epoch": 3.67, + "learning_rate": 1.1828034352711284e-05, + "loss": 0.0401, + "step": 27986 + }, + { + "epoch": 3.67, + "learning_rate": 1.182751296132776e-05, + "loss": 0.0446, + "step": 27987 + }, + { + "epoch": 3.67, + "learning_rate": 1.1826991564804453e-05, + "loss": 0.0341, + "step": 27988 + }, + { + "epoch": 3.67, + "learning_rate": 1.1826470163142827e-05, + "loss": 0.0333, + "step": 27989 + }, + { + "epoch": 3.67, + "learning_rate": 1.1825948756344354e-05, + "loss": 0.041, + "step": 27990 + }, + { + "epoch": 3.67, + "learning_rate": 1.1825427344410495e-05, + "loss": 0.0361, + "step": 27991 + }, + { + "epoch": 3.67, + "learning_rate": 1.1824905927342722e-05, + "loss": 0.0417, + "step": 27992 + }, + { + "epoch": 3.67, + "learning_rate": 1.1824384505142497e-05, + "loss": 0.0375, + "step": 27993 + }, + { + "epoch": 3.67, + "learning_rate": 1.1823863077811288e-05, + "loss": 0.0385, + "step": 27994 + }, + { + "epoch": 3.67, + "learning_rate": 1.1823341645350557e-05, + "loss": 0.0392, + "step": 27995 + }, + { + "epoch": 3.67, + "learning_rate": 1.182282020776178e-05, + "loss": 0.0326, + "step": 27996 + }, + { + "epoch": 3.67, + "learning_rate": 1.1822298765046416e-05, + "loss": 0.0347, + "step": 27997 + }, + { + "epoch": 3.67, + "learning_rate": 1.1821777317205936e-05, + "loss": 0.0355, + "step": 27998 + }, + { + "epoch": 3.67, + "learning_rate": 1.18212558642418e-05, + "loss": 0.0375, + "step": 27999 + }, + { + "epoch": 3.67, + "learning_rate": 1.1820734406155481e-05, + "loss": 0.0382, + "step": 28000 + }, + { + "epoch": 3.67, + "learning_rate": 1.1820212942948443e-05, + "loss": 0.043, + "step": 28001 + }, + { + "epoch": 3.67, + "learning_rate": 1.1819691474622154e-05, + "loss": 0.0349, + "step": 28002 + }, + { + "epoch": 3.67, + "learning_rate": 1.181917000117808e-05, + "loss": 0.0321, + "step": 28003 + }, + { + "epoch": 3.67, + "learning_rate": 1.1818648522617689e-05, + "loss": 0.0419, + "step": 28004 + }, + { + "epoch": 3.67, + "learning_rate": 1.1818127038942441e-05, + "loss": 0.0365, + "step": 28005 + }, + { + "epoch": 3.67, + "learning_rate": 1.181760555015381e-05, + "loss": 0.0404, + "step": 28006 + }, + { + "epoch": 3.67, + "learning_rate": 1.181708405625326e-05, + "loss": 0.0394, + "step": 28007 + }, + { + "epoch": 3.67, + "learning_rate": 1.1816562557242262e-05, + "loss": 0.0323, + "step": 28008 + }, + { + "epoch": 3.67, + "learning_rate": 1.1816041053122272e-05, + "loss": 0.0337, + "step": 28009 + }, + { + "epoch": 3.67, + "learning_rate": 1.1815519543894768e-05, + "loss": 0.0383, + "step": 28010 + }, + { + "epoch": 3.67, + "learning_rate": 1.1814998029561207e-05, + "loss": 0.0318, + "step": 28011 + }, + { + "epoch": 3.67, + "learning_rate": 1.1814476510123064e-05, + "loss": 0.0387, + "step": 28012 + }, + { + "epoch": 3.67, + "learning_rate": 1.1813954985581804e-05, + "loss": 0.038, + "step": 28013 + }, + { + "epoch": 3.67, + "learning_rate": 1.1813433455938887e-05, + "loss": 0.0361, + "step": 28014 + }, + { + "epoch": 3.67, + "learning_rate": 1.1812911921195791e-05, + "loss": 0.0324, + "step": 28015 + }, + { + "epoch": 3.67, + "learning_rate": 1.1812390381353974e-05, + "loss": 0.0325, + "step": 28016 + }, + { + "epoch": 3.67, + "learning_rate": 1.1811868836414903e-05, + "loss": 0.0335, + "step": 28017 + }, + { + "epoch": 3.67, + "learning_rate": 1.181134728638005e-05, + "loss": 0.0361, + "step": 28018 + }, + { + "epoch": 3.67, + "learning_rate": 1.1810825731250881e-05, + "loss": 0.0331, + "step": 28019 + }, + { + "epoch": 3.67, + "learning_rate": 1.181030417102886e-05, + "loss": 0.0336, + "step": 28020 + }, + { + "epoch": 3.67, + "learning_rate": 1.1809782605715453e-05, + "loss": 0.0499, + "step": 28021 + }, + { + "epoch": 3.67, + "learning_rate": 1.1809261035312127e-05, + "loss": 0.0408, + "step": 28022 + }, + { + "epoch": 3.67, + "learning_rate": 1.1808739459820356e-05, + "loss": 0.0371, + "step": 28023 + }, + { + "epoch": 3.67, + "learning_rate": 1.1808217879241598e-05, + "loss": 0.0386, + "step": 28024 + }, + { + "epoch": 3.67, + "learning_rate": 1.1807696293577326e-05, + "loss": 0.0371, + "step": 28025 + }, + { + "epoch": 3.67, + "learning_rate": 1.1807174702829003e-05, + "loss": 0.0288, + "step": 28026 + }, + { + "epoch": 3.67, + "learning_rate": 1.1806653106998098e-05, + "loss": 0.0342, + "step": 28027 + }, + { + "epoch": 3.67, + "learning_rate": 1.1806131506086076e-05, + "loss": 0.0366, + "step": 28028 + }, + { + "epoch": 3.67, + "learning_rate": 1.1805609900094409e-05, + "loss": 0.0388, + "step": 28029 + }, + { + "epoch": 3.67, + "learning_rate": 1.1805088289024556e-05, + "loss": 0.0368, + "step": 28030 + }, + { + "epoch": 3.67, + "learning_rate": 1.1804566672877988e-05, + "loss": 0.0382, + "step": 28031 + }, + { + "epoch": 3.67, + "learning_rate": 1.1804045051656176e-05, + "loss": 0.0438, + "step": 28032 + }, + { + "epoch": 3.67, + "learning_rate": 1.180352342536058e-05, + "loss": 0.0299, + "step": 28033 + }, + { + "epoch": 3.67, + "learning_rate": 1.1803001793992672e-05, + "loss": 0.0413, + "step": 28034 + }, + { + "epoch": 3.67, + "learning_rate": 1.1802480157553916e-05, + "loss": 0.0358, + "step": 28035 + }, + { + "epoch": 3.67, + "learning_rate": 1.1801958516045784e-05, + "loss": 0.0294, + "step": 28036 + }, + { + "epoch": 3.67, + "learning_rate": 1.1801436869469737e-05, + "loss": 0.0397, + "step": 28037 + }, + { + "epoch": 3.67, + "learning_rate": 1.1800915217827247e-05, + "loss": 0.035, + "step": 28038 + }, + { + "epoch": 3.67, + "learning_rate": 1.1800393561119773e-05, + "loss": 0.0434, + "step": 28039 + }, + { + "epoch": 3.67, + "learning_rate": 1.1799871899348796e-05, + "loss": 0.0355, + "step": 28040 + }, + { + "epoch": 3.67, + "learning_rate": 1.179935023251577e-05, + "loss": 0.0336, + "step": 28041 + }, + { + "epoch": 3.67, + "learning_rate": 1.179882856062217e-05, + "loss": 0.036, + "step": 28042 + }, + { + "epoch": 3.67, + "learning_rate": 1.1798306883669457e-05, + "loss": 0.037, + "step": 28043 + }, + { + "epoch": 3.67, + "learning_rate": 1.1797785201659105e-05, + "loss": 0.0362, + "step": 28044 + }, + { + "epoch": 3.67, + "learning_rate": 1.1797263514592573e-05, + "loss": 0.0412, + "step": 28045 + }, + { + "epoch": 3.67, + "learning_rate": 1.179674182247134e-05, + "loss": 0.0416, + "step": 28046 + }, + { + "epoch": 3.67, + "learning_rate": 1.1796220125296863e-05, + "loss": 0.0345, + "step": 28047 + }, + { + "epoch": 3.67, + "learning_rate": 1.1795698423070613e-05, + "loss": 0.034, + "step": 28048 + }, + { + "epoch": 3.67, + "learning_rate": 1.1795176715794055e-05, + "loss": 0.0382, + "step": 28049 + }, + { + "epoch": 3.67, + "learning_rate": 1.179465500346866e-05, + "loss": 0.0376, + "step": 28050 + }, + { + "epoch": 3.67, + "learning_rate": 1.1794133286095893e-05, + "loss": 0.0397, + "step": 28051 + }, + { + "epoch": 3.67, + "learning_rate": 1.1793611563677221e-05, + "loss": 0.0327, + "step": 28052 + }, + { + "epoch": 3.67, + "learning_rate": 1.1793089836214113e-05, + "loss": 0.0372, + "step": 28053 + }, + { + "epoch": 3.67, + "learning_rate": 1.1792568103708033e-05, + "loss": 0.0339, + "step": 28054 + }, + { + "epoch": 3.68, + "learning_rate": 1.1792046366160452e-05, + "loss": 0.0404, + "step": 28055 + }, + { + "epoch": 3.68, + "learning_rate": 1.1791524623572836e-05, + "loss": 0.0349, + "step": 28056 + }, + { + "epoch": 3.68, + "learning_rate": 1.1791002875946656e-05, + "loss": 0.0388, + "step": 28057 + }, + { + "epoch": 3.68, + "learning_rate": 1.179048112328337e-05, + "loss": 0.0404, + "step": 28058 + }, + { + "epoch": 3.68, + "learning_rate": 1.1789959365584455e-05, + "loss": 0.0384, + "step": 28059 + }, + { + "epoch": 3.68, + "learning_rate": 1.1789437602851372e-05, + "loss": 0.0328, + "step": 28060 + }, + { + "epoch": 3.68, + "learning_rate": 1.1788915835085594e-05, + "loss": 0.0327, + "step": 28061 + }, + { + "epoch": 3.68, + "learning_rate": 1.1788394062288581e-05, + "loss": 0.0395, + "step": 28062 + }, + { + "epoch": 3.68, + "learning_rate": 1.1787872284461812e-05, + "loss": 0.04, + "step": 28063 + }, + { + "epoch": 3.68, + "learning_rate": 1.1787350501606743e-05, + "loss": 0.0302, + "step": 28064 + }, + { + "epoch": 3.68, + "learning_rate": 1.1786828713724847e-05, + "loss": 0.0327, + "step": 28065 + }, + { + "epoch": 3.68, + "learning_rate": 1.1786306920817587e-05, + "loss": 0.0431, + "step": 28066 + }, + { + "epoch": 3.68, + "learning_rate": 1.1785785122886438e-05, + "loss": 0.0397, + "step": 28067 + }, + { + "epoch": 3.68, + "learning_rate": 1.178526331993286e-05, + "loss": 0.0437, + "step": 28068 + }, + { + "epoch": 3.68, + "learning_rate": 1.178474151195833e-05, + "loss": 0.0345, + "step": 28069 + }, + { + "epoch": 3.68, + "learning_rate": 1.1784219698964304e-05, + "loss": 0.0354, + "step": 28070 + }, + { + "epoch": 3.68, + "learning_rate": 1.1783697880952253e-05, + "loss": 0.0362, + "step": 28071 + }, + { + "epoch": 3.68, + "learning_rate": 1.178317605792365e-05, + "loss": 0.0358, + "step": 28072 + }, + { + "epoch": 3.68, + "learning_rate": 1.1782654229879962e-05, + "loss": 0.0354, + "step": 28073 + }, + { + "epoch": 3.68, + "learning_rate": 1.1782132396822652e-05, + "loss": 0.0448, + "step": 28074 + }, + { + "epoch": 3.68, + "learning_rate": 1.1781610558753186e-05, + "loss": 0.0361, + "step": 28075 + }, + { + "epoch": 3.68, + "learning_rate": 1.1781088715673042e-05, + "loss": 0.0367, + "step": 28076 + }, + { + "epoch": 3.68, + "learning_rate": 1.1780566867583677e-05, + "loss": 0.0273, + "step": 28077 + }, + { + "epoch": 3.68, + "learning_rate": 1.1780045014486561e-05, + "loss": 0.0375, + "step": 28078 + }, + { + "epoch": 3.68, + "learning_rate": 1.1779523156383164e-05, + "loss": 0.0269, + "step": 28079 + }, + { + "epoch": 3.68, + "learning_rate": 1.1779001293274954e-05, + "loss": 0.0403, + "step": 28080 + }, + { + "epoch": 3.68, + "learning_rate": 1.1778479425163399e-05, + "loss": 0.036, + "step": 28081 + }, + { + "epoch": 3.68, + "learning_rate": 1.1777957552049961e-05, + "loss": 0.0365, + "step": 28082 + }, + { + "epoch": 3.68, + "learning_rate": 1.1777435673936116e-05, + "loss": 0.046, + "step": 28083 + }, + { + "epoch": 3.68, + "learning_rate": 1.1776913790823325e-05, + "loss": 0.0343, + "step": 28084 + }, + { + "epoch": 3.68, + "learning_rate": 1.177639190271306e-05, + "loss": 0.0333, + "step": 28085 + }, + { + "epoch": 3.68, + "learning_rate": 1.1775870009606788e-05, + "loss": 0.0359, + "step": 28086 + }, + { + "epoch": 3.68, + "learning_rate": 1.1775348111505975e-05, + "loss": 0.0327, + "step": 28087 + }, + { + "epoch": 3.68, + "learning_rate": 1.1774826208412091e-05, + "loss": 0.0349, + "step": 28088 + }, + { + "epoch": 3.68, + "learning_rate": 1.1774304300326602e-05, + "loss": 0.0385, + "step": 28089 + }, + { + "epoch": 3.68, + "learning_rate": 1.1773782387250978e-05, + "loss": 0.0353, + "step": 28090 + }, + { + "epoch": 3.68, + "learning_rate": 1.1773260469186684e-05, + "loss": 0.0319, + "step": 28091 + }, + { + "epoch": 3.68, + "learning_rate": 1.1772738546135191e-05, + "loss": 0.0353, + "step": 28092 + }, + { + "epoch": 3.68, + "learning_rate": 1.1772216618097962e-05, + "loss": 0.0376, + "step": 28093 + }, + { + "epoch": 3.68, + "learning_rate": 1.1771694685076471e-05, + "loss": 0.0347, + "step": 28094 + }, + { + "epoch": 3.68, + "learning_rate": 1.1771172747072182e-05, + "loss": 0.0305, + "step": 28095 + }, + { + "epoch": 3.68, + "learning_rate": 1.1770650804086563e-05, + "loss": 0.0424, + "step": 28096 + }, + { + "epoch": 3.68, + "learning_rate": 1.1770128856121087e-05, + "loss": 0.0359, + "step": 28097 + }, + { + "epoch": 3.68, + "learning_rate": 1.1769606903177213e-05, + "loss": 0.0392, + "step": 28098 + }, + { + "epoch": 3.68, + "learning_rate": 1.1769084945256414e-05, + "loss": 0.0386, + "step": 28099 + }, + { + "epoch": 3.68, + "learning_rate": 1.1768562982360161e-05, + "loss": 0.0396, + "step": 28100 + }, + { + "epoch": 3.68, + "learning_rate": 1.1768041014489919e-05, + "loss": 0.0338, + "step": 28101 + }, + { + "epoch": 3.68, + "learning_rate": 1.1767519041647152e-05, + "loss": 0.036, + "step": 28102 + }, + { + "epoch": 3.68, + "learning_rate": 1.1766997063833336e-05, + "loss": 0.0392, + "step": 28103 + }, + { + "epoch": 3.68, + "learning_rate": 1.1766475081049931e-05, + "loss": 0.0362, + "step": 28104 + }, + { + "epoch": 3.68, + "learning_rate": 1.1765953093298413e-05, + "loss": 0.0327, + "step": 28105 + }, + { + "epoch": 3.68, + "learning_rate": 1.1765431100580242e-05, + "loss": 0.0297, + "step": 28106 + }, + { + "epoch": 3.68, + "learning_rate": 1.1764909102896893e-05, + "loss": 0.0348, + "step": 28107 + }, + { + "epoch": 3.68, + "learning_rate": 1.176438710024983e-05, + "loss": 0.0375, + "step": 28108 + }, + { + "epoch": 3.68, + "learning_rate": 1.1763865092640522e-05, + "loss": 0.0305, + "step": 28109 + }, + { + "epoch": 3.68, + "learning_rate": 1.1763343080070437e-05, + "loss": 0.0401, + "step": 28110 + }, + { + "epoch": 3.68, + "learning_rate": 1.1762821062541046e-05, + "loss": 0.0462, + "step": 28111 + }, + { + "epoch": 3.68, + "learning_rate": 1.1762299040053811e-05, + "loss": 0.0322, + "step": 28112 + }, + { + "epoch": 3.68, + "learning_rate": 1.1761777012610207e-05, + "loss": 0.0338, + "step": 28113 + }, + { + "epoch": 3.68, + "learning_rate": 1.1761254980211697e-05, + "loss": 0.0391, + "step": 28114 + }, + { + "epoch": 3.68, + "learning_rate": 1.1760732942859751e-05, + "loss": 0.0357, + "step": 28115 + }, + { + "epoch": 3.68, + "learning_rate": 1.1760210900555838e-05, + "loss": 0.0434, + "step": 28116 + }, + { + "epoch": 3.68, + "learning_rate": 1.1759688853301426e-05, + "loss": 0.0373, + "step": 28117 + }, + { + "epoch": 3.68, + "learning_rate": 1.1759166801097982e-05, + "loss": 0.0365, + "step": 28118 + }, + { + "epoch": 3.68, + "learning_rate": 1.1758644743946974e-05, + "loss": 0.035, + "step": 28119 + }, + { + "epoch": 3.68, + "learning_rate": 1.1758122681849873e-05, + "loss": 0.0321, + "step": 28120 + }, + { + "epoch": 3.68, + "learning_rate": 1.1757600614808144e-05, + "loss": 0.0354, + "step": 28121 + }, + { + "epoch": 3.68, + "learning_rate": 1.175707854282326e-05, + "loss": 0.0376, + "step": 28122 + }, + { + "epoch": 3.68, + "learning_rate": 1.1756556465896684e-05, + "loss": 0.0382, + "step": 28123 + }, + { + "epoch": 3.68, + "learning_rate": 1.1756034384029888e-05, + "loss": 0.0383, + "step": 28124 + }, + { + "epoch": 3.68, + "learning_rate": 1.1755512297224336e-05, + "loss": 0.0408, + "step": 28125 + }, + { + "epoch": 3.68, + "learning_rate": 1.1754990205481502e-05, + "loss": 0.0407, + "step": 28126 + }, + { + "epoch": 3.68, + "learning_rate": 1.1754468108802847e-05, + "loss": 0.038, + "step": 28127 + }, + { + "epoch": 3.68, + "learning_rate": 1.1753946007189851e-05, + "loss": 0.0374, + "step": 28128 + }, + { + "epoch": 3.68, + "learning_rate": 1.175342390064397e-05, + "loss": 0.0333, + "step": 28129 + }, + { + "epoch": 3.68, + "learning_rate": 1.1752901789166682e-05, + "loss": 0.0366, + "step": 28130 + }, + { + "epoch": 3.68, + "learning_rate": 1.1752379672759444e-05, + "loss": 0.0382, + "step": 28131 + }, + { + "epoch": 3.69, + "learning_rate": 1.1751857551423738e-05, + "loss": 0.0399, + "step": 28132 + }, + { + "epoch": 3.69, + "learning_rate": 1.1751335425161022e-05, + "loss": 0.0329, + "step": 28133 + }, + { + "epoch": 3.69, + "learning_rate": 1.175081329397277e-05, + "loss": 0.0381, + "step": 28134 + }, + { + "epoch": 3.69, + "learning_rate": 1.175029115786045e-05, + "loss": 0.0351, + "step": 28135 + }, + { + "epoch": 3.69, + "learning_rate": 1.1749769016825528e-05, + "loss": 0.0357, + "step": 28136 + }, + { + "epoch": 3.69, + "learning_rate": 1.1749246870869473e-05, + "loss": 0.0412, + "step": 28137 + }, + { + "epoch": 3.69, + "learning_rate": 1.1748724719993755e-05, + "loss": 0.0378, + "step": 28138 + }, + { + "epoch": 3.69, + "learning_rate": 1.1748202564199841e-05, + "loss": 0.0387, + "step": 28139 + }, + { + "epoch": 3.69, + "learning_rate": 1.1747680403489203e-05, + "loss": 0.0435, + "step": 28140 + }, + { + "epoch": 3.69, + "learning_rate": 1.1747158237863308e-05, + "loss": 0.0372, + "step": 28141 + }, + { + "epoch": 3.69, + "learning_rate": 1.1746636067323617e-05, + "loss": 0.0318, + "step": 28142 + }, + { + "epoch": 3.69, + "learning_rate": 1.1746113891871612e-05, + "loss": 0.0277, + "step": 28143 + }, + { + "epoch": 3.69, + "learning_rate": 1.1745591711508748e-05, + "loss": 0.0369, + "step": 28144 + }, + { + "epoch": 3.69, + "learning_rate": 1.1745069526236507e-05, + "loss": 0.0377, + "step": 28145 + }, + { + "epoch": 3.69, + "learning_rate": 1.1744547336056347e-05, + "loss": 0.0422, + "step": 28146 + }, + { + "epoch": 3.69, + "learning_rate": 1.1744025140969742e-05, + "loss": 0.0369, + "step": 28147 + }, + { + "epoch": 3.69, + "learning_rate": 1.1743502940978155e-05, + "loss": 0.0406, + "step": 28148 + }, + { + "epoch": 3.69, + "learning_rate": 1.1742980736083066e-05, + "loss": 0.0365, + "step": 28149 + }, + { + "epoch": 3.69, + "learning_rate": 1.174245852628593e-05, + "loss": 0.0392, + "step": 28150 + }, + { + "epoch": 3.69, + "learning_rate": 1.1741936311588229e-05, + "loss": 0.036, + "step": 28151 + }, + { + "epoch": 3.69, + "learning_rate": 1.174141409199142e-05, + "loss": 0.0363, + "step": 28152 + }, + { + "epoch": 3.69, + "learning_rate": 1.1740891867496975e-05, + "loss": 0.0337, + "step": 28153 + }, + { + "epoch": 3.69, + "learning_rate": 1.1740369638106367e-05, + "loss": 0.0316, + "step": 28154 + }, + { + "epoch": 3.69, + "learning_rate": 1.1739847403821063e-05, + "loss": 0.0292, + "step": 28155 + }, + { + "epoch": 3.69, + "learning_rate": 1.173932516464253e-05, + "loss": 0.0366, + "step": 28156 + }, + { + "epoch": 3.69, + "learning_rate": 1.1738802920572237e-05, + "loss": 0.0425, + "step": 28157 + }, + { + "epoch": 3.69, + "learning_rate": 1.1738280671611654e-05, + "loss": 0.0357, + "step": 28158 + }, + { + "epoch": 3.69, + "learning_rate": 1.1737758417762248e-05, + "loss": 0.0362, + "step": 28159 + }, + { + "epoch": 3.69, + "learning_rate": 1.173723615902549e-05, + "loss": 0.038, + "step": 28160 + }, + { + "epoch": 3.69, + "learning_rate": 1.1736713895402846e-05, + "loss": 0.0358, + "step": 28161 + }, + { + "epoch": 3.69, + "learning_rate": 1.1736191626895793e-05, + "loss": 0.0367, + "step": 28162 + }, + { + "epoch": 3.69, + "learning_rate": 1.1735669353505786e-05, + "loss": 0.0399, + "step": 28163 + }, + { + "epoch": 3.69, + "learning_rate": 1.1735147075234304e-05, + "loss": 0.0349, + "step": 28164 + }, + { + "epoch": 3.69, + "learning_rate": 1.1734624792082813e-05, + "loss": 0.0385, + "step": 28165 + }, + { + "epoch": 3.69, + "learning_rate": 1.1734102504052785e-05, + "loss": 0.0394, + "step": 28166 + }, + { + "epoch": 3.69, + "learning_rate": 1.1733580211145684e-05, + "loss": 0.033, + "step": 28167 + }, + { + "epoch": 3.69, + "learning_rate": 1.173305791336298e-05, + "loss": 0.0422, + "step": 28168 + }, + { + "epoch": 3.69, + "learning_rate": 1.1732535610706144e-05, + "loss": 0.0367, + "step": 28169 + }, + { + "epoch": 3.69, + "learning_rate": 1.1732013303176642e-05, + "loss": 0.0369, + "step": 28170 + }, + { + "epoch": 3.69, + "learning_rate": 1.1731490990775947e-05, + "loss": 0.0361, + "step": 28171 + }, + { + "epoch": 3.69, + "learning_rate": 1.1730968673505525e-05, + "loss": 0.0452, + "step": 28172 + }, + { + "epoch": 3.69, + "learning_rate": 1.1730446351366844e-05, + "loss": 0.0425, + "step": 28173 + }, + { + "epoch": 3.69, + "learning_rate": 1.1729924024361379e-05, + "loss": 0.0353, + "step": 28174 + }, + { + "epoch": 3.69, + "learning_rate": 1.172940169249059e-05, + "loss": 0.0442, + "step": 28175 + }, + { + "epoch": 3.69, + "learning_rate": 1.1728879355755955e-05, + "loss": 0.0374, + "step": 28176 + }, + { + "epoch": 3.69, + "learning_rate": 1.1728357014158935e-05, + "loss": 0.0374, + "step": 28177 + }, + { + "epoch": 3.69, + "learning_rate": 1.1727834667701008e-05, + "loss": 0.0351, + "step": 28178 + }, + { + "epoch": 3.69, + "learning_rate": 1.1727312316383634e-05, + "loss": 0.0427, + "step": 28179 + }, + { + "epoch": 3.69, + "learning_rate": 1.1726789960208286e-05, + "loss": 0.0373, + "step": 28180 + }, + { + "epoch": 3.69, + "learning_rate": 1.1726267599176436e-05, + "loss": 0.0419, + "step": 28181 + }, + { + "epoch": 3.69, + "learning_rate": 1.1725745233289547e-05, + "loss": 0.0346, + "step": 28182 + }, + { + "epoch": 3.69, + "learning_rate": 1.1725222862549094e-05, + "loss": 0.0373, + "step": 28183 + }, + { + "epoch": 3.69, + "learning_rate": 1.1724700486956542e-05, + "loss": 0.0367, + "step": 28184 + }, + { + "epoch": 3.69, + "learning_rate": 1.1724178106513365e-05, + "loss": 0.035, + "step": 28185 + }, + { + "epoch": 3.69, + "learning_rate": 1.1723655721221024e-05, + "loss": 0.0444, + "step": 28186 + }, + { + "epoch": 3.69, + "learning_rate": 1.1723133331080996e-05, + "loss": 0.0252, + "step": 28187 + }, + { + "epoch": 3.69, + "learning_rate": 1.1722610936094744e-05, + "loss": 0.0396, + "step": 28188 + }, + { + "epoch": 3.69, + "learning_rate": 1.1722088536263746e-05, + "loss": 0.0356, + "step": 28189 + }, + { + "epoch": 3.69, + "learning_rate": 1.1721566131589462e-05, + "loss": 0.0332, + "step": 28190 + }, + { + "epoch": 3.69, + "learning_rate": 1.1721043722073366e-05, + "loss": 0.0368, + "step": 28191 + }, + { + "epoch": 3.69, + "learning_rate": 1.1720521307716925e-05, + "loss": 0.0349, + "step": 28192 + }, + { + "epoch": 3.69, + "learning_rate": 1.1719998888521611e-05, + "loss": 0.0451, + "step": 28193 + }, + { + "epoch": 3.69, + "learning_rate": 1.171947646448889e-05, + "loss": 0.0376, + "step": 28194 + }, + { + "epoch": 3.69, + "learning_rate": 1.1718954035620236e-05, + "loss": 0.0362, + "step": 28195 + }, + { + "epoch": 3.69, + "learning_rate": 1.1718431601917112e-05, + "loss": 0.0396, + "step": 28196 + }, + { + "epoch": 3.69, + "learning_rate": 1.1717909163380993e-05, + "loss": 0.038, + "step": 28197 + }, + { + "epoch": 3.69, + "learning_rate": 1.1717386720013344e-05, + "loss": 0.0345, + "step": 28198 + }, + { + "epoch": 3.69, + "learning_rate": 1.171686427181564e-05, + "loss": 0.0319, + "step": 28199 + }, + { + "epoch": 3.69, + "learning_rate": 1.1716341818789342e-05, + "loss": 0.04, + "step": 28200 + }, + { + "epoch": 3.69, + "learning_rate": 1.1715819360935927e-05, + "loss": 0.0345, + "step": 28201 + }, + { + "epoch": 3.69, + "learning_rate": 1.1715296898256858e-05, + "loss": 0.0411, + "step": 28202 + }, + { + "epoch": 3.69, + "learning_rate": 1.1714774430753612e-05, + "loss": 0.0271, + "step": 28203 + }, + { + "epoch": 3.69, + "learning_rate": 1.1714251958427654e-05, + "loss": 0.0345, + "step": 28204 + }, + { + "epoch": 3.69, + "learning_rate": 1.171372948128045e-05, + "loss": 0.0413, + "step": 28205 + }, + { + "epoch": 3.69, + "learning_rate": 1.1713206999313477e-05, + "loss": 0.0377, + "step": 28206 + }, + { + "epoch": 3.69, + "learning_rate": 1.1712684512528198e-05, + "loss": 0.0388, + "step": 28207 + }, + { + "epoch": 3.7, + "learning_rate": 1.1712162020926085e-05, + "loss": 0.0339, + "step": 28208 + }, + { + "epoch": 3.7, + "learning_rate": 1.1711639524508608e-05, + "loss": 0.0335, + "step": 28209 + }, + { + "epoch": 3.7, + "learning_rate": 1.1711117023277238e-05, + "loss": 0.0453, + "step": 28210 + }, + { + "epoch": 3.7, + "learning_rate": 1.171059451723344e-05, + "loss": 0.0382, + "step": 28211 + }, + { + "epoch": 3.7, + "learning_rate": 1.1710072006378688e-05, + "loss": 0.0385, + "step": 28212 + }, + { + "epoch": 3.7, + "learning_rate": 1.1709549490714446e-05, + "loss": 0.0432, + "step": 28213 + }, + { + "epoch": 3.7, + "learning_rate": 1.170902697024219e-05, + "loss": 0.0416, + "step": 28214 + }, + { + "epoch": 3.7, + "learning_rate": 1.1708504444963386e-05, + "loss": 0.0353, + "step": 28215 + }, + { + "epoch": 3.7, + "learning_rate": 1.1707981914879507e-05, + "loss": 0.0433, + "step": 28216 + }, + { + "epoch": 3.7, + "learning_rate": 1.1707459379992016e-05, + "loss": 0.0388, + "step": 28217 + }, + { + "epoch": 3.7, + "learning_rate": 1.170693684030239e-05, + "loss": 0.0387, + "step": 28218 + }, + { + "epoch": 3.7, + "learning_rate": 1.1706414295812088e-05, + "loss": 0.0371, + "step": 28219 + }, + { + "epoch": 3.7, + "learning_rate": 1.1705891746522593e-05, + "loss": 0.0338, + "step": 28220 + }, + { + "epoch": 3.7, + "learning_rate": 1.1705369192435367e-05, + "loss": 0.0387, + "step": 28221 + }, + { + "epoch": 3.7, + "learning_rate": 1.1704846633551879e-05, + "loss": 0.0285, + "step": 28222 + }, + { + "epoch": 3.7, + "learning_rate": 1.1704324069873604e-05, + "loss": 0.0376, + "step": 28223 + }, + { + "epoch": 3.7, + "learning_rate": 1.1703801501402006e-05, + "loss": 0.0391, + "step": 28224 + }, + { + "epoch": 3.7, + "learning_rate": 1.1703278928138558e-05, + "loss": 0.0384, + "step": 28225 + }, + { + "epoch": 3.7, + "learning_rate": 1.1702756350084728e-05, + "loss": 0.0326, + "step": 28226 + }, + { + "epoch": 3.7, + "learning_rate": 1.1702233767241986e-05, + "loss": 0.0466, + "step": 28227 + }, + { + "epoch": 3.7, + "learning_rate": 1.1701711179611804e-05, + "loss": 0.0353, + "step": 28228 + }, + { + "epoch": 3.7, + "learning_rate": 1.170118858719565e-05, + "loss": 0.0321, + "step": 28229 + }, + { + "epoch": 3.7, + "learning_rate": 1.1700665989994987e-05, + "loss": 0.039, + "step": 28230 + }, + { + "epoch": 3.7, + "learning_rate": 1.17001433880113e-05, + "loss": 0.0365, + "step": 28231 + }, + { + "epoch": 3.7, + "learning_rate": 1.1699620781246047e-05, + "loss": 0.0376, + "step": 28232 + }, + { + "epoch": 3.7, + "learning_rate": 1.1699098169700702e-05, + "loss": 0.04, + "step": 28233 + }, + { + "epoch": 3.7, + "learning_rate": 1.1698575553376731e-05, + "loss": 0.0386, + "step": 28234 + }, + { + "epoch": 3.7, + "learning_rate": 1.169805293227561e-05, + "loss": 0.0333, + "step": 28235 + }, + { + "epoch": 3.7, + "learning_rate": 1.1697530306398803e-05, + "loss": 0.0401, + "step": 28236 + }, + { + "epoch": 3.7, + "learning_rate": 1.1697007675747787e-05, + "loss": 0.0292, + "step": 28237 + }, + { + "epoch": 3.7, + "learning_rate": 1.1696485040324021e-05, + "loss": 0.0366, + "step": 28238 + }, + { + "epoch": 3.7, + "learning_rate": 1.1695962400128988e-05, + "loss": 0.031, + "step": 28239 + }, + { + "epoch": 3.7, + "learning_rate": 1.1695439755164143e-05, + "loss": 0.0299, + "step": 28240 + }, + { + "epoch": 3.7, + "learning_rate": 1.1694917105430972e-05, + "loss": 0.0339, + "step": 28241 + }, + { + "epoch": 3.7, + "learning_rate": 1.1694394450930933e-05, + "loss": 0.0327, + "step": 28242 + }, + { + "epoch": 3.7, + "learning_rate": 1.1693871791665501e-05, + "loss": 0.0382, + "step": 28243 + }, + { + "epoch": 3.7, + "learning_rate": 1.1693349127636142e-05, + "loss": 0.0328, + "step": 28244 + }, + { + "epoch": 3.7, + "learning_rate": 1.1692826458844331e-05, + "loss": 0.04, + "step": 28245 + }, + { + "epoch": 3.7, + "learning_rate": 1.1692303785291535e-05, + "loss": 0.0378, + "step": 28246 + }, + { + "epoch": 3.7, + "learning_rate": 1.1691781106979223e-05, + "loss": 0.0368, + "step": 28247 + }, + { + "epoch": 3.7, + "learning_rate": 1.1691258423908871e-05, + "loss": 0.0378, + "step": 28248 + }, + { + "epoch": 3.7, + "learning_rate": 1.1690735736081942e-05, + "loss": 0.0345, + "step": 28249 + }, + { + "epoch": 3.7, + "learning_rate": 1.169021304349991e-05, + "loss": 0.0381, + "step": 28250 + }, + { + "epoch": 3.7, + "learning_rate": 1.168969034616424e-05, + "loss": 0.037, + "step": 28251 + }, + { + "epoch": 3.7, + "learning_rate": 1.168916764407641e-05, + "loss": 0.0411, + "step": 28252 + }, + { + "epoch": 3.7, + "learning_rate": 1.1688644937237886e-05, + "loss": 0.0355, + "step": 28253 + }, + { + "epoch": 3.7, + "learning_rate": 1.1688122225650139e-05, + "loss": 0.0454, + "step": 28254 + }, + { + "epoch": 3.7, + "learning_rate": 1.1687599509314635e-05, + "loss": 0.0358, + "step": 28255 + }, + { + "epoch": 3.7, + "learning_rate": 1.168707678823285e-05, + "loss": 0.0365, + "step": 28256 + }, + { + "epoch": 3.7, + "learning_rate": 1.1686554062406245e-05, + "loss": 0.0321, + "step": 28257 + }, + { + "epoch": 3.7, + "learning_rate": 1.1686031331836305e-05, + "loss": 0.0391, + "step": 28258 + }, + { + "epoch": 3.7, + "learning_rate": 1.1685508596524487e-05, + "loss": 0.0414, + "step": 28259 + }, + { + "epoch": 3.7, + "learning_rate": 1.1684985856472269e-05, + "loss": 0.0433, + "step": 28260 + }, + { + "epoch": 3.7, + "learning_rate": 1.1684463111681115e-05, + "loss": 0.0405, + "step": 28261 + }, + { + "epoch": 3.7, + "learning_rate": 1.1683940362152497e-05, + "loss": 0.0402, + "step": 28262 + }, + { + "epoch": 3.7, + "learning_rate": 1.1683417607887888e-05, + "loss": 0.0362, + "step": 28263 + }, + { + "epoch": 3.7, + "learning_rate": 1.1682894848888759e-05, + "loss": 0.038, + "step": 28264 + }, + { + "epoch": 3.7, + "learning_rate": 1.1682372085156574e-05, + "loss": 0.0328, + "step": 28265 + }, + { + "epoch": 3.7, + "learning_rate": 1.168184931669281e-05, + "loss": 0.0317, + "step": 28266 + }, + { + "epoch": 3.7, + "learning_rate": 1.1681326543498933e-05, + "loss": 0.0339, + "step": 28267 + }, + { + "epoch": 3.7, + "learning_rate": 1.1680803765576416e-05, + "loss": 0.035, + "step": 28268 + }, + { + "epoch": 3.7, + "learning_rate": 1.1680280982926727e-05, + "loss": 0.046, + "step": 28269 + }, + { + "epoch": 3.7, + "learning_rate": 1.1679758195551337e-05, + "loss": 0.0419, + "step": 28270 + }, + { + "epoch": 3.7, + "learning_rate": 1.1679235403451716e-05, + "loss": 0.0435, + "step": 28271 + }, + { + "epoch": 3.7, + "learning_rate": 1.1678712606629335e-05, + "loss": 0.0378, + "step": 28272 + }, + { + "epoch": 3.7, + "learning_rate": 1.1678189805085664e-05, + "loss": 0.0345, + "step": 28273 + }, + { + "epoch": 3.7, + "learning_rate": 1.1677666998822175e-05, + "loss": 0.0387, + "step": 28274 + }, + { + "epoch": 3.7, + "learning_rate": 1.1677144187840336e-05, + "loss": 0.0343, + "step": 28275 + }, + { + "epoch": 3.7, + "learning_rate": 1.1676621372141618e-05, + "loss": 0.0333, + "step": 28276 + }, + { + "epoch": 3.7, + "learning_rate": 1.1676098551727493e-05, + "loss": 0.0339, + "step": 28277 + }, + { + "epoch": 3.7, + "learning_rate": 1.1675575726599428e-05, + "loss": 0.0406, + "step": 28278 + }, + { + "epoch": 3.7, + "learning_rate": 1.1675052896758899e-05, + "loss": 0.033, + "step": 28279 + }, + { + "epoch": 3.7, + "learning_rate": 1.1674530062207367e-05, + "loss": 0.0373, + "step": 28280 + }, + { + "epoch": 3.7, + "learning_rate": 1.1674007222946315e-05, + "loss": 0.0366, + "step": 28281 + }, + { + "epoch": 3.7, + "learning_rate": 1.1673484378977202e-05, + "loss": 0.0377, + "step": 28282 + }, + { + "epoch": 3.7, + "learning_rate": 1.1672961530301505e-05, + "loss": 0.0394, + "step": 28283 + }, + { + "epoch": 3.71, + "learning_rate": 1.1672438676920692e-05, + "loss": 0.0379, + "step": 28284 + }, + { + "epoch": 3.71, + "learning_rate": 1.1671915818836237e-05, + "loss": 0.0429, + "step": 28285 + }, + { + "epoch": 3.71, + "learning_rate": 1.1671392956049606e-05, + "loss": 0.0297, + "step": 28286 + }, + { + "epoch": 3.71, + "learning_rate": 1.167087008856227e-05, + "loss": 0.0362, + "step": 28287 + }, + { + "epoch": 3.71, + "learning_rate": 1.1670347216375704e-05, + "loss": 0.0398, + "step": 28288 + }, + { + "epoch": 3.71, + "learning_rate": 1.1669824339491373e-05, + "loss": 0.0371, + "step": 28289 + }, + { + "epoch": 3.71, + "learning_rate": 1.1669301457910751e-05, + "loss": 0.0346, + "step": 28290 + }, + { + "epoch": 3.71, + "learning_rate": 1.1668778571635305e-05, + "loss": 0.0322, + "step": 28291 + }, + { + "epoch": 3.71, + "learning_rate": 1.1668255680666511e-05, + "loss": 0.0336, + "step": 28292 + }, + { + "epoch": 3.71, + "learning_rate": 1.1667732785005837e-05, + "loss": 0.0413, + "step": 28293 + }, + { + "epoch": 3.71, + "learning_rate": 1.1667209884654753e-05, + "loss": 0.0433, + "step": 28294 + }, + { + "epoch": 3.71, + "learning_rate": 1.1666686979614728e-05, + "loss": 0.0355, + "step": 28295 + }, + { + "epoch": 3.71, + "learning_rate": 1.1666164069887235e-05, + "loss": 0.0379, + "step": 28296 + }, + { + "epoch": 3.71, + "learning_rate": 1.1665641155473746e-05, + "loss": 0.0438, + "step": 28297 + }, + { + "epoch": 3.71, + "learning_rate": 1.166511823637573e-05, + "loss": 0.0383, + "step": 28298 + }, + { + "epoch": 3.71, + "learning_rate": 1.1664595312594656e-05, + "loss": 0.0374, + "step": 28299 + }, + { + "epoch": 3.71, + "learning_rate": 1.1664072384131999e-05, + "loss": 0.044, + "step": 28300 + }, + { + "epoch": 3.71, + "learning_rate": 1.1663549450989222e-05, + "loss": 0.0403, + "step": 28301 + }, + { + "epoch": 3.71, + "learning_rate": 1.1663026513167806e-05, + "loss": 0.0387, + "step": 28302 + }, + { + "epoch": 3.71, + "learning_rate": 1.1662503570669213e-05, + "loss": 0.0391, + "step": 28303 + }, + { + "epoch": 3.71, + "learning_rate": 1.166198062349492e-05, + "loss": 0.0326, + "step": 28304 + }, + { + "epoch": 3.71, + "learning_rate": 1.1661457671646394e-05, + "loss": 0.0399, + "step": 28305 + }, + { + "epoch": 3.71, + "learning_rate": 1.1660934715125105e-05, + "loss": 0.0364, + "step": 28306 + }, + { + "epoch": 3.71, + "learning_rate": 1.1660411753932528e-05, + "loss": 0.0376, + "step": 28307 + }, + { + "epoch": 3.71, + "learning_rate": 1.1659888788070129e-05, + "loss": 0.0404, + "step": 28308 + }, + { + "epoch": 3.71, + "learning_rate": 1.1659365817539383e-05, + "loss": 0.038, + "step": 28309 + }, + { + "epoch": 3.71, + "learning_rate": 1.1658842842341758e-05, + "loss": 0.0407, + "step": 28310 + }, + { + "epoch": 3.71, + "learning_rate": 1.1658319862478727e-05, + "loss": 0.0339, + "step": 28311 + }, + { + "epoch": 3.71, + "learning_rate": 1.1657796877951757e-05, + "loss": 0.0341, + "step": 28312 + }, + { + "epoch": 3.71, + "learning_rate": 1.1657273888762326e-05, + "loss": 0.0395, + "step": 28313 + }, + { + "epoch": 3.71, + "learning_rate": 1.1656750894911896e-05, + "loss": 0.0371, + "step": 28314 + }, + { + "epoch": 3.71, + "learning_rate": 1.1656227896401947e-05, + "loss": 0.0313, + "step": 28315 + }, + { + "epoch": 3.71, + "learning_rate": 1.1655704893233941e-05, + "loss": 0.0407, + "step": 28316 + }, + { + "epoch": 3.71, + "learning_rate": 1.1655181885409355e-05, + "loss": 0.0311, + "step": 28317 + }, + { + "epoch": 3.71, + "learning_rate": 1.1654658872929655e-05, + "loss": 0.0422, + "step": 28318 + }, + { + "epoch": 3.71, + "learning_rate": 1.1654135855796323e-05, + "loss": 0.0296, + "step": 28319 + }, + { + "epoch": 3.71, + "learning_rate": 1.1653612834010814e-05, + "loss": 0.0435, + "step": 28320 + }, + { + "epoch": 3.71, + "learning_rate": 1.165308980757461e-05, + "loss": 0.0416, + "step": 28321 + }, + { + "epoch": 3.71, + "learning_rate": 1.1652566776489178e-05, + "loss": 0.0319, + "step": 28322 + }, + { + "epoch": 3.71, + "learning_rate": 1.1652043740755992e-05, + "loss": 0.0431, + "step": 28323 + }, + { + "epoch": 3.71, + "learning_rate": 1.1651520700376519e-05, + "loss": 0.0414, + "step": 28324 + }, + { + "epoch": 3.71, + "learning_rate": 1.1650997655352235e-05, + "loss": 0.0428, + "step": 28325 + }, + { + "epoch": 3.71, + "learning_rate": 1.1650474605684606e-05, + "loss": 0.0324, + "step": 28326 + }, + { + "epoch": 3.71, + "learning_rate": 1.1649951551375102e-05, + "loss": 0.034, + "step": 28327 + }, + { + "epoch": 3.71, + "learning_rate": 1.1649428492425201e-05, + "loss": 0.0351, + "step": 28328 + }, + { + "epoch": 3.71, + "learning_rate": 1.164890542883637e-05, + "loss": 0.0428, + "step": 28329 + }, + { + "epoch": 3.71, + "learning_rate": 1.164838236061008e-05, + "loss": 0.0334, + "step": 28330 + }, + { + "epoch": 3.71, + "learning_rate": 1.16478592877478e-05, + "loss": 0.0378, + "step": 28331 + }, + { + "epoch": 3.71, + "learning_rate": 1.1647336210251007e-05, + "loss": 0.0415, + "step": 28332 + }, + { + "epoch": 3.71, + "learning_rate": 1.1646813128121164e-05, + "loss": 0.0306, + "step": 28333 + }, + { + "epoch": 3.71, + "learning_rate": 1.1646290041359751e-05, + "loss": 0.0342, + "step": 28334 + }, + { + "epoch": 3.71, + "learning_rate": 1.1645766949968233e-05, + "loss": 0.0378, + "step": 28335 + }, + { + "epoch": 3.71, + "learning_rate": 1.1645243853948084e-05, + "loss": 0.0448, + "step": 28336 + }, + { + "epoch": 3.71, + "learning_rate": 1.1644720753300776e-05, + "loss": 0.0313, + "step": 28337 + }, + { + "epoch": 3.71, + "learning_rate": 1.1644197648027777e-05, + "loss": 0.0322, + "step": 28338 + }, + { + "epoch": 3.71, + "learning_rate": 1.1643674538130555e-05, + "loss": 0.0369, + "step": 28339 + }, + { + "epoch": 3.71, + "learning_rate": 1.1643151423610592e-05, + "loss": 0.0359, + "step": 28340 + }, + { + "epoch": 3.71, + "learning_rate": 1.1642628304469351e-05, + "loss": 0.0334, + "step": 28341 + }, + { + "epoch": 3.71, + "learning_rate": 1.1642105180708308e-05, + "loss": 0.0346, + "step": 28342 + }, + { + "epoch": 3.71, + "learning_rate": 1.1641582052328928e-05, + "loss": 0.0422, + "step": 28343 + }, + { + "epoch": 3.71, + "learning_rate": 1.1641058919332686e-05, + "loss": 0.0346, + "step": 28344 + }, + { + "epoch": 3.71, + "learning_rate": 1.1640535781721054e-05, + "loss": 0.0367, + "step": 28345 + }, + { + "epoch": 3.71, + "learning_rate": 1.1640012639495504e-05, + "loss": 0.0338, + "step": 28346 + }, + { + "epoch": 3.71, + "learning_rate": 1.1639489492657506e-05, + "loss": 0.0398, + "step": 28347 + }, + { + "epoch": 3.71, + "learning_rate": 1.1638966341208526e-05, + "loss": 0.0349, + "step": 28348 + }, + { + "epoch": 3.71, + "learning_rate": 1.1638443185150048e-05, + "loss": 0.0372, + "step": 28349 + }, + { + "epoch": 3.71, + "learning_rate": 1.1637920024483532e-05, + "loss": 0.0343, + "step": 28350 + }, + { + "epoch": 3.71, + "learning_rate": 1.163739685921045e-05, + "loss": 0.0434, + "step": 28351 + }, + { + "epoch": 3.71, + "learning_rate": 1.163687368933228e-05, + "loss": 0.0311, + "step": 28352 + }, + { + "epoch": 3.71, + "learning_rate": 1.1636350514850489e-05, + "loss": 0.037, + "step": 28353 + }, + { + "epoch": 3.71, + "learning_rate": 1.1635827335766549e-05, + "loss": 0.0321, + "step": 28354 + }, + { + "epoch": 3.71, + "learning_rate": 1.1635304152081934e-05, + "loss": 0.0359, + "step": 28355 + }, + { + "epoch": 3.71, + "learning_rate": 1.163478096379811e-05, + "loss": 0.0377, + "step": 28356 + }, + { + "epoch": 3.71, + "learning_rate": 1.1634257770916555e-05, + "loss": 0.0274, + "step": 28357 + }, + { + "epoch": 3.71, + "learning_rate": 1.1633734573438735e-05, + "loss": 0.0409, + "step": 28358 + }, + { + "epoch": 3.71, + "learning_rate": 1.1633211371366125e-05, + "loss": 0.0401, + "step": 28359 + }, + { + "epoch": 3.71, + "learning_rate": 1.1632688164700192e-05, + "loss": 0.0402, + "step": 28360 + }, + { + "epoch": 3.72, + "learning_rate": 1.1632164953442413e-05, + "loss": 0.0351, + "step": 28361 + }, + { + "epoch": 3.72, + "learning_rate": 1.1631641737594255e-05, + "loss": 0.0363, + "step": 28362 + }, + { + "epoch": 3.72, + "learning_rate": 1.1631118517157196e-05, + "loss": 0.0358, + "step": 28363 + }, + { + "epoch": 3.72, + "learning_rate": 1.1630595292132698e-05, + "loss": 0.0395, + "step": 28364 + }, + { + "epoch": 3.72, + "learning_rate": 1.1630072062522242e-05, + "loss": 0.0402, + "step": 28365 + }, + { + "epoch": 3.72, + "learning_rate": 1.162954882832729e-05, + "loss": 0.0341, + "step": 28366 + }, + { + "epoch": 3.72, + "learning_rate": 1.162902558954932e-05, + "loss": 0.0359, + "step": 28367 + }, + { + "epoch": 3.72, + "learning_rate": 1.1628502346189805e-05, + "loss": 0.0314, + "step": 28368 + }, + { + "epoch": 3.72, + "learning_rate": 1.1627979098250213e-05, + "loss": 0.0409, + "step": 28369 + }, + { + "epoch": 3.72, + "learning_rate": 1.1627455845732017e-05, + "loss": 0.0377, + "step": 28370 + }, + { + "epoch": 3.72, + "learning_rate": 1.1626932588636686e-05, + "loss": 0.0379, + "step": 28371 + }, + { + "epoch": 3.72, + "learning_rate": 1.1626409326965692e-05, + "loss": 0.0357, + "step": 28372 + }, + { + "epoch": 3.72, + "learning_rate": 1.1625886060720512e-05, + "loss": 0.0371, + "step": 28373 + }, + { + "epoch": 3.72, + "learning_rate": 1.1625362789902615e-05, + "loss": 0.0366, + "step": 28374 + }, + { + "epoch": 3.72, + "learning_rate": 1.1624839514513468e-05, + "loss": 0.0309, + "step": 28375 + }, + { + "epoch": 3.72, + "learning_rate": 1.1624316234554548e-05, + "loss": 0.0381, + "step": 28376 + }, + { + "epoch": 3.72, + "learning_rate": 1.1623792950027324e-05, + "loss": 0.0359, + "step": 28377 + }, + { + "epoch": 3.72, + "learning_rate": 1.1623269660933273e-05, + "loss": 0.032, + "step": 28378 + }, + { + "epoch": 3.72, + "learning_rate": 1.1622746367273857e-05, + "loss": 0.0328, + "step": 28379 + }, + { + "epoch": 3.72, + "learning_rate": 1.1622223069050558e-05, + "loss": 0.0439, + "step": 28380 + }, + { + "epoch": 3.72, + "learning_rate": 1.162169976626484e-05, + "loss": 0.0313, + "step": 28381 + }, + { + "epoch": 3.72, + "learning_rate": 1.162117645891818e-05, + "loss": 0.0408, + "step": 28382 + }, + { + "epoch": 3.72, + "learning_rate": 1.1620653147012045e-05, + "loss": 0.0362, + "step": 28383 + }, + { + "epoch": 3.72, + "learning_rate": 1.1620129830547913e-05, + "loss": 0.0263, + "step": 28384 + }, + { + "epoch": 3.72, + "learning_rate": 1.1619606509527248e-05, + "loss": 0.0325, + "step": 28385 + }, + { + "epoch": 3.72, + "learning_rate": 1.161908318395153e-05, + "loss": 0.0387, + "step": 28386 + }, + { + "epoch": 3.72, + "learning_rate": 1.1618559853822222e-05, + "loss": 0.0448, + "step": 28387 + }, + { + "epoch": 3.72, + "learning_rate": 1.1618036519140803e-05, + "loss": 0.0331, + "step": 28388 + }, + { + "epoch": 3.72, + "learning_rate": 1.1617513179908742e-05, + "loss": 0.038, + "step": 28389 + }, + { + "epoch": 3.72, + "learning_rate": 1.1616989836127512e-05, + "loss": 0.0337, + "step": 28390 + }, + { + "epoch": 3.72, + "learning_rate": 1.1616466487798583e-05, + "loss": 0.0338, + "step": 28391 + }, + { + "epoch": 3.72, + "learning_rate": 1.1615943134923429e-05, + "loss": 0.0452, + "step": 28392 + }, + { + "epoch": 3.72, + "learning_rate": 1.161541977750352e-05, + "loss": 0.0335, + "step": 28393 + }, + { + "epoch": 3.72, + "learning_rate": 1.1614896415540334e-05, + "loss": 0.0371, + "step": 28394 + }, + { + "epoch": 3.72, + "learning_rate": 1.161437304903533e-05, + "loss": 0.0335, + "step": 28395 + }, + { + "epoch": 3.72, + "learning_rate": 1.1613849677989992e-05, + "loss": 0.0378, + "step": 28396 + }, + { + "epoch": 3.72, + "learning_rate": 1.1613326302405789e-05, + "loss": 0.0387, + "step": 28397 + }, + { + "epoch": 3.72, + "learning_rate": 1.161280292228419e-05, + "loss": 0.0339, + "step": 28398 + }, + { + "epoch": 3.72, + "learning_rate": 1.1612279537626667e-05, + "loss": 0.0364, + "step": 28399 + }, + { + "epoch": 3.72, + "learning_rate": 1.1611756148434695e-05, + "loss": 0.0378, + "step": 28400 + }, + { + "epoch": 3.72, + "learning_rate": 1.1611232754709747e-05, + "loss": 0.0373, + "step": 28401 + }, + { + "epoch": 3.72, + "learning_rate": 1.1610709356453289e-05, + "loss": 0.0381, + "step": 28402 + }, + { + "epoch": 3.72, + "learning_rate": 1.16101859536668e-05, + "loss": 0.039, + "step": 28403 + }, + { + "epoch": 3.72, + "learning_rate": 1.1609662546351746e-05, + "loss": 0.0425, + "step": 28404 + }, + { + "epoch": 3.72, + "learning_rate": 1.1609139134509602e-05, + "loss": 0.0379, + "step": 28405 + }, + { + "epoch": 3.72, + "learning_rate": 1.160861571814184e-05, + "loss": 0.0375, + "step": 28406 + }, + { + "epoch": 3.72, + "learning_rate": 1.1608092297249936e-05, + "loss": 0.0389, + "step": 28407 + }, + { + "epoch": 3.72, + "learning_rate": 1.1607568871835352e-05, + "loss": 0.0336, + "step": 28408 + }, + { + "epoch": 3.72, + "learning_rate": 1.1607045441899571e-05, + "loss": 0.0427, + "step": 28409 + }, + { + "epoch": 3.72, + "learning_rate": 1.1606522007444056e-05, + "loss": 0.0297, + "step": 28410 + }, + { + "epoch": 3.72, + "learning_rate": 1.1605998568470287e-05, + "loss": 0.039, + "step": 28411 + }, + { + "epoch": 3.72, + "learning_rate": 1.1605475124979731e-05, + "loss": 0.0425, + "step": 28412 + }, + { + "epoch": 3.72, + "learning_rate": 1.1604951676973862e-05, + "loss": 0.0325, + "step": 28413 + }, + { + "epoch": 3.72, + "learning_rate": 1.1604428224454151e-05, + "loss": 0.0453, + "step": 28414 + }, + { + "epoch": 3.72, + "learning_rate": 1.1603904767422072e-05, + "loss": 0.04, + "step": 28415 + }, + { + "epoch": 3.72, + "learning_rate": 1.1603381305879096e-05, + "loss": 0.0272, + "step": 28416 + }, + { + "epoch": 3.72, + "learning_rate": 1.1602857839826695e-05, + "loss": 0.0391, + "step": 28417 + }, + { + "epoch": 3.72, + "learning_rate": 1.1602334369266343e-05, + "loss": 0.0359, + "step": 28418 + }, + { + "epoch": 3.72, + "learning_rate": 1.160181089419951e-05, + "loss": 0.0455, + "step": 28419 + }, + { + "epoch": 3.72, + "learning_rate": 1.160128741462767e-05, + "loss": 0.0406, + "step": 28420 + }, + { + "epoch": 3.72, + "learning_rate": 1.1600763930552293e-05, + "loss": 0.0441, + "step": 28421 + }, + { + "epoch": 3.72, + "learning_rate": 1.1600240441974852e-05, + "loss": 0.0343, + "step": 28422 + }, + { + "epoch": 3.72, + "learning_rate": 1.1599716948896822e-05, + "loss": 0.0397, + "step": 28423 + }, + { + "epoch": 3.72, + "learning_rate": 1.1599193451319674e-05, + "loss": 0.0367, + "step": 28424 + }, + { + "epoch": 3.72, + "learning_rate": 1.1598669949244879e-05, + "loss": 0.0373, + "step": 28425 + }, + { + "epoch": 3.72, + "learning_rate": 1.1598146442673908e-05, + "loss": 0.0407, + "step": 28426 + }, + { + "epoch": 3.72, + "learning_rate": 1.1597622931608234e-05, + "loss": 0.0367, + "step": 28427 + }, + { + "epoch": 3.72, + "learning_rate": 1.1597099416049336e-05, + "loss": 0.0312, + "step": 28428 + }, + { + "epoch": 3.72, + "learning_rate": 1.1596575895998677e-05, + "loss": 0.0377, + "step": 28429 + }, + { + "epoch": 3.72, + "learning_rate": 1.1596052371457735e-05, + "loss": 0.0357, + "step": 28430 + }, + { + "epoch": 3.72, + "learning_rate": 1.1595528842427979e-05, + "loss": 0.0406, + "step": 28431 + }, + { + "epoch": 3.72, + "learning_rate": 1.1595005308910885e-05, + "loss": 0.0333, + "step": 28432 + }, + { + "epoch": 3.72, + "learning_rate": 1.1594481770907921e-05, + "loss": 0.0333, + "step": 28433 + }, + { + "epoch": 3.72, + "learning_rate": 1.1593958228420565e-05, + "loss": 0.0374, + "step": 28434 + }, + { + "epoch": 3.72, + "learning_rate": 1.1593434681450284e-05, + "loss": 0.0341, + "step": 28435 + }, + { + "epoch": 3.72, + "learning_rate": 1.1592911129998552e-05, + "loss": 0.0369, + "step": 28436 + }, + { + "epoch": 3.73, + "learning_rate": 1.1592387574066843e-05, + "loss": 0.0395, + "step": 28437 + }, + { + "epoch": 3.73, + "learning_rate": 1.159186401365663e-05, + "loss": 0.0425, + "step": 28438 + }, + { + "epoch": 3.73, + "learning_rate": 1.1591340448769385e-05, + "loss": 0.0347, + "step": 28439 + }, + { + "epoch": 3.73, + "learning_rate": 1.1590816879406578e-05, + "loss": 0.0306, + "step": 28440 + }, + { + "epoch": 3.73, + "learning_rate": 1.1590293305569686e-05, + "loss": 0.0358, + "step": 28441 + }, + { + "epoch": 3.73, + "learning_rate": 1.1589769727260173e-05, + "loss": 0.0449, + "step": 28442 + }, + { + "epoch": 3.73, + "learning_rate": 1.1589246144479523e-05, + "loss": 0.0412, + "step": 28443 + }, + { + "epoch": 3.73, + "learning_rate": 1.15887225572292e-05, + "loss": 0.0424, + "step": 28444 + }, + { + "epoch": 3.73, + "learning_rate": 1.1588198965510683e-05, + "loss": 0.0362, + "step": 28445 + }, + { + "epoch": 3.73, + "learning_rate": 1.1587675369325436e-05, + "loss": 0.0373, + "step": 28446 + }, + { + "epoch": 3.73, + "learning_rate": 1.158715176867494e-05, + "loss": 0.0334, + "step": 28447 + }, + { + "epoch": 3.73, + "learning_rate": 1.158662816356066e-05, + "loss": 0.0373, + "step": 28448 + }, + { + "epoch": 3.73, + "learning_rate": 1.158610455398408e-05, + "loss": 0.0387, + "step": 28449 + }, + { + "epoch": 3.73, + "learning_rate": 1.158558093994666e-05, + "loss": 0.0404, + "step": 28450 + }, + { + "epoch": 3.73, + "learning_rate": 1.158505732144988e-05, + "loss": 0.0409, + "step": 28451 + }, + { + "epoch": 3.73, + "learning_rate": 1.158453369849521e-05, + "loss": 0.0339, + "step": 28452 + }, + { + "epoch": 3.73, + "learning_rate": 1.158401007108412e-05, + "loss": 0.0371, + "step": 28453 + }, + { + "epoch": 3.73, + "learning_rate": 1.158348643921809e-05, + "loss": 0.0416, + "step": 28454 + }, + { + "epoch": 3.73, + "learning_rate": 1.158296280289859e-05, + "loss": 0.0345, + "step": 28455 + }, + { + "epoch": 3.73, + "learning_rate": 1.1582439162127088e-05, + "loss": 0.031, + "step": 28456 + }, + { + "epoch": 3.73, + "learning_rate": 1.158191551690506e-05, + "loss": 0.0313, + "step": 28457 + }, + { + "epoch": 3.73, + "learning_rate": 1.1581391867233983e-05, + "loss": 0.0386, + "step": 28458 + }, + { + "epoch": 3.73, + "learning_rate": 1.1580868213115321e-05, + "loss": 0.0368, + "step": 28459 + }, + { + "epoch": 3.73, + "learning_rate": 1.1580344554550553e-05, + "loss": 0.0371, + "step": 28460 + }, + { + "epoch": 3.73, + "learning_rate": 1.157982089154115e-05, + "loss": 0.032, + "step": 28461 + }, + { + "epoch": 3.73, + "learning_rate": 1.1579297224088586e-05, + "loss": 0.048, + "step": 28462 + }, + { + "epoch": 3.73, + "learning_rate": 1.1578773552194332e-05, + "loss": 0.0362, + "step": 28463 + }, + { + "epoch": 3.73, + "learning_rate": 1.157824987585986e-05, + "loss": 0.0358, + "step": 28464 + }, + { + "epoch": 3.73, + "learning_rate": 1.1577726195086645e-05, + "loss": 0.0405, + "step": 28465 + }, + { + "epoch": 3.73, + "learning_rate": 1.1577202509876162e-05, + "loss": 0.0336, + "step": 28466 + }, + { + "epoch": 3.73, + "learning_rate": 1.1576678820229876e-05, + "loss": 0.0341, + "step": 28467 + }, + { + "epoch": 3.73, + "learning_rate": 1.157615512614927e-05, + "loss": 0.0329, + "step": 28468 + }, + { + "epoch": 3.73, + "learning_rate": 1.157563142763581e-05, + "loss": 0.0359, + "step": 28469 + }, + { + "epoch": 3.73, + "learning_rate": 1.1575107724690969e-05, + "loss": 0.037, + "step": 28470 + }, + { + "epoch": 3.73, + "learning_rate": 1.1574584017316219e-05, + "loss": 0.0317, + "step": 28471 + }, + { + "epoch": 3.73, + "learning_rate": 1.157406030551304e-05, + "loss": 0.0333, + "step": 28472 + }, + { + "epoch": 3.73, + "learning_rate": 1.1573536589282899e-05, + "loss": 0.0303, + "step": 28473 + }, + { + "epoch": 3.73, + "learning_rate": 1.157301286862727e-05, + "loss": 0.0407, + "step": 28474 + }, + { + "epoch": 3.73, + "learning_rate": 1.1572489143547624e-05, + "loss": 0.0419, + "step": 28475 + }, + { + "epoch": 3.73, + "learning_rate": 1.1571965414045439e-05, + "loss": 0.0417, + "step": 28476 + }, + { + "epoch": 3.73, + "learning_rate": 1.1571441680122182e-05, + "loss": 0.0404, + "step": 28477 + }, + { + "epoch": 3.73, + "learning_rate": 1.1570917941779332e-05, + "loss": 0.0359, + "step": 28478 + }, + { + "epoch": 3.73, + "learning_rate": 1.1570394199018357e-05, + "loss": 0.0391, + "step": 28479 + }, + { + "epoch": 3.73, + "learning_rate": 1.1569870451840732e-05, + "loss": 0.0428, + "step": 28480 + }, + { + "epoch": 3.73, + "learning_rate": 1.156934670024793e-05, + "loss": 0.042, + "step": 28481 + }, + { + "epoch": 3.73, + "learning_rate": 1.1568822944241425e-05, + "loss": 0.0407, + "step": 28482 + }, + { + "epoch": 3.73, + "learning_rate": 1.1568299183822688e-05, + "loss": 0.0329, + "step": 28483 + }, + { + "epoch": 3.73, + "learning_rate": 1.1567775418993195e-05, + "loss": 0.0375, + "step": 28484 + }, + { + "epoch": 3.73, + "learning_rate": 1.1567251649754416e-05, + "loss": 0.035, + "step": 28485 + }, + { + "epoch": 3.73, + "learning_rate": 1.1566727876107824e-05, + "loss": 0.0351, + "step": 28486 + }, + { + "epoch": 3.73, + "learning_rate": 1.1566204098054894e-05, + "loss": 0.039, + "step": 28487 + }, + { + "epoch": 3.73, + "learning_rate": 1.1565680315597099e-05, + "loss": 0.0401, + "step": 28488 + }, + { + "epoch": 3.73, + "learning_rate": 1.1565156528735913e-05, + "loss": 0.0293, + "step": 28489 + }, + { + "epoch": 3.73, + "learning_rate": 1.1564632737472805e-05, + "loss": 0.0401, + "step": 28490 + }, + { + "epoch": 3.73, + "learning_rate": 1.1564108941809251e-05, + "loss": 0.0371, + "step": 28491 + }, + { + "epoch": 3.73, + "learning_rate": 1.1563585141746723e-05, + "loss": 0.0437, + "step": 28492 + }, + { + "epoch": 3.73, + "learning_rate": 1.1563061337286696e-05, + "loss": 0.0336, + "step": 28493 + }, + { + "epoch": 3.73, + "learning_rate": 1.1562537528430645e-05, + "loss": 0.0358, + "step": 28494 + }, + { + "epoch": 3.73, + "learning_rate": 1.1562013715180039e-05, + "loss": 0.0327, + "step": 28495 + }, + { + "epoch": 3.73, + "learning_rate": 1.156148989753635e-05, + "loss": 0.0317, + "step": 28496 + }, + { + "epoch": 3.73, + "learning_rate": 1.1560966075501055e-05, + "loss": 0.0361, + "step": 28497 + }, + { + "epoch": 3.73, + "learning_rate": 1.1560442249075625e-05, + "loss": 0.0337, + "step": 28498 + }, + { + "epoch": 3.73, + "learning_rate": 1.1559918418261536e-05, + "loss": 0.0416, + "step": 28499 + }, + { + "epoch": 3.73, + "learning_rate": 1.1559394583060259e-05, + "loss": 0.0404, + "step": 28500 + }, + { + "epoch": 3.73, + "learning_rate": 1.155887074347327e-05, + "loss": 0.0354, + "step": 28501 + }, + { + "epoch": 3.73, + "learning_rate": 1.1558346899502037e-05, + "loss": 0.0377, + "step": 28502 + }, + { + "epoch": 3.73, + "learning_rate": 1.1557823051148036e-05, + "loss": 0.0373, + "step": 28503 + }, + { + "epoch": 3.73, + "learning_rate": 1.1557299198412744e-05, + "loss": 0.0377, + "step": 28504 + }, + { + "epoch": 3.73, + "learning_rate": 1.1556775341297627e-05, + "loss": 0.0388, + "step": 28505 + }, + { + "epoch": 3.73, + "learning_rate": 1.1556251479804165e-05, + "loss": 0.0385, + "step": 28506 + }, + { + "epoch": 3.73, + "learning_rate": 1.155572761393383e-05, + "loss": 0.0395, + "step": 28507 + }, + { + "epoch": 3.73, + "learning_rate": 1.1555203743688091e-05, + "loss": 0.0414, + "step": 28508 + }, + { + "epoch": 3.73, + "learning_rate": 1.1554679869068424e-05, + "loss": 0.0255, + "step": 28509 + }, + { + "epoch": 3.73, + "learning_rate": 1.1554155990076304e-05, + "loss": 0.0456, + "step": 28510 + }, + { + "epoch": 3.73, + "learning_rate": 1.1553632106713202e-05, + "loss": 0.0355, + "step": 28511 + }, + { + "epoch": 3.73, + "learning_rate": 1.1553108218980595e-05, + "loss": 0.0386, + "step": 28512 + }, + { + "epoch": 3.74, + "learning_rate": 1.155258432687995e-05, + "loss": 0.0293, + "step": 28513 + }, + { + "epoch": 3.74, + "learning_rate": 1.1552060430412746e-05, + "loss": 0.0351, + "step": 28514 + }, + { + "epoch": 3.74, + "learning_rate": 1.1551536529580453e-05, + "loss": 0.0413, + "step": 28515 + }, + { + "epoch": 3.74, + "learning_rate": 1.1551012624384551e-05, + "loss": 0.0326, + "step": 28516 + }, + { + "epoch": 3.74, + "learning_rate": 1.1550488714826506e-05, + "loss": 0.0379, + "step": 28517 + }, + { + "epoch": 3.74, + "learning_rate": 1.1549964800907793e-05, + "loss": 0.0353, + "step": 28518 + }, + { + "epoch": 3.74, + "learning_rate": 1.1549440882629886e-05, + "loss": 0.0383, + "step": 28519 + }, + { + "epoch": 3.74, + "learning_rate": 1.1548916959994263e-05, + "loss": 0.0362, + "step": 28520 + }, + { + "epoch": 3.74, + "learning_rate": 1.154839303300239e-05, + "loss": 0.0357, + "step": 28521 + }, + { + "epoch": 3.74, + "learning_rate": 1.1547869101655745e-05, + "loss": 0.0366, + "step": 28522 + }, + { + "epoch": 3.74, + "learning_rate": 1.1547345165955801e-05, + "loss": 0.0327, + "step": 28523 + }, + { + "epoch": 3.74, + "learning_rate": 1.1546821225904028e-05, + "loss": 0.0359, + "step": 28524 + }, + { + "epoch": 3.74, + "learning_rate": 1.1546297281501906e-05, + "loss": 0.0357, + "step": 28525 + }, + { + "epoch": 3.74, + "learning_rate": 1.1545773332750905e-05, + "loss": 0.0362, + "step": 28526 + }, + { + "epoch": 3.74, + "learning_rate": 1.15452493796525e-05, + "loss": 0.0335, + "step": 28527 + }, + { + "epoch": 3.74, + "learning_rate": 1.154472542220816e-05, + "loss": 0.0381, + "step": 28528 + }, + { + "epoch": 3.74, + "learning_rate": 1.1544201460419364e-05, + "loss": 0.0446, + "step": 28529 + }, + { + "epoch": 3.74, + "learning_rate": 1.1543677494287582e-05, + "loss": 0.0415, + "step": 28530 + }, + { + "epoch": 3.74, + "learning_rate": 1.1543153523814292e-05, + "loss": 0.0393, + "step": 28531 + }, + { + "epoch": 3.74, + "learning_rate": 1.1542629549000964e-05, + "loss": 0.0407, + "step": 28532 + }, + { + "epoch": 3.74, + "learning_rate": 1.1542105569849072e-05, + "loss": 0.0311, + "step": 28533 + }, + { + "epoch": 3.74, + "learning_rate": 1.154158158636009e-05, + "loss": 0.03, + "step": 28534 + }, + { + "epoch": 3.74, + "learning_rate": 1.1541057598535493e-05, + "loss": 0.0422, + "step": 28535 + }, + { + "epoch": 3.74, + "learning_rate": 1.1540533606376751e-05, + "loss": 0.0411, + "step": 28536 + }, + { + "epoch": 3.74, + "learning_rate": 1.1540009609885343e-05, + "loss": 0.0309, + "step": 28537 + }, + { + "epoch": 3.74, + "learning_rate": 1.153948560906274e-05, + "loss": 0.0452, + "step": 28538 + }, + { + "epoch": 3.74, + "learning_rate": 1.1538961603910413e-05, + "loss": 0.0389, + "step": 28539 + }, + { + "epoch": 3.74, + "learning_rate": 1.153843759442984e-05, + "loss": 0.0359, + "step": 28540 + }, + { + "epoch": 3.74, + "learning_rate": 1.1537913580622494e-05, + "loss": 0.0326, + "step": 28541 + }, + { + "epoch": 3.74, + "learning_rate": 1.1537389562489848e-05, + "loss": 0.0337, + "step": 28542 + }, + { + "epoch": 3.74, + "learning_rate": 1.1536865540033373e-05, + "loss": 0.0408, + "step": 28543 + }, + { + "epoch": 3.74, + "learning_rate": 1.153634151325455e-05, + "loss": 0.0449, + "step": 28544 + }, + { + "epoch": 3.74, + "learning_rate": 1.1535817482154844e-05, + "loss": 0.0435, + "step": 28545 + }, + { + "epoch": 3.74, + "learning_rate": 1.1535293446735734e-05, + "loss": 0.0322, + "step": 28546 + }, + { + "epoch": 3.74, + "learning_rate": 1.1534769406998694e-05, + "loss": 0.04, + "step": 28547 + }, + { + "epoch": 3.74, + "learning_rate": 1.1534245362945199e-05, + "loss": 0.0328, + "step": 28548 + }, + { + "epoch": 3.74, + "learning_rate": 1.1533721314576715e-05, + "loss": 0.0336, + "step": 28549 + }, + { + "epoch": 3.74, + "learning_rate": 1.1533197261894729e-05, + "loss": 0.0388, + "step": 28550 + }, + { + "epoch": 3.74, + "learning_rate": 1.1532673204900702e-05, + "loss": 0.0369, + "step": 28551 + }, + { + "epoch": 3.74, + "learning_rate": 1.1532149143596115e-05, + "loss": 0.0331, + "step": 28552 + }, + { + "epoch": 3.74, + "learning_rate": 1.1531625077982435e-05, + "loss": 0.0402, + "step": 28553 + }, + { + "epoch": 3.74, + "learning_rate": 1.153110100806115e-05, + "loss": 0.0324, + "step": 28554 + }, + { + "epoch": 3.74, + "learning_rate": 1.153057693383372e-05, + "loss": 0.0372, + "step": 28555 + }, + { + "epoch": 3.74, + "learning_rate": 1.1530052855301626e-05, + "loss": 0.044, + "step": 28556 + }, + { + "epoch": 3.74, + "learning_rate": 1.1529528772466333e-05, + "loss": 0.04, + "step": 28557 + }, + { + "epoch": 3.74, + "learning_rate": 1.1529004685329328e-05, + "loss": 0.0392, + "step": 28558 + }, + { + "epoch": 3.74, + "learning_rate": 1.1528480593892077e-05, + "loss": 0.0446, + "step": 28559 + }, + { + "epoch": 3.74, + "learning_rate": 1.1527956498156058e-05, + "loss": 0.0417, + "step": 28560 + }, + { + "epoch": 3.74, + "learning_rate": 1.1527432398122743e-05, + "loss": 0.0314, + "step": 28561 + }, + { + "epoch": 3.74, + "learning_rate": 1.1526908293793601e-05, + "loss": 0.0337, + "step": 28562 + }, + { + "epoch": 3.74, + "learning_rate": 1.1526384185170115e-05, + "loss": 0.0393, + "step": 28563 + }, + { + "epoch": 3.74, + "learning_rate": 1.1525860072253752e-05, + "loss": 0.032, + "step": 28564 + }, + { + "epoch": 3.74, + "learning_rate": 1.1525335955045991e-05, + "loss": 0.0353, + "step": 28565 + }, + { + "epoch": 3.74, + "learning_rate": 1.1524811833548302e-05, + "loss": 0.0391, + "step": 28566 + }, + { + "epoch": 3.74, + "learning_rate": 1.1524287707762162e-05, + "loss": 0.0325, + "step": 28567 + }, + { + "epoch": 3.74, + "learning_rate": 1.1523763577689042e-05, + "loss": 0.0387, + "step": 28568 + }, + { + "epoch": 3.74, + "learning_rate": 1.1523239443330423e-05, + "loss": 0.0327, + "step": 28569 + }, + { + "epoch": 3.74, + "learning_rate": 1.152271530468777e-05, + "loss": 0.0357, + "step": 28570 + }, + { + "epoch": 3.74, + "learning_rate": 1.1522191161762564e-05, + "loss": 0.0323, + "step": 28571 + }, + { + "epoch": 3.74, + "learning_rate": 1.1521667014556275e-05, + "loss": 0.0342, + "step": 28572 + }, + { + "epoch": 3.74, + "learning_rate": 1.1521142863070378e-05, + "loss": 0.0331, + "step": 28573 + }, + { + "epoch": 3.74, + "learning_rate": 1.1520618707306345e-05, + "loss": 0.0356, + "step": 28574 + }, + { + "epoch": 3.74, + "learning_rate": 1.1520094547265658e-05, + "loss": 0.0387, + "step": 28575 + }, + { + "epoch": 3.74, + "learning_rate": 1.1519570382949786e-05, + "loss": 0.0318, + "step": 28576 + }, + { + "epoch": 3.74, + "learning_rate": 1.1519046214360202e-05, + "loss": 0.0288, + "step": 28577 + }, + { + "epoch": 3.74, + "learning_rate": 1.1518522041498379e-05, + "loss": 0.03, + "step": 28578 + }, + { + "epoch": 3.74, + "learning_rate": 1.1517997864365795e-05, + "loss": 0.0318, + "step": 28579 + }, + { + "epoch": 3.74, + "learning_rate": 1.1517473682963923e-05, + "loss": 0.0411, + "step": 28580 + }, + { + "epoch": 3.74, + "learning_rate": 1.1516949497294238e-05, + "loss": 0.0324, + "step": 28581 + }, + { + "epoch": 3.74, + "learning_rate": 1.1516425307358213e-05, + "loss": 0.0408, + "step": 28582 + }, + { + "epoch": 3.74, + "learning_rate": 1.1515901113157319e-05, + "loss": 0.0397, + "step": 28583 + }, + { + "epoch": 3.74, + "learning_rate": 1.1515376914693037e-05, + "loss": 0.0379, + "step": 28584 + }, + { + "epoch": 3.74, + "learning_rate": 1.151485271196684e-05, + "loss": 0.0323, + "step": 28585 + }, + { + "epoch": 3.74, + "learning_rate": 1.1514328504980198e-05, + "loss": 0.0357, + "step": 28586 + }, + { + "epoch": 3.74, + "learning_rate": 1.1513804293734588e-05, + "loss": 0.041, + "step": 28587 + }, + { + "epoch": 3.74, + "learning_rate": 1.1513280078231486e-05, + "loss": 0.0389, + "step": 28588 + }, + { + "epoch": 3.74, + "learning_rate": 1.1512755858472362e-05, + "loss": 0.0369, + "step": 28589 + }, + { + "epoch": 3.75, + "learning_rate": 1.1512231634458692e-05, + "loss": 0.042, + "step": 28590 + }, + { + "epoch": 3.75, + "learning_rate": 1.1511707406191952e-05, + "loss": 0.0357, + "step": 28591 + }, + { + "epoch": 3.75, + "learning_rate": 1.1511183173673617e-05, + "loss": 0.0379, + "step": 28592 + }, + { + "epoch": 3.75, + "learning_rate": 1.1510658936905159e-05, + "loss": 0.0355, + "step": 28593 + }, + { + "epoch": 3.75, + "learning_rate": 1.1510134695888053e-05, + "loss": 0.0312, + "step": 28594 + }, + { + "epoch": 3.75, + "learning_rate": 1.1509610450623772e-05, + "loss": 0.036, + "step": 28595 + }, + { + "epoch": 3.75, + "learning_rate": 1.1509086201113795e-05, + "loss": 0.0325, + "step": 28596 + }, + { + "epoch": 3.75, + "learning_rate": 1.150856194735959e-05, + "loss": 0.0317, + "step": 28597 + }, + { + "epoch": 3.75, + "learning_rate": 1.1508037689362638e-05, + "loss": 0.0393, + "step": 28598 + }, + { + "epoch": 3.75, + "learning_rate": 1.1507513427124409e-05, + "loss": 0.0383, + "step": 28599 + }, + { + "epoch": 3.75, + "learning_rate": 1.1506989160646378e-05, + "loss": 0.0286, + "step": 28600 + }, + { + "epoch": 3.75, + "learning_rate": 1.150646488993002e-05, + "loss": 0.0383, + "step": 28601 + }, + { + "epoch": 3.75, + "learning_rate": 1.1505940614976813e-05, + "loss": 0.0318, + "step": 28602 + }, + { + "epoch": 3.75, + "learning_rate": 1.1505416335788225e-05, + "loss": 0.034, + "step": 28603 + }, + { + "epoch": 3.75, + "learning_rate": 1.1504892052365735e-05, + "loss": 0.0303, + "step": 28604 + }, + { + "epoch": 3.75, + "learning_rate": 1.1504367764710817e-05, + "loss": 0.0323, + "step": 28605 + }, + { + "epoch": 3.75, + "learning_rate": 1.1503843472824942e-05, + "loss": 0.0415, + "step": 28606 + }, + { + "epoch": 3.75, + "learning_rate": 1.1503319176709588e-05, + "loss": 0.0361, + "step": 28607 + }, + { + "epoch": 3.75, + "learning_rate": 1.150279487636623e-05, + "loss": 0.0409, + "step": 28608 + }, + { + "epoch": 3.75, + "learning_rate": 1.1502270571796343e-05, + "loss": 0.0488, + "step": 28609 + }, + { + "epoch": 3.75, + "learning_rate": 1.1501746263001399e-05, + "loss": 0.0404, + "step": 28610 + }, + { + "epoch": 3.75, + "learning_rate": 1.1501221949982873e-05, + "loss": 0.0347, + "step": 28611 + }, + { + "epoch": 3.75, + "learning_rate": 1.1500697632742238e-05, + "loss": 0.0334, + "step": 28612 + }, + { + "epoch": 3.75, + "learning_rate": 1.1500173311280976e-05, + "loss": 0.0361, + "step": 28613 + }, + { + "epoch": 3.75, + "learning_rate": 1.1499648985600552e-05, + "loss": 0.0302, + "step": 28614 + }, + { + "epoch": 3.75, + "learning_rate": 1.1499124655702447e-05, + "loss": 0.0341, + "step": 28615 + }, + { + "epoch": 3.75, + "learning_rate": 1.1498600321588134e-05, + "loss": 0.0327, + "step": 28616 + }, + { + "epoch": 3.75, + "learning_rate": 1.1498075983259085e-05, + "loss": 0.0336, + "step": 28617 + }, + { + "epoch": 3.75, + "learning_rate": 1.1497551640716778e-05, + "loss": 0.0342, + "step": 28618 + }, + { + "epoch": 3.75, + "learning_rate": 1.1497027293962689e-05, + "loss": 0.04, + "step": 28619 + }, + { + "epoch": 3.75, + "learning_rate": 1.1496502942998288e-05, + "loss": 0.0406, + "step": 28620 + }, + { + "epoch": 3.75, + "learning_rate": 1.1495978587825054e-05, + "loss": 0.0402, + "step": 28621 + }, + { + "epoch": 3.75, + "learning_rate": 1.1495454228444458e-05, + "loss": 0.0301, + "step": 28622 + }, + { + "epoch": 3.75, + "learning_rate": 1.1494929864857978e-05, + "loss": 0.0354, + "step": 28623 + }, + { + "epoch": 3.75, + "learning_rate": 1.1494405497067087e-05, + "loss": 0.0369, + "step": 28624 + }, + { + "epoch": 3.75, + "learning_rate": 1.149388112507326e-05, + "loss": 0.0304, + "step": 28625 + }, + { + "epoch": 3.75, + "learning_rate": 1.1493356748877974e-05, + "loss": 0.0421, + "step": 28626 + }, + { + "epoch": 3.75, + "learning_rate": 1.1492832368482699e-05, + "loss": 0.0316, + "step": 28627 + }, + { + "epoch": 3.75, + "learning_rate": 1.149230798388891e-05, + "loss": 0.0339, + "step": 28628 + }, + { + "epoch": 3.75, + "learning_rate": 1.1491783595098087e-05, + "loss": 0.0368, + "step": 28629 + }, + { + "epoch": 3.75, + "learning_rate": 1.1491259202111705e-05, + "loss": 0.0288, + "step": 28630 + }, + { + "epoch": 3.75, + "learning_rate": 1.1490734804931232e-05, + "loss": 0.0408, + "step": 28631 + }, + { + "epoch": 3.75, + "learning_rate": 1.1490210403558151e-05, + "loss": 0.0392, + "step": 28632 + }, + { + "epoch": 3.75, + "learning_rate": 1.1489685997993929e-05, + "loss": 0.0375, + "step": 28633 + }, + { + "epoch": 3.75, + "learning_rate": 1.1489161588240043e-05, + "loss": 0.035, + "step": 28634 + }, + { + "epoch": 3.75, + "learning_rate": 1.1488637174297973e-05, + "loss": 0.0421, + "step": 28635 + }, + { + "epoch": 3.75, + "learning_rate": 1.148811275616919e-05, + "loss": 0.041, + "step": 28636 + }, + { + "epoch": 3.75, + "learning_rate": 1.1487588333855167e-05, + "loss": 0.0385, + "step": 28637 + }, + { + "epoch": 3.75, + "learning_rate": 1.1487063907357383e-05, + "loss": 0.0302, + "step": 28638 + }, + { + "epoch": 3.75, + "learning_rate": 1.1486539476677306e-05, + "loss": 0.0329, + "step": 28639 + }, + { + "epoch": 3.75, + "learning_rate": 1.148601504181642e-05, + "loss": 0.0319, + "step": 28640 + }, + { + "epoch": 3.75, + "learning_rate": 1.1485490602776197e-05, + "loss": 0.0441, + "step": 28641 + }, + { + "epoch": 3.75, + "learning_rate": 1.1484966159558112e-05, + "loss": 0.0343, + "step": 28642 + }, + { + "epoch": 3.75, + "learning_rate": 1.1484441712163633e-05, + "loss": 0.0383, + "step": 28643 + }, + { + "epoch": 3.75, + "learning_rate": 1.1483917260594243e-05, + "loss": 0.0348, + "step": 28644 + }, + { + "epoch": 3.75, + "learning_rate": 1.1483392804851417e-05, + "loss": 0.0431, + "step": 28645 + }, + { + "epoch": 3.75, + "learning_rate": 1.1482868344936626e-05, + "loss": 0.0329, + "step": 28646 + }, + { + "epoch": 3.75, + "learning_rate": 1.1482343880851347e-05, + "loss": 0.0408, + "step": 28647 + }, + { + "epoch": 3.75, + "learning_rate": 1.1481819412597054e-05, + "loss": 0.0403, + "step": 28648 + }, + { + "epoch": 3.75, + "learning_rate": 1.1481294940175224e-05, + "loss": 0.0388, + "step": 28649 + }, + { + "epoch": 3.75, + "learning_rate": 1.1480770463587329e-05, + "loss": 0.0396, + "step": 28650 + }, + { + "epoch": 3.75, + "learning_rate": 1.148024598283485e-05, + "loss": 0.0358, + "step": 28651 + }, + { + "epoch": 3.75, + "learning_rate": 1.1479721497919253e-05, + "loss": 0.033, + "step": 28652 + }, + { + "epoch": 3.75, + "learning_rate": 1.1479197008842023e-05, + "loss": 0.0451, + "step": 28653 + }, + { + "epoch": 3.75, + "learning_rate": 1.1478672515604628e-05, + "loss": 0.0352, + "step": 28654 + }, + { + "epoch": 3.75, + "learning_rate": 1.1478148018208544e-05, + "loss": 0.036, + "step": 28655 + }, + { + "epoch": 3.75, + "learning_rate": 1.1477623516655247e-05, + "loss": 0.0378, + "step": 28656 + }, + { + "epoch": 3.75, + "learning_rate": 1.1477099010946215e-05, + "loss": 0.0434, + "step": 28657 + }, + { + "epoch": 3.75, + "learning_rate": 1.147657450108292e-05, + "loss": 0.0368, + "step": 28658 + }, + { + "epoch": 3.75, + "learning_rate": 1.147604998706684e-05, + "loss": 0.0369, + "step": 28659 + }, + { + "epoch": 3.75, + "learning_rate": 1.1475525468899442e-05, + "loss": 0.036, + "step": 28660 + }, + { + "epoch": 3.75, + "learning_rate": 1.1475000946582212e-05, + "loss": 0.0351, + "step": 28661 + }, + { + "epoch": 3.75, + "learning_rate": 1.1474476420116618e-05, + "loss": 0.0344, + "step": 28662 + }, + { + "epoch": 3.75, + "learning_rate": 1.147395188950414e-05, + "loss": 0.0375, + "step": 28663 + }, + { + "epoch": 3.75, + "learning_rate": 1.147342735474625e-05, + "loss": 0.0376, + "step": 28664 + }, + { + "epoch": 3.75, + "learning_rate": 1.1472902815844422e-05, + "loss": 0.0358, + "step": 28665 + }, + { + "epoch": 3.76, + "learning_rate": 1.1472378272800132e-05, + "loss": 0.0524, + "step": 28666 + }, + { + "epoch": 3.76, + "learning_rate": 1.1471853725614863e-05, + "loss": 0.0365, + "step": 28667 + }, + { + "epoch": 3.76, + "learning_rate": 1.147132917429008e-05, + "loss": 0.0409, + "step": 28668 + }, + { + "epoch": 3.76, + "learning_rate": 1.147080461882726e-05, + "loss": 0.0286, + "step": 28669 + }, + { + "epoch": 3.76, + "learning_rate": 1.1470280059227886e-05, + "loss": 0.0427, + "step": 28670 + }, + { + "epoch": 3.76, + "learning_rate": 1.1469755495493422e-05, + "loss": 0.0394, + "step": 28671 + }, + { + "epoch": 3.76, + "learning_rate": 1.146923092762535e-05, + "loss": 0.0358, + "step": 28672 + }, + { + "epoch": 3.76, + "learning_rate": 1.1468706355625145e-05, + "loss": 0.0306, + "step": 28673 + }, + { + "epoch": 3.76, + "learning_rate": 1.1468181779494283e-05, + "loss": 0.0351, + "step": 28674 + }, + { + "epoch": 3.76, + "learning_rate": 1.1467657199234237e-05, + "loss": 0.0378, + "step": 28675 + }, + { + "epoch": 3.76, + "learning_rate": 1.1467132614846483e-05, + "loss": 0.0394, + "step": 28676 + }, + { + "epoch": 3.76, + "learning_rate": 1.1466608026332496e-05, + "loss": 0.0352, + "step": 28677 + }, + { + "epoch": 3.76, + "learning_rate": 1.1466083433693753e-05, + "loss": 0.038, + "step": 28678 + }, + { + "epoch": 3.76, + "learning_rate": 1.1465558836931728e-05, + "loss": 0.0385, + "step": 28679 + }, + { + "epoch": 3.76, + "learning_rate": 1.1465034236047898e-05, + "loss": 0.0426, + "step": 28680 + }, + { + "epoch": 3.76, + "learning_rate": 1.1464509631043737e-05, + "loss": 0.0322, + "step": 28681 + }, + { + "epoch": 3.76, + "learning_rate": 1.1463985021920721e-05, + "loss": 0.0292, + "step": 28682 + }, + { + "epoch": 3.76, + "learning_rate": 1.146346040868032e-05, + "loss": 0.037, + "step": 28683 + }, + { + "epoch": 3.76, + "learning_rate": 1.1462935791324021e-05, + "loss": 0.0402, + "step": 28684 + }, + { + "epoch": 3.76, + "learning_rate": 1.146241116985329e-05, + "loss": 0.0393, + "step": 28685 + }, + { + "epoch": 3.76, + "learning_rate": 1.1461886544269607e-05, + "loss": 0.0392, + "step": 28686 + }, + { + "epoch": 3.76, + "learning_rate": 1.1461361914574445e-05, + "loss": 0.0346, + "step": 28687 + }, + { + "epoch": 3.76, + "learning_rate": 1.1460837280769279e-05, + "loss": 0.0323, + "step": 28688 + }, + { + "epoch": 3.76, + "learning_rate": 1.1460312642855586e-05, + "loss": 0.0331, + "step": 28689 + }, + { + "epoch": 3.76, + "learning_rate": 1.1459788000834841e-05, + "loss": 0.0389, + "step": 28690 + }, + { + "epoch": 3.76, + "learning_rate": 1.1459263354708526e-05, + "loss": 0.0382, + "step": 28691 + }, + { + "epoch": 3.76, + "learning_rate": 1.1458738704478105e-05, + "loss": 0.0376, + "step": 28692 + }, + { + "epoch": 3.76, + "learning_rate": 1.145821405014506e-05, + "loss": 0.0376, + "step": 28693 + }, + { + "epoch": 3.76, + "learning_rate": 1.1457689391710865e-05, + "loss": 0.0459, + "step": 28694 + }, + { + "epoch": 3.76, + "learning_rate": 1.1457164729176997e-05, + "loss": 0.0399, + "step": 28695 + }, + { + "epoch": 3.76, + "learning_rate": 1.145664006254493e-05, + "loss": 0.0396, + "step": 28696 + }, + { + "epoch": 3.76, + "learning_rate": 1.1456115391816145e-05, + "loss": 0.0412, + "step": 28697 + }, + { + "epoch": 3.76, + "learning_rate": 1.1455590716992106e-05, + "loss": 0.0398, + "step": 28698 + }, + { + "epoch": 3.76, + "learning_rate": 1.14550660380743e-05, + "loss": 0.0287, + "step": 28699 + }, + { + "epoch": 3.76, + "learning_rate": 1.1454541355064198e-05, + "loss": 0.0413, + "step": 28700 + }, + { + "epoch": 3.76, + "learning_rate": 1.1454016667963275e-05, + "loss": 0.033, + "step": 28701 + }, + { + "epoch": 3.76, + "learning_rate": 1.1453491976773008e-05, + "loss": 0.0346, + "step": 28702 + }, + { + "epoch": 3.76, + "learning_rate": 1.1452967281494873e-05, + "loss": 0.0376, + "step": 28703 + }, + { + "epoch": 3.76, + "learning_rate": 1.1452442582130341e-05, + "loss": 0.0365, + "step": 28704 + }, + { + "epoch": 3.76, + "learning_rate": 1.1451917878680896e-05, + "loss": 0.0294, + "step": 28705 + }, + { + "epoch": 3.76, + "learning_rate": 1.1451393171148007e-05, + "loss": 0.0423, + "step": 28706 + }, + { + "epoch": 3.76, + "learning_rate": 1.1450868459533153e-05, + "loss": 0.0335, + "step": 28707 + }, + { + "epoch": 3.76, + "learning_rate": 1.1450343743837809e-05, + "loss": 0.0343, + "step": 28708 + }, + { + "epoch": 3.76, + "learning_rate": 1.1449819024063449e-05, + "loss": 0.0369, + "step": 28709 + }, + { + "epoch": 3.76, + "learning_rate": 1.144929430021155e-05, + "loss": 0.0329, + "step": 28710 + }, + { + "epoch": 3.76, + "learning_rate": 1.144876957228359e-05, + "loss": 0.039, + "step": 28711 + }, + { + "epoch": 3.76, + "learning_rate": 1.1448244840281041e-05, + "loss": 0.037, + "step": 28712 + }, + { + "epoch": 3.76, + "learning_rate": 1.1447720104205382e-05, + "loss": 0.0319, + "step": 28713 + }, + { + "epoch": 3.76, + "learning_rate": 1.1447195364058089e-05, + "loss": 0.0361, + "step": 28714 + }, + { + "epoch": 3.76, + "learning_rate": 1.144667061984063e-05, + "loss": 0.0458, + "step": 28715 + }, + { + "epoch": 3.76, + "learning_rate": 1.144614587155449e-05, + "loss": 0.0352, + "step": 28716 + }, + { + "epoch": 3.76, + "learning_rate": 1.1445621119201143e-05, + "loss": 0.0374, + "step": 28717 + }, + { + "epoch": 3.76, + "learning_rate": 1.1445096362782064e-05, + "loss": 0.0351, + "step": 28718 + }, + { + "epoch": 3.76, + "learning_rate": 1.1444571602298726e-05, + "loss": 0.0321, + "step": 28719 + }, + { + "epoch": 3.76, + "learning_rate": 1.1444046837752612e-05, + "loss": 0.0342, + "step": 28720 + }, + { + "epoch": 3.76, + "learning_rate": 1.1443522069145185e-05, + "loss": 0.034, + "step": 28721 + }, + { + "epoch": 3.76, + "learning_rate": 1.1442997296477937e-05, + "loss": 0.0319, + "step": 28722 + }, + { + "epoch": 3.76, + "learning_rate": 1.1442472519752333e-05, + "loss": 0.0294, + "step": 28723 + }, + { + "epoch": 3.76, + "learning_rate": 1.1441947738969853e-05, + "loss": 0.0301, + "step": 28724 + }, + { + "epoch": 3.76, + "learning_rate": 1.1441422954131968e-05, + "loss": 0.0339, + "step": 28725 + }, + { + "epoch": 3.76, + "learning_rate": 1.1440898165240162e-05, + "loss": 0.0371, + "step": 28726 + }, + { + "epoch": 3.76, + "learning_rate": 1.1440373372295903e-05, + "loss": 0.0339, + "step": 28727 + }, + { + "epoch": 3.76, + "learning_rate": 1.1439848575300675e-05, + "loss": 0.0352, + "step": 28728 + }, + { + "epoch": 3.76, + "learning_rate": 1.1439323774255946e-05, + "loss": 0.0399, + "step": 28729 + }, + { + "epoch": 3.76, + "learning_rate": 1.1438798969163193e-05, + "loss": 0.0346, + "step": 28730 + }, + { + "epoch": 3.76, + "learning_rate": 1.1438274160023901e-05, + "loss": 0.0455, + "step": 28731 + }, + { + "epoch": 3.76, + "learning_rate": 1.1437749346839536e-05, + "loss": 0.0336, + "step": 28732 + }, + { + "epoch": 3.76, + "learning_rate": 1.1437224529611577e-05, + "loss": 0.0303, + "step": 28733 + }, + { + "epoch": 3.76, + "learning_rate": 1.1436699708341501e-05, + "loss": 0.0318, + "step": 28734 + }, + { + "epoch": 3.76, + "learning_rate": 1.1436174883030787e-05, + "loss": 0.0355, + "step": 28735 + }, + { + "epoch": 3.76, + "learning_rate": 1.1435650053680902e-05, + "loss": 0.0381, + "step": 28736 + }, + { + "epoch": 3.76, + "learning_rate": 1.143512522029333e-05, + "loss": 0.0352, + "step": 28737 + }, + { + "epoch": 3.76, + "learning_rate": 1.1434600382869545e-05, + "loss": 0.0409, + "step": 28738 + }, + { + "epoch": 3.76, + "learning_rate": 1.1434075541411025e-05, + "loss": 0.0397, + "step": 28739 + }, + { + "epoch": 3.76, + "learning_rate": 1.1433550695919242e-05, + "loss": 0.0475, + "step": 28740 + }, + { + "epoch": 3.76, + "learning_rate": 1.1433025846395676e-05, + "loss": 0.0383, + "step": 28741 + }, + { + "epoch": 3.76, + "learning_rate": 1.1432500992841798e-05, + "loss": 0.0365, + "step": 28742 + }, + { + "epoch": 3.77, + "learning_rate": 1.1431976135259086e-05, + "loss": 0.0334, + "step": 28743 + }, + { + "epoch": 3.77, + "learning_rate": 1.143145127364902e-05, + "loss": 0.043, + "step": 28744 + }, + { + "epoch": 3.77, + "learning_rate": 1.1430926408013074e-05, + "loss": 0.0356, + "step": 28745 + }, + { + "epoch": 3.77, + "learning_rate": 1.1430401538352722e-05, + "loss": 0.0476, + "step": 28746 + }, + { + "epoch": 3.77, + "learning_rate": 1.1429876664669445e-05, + "loss": 0.0308, + "step": 28747 + }, + { + "epoch": 3.77, + "learning_rate": 1.142935178696471e-05, + "loss": 0.0335, + "step": 28748 + }, + { + "epoch": 3.77, + "learning_rate": 1.1428826905240005e-05, + "loss": 0.0414, + "step": 28749 + }, + { + "epoch": 3.77, + "learning_rate": 1.1428302019496797e-05, + "loss": 0.033, + "step": 28750 + }, + { + "epoch": 3.77, + "learning_rate": 1.142777712973657e-05, + "loss": 0.034, + "step": 28751 + }, + { + "epoch": 3.77, + "learning_rate": 1.1427252235960792e-05, + "loss": 0.0396, + "step": 28752 + }, + { + "epoch": 3.77, + "learning_rate": 1.1426727338170944e-05, + "loss": 0.039, + "step": 28753 + }, + { + "epoch": 3.77, + "learning_rate": 1.14262024363685e-05, + "loss": 0.0378, + "step": 28754 + }, + { + "epoch": 3.77, + "learning_rate": 1.142567753055494e-05, + "loss": 0.0337, + "step": 28755 + }, + { + "epoch": 3.77, + "learning_rate": 1.1425152620731736e-05, + "loss": 0.0296, + "step": 28756 + }, + { + "epoch": 3.77, + "learning_rate": 1.1424627706900367e-05, + "loss": 0.0411, + "step": 28757 + }, + { + "epoch": 3.77, + "learning_rate": 1.142410278906231e-05, + "loss": 0.0293, + "step": 28758 + }, + { + "epoch": 3.77, + "learning_rate": 1.1423577867219036e-05, + "loss": 0.0372, + "step": 28759 + }, + { + "epoch": 3.77, + "learning_rate": 1.1423052941372029e-05, + "loss": 0.0302, + "step": 28760 + }, + { + "epoch": 3.77, + "learning_rate": 1.1422528011522758e-05, + "loss": 0.0385, + "step": 28761 + }, + { + "epoch": 3.77, + "learning_rate": 1.1422003077672706e-05, + "loss": 0.0343, + "step": 28762 + }, + { + "epoch": 3.77, + "learning_rate": 1.1421478139823343e-05, + "loss": 0.0451, + "step": 28763 + }, + { + "epoch": 3.77, + "learning_rate": 1.1420953197976153e-05, + "loss": 0.0309, + "step": 28764 + }, + { + "epoch": 3.77, + "learning_rate": 1.14204282521326e-05, + "loss": 0.0363, + "step": 28765 + }, + { + "epoch": 3.77, + "learning_rate": 1.1419903302294175e-05, + "loss": 0.0403, + "step": 28766 + }, + { + "epoch": 3.77, + "learning_rate": 1.1419378348462344e-05, + "loss": 0.0353, + "step": 28767 + }, + { + "epoch": 3.77, + "learning_rate": 1.1418853390638591e-05, + "loss": 0.0384, + "step": 28768 + }, + { + "epoch": 3.77, + "learning_rate": 1.1418328428824386e-05, + "loss": 0.035, + "step": 28769 + }, + { + "epoch": 3.77, + "learning_rate": 1.1417803463021207e-05, + "loss": 0.0341, + "step": 28770 + }, + { + "epoch": 3.77, + "learning_rate": 1.141727849323053e-05, + "loss": 0.0422, + "step": 28771 + }, + { + "epoch": 3.77, + "learning_rate": 1.1416753519453836e-05, + "loss": 0.0441, + "step": 28772 + }, + { + "epoch": 3.77, + "learning_rate": 1.1416228541692595e-05, + "loss": 0.0374, + "step": 28773 + }, + { + "epoch": 3.77, + "learning_rate": 1.1415703559948289e-05, + "loss": 0.0308, + "step": 28774 + }, + { + "epoch": 3.77, + "learning_rate": 1.141517857422239e-05, + "loss": 0.0351, + "step": 28775 + }, + { + "epoch": 3.77, + "learning_rate": 1.141465358451638e-05, + "loss": 0.0395, + "step": 28776 + }, + { + "epoch": 3.77, + "learning_rate": 1.141412859083173e-05, + "loss": 0.0373, + "step": 28777 + }, + { + "epoch": 3.77, + "learning_rate": 1.1413603593169917e-05, + "loss": 0.0353, + "step": 28778 + }, + { + "epoch": 3.77, + "learning_rate": 1.1413078591532422e-05, + "loss": 0.0342, + "step": 28779 + }, + { + "epoch": 3.77, + "learning_rate": 1.1412553585920717e-05, + "loss": 0.0321, + "step": 28780 + }, + { + "epoch": 3.77, + "learning_rate": 1.1412028576336281e-05, + "loss": 0.0391, + "step": 28781 + }, + { + "epoch": 3.77, + "learning_rate": 1.141150356278059e-05, + "loss": 0.0371, + "step": 28782 + }, + { + "epoch": 3.77, + "learning_rate": 1.141097854525512e-05, + "loss": 0.0388, + "step": 28783 + }, + { + "epoch": 3.77, + "learning_rate": 1.1410453523761349e-05, + "loss": 0.0339, + "step": 28784 + }, + { + "epoch": 3.77, + "learning_rate": 1.1409928498300752e-05, + "loss": 0.0317, + "step": 28785 + }, + { + "epoch": 3.77, + "learning_rate": 1.1409403468874804e-05, + "loss": 0.0366, + "step": 28786 + }, + { + "epoch": 3.77, + "learning_rate": 1.1408878435484986e-05, + "loss": 0.0352, + "step": 28787 + }, + { + "epoch": 3.77, + "learning_rate": 1.1408353398132772e-05, + "loss": 0.041, + "step": 28788 + }, + { + "epoch": 3.77, + "learning_rate": 1.1407828356819642e-05, + "loss": 0.0356, + "step": 28789 + }, + { + "epoch": 3.77, + "learning_rate": 1.1407303311547066e-05, + "loss": 0.0382, + "step": 28790 + }, + { + "epoch": 3.77, + "learning_rate": 1.1406778262316524e-05, + "loss": 0.037, + "step": 28791 + }, + { + "epoch": 3.77, + "learning_rate": 1.1406253209129495e-05, + "loss": 0.0381, + "step": 28792 + }, + { + "epoch": 3.77, + "learning_rate": 1.1405728151987453e-05, + "loss": 0.0331, + "step": 28793 + }, + { + "epoch": 3.77, + "learning_rate": 1.1405203090891877e-05, + "loss": 0.0357, + "step": 28794 + }, + { + "epoch": 3.77, + "learning_rate": 1.140467802584424e-05, + "loss": 0.0377, + "step": 28795 + }, + { + "epoch": 3.77, + "learning_rate": 1.1404152956846022e-05, + "loss": 0.0325, + "step": 28796 + }, + { + "epoch": 3.77, + "learning_rate": 1.1403627883898699e-05, + "loss": 0.0369, + "step": 28797 + }, + { + "epoch": 3.77, + "learning_rate": 1.1403102807003745e-05, + "loss": 0.0407, + "step": 28798 + }, + { + "epoch": 3.77, + "learning_rate": 1.140257772616264e-05, + "loss": 0.0419, + "step": 28799 + }, + { + "epoch": 3.77, + "learning_rate": 1.1402052641376862e-05, + "loss": 0.0399, + "step": 28800 + }, + { + "epoch": 3.77, + "learning_rate": 1.1401527552647881e-05, + "loss": 0.032, + "step": 28801 + }, + { + "epoch": 3.77, + "learning_rate": 1.1401002459977187e-05, + "loss": 0.0344, + "step": 28802 + }, + { + "epoch": 3.77, + "learning_rate": 1.1400477363366237e-05, + "loss": 0.036, + "step": 28803 + }, + { + "epoch": 3.77, + "learning_rate": 1.1399952262816527e-05, + "loss": 0.033, + "step": 28804 + }, + { + "epoch": 3.77, + "learning_rate": 1.1399427158329523e-05, + "loss": 0.043, + "step": 28805 + }, + { + "epoch": 3.77, + "learning_rate": 1.1398902049906706e-05, + "loss": 0.0327, + "step": 28806 + }, + { + "epoch": 3.77, + "learning_rate": 1.139837693754955e-05, + "loss": 0.0339, + "step": 28807 + }, + { + "epoch": 3.77, + "learning_rate": 1.1397851821259533e-05, + "loss": 0.0325, + "step": 28808 + }, + { + "epoch": 3.77, + "learning_rate": 1.1397326701038135e-05, + "loss": 0.0273, + "step": 28809 + }, + { + "epoch": 3.77, + "learning_rate": 1.1396801576886827e-05, + "loss": 0.0379, + "step": 28810 + }, + { + "epoch": 3.77, + "learning_rate": 1.139627644880709e-05, + "loss": 0.0359, + "step": 28811 + }, + { + "epoch": 3.77, + "learning_rate": 1.1395751316800401e-05, + "loss": 0.0325, + "step": 28812 + }, + { + "epoch": 3.77, + "learning_rate": 1.1395226180868234e-05, + "loss": 0.0365, + "step": 28813 + }, + { + "epoch": 3.77, + "learning_rate": 1.1394701041012069e-05, + "loss": 0.0324, + "step": 28814 + }, + { + "epoch": 3.77, + "learning_rate": 1.139417589723338e-05, + "loss": 0.0372, + "step": 28815 + }, + { + "epoch": 3.77, + "learning_rate": 1.1393650749533645e-05, + "loss": 0.0382, + "step": 28816 + }, + { + "epoch": 3.77, + "learning_rate": 1.1393125597914345e-05, + "loss": 0.0338, + "step": 28817 + }, + { + "epoch": 3.77, + "learning_rate": 1.1392600442376951e-05, + "loss": 0.0326, + "step": 28818 + }, + { + "epoch": 3.78, + "learning_rate": 1.139207528292294e-05, + "loss": 0.0385, + "step": 28819 + }, + { + "epoch": 3.78, + "learning_rate": 1.1391550119553796e-05, + "loss": 0.0328, + "step": 28820 + }, + { + "epoch": 3.78, + "learning_rate": 1.1391024952270991e-05, + "loss": 0.0368, + "step": 28821 + }, + { + "epoch": 3.78, + "learning_rate": 1.1390499781075999e-05, + "loss": 0.038, + "step": 28822 + }, + { + "epoch": 3.78, + "learning_rate": 1.1389974605970305e-05, + "loss": 0.0388, + "step": 28823 + }, + { + "epoch": 3.78, + "learning_rate": 1.1389449426955377e-05, + "loss": 0.0346, + "step": 28824 + }, + { + "epoch": 3.78, + "learning_rate": 1.1388924244032697e-05, + "loss": 0.0375, + "step": 28825 + }, + { + "epoch": 3.78, + "learning_rate": 1.1388399057203743e-05, + "loss": 0.0383, + "step": 28826 + }, + { + "epoch": 3.78, + "learning_rate": 1.1387873866469991e-05, + "loss": 0.0358, + "step": 28827 + }, + { + "epoch": 3.78, + "learning_rate": 1.1387348671832918e-05, + "loss": 0.0376, + "step": 28828 + }, + { + "epoch": 3.78, + "learning_rate": 1.1386823473294002e-05, + "loss": 0.0367, + "step": 28829 + }, + { + "epoch": 3.78, + "learning_rate": 1.1386298270854712e-05, + "loss": 0.032, + "step": 28830 + }, + { + "epoch": 3.78, + "learning_rate": 1.138577306451654e-05, + "loss": 0.0338, + "step": 28831 + }, + { + "epoch": 3.78, + "learning_rate": 1.138524785428095e-05, + "loss": 0.0383, + "step": 28832 + }, + { + "epoch": 3.78, + "learning_rate": 1.1384722640149425e-05, + "loss": 0.0352, + "step": 28833 + }, + { + "epoch": 3.78, + "learning_rate": 1.1384197422123444e-05, + "loss": 0.0333, + "step": 28834 + }, + { + "epoch": 3.78, + "learning_rate": 1.1383672200204478e-05, + "loss": 0.0424, + "step": 28835 + }, + { + "epoch": 3.78, + "learning_rate": 1.1383146974394007e-05, + "loss": 0.0412, + "step": 28836 + }, + { + "epoch": 3.78, + "learning_rate": 1.1382621744693514e-05, + "loss": 0.033, + "step": 28837 + }, + { + "epoch": 3.78, + "learning_rate": 1.1382096511104468e-05, + "loss": 0.0398, + "step": 28838 + }, + { + "epoch": 3.78, + "learning_rate": 1.138157127362835e-05, + "loss": 0.0291, + "step": 28839 + }, + { + "epoch": 3.78, + "learning_rate": 1.1381046032266637e-05, + "loss": 0.0416, + "step": 28840 + }, + { + "epoch": 3.78, + "learning_rate": 1.1380520787020799e-05, + "loss": 0.0341, + "step": 28841 + }, + { + "epoch": 3.78, + "learning_rate": 1.1379995537892328e-05, + "loss": 0.0329, + "step": 28842 + }, + { + "epoch": 3.78, + "learning_rate": 1.1379470284882687e-05, + "loss": 0.0306, + "step": 28843 + }, + { + "epoch": 3.78, + "learning_rate": 1.1378945027993365e-05, + "loss": 0.0409, + "step": 28844 + }, + { + "epoch": 3.78, + "learning_rate": 1.137841976722583e-05, + "loss": 0.0304, + "step": 28845 + }, + { + "epoch": 3.78, + "learning_rate": 1.1377894502581563e-05, + "loss": 0.0373, + "step": 28846 + }, + { + "epoch": 3.78, + "learning_rate": 1.1377369234062041e-05, + "loss": 0.0305, + "step": 28847 + }, + { + "epoch": 3.78, + "learning_rate": 1.1376843961668745e-05, + "loss": 0.0442, + "step": 28848 + }, + { + "epoch": 3.78, + "learning_rate": 1.1376318685403144e-05, + "loss": 0.0366, + "step": 28849 + }, + { + "epoch": 3.78, + "learning_rate": 1.1375793405266725e-05, + "loss": 0.036, + "step": 28850 + }, + { + "epoch": 3.78, + "learning_rate": 1.1375268121260954e-05, + "loss": 0.0371, + "step": 28851 + }, + { + "epoch": 3.78, + "learning_rate": 1.1374742833387317e-05, + "loss": 0.0374, + "step": 28852 + }, + { + "epoch": 3.78, + "learning_rate": 1.137421754164729e-05, + "loss": 0.0347, + "step": 28853 + }, + { + "epoch": 3.78, + "learning_rate": 1.137369224604235e-05, + "loss": 0.0379, + "step": 28854 + }, + { + "epoch": 3.78, + "learning_rate": 1.1373166946573972e-05, + "loss": 0.0311, + "step": 28855 + }, + { + "epoch": 3.78, + "learning_rate": 1.1372641643243636e-05, + "loss": 0.0338, + "step": 28856 + }, + { + "epoch": 3.78, + "learning_rate": 1.1372116336052817e-05, + "loss": 0.0313, + "step": 28857 + }, + { + "epoch": 3.78, + "learning_rate": 1.1371591025002996e-05, + "loss": 0.033, + "step": 28858 + }, + { + "epoch": 3.78, + "learning_rate": 1.1371065710095648e-05, + "loss": 0.0364, + "step": 28859 + }, + { + "epoch": 3.78, + "learning_rate": 1.137054039133225e-05, + "loss": 0.0334, + "step": 28860 + }, + { + "epoch": 3.78, + "learning_rate": 1.1370015068714281e-05, + "loss": 0.0336, + "step": 28861 + }, + { + "epoch": 3.78, + "learning_rate": 1.1369489742243217e-05, + "loss": 0.0384, + "step": 28862 + }, + { + "epoch": 3.78, + "learning_rate": 1.1368964411920534e-05, + "loss": 0.0462, + "step": 28863 + }, + { + "epoch": 3.78, + "learning_rate": 1.1368439077747712e-05, + "loss": 0.0322, + "step": 28864 + }, + { + "epoch": 3.78, + "learning_rate": 1.1367913739726233e-05, + "loss": 0.031, + "step": 28865 + }, + { + "epoch": 3.78, + "learning_rate": 1.1367388397857563e-05, + "loss": 0.0444, + "step": 28866 + }, + { + "epoch": 3.78, + "learning_rate": 1.136686305214319e-05, + "loss": 0.0417, + "step": 28867 + }, + { + "epoch": 3.78, + "learning_rate": 1.1366337702584585e-05, + "loss": 0.0384, + "step": 28868 + }, + { + "epoch": 3.78, + "learning_rate": 1.1365812349183229e-05, + "loss": 0.0386, + "step": 28869 + }, + { + "epoch": 3.78, + "learning_rate": 1.1365286991940599e-05, + "loss": 0.0354, + "step": 28870 + }, + { + "epoch": 3.78, + "learning_rate": 1.1364761630858174e-05, + "loss": 0.0401, + "step": 28871 + }, + { + "epoch": 3.78, + "learning_rate": 1.1364236265937428e-05, + "loss": 0.0297, + "step": 28872 + }, + { + "epoch": 3.78, + "learning_rate": 1.1363710897179839e-05, + "loss": 0.0355, + "step": 28873 + }, + { + "epoch": 3.78, + "learning_rate": 1.1363185524586883e-05, + "loss": 0.0321, + "step": 28874 + }, + { + "epoch": 3.78, + "learning_rate": 1.1362660148160045e-05, + "loss": 0.0307, + "step": 28875 + }, + { + "epoch": 3.78, + "learning_rate": 1.1362134767900797e-05, + "loss": 0.0345, + "step": 28876 + }, + { + "epoch": 3.78, + "learning_rate": 1.1361609383810616e-05, + "loss": 0.0347, + "step": 28877 + }, + { + "epoch": 3.78, + "learning_rate": 1.1361083995890982e-05, + "loss": 0.0353, + "step": 28878 + }, + { + "epoch": 3.78, + "learning_rate": 1.1360558604143372e-05, + "loss": 0.0371, + "step": 28879 + }, + { + "epoch": 3.78, + "learning_rate": 1.1360033208569263e-05, + "loss": 0.0329, + "step": 28880 + }, + { + "epoch": 3.78, + "learning_rate": 1.1359507809170132e-05, + "loss": 0.0348, + "step": 28881 + }, + { + "epoch": 3.78, + "learning_rate": 1.135898240594746e-05, + "loss": 0.0348, + "step": 28882 + }, + { + "epoch": 3.78, + "learning_rate": 1.135845699890272e-05, + "loss": 0.0354, + "step": 28883 + }, + { + "epoch": 3.78, + "learning_rate": 1.1357931588037396e-05, + "loss": 0.0301, + "step": 28884 + }, + { + "epoch": 3.78, + "learning_rate": 1.1357406173352955e-05, + "loss": 0.0344, + "step": 28885 + }, + { + "epoch": 3.78, + "learning_rate": 1.1356880754850887e-05, + "loss": 0.0402, + "step": 28886 + }, + { + "epoch": 3.78, + "learning_rate": 1.1356355332532662e-05, + "loss": 0.0405, + "step": 28887 + }, + { + "epoch": 3.78, + "learning_rate": 1.1355829906399762e-05, + "loss": 0.0291, + "step": 28888 + }, + { + "epoch": 3.78, + "learning_rate": 1.1355304476453662e-05, + "loss": 0.0415, + "step": 28889 + }, + { + "epoch": 3.78, + "learning_rate": 1.1354779042695839e-05, + "loss": 0.0385, + "step": 28890 + }, + { + "epoch": 3.78, + "learning_rate": 1.1354253605127772e-05, + "loss": 0.0372, + "step": 28891 + }, + { + "epoch": 3.78, + "learning_rate": 1.135372816375094e-05, + "loss": 0.0336, + "step": 28892 + }, + { + "epoch": 3.78, + "learning_rate": 1.1353202718566818e-05, + "loss": 0.0393, + "step": 28893 + }, + { + "epoch": 3.78, + "learning_rate": 1.1352677269576888e-05, + "loss": 0.042, + "step": 28894 + }, + { + "epoch": 3.79, + "learning_rate": 1.1352151816782621e-05, + "loss": 0.0428, + "step": 28895 + }, + { + "epoch": 3.79, + "learning_rate": 1.1351626360185502e-05, + "loss": 0.0335, + "step": 28896 + }, + { + "epoch": 3.79, + "learning_rate": 1.1351100899787005e-05, + "loss": 0.0417, + "step": 28897 + }, + { + "epoch": 3.79, + "learning_rate": 1.1350575435588612e-05, + "loss": 0.036, + "step": 28898 + }, + { + "epoch": 3.79, + "learning_rate": 1.1350049967591793e-05, + "loss": 0.0393, + "step": 28899 + }, + { + "epoch": 3.79, + "learning_rate": 1.134952449579803e-05, + "loss": 0.0337, + "step": 28900 + }, + { + "epoch": 3.79, + "learning_rate": 1.1348999020208802e-05, + "loss": 0.0393, + "step": 28901 + }, + { + "epoch": 3.79, + "learning_rate": 1.1348473540825591e-05, + "loss": 0.0315, + "step": 28902 + }, + { + "epoch": 3.79, + "learning_rate": 1.1347948057649865e-05, + "loss": 0.0393, + "step": 28903 + }, + { + "epoch": 3.79, + "learning_rate": 1.1347422570683106e-05, + "loss": 0.0318, + "step": 28904 + }, + { + "epoch": 3.79, + "learning_rate": 1.1346897079926798e-05, + "loss": 0.0274, + "step": 28905 + }, + { + "epoch": 3.79, + "learning_rate": 1.1346371585382407e-05, + "loss": 0.0381, + "step": 28906 + }, + { + "epoch": 3.79, + "learning_rate": 1.1345846087051422e-05, + "loss": 0.0399, + "step": 28907 + }, + { + "epoch": 3.79, + "learning_rate": 1.1345320584935315e-05, + "loss": 0.0424, + "step": 28908 + }, + { + "epoch": 3.79, + "learning_rate": 1.1344795079035568e-05, + "loss": 0.0344, + "step": 28909 + }, + { + "epoch": 3.79, + "learning_rate": 1.1344269569353654e-05, + "loss": 0.0354, + "step": 28910 + }, + { + "epoch": 3.79, + "learning_rate": 1.1343744055891056e-05, + "loss": 0.0367, + "step": 28911 + }, + { + "epoch": 3.79, + "learning_rate": 1.1343218538649245e-05, + "loss": 0.0392, + "step": 28912 + }, + { + "epoch": 3.79, + "learning_rate": 1.1342693017629707e-05, + "loss": 0.0366, + "step": 28913 + }, + { + "epoch": 3.79, + "learning_rate": 1.1342167492833916e-05, + "loss": 0.0397, + "step": 28914 + }, + { + "epoch": 3.79, + "learning_rate": 1.134164196426335e-05, + "loss": 0.0349, + "step": 28915 + }, + { + "epoch": 3.79, + "learning_rate": 1.1341116431919485e-05, + "loss": 0.0289, + "step": 28916 + }, + { + "epoch": 3.79, + "learning_rate": 1.1340590895803805e-05, + "loss": 0.039, + "step": 28917 + }, + { + "epoch": 3.79, + "learning_rate": 1.1340065355917782e-05, + "loss": 0.0386, + "step": 28918 + }, + { + "epoch": 3.79, + "learning_rate": 1.1339539812262899e-05, + "loss": 0.0365, + "step": 28919 + }, + { + "epoch": 3.79, + "learning_rate": 1.1339014264840629e-05, + "loss": 0.0357, + "step": 28920 + }, + { + "epoch": 3.79, + "learning_rate": 1.1338488713652453e-05, + "loss": 0.0351, + "step": 28921 + }, + { + "epoch": 3.79, + "learning_rate": 1.1337963158699852e-05, + "loss": 0.029, + "step": 28922 + }, + { + "epoch": 3.79, + "learning_rate": 1.1337437599984298e-05, + "loss": 0.0317, + "step": 28923 + }, + { + "epoch": 3.79, + "learning_rate": 1.133691203750727e-05, + "loss": 0.0379, + "step": 28924 + }, + { + "epoch": 3.79, + "learning_rate": 1.133638647127025e-05, + "loss": 0.033, + "step": 28925 + }, + { + "epoch": 3.79, + "learning_rate": 1.1335860901274718e-05, + "loss": 0.0385, + "step": 28926 + }, + { + "epoch": 3.79, + "learning_rate": 1.1335335327522144e-05, + "loss": 0.0374, + "step": 28927 + }, + { + "epoch": 3.79, + "learning_rate": 1.133480975001401e-05, + "loss": 0.0437, + "step": 28928 + }, + { + "epoch": 3.79, + "learning_rate": 1.1334284168751796e-05, + "loss": 0.0372, + "step": 28929 + }, + { + "epoch": 3.79, + "learning_rate": 1.133375858373698e-05, + "loss": 0.0364, + "step": 28930 + }, + { + "epoch": 3.79, + "learning_rate": 1.1333232994971039e-05, + "loss": 0.0396, + "step": 28931 + }, + { + "epoch": 3.79, + "learning_rate": 1.1332707402455452e-05, + "loss": 0.0382, + "step": 28932 + }, + { + "epoch": 3.79, + "learning_rate": 1.1332181806191693e-05, + "loss": 0.0377, + "step": 28933 + }, + { + "epoch": 3.79, + "learning_rate": 1.1331656206181245e-05, + "loss": 0.0396, + "step": 28934 + }, + { + "epoch": 3.79, + "learning_rate": 1.1331130602425585e-05, + "loss": 0.0364, + "step": 28935 + }, + { + "epoch": 3.79, + "learning_rate": 1.1330604994926193e-05, + "loss": 0.038, + "step": 28936 + }, + { + "epoch": 3.79, + "learning_rate": 1.1330079383684542e-05, + "loss": 0.0411, + "step": 28937 + }, + { + "epoch": 3.79, + "learning_rate": 1.1329553768702117e-05, + "loss": 0.0311, + "step": 28938 + }, + { + "epoch": 3.79, + "learning_rate": 1.1329028149980388e-05, + "loss": 0.0376, + "step": 28939 + }, + { + "epoch": 3.79, + "learning_rate": 1.1328502527520844e-05, + "loss": 0.0331, + "step": 28940 + }, + { + "epoch": 3.79, + "learning_rate": 1.1327976901324954e-05, + "loss": 0.0347, + "step": 28941 + }, + { + "epoch": 3.79, + "learning_rate": 1.1327451271394198e-05, + "loss": 0.0276, + "step": 28942 + }, + { + "epoch": 3.79, + "learning_rate": 1.1326925637730059e-05, + "loss": 0.0364, + "step": 28943 + }, + { + "epoch": 3.79, + "learning_rate": 1.1326400000334013e-05, + "loss": 0.0346, + "step": 28944 + }, + { + "epoch": 3.79, + "learning_rate": 1.1325874359207533e-05, + "loss": 0.0299, + "step": 28945 + }, + { + "epoch": 3.79, + "learning_rate": 1.1325348714352105e-05, + "loss": 0.031, + "step": 28946 + }, + { + "epoch": 3.79, + "learning_rate": 1.1324823065769206e-05, + "loss": 0.0402, + "step": 28947 + }, + { + "epoch": 3.79, + "learning_rate": 1.132429741346031e-05, + "loss": 0.0283, + "step": 28948 + }, + { + "epoch": 3.79, + "learning_rate": 1.1323771757426903e-05, + "loss": 0.0388, + "step": 28949 + }, + { + "epoch": 3.79, + "learning_rate": 1.132324609767045e-05, + "loss": 0.038, + "step": 28950 + }, + { + "epoch": 3.79, + "learning_rate": 1.1322720434192445e-05, + "loss": 0.034, + "step": 28951 + }, + { + "epoch": 3.79, + "learning_rate": 1.1322194766994354e-05, + "loss": 0.0386, + "step": 28952 + }, + { + "epoch": 3.79, + "learning_rate": 1.1321669096077665e-05, + "loss": 0.0433, + "step": 28953 + }, + { + "epoch": 3.79, + "learning_rate": 1.132114342144385e-05, + "loss": 0.0382, + "step": 28954 + }, + { + "epoch": 3.79, + "learning_rate": 1.1320617743094391e-05, + "loss": 0.0362, + "step": 28955 + }, + { + "epoch": 3.79, + "learning_rate": 1.132009206103076e-05, + "loss": 0.035, + "step": 28956 + }, + { + "epoch": 3.79, + "learning_rate": 1.1319566375254447e-05, + "loss": 0.0304, + "step": 28957 + }, + { + "epoch": 3.79, + "learning_rate": 1.131904068576692e-05, + "loss": 0.0362, + "step": 28958 + }, + { + "epoch": 3.79, + "learning_rate": 1.1318514992569664e-05, + "loss": 0.0339, + "step": 28959 + }, + { + "epoch": 3.79, + "learning_rate": 1.131798929566415e-05, + "loss": 0.0358, + "step": 28960 + }, + { + "epoch": 3.79, + "learning_rate": 1.1317463595051865e-05, + "loss": 0.04, + "step": 28961 + }, + { + "epoch": 3.79, + "learning_rate": 1.1316937890734284e-05, + "loss": 0.0344, + "step": 28962 + }, + { + "epoch": 3.79, + "learning_rate": 1.1316412182712886e-05, + "loss": 0.0316, + "step": 28963 + }, + { + "epoch": 3.79, + "learning_rate": 1.1315886470989146e-05, + "loss": 0.034, + "step": 28964 + }, + { + "epoch": 3.79, + "learning_rate": 1.1315360755564547e-05, + "loss": 0.031, + "step": 28965 + }, + { + "epoch": 3.79, + "learning_rate": 1.1314835036440567e-05, + "loss": 0.032, + "step": 28966 + }, + { + "epoch": 3.79, + "learning_rate": 1.131430931361868e-05, + "loss": 0.0387, + "step": 28967 + }, + { + "epoch": 3.79, + "learning_rate": 1.1313783587100369e-05, + "loss": 0.035, + "step": 28968 + }, + { + "epoch": 3.79, + "learning_rate": 1.1313257856887111e-05, + "loss": 0.0326, + "step": 28969 + }, + { + "epoch": 3.79, + "learning_rate": 1.131273212298039e-05, + "loss": 0.0365, + "step": 28970 + }, + { + "epoch": 3.79, + "learning_rate": 1.1312206385381676e-05, + "loss": 0.0328, + "step": 28971 + }, + { + "epoch": 3.8, + "learning_rate": 1.1311680644092451e-05, + "loss": 0.0419, + "step": 28972 + }, + { + "epoch": 3.8, + "learning_rate": 1.1311154899114195e-05, + "loss": 0.0436, + "step": 28973 + }, + { + "epoch": 3.8, + "learning_rate": 1.1310629150448387e-05, + "loss": 0.0333, + "step": 28974 + }, + { + "epoch": 3.8, + "learning_rate": 1.1310103398096503e-05, + "loss": 0.0311, + "step": 28975 + }, + { + "epoch": 3.8, + "learning_rate": 1.1309577642060023e-05, + "loss": 0.0415, + "step": 28976 + }, + { + "epoch": 3.8, + "learning_rate": 1.1309051882340426e-05, + "loss": 0.0355, + "step": 28977 + }, + { + "epoch": 3.8, + "learning_rate": 1.1308526118939189e-05, + "loss": 0.0326, + "step": 28978 + }, + { + "epoch": 3.8, + "learning_rate": 1.1308000351857794e-05, + "loss": 0.0402, + "step": 28979 + }, + { + "epoch": 3.8, + "learning_rate": 1.1307474581097718e-05, + "loss": 0.0364, + "step": 28980 + }, + { + "epoch": 3.8, + "learning_rate": 1.1306948806660436e-05, + "loss": 0.0348, + "step": 28981 + }, + { + "epoch": 3.8, + "learning_rate": 1.1306423028547431e-05, + "loss": 0.0357, + "step": 28982 + }, + { + "epoch": 3.8, + "learning_rate": 1.1305897246760182e-05, + "loss": 0.0352, + "step": 28983 + }, + { + "epoch": 3.8, + "learning_rate": 1.1305371461300168e-05, + "loss": 0.0391, + "step": 28984 + }, + { + "epoch": 3.8, + "learning_rate": 1.1304845672168866e-05, + "loss": 0.0406, + "step": 28985 + }, + { + "epoch": 3.8, + "learning_rate": 1.130431987936775e-05, + "loss": 0.034, + "step": 28986 + }, + { + "epoch": 3.8, + "learning_rate": 1.130379408289831e-05, + "loss": 0.046, + "step": 28987 + }, + { + "epoch": 3.8, + "learning_rate": 1.1303268282762017e-05, + "loss": 0.0316, + "step": 28988 + }, + { + "epoch": 3.8, + "learning_rate": 1.130274247896035e-05, + "loss": 0.0386, + "step": 28989 + }, + { + "epoch": 3.8, + "learning_rate": 1.1302216671494788e-05, + "loss": 0.0351, + "step": 28990 + }, + { + "epoch": 3.8, + "learning_rate": 1.1301690860366814e-05, + "loss": 0.0385, + "step": 28991 + }, + { + "epoch": 3.8, + "learning_rate": 1.1301165045577901e-05, + "loss": 0.0337, + "step": 28992 + }, + { + "epoch": 3.8, + "learning_rate": 1.1300639227129535e-05, + "loss": 0.0341, + "step": 28993 + }, + { + "epoch": 3.8, + "learning_rate": 1.1300113405023182e-05, + "loss": 0.043, + "step": 28994 + }, + { + "epoch": 3.8, + "learning_rate": 1.1299587579260337e-05, + "loss": 0.0385, + "step": 28995 + }, + { + "epoch": 3.8, + "learning_rate": 1.1299061749842467e-05, + "loss": 0.0293, + "step": 28996 + }, + { + "epoch": 3.8, + "learning_rate": 1.1298535916771061e-05, + "loss": 0.0425, + "step": 28997 + }, + { + "epoch": 3.8, + "learning_rate": 1.1298010080047585e-05, + "loss": 0.0301, + "step": 28998 + }, + { + "epoch": 3.8, + "learning_rate": 1.1297484239673528e-05, + "loss": 0.0339, + "step": 28999 + }, + { + "epoch": 3.8, + "learning_rate": 1.1296958395650366e-05, + "loss": 0.0445, + "step": 29000 + }, + { + "epoch": 3.8, + "learning_rate": 1.1296432547979576e-05, + "loss": 0.0377, + "step": 29001 + }, + { + "epoch": 3.8, + "learning_rate": 1.129590669666264e-05, + "loss": 0.0332, + "step": 29002 + }, + { + "epoch": 3.8, + "learning_rate": 1.1295380841701036e-05, + "loss": 0.0399, + "step": 29003 + }, + { + "epoch": 3.8, + "learning_rate": 1.1294854983096239e-05, + "loss": 0.0326, + "step": 29004 + }, + { + "epoch": 3.8, + "learning_rate": 1.1294329120849733e-05, + "loss": 0.0269, + "step": 29005 + }, + { + "epoch": 3.8, + "learning_rate": 1.1293803254962995e-05, + "loss": 0.043, + "step": 29006 + }, + { + "epoch": 3.8, + "learning_rate": 1.1293277385437504e-05, + "loss": 0.0391, + "step": 29007 + }, + { + "epoch": 3.8, + "learning_rate": 1.1292751512274743e-05, + "loss": 0.0386, + "step": 29008 + }, + { + "epoch": 3.8, + "learning_rate": 1.1292225635476184e-05, + "loss": 0.0374, + "step": 29009 + }, + { + "epoch": 3.8, + "learning_rate": 1.1291699755043309e-05, + "loss": 0.0354, + "step": 29010 + }, + { + "epoch": 3.8, + "learning_rate": 1.1291173870977597e-05, + "loss": 0.0434, + "step": 29011 + }, + { + "epoch": 3.8, + "learning_rate": 1.129064798328053e-05, + "loss": 0.0378, + "step": 29012 + }, + { + "epoch": 3.8, + "learning_rate": 1.1290122091953579e-05, + "loss": 0.0346, + "step": 29013 + }, + { + "epoch": 3.8, + "learning_rate": 1.1289596196998234e-05, + "loss": 0.034, + "step": 29014 + }, + { + "epoch": 3.8, + "learning_rate": 1.1289070298415965e-05, + "loss": 0.0397, + "step": 29015 + }, + { + "epoch": 3.8, + "learning_rate": 1.1288544396208256e-05, + "loss": 0.0289, + "step": 29016 + }, + { + "epoch": 3.8, + "learning_rate": 1.1288018490376582e-05, + "loss": 0.0338, + "step": 29017 + }, + { + "epoch": 3.8, + "learning_rate": 1.1287492580922427e-05, + "loss": 0.0297, + "step": 29018 + }, + { + "epoch": 3.8, + "learning_rate": 1.1286966667847268e-05, + "loss": 0.0305, + "step": 29019 + }, + { + "epoch": 3.8, + "learning_rate": 1.1286440751152586e-05, + "loss": 0.0404, + "step": 29020 + }, + { + "epoch": 3.8, + "learning_rate": 1.128591483083985e-05, + "loss": 0.0344, + "step": 29021 + }, + { + "epoch": 3.8, + "learning_rate": 1.1285388906910554e-05, + "loss": 0.0371, + "step": 29022 + }, + { + "epoch": 3.8, + "learning_rate": 1.1284862979366167e-05, + "loss": 0.03, + "step": 29023 + }, + { + "epoch": 3.8, + "learning_rate": 1.1284337048208173e-05, + "loss": 0.0367, + "step": 29024 + }, + { + "epoch": 3.8, + "learning_rate": 1.1283811113438051e-05, + "loss": 0.0362, + "step": 29025 + }, + { + "epoch": 3.8, + "learning_rate": 1.1283285175057275e-05, + "loss": 0.0303, + "step": 29026 + }, + { + "epoch": 3.8, + "learning_rate": 1.1282759233067329e-05, + "loss": 0.037, + "step": 29027 + }, + { + "epoch": 3.8, + "learning_rate": 1.1282233287469691e-05, + "loss": 0.0326, + "step": 29028 + }, + { + "epoch": 3.8, + "learning_rate": 1.128170733826584e-05, + "loss": 0.0347, + "step": 29029 + }, + { + "epoch": 3.8, + "learning_rate": 1.1281181385457253e-05, + "loss": 0.0404, + "step": 29030 + }, + { + "epoch": 3.8, + "learning_rate": 1.1280655429045417e-05, + "loss": 0.0317, + "step": 29031 + }, + { + "epoch": 3.8, + "learning_rate": 1.12801294690318e-05, + "loss": 0.0346, + "step": 29032 + }, + { + "epoch": 3.8, + "learning_rate": 1.1279603505417893e-05, + "loss": 0.0304, + "step": 29033 + }, + { + "epoch": 3.8, + "learning_rate": 1.1279077538205166e-05, + "loss": 0.0358, + "step": 29034 + }, + { + "epoch": 3.8, + "learning_rate": 1.1278551567395101e-05, + "loss": 0.0349, + "step": 29035 + }, + { + "epoch": 3.8, + "learning_rate": 1.1278025592989179e-05, + "loss": 0.0389, + "step": 29036 + }, + { + "epoch": 3.8, + "learning_rate": 1.1277499614988878e-05, + "loss": 0.0313, + "step": 29037 + }, + { + "epoch": 3.8, + "learning_rate": 1.1276973633395676e-05, + "loss": 0.0457, + "step": 29038 + }, + { + "epoch": 3.8, + "learning_rate": 1.1276447648211058e-05, + "loss": 0.0381, + "step": 29039 + }, + { + "epoch": 3.8, + "learning_rate": 1.1275921659436494e-05, + "loss": 0.0402, + "step": 29040 + }, + { + "epoch": 3.8, + "learning_rate": 1.127539566707347e-05, + "loss": 0.0362, + "step": 29041 + }, + { + "epoch": 3.8, + "learning_rate": 1.1274869671123465e-05, + "loss": 0.0352, + "step": 29042 + }, + { + "epoch": 3.8, + "learning_rate": 1.1274343671587956e-05, + "loss": 0.0329, + "step": 29043 + }, + { + "epoch": 3.8, + "learning_rate": 1.127381766846842e-05, + "loss": 0.0385, + "step": 29044 + }, + { + "epoch": 3.8, + "learning_rate": 1.1273291661766343e-05, + "loss": 0.0341, + "step": 29045 + }, + { + "epoch": 3.8, + "learning_rate": 1.1272765651483202e-05, + "loss": 0.0399, + "step": 29046 + }, + { + "epoch": 3.8, + "learning_rate": 1.1272239637620472e-05, + "loss": 0.039, + "step": 29047 + }, + { + "epoch": 3.81, + "learning_rate": 1.1271713620179636e-05, + "loss": 0.0318, + "step": 29048 + }, + { + "epoch": 3.81, + "learning_rate": 1.1271187599162177e-05, + "loss": 0.0336, + "step": 29049 + }, + { + "epoch": 3.81, + "learning_rate": 1.1270661574569569e-05, + "loss": 0.031, + "step": 29050 + }, + { + "epoch": 3.81, + "learning_rate": 1.1270135546403292e-05, + "loss": 0.0353, + "step": 29051 + }, + { + "epoch": 3.81, + "learning_rate": 1.1269609514664828e-05, + "loss": 0.0313, + "step": 29052 + }, + { + "epoch": 3.81, + "learning_rate": 1.126908347935565e-05, + "loss": 0.0377, + "step": 29053 + }, + { + "epoch": 3.81, + "learning_rate": 1.1268557440477248e-05, + "loss": 0.0378, + "step": 29054 + }, + { + "epoch": 3.81, + "learning_rate": 1.1268031398031091e-05, + "loss": 0.0315, + "step": 29055 + }, + { + "epoch": 3.81, + "learning_rate": 1.1267505352018669e-05, + "loss": 0.0385, + "step": 29056 + }, + { + "epoch": 3.81, + "learning_rate": 1.1266979302441452e-05, + "loss": 0.0385, + "step": 29057 + }, + { + "epoch": 3.81, + "learning_rate": 1.1266453249300925e-05, + "loss": 0.0323, + "step": 29058 + }, + { + "epoch": 3.81, + "learning_rate": 1.1265927192598562e-05, + "loss": 0.0326, + "step": 29059 + }, + { + "epoch": 3.81, + "learning_rate": 1.1265401132335852e-05, + "loss": 0.0332, + "step": 29060 + }, + { + "epoch": 3.81, + "learning_rate": 1.1264875068514263e-05, + "loss": 0.0356, + "step": 29061 + }, + { + "epoch": 3.81, + "learning_rate": 1.1264349001135284e-05, + "loss": 0.0352, + "step": 29062 + }, + { + "epoch": 3.81, + "learning_rate": 1.126382293020039e-05, + "loss": 0.0357, + "step": 29063 + }, + { + "epoch": 3.81, + "learning_rate": 1.126329685571106e-05, + "loss": 0.0316, + "step": 29064 + }, + { + "epoch": 3.81, + "learning_rate": 1.1262770777668775e-05, + "loss": 0.039, + "step": 29065 + }, + { + "epoch": 3.81, + "learning_rate": 1.1262244696075016e-05, + "loss": 0.0311, + "step": 29066 + }, + { + "epoch": 3.81, + "learning_rate": 1.1261718610931258e-05, + "loss": 0.0419, + "step": 29067 + }, + { + "epoch": 3.81, + "learning_rate": 1.1261192522238988e-05, + "loss": 0.0357, + "step": 29068 + }, + { + "epoch": 3.81, + "learning_rate": 1.1260666429999678e-05, + "loss": 0.0387, + "step": 29069 + }, + { + "epoch": 3.81, + "learning_rate": 1.126014033421481e-05, + "loss": 0.0281, + "step": 29070 + }, + { + "epoch": 3.81, + "learning_rate": 1.1259614234885868e-05, + "loss": 0.037, + "step": 29071 + }, + { + "epoch": 3.81, + "learning_rate": 1.1259088132014324e-05, + "loss": 0.0317, + "step": 29072 + }, + { + "epoch": 3.81, + "learning_rate": 1.1258562025601664e-05, + "loss": 0.0338, + "step": 29073 + }, + { + "epoch": 3.81, + "learning_rate": 1.1258035915649366e-05, + "loss": 0.0408, + "step": 29074 + }, + { + "epoch": 3.81, + "learning_rate": 1.1257509802158908e-05, + "loss": 0.0356, + "step": 29075 + }, + { + "epoch": 3.81, + "learning_rate": 1.1256983685131769e-05, + "loss": 0.0302, + "step": 29076 + }, + { + "epoch": 3.81, + "learning_rate": 1.1256457564569434e-05, + "loss": 0.0411, + "step": 29077 + }, + { + "epoch": 3.81, + "learning_rate": 1.1255931440473375e-05, + "loss": 0.0289, + "step": 29078 + }, + { + "epoch": 3.81, + "learning_rate": 1.1255405312845081e-05, + "loss": 0.0337, + "step": 29079 + }, + { + "epoch": 3.81, + "learning_rate": 1.125487918168602e-05, + "loss": 0.0341, + "step": 29080 + }, + { + "epoch": 3.81, + "learning_rate": 1.125435304699768e-05, + "loss": 0.0358, + "step": 29081 + }, + { + "epoch": 3.81, + "learning_rate": 1.1253826908781542e-05, + "loss": 0.0323, + "step": 29082 + }, + { + "epoch": 3.81, + "learning_rate": 1.125330076703908e-05, + "loss": 0.0315, + "step": 29083 + }, + { + "epoch": 3.81, + "learning_rate": 1.1252774621771778e-05, + "loss": 0.0355, + "step": 29084 + }, + { + "epoch": 3.81, + "learning_rate": 1.1252248472981116e-05, + "loss": 0.0407, + "step": 29085 + }, + { + "epoch": 3.81, + "learning_rate": 1.1251722320668567e-05, + "loss": 0.0342, + "step": 29086 + }, + { + "epoch": 3.81, + "learning_rate": 1.1251196164835616e-05, + "loss": 0.037, + "step": 29087 + }, + { + "epoch": 3.81, + "learning_rate": 1.1250670005483744e-05, + "loss": 0.0371, + "step": 29088 + }, + { + "epoch": 3.81, + "learning_rate": 1.1250143842614432e-05, + "loss": 0.0275, + "step": 29089 + }, + { + "epoch": 3.81, + "learning_rate": 1.1249617676229151e-05, + "loss": 0.0344, + "step": 29090 + }, + { + "epoch": 3.81, + "learning_rate": 1.124909150632939e-05, + "loss": 0.0366, + "step": 29091 + }, + { + "epoch": 3.81, + "learning_rate": 1.1248565332916625e-05, + "loss": 0.0367, + "step": 29092 + }, + { + "epoch": 3.81, + "learning_rate": 1.124803915599234e-05, + "loss": 0.0375, + "step": 29093 + }, + { + "epoch": 3.81, + "learning_rate": 1.1247512975558008e-05, + "loss": 0.0291, + "step": 29094 + }, + { + "epoch": 3.81, + "learning_rate": 1.1246986791615111e-05, + "loss": 0.0335, + "step": 29095 + }, + { + "epoch": 3.81, + "learning_rate": 1.1246460604165135e-05, + "loss": 0.0328, + "step": 29096 + }, + { + "epoch": 3.81, + "learning_rate": 1.124593441320955e-05, + "loss": 0.029, + "step": 29097 + }, + { + "epoch": 3.81, + "learning_rate": 1.1245408218749843e-05, + "loss": 0.0374, + "step": 29098 + }, + { + "epoch": 3.81, + "learning_rate": 1.1244882020787488e-05, + "loss": 0.0319, + "step": 29099 + }, + { + "epoch": 3.81, + "learning_rate": 1.1244355819323976e-05, + "loss": 0.0343, + "step": 29100 + }, + { + "epoch": 3.81, + "learning_rate": 1.1243829614360772e-05, + "loss": 0.043, + "step": 29101 + }, + { + "epoch": 3.81, + "learning_rate": 1.124330340589937e-05, + "loss": 0.0357, + "step": 29102 + }, + { + "epoch": 3.81, + "learning_rate": 1.1242777193941236e-05, + "loss": 0.0383, + "step": 29103 + }, + { + "epoch": 3.81, + "learning_rate": 1.1242250978487865e-05, + "loss": 0.0326, + "step": 29104 + }, + { + "epoch": 3.81, + "learning_rate": 1.1241724759540724e-05, + "loss": 0.0407, + "step": 29105 + }, + { + "epoch": 3.81, + "learning_rate": 1.1241198537101303e-05, + "loss": 0.0305, + "step": 29106 + }, + { + "epoch": 3.81, + "learning_rate": 1.1240672311171073e-05, + "loss": 0.0389, + "step": 29107 + }, + { + "epoch": 3.81, + "learning_rate": 1.1240146081751519e-05, + "loss": 0.0318, + "step": 29108 + }, + { + "epoch": 3.81, + "learning_rate": 1.1239619848844119e-05, + "loss": 0.0356, + "step": 29109 + }, + { + "epoch": 3.81, + "learning_rate": 1.1239093612450359e-05, + "loss": 0.0402, + "step": 29110 + }, + { + "epoch": 3.81, + "learning_rate": 1.123856737257171e-05, + "loss": 0.0336, + "step": 29111 + }, + { + "epoch": 3.81, + "learning_rate": 1.1238041129209655e-05, + "loss": 0.0322, + "step": 29112 + }, + { + "epoch": 3.81, + "learning_rate": 1.123751488236568e-05, + "loss": 0.0326, + "step": 29113 + }, + { + "epoch": 3.81, + "learning_rate": 1.1236988632041257e-05, + "loss": 0.0373, + "step": 29114 + }, + { + "epoch": 3.81, + "learning_rate": 1.123646237823787e-05, + "loss": 0.0384, + "step": 29115 + }, + { + "epoch": 3.81, + "learning_rate": 1.1235936120956998e-05, + "loss": 0.034, + "step": 29116 + }, + { + "epoch": 3.81, + "learning_rate": 1.1235409860200124e-05, + "loss": 0.0328, + "step": 29117 + }, + { + "epoch": 3.81, + "learning_rate": 1.1234883595968722e-05, + "loss": 0.0306, + "step": 29118 + }, + { + "epoch": 3.81, + "learning_rate": 1.1234357328264277e-05, + "loss": 0.0382, + "step": 29119 + }, + { + "epoch": 3.81, + "learning_rate": 1.1233831057088266e-05, + "loss": 0.0357, + "step": 29120 + }, + { + "epoch": 3.81, + "learning_rate": 1.1233304782442176e-05, + "loss": 0.0333, + "step": 29121 + }, + { + "epoch": 3.81, + "learning_rate": 1.1232778504327478e-05, + "loss": 0.0347, + "step": 29122 + }, + { + "epoch": 3.81, + "learning_rate": 1.1232252222745658e-05, + "loss": 0.035, + "step": 29123 + }, + { + "epoch": 3.82, + "learning_rate": 1.1231725937698193e-05, + "loss": 0.0313, + "step": 29124 + }, + { + "epoch": 3.82, + "learning_rate": 1.1231199649186563e-05, + "loss": 0.0359, + "step": 29125 + }, + { + "epoch": 3.82, + "learning_rate": 1.1230673357212251e-05, + "loss": 0.0357, + "step": 29126 + }, + { + "epoch": 3.82, + "learning_rate": 1.1230147061776737e-05, + "loss": 0.0404, + "step": 29127 + }, + { + "epoch": 3.82, + "learning_rate": 1.12296207628815e-05, + "loss": 0.0444, + "step": 29128 + }, + { + "epoch": 3.82, + "learning_rate": 1.1229094460528018e-05, + "loss": 0.0379, + "step": 29129 + }, + { + "epoch": 3.82, + "learning_rate": 1.1228568154717771e-05, + "loss": 0.0469, + "step": 29130 + }, + { + "epoch": 3.82, + "learning_rate": 1.1228041845452249e-05, + "loss": 0.0336, + "step": 29131 + }, + { + "epoch": 3.82, + "learning_rate": 1.1227515532732917e-05, + "loss": 0.0329, + "step": 29132 + }, + { + "epoch": 3.82, + "learning_rate": 1.1226989216561265e-05, + "loss": 0.0358, + "step": 29133 + }, + { + "epoch": 3.82, + "learning_rate": 1.1226462896938776e-05, + "loss": 0.037, + "step": 29134 + }, + { + "epoch": 3.82, + "learning_rate": 1.122593657386692e-05, + "loss": 0.0343, + "step": 29135 + }, + { + "epoch": 3.82, + "learning_rate": 1.1225410247347187e-05, + "loss": 0.0346, + "step": 29136 + }, + { + "epoch": 3.82, + "learning_rate": 1.1224883917381048e-05, + "loss": 0.039, + "step": 29137 + }, + { + "epoch": 3.82, + "learning_rate": 1.1224357583969992e-05, + "loss": 0.0342, + "step": 29138 + }, + { + "epoch": 3.82, + "learning_rate": 1.1223831247115493e-05, + "loss": 0.0345, + "step": 29139 + }, + { + "epoch": 3.82, + "learning_rate": 1.1223304906819037e-05, + "loss": 0.0336, + "step": 29140 + }, + { + "epoch": 3.82, + "learning_rate": 1.1222778563082096e-05, + "loss": 0.0373, + "step": 29141 + }, + { + "epoch": 3.82, + "learning_rate": 1.122225221590616e-05, + "loss": 0.0377, + "step": 29142 + }, + { + "epoch": 3.82, + "learning_rate": 1.1221725865292704e-05, + "loss": 0.0305, + "step": 29143 + }, + { + "epoch": 3.82, + "learning_rate": 1.1221199511243208e-05, + "loss": 0.0416, + "step": 29144 + }, + { + "epoch": 3.82, + "learning_rate": 1.1220673153759154e-05, + "loss": 0.0369, + "step": 29145 + }, + { + "epoch": 3.82, + "learning_rate": 1.1220146792842021e-05, + "loss": 0.0361, + "step": 29146 + }, + { + "epoch": 3.82, + "learning_rate": 1.121962042849329e-05, + "loss": 0.0336, + "step": 29147 + }, + { + "epoch": 3.82, + "learning_rate": 1.1219094060714442e-05, + "loss": 0.0353, + "step": 29148 + }, + { + "epoch": 3.82, + "learning_rate": 1.1218567689506956e-05, + "loss": 0.0453, + "step": 29149 + }, + { + "epoch": 3.82, + "learning_rate": 1.1218041314872317e-05, + "loss": 0.0315, + "step": 29150 + }, + { + "epoch": 3.82, + "learning_rate": 1.1217514936811998e-05, + "loss": 0.0319, + "step": 29151 + }, + { + "epoch": 3.82, + "learning_rate": 1.1216988555327484e-05, + "loss": 0.035, + "step": 29152 + }, + { + "epoch": 3.82, + "learning_rate": 1.1216462170420252e-05, + "loss": 0.037, + "step": 29153 + }, + { + "epoch": 3.82, + "learning_rate": 1.121593578209179e-05, + "loss": 0.037, + "step": 29154 + }, + { + "epoch": 3.82, + "learning_rate": 1.121540939034357e-05, + "loss": 0.0373, + "step": 29155 + }, + { + "epoch": 3.82, + "learning_rate": 1.1214882995177074e-05, + "loss": 0.0326, + "step": 29156 + }, + { + "epoch": 3.82, + "learning_rate": 1.1214356596593789e-05, + "loss": 0.0364, + "step": 29157 + }, + { + "epoch": 3.82, + "learning_rate": 1.1213830194595186e-05, + "loss": 0.0368, + "step": 29158 + }, + { + "epoch": 3.82, + "learning_rate": 1.1213303789182755e-05, + "loss": 0.0371, + "step": 29159 + }, + { + "epoch": 3.82, + "learning_rate": 1.1212777380357969e-05, + "loss": 0.0372, + "step": 29160 + }, + { + "epoch": 3.82, + "learning_rate": 1.1212250968122313e-05, + "loss": 0.0378, + "step": 29161 + }, + { + "epoch": 3.82, + "learning_rate": 1.1211724552477264e-05, + "loss": 0.0329, + "step": 29162 + }, + { + "epoch": 3.82, + "learning_rate": 1.1211198133424304e-05, + "loss": 0.0381, + "step": 29163 + }, + { + "epoch": 3.82, + "learning_rate": 1.1210671710964914e-05, + "loss": 0.0414, + "step": 29164 + }, + { + "epoch": 3.82, + "learning_rate": 1.1210145285100576e-05, + "loss": 0.0309, + "step": 29165 + }, + { + "epoch": 3.82, + "learning_rate": 1.1209618855832767e-05, + "loss": 0.0369, + "step": 29166 + }, + { + "epoch": 3.82, + "learning_rate": 1.120909242316297e-05, + "loss": 0.0266, + "step": 29167 + }, + { + "epoch": 3.82, + "learning_rate": 1.1208565987092661e-05, + "loss": 0.0351, + "step": 29168 + }, + { + "epoch": 3.82, + "learning_rate": 1.1208039547623331e-05, + "loss": 0.037, + "step": 29169 + }, + { + "epoch": 3.82, + "learning_rate": 1.120751310475645e-05, + "loss": 0.0362, + "step": 29170 + }, + { + "epoch": 3.82, + "learning_rate": 1.1206986658493506e-05, + "loss": 0.0393, + "step": 29171 + }, + { + "epoch": 3.82, + "learning_rate": 1.1206460208835974e-05, + "loss": 0.0399, + "step": 29172 + }, + { + "epoch": 3.82, + "learning_rate": 1.1205933755785336e-05, + "loss": 0.0299, + "step": 29173 + }, + { + "epoch": 3.82, + "learning_rate": 1.1205407299343074e-05, + "loss": 0.0355, + "step": 29174 + }, + { + "epoch": 3.82, + "learning_rate": 1.1204880839510672e-05, + "loss": 0.0344, + "step": 29175 + }, + { + "epoch": 3.82, + "learning_rate": 1.1204354376289603e-05, + "loss": 0.0347, + "step": 29176 + }, + { + "epoch": 3.82, + "learning_rate": 1.120382790968135e-05, + "loss": 0.0317, + "step": 29177 + }, + { + "epoch": 3.82, + "learning_rate": 1.12033014396874e-05, + "loss": 0.0356, + "step": 29178 + }, + { + "epoch": 3.82, + "learning_rate": 1.1202774966309224e-05, + "loss": 0.0379, + "step": 29179 + }, + { + "epoch": 3.82, + "learning_rate": 1.120224848954831e-05, + "loss": 0.0353, + "step": 29180 + }, + { + "epoch": 3.82, + "learning_rate": 1.1201722009406135e-05, + "loss": 0.0366, + "step": 29181 + }, + { + "epoch": 3.82, + "learning_rate": 1.1201195525884183e-05, + "loss": 0.0403, + "step": 29182 + }, + { + "epoch": 3.82, + "learning_rate": 1.1200669038983932e-05, + "loss": 0.0304, + "step": 29183 + }, + { + "epoch": 3.82, + "learning_rate": 1.1200142548706862e-05, + "loss": 0.0313, + "step": 29184 + }, + { + "epoch": 3.82, + "learning_rate": 1.1199616055054454e-05, + "loss": 0.0341, + "step": 29185 + }, + { + "epoch": 3.82, + "learning_rate": 1.1199089558028193e-05, + "loss": 0.0307, + "step": 29186 + }, + { + "epoch": 3.82, + "learning_rate": 1.1198563057629552e-05, + "loss": 0.0367, + "step": 29187 + }, + { + "epoch": 3.82, + "learning_rate": 1.1198036553860021e-05, + "loss": 0.0378, + "step": 29188 + }, + { + "epoch": 3.82, + "learning_rate": 1.1197510046721073e-05, + "loss": 0.0344, + "step": 29189 + }, + { + "epoch": 3.82, + "learning_rate": 1.1196983536214192e-05, + "loss": 0.0356, + "step": 29190 + }, + { + "epoch": 3.82, + "learning_rate": 1.119645702234086e-05, + "loss": 0.0321, + "step": 29191 + }, + { + "epoch": 3.82, + "learning_rate": 1.1195930505102556e-05, + "loss": 0.0364, + "step": 29192 + }, + { + "epoch": 3.82, + "learning_rate": 1.1195403984500761e-05, + "loss": 0.0457, + "step": 29193 + }, + { + "epoch": 3.82, + "learning_rate": 1.1194877460536957e-05, + "loss": 0.0309, + "step": 29194 + }, + { + "epoch": 3.82, + "learning_rate": 1.1194350933212621e-05, + "loss": 0.0259, + "step": 29195 + }, + { + "epoch": 3.82, + "learning_rate": 1.1193824402529237e-05, + "loss": 0.0323, + "step": 29196 + }, + { + "epoch": 3.82, + "learning_rate": 1.1193297868488287e-05, + "loss": 0.034, + "step": 29197 + }, + { + "epoch": 3.82, + "learning_rate": 1.1192771331091248e-05, + "loss": 0.0406, + "step": 29198 + }, + { + "epoch": 3.82, + "learning_rate": 1.1192244790339605e-05, + "loss": 0.0336, + "step": 29199 + }, + { + "epoch": 3.82, + "learning_rate": 1.1191718246234837e-05, + "loss": 0.0398, + "step": 29200 + }, + { + "epoch": 3.83, + "learning_rate": 1.1191191698778425e-05, + "loss": 0.0367, + "step": 29201 + }, + { + "epoch": 3.83, + "learning_rate": 1.1190665147971848e-05, + "loss": 0.0336, + "step": 29202 + }, + { + "epoch": 3.83, + "learning_rate": 1.1190138593816593e-05, + "loss": 0.0297, + "step": 29203 + }, + { + "epoch": 3.83, + "learning_rate": 1.1189612036314132e-05, + "loss": 0.0408, + "step": 29204 + }, + { + "epoch": 3.83, + "learning_rate": 1.1189085475465955e-05, + "loss": 0.0404, + "step": 29205 + }, + { + "epoch": 3.83, + "learning_rate": 1.1188558911273535e-05, + "loss": 0.0425, + "step": 29206 + }, + { + "epoch": 3.83, + "learning_rate": 1.1188032343738355e-05, + "loss": 0.0373, + "step": 29207 + }, + { + "epoch": 3.83, + "learning_rate": 1.11875057728619e-05, + "loss": 0.0389, + "step": 29208 + }, + { + "epoch": 3.83, + "learning_rate": 1.1186979198645649e-05, + "loss": 0.0344, + "step": 29209 + }, + { + "epoch": 3.83, + "learning_rate": 1.118645262109108e-05, + "loss": 0.0318, + "step": 29210 + }, + { + "epoch": 3.83, + "learning_rate": 1.1185926040199679e-05, + "loss": 0.0347, + "step": 29211 + }, + { + "epoch": 3.83, + "learning_rate": 1.1185399455972918e-05, + "loss": 0.0394, + "step": 29212 + }, + { + "epoch": 3.83, + "learning_rate": 1.1184872868412291e-05, + "loss": 0.0337, + "step": 29213 + }, + { + "epoch": 3.83, + "learning_rate": 1.118434627751927e-05, + "loss": 0.0445, + "step": 29214 + }, + { + "epoch": 3.83, + "learning_rate": 1.1183819683295339e-05, + "loss": 0.0365, + "step": 29215 + }, + { + "epoch": 3.83, + "learning_rate": 1.1183293085741975e-05, + "loss": 0.0376, + "step": 29216 + }, + { + "epoch": 3.83, + "learning_rate": 1.1182766484860662e-05, + "loss": 0.0359, + "step": 29217 + }, + { + "epoch": 3.83, + "learning_rate": 1.1182239880652884e-05, + "loss": 0.0375, + "step": 29218 + }, + { + "epoch": 3.83, + "learning_rate": 1.1181713273120121e-05, + "loss": 0.0311, + "step": 29219 + }, + { + "epoch": 3.83, + "learning_rate": 1.1181186662263847e-05, + "loss": 0.0299, + "step": 29220 + }, + { + "epoch": 3.83, + "learning_rate": 1.1180660048085552e-05, + "loss": 0.0347, + "step": 29221 + }, + { + "epoch": 3.83, + "learning_rate": 1.1180133430586715e-05, + "loss": 0.0375, + "step": 29222 + }, + { + "epoch": 3.83, + "learning_rate": 1.117960680976881e-05, + "loss": 0.035, + "step": 29223 + }, + { + "epoch": 3.83, + "learning_rate": 1.1179080185633328e-05, + "loss": 0.0347, + "step": 29224 + }, + { + "epoch": 3.83, + "learning_rate": 1.1178553558181744e-05, + "loss": 0.043, + "step": 29225 + }, + { + "epoch": 3.83, + "learning_rate": 1.1178026927415544e-05, + "loss": 0.0387, + "step": 29226 + }, + { + "epoch": 3.83, + "learning_rate": 1.1177500293336203e-05, + "loss": 0.0401, + "step": 29227 + }, + { + "epoch": 3.83, + "learning_rate": 1.1176973655945207e-05, + "loss": 0.039, + "step": 29228 + }, + { + "epoch": 3.83, + "learning_rate": 1.1176447015244032e-05, + "loss": 0.0384, + "step": 29229 + }, + { + "epoch": 3.83, + "learning_rate": 1.1175920371234165e-05, + "loss": 0.0378, + "step": 29230 + }, + { + "epoch": 3.83, + "learning_rate": 1.1175393723917085e-05, + "loss": 0.0373, + "step": 29231 + }, + { + "epoch": 3.83, + "learning_rate": 1.1174867073294273e-05, + "loss": 0.0288, + "step": 29232 + }, + { + "epoch": 3.83, + "learning_rate": 1.1174340419367207e-05, + "loss": 0.0355, + "step": 29233 + }, + { + "epoch": 3.83, + "learning_rate": 1.1173813762137373e-05, + "loss": 0.0329, + "step": 29234 + }, + { + "epoch": 3.83, + "learning_rate": 1.117328710160625e-05, + "loss": 0.0304, + "step": 29235 + }, + { + "epoch": 3.83, + "learning_rate": 1.1172760437775321e-05, + "loss": 0.0333, + "step": 29236 + }, + { + "epoch": 3.83, + "learning_rate": 1.1172233770646064e-05, + "loss": 0.0371, + "step": 29237 + }, + { + "epoch": 3.83, + "learning_rate": 1.1171707100219962e-05, + "loss": 0.0347, + "step": 29238 + }, + { + "epoch": 3.83, + "learning_rate": 1.1171180426498496e-05, + "loss": 0.0367, + "step": 29239 + }, + { + "epoch": 3.83, + "learning_rate": 1.1170653749483149e-05, + "loss": 0.0357, + "step": 29240 + }, + { + "epoch": 3.83, + "learning_rate": 1.1170127069175397e-05, + "loss": 0.0352, + "step": 29241 + }, + { + "epoch": 3.83, + "learning_rate": 1.1169600385576728e-05, + "loss": 0.0387, + "step": 29242 + }, + { + "epoch": 3.83, + "learning_rate": 1.1169073698688622e-05, + "loss": 0.0364, + "step": 29243 + }, + { + "epoch": 3.83, + "learning_rate": 1.1168547008512556e-05, + "loss": 0.037, + "step": 29244 + }, + { + "epoch": 3.83, + "learning_rate": 1.1168020315050012e-05, + "loss": 0.034, + "step": 29245 + }, + { + "epoch": 3.83, + "learning_rate": 1.1167493618302476e-05, + "loss": 0.0434, + "step": 29246 + }, + { + "epoch": 3.83, + "learning_rate": 1.1166966918271427e-05, + "loss": 0.0353, + "step": 29247 + }, + { + "epoch": 3.83, + "learning_rate": 1.1166440214958342e-05, + "loss": 0.0318, + "step": 29248 + }, + { + "epoch": 3.83, + "learning_rate": 1.1165913508364709e-05, + "loss": 0.033, + "step": 29249 + }, + { + "epoch": 3.83, + "learning_rate": 1.1165386798492005e-05, + "loss": 0.0382, + "step": 29250 + }, + { + "epoch": 3.83, + "learning_rate": 1.1164860085341713e-05, + "loss": 0.036, + "step": 29251 + }, + { + "epoch": 3.83, + "learning_rate": 1.1164333368915315e-05, + "loss": 0.038, + "step": 29252 + }, + { + "epoch": 3.83, + "learning_rate": 1.1163806649214292e-05, + "loss": 0.0429, + "step": 29253 + }, + { + "epoch": 3.83, + "learning_rate": 1.1163279926240123e-05, + "loss": 0.035, + "step": 29254 + }, + { + "epoch": 3.83, + "learning_rate": 1.1162753199994292e-05, + "loss": 0.0368, + "step": 29255 + }, + { + "epoch": 3.83, + "learning_rate": 1.1162226470478276e-05, + "loss": 0.0386, + "step": 29256 + }, + { + "epoch": 3.83, + "learning_rate": 1.1161699737693566e-05, + "loss": 0.04, + "step": 29257 + }, + { + "epoch": 3.83, + "learning_rate": 1.1161173001641635e-05, + "loss": 0.0265, + "step": 29258 + }, + { + "epoch": 3.83, + "learning_rate": 1.1160646262323965e-05, + "loss": 0.0393, + "step": 29259 + }, + { + "epoch": 3.83, + "learning_rate": 1.1160119519742041e-05, + "loss": 0.0357, + "step": 29260 + }, + { + "epoch": 3.83, + "learning_rate": 1.1159592773897341e-05, + "loss": 0.0387, + "step": 29261 + }, + { + "epoch": 3.83, + "learning_rate": 1.115906602479135e-05, + "loss": 0.0387, + "step": 29262 + }, + { + "epoch": 3.83, + "learning_rate": 1.1158539272425544e-05, + "loss": 0.0319, + "step": 29263 + }, + { + "epoch": 3.83, + "learning_rate": 1.1158012516801413e-05, + "loss": 0.0365, + "step": 29264 + }, + { + "epoch": 3.83, + "learning_rate": 1.115748575792043e-05, + "loss": 0.0403, + "step": 29265 + }, + { + "epoch": 3.83, + "learning_rate": 1.1156958995784081e-05, + "loss": 0.0314, + "step": 29266 + }, + { + "epoch": 3.83, + "learning_rate": 1.1156432230393846e-05, + "loss": 0.0368, + "step": 29267 + }, + { + "epoch": 3.83, + "learning_rate": 1.1155905461751209e-05, + "loss": 0.0306, + "step": 29268 + }, + { + "epoch": 3.83, + "learning_rate": 1.1155378689857645e-05, + "loss": 0.0354, + "step": 29269 + }, + { + "epoch": 3.83, + "learning_rate": 1.1154851914714645e-05, + "loss": 0.0317, + "step": 29270 + }, + { + "epoch": 3.83, + "learning_rate": 1.1154325136323682e-05, + "loss": 0.034, + "step": 29271 + }, + { + "epoch": 3.83, + "learning_rate": 1.1153798354686242e-05, + "loss": 0.0278, + "step": 29272 + }, + { + "epoch": 3.83, + "learning_rate": 1.1153271569803805e-05, + "loss": 0.0344, + "step": 29273 + }, + { + "epoch": 3.83, + "learning_rate": 1.1152744781677858e-05, + "loss": 0.0393, + "step": 29274 + }, + { + "epoch": 3.83, + "learning_rate": 1.1152217990309873e-05, + "loss": 0.0294, + "step": 29275 + }, + { + "epoch": 3.83, + "learning_rate": 1.1151691195701337e-05, + "loss": 0.0361, + "step": 29276 + }, + { + "epoch": 3.84, + "learning_rate": 1.115116439785373e-05, + "loss": 0.0296, + "step": 29277 + }, + { + "epoch": 3.84, + "learning_rate": 1.1150637596768534e-05, + "loss": 0.0306, + "step": 29278 + }, + { + "epoch": 3.84, + "learning_rate": 1.1150110792447233e-05, + "loss": 0.0429, + "step": 29279 + }, + { + "epoch": 3.84, + "learning_rate": 1.1149583984891308e-05, + "loss": 0.0325, + "step": 29280 + }, + { + "epoch": 3.84, + "learning_rate": 1.1149057174102236e-05, + "loss": 0.029, + "step": 29281 + }, + { + "epoch": 3.84, + "learning_rate": 1.1148530360081502e-05, + "loss": 0.0386, + "step": 29282 + }, + { + "epoch": 3.84, + "learning_rate": 1.1148003542830587e-05, + "loss": 0.0343, + "step": 29283 + }, + { + "epoch": 3.84, + "learning_rate": 1.1147476722350974e-05, + "loss": 0.0335, + "step": 29284 + }, + { + "epoch": 3.84, + "learning_rate": 1.1146949898644146e-05, + "loss": 0.0391, + "step": 29285 + }, + { + "epoch": 3.84, + "learning_rate": 1.1146423071711581e-05, + "loss": 0.0321, + "step": 29286 + }, + { + "epoch": 3.84, + "learning_rate": 1.1145896241554764e-05, + "loss": 0.0393, + "step": 29287 + }, + { + "epoch": 3.84, + "learning_rate": 1.1145369408175172e-05, + "loss": 0.037, + "step": 29288 + }, + { + "epoch": 3.84, + "learning_rate": 1.114484257157429e-05, + "loss": 0.036, + "step": 29289 + }, + { + "epoch": 3.84, + "learning_rate": 1.1144315731753598e-05, + "loss": 0.0462, + "step": 29290 + }, + { + "epoch": 3.84, + "learning_rate": 1.1143788888714584e-05, + "loss": 0.0351, + "step": 29291 + }, + { + "epoch": 3.84, + "learning_rate": 1.114326204245872e-05, + "loss": 0.0369, + "step": 29292 + }, + { + "epoch": 3.84, + "learning_rate": 1.1142735192987497e-05, + "loss": 0.0363, + "step": 29293 + }, + { + "epoch": 3.84, + "learning_rate": 1.1142208340302385e-05, + "loss": 0.0398, + "step": 29294 + }, + { + "epoch": 3.84, + "learning_rate": 1.1141681484404878e-05, + "loss": 0.0422, + "step": 29295 + }, + { + "epoch": 3.84, + "learning_rate": 1.114115462529645e-05, + "loss": 0.0386, + "step": 29296 + }, + { + "epoch": 3.84, + "learning_rate": 1.114062776297859e-05, + "loss": 0.0414, + "step": 29297 + }, + { + "epoch": 3.84, + "learning_rate": 1.114010089745277e-05, + "loss": 0.037, + "step": 29298 + }, + { + "epoch": 3.84, + "learning_rate": 1.113957402872048e-05, + "loss": 0.0428, + "step": 29299 + }, + { + "epoch": 3.84, + "learning_rate": 1.1139047156783196e-05, + "loss": 0.0409, + "step": 29300 + }, + { + "epoch": 3.84, + "learning_rate": 1.1138520281642407e-05, + "loss": 0.0353, + "step": 29301 + }, + { + "epoch": 3.84, + "learning_rate": 1.1137993403299584e-05, + "loss": 0.0329, + "step": 29302 + }, + { + "epoch": 3.84, + "learning_rate": 1.1137466521756219e-05, + "loss": 0.0343, + "step": 29303 + }, + { + "epoch": 3.84, + "learning_rate": 1.1136939637013793e-05, + "loss": 0.0312, + "step": 29304 + }, + { + "epoch": 3.84, + "learning_rate": 1.113641274907378e-05, + "loss": 0.0376, + "step": 29305 + }, + { + "epoch": 3.84, + "learning_rate": 1.1135885857937667e-05, + "loss": 0.0349, + "step": 29306 + }, + { + "epoch": 3.84, + "learning_rate": 1.1135358963606938e-05, + "loss": 0.0358, + "step": 29307 + }, + { + "epoch": 3.84, + "learning_rate": 1.1134832066083072e-05, + "loss": 0.0322, + "step": 29308 + }, + { + "epoch": 3.84, + "learning_rate": 1.113430516536755e-05, + "loss": 0.032, + "step": 29309 + }, + { + "epoch": 3.84, + "learning_rate": 1.1133778261461854e-05, + "loss": 0.0395, + "step": 29310 + }, + { + "epoch": 3.84, + "learning_rate": 1.1133251354367467e-05, + "loss": 0.0337, + "step": 29311 + }, + { + "epoch": 3.84, + "learning_rate": 1.1132724444085874e-05, + "loss": 0.0314, + "step": 29312 + }, + { + "epoch": 3.84, + "learning_rate": 1.1132197530618552e-05, + "loss": 0.031, + "step": 29313 + }, + { + "epoch": 3.84, + "learning_rate": 1.1131670613966987e-05, + "loss": 0.0402, + "step": 29314 + }, + { + "epoch": 3.84, + "learning_rate": 1.1131143694132655e-05, + "loss": 0.03, + "step": 29315 + }, + { + "epoch": 3.84, + "learning_rate": 1.1130616771117042e-05, + "loss": 0.0327, + "step": 29316 + }, + { + "epoch": 3.84, + "learning_rate": 1.113008984492163e-05, + "loss": 0.0388, + "step": 29317 + }, + { + "epoch": 3.84, + "learning_rate": 1.1129562915547903e-05, + "loss": 0.0357, + "step": 29318 + }, + { + "epoch": 3.84, + "learning_rate": 1.1129035982997338e-05, + "loss": 0.0469, + "step": 29319 + }, + { + "epoch": 3.84, + "learning_rate": 1.1128509047271423e-05, + "loss": 0.029, + "step": 29320 + }, + { + "epoch": 3.84, + "learning_rate": 1.112798210837163e-05, + "loss": 0.0351, + "step": 29321 + }, + { + "epoch": 3.84, + "learning_rate": 1.1127455166299452e-05, + "loss": 0.0434, + "step": 29322 + }, + { + "epoch": 3.84, + "learning_rate": 1.1126928221056365e-05, + "loss": 0.0328, + "step": 29323 + }, + { + "epoch": 3.84, + "learning_rate": 1.1126401272643851e-05, + "loss": 0.0408, + "step": 29324 + }, + { + "epoch": 3.84, + "learning_rate": 1.1125874321063397e-05, + "loss": 0.0287, + "step": 29325 + }, + { + "epoch": 3.84, + "learning_rate": 1.112534736631648e-05, + "loss": 0.0376, + "step": 29326 + }, + { + "epoch": 3.84, + "learning_rate": 1.1124820408404581e-05, + "loss": 0.0375, + "step": 29327 + }, + { + "epoch": 3.84, + "learning_rate": 1.1124293447329186e-05, + "loss": 0.0357, + "step": 29328 + }, + { + "epoch": 3.84, + "learning_rate": 1.1123766483091777e-05, + "loss": 0.0388, + "step": 29329 + }, + { + "epoch": 3.84, + "learning_rate": 1.1123239515693833e-05, + "loss": 0.0372, + "step": 29330 + }, + { + "epoch": 3.84, + "learning_rate": 1.112271254513684e-05, + "loss": 0.0281, + "step": 29331 + }, + { + "epoch": 3.84, + "learning_rate": 1.1122185571422272e-05, + "loss": 0.0335, + "step": 29332 + }, + { + "epoch": 3.84, + "learning_rate": 1.1121658594551624e-05, + "loss": 0.0257, + "step": 29333 + }, + { + "epoch": 3.84, + "learning_rate": 1.1121131614526368e-05, + "loss": 0.031, + "step": 29334 + }, + { + "epoch": 3.84, + "learning_rate": 1.1120604631347989e-05, + "loss": 0.0422, + "step": 29335 + }, + { + "epoch": 3.84, + "learning_rate": 1.1120077645017969e-05, + "loss": 0.0312, + "step": 29336 + }, + { + "epoch": 3.84, + "learning_rate": 1.1119550655537793e-05, + "loss": 0.045, + "step": 29337 + }, + { + "epoch": 3.84, + "learning_rate": 1.1119023662908934e-05, + "loss": 0.0326, + "step": 29338 + }, + { + "epoch": 3.84, + "learning_rate": 1.1118496667132886e-05, + "loss": 0.035, + "step": 29339 + }, + { + "epoch": 3.84, + "learning_rate": 1.1117969668211122e-05, + "loss": 0.041, + "step": 29340 + }, + { + "epoch": 3.84, + "learning_rate": 1.111744266614513e-05, + "loss": 0.0408, + "step": 29341 + }, + { + "epoch": 3.84, + "learning_rate": 1.1116915660936391e-05, + "loss": 0.0389, + "step": 29342 + }, + { + "epoch": 3.84, + "learning_rate": 1.1116388652586385e-05, + "loss": 0.0387, + "step": 29343 + }, + { + "epoch": 3.84, + "learning_rate": 1.1115861641096595e-05, + "loss": 0.0319, + "step": 29344 + }, + { + "epoch": 3.84, + "learning_rate": 1.1115334626468503e-05, + "loss": 0.0295, + "step": 29345 + }, + { + "epoch": 3.84, + "learning_rate": 1.1114807608703593e-05, + "loss": 0.0321, + "step": 29346 + }, + { + "epoch": 3.84, + "learning_rate": 1.1114280587803345e-05, + "loss": 0.032, + "step": 29347 + }, + { + "epoch": 3.84, + "learning_rate": 1.1113753563769244e-05, + "loss": 0.035, + "step": 29348 + }, + { + "epoch": 3.84, + "learning_rate": 1.1113226536602765e-05, + "loss": 0.0363, + "step": 29349 + }, + { + "epoch": 3.84, + "learning_rate": 1.1112699506305403e-05, + "loss": 0.0379, + "step": 29350 + }, + { + "epoch": 3.84, + "learning_rate": 1.1112172472878628e-05, + "loss": 0.0368, + "step": 29351 + }, + { + "epoch": 3.84, + "learning_rate": 1.1111645436323932e-05, + "loss": 0.0321, + "step": 29352 + }, + { + "epoch": 3.85, + "learning_rate": 1.1111118396642788e-05, + "loss": 0.0389, + "step": 29353 + }, + { + "epoch": 3.85, + "learning_rate": 1.1110591353836684e-05, + "loss": 0.0373, + "step": 29354 + }, + { + "epoch": 3.85, + "learning_rate": 1.11100643079071e-05, + "loss": 0.0332, + "step": 29355 + }, + { + "epoch": 3.85, + "learning_rate": 1.110953725885552e-05, + "loss": 0.0358, + "step": 29356 + }, + { + "epoch": 3.85, + "learning_rate": 1.1109010206683426e-05, + "loss": 0.0353, + "step": 29357 + }, + { + "epoch": 3.85, + "learning_rate": 1.1108483151392302e-05, + "loss": 0.0367, + "step": 29358 + }, + { + "epoch": 3.85, + "learning_rate": 1.1107956092983625e-05, + "loss": 0.031, + "step": 29359 + }, + { + "epoch": 3.85, + "learning_rate": 1.110742903145888e-05, + "loss": 0.0306, + "step": 29360 + }, + { + "epoch": 3.85, + "learning_rate": 1.110690196681955e-05, + "loss": 0.0342, + "step": 29361 + }, + { + "epoch": 3.85, + "learning_rate": 1.1106374899067121e-05, + "loss": 0.0405, + "step": 29362 + }, + { + "epoch": 3.85, + "learning_rate": 1.1105847828203068e-05, + "loss": 0.0344, + "step": 29363 + }, + { + "epoch": 3.85, + "learning_rate": 1.1105320754228877e-05, + "loss": 0.0327, + "step": 29364 + }, + { + "epoch": 3.85, + "learning_rate": 1.1104793677146029e-05, + "loss": 0.0355, + "step": 29365 + }, + { + "epoch": 3.85, + "learning_rate": 1.110426659695601e-05, + "loss": 0.0322, + "step": 29366 + }, + { + "epoch": 3.85, + "learning_rate": 1.1103739513660302e-05, + "loss": 0.0404, + "step": 29367 + }, + { + "epoch": 3.85, + "learning_rate": 1.110321242726038e-05, + "loss": 0.0378, + "step": 29368 + }, + { + "epoch": 3.85, + "learning_rate": 1.1102685337757737e-05, + "loss": 0.0354, + "step": 29369 + }, + { + "epoch": 3.85, + "learning_rate": 1.1102158245153846e-05, + "loss": 0.0417, + "step": 29370 + }, + { + "epoch": 3.85, + "learning_rate": 1.1101631149450195e-05, + "loss": 0.0346, + "step": 29371 + }, + { + "epoch": 3.85, + "learning_rate": 1.1101104050648266e-05, + "loss": 0.036, + "step": 29372 + }, + { + "epoch": 3.85, + "learning_rate": 1.1100576948749542e-05, + "loss": 0.0367, + "step": 29373 + }, + { + "epoch": 3.85, + "learning_rate": 1.11000498437555e-05, + "loss": 0.0348, + "step": 29374 + }, + { + "epoch": 3.85, + "learning_rate": 1.109952273566763e-05, + "loss": 0.036, + "step": 29375 + }, + { + "epoch": 3.85, + "learning_rate": 1.1098995624487408e-05, + "loss": 0.0284, + "step": 29376 + }, + { + "epoch": 3.85, + "learning_rate": 1.109846851021632e-05, + "loss": 0.0315, + "step": 29377 + }, + { + "epoch": 3.85, + "learning_rate": 1.109794139285585e-05, + "loss": 0.0361, + "step": 29378 + }, + { + "epoch": 3.85, + "learning_rate": 1.1097414272407476e-05, + "loss": 0.0373, + "step": 29379 + }, + { + "epoch": 3.85, + "learning_rate": 1.1096887148872684e-05, + "loss": 0.0348, + "step": 29380 + }, + { + "epoch": 3.85, + "learning_rate": 1.1096360022252955e-05, + "loss": 0.0368, + "step": 29381 + }, + { + "epoch": 3.85, + "learning_rate": 1.109583289254977e-05, + "loss": 0.0304, + "step": 29382 + }, + { + "epoch": 3.85, + "learning_rate": 1.1095305759764616e-05, + "loss": 0.0435, + "step": 29383 + }, + { + "epoch": 3.85, + "learning_rate": 1.1094778623898972e-05, + "loss": 0.0306, + "step": 29384 + }, + { + "epoch": 3.85, + "learning_rate": 1.1094251484954322e-05, + "loss": 0.0335, + "step": 29385 + }, + { + "epoch": 3.85, + "learning_rate": 1.1093724342932147e-05, + "loss": 0.0307, + "step": 29386 + }, + { + "epoch": 3.85, + "learning_rate": 1.109319719783393e-05, + "loss": 0.0361, + "step": 29387 + }, + { + "epoch": 3.85, + "learning_rate": 1.1092670049661155e-05, + "loss": 0.0377, + "step": 29388 + }, + { + "epoch": 3.85, + "learning_rate": 1.10921428984153e-05, + "loss": 0.0387, + "step": 29389 + }, + { + "epoch": 3.85, + "learning_rate": 1.109161574409786e-05, + "loss": 0.0385, + "step": 29390 + }, + { + "epoch": 3.85, + "learning_rate": 1.1091088586710302e-05, + "loss": 0.0395, + "step": 29391 + }, + { + "epoch": 3.85, + "learning_rate": 1.1090561426254116e-05, + "loss": 0.045, + "step": 29392 + }, + { + "epoch": 3.85, + "learning_rate": 1.1090034262730784e-05, + "loss": 0.029, + "step": 29393 + }, + { + "epoch": 3.85, + "learning_rate": 1.1089507096141792e-05, + "loss": 0.0266, + "step": 29394 + }, + { + "epoch": 3.85, + "learning_rate": 1.1088979926488616e-05, + "loss": 0.0359, + "step": 29395 + }, + { + "epoch": 3.85, + "learning_rate": 1.1088452753772745e-05, + "loss": 0.0327, + "step": 29396 + }, + { + "epoch": 3.85, + "learning_rate": 1.1087925577995655e-05, + "loss": 0.037, + "step": 29397 + }, + { + "epoch": 3.85, + "learning_rate": 1.1087398399158833e-05, + "loss": 0.031, + "step": 29398 + }, + { + "epoch": 3.85, + "learning_rate": 1.108687121726376e-05, + "loss": 0.0363, + "step": 29399 + }, + { + "epoch": 3.85, + "learning_rate": 1.1086344032311925e-05, + "loss": 0.0434, + "step": 29400 + }, + { + "epoch": 3.85, + "learning_rate": 1.10858168443048e-05, + "loss": 0.0369, + "step": 29401 + }, + { + "epoch": 3.85, + "learning_rate": 1.1085289653243878e-05, + "loss": 0.0412, + "step": 29402 + }, + { + "epoch": 3.85, + "learning_rate": 1.108476245913063e-05, + "loss": 0.0378, + "step": 29403 + }, + { + "epoch": 3.85, + "learning_rate": 1.1084235261966549e-05, + "loss": 0.0317, + "step": 29404 + }, + { + "epoch": 3.85, + "learning_rate": 1.1083708061753114e-05, + "loss": 0.0385, + "step": 29405 + }, + { + "epoch": 3.85, + "learning_rate": 1.1083180858491808e-05, + "loss": 0.0343, + "step": 29406 + }, + { + "epoch": 3.85, + "learning_rate": 1.1082653652184112e-05, + "loss": 0.0373, + "step": 29407 + }, + { + "epoch": 3.85, + "learning_rate": 1.1082126442831511e-05, + "loss": 0.0342, + "step": 29408 + }, + { + "epoch": 3.85, + "learning_rate": 1.1081599230435487e-05, + "loss": 0.031, + "step": 29409 + }, + { + "epoch": 3.85, + "learning_rate": 1.1081072014997524e-05, + "loss": 0.0416, + "step": 29410 + }, + { + "epoch": 3.85, + "learning_rate": 1.1080544796519101e-05, + "loss": 0.0306, + "step": 29411 + }, + { + "epoch": 3.85, + "learning_rate": 1.1080017575001705e-05, + "loss": 0.0343, + "step": 29412 + }, + { + "epoch": 3.85, + "learning_rate": 1.107949035044682e-05, + "loss": 0.0425, + "step": 29413 + }, + { + "epoch": 3.85, + "learning_rate": 1.107896312285592e-05, + "loss": 0.032, + "step": 29414 + }, + { + "epoch": 3.85, + "learning_rate": 1.1078435892230496e-05, + "loss": 0.0407, + "step": 29415 + }, + { + "epoch": 3.85, + "learning_rate": 1.107790865857203e-05, + "loss": 0.0394, + "step": 29416 + }, + { + "epoch": 3.85, + "learning_rate": 1.1077381421882003e-05, + "loss": 0.0357, + "step": 29417 + }, + { + "epoch": 3.85, + "learning_rate": 1.1076854182161897e-05, + "loss": 0.0334, + "step": 29418 + }, + { + "epoch": 3.85, + "learning_rate": 1.1076326939413198e-05, + "loss": 0.0396, + "step": 29419 + }, + { + "epoch": 3.85, + "learning_rate": 1.1075799693637382e-05, + "loss": 0.0341, + "step": 29420 + }, + { + "epoch": 3.85, + "learning_rate": 1.1075272444835941e-05, + "loss": 0.0357, + "step": 29421 + }, + { + "epoch": 3.85, + "learning_rate": 1.1074745193010352e-05, + "loss": 0.0337, + "step": 29422 + }, + { + "epoch": 3.85, + "learning_rate": 1.1074217938162101e-05, + "loss": 0.0355, + "step": 29423 + }, + { + "epoch": 3.85, + "learning_rate": 1.1073690680292668e-05, + "loss": 0.0315, + "step": 29424 + }, + { + "epoch": 3.85, + "learning_rate": 1.1073163419403537e-05, + "loss": 0.0371, + "step": 29425 + }, + { + "epoch": 3.85, + "learning_rate": 1.107263615549619e-05, + "loss": 0.0437, + "step": 29426 + }, + { + "epoch": 3.85, + "learning_rate": 1.1072108888572113e-05, + "loss": 0.0404, + "step": 29427 + }, + { + "epoch": 3.85, + "learning_rate": 1.1071581618632785e-05, + "loss": 0.0299, + "step": 29428 + }, + { + "epoch": 3.85, + "learning_rate": 1.1071054345679691e-05, + "loss": 0.037, + "step": 29429 + }, + { + "epoch": 3.86, + "learning_rate": 1.1070527069714316e-05, + "loss": 0.0331, + "step": 29430 + }, + { + "epoch": 3.86, + "learning_rate": 1.106999979073814e-05, + "loss": 0.0398, + "step": 29431 + }, + { + "epoch": 3.86, + "learning_rate": 1.1069472508752643e-05, + "loss": 0.0313, + "step": 29432 + }, + { + "epoch": 3.86, + "learning_rate": 1.1068945223759313e-05, + "loss": 0.0329, + "step": 29433 + }, + { + "epoch": 3.86, + "learning_rate": 1.1068417935759634e-05, + "loss": 0.0392, + "step": 29434 + }, + { + "epoch": 3.86, + "learning_rate": 1.1067890644755087e-05, + "loss": 0.0357, + "step": 29435 + }, + { + "epoch": 3.86, + "learning_rate": 1.106736335074715e-05, + "loss": 0.028, + "step": 29436 + }, + { + "epoch": 3.86, + "learning_rate": 1.1066836053737312e-05, + "loss": 0.0347, + "step": 29437 + }, + { + "epoch": 3.86, + "learning_rate": 1.1066308753727057e-05, + "loss": 0.0329, + "step": 29438 + }, + { + "epoch": 3.86, + "learning_rate": 1.1065781450717862e-05, + "loss": 0.0385, + "step": 29439 + }, + { + "epoch": 3.86, + "learning_rate": 1.1065254144711217e-05, + "loss": 0.0385, + "step": 29440 + }, + { + "epoch": 3.86, + "learning_rate": 1.1064726835708597e-05, + "loss": 0.0377, + "step": 29441 + }, + { + "epoch": 3.86, + "learning_rate": 1.1064199523711495e-05, + "loss": 0.0334, + "step": 29442 + }, + { + "epoch": 3.86, + "learning_rate": 1.1063672208721384e-05, + "loss": 0.0336, + "step": 29443 + }, + { + "epoch": 3.86, + "learning_rate": 1.1063144890739754e-05, + "loss": 0.0362, + "step": 29444 + }, + { + "epoch": 3.86, + "learning_rate": 1.1062617569768084e-05, + "loss": 0.0288, + "step": 29445 + }, + { + "epoch": 3.86, + "learning_rate": 1.1062090245807862e-05, + "loss": 0.0392, + "step": 29446 + }, + { + "epoch": 3.86, + "learning_rate": 1.1061562918860562e-05, + "loss": 0.0296, + "step": 29447 + }, + { + "epoch": 3.86, + "learning_rate": 1.1061035588927679e-05, + "loss": 0.0341, + "step": 29448 + }, + { + "epoch": 3.86, + "learning_rate": 1.1060508256010684e-05, + "loss": 0.0312, + "step": 29449 + }, + { + "epoch": 3.86, + "learning_rate": 1.1059980920111068e-05, + "loss": 0.0302, + "step": 29450 + }, + { + "epoch": 3.86, + "learning_rate": 1.1059453581230315e-05, + "loss": 0.0332, + "step": 29451 + }, + { + "epoch": 3.86, + "learning_rate": 1.1058926239369903e-05, + "loss": 0.0345, + "step": 29452 + }, + { + "epoch": 3.86, + "learning_rate": 1.1058398894531318e-05, + "loss": 0.0286, + "step": 29453 + }, + { + "epoch": 3.86, + "learning_rate": 1.1057871546716039e-05, + "loss": 0.0329, + "step": 29454 + }, + { + "epoch": 3.86, + "learning_rate": 1.1057344195925558e-05, + "loss": 0.0344, + "step": 29455 + }, + { + "epoch": 3.86, + "learning_rate": 1.105681684216135e-05, + "loss": 0.0404, + "step": 29456 + }, + { + "epoch": 3.86, + "learning_rate": 1.1056289485424903e-05, + "loss": 0.0384, + "step": 29457 + }, + { + "epoch": 3.86, + "learning_rate": 1.1055762125717693e-05, + "loss": 0.0367, + "step": 29458 + }, + { + "epoch": 3.86, + "learning_rate": 1.1055234763041215e-05, + "loss": 0.0318, + "step": 29459 + }, + { + "epoch": 3.86, + "learning_rate": 1.105470739739694e-05, + "loss": 0.0373, + "step": 29460 + }, + { + "epoch": 3.86, + "learning_rate": 1.1054180028786359e-05, + "loss": 0.0339, + "step": 29461 + }, + { + "epoch": 3.86, + "learning_rate": 1.1053652657210954e-05, + "loss": 0.0288, + "step": 29462 + }, + { + "epoch": 3.86, + "learning_rate": 1.1053125282672204e-05, + "loss": 0.039, + "step": 29463 + }, + { + "epoch": 3.86, + "learning_rate": 1.1052597905171596e-05, + "loss": 0.0368, + "step": 29464 + }, + { + "epoch": 3.86, + "learning_rate": 1.1052070524710614e-05, + "loss": 0.0338, + "step": 29465 + }, + { + "epoch": 3.86, + "learning_rate": 1.1051543141290739e-05, + "loss": 0.0338, + "step": 29466 + }, + { + "epoch": 3.86, + "learning_rate": 1.1051015754913455e-05, + "loss": 0.0374, + "step": 29467 + }, + { + "epoch": 3.86, + "learning_rate": 1.1050488365580246e-05, + "loss": 0.0343, + "step": 29468 + }, + { + "epoch": 3.86, + "learning_rate": 1.1049960973292591e-05, + "loss": 0.0399, + "step": 29469 + }, + { + "epoch": 3.86, + "learning_rate": 1.1049433578051978e-05, + "loss": 0.0311, + "step": 29470 + }, + { + "epoch": 3.86, + "learning_rate": 1.1048906179859892e-05, + "loss": 0.0263, + "step": 29471 + }, + { + "epoch": 3.86, + "learning_rate": 1.104837877871781e-05, + "loss": 0.0316, + "step": 29472 + }, + { + "epoch": 3.86, + "learning_rate": 1.1047851374627218e-05, + "loss": 0.0367, + "step": 29473 + }, + { + "epoch": 3.86, + "learning_rate": 1.1047323967589603e-05, + "loss": 0.0378, + "step": 29474 + }, + { + "epoch": 3.86, + "learning_rate": 1.1046796557606443e-05, + "loss": 0.039, + "step": 29475 + }, + { + "epoch": 3.86, + "learning_rate": 1.1046269144679224e-05, + "loss": 0.0371, + "step": 29476 + }, + { + "epoch": 3.86, + "learning_rate": 1.1045741728809429e-05, + "loss": 0.0298, + "step": 29477 + }, + { + "epoch": 3.86, + "learning_rate": 1.1045214309998542e-05, + "loss": 0.0414, + "step": 29478 + }, + { + "epoch": 3.86, + "learning_rate": 1.1044686888248044e-05, + "loss": 0.0401, + "step": 29479 + }, + { + "epoch": 3.86, + "learning_rate": 1.1044159463559419e-05, + "loss": 0.0295, + "step": 29480 + }, + { + "epoch": 3.86, + "learning_rate": 1.1043632035934151e-05, + "loss": 0.044, + "step": 29481 + }, + { + "epoch": 3.86, + "learning_rate": 1.1043104605373726e-05, + "loss": 0.0338, + "step": 29482 + }, + { + "epoch": 3.86, + "learning_rate": 1.1042577171879623e-05, + "loss": 0.038, + "step": 29483 + }, + { + "epoch": 3.86, + "learning_rate": 1.104204973545333e-05, + "loss": 0.0417, + "step": 29484 + }, + { + "epoch": 3.86, + "learning_rate": 1.1041522296096323e-05, + "loss": 0.0475, + "step": 29485 + }, + { + "epoch": 3.86, + "learning_rate": 1.1040994853810092e-05, + "loss": 0.0393, + "step": 29486 + }, + { + "epoch": 3.86, + "learning_rate": 1.1040467408596119e-05, + "loss": 0.0386, + "step": 29487 + }, + { + "epoch": 3.86, + "learning_rate": 1.1039939960455888e-05, + "loss": 0.0373, + "step": 29488 + }, + { + "epoch": 3.86, + "learning_rate": 1.1039412509390879e-05, + "loss": 0.0285, + "step": 29489 + }, + { + "epoch": 3.86, + "learning_rate": 1.1038885055402578e-05, + "loss": 0.0303, + "step": 29490 + }, + { + "epoch": 3.86, + "learning_rate": 1.1038357598492469e-05, + "loss": 0.0315, + "step": 29491 + }, + { + "epoch": 3.86, + "learning_rate": 1.1037830138662033e-05, + "loss": 0.0345, + "step": 29492 + }, + { + "epoch": 3.86, + "learning_rate": 1.1037302675912758e-05, + "loss": 0.0265, + "step": 29493 + }, + { + "epoch": 3.86, + "learning_rate": 1.103677521024612e-05, + "loss": 0.0393, + "step": 29494 + }, + { + "epoch": 3.86, + "learning_rate": 1.1036247741663612e-05, + "loss": 0.0298, + "step": 29495 + }, + { + "epoch": 3.86, + "learning_rate": 1.103572027016671e-05, + "loss": 0.0332, + "step": 29496 + }, + { + "epoch": 3.86, + "learning_rate": 1.1035192795756897e-05, + "loss": 0.0298, + "step": 29497 + }, + { + "epoch": 3.86, + "learning_rate": 1.1034665318435662e-05, + "loss": 0.0396, + "step": 29498 + }, + { + "epoch": 3.86, + "learning_rate": 1.1034137838204488e-05, + "loss": 0.0381, + "step": 29499 + }, + { + "epoch": 3.86, + "learning_rate": 1.1033610355064854e-05, + "loss": 0.0374, + "step": 29500 + }, + { + "epoch": 3.86, + "learning_rate": 1.1033082869018245e-05, + "loss": 0.0328, + "step": 29501 + }, + { + "epoch": 3.86, + "learning_rate": 1.1032555380066149e-05, + "loss": 0.0424, + "step": 29502 + }, + { + "epoch": 3.86, + "learning_rate": 1.1032027888210044e-05, + "loss": 0.0291, + "step": 29503 + }, + { + "epoch": 3.86, + "learning_rate": 1.1031500393451413e-05, + "loss": 0.0353, + "step": 29504 + }, + { + "epoch": 3.86, + "learning_rate": 1.1030972895791747e-05, + "loss": 0.0339, + "step": 29505 + }, + { + "epoch": 3.87, + "learning_rate": 1.103044539523252e-05, + "loss": 0.0345, + "step": 29506 + }, + { + "epoch": 3.87, + "learning_rate": 1.1029917891775224e-05, + "loss": 0.0374, + "step": 29507 + }, + { + "epoch": 3.87, + "learning_rate": 1.1029390385421335e-05, + "loss": 0.0294, + "step": 29508 + }, + { + "epoch": 3.87, + "learning_rate": 1.1028862876172344e-05, + "loss": 0.0304, + "step": 29509 + }, + { + "epoch": 3.87, + "learning_rate": 1.1028335364029727e-05, + "loss": 0.0347, + "step": 29510 + }, + { + "epoch": 3.87, + "learning_rate": 1.1027807848994976e-05, + "loss": 0.0349, + "step": 29511 + }, + { + "epoch": 3.87, + "learning_rate": 1.1027280331069565e-05, + "loss": 0.0405, + "step": 29512 + }, + { + "epoch": 3.87, + "learning_rate": 1.1026752810254988e-05, + "loss": 0.0382, + "step": 29513 + }, + { + "epoch": 3.87, + "learning_rate": 1.1026225286552721e-05, + "loss": 0.0327, + "step": 29514 + }, + { + "epoch": 3.87, + "learning_rate": 1.1025697759964249e-05, + "loss": 0.0379, + "step": 29515 + }, + { + "epoch": 3.87, + "learning_rate": 1.102517023049106e-05, + "loss": 0.0481, + "step": 29516 + }, + { + "epoch": 3.87, + "learning_rate": 1.102464269813463e-05, + "loss": 0.0298, + "step": 29517 + }, + { + "epoch": 3.87, + "learning_rate": 1.102411516289645e-05, + "loss": 0.0418, + "step": 29518 + }, + { + "epoch": 3.87, + "learning_rate": 1.1023587624777999e-05, + "loss": 0.0356, + "step": 29519 + }, + { + "epoch": 3.87, + "learning_rate": 1.1023060083780766e-05, + "loss": 0.0377, + "step": 29520 + }, + { + "epoch": 3.87, + "learning_rate": 1.1022532539906225e-05, + "loss": 0.0258, + "step": 29521 + }, + { + "epoch": 3.87, + "learning_rate": 1.1022004993155873e-05, + "loss": 0.0314, + "step": 29522 + }, + { + "epoch": 3.87, + "learning_rate": 1.102147744353118e-05, + "loss": 0.0361, + "step": 29523 + }, + { + "epoch": 3.87, + "learning_rate": 1.102094989103364e-05, + "loss": 0.0316, + "step": 29524 + }, + { + "epoch": 3.87, + "learning_rate": 1.102042233566473e-05, + "loss": 0.0275, + "step": 29525 + }, + { + "epoch": 3.87, + "learning_rate": 1.1019894777425937e-05, + "loss": 0.0395, + "step": 29526 + }, + { + "epoch": 3.87, + "learning_rate": 1.1019367216318747e-05, + "loss": 0.0358, + "step": 29527 + }, + { + "epoch": 3.87, + "learning_rate": 1.1018839652344639e-05, + "loss": 0.0377, + "step": 29528 + }, + { + "epoch": 3.87, + "learning_rate": 1.1018312085505096e-05, + "loss": 0.0368, + "step": 29529 + }, + { + "epoch": 3.87, + "learning_rate": 1.1017784515801609e-05, + "loss": 0.0324, + "step": 29530 + }, + { + "epoch": 3.87, + "learning_rate": 1.1017256943235653e-05, + "loss": 0.0318, + "step": 29531 + }, + { + "epoch": 3.87, + "learning_rate": 1.1016729367808721e-05, + "loss": 0.035, + "step": 29532 + }, + { + "epoch": 3.87, + "learning_rate": 1.1016201789522287e-05, + "loss": 0.0345, + "step": 29533 + }, + { + "epoch": 3.87, + "learning_rate": 1.1015674208377843e-05, + "loss": 0.0331, + "step": 29534 + }, + { + "epoch": 3.87, + "learning_rate": 1.1015146624376867e-05, + "loss": 0.0378, + "step": 29535 + }, + { + "epoch": 3.87, + "learning_rate": 1.1014619037520846e-05, + "loss": 0.0335, + "step": 29536 + }, + { + "epoch": 3.87, + "learning_rate": 1.1014091447811263e-05, + "loss": 0.0397, + "step": 29537 + }, + { + "epoch": 3.87, + "learning_rate": 1.1013563855249603e-05, + "loss": 0.0403, + "step": 29538 + }, + { + "epoch": 3.87, + "learning_rate": 1.101303625983735e-05, + "loss": 0.0372, + "step": 29539 + }, + { + "epoch": 3.87, + "learning_rate": 1.101250866157598e-05, + "loss": 0.0344, + "step": 29540 + }, + { + "epoch": 3.87, + "learning_rate": 1.1011981060466989e-05, + "loss": 0.037, + "step": 29541 + }, + { + "epoch": 3.87, + "learning_rate": 1.1011453456511854e-05, + "loss": 0.0402, + "step": 29542 + }, + { + "epoch": 3.87, + "learning_rate": 1.101092584971206e-05, + "loss": 0.0376, + "step": 29543 + }, + { + "epoch": 3.87, + "learning_rate": 1.1010398240069088e-05, + "loss": 0.0385, + "step": 29544 + }, + { + "epoch": 3.87, + "learning_rate": 1.100987062758443e-05, + "loss": 0.0358, + "step": 29545 + }, + { + "epoch": 3.87, + "learning_rate": 1.100934301225956e-05, + "loss": 0.0349, + "step": 29546 + }, + { + "epoch": 3.87, + "learning_rate": 1.1008815394095971e-05, + "loss": 0.0369, + "step": 29547 + }, + { + "epoch": 3.87, + "learning_rate": 1.1008287773095138e-05, + "loss": 0.0361, + "step": 29548 + }, + { + "epoch": 3.87, + "learning_rate": 1.1007760149258553e-05, + "loss": 0.0326, + "step": 29549 + }, + { + "epoch": 3.87, + "learning_rate": 1.1007232522587695e-05, + "loss": 0.0346, + "step": 29550 + }, + { + "epoch": 3.87, + "learning_rate": 1.1006704893084047e-05, + "loss": 0.0307, + "step": 29551 + }, + { + "epoch": 3.87, + "learning_rate": 1.1006177260749096e-05, + "loss": 0.0394, + "step": 29552 + }, + { + "epoch": 3.87, + "learning_rate": 1.1005649625584327e-05, + "loss": 0.0293, + "step": 29553 + }, + { + "epoch": 3.87, + "learning_rate": 1.100512198759122e-05, + "loss": 0.0377, + "step": 29554 + }, + { + "epoch": 3.87, + "learning_rate": 1.1004594346771262e-05, + "loss": 0.0369, + "step": 29555 + }, + { + "epoch": 3.87, + "learning_rate": 1.1004066703125935e-05, + "loss": 0.0292, + "step": 29556 + }, + { + "epoch": 3.87, + "learning_rate": 1.1003539056656725e-05, + "loss": 0.0383, + "step": 29557 + }, + { + "epoch": 3.87, + "learning_rate": 1.1003011407365113e-05, + "loss": 0.0357, + "step": 29558 + }, + { + "epoch": 3.87, + "learning_rate": 1.1002483755252586e-05, + "loss": 0.0384, + "step": 29559 + }, + { + "epoch": 3.87, + "learning_rate": 1.100195610032063e-05, + "loss": 0.0356, + "step": 29560 + }, + { + "epoch": 3.87, + "learning_rate": 1.1001428442570721e-05, + "loss": 0.0331, + "step": 29561 + }, + { + "epoch": 3.87, + "learning_rate": 1.100090078200435e-05, + "loss": 0.038, + "step": 29562 + }, + { + "epoch": 3.87, + "learning_rate": 1.1000373118622999e-05, + "loss": 0.033, + "step": 29563 + }, + { + "epoch": 3.87, + "learning_rate": 1.099984545242815e-05, + "loss": 0.031, + "step": 29564 + }, + { + "epoch": 3.87, + "learning_rate": 1.0999317783421291e-05, + "loss": 0.041, + "step": 29565 + }, + { + "epoch": 3.87, + "learning_rate": 1.0998790111603906e-05, + "loss": 0.0335, + "step": 29566 + }, + { + "epoch": 3.87, + "learning_rate": 1.0998262436977469e-05, + "loss": 0.0341, + "step": 29567 + }, + { + "epoch": 3.87, + "learning_rate": 1.099773475954348e-05, + "loss": 0.0277, + "step": 29568 + }, + { + "epoch": 3.87, + "learning_rate": 1.0997207079303412e-05, + "loss": 0.0369, + "step": 29569 + }, + { + "epoch": 3.87, + "learning_rate": 1.0996679396258754e-05, + "loss": 0.0389, + "step": 29570 + }, + { + "epoch": 3.87, + "learning_rate": 1.0996151710410986e-05, + "loss": 0.0393, + "step": 29571 + }, + { + "epoch": 3.87, + "learning_rate": 1.0995624021761594e-05, + "loss": 0.0327, + "step": 29572 + }, + { + "epoch": 3.87, + "learning_rate": 1.0995096330312063e-05, + "loss": 0.0371, + "step": 29573 + }, + { + "epoch": 3.87, + "learning_rate": 1.0994568636063878e-05, + "loss": 0.0382, + "step": 29574 + }, + { + "epoch": 3.87, + "learning_rate": 1.099404093901852e-05, + "loss": 0.0387, + "step": 29575 + }, + { + "epoch": 3.87, + "learning_rate": 1.0993513239177476e-05, + "loss": 0.0359, + "step": 29576 + }, + { + "epoch": 3.87, + "learning_rate": 1.0992985536542231e-05, + "loss": 0.0376, + "step": 29577 + }, + { + "epoch": 3.87, + "learning_rate": 1.0992457831114263e-05, + "loss": 0.0389, + "step": 29578 + }, + { + "epoch": 3.87, + "learning_rate": 1.0991930122895063e-05, + "loss": 0.0329, + "step": 29579 + }, + { + "epoch": 3.87, + "learning_rate": 1.099140241188611e-05, + "loss": 0.0308, + "step": 29580 + }, + { + "epoch": 3.87, + "learning_rate": 1.099087469808889e-05, + "loss": 0.0434, + "step": 29581 + }, + { + "epoch": 3.88, + "learning_rate": 1.099034698150489e-05, + "loss": 0.0315, + "step": 29582 + }, + { + "epoch": 3.88, + "learning_rate": 1.0989819262135592e-05, + "loss": 0.0339, + "step": 29583 + }, + { + "epoch": 3.88, + "learning_rate": 1.0989291539982478e-05, + "loss": 0.0346, + "step": 29584 + }, + { + "epoch": 3.88, + "learning_rate": 1.0988763815047036e-05, + "loss": 0.0374, + "step": 29585 + }, + { + "epoch": 3.88, + "learning_rate": 1.0988236087330748e-05, + "loss": 0.0443, + "step": 29586 + }, + { + "epoch": 3.88, + "learning_rate": 1.0987708356835101e-05, + "loss": 0.0363, + "step": 29587 + }, + { + "epoch": 3.88, + "learning_rate": 1.0987180623561574e-05, + "loss": 0.0397, + "step": 29588 + }, + { + "epoch": 3.88, + "learning_rate": 1.0986652887511654e-05, + "loss": 0.0344, + "step": 29589 + }, + { + "epoch": 3.88, + "learning_rate": 1.0986125148686827e-05, + "loss": 0.0356, + "step": 29590 + }, + { + "epoch": 3.88, + "learning_rate": 1.0985597407088573e-05, + "loss": 0.037, + "step": 29591 + }, + { + "epoch": 3.88, + "learning_rate": 1.098506966271838e-05, + "loss": 0.0353, + "step": 29592 + }, + { + "epoch": 3.88, + "learning_rate": 1.0984541915577733e-05, + "loss": 0.0337, + "step": 29593 + }, + { + "epoch": 3.88, + "learning_rate": 1.0984014165668108e-05, + "loss": 0.0351, + "step": 29594 + }, + { + "epoch": 3.88, + "learning_rate": 1.0983486412991003e-05, + "loss": 0.0316, + "step": 29595 + }, + { + "epoch": 3.88, + "learning_rate": 1.098295865754789e-05, + "loss": 0.0372, + "step": 29596 + }, + { + "epoch": 3.88, + "learning_rate": 1.0982430899340265e-05, + "loss": 0.0361, + "step": 29597 + }, + { + "epoch": 3.88, + "learning_rate": 1.0981903138369599e-05, + "loss": 0.0403, + "step": 29598 + }, + { + "epoch": 3.88, + "learning_rate": 1.0981375374637382e-05, + "loss": 0.032, + "step": 29599 + }, + { + "epoch": 3.88, + "learning_rate": 1.09808476081451e-05, + "loss": 0.035, + "step": 29600 + }, + { + "epoch": 3.88, + "learning_rate": 1.0980319838894237e-05, + "loss": 0.0315, + "step": 29601 + }, + { + "epoch": 3.88, + "learning_rate": 1.097979206688628e-05, + "loss": 0.0322, + "step": 29602 + }, + { + "epoch": 3.88, + "learning_rate": 1.0979264292122705e-05, + "loss": 0.0389, + "step": 29603 + }, + { + "epoch": 3.88, + "learning_rate": 1.0978736514605007e-05, + "loss": 0.0365, + "step": 29604 + }, + { + "epoch": 3.88, + "learning_rate": 1.0978208734334656e-05, + "loss": 0.0406, + "step": 29605 + }, + { + "epoch": 3.88, + "learning_rate": 1.0977680951313152e-05, + "loss": 0.0363, + "step": 29606 + }, + { + "epoch": 3.88, + "learning_rate": 1.097715316554197e-05, + "loss": 0.0401, + "step": 29607 + }, + { + "epoch": 3.88, + "learning_rate": 1.0976625377022599e-05, + "loss": 0.0401, + "step": 29608 + }, + { + "epoch": 3.88, + "learning_rate": 1.097609758575652e-05, + "loss": 0.0315, + "step": 29609 + }, + { + "epoch": 3.88, + "learning_rate": 1.0975569791745217e-05, + "loss": 0.0407, + "step": 29610 + }, + { + "epoch": 3.88, + "learning_rate": 1.0975041994990177e-05, + "loss": 0.039, + "step": 29611 + }, + { + "epoch": 3.88, + "learning_rate": 1.0974514195492883e-05, + "loss": 0.0282, + "step": 29612 + }, + { + "epoch": 3.88, + "learning_rate": 1.097398639325482e-05, + "loss": 0.0434, + "step": 29613 + }, + { + "epoch": 3.88, + "learning_rate": 1.0973458588277473e-05, + "loss": 0.0336, + "step": 29614 + }, + { + "epoch": 3.88, + "learning_rate": 1.0972930780562323e-05, + "loss": 0.0359, + "step": 29615 + }, + { + "epoch": 3.88, + "learning_rate": 1.0972402970110859e-05, + "loss": 0.0327, + "step": 29616 + }, + { + "epoch": 3.88, + "learning_rate": 1.0971875156924564e-05, + "loss": 0.0328, + "step": 29617 + }, + { + "epoch": 3.88, + "learning_rate": 1.097134734100492e-05, + "loss": 0.0365, + "step": 29618 + }, + { + "epoch": 3.88, + "learning_rate": 1.0970819522353413e-05, + "loss": 0.033, + "step": 29619 + }, + { + "epoch": 3.88, + "learning_rate": 1.0970291700971529e-05, + "loss": 0.0336, + "step": 29620 + }, + { + "epoch": 3.88, + "learning_rate": 1.0969763876860753e-05, + "loss": 0.0389, + "step": 29621 + }, + { + "epoch": 3.88, + "learning_rate": 1.0969236050022564e-05, + "loss": 0.0383, + "step": 29622 + }, + { + "epoch": 3.88, + "learning_rate": 1.0968708220458453e-05, + "loss": 0.0309, + "step": 29623 + }, + { + "epoch": 3.88, + "learning_rate": 1.09681803881699e-05, + "loss": 0.0405, + "step": 29624 + }, + { + "epoch": 3.88, + "learning_rate": 1.0967652553158394e-05, + "loss": 0.0381, + "step": 29625 + }, + { + "epoch": 3.88, + "learning_rate": 1.0967124715425414e-05, + "loss": 0.044, + "step": 29626 + }, + { + "epoch": 3.88, + "learning_rate": 1.0966596874972448e-05, + "loss": 0.0302, + "step": 29627 + }, + { + "epoch": 3.88, + "learning_rate": 1.0966069031800978e-05, + "loss": 0.0374, + "step": 29628 + }, + { + "epoch": 3.88, + "learning_rate": 1.0965541185912494e-05, + "loss": 0.033, + "step": 29629 + }, + { + "epoch": 3.88, + "learning_rate": 1.0965013337308474e-05, + "loss": 0.0401, + "step": 29630 + }, + { + "epoch": 3.88, + "learning_rate": 1.0964485485990409e-05, + "loss": 0.0329, + "step": 29631 + }, + { + "epoch": 3.88, + "learning_rate": 1.0963957631959774e-05, + "loss": 0.0328, + "step": 29632 + }, + { + "epoch": 3.88, + "learning_rate": 1.0963429775218064e-05, + "loss": 0.0387, + "step": 29633 + }, + { + "epoch": 3.88, + "learning_rate": 1.0962901915766758e-05, + "loss": 0.0316, + "step": 29634 + }, + { + "epoch": 3.88, + "learning_rate": 1.0962374053607344e-05, + "loss": 0.036, + "step": 29635 + }, + { + "epoch": 3.88, + "learning_rate": 1.09618461887413e-05, + "loss": 0.0369, + "step": 29636 + }, + { + "epoch": 3.88, + "learning_rate": 1.0961318321170118e-05, + "loss": 0.0306, + "step": 29637 + }, + { + "epoch": 3.88, + "learning_rate": 1.0960790450895275e-05, + "loss": 0.0375, + "step": 29638 + }, + { + "epoch": 3.88, + "learning_rate": 1.0960262577918266e-05, + "loss": 0.0311, + "step": 29639 + }, + { + "epoch": 3.88, + "learning_rate": 1.0959734702240565e-05, + "loss": 0.0313, + "step": 29640 + }, + { + "epoch": 3.88, + "learning_rate": 1.0959206823863662e-05, + "loss": 0.0298, + "step": 29641 + }, + { + "epoch": 3.88, + "learning_rate": 1.0958678942789043e-05, + "loss": 0.038, + "step": 29642 + }, + { + "epoch": 3.88, + "learning_rate": 1.095815105901819e-05, + "loss": 0.028, + "step": 29643 + }, + { + "epoch": 3.88, + "learning_rate": 1.0957623172552588e-05, + "loss": 0.0278, + "step": 29644 + }, + { + "epoch": 3.88, + "learning_rate": 1.0957095283393719e-05, + "loss": 0.0356, + "step": 29645 + }, + { + "epoch": 3.88, + "learning_rate": 1.0956567391543073e-05, + "loss": 0.0336, + "step": 29646 + }, + { + "epoch": 3.88, + "learning_rate": 1.0956039497002133e-05, + "loss": 0.0378, + "step": 29647 + }, + { + "epoch": 3.88, + "learning_rate": 1.0955511599772382e-05, + "loss": 0.0389, + "step": 29648 + }, + { + "epoch": 3.88, + "learning_rate": 1.0954983699855304e-05, + "loss": 0.0409, + "step": 29649 + }, + { + "epoch": 3.88, + "learning_rate": 1.0954455797252387e-05, + "loss": 0.0294, + "step": 29650 + }, + { + "epoch": 3.88, + "learning_rate": 1.0953927891965112e-05, + "loss": 0.0313, + "step": 29651 + }, + { + "epoch": 3.88, + "learning_rate": 1.095339998399497e-05, + "loss": 0.0342, + "step": 29652 + }, + { + "epoch": 3.88, + "learning_rate": 1.0952872073343438e-05, + "loss": 0.0374, + "step": 29653 + }, + { + "epoch": 3.88, + "learning_rate": 1.0952344160012004e-05, + "loss": 0.0347, + "step": 29654 + }, + { + "epoch": 3.88, + "learning_rate": 1.0951816244002151e-05, + "loss": 0.0342, + "step": 29655 + }, + { + "epoch": 3.88, + "learning_rate": 1.095128832531537e-05, + "loss": 0.0333, + "step": 29656 + }, + { + "epoch": 3.88, + "learning_rate": 1.0950760403953137e-05, + "loss": 0.0422, + "step": 29657 + }, + { + "epoch": 3.88, + "learning_rate": 1.0950232479916945e-05, + "loss": 0.0271, + "step": 29658 + }, + { + "epoch": 3.89, + "learning_rate": 1.0949704553208272e-05, + "loss": 0.0297, + "step": 29659 + }, + { + "epoch": 3.89, + "learning_rate": 1.0949176623828606e-05, + "loss": 0.0324, + "step": 29660 + }, + { + "epoch": 3.89, + "learning_rate": 1.0948648691779428e-05, + "loss": 0.0332, + "step": 29661 + }, + { + "epoch": 3.89, + "learning_rate": 1.0948120757062233e-05, + "loss": 0.0399, + "step": 29662 + }, + { + "epoch": 3.89, + "learning_rate": 1.0947592819678494e-05, + "loss": 0.0335, + "step": 29663 + }, + { + "epoch": 3.89, + "learning_rate": 1.0947064879629703e-05, + "loss": 0.0307, + "step": 29664 + }, + { + "epoch": 3.89, + "learning_rate": 1.094653693691734e-05, + "loss": 0.037, + "step": 29665 + }, + { + "epoch": 3.89, + "learning_rate": 1.0946008991542894e-05, + "loss": 0.0354, + "step": 29666 + }, + { + "epoch": 3.89, + "learning_rate": 1.094548104350785e-05, + "loss": 0.0303, + "step": 29667 + }, + { + "epoch": 3.89, + "learning_rate": 1.0944953092813685e-05, + "loss": 0.0377, + "step": 29668 + }, + { + "epoch": 3.89, + "learning_rate": 1.0944425139461896e-05, + "loss": 0.0348, + "step": 29669 + }, + { + "epoch": 3.89, + "learning_rate": 1.0943897183453957e-05, + "loss": 0.0355, + "step": 29670 + }, + { + "epoch": 3.89, + "learning_rate": 1.0943369224791361e-05, + "loss": 0.0316, + "step": 29671 + }, + { + "epoch": 3.89, + "learning_rate": 1.0942841263475587e-05, + "loss": 0.0298, + "step": 29672 + }, + { + "epoch": 3.89, + "learning_rate": 1.0942313299508125e-05, + "loss": 0.0305, + "step": 29673 + }, + { + "epoch": 3.89, + "learning_rate": 1.0941785332890453e-05, + "loss": 0.0341, + "step": 29674 + }, + { + "epoch": 3.89, + "learning_rate": 1.0941257363624065e-05, + "loss": 0.0343, + "step": 29675 + }, + { + "epoch": 3.89, + "learning_rate": 1.0940729391710433e-05, + "loss": 0.0295, + "step": 29676 + }, + { + "epoch": 3.89, + "learning_rate": 1.0940201417151057e-05, + "loss": 0.0383, + "step": 29677 + }, + { + "epoch": 3.89, + "learning_rate": 1.0939673439947411e-05, + "loss": 0.0316, + "step": 29678 + }, + { + "epoch": 3.89, + "learning_rate": 1.0939145460100987e-05, + "loss": 0.0389, + "step": 29679 + }, + { + "epoch": 3.89, + "learning_rate": 1.0938617477613263e-05, + "loss": 0.0338, + "step": 29680 + }, + { + "epoch": 3.89, + "learning_rate": 1.0938089492485726e-05, + "loss": 0.0341, + "step": 29681 + }, + { + "epoch": 3.89, + "learning_rate": 1.0937561504719864e-05, + "loss": 0.0314, + "step": 29682 + }, + { + "epoch": 3.89, + "learning_rate": 1.0937033514317163e-05, + "loss": 0.0306, + "step": 29683 + }, + { + "epoch": 3.89, + "learning_rate": 1.0936505521279102e-05, + "loss": 0.0379, + "step": 29684 + }, + { + "epoch": 3.89, + "learning_rate": 1.0935977525607168e-05, + "loss": 0.0384, + "step": 29685 + }, + { + "epoch": 3.89, + "learning_rate": 1.093544952730285e-05, + "loss": 0.0253, + "step": 29686 + }, + { + "epoch": 3.89, + "learning_rate": 1.093492152636763e-05, + "loss": 0.0373, + "step": 29687 + }, + { + "epoch": 3.89, + "learning_rate": 1.0934393522802991e-05, + "loss": 0.0352, + "step": 29688 + }, + { + "epoch": 3.89, + "learning_rate": 1.093386551661042e-05, + "loss": 0.0389, + "step": 29689 + }, + { + "epoch": 3.89, + "learning_rate": 1.0933337507791404e-05, + "loss": 0.0369, + "step": 29690 + }, + { + "epoch": 3.89, + "learning_rate": 1.0932809496347425e-05, + "loss": 0.0409, + "step": 29691 + }, + { + "epoch": 3.89, + "learning_rate": 1.0932281482279969e-05, + "loss": 0.0327, + "step": 29692 + }, + { + "epoch": 3.89, + "learning_rate": 1.0931753465590519e-05, + "loss": 0.042, + "step": 29693 + }, + { + "epoch": 3.89, + "learning_rate": 1.0931225446280565e-05, + "loss": 0.0389, + "step": 29694 + }, + { + "epoch": 3.89, + "learning_rate": 1.0930697424351586e-05, + "loss": 0.0387, + "step": 29695 + }, + { + "epoch": 3.89, + "learning_rate": 1.0930169399805075e-05, + "loss": 0.0361, + "step": 29696 + }, + { + "epoch": 3.89, + "learning_rate": 1.0929641372642508e-05, + "loss": 0.034, + "step": 29697 + }, + { + "epoch": 3.89, + "learning_rate": 1.0929113342865373e-05, + "loss": 0.0366, + "step": 29698 + }, + { + "epoch": 3.89, + "learning_rate": 1.092858531047516e-05, + "loss": 0.0404, + "step": 29699 + }, + { + "epoch": 3.89, + "learning_rate": 1.092805727547335e-05, + "loss": 0.0306, + "step": 29700 + }, + { + "epoch": 3.89, + "learning_rate": 1.0927529237861425e-05, + "loss": 0.0372, + "step": 29701 + }, + { + "epoch": 3.89, + "learning_rate": 1.0927001197640876e-05, + "loss": 0.0346, + "step": 29702 + }, + { + "epoch": 3.89, + "learning_rate": 1.0926473154813185e-05, + "loss": 0.0291, + "step": 29703 + }, + { + "epoch": 3.89, + "learning_rate": 1.0925945109379836e-05, + "loss": 0.0326, + "step": 29704 + }, + { + "epoch": 3.89, + "learning_rate": 1.0925417061342318e-05, + "loss": 0.033, + "step": 29705 + }, + { + "epoch": 3.89, + "learning_rate": 1.0924889010702111e-05, + "loss": 0.0364, + "step": 29706 + }, + { + "epoch": 3.89, + "learning_rate": 1.0924360957460706e-05, + "loss": 0.0338, + "step": 29707 + }, + { + "epoch": 3.89, + "learning_rate": 1.0923832901619584e-05, + "loss": 0.038, + "step": 29708 + }, + { + "epoch": 3.89, + "learning_rate": 1.0923304843180232e-05, + "loss": 0.0361, + "step": 29709 + }, + { + "epoch": 3.89, + "learning_rate": 1.092277678214413e-05, + "loss": 0.0344, + "step": 29710 + }, + { + "epoch": 3.89, + "learning_rate": 1.0922248718512774e-05, + "loss": 0.0343, + "step": 29711 + }, + { + "epoch": 3.89, + "learning_rate": 1.0921720652287637e-05, + "loss": 0.0318, + "step": 29712 + }, + { + "epoch": 3.89, + "learning_rate": 1.0921192583470216e-05, + "loss": 0.0326, + "step": 29713 + }, + { + "epoch": 3.89, + "learning_rate": 1.0920664512061983e-05, + "loss": 0.0341, + "step": 29714 + }, + { + "epoch": 3.89, + "learning_rate": 1.0920136438064433e-05, + "loss": 0.0422, + "step": 29715 + }, + { + "epoch": 3.89, + "learning_rate": 1.0919608361479049e-05, + "loss": 0.0301, + "step": 29716 + }, + { + "epoch": 3.89, + "learning_rate": 1.0919080282307315e-05, + "loss": 0.0307, + "step": 29717 + }, + { + "epoch": 3.89, + "learning_rate": 1.0918552200550716e-05, + "loss": 0.0301, + "step": 29718 + }, + { + "epoch": 3.89, + "learning_rate": 1.0918024116210739e-05, + "loss": 0.0257, + "step": 29719 + }, + { + "epoch": 3.89, + "learning_rate": 1.0917496029288865e-05, + "loss": 0.0319, + "step": 29720 + }, + { + "epoch": 3.89, + "learning_rate": 1.0916967939786588e-05, + "loss": 0.0344, + "step": 29721 + }, + { + "epoch": 3.89, + "learning_rate": 1.0916439847705382e-05, + "loss": 0.0399, + "step": 29722 + }, + { + "epoch": 3.89, + "learning_rate": 1.0915911753046742e-05, + "loss": 0.032, + "step": 29723 + }, + { + "epoch": 3.89, + "learning_rate": 1.0915383655812146e-05, + "loss": 0.0333, + "step": 29724 + }, + { + "epoch": 3.89, + "learning_rate": 1.0914855556003082e-05, + "loss": 0.0381, + "step": 29725 + }, + { + "epoch": 3.89, + "learning_rate": 1.0914327453621036e-05, + "loss": 0.0383, + "step": 29726 + }, + { + "epoch": 3.89, + "learning_rate": 1.0913799348667495e-05, + "loss": 0.0337, + "step": 29727 + }, + { + "epoch": 3.89, + "learning_rate": 1.091327124114394e-05, + "loss": 0.036, + "step": 29728 + }, + { + "epoch": 3.89, + "learning_rate": 1.0912743131051857e-05, + "loss": 0.0357, + "step": 29729 + }, + { + "epoch": 3.89, + "learning_rate": 1.0912215018392736e-05, + "loss": 0.035, + "step": 29730 + }, + { + "epoch": 3.89, + "learning_rate": 1.0911686903168053e-05, + "loss": 0.0368, + "step": 29731 + }, + { + "epoch": 3.89, + "learning_rate": 1.0911158785379306e-05, + "loss": 0.0319, + "step": 29732 + }, + { + "epoch": 3.89, + "learning_rate": 1.0910630665027968e-05, + "loss": 0.0415, + "step": 29733 + }, + { + "epoch": 3.89, + "learning_rate": 1.0910102542115535e-05, + "loss": 0.0381, + "step": 29734 + }, + { + "epoch": 3.9, + "learning_rate": 1.0909574416643483e-05, + "loss": 0.0309, + "step": 29735 + }, + { + "epoch": 3.9, + "learning_rate": 1.0909046288613303e-05, + "loss": 0.0352, + "step": 29736 + }, + { + "epoch": 3.9, + "learning_rate": 1.0908518158026478e-05, + "loss": 0.0343, + "step": 29737 + }, + { + "epoch": 3.9, + "learning_rate": 1.0907990024884495e-05, + "loss": 0.0342, + "step": 29738 + }, + { + "epoch": 3.9, + "learning_rate": 1.0907461889188837e-05, + "loss": 0.0378, + "step": 29739 + }, + { + "epoch": 3.9, + "learning_rate": 1.0906933750940994e-05, + "loss": 0.0289, + "step": 29740 + }, + { + "epoch": 3.9, + "learning_rate": 1.0906405610142445e-05, + "loss": 0.0343, + "step": 29741 + }, + { + "epoch": 3.9, + "learning_rate": 1.0905877466794679e-05, + "loss": 0.0301, + "step": 29742 + }, + { + "epoch": 3.9, + "learning_rate": 1.0905349320899179e-05, + "loss": 0.0316, + "step": 29743 + }, + { + "epoch": 3.9, + "learning_rate": 1.0904821172457436e-05, + "loss": 0.0385, + "step": 29744 + }, + { + "epoch": 3.9, + "learning_rate": 1.090429302147093e-05, + "loss": 0.0312, + "step": 29745 + }, + { + "epoch": 3.9, + "learning_rate": 1.0903764867941146e-05, + "loss": 0.0318, + "step": 29746 + }, + { + "epoch": 3.9, + "learning_rate": 1.0903236711869573e-05, + "loss": 0.0366, + "step": 29747 + }, + { + "epoch": 3.9, + "learning_rate": 1.0902708553257698e-05, + "loss": 0.0305, + "step": 29748 + }, + { + "epoch": 3.9, + "learning_rate": 1.0902180392107e-05, + "loss": 0.0292, + "step": 29749 + }, + { + "epoch": 3.9, + "learning_rate": 1.0901652228418968e-05, + "loss": 0.0264, + "step": 29750 + }, + { + "epoch": 3.9, + "learning_rate": 1.0901124062195087e-05, + "loss": 0.0358, + "step": 29751 + }, + { + "epoch": 3.9, + "learning_rate": 1.0900595893436843e-05, + "loss": 0.034, + "step": 29752 + }, + { + "epoch": 3.9, + "learning_rate": 1.0900067722145722e-05, + "loss": 0.0279, + "step": 29753 + }, + { + "epoch": 3.9, + "learning_rate": 1.0899539548323205e-05, + "loss": 0.0381, + "step": 29754 + }, + { + "epoch": 3.9, + "learning_rate": 1.0899011371970785e-05, + "loss": 0.0313, + "step": 29755 + }, + { + "epoch": 3.9, + "learning_rate": 1.0898483193089942e-05, + "loss": 0.0392, + "step": 29756 + }, + { + "epoch": 3.9, + "learning_rate": 1.0897955011682165e-05, + "loss": 0.0362, + "step": 29757 + }, + { + "epoch": 3.9, + "learning_rate": 1.0897426827748932e-05, + "loss": 0.0372, + "step": 29758 + }, + { + "epoch": 3.9, + "learning_rate": 1.0896898641291738e-05, + "loss": 0.0322, + "step": 29759 + }, + { + "epoch": 3.9, + "learning_rate": 1.0896370452312061e-05, + "loss": 0.0383, + "step": 29760 + }, + { + "epoch": 3.9, + "learning_rate": 1.0895842260811394e-05, + "loss": 0.0319, + "step": 29761 + }, + { + "epoch": 3.9, + "learning_rate": 1.0895314066791217e-05, + "loss": 0.0291, + "step": 29762 + }, + { + "epoch": 3.9, + "learning_rate": 1.0894785870253016e-05, + "loss": 0.0374, + "step": 29763 + }, + { + "epoch": 3.9, + "learning_rate": 1.0894257671198276e-05, + "loss": 0.0293, + "step": 29764 + }, + { + "epoch": 3.9, + "learning_rate": 1.0893729469628486e-05, + "loss": 0.0341, + "step": 29765 + }, + { + "epoch": 3.9, + "learning_rate": 1.0893201265545128e-05, + "loss": 0.0395, + "step": 29766 + }, + { + "epoch": 3.9, + "learning_rate": 1.089267305894969e-05, + "loss": 0.0362, + "step": 29767 + }, + { + "epoch": 3.9, + "learning_rate": 1.0892144849843657e-05, + "loss": 0.0404, + "step": 29768 + }, + { + "epoch": 3.9, + "learning_rate": 1.0891616638228516e-05, + "loss": 0.0362, + "step": 29769 + }, + { + "epoch": 3.9, + "learning_rate": 1.0891088424105745e-05, + "loss": 0.0349, + "step": 29770 + }, + { + "epoch": 3.9, + "learning_rate": 1.0890560207476839e-05, + "loss": 0.037, + "step": 29771 + }, + { + "epoch": 3.9, + "learning_rate": 1.0890031988343282e-05, + "loss": 0.0346, + "step": 29772 + }, + { + "epoch": 3.9, + "learning_rate": 1.0889503766706553e-05, + "loss": 0.0425, + "step": 29773 + }, + { + "epoch": 3.9, + "learning_rate": 1.0888975542568144e-05, + "loss": 0.0358, + "step": 29774 + }, + { + "epoch": 3.9, + "learning_rate": 1.0888447315929538e-05, + "loss": 0.0295, + "step": 29775 + }, + { + "epoch": 3.9, + "learning_rate": 1.0887919086792222e-05, + "loss": 0.0358, + "step": 29776 + }, + { + "epoch": 3.9, + "learning_rate": 1.088739085515768e-05, + "loss": 0.0389, + "step": 29777 + }, + { + "epoch": 3.9, + "learning_rate": 1.0886862621027401e-05, + "loss": 0.0344, + "step": 29778 + }, + { + "epoch": 3.9, + "learning_rate": 1.0886334384402866e-05, + "loss": 0.0351, + "step": 29779 + }, + { + "epoch": 3.9, + "learning_rate": 1.0885806145285563e-05, + "loss": 0.0323, + "step": 29780 + }, + { + "epoch": 3.9, + "learning_rate": 1.0885277903676974e-05, + "loss": 0.0353, + "step": 29781 + }, + { + "epoch": 3.9, + "learning_rate": 1.0884749659578593e-05, + "loss": 0.0338, + "step": 29782 + }, + { + "epoch": 3.9, + "learning_rate": 1.08842214129919e-05, + "loss": 0.0328, + "step": 29783 + }, + { + "epoch": 3.9, + "learning_rate": 1.0883693163918381e-05, + "loss": 0.036, + "step": 29784 + }, + { + "epoch": 3.9, + "learning_rate": 1.0883164912359519e-05, + "loss": 0.0378, + "step": 29785 + }, + { + "epoch": 3.9, + "learning_rate": 1.0882636658316807e-05, + "loss": 0.0354, + "step": 29786 + }, + { + "epoch": 3.9, + "learning_rate": 1.0882108401791722e-05, + "loss": 0.0342, + "step": 29787 + }, + { + "epoch": 3.9, + "learning_rate": 1.088158014278576e-05, + "loss": 0.0318, + "step": 29788 + }, + { + "epoch": 3.9, + "learning_rate": 1.0881051881300398e-05, + "loss": 0.0366, + "step": 29789 + }, + { + "epoch": 3.9, + "learning_rate": 1.0880523617337122e-05, + "loss": 0.0278, + "step": 29790 + }, + { + "epoch": 3.9, + "learning_rate": 1.087999535089742e-05, + "loss": 0.0323, + "step": 29791 + }, + { + "epoch": 3.9, + "learning_rate": 1.087946708198278e-05, + "loss": 0.0344, + "step": 29792 + }, + { + "epoch": 3.9, + "learning_rate": 1.0878938810594688e-05, + "loss": 0.0351, + "step": 29793 + }, + { + "epoch": 3.9, + "learning_rate": 1.0878410536734624e-05, + "loss": 0.0317, + "step": 29794 + }, + { + "epoch": 3.9, + "learning_rate": 1.087788226040408e-05, + "loss": 0.0323, + "step": 29795 + }, + { + "epoch": 3.9, + "learning_rate": 1.0877353981604534e-05, + "loss": 0.0368, + "step": 29796 + }, + { + "epoch": 3.9, + "learning_rate": 1.0876825700337482e-05, + "loss": 0.0352, + "step": 29797 + }, + { + "epoch": 3.9, + "learning_rate": 1.0876297416604402e-05, + "loss": 0.0335, + "step": 29798 + }, + { + "epoch": 3.9, + "learning_rate": 1.0875769130406784e-05, + "loss": 0.0315, + "step": 29799 + }, + { + "epoch": 3.9, + "learning_rate": 1.0875240841746108e-05, + "loss": 0.0377, + "step": 29800 + }, + { + "epoch": 3.9, + "learning_rate": 1.087471255062387e-05, + "loss": 0.0379, + "step": 29801 + }, + { + "epoch": 3.9, + "learning_rate": 1.0874184257041543e-05, + "loss": 0.04, + "step": 29802 + }, + { + "epoch": 3.9, + "learning_rate": 1.0873655961000621e-05, + "loss": 0.0441, + "step": 29803 + }, + { + "epoch": 3.9, + "learning_rate": 1.087312766250259e-05, + "loss": 0.0294, + "step": 29804 + }, + { + "epoch": 3.9, + "learning_rate": 1.0872599361548935e-05, + "loss": 0.0387, + "step": 29805 + }, + { + "epoch": 3.9, + "learning_rate": 1.0872071058141139e-05, + "loss": 0.0312, + "step": 29806 + }, + { + "epoch": 3.9, + "learning_rate": 1.0871542752280688e-05, + "loss": 0.0371, + "step": 29807 + }, + { + "epoch": 3.9, + "learning_rate": 1.087101444396907e-05, + "loss": 0.0338, + "step": 29808 + }, + { + "epoch": 3.9, + "learning_rate": 1.0870486133207771e-05, + "loss": 0.0371, + "step": 29809 + }, + { + "epoch": 3.9, + "learning_rate": 1.0869957819998278e-05, + "loss": 0.0342, + "step": 29810 + }, + { + "epoch": 3.91, + "learning_rate": 1.086942950434207e-05, + "loss": 0.0386, + "step": 29811 + }, + { + "epoch": 3.91, + "learning_rate": 1.0868901186240644e-05, + "loss": 0.0329, + "step": 29812 + }, + { + "epoch": 3.91, + "learning_rate": 1.0868372865695476e-05, + "loss": 0.0346, + "step": 29813 + }, + { + "epoch": 3.91, + "learning_rate": 1.0867844542708056e-05, + "loss": 0.0325, + "step": 29814 + }, + { + "epoch": 3.91, + "learning_rate": 1.0867316217279869e-05, + "loss": 0.0376, + "step": 29815 + }, + { + "epoch": 3.91, + "learning_rate": 1.0866787889412405e-05, + "loss": 0.0341, + "step": 29816 + }, + { + "epoch": 3.91, + "learning_rate": 1.0866259559107142e-05, + "loss": 0.037, + "step": 29817 + }, + { + "epoch": 3.91, + "learning_rate": 1.086573122636557e-05, + "loss": 0.0382, + "step": 29818 + }, + { + "epoch": 3.91, + "learning_rate": 1.0865202891189176e-05, + "loss": 0.032, + "step": 29819 + }, + { + "epoch": 3.91, + "learning_rate": 1.0864674553579447e-05, + "loss": 0.0308, + "step": 29820 + }, + { + "epoch": 3.91, + "learning_rate": 1.0864146213537862e-05, + "loss": 0.0297, + "step": 29821 + }, + { + "epoch": 3.91, + "learning_rate": 1.0863617871065918e-05, + "loss": 0.0348, + "step": 29822 + }, + { + "epoch": 3.91, + "learning_rate": 1.086308952616509e-05, + "loss": 0.0333, + "step": 29823 + }, + { + "epoch": 3.91, + "learning_rate": 1.0862561178836871e-05, + "loss": 0.0397, + "step": 29824 + }, + { + "epoch": 3.91, + "learning_rate": 1.0862032829082745e-05, + "loss": 0.0323, + "step": 29825 + }, + { + "epoch": 3.91, + "learning_rate": 1.0861504476904196e-05, + "loss": 0.0297, + "step": 29826 + }, + { + "epoch": 3.91, + "learning_rate": 1.0860976122302714e-05, + "loss": 0.0355, + "step": 29827 + }, + { + "epoch": 3.91, + "learning_rate": 1.086044776527978e-05, + "loss": 0.0422, + "step": 29828 + }, + { + "epoch": 3.91, + "learning_rate": 1.0859919405836881e-05, + "loss": 0.0388, + "step": 29829 + }, + { + "epoch": 3.91, + "learning_rate": 1.0859391043975506e-05, + "loss": 0.0372, + "step": 29830 + }, + { + "epoch": 3.91, + "learning_rate": 1.0858862679697141e-05, + "loss": 0.0293, + "step": 29831 + }, + { + "epoch": 3.91, + "learning_rate": 1.0858334313003269e-05, + "loss": 0.0327, + "step": 29832 + }, + { + "epoch": 3.91, + "learning_rate": 1.0857805943895378e-05, + "loss": 0.0379, + "step": 29833 + }, + { + "epoch": 3.91, + "learning_rate": 1.0857277572374954e-05, + "loss": 0.0362, + "step": 29834 + }, + { + "epoch": 3.91, + "learning_rate": 1.0856749198443482e-05, + "loss": 0.033, + "step": 29835 + }, + { + "epoch": 3.91, + "learning_rate": 1.0856220822102448e-05, + "loss": 0.0414, + "step": 29836 + }, + { + "epoch": 3.91, + "learning_rate": 1.0855692443353341e-05, + "loss": 0.0325, + "step": 29837 + }, + { + "epoch": 3.91, + "learning_rate": 1.085516406219764e-05, + "loss": 0.03, + "step": 29838 + }, + { + "epoch": 3.91, + "learning_rate": 1.085463567863684e-05, + "loss": 0.0345, + "step": 29839 + }, + { + "epoch": 3.91, + "learning_rate": 1.0854107292672419e-05, + "loss": 0.0328, + "step": 29840 + }, + { + "epoch": 3.91, + "learning_rate": 1.0853578904305868e-05, + "loss": 0.0368, + "step": 29841 + }, + { + "epoch": 3.91, + "learning_rate": 1.0853050513538672e-05, + "loss": 0.0385, + "step": 29842 + }, + { + "epoch": 3.91, + "learning_rate": 1.085252212037232e-05, + "loss": 0.0341, + "step": 29843 + }, + { + "epoch": 3.91, + "learning_rate": 1.0851993724808291e-05, + "loss": 0.0286, + "step": 29844 + }, + { + "epoch": 3.91, + "learning_rate": 1.0851465326848075e-05, + "loss": 0.0347, + "step": 29845 + }, + { + "epoch": 3.91, + "learning_rate": 1.0850936926493159e-05, + "loss": 0.0344, + "step": 29846 + }, + { + "epoch": 3.91, + "learning_rate": 1.085040852374503e-05, + "loss": 0.037, + "step": 29847 + }, + { + "epoch": 3.91, + "learning_rate": 1.084988011860517e-05, + "loss": 0.0321, + "step": 29848 + }, + { + "epoch": 3.91, + "learning_rate": 1.084935171107507e-05, + "loss": 0.0373, + "step": 29849 + }, + { + "epoch": 3.91, + "learning_rate": 1.084882330115621e-05, + "loss": 0.0363, + "step": 29850 + }, + { + "epoch": 3.91, + "learning_rate": 1.0848294888850082e-05, + "loss": 0.0364, + "step": 29851 + }, + { + "epoch": 3.91, + "learning_rate": 1.0847766474158168e-05, + "loss": 0.0347, + "step": 29852 + }, + { + "epoch": 3.91, + "learning_rate": 1.084723805708196e-05, + "loss": 0.0369, + "step": 29853 + }, + { + "epoch": 3.91, + "learning_rate": 1.0846709637622933e-05, + "loss": 0.0392, + "step": 29854 + }, + { + "epoch": 3.91, + "learning_rate": 1.0846181215782584e-05, + "loss": 0.0377, + "step": 29855 + }, + { + "epoch": 3.91, + "learning_rate": 1.0845652791562398e-05, + "loss": 0.04, + "step": 29856 + }, + { + "epoch": 3.91, + "learning_rate": 1.0845124364963855e-05, + "loss": 0.029, + "step": 29857 + }, + { + "epoch": 3.91, + "learning_rate": 1.0844595935988446e-05, + "loss": 0.0328, + "step": 29858 + }, + { + "epoch": 3.91, + "learning_rate": 1.0844067504637656e-05, + "loss": 0.0343, + "step": 29859 + }, + { + "epoch": 3.91, + "learning_rate": 1.0843539070912974e-05, + "loss": 0.0403, + "step": 29860 + }, + { + "epoch": 3.91, + "learning_rate": 1.0843010634815878e-05, + "loss": 0.0287, + "step": 29861 + }, + { + "epoch": 3.91, + "learning_rate": 1.084248219634786e-05, + "loss": 0.0317, + "step": 29862 + }, + { + "epoch": 3.91, + "learning_rate": 1.0841953755510407e-05, + "loss": 0.03, + "step": 29863 + }, + { + "epoch": 3.91, + "learning_rate": 1.0841425312305008e-05, + "loss": 0.0342, + "step": 29864 + }, + { + "epoch": 3.91, + "learning_rate": 1.0840896866733142e-05, + "loss": 0.0348, + "step": 29865 + }, + { + "epoch": 3.91, + "learning_rate": 1.0840368418796299e-05, + "loss": 0.0343, + "step": 29866 + }, + { + "epoch": 3.91, + "learning_rate": 1.083983996849596e-05, + "loss": 0.0397, + "step": 29867 + }, + { + "epoch": 3.91, + "learning_rate": 1.0839311515833622e-05, + "loss": 0.0379, + "step": 29868 + }, + { + "epoch": 3.91, + "learning_rate": 1.083878306081076e-05, + "loss": 0.0343, + "step": 29869 + }, + { + "epoch": 3.91, + "learning_rate": 1.083825460342887e-05, + "loss": 0.0389, + "step": 29870 + }, + { + "epoch": 3.91, + "learning_rate": 1.0837726143689432e-05, + "loss": 0.0355, + "step": 29871 + }, + { + "epoch": 3.91, + "learning_rate": 1.0837197681593932e-05, + "loss": 0.0349, + "step": 29872 + }, + { + "epoch": 3.91, + "learning_rate": 1.0836669217143858e-05, + "loss": 0.0324, + "step": 29873 + }, + { + "epoch": 3.91, + "learning_rate": 1.0836140750340698e-05, + "loss": 0.0326, + "step": 29874 + }, + { + "epoch": 3.91, + "learning_rate": 1.0835612281185937e-05, + "loss": 0.0304, + "step": 29875 + }, + { + "epoch": 3.91, + "learning_rate": 1.0835083809681057e-05, + "loss": 0.0293, + "step": 29876 + }, + { + "epoch": 3.91, + "learning_rate": 1.0834555335827555e-05, + "loss": 0.0343, + "step": 29877 + }, + { + "epoch": 3.91, + "learning_rate": 1.0834026859626905e-05, + "loss": 0.0274, + "step": 29878 + }, + { + "epoch": 3.91, + "learning_rate": 1.08334983810806e-05, + "loss": 0.0334, + "step": 29879 + }, + { + "epoch": 3.91, + "learning_rate": 1.0832969900190128e-05, + "loss": 0.0393, + "step": 29880 + }, + { + "epoch": 3.91, + "learning_rate": 1.0832441416956971e-05, + "loss": 0.0376, + "step": 29881 + }, + { + "epoch": 3.91, + "learning_rate": 1.0831912931382616e-05, + "loss": 0.0331, + "step": 29882 + }, + { + "epoch": 3.91, + "learning_rate": 1.083138444346855e-05, + "loss": 0.0346, + "step": 29883 + }, + { + "epoch": 3.91, + "learning_rate": 1.0830855953216259e-05, + "loss": 0.0328, + "step": 29884 + }, + { + "epoch": 3.91, + "learning_rate": 1.0830327460627233e-05, + "loss": 0.0378, + "step": 29885 + }, + { + "epoch": 3.91, + "learning_rate": 1.0829798965702951e-05, + "loss": 0.0372, + "step": 29886 + }, + { + "epoch": 3.91, + "learning_rate": 1.0829270468444908e-05, + "loss": 0.0313, + "step": 29887 + }, + { + "epoch": 3.92, + "learning_rate": 1.0828741968854583e-05, + "loss": 0.0385, + "step": 29888 + }, + { + "epoch": 3.92, + "learning_rate": 1.0828213466933466e-05, + "loss": 0.0327, + "step": 29889 + }, + { + "epoch": 3.92, + "learning_rate": 1.0827684962683042e-05, + "loss": 0.0339, + "step": 29890 + }, + { + "epoch": 3.92, + "learning_rate": 1.08271564561048e-05, + "loss": 0.0297, + "step": 29891 + }, + { + "epoch": 3.92, + "learning_rate": 1.0826627947200224e-05, + "loss": 0.032, + "step": 29892 + }, + { + "epoch": 3.92, + "learning_rate": 1.0826099435970799e-05, + "loss": 0.0355, + "step": 29893 + }, + { + "epoch": 3.92, + "learning_rate": 1.0825570922418015e-05, + "loss": 0.0363, + "step": 29894 + }, + { + "epoch": 3.92, + "learning_rate": 1.0825042406543359e-05, + "loss": 0.0327, + "step": 29895 + }, + { + "epoch": 3.92, + "learning_rate": 1.0824513888348309e-05, + "loss": 0.0322, + "step": 29896 + }, + { + "epoch": 3.92, + "learning_rate": 1.0823985367834361e-05, + "loss": 0.0366, + "step": 29897 + }, + { + "epoch": 3.92, + "learning_rate": 1.0823456845003e-05, + "loss": 0.0321, + "step": 29898 + }, + { + "epoch": 3.92, + "learning_rate": 1.082292831985571e-05, + "loss": 0.0301, + "step": 29899 + }, + { + "epoch": 3.92, + "learning_rate": 1.0822399792393975e-05, + "loss": 0.0329, + "step": 29900 + }, + { + "epoch": 3.92, + "learning_rate": 1.0821871262619285e-05, + "loss": 0.0347, + "step": 29901 + }, + { + "epoch": 3.92, + "learning_rate": 1.0821342730533129e-05, + "loss": 0.0317, + "step": 29902 + }, + { + "epoch": 3.92, + "learning_rate": 1.0820814196136989e-05, + "loss": 0.036, + "step": 29903 + }, + { + "epoch": 3.92, + "learning_rate": 1.0820285659432352e-05, + "loss": 0.0329, + "step": 29904 + }, + { + "epoch": 3.92, + "learning_rate": 1.0819757120420702e-05, + "loss": 0.0358, + "step": 29905 + }, + { + "epoch": 3.92, + "learning_rate": 1.0819228579103534e-05, + "loss": 0.038, + "step": 29906 + }, + { + "epoch": 3.92, + "learning_rate": 1.0818700035482326e-05, + "loss": 0.036, + "step": 29907 + }, + { + "epoch": 3.92, + "learning_rate": 1.081817148955857e-05, + "loss": 0.0407, + "step": 29908 + }, + { + "epoch": 3.92, + "learning_rate": 1.081764294133375e-05, + "loss": 0.0332, + "step": 29909 + }, + { + "epoch": 3.92, + "learning_rate": 1.0817114390809354e-05, + "loss": 0.0395, + "step": 29910 + }, + { + "epoch": 3.92, + "learning_rate": 1.0816585837986862e-05, + "loss": 0.0352, + "step": 29911 + }, + { + "epoch": 3.92, + "learning_rate": 1.0816057282867771e-05, + "loss": 0.0298, + "step": 29912 + }, + { + "epoch": 3.92, + "learning_rate": 1.081552872545356e-05, + "loss": 0.0391, + "step": 29913 + }, + { + "epoch": 3.92, + "learning_rate": 1.0815000165745719e-05, + "loss": 0.0366, + "step": 29914 + }, + { + "epoch": 3.92, + "learning_rate": 1.0814471603745734e-05, + "loss": 0.0342, + "step": 29915 + }, + { + "epoch": 3.92, + "learning_rate": 1.0813943039455089e-05, + "loss": 0.0424, + "step": 29916 + }, + { + "epoch": 3.92, + "learning_rate": 1.0813414472875273e-05, + "loss": 0.0344, + "step": 29917 + }, + { + "epoch": 3.92, + "learning_rate": 1.081288590400777e-05, + "loss": 0.0361, + "step": 29918 + }, + { + "epoch": 3.92, + "learning_rate": 1.0812357332854072e-05, + "loss": 0.0348, + "step": 29919 + }, + { + "epoch": 3.92, + "learning_rate": 1.0811828759415661e-05, + "loss": 0.0362, + "step": 29920 + }, + { + "epoch": 3.92, + "learning_rate": 1.0811300183694027e-05, + "loss": 0.0308, + "step": 29921 + }, + { + "epoch": 3.92, + "learning_rate": 1.081077160569065e-05, + "loss": 0.0354, + "step": 29922 + }, + { + "epoch": 3.92, + "learning_rate": 1.0810243025407026e-05, + "loss": 0.033, + "step": 29923 + }, + { + "epoch": 3.92, + "learning_rate": 1.0809714442844635e-05, + "loss": 0.0308, + "step": 29924 + }, + { + "epoch": 3.92, + "learning_rate": 1.0809185858004965e-05, + "loss": 0.0373, + "step": 29925 + }, + { + "epoch": 3.92, + "learning_rate": 1.0808657270889503e-05, + "loss": 0.0395, + "step": 29926 + }, + { + "epoch": 3.92, + "learning_rate": 1.0808128681499732e-05, + "loss": 0.0425, + "step": 29927 + }, + { + "epoch": 3.92, + "learning_rate": 1.0807600089837144e-05, + "loss": 0.0353, + "step": 29928 + }, + { + "epoch": 3.92, + "learning_rate": 1.0807071495903228e-05, + "loss": 0.0233, + "step": 29929 + }, + { + "epoch": 3.92, + "learning_rate": 1.080654289969946e-05, + "loss": 0.0371, + "step": 29930 + }, + { + "epoch": 3.92, + "learning_rate": 1.0806014301227342e-05, + "loss": 0.0367, + "step": 29931 + }, + { + "epoch": 3.92, + "learning_rate": 1.0805485700488344e-05, + "loss": 0.0302, + "step": 29932 + }, + { + "epoch": 3.92, + "learning_rate": 1.0804957097483961e-05, + "loss": 0.0358, + "step": 29933 + }, + { + "epoch": 3.92, + "learning_rate": 1.080442849221568e-05, + "loss": 0.0336, + "step": 29934 + }, + { + "epoch": 3.92, + "learning_rate": 1.0803899884684989e-05, + "loss": 0.0336, + "step": 29935 + }, + { + "epoch": 3.92, + "learning_rate": 1.0803371274893369e-05, + "loss": 0.0317, + "step": 29936 + }, + { + "epoch": 3.92, + "learning_rate": 1.0802842662842312e-05, + "loss": 0.0354, + "step": 29937 + }, + { + "epoch": 3.92, + "learning_rate": 1.08023140485333e-05, + "loss": 0.0328, + "step": 29938 + }, + { + "epoch": 3.92, + "learning_rate": 1.0801785431967829e-05, + "loss": 0.0345, + "step": 29939 + }, + { + "epoch": 3.92, + "learning_rate": 1.0801256813147371e-05, + "loss": 0.0372, + "step": 29940 + }, + { + "epoch": 3.92, + "learning_rate": 1.0800728192073426e-05, + "loss": 0.0346, + "step": 29941 + }, + { + "epoch": 3.92, + "learning_rate": 1.0800199568747477e-05, + "loss": 0.0359, + "step": 29942 + }, + { + "epoch": 3.92, + "learning_rate": 1.0799670943171006e-05, + "loss": 0.0321, + "step": 29943 + }, + { + "epoch": 3.92, + "learning_rate": 1.07991423153455e-05, + "loss": 0.0383, + "step": 29944 + }, + { + "epoch": 3.92, + "learning_rate": 1.0798613685272454e-05, + "loss": 0.0311, + "step": 29945 + }, + { + "epoch": 3.92, + "learning_rate": 1.0798085052953349e-05, + "loss": 0.033, + "step": 29946 + }, + { + "epoch": 3.92, + "learning_rate": 1.0797556418389672e-05, + "loss": 0.0407, + "step": 29947 + }, + { + "epoch": 3.92, + "learning_rate": 1.0797027781582909e-05, + "loss": 0.0434, + "step": 29948 + }, + { + "epoch": 3.92, + "learning_rate": 1.0796499142534547e-05, + "loss": 0.0307, + "step": 29949 + }, + { + "epoch": 3.92, + "learning_rate": 1.0795970501246076e-05, + "loss": 0.0306, + "step": 29950 + }, + { + "epoch": 3.92, + "learning_rate": 1.0795441857718978e-05, + "loss": 0.0341, + "step": 29951 + }, + { + "epoch": 3.92, + "learning_rate": 1.0794913211954745e-05, + "loss": 0.0378, + "step": 29952 + }, + { + "epoch": 3.92, + "learning_rate": 1.079438456395486e-05, + "loss": 0.0347, + "step": 29953 + }, + { + "epoch": 3.92, + "learning_rate": 1.079385591372081e-05, + "loss": 0.0333, + "step": 29954 + }, + { + "epoch": 3.92, + "learning_rate": 1.0793327261254081e-05, + "loss": 0.0332, + "step": 29955 + }, + { + "epoch": 3.92, + "learning_rate": 1.0792798606556162e-05, + "loss": 0.0404, + "step": 29956 + }, + { + "epoch": 3.92, + "learning_rate": 1.0792269949628543e-05, + "loss": 0.0387, + "step": 29957 + }, + { + "epoch": 3.92, + "learning_rate": 1.0791741290472701e-05, + "loss": 0.0334, + "step": 29958 + }, + { + "epoch": 3.92, + "learning_rate": 1.0791212629090134e-05, + "loss": 0.0334, + "step": 29959 + }, + { + "epoch": 3.92, + "learning_rate": 1.0790683965482321e-05, + "loss": 0.042, + "step": 29960 + }, + { + "epoch": 3.92, + "learning_rate": 1.079015529965075e-05, + "loss": 0.0338, + "step": 29961 + }, + { + "epoch": 3.92, + "learning_rate": 1.078962663159691e-05, + "loss": 0.0347, + "step": 29962 + }, + { + "epoch": 3.92, + "learning_rate": 1.078909796132229e-05, + "loss": 0.037, + "step": 29963 + }, + { + "epoch": 3.93, + "learning_rate": 1.0788569288828373e-05, + "loss": 0.0337, + "step": 29964 + }, + { + "epoch": 3.93, + "learning_rate": 1.0788040614116646e-05, + "loss": 0.0318, + "step": 29965 + }, + { + "epoch": 3.93, + "learning_rate": 1.0787511937188594e-05, + "loss": 0.0347, + "step": 29966 + }, + { + "epoch": 3.93, + "learning_rate": 1.0786983258045711e-05, + "loss": 0.0402, + "step": 29967 + }, + { + "epoch": 3.93, + "learning_rate": 1.0786454576689478e-05, + "loss": 0.0294, + "step": 29968 + }, + { + "epoch": 3.93, + "learning_rate": 1.0785925893121385e-05, + "loss": 0.0264, + "step": 29969 + }, + { + "epoch": 3.93, + "learning_rate": 1.0785397207342915e-05, + "loss": 0.0443, + "step": 29970 + }, + { + "epoch": 3.93, + "learning_rate": 1.0784868519355558e-05, + "loss": 0.0299, + "step": 29971 + }, + { + "epoch": 3.93, + "learning_rate": 1.0784339829160799e-05, + "loss": 0.0386, + "step": 29972 + }, + { + "epoch": 3.93, + "learning_rate": 1.0783811136760127e-05, + "loss": 0.0327, + "step": 29973 + }, + { + "epoch": 3.93, + "learning_rate": 1.0783282442155027e-05, + "loss": 0.0347, + "step": 29974 + }, + { + "epoch": 3.93, + "learning_rate": 1.078275374534699e-05, + "loss": 0.0277, + "step": 29975 + }, + { + "epoch": 3.93, + "learning_rate": 1.0782225046337496e-05, + "loss": 0.0394, + "step": 29976 + }, + { + "epoch": 3.93, + "learning_rate": 1.078169634512804e-05, + "loss": 0.0336, + "step": 29977 + }, + { + "epoch": 3.93, + "learning_rate": 1.0781167641720098e-05, + "loss": 0.0407, + "step": 29978 + }, + { + "epoch": 3.93, + "learning_rate": 1.078063893611517e-05, + "loss": 0.0301, + "step": 29979 + }, + { + "epoch": 3.93, + "learning_rate": 1.0780110228314733e-05, + "loss": 0.0376, + "step": 29980 + }, + { + "epoch": 3.93, + "learning_rate": 1.077958151832028e-05, + "loss": 0.0367, + "step": 29981 + }, + { + "epoch": 3.93, + "learning_rate": 1.0779052806133292e-05, + "loss": 0.0306, + "step": 29982 + }, + { + "epoch": 3.93, + "learning_rate": 1.077852409175526e-05, + "loss": 0.0353, + "step": 29983 + }, + { + "epoch": 3.93, + "learning_rate": 1.0777995375187673e-05, + "loss": 0.0336, + "step": 29984 + }, + { + "epoch": 3.93, + "learning_rate": 1.0777466656432015e-05, + "loss": 0.0322, + "step": 29985 + }, + { + "epoch": 3.93, + "learning_rate": 1.0776937935489773e-05, + "loss": 0.0349, + "step": 29986 + }, + { + "epoch": 3.93, + "learning_rate": 1.0776409212362433e-05, + "loss": 0.0297, + "step": 29987 + }, + { + "epoch": 3.93, + "learning_rate": 1.0775880487051485e-05, + "loss": 0.036, + "step": 29988 + }, + { + "epoch": 3.93, + "learning_rate": 1.0775351759558414e-05, + "loss": 0.0315, + "step": 29989 + }, + { + "epoch": 3.93, + "learning_rate": 1.0774823029884708e-05, + "loss": 0.0365, + "step": 29990 + }, + { + "epoch": 3.93, + "learning_rate": 1.0774294298031852e-05, + "loss": 0.0337, + "step": 29991 + }, + { + "epoch": 3.93, + "learning_rate": 1.077376556400134e-05, + "loss": 0.0327, + "step": 29992 + }, + { + "epoch": 3.93, + "learning_rate": 1.0773236827794644e-05, + "loss": 0.0377, + "step": 29993 + }, + { + "epoch": 3.93, + "learning_rate": 1.0772708089413268e-05, + "loss": 0.0319, + "step": 29994 + }, + { + "epoch": 3.93, + "learning_rate": 1.0772179348858689e-05, + "loss": 0.0363, + "step": 29995 + }, + { + "epoch": 3.93, + "learning_rate": 1.0771650606132398e-05, + "loss": 0.0382, + "step": 29996 + }, + { + "epoch": 3.93, + "learning_rate": 1.077112186123588e-05, + "loss": 0.0289, + "step": 29997 + }, + { + "epoch": 3.93, + "learning_rate": 1.0770593114170621e-05, + "loss": 0.0321, + "step": 29998 + }, + { + "epoch": 3.93, + "learning_rate": 1.0770064364938111e-05, + "loss": 0.0305, + "step": 29999 + }, + { + "epoch": 3.93, + "learning_rate": 1.0769535613539838e-05, + "loss": 0.0384, + "step": 30000 + }, + { + "epoch": 3.93, + "learning_rate": 1.0769006859977284e-05, + "loss": 0.0313, + "step": 30001 + }, + { + "epoch": 3.93, + "learning_rate": 1.0768478104251943e-05, + "loss": 0.0344, + "step": 30002 + }, + { + "epoch": 3.93, + "learning_rate": 1.0767949346365296e-05, + "loss": 0.0324, + "step": 30003 + }, + { + "epoch": 3.93, + "learning_rate": 1.0767420586318833e-05, + "loss": 0.0253, + "step": 30004 + }, + { + "epoch": 3.93, + "learning_rate": 1.0766891824114039e-05, + "loss": 0.031, + "step": 30005 + }, + { + "epoch": 3.93, + "learning_rate": 1.0766363059752402e-05, + "loss": 0.0343, + "step": 30006 + }, + { + "epoch": 3.93, + "learning_rate": 1.0765834293235413e-05, + "loss": 0.038, + "step": 30007 + }, + { + "epoch": 3.93, + "learning_rate": 1.0765305524564554e-05, + "loss": 0.0321, + "step": 30008 + }, + { + "epoch": 3.93, + "learning_rate": 1.0764776753741313e-05, + "loss": 0.0347, + "step": 30009 + }, + { + "epoch": 3.93, + "learning_rate": 1.0764247980767177e-05, + "loss": 0.0418, + "step": 30010 + }, + { + "epoch": 3.93, + "learning_rate": 1.0763719205643638e-05, + "loss": 0.0362, + "step": 30011 + }, + { + "epoch": 3.93, + "learning_rate": 1.0763190428372176e-05, + "loss": 0.0377, + "step": 30012 + }, + { + "epoch": 3.93, + "learning_rate": 1.0762661648954287e-05, + "loss": 0.0353, + "step": 30013 + }, + { + "epoch": 3.93, + "learning_rate": 1.0762132867391448e-05, + "loss": 0.0407, + "step": 30014 + }, + { + "epoch": 3.93, + "learning_rate": 1.076160408368515e-05, + "loss": 0.0396, + "step": 30015 + }, + { + "epoch": 3.93, + "learning_rate": 1.0761075297836882e-05, + "loss": 0.0336, + "step": 30016 + }, + { + "epoch": 3.93, + "learning_rate": 1.0760546509848134e-05, + "loss": 0.0374, + "step": 30017 + }, + { + "epoch": 3.93, + "learning_rate": 1.0760017719720385e-05, + "loss": 0.03, + "step": 30018 + }, + { + "epoch": 3.93, + "learning_rate": 1.075948892745513e-05, + "loss": 0.0326, + "step": 30019 + }, + { + "epoch": 3.93, + "learning_rate": 1.0758960133053848e-05, + "loss": 0.0328, + "step": 30020 + }, + { + "epoch": 3.93, + "learning_rate": 1.0758431336518038e-05, + "loss": 0.0293, + "step": 30021 + }, + { + "epoch": 3.93, + "learning_rate": 1.0757902537849174e-05, + "loss": 0.044, + "step": 30022 + }, + { + "epoch": 3.93, + "learning_rate": 1.0757373737048752e-05, + "loss": 0.0329, + "step": 30023 + }, + { + "epoch": 3.93, + "learning_rate": 1.0756844934118258e-05, + "loss": 0.0315, + "step": 30024 + }, + { + "epoch": 3.93, + "learning_rate": 1.0756316129059177e-05, + "loss": 0.0393, + "step": 30025 + }, + { + "epoch": 3.93, + "learning_rate": 1.0755787321872995e-05, + "loss": 0.0332, + "step": 30026 + }, + { + "epoch": 3.93, + "learning_rate": 1.0755258512561202e-05, + "loss": 0.0305, + "step": 30027 + }, + { + "epoch": 3.93, + "learning_rate": 1.0754729701125288e-05, + "loss": 0.0352, + "step": 30028 + }, + { + "epoch": 3.93, + "learning_rate": 1.0754200887566734e-05, + "loss": 0.0335, + "step": 30029 + }, + { + "epoch": 3.93, + "learning_rate": 1.075367207188703e-05, + "loss": 0.0337, + "step": 30030 + }, + { + "epoch": 3.93, + "learning_rate": 1.0753143254087663e-05, + "loss": 0.0391, + "step": 30031 + }, + { + "epoch": 3.93, + "learning_rate": 1.0752614434170123e-05, + "loss": 0.0291, + "step": 30032 + }, + { + "epoch": 3.93, + "learning_rate": 1.0752085612135894e-05, + "loss": 0.0291, + "step": 30033 + }, + { + "epoch": 3.93, + "learning_rate": 1.0751556787986465e-05, + "loss": 0.0395, + "step": 30034 + }, + { + "epoch": 3.93, + "learning_rate": 1.0751027961723322e-05, + "loss": 0.0349, + "step": 30035 + }, + { + "epoch": 3.93, + "learning_rate": 1.0750499133347953e-05, + "loss": 0.0371, + "step": 30036 + }, + { + "epoch": 3.93, + "learning_rate": 1.0749970302861843e-05, + "loss": 0.0368, + "step": 30037 + }, + { + "epoch": 3.93, + "learning_rate": 1.0749441470266487e-05, + "loss": 0.0334, + "step": 30038 + }, + { + "epoch": 3.93, + "learning_rate": 1.074891263556336e-05, + "loss": 0.031, + "step": 30039 + }, + { + "epoch": 3.94, + "learning_rate": 1.074838379875396e-05, + "loss": 0.0318, + "step": 30040 + }, + { + "epoch": 3.94, + "learning_rate": 1.074785495983977e-05, + "loss": 0.04, + "step": 30041 + }, + { + "epoch": 3.94, + "learning_rate": 1.0747326118822277e-05, + "loss": 0.0359, + "step": 30042 + }, + { + "epoch": 3.94, + "learning_rate": 1.0746797275702968e-05, + "loss": 0.0264, + "step": 30043 + }, + { + "epoch": 3.94, + "learning_rate": 1.0746268430483335e-05, + "loss": 0.0344, + "step": 30044 + }, + { + "epoch": 3.94, + "learning_rate": 1.0745739583164858e-05, + "loss": 0.0381, + "step": 30045 + }, + { + "epoch": 3.94, + "learning_rate": 1.074521073374903e-05, + "loss": 0.0359, + "step": 30046 + }, + { + "epoch": 3.94, + "learning_rate": 1.0744681882237336e-05, + "loss": 0.0361, + "step": 30047 + }, + { + "epoch": 3.94, + "learning_rate": 1.0744153028631263e-05, + "loss": 0.0278, + "step": 30048 + }, + { + "epoch": 3.94, + "learning_rate": 1.07436241729323e-05, + "loss": 0.0368, + "step": 30049 + }, + { + "epoch": 3.94, + "learning_rate": 1.0743095315141935e-05, + "loss": 0.0327, + "step": 30050 + }, + { + "epoch": 3.94, + "learning_rate": 1.0742566455261654e-05, + "loss": 0.0326, + "step": 30051 + }, + { + "epoch": 3.94, + "learning_rate": 1.0742037593292945e-05, + "loss": 0.0404, + "step": 30052 + }, + { + "epoch": 3.94, + "learning_rate": 1.074150872923729e-05, + "loss": 0.0446, + "step": 30053 + }, + { + "epoch": 3.94, + "learning_rate": 1.0740979863096184e-05, + "loss": 0.0374, + "step": 30054 + }, + { + "epoch": 3.94, + "learning_rate": 1.0740450994871115e-05, + "loss": 0.0326, + "step": 30055 + }, + { + "epoch": 3.94, + "learning_rate": 1.0739922124563563e-05, + "loss": 0.0363, + "step": 30056 + }, + { + "epoch": 3.94, + "learning_rate": 1.073939325217502e-05, + "loss": 0.0333, + "step": 30057 + }, + { + "epoch": 3.94, + "learning_rate": 1.0738864377706973e-05, + "loss": 0.037, + "step": 30058 + }, + { + "epoch": 3.94, + "learning_rate": 1.073833550116091e-05, + "loss": 0.033, + "step": 30059 + }, + { + "epoch": 3.94, + "learning_rate": 1.0737806622538315e-05, + "loss": 0.0378, + "step": 30060 + }, + { + "epoch": 3.94, + "learning_rate": 1.0737277741840683e-05, + "loss": 0.032, + "step": 30061 + }, + { + "epoch": 3.94, + "learning_rate": 1.0736748859069492e-05, + "loss": 0.0324, + "step": 30062 + }, + { + "epoch": 3.94, + "learning_rate": 1.0736219974226236e-05, + "loss": 0.0333, + "step": 30063 + }, + { + "epoch": 3.94, + "learning_rate": 1.0735691087312402e-05, + "loss": 0.0379, + "step": 30064 + }, + { + "epoch": 3.94, + "learning_rate": 1.0735162198329474e-05, + "loss": 0.0441, + "step": 30065 + }, + { + "epoch": 3.94, + "learning_rate": 1.0734633307278941e-05, + "loss": 0.0334, + "step": 30066 + }, + { + "epoch": 3.94, + "learning_rate": 1.0734104414162291e-05, + "loss": 0.0316, + "step": 30067 + }, + { + "epoch": 3.94, + "learning_rate": 1.0733575518981014e-05, + "loss": 0.0319, + "step": 30068 + }, + { + "epoch": 3.94, + "learning_rate": 1.0733046621736594e-05, + "loss": 0.0334, + "step": 30069 + }, + { + "epoch": 3.94, + "learning_rate": 1.0732517722430517e-05, + "loss": 0.0376, + "step": 30070 + }, + { + "epoch": 3.94, + "learning_rate": 1.0731988821064274e-05, + "loss": 0.0348, + "step": 30071 + }, + { + "epoch": 3.94, + "learning_rate": 1.0731459917639356e-05, + "loss": 0.0343, + "step": 30072 + }, + { + "epoch": 3.94, + "learning_rate": 1.073093101215724e-05, + "loss": 0.0411, + "step": 30073 + }, + { + "epoch": 3.94, + "learning_rate": 1.0730402104619421e-05, + "loss": 0.0277, + "step": 30074 + }, + { + "epoch": 3.94, + "learning_rate": 1.0729873195027384e-05, + "loss": 0.0385, + "step": 30075 + }, + { + "epoch": 3.94, + "learning_rate": 1.072934428338262e-05, + "loss": 0.0396, + "step": 30076 + }, + { + "epoch": 3.94, + "learning_rate": 1.0728815369686614e-05, + "loss": 0.0384, + "step": 30077 + }, + { + "epoch": 3.94, + "learning_rate": 1.0728286453940852e-05, + "loss": 0.0356, + "step": 30078 + }, + { + "epoch": 3.94, + "learning_rate": 1.0727757536146824e-05, + "loss": 0.0412, + "step": 30079 + }, + { + "epoch": 3.94, + "learning_rate": 1.0727228616306015e-05, + "loss": 0.0305, + "step": 30080 + }, + { + "epoch": 3.94, + "learning_rate": 1.0726699694419915e-05, + "loss": 0.0324, + "step": 30081 + }, + { + "epoch": 3.94, + "learning_rate": 1.0726170770490014e-05, + "loss": 0.0325, + "step": 30082 + }, + { + "epoch": 3.94, + "learning_rate": 1.0725641844517793e-05, + "loss": 0.0322, + "step": 30083 + }, + { + "epoch": 3.94, + "learning_rate": 1.0725112916504743e-05, + "loss": 0.0336, + "step": 30084 + }, + { + "epoch": 3.94, + "learning_rate": 1.0724583986452355e-05, + "loss": 0.0262, + "step": 30085 + }, + { + "epoch": 3.94, + "learning_rate": 1.0724055054362108e-05, + "loss": 0.0329, + "step": 30086 + }, + { + "epoch": 3.94, + "learning_rate": 1.0723526120235496e-05, + "loss": 0.0398, + "step": 30087 + }, + { + "epoch": 3.94, + "learning_rate": 1.0722997184074007e-05, + "loss": 0.0385, + "step": 30088 + }, + { + "epoch": 3.94, + "learning_rate": 1.0722468245879128e-05, + "loss": 0.0304, + "step": 30089 + }, + { + "epoch": 3.94, + "learning_rate": 1.0721939305652342e-05, + "loss": 0.0432, + "step": 30090 + }, + { + "epoch": 3.94, + "learning_rate": 1.0721410363395143e-05, + "loss": 0.0325, + "step": 30091 + }, + { + "epoch": 3.94, + "learning_rate": 1.0720881419109013e-05, + "loss": 0.0356, + "step": 30092 + }, + { + "epoch": 3.94, + "learning_rate": 1.0720352472795446e-05, + "loss": 0.0318, + "step": 30093 + }, + { + "epoch": 3.94, + "learning_rate": 1.0719823524455923e-05, + "loss": 0.0358, + "step": 30094 + }, + { + "epoch": 3.94, + "learning_rate": 1.0719294574091938e-05, + "loss": 0.033, + "step": 30095 + }, + { + "epoch": 3.94, + "learning_rate": 1.0718765621704972e-05, + "loss": 0.0302, + "step": 30096 + }, + { + "epoch": 3.94, + "learning_rate": 1.0718236667296517e-05, + "loss": 0.0346, + "step": 30097 + }, + { + "epoch": 3.94, + "learning_rate": 1.0717707710868059e-05, + "loss": 0.0361, + "step": 30098 + }, + { + "epoch": 3.94, + "learning_rate": 1.071717875242109e-05, + "loss": 0.0338, + "step": 30099 + }, + { + "epoch": 3.94, + "learning_rate": 1.071664979195709e-05, + "loss": 0.0373, + "step": 30100 + }, + { + "epoch": 3.94, + "learning_rate": 1.0716120829477554e-05, + "loss": 0.0372, + "step": 30101 + }, + { + "epoch": 3.94, + "learning_rate": 1.071559186498396e-05, + "loss": 0.0277, + "step": 30102 + }, + { + "epoch": 3.94, + "learning_rate": 1.0715062898477808e-05, + "loss": 0.0313, + "step": 30103 + }, + { + "epoch": 3.94, + "learning_rate": 1.0714533929960577e-05, + "loss": 0.0347, + "step": 30104 + }, + { + "epoch": 3.94, + "learning_rate": 1.0714004959433759e-05, + "loss": 0.0302, + "step": 30105 + }, + { + "epoch": 3.94, + "learning_rate": 1.0713475986898837e-05, + "loss": 0.0343, + "step": 30106 + }, + { + "epoch": 3.94, + "learning_rate": 1.0712947012357305e-05, + "loss": 0.0288, + "step": 30107 + }, + { + "epoch": 3.94, + "learning_rate": 1.0712418035810646e-05, + "loss": 0.031, + "step": 30108 + }, + { + "epoch": 3.94, + "learning_rate": 1.071188905726035e-05, + "loss": 0.0301, + "step": 30109 + }, + { + "epoch": 3.94, + "learning_rate": 1.0711360076707904e-05, + "loss": 0.0298, + "step": 30110 + }, + { + "epoch": 3.94, + "learning_rate": 1.0710831094154794e-05, + "loss": 0.0337, + "step": 30111 + }, + { + "epoch": 3.94, + "learning_rate": 1.0710302109602513e-05, + "loss": 0.0354, + "step": 30112 + }, + { + "epoch": 3.94, + "learning_rate": 1.070977312305254e-05, + "loss": 0.0356, + "step": 30113 + }, + { + "epoch": 3.94, + "learning_rate": 1.0709244134506372e-05, + "loss": 0.0363, + "step": 30114 + }, + { + "epoch": 3.94, + "learning_rate": 1.0708715143965491e-05, + "loss": 0.0386, + "step": 30115 + }, + { + "epoch": 3.94, + "learning_rate": 1.0708186151431388e-05, + "loss": 0.0309, + "step": 30116 + }, + { + "epoch": 3.95, + "learning_rate": 1.0707657156905547e-05, + "loss": 0.0387, + "step": 30117 + }, + { + "epoch": 3.95, + "learning_rate": 1.0707128160389457e-05, + "loss": 0.0368, + "step": 30118 + }, + { + "epoch": 3.95, + "learning_rate": 1.0706599161884607e-05, + "loss": 0.0391, + "step": 30119 + }, + { + "epoch": 3.95, + "learning_rate": 1.070607016139249e-05, + "loss": 0.0389, + "step": 30120 + }, + { + "epoch": 3.95, + "learning_rate": 1.0705541158914584e-05, + "loss": 0.0317, + "step": 30121 + }, + { + "epoch": 3.95, + "learning_rate": 1.0705012154452381e-05, + "loss": 0.0316, + "step": 30122 + }, + { + "epoch": 3.95, + "learning_rate": 1.070448314800737e-05, + "loss": 0.0305, + "step": 30123 + }, + { + "epoch": 3.95, + "learning_rate": 1.0703954139581035e-05, + "loss": 0.0321, + "step": 30124 + }, + { + "epoch": 3.95, + "learning_rate": 1.0703425129174867e-05, + "loss": 0.0325, + "step": 30125 + }, + { + "epoch": 3.95, + "learning_rate": 1.0702896116790357e-05, + "loss": 0.0341, + "step": 30126 + }, + { + "epoch": 3.95, + "learning_rate": 1.0702367102428986e-05, + "loss": 0.0435, + "step": 30127 + }, + { + "epoch": 3.95, + "learning_rate": 1.0701838086092244e-05, + "loss": 0.0369, + "step": 30128 + }, + { + "epoch": 3.95, + "learning_rate": 1.0701309067781623e-05, + "loss": 0.0387, + "step": 30129 + }, + { + "epoch": 3.95, + "learning_rate": 1.0700780047498607e-05, + "loss": 0.0299, + "step": 30130 + }, + { + "epoch": 3.95, + "learning_rate": 1.0700251025244681e-05, + "loss": 0.0306, + "step": 30131 + }, + { + "epoch": 3.95, + "learning_rate": 1.069972200102134e-05, + "loss": 0.0344, + "step": 30132 + }, + { + "epoch": 3.95, + "learning_rate": 1.0699192974830069e-05, + "loss": 0.0331, + "step": 30133 + }, + { + "epoch": 3.95, + "learning_rate": 1.069866394667235e-05, + "loss": 0.0337, + "step": 30134 + }, + { + "epoch": 3.95, + "learning_rate": 1.0698134916549678e-05, + "loss": 0.0292, + "step": 30135 + }, + { + "epoch": 3.95, + "learning_rate": 1.069760588446354e-05, + "loss": 0.0339, + "step": 30136 + }, + { + "epoch": 3.95, + "learning_rate": 1.0697076850415422e-05, + "loss": 0.0342, + "step": 30137 + }, + { + "epoch": 3.95, + "learning_rate": 1.0696547814406812e-05, + "loss": 0.0358, + "step": 30138 + }, + { + "epoch": 3.95, + "learning_rate": 1.06960187764392e-05, + "loss": 0.0351, + "step": 30139 + }, + { + "epoch": 3.95, + "learning_rate": 1.0695489736514068e-05, + "loss": 0.0345, + "step": 30140 + }, + { + "epoch": 3.95, + "learning_rate": 1.0694960694632915e-05, + "loss": 0.0322, + "step": 30141 + }, + { + "epoch": 3.95, + "learning_rate": 1.0694431650797216e-05, + "loss": 0.0359, + "step": 30142 + }, + { + "epoch": 3.95, + "learning_rate": 1.0693902605008468e-05, + "loss": 0.0295, + "step": 30143 + }, + { + "epoch": 3.95, + "learning_rate": 1.0693373557268156e-05, + "loss": 0.0318, + "step": 30144 + }, + { + "epoch": 3.95, + "learning_rate": 1.0692844507577769e-05, + "loss": 0.0373, + "step": 30145 + }, + { + "epoch": 3.95, + "learning_rate": 1.069231545593879e-05, + "loss": 0.0335, + "step": 30146 + }, + { + "epoch": 3.95, + "learning_rate": 1.069178640235271e-05, + "loss": 0.0317, + "step": 30147 + }, + { + "epoch": 3.95, + "learning_rate": 1.069125734682102e-05, + "loss": 0.0376, + "step": 30148 + }, + { + "epoch": 3.95, + "learning_rate": 1.0690728289345205e-05, + "loss": 0.031, + "step": 30149 + }, + { + "epoch": 3.95, + "learning_rate": 1.0690199229926757e-05, + "loss": 0.0365, + "step": 30150 + }, + { + "epoch": 3.95, + "learning_rate": 1.0689670168567156e-05, + "loss": 0.0331, + "step": 30151 + }, + { + "epoch": 3.95, + "learning_rate": 1.0689141105267896e-05, + "loss": 0.0284, + "step": 30152 + }, + { + "epoch": 3.95, + "learning_rate": 1.0688612040030461e-05, + "loss": 0.0263, + "step": 30153 + }, + { + "epoch": 3.95, + "learning_rate": 1.0688082972856343e-05, + "loss": 0.0311, + "step": 30154 + }, + { + "epoch": 3.95, + "learning_rate": 1.068755390374703e-05, + "loss": 0.0282, + "step": 30155 + }, + { + "epoch": 3.95, + "learning_rate": 1.0687024832704008e-05, + "loss": 0.0353, + "step": 30156 + }, + { + "epoch": 3.95, + "learning_rate": 1.0686495759728761e-05, + "loss": 0.0291, + "step": 30157 + }, + { + "epoch": 3.95, + "learning_rate": 1.0685966684822786e-05, + "loss": 0.0377, + "step": 30158 + }, + { + "epoch": 3.95, + "learning_rate": 1.0685437607987563e-05, + "loss": 0.0353, + "step": 30159 + }, + { + "epoch": 3.95, + "learning_rate": 1.0684908529224586e-05, + "loss": 0.0339, + "step": 30160 + }, + { + "epoch": 3.95, + "learning_rate": 1.068437944853534e-05, + "loss": 0.0275, + "step": 30161 + }, + { + "epoch": 3.95, + "learning_rate": 1.068385036592131e-05, + "loss": 0.024, + "step": 30162 + }, + { + "epoch": 3.95, + "learning_rate": 1.0683321281383992e-05, + "loss": 0.043, + "step": 30163 + }, + { + "epoch": 3.95, + "learning_rate": 1.0682792194924867e-05, + "loss": 0.0375, + "step": 30164 + }, + { + "epoch": 3.95, + "learning_rate": 1.0682263106545425e-05, + "loss": 0.0347, + "step": 30165 + }, + { + "epoch": 3.95, + "learning_rate": 1.0681734016247157e-05, + "loss": 0.0306, + "step": 30166 + }, + { + "epoch": 3.95, + "learning_rate": 1.068120492403154e-05, + "loss": 0.0361, + "step": 30167 + }, + { + "epoch": 3.95, + "learning_rate": 1.068067582990008e-05, + "loss": 0.0356, + "step": 30168 + }, + { + "epoch": 3.95, + "learning_rate": 1.068014673385425e-05, + "loss": 0.0335, + "step": 30169 + }, + { + "epoch": 3.95, + "learning_rate": 1.0679617635895547e-05, + "loss": 0.0326, + "step": 30170 + }, + { + "epoch": 3.95, + "learning_rate": 1.0679088536025452e-05, + "loss": 0.0323, + "step": 30171 + }, + { + "epoch": 3.95, + "learning_rate": 1.067855943424546e-05, + "loss": 0.0405, + "step": 30172 + }, + { + "epoch": 3.95, + "learning_rate": 1.0678030330557053e-05, + "loss": 0.0392, + "step": 30173 + }, + { + "epoch": 3.95, + "learning_rate": 1.0677501224961722e-05, + "loss": 0.0329, + "step": 30174 + }, + { + "epoch": 3.95, + "learning_rate": 1.0676972117460957e-05, + "loss": 0.0315, + "step": 30175 + }, + { + "epoch": 3.95, + "learning_rate": 1.0676443008056243e-05, + "loss": 0.0335, + "step": 30176 + }, + { + "epoch": 3.95, + "learning_rate": 1.0675913896749071e-05, + "loss": 0.0338, + "step": 30177 + }, + { + "epoch": 3.95, + "learning_rate": 1.067538478354092e-05, + "loss": 0.0339, + "step": 30178 + }, + { + "epoch": 3.95, + "learning_rate": 1.0674855668433293e-05, + "loss": 0.0296, + "step": 30179 + }, + { + "epoch": 3.95, + "learning_rate": 1.0674326551427668e-05, + "loss": 0.0398, + "step": 30180 + }, + { + "epoch": 3.95, + "learning_rate": 1.0673797432525536e-05, + "loss": 0.0358, + "step": 30181 + }, + { + "epoch": 3.95, + "learning_rate": 1.0673268311728385e-05, + "loss": 0.0339, + "step": 30182 + }, + { + "epoch": 3.95, + "learning_rate": 1.0672739189037702e-05, + "loss": 0.038, + "step": 30183 + }, + { + "epoch": 3.95, + "learning_rate": 1.0672210064454974e-05, + "loss": 0.0373, + "step": 30184 + }, + { + "epoch": 3.95, + "learning_rate": 1.0671680937981692e-05, + "loss": 0.0281, + "step": 30185 + }, + { + "epoch": 3.95, + "learning_rate": 1.0671151809619344e-05, + "loss": 0.0379, + "step": 30186 + }, + { + "epoch": 3.95, + "learning_rate": 1.0670622679369417e-05, + "loss": 0.0395, + "step": 30187 + }, + { + "epoch": 3.95, + "learning_rate": 1.06700935472334e-05, + "loss": 0.036, + "step": 30188 + }, + { + "epoch": 3.95, + "learning_rate": 1.0669564413212779e-05, + "loss": 0.0332, + "step": 30189 + }, + { + "epoch": 3.95, + "learning_rate": 1.0669035277309044e-05, + "loss": 0.034, + "step": 30190 + }, + { + "epoch": 3.95, + "learning_rate": 1.0668506139523685e-05, + "loss": 0.0445, + "step": 30191 + }, + { + "epoch": 3.95, + "learning_rate": 1.0667976999858185e-05, + "loss": 0.0368, + "step": 30192 + }, + { + "epoch": 3.96, + "learning_rate": 1.0667447858314038e-05, + "loss": 0.039, + "step": 30193 + }, + { + "epoch": 3.96, + "learning_rate": 1.0666918714892728e-05, + "loss": 0.0311, + "step": 30194 + }, + { + "epoch": 3.96, + "learning_rate": 1.0666389569595745e-05, + "loss": 0.0377, + "step": 30195 + }, + { + "epoch": 3.96, + "learning_rate": 1.0665860422424577e-05, + "loss": 0.0382, + "step": 30196 + }, + { + "epoch": 3.96, + "learning_rate": 1.066533127338071e-05, + "loss": 0.0345, + "step": 30197 + }, + { + "epoch": 3.96, + "learning_rate": 1.0664802122465638e-05, + "loss": 0.0327, + "step": 30198 + }, + { + "epoch": 3.96, + "learning_rate": 1.0664272969680842e-05, + "loss": 0.0268, + "step": 30199 + }, + { + "epoch": 3.96, + "learning_rate": 1.0663743815027816e-05, + "loss": 0.0388, + "step": 30200 + }, + { + "epoch": 3.96, + "learning_rate": 1.0663214658508043e-05, + "loss": 0.0393, + "step": 30201 + }, + { + "epoch": 3.96, + "learning_rate": 1.0662685500123017e-05, + "loss": 0.0302, + "step": 30202 + }, + { + "epoch": 3.96, + "learning_rate": 1.0662156339874224e-05, + "loss": 0.0344, + "step": 30203 + }, + { + "epoch": 3.96, + "learning_rate": 1.0661627177763149e-05, + "loss": 0.0296, + "step": 30204 + }, + { + "epoch": 3.96, + "learning_rate": 1.0661098013791285e-05, + "loss": 0.0328, + "step": 30205 + }, + { + "epoch": 3.96, + "learning_rate": 1.0660568847960116e-05, + "loss": 0.0278, + "step": 30206 + }, + { + "epoch": 3.96, + "learning_rate": 1.066003968027113e-05, + "loss": 0.0307, + "step": 30207 + }, + { + "epoch": 3.96, + "learning_rate": 1.0659510510725823e-05, + "loss": 0.0357, + "step": 30208 + }, + { + "epoch": 3.96, + "learning_rate": 1.0658981339325675e-05, + "loss": 0.0301, + "step": 30209 + }, + { + "epoch": 3.96, + "learning_rate": 1.0658452166072177e-05, + "loss": 0.0346, + "step": 30210 + }, + { + "epoch": 3.96, + "learning_rate": 1.0657922990966818e-05, + "loss": 0.036, + "step": 30211 + }, + { + "epoch": 3.96, + "learning_rate": 1.0657393814011085e-05, + "loss": 0.0319, + "step": 30212 + }, + { + "epoch": 3.96, + "learning_rate": 1.0656864635206468e-05, + "loss": 0.0314, + "step": 30213 + }, + { + "epoch": 3.96, + "learning_rate": 1.0656335454554453e-05, + "loss": 0.0365, + "step": 30214 + }, + { + "epoch": 3.96, + "learning_rate": 1.0655806272056532e-05, + "loss": 0.0335, + "step": 30215 + }, + { + "epoch": 3.96, + "learning_rate": 1.0655277087714188e-05, + "loss": 0.0334, + "step": 30216 + }, + { + "epoch": 3.96, + "learning_rate": 1.0654747901528913e-05, + "loss": 0.0367, + "step": 30217 + }, + { + "epoch": 3.96, + "learning_rate": 1.0654218713502194e-05, + "loss": 0.0369, + "step": 30218 + }, + { + "epoch": 3.96, + "learning_rate": 1.065368952363552e-05, + "loss": 0.0353, + "step": 30219 + }, + { + "epoch": 3.96, + "learning_rate": 1.065316033193038e-05, + "loss": 0.0311, + "step": 30220 + }, + { + "epoch": 3.96, + "learning_rate": 1.065263113838826e-05, + "loss": 0.0299, + "step": 30221 + }, + { + "epoch": 3.96, + "learning_rate": 1.0652101943010647e-05, + "loss": 0.041, + "step": 30222 + }, + { + "epoch": 3.96, + "learning_rate": 1.0651572745799037e-05, + "loss": 0.0297, + "step": 30223 + }, + { + "epoch": 3.96, + "learning_rate": 1.0651043546754912e-05, + "loss": 0.0376, + "step": 30224 + }, + { + "epoch": 3.96, + "learning_rate": 1.065051434587976e-05, + "loss": 0.0396, + "step": 30225 + }, + { + "epoch": 3.96, + "learning_rate": 1.0649985143175072e-05, + "loss": 0.0312, + "step": 30226 + }, + { + "epoch": 3.96, + "learning_rate": 1.0649455938642338e-05, + "loss": 0.0422, + "step": 30227 + }, + { + "epoch": 3.96, + "learning_rate": 1.0648926732283038e-05, + "loss": 0.0318, + "step": 30228 + }, + { + "epoch": 3.96, + "learning_rate": 1.0648397524098673e-05, + "loss": 0.034, + "step": 30229 + }, + { + "epoch": 3.96, + "learning_rate": 1.0647868314090719e-05, + "loss": 0.0349, + "step": 30230 + }, + { + "epoch": 3.96, + "learning_rate": 1.0647339102260674e-05, + "loss": 0.0307, + "step": 30231 + }, + { + "epoch": 3.96, + "learning_rate": 1.064680988861002e-05, + "loss": 0.038, + "step": 30232 + }, + { + "epoch": 3.96, + "learning_rate": 1.0646280673140246e-05, + "loss": 0.038, + "step": 30233 + }, + { + "epoch": 3.96, + "learning_rate": 1.0645751455852842e-05, + "loss": 0.0331, + "step": 30234 + }, + { + "epoch": 3.96, + "learning_rate": 1.0645222236749298e-05, + "loss": 0.039, + "step": 30235 + }, + { + "epoch": 3.96, + "learning_rate": 1.0644693015831102e-05, + "loss": 0.0419, + "step": 30236 + }, + { + "epoch": 3.96, + "learning_rate": 1.064416379309974e-05, + "loss": 0.0279, + "step": 30237 + }, + { + "epoch": 3.96, + "learning_rate": 1.0643634568556702e-05, + "loss": 0.0387, + "step": 30238 + }, + { + "epoch": 3.96, + "learning_rate": 1.0643105342203474e-05, + "loss": 0.0397, + "step": 30239 + }, + { + "epoch": 3.96, + "learning_rate": 1.064257611404155e-05, + "loss": 0.0437, + "step": 30240 + }, + { + "epoch": 3.96, + "learning_rate": 1.0642046884072412e-05, + "loss": 0.0335, + "step": 30241 + }, + { + "epoch": 3.96, + "learning_rate": 1.0641517652297555e-05, + "loss": 0.0343, + "step": 30242 + }, + { + "epoch": 3.96, + "learning_rate": 1.0640988418718461e-05, + "loss": 0.0322, + "step": 30243 + }, + { + "epoch": 3.96, + "learning_rate": 1.064045918333662e-05, + "loss": 0.0344, + "step": 30244 + }, + { + "epoch": 3.96, + "learning_rate": 1.0639929946153523e-05, + "loss": 0.0315, + "step": 30245 + }, + { + "epoch": 3.96, + "learning_rate": 1.0639400707170659e-05, + "loss": 0.0289, + "step": 30246 + }, + { + "epoch": 3.96, + "learning_rate": 1.0638871466389514e-05, + "loss": 0.0395, + "step": 30247 + }, + { + "epoch": 3.96, + "learning_rate": 1.0638342223811576e-05, + "loss": 0.0362, + "step": 30248 + }, + { + "epoch": 3.96, + "learning_rate": 1.0637812979438332e-05, + "loss": 0.0354, + "step": 30249 + }, + { + "epoch": 3.96, + "learning_rate": 1.0637283733271275e-05, + "loss": 0.0356, + "step": 30250 + }, + { + "epoch": 3.96, + "learning_rate": 1.0636754485311891e-05, + "loss": 0.0329, + "step": 30251 + }, + { + "epoch": 3.96, + "learning_rate": 1.0636225235561672e-05, + "loss": 0.0294, + "step": 30252 + }, + { + "epoch": 3.96, + "learning_rate": 1.06356959840221e-05, + "loss": 0.036, + "step": 30253 + }, + { + "epoch": 3.96, + "learning_rate": 1.0635166730694667e-05, + "loss": 0.0315, + "step": 30254 + }, + { + "epoch": 3.96, + "learning_rate": 1.0634637475580862e-05, + "loss": 0.0271, + "step": 30255 + }, + { + "epoch": 3.96, + "learning_rate": 1.0634108218682176e-05, + "loss": 0.0315, + "step": 30256 + }, + { + "epoch": 3.96, + "learning_rate": 1.063357896000009e-05, + "loss": 0.0359, + "step": 30257 + }, + { + "epoch": 3.96, + "learning_rate": 1.0633049699536097e-05, + "loss": 0.0286, + "step": 30258 + }, + { + "epoch": 3.96, + "learning_rate": 1.063252043729169e-05, + "loss": 0.0422, + "step": 30259 + }, + { + "epoch": 3.96, + "learning_rate": 1.0631991173268348e-05, + "loss": 0.0368, + "step": 30260 + }, + { + "epoch": 3.96, + "learning_rate": 1.0631461907467567e-05, + "loss": 0.0342, + "step": 30261 + }, + { + "epoch": 3.96, + "learning_rate": 1.0630932639890833e-05, + "loss": 0.0288, + "step": 30262 + }, + { + "epoch": 3.96, + "learning_rate": 1.0630403370539635e-05, + "loss": 0.0342, + "step": 30263 + }, + { + "epoch": 3.96, + "learning_rate": 1.0629874099415458e-05, + "loss": 0.0301, + "step": 30264 + }, + { + "epoch": 3.96, + "learning_rate": 1.0629344826519799e-05, + "loss": 0.0357, + "step": 30265 + }, + { + "epoch": 3.96, + "learning_rate": 1.0628815551854135e-05, + "loss": 0.034, + "step": 30266 + }, + { + "epoch": 3.96, + "learning_rate": 1.0628286275419966e-05, + "loss": 0.0374, + "step": 30267 + }, + { + "epoch": 3.96, + "learning_rate": 1.062775699721877e-05, + "loss": 0.0264, + "step": 30268 + }, + { + "epoch": 3.97, + "learning_rate": 1.0627227717252046e-05, + "loss": 0.0278, + "step": 30269 + }, + { + "epoch": 3.97, + "learning_rate": 1.0626698435521274e-05, + "loss": 0.0396, + "step": 30270 + }, + { + "epoch": 3.97, + "learning_rate": 1.0626169152027947e-05, + "loss": 0.0295, + "step": 30271 + }, + { + "epoch": 3.97, + "learning_rate": 1.0625639866773554e-05, + "loss": 0.0344, + "step": 30272 + }, + { + "epoch": 3.97, + "learning_rate": 1.0625110579759582e-05, + "loss": 0.0328, + "step": 30273 + }, + { + "epoch": 3.97, + "learning_rate": 1.062458129098752e-05, + "loss": 0.0373, + "step": 30274 + }, + { + "epoch": 3.97, + "learning_rate": 1.0624052000458854e-05, + "loss": 0.0309, + "step": 30275 + }, + { + "epoch": 3.97, + "learning_rate": 1.0623522708175076e-05, + "loss": 0.0326, + "step": 30276 + }, + { + "epoch": 3.97, + "learning_rate": 1.0622993414137675e-05, + "loss": 0.0302, + "step": 30277 + }, + { + "epoch": 3.97, + "learning_rate": 1.0622464118348135e-05, + "loss": 0.0375, + "step": 30278 + }, + { + "epoch": 3.97, + "learning_rate": 1.0621934820807949e-05, + "loss": 0.0343, + "step": 30279 + }, + { + "epoch": 3.97, + "learning_rate": 1.0621405521518606e-05, + "loss": 0.0342, + "step": 30280 + }, + { + "epoch": 3.97, + "learning_rate": 1.0620876220481593e-05, + "loss": 0.0315, + "step": 30281 + }, + { + "epoch": 3.97, + "learning_rate": 1.0620346917698397e-05, + "loss": 0.0313, + "step": 30282 + }, + { + "epoch": 3.97, + "learning_rate": 1.061981761317051e-05, + "loss": 0.0331, + "step": 30283 + }, + { + "epoch": 3.97, + "learning_rate": 1.0619288306899418e-05, + "loss": 0.0315, + "step": 30284 + }, + { + "epoch": 3.97, + "learning_rate": 1.0618758998886611e-05, + "loss": 0.0401, + "step": 30285 + }, + { + "epoch": 3.97, + "learning_rate": 1.0618229689133579e-05, + "loss": 0.0337, + "step": 30286 + }, + { + "epoch": 3.97, + "learning_rate": 1.0617700377641806e-05, + "loss": 0.0319, + "step": 30287 + }, + { + "epoch": 3.97, + "learning_rate": 1.0617171064412786e-05, + "loss": 0.0323, + "step": 30288 + }, + { + "epoch": 3.97, + "learning_rate": 1.0616641749448002e-05, + "loss": 0.0366, + "step": 30289 + }, + { + "epoch": 3.97, + "learning_rate": 1.0616112432748951e-05, + "loss": 0.0313, + "step": 30290 + }, + { + "epoch": 3.97, + "learning_rate": 1.0615583114317112e-05, + "loss": 0.0369, + "step": 30291 + }, + { + "epoch": 3.97, + "learning_rate": 1.061505379415398e-05, + "loss": 0.0339, + "step": 30292 + }, + { + "epoch": 3.97, + "learning_rate": 1.0614524472261038e-05, + "loss": 0.0323, + "step": 30293 + }, + { + "epoch": 3.97, + "learning_rate": 1.0613995148639786e-05, + "loss": 0.0325, + "step": 30294 + }, + { + "epoch": 3.97, + "learning_rate": 1.06134658232917e-05, + "loss": 0.0367, + "step": 30295 + }, + { + "epoch": 3.97, + "learning_rate": 1.0612936496218279e-05, + "loss": 0.0376, + "step": 30296 + }, + { + "epoch": 3.97, + "learning_rate": 1.0612407167421001e-05, + "loss": 0.0275, + "step": 30297 + }, + { + "epoch": 3.97, + "learning_rate": 1.0611877836901364e-05, + "loss": 0.0329, + "step": 30298 + }, + { + "epoch": 3.97, + "learning_rate": 1.061134850466085e-05, + "loss": 0.0358, + "step": 30299 + }, + { + "epoch": 3.97, + "learning_rate": 1.0610819170700953e-05, + "loss": 0.0328, + "step": 30300 + }, + { + "epoch": 3.97, + "learning_rate": 1.0610289835023162e-05, + "loss": 0.0349, + "step": 30301 + }, + { + "epoch": 3.97, + "learning_rate": 1.060976049762896e-05, + "loss": 0.0325, + "step": 30302 + }, + { + "epoch": 3.97, + "learning_rate": 1.060923115851984e-05, + "loss": 0.0328, + "step": 30303 + }, + { + "epoch": 3.97, + "learning_rate": 1.060870181769729e-05, + "loss": 0.0377, + "step": 30304 + }, + { + "epoch": 3.97, + "learning_rate": 1.06081724751628e-05, + "loss": 0.0378, + "step": 30305 + }, + { + "epoch": 3.97, + "learning_rate": 1.0607643130917854e-05, + "loss": 0.0383, + "step": 30306 + }, + { + "epoch": 3.97, + "learning_rate": 1.0607113784963948e-05, + "loss": 0.0335, + "step": 30307 + }, + { + "epoch": 3.97, + "learning_rate": 1.0606584437302562e-05, + "loss": 0.0398, + "step": 30308 + }, + { + "epoch": 3.97, + "learning_rate": 1.0606055087935192e-05, + "loss": 0.0268, + "step": 30309 + }, + { + "epoch": 3.97, + "learning_rate": 1.0605525736863324e-05, + "loss": 0.0341, + "step": 30310 + }, + { + "epoch": 3.97, + "learning_rate": 1.0604996384088449e-05, + "loss": 0.0364, + "step": 30311 + }, + { + "epoch": 3.97, + "learning_rate": 1.0604467029612051e-05, + "loss": 0.0402, + "step": 30312 + }, + { + "epoch": 3.97, + "learning_rate": 1.0603937673435625e-05, + "loss": 0.0351, + "step": 30313 + }, + { + "epoch": 3.97, + "learning_rate": 1.0603408315560655e-05, + "loss": 0.0368, + "step": 30314 + }, + { + "epoch": 3.97, + "learning_rate": 1.060287895598863e-05, + "loss": 0.0344, + "step": 30315 + }, + { + "epoch": 3.97, + "learning_rate": 1.060234959472104e-05, + "loss": 0.03, + "step": 30316 + }, + { + "epoch": 3.97, + "learning_rate": 1.0601820231759377e-05, + "loss": 0.0263, + "step": 30317 + }, + { + "epoch": 3.97, + "learning_rate": 1.0601290867105123e-05, + "loss": 0.0342, + "step": 30318 + }, + { + "epoch": 3.97, + "learning_rate": 1.0600761500759772e-05, + "loss": 0.0363, + "step": 30319 + }, + { + "epoch": 3.97, + "learning_rate": 1.0600232132724811e-05, + "loss": 0.0384, + "step": 30320 + }, + { + "epoch": 3.97, + "learning_rate": 1.0599702763001729e-05, + "loss": 0.0383, + "step": 30321 + }, + { + "epoch": 3.97, + "learning_rate": 1.0599173391592015e-05, + "loss": 0.0324, + "step": 30322 + }, + { + "epoch": 3.97, + "learning_rate": 1.0598644018497157e-05, + "loss": 0.0374, + "step": 30323 + }, + { + "epoch": 3.97, + "learning_rate": 1.0598114643718647e-05, + "loss": 0.0368, + "step": 30324 + }, + { + "epoch": 3.97, + "learning_rate": 1.0597585267257969e-05, + "loss": 0.0313, + "step": 30325 + }, + { + "epoch": 3.97, + "learning_rate": 1.0597055889116615e-05, + "loss": 0.0315, + "step": 30326 + }, + { + "epoch": 3.97, + "learning_rate": 1.0596526509296071e-05, + "loss": 0.0348, + "step": 30327 + }, + { + "epoch": 3.97, + "learning_rate": 1.0595997127797832e-05, + "loss": 0.0326, + "step": 30328 + }, + { + "epoch": 3.97, + "learning_rate": 1.0595467744623383e-05, + "loss": 0.0326, + "step": 30329 + }, + { + "epoch": 3.97, + "learning_rate": 1.059493835977421e-05, + "loss": 0.0317, + "step": 30330 + }, + { + "epoch": 3.97, + "learning_rate": 1.0594408973251803e-05, + "loss": 0.0291, + "step": 30331 + }, + { + "epoch": 3.97, + "learning_rate": 1.0593879585057655e-05, + "loss": 0.0371, + "step": 30332 + }, + { + "epoch": 3.97, + "learning_rate": 1.0593350195193253e-05, + "loss": 0.0299, + "step": 30333 + }, + { + "epoch": 3.97, + "learning_rate": 1.0592820803660084e-05, + "loss": 0.035, + "step": 30334 + }, + { + "epoch": 3.97, + "learning_rate": 1.059229141045964e-05, + "loss": 0.0311, + "step": 30335 + }, + { + "epoch": 3.97, + "learning_rate": 1.0591762015593406e-05, + "loss": 0.0356, + "step": 30336 + }, + { + "epoch": 3.97, + "learning_rate": 1.059123261906287e-05, + "loss": 0.0402, + "step": 30337 + }, + { + "epoch": 3.97, + "learning_rate": 1.059070322086953e-05, + "loss": 0.0331, + "step": 30338 + }, + { + "epoch": 3.97, + "learning_rate": 1.0590173821014863e-05, + "loss": 0.0303, + "step": 30339 + }, + { + "epoch": 3.97, + "learning_rate": 1.0589644419500367e-05, + "loss": 0.0352, + "step": 30340 + }, + { + "epoch": 3.97, + "learning_rate": 1.0589115016327527e-05, + "loss": 0.0367, + "step": 30341 + }, + { + "epoch": 3.97, + "learning_rate": 1.0588585611497831e-05, + "loss": 0.0353, + "step": 30342 + }, + { + "epoch": 3.97, + "learning_rate": 1.0588056205012769e-05, + "loss": 0.0295, + "step": 30343 + }, + { + "epoch": 3.97, + "learning_rate": 1.058752679687383e-05, + "loss": 0.0351, + "step": 30344 + }, + { + "epoch": 3.97, + "learning_rate": 1.0586997387082507e-05, + "loss": 0.032, + "step": 30345 + }, + { + "epoch": 3.98, + "learning_rate": 1.0586467975640283e-05, + "loss": 0.0338, + "step": 30346 + }, + { + "epoch": 3.98, + "learning_rate": 1.058593856254865e-05, + "loss": 0.0335, + "step": 30347 + }, + { + "epoch": 3.98, + "learning_rate": 1.058540914780909e-05, + "loss": 0.0305, + "step": 30348 + }, + { + "epoch": 3.98, + "learning_rate": 1.0584879731423106e-05, + "loss": 0.0373, + "step": 30349 + }, + { + "epoch": 3.98, + "learning_rate": 1.0584350313392173e-05, + "loss": 0.0373, + "step": 30350 + }, + { + "epoch": 3.98, + "learning_rate": 1.0583820893717791e-05, + "loss": 0.0265, + "step": 30351 + }, + { + "epoch": 3.98, + "learning_rate": 1.058329147240144e-05, + "loss": 0.0365, + "step": 30352 + }, + { + "epoch": 3.98, + "learning_rate": 1.0582762049444612e-05, + "loss": 0.0424, + "step": 30353 + }, + { + "epoch": 3.98, + "learning_rate": 1.05822326248488e-05, + "loss": 0.0395, + "step": 30354 + }, + { + "epoch": 3.98, + "learning_rate": 1.058170319861549e-05, + "loss": 0.0327, + "step": 30355 + }, + { + "epoch": 3.98, + "learning_rate": 1.0581173770746169e-05, + "loss": 0.0392, + "step": 30356 + }, + { + "epoch": 3.98, + "learning_rate": 1.0580644341242329e-05, + "loss": 0.0278, + "step": 30357 + }, + { + "epoch": 3.98, + "learning_rate": 1.0580114910105457e-05, + "loss": 0.0336, + "step": 30358 + }, + { + "epoch": 3.98, + "learning_rate": 1.057958547733704e-05, + "loss": 0.0341, + "step": 30359 + }, + { + "epoch": 3.98, + "learning_rate": 1.0579056042938571e-05, + "loss": 0.0335, + "step": 30360 + }, + { + "epoch": 3.98, + "learning_rate": 1.0578526606911541e-05, + "loss": 0.0361, + "step": 30361 + }, + { + "epoch": 3.98, + "learning_rate": 1.0577997169257433e-05, + "loss": 0.0316, + "step": 30362 + }, + { + "epoch": 3.98, + "learning_rate": 1.0577467729977739e-05, + "loss": 0.0281, + "step": 30363 + }, + { + "epoch": 3.98, + "learning_rate": 1.0576938289073948e-05, + "loss": 0.0372, + "step": 30364 + }, + { + "epoch": 3.98, + "learning_rate": 1.0576408846547548e-05, + "loss": 0.0343, + "step": 30365 + }, + { + "epoch": 3.98, + "learning_rate": 1.0575879402400031e-05, + "loss": 0.0365, + "step": 30366 + }, + { + "epoch": 3.98, + "learning_rate": 1.057534995663288e-05, + "loss": 0.0308, + "step": 30367 + }, + { + "epoch": 3.98, + "learning_rate": 1.0574820509247594e-05, + "loss": 0.0385, + "step": 30368 + }, + { + "epoch": 3.98, + "learning_rate": 1.0574291060245649e-05, + "loss": 0.0321, + "step": 30369 + }, + { + "epoch": 3.98, + "learning_rate": 1.0573761609628545e-05, + "loss": 0.0303, + "step": 30370 + }, + { + "epoch": 3.98, + "learning_rate": 1.0573232157397766e-05, + "loss": 0.0351, + "step": 30371 + }, + { + "epoch": 3.98, + "learning_rate": 1.0572702703554805e-05, + "loss": 0.0327, + "step": 30372 + }, + { + "epoch": 3.98, + "learning_rate": 1.0572173248101142e-05, + "loss": 0.033, + "step": 30373 + }, + { + "epoch": 3.98, + "learning_rate": 1.0571643791038279e-05, + "loss": 0.0316, + "step": 30374 + }, + { + "epoch": 3.98, + "learning_rate": 1.0571114332367691e-05, + "loss": 0.034, + "step": 30375 + }, + { + "epoch": 3.98, + "learning_rate": 1.0570584872090882e-05, + "loss": 0.0292, + "step": 30376 + }, + { + "epoch": 3.98, + "learning_rate": 1.0570055410209328e-05, + "loss": 0.0342, + "step": 30377 + }, + { + "epoch": 3.98, + "learning_rate": 1.0569525946724528e-05, + "loss": 0.027, + "step": 30378 + }, + { + "epoch": 3.98, + "learning_rate": 1.0568996481637963e-05, + "loss": 0.0303, + "step": 30379 + }, + { + "epoch": 3.98, + "learning_rate": 1.0568467014951125e-05, + "loss": 0.0363, + "step": 30380 + }, + { + "epoch": 3.98, + "learning_rate": 1.0567937546665505e-05, + "loss": 0.0336, + "step": 30381 + }, + { + "epoch": 3.98, + "learning_rate": 1.0567408076782593e-05, + "loss": 0.04, + "step": 30382 + }, + { + "epoch": 3.98, + "learning_rate": 1.0566878605303875e-05, + "loss": 0.0333, + "step": 30383 + }, + { + "epoch": 3.98, + "learning_rate": 1.0566349132230838e-05, + "loss": 0.0338, + "step": 30384 + }, + { + "epoch": 3.98, + "learning_rate": 1.056581965756498e-05, + "loss": 0.0334, + "step": 30385 + }, + { + "epoch": 3.98, + "learning_rate": 1.056529018130778e-05, + "loss": 0.0291, + "step": 30386 + }, + { + "epoch": 3.98, + "learning_rate": 1.056476070346073e-05, + "loss": 0.0293, + "step": 30387 + }, + { + "epoch": 3.98, + "learning_rate": 1.0564231224025323e-05, + "loss": 0.0298, + "step": 30388 + }, + { + "epoch": 3.98, + "learning_rate": 1.0563701743003047e-05, + "loss": 0.0348, + "step": 30389 + }, + { + "epoch": 3.98, + "learning_rate": 1.056317226039539e-05, + "loss": 0.0327, + "step": 30390 + }, + { + "epoch": 3.98, + "learning_rate": 1.0562642776203838e-05, + "loss": 0.0393, + "step": 30391 + }, + { + "epoch": 3.98, + "learning_rate": 1.0562113290429883e-05, + "loss": 0.0335, + "step": 30392 + }, + { + "epoch": 3.98, + "learning_rate": 1.0561583803075018e-05, + "loss": 0.0302, + "step": 30393 + }, + { + "epoch": 3.98, + "learning_rate": 1.0561054314140726e-05, + "loss": 0.0309, + "step": 30394 + }, + { + "epoch": 3.98, + "learning_rate": 1.0560524823628503e-05, + "loss": 0.0335, + "step": 30395 + }, + { + "epoch": 3.98, + "learning_rate": 1.0559995331539827e-05, + "loss": 0.0283, + "step": 30396 + }, + { + "epoch": 3.98, + "learning_rate": 1.0559465837876198e-05, + "loss": 0.0343, + "step": 30397 + }, + { + "epoch": 3.98, + "learning_rate": 1.05589363426391e-05, + "loss": 0.0323, + "step": 30398 + }, + { + "epoch": 3.98, + "learning_rate": 1.0558406845830023e-05, + "loss": 0.0327, + "step": 30399 + }, + { + "epoch": 3.98, + "learning_rate": 1.0557877347450457e-05, + "loss": 0.0305, + "step": 30400 + }, + { + "epoch": 3.98, + "learning_rate": 1.0557347847501888e-05, + "loss": 0.0423, + "step": 30401 + }, + { + "epoch": 3.98, + "learning_rate": 1.0556818345985811e-05, + "loss": 0.0373, + "step": 30402 + }, + { + "epoch": 3.98, + "learning_rate": 1.0556288842903712e-05, + "loss": 0.0343, + "step": 30403 + }, + { + "epoch": 3.98, + "learning_rate": 1.0555759338257079e-05, + "loss": 0.0344, + "step": 30404 + }, + { + "epoch": 3.98, + "learning_rate": 1.0555229832047402e-05, + "loss": 0.0332, + "step": 30405 + }, + { + "epoch": 3.98, + "learning_rate": 1.0554700324276175e-05, + "loss": 0.0377, + "step": 30406 + }, + { + "epoch": 3.98, + "learning_rate": 1.0554170814944877e-05, + "loss": 0.0298, + "step": 30407 + }, + { + "epoch": 3.98, + "learning_rate": 1.0553641304055006e-05, + "loss": 0.0352, + "step": 30408 + }, + { + "epoch": 3.98, + "learning_rate": 1.0553111791608048e-05, + "loss": 0.0346, + "step": 30409 + }, + { + "epoch": 3.98, + "learning_rate": 1.0552582277605494e-05, + "loss": 0.0298, + "step": 30410 + }, + { + "epoch": 3.98, + "learning_rate": 1.055205276204883e-05, + "loss": 0.0388, + "step": 30411 + }, + { + "epoch": 3.98, + "learning_rate": 1.055152324493955e-05, + "loss": 0.0278, + "step": 30412 + }, + { + "epoch": 3.98, + "learning_rate": 1.0550993726279133e-05, + "loss": 0.0337, + "step": 30413 + }, + { + "epoch": 3.98, + "learning_rate": 1.0550464206069082e-05, + "loss": 0.0302, + "step": 30414 + }, + { + "epoch": 3.98, + "learning_rate": 1.054993468431088e-05, + "loss": 0.0321, + "step": 30415 + }, + { + "epoch": 3.98, + "learning_rate": 1.0549405161006013e-05, + "loss": 0.0309, + "step": 30416 + }, + { + "epoch": 3.98, + "learning_rate": 1.0548875636155974e-05, + "loss": 0.0352, + "step": 30417 + }, + { + "epoch": 3.98, + "learning_rate": 1.0548346109762252e-05, + "loss": 0.031, + "step": 30418 + }, + { + "epoch": 3.98, + "learning_rate": 1.0547816581826333e-05, + "loss": 0.0305, + "step": 30419 + }, + { + "epoch": 3.98, + "learning_rate": 1.0547287052349713e-05, + "loss": 0.041, + "step": 30420 + }, + { + "epoch": 3.98, + "learning_rate": 1.0546757521333874e-05, + "loss": 0.0395, + "step": 30421 + }, + { + "epoch": 3.99, + "learning_rate": 1.0546227988780312e-05, + "loss": 0.0358, + "step": 30422 + }, + { + "epoch": 3.99, + "learning_rate": 1.0545698454690512e-05, + "loss": 0.0376, + "step": 30423 + }, + { + "epoch": 3.99, + "learning_rate": 1.0545168919065962e-05, + "loss": 0.0358, + "step": 30424 + }, + { + "epoch": 3.99, + "learning_rate": 1.0544639381908154e-05, + "loss": 0.0276, + "step": 30425 + }, + { + "epoch": 3.99, + "learning_rate": 1.0544109843218577e-05, + "loss": 0.0344, + "step": 30426 + }, + { + "epoch": 3.99, + "learning_rate": 1.0543580302998722e-05, + "loss": 0.0389, + "step": 30427 + }, + { + "epoch": 3.99, + "learning_rate": 1.0543050761250076e-05, + "loss": 0.0316, + "step": 30428 + }, + { + "epoch": 3.99, + "learning_rate": 1.0542521217974124e-05, + "loss": 0.0357, + "step": 30429 + }, + { + "epoch": 3.99, + "learning_rate": 1.0541991673172364e-05, + "loss": 0.0368, + "step": 30430 + }, + { + "epoch": 3.99, + "learning_rate": 1.0541462126846282e-05, + "loss": 0.0268, + "step": 30431 + }, + { + "epoch": 3.99, + "learning_rate": 1.0540932578997366e-05, + "loss": 0.0344, + "step": 30432 + }, + { + "epoch": 3.99, + "learning_rate": 1.0540403029627107e-05, + "loss": 0.0318, + "step": 30433 + }, + { + "epoch": 3.99, + "learning_rate": 1.0539873478736991e-05, + "loss": 0.0404, + "step": 30434 + }, + { + "epoch": 3.99, + "learning_rate": 1.053934392632851e-05, + "loss": 0.0296, + "step": 30435 + }, + { + "epoch": 3.99, + "learning_rate": 1.053881437240315e-05, + "loss": 0.0332, + "step": 30436 + }, + { + "epoch": 3.99, + "learning_rate": 1.053828481696241e-05, + "loss": 0.0355, + "step": 30437 + }, + { + "epoch": 3.99, + "learning_rate": 1.0537755260007767e-05, + "loss": 0.029, + "step": 30438 + }, + { + "epoch": 3.99, + "learning_rate": 1.053722570154072e-05, + "loss": 0.0298, + "step": 30439 + }, + { + "epoch": 3.99, + "learning_rate": 1.053669614156275e-05, + "loss": 0.0318, + "step": 30440 + }, + { + "epoch": 3.99, + "learning_rate": 1.0536166580075353e-05, + "loss": 0.0304, + "step": 30441 + }, + { + "epoch": 3.99, + "learning_rate": 1.0535637017080015e-05, + "loss": 0.0393, + "step": 30442 + }, + { + "epoch": 3.99, + "learning_rate": 1.0535107452578231e-05, + "loss": 0.039, + "step": 30443 + }, + { + "epoch": 3.99, + "learning_rate": 1.0534577886571479e-05, + "loss": 0.0377, + "step": 30444 + }, + { + "epoch": 3.99, + "learning_rate": 1.0534048319061259e-05, + "loss": 0.035, + "step": 30445 + }, + { + "epoch": 3.99, + "learning_rate": 1.0533518750049056e-05, + "loss": 0.0341, + "step": 30446 + }, + { + "epoch": 3.99, + "learning_rate": 1.0532989179536362e-05, + "loss": 0.0319, + "step": 30447 + }, + { + "epoch": 3.99, + "learning_rate": 1.0532459607524662e-05, + "loss": 0.033, + "step": 30448 + }, + { + "epoch": 3.99, + "learning_rate": 1.0531930034015446e-05, + "loss": 0.0387, + "step": 30449 + }, + { + "epoch": 3.99, + "learning_rate": 1.0531400459010207e-05, + "loss": 0.0336, + "step": 30450 + }, + { + "epoch": 3.99, + "learning_rate": 1.0530870882510432e-05, + "loss": 0.0293, + "step": 30451 + }, + { + "epoch": 3.99, + "learning_rate": 1.0530341304517613e-05, + "loss": 0.0385, + "step": 30452 + }, + { + "epoch": 3.99, + "learning_rate": 1.0529811725033234e-05, + "loss": 0.0328, + "step": 30453 + }, + { + "epoch": 3.99, + "learning_rate": 1.052928214405879e-05, + "loss": 0.0368, + "step": 30454 + }, + { + "epoch": 3.99, + "learning_rate": 1.0528752561595768e-05, + "loss": 0.038, + "step": 30455 + }, + { + "epoch": 3.99, + "learning_rate": 1.0528222977645659e-05, + "loss": 0.0335, + "step": 30456 + }, + { + "epoch": 3.99, + "learning_rate": 1.0527693392209946e-05, + "loss": 0.0315, + "step": 30457 + }, + { + "epoch": 3.99, + "learning_rate": 1.0527163805290129e-05, + "loss": 0.0247, + "step": 30458 + }, + { + "epoch": 3.99, + "learning_rate": 1.0526634216887689e-05, + "loss": 0.0304, + "step": 30459 + }, + { + "epoch": 3.99, + "learning_rate": 1.0526104627004118e-05, + "loss": 0.0303, + "step": 30460 + }, + { + "epoch": 3.99, + "learning_rate": 1.0525575035640906e-05, + "loss": 0.0393, + "step": 30461 + }, + { + "epoch": 3.99, + "learning_rate": 1.0525045442799544e-05, + "loss": 0.0249, + "step": 30462 + }, + { + "epoch": 3.99, + "learning_rate": 1.0524515848481518e-05, + "loss": 0.0398, + "step": 30463 + }, + { + "epoch": 3.99, + "learning_rate": 1.052398625268832e-05, + "loss": 0.0339, + "step": 30464 + }, + { + "epoch": 3.99, + "learning_rate": 1.0523456655421438e-05, + "loss": 0.0324, + "step": 30465 + }, + { + "epoch": 3.99, + "learning_rate": 1.0522927056682362e-05, + "loss": 0.0357, + "step": 30466 + }, + { + "epoch": 3.99, + "learning_rate": 1.0522397456472582e-05, + "loss": 0.0377, + "step": 30467 + }, + { + "epoch": 3.99, + "learning_rate": 1.0521867854793587e-05, + "loss": 0.0394, + "step": 30468 + }, + { + "epoch": 3.99, + "learning_rate": 1.0521338251646864e-05, + "loss": 0.0306, + "step": 30469 + }, + { + "epoch": 3.99, + "learning_rate": 1.0520808647033907e-05, + "loss": 0.027, + "step": 30470 + }, + { + "epoch": 3.99, + "learning_rate": 1.0520279040956205e-05, + "loss": 0.0355, + "step": 30471 + }, + { + "epoch": 3.99, + "learning_rate": 1.0519749433415242e-05, + "loss": 0.0262, + "step": 30472 + }, + { + "epoch": 3.99, + "learning_rate": 1.0519219824412515e-05, + "loss": 0.0347, + "step": 30473 + }, + { + "epoch": 3.99, + "learning_rate": 1.0518690213949507e-05, + "loss": 0.0299, + "step": 30474 + }, + { + "epoch": 3.99, + "learning_rate": 1.0518160602027714e-05, + "loss": 0.0391, + "step": 30475 + }, + { + "epoch": 3.99, + "learning_rate": 1.0517630988648619e-05, + "loss": 0.0352, + "step": 30476 + }, + { + "epoch": 3.99, + "learning_rate": 1.0517101373813716e-05, + "loss": 0.0387, + "step": 30477 + }, + { + "epoch": 3.99, + "learning_rate": 1.0516571757524491e-05, + "loss": 0.0327, + "step": 30478 + }, + { + "epoch": 3.99, + "learning_rate": 1.051604213978244e-05, + "loss": 0.0296, + "step": 30479 + }, + { + "epoch": 3.99, + "learning_rate": 1.0515512520589042e-05, + "loss": 0.0359, + "step": 30480 + }, + { + "epoch": 3.99, + "learning_rate": 1.0514982899945795e-05, + "loss": 0.0353, + "step": 30481 + }, + { + "epoch": 3.99, + "learning_rate": 1.0514453277854187e-05, + "loss": 0.0266, + "step": 30482 + }, + { + "epoch": 3.99, + "learning_rate": 1.0513923654315707e-05, + "loss": 0.0352, + "step": 30483 + }, + { + "epoch": 3.99, + "learning_rate": 1.051339402933184e-05, + "loss": 0.0325, + "step": 30484 + }, + { + "epoch": 3.99, + "learning_rate": 1.0512864402904083e-05, + "loss": 0.037, + "step": 30485 + }, + { + "epoch": 3.99, + "learning_rate": 1.0512334775033921e-05, + "loss": 0.0309, + "step": 30486 + }, + { + "epoch": 3.99, + "learning_rate": 1.0511805145722849e-05, + "loss": 0.0369, + "step": 30487 + }, + { + "epoch": 3.99, + "learning_rate": 1.0511275514972348e-05, + "loss": 0.0389, + "step": 30488 + }, + { + "epoch": 3.99, + "learning_rate": 1.051074588278391e-05, + "loss": 0.0306, + "step": 30489 + }, + { + "epoch": 3.99, + "learning_rate": 1.051021624915903e-05, + "loss": 0.0368, + "step": 30490 + }, + { + "epoch": 3.99, + "learning_rate": 1.050968661409919e-05, + "loss": 0.0377, + "step": 30491 + }, + { + "epoch": 3.99, + "learning_rate": 1.0509156977605887e-05, + "loss": 0.0258, + "step": 30492 + }, + { + "epoch": 3.99, + "learning_rate": 1.0508627339680609e-05, + "loss": 0.0346, + "step": 30493 + }, + { + "epoch": 3.99, + "learning_rate": 1.0508097700324842e-05, + "loss": 0.0389, + "step": 30494 + }, + { + "epoch": 3.99, + "learning_rate": 1.0507568059540072e-05, + "loss": 0.0301, + "step": 30495 + }, + { + "epoch": 3.99, + "learning_rate": 1.0507038417327802e-05, + "loss": 0.0373, + "step": 30496 + }, + { + "epoch": 3.99, + "learning_rate": 1.0506508773689508e-05, + "loss": 0.0299, + "step": 30497 + }, + { + "epoch": 4.0, + "learning_rate": 1.050597912862669e-05, + "loss": 0.0327, + "step": 30498 + }, + { + "epoch": 4.0, + "learning_rate": 1.0505449482140825e-05, + "loss": 0.0318, + "step": 30499 + }, + { + "epoch": 4.0, + "learning_rate": 1.0504919834233416e-05, + "loss": 0.0295, + "step": 30500 + }, + { + "epoch": 4.0, + "learning_rate": 1.0504390184905946e-05, + "loss": 0.0314, + "step": 30501 + }, + { + "epoch": 4.0, + "learning_rate": 1.0503860534159907e-05, + "loss": 0.0359, + "step": 30502 + }, + { + "epoch": 4.0, + "learning_rate": 1.0503330881996782e-05, + "loss": 0.0402, + "step": 30503 + }, + { + "epoch": 4.0, + "learning_rate": 1.0502801228418072e-05, + "loss": 0.0326, + "step": 30504 + }, + { + "epoch": 4.0, + "learning_rate": 1.0502271573425256e-05, + "loss": 0.0446, + "step": 30505 + }, + { + "epoch": 4.0, + "learning_rate": 1.0501741917019827e-05, + "loss": 0.0333, + "step": 30506 + }, + { + "epoch": 4.0, + "learning_rate": 1.0501212259203278e-05, + "loss": 0.0318, + "step": 30507 + }, + { + "epoch": 4.0, + "learning_rate": 1.0500682599977096e-05, + "loss": 0.0335, + "step": 30508 + }, + { + "epoch": 4.0, + "learning_rate": 1.0500152939342771e-05, + "loss": 0.0304, + "step": 30509 + }, + { + "epoch": 4.0, + "learning_rate": 1.0499623277301792e-05, + "loss": 0.0352, + "step": 30510 + }, + { + "epoch": 4.0, + "learning_rate": 1.049909361385565e-05, + "loss": 0.0294, + "step": 30511 + }, + { + "epoch": 4.0, + "learning_rate": 1.0498563949005833e-05, + "loss": 0.0295, + "step": 30512 + }, + { + "epoch": 4.0, + "learning_rate": 1.0498034282753831e-05, + "loss": 0.0349, + "step": 30513 + }, + { + "epoch": 4.0, + "learning_rate": 1.0497504615101135e-05, + "loss": 0.0294, + "step": 30514 + }, + { + "epoch": 4.0, + "learning_rate": 1.0496974946049234e-05, + "loss": 0.0325, + "step": 30515 + }, + { + "epoch": 4.0, + "learning_rate": 1.0496445275599616e-05, + "loss": 0.0333, + "step": 30516 + }, + { + "epoch": 4.0, + "learning_rate": 1.0495915603753772e-05, + "loss": 0.0347, + "step": 30517 + }, + { + "epoch": 4.0, + "learning_rate": 1.0495385930513193e-05, + "loss": 0.034, + "step": 30518 + }, + { + "epoch": 4.0, + "learning_rate": 1.0494856255879368e-05, + "loss": 0.0327, + "step": 30519 + }, + { + "epoch": 4.0, + "learning_rate": 1.0494326579853787e-05, + "loss": 0.0345, + "step": 30520 + }, + { + "epoch": 4.0, + "learning_rate": 1.0493796902437937e-05, + "loss": 0.0317, + "step": 30521 + }, + { + "epoch": 4.0, + "learning_rate": 1.0493267223633308e-05, + "loss": 0.031, + "step": 30522 + }, + { + "epoch": 4.0, + "learning_rate": 1.0492737543441394e-05, + "loss": 0.03, + "step": 30523 + }, + { + "epoch": 4.0, + "learning_rate": 1.049220786186368e-05, + "loss": 0.0429, + "step": 30524 + }, + { + "epoch": 4.0, + "learning_rate": 1.0491678178901659e-05, + "loss": 0.0322, + "step": 30525 + }, + { + "epoch": 4.0, + "learning_rate": 1.0491148494556817e-05, + "loss": 0.032, + "step": 30526 + }, + { + "epoch": 4.0, + "learning_rate": 1.049061880883065e-05, + "loss": 0.0342, + "step": 30527 + }, + { + "epoch": 4.0, + "learning_rate": 1.0490089121724638e-05, + "loss": 0.0307, + "step": 30528 + }, + { + "epoch": 4.0, + "learning_rate": 1.0489559433240281e-05, + "loss": 0.0327, + "step": 30529 + }, + { + "epoch": 4.0, + "learning_rate": 1.0489029743379064e-05, + "loss": 0.0353, + "step": 30530 + }, + { + "epoch": 4.0, + "learning_rate": 1.0488500052142474e-05, + "loss": 0.0365, + "step": 30531 + }, + { + "epoch": 4.0, + "learning_rate": 1.0487970359532008e-05, + "loss": 0.0308, + "step": 30532 + }, + { + "epoch": 4.0, + "learning_rate": 1.0487440665549149e-05, + "loss": 0.0352, + "step": 30533 + }, + { + "epoch": 4.0, + "learning_rate": 1.0486910970195387e-05, + "loss": 0.0319, + "step": 30534 + }, + { + "epoch": 4.0, + "learning_rate": 1.0486381273472215e-05, + "loss": 0.0367, + "step": 30535 + }, + { + "epoch": 4.0, + "learning_rate": 1.0485851575381126e-05, + "loss": 0.0265, + "step": 30536 + }, + { + "epoch": 4.0, + "learning_rate": 1.04853218759236e-05, + "loss": 0.0243, + "step": 30537 + }, + { + "epoch": 4.0, + "learning_rate": 1.0484792175101136e-05, + "loss": 0.0229, + "step": 30538 + }, + { + "epoch": 4.0, + "learning_rate": 1.0484262472915215e-05, + "loss": 0.0265, + "step": 30539 + }, + { + "epoch": 4.0, + "learning_rate": 1.0483732769367334e-05, + "loss": 0.0227, + "step": 30540 + }, + { + "epoch": 4.0, + "learning_rate": 1.048320306445898e-05, + "loss": 0.0289, + "step": 30541 + }, + { + "epoch": 4.0, + "learning_rate": 1.0482673358191644e-05, + "loss": 0.0266, + "step": 30542 + }, + { + "epoch": 4.0, + "learning_rate": 1.0482143650566815e-05, + "loss": 0.0244, + "step": 30543 + }, + { + "epoch": 4.0, + "learning_rate": 1.0481613941585982e-05, + "loss": 0.0263, + "step": 30544 + }, + { + "epoch": 4.0, + "learning_rate": 1.0481084231250633e-05, + "loss": 0.0235, + "step": 30545 + }, + { + "epoch": 4.0, + "learning_rate": 1.0480554519562265e-05, + "loss": 0.0276, + "step": 30546 + }, + { + "epoch": 4.0, + "learning_rate": 1.048002480652236e-05, + "loss": 0.0269, + "step": 30547 + }, + { + "epoch": 4.0, + "learning_rate": 1.0479495092132414e-05, + "loss": 0.0257, + "step": 30548 + }, + { + "epoch": 4.0, + "learning_rate": 1.047896537639391e-05, + "loss": 0.024, + "step": 30549 + }, + { + "epoch": 4.0, + "learning_rate": 1.0478435659308342e-05, + "loss": 0.0233, + "step": 30550 + }, + { + "epoch": 4.0, + "learning_rate": 1.04779059408772e-05, + "loss": 0.0278, + "step": 30551 + }, + { + "epoch": 4.0, + "learning_rate": 1.0477376221101974e-05, + "loss": 0.0197, + "step": 30552 + }, + { + "epoch": 4.0, + "learning_rate": 1.047684649998415e-05, + "loss": 0.0267, + "step": 30553 + }, + { + "epoch": 4.0, + "learning_rate": 1.0476316777525222e-05, + "loss": 0.0253, + "step": 30554 + }, + { + "epoch": 4.0, + "learning_rate": 1.0475787053726678e-05, + "loss": 0.0219, + "step": 30555 + }, + { + "epoch": 4.0, + "learning_rate": 1.0475257328590006e-05, + "loss": 0.0233, + "step": 30556 + }, + { + "epoch": 4.0, + "learning_rate": 1.0474727602116705e-05, + "loss": 0.0213, + "step": 30557 + }, + { + "epoch": 4.0, + "learning_rate": 1.0474197874308252e-05, + "loss": 0.0222, + "step": 30558 + }, + { + "epoch": 4.0, + "learning_rate": 1.0473668145166147e-05, + "loss": 0.0218, + "step": 30559 + }, + { + "epoch": 4.0, + "learning_rate": 1.047313841469187e-05, + "loss": 0.0251, + "step": 30560 + }, + { + "epoch": 4.0, + "learning_rate": 1.047260868288692e-05, + "loss": 0.0235, + "step": 30561 + }, + { + "epoch": 4.0, + "learning_rate": 1.0472078949752784e-05, + "loss": 0.0241, + "step": 30562 + }, + { + "epoch": 4.0, + "learning_rate": 1.047154921529095e-05, + "loss": 0.0248, + "step": 30563 + }, + { + "epoch": 4.0, + "learning_rate": 1.0471019479502909e-05, + "loss": 0.0257, + "step": 30564 + }, + { + "epoch": 4.0, + "learning_rate": 1.047048974239015e-05, + "loss": 0.0234, + "step": 30565 + }, + { + "epoch": 4.0, + "learning_rate": 1.0469960003954162e-05, + "loss": 0.0238, + "step": 30566 + }, + { + "epoch": 4.0, + "learning_rate": 1.0469430264196441e-05, + "loss": 0.0258, + "step": 30567 + }, + { + "epoch": 4.0, + "learning_rate": 1.046890052311847e-05, + "loss": 0.019, + "step": 30568 + }, + { + "epoch": 4.0, + "learning_rate": 1.0468370780721744e-05, + "loss": 0.0236, + "step": 30569 + }, + { + "epoch": 4.0, + "learning_rate": 1.0467841037007746e-05, + "loss": 0.0215, + "step": 30570 + }, + { + "epoch": 4.0, + "learning_rate": 1.046731129197797e-05, + "loss": 0.0223, + "step": 30571 + }, + { + "epoch": 4.0, + "learning_rate": 1.0466781545633909e-05, + "loss": 0.025, + "step": 30572 + }, + { + "epoch": 4.0, + "learning_rate": 1.0466251797977047e-05, + "loss": 0.0186, + "step": 30573 + }, + { + "epoch": 4.0, + "learning_rate": 1.0465722049008878e-05, + "loss": 0.0271, + "step": 30574 + }, + { + "epoch": 4.01, + "learning_rate": 1.0465192298730891e-05, + "loss": 0.0178, + "step": 30575 + }, + { + "epoch": 4.01, + "learning_rate": 1.0464662547144576e-05, + "loss": 0.0213, + "step": 30576 + }, + { + "epoch": 4.01, + "learning_rate": 1.046413279425142e-05, + "loss": 0.0291, + "step": 30577 + }, + { + "epoch": 4.01, + "learning_rate": 1.0463603040052914e-05, + "loss": 0.0166, + "step": 30578 + }, + { + "epoch": 4.01, + "learning_rate": 1.0463073284550552e-05, + "loss": 0.0228, + "step": 30579 + }, + { + "epoch": 4.01, + "learning_rate": 1.0462543527745824e-05, + "loss": 0.022, + "step": 30580 + }, + { + "epoch": 4.01, + "learning_rate": 1.046201376964021e-05, + "loss": 0.0187, + "step": 30581 + }, + { + "epoch": 4.01, + "learning_rate": 1.046148401023521e-05, + "loss": 0.019, + "step": 30582 + }, + { + "epoch": 4.01, + "learning_rate": 1.0460954249532311e-05, + "loss": 0.0197, + "step": 30583 + }, + { + "epoch": 4.01, + "learning_rate": 1.0460424487533006e-05, + "loss": 0.0324, + "step": 30584 + }, + { + "epoch": 4.01, + "learning_rate": 1.0459894724238777e-05, + "loss": 0.0283, + "step": 30585 + }, + { + "epoch": 4.01, + "learning_rate": 1.045936495965112e-05, + "loss": 0.0169, + "step": 30586 + }, + { + "epoch": 4.01, + "learning_rate": 1.0458835193771524e-05, + "loss": 0.0241, + "step": 30587 + }, + { + "epoch": 4.01, + "learning_rate": 1.0458305426601478e-05, + "loss": 0.0232, + "step": 30588 + }, + { + "epoch": 4.01, + "learning_rate": 1.045777565814247e-05, + "loss": 0.0142, + "step": 30589 + }, + { + "epoch": 4.01, + "learning_rate": 1.0457245888395997e-05, + "loss": 0.0241, + "step": 30590 + }, + { + "epoch": 4.01, + "learning_rate": 1.0456716117363542e-05, + "loss": 0.0218, + "step": 30591 + }, + { + "epoch": 4.01, + "learning_rate": 1.0456186345046598e-05, + "loss": 0.0246, + "step": 30592 + }, + { + "epoch": 4.01, + "learning_rate": 1.0455656571446651e-05, + "loss": 0.0141, + "step": 30593 + }, + { + "epoch": 4.01, + "learning_rate": 1.0455126796565198e-05, + "loss": 0.0293, + "step": 30594 + }, + { + "epoch": 4.01, + "learning_rate": 1.0454597020403723e-05, + "loss": 0.0205, + "step": 30595 + }, + { + "epoch": 4.01, + "learning_rate": 1.0454067242963718e-05, + "loss": 0.0233, + "step": 30596 + }, + { + "epoch": 4.01, + "learning_rate": 1.0453537464246677e-05, + "loss": 0.022, + "step": 30597 + }, + { + "epoch": 4.01, + "learning_rate": 1.0453007684254081e-05, + "loss": 0.0203, + "step": 30598 + }, + { + "epoch": 4.01, + "learning_rate": 1.0452477902987428e-05, + "loss": 0.0202, + "step": 30599 + }, + { + "epoch": 4.01, + "learning_rate": 1.0451948120448204e-05, + "loss": 0.0267, + "step": 30600 + }, + { + "epoch": 4.01, + "learning_rate": 1.0451418336637899e-05, + "loss": 0.0217, + "step": 30601 + }, + { + "epoch": 4.01, + "learning_rate": 1.0450888551558005e-05, + "loss": 0.0232, + "step": 30602 + }, + { + "epoch": 4.01, + "learning_rate": 1.0450358765210014e-05, + "loss": 0.0254, + "step": 30603 + }, + { + "epoch": 4.01, + "learning_rate": 1.0449828977595406e-05, + "loss": 0.0233, + "step": 30604 + }, + { + "epoch": 4.01, + "learning_rate": 1.0449299188715682e-05, + "loss": 0.0214, + "step": 30605 + }, + { + "epoch": 4.01, + "learning_rate": 1.0448769398572327e-05, + "loss": 0.0219, + "step": 30606 + }, + { + "epoch": 4.01, + "learning_rate": 1.0448239607166835e-05, + "loss": 0.0261, + "step": 30607 + }, + { + "epoch": 4.01, + "learning_rate": 1.044770981450069e-05, + "loss": 0.0185, + "step": 30608 + }, + { + "epoch": 4.01, + "learning_rate": 1.0447180020575387e-05, + "loss": 0.0225, + "step": 30609 + }, + { + "epoch": 4.01, + "learning_rate": 1.0446650225392409e-05, + "loss": 0.0185, + "step": 30610 + }, + { + "epoch": 4.01, + "learning_rate": 1.0446120428953257e-05, + "loss": 0.0283, + "step": 30611 + }, + { + "epoch": 4.01, + "learning_rate": 1.0445590631259413e-05, + "loss": 0.0209, + "step": 30612 + }, + { + "epoch": 4.01, + "learning_rate": 1.044506083231237e-05, + "loss": 0.0221, + "step": 30613 + }, + { + "epoch": 4.01, + "learning_rate": 1.0444531032113616e-05, + "loss": 0.0258, + "step": 30614 + }, + { + "epoch": 4.01, + "learning_rate": 1.0444001230664642e-05, + "loss": 0.0224, + "step": 30615 + }, + { + "epoch": 4.01, + "learning_rate": 1.0443471427966935e-05, + "loss": 0.0214, + "step": 30616 + }, + { + "epoch": 4.01, + "learning_rate": 1.0442941624021991e-05, + "loss": 0.0248, + "step": 30617 + }, + { + "epoch": 4.01, + "learning_rate": 1.0442411818831302e-05, + "loss": 0.0216, + "step": 30618 + }, + { + "epoch": 4.01, + "learning_rate": 1.0441882012396347e-05, + "loss": 0.0289, + "step": 30619 + }, + { + "epoch": 4.01, + "learning_rate": 1.0441352204718627e-05, + "loss": 0.0195, + "step": 30620 + }, + { + "epoch": 4.01, + "learning_rate": 1.0440822395799621e-05, + "loss": 0.0224, + "step": 30621 + }, + { + "epoch": 4.01, + "learning_rate": 1.0440292585640832e-05, + "loss": 0.0246, + "step": 30622 + }, + { + "epoch": 4.01, + "learning_rate": 1.043976277424374e-05, + "loss": 0.0215, + "step": 30623 + }, + { + "epoch": 4.01, + "learning_rate": 1.043923296160984e-05, + "loss": 0.0232, + "step": 30624 + }, + { + "epoch": 4.01, + "learning_rate": 1.043870314774062e-05, + "loss": 0.0198, + "step": 30625 + }, + { + "epoch": 4.01, + "learning_rate": 1.043817333263757e-05, + "loss": 0.0266, + "step": 30626 + }, + { + "epoch": 4.01, + "learning_rate": 1.0437643516302183e-05, + "loss": 0.0194, + "step": 30627 + }, + { + "epoch": 4.01, + "learning_rate": 1.0437113698735946e-05, + "loss": 0.0227, + "step": 30628 + }, + { + "epoch": 4.01, + "learning_rate": 1.043658387994035e-05, + "loss": 0.0282, + "step": 30629 + }, + { + "epoch": 4.01, + "learning_rate": 1.0436054059916885e-05, + "loss": 0.0228, + "step": 30630 + }, + { + "epoch": 4.01, + "learning_rate": 1.0435524238667038e-05, + "loss": 0.0261, + "step": 30631 + }, + { + "epoch": 4.01, + "learning_rate": 1.0434994416192308e-05, + "loss": 0.0256, + "step": 30632 + }, + { + "epoch": 4.01, + "learning_rate": 1.0434464592494176e-05, + "loss": 0.0199, + "step": 30633 + }, + { + "epoch": 4.01, + "learning_rate": 1.0433934767574136e-05, + "loss": 0.024, + "step": 30634 + }, + { + "epoch": 4.01, + "learning_rate": 1.0433404941433678e-05, + "loss": 0.0181, + "step": 30635 + }, + { + "epoch": 4.01, + "learning_rate": 1.043287511407429e-05, + "loss": 0.0244, + "step": 30636 + }, + { + "epoch": 4.01, + "learning_rate": 1.0432345285497466e-05, + "loss": 0.0219, + "step": 30637 + }, + { + "epoch": 4.01, + "learning_rate": 1.0431815455704695e-05, + "loss": 0.026, + "step": 30638 + }, + { + "epoch": 4.01, + "learning_rate": 1.0431285624697463e-05, + "loss": 0.0232, + "step": 30639 + }, + { + "epoch": 4.01, + "learning_rate": 1.0430755792477263e-05, + "loss": 0.0249, + "step": 30640 + }, + { + "epoch": 4.01, + "learning_rate": 1.0430225959045588e-05, + "loss": 0.0211, + "step": 30641 + }, + { + "epoch": 4.01, + "learning_rate": 1.042969612440392e-05, + "loss": 0.0241, + "step": 30642 + }, + { + "epoch": 4.01, + "learning_rate": 1.0429166288553763e-05, + "loss": 0.0222, + "step": 30643 + }, + { + "epoch": 4.01, + "learning_rate": 1.0428636451496591e-05, + "loss": 0.0187, + "step": 30644 + }, + { + "epoch": 4.01, + "learning_rate": 1.0428106613233908e-05, + "loss": 0.0287, + "step": 30645 + }, + { + "epoch": 4.01, + "learning_rate": 1.0427576773767194e-05, + "loss": 0.024, + "step": 30646 + }, + { + "epoch": 4.01, + "learning_rate": 1.0427046933097947e-05, + "loss": 0.0202, + "step": 30647 + }, + { + "epoch": 4.01, + "learning_rate": 1.0426517091227646e-05, + "loss": 0.0215, + "step": 30648 + }, + { + "epoch": 4.01, + "learning_rate": 1.0425987248157795e-05, + "loss": 0.0219, + "step": 30649 + }, + { + "epoch": 4.01, + "learning_rate": 1.0425457403889874e-05, + "loss": 0.0192, + "step": 30650 + }, + { + "epoch": 4.02, + "learning_rate": 1.042492755842538e-05, + "loss": 0.0281, + "step": 30651 + }, + { + "epoch": 4.02, + "learning_rate": 1.0424397711765797e-05, + "loss": 0.0188, + "step": 30652 + }, + { + "epoch": 4.02, + "learning_rate": 1.0423867863912618e-05, + "loss": 0.0209, + "step": 30653 + }, + { + "epoch": 4.02, + "learning_rate": 1.0423338014867332e-05, + "loss": 0.023, + "step": 30654 + }, + { + "epoch": 4.02, + "learning_rate": 1.0422808164631434e-05, + "loss": 0.0214, + "step": 30655 + }, + { + "epoch": 4.02, + "learning_rate": 1.0422278313206409e-05, + "loss": 0.0246, + "step": 30656 + }, + { + "epoch": 4.02, + "learning_rate": 1.0421748460593749e-05, + "loss": 0.0258, + "step": 30657 + }, + { + "epoch": 4.02, + "learning_rate": 1.0421218606794942e-05, + "loss": 0.0181, + "step": 30658 + }, + { + "epoch": 4.02, + "learning_rate": 1.0420688751811483e-05, + "loss": 0.0213, + "step": 30659 + }, + { + "epoch": 4.02, + "learning_rate": 1.0420158895644857e-05, + "loss": 0.0255, + "step": 30660 + }, + { + "epoch": 4.02, + "learning_rate": 1.0419629038296557e-05, + "loss": 0.0248, + "step": 30661 + }, + { + "epoch": 4.02, + "learning_rate": 1.0419099179768075e-05, + "loss": 0.0208, + "step": 30662 + }, + { + "epoch": 4.02, + "learning_rate": 1.0418569320060897e-05, + "loss": 0.0202, + "step": 30663 + }, + { + "epoch": 4.02, + "learning_rate": 1.0418039459176512e-05, + "loss": 0.0147, + "step": 30664 + }, + { + "epoch": 4.02, + "learning_rate": 1.0417509597116418e-05, + "loss": 0.0247, + "step": 30665 + }, + { + "epoch": 4.02, + "learning_rate": 1.04169797338821e-05, + "loss": 0.0213, + "step": 30666 + }, + { + "epoch": 4.02, + "learning_rate": 1.0416449869475048e-05, + "loss": 0.0255, + "step": 30667 + }, + { + "epoch": 4.02, + "learning_rate": 1.041592000389675e-05, + "loss": 0.0219, + "step": 30668 + }, + { + "epoch": 4.02, + "learning_rate": 1.0415390137148701e-05, + "loss": 0.0224, + "step": 30669 + }, + { + "epoch": 4.02, + "learning_rate": 1.0414860269232392e-05, + "loss": 0.0182, + "step": 30670 + }, + { + "epoch": 4.02, + "learning_rate": 1.0414330400149306e-05, + "loss": 0.0239, + "step": 30671 + }, + { + "epoch": 4.02, + "learning_rate": 1.0413800529900944e-05, + "loss": 0.0259, + "step": 30672 + }, + { + "epoch": 4.02, + "learning_rate": 1.0413270658488785e-05, + "loss": 0.0286, + "step": 30673 + }, + { + "epoch": 4.02, + "learning_rate": 1.0412740785914328e-05, + "loss": 0.0227, + "step": 30674 + }, + { + "epoch": 4.02, + "learning_rate": 1.0412210912179053e-05, + "loss": 0.0245, + "step": 30675 + }, + { + "epoch": 4.02, + "learning_rate": 1.0411681037284463e-05, + "loss": 0.02, + "step": 30676 + }, + { + "epoch": 4.02, + "learning_rate": 1.0411151161232041e-05, + "loss": 0.023, + "step": 30677 + }, + { + "epoch": 4.02, + "learning_rate": 1.0410621284023277e-05, + "loss": 0.0259, + "step": 30678 + }, + { + "epoch": 4.02, + "learning_rate": 1.0410091405659661e-05, + "loss": 0.022, + "step": 30679 + }, + { + "epoch": 4.02, + "learning_rate": 1.0409561526142687e-05, + "loss": 0.0219, + "step": 30680 + }, + { + "epoch": 4.02, + "learning_rate": 1.0409031645473842e-05, + "loss": 0.0241, + "step": 30681 + }, + { + "epoch": 4.02, + "learning_rate": 1.0408501763654617e-05, + "loss": 0.0256, + "step": 30682 + }, + { + "epoch": 4.02, + "learning_rate": 1.0407971880686505e-05, + "loss": 0.018, + "step": 30683 + }, + { + "epoch": 4.02, + "learning_rate": 1.040744199657099e-05, + "loss": 0.024, + "step": 30684 + }, + { + "epoch": 4.02, + "learning_rate": 1.040691211130957e-05, + "loss": 0.0254, + "step": 30685 + }, + { + "epoch": 4.02, + "learning_rate": 1.0406382224903729e-05, + "loss": 0.0227, + "step": 30686 + }, + { + "epoch": 4.02, + "learning_rate": 1.0405852337354961e-05, + "loss": 0.0226, + "step": 30687 + }, + { + "epoch": 4.02, + "learning_rate": 1.0405322448664753e-05, + "loss": 0.019, + "step": 30688 + }, + { + "epoch": 4.02, + "learning_rate": 1.04047925588346e-05, + "loss": 0.0228, + "step": 30689 + }, + { + "epoch": 4.02, + "learning_rate": 1.0404262667865988e-05, + "loss": 0.0192, + "step": 30690 + }, + { + "epoch": 4.02, + "learning_rate": 1.0403732775760406e-05, + "loss": 0.022, + "step": 30691 + }, + { + "epoch": 4.02, + "learning_rate": 1.040320288251935e-05, + "loss": 0.0229, + "step": 30692 + }, + { + "epoch": 4.02, + "learning_rate": 1.040267298814431e-05, + "loss": 0.0213, + "step": 30693 + }, + { + "epoch": 4.02, + "learning_rate": 1.0402143092636769e-05, + "loss": 0.0213, + "step": 30694 + }, + { + "epoch": 4.02, + "learning_rate": 1.0401613195998226e-05, + "loss": 0.0187, + "step": 30695 + }, + { + "epoch": 4.02, + "learning_rate": 1.0401083298230163e-05, + "loss": 0.0222, + "step": 30696 + }, + { + "epoch": 4.02, + "learning_rate": 1.0400553399334077e-05, + "loss": 0.0265, + "step": 30697 + }, + { + "epoch": 4.02, + "learning_rate": 1.0400023499311453e-05, + "loss": 0.0223, + "step": 30698 + }, + { + "epoch": 4.02, + "learning_rate": 1.0399493598163788e-05, + "loss": 0.0293, + "step": 30699 + }, + { + "epoch": 4.02, + "learning_rate": 1.0398963695892568e-05, + "loss": 0.0258, + "step": 30700 + }, + { + "epoch": 4.02, + "learning_rate": 1.0398433792499283e-05, + "loss": 0.0223, + "step": 30701 + }, + { + "epoch": 4.02, + "learning_rate": 1.0397903887985424e-05, + "loss": 0.0198, + "step": 30702 + }, + { + "epoch": 4.02, + "learning_rate": 1.039737398235248e-05, + "loss": 0.0203, + "step": 30703 + }, + { + "epoch": 4.02, + "learning_rate": 1.0396844075601947e-05, + "loss": 0.0177, + "step": 30704 + }, + { + "epoch": 4.02, + "learning_rate": 1.0396314167735306e-05, + "loss": 0.0181, + "step": 30705 + }, + { + "epoch": 4.02, + "learning_rate": 1.0395784258754058e-05, + "loss": 0.0232, + "step": 30706 + }, + { + "epoch": 4.02, + "learning_rate": 1.0395254348659684e-05, + "loss": 0.0194, + "step": 30707 + }, + { + "epoch": 4.02, + "learning_rate": 1.0394724437453679e-05, + "loss": 0.0329, + "step": 30708 + }, + { + "epoch": 4.02, + "learning_rate": 1.039419452513753e-05, + "loss": 0.0233, + "step": 30709 + }, + { + "epoch": 4.02, + "learning_rate": 1.0393664611712736e-05, + "loss": 0.0236, + "step": 30710 + }, + { + "epoch": 4.02, + "learning_rate": 1.0393134697180776e-05, + "loss": 0.0173, + "step": 30711 + }, + { + "epoch": 4.02, + "learning_rate": 1.039260478154315e-05, + "loss": 0.0223, + "step": 30712 + }, + { + "epoch": 4.02, + "learning_rate": 1.0392074864801338e-05, + "loss": 0.0212, + "step": 30713 + }, + { + "epoch": 4.02, + "learning_rate": 1.0391544946956842e-05, + "loss": 0.0222, + "step": 30714 + }, + { + "epoch": 4.02, + "learning_rate": 1.0391015028011144e-05, + "loss": 0.019, + "step": 30715 + }, + { + "epoch": 4.02, + "learning_rate": 1.0390485107965737e-05, + "loss": 0.0277, + "step": 30716 + }, + { + "epoch": 4.02, + "learning_rate": 1.0389955186822114e-05, + "loss": 0.0255, + "step": 30717 + }, + { + "epoch": 4.02, + "learning_rate": 1.0389425264581758e-05, + "loss": 0.0193, + "step": 30718 + }, + { + "epoch": 4.02, + "learning_rate": 1.0388895341246169e-05, + "loss": 0.0309, + "step": 30719 + }, + { + "epoch": 4.02, + "learning_rate": 1.038836541681683e-05, + "loss": 0.0225, + "step": 30720 + }, + { + "epoch": 4.02, + "learning_rate": 1.0387835491295235e-05, + "loss": 0.0205, + "step": 30721 + }, + { + "epoch": 4.02, + "learning_rate": 1.0387305564682873e-05, + "loss": 0.019, + "step": 30722 + }, + { + "epoch": 4.02, + "learning_rate": 1.0386775636981235e-05, + "loss": 0.0197, + "step": 30723 + }, + { + "epoch": 4.02, + "learning_rate": 1.0386245708191808e-05, + "loss": 0.0203, + "step": 30724 + }, + { + "epoch": 4.02, + "learning_rate": 1.0385715778316087e-05, + "loss": 0.0227, + "step": 30725 + }, + { + "epoch": 4.02, + "learning_rate": 1.0385185847355563e-05, + "loss": 0.0198, + "step": 30726 + }, + { + "epoch": 4.03, + "learning_rate": 1.0384655915311725e-05, + "loss": 0.0228, + "step": 30727 + }, + { + "epoch": 4.03, + "learning_rate": 1.0384125982186059e-05, + "loss": 0.0216, + "step": 30728 + }, + { + "epoch": 4.03, + "learning_rate": 1.0383596047980065e-05, + "loss": 0.0252, + "step": 30729 + }, + { + "epoch": 4.03, + "learning_rate": 1.0383066112695219e-05, + "loss": 0.0252, + "step": 30730 + }, + { + "epoch": 4.03, + "learning_rate": 1.0382536176333026e-05, + "loss": 0.0274, + "step": 30731 + }, + { + "epoch": 4.03, + "learning_rate": 1.038200623889497e-05, + "loss": 0.0214, + "step": 30732 + }, + { + "epoch": 4.03, + "learning_rate": 1.0381476300382542e-05, + "loss": 0.0219, + "step": 30733 + }, + { + "epoch": 4.03, + "learning_rate": 1.0380946360797231e-05, + "loss": 0.024, + "step": 30734 + }, + { + "epoch": 4.03, + "learning_rate": 1.0380416420140529e-05, + "loss": 0.0231, + "step": 30735 + }, + { + "epoch": 4.03, + "learning_rate": 1.0379886478413924e-05, + "loss": 0.025, + "step": 30736 + }, + { + "epoch": 4.03, + "learning_rate": 1.0379356535618911e-05, + "loss": 0.0254, + "step": 30737 + }, + { + "epoch": 4.03, + "learning_rate": 1.0378826591756979e-05, + "loss": 0.0234, + "step": 30738 + }, + { + "epoch": 4.03, + "learning_rate": 1.0378296646829615e-05, + "loss": 0.0208, + "step": 30739 + }, + { + "epoch": 4.03, + "learning_rate": 1.0377766700838313e-05, + "loss": 0.0245, + "step": 30740 + }, + { + "epoch": 4.03, + "learning_rate": 1.037723675378456e-05, + "loss": 0.0231, + "step": 30741 + }, + { + "epoch": 4.03, + "learning_rate": 1.0376706805669852e-05, + "loss": 0.0216, + "step": 30742 + }, + { + "epoch": 4.03, + "learning_rate": 1.0376176856495672e-05, + "loss": 0.0287, + "step": 30743 + }, + { + "epoch": 4.03, + "learning_rate": 1.0375646906263519e-05, + "loss": 0.0282, + "step": 30744 + }, + { + "epoch": 4.03, + "learning_rate": 1.0375116954974878e-05, + "loss": 0.0222, + "step": 30745 + }, + { + "epoch": 4.03, + "learning_rate": 1.0374587002631238e-05, + "loss": 0.0257, + "step": 30746 + }, + { + "epoch": 4.03, + "learning_rate": 1.0374057049234094e-05, + "loss": 0.0231, + "step": 30747 + }, + { + "epoch": 4.03, + "learning_rate": 1.0373527094784937e-05, + "loss": 0.0313, + "step": 30748 + }, + { + "epoch": 4.03, + "learning_rate": 1.037299713928525e-05, + "loss": 0.0236, + "step": 30749 + }, + { + "epoch": 4.03, + "learning_rate": 1.0372467182736532e-05, + "loss": 0.0237, + "step": 30750 + }, + { + "epoch": 4.03, + "learning_rate": 1.0371937225140269e-05, + "loss": 0.017, + "step": 30751 + }, + { + "epoch": 4.03, + "learning_rate": 1.037140726649795e-05, + "loss": 0.0307, + "step": 30752 + }, + { + "epoch": 4.03, + "learning_rate": 1.0370877306811071e-05, + "loss": 0.0206, + "step": 30753 + }, + { + "epoch": 4.03, + "learning_rate": 1.037034734608112e-05, + "loss": 0.0216, + "step": 30754 + }, + { + "epoch": 4.03, + "learning_rate": 1.0369817384309584e-05, + "loss": 0.0228, + "step": 30755 + }, + { + "epoch": 4.03, + "learning_rate": 1.0369287421497957e-05, + "loss": 0.023, + "step": 30756 + }, + { + "epoch": 4.03, + "learning_rate": 1.0368757457647726e-05, + "loss": 0.0248, + "step": 30757 + }, + { + "epoch": 4.03, + "learning_rate": 1.036822749276039e-05, + "loss": 0.0225, + "step": 30758 + }, + { + "epoch": 4.03, + "learning_rate": 1.036769752683743e-05, + "loss": 0.0213, + "step": 30759 + }, + { + "epoch": 4.03, + "learning_rate": 1.0367167559880344e-05, + "loss": 0.0255, + "step": 30760 + }, + { + "epoch": 4.03, + "learning_rate": 1.0366637591890616e-05, + "loss": 0.0242, + "step": 30761 + }, + { + "epoch": 4.03, + "learning_rate": 1.0366107622869738e-05, + "loss": 0.0268, + "step": 30762 + }, + { + "epoch": 4.03, + "learning_rate": 1.0365577652819204e-05, + "loss": 0.0209, + "step": 30763 + }, + { + "epoch": 4.03, + "learning_rate": 1.03650476817405e-05, + "loss": 0.0203, + "step": 30764 + }, + { + "epoch": 4.03, + "learning_rate": 1.0364517709635122e-05, + "loss": 0.0216, + "step": 30765 + }, + { + "epoch": 4.03, + "learning_rate": 1.0363987736504555e-05, + "loss": 0.0227, + "step": 30766 + }, + { + "epoch": 4.03, + "learning_rate": 1.0363457762350295e-05, + "loss": 0.0213, + "step": 30767 + }, + { + "epoch": 4.03, + "learning_rate": 1.0362927787173824e-05, + "loss": 0.0256, + "step": 30768 + }, + { + "epoch": 4.03, + "learning_rate": 1.0362397810976641e-05, + "loss": 0.0191, + "step": 30769 + }, + { + "epoch": 4.03, + "learning_rate": 1.0361867833760234e-05, + "loss": 0.0273, + "step": 30770 + }, + { + "epoch": 4.03, + "learning_rate": 1.0361337855526094e-05, + "loss": 0.0224, + "step": 30771 + }, + { + "epoch": 4.03, + "learning_rate": 1.036080787627571e-05, + "loss": 0.0244, + "step": 30772 + }, + { + "epoch": 4.03, + "learning_rate": 1.036027789601057e-05, + "loss": 0.0227, + "step": 30773 + }, + { + "epoch": 4.03, + "learning_rate": 1.0359747914732168e-05, + "loss": 0.0249, + "step": 30774 + }, + { + "epoch": 4.03, + "learning_rate": 1.0359217932441999e-05, + "loss": 0.0224, + "step": 30775 + }, + { + "epoch": 4.03, + "learning_rate": 1.0358687949141543e-05, + "loss": 0.0215, + "step": 30776 + }, + { + "epoch": 4.03, + "learning_rate": 1.03581579648323e-05, + "loss": 0.0292, + "step": 30777 + }, + { + "epoch": 4.03, + "learning_rate": 1.0357627979515757e-05, + "loss": 0.0282, + "step": 30778 + }, + { + "epoch": 4.03, + "learning_rate": 1.0357097993193401e-05, + "loss": 0.0183, + "step": 30779 + }, + { + "epoch": 4.03, + "learning_rate": 1.0356568005866727e-05, + "loss": 0.0165, + "step": 30780 + }, + { + "epoch": 4.03, + "learning_rate": 1.0356038017537227e-05, + "loss": 0.0286, + "step": 30781 + }, + { + "epoch": 4.03, + "learning_rate": 1.0355508028206386e-05, + "loss": 0.0168, + "step": 30782 + }, + { + "epoch": 4.03, + "learning_rate": 1.0354978037875698e-05, + "loss": 0.0251, + "step": 30783 + }, + { + "epoch": 4.03, + "learning_rate": 1.0354448046546654e-05, + "loss": 0.0258, + "step": 30784 + }, + { + "epoch": 4.03, + "learning_rate": 1.0353918054220746e-05, + "loss": 0.021, + "step": 30785 + }, + { + "epoch": 4.03, + "learning_rate": 1.0353388060899459e-05, + "loss": 0.0267, + "step": 30786 + }, + { + "epoch": 4.03, + "learning_rate": 1.0352858066584288e-05, + "loss": 0.0241, + "step": 30787 + }, + { + "epoch": 4.03, + "learning_rate": 1.0352328071276722e-05, + "loss": 0.0253, + "step": 30788 + }, + { + "epoch": 4.03, + "learning_rate": 1.0351798074978255e-05, + "loss": 0.0253, + "step": 30789 + }, + { + "epoch": 4.03, + "learning_rate": 1.0351268077690371e-05, + "loss": 0.0177, + "step": 30790 + }, + { + "epoch": 4.03, + "learning_rate": 1.0350738079414565e-05, + "loss": 0.0234, + "step": 30791 + }, + { + "epoch": 4.03, + "learning_rate": 1.0350208080152329e-05, + "loss": 0.0181, + "step": 30792 + }, + { + "epoch": 4.03, + "learning_rate": 1.0349678079905148e-05, + "loss": 0.0192, + "step": 30793 + }, + { + "epoch": 4.03, + "learning_rate": 1.0349148078674518e-05, + "loss": 0.0232, + "step": 30794 + }, + { + "epoch": 4.03, + "learning_rate": 1.0348618076461925e-05, + "loss": 0.0213, + "step": 30795 + }, + { + "epoch": 4.03, + "learning_rate": 1.0348088073268867e-05, + "loss": 0.0256, + "step": 30796 + }, + { + "epoch": 4.03, + "learning_rate": 1.0347558069096825e-05, + "loss": 0.0232, + "step": 30797 + }, + { + "epoch": 4.03, + "learning_rate": 1.03470280639473e-05, + "loss": 0.0262, + "step": 30798 + }, + { + "epoch": 4.03, + "learning_rate": 1.0346498057821772e-05, + "loss": 0.0237, + "step": 30799 + }, + { + "epoch": 4.03, + "learning_rate": 1.0345968050721742e-05, + "loss": 0.0231, + "step": 30800 + }, + { + "epoch": 4.03, + "learning_rate": 1.0345438042648688e-05, + "loss": 0.0237, + "step": 30801 + }, + { + "epoch": 4.03, + "learning_rate": 1.0344908033604113e-05, + "loss": 0.0247, + "step": 30802 + }, + { + "epoch": 4.03, + "learning_rate": 1.03443780235895e-05, + "loss": 0.0186, + "step": 30803 + }, + { + "epoch": 4.04, + "learning_rate": 1.0343848012606346e-05, + "loss": 0.02, + "step": 30804 + }, + { + "epoch": 4.04, + "learning_rate": 1.0343318000656134e-05, + "loss": 0.0234, + "step": 30805 + }, + { + "epoch": 4.04, + "learning_rate": 1.034278798774036e-05, + "loss": 0.0228, + "step": 30806 + }, + { + "epoch": 4.04, + "learning_rate": 1.0342257973860512e-05, + "loss": 0.0261, + "step": 30807 + }, + { + "epoch": 4.04, + "learning_rate": 1.0341727959018082e-05, + "loss": 0.0199, + "step": 30808 + }, + { + "epoch": 4.04, + "learning_rate": 1.0341197943214562e-05, + "loss": 0.0232, + "step": 30809 + }, + { + "epoch": 4.04, + "learning_rate": 1.0340667926451438e-05, + "loss": 0.0161, + "step": 30810 + }, + { + "epoch": 4.04, + "learning_rate": 1.0340137908730208e-05, + "loss": 0.0228, + "step": 30811 + }, + { + "epoch": 4.04, + "learning_rate": 1.0339607890052352e-05, + "loss": 0.0234, + "step": 30812 + }, + { + "epoch": 4.04, + "learning_rate": 1.0339077870419372e-05, + "loss": 0.0221, + "step": 30813 + }, + { + "epoch": 4.04, + "learning_rate": 1.0338547849832752e-05, + "loss": 0.0293, + "step": 30814 + }, + { + "epoch": 4.04, + "learning_rate": 1.0338017828293985e-05, + "loss": 0.0207, + "step": 30815 + }, + { + "epoch": 4.04, + "learning_rate": 1.0337487805804559e-05, + "loss": 0.024, + "step": 30816 + }, + { + "epoch": 4.04, + "learning_rate": 1.0336957782365966e-05, + "loss": 0.0215, + "step": 30817 + }, + { + "epoch": 4.04, + "learning_rate": 1.0336427757979698e-05, + "loss": 0.0205, + "step": 30818 + }, + { + "epoch": 4.04, + "learning_rate": 1.0335897732647247e-05, + "loss": 0.022, + "step": 30819 + }, + { + "epoch": 4.04, + "learning_rate": 1.0335367706370101e-05, + "loss": 0.0262, + "step": 30820 + }, + { + "epoch": 4.04, + "learning_rate": 1.033483767914975e-05, + "loss": 0.0182, + "step": 30821 + }, + { + "epoch": 4.04, + "learning_rate": 1.0334307650987684e-05, + "loss": 0.0248, + "step": 30822 + }, + { + "epoch": 4.04, + "learning_rate": 1.03337776218854e-05, + "loss": 0.0227, + "step": 30823 + }, + { + "epoch": 4.04, + "learning_rate": 1.033324759184438e-05, + "loss": 0.0226, + "step": 30824 + }, + { + "epoch": 4.04, + "learning_rate": 1.0332717560866122e-05, + "loss": 0.0212, + "step": 30825 + }, + { + "epoch": 4.04, + "learning_rate": 1.0332187528952111e-05, + "loss": 0.0231, + "step": 30826 + }, + { + "epoch": 4.04, + "learning_rate": 1.0331657496103842e-05, + "loss": 0.0238, + "step": 30827 + }, + { + "epoch": 4.04, + "learning_rate": 1.0331127462322803e-05, + "loss": 0.0267, + "step": 30828 + }, + { + "epoch": 4.04, + "learning_rate": 1.0330597427610486e-05, + "loss": 0.0185, + "step": 30829 + }, + { + "epoch": 4.04, + "learning_rate": 1.0330067391968382e-05, + "loss": 0.0211, + "step": 30830 + }, + { + "epoch": 4.04, + "learning_rate": 1.032953735539798e-05, + "loss": 0.0161, + "step": 30831 + }, + { + "epoch": 4.04, + "learning_rate": 1.0329007317900774e-05, + "loss": 0.0231, + "step": 30832 + }, + { + "epoch": 4.04, + "learning_rate": 1.0328477279478248e-05, + "loss": 0.0257, + "step": 30833 + }, + { + "epoch": 4.04, + "learning_rate": 1.0327947240131901e-05, + "loss": 0.0218, + "step": 30834 + }, + { + "epoch": 4.04, + "learning_rate": 1.032741719986322e-05, + "loss": 0.0229, + "step": 30835 + }, + { + "epoch": 4.04, + "learning_rate": 1.0326887158673693e-05, + "loss": 0.0205, + "step": 30836 + }, + { + "epoch": 4.04, + "learning_rate": 1.0326357116564814e-05, + "loss": 0.0247, + "step": 30837 + }, + { + "epoch": 4.04, + "learning_rate": 1.0325827073538075e-05, + "loss": 0.0228, + "step": 30838 + }, + { + "epoch": 4.04, + "learning_rate": 1.0325297029594961e-05, + "loss": 0.02, + "step": 30839 + }, + { + "epoch": 4.04, + "learning_rate": 1.0324766984736969e-05, + "loss": 0.0229, + "step": 30840 + }, + { + "epoch": 4.04, + "learning_rate": 1.0324236938965586e-05, + "loss": 0.0238, + "step": 30841 + }, + { + "epoch": 4.04, + "learning_rate": 1.0323706892282308e-05, + "loss": 0.0285, + "step": 30842 + }, + { + "epoch": 4.04, + "learning_rate": 1.0323176844688617e-05, + "loss": 0.0189, + "step": 30843 + }, + { + "epoch": 4.04, + "learning_rate": 1.032264679618601e-05, + "loss": 0.0227, + "step": 30844 + }, + { + "epoch": 4.04, + "learning_rate": 1.0322116746775974e-05, + "loss": 0.0171, + "step": 30845 + }, + { + "epoch": 4.04, + "learning_rate": 1.0321586696460005e-05, + "loss": 0.0251, + "step": 30846 + }, + { + "epoch": 4.04, + "learning_rate": 1.0321056645239589e-05, + "loss": 0.018, + "step": 30847 + }, + { + "epoch": 4.04, + "learning_rate": 1.0320526593116219e-05, + "loss": 0.021, + "step": 30848 + }, + { + "epoch": 4.04, + "learning_rate": 1.0319996540091384e-05, + "loss": 0.0253, + "step": 30849 + }, + { + "epoch": 4.04, + "learning_rate": 1.0319466486166576e-05, + "loss": 0.0222, + "step": 30850 + }, + { + "epoch": 4.04, + "learning_rate": 1.0318936431343286e-05, + "loss": 0.0224, + "step": 30851 + }, + { + "epoch": 4.04, + "learning_rate": 1.0318406375623002e-05, + "loss": 0.0238, + "step": 30852 + }, + { + "epoch": 4.04, + "learning_rate": 1.031787631900722e-05, + "loss": 0.0194, + "step": 30853 + }, + { + "epoch": 4.04, + "learning_rate": 1.0317346261497425e-05, + "loss": 0.0231, + "step": 30854 + }, + { + "epoch": 4.04, + "learning_rate": 1.0316816203095112e-05, + "loss": 0.021, + "step": 30855 + }, + { + "epoch": 4.04, + "learning_rate": 1.0316286143801771e-05, + "loss": 0.0254, + "step": 30856 + }, + { + "epoch": 4.04, + "learning_rate": 1.0315756083618893e-05, + "loss": 0.0232, + "step": 30857 + }, + { + "epoch": 4.04, + "learning_rate": 1.0315226022547966e-05, + "loss": 0.0195, + "step": 30858 + }, + { + "epoch": 4.04, + "learning_rate": 1.0314695960590485e-05, + "loss": 0.0228, + "step": 30859 + }, + { + "epoch": 4.04, + "learning_rate": 1.0314165897747935e-05, + "loss": 0.0194, + "step": 30860 + }, + { + "epoch": 4.04, + "learning_rate": 1.0313635834021811e-05, + "loss": 0.0228, + "step": 30861 + }, + { + "epoch": 4.04, + "learning_rate": 1.0313105769413602e-05, + "loss": 0.0184, + "step": 30862 + }, + { + "epoch": 4.04, + "learning_rate": 1.0312575703924803e-05, + "loss": 0.0217, + "step": 30863 + }, + { + "epoch": 4.04, + "learning_rate": 1.0312045637556897e-05, + "loss": 0.0243, + "step": 30864 + }, + { + "epoch": 4.04, + "learning_rate": 1.0311515570311385e-05, + "loss": 0.0292, + "step": 30865 + }, + { + "epoch": 4.04, + "learning_rate": 1.0310985502189745e-05, + "loss": 0.0209, + "step": 30866 + }, + { + "epoch": 4.04, + "learning_rate": 1.0310455433193482e-05, + "loss": 0.0177, + "step": 30867 + }, + { + "epoch": 4.04, + "learning_rate": 1.0309925363324076e-05, + "loss": 0.0191, + "step": 30868 + }, + { + "epoch": 4.04, + "learning_rate": 1.0309395292583021e-05, + "loss": 0.0261, + "step": 30869 + }, + { + "epoch": 4.04, + "learning_rate": 1.030886522097181e-05, + "loss": 0.0221, + "step": 30870 + }, + { + "epoch": 4.04, + "learning_rate": 1.030833514849193e-05, + "loss": 0.017, + "step": 30871 + }, + { + "epoch": 4.04, + "learning_rate": 1.0307805075144874e-05, + "loss": 0.0197, + "step": 30872 + }, + { + "epoch": 4.04, + "learning_rate": 1.0307275000932133e-05, + "loss": 0.0218, + "step": 30873 + }, + { + "epoch": 4.04, + "learning_rate": 1.03067449258552e-05, + "loss": 0.0194, + "step": 30874 + }, + { + "epoch": 4.04, + "learning_rate": 1.0306214849915561e-05, + "loss": 0.0248, + "step": 30875 + }, + { + "epoch": 4.04, + "learning_rate": 1.030568477311471e-05, + "loss": 0.0287, + "step": 30876 + }, + { + "epoch": 4.04, + "learning_rate": 1.0305154695454133e-05, + "loss": 0.0238, + "step": 30877 + }, + { + "epoch": 4.04, + "learning_rate": 1.0304624616935331e-05, + "loss": 0.0269, + "step": 30878 + }, + { + "epoch": 4.04, + "learning_rate": 1.0304094537559784e-05, + "loss": 0.0216, + "step": 30879 + }, + { + "epoch": 4.05, + "learning_rate": 1.030356445732899e-05, + "loss": 0.0207, + "step": 30880 + }, + { + "epoch": 4.05, + "learning_rate": 1.0303034376244435e-05, + "loss": 0.0246, + "step": 30881 + }, + { + "epoch": 4.05, + "learning_rate": 1.0302504294307614e-05, + "loss": 0.0206, + "step": 30882 + }, + { + "epoch": 4.05, + "learning_rate": 1.0301974211520012e-05, + "loss": 0.0175, + "step": 30883 + }, + { + "epoch": 4.05, + "learning_rate": 1.0301444127883127e-05, + "loss": 0.0199, + "step": 30884 + }, + { + "epoch": 4.05, + "learning_rate": 1.0300914043398444e-05, + "loss": 0.0242, + "step": 30885 + }, + { + "epoch": 4.05, + "learning_rate": 1.030038395806746e-05, + "loss": 0.0225, + "step": 30886 + }, + { + "epoch": 4.05, + "learning_rate": 1.029985387189166e-05, + "loss": 0.0229, + "step": 30887 + }, + { + "epoch": 4.05, + "learning_rate": 1.0299323784872534e-05, + "loss": 0.0217, + "step": 30888 + }, + { + "epoch": 4.05, + "learning_rate": 1.0298793697011577e-05, + "loss": 0.0229, + "step": 30889 + }, + { + "epoch": 4.05, + "learning_rate": 1.0298263608310284e-05, + "loss": 0.0247, + "step": 30890 + }, + { + "epoch": 4.05, + "learning_rate": 1.0297733518770135e-05, + "loss": 0.0207, + "step": 30891 + }, + { + "epoch": 4.05, + "learning_rate": 1.0297203428392627e-05, + "loss": 0.0258, + "step": 30892 + }, + { + "epoch": 4.05, + "learning_rate": 1.0296673337179251e-05, + "loss": 0.0226, + "step": 30893 + }, + { + "epoch": 4.05, + "learning_rate": 1.0296143245131498e-05, + "loss": 0.026, + "step": 30894 + }, + { + "epoch": 4.05, + "learning_rate": 1.0295613152250856e-05, + "loss": 0.0211, + "step": 30895 + }, + { + "epoch": 4.05, + "learning_rate": 1.0295083058538819e-05, + "loss": 0.022, + "step": 30896 + }, + { + "epoch": 4.05, + "learning_rate": 1.0294552963996878e-05, + "loss": 0.0206, + "step": 30897 + }, + { + "epoch": 4.05, + "learning_rate": 1.0294022868626521e-05, + "loss": 0.0328, + "step": 30898 + }, + { + "epoch": 4.05, + "learning_rate": 1.0293492772429238e-05, + "loss": 0.0199, + "step": 30899 + }, + { + "epoch": 4.05, + "learning_rate": 1.0292962675406526e-05, + "loss": 0.0192, + "step": 30900 + }, + { + "epoch": 4.05, + "learning_rate": 1.0292432577559872e-05, + "loss": 0.0164, + "step": 30901 + }, + { + "epoch": 4.05, + "learning_rate": 1.0291902478890766e-05, + "loss": 0.0194, + "step": 30902 + }, + { + "epoch": 4.05, + "learning_rate": 1.0291372379400698e-05, + "loss": 0.0271, + "step": 30903 + }, + { + "epoch": 4.05, + "learning_rate": 1.029084227909116e-05, + "loss": 0.0215, + "step": 30904 + }, + { + "epoch": 4.05, + "learning_rate": 1.0290312177963648e-05, + "loss": 0.0181, + "step": 30905 + }, + { + "epoch": 4.05, + "learning_rate": 1.0289782076019645e-05, + "loss": 0.0183, + "step": 30906 + }, + { + "epoch": 4.05, + "learning_rate": 1.0289251973260647e-05, + "loss": 0.0226, + "step": 30907 + }, + { + "epoch": 4.05, + "learning_rate": 1.0288721869688144e-05, + "loss": 0.0233, + "step": 30908 + }, + { + "epoch": 4.05, + "learning_rate": 1.0288191765303624e-05, + "loss": 0.0173, + "step": 30909 + }, + { + "epoch": 4.05, + "learning_rate": 1.028766166010858e-05, + "loss": 0.0221, + "step": 30910 + }, + { + "epoch": 4.05, + "learning_rate": 1.0287131554104504e-05, + "loss": 0.0207, + "step": 30911 + }, + { + "epoch": 4.05, + "learning_rate": 1.0286601447292885e-05, + "loss": 0.019, + "step": 30912 + }, + { + "epoch": 4.05, + "learning_rate": 1.0286071339675215e-05, + "loss": 0.0219, + "step": 30913 + }, + { + "epoch": 4.05, + "learning_rate": 1.0285541231252987e-05, + "loss": 0.023, + "step": 30914 + }, + { + "epoch": 4.05, + "learning_rate": 1.0285011122027685e-05, + "loss": 0.0214, + "step": 30915 + }, + { + "epoch": 4.05, + "learning_rate": 1.0284481012000806e-05, + "loss": 0.0273, + "step": 30916 + }, + { + "epoch": 4.05, + "learning_rate": 1.028395090117384e-05, + "loss": 0.0205, + "step": 30917 + }, + { + "epoch": 4.05, + "learning_rate": 1.0283420789548279e-05, + "loss": 0.0252, + "step": 30918 + }, + { + "epoch": 4.05, + "learning_rate": 1.028289067712561e-05, + "loss": 0.021, + "step": 30919 + }, + { + "epoch": 4.05, + "learning_rate": 1.0282360563907328e-05, + "loss": 0.0298, + "step": 30920 + }, + { + "epoch": 4.05, + "learning_rate": 1.0281830449894917e-05, + "loss": 0.0248, + "step": 30921 + }, + { + "epoch": 4.05, + "learning_rate": 1.0281300335089877e-05, + "loss": 0.0211, + "step": 30922 + }, + { + "epoch": 4.05, + "learning_rate": 1.0280770219493694e-05, + "loss": 0.022, + "step": 30923 + }, + { + "epoch": 4.05, + "learning_rate": 1.0280240103107862e-05, + "loss": 0.0211, + "step": 30924 + }, + { + "epoch": 4.05, + "learning_rate": 1.0279709985933866e-05, + "loss": 0.0263, + "step": 30925 + }, + { + "epoch": 4.05, + "learning_rate": 1.0279179867973202e-05, + "loss": 0.0242, + "step": 30926 + }, + { + "epoch": 4.05, + "learning_rate": 1.0278649749227358e-05, + "loss": 0.0206, + "step": 30927 + }, + { + "epoch": 4.05, + "learning_rate": 1.027811962969783e-05, + "loss": 0.0188, + "step": 30928 + }, + { + "epoch": 4.05, + "learning_rate": 1.0277589509386102e-05, + "loss": 0.0222, + "step": 30929 + }, + { + "epoch": 4.05, + "learning_rate": 1.027705938829367e-05, + "loss": 0.0227, + "step": 30930 + }, + { + "epoch": 4.05, + "learning_rate": 1.0276529266422021e-05, + "loss": 0.0237, + "step": 30931 + }, + { + "epoch": 4.05, + "learning_rate": 1.027599914377265e-05, + "loss": 0.0272, + "step": 30932 + }, + { + "epoch": 4.05, + "learning_rate": 1.0275469020347046e-05, + "loss": 0.0262, + "step": 30933 + }, + { + "epoch": 4.05, + "learning_rate": 1.0274938896146699e-05, + "loss": 0.0204, + "step": 30934 + }, + { + "epoch": 4.05, + "learning_rate": 1.0274408771173103e-05, + "loss": 0.0215, + "step": 30935 + }, + { + "epoch": 4.05, + "learning_rate": 1.0273878645427747e-05, + "loss": 0.0215, + "step": 30936 + }, + { + "epoch": 4.05, + "learning_rate": 1.0273348518912118e-05, + "loss": 0.0174, + "step": 30937 + }, + { + "epoch": 4.05, + "learning_rate": 1.0272818391627713e-05, + "loss": 0.0315, + "step": 30938 + }, + { + "epoch": 4.05, + "learning_rate": 1.0272288263576022e-05, + "loss": 0.023, + "step": 30939 + }, + { + "epoch": 4.05, + "learning_rate": 1.0271758134758536e-05, + "loss": 0.0193, + "step": 30940 + }, + { + "epoch": 4.05, + "learning_rate": 1.0271228005176743e-05, + "loss": 0.0211, + "step": 30941 + }, + { + "epoch": 4.05, + "learning_rate": 1.0270697874832136e-05, + "loss": 0.0205, + "step": 30942 + }, + { + "epoch": 4.05, + "learning_rate": 1.0270167743726205e-05, + "loss": 0.0186, + "step": 30943 + }, + { + "epoch": 4.05, + "learning_rate": 1.026963761186044e-05, + "loss": 0.0174, + "step": 30944 + }, + { + "epoch": 4.05, + "learning_rate": 1.0269107479236338e-05, + "loss": 0.0261, + "step": 30945 + }, + { + "epoch": 4.05, + "learning_rate": 1.0268577345855381e-05, + "loss": 0.0243, + "step": 30946 + }, + { + "epoch": 4.05, + "learning_rate": 1.0268047211719067e-05, + "loss": 0.0238, + "step": 30947 + }, + { + "epoch": 4.05, + "learning_rate": 1.0267517076828883e-05, + "loss": 0.0249, + "step": 30948 + }, + { + "epoch": 4.05, + "learning_rate": 1.0266986941186323e-05, + "loss": 0.0187, + "step": 30949 + }, + { + "epoch": 4.05, + "learning_rate": 1.0266456804792875e-05, + "loss": 0.0238, + "step": 30950 + }, + { + "epoch": 4.05, + "learning_rate": 1.0265926667650035e-05, + "loss": 0.0214, + "step": 30951 + }, + { + "epoch": 4.05, + "learning_rate": 1.0265396529759287e-05, + "loss": 0.023, + "step": 30952 + }, + { + "epoch": 4.05, + "learning_rate": 1.0264866391122126e-05, + "loss": 0.0179, + "step": 30953 + }, + { + "epoch": 4.05, + "learning_rate": 1.0264336251740041e-05, + "loss": 0.0302, + "step": 30954 + }, + { + "epoch": 4.05, + "learning_rate": 1.0263806111614527e-05, + "loss": 0.0242, + "step": 30955 + }, + { + "epoch": 4.06, + "learning_rate": 1.0263275970747072e-05, + "loss": 0.0228, + "step": 30956 + }, + { + "epoch": 4.06, + "learning_rate": 1.0262745829139165e-05, + "loss": 0.0241, + "step": 30957 + }, + { + "epoch": 4.06, + "learning_rate": 1.0262215686792302e-05, + "loss": 0.0231, + "step": 30958 + }, + { + "epoch": 4.06, + "learning_rate": 1.0261685543707966e-05, + "loss": 0.022, + "step": 30959 + }, + { + "epoch": 4.06, + "learning_rate": 1.026115539988766e-05, + "loss": 0.024, + "step": 30960 + }, + { + "epoch": 4.06, + "learning_rate": 1.0260625255332866e-05, + "loss": 0.0173, + "step": 30961 + }, + { + "epoch": 4.06, + "learning_rate": 1.0260095110045079e-05, + "loss": 0.0249, + "step": 30962 + }, + { + "epoch": 4.06, + "learning_rate": 1.0259564964025786e-05, + "loss": 0.0197, + "step": 30963 + }, + { + "epoch": 4.06, + "learning_rate": 1.025903481727648e-05, + "loss": 0.0217, + "step": 30964 + }, + { + "epoch": 4.06, + "learning_rate": 1.0258504669798653e-05, + "loss": 0.0221, + "step": 30965 + }, + { + "epoch": 4.06, + "learning_rate": 1.0257974521593796e-05, + "loss": 0.0305, + "step": 30966 + }, + { + "epoch": 4.06, + "learning_rate": 1.0257444372663399e-05, + "loss": 0.0252, + "step": 30967 + }, + { + "epoch": 4.06, + "learning_rate": 1.0256914223008955e-05, + "loss": 0.0286, + "step": 30968 + }, + { + "epoch": 4.06, + "learning_rate": 1.0256384072631952e-05, + "loss": 0.0269, + "step": 30969 + }, + { + "epoch": 4.06, + "learning_rate": 1.0255853921533882e-05, + "loss": 0.0193, + "step": 30970 + }, + { + "epoch": 4.06, + "learning_rate": 1.0255323769716237e-05, + "loss": 0.0247, + "step": 30971 + }, + { + "epoch": 4.06, + "learning_rate": 1.0254793617180507e-05, + "loss": 0.0203, + "step": 30972 + }, + { + "epoch": 4.06, + "learning_rate": 1.0254263463928185e-05, + "loss": 0.0229, + "step": 30973 + }, + { + "epoch": 4.06, + "learning_rate": 1.0253733309960758e-05, + "loss": 0.0238, + "step": 30974 + }, + { + "epoch": 4.06, + "learning_rate": 1.025320315527972e-05, + "loss": 0.0233, + "step": 30975 + }, + { + "epoch": 4.06, + "learning_rate": 1.0252672999886564e-05, + "loss": 0.0186, + "step": 30976 + }, + { + "epoch": 4.06, + "learning_rate": 1.0252142843782778e-05, + "loss": 0.0232, + "step": 30977 + }, + { + "epoch": 4.06, + "learning_rate": 1.0251612686969853e-05, + "loss": 0.0221, + "step": 30978 + }, + { + "epoch": 4.06, + "learning_rate": 1.025108252944928e-05, + "loss": 0.0268, + "step": 30979 + }, + { + "epoch": 4.06, + "learning_rate": 1.025055237122255e-05, + "loss": 0.0235, + "step": 30980 + }, + { + "epoch": 4.06, + "learning_rate": 1.0250022212291158e-05, + "loss": 0.023, + "step": 30981 + }, + { + "epoch": 4.06, + "learning_rate": 1.024949205265659e-05, + "loss": 0.0226, + "step": 30982 + }, + { + "epoch": 4.06, + "learning_rate": 1.024896189232034e-05, + "loss": 0.0267, + "step": 30983 + }, + { + "epoch": 4.06, + "learning_rate": 1.0248431731283896e-05, + "loss": 0.0221, + "step": 30984 + }, + { + "epoch": 4.06, + "learning_rate": 1.0247901569548754e-05, + "loss": 0.0189, + "step": 30985 + }, + { + "epoch": 4.06, + "learning_rate": 1.0247371407116395e-05, + "loss": 0.0216, + "step": 30986 + }, + { + "epoch": 4.06, + "learning_rate": 1.0246841243988324e-05, + "loss": 0.0235, + "step": 30987 + }, + { + "epoch": 4.06, + "learning_rate": 1.0246311080166023e-05, + "loss": 0.0194, + "step": 30988 + }, + { + "epoch": 4.06, + "learning_rate": 1.0245780915650986e-05, + "loss": 0.0241, + "step": 30989 + }, + { + "epoch": 4.06, + "learning_rate": 1.0245250750444704e-05, + "loss": 0.0208, + "step": 30990 + }, + { + "epoch": 4.06, + "learning_rate": 1.0244720584548667e-05, + "loss": 0.0226, + "step": 30991 + }, + { + "epoch": 4.06, + "learning_rate": 1.0244190417964361e-05, + "loss": 0.0237, + "step": 30992 + }, + { + "epoch": 4.06, + "learning_rate": 1.024366025069329e-05, + "loss": 0.0241, + "step": 30993 + }, + { + "epoch": 4.06, + "learning_rate": 1.0243130082736934e-05, + "loss": 0.0193, + "step": 30994 + }, + { + "epoch": 4.06, + "learning_rate": 1.0242599914096789e-05, + "loss": 0.0198, + "step": 30995 + }, + { + "epoch": 4.06, + "learning_rate": 1.0242069744774344e-05, + "loss": 0.0213, + "step": 30996 + }, + { + "epoch": 4.06, + "learning_rate": 1.024153957477109e-05, + "loss": 0.0261, + "step": 30997 + }, + { + "epoch": 4.06, + "learning_rate": 1.0241009404088518e-05, + "loss": 0.0217, + "step": 30998 + }, + { + "epoch": 4.06, + "learning_rate": 1.024047923272812e-05, + "loss": 0.0184, + "step": 30999 + }, + { + "epoch": 4.06, + "learning_rate": 1.0239949060691387e-05, + "loss": 0.0206, + "step": 31000 + }, + { + "epoch": 4.06, + "learning_rate": 1.0239418887979812e-05, + "loss": 0.0217, + "step": 31001 + }, + { + "epoch": 4.06, + "learning_rate": 1.0238888714594885e-05, + "loss": 0.0205, + "step": 31002 + }, + { + "epoch": 4.06, + "learning_rate": 1.0238358540538092e-05, + "loss": 0.0241, + "step": 31003 + }, + { + "epoch": 4.06, + "learning_rate": 1.0237828365810931e-05, + "loss": 0.0235, + "step": 31004 + }, + { + "epoch": 4.06, + "learning_rate": 1.0237298190414891e-05, + "loss": 0.0193, + "step": 31005 + }, + { + "epoch": 4.06, + "learning_rate": 1.0236768014351461e-05, + "loss": 0.0177, + "step": 31006 + }, + { + "epoch": 4.06, + "learning_rate": 1.0236237837622133e-05, + "loss": 0.025, + "step": 31007 + }, + { + "epoch": 4.06, + "learning_rate": 1.02357076602284e-05, + "loss": 0.0207, + "step": 31008 + }, + { + "epoch": 4.06, + "learning_rate": 1.0235177482171752e-05, + "loss": 0.0257, + "step": 31009 + }, + { + "epoch": 4.06, + "learning_rate": 1.023464730345368e-05, + "loss": 0.0198, + "step": 31010 + }, + { + "epoch": 4.06, + "learning_rate": 1.0234117124075674e-05, + "loss": 0.0245, + "step": 31011 + }, + { + "epoch": 4.06, + "learning_rate": 1.023358694403923e-05, + "loss": 0.022, + "step": 31012 + }, + { + "epoch": 4.06, + "learning_rate": 1.0233056763345828e-05, + "loss": 0.0244, + "step": 31013 + }, + { + "epoch": 4.06, + "learning_rate": 1.0232526581996968e-05, + "loss": 0.0255, + "step": 31014 + }, + { + "epoch": 4.06, + "learning_rate": 1.0231996399994142e-05, + "loss": 0.0208, + "step": 31015 + }, + { + "epoch": 4.06, + "learning_rate": 1.023146621733884e-05, + "loss": 0.0236, + "step": 31016 + }, + { + "epoch": 4.06, + "learning_rate": 1.0230936034032548e-05, + "loss": 0.019, + "step": 31017 + }, + { + "epoch": 4.06, + "learning_rate": 1.0230405850076762e-05, + "loss": 0.0191, + "step": 31018 + }, + { + "epoch": 4.06, + "learning_rate": 1.022987566547297e-05, + "loss": 0.024, + "step": 31019 + }, + { + "epoch": 4.06, + "learning_rate": 1.0229345480222668e-05, + "loss": 0.025, + "step": 31020 + }, + { + "epoch": 4.06, + "learning_rate": 1.0228815294327343e-05, + "loss": 0.0234, + "step": 31021 + }, + { + "epoch": 4.06, + "learning_rate": 1.0228285107788487e-05, + "loss": 0.0254, + "step": 31022 + }, + { + "epoch": 4.06, + "learning_rate": 1.022775492060759e-05, + "loss": 0.0246, + "step": 31023 + }, + { + "epoch": 4.06, + "learning_rate": 1.0227224732786144e-05, + "loss": 0.0216, + "step": 31024 + }, + { + "epoch": 4.06, + "learning_rate": 1.0226694544325644e-05, + "loss": 0.0249, + "step": 31025 + }, + { + "epoch": 4.06, + "learning_rate": 1.0226164355227575e-05, + "loss": 0.0258, + "step": 31026 + }, + { + "epoch": 4.06, + "learning_rate": 1.0225634165493435e-05, + "loss": 0.0171, + "step": 31027 + }, + { + "epoch": 4.06, + "learning_rate": 1.0225103975124706e-05, + "loss": 0.0267, + "step": 31028 + }, + { + "epoch": 4.06, + "learning_rate": 1.0224573784122887e-05, + "loss": 0.0232, + "step": 31029 + }, + { + "epoch": 4.06, + "learning_rate": 1.0224043592489462e-05, + "loss": 0.0232, + "step": 31030 + }, + { + "epoch": 4.06, + "learning_rate": 1.022351340022593e-05, + "loss": 0.0253, + "step": 31031 + }, + { + "epoch": 4.06, + "learning_rate": 1.0222983207333776e-05, + "loss": 0.0279, + "step": 31032 + }, + { + "epoch": 4.07, + "learning_rate": 1.0222453013814498e-05, + "loss": 0.025, + "step": 31033 + }, + { + "epoch": 4.07, + "learning_rate": 1.0221922819669578e-05, + "loss": 0.0219, + "step": 31034 + }, + { + "epoch": 4.07, + "learning_rate": 1.0221392624900515e-05, + "loss": 0.0205, + "step": 31035 + }, + { + "epoch": 4.07, + "learning_rate": 1.0220862429508795e-05, + "loss": 0.0238, + "step": 31036 + }, + { + "epoch": 4.07, + "learning_rate": 1.0220332233495913e-05, + "loss": 0.0178, + "step": 31037 + }, + { + "epoch": 4.07, + "learning_rate": 1.0219802036863357e-05, + "loss": 0.0207, + "step": 31038 + }, + { + "epoch": 4.07, + "learning_rate": 1.021927183961262e-05, + "loss": 0.0259, + "step": 31039 + }, + { + "epoch": 4.07, + "learning_rate": 1.0218741641745193e-05, + "loss": 0.0209, + "step": 31040 + }, + { + "epoch": 4.07, + "learning_rate": 1.0218211443262565e-05, + "loss": 0.0295, + "step": 31041 + }, + { + "epoch": 4.07, + "learning_rate": 1.021768124416623e-05, + "loss": 0.0198, + "step": 31042 + }, + { + "epoch": 4.07, + "learning_rate": 1.0217151044457678e-05, + "loss": 0.0242, + "step": 31043 + }, + { + "epoch": 4.07, + "learning_rate": 1.0216620844138403e-05, + "loss": 0.0217, + "step": 31044 + }, + { + "epoch": 4.07, + "learning_rate": 1.0216090643209888e-05, + "loss": 0.0163, + "step": 31045 + }, + { + "epoch": 4.07, + "learning_rate": 1.0215560441673633e-05, + "loss": 0.0233, + "step": 31046 + }, + { + "epoch": 4.07, + "learning_rate": 1.0215030239531126e-05, + "loss": 0.0217, + "step": 31047 + }, + { + "epoch": 4.07, + "learning_rate": 1.0214500036783856e-05, + "loss": 0.027, + "step": 31048 + }, + { + "epoch": 4.07, + "learning_rate": 1.0213969833433318e-05, + "loss": 0.0163, + "step": 31049 + }, + { + "epoch": 4.07, + "learning_rate": 1.0213439629481001e-05, + "loss": 0.0252, + "step": 31050 + }, + { + "epoch": 4.07, + "learning_rate": 1.0212909424928396e-05, + "loss": 0.0238, + "step": 31051 + }, + { + "epoch": 4.07, + "learning_rate": 1.0212379219776993e-05, + "loss": 0.0197, + "step": 31052 + }, + { + "epoch": 4.07, + "learning_rate": 1.0211849014028285e-05, + "loss": 0.0184, + "step": 31053 + }, + { + "epoch": 4.07, + "learning_rate": 1.0211318807683768e-05, + "loss": 0.0305, + "step": 31054 + }, + { + "epoch": 4.07, + "learning_rate": 1.0210788600744924e-05, + "loss": 0.0223, + "step": 31055 + }, + { + "epoch": 4.07, + "learning_rate": 1.0210258393213248e-05, + "loss": 0.023, + "step": 31056 + }, + { + "epoch": 4.07, + "learning_rate": 1.0209728185090231e-05, + "loss": 0.0169, + "step": 31057 + }, + { + "epoch": 4.07, + "learning_rate": 1.0209197976377366e-05, + "loss": 0.0268, + "step": 31058 + }, + { + "epoch": 4.07, + "learning_rate": 1.0208667767076141e-05, + "loss": 0.0258, + "step": 31059 + }, + { + "epoch": 4.07, + "learning_rate": 1.0208137557188052e-05, + "loss": 0.0206, + "step": 31060 + }, + { + "epoch": 4.07, + "learning_rate": 1.0207607346714585e-05, + "loss": 0.0229, + "step": 31061 + }, + { + "epoch": 4.07, + "learning_rate": 1.0207077135657236e-05, + "loss": 0.0225, + "step": 31062 + }, + { + "epoch": 4.07, + "learning_rate": 1.0206546924017489e-05, + "loss": 0.0187, + "step": 31063 + }, + { + "epoch": 4.07, + "learning_rate": 1.0206016711796842e-05, + "loss": 0.024, + "step": 31064 + }, + { + "epoch": 4.07, + "learning_rate": 1.0205486498996787e-05, + "loss": 0.0206, + "step": 31065 + }, + { + "epoch": 4.07, + "learning_rate": 1.0204956285618808e-05, + "loss": 0.0267, + "step": 31066 + }, + { + "epoch": 4.07, + "learning_rate": 1.0204426071664404e-05, + "loss": 0.0205, + "step": 31067 + }, + { + "epoch": 4.07, + "learning_rate": 1.0203895857135057e-05, + "loss": 0.0224, + "step": 31068 + }, + { + "epoch": 4.07, + "learning_rate": 1.0203365642032269e-05, + "loss": 0.0273, + "step": 31069 + }, + { + "epoch": 4.07, + "learning_rate": 1.0202835426357523e-05, + "loss": 0.0206, + "step": 31070 + }, + { + "epoch": 4.07, + "learning_rate": 1.0202305210112315e-05, + "loss": 0.021, + "step": 31071 + }, + { + "epoch": 4.07, + "learning_rate": 1.0201774993298134e-05, + "loss": 0.0169, + "step": 31072 + }, + { + "epoch": 4.07, + "learning_rate": 1.0201244775916473e-05, + "loss": 0.0221, + "step": 31073 + }, + { + "epoch": 4.07, + "learning_rate": 1.0200714557968817e-05, + "loss": 0.0262, + "step": 31074 + }, + { + "epoch": 4.07, + "learning_rate": 1.0200184339456667e-05, + "loss": 0.0189, + "step": 31075 + }, + { + "epoch": 4.07, + "learning_rate": 1.0199654120381507e-05, + "loss": 0.0292, + "step": 31076 + }, + { + "epoch": 4.07, + "learning_rate": 1.0199123900744831e-05, + "loss": 0.0197, + "step": 31077 + }, + { + "epoch": 4.07, + "learning_rate": 1.019859368054813e-05, + "loss": 0.0241, + "step": 31078 + }, + { + "epoch": 4.07, + "learning_rate": 1.0198063459792893e-05, + "loss": 0.0271, + "step": 31079 + }, + { + "epoch": 4.07, + "learning_rate": 1.0197533238480614e-05, + "loss": 0.0226, + "step": 31080 + }, + { + "epoch": 4.07, + "learning_rate": 1.0197003016612785e-05, + "loss": 0.0268, + "step": 31081 + }, + { + "epoch": 4.07, + "learning_rate": 1.0196472794190895e-05, + "loss": 0.0195, + "step": 31082 + }, + { + "epoch": 4.07, + "learning_rate": 1.0195942571216433e-05, + "loss": 0.022, + "step": 31083 + }, + { + "epoch": 4.07, + "learning_rate": 1.0195412347690896e-05, + "loss": 0.0247, + "step": 31084 + }, + { + "epoch": 4.07, + "learning_rate": 1.0194882123615768e-05, + "loss": 0.0231, + "step": 31085 + }, + { + "epoch": 4.07, + "learning_rate": 1.019435189899255e-05, + "loss": 0.0229, + "step": 31086 + }, + { + "epoch": 4.07, + "learning_rate": 1.0193821673822724e-05, + "loss": 0.0191, + "step": 31087 + }, + { + "epoch": 4.07, + "learning_rate": 1.0193291448107786e-05, + "loss": 0.0225, + "step": 31088 + }, + { + "epoch": 4.07, + "learning_rate": 1.0192761221849225e-05, + "loss": 0.0215, + "step": 31089 + }, + { + "epoch": 4.07, + "learning_rate": 1.0192230995048535e-05, + "loss": 0.0187, + "step": 31090 + }, + { + "epoch": 4.07, + "learning_rate": 1.0191700767707205e-05, + "loss": 0.0202, + "step": 31091 + }, + { + "epoch": 4.07, + "learning_rate": 1.0191170539826726e-05, + "loss": 0.0212, + "step": 31092 + }, + { + "epoch": 4.07, + "learning_rate": 1.019064031140859e-05, + "loss": 0.0204, + "step": 31093 + }, + { + "epoch": 4.07, + "learning_rate": 1.0190110082454292e-05, + "loss": 0.0274, + "step": 31094 + }, + { + "epoch": 4.07, + "learning_rate": 1.0189579852965313e-05, + "loss": 0.0242, + "step": 31095 + }, + { + "epoch": 4.07, + "learning_rate": 1.0189049622943157e-05, + "loss": 0.0223, + "step": 31096 + }, + { + "epoch": 4.07, + "learning_rate": 1.0188519392389305e-05, + "loss": 0.0242, + "step": 31097 + }, + { + "epoch": 4.07, + "learning_rate": 1.0187989161305254e-05, + "loss": 0.0225, + "step": 31098 + }, + { + "epoch": 4.07, + "learning_rate": 1.0187458929692492e-05, + "loss": 0.0219, + "step": 31099 + }, + { + "epoch": 4.07, + "learning_rate": 1.0186928697552514e-05, + "loss": 0.0194, + "step": 31100 + }, + { + "epoch": 4.07, + "learning_rate": 1.0186398464886807e-05, + "loss": 0.0223, + "step": 31101 + }, + { + "epoch": 4.07, + "learning_rate": 1.0185868231696865e-05, + "loss": 0.026, + "step": 31102 + }, + { + "epoch": 4.07, + "learning_rate": 1.0185337997984176e-05, + "loss": 0.0191, + "step": 31103 + }, + { + "epoch": 4.07, + "learning_rate": 1.0184807763750237e-05, + "loss": 0.0195, + "step": 31104 + }, + { + "epoch": 4.07, + "learning_rate": 1.0184277528996538e-05, + "loss": 0.0227, + "step": 31105 + }, + { + "epoch": 4.07, + "learning_rate": 1.0183747293724564e-05, + "loss": 0.02, + "step": 31106 + }, + { + "epoch": 4.07, + "learning_rate": 1.0183217057935813e-05, + "loss": 0.0204, + "step": 31107 + }, + { + "epoch": 4.07, + "learning_rate": 1.018268682163177e-05, + "loss": 0.0209, + "step": 31108 + }, + { + "epoch": 4.08, + "learning_rate": 1.0182156584813936e-05, + "loss": 0.02, + "step": 31109 + }, + { + "epoch": 4.08, + "learning_rate": 1.0181626347483791e-05, + "loss": 0.0223, + "step": 31110 + }, + { + "epoch": 4.08, + "learning_rate": 1.0181096109642837e-05, + "loss": 0.0245, + "step": 31111 + }, + { + "epoch": 4.08, + "learning_rate": 1.0180565871292552e-05, + "loss": 0.0258, + "step": 31112 + }, + { + "epoch": 4.08, + "learning_rate": 1.0180035632434441e-05, + "loss": 0.0194, + "step": 31113 + }, + { + "epoch": 4.08, + "learning_rate": 1.0179505393069988e-05, + "loss": 0.0208, + "step": 31114 + }, + { + "epoch": 4.08, + "learning_rate": 1.0178975153200688e-05, + "loss": 0.0245, + "step": 31115 + }, + { + "epoch": 4.08, + "learning_rate": 1.0178444912828025e-05, + "loss": 0.0232, + "step": 31116 + }, + { + "epoch": 4.08, + "learning_rate": 1.0177914671953496e-05, + "loss": 0.0231, + "step": 31117 + }, + { + "epoch": 4.08, + "learning_rate": 1.0177384430578594e-05, + "loss": 0.0211, + "step": 31118 + }, + { + "epoch": 4.08, + "learning_rate": 1.017685418870481e-05, + "loss": 0.0203, + "step": 31119 + }, + { + "epoch": 4.08, + "learning_rate": 1.0176323946333629e-05, + "loss": 0.017, + "step": 31120 + }, + { + "epoch": 4.08, + "learning_rate": 1.0175793703466547e-05, + "loss": 0.0232, + "step": 31121 + }, + { + "epoch": 4.08, + "learning_rate": 1.0175263460105056e-05, + "loss": 0.0189, + "step": 31122 + }, + { + "epoch": 4.08, + "learning_rate": 1.0174733216250643e-05, + "loss": 0.0274, + "step": 31123 + }, + { + "epoch": 4.08, + "learning_rate": 1.0174202971904803e-05, + "loss": 0.0218, + "step": 31124 + }, + { + "epoch": 4.08, + "learning_rate": 1.0173672727069027e-05, + "loss": 0.0257, + "step": 31125 + }, + { + "epoch": 4.08, + "learning_rate": 1.0173142481744807e-05, + "loss": 0.0221, + "step": 31126 + }, + { + "epoch": 4.08, + "learning_rate": 1.0172612235933632e-05, + "loss": 0.0284, + "step": 31127 + }, + { + "epoch": 4.08, + "learning_rate": 1.0172081989636992e-05, + "loss": 0.0205, + "step": 31128 + }, + { + "epoch": 4.08, + "learning_rate": 1.0171551742856382e-05, + "loss": 0.022, + "step": 31129 + }, + { + "epoch": 4.08, + "learning_rate": 1.0171021495593296e-05, + "loss": 0.0217, + "step": 31130 + }, + { + "epoch": 4.08, + "learning_rate": 1.0170491247849217e-05, + "loss": 0.0247, + "step": 31131 + }, + { + "epoch": 4.08, + "learning_rate": 1.0169960999625643e-05, + "loss": 0.0245, + "step": 31132 + }, + { + "epoch": 4.08, + "learning_rate": 1.0169430750924059e-05, + "loss": 0.0208, + "step": 31133 + }, + { + "epoch": 4.08, + "learning_rate": 1.0168900501745962e-05, + "loss": 0.0184, + "step": 31134 + }, + { + "epoch": 4.08, + "learning_rate": 1.016837025209284e-05, + "loss": 0.0189, + "step": 31135 + }, + { + "epoch": 4.08, + "learning_rate": 1.016784000196619e-05, + "loss": 0.0153, + "step": 31136 + }, + { + "epoch": 4.08, + "learning_rate": 1.0167309751367494e-05, + "loss": 0.0228, + "step": 31137 + }, + { + "epoch": 4.08, + "learning_rate": 1.0166779500298252e-05, + "loss": 0.0182, + "step": 31138 + }, + { + "epoch": 4.08, + "learning_rate": 1.0166249248759948e-05, + "loss": 0.0259, + "step": 31139 + }, + { + "epoch": 4.08, + "learning_rate": 1.0165718996754079e-05, + "loss": 0.0251, + "step": 31140 + }, + { + "epoch": 4.08, + "learning_rate": 1.0165188744282133e-05, + "loss": 0.0234, + "step": 31141 + }, + { + "epoch": 4.08, + "learning_rate": 1.0164658491345604e-05, + "loss": 0.0173, + "step": 31142 + }, + { + "epoch": 4.08, + "learning_rate": 1.0164128237945981e-05, + "loss": 0.0178, + "step": 31143 + }, + { + "epoch": 4.08, + "learning_rate": 1.0163597984084754e-05, + "loss": 0.0183, + "step": 31144 + }, + { + "epoch": 4.08, + "learning_rate": 1.0163067729763418e-05, + "loss": 0.0251, + "step": 31145 + }, + { + "epoch": 4.08, + "learning_rate": 1.0162537474983465e-05, + "loss": 0.0187, + "step": 31146 + }, + { + "epoch": 4.08, + "learning_rate": 1.0162007219746382e-05, + "loss": 0.0256, + "step": 31147 + }, + { + "epoch": 4.08, + "learning_rate": 1.0161476964053661e-05, + "loss": 0.0237, + "step": 31148 + }, + { + "epoch": 4.08, + "learning_rate": 1.0160946707906798e-05, + "loss": 0.0228, + "step": 31149 + }, + { + "epoch": 4.08, + "learning_rate": 1.0160416451307276e-05, + "loss": 0.0238, + "step": 31150 + }, + { + "epoch": 4.08, + "learning_rate": 1.0159886194256594e-05, + "loss": 0.0218, + "step": 31151 + }, + { + "epoch": 4.08, + "learning_rate": 1.0159355936756242e-05, + "loss": 0.0208, + "step": 31152 + }, + { + "epoch": 4.08, + "learning_rate": 1.015882567880771e-05, + "loss": 0.0192, + "step": 31153 + }, + { + "epoch": 4.08, + "learning_rate": 1.0158295420412488e-05, + "loss": 0.0201, + "step": 31154 + }, + { + "epoch": 4.08, + "learning_rate": 1.0157765161572067e-05, + "loss": 0.0166, + "step": 31155 + }, + { + "epoch": 4.08, + "learning_rate": 1.015723490228794e-05, + "loss": 0.0228, + "step": 31156 + }, + { + "epoch": 4.08, + "learning_rate": 1.01567046425616e-05, + "loss": 0.0254, + "step": 31157 + }, + { + "epoch": 4.08, + "learning_rate": 1.0156174382394537e-05, + "loss": 0.0226, + "step": 31158 + }, + { + "epoch": 4.08, + "learning_rate": 1.0155644121788242e-05, + "loss": 0.0187, + "step": 31159 + }, + { + "epoch": 4.08, + "learning_rate": 1.0155113860744206e-05, + "loss": 0.0209, + "step": 31160 + }, + { + "epoch": 4.08, + "learning_rate": 1.0154583599263919e-05, + "loss": 0.0244, + "step": 31161 + }, + { + "epoch": 4.08, + "learning_rate": 1.0154053337348872e-05, + "loss": 0.0208, + "step": 31162 + }, + { + "epoch": 4.08, + "learning_rate": 1.0153523075000564e-05, + "loss": 0.0222, + "step": 31163 + }, + { + "epoch": 4.08, + "learning_rate": 1.0152992812220477e-05, + "loss": 0.0203, + "step": 31164 + }, + { + "epoch": 4.08, + "learning_rate": 1.0152462549010104e-05, + "loss": 0.0218, + "step": 31165 + }, + { + "epoch": 4.08, + "learning_rate": 1.015193228537094e-05, + "loss": 0.0239, + "step": 31166 + }, + { + "epoch": 4.08, + "learning_rate": 1.0151402021304477e-05, + "loss": 0.0235, + "step": 31167 + }, + { + "epoch": 4.08, + "learning_rate": 1.0150871756812202e-05, + "loss": 0.0276, + "step": 31168 + }, + { + "epoch": 4.08, + "learning_rate": 1.0150341491895606e-05, + "loss": 0.0258, + "step": 31169 + }, + { + "epoch": 4.08, + "learning_rate": 1.0149811226556186e-05, + "loss": 0.0243, + "step": 31170 + }, + { + "epoch": 4.08, + "learning_rate": 1.0149280960795427e-05, + "loss": 0.0184, + "step": 31171 + }, + { + "epoch": 4.08, + "learning_rate": 1.0148750694614823e-05, + "loss": 0.0229, + "step": 31172 + }, + { + "epoch": 4.08, + "learning_rate": 1.0148220428015869e-05, + "loss": 0.0285, + "step": 31173 + }, + { + "epoch": 4.08, + "learning_rate": 1.0147690161000052e-05, + "loss": 0.0224, + "step": 31174 + }, + { + "epoch": 4.08, + "learning_rate": 1.0147159893568862e-05, + "loss": 0.0245, + "step": 31175 + }, + { + "epoch": 4.08, + "learning_rate": 1.0146629625723795e-05, + "loss": 0.0238, + "step": 31176 + }, + { + "epoch": 4.08, + "learning_rate": 1.0146099357466338e-05, + "loss": 0.0195, + "step": 31177 + }, + { + "epoch": 4.08, + "learning_rate": 1.0145569088797986e-05, + "loss": 0.0213, + "step": 31178 + }, + { + "epoch": 4.08, + "learning_rate": 1.0145038819720228e-05, + "loss": 0.0204, + "step": 31179 + }, + { + "epoch": 4.08, + "learning_rate": 1.0144508550234557e-05, + "loss": 0.0257, + "step": 31180 + }, + { + "epoch": 4.08, + "learning_rate": 1.0143978280342464e-05, + "loss": 0.0233, + "step": 31181 + }, + { + "epoch": 4.08, + "learning_rate": 1.014344801004544e-05, + "loss": 0.0167, + "step": 31182 + }, + { + "epoch": 4.08, + "learning_rate": 1.014291773934497e-05, + "loss": 0.026, + "step": 31183 + }, + { + "epoch": 4.08, + "learning_rate": 1.0142387468242558e-05, + "loss": 0.0276, + "step": 31184 + }, + { + "epoch": 4.09, + "learning_rate": 1.0141857196739686e-05, + "loss": 0.0235, + "step": 31185 + }, + { + "epoch": 4.09, + "learning_rate": 1.0141326924837852e-05, + "loss": 0.0154, + "step": 31186 + }, + { + "epoch": 4.09, + "learning_rate": 1.0140796652538538e-05, + "loss": 0.023, + "step": 31187 + }, + { + "epoch": 4.09, + "learning_rate": 1.0140266379843244e-05, + "loss": 0.0221, + "step": 31188 + }, + { + "epoch": 4.09, + "learning_rate": 1.0139736106753457e-05, + "loss": 0.0185, + "step": 31189 + }, + { + "epoch": 4.09, + "learning_rate": 1.0139205833270669e-05, + "loss": 0.0185, + "step": 31190 + }, + { + "epoch": 4.09, + "learning_rate": 1.0138675559396376e-05, + "loss": 0.0229, + "step": 31191 + }, + { + "epoch": 4.09, + "learning_rate": 1.0138145285132061e-05, + "loss": 0.0184, + "step": 31192 + }, + { + "epoch": 4.09, + "learning_rate": 1.0137615010479223e-05, + "loss": 0.0185, + "step": 31193 + }, + { + "epoch": 4.09, + "learning_rate": 1.0137084735439346e-05, + "loss": 0.0233, + "step": 31194 + }, + { + "epoch": 4.09, + "learning_rate": 1.013655446001393e-05, + "loss": 0.0244, + "step": 31195 + }, + { + "epoch": 4.09, + "learning_rate": 1.0136024184204458e-05, + "loss": 0.026, + "step": 31196 + }, + { + "epoch": 4.09, + "learning_rate": 1.013549390801243e-05, + "loss": 0.0273, + "step": 31197 + }, + { + "epoch": 4.09, + "learning_rate": 1.0134963631439329e-05, + "loss": 0.0215, + "step": 31198 + }, + { + "epoch": 4.09, + "learning_rate": 1.013443335448665e-05, + "loss": 0.0215, + "step": 31199 + }, + { + "epoch": 4.09, + "learning_rate": 1.0133903077155885e-05, + "loss": 0.0223, + "step": 31200 + }, + { + "epoch": 4.09, + "learning_rate": 1.0133372799448527e-05, + "loss": 0.025, + "step": 31201 + }, + { + "epoch": 4.09, + "learning_rate": 1.0132842521366059e-05, + "loss": 0.0193, + "step": 31202 + }, + { + "epoch": 4.09, + "learning_rate": 1.0132312242909985e-05, + "loss": 0.0207, + "step": 31203 + }, + { + "epoch": 4.09, + "learning_rate": 1.0131781964081789e-05, + "loss": 0.0244, + "step": 31204 + }, + { + "epoch": 4.09, + "learning_rate": 1.0131251684882959e-05, + "loss": 0.0207, + "step": 31205 + }, + { + "epoch": 4.09, + "learning_rate": 1.0130721405314991e-05, + "loss": 0.0214, + "step": 31206 + }, + { + "epoch": 4.09, + "learning_rate": 1.0130191125379381e-05, + "loss": 0.0255, + "step": 31207 + }, + { + "epoch": 4.09, + "learning_rate": 1.0129660845077612e-05, + "loss": 0.0227, + "step": 31208 + }, + { + "epoch": 4.09, + "learning_rate": 1.0129130564411178e-05, + "loss": 0.0215, + "step": 31209 + }, + { + "epoch": 4.09, + "learning_rate": 1.0128600283381574e-05, + "loss": 0.0197, + "step": 31210 + }, + { + "epoch": 4.09, + "learning_rate": 1.0128070001990287e-05, + "loss": 0.0169, + "step": 31211 + }, + { + "epoch": 4.09, + "learning_rate": 1.0127539720238807e-05, + "loss": 0.021, + "step": 31212 + }, + { + "epoch": 4.09, + "learning_rate": 1.0127009438128632e-05, + "loss": 0.027, + "step": 31213 + }, + { + "epoch": 4.09, + "learning_rate": 1.012647915566125e-05, + "loss": 0.0201, + "step": 31214 + }, + { + "epoch": 4.09, + "learning_rate": 1.0125948872838152e-05, + "loss": 0.0229, + "step": 31215 + }, + { + "epoch": 4.09, + "learning_rate": 1.0125418589660826e-05, + "loss": 0.021, + "step": 31216 + }, + { + "epoch": 4.09, + "learning_rate": 1.012488830613077e-05, + "loss": 0.0272, + "step": 31217 + }, + { + "epoch": 4.09, + "learning_rate": 1.0124358022249474e-05, + "loss": 0.0229, + "step": 31218 + }, + { + "epoch": 4.09, + "learning_rate": 1.0123827738018426e-05, + "loss": 0.0245, + "step": 31219 + }, + { + "epoch": 4.09, + "learning_rate": 1.0123297453439119e-05, + "loss": 0.0221, + "step": 31220 + }, + { + "epoch": 4.09, + "learning_rate": 1.0122767168513041e-05, + "loss": 0.0216, + "step": 31221 + }, + { + "epoch": 4.09, + "learning_rate": 1.0122236883241691e-05, + "loss": 0.0173, + "step": 31222 + }, + { + "epoch": 4.09, + "learning_rate": 1.0121706597626555e-05, + "loss": 0.0286, + "step": 31223 + }, + { + "epoch": 4.09, + "learning_rate": 1.0121176311669127e-05, + "loss": 0.0225, + "step": 31224 + }, + { + "epoch": 4.09, + "learning_rate": 1.0120646025370895e-05, + "loss": 0.024, + "step": 31225 + }, + { + "epoch": 4.09, + "learning_rate": 1.0120115738733354e-05, + "loss": 0.0243, + "step": 31226 + }, + { + "epoch": 4.09, + "learning_rate": 1.0119585451757991e-05, + "loss": 0.0193, + "step": 31227 + }, + { + "epoch": 4.09, + "learning_rate": 1.0119055164446305e-05, + "loss": 0.0279, + "step": 31228 + }, + { + "epoch": 4.09, + "learning_rate": 1.011852487679978e-05, + "loss": 0.0219, + "step": 31229 + }, + { + "epoch": 4.09, + "learning_rate": 1.011799458881991e-05, + "loss": 0.0257, + "step": 31230 + }, + { + "epoch": 4.09, + "learning_rate": 1.0117464300508188e-05, + "loss": 0.0248, + "step": 31231 + }, + { + "epoch": 4.09, + "learning_rate": 1.0116934011866103e-05, + "loss": 0.0186, + "step": 31232 + }, + { + "epoch": 4.09, + "learning_rate": 1.0116403722895144e-05, + "loss": 0.0235, + "step": 31233 + }, + { + "epoch": 4.09, + "learning_rate": 1.011587343359681e-05, + "loss": 0.0195, + "step": 31234 + }, + { + "epoch": 4.09, + "learning_rate": 1.011534314397259e-05, + "loss": 0.0195, + "step": 31235 + }, + { + "epoch": 4.09, + "learning_rate": 1.0114812854023967e-05, + "loss": 0.0239, + "step": 31236 + }, + { + "epoch": 4.09, + "learning_rate": 1.0114282563752442e-05, + "loss": 0.0263, + "step": 31237 + }, + { + "epoch": 4.09, + "learning_rate": 1.0113752273159504e-05, + "loss": 0.0269, + "step": 31238 + }, + { + "epoch": 4.09, + "learning_rate": 1.0113221982246645e-05, + "loss": 0.0208, + "step": 31239 + }, + { + "epoch": 4.09, + "learning_rate": 1.0112691691015353e-05, + "loss": 0.022, + "step": 31240 + }, + { + "epoch": 4.09, + "learning_rate": 1.0112161399467124e-05, + "loss": 0.024, + "step": 31241 + }, + { + "epoch": 4.09, + "learning_rate": 1.0111631107603445e-05, + "loss": 0.0241, + "step": 31242 + }, + { + "epoch": 4.09, + "learning_rate": 1.011110081542581e-05, + "loss": 0.0232, + "step": 31243 + }, + { + "epoch": 4.09, + "learning_rate": 1.0110570522935708e-05, + "loss": 0.0217, + "step": 31244 + }, + { + "epoch": 4.09, + "learning_rate": 1.0110040230134636e-05, + "loss": 0.0211, + "step": 31245 + }, + { + "epoch": 4.09, + "learning_rate": 1.010950993702408e-05, + "loss": 0.0236, + "step": 31246 + }, + { + "epoch": 4.09, + "learning_rate": 1.0108979643605534e-05, + "loss": 0.0236, + "step": 31247 + }, + { + "epoch": 4.09, + "learning_rate": 1.0108449349880485e-05, + "loss": 0.0234, + "step": 31248 + }, + { + "epoch": 4.09, + "learning_rate": 1.0107919055850433e-05, + "loss": 0.0216, + "step": 31249 + }, + { + "epoch": 4.09, + "learning_rate": 1.0107388761516861e-05, + "loss": 0.0244, + "step": 31250 + }, + { + "epoch": 4.09, + "learning_rate": 1.0106858466881262e-05, + "loss": 0.0177, + "step": 31251 + }, + { + "epoch": 4.09, + "learning_rate": 1.0106328171945134e-05, + "loss": 0.0242, + "step": 31252 + }, + { + "epoch": 4.09, + "learning_rate": 1.0105797876709962e-05, + "loss": 0.019, + "step": 31253 + }, + { + "epoch": 4.09, + "learning_rate": 1.0105267581177238e-05, + "loss": 0.0183, + "step": 31254 + }, + { + "epoch": 4.09, + "learning_rate": 1.0104737285348454e-05, + "loss": 0.0299, + "step": 31255 + }, + { + "epoch": 4.09, + "learning_rate": 1.0104206989225105e-05, + "loss": 0.0197, + "step": 31256 + }, + { + "epoch": 4.09, + "learning_rate": 1.0103676692808677e-05, + "loss": 0.0193, + "step": 31257 + }, + { + "epoch": 4.09, + "learning_rate": 1.0103146396100667e-05, + "loss": 0.0303, + "step": 31258 + }, + { + "epoch": 4.09, + "learning_rate": 1.0102616099102558e-05, + "loss": 0.023, + "step": 31259 + }, + { + "epoch": 4.09, + "learning_rate": 1.010208580181585e-05, + "loss": 0.0193, + "step": 31260 + }, + { + "epoch": 4.09, + "learning_rate": 1.0101555504242029e-05, + "loss": 0.0244, + "step": 31261 + }, + { + "epoch": 4.1, + "learning_rate": 1.0101025206382593e-05, + "loss": 0.0242, + "step": 31262 + }, + { + "epoch": 4.1, + "learning_rate": 1.0100494908239024e-05, + "loss": 0.0167, + "step": 31263 + }, + { + "epoch": 4.1, + "learning_rate": 1.0099964609812822e-05, + "loss": 0.0209, + "step": 31264 + }, + { + "epoch": 4.1, + "learning_rate": 1.0099434311105472e-05, + "loss": 0.0286, + "step": 31265 + }, + { + "epoch": 4.1, + "learning_rate": 1.0098904012118472e-05, + "loss": 0.0223, + "step": 31266 + }, + { + "epoch": 4.1, + "learning_rate": 1.0098373712853304e-05, + "loss": 0.0207, + "step": 31267 + }, + { + "epoch": 4.1, + "learning_rate": 1.0097843413311473e-05, + "loss": 0.0263, + "step": 31268 + }, + { + "epoch": 4.1, + "learning_rate": 1.0097313113494456e-05, + "loss": 0.0277, + "step": 31269 + }, + { + "epoch": 4.1, + "learning_rate": 1.009678281340375e-05, + "loss": 0.0205, + "step": 31270 + }, + { + "epoch": 4.1, + "learning_rate": 1.0096252513040852e-05, + "loss": 0.0211, + "step": 31271 + }, + { + "epoch": 4.1, + "learning_rate": 1.0095722212407251e-05, + "loss": 0.0239, + "step": 31272 + }, + { + "epoch": 4.1, + "learning_rate": 1.0095191911504431e-05, + "loss": 0.0172, + "step": 31273 + }, + { + "epoch": 4.1, + "learning_rate": 1.0094661610333891e-05, + "loss": 0.0238, + "step": 31274 + }, + { + "epoch": 4.1, + "learning_rate": 1.0094131308897122e-05, + "loss": 0.0252, + "step": 31275 + }, + { + "epoch": 4.1, + "learning_rate": 1.0093601007195608e-05, + "loss": 0.0223, + "step": 31276 + }, + { + "epoch": 4.1, + "learning_rate": 1.0093070705230851e-05, + "loss": 0.0287, + "step": 31277 + }, + { + "epoch": 4.1, + "learning_rate": 1.0092540403004336e-05, + "loss": 0.0245, + "step": 31278 + }, + { + "epoch": 4.1, + "learning_rate": 1.0092010100517557e-05, + "loss": 0.0221, + "step": 31279 + }, + { + "epoch": 4.1, + "learning_rate": 1.0091479797772005e-05, + "loss": 0.0238, + "step": 31280 + }, + { + "epoch": 4.1, + "learning_rate": 1.0090949494769169e-05, + "loss": 0.0265, + "step": 31281 + }, + { + "epoch": 4.1, + "learning_rate": 1.0090419191510544e-05, + "loss": 0.0204, + "step": 31282 + }, + { + "epoch": 4.1, + "learning_rate": 1.008988888799762e-05, + "loss": 0.0237, + "step": 31283 + }, + { + "epoch": 4.1, + "learning_rate": 1.0089358584231887e-05, + "loss": 0.0293, + "step": 31284 + }, + { + "epoch": 4.1, + "learning_rate": 1.0088828280214839e-05, + "loss": 0.0226, + "step": 31285 + }, + { + "epoch": 4.1, + "learning_rate": 1.0088297975947965e-05, + "loss": 0.0187, + "step": 31286 + }, + { + "epoch": 4.1, + "learning_rate": 1.0087767671432758e-05, + "loss": 0.023, + "step": 31287 + }, + { + "epoch": 4.1, + "learning_rate": 1.0087237366670709e-05, + "loss": 0.0232, + "step": 31288 + }, + { + "epoch": 4.1, + "learning_rate": 1.008670706166331e-05, + "loss": 0.0293, + "step": 31289 + }, + { + "epoch": 4.1, + "learning_rate": 1.0086176756412051e-05, + "loss": 0.0228, + "step": 31290 + }, + { + "epoch": 4.1, + "learning_rate": 1.0085646450918424e-05, + "loss": 0.0192, + "step": 31291 + }, + { + "epoch": 4.1, + "learning_rate": 1.0085116145183922e-05, + "loss": 0.0205, + "step": 31292 + }, + { + "epoch": 4.1, + "learning_rate": 1.0084585839210038e-05, + "loss": 0.0181, + "step": 31293 + }, + { + "epoch": 4.1, + "learning_rate": 1.0084055532998258e-05, + "loss": 0.0222, + "step": 31294 + }, + { + "epoch": 4.1, + "learning_rate": 1.0083525226550074e-05, + "loss": 0.0214, + "step": 31295 + }, + { + "epoch": 4.1, + "learning_rate": 1.0082994919866986e-05, + "loss": 0.0257, + "step": 31296 + }, + { + "epoch": 4.1, + "learning_rate": 1.0082464612950475e-05, + "loss": 0.0231, + "step": 31297 + }, + { + "epoch": 4.1, + "learning_rate": 1.0081934305802037e-05, + "loss": 0.0188, + "step": 31298 + }, + { + "epoch": 4.1, + "learning_rate": 1.0081403998423162e-05, + "loss": 0.0235, + "step": 31299 + }, + { + "epoch": 4.1, + "learning_rate": 1.0080873690815345e-05, + "loss": 0.0225, + "step": 31300 + }, + { + "epoch": 4.1, + "learning_rate": 1.0080343382980074e-05, + "loss": 0.024, + "step": 31301 + }, + { + "epoch": 4.1, + "learning_rate": 1.0079813074918843e-05, + "loss": 0.0216, + "step": 31302 + }, + { + "epoch": 4.1, + "learning_rate": 1.0079282766633138e-05, + "loss": 0.0261, + "step": 31303 + }, + { + "epoch": 4.1, + "learning_rate": 1.0078752458124459e-05, + "loss": 0.0226, + "step": 31304 + }, + { + "epoch": 4.1, + "learning_rate": 1.007822214939429e-05, + "loss": 0.0193, + "step": 31305 + }, + { + "epoch": 4.1, + "learning_rate": 1.0077691840444127e-05, + "loss": 0.0238, + "step": 31306 + }, + { + "epoch": 4.1, + "learning_rate": 1.0077161531275459e-05, + "loss": 0.0277, + "step": 31307 + }, + { + "epoch": 4.1, + "learning_rate": 1.0076631221889777e-05, + "loss": 0.0227, + "step": 31308 + }, + { + "epoch": 4.1, + "learning_rate": 1.0076100912288576e-05, + "loss": 0.0237, + "step": 31309 + }, + { + "epoch": 4.1, + "learning_rate": 1.0075570602473346e-05, + "loss": 0.0203, + "step": 31310 + }, + { + "epoch": 4.1, + "learning_rate": 1.0075040292445574e-05, + "loss": 0.0193, + "step": 31311 + }, + { + "epoch": 4.1, + "learning_rate": 1.0074509982206759e-05, + "loss": 0.0231, + "step": 31312 + }, + { + "epoch": 4.1, + "learning_rate": 1.0073979671758388e-05, + "loss": 0.0261, + "step": 31313 + }, + { + "epoch": 4.1, + "learning_rate": 1.007344936110195e-05, + "loss": 0.0274, + "step": 31314 + }, + { + "epoch": 4.1, + "learning_rate": 1.007291905023894e-05, + "loss": 0.0269, + "step": 31315 + }, + { + "epoch": 4.1, + "learning_rate": 1.0072388739170851e-05, + "loss": 0.0283, + "step": 31316 + }, + { + "epoch": 4.1, + "learning_rate": 1.0071858427899172e-05, + "loss": 0.0192, + "step": 31317 + }, + { + "epoch": 4.1, + "learning_rate": 1.0071328116425396e-05, + "loss": 0.0259, + "step": 31318 + }, + { + "epoch": 4.1, + "learning_rate": 1.0070797804751011e-05, + "loss": 0.0239, + "step": 31319 + }, + { + "epoch": 4.1, + "learning_rate": 1.0070267492877512e-05, + "loss": 0.022, + "step": 31320 + }, + { + "epoch": 4.1, + "learning_rate": 1.0069737180806393e-05, + "loss": 0.0216, + "step": 31321 + }, + { + "epoch": 4.1, + "learning_rate": 1.0069206868539139e-05, + "loss": 0.0229, + "step": 31322 + }, + { + "epoch": 4.1, + "learning_rate": 1.0068676556077243e-05, + "loss": 0.0215, + "step": 31323 + }, + { + "epoch": 4.1, + "learning_rate": 1.00681462434222e-05, + "loss": 0.0191, + "step": 31324 + }, + { + "epoch": 4.1, + "learning_rate": 1.0067615930575498e-05, + "loss": 0.0226, + "step": 31325 + }, + { + "epoch": 4.1, + "learning_rate": 1.006708561753863e-05, + "loss": 0.0269, + "step": 31326 + }, + { + "epoch": 4.1, + "learning_rate": 1.0066555304313088e-05, + "loss": 0.0181, + "step": 31327 + }, + { + "epoch": 4.1, + "learning_rate": 1.0066024990900362e-05, + "loss": 0.0232, + "step": 31328 + }, + { + "epoch": 4.1, + "learning_rate": 1.0065494677301947e-05, + "loss": 0.0216, + "step": 31329 + }, + { + "epoch": 4.1, + "learning_rate": 1.0064964363519325e-05, + "loss": 0.0225, + "step": 31330 + }, + { + "epoch": 4.1, + "learning_rate": 1.0064434049554002e-05, + "loss": 0.0255, + "step": 31331 + }, + { + "epoch": 4.1, + "learning_rate": 1.0063903735407456e-05, + "loss": 0.0197, + "step": 31332 + }, + { + "epoch": 4.1, + "learning_rate": 1.0063373421081188e-05, + "loss": 0.0225, + "step": 31333 + }, + { + "epoch": 4.1, + "learning_rate": 1.0062843106576683e-05, + "loss": 0.0267, + "step": 31334 + }, + { + "epoch": 4.1, + "learning_rate": 1.0062312791895434e-05, + "loss": 0.0225, + "step": 31335 + }, + { + "epoch": 4.1, + "learning_rate": 1.0061782477038937e-05, + "loss": 0.0213, + "step": 31336 + }, + { + "epoch": 4.1, + "learning_rate": 1.0061252162008678e-05, + "loss": 0.0265, + "step": 31337 + }, + { + "epoch": 4.11, + "learning_rate": 1.0060721846806152e-05, + "loss": 0.0239, + "step": 31338 + }, + { + "epoch": 4.11, + "learning_rate": 1.0060191531432848e-05, + "loss": 0.0226, + "step": 31339 + }, + { + "epoch": 4.11, + "learning_rate": 1.005966121589026e-05, + "loss": 0.0213, + "step": 31340 + }, + { + "epoch": 4.11, + "learning_rate": 1.0059130900179874e-05, + "loss": 0.0269, + "step": 31341 + }, + { + "epoch": 4.11, + "learning_rate": 1.0058600584303189e-05, + "loss": 0.0263, + "step": 31342 + }, + { + "epoch": 4.11, + "learning_rate": 1.0058070268261693e-05, + "loss": 0.0225, + "step": 31343 + }, + { + "epoch": 4.11, + "learning_rate": 1.0057539952056876e-05, + "loss": 0.0216, + "step": 31344 + }, + { + "epoch": 4.11, + "learning_rate": 1.0057009635690231e-05, + "loss": 0.0227, + "step": 31345 + }, + { + "epoch": 4.11, + "learning_rate": 1.0056479319163253e-05, + "loss": 0.0237, + "step": 31346 + }, + { + "epoch": 4.11, + "learning_rate": 1.0055949002477423e-05, + "loss": 0.0233, + "step": 31347 + }, + { + "epoch": 4.11, + "learning_rate": 1.0055418685634246e-05, + "loss": 0.0168, + "step": 31348 + }, + { + "epoch": 4.11, + "learning_rate": 1.0054888368635203e-05, + "loss": 0.0212, + "step": 31349 + }, + { + "epoch": 4.11, + "learning_rate": 1.0054358051481792e-05, + "loss": 0.023, + "step": 31350 + }, + { + "epoch": 4.11, + "learning_rate": 1.0053827734175501e-05, + "loss": 0.0223, + "step": 31351 + }, + { + "epoch": 4.11, + "learning_rate": 1.005329741671782e-05, + "loss": 0.0221, + "step": 31352 + }, + { + "epoch": 4.11, + "learning_rate": 1.0052767099110245e-05, + "loss": 0.0198, + "step": 31353 + }, + { + "epoch": 4.11, + "learning_rate": 1.0052236781354265e-05, + "loss": 0.0183, + "step": 31354 + }, + { + "epoch": 4.11, + "learning_rate": 1.0051706463451372e-05, + "loss": 0.0271, + "step": 31355 + }, + { + "epoch": 4.11, + "learning_rate": 1.0051176145403057e-05, + "loss": 0.0218, + "step": 31356 + }, + { + "epoch": 4.11, + "learning_rate": 1.0050645827210812e-05, + "loss": 0.0251, + "step": 31357 + }, + { + "epoch": 4.11, + "learning_rate": 1.005011550887613e-05, + "loss": 0.0217, + "step": 31358 + }, + { + "epoch": 4.11, + "learning_rate": 1.00495851904005e-05, + "loss": 0.0184, + "step": 31359 + }, + { + "epoch": 4.11, + "learning_rate": 1.004905487178541e-05, + "loss": 0.0174, + "step": 31360 + }, + { + "epoch": 4.11, + "learning_rate": 1.0048524553032361e-05, + "loss": 0.0205, + "step": 31361 + }, + { + "epoch": 4.11, + "learning_rate": 1.0047994234142837e-05, + "loss": 0.023, + "step": 31362 + }, + { + "epoch": 4.11, + "learning_rate": 1.0047463915118332e-05, + "loss": 0.0209, + "step": 31363 + }, + { + "epoch": 4.11, + "learning_rate": 1.0046933595960337e-05, + "loss": 0.0213, + "step": 31364 + }, + { + "epoch": 4.11, + "learning_rate": 1.0046403276670347e-05, + "loss": 0.0206, + "step": 31365 + }, + { + "epoch": 4.11, + "learning_rate": 1.0045872957249847e-05, + "loss": 0.0244, + "step": 31366 + }, + { + "epoch": 4.11, + "learning_rate": 1.0045342637700333e-05, + "loss": 0.0216, + "step": 31367 + }, + { + "epoch": 4.11, + "learning_rate": 1.004481231802329e-05, + "loss": 0.0212, + "step": 31368 + }, + { + "epoch": 4.11, + "learning_rate": 1.0044281998220222e-05, + "loss": 0.0212, + "step": 31369 + }, + { + "epoch": 4.11, + "learning_rate": 1.004375167829261e-05, + "loss": 0.0232, + "step": 31370 + }, + { + "epoch": 4.11, + "learning_rate": 1.0043221358241953e-05, + "loss": 0.0238, + "step": 31371 + }, + { + "epoch": 4.11, + "learning_rate": 1.0042691038069731e-05, + "loss": 0.0221, + "step": 31372 + }, + { + "epoch": 4.11, + "learning_rate": 1.004216071777745e-05, + "loss": 0.0175, + "step": 31373 + }, + { + "epoch": 4.11, + "learning_rate": 1.0041630397366585e-05, + "loss": 0.0222, + "step": 31374 + }, + { + "epoch": 4.11, + "learning_rate": 1.0041100076838646e-05, + "loss": 0.0226, + "step": 31375 + }, + { + "epoch": 4.11, + "learning_rate": 1.0040569756195112e-05, + "loss": 0.0216, + "step": 31376 + }, + { + "epoch": 4.11, + "learning_rate": 1.0040039435437473e-05, + "loss": 0.0247, + "step": 31377 + }, + { + "epoch": 4.11, + "learning_rate": 1.0039509114567234e-05, + "loss": 0.0253, + "step": 31378 + }, + { + "epoch": 4.11, + "learning_rate": 1.0038978793585871e-05, + "loss": 0.018, + "step": 31379 + }, + { + "epoch": 4.11, + "learning_rate": 1.0038448472494883e-05, + "loss": 0.0191, + "step": 31380 + }, + { + "epoch": 4.11, + "learning_rate": 1.0037918151295763e-05, + "loss": 0.0231, + "step": 31381 + }, + { + "epoch": 4.11, + "learning_rate": 1.003738782999e-05, + "loss": 0.0221, + "step": 31382 + }, + { + "epoch": 4.11, + "learning_rate": 1.0036857508579086e-05, + "loss": 0.0202, + "step": 31383 + }, + { + "epoch": 4.11, + "learning_rate": 1.0036327187064512e-05, + "loss": 0.0224, + "step": 31384 + }, + { + "epoch": 4.11, + "learning_rate": 1.0035796865447764e-05, + "loss": 0.0204, + "step": 31385 + }, + { + "epoch": 4.11, + "learning_rate": 1.0035266543730346e-05, + "loss": 0.0253, + "step": 31386 + }, + { + "epoch": 4.11, + "learning_rate": 1.0034736221913741e-05, + "loss": 0.0213, + "step": 31387 + }, + { + "epoch": 4.11, + "learning_rate": 1.0034205899999445e-05, + "loss": 0.019, + "step": 31388 + }, + { + "epoch": 4.11, + "learning_rate": 1.0033675577988942e-05, + "loss": 0.0238, + "step": 31389 + }, + { + "epoch": 4.11, + "learning_rate": 1.0033145255883729e-05, + "loss": 0.026, + "step": 31390 + }, + { + "epoch": 4.11, + "learning_rate": 1.0032614933685298e-05, + "loss": 0.0262, + "step": 31391 + }, + { + "epoch": 4.11, + "learning_rate": 1.003208461139514e-05, + "loss": 0.0294, + "step": 31392 + }, + { + "epoch": 4.11, + "learning_rate": 1.0031554289014745e-05, + "loss": 0.0235, + "step": 31393 + }, + { + "epoch": 4.11, + "learning_rate": 1.0031023966545605e-05, + "loss": 0.0242, + "step": 31394 + }, + { + "epoch": 4.11, + "learning_rate": 1.0030493643989212e-05, + "loss": 0.0213, + "step": 31395 + }, + { + "epoch": 4.11, + "learning_rate": 1.0029963321347056e-05, + "loss": 0.023, + "step": 31396 + }, + { + "epoch": 4.11, + "learning_rate": 1.002943299862063e-05, + "loss": 0.0239, + "step": 31397 + }, + { + "epoch": 4.11, + "learning_rate": 1.0028902675811427e-05, + "loss": 0.0253, + "step": 31398 + }, + { + "epoch": 4.11, + "learning_rate": 1.0028372352920937e-05, + "loss": 0.0208, + "step": 31399 + }, + { + "epoch": 4.11, + "learning_rate": 1.002784202995065e-05, + "loss": 0.025, + "step": 31400 + }, + { + "epoch": 4.11, + "learning_rate": 1.0027311706902056e-05, + "loss": 0.0239, + "step": 31401 + }, + { + "epoch": 4.11, + "learning_rate": 1.002678138377665e-05, + "loss": 0.0258, + "step": 31402 + }, + { + "epoch": 4.11, + "learning_rate": 1.0026251060575929e-05, + "loss": 0.0199, + "step": 31403 + }, + { + "epoch": 4.11, + "learning_rate": 1.0025720737301375e-05, + "loss": 0.0238, + "step": 31404 + }, + { + "epoch": 4.11, + "learning_rate": 1.0025190413954482e-05, + "loss": 0.0199, + "step": 31405 + }, + { + "epoch": 4.11, + "learning_rate": 1.0024660090536743e-05, + "loss": 0.0206, + "step": 31406 + }, + { + "epoch": 4.11, + "learning_rate": 1.0024129767049646e-05, + "loss": 0.0234, + "step": 31407 + }, + { + "epoch": 4.11, + "learning_rate": 1.0023599443494687e-05, + "loss": 0.02, + "step": 31408 + }, + { + "epoch": 4.11, + "learning_rate": 1.0023069119873357e-05, + "loss": 0.0179, + "step": 31409 + }, + { + "epoch": 4.11, + "learning_rate": 1.0022538796187145e-05, + "loss": 0.0201, + "step": 31410 + }, + { + "epoch": 4.11, + "learning_rate": 1.0022008472437546e-05, + "loss": 0.0215, + "step": 31411 + }, + { + "epoch": 4.11, + "learning_rate": 1.0021478148626044e-05, + "loss": 0.019, + "step": 31412 + }, + { + "epoch": 4.11, + "learning_rate": 1.0020947824754142e-05, + "loss": 0.0202, + "step": 31413 + }, + { + "epoch": 4.12, + "learning_rate": 1.0020417500823322e-05, + "loss": 0.0197, + "step": 31414 + }, + { + "epoch": 4.12, + "learning_rate": 1.0019887176835082e-05, + "loss": 0.0237, + "step": 31415 + }, + { + "epoch": 4.12, + "learning_rate": 1.0019356852790907e-05, + "loss": 0.0262, + "step": 31416 + }, + { + "epoch": 4.12, + "learning_rate": 1.0018826528692293e-05, + "loss": 0.0221, + "step": 31417 + }, + { + "epoch": 4.12, + "learning_rate": 1.001829620454073e-05, + "loss": 0.019, + "step": 31418 + }, + { + "epoch": 4.12, + "learning_rate": 1.001776588033771e-05, + "loss": 0.0231, + "step": 31419 + }, + { + "epoch": 4.12, + "learning_rate": 1.0017235556084726e-05, + "loss": 0.0264, + "step": 31420 + }, + { + "epoch": 4.12, + "learning_rate": 1.0016705231783265e-05, + "loss": 0.0186, + "step": 31421 + }, + { + "epoch": 4.12, + "learning_rate": 1.0016174907434826e-05, + "loss": 0.0197, + "step": 31422 + }, + { + "epoch": 4.12, + "learning_rate": 1.001564458304089e-05, + "loss": 0.0204, + "step": 31423 + }, + { + "epoch": 4.12, + "learning_rate": 1.0015114258602959e-05, + "loss": 0.0214, + "step": 31424 + }, + { + "epoch": 4.12, + "learning_rate": 1.0014583934122518e-05, + "loss": 0.0213, + "step": 31425 + }, + { + "epoch": 4.12, + "learning_rate": 1.0014053609601061e-05, + "loss": 0.019, + "step": 31426 + }, + { + "epoch": 4.12, + "learning_rate": 1.001352328504008e-05, + "loss": 0.0218, + "step": 31427 + }, + { + "epoch": 4.12, + "learning_rate": 1.001299296044106e-05, + "loss": 0.0244, + "step": 31428 + }, + { + "epoch": 4.12, + "learning_rate": 1.0012462635805504e-05, + "loss": 0.0218, + "step": 31429 + }, + { + "epoch": 4.12, + "learning_rate": 1.0011932311134897e-05, + "loss": 0.0209, + "step": 31430 + }, + { + "epoch": 4.12, + "learning_rate": 1.001140198643073e-05, + "loss": 0.0195, + "step": 31431 + }, + { + "epoch": 4.12, + "learning_rate": 1.0010871661694496e-05, + "loss": 0.0228, + "step": 31432 + }, + { + "epoch": 4.12, + "learning_rate": 1.0010341336927685e-05, + "loss": 0.0212, + "step": 31433 + }, + { + "epoch": 4.12, + "learning_rate": 1.0009811012131789e-05, + "loss": 0.0195, + "step": 31434 + }, + { + "epoch": 4.12, + "learning_rate": 1.00092806873083e-05, + "loss": 0.0211, + "step": 31435 + }, + { + "epoch": 4.12, + "learning_rate": 1.0008750362458713e-05, + "loss": 0.0193, + "step": 31436 + }, + { + "epoch": 4.12, + "learning_rate": 1.0008220037584514e-05, + "loss": 0.0209, + "step": 31437 + }, + { + "epoch": 4.12, + "learning_rate": 1.0007689712687197e-05, + "loss": 0.0239, + "step": 31438 + }, + { + "epoch": 4.12, + "learning_rate": 1.0007159387768249e-05, + "loss": 0.0206, + "step": 31439 + }, + { + "epoch": 4.12, + "learning_rate": 1.000662906282917e-05, + "loss": 0.0235, + "step": 31440 + }, + { + "epoch": 4.12, + "learning_rate": 1.0006098737871446e-05, + "loss": 0.0185, + "step": 31441 + }, + { + "epoch": 4.12, + "learning_rate": 1.000556841289657e-05, + "loss": 0.019, + "step": 31442 + }, + { + "epoch": 4.12, + "learning_rate": 1.0005038087906034e-05, + "loss": 0.0194, + "step": 31443 + }, + { + "epoch": 4.12, + "learning_rate": 1.0004507762901325e-05, + "loss": 0.0259, + "step": 31444 + }, + { + "epoch": 4.12, + "learning_rate": 1.0003977437883941e-05, + "loss": 0.0188, + "step": 31445 + }, + { + "epoch": 4.12, + "learning_rate": 1.0003447112855368e-05, + "loss": 0.0201, + "step": 31446 + }, + { + "epoch": 4.12, + "learning_rate": 1.0002916787817105e-05, + "loss": 0.0209, + "step": 31447 + }, + { + "epoch": 4.12, + "learning_rate": 1.0002386462770636e-05, + "loss": 0.0274, + "step": 31448 + }, + { + "epoch": 4.12, + "learning_rate": 1.0001856137717457e-05, + "loss": 0.0217, + "step": 31449 + }, + { + "epoch": 4.12, + "learning_rate": 1.0001325812659053e-05, + "loss": 0.0238, + "step": 31450 + }, + { + "epoch": 4.12, + "learning_rate": 1.0000795487596924e-05, + "loss": 0.0256, + "step": 31451 + }, + { + "epoch": 4.12, + "learning_rate": 1.0000265162532557e-05, + "loss": 0.0222, + "step": 31452 + }, + { + "epoch": 4.12, + "learning_rate": 9.999734837467446e-06, + "loss": 0.0185, + "step": 31453 + }, + { + "epoch": 4.12, + "learning_rate": 9.999204512403078e-06, + "loss": 0.0205, + "step": 31454 + }, + { + "epoch": 4.12, + "learning_rate": 9.99867418734095e-06, + "loss": 0.0242, + "step": 31455 + }, + { + "epoch": 4.12, + "learning_rate": 9.998143862282548e-06, + "loss": 0.0267, + "step": 31456 + }, + { + "epoch": 4.12, + "learning_rate": 9.997613537229367e-06, + "loss": 0.0268, + "step": 31457 + }, + { + "epoch": 4.12, + "learning_rate": 9.997083212182895e-06, + "loss": 0.0164, + "step": 31458 + }, + { + "epoch": 4.12, + "learning_rate": 9.996552887144632e-06, + "loss": 0.02, + "step": 31459 + }, + { + "epoch": 4.12, + "learning_rate": 9.99602256211606e-06, + "loss": 0.021, + "step": 31460 + }, + { + "epoch": 4.12, + "learning_rate": 9.995492237098677e-06, + "loss": 0.0226, + "step": 31461 + }, + { + "epoch": 4.12, + "learning_rate": 9.99496191209397e-06, + "loss": 0.0251, + "step": 31462 + }, + { + "epoch": 4.12, + "learning_rate": 9.994431587103434e-06, + "loss": 0.0216, + "step": 31463 + }, + { + "epoch": 4.12, + "learning_rate": 9.993901262128557e-06, + "loss": 0.0227, + "step": 31464 + }, + { + "epoch": 4.12, + "learning_rate": 9.993370937170832e-06, + "loss": 0.0239, + "step": 31465 + }, + { + "epoch": 4.12, + "learning_rate": 9.992840612231753e-06, + "loss": 0.0197, + "step": 31466 + }, + { + "epoch": 4.12, + "learning_rate": 9.992310287312808e-06, + "loss": 0.0223, + "step": 31467 + }, + { + "epoch": 4.12, + "learning_rate": 9.991779962415491e-06, + "loss": 0.0229, + "step": 31468 + }, + { + "epoch": 4.12, + "learning_rate": 9.99124963754129e-06, + "loss": 0.0206, + "step": 31469 + }, + { + "epoch": 4.12, + "learning_rate": 9.9907193126917e-06, + "loss": 0.0231, + "step": 31470 + }, + { + "epoch": 4.12, + "learning_rate": 9.990188987868211e-06, + "loss": 0.0206, + "step": 31471 + }, + { + "epoch": 4.12, + "learning_rate": 9.989658663072318e-06, + "loss": 0.0224, + "step": 31472 + }, + { + "epoch": 4.12, + "learning_rate": 9.989128338305505e-06, + "loss": 0.0191, + "step": 31473 + }, + { + "epoch": 4.12, + "learning_rate": 9.988598013569272e-06, + "loss": 0.0221, + "step": 31474 + }, + { + "epoch": 4.12, + "learning_rate": 9.988067688865106e-06, + "loss": 0.0207, + "step": 31475 + }, + { + "epoch": 4.12, + "learning_rate": 9.9875373641945e-06, + "loss": 0.0238, + "step": 31476 + }, + { + "epoch": 4.12, + "learning_rate": 9.987007039558941e-06, + "loss": 0.0235, + "step": 31477 + }, + { + "epoch": 4.12, + "learning_rate": 9.986476714959926e-06, + "loss": 0.0265, + "step": 31478 + }, + { + "epoch": 4.12, + "learning_rate": 9.985946390398944e-06, + "loss": 0.0218, + "step": 31479 + }, + { + "epoch": 4.12, + "learning_rate": 9.985416065877484e-06, + "loss": 0.0201, + "step": 31480 + }, + { + "epoch": 4.12, + "learning_rate": 9.984885741397043e-06, + "loss": 0.0163, + "step": 31481 + }, + { + "epoch": 4.12, + "learning_rate": 9.984355416959111e-06, + "loss": 0.02, + "step": 31482 + }, + { + "epoch": 4.12, + "learning_rate": 9.983825092565178e-06, + "loss": 0.0252, + "step": 31483 + }, + { + "epoch": 4.12, + "learning_rate": 9.983294768216737e-06, + "loss": 0.0249, + "step": 31484 + }, + { + "epoch": 4.12, + "learning_rate": 9.982764443915277e-06, + "loss": 0.0225, + "step": 31485 + }, + { + "epoch": 4.12, + "learning_rate": 9.982234119662291e-06, + "loss": 0.019, + "step": 31486 + }, + { + "epoch": 4.12, + "learning_rate": 9.981703795459274e-06, + "loss": 0.0247, + "step": 31487 + }, + { + "epoch": 4.12, + "learning_rate": 9.98117347130771e-06, + "loss": 0.0223, + "step": 31488 + }, + { + "epoch": 4.12, + "learning_rate": 9.980643147209098e-06, + "loss": 0.0225, + "step": 31489 + }, + { + "epoch": 4.12, + "learning_rate": 9.980112823164923e-06, + "loss": 0.0279, + "step": 31490 + }, + { + "epoch": 4.13, + "learning_rate": 9.97958249917668e-06, + "loss": 0.0207, + "step": 31491 + }, + { + "epoch": 4.13, + "learning_rate": 9.979052175245858e-06, + "loss": 0.0234, + "step": 31492 + }, + { + "epoch": 4.13, + "learning_rate": 9.978521851373956e-06, + "loss": 0.0157, + "step": 31493 + }, + { + "epoch": 4.13, + "learning_rate": 9.977991527562457e-06, + "loss": 0.0224, + "step": 31494 + }, + { + "epoch": 4.13, + "learning_rate": 9.977461203812858e-06, + "loss": 0.0244, + "step": 31495 + }, + { + "epoch": 4.13, + "learning_rate": 9.976930880126644e-06, + "loss": 0.0262, + "step": 31496 + }, + { + "epoch": 4.13, + "learning_rate": 9.976400556505317e-06, + "loss": 0.0168, + "step": 31497 + }, + { + "epoch": 4.13, + "learning_rate": 9.975870232950357e-06, + "loss": 0.0211, + "step": 31498 + }, + { + "epoch": 4.13, + "learning_rate": 9.975339909463264e-06, + "loss": 0.0206, + "step": 31499 + }, + { + "epoch": 4.13, + "learning_rate": 9.974809586045523e-06, + "loss": 0.0197, + "step": 31500 + }, + { + "epoch": 4.13, + "learning_rate": 9.97427926269863e-06, + "loss": 0.0174, + "step": 31501 + }, + { + "epoch": 4.13, + "learning_rate": 9.973748939424071e-06, + "loss": 0.0215, + "step": 31502 + }, + { + "epoch": 4.13, + "learning_rate": 9.973218616223348e-06, + "loss": 0.0259, + "step": 31503 + }, + { + "epoch": 4.13, + "learning_rate": 9.972688293097944e-06, + "loss": 0.0255, + "step": 31504 + }, + { + "epoch": 4.13, + "learning_rate": 9.972157970049355e-06, + "loss": 0.0239, + "step": 31505 + }, + { + "epoch": 4.13, + "learning_rate": 9.971627647079067e-06, + "loss": 0.0208, + "step": 31506 + }, + { + "epoch": 4.13, + "learning_rate": 9.971097324188574e-06, + "loss": 0.0228, + "step": 31507 + }, + { + "epoch": 4.13, + "learning_rate": 9.970567001379372e-06, + "loss": 0.0231, + "step": 31508 + }, + { + "epoch": 4.13, + "learning_rate": 9.970036678652946e-06, + "loss": 0.027, + "step": 31509 + }, + { + "epoch": 4.13, + "learning_rate": 9.969506356010794e-06, + "loss": 0.021, + "step": 31510 + }, + { + "epoch": 4.13, + "learning_rate": 9.9689760334544e-06, + "loss": 0.0273, + "step": 31511 + }, + { + "epoch": 4.13, + "learning_rate": 9.96844571098526e-06, + "loss": 0.0221, + "step": 31512 + }, + { + "epoch": 4.13, + "learning_rate": 9.967915388604863e-06, + "loss": 0.0235, + "step": 31513 + }, + { + "epoch": 4.13, + "learning_rate": 9.967385066314704e-06, + "loss": 0.0181, + "step": 31514 + }, + { + "epoch": 4.13, + "learning_rate": 9.966854744116271e-06, + "loss": 0.0218, + "step": 31515 + }, + { + "epoch": 4.13, + "learning_rate": 9.966324422011061e-06, + "loss": 0.0217, + "step": 31516 + }, + { + "epoch": 4.13, + "learning_rate": 9.965794100000558e-06, + "loss": 0.0217, + "step": 31517 + }, + { + "epoch": 4.13, + "learning_rate": 9.96526377808626e-06, + "loss": 0.0206, + "step": 31518 + }, + { + "epoch": 4.13, + "learning_rate": 9.964733456269655e-06, + "loss": 0.0224, + "step": 31519 + }, + { + "epoch": 4.13, + "learning_rate": 9.964203134552237e-06, + "loss": 0.019, + "step": 31520 + }, + { + "epoch": 4.13, + "learning_rate": 9.963672812935493e-06, + "loss": 0.0288, + "step": 31521 + }, + { + "epoch": 4.13, + "learning_rate": 9.963142491420919e-06, + "loss": 0.023, + "step": 31522 + }, + { + "epoch": 4.13, + "learning_rate": 9.962612170010003e-06, + "loss": 0.0271, + "step": 31523 + }, + { + "epoch": 4.13, + "learning_rate": 9.962081848704239e-06, + "loss": 0.0192, + "step": 31524 + }, + { + "epoch": 4.13, + "learning_rate": 9.961551527505117e-06, + "loss": 0.0265, + "step": 31525 + }, + { + "epoch": 4.13, + "learning_rate": 9.96102120641413e-06, + "loss": 0.0254, + "step": 31526 + }, + { + "epoch": 4.13, + "learning_rate": 9.960490885432769e-06, + "loss": 0.0267, + "step": 31527 + }, + { + "epoch": 4.13, + "learning_rate": 9.959960564562528e-06, + "loss": 0.0197, + "step": 31528 + }, + { + "epoch": 4.13, + "learning_rate": 9.959430243804892e-06, + "loss": 0.026, + "step": 31529 + }, + { + "epoch": 4.13, + "learning_rate": 9.958899923161358e-06, + "loss": 0.0215, + "step": 31530 + }, + { + "epoch": 4.13, + "learning_rate": 9.958369602633416e-06, + "loss": 0.0188, + "step": 31531 + }, + { + "epoch": 4.13, + "learning_rate": 9.957839282222555e-06, + "loss": 0.0234, + "step": 31532 + }, + { + "epoch": 4.13, + "learning_rate": 9.957308961930272e-06, + "loss": 0.0217, + "step": 31533 + }, + { + "epoch": 4.13, + "learning_rate": 9.956778641758054e-06, + "loss": 0.0222, + "step": 31534 + }, + { + "epoch": 4.13, + "learning_rate": 9.95624832170739e-06, + "loss": 0.0202, + "step": 31535 + }, + { + "epoch": 4.13, + "learning_rate": 9.955718001779778e-06, + "loss": 0.0219, + "step": 31536 + }, + { + "epoch": 4.13, + "learning_rate": 9.955187681976709e-06, + "loss": 0.0179, + "step": 31537 + }, + { + "epoch": 4.13, + "learning_rate": 9.95465736229967e-06, + "loss": 0.0202, + "step": 31538 + }, + { + "epoch": 4.13, + "learning_rate": 9.954127042750157e-06, + "loss": 0.0234, + "step": 31539 + }, + { + "epoch": 4.13, + "learning_rate": 9.953596723329657e-06, + "loss": 0.0266, + "step": 31540 + }, + { + "epoch": 4.13, + "learning_rate": 9.953066404039666e-06, + "loss": 0.0258, + "step": 31541 + }, + { + "epoch": 4.13, + "learning_rate": 9.95253608488167e-06, + "loss": 0.0242, + "step": 31542 + }, + { + "epoch": 4.13, + "learning_rate": 9.952005765857165e-06, + "loss": 0.026, + "step": 31543 + }, + { + "epoch": 4.13, + "learning_rate": 9.951475446967644e-06, + "loss": 0.0248, + "step": 31544 + }, + { + "epoch": 4.13, + "learning_rate": 9.950945128214593e-06, + "loss": 0.0241, + "step": 31545 + }, + { + "epoch": 4.13, + "learning_rate": 9.950414809599502e-06, + "loss": 0.0183, + "step": 31546 + }, + { + "epoch": 4.13, + "learning_rate": 9.949884491123873e-06, + "loss": 0.0288, + "step": 31547 + }, + { + "epoch": 4.13, + "learning_rate": 9.94935417278919e-06, + "loss": 0.0239, + "step": 31548 + }, + { + "epoch": 4.13, + "learning_rate": 9.948823854596945e-06, + "loss": 0.0214, + "step": 31549 + }, + { + "epoch": 4.13, + "learning_rate": 9.94829353654863e-06, + "loss": 0.0206, + "step": 31550 + }, + { + "epoch": 4.13, + "learning_rate": 9.947763218645736e-06, + "loss": 0.0193, + "step": 31551 + }, + { + "epoch": 4.13, + "learning_rate": 9.947232900889758e-06, + "loss": 0.0174, + "step": 31552 + }, + { + "epoch": 4.13, + "learning_rate": 9.946702583282181e-06, + "loss": 0.0193, + "step": 31553 + }, + { + "epoch": 4.13, + "learning_rate": 9.946172265824504e-06, + "loss": 0.0256, + "step": 31554 + }, + { + "epoch": 4.13, + "learning_rate": 9.945641948518212e-06, + "loss": 0.0184, + "step": 31555 + }, + { + "epoch": 4.13, + "learning_rate": 9.945111631364802e-06, + "loss": 0.0194, + "step": 31556 + }, + { + "epoch": 4.13, + "learning_rate": 9.944581314365756e-06, + "loss": 0.015, + "step": 31557 + }, + { + "epoch": 4.13, + "learning_rate": 9.944050997522576e-06, + "loss": 0.0207, + "step": 31558 + }, + { + "epoch": 4.13, + "learning_rate": 9.94352068083675e-06, + "loss": 0.0226, + "step": 31559 + }, + { + "epoch": 4.13, + "learning_rate": 9.94299036430977e-06, + "loss": 0.0213, + "step": 31560 + }, + { + "epoch": 4.13, + "learning_rate": 9.942460047943126e-06, + "loss": 0.0185, + "step": 31561 + }, + { + "epoch": 4.13, + "learning_rate": 9.941929731738312e-06, + "loss": 0.023, + "step": 31562 + }, + { + "epoch": 4.13, + "learning_rate": 9.941399415696813e-06, + "loss": 0.0203, + "step": 31563 + }, + { + "epoch": 4.13, + "learning_rate": 9.94086909982013e-06, + "loss": 0.0227, + "step": 31564 + }, + { + "epoch": 4.13, + "learning_rate": 9.940338784109745e-06, + "loss": 0.0239, + "step": 31565 + }, + { + "epoch": 4.13, + "learning_rate": 9.939808468567155e-06, + "loss": 0.0165, + "step": 31566 + }, + { + "epoch": 4.14, + "learning_rate": 9.939278153193853e-06, + "loss": 0.0278, + "step": 31567 + }, + { + "epoch": 4.14, + "learning_rate": 9.938747837991323e-06, + "loss": 0.0239, + "step": 31568 + }, + { + "epoch": 4.14, + "learning_rate": 9.938217522961065e-06, + "loss": 0.0206, + "step": 31569 + }, + { + "epoch": 4.14, + "learning_rate": 9.937687208104568e-06, + "loss": 0.0195, + "step": 31570 + }, + { + "epoch": 4.14, + "learning_rate": 9.93715689342332e-06, + "loss": 0.0219, + "step": 31571 + }, + { + "epoch": 4.14, + "learning_rate": 9.936626578918815e-06, + "loss": 0.0188, + "step": 31572 + }, + { + "epoch": 4.14, + "learning_rate": 9.936096264592546e-06, + "loss": 0.0262, + "step": 31573 + }, + { + "epoch": 4.14, + "learning_rate": 9.935565950446002e-06, + "loss": 0.0196, + "step": 31574 + }, + { + "epoch": 4.14, + "learning_rate": 9.935035636480678e-06, + "loss": 0.0234, + "step": 31575 + }, + { + "epoch": 4.14, + "learning_rate": 9.934505322698058e-06, + "loss": 0.0255, + "step": 31576 + }, + { + "epoch": 4.14, + "learning_rate": 9.933975009099643e-06, + "loss": 0.0216, + "step": 31577 + }, + { + "epoch": 4.14, + "learning_rate": 9.933444695686917e-06, + "loss": 0.023, + "step": 31578 + }, + { + "epoch": 4.14, + "learning_rate": 9.93291438246137e-06, + "loss": 0.0225, + "step": 31579 + }, + { + "epoch": 4.14, + "learning_rate": 9.932384069424502e-06, + "loss": 0.0202, + "step": 31580 + }, + { + "epoch": 4.14, + "learning_rate": 9.931853756577803e-06, + "loss": 0.0244, + "step": 31581 + }, + { + "epoch": 4.14, + "learning_rate": 9.931323443922758e-06, + "loss": 0.0217, + "step": 31582 + }, + { + "epoch": 4.14, + "learning_rate": 9.930793131460866e-06, + "loss": 0.0269, + "step": 31583 + }, + { + "epoch": 4.14, + "learning_rate": 9.93026281919361e-06, + "loss": 0.023, + "step": 31584 + }, + { + "epoch": 4.14, + "learning_rate": 9.92973250712249e-06, + "loss": 0.0238, + "step": 31585 + }, + { + "epoch": 4.14, + "learning_rate": 9.92920219524899e-06, + "loss": 0.0242, + "step": 31586 + }, + { + "epoch": 4.14, + "learning_rate": 9.928671883574607e-06, + "loss": 0.0232, + "step": 31587 + }, + { + "epoch": 4.14, + "learning_rate": 9.928141572100831e-06, + "loss": 0.0191, + "step": 31588 + }, + { + "epoch": 4.14, + "learning_rate": 9.927611260829149e-06, + "loss": 0.0244, + "step": 31589 + }, + { + "epoch": 4.14, + "learning_rate": 9.92708094976106e-06, + "loss": 0.0242, + "step": 31590 + }, + { + "epoch": 4.14, + "learning_rate": 9.926550638898052e-06, + "loss": 0.0234, + "step": 31591 + }, + { + "epoch": 4.14, + "learning_rate": 9.926020328241617e-06, + "loss": 0.0174, + "step": 31592 + }, + { + "epoch": 4.14, + "learning_rate": 9.925490017793243e-06, + "loss": 0.0203, + "step": 31593 + }, + { + "epoch": 4.14, + "learning_rate": 9.924959707554429e-06, + "loss": 0.0192, + "step": 31594 + }, + { + "epoch": 4.14, + "learning_rate": 9.924429397526659e-06, + "loss": 0.0227, + "step": 31595 + }, + { + "epoch": 4.14, + "learning_rate": 9.923899087711428e-06, + "loss": 0.0242, + "step": 31596 + }, + { + "epoch": 4.14, + "learning_rate": 9.923368778110225e-06, + "loss": 0.0189, + "step": 31597 + }, + { + "epoch": 4.14, + "learning_rate": 9.922838468724546e-06, + "loss": 0.0284, + "step": 31598 + }, + { + "epoch": 4.14, + "learning_rate": 9.922308159555878e-06, + "loss": 0.0196, + "step": 31599 + }, + { + "epoch": 4.14, + "learning_rate": 9.921777850605711e-06, + "loss": 0.0271, + "step": 31600 + }, + { + "epoch": 4.14, + "learning_rate": 9.921247541875543e-06, + "loss": 0.0227, + "step": 31601 + }, + { + "epoch": 4.14, + "learning_rate": 9.920717233366863e-06, + "loss": 0.0184, + "step": 31602 + }, + { + "epoch": 4.14, + "learning_rate": 9.92018692508116e-06, + "loss": 0.0229, + "step": 31603 + }, + { + "epoch": 4.14, + "learning_rate": 9.91965661701993e-06, + "loss": 0.0206, + "step": 31604 + }, + { + "epoch": 4.14, + "learning_rate": 9.919126309184657e-06, + "loss": 0.0227, + "step": 31605 + }, + { + "epoch": 4.14, + "learning_rate": 9.918596001576841e-06, + "loss": 0.0206, + "step": 31606 + }, + { + "epoch": 4.14, + "learning_rate": 9.918065694197966e-06, + "loss": 0.0206, + "step": 31607 + }, + { + "epoch": 4.14, + "learning_rate": 9.917535387049529e-06, + "loss": 0.0187, + "step": 31608 + }, + { + "epoch": 4.14, + "learning_rate": 9.91700508013302e-06, + "loss": 0.0241, + "step": 31609 + }, + { + "epoch": 4.14, + "learning_rate": 9.916474773449928e-06, + "loss": 0.0229, + "step": 31610 + }, + { + "epoch": 4.14, + "learning_rate": 9.915944467001742e-06, + "loss": 0.018, + "step": 31611 + }, + { + "epoch": 4.14, + "learning_rate": 9.915414160789964e-06, + "loss": 0.0205, + "step": 31612 + }, + { + "epoch": 4.14, + "learning_rate": 9.914883854816078e-06, + "loss": 0.017, + "step": 31613 + }, + { + "epoch": 4.14, + "learning_rate": 9.914353549081577e-06, + "loss": 0.0197, + "step": 31614 + }, + { + "epoch": 4.14, + "learning_rate": 9.91382324358795e-06, + "loss": 0.0248, + "step": 31615 + }, + { + "epoch": 4.14, + "learning_rate": 9.913292938336692e-06, + "loss": 0.0225, + "step": 31616 + }, + { + "epoch": 4.14, + "learning_rate": 9.912762633329295e-06, + "loss": 0.0233, + "step": 31617 + }, + { + "epoch": 4.14, + "learning_rate": 9.912232328567245e-06, + "loss": 0.0249, + "step": 31618 + }, + { + "epoch": 4.14, + "learning_rate": 9.91170202405204e-06, + "loss": 0.0261, + "step": 31619 + }, + { + "epoch": 4.14, + "learning_rate": 9.911171719785166e-06, + "loss": 0.0212, + "step": 31620 + }, + { + "epoch": 4.14, + "learning_rate": 9.910641415768118e-06, + "loss": 0.021, + "step": 31621 + }, + { + "epoch": 4.14, + "learning_rate": 9.910111112002381e-06, + "loss": 0.0248, + "step": 31622 + }, + { + "epoch": 4.14, + "learning_rate": 9.909580808489458e-06, + "loss": 0.0201, + "step": 31623 + }, + { + "epoch": 4.14, + "learning_rate": 9.909050505230831e-06, + "loss": 0.0226, + "step": 31624 + }, + { + "epoch": 4.14, + "learning_rate": 9.908520202227998e-06, + "loss": 0.0219, + "step": 31625 + }, + { + "epoch": 4.14, + "learning_rate": 9.907989899482444e-06, + "loss": 0.0244, + "step": 31626 + }, + { + "epoch": 4.14, + "learning_rate": 9.907459596995666e-06, + "loss": 0.0226, + "step": 31627 + }, + { + "epoch": 4.14, + "learning_rate": 9.906929294769152e-06, + "loss": 0.0222, + "step": 31628 + }, + { + "epoch": 4.14, + "learning_rate": 9.906398992804395e-06, + "loss": 0.0217, + "step": 31629 + }, + { + "epoch": 4.14, + "learning_rate": 9.905868691102884e-06, + "loss": 0.02, + "step": 31630 + }, + { + "epoch": 4.14, + "learning_rate": 9.905338389666112e-06, + "loss": 0.0223, + "step": 31631 + }, + { + "epoch": 4.14, + "learning_rate": 9.904808088495574e-06, + "loss": 0.0222, + "step": 31632 + }, + { + "epoch": 4.14, + "learning_rate": 9.904277787592754e-06, + "loss": 0.0223, + "step": 31633 + }, + { + "epoch": 4.14, + "learning_rate": 9.903747486959148e-06, + "loss": 0.022, + "step": 31634 + }, + { + "epoch": 4.14, + "learning_rate": 9.90321718659625e-06, + "loss": 0.0244, + "step": 31635 + }, + { + "epoch": 4.14, + "learning_rate": 9.902686886505548e-06, + "loss": 0.0212, + "step": 31636 + }, + { + "epoch": 4.14, + "learning_rate": 9.902156586688532e-06, + "loss": 0.0271, + "step": 31637 + }, + { + "epoch": 4.14, + "learning_rate": 9.901626287146698e-06, + "loss": 0.0233, + "step": 31638 + }, + { + "epoch": 4.14, + "learning_rate": 9.901095987881533e-06, + "loss": 0.0239, + "step": 31639 + }, + { + "epoch": 4.14, + "learning_rate": 9.900565688894531e-06, + "loss": 0.025, + "step": 31640 + }, + { + "epoch": 4.14, + "learning_rate": 9.900035390187183e-06, + "loss": 0.0229, + "step": 31641 + }, + { + "epoch": 4.14, + "learning_rate": 9.899505091760981e-06, + "loss": 0.0263, + "step": 31642 + }, + { + "epoch": 4.15, + "learning_rate": 9.898974793617413e-06, + "loss": 0.0203, + "step": 31643 + }, + { + "epoch": 4.15, + "learning_rate": 9.898444495757971e-06, + "loss": 0.0198, + "step": 31644 + }, + { + "epoch": 4.15, + "learning_rate": 9.897914198184151e-06, + "loss": 0.022, + "step": 31645 + }, + { + "epoch": 4.15, + "learning_rate": 9.897383900897444e-06, + "loss": 0.0217, + "step": 31646 + }, + { + "epoch": 4.15, + "learning_rate": 9.896853603899336e-06, + "loss": 0.0239, + "step": 31647 + }, + { + "epoch": 4.15, + "learning_rate": 9.896323307191325e-06, + "loss": 0.0181, + "step": 31648 + }, + { + "epoch": 4.15, + "learning_rate": 9.895793010774898e-06, + "loss": 0.0196, + "step": 31649 + }, + { + "epoch": 4.15, + "learning_rate": 9.89526271465155e-06, + "loss": 0.0271, + "step": 31650 + }, + { + "epoch": 4.15, + "learning_rate": 9.894732418822765e-06, + "loss": 0.025, + "step": 31651 + }, + { + "epoch": 4.15, + "learning_rate": 9.894202123290041e-06, + "loss": 0.0254, + "step": 31652 + }, + { + "epoch": 4.15, + "learning_rate": 9.89367182805487e-06, + "loss": 0.0229, + "step": 31653 + }, + { + "epoch": 4.15, + "learning_rate": 9.89314153311874e-06, + "loss": 0.0248, + "step": 31654 + }, + { + "epoch": 4.15, + "learning_rate": 9.89261123848314e-06, + "loss": 0.0292, + "step": 31655 + }, + { + "epoch": 4.15, + "learning_rate": 9.89208094414957e-06, + "loss": 0.0226, + "step": 31656 + }, + { + "epoch": 4.15, + "learning_rate": 9.891550650119515e-06, + "loss": 0.0221, + "step": 31657 + }, + { + "epoch": 4.15, + "learning_rate": 9.891020356394468e-06, + "loss": 0.0216, + "step": 31658 + }, + { + "epoch": 4.15, + "learning_rate": 9.890490062975924e-06, + "loss": 0.0212, + "step": 31659 + }, + { + "epoch": 4.15, + "learning_rate": 9.889959769865367e-06, + "loss": 0.0206, + "step": 31660 + }, + { + "epoch": 4.15, + "learning_rate": 9.889429477064294e-06, + "loss": 0.0196, + "step": 31661 + }, + { + "epoch": 4.15, + "learning_rate": 9.888899184574193e-06, + "loss": 0.0274, + "step": 31662 + }, + { + "epoch": 4.15, + "learning_rate": 9.88836889239656e-06, + "loss": 0.0287, + "step": 31663 + }, + { + "epoch": 4.15, + "learning_rate": 9.887838600532881e-06, + "loss": 0.0204, + "step": 31664 + }, + { + "epoch": 4.15, + "learning_rate": 9.887308308984652e-06, + "loss": 0.022, + "step": 31665 + }, + { + "epoch": 4.15, + "learning_rate": 9.886778017753356e-06, + "loss": 0.0206, + "step": 31666 + }, + { + "epoch": 4.15, + "learning_rate": 9.886247726840497e-06, + "loss": 0.0218, + "step": 31667 + }, + { + "epoch": 4.15, + "learning_rate": 9.885717436247559e-06, + "loss": 0.0166, + "step": 31668 + }, + { + "epoch": 4.15, + "learning_rate": 9.885187145976034e-06, + "loss": 0.023, + "step": 31669 + }, + { + "epoch": 4.15, + "learning_rate": 9.884656856027414e-06, + "loss": 0.0231, + "step": 31670 + }, + { + "epoch": 4.15, + "learning_rate": 9.884126566403194e-06, + "loss": 0.0193, + "step": 31671 + }, + { + "epoch": 4.15, + "learning_rate": 9.883596277104857e-06, + "loss": 0.0203, + "step": 31672 + }, + { + "epoch": 4.15, + "learning_rate": 9.8830659881339e-06, + "loss": 0.0215, + "step": 31673 + }, + { + "epoch": 4.15, + "learning_rate": 9.882535699491817e-06, + "loss": 0.0202, + "step": 31674 + }, + { + "epoch": 4.15, + "learning_rate": 9.882005411180093e-06, + "loss": 0.0193, + "step": 31675 + }, + { + "epoch": 4.15, + "learning_rate": 9.881475123200224e-06, + "loss": 0.0189, + "step": 31676 + }, + { + "epoch": 4.15, + "learning_rate": 9.880944835553697e-06, + "loss": 0.025, + "step": 31677 + }, + { + "epoch": 4.15, + "learning_rate": 9.880414548242008e-06, + "loss": 0.0209, + "step": 31678 + }, + { + "epoch": 4.15, + "learning_rate": 9.87988426126665e-06, + "loss": 0.0245, + "step": 31679 + }, + { + "epoch": 4.15, + "learning_rate": 9.879353974629107e-06, + "loss": 0.021, + "step": 31680 + }, + { + "epoch": 4.15, + "learning_rate": 9.878823688330874e-06, + "loss": 0.0208, + "step": 31681 + }, + { + "epoch": 4.15, + "learning_rate": 9.878293402373448e-06, + "loss": 0.0178, + "step": 31682 + }, + { + "epoch": 4.15, + "learning_rate": 9.87776311675831e-06, + "loss": 0.0229, + "step": 31683 + }, + { + "epoch": 4.15, + "learning_rate": 9.877232831486962e-06, + "loss": 0.0216, + "step": 31684 + }, + { + "epoch": 4.15, + "learning_rate": 9.876702546560886e-06, + "loss": 0.0214, + "step": 31685 + }, + { + "epoch": 4.15, + "learning_rate": 9.87617226198158e-06, + "loss": 0.0215, + "step": 31686 + }, + { + "epoch": 4.15, + "learning_rate": 9.87564197775053e-06, + "loss": 0.0254, + "step": 31687 + }, + { + "epoch": 4.15, + "learning_rate": 9.87511169386923e-06, + "loss": 0.0212, + "step": 31688 + }, + { + "epoch": 4.15, + "learning_rate": 9.874581410339172e-06, + "loss": 0.0251, + "step": 31689 + }, + { + "epoch": 4.15, + "learning_rate": 9.874051127161851e-06, + "loss": 0.0226, + "step": 31690 + }, + { + "epoch": 4.15, + "learning_rate": 9.873520844338752e-06, + "loss": 0.0224, + "step": 31691 + }, + { + "epoch": 4.15, + "learning_rate": 9.87299056187137e-06, + "loss": 0.0223, + "step": 31692 + }, + { + "epoch": 4.15, + "learning_rate": 9.872460279761194e-06, + "loss": 0.0242, + "step": 31693 + }, + { + "epoch": 4.15, + "learning_rate": 9.871929998009716e-06, + "loss": 0.0245, + "step": 31694 + }, + { + "epoch": 4.15, + "learning_rate": 9.871399716618431e-06, + "loss": 0.0194, + "step": 31695 + }, + { + "epoch": 4.15, + "learning_rate": 9.870869435588825e-06, + "loss": 0.0223, + "step": 31696 + }, + { + "epoch": 4.15, + "learning_rate": 9.870339154922393e-06, + "loss": 0.0207, + "step": 31697 + }, + { + "epoch": 4.15, + "learning_rate": 9.869808874620624e-06, + "loss": 0.0225, + "step": 31698 + }, + { + "epoch": 4.15, + "learning_rate": 9.869278594685007e-06, + "loss": 0.0251, + "step": 31699 + }, + { + "epoch": 4.15, + "learning_rate": 9.868748315117043e-06, + "loss": 0.0272, + "step": 31700 + }, + { + "epoch": 4.15, + "learning_rate": 9.868218035918216e-06, + "loss": 0.0254, + "step": 31701 + }, + { + "epoch": 4.15, + "learning_rate": 9.867687757090018e-06, + "loss": 0.0197, + "step": 31702 + }, + { + "epoch": 4.15, + "learning_rate": 9.867157478633943e-06, + "loss": 0.0246, + "step": 31703 + }, + { + "epoch": 4.15, + "learning_rate": 9.866627200551478e-06, + "loss": 0.0219, + "step": 31704 + }, + { + "epoch": 4.15, + "learning_rate": 9.866096922844118e-06, + "loss": 0.0234, + "step": 31705 + }, + { + "epoch": 4.15, + "learning_rate": 9.865566645513353e-06, + "loss": 0.0232, + "step": 31706 + }, + { + "epoch": 4.15, + "learning_rate": 9.865036368560676e-06, + "loss": 0.021, + "step": 31707 + }, + { + "epoch": 4.15, + "learning_rate": 9.864506091987576e-06, + "loss": 0.0243, + "step": 31708 + }, + { + "epoch": 4.15, + "learning_rate": 9.863975815795543e-06, + "loss": 0.0231, + "step": 31709 + }, + { + "epoch": 4.15, + "learning_rate": 9.863445539986071e-06, + "loss": 0.0211, + "step": 31710 + }, + { + "epoch": 4.15, + "learning_rate": 9.862915264560654e-06, + "loss": 0.0153, + "step": 31711 + }, + { + "epoch": 4.15, + "learning_rate": 9.862384989520779e-06, + "loss": 0.0196, + "step": 31712 + }, + { + "epoch": 4.15, + "learning_rate": 9.86185471486794e-06, + "loss": 0.0242, + "step": 31713 + }, + { + "epoch": 4.15, + "learning_rate": 9.861324440603628e-06, + "loss": 0.0281, + "step": 31714 + }, + { + "epoch": 4.15, + "learning_rate": 9.860794166729334e-06, + "loss": 0.0193, + "step": 31715 + }, + { + "epoch": 4.15, + "learning_rate": 9.860263893246546e-06, + "loss": 0.021, + "step": 31716 + }, + { + "epoch": 4.15, + "learning_rate": 9.85973362015676e-06, + "loss": 0.0233, + "step": 31717 + }, + { + "epoch": 4.15, + "learning_rate": 9.859203347461465e-06, + "loss": 0.0259, + "step": 31718 + }, + { + "epoch": 4.15, + "learning_rate": 9.858673075162153e-06, + "loss": 0.0201, + "step": 31719 + }, + { + "epoch": 4.16, + "learning_rate": 9.858142803260314e-06, + "loss": 0.0212, + "step": 31720 + }, + { + "epoch": 4.16, + "learning_rate": 9.857612531757445e-06, + "loss": 0.0202, + "step": 31721 + }, + { + "epoch": 4.16, + "learning_rate": 9.85708226065503e-06, + "loss": 0.0235, + "step": 31722 + }, + { + "epoch": 4.16, + "learning_rate": 9.856551989954564e-06, + "loss": 0.0244, + "step": 31723 + }, + { + "epoch": 4.16, + "learning_rate": 9.85602171965754e-06, + "loss": 0.0212, + "step": 31724 + }, + { + "epoch": 4.16, + "learning_rate": 9.855491449765445e-06, + "loss": 0.0257, + "step": 31725 + }, + { + "epoch": 4.16, + "learning_rate": 9.854961180279773e-06, + "loss": 0.0231, + "step": 31726 + }, + { + "epoch": 4.16, + "learning_rate": 9.854430911202016e-06, + "loss": 0.0206, + "step": 31727 + }, + { + "epoch": 4.16, + "learning_rate": 9.853900642533665e-06, + "loss": 0.0234, + "step": 31728 + }, + { + "epoch": 4.16, + "learning_rate": 9.853370374276208e-06, + "loss": 0.0247, + "step": 31729 + }, + { + "epoch": 4.16, + "learning_rate": 9.852840106431143e-06, + "loss": 0.0218, + "step": 31730 + }, + { + "epoch": 4.16, + "learning_rate": 9.852309838999948e-06, + "loss": 0.0255, + "step": 31731 + }, + { + "epoch": 4.16, + "learning_rate": 9.851779571984133e-06, + "loss": 0.0156, + "step": 31732 + }, + { + "epoch": 4.16, + "learning_rate": 9.851249305385175e-06, + "loss": 0.023, + "step": 31733 + }, + { + "epoch": 4.16, + "learning_rate": 9.850719039204575e-06, + "loss": 0.0231, + "step": 31734 + }, + { + "epoch": 4.16, + "learning_rate": 9.850188773443817e-06, + "loss": 0.0201, + "step": 31735 + }, + { + "epoch": 4.16, + "learning_rate": 9.849658508104397e-06, + "loss": 0.0275, + "step": 31736 + }, + { + "epoch": 4.16, + "learning_rate": 9.849128243187803e-06, + "loss": 0.0221, + "step": 31737 + }, + { + "epoch": 4.16, + "learning_rate": 9.848597978695526e-06, + "loss": 0.0242, + "step": 31738 + }, + { + "epoch": 4.16, + "learning_rate": 9.848067714629063e-06, + "loss": 0.0221, + "step": 31739 + }, + { + "epoch": 4.16, + "learning_rate": 9.847537450989897e-06, + "loss": 0.0193, + "step": 31740 + }, + { + "epoch": 4.16, + "learning_rate": 9.847007187779528e-06, + "loss": 0.0212, + "step": 31741 + }, + { + "epoch": 4.16, + "learning_rate": 9.846476924999439e-06, + "loss": 0.0211, + "step": 31742 + }, + { + "epoch": 4.16, + "learning_rate": 9.845946662651128e-06, + "loss": 0.0181, + "step": 31743 + }, + { + "epoch": 4.16, + "learning_rate": 9.845416400736083e-06, + "loss": 0.0198, + "step": 31744 + }, + { + "epoch": 4.16, + "learning_rate": 9.844886139255797e-06, + "loss": 0.0178, + "step": 31745 + }, + { + "epoch": 4.16, + "learning_rate": 9.84435587821176e-06, + "loss": 0.0246, + "step": 31746 + }, + { + "epoch": 4.16, + "learning_rate": 9.843825617605467e-06, + "loss": 0.0238, + "step": 31747 + }, + { + "epoch": 4.16, + "learning_rate": 9.8432953574384e-06, + "loss": 0.0229, + "step": 31748 + }, + { + "epoch": 4.16, + "learning_rate": 9.842765097712063e-06, + "loss": 0.0218, + "step": 31749 + }, + { + "epoch": 4.16, + "learning_rate": 9.842234838427936e-06, + "loss": 0.0227, + "step": 31750 + }, + { + "epoch": 4.16, + "learning_rate": 9.841704579587519e-06, + "loss": 0.0251, + "step": 31751 + }, + { + "epoch": 4.16, + "learning_rate": 9.841174321192295e-06, + "loss": 0.0175, + "step": 31752 + }, + { + "epoch": 4.16, + "learning_rate": 9.84064406324376e-06, + "loss": 0.0193, + "step": 31753 + }, + { + "epoch": 4.16, + "learning_rate": 9.840113805743406e-06, + "loss": 0.0173, + "step": 31754 + }, + { + "epoch": 4.16, + "learning_rate": 9.839583548692725e-06, + "loss": 0.0218, + "step": 31755 + }, + { + "epoch": 4.16, + "learning_rate": 9.839053292093207e-06, + "loss": 0.026, + "step": 31756 + }, + { + "epoch": 4.16, + "learning_rate": 9.838523035946342e-06, + "loss": 0.0182, + "step": 31757 + }, + { + "epoch": 4.16, + "learning_rate": 9.837992780253621e-06, + "loss": 0.0187, + "step": 31758 + }, + { + "epoch": 4.16, + "learning_rate": 9.837462525016538e-06, + "loss": 0.0258, + "step": 31759 + }, + { + "epoch": 4.16, + "learning_rate": 9.836932270236585e-06, + "loss": 0.0192, + "step": 31760 + }, + { + "epoch": 4.16, + "learning_rate": 9.836402015915248e-06, + "loss": 0.0229, + "step": 31761 + }, + { + "epoch": 4.16, + "learning_rate": 9.835871762054024e-06, + "loss": 0.0216, + "step": 31762 + }, + { + "epoch": 4.16, + "learning_rate": 9.8353415086544e-06, + "loss": 0.0264, + "step": 31763 + }, + { + "epoch": 4.16, + "learning_rate": 9.834811255717867e-06, + "loss": 0.0196, + "step": 31764 + }, + { + "epoch": 4.16, + "learning_rate": 9.834281003245925e-06, + "loss": 0.0249, + "step": 31765 + }, + { + "epoch": 4.16, + "learning_rate": 9.833750751240054e-06, + "loss": 0.022, + "step": 31766 + }, + { + "epoch": 4.16, + "learning_rate": 9.833220499701752e-06, + "loss": 0.0232, + "step": 31767 + }, + { + "epoch": 4.16, + "learning_rate": 9.832690248632508e-06, + "loss": 0.0218, + "step": 31768 + }, + { + "epoch": 4.16, + "learning_rate": 9.832159998033814e-06, + "loss": 0.0192, + "step": 31769 + }, + { + "epoch": 4.16, + "learning_rate": 9.831629747907162e-06, + "loss": 0.0222, + "step": 31770 + }, + { + "epoch": 4.16, + "learning_rate": 9.831099498254042e-06, + "loss": 0.021, + "step": 31771 + }, + { + "epoch": 4.16, + "learning_rate": 9.830569249075946e-06, + "loss": 0.0276, + "step": 31772 + }, + { + "epoch": 4.16, + "learning_rate": 9.830039000374362e-06, + "loss": 0.0163, + "step": 31773 + }, + { + "epoch": 4.16, + "learning_rate": 9.829508752150786e-06, + "loss": 0.0195, + "step": 31774 + }, + { + "epoch": 4.16, + "learning_rate": 9.828978504406706e-06, + "loss": 0.0185, + "step": 31775 + }, + { + "epoch": 4.16, + "learning_rate": 9.828448257143618e-06, + "loss": 0.0235, + "step": 31776 + }, + { + "epoch": 4.16, + "learning_rate": 9.827918010363008e-06, + "loss": 0.0228, + "step": 31777 + }, + { + "epoch": 4.16, + "learning_rate": 9.827387764066371e-06, + "loss": 0.0206, + "step": 31778 + }, + { + "epoch": 4.16, + "learning_rate": 9.826857518255194e-06, + "loss": 0.0232, + "step": 31779 + }, + { + "epoch": 4.16, + "learning_rate": 9.826327272930975e-06, + "loss": 0.0243, + "step": 31780 + }, + { + "epoch": 4.16, + "learning_rate": 9.8257970280952e-06, + "loss": 0.0221, + "step": 31781 + }, + { + "epoch": 4.16, + "learning_rate": 9.82526678374936e-06, + "loss": 0.0229, + "step": 31782 + }, + { + "epoch": 4.16, + "learning_rate": 9.82473653989495e-06, + "loss": 0.0254, + "step": 31783 + }, + { + "epoch": 4.16, + "learning_rate": 9.824206296533458e-06, + "loss": 0.0198, + "step": 31784 + }, + { + "epoch": 4.16, + "learning_rate": 9.823676053666376e-06, + "loss": 0.0183, + "step": 31785 + }, + { + "epoch": 4.16, + "learning_rate": 9.823145811295195e-06, + "loss": 0.0174, + "step": 31786 + }, + { + "epoch": 4.16, + "learning_rate": 9.822615569421406e-06, + "loss": 0.0276, + "step": 31787 + }, + { + "epoch": 4.16, + "learning_rate": 9.822085328046502e-06, + "loss": 0.0201, + "step": 31788 + }, + { + "epoch": 4.16, + "learning_rate": 9.821555087171977e-06, + "loss": 0.0191, + "step": 31789 + }, + { + "epoch": 4.16, + "learning_rate": 9.821024846799316e-06, + "loss": 0.0242, + "step": 31790 + }, + { + "epoch": 4.16, + "learning_rate": 9.820494606930015e-06, + "loss": 0.0198, + "step": 31791 + }, + { + "epoch": 4.16, + "learning_rate": 9.81996436756556e-06, + "loss": 0.0222, + "step": 31792 + }, + { + "epoch": 4.16, + "learning_rate": 9.81943412870745e-06, + "loss": 0.0262, + "step": 31793 + }, + { + "epoch": 4.16, + "learning_rate": 9.81890389035717e-06, + "loss": 0.025, + "step": 31794 + }, + { + "epoch": 4.16, + "learning_rate": 9.818373652516214e-06, + "loss": 0.0194, + "step": 31795 + }, + { + "epoch": 4.17, + "learning_rate": 9.817843415186071e-06, + "loss": 0.0238, + "step": 31796 + }, + { + "epoch": 4.17, + "learning_rate": 9.817313178368229e-06, + "loss": 0.0243, + "step": 31797 + }, + { + "epoch": 4.17, + "learning_rate": 9.816782942064189e-06, + "loss": 0.0237, + "step": 31798 + }, + { + "epoch": 4.17, + "learning_rate": 9.816252706275438e-06, + "loss": 0.0221, + "step": 31799 + }, + { + "epoch": 4.17, + "learning_rate": 9.815722471003465e-06, + "loss": 0.0265, + "step": 31800 + }, + { + "epoch": 4.17, + "learning_rate": 9.815192236249766e-06, + "loss": 0.0239, + "step": 31801 + }, + { + "epoch": 4.17, + "learning_rate": 9.814662002015825e-06, + "loss": 0.0248, + "step": 31802 + }, + { + "epoch": 4.17, + "learning_rate": 9.814131768303139e-06, + "loss": 0.0223, + "step": 31803 + }, + { + "epoch": 4.17, + "learning_rate": 9.813601535113198e-06, + "loss": 0.02, + "step": 31804 + }, + { + "epoch": 4.17, + "learning_rate": 9.81307130244749e-06, + "loss": 0.021, + "step": 31805 + }, + { + "epoch": 4.17, + "learning_rate": 9.812541070307513e-06, + "loss": 0.0212, + "step": 31806 + }, + { + "epoch": 4.17, + "learning_rate": 9.81201083869475e-06, + "loss": 0.0233, + "step": 31807 + }, + { + "epoch": 4.17, + "learning_rate": 9.811480607610697e-06, + "loss": 0.0202, + "step": 31808 + }, + { + "epoch": 4.17, + "learning_rate": 9.810950377056845e-06, + "loss": 0.018, + "step": 31809 + }, + { + "epoch": 4.17, + "learning_rate": 9.810420147034687e-06, + "loss": 0.0232, + "step": 31810 + }, + { + "epoch": 4.17, + "learning_rate": 9.80988991754571e-06, + "loss": 0.0235, + "step": 31811 + }, + { + "epoch": 4.17, + "learning_rate": 9.809359688591412e-06, + "loss": 0.0216, + "step": 31812 + }, + { + "epoch": 4.17, + "learning_rate": 9.808829460173276e-06, + "loss": 0.0208, + "step": 31813 + }, + { + "epoch": 4.17, + "learning_rate": 9.8082992322928e-06, + "loss": 0.0211, + "step": 31814 + }, + { + "epoch": 4.17, + "learning_rate": 9.807769004951468e-06, + "loss": 0.0176, + "step": 31815 + }, + { + "epoch": 4.17, + "learning_rate": 9.80723877815078e-06, + "loss": 0.0247, + "step": 31816 + }, + { + "epoch": 4.17, + "learning_rate": 9.806708551892219e-06, + "loss": 0.0179, + "step": 31817 + }, + { + "epoch": 4.17, + "learning_rate": 9.80617832617728e-06, + "loss": 0.0202, + "step": 31818 + }, + { + "epoch": 4.17, + "learning_rate": 9.805648101007452e-06, + "loss": 0.0272, + "step": 31819 + }, + { + "epoch": 4.17, + "learning_rate": 9.805117876384233e-06, + "loss": 0.018, + "step": 31820 + }, + { + "epoch": 4.17, + "learning_rate": 9.804587652309106e-06, + "loss": 0.0178, + "step": 31821 + }, + { + "epoch": 4.17, + "learning_rate": 9.80405742878357e-06, + "loss": 0.0169, + "step": 31822 + }, + { + "epoch": 4.17, + "learning_rate": 9.803527205809108e-06, + "loss": 0.0211, + "step": 31823 + }, + { + "epoch": 4.17, + "learning_rate": 9.802996983387217e-06, + "loss": 0.0274, + "step": 31824 + }, + { + "epoch": 4.17, + "learning_rate": 9.80246676151939e-06, + "loss": 0.0206, + "step": 31825 + }, + { + "epoch": 4.17, + "learning_rate": 9.801936540207109e-06, + "loss": 0.0245, + "step": 31826 + }, + { + "epoch": 4.17, + "learning_rate": 9.801406319451875e-06, + "loss": 0.0189, + "step": 31827 + }, + { + "epoch": 4.17, + "learning_rate": 9.800876099255174e-06, + "loss": 0.0204, + "step": 31828 + }, + { + "epoch": 4.17, + "learning_rate": 9.800345879618498e-06, + "loss": 0.0227, + "step": 31829 + }, + { + "epoch": 4.17, + "learning_rate": 9.799815660543336e-06, + "loss": 0.0226, + "step": 31830 + }, + { + "epoch": 4.17, + "learning_rate": 9.799285442031184e-06, + "loss": 0.0235, + "step": 31831 + }, + { + "epoch": 4.17, + "learning_rate": 9.798755224083529e-06, + "loss": 0.0215, + "step": 31832 + }, + { + "epoch": 4.17, + "learning_rate": 9.798225006701868e-06, + "loss": 0.0219, + "step": 31833 + }, + { + "epoch": 4.17, + "learning_rate": 9.797694789887686e-06, + "loss": 0.0179, + "step": 31834 + }, + { + "epoch": 4.17, + "learning_rate": 9.797164573642479e-06, + "loss": 0.0215, + "step": 31835 + }, + { + "epoch": 4.17, + "learning_rate": 9.796634357967733e-06, + "loss": 0.0192, + "step": 31836 + }, + { + "epoch": 4.17, + "learning_rate": 9.796104142864946e-06, + "loss": 0.0252, + "step": 31837 + }, + { + "epoch": 4.17, + "learning_rate": 9.7955739283356e-06, + "loss": 0.0231, + "step": 31838 + }, + { + "epoch": 4.17, + "learning_rate": 9.795043714381196e-06, + "loss": 0.0237, + "step": 31839 + }, + { + "epoch": 4.17, + "learning_rate": 9.794513501003218e-06, + "loss": 0.0173, + "step": 31840 + }, + { + "epoch": 4.17, + "learning_rate": 9.793983288203158e-06, + "loss": 0.0218, + "step": 31841 + }, + { + "epoch": 4.17, + "learning_rate": 9.793453075982511e-06, + "loss": 0.0195, + "step": 31842 + }, + { + "epoch": 4.17, + "learning_rate": 9.79292286434277e-06, + "loss": 0.022, + "step": 31843 + }, + { + "epoch": 4.17, + "learning_rate": 9.792392653285417e-06, + "loss": 0.0239, + "step": 31844 + }, + { + "epoch": 4.17, + "learning_rate": 9.791862442811953e-06, + "loss": 0.0236, + "step": 31845 + }, + { + "epoch": 4.17, + "learning_rate": 9.791332232923862e-06, + "loss": 0.025, + "step": 31846 + }, + { + "epoch": 4.17, + "learning_rate": 9.790802023622636e-06, + "loss": 0.0242, + "step": 31847 + }, + { + "epoch": 4.17, + "learning_rate": 9.790271814909772e-06, + "loss": 0.022, + "step": 31848 + }, + { + "epoch": 4.17, + "learning_rate": 9.789741606786757e-06, + "loss": 0.0221, + "step": 31849 + }, + { + "epoch": 4.17, + "learning_rate": 9.789211399255081e-06, + "loss": 0.0238, + "step": 31850 + }, + { + "epoch": 4.17, + "learning_rate": 9.788681192316237e-06, + "loss": 0.0235, + "step": 31851 + }, + { + "epoch": 4.17, + "learning_rate": 9.788150985971715e-06, + "loss": 0.0212, + "step": 31852 + }, + { + "epoch": 4.17, + "learning_rate": 9.787620780223007e-06, + "loss": 0.0212, + "step": 31853 + }, + { + "epoch": 4.17, + "learning_rate": 9.787090575071607e-06, + "loss": 0.0258, + "step": 31854 + }, + { + "epoch": 4.17, + "learning_rate": 9.786560370519002e-06, + "loss": 0.0243, + "step": 31855 + }, + { + "epoch": 4.17, + "learning_rate": 9.786030166566684e-06, + "loss": 0.024, + "step": 31856 + }, + { + "epoch": 4.17, + "learning_rate": 9.785499963216146e-06, + "loss": 0.0221, + "step": 31857 + }, + { + "epoch": 4.17, + "learning_rate": 9.78496976046888e-06, + "loss": 0.0212, + "step": 31858 + }, + { + "epoch": 4.17, + "learning_rate": 9.784439558326369e-06, + "loss": 0.0232, + "step": 31859 + }, + { + "epoch": 4.17, + "learning_rate": 9.783909356790114e-06, + "loss": 0.0195, + "step": 31860 + }, + { + "epoch": 4.17, + "learning_rate": 9.783379155861604e-06, + "loss": 0.0204, + "step": 31861 + }, + { + "epoch": 4.17, + "learning_rate": 9.782848955542323e-06, + "loss": 0.0222, + "step": 31862 + }, + { + "epoch": 4.17, + "learning_rate": 9.78231875583377e-06, + "loss": 0.0204, + "step": 31863 + }, + { + "epoch": 4.17, + "learning_rate": 9.781788556737437e-06, + "loss": 0.0193, + "step": 31864 + }, + { + "epoch": 4.17, + "learning_rate": 9.781258358254808e-06, + "loss": 0.0196, + "step": 31865 + }, + { + "epoch": 4.17, + "learning_rate": 9.780728160387384e-06, + "loss": 0.0218, + "step": 31866 + }, + { + "epoch": 4.17, + "learning_rate": 9.780197963136646e-06, + "loss": 0.025, + "step": 31867 + }, + { + "epoch": 4.17, + "learning_rate": 9.77966776650409e-06, + "loss": 0.0253, + "step": 31868 + }, + { + "epoch": 4.17, + "learning_rate": 9.779137570491209e-06, + "loss": 0.022, + "step": 31869 + }, + { + "epoch": 4.17, + "learning_rate": 9.778607375099488e-06, + "loss": 0.0207, + "step": 31870 + }, + { + "epoch": 4.17, + "learning_rate": 9.778077180330427e-06, + "loss": 0.022, + "step": 31871 + }, + { + "epoch": 4.18, + "learning_rate": 9.777546986185507e-06, + "loss": 0.021, + "step": 31872 + }, + { + "epoch": 4.18, + "learning_rate": 9.777016792666222e-06, + "loss": 0.0216, + "step": 31873 + }, + { + "epoch": 4.18, + "learning_rate": 9.77648659977407e-06, + "loss": 0.0222, + "step": 31874 + }, + { + "epoch": 4.18, + "learning_rate": 9.77595640751054e-06, + "loss": 0.0197, + "step": 31875 + }, + { + "epoch": 4.18, + "learning_rate": 9.775426215877117e-06, + "loss": 0.0232, + "step": 31876 + }, + { + "epoch": 4.18, + "learning_rate": 9.774896024875297e-06, + "loss": 0.0235, + "step": 31877 + }, + { + "epoch": 4.18, + "learning_rate": 9.77436583450657e-06, + "loss": 0.02, + "step": 31878 + }, + { + "epoch": 4.18, + "learning_rate": 9.773835644772427e-06, + "loss": 0.0241, + "step": 31879 + }, + { + "epoch": 4.18, + "learning_rate": 9.77330545567436e-06, + "loss": 0.0287, + "step": 31880 + }, + { + "epoch": 4.18, + "learning_rate": 9.772775267213859e-06, + "loss": 0.0254, + "step": 31881 + }, + { + "epoch": 4.18, + "learning_rate": 9.772245079392413e-06, + "loss": 0.0264, + "step": 31882 + }, + { + "epoch": 4.18, + "learning_rate": 9.771714892211516e-06, + "loss": 0.0148, + "step": 31883 + }, + { + "epoch": 4.18, + "learning_rate": 9.771184705672659e-06, + "loss": 0.0157, + "step": 31884 + }, + { + "epoch": 4.18, + "learning_rate": 9.770654519777333e-06, + "loss": 0.0167, + "step": 31885 + }, + { + "epoch": 4.18, + "learning_rate": 9.77012433452703e-06, + "loss": 0.0249, + "step": 31886 + }, + { + "epoch": 4.18, + "learning_rate": 9.769594149923242e-06, + "loss": 0.0274, + "step": 31887 + }, + { + "epoch": 4.18, + "learning_rate": 9.769063965967455e-06, + "loss": 0.0255, + "step": 31888 + }, + { + "epoch": 4.18, + "learning_rate": 9.768533782661164e-06, + "loss": 0.0235, + "step": 31889 + }, + { + "epoch": 4.18, + "learning_rate": 9.768003600005861e-06, + "loss": 0.0244, + "step": 31890 + }, + { + "epoch": 4.18, + "learning_rate": 9.767473418003033e-06, + "loss": 0.0216, + "step": 31891 + }, + { + "epoch": 4.18, + "learning_rate": 9.766943236654176e-06, + "loss": 0.0258, + "step": 31892 + }, + { + "epoch": 4.18, + "learning_rate": 9.766413055960777e-06, + "loss": 0.021, + "step": 31893 + }, + { + "epoch": 4.18, + "learning_rate": 9.765882875924331e-06, + "loss": 0.0198, + "step": 31894 + }, + { + "epoch": 4.18, + "learning_rate": 9.765352696546323e-06, + "loss": 0.0146, + "step": 31895 + }, + { + "epoch": 4.18, + "learning_rate": 9.76482251782825e-06, + "loss": 0.0223, + "step": 31896 + }, + { + "epoch": 4.18, + "learning_rate": 9.7642923397716e-06, + "loss": 0.0215, + "step": 31897 + }, + { + "epoch": 4.18, + "learning_rate": 9.763762162377868e-06, + "loss": 0.0261, + "step": 31898 + }, + { + "epoch": 4.18, + "learning_rate": 9.763231985648542e-06, + "loss": 0.0299, + "step": 31899 + }, + { + "epoch": 4.18, + "learning_rate": 9.762701809585114e-06, + "loss": 0.0241, + "step": 31900 + }, + { + "epoch": 4.18, + "learning_rate": 9.762171634189072e-06, + "loss": 0.0254, + "step": 31901 + }, + { + "epoch": 4.18, + "learning_rate": 9.761641459461911e-06, + "loss": 0.0258, + "step": 31902 + }, + { + "epoch": 4.18, + "learning_rate": 9.761111285405121e-06, + "loss": 0.0212, + "step": 31903 + }, + { + "epoch": 4.18, + "learning_rate": 9.760581112020191e-06, + "loss": 0.0228, + "step": 31904 + }, + { + "epoch": 4.18, + "learning_rate": 9.760050939308616e-06, + "loss": 0.0205, + "step": 31905 + }, + { + "epoch": 4.18, + "learning_rate": 9.759520767271881e-06, + "loss": 0.0257, + "step": 31906 + }, + { + "epoch": 4.18, + "learning_rate": 9.758990595911484e-06, + "loss": 0.0254, + "step": 31907 + }, + { + "epoch": 4.18, + "learning_rate": 9.758460425228915e-06, + "loss": 0.0191, + "step": 31908 + }, + { + "epoch": 4.18, + "learning_rate": 9.757930255225661e-06, + "loss": 0.024, + "step": 31909 + }, + { + "epoch": 4.18, + "learning_rate": 9.757400085903214e-06, + "loss": 0.0234, + "step": 31910 + }, + { + "epoch": 4.18, + "learning_rate": 9.75686991726307e-06, + "loss": 0.0212, + "step": 31911 + }, + { + "epoch": 4.18, + "learning_rate": 9.756339749306714e-06, + "loss": 0.0233, + "step": 31912 + }, + { + "epoch": 4.18, + "learning_rate": 9.75580958203564e-06, + "loss": 0.0239, + "step": 31913 + }, + { + "epoch": 4.18, + "learning_rate": 9.755279415451338e-06, + "loss": 0.0225, + "step": 31914 + }, + { + "epoch": 4.18, + "learning_rate": 9.754749249555302e-06, + "loss": 0.0249, + "step": 31915 + }, + { + "epoch": 4.18, + "learning_rate": 9.754219084349017e-06, + "loss": 0.0189, + "step": 31916 + }, + { + "epoch": 4.18, + "learning_rate": 9.753688919833977e-06, + "loss": 0.021, + "step": 31917 + }, + { + "epoch": 4.18, + "learning_rate": 9.753158756011676e-06, + "loss": 0.018, + "step": 31918 + }, + { + "epoch": 4.18, + "learning_rate": 9.752628592883605e-06, + "loss": 0.0243, + "step": 31919 + }, + { + "epoch": 4.18, + "learning_rate": 9.75209843045125e-06, + "loss": 0.0215, + "step": 31920 + }, + { + "epoch": 4.18, + "learning_rate": 9.751568268716108e-06, + "loss": 0.0204, + "step": 31921 + }, + { + "epoch": 4.18, + "learning_rate": 9.751038107679664e-06, + "loss": 0.0202, + "step": 31922 + }, + { + "epoch": 4.18, + "learning_rate": 9.750507947343414e-06, + "loss": 0.0238, + "step": 31923 + }, + { + "epoch": 4.18, + "learning_rate": 9.749977787708845e-06, + "loss": 0.0195, + "step": 31924 + }, + { + "epoch": 4.18, + "learning_rate": 9.749447628777451e-06, + "loss": 0.0234, + "step": 31925 + }, + { + "epoch": 4.18, + "learning_rate": 9.748917470550725e-06, + "loss": 0.0187, + "step": 31926 + }, + { + "epoch": 4.18, + "learning_rate": 9.74838731303015e-06, + "loss": 0.0185, + "step": 31927 + }, + { + "epoch": 4.18, + "learning_rate": 9.747857156217224e-06, + "loss": 0.0236, + "step": 31928 + }, + { + "epoch": 4.18, + "learning_rate": 9.747327000113438e-06, + "loss": 0.0173, + "step": 31929 + }, + { + "epoch": 4.18, + "learning_rate": 9.74679684472028e-06, + "loss": 0.0191, + "step": 31930 + }, + { + "epoch": 4.18, + "learning_rate": 9.746266690039245e-06, + "loss": 0.0241, + "step": 31931 + }, + { + "epoch": 4.18, + "learning_rate": 9.745736536071818e-06, + "loss": 0.0187, + "step": 31932 + }, + { + "epoch": 4.18, + "learning_rate": 9.745206382819496e-06, + "loss": 0.0225, + "step": 31933 + }, + { + "epoch": 4.18, + "learning_rate": 9.744676230283767e-06, + "loss": 0.0246, + "step": 31934 + }, + { + "epoch": 4.18, + "learning_rate": 9.744146078466121e-06, + "loss": 0.025, + "step": 31935 + }, + { + "epoch": 4.18, + "learning_rate": 9.743615927368053e-06, + "loss": 0.017, + "step": 31936 + }, + { + "epoch": 4.18, + "learning_rate": 9.74308577699105e-06, + "loss": 0.0167, + "step": 31937 + }, + { + "epoch": 4.18, + "learning_rate": 9.742555627336605e-06, + "loss": 0.0224, + "step": 31938 + }, + { + "epoch": 4.18, + "learning_rate": 9.742025478406203e-06, + "loss": 0.0222, + "step": 31939 + }, + { + "epoch": 4.18, + "learning_rate": 9.741495330201349e-06, + "loss": 0.0222, + "step": 31940 + }, + { + "epoch": 4.18, + "learning_rate": 9.74096518272352e-06, + "loss": 0.024, + "step": 31941 + }, + { + "epoch": 4.18, + "learning_rate": 9.740435035974218e-06, + "loss": 0.0216, + "step": 31942 + }, + { + "epoch": 4.18, + "learning_rate": 9.739904889954925e-06, + "loss": 0.0153, + "step": 31943 + }, + { + "epoch": 4.18, + "learning_rate": 9.739374744667138e-06, + "loss": 0.0209, + "step": 31944 + }, + { + "epoch": 4.18, + "learning_rate": 9.738844600112343e-06, + "loss": 0.0167, + "step": 31945 + }, + { + "epoch": 4.18, + "learning_rate": 9.738314456292035e-06, + "loss": 0.0171, + "step": 31946 + }, + { + "epoch": 4.18, + "learning_rate": 9.737784313207703e-06, + "loss": 0.0143, + "step": 31947 + }, + { + "epoch": 4.18, + "learning_rate": 9.737254170860838e-06, + "loss": 0.024, + "step": 31948 + }, + { + "epoch": 4.19, + "learning_rate": 9.736724029252933e-06, + "loss": 0.0277, + "step": 31949 + }, + { + "epoch": 4.19, + "learning_rate": 9.736193888385476e-06, + "loss": 0.0254, + "step": 31950 + }, + { + "epoch": 4.19, + "learning_rate": 9.73566374825996e-06, + "loss": 0.0284, + "step": 31951 + }, + { + "epoch": 4.19, + "learning_rate": 9.735133608877879e-06, + "loss": 0.0198, + "step": 31952 + }, + { + "epoch": 4.19, + "learning_rate": 9.734603470240717e-06, + "loss": 0.0216, + "step": 31953 + }, + { + "epoch": 4.19, + "learning_rate": 9.734073332349969e-06, + "loss": 0.0238, + "step": 31954 + }, + { + "epoch": 4.19, + "learning_rate": 9.733543195207127e-06, + "loss": 0.0211, + "step": 31955 + }, + { + "epoch": 4.19, + "learning_rate": 9.733013058813679e-06, + "loss": 0.0189, + "step": 31956 + }, + { + "epoch": 4.19, + "learning_rate": 9.73248292317112e-06, + "loss": 0.0226, + "step": 31957 + }, + { + "epoch": 4.19, + "learning_rate": 9.731952788280938e-06, + "loss": 0.025, + "step": 31958 + }, + { + "epoch": 4.19, + "learning_rate": 9.731422654144624e-06, + "loss": 0.0206, + "step": 31959 + }, + { + "epoch": 4.19, + "learning_rate": 9.730892520763669e-06, + "loss": 0.0224, + "step": 31960 + }, + { + "epoch": 4.19, + "learning_rate": 9.73036238813956e-06, + "loss": 0.0221, + "step": 31961 + }, + { + "epoch": 4.19, + "learning_rate": 9.729832256273796e-06, + "loss": 0.03, + "step": 31962 + }, + { + "epoch": 4.19, + "learning_rate": 9.729302125167867e-06, + "loss": 0.0185, + "step": 31963 + }, + { + "epoch": 4.19, + "learning_rate": 9.728771994823259e-06, + "loss": 0.0222, + "step": 31964 + }, + { + "epoch": 4.19, + "learning_rate": 9.728241865241467e-06, + "loss": 0.0223, + "step": 31965 + }, + { + "epoch": 4.19, + "learning_rate": 9.72771173642398e-06, + "loss": 0.0224, + "step": 31966 + }, + { + "epoch": 4.19, + "learning_rate": 9.727181608372289e-06, + "loss": 0.0232, + "step": 31967 + }, + { + "epoch": 4.19, + "learning_rate": 9.726651481087884e-06, + "loss": 0.0235, + "step": 31968 + }, + { + "epoch": 4.19, + "learning_rate": 9.726121354572258e-06, + "loss": 0.0246, + "step": 31969 + }, + { + "epoch": 4.19, + "learning_rate": 9.725591228826902e-06, + "loss": 0.0211, + "step": 31970 + }, + { + "epoch": 4.19, + "learning_rate": 9.725061103853303e-06, + "loss": 0.019, + "step": 31971 + }, + { + "epoch": 4.19, + "learning_rate": 9.724530979652955e-06, + "loss": 0.0224, + "step": 31972 + }, + { + "epoch": 4.19, + "learning_rate": 9.724000856227353e-06, + "loss": 0.0215, + "step": 31973 + }, + { + "epoch": 4.19, + "learning_rate": 9.72347073357798e-06, + "loss": 0.0212, + "step": 31974 + }, + { + "epoch": 4.19, + "learning_rate": 9.722940611706333e-06, + "loss": 0.0212, + "step": 31975 + }, + { + "epoch": 4.19, + "learning_rate": 9.722410490613902e-06, + "loss": 0.0186, + "step": 31976 + }, + { + "epoch": 4.19, + "learning_rate": 9.721880370302174e-06, + "loss": 0.0274, + "step": 31977 + }, + { + "epoch": 4.19, + "learning_rate": 9.721350250772645e-06, + "loss": 0.026, + "step": 31978 + }, + { + "epoch": 4.19, + "learning_rate": 9.720820132026802e-06, + "loss": 0.0227, + "step": 31979 + }, + { + "epoch": 4.19, + "learning_rate": 9.72029001406614e-06, + "loss": 0.0231, + "step": 31980 + }, + { + "epoch": 4.19, + "learning_rate": 9.719759896892143e-06, + "loss": 0.0192, + "step": 31981 + }, + { + "epoch": 4.19, + "learning_rate": 9.719229780506311e-06, + "loss": 0.0248, + "step": 31982 + }, + { + "epoch": 4.19, + "learning_rate": 9.718699664910123e-06, + "loss": 0.019, + "step": 31983 + }, + { + "epoch": 4.19, + "learning_rate": 9.718169550105085e-06, + "loss": 0.0208, + "step": 31984 + }, + { + "epoch": 4.19, + "learning_rate": 9.717639436092675e-06, + "loss": 0.0201, + "step": 31985 + }, + { + "epoch": 4.19, + "learning_rate": 9.717109322874392e-06, + "loss": 0.0226, + "step": 31986 + }, + { + "epoch": 4.19, + "learning_rate": 9.716579210451724e-06, + "loss": 0.0263, + "step": 31987 + }, + { + "epoch": 4.19, + "learning_rate": 9.716049098826162e-06, + "loss": 0.0183, + "step": 31988 + }, + { + "epoch": 4.19, + "learning_rate": 9.715518987999196e-06, + "loss": 0.0222, + "step": 31989 + }, + { + "epoch": 4.19, + "learning_rate": 9.714988877972316e-06, + "loss": 0.0257, + "step": 31990 + }, + { + "epoch": 4.19, + "learning_rate": 9.71445876874702e-06, + "loss": 0.0225, + "step": 31991 + }, + { + "epoch": 4.19, + "learning_rate": 9.713928660324788e-06, + "loss": 0.0218, + "step": 31992 + }, + { + "epoch": 4.19, + "learning_rate": 9.713398552707115e-06, + "loss": 0.0209, + "step": 31993 + }, + { + "epoch": 4.19, + "learning_rate": 9.712868445895498e-06, + "loss": 0.0188, + "step": 31994 + }, + { + "epoch": 4.19, + "learning_rate": 9.712338339891422e-06, + "loss": 0.016, + "step": 31995 + }, + { + "epoch": 4.19, + "learning_rate": 9.71180823469638e-06, + "loss": 0.0226, + "step": 31996 + }, + { + "epoch": 4.19, + "learning_rate": 9.71127813031186e-06, + "loss": 0.0234, + "step": 31997 + }, + { + "epoch": 4.19, + "learning_rate": 9.710748026739354e-06, + "loss": 0.0175, + "step": 31998 + }, + { + "epoch": 4.19, + "learning_rate": 9.710217923980358e-06, + "loss": 0.023, + "step": 31999 + }, + { + "epoch": 4.19, + "learning_rate": 9.709687822036355e-06, + "loss": 0.0299, + "step": 32000 + }, + { + "epoch": 4.19, + "learning_rate": 9.709157720908843e-06, + "loss": 0.0255, + "step": 32001 + }, + { + "epoch": 4.19, + "learning_rate": 9.708627620599305e-06, + "loss": 0.0256, + "step": 32002 + }, + { + "epoch": 4.19, + "learning_rate": 9.708097521109241e-06, + "loss": 0.0255, + "step": 32003 + }, + { + "epoch": 4.19, + "learning_rate": 9.70756742244013e-06, + "loss": 0.0276, + "step": 32004 + }, + { + "epoch": 4.19, + "learning_rate": 9.707037324593476e-06, + "loss": 0.0196, + "step": 32005 + }, + { + "epoch": 4.19, + "learning_rate": 9.706507227570762e-06, + "loss": 0.0197, + "step": 32006 + }, + { + "epoch": 4.19, + "learning_rate": 9.705977131373482e-06, + "loss": 0.0196, + "step": 32007 + }, + { + "epoch": 4.19, + "learning_rate": 9.705447036003124e-06, + "loss": 0.019, + "step": 32008 + }, + { + "epoch": 4.19, + "learning_rate": 9.704916941461183e-06, + "loss": 0.0202, + "step": 32009 + }, + { + "epoch": 4.19, + "learning_rate": 9.704386847749146e-06, + "loss": 0.0231, + "step": 32010 + }, + { + "epoch": 4.19, + "learning_rate": 9.703856754868504e-06, + "loss": 0.0206, + "step": 32011 + }, + { + "epoch": 4.19, + "learning_rate": 9.703326662820752e-06, + "loss": 0.0143, + "step": 32012 + }, + { + "epoch": 4.19, + "learning_rate": 9.702796571607376e-06, + "loss": 0.0216, + "step": 32013 + }, + { + "epoch": 4.19, + "learning_rate": 9.702266481229869e-06, + "loss": 0.0195, + "step": 32014 + }, + { + "epoch": 4.19, + "learning_rate": 9.70173639168972e-06, + "loss": 0.0239, + "step": 32015 + }, + { + "epoch": 4.19, + "learning_rate": 9.701206302988421e-06, + "loss": 0.0252, + "step": 32016 + }, + { + "epoch": 4.19, + "learning_rate": 9.700676215127467e-06, + "loss": 0.016, + "step": 32017 + }, + { + "epoch": 4.19, + "learning_rate": 9.700146128108344e-06, + "loss": 0.0199, + "step": 32018 + }, + { + "epoch": 4.19, + "learning_rate": 9.699616041932544e-06, + "loss": 0.0225, + "step": 32019 + }, + { + "epoch": 4.19, + "learning_rate": 9.699085956601559e-06, + "loss": 0.0178, + "step": 32020 + }, + { + "epoch": 4.19, + "learning_rate": 9.698555872116877e-06, + "loss": 0.0219, + "step": 32021 + }, + { + "epoch": 4.19, + "learning_rate": 9.698025788479992e-06, + "loss": 0.0227, + "step": 32022 + }, + { + "epoch": 4.19, + "learning_rate": 9.697495705692393e-06, + "loss": 0.0249, + "step": 32023 + }, + { + "epoch": 4.19, + "learning_rate": 9.69696562375557e-06, + "loss": 0.0178, + "step": 32024 + }, + { + "epoch": 4.2, + "learning_rate": 9.696435542671016e-06, + "loss": 0.0225, + "step": 32025 + }, + { + "epoch": 4.2, + "learning_rate": 9.695905462440218e-06, + "loss": 0.0205, + "step": 32026 + }, + { + "epoch": 4.2, + "learning_rate": 9.69537538306467e-06, + "loss": 0.0257, + "step": 32027 + }, + { + "epoch": 4.2, + "learning_rate": 9.694845304545867e-06, + "loss": 0.0156, + "step": 32028 + }, + { + "epoch": 4.2, + "learning_rate": 9.694315226885292e-06, + "loss": 0.0204, + "step": 32029 + }, + { + "epoch": 4.2, + "learning_rate": 9.693785150084442e-06, + "loss": 0.0246, + "step": 32030 + }, + { + "epoch": 4.2, + "learning_rate": 9.693255074144803e-06, + "loss": 0.0212, + "step": 32031 + }, + { + "epoch": 4.2, + "learning_rate": 9.69272499906787e-06, + "loss": 0.0237, + "step": 32032 + }, + { + "epoch": 4.2, + "learning_rate": 9.692194924855128e-06, + "loss": 0.0189, + "step": 32033 + }, + { + "epoch": 4.2, + "learning_rate": 9.691664851508072e-06, + "loss": 0.0194, + "step": 32034 + }, + { + "epoch": 4.2, + "learning_rate": 9.691134779028195e-06, + "loss": 0.0248, + "step": 32035 + }, + { + "epoch": 4.2, + "learning_rate": 9.690604707416984e-06, + "loss": 0.0212, + "step": 32036 + }, + { + "epoch": 4.2, + "learning_rate": 9.690074636675926e-06, + "loss": 0.0232, + "step": 32037 + }, + { + "epoch": 4.2, + "learning_rate": 9.689544566806521e-06, + "loss": 0.0235, + "step": 32038 + }, + { + "epoch": 4.2, + "learning_rate": 9.689014497810253e-06, + "loss": 0.0207, + "step": 32039 + }, + { + "epoch": 4.2, + "learning_rate": 9.688484429688616e-06, + "loss": 0.0197, + "step": 32040 + }, + { + "epoch": 4.2, + "learning_rate": 9.687954362443105e-06, + "loss": 0.0197, + "step": 32041 + }, + { + "epoch": 4.2, + "learning_rate": 9.6874242960752e-06, + "loss": 0.0203, + "step": 32042 + }, + { + "epoch": 4.2, + "learning_rate": 9.686894230586401e-06, + "loss": 0.0203, + "step": 32043 + }, + { + "epoch": 4.2, + "learning_rate": 9.686364165978192e-06, + "loss": 0.0235, + "step": 32044 + }, + { + "epoch": 4.2, + "learning_rate": 9.68583410225207e-06, + "loss": 0.0241, + "step": 32045 + }, + { + "epoch": 4.2, + "learning_rate": 9.68530403940952e-06, + "loss": 0.0256, + "step": 32046 + }, + { + "epoch": 4.2, + "learning_rate": 9.68477397745204e-06, + "loss": 0.0208, + "step": 32047 + }, + { + "epoch": 4.2, + "learning_rate": 9.684243916381109e-06, + "loss": 0.0208, + "step": 32048 + }, + { + "epoch": 4.2, + "learning_rate": 9.68371385619823e-06, + "loss": 0.0242, + "step": 32049 + }, + { + "epoch": 4.2, + "learning_rate": 9.683183796904888e-06, + "loss": 0.0206, + "step": 32050 + }, + { + "epoch": 4.2, + "learning_rate": 9.682653738502576e-06, + "loss": 0.016, + "step": 32051 + }, + { + "epoch": 4.2, + "learning_rate": 9.682123680992783e-06, + "loss": 0.0256, + "step": 32052 + }, + { + "epoch": 4.2, + "learning_rate": 9.681593624377001e-06, + "loss": 0.0206, + "step": 32053 + }, + { + "epoch": 4.2, + "learning_rate": 9.68106356865672e-06, + "loss": 0.0191, + "step": 32054 + }, + { + "epoch": 4.2, + "learning_rate": 9.680533513833429e-06, + "loss": 0.0237, + "step": 32055 + }, + { + "epoch": 4.2, + "learning_rate": 9.680003459908621e-06, + "loss": 0.0238, + "step": 32056 + }, + { + "epoch": 4.2, + "learning_rate": 9.679473406883786e-06, + "loss": 0.0221, + "step": 32057 + }, + { + "epoch": 4.2, + "learning_rate": 9.678943354760418e-06, + "loss": 0.0305, + "step": 32058 + }, + { + "epoch": 4.2, + "learning_rate": 9.678413303539997e-06, + "loss": 0.0207, + "step": 32059 + }, + { + "epoch": 4.2, + "learning_rate": 9.677883253224026e-06, + "loss": 0.0209, + "step": 32060 + }, + { + "epoch": 4.2, + "learning_rate": 9.677353203813992e-06, + "loss": 0.0201, + "step": 32061 + }, + { + "epoch": 4.2, + "learning_rate": 9.676823155311385e-06, + "loss": 0.0206, + "step": 32062 + }, + { + "epoch": 4.2, + "learning_rate": 9.676293107717697e-06, + "loss": 0.0226, + "step": 32063 + }, + { + "epoch": 4.2, + "learning_rate": 9.675763061034416e-06, + "loss": 0.0225, + "step": 32064 + }, + { + "epoch": 4.2, + "learning_rate": 9.675233015263033e-06, + "loss": 0.0194, + "step": 32065 + }, + { + "epoch": 4.2, + "learning_rate": 9.674702970405042e-06, + "loss": 0.0179, + "step": 32066 + }, + { + "epoch": 4.2, + "learning_rate": 9.67417292646193e-06, + "loss": 0.0216, + "step": 32067 + }, + { + "epoch": 4.2, + "learning_rate": 9.673642883435191e-06, + "loss": 0.0274, + "step": 32068 + }, + { + "epoch": 4.2, + "learning_rate": 9.67311284132631e-06, + "loss": 0.021, + "step": 32069 + }, + { + "epoch": 4.2, + "learning_rate": 9.672582800136782e-06, + "loss": 0.0223, + "step": 32070 + }, + { + "epoch": 4.2, + "learning_rate": 9.6720527598681e-06, + "loss": 0.0238, + "step": 32071 + }, + { + "epoch": 4.2, + "learning_rate": 9.671522720521754e-06, + "loss": 0.0257, + "step": 32072 + }, + { + "epoch": 4.2, + "learning_rate": 9.67099268209923e-06, + "loss": 0.0259, + "step": 32073 + }, + { + "epoch": 4.2, + "learning_rate": 9.670462644602023e-06, + "loss": 0.0209, + "step": 32074 + }, + { + "epoch": 4.2, + "learning_rate": 9.669932608031622e-06, + "loss": 0.0212, + "step": 32075 + }, + { + "epoch": 4.2, + "learning_rate": 9.669402572389515e-06, + "loss": 0.0224, + "step": 32076 + }, + { + "epoch": 4.2, + "learning_rate": 9.6688725376772e-06, + "loss": 0.021, + "step": 32077 + }, + { + "epoch": 4.2, + "learning_rate": 9.668342503896162e-06, + "loss": 0.0247, + "step": 32078 + }, + { + "epoch": 4.2, + "learning_rate": 9.667812471047892e-06, + "loss": 0.0248, + "step": 32079 + }, + { + "epoch": 4.2, + "learning_rate": 9.667282439133883e-06, + "loss": 0.0224, + "step": 32080 + }, + { + "epoch": 4.2, + "learning_rate": 9.66675240815562e-06, + "loss": 0.0255, + "step": 32081 + }, + { + "epoch": 4.2, + "learning_rate": 9.666222378114604e-06, + "loss": 0.0225, + "step": 32082 + }, + { + "epoch": 4.2, + "learning_rate": 9.665692349012317e-06, + "loss": 0.0233, + "step": 32083 + }, + { + "epoch": 4.2, + "learning_rate": 9.665162320850252e-06, + "loss": 0.0208, + "step": 32084 + }, + { + "epoch": 4.2, + "learning_rate": 9.664632293629904e-06, + "loss": 0.0175, + "step": 32085 + }, + { + "epoch": 4.2, + "learning_rate": 9.664102267352755e-06, + "loss": 0.0235, + "step": 32086 + }, + { + "epoch": 4.2, + "learning_rate": 9.663572242020305e-06, + "loss": 0.0207, + "step": 32087 + }, + { + "epoch": 4.2, + "learning_rate": 9.663042217634037e-06, + "loss": 0.0207, + "step": 32088 + }, + { + "epoch": 4.2, + "learning_rate": 9.662512194195446e-06, + "loss": 0.0234, + "step": 32089 + }, + { + "epoch": 4.2, + "learning_rate": 9.66198217170602e-06, + "loss": 0.0219, + "step": 32090 + }, + { + "epoch": 4.2, + "learning_rate": 9.661452150167252e-06, + "loss": 0.0184, + "step": 32091 + }, + { + "epoch": 4.2, + "learning_rate": 9.66092212958063e-06, + "loss": 0.0241, + "step": 32092 + }, + { + "epoch": 4.2, + "learning_rate": 9.66039210994765e-06, + "loss": 0.0221, + "step": 32093 + }, + { + "epoch": 4.2, + "learning_rate": 9.659862091269796e-06, + "loss": 0.0183, + "step": 32094 + }, + { + "epoch": 4.2, + "learning_rate": 9.659332073548564e-06, + "loss": 0.0229, + "step": 32095 + }, + { + "epoch": 4.2, + "learning_rate": 9.658802056785441e-06, + "loss": 0.024, + "step": 32096 + }, + { + "epoch": 4.2, + "learning_rate": 9.658272040981923e-06, + "loss": 0.023, + "step": 32097 + }, + { + "epoch": 4.2, + "learning_rate": 9.657742026139492e-06, + "loss": 0.0234, + "step": 32098 + }, + { + "epoch": 4.2, + "learning_rate": 9.657212012259644e-06, + "loss": 0.0185, + "step": 32099 + }, + { + "epoch": 4.2, + "learning_rate": 9.65668199934387e-06, + "loss": 0.0272, + "step": 32100 + }, + { + "epoch": 4.21, + "learning_rate": 9.656151987393659e-06, + "loss": 0.0222, + "step": 32101 + }, + { + "epoch": 4.21, + "learning_rate": 9.655621976410505e-06, + "loss": 0.0238, + "step": 32102 + }, + { + "epoch": 4.21, + "learning_rate": 9.65509196639589e-06, + "loss": 0.0243, + "step": 32103 + }, + { + "epoch": 4.21, + "learning_rate": 9.654561957351314e-06, + "loss": 0.0205, + "step": 32104 + }, + { + "epoch": 4.21, + "learning_rate": 9.654031949278261e-06, + "loss": 0.024, + "step": 32105 + }, + { + "epoch": 4.21, + "learning_rate": 9.65350194217823e-06, + "loss": 0.0192, + "step": 32106 + }, + { + "epoch": 4.21, + "learning_rate": 9.652971936052703e-06, + "loss": 0.0244, + "step": 32107 + }, + { + "epoch": 4.21, + "learning_rate": 9.652441930903176e-06, + "loss": 0.0276, + "step": 32108 + }, + { + "epoch": 4.21, + "learning_rate": 9.651911926731136e-06, + "loss": 0.0167, + "step": 32109 + }, + { + "epoch": 4.21, + "learning_rate": 9.651381923538076e-06, + "loss": 0.023, + "step": 32110 + }, + { + "epoch": 4.21, + "learning_rate": 9.650851921325486e-06, + "loss": 0.0201, + "step": 32111 + }, + { + "epoch": 4.21, + "learning_rate": 9.650321920094857e-06, + "loss": 0.0193, + "step": 32112 + }, + { + "epoch": 4.21, + "learning_rate": 9.649791919847678e-06, + "loss": 0.0168, + "step": 32113 + }, + { + "epoch": 4.21, + "learning_rate": 9.649261920585437e-06, + "loss": 0.025, + "step": 32114 + }, + { + "epoch": 4.21, + "learning_rate": 9.64873192230963e-06, + "loss": 0.0245, + "step": 32115 + }, + { + "epoch": 4.21, + "learning_rate": 9.648201925021749e-06, + "loss": 0.0231, + "step": 32116 + }, + { + "epoch": 4.21, + "learning_rate": 9.64767192872328e-06, + "loss": 0.0233, + "step": 32117 + }, + { + "epoch": 4.21, + "learning_rate": 9.647141933415715e-06, + "loss": 0.02, + "step": 32118 + }, + { + "epoch": 4.21, + "learning_rate": 9.646611939100545e-06, + "loss": 0.0196, + "step": 32119 + }, + { + "epoch": 4.21, + "learning_rate": 9.646081945779258e-06, + "loss": 0.0212, + "step": 32120 + }, + { + "epoch": 4.21, + "learning_rate": 9.645551953453347e-06, + "loss": 0.0251, + "step": 32121 + }, + { + "epoch": 4.21, + "learning_rate": 9.645021962124304e-06, + "loss": 0.0222, + "step": 32122 + }, + { + "epoch": 4.21, + "learning_rate": 9.644491971793619e-06, + "loss": 0.0219, + "step": 32123 + }, + { + "epoch": 4.21, + "learning_rate": 9.643961982462776e-06, + "loss": 0.0194, + "step": 32124 + }, + { + "epoch": 4.21, + "learning_rate": 9.643431994133273e-06, + "loss": 0.0198, + "step": 32125 + }, + { + "epoch": 4.21, + "learning_rate": 9.642902006806598e-06, + "loss": 0.0191, + "step": 32126 + }, + { + "epoch": 4.21, + "learning_rate": 9.642372020484248e-06, + "loss": 0.0176, + "step": 32127 + }, + { + "epoch": 4.21, + "learning_rate": 9.641842035167703e-06, + "loss": 0.0236, + "step": 32128 + }, + { + "epoch": 4.21, + "learning_rate": 9.641312050858458e-06, + "loss": 0.0199, + "step": 32129 + }, + { + "epoch": 4.21, + "learning_rate": 9.640782067558006e-06, + "loss": 0.0241, + "step": 32130 + }, + { + "epoch": 4.21, + "learning_rate": 9.640252085267834e-06, + "loss": 0.0192, + "step": 32131 + }, + { + "epoch": 4.21, + "learning_rate": 9.639722103989432e-06, + "loss": 0.0187, + "step": 32132 + }, + { + "epoch": 4.21, + "learning_rate": 9.639192123724297e-06, + "loss": 0.023, + "step": 32133 + }, + { + "epoch": 4.21, + "learning_rate": 9.638662144473911e-06, + "loss": 0.0205, + "step": 32134 + }, + { + "epoch": 4.21, + "learning_rate": 9.638132166239766e-06, + "loss": 0.025, + "step": 32135 + }, + { + "epoch": 4.21, + "learning_rate": 9.637602189023359e-06, + "loss": 0.0237, + "step": 32136 + }, + { + "epoch": 4.21, + "learning_rate": 9.637072212826177e-06, + "loss": 0.0263, + "step": 32137 + }, + { + "epoch": 4.21, + "learning_rate": 9.636542237649708e-06, + "loss": 0.0205, + "step": 32138 + }, + { + "epoch": 4.21, + "learning_rate": 9.636012263495448e-06, + "loss": 0.0261, + "step": 32139 + }, + { + "epoch": 4.21, + "learning_rate": 9.635482290364882e-06, + "loss": 0.0253, + "step": 32140 + }, + { + "epoch": 4.21, + "learning_rate": 9.634952318259501e-06, + "loss": 0.0289, + "step": 32141 + }, + { + "epoch": 4.21, + "learning_rate": 9.6344223471808e-06, + "loss": 0.0245, + "step": 32142 + }, + { + "epoch": 4.21, + "learning_rate": 9.633892377130265e-06, + "loss": 0.0192, + "step": 32143 + }, + { + "epoch": 4.21, + "learning_rate": 9.633362408109391e-06, + "loss": 0.0223, + "step": 32144 + }, + { + "epoch": 4.21, + "learning_rate": 9.632832440119662e-06, + "loss": 0.0244, + "step": 32145 + }, + { + "epoch": 4.21, + "learning_rate": 9.63230247316257e-06, + "loss": 0.0223, + "step": 32146 + }, + { + "epoch": 4.21, + "learning_rate": 9.631772507239612e-06, + "loss": 0.0232, + "step": 32147 + }, + { + "epoch": 4.21, + "learning_rate": 9.631242542352274e-06, + "loss": 0.0224, + "step": 32148 + }, + { + "epoch": 4.21, + "learning_rate": 9.630712578502045e-06, + "loss": 0.0214, + "step": 32149 + }, + { + "epoch": 4.21, + "learning_rate": 9.63018261569042e-06, + "loss": 0.0188, + "step": 32150 + }, + { + "epoch": 4.21, + "learning_rate": 9.629652653918884e-06, + "loss": 0.0217, + "step": 32151 + }, + { + "epoch": 4.21, + "learning_rate": 9.629122693188932e-06, + "loss": 0.0173, + "step": 32152 + }, + { + "epoch": 4.21, + "learning_rate": 9.628592733502051e-06, + "loss": 0.0229, + "step": 32153 + }, + { + "epoch": 4.21, + "learning_rate": 9.628062774859736e-06, + "loss": 0.0277, + "step": 32154 + }, + { + "epoch": 4.21, + "learning_rate": 9.627532817263473e-06, + "loss": 0.019, + "step": 32155 + }, + { + "epoch": 4.21, + "learning_rate": 9.627002860714752e-06, + "loss": 0.0173, + "step": 32156 + }, + { + "epoch": 4.21, + "learning_rate": 9.626472905215065e-06, + "loss": 0.0174, + "step": 32157 + }, + { + "epoch": 4.21, + "learning_rate": 9.625942950765906e-06, + "loss": 0.0237, + "step": 32158 + }, + { + "epoch": 4.21, + "learning_rate": 9.625412997368762e-06, + "loss": 0.0178, + "step": 32159 + }, + { + "epoch": 4.21, + "learning_rate": 9.624883045025126e-06, + "loss": 0.0231, + "step": 32160 + }, + { + "epoch": 4.21, + "learning_rate": 9.624353093736483e-06, + "loss": 0.0175, + "step": 32161 + }, + { + "epoch": 4.21, + "learning_rate": 9.62382314350433e-06, + "loss": 0.022, + "step": 32162 + }, + { + "epoch": 4.21, + "learning_rate": 9.623293194330152e-06, + "loss": 0.0234, + "step": 32163 + }, + { + "epoch": 4.21, + "learning_rate": 9.622763246215441e-06, + "loss": 0.0246, + "step": 32164 + }, + { + "epoch": 4.21, + "learning_rate": 9.622233299161692e-06, + "loss": 0.0243, + "step": 32165 + }, + { + "epoch": 4.21, + "learning_rate": 9.621703353170389e-06, + "loss": 0.0216, + "step": 32166 + }, + { + "epoch": 4.21, + "learning_rate": 9.621173408243028e-06, + "loss": 0.0233, + "step": 32167 + }, + { + "epoch": 4.21, + "learning_rate": 9.62064346438109e-06, + "loss": 0.0179, + "step": 32168 + }, + { + "epoch": 4.21, + "learning_rate": 9.620113521586077e-06, + "loss": 0.0227, + "step": 32169 + }, + { + "epoch": 4.21, + "learning_rate": 9.619583579859472e-06, + "loss": 0.0193, + "step": 32170 + }, + { + "epoch": 4.21, + "learning_rate": 9.619053639202772e-06, + "loss": 0.023, + "step": 32171 + }, + { + "epoch": 4.21, + "learning_rate": 9.61852369961746e-06, + "loss": 0.0191, + "step": 32172 + }, + { + "epoch": 4.21, + "learning_rate": 9.617993761105034e-06, + "loss": 0.0237, + "step": 32173 + }, + { + "epoch": 4.21, + "learning_rate": 9.617463823666975e-06, + "loss": 0.0214, + "step": 32174 + }, + { + "epoch": 4.21, + "learning_rate": 9.616933887304783e-06, + "loss": 0.0226, + "step": 32175 + }, + { + "epoch": 4.21, + "learning_rate": 9.61640395201994e-06, + "loss": 0.0259, + "step": 32176 + }, + { + "epoch": 4.21, + "learning_rate": 9.615874017813943e-06, + "loss": 0.0232, + "step": 32177 + }, + { + "epoch": 4.22, + "learning_rate": 9.61534408468828e-06, + "loss": 0.0219, + "step": 32178 + }, + { + "epoch": 4.22, + "learning_rate": 9.614814152644437e-06, + "loss": 0.0177, + "step": 32179 + }, + { + "epoch": 4.22, + "learning_rate": 9.614284221683911e-06, + "loss": 0.0199, + "step": 32180 + }, + { + "epoch": 4.22, + "learning_rate": 9.613754291808194e-06, + "loss": 0.0235, + "step": 32181 + }, + { + "epoch": 4.22, + "learning_rate": 9.613224363018769e-06, + "loss": 0.024, + "step": 32182 + }, + { + "epoch": 4.22, + "learning_rate": 9.612694435317132e-06, + "loss": 0.0151, + "step": 32183 + }, + { + "epoch": 4.22, + "learning_rate": 9.612164508704768e-06, + "loss": 0.0218, + "step": 32184 + }, + { + "epoch": 4.22, + "learning_rate": 9.611634583183173e-06, + "loss": 0.0292, + "step": 32185 + }, + { + "epoch": 4.22, + "learning_rate": 9.611104658753836e-06, + "loss": 0.0216, + "step": 32186 + }, + { + "epoch": 4.22, + "learning_rate": 9.610574735418243e-06, + "loss": 0.0219, + "step": 32187 + }, + { + "epoch": 4.22, + "learning_rate": 9.610044813177892e-06, + "loss": 0.0222, + "step": 32188 + }, + { + "epoch": 4.22, + "learning_rate": 9.609514892034266e-06, + "loss": 0.0194, + "step": 32189 + }, + { + "epoch": 4.22, + "learning_rate": 9.608984971988857e-06, + "loss": 0.0209, + "step": 32190 + }, + { + "epoch": 4.22, + "learning_rate": 9.608455053043158e-06, + "loss": 0.0218, + "step": 32191 + }, + { + "epoch": 4.22, + "learning_rate": 9.607925135198662e-06, + "loss": 0.0201, + "step": 32192 + }, + { + "epoch": 4.22, + "learning_rate": 9.607395218456853e-06, + "loss": 0.0208, + "step": 32193 + }, + { + "epoch": 4.22, + "learning_rate": 9.606865302819226e-06, + "loss": 0.0225, + "step": 32194 + }, + { + "epoch": 4.22, + "learning_rate": 9.606335388287268e-06, + "loss": 0.0218, + "step": 32195 + }, + { + "epoch": 4.22, + "learning_rate": 9.605805474862472e-06, + "loss": 0.0175, + "step": 32196 + }, + { + "epoch": 4.22, + "learning_rate": 9.605275562546325e-06, + "loss": 0.0215, + "step": 32197 + }, + { + "epoch": 4.22, + "learning_rate": 9.604745651340321e-06, + "loss": 0.0195, + "step": 32198 + }, + { + "epoch": 4.22, + "learning_rate": 9.604215741245948e-06, + "loss": 0.0215, + "step": 32199 + }, + { + "epoch": 4.22, + "learning_rate": 9.603685832264696e-06, + "loss": 0.0217, + "step": 32200 + }, + { + "epoch": 4.22, + "learning_rate": 9.603155924398055e-06, + "loss": 0.0222, + "step": 32201 + }, + { + "epoch": 4.22, + "learning_rate": 9.602626017647521e-06, + "loss": 0.0186, + "step": 32202 + }, + { + "epoch": 4.22, + "learning_rate": 9.602096112014578e-06, + "loss": 0.0202, + "step": 32203 + }, + { + "epoch": 4.22, + "learning_rate": 9.60156620750072e-06, + "loss": 0.0189, + "step": 32204 + }, + { + "epoch": 4.22, + "learning_rate": 9.601036304107436e-06, + "loss": 0.0224, + "step": 32205 + }, + { + "epoch": 4.22, + "learning_rate": 9.600506401836213e-06, + "loss": 0.0188, + "step": 32206 + }, + { + "epoch": 4.22, + "learning_rate": 9.59997650068855e-06, + "loss": 0.0248, + "step": 32207 + }, + { + "epoch": 4.22, + "learning_rate": 9.599446600665928e-06, + "loss": 0.0249, + "step": 32208 + }, + { + "epoch": 4.22, + "learning_rate": 9.598916701769842e-06, + "loss": 0.0213, + "step": 32209 + }, + { + "epoch": 4.22, + "learning_rate": 9.59838680400178e-06, + "loss": 0.0192, + "step": 32210 + }, + { + "epoch": 4.22, + "learning_rate": 9.597856907363236e-06, + "loss": 0.0194, + "step": 32211 + }, + { + "epoch": 4.22, + "learning_rate": 9.597327011855694e-06, + "loss": 0.0219, + "step": 32212 + }, + { + "epoch": 4.22, + "learning_rate": 9.59679711748065e-06, + "loss": 0.0205, + "step": 32213 + }, + { + "epoch": 4.22, + "learning_rate": 9.596267224239594e-06, + "loss": 0.0186, + "step": 32214 + }, + { + "epoch": 4.22, + "learning_rate": 9.595737332134016e-06, + "loss": 0.0274, + "step": 32215 + }, + { + "epoch": 4.22, + "learning_rate": 9.595207441165403e-06, + "loss": 0.0204, + "step": 32216 + }, + { + "epoch": 4.22, + "learning_rate": 9.59467755133525e-06, + "loss": 0.027, + "step": 32217 + }, + { + "epoch": 4.22, + "learning_rate": 9.594147662645042e-06, + "loss": 0.0245, + "step": 32218 + }, + { + "epoch": 4.22, + "learning_rate": 9.593617775096276e-06, + "loss": 0.0198, + "step": 32219 + }, + { + "epoch": 4.22, + "learning_rate": 9.593087888690435e-06, + "loss": 0.0236, + "step": 32220 + }, + { + "epoch": 4.22, + "learning_rate": 9.592558003429012e-06, + "loss": 0.0235, + "step": 32221 + }, + { + "epoch": 4.22, + "learning_rate": 9.5920281193135e-06, + "loss": 0.0218, + "step": 32222 + }, + { + "epoch": 4.22, + "learning_rate": 9.591498236345383e-06, + "loss": 0.0183, + "step": 32223 + }, + { + "epoch": 4.22, + "learning_rate": 9.59096835452616e-06, + "loss": 0.0202, + "step": 32224 + }, + { + "epoch": 4.22, + "learning_rate": 9.590438473857315e-06, + "loss": 0.0233, + "step": 32225 + }, + { + "epoch": 4.22, + "learning_rate": 9.58990859434034e-06, + "loss": 0.0195, + "step": 32226 + }, + { + "epoch": 4.22, + "learning_rate": 9.589378715976726e-06, + "loss": 0.0245, + "step": 32227 + }, + { + "epoch": 4.22, + "learning_rate": 9.588848838767964e-06, + "loss": 0.0236, + "step": 32228 + }, + { + "epoch": 4.22, + "learning_rate": 9.58831896271554e-06, + "loss": 0.0227, + "step": 32229 + }, + { + "epoch": 4.22, + "learning_rate": 9.587789087820948e-06, + "loss": 0.0185, + "step": 32230 + }, + { + "epoch": 4.22, + "learning_rate": 9.587259214085678e-06, + "loss": 0.0237, + "step": 32231 + }, + { + "epoch": 4.22, + "learning_rate": 9.586729341511221e-06, + "loss": 0.021, + "step": 32232 + }, + { + "epoch": 4.22, + "learning_rate": 9.586199470099063e-06, + "loss": 0.0224, + "step": 32233 + }, + { + "epoch": 4.22, + "learning_rate": 9.585669599850694e-06, + "loss": 0.02, + "step": 32234 + }, + { + "epoch": 4.22, + "learning_rate": 9.58513973076761e-06, + "loss": 0.0233, + "step": 32235 + }, + { + "epoch": 4.22, + "learning_rate": 9.5846098628513e-06, + "loss": 0.0178, + "step": 32236 + }, + { + "epoch": 4.22, + "learning_rate": 9.584079996103251e-06, + "loss": 0.0232, + "step": 32237 + }, + { + "epoch": 4.22, + "learning_rate": 9.583550130524957e-06, + "loss": 0.0238, + "step": 32238 + }, + { + "epoch": 4.22, + "learning_rate": 9.583020266117904e-06, + "loss": 0.0223, + "step": 32239 + }, + { + "epoch": 4.22, + "learning_rate": 9.582490402883586e-06, + "loss": 0.0267, + "step": 32240 + }, + { + "epoch": 4.22, + "learning_rate": 9.58196054082349e-06, + "loss": 0.0241, + "step": 32241 + }, + { + "epoch": 4.22, + "learning_rate": 9.581430679939108e-06, + "loss": 0.0266, + "step": 32242 + }, + { + "epoch": 4.22, + "learning_rate": 9.58090082023193e-06, + "loss": 0.0257, + "step": 32243 + }, + { + "epoch": 4.22, + "learning_rate": 9.580370961703446e-06, + "loss": 0.0261, + "step": 32244 + }, + { + "epoch": 4.22, + "learning_rate": 9.579841104355145e-06, + "loss": 0.0208, + "step": 32245 + }, + { + "epoch": 4.22, + "learning_rate": 9.579311248188519e-06, + "loss": 0.0211, + "step": 32246 + }, + { + "epoch": 4.22, + "learning_rate": 9.57878139320506e-06, + "loss": 0.022, + "step": 32247 + }, + { + "epoch": 4.22, + "learning_rate": 9.578251539406256e-06, + "loss": 0.0221, + "step": 32248 + }, + { + "epoch": 4.22, + "learning_rate": 9.577721686793594e-06, + "loss": 0.0251, + "step": 32249 + }, + { + "epoch": 4.22, + "learning_rate": 9.577191835368569e-06, + "loss": 0.0247, + "step": 32250 + }, + { + "epoch": 4.22, + "learning_rate": 9.57666198513267e-06, + "loss": 0.0262, + "step": 32251 + }, + { + "epoch": 4.22, + "learning_rate": 9.576132136087385e-06, + "loss": 0.0216, + "step": 32252 + }, + { + "epoch": 4.22, + "learning_rate": 9.57560228823421e-06, + "loss": 0.0201, + "step": 32253 + }, + { + "epoch": 4.23, + "learning_rate": 9.575072441574626e-06, + "loss": 0.0214, + "step": 32254 + }, + { + "epoch": 4.23, + "learning_rate": 9.574542596110127e-06, + "loss": 0.0253, + "step": 32255 + }, + { + "epoch": 4.23, + "learning_rate": 9.574012751842207e-06, + "loss": 0.0213, + "step": 32256 + }, + { + "epoch": 4.23, + "learning_rate": 9.573482908772355e-06, + "loss": 0.0237, + "step": 32257 + }, + { + "epoch": 4.23, + "learning_rate": 9.572953066902058e-06, + "loss": 0.0218, + "step": 32258 + }, + { + "epoch": 4.23, + "learning_rate": 9.572423226232808e-06, + "loss": 0.0265, + "step": 32259 + }, + { + "epoch": 4.23, + "learning_rate": 9.571893386766096e-06, + "loss": 0.0188, + "step": 32260 + }, + { + "epoch": 4.23, + "learning_rate": 9.57136354850341e-06, + "loss": 0.0215, + "step": 32261 + }, + { + "epoch": 4.23, + "learning_rate": 9.57083371144624e-06, + "loss": 0.024, + "step": 32262 + }, + { + "epoch": 4.23, + "learning_rate": 9.570303875596081e-06, + "loss": 0.0178, + "step": 32263 + }, + { + "epoch": 4.23, + "learning_rate": 9.569774040954417e-06, + "loss": 0.025, + "step": 32264 + }, + { + "epoch": 4.23, + "learning_rate": 9.56924420752274e-06, + "loss": 0.0215, + "step": 32265 + }, + { + "epoch": 4.23, + "learning_rate": 9.568714375302539e-06, + "loss": 0.0205, + "step": 32266 + }, + { + "epoch": 4.23, + "learning_rate": 9.568184544295309e-06, + "loss": 0.0226, + "step": 32267 + }, + { + "epoch": 4.23, + "learning_rate": 9.567654714502534e-06, + "loss": 0.0239, + "step": 32268 + }, + { + "epoch": 4.23, + "learning_rate": 9.567124885925711e-06, + "loss": 0.0179, + "step": 32269 + }, + { + "epoch": 4.23, + "learning_rate": 9.566595058566325e-06, + "loss": 0.0172, + "step": 32270 + }, + { + "epoch": 4.23, + "learning_rate": 9.566065232425865e-06, + "loss": 0.0159, + "step": 32271 + }, + { + "epoch": 4.23, + "learning_rate": 9.565535407505827e-06, + "loss": 0.0235, + "step": 32272 + }, + { + "epoch": 4.23, + "learning_rate": 9.565005583807695e-06, + "loss": 0.0246, + "step": 32273 + }, + { + "epoch": 4.23, + "learning_rate": 9.564475761332964e-06, + "loss": 0.0211, + "step": 32274 + }, + { + "epoch": 4.23, + "learning_rate": 9.56394594008312e-06, + "loss": 0.0248, + "step": 32275 + }, + { + "epoch": 4.23, + "learning_rate": 9.563416120059655e-06, + "loss": 0.0227, + "step": 32276 + }, + { + "epoch": 4.23, + "learning_rate": 9.562886301264055e-06, + "loss": 0.0199, + "step": 32277 + }, + { + "epoch": 4.23, + "learning_rate": 9.562356483697819e-06, + "loss": 0.0187, + "step": 32278 + }, + { + "epoch": 4.23, + "learning_rate": 9.561826667362431e-06, + "loss": 0.024, + "step": 32279 + }, + { + "epoch": 4.23, + "learning_rate": 9.561296852259383e-06, + "loss": 0.0206, + "step": 32280 + }, + { + "epoch": 4.23, + "learning_rate": 9.560767038390162e-06, + "loss": 0.0206, + "step": 32281 + }, + { + "epoch": 4.23, + "learning_rate": 9.560237225756264e-06, + "loss": 0.0233, + "step": 32282 + }, + { + "epoch": 4.23, + "learning_rate": 9.559707414359172e-06, + "loss": 0.021, + "step": 32283 + }, + { + "epoch": 4.23, + "learning_rate": 9.559177604200382e-06, + "loss": 0.0234, + "step": 32284 + }, + { + "epoch": 4.23, + "learning_rate": 9.558647795281378e-06, + "loss": 0.0244, + "step": 32285 + }, + { + "epoch": 4.23, + "learning_rate": 9.558117987603656e-06, + "loss": 0.017, + "step": 32286 + }, + { + "epoch": 4.23, + "learning_rate": 9.557588181168704e-06, + "loss": 0.0245, + "step": 32287 + }, + { + "epoch": 4.23, + "learning_rate": 9.557058375978009e-06, + "loss": 0.0232, + "step": 32288 + }, + { + "epoch": 4.23, + "learning_rate": 9.556528572033063e-06, + "loss": 0.0211, + "step": 32289 + }, + { + "epoch": 4.23, + "learning_rate": 9.555998769335361e-06, + "loss": 0.0197, + "step": 32290 + }, + { + "epoch": 4.23, + "learning_rate": 9.555468967886388e-06, + "loss": 0.0209, + "step": 32291 + }, + { + "epoch": 4.23, + "learning_rate": 9.554939167687632e-06, + "loss": 0.0258, + "step": 32292 + }, + { + "epoch": 4.23, + "learning_rate": 9.55440936874059e-06, + "loss": 0.0303, + "step": 32293 + }, + { + "epoch": 4.23, + "learning_rate": 9.553879571046744e-06, + "loss": 0.0242, + "step": 32294 + }, + { + "epoch": 4.23, + "learning_rate": 9.553349774607593e-06, + "loss": 0.0206, + "step": 32295 + }, + { + "epoch": 4.23, + "learning_rate": 9.552819979424617e-06, + "loss": 0.0231, + "step": 32296 + }, + { + "epoch": 4.23, + "learning_rate": 9.552290185499315e-06, + "loss": 0.0242, + "step": 32297 + }, + { + "epoch": 4.23, + "learning_rate": 9.551760392833169e-06, + "loss": 0.0192, + "step": 32298 + }, + { + "epoch": 4.23, + "learning_rate": 9.551230601427672e-06, + "loss": 0.0189, + "step": 32299 + }, + { + "epoch": 4.23, + "learning_rate": 9.550700811284318e-06, + "loss": 0.0189, + "step": 32300 + }, + { + "epoch": 4.23, + "learning_rate": 9.550171022404595e-06, + "loss": 0.0243, + "step": 32301 + }, + { + "epoch": 4.23, + "learning_rate": 9.54964123478999e-06, + "loss": 0.0196, + "step": 32302 + }, + { + "epoch": 4.23, + "learning_rate": 9.549111448441999e-06, + "loss": 0.0247, + "step": 32303 + }, + { + "epoch": 4.23, + "learning_rate": 9.548581663362103e-06, + "loss": 0.021, + "step": 32304 + }, + { + "epoch": 4.23, + "learning_rate": 9.548051879551801e-06, + "loss": 0.0246, + "step": 32305 + }, + { + "epoch": 4.23, + "learning_rate": 9.547522097012576e-06, + "loss": 0.0268, + "step": 32306 + }, + { + "epoch": 4.23, + "learning_rate": 9.54699231574592e-06, + "loss": 0.0205, + "step": 32307 + }, + { + "epoch": 4.23, + "learning_rate": 9.546462535753328e-06, + "loss": 0.0244, + "step": 32308 + }, + { + "epoch": 4.23, + "learning_rate": 9.545932757036283e-06, + "loss": 0.0223, + "step": 32309 + }, + { + "epoch": 4.23, + "learning_rate": 9.545402979596277e-06, + "loss": 0.0271, + "step": 32310 + }, + { + "epoch": 4.23, + "learning_rate": 9.544873203434805e-06, + "loss": 0.0237, + "step": 32311 + }, + { + "epoch": 4.23, + "learning_rate": 9.544343428553349e-06, + "loss": 0.0162, + "step": 32312 + }, + { + "epoch": 4.23, + "learning_rate": 9.543813654953407e-06, + "loss": 0.0244, + "step": 32313 + }, + { + "epoch": 4.23, + "learning_rate": 9.54328388263646e-06, + "loss": 0.0241, + "step": 32314 + }, + { + "epoch": 4.23, + "learning_rate": 9.542754111604006e-06, + "loss": 0.0249, + "step": 32315 + }, + { + "epoch": 4.23, + "learning_rate": 9.542224341857533e-06, + "loss": 0.0252, + "step": 32316 + }, + { + "epoch": 4.23, + "learning_rate": 9.541694573398526e-06, + "loss": 0.0213, + "step": 32317 + }, + { + "epoch": 4.23, + "learning_rate": 9.541164806228481e-06, + "loss": 0.0206, + "step": 32318 + }, + { + "epoch": 4.23, + "learning_rate": 9.540635040348884e-06, + "loss": 0.024, + "step": 32319 + }, + { + "epoch": 4.23, + "learning_rate": 9.54010527576123e-06, + "loss": 0.0237, + "step": 32320 + }, + { + "epoch": 4.23, + "learning_rate": 9.539575512466996e-06, + "loss": 0.0187, + "step": 32321 + }, + { + "epoch": 4.23, + "learning_rate": 9.539045750467689e-06, + "loss": 0.0161, + "step": 32322 + }, + { + "epoch": 4.23, + "learning_rate": 9.53851598976479e-06, + "loss": 0.0251, + "step": 32323 + }, + { + "epoch": 4.23, + "learning_rate": 9.537986230359791e-06, + "loss": 0.0219, + "step": 32324 + }, + { + "epoch": 4.23, + "learning_rate": 9.537456472254181e-06, + "loss": 0.028, + "step": 32325 + }, + { + "epoch": 4.23, + "learning_rate": 9.53692671544945e-06, + "loss": 0.0224, + "step": 32326 + }, + { + "epoch": 4.23, + "learning_rate": 9.536396959947087e-06, + "loss": 0.0271, + "step": 32327 + }, + { + "epoch": 4.23, + "learning_rate": 9.535867205748583e-06, + "loss": 0.0218, + "step": 32328 + }, + { + "epoch": 4.23, + "learning_rate": 9.53533745285543e-06, + "loss": 0.0243, + "step": 32329 + }, + { + "epoch": 4.24, + "learning_rate": 9.534807701269114e-06, + "loss": 0.0204, + "step": 32330 + }, + { + "epoch": 4.24, + "learning_rate": 9.534277950991127e-06, + "loss": 0.0259, + "step": 32331 + }, + { + "epoch": 4.24, + "learning_rate": 9.533748202022955e-06, + "loss": 0.0245, + "step": 32332 + }, + { + "epoch": 4.24, + "learning_rate": 9.533218454366093e-06, + "loss": 0.0238, + "step": 32333 + }, + { + "epoch": 4.24, + "learning_rate": 9.532688708022032e-06, + "loss": 0.0206, + "step": 32334 + }, + { + "epoch": 4.24, + "learning_rate": 9.532158962992258e-06, + "loss": 0.0203, + "step": 32335 + }, + { + "epoch": 4.24, + "learning_rate": 9.53162921927826e-06, + "loss": 0.0197, + "step": 32336 + }, + { + "epoch": 4.24, + "learning_rate": 9.531099476881533e-06, + "loss": 0.0211, + "step": 32337 + }, + { + "epoch": 4.24, + "learning_rate": 9.53056973580356e-06, + "loss": 0.0282, + "step": 32338 + }, + { + "epoch": 4.24, + "learning_rate": 9.53003999604584e-06, + "loss": 0.0242, + "step": 32339 + }, + { + "epoch": 4.24, + "learning_rate": 9.529510257609853e-06, + "loss": 0.0187, + "step": 32340 + }, + { + "epoch": 4.24, + "learning_rate": 9.528980520497096e-06, + "loss": 0.0209, + "step": 32341 + }, + { + "epoch": 4.24, + "learning_rate": 9.528450784709054e-06, + "loss": 0.0223, + "step": 32342 + }, + { + "epoch": 4.24, + "learning_rate": 9.527921050247217e-06, + "loss": 0.0239, + "step": 32343 + }, + { + "epoch": 4.24, + "learning_rate": 9.527391317113079e-06, + "loss": 0.0249, + "step": 32344 + }, + { + "epoch": 4.24, + "learning_rate": 9.52686158530813e-06, + "loss": 0.0186, + "step": 32345 + }, + { + "epoch": 4.24, + "learning_rate": 9.526331854833855e-06, + "loss": 0.0179, + "step": 32346 + }, + { + "epoch": 4.24, + "learning_rate": 9.52580212569175e-06, + "loss": 0.0217, + "step": 32347 + }, + { + "epoch": 4.24, + "learning_rate": 9.525272397883298e-06, + "loss": 0.0238, + "step": 32348 + }, + { + "epoch": 4.24, + "learning_rate": 9.524742671409995e-06, + "loss": 0.023, + "step": 32349 + }, + { + "epoch": 4.24, + "learning_rate": 9.524212946273325e-06, + "loss": 0.0256, + "step": 32350 + }, + { + "epoch": 4.24, + "learning_rate": 9.523683222474781e-06, + "loss": 0.0252, + "step": 32351 + }, + { + "epoch": 4.24, + "learning_rate": 9.523153500015855e-06, + "loss": 0.0183, + "step": 32352 + }, + { + "epoch": 4.24, + "learning_rate": 9.522623778898031e-06, + "loss": 0.0236, + "step": 32353 + }, + { + "epoch": 4.24, + "learning_rate": 9.522094059122802e-06, + "loss": 0.0235, + "step": 32354 + }, + { + "epoch": 4.24, + "learning_rate": 9.52156434069166e-06, + "loss": 0.0207, + "step": 32355 + }, + { + "epoch": 4.24, + "learning_rate": 9.521034623606092e-06, + "loss": 0.0288, + "step": 32356 + }, + { + "epoch": 4.24, + "learning_rate": 9.52050490786759e-06, + "loss": 0.0228, + "step": 32357 + }, + { + "epoch": 4.24, + "learning_rate": 9.519975193477641e-06, + "loss": 0.0234, + "step": 32358 + }, + { + "epoch": 4.24, + "learning_rate": 9.519445480437737e-06, + "loss": 0.0248, + "step": 32359 + }, + { + "epoch": 4.24, + "learning_rate": 9.518915768749368e-06, + "loss": 0.0181, + "step": 32360 + }, + { + "epoch": 4.24, + "learning_rate": 9.518386058414021e-06, + "loss": 0.0193, + "step": 32361 + }, + { + "epoch": 4.24, + "learning_rate": 9.51785634943319e-06, + "loss": 0.0249, + "step": 32362 + }, + { + "epoch": 4.24, + "learning_rate": 9.51732664180836e-06, + "loss": 0.024, + "step": 32363 + }, + { + "epoch": 4.24, + "learning_rate": 9.516796935541025e-06, + "loss": 0.0247, + "step": 32364 + }, + { + "epoch": 4.24, + "learning_rate": 9.516267230632666e-06, + "loss": 0.0214, + "step": 32365 + }, + { + "epoch": 4.24, + "learning_rate": 9.515737527084787e-06, + "loss": 0.0208, + "step": 32366 + }, + { + "epoch": 4.24, + "learning_rate": 9.515207824898868e-06, + "loss": 0.0186, + "step": 32367 + }, + { + "epoch": 4.24, + "learning_rate": 9.514678124076404e-06, + "loss": 0.0216, + "step": 32368 + }, + { + "epoch": 4.24, + "learning_rate": 9.514148424618879e-06, + "loss": 0.0205, + "step": 32369 + }, + { + "epoch": 4.24, + "learning_rate": 9.513618726527786e-06, + "loss": 0.0238, + "step": 32370 + }, + { + "epoch": 4.24, + "learning_rate": 9.513089029804615e-06, + "loss": 0.0227, + "step": 32371 + }, + { + "epoch": 4.24, + "learning_rate": 9.512559334450855e-06, + "loss": 0.0262, + "step": 32372 + }, + { + "epoch": 4.24, + "learning_rate": 9.512029640467997e-06, + "loss": 0.0199, + "step": 32373 + }, + { + "epoch": 4.24, + "learning_rate": 9.511499947857528e-06, + "loss": 0.0209, + "step": 32374 + }, + { + "epoch": 4.24, + "learning_rate": 9.510970256620941e-06, + "loss": 0.0155, + "step": 32375 + }, + { + "epoch": 4.24, + "learning_rate": 9.51044056675972e-06, + "loss": 0.0252, + "step": 32376 + }, + { + "epoch": 4.24, + "learning_rate": 9.509910878275362e-06, + "loss": 0.0239, + "step": 32377 + }, + { + "epoch": 4.24, + "learning_rate": 9.509381191169352e-06, + "loss": 0.0243, + "step": 32378 + }, + { + "epoch": 4.24, + "learning_rate": 9.508851505443185e-06, + "loss": 0.0191, + "step": 32379 + }, + { + "epoch": 4.24, + "learning_rate": 9.508321821098344e-06, + "loss": 0.0205, + "step": 32380 + }, + { + "epoch": 4.24, + "learning_rate": 9.507792138136324e-06, + "loss": 0.0178, + "step": 32381 + }, + { + "epoch": 4.24, + "learning_rate": 9.50726245655861e-06, + "loss": 0.0213, + "step": 32382 + }, + { + "epoch": 4.24, + "learning_rate": 9.506732776366697e-06, + "loss": 0.0229, + "step": 32383 + }, + { + "epoch": 4.24, + "learning_rate": 9.506203097562068e-06, + "loss": 0.0253, + "step": 32384 + }, + { + "epoch": 4.24, + "learning_rate": 9.50567342014622e-06, + "loss": 0.0208, + "step": 32385 + }, + { + "epoch": 4.24, + "learning_rate": 9.505143744120632e-06, + "loss": 0.0214, + "step": 32386 + }, + { + "epoch": 4.24, + "learning_rate": 9.504614069486809e-06, + "loss": 0.0224, + "step": 32387 + }, + { + "epoch": 4.24, + "learning_rate": 9.504084396246228e-06, + "loss": 0.022, + "step": 32388 + }, + { + "epoch": 4.24, + "learning_rate": 9.503554724400387e-06, + "loss": 0.0198, + "step": 32389 + }, + { + "epoch": 4.24, + "learning_rate": 9.503025053950768e-06, + "loss": 0.0274, + "step": 32390 + }, + { + "epoch": 4.24, + "learning_rate": 9.502495384898868e-06, + "loss": 0.0208, + "step": 32391 + }, + { + "epoch": 4.24, + "learning_rate": 9.50196571724617e-06, + "loss": 0.0228, + "step": 32392 + }, + { + "epoch": 4.24, + "learning_rate": 9.501436050994169e-06, + "loss": 0.0214, + "step": 32393 + }, + { + "epoch": 4.24, + "learning_rate": 9.500906386144354e-06, + "loss": 0.021, + "step": 32394 + }, + { + "epoch": 4.24, + "learning_rate": 9.50037672269821e-06, + "loss": 0.0213, + "step": 32395 + }, + { + "epoch": 4.24, + "learning_rate": 9.499847060657234e-06, + "loss": 0.0178, + "step": 32396 + }, + { + "epoch": 4.24, + "learning_rate": 9.499317400022906e-06, + "loss": 0.0184, + "step": 32397 + }, + { + "epoch": 4.24, + "learning_rate": 9.498787740796722e-06, + "loss": 0.0225, + "step": 32398 + }, + { + "epoch": 4.24, + "learning_rate": 9.498258082980176e-06, + "loss": 0.0208, + "step": 32399 + }, + { + "epoch": 4.24, + "learning_rate": 9.497728426574748e-06, + "loss": 0.0178, + "step": 32400 + }, + { + "epoch": 4.24, + "learning_rate": 9.497198771581931e-06, + "loss": 0.0225, + "step": 32401 + }, + { + "epoch": 4.24, + "learning_rate": 9.49666911800322e-06, + "loss": 0.0211, + "step": 32402 + }, + { + "epoch": 4.24, + "learning_rate": 9.496139465840098e-06, + "loss": 0.0194, + "step": 32403 + }, + { + "epoch": 4.24, + "learning_rate": 9.495609815094058e-06, + "loss": 0.0232, + "step": 32404 + }, + { + "epoch": 4.24, + "learning_rate": 9.495080165766588e-06, + "loss": 0.0271, + "step": 32405 + }, + { + "epoch": 4.24, + "learning_rate": 9.494550517859178e-06, + "loss": 0.0202, + "step": 32406 + }, + { + "epoch": 4.25, + "learning_rate": 9.494020871373317e-06, + "loss": 0.0206, + "step": 32407 + }, + { + "epoch": 4.25, + "learning_rate": 9.493491226310493e-06, + "loss": 0.0243, + "step": 32408 + }, + { + "epoch": 4.25, + "learning_rate": 9.4929615826722e-06, + "loss": 0.0251, + "step": 32409 + }, + { + "epoch": 4.25, + "learning_rate": 9.492431940459928e-06, + "loss": 0.017, + "step": 32410 + }, + { + "epoch": 4.25, + "learning_rate": 9.491902299675161e-06, + "loss": 0.0185, + "step": 32411 + }, + { + "epoch": 4.25, + "learning_rate": 9.491372660319396e-06, + "loss": 0.0245, + "step": 32412 + }, + { + "epoch": 4.25, + "learning_rate": 9.490843022394115e-06, + "loss": 0.0196, + "step": 32413 + }, + { + "epoch": 4.25, + "learning_rate": 9.490313385900813e-06, + "loss": 0.0235, + "step": 32414 + }, + { + "epoch": 4.25, + "learning_rate": 9.489783750840974e-06, + "loss": 0.0195, + "step": 32415 + }, + { + "epoch": 4.25, + "learning_rate": 9.489254117216092e-06, + "loss": 0.0225, + "step": 32416 + }, + { + "epoch": 4.25, + "learning_rate": 9.488724485027659e-06, + "loss": 0.0187, + "step": 32417 + }, + { + "epoch": 4.25, + "learning_rate": 9.488194854277158e-06, + "loss": 0.024, + "step": 32418 + }, + { + "epoch": 4.25, + "learning_rate": 9.487665224966079e-06, + "loss": 0.02, + "step": 32419 + }, + { + "epoch": 4.25, + "learning_rate": 9.487135597095918e-06, + "loss": 0.0228, + "step": 32420 + }, + { + "epoch": 4.25, + "learning_rate": 9.48660597066816e-06, + "loss": 0.0204, + "step": 32421 + }, + { + "epoch": 4.25, + "learning_rate": 9.486076345684295e-06, + "loss": 0.0222, + "step": 32422 + }, + { + "epoch": 4.25, + "learning_rate": 9.485546722145817e-06, + "loss": 0.0194, + "step": 32423 + }, + { + "epoch": 4.25, + "learning_rate": 9.485017100054206e-06, + "loss": 0.0193, + "step": 32424 + }, + { + "epoch": 4.25, + "learning_rate": 9.484487479410961e-06, + "loss": 0.0205, + "step": 32425 + }, + { + "epoch": 4.25, + "learning_rate": 9.483957860217566e-06, + "loss": 0.0253, + "step": 32426 + }, + { + "epoch": 4.25, + "learning_rate": 9.483428242475512e-06, + "loss": 0.0238, + "step": 32427 + }, + { + "epoch": 4.25, + "learning_rate": 9.482898626186287e-06, + "loss": 0.0207, + "step": 32428 + }, + { + "epoch": 4.25, + "learning_rate": 9.482369011351386e-06, + "loss": 0.0239, + "step": 32429 + }, + { + "epoch": 4.25, + "learning_rate": 9.481839397972287e-06, + "loss": 0.0262, + "step": 32430 + }, + { + "epoch": 4.25, + "learning_rate": 9.481309786050493e-06, + "loss": 0.0211, + "step": 32431 + }, + { + "epoch": 4.25, + "learning_rate": 9.480780175587486e-06, + "loss": 0.0231, + "step": 32432 + }, + { + "epoch": 4.25, + "learning_rate": 9.48025056658476e-06, + "loss": 0.0226, + "step": 32433 + }, + { + "epoch": 4.25, + "learning_rate": 9.479720959043798e-06, + "loss": 0.0252, + "step": 32434 + }, + { + "epoch": 4.25, + "learning_rate": 9.479191352966096e-06, + "loss": 0.0199, + "step": 32435 + }, + { + "epoch": 4.25, + "learning_rate": 9.47866174835314e-06, + "loss": 0.0182, + "step": 32436 + }, + { + "epoch": 4.25, + "learning_rate": 9.478132145206417e-06, + "loss": 0.0188, + "step": 32437 + }, + { + "epoch": 4.25, + "learning_rate": 9.477602543527423e-06, + "loss": 0.0212, + "step": 32438 + }, + { + "epoch": 4.25, + "learning_rate": 9.477072943317642e-06, + "loss": 0.0176, + "step": 32439 + }, + { + "epoch": 4.25, + "learning_rate": 9.476543344578567e-06, + "loss": 0.0276, + "step": 32440 + }, + { + "epoch": 4.25, + "learning_rate": 9.476013747311683e-06, + "loss": 0.0247, + "step": 32441 + }, + { + "epoch": 4.25, + "learning_rate": 9.475484151518484e-06, + "loss": 0.0259, + "step": 32442 + }, + { + "epoch": 4.25, + "learning_rate": 9.474954557200457e-06, + "loss": 0.025, + "step": 32443 + }, + { + "epoch": 4.25, + "learning_rate": 9.474424964359095e-06, + "loss": 0.0233, + "step": 32444 + }, + { + "epoch": 4.25, + "learning_rate": 9.473895372995884e-06, + "loss": 0.0205, + "step": 32445 + }, + { + "epoch": 4.25, + "learning_rate": 9.473365783112315e-06, + "loss": 0.0236, + "step": 32446 + }, + { + "epoch": 4.25, + "learning_rate": 9.472836194709875e-06, + "loss": 0.0238, + "step": 32447 + }, + { + "epoch": 4.25, + "learning_rate": 9.472306607790057e-06, + "loss": 0.0236, + "step": 32448 + }, + { + "epoch": 4.25, + "learning_rate": 9.471777022354346e-06, + "loss": 0.0232, + "step": 32449 + }, + { + "epoch": 4.25, + "learning_rate": 9.471247438404237e-06, + "loss": 0.0239, + "step": 32450 + }, + { + "epoch": 4.25, + "learning_rate": 9.470717855941214e-06, + "loss": 0.0202, + "step": 32451 + }, + { + "epoch": 4.25, + "learning_rate": 9.470188274966768e-06, + "loss": 0.0194, + "step": 32452 + }, + { + "epoch": 4.25, + "learning_rate": 9.469658695482389e-06, + "loss": 0.0152, + "step": 32453 + }, + { + "epoch": 4.25, + "learning_rate": 9.46912911748957e-06, + "loss": 0.0208, + "step": 32454 + }, + { + "epoch": 4.25, + "learning_rate": 9.468599540989794e-06, + "loss": 0.023, + "step": 32455 + }, + { + "epoch": 4.25, + "learning_rate": 9.468069965984557e-06, + "loss": 0.0209, + "step": 32456 + }, + { + "epoch": 4.25, + "learning_rate": 9.467540392475343e-06, + "loss": 0.0211, + "step": 32457 + }, + { + "epoch": 4.25, + "learning_rate": 9.467010820463643e-06, + "loss": 0.0233, + "step": 32458 + }, + { + "epoch": 4.25, + "learning_rate": 9.466481249950947e-06, + "loss": 0.0226, + "step": 32459 + }, + { + "epoch": 4.25, + "learning_rate": 9.465951680938743e-06, + "loss": 0.0237, + "step": 32460 + }, + { + "epoch": 4.25, + "learning_rate": 9.465422113428524e-06, + "loss": 0.0182, + "step": 32461 + }, + { + "epoch": 4.25, + "learning_rate": 9.464892547421774e-06, + "loss": 0.0158, + "step": 32462 + }, + { + "epoch": 4.25, + "learning_rate": 9.464362982919985e-06, + "loss": 0.0191, + "step": 32463 + }, + { + "epoch": 4.25, + "learning_rate": 9.46383341992465e-06, + "loss": 0.0203, + "step": 32464 + }, + { + "epoch": 4.25, + "learning_rate": 9.463303858437251e-06, + "loss": 0.0249, + "step": 32465 + }, + { + "epoch": 4.25, + "learning_rate": 9.462774298459283e-06, + "loss": 0.0221, + "step": 32466 + }, + { + "epoch": 4.25, + "learning_rate": 9.462244739992236e-06, + "loss": 0.0167, + "step": 32467 + }, + { + "epoch": 4.25, + "learning_rate": 9.461715183037595e-06, + "loss": 0.0206, + "step": 32468 + }, + { + "epoch": 4.25, + "learning_rate": 9.461185627596852e-06, + "loss": 0.0229, + "step": 32469 + }, + { + "epoch": 4.25, + "learning_rate": 9.460656073671494e-06, + "loss": 0.0226, + "step": 32470 + }, + { + "epoch": 4.25, + "learning_rate": 9.460126521263016e-06, + "loss": 0.0203, + "step": 32471 + }, + { + "epoch": 4.25, + "learning_rate": 9.4595969703729e-06, + "loss": 0.0231, + "step": 32472 + }, + { + "epoch": 4.25, + "learning_rate": 9.459067421002638e-06, + "loss": 0.0205, + "step": 32473 + }, + { + "epoch": 4.25, + "learning_rate": 9.458537873153718e-06, + "loss": 0.0207, + "step": 32474 + }, + { + "epoch": 4.25, + "learning_rate": 9.458008326827637e-06, + "loss": 0.0219, + "step": 32475 + }, + { + "epoch": 4.25, + "learning_rate": 9.457478782025874e-06, + "loss": 0.0203, + "step": 32476 + }, + { + "epoch": 4.25, + "learning_rate": 9.45694923874993e-06, + "loss": 0.0238, + "step": 32477 + }, + { + "epoch": 4.25, + "learning_rate": 9.456419697001281e-06, + "loss": 0.0235, + "step": 32478 + }, + { + "epoch": 4.25, + "learning_rate": 9.455890156781426e-06, + "loss": 0.0206, + "step": 32479 + }, + { + "epoch": 4.25, + "learning_rate": 9.455360618091849e-06, + "loss": 0.0209, + "step": 32480 + }, + { + "epoch": 4.25, + "learning_rate": 9.45483108093404e-06, + "loss": 0.0155, + "step": 32481 + }, + { + "epoch": 4.25, + "learning_rate": 9.454301545309493e-06, + "loss": 0.0236, + "step": 32482 + }, + { + "epoch": 4.26, + "learning_rate": 9.453772011219693e-06, + "loss": 0.0251, + "step": 32483 + }, + { + "epoch": 4.26, + "learning_rate": 9.453242478666131e-06, + "loss": 0.0243, + "step": 32484 + }, + { + "epoch": 4.26, + "learning_rate": 9.45271294765029e-06, + "loss": 0.0179, + "step": 32485 + }, + { + "epoch": 4.26, + "learning_rate": 9.452183418173669e-06, + "loss": 0.0224, + "step": 32486 + }, + { + "epoch": 4.26, + "learning_rate": 9.451653890237751e-06, + "loss": 0.0228, + "step": 32487 + }, + { + "epoch": 4.26, + "learning_rate": 9.45112436384403e-06, + "loss": 0.0216, + "step": 32488 + }, + { + "epoch": 4.26, + "learning_rate": 9.450594838993989e-06, + "loss": 0.0216, + "step": 32489 + }, + { + "epoch": 4.26, + "learning_rate": 9.450065315689126e-06, + "loss": 0.0201, + "step": 32490 + }, + { + "epoch": 4.26, + "learning_rate": 9.449535793930921e-06, + "loss": 0.02, + "step": 32491 + }, + { + "epoch": 4.26, + "learning_rate": 9.449006273720869e-06, + "loss": 0.0228, + "step": 32492 + }, + { + "epoch": 4.26, + "learning_rate": 9.448476755060457e-06, + "loss": 0.0197, + "step": 32493 + }, + { + "epoch": 4.26, + "learning_rate": 9.447947237951172e-06, + "loss": 0.0215, + "step": 32494 + }, + { + "epoch": 4.26, + "learning_rate": 9.44741772239451e-06, + "loss": 0.0191, + "step": 32495 + }, + { + "epoch": 4.26, + "learning_rate": 9.446888208391952e-06, + "loss": 0.0247, + "step": 32496 + }, + { + "epoch": 4.26, + "learning_rate": 9.446358695944994e-06, + "loss": 0.0248, + "step": 32497 + }, + { + "epoch": 4.26, + "learning_rate": 9.445829185055125e-06, + "loss": 0.0244, + "step": 32498 + }, + { + "epoch": 4.26, + "learning_rate": 9.445299675723828e-06, + "loss": 0.0226, + "step": 32499 + }, + { + "epoch": 4.26, + "learning_rate": 9.4447701679526e-06, + "loss": 0.0192, + "step": 32500 + }, + { + "epoch": 4.26, + "learning_rate": 9.444240661742923e-06, + "loss": 0.0213, + "step": 32501 + }, + { + "epoch": 4.26, + "learning_rate": 9.44371115709629e-06, + "loss": 0.0177, + "step": 32502 + }, + { + "epoch": 4.26, + "learning_rate": 9.443181654014192e-06, + "loss": 0.0202, + "step": 32503 + }, + { + "epoch": 4.26, + "learning_rate": 9.442652152498113e-06, + "loss": 0.0204, + "step": 32504 + }, + { + "epoch": 4.26, + "learning_rate": 9.442122652549548e-06, + "loss": 0.0194, + "step": 32505 + }, + { + "epoch": 4.26, + "learning_rate": 9.441593154169982e-06, + "loss": 0.0245, + "step": 32506 + }, + { + "epoch": 4.26, + "learning_rate": 9.441063657360902e-06, + "loss": 0.0206, + "step": 32507 + }, + { + "epoch": 4.26, + "learning_rate": 9.440534162123804e-06, + "loss": 0.026, + "step": 32508 + }, + { + "epoch": 4.26, + "learning_rate": 9.440004668460176e-06, + "loss": 0.0234, + "step": 32509 + }, + { + "epoch": 4.26, + "learning_rate": 9.439475176371502e-06, + "loss": 0.0202, + "step": 32510 + }, + { + "epoch": 4.26, + "learning_rate": 9.438945685859276e-06, + "loss": 0.0205, + "step": 32511 + }, + { + "epoch": 4.26, + "learning_rate": 9.438416196924984e-06, + "loss": 0.0207, + "step": 32512 + }, + { + "epoch": 4.26, + "learning_rate": 9.437886709570118e-06, + "loss": 0.0241, + "step": 32513 + }, + { + "epoch": 4.26, + "learning_rate": 9.437357223796165e-06, + "loss": 0.0234, + "step": 32514 + }, + { + "epoch": 4.26, + "learning_rate": 9.436827739604617e-06, + "loss": 0.0226, + "step": 32515 + }, + { + "epoch": 4.26, + "learning_rate": 9.436298256996958e-06, + "loss": 0.0174, + "step": 32516 + }, + { + "epoch": 4.26, + "learning_rate": 9.435768775974677e-06, + "loss": 0.0257, + "step": 32517 + }, + { + "epoch": 4.26, + "learning_rate": 9.43523929653927e-06, + "loss": 0.0225, + "step": 32518 + }, + { + "epoch": 4.26, + "learning_rate": 9.434709818692224e-06, + "loss": 0.0208, + "step": 32519 + }, + { + "epoch": 4.26, + "learning_rate": 9.434180342435025e-06, + "loss": 0.0222, + "step": 32520 + }, + { + "epoch": 4.26, + "learning_rate": 9.433650867769163e-06, + "loss": 0.0273, + "step": 32521 + }, + { + "epoch": 4.26, + "learning_rate": 9.433121394696129e-06, + "loss": 0.0226, + "step": 32522 + }, + { + "epoch": 4.26, + "learning_rate": 9.43259192321741e-06, + "loss": 0.0202, + "step": 32523 + }, + { + "epoch": 4.26, + "learning_rate": 9.432062453334497e-06, + "loss": 0.0176, + "step": 32524 + }, + { + "epoch": 4.26, + "learning_rate": 9.431532985048877e-06, + "loss": 0.0212, + "step": 32525 + }, + { + "epoch": 4.26, + "learning_rate": 9.431003518362042e-06, + "loss": 0.0219, + "step": 32526 + }, + { + "epoch": 4.26, + "learning_rate": 9.430474053275478e-06, + "loss": 0.0231, + "step": 32527 + }, + { + "epoch": 4.26, + "learning_rate": 9.429944589790672e-06, + "loss": 0.0264, + "step": 32528 + }, + { + "epoch": 4.26, + "learning_rate": 9.429415127909122e-06, + "loss": 0.0169, + "step": 32529 + }, + { + "epoch": 4.26, + "learning_rate": 9.428885667632309e-06, + "loss": 0.0213, + "step": 32530 + }, + { + "epoch": 4.26, + "learning_rate": 9.428356208961724e-06, + "loss": 0.0207, + "step": 32531 + }, + { + "epoch": 4.26, + "learning_rate": 9.42782675189886e-06, + "loss": 0.0212, + "step": 32532 + }, + { + "epoch": 4.26, + "learning_rate": 9.4272972964452e-06, + "loss": 0.0219, + "step": 32533 + }, + { + "epoch": 4.26, + "learning_rate": 9.426767842602237e-06, + "loss": 0.0246, + "step": 32534 + }, + { + "epoch": 4.26, + "learning_rate": 9.426238390371456e-06, + "loss": 0.0196, + "step": 32535 + }, + { + "epoch": 4.26, + "learning_rate": 9.425708939754355e-06, + "loss": 0.0194, + "step": 32536 + }, + { + "epoch": 4.26, + "learning_rate": 9.425179490752413e-06, + "loss": 0.0219, + "step": 32537 + }, + { + "epoch": 4.26, + "learning_rate": 9.424650043367121e-06, + "loss": 0.0176, + "step": 32538 + }, + { + "epoch": 4.26, + "learning_rate": 9.42412059759997e-06, + "loss": 0.0296, + "step": 32539 + }, + { + "epoch": 4.26, + "learning_rate": 9.423591153452454e-06, + "loss": 0.0189, + "step": 32540 + }, + { + "epoch": 4.26, + "learning_rate": 9.423061710926053e-06, + "loss": 0.0238, + "step": 32541 + }, + { + "epoch": 4.26, + "learning_rate": 9.422532270022265e-06, + "loss": 0.0193, + "step": 32542 + }, + { + "epoch": 4.26, + "learning_rate": 9.422002830742568e-06, + "loss": 0.0257, + "step": 32543 + }, + { + "epoch": 4.26, + "learning_rate": 9.421473393088462e-06, + "loss": 0.0239, + "step": 32544 + }, + { + "epoch": 4.26, + "learning_rate": 9.420943957061432e-06, + "loss": 0.0211, + "step": 32545 + }, + { + "epoch": 4.26, + "learning_rate": 9.420414522662963e-06, + "loss": 0.0221, + "step": 32546 + }, + { + "epoch": 4.26, + "learning_rate": 9.41988508989455e-06, + "loss": 0.0238, + "step": 32547 + }, + { + "epoch": 4.26, + "learning_rate": 9.419355658757676e-06, + "loss": 0.021, + "step": 32548 + }, + { + "epoch": 4.26, + "learning_rate": 9.418826229253838e-06, + "loss": 0.0192, + "step": 32549 + }, + { + "epoch": 4.26, + "learning_rate": 9.418296801384514e-06, + "loss": 0.0206, + "step": 32550 + }, + { + "epoch": 4.26, + "learning_rate": 9.417767375151202e-06, + "loss": 0.0179, + "step": 32551 + }, + { + "epoch": 4.26, + "learning_rate": 9.417237950555386e-06, + "loss": 0.022, + "step": 32552 + }, + { + "epoch": 4.26, + "learning_rate": 9.416708527598562e-06, + "loss": 0.0204, + "step": 32553 + }, + { + "epoch": 4.26, + "learning_rate": 9.416179106282214e-06, + "loss": 0.0214, + "step": 32554 + }, + { + "epoch": 4.26, + "learning_rate": 9.41564968660783e-06, + "loss": 0.0186, + "step": 32555 + }, + { + "epoch": 4.26, + "learning_rate": 9.415120268576899e-06, + "loss": 0.0235, + "step": 32556 + }, + { + "epoch": 4.26, + "learning_rate": 9.414590852190913e-06, + "loss": 0.0187, + "step": 32557 + }, + { + "epoch": 4.26, + "learning_rate": 9.414061437451357e-06, + "loss": 0.022, + "step": 32558 + }, + { + "epoch": 4.26, + "learning_rate": 9.413532024359722e-06, + "loss": 0.0264, + "step": 32559 + }, + { + "epoch": 4.27, + "learning_rate": 9.413002612917498e-06, + "loss": 0.0226, + "step": 32560 + }, + { + "epoch": 4.27, + "learning_rate": 9.41247320312617e-06, + "loss": 0.0233, + "step": 32561 + }, + { + "epoch": 4.27, + "learning_rate": 9.411943794987231e-06, + "loss": 0.0191, + "step": 32562 + }, + { + "epoch": 4.27, + "learning_rate": 9.411414388502172e-06, + "loss": 0.0202, + "step": 32563 + }, + { + "epoch": 4.27, + "learning_rate": 9.410884983672477e-06, + "loss": 0.0204, + "step": 32564 + }, + { + "epoch": 4.27, + "learning_rate": 9.410355580499637e-06, + "loss": 0.0185, + "step": 32565 + }, + { + "epoch": 4.27, + "learning_rate": 9.40982617898514e-06, + "loss": 0.0248, + "step": 32566 + }, + { + "epoch": 4.27, + "learning_rate": 9.409296779130475e-06, + "loss": 0.0211, + "step": 32567 + }, + { + "epoch": 4.27, + "learning_rate": 9.408767380937132e-06, + "loss": 0.0228, + "step": 32568 + }, + { + "epoch": 4.27, + "learning_rate": 9.4082379844066e-06, + "loss": 0.0209, + "step": 32569 + }, + { + "epoch": 4.27, + "learning_rate": 9.407708589540366e-06, + "loss": 0.0203, + "step": 32570 + }, + { + "epoch": 4.27, + "learning_rate": 9.40717919633992e-06, + "loss": 0.0255, + "step": 32571 + }, + { + "epoch": 4.27, + "learning_rate": 9.406649804806749e-06, + "loss": 0.0204, + "step": 32572 + }, + { + "epoch": 4.27, + "learning_rate": 9.406120414942345e-06, + "loss": 0.0265, + "step": 32573 + }, + { + "epoch": 4.27, + "learning_rate": 9.405591026748198e-06, + "loss": 0.0219, + "step": 32574 + }, + { + "epoch": 4.27, + "learning_rate": 9.405061640225792e-06, + "loss": 0.0205, + "step": 32575 + }, + { + "epoch": 4.27, + "learning_rate": 9.404532255376622e-06, + "loss": 0.0166, + "step": 32576 + }, + { + "epoch": 4.27, + "learning_rate": 9.404002872202171e-06, + "loss": 0.0231, + "step": 32577 + }, + { + "epoch": 4.27, + "learning_rate": 9.40347349070393e-06, + "loss": 0.0236, + "step": 32578 + }, + { + "epoch": 4.27, + "learning_rate": 9.402944110883389e-06, + "loss": 0.0247, + "step": 32579 + }, + { + "epoch": 4.27, + "learning_rate": 9.402414732742036e-06, + "loss": 0.0204, + "step": 32580 + }, + { + "epoch": 4.27, + "learning_rate": 9.401885356281358e-06, + "loss": 0.0202, + "step": 32581 + }, + { + "epoch": 4.27, + "learning_rate": 9.401355981502846e-06, + "loss": 0.0181, + "step": 32582 + }, + { + "epoch": 4.27, + "learning_rate": 9.400826608407985e-06, + "loss": 0.0192, + "step": 32583 + }, + { + "epoch": 4.27, + "learning_rate": 9.400297236998274e-06, + "loss": 0.0183, + "step": 32584 + }, + { + "epoch": 4.27, + "learning_rate": 9.39976786727519e-06, + "loss": 0.0209, + "step": 32585 + }, + { + "epoch": 4.27, + "learning_rate": 9.399238499240232e-06, + "loss": 0.0228, + "step": 32586 + }, + { + "epoch": 4.27, + "learning_rate": 9.39870913289488e-06, + "loss": 0.0221, + "step": 32587 + }, + { + "epoch": 4.27, + "learning_rate": 9.398179768240628e-06, + "loss": 0.0199, + "step": 32588 + }, + { + "epoch": 4.27, + "learning_rate": 9.397650405278963e-06, + "loss": 0.0199, + "step": 32589 + }, + { + "epoch": 4.27, + "learning_rate": 9.397121044011374e-06, + "loss": 0.025, + "step": 32590 + }, + { + "epoch": 4.27, + "learning_rate": 9.396591684439352e-06, + "loss": 0.0202, + "step": 32591 + }, + { + "epoch": 4.27, + "learning_rate": 9.396062326564379e-06, + "loss": 0.0253, + "step": 32592 + }, + { + "epoch": 4.27, + "learning_rate": 9.395532970387952e-06, + "loss": 0.0197, + "step": 32593 + }, + { + "epoch": 4.27, + "learning_rate": 9.395003615911551e-06, + "loss": 0.0209, + "step": 32594 + }, + { + "epoch": 4.27, + "learning_rate": 9.394474263136677e-06, + "loss": 0.0212, + "step": 32595 + }, + { + "epoch": 4.27, + "learning_rate": 9.393944912064808e-06, + "loss": 0.0223, + "step": 32596 + }, + { + "epoch": 4.27, + "learning_rate": 9.39341556269744e-06, + "loss": 0.0243, + "step": 32597 + }, + { + "epoch": 4.27, + "learning_rate": 9.392886215036057e-06, + "loss": 0.0262, + "step": 32598 + }, + { + "epoch": 4.27, + "learning_rate": 9.39235686908215e-06, + "loss": 0.0246, + "step": 32599 + }, + { + "epoch": 4.27, + "learning_rate": 9.391827524837206e-06, + "loss": 0.0222, + "step": 32600 + }, + { + "epoch": 4.27, + "learning_rate": 9.391298182302716e-06, + "loss": 0.0243, + "step": 32601 + }, + { + "epoch": 4.27, + "learning_rate": 9.390768841480164e-06, + "loss": 0.0204, + "step": 32602 + }, + { + "epoch": 4.27, + "learning_rate": 9.390239502371043e-06, + "loss": 0.0179, + "step": 32603 + }, + { + "epoch": 4.27, + "learning_rate": 9.389710164976844e-06, + "loss": 0.0213, + "step": 32604 + }, + { + "epoch": 4.27, + "learning_rate": 9.389180829299047e-06, + "loss": 0.0211, + "step": 32605 + }, + { + "epoch": 4.27, + "learning_rate": 9.38865149533915e-06, + "loss": 0.014, + "step": 32606 + }, + { + "epoch": 4.27, + "learning_rate": 9.38812216309864e-06, + "loss": 0.0238, + "step": 32607 + }, + { + "epoch": 4.27, + "learning_rate": 9.387592832579e-06, + "loss": 0.0188, + "step": 32608 + }, + { + "epoch": 4.27, + "learning_rate": 9.387063503781725e-06, + "loss": 0.0241, + "step": 32609 + }, + { + "epoch": 4.27, + "learning_rate": 9.386534176708302e-06, + "loss": 0.0204, + "step": 32610 + }, + { + "epoch": 4.27, + "learning_rate": 9.386004851360217e-06, + "loss": 0.0202, + "step": 32611 + }, + { + "epoch": 4.27, + "learning_rate": 9.385475527738963e-06, + "loss": 0.0186, + "step": 32612 + }, + { + "epoch": 4.27, + "learning_rate": 9.384946205846024e-06, + "loss": 0.0282, + "step": 32613 + }, + { + "epoch": 4.27, + "learning_rate": 9.384416885682893e-06, + "loss": 0.0224, + "step": 32614 + }, + { + "epoch": 4.27, + "learning_rate": 9.383887567251056e-06, + "loss": 0.0233, + "step": 32615 + }, + { + "epoch": 4.27, + "learning_rate": 9.383358250551998e-06, + "loss": 0.0262, + "step": 32616 + }, + { + "epoch": 4.27, + "learning_rate": 9.382828935587216e-06, + "loss": 0.0233, + "step": 32617 + }, + { + "epoch": 4.27, + "learning_rate": 9.382299622358197e-06, + "loss": 0.0235, + "step": 32618 + }, + { + "epoch": 4.27, + "learning_rate": 9.381770310866423e-06, + "loss": 0.0162, + "step": 32619 + }, + { + "epoch": 4.27, + "learning_rate": 9.38124100111339e-06, + "loss": 0.0213, + "step": 32620 + }, + { + "epoch": 4.27, + "learning_rate": 9.380711693100584e-06, + "loss": 0.0158, + "step": 32621 + }, + { + "epoch": 4.27, + "learning_rate": 9.380182386829494e-06, + "loss": 0.0186, + "step": 32622 + }, + { + "epoch": 4.27, + "learning_rate": 9.379653082301605e-06, + "loss": 0.0227, + "step": 32623 + }, + { + "epoch": 4.27, + "learning_rate": 9.37912377951841e-06, + "loss": 0.0239, + "step": 32624 + }, + { + "epoch": 4.27, + "learning_rate": 9.378594478481398e-06, + "loss": 0.0216, + "step": 32625 + }, + { + "epoch": 4.27, + "learning_rate": 9.378065179192055e-06, + "loss": 0.0205, + "step": 32626 + }, + { + "epoch": 4.27, + "learning_rate": 9.377535881651865e-06, + "loss": 0.0204, + "step": 32627 + }, + { + "epoch": 4.27, + "learning_rate": 9.37700658586233e-06, + "loss": 0.02, + "step": 32628 + }, + { + "epoch": 4.27, + "learning_rate": 9.376477291824927e-06, + "loss": 0.0194, + "step": 32629 + }, + { + "epoch": 4.27, + "learning_rate": 9.375947999541151e-06, + "loss": 0.0204, + "step": 32630 + }, + { + "epoch": 4.27, + "learning_rate": 9.375418709012485e-06, + "loss": 0.0202, + "step": 32631 + }, + { + "epoch": 4.27, + "learning_rate": 9.37488942024042e-06, + "loss": 0.0237, + "step": 32632 + }, + { + "epoch": 4.27, + "learning_rate": 9.37436013322645e-06, + "loss": 0.0177, + "step": 32633 + }, + { + "epoch": 4.27, + "learning_rate": 9.373830847972055e-06, + "loss": 0.0204, + "step": 32634 + }, + { + "epoch": 4.27, + "learning_rate": 9.37330156447873e-06, + "loss": 0.0216, + "step": 32635 + }, + { + "epoch": 4.28, + "learning_rate": 9.372772282747958e-06, + "loss": 0.0202, + "step": 32636 + }, + { + "epoch": 4.28, + "learning_rate": 9.372243002781233e-06, + "loss": 0.0197, + "step": 32637 + }, + { + "epoch": 4.28, + "learning_rate": 9.371713724580036e-06, + "loss": 0.0233, + "step": 32638 + }, + { + "epoch": 4.28, + "learning_rate": 9.371184448145866e-06, + "loss": 0.0228, + "step": 32639 + }, + { + "epoch": 4.28, + "learning_rate": 9.370655173480205e-06, + "loss": 0.0219, + "step": 32640 + }, + { + "epoch": 4.28, + "learning_rate": 9.370125900584544e-06, + "loss": 0.0183, + "step": 32641 + }, + { + "epoch": 4.28, + "learning_rate": 9.369596629460368e-06, + "loss": 0.0195, + "step": 32642 + }, + { + "epoch": 4.28, + "learning_rate": 9.36906736010917e-06, + "loss": 0.0189, + "step": 32643 + }, + { + "epoch": 4.28, + "learning_rate": 9.368538092532435e-06, + "loss": 0.0199, + "step": 32644 + }, + { + "epoch": 4.28, + "learning_rate": 9.368008826731656e-06, + "loss": 0.0301, + "step": 32645 + }, + { + "epoch": 4.28, + "learning_rate": 9.367479562708314e-06, + "loss": 0.0165, + "step": 32646 + }, + { + "epoch": 4.28, + "learning_rate": 9.366950300463905e-06, + "loss": 0.0217, + "step": 32647 + }, + { + "epoch": 4.28, + "learning_rate": 9.36642103999991e-06, + "loss": 0.02, + "step": 32648 + }, + { + "epoch": 4.28, + "learning_rate": 9.365891781317827e-06, + "loss": 0.0187, + "step": 32649 + }, + { + "epoch": 4.28, + "learning_rate": 9.365362524419137e-06, + "loss": 0.021, + "step": 32650 + }, + { + "epoch": 4.28, + "learning_rate": 9.364833269305334e-06, + "loss": 0.0219, + "step": 32651 + }, + { + "epoch": 4.28, + "learning_rate": 9.364304015977901e-06, + "loss": 0.0245, + "step": 32652 + }, + { + "epoch": 4.28, + "learning_rate": 9.36377476443833e-06, + "loss": 0.0225, + "step": 32653 + }, + { + "epoch": 4.28, + "learning_rate": 9.36324551468811e-06, + "loss": 0.024, + "step": 32654 + }, + { + "epoch": 4.28, + "learning_rate": 9.362716266728727e-06, + "loss": 0.0236, + "step": 32655 + }, + { + "epoch": 4.28, + "learning_rate": 9.362187020561672e-06, + "loss": 0.027, + "step": 32656 + }, + { + "epoch": 4.28, + "learning_rate": 9.361657776188429e-06, + "loss": 0.021, + "step": 32657 + }, + { + "epoch": 4.28, + "learning_rate": 9.361128533610493e-06, + "loss": 0.0269, + "step": 32658 + }, + { + "epoch": 4.28, + "learning_rate": 9.360599292829343e-06, + "loss": 0.0193, + "step": 32659 + }, + { + "epoch": 4.28, + "learning_rate": 9.360070053846477e-06, + "loss": 0.0187, + "step": 32660 + }, + { + "epoch": 4.28, + "learning_rate": 9.359540816663379e-06, + "loss": 0.0237, + "step": 32661 + }, + { + "epoch": 4.28, + "learning_rate": 9.359011581281542e-06, + "loss": 0.0255, + "step": 32662 + }, + { + "epoch": 4.28, + "learning_rate": 9.358482347702449e-06, + "loss": 0.0206, + "step": 32663 + }, + { + "epoch": 4.28, + "learning_rate": 9.35795311592759e-06, + "loss": 0.0209, + "step": 32664 + }, + { + "epoch": 4.28, + "learning_rate": 9.357423885958452e-06, + "loss": 0.0214, + "step": 32665 + }, + { + "epoch": 4.28, + "learning_rate": 9.356894657796527e-06, + "loss": 0.0217, + "step": 32666 + }, + { + "epoch": 4.28, + "learning_rate": 9.356365431443301e-06, + "loss": 0.0255, + "step": 32667 + }, + { + "epoch": 4.28, + "learning_rate": 9.355836206900264e-06, + "loss": 0.0211, + "step": 32668 + }, + { + "epoch": 4.28, + "learning_rate": 9.355306984168903e-06, + "loss": 0.0189, + "step": 32669 + }, + { + "epoch": 4.28, + "learning_rate": 9.354777763250702e-06, + "loss": 0.0219, + "step": 32670 + }, + { + "epoch": 4.28, + "learning_rate": 9.354248544147158e-06, + "loss": 0.022, + "step": 32671 + }, + { + "epoch": 4.28, + "learning_rate": 9.353719326859757e-06, + "loss": 0.019, + "step": 32672 + }, + { + "epoch": 4.28, + "learning_rate": 9.353190111389984e-06, + "loss": 0.0245, + "step": 32673 + }, + { + "epoch": 4.28, + "learning_rate": 9.35266089773933e-06, + "loss": 0.0308, + "step": 32674 + }, + { + "epoch": 4.28, + "learning_rate": 9.352131685909284e-06, + "loss": 0.0181, + "step": 32675 + }, + { + "epoch": 4.28, + "learning_rate": 9.351602475901332e-06, + "loss": 0.0201, + "step": 32676 + }, + { + "epoch": 4.28, + "learning_rate": 9.351073267716964e-06, + "loss": 0.021, + "step": 32677 + }, + { + "epoch": 4.28, + "learning_rate": 9.350544061357667e-06, + "loss": 0.0272, + "step": 32678 + }, + { + "epoch": 4.28, + "learning_rate": 9.350014856824931e-06, + "loss": 0.0186, + "step": 32679 + }, + { + "epoch": 4.28, + "learning_rate": 9.349485654120245e-06, + "loss": 0.025, + "step": 32680 + }, + { + "epoch": 4.28, + "learning_rate": 9.34895645324509e-06, + "loss": 0.0151, + "step": 32681 + }, + { + "epoch": 4.28, + "learning_rate": 9.348427254200963e-06, + "loss": 0.0233, + "step": 32682 + }, + { + "epoch": 4.28, + "learning_rate": 9.347898056989353e-06, + "loss": 0.0227, + "step": 32683 + }, + { + "epoch": 4.28, + "learning_rate": 9.347368861611743e-06, + "loss": 0.0246, + "step": 32684 + }, + { + "epoch": 4.28, + "learning_rate": 9.346839668069626e-06, + "loss": 0.0196, + "step": 32685 + }, + { + "epoch": 4.28, + "learning_rate": 9.346310476364483e-06, + "loss": 0.0205, + "step": 32686 + }, + { + "epoch": 4.28, + "learning_rate": 9.345781286497811e-06, + "loss": 0.0218, + "step": 32687 + }, + { + "epoch": 4.28, + "learning_rate": 9.34525209847109e-06, + "loss": 0.0235, + "step": 32688 + }, + { + "epoch": 4.28, + "learning_rate": 9.344722912285815e-06, + "loss": 0.0203, + "step": 32689 + }, + { + "epoch": 4.28, + "learning_rate": 9.344193727943473e-06, + "loss": 0.025, + "step": 32690 + }, + { + "epoch": 4.28, + "learning_rate": 9.34366454544555e-06, + "loss": 0.0249, + "step": 32691 + }, + { + "epoch": 4.28, + "learning_rate": 9.343135364793534e-06, + "loss": 0.0247, + "step": 32692 + }, + { + "epoch": 4.28, + "learning_rate": 9.342606185988917e-06, + "loss": 0.0195, + "step": 32693 + }, + { + "epoch": 4.28, + "learning_rate": 9.342077009033183e-06, + "loss": 0.0203, + "step": 32694 + }, + { + "epoch": 4.28, + "learning_rate": 9.341547833927825e-06, + "loss": 0.0188, + "step": 32695 + }, + { + "epoch": 4.28, + "learning_rate": 9.341018660674328e-06, + "loss": 0.0255, + "step": 32696 + }, + { + "epoch": 4.28, + "learning_rate": 9.340489489274179e-06, + "loss": 0.0204, + "step": 32697 + }, + { + "epoch": 4.28, + "learning_rate": 9.339960319728872e-06, + "loss": 0.0247, + "step": 32698 + }, + { + "epoch": 4.28, + "learning_rate": 9.339431152039889e-06, + "loss": 0.0196, + "step": 32699 + }, + { + "epoch": 4.28, + "learning_rate": 9.338901986208721e-06, + "loss": 0.0228, + "step": 32700 + }, + { + "epoch": 4.28, + "learning_rate": 9.338372822236856e-06, + "loss": 0.0228, + "step": 32701 + }, + { + "epoch": 4.28, + "learning_rate": 9.337843660125783e-06, + "loss": 0.0226, + "step": 32702 + }, + { + "epoch": 4.28, + "learning_rate": 9.337314499876983e-06, + "loss": 0.0265, + "step": 32703 + }, + { + "epoch": 4.28, + "learning_rate": 9.336785341491957e-06, + "loss": 0.0198, + "step": 32704 + }, + { + "epoch": 4.28, + "learning_rate": 9.336256184972186e-06, + "loss": 0.0278, + "step": 32705 + }, + { + "epoch": 4.28, + "learning_rate": 9.33572703031916e-06, + "loss": 0.0236, + "step": 32706 + }, + { + "epoch": 4.28, + "learning_rate": 9.335197877534365e-06, + "loss": 0.0242, + "step": 32707 + }, + { + "epoch": 4.28, + "learning_rate": 9.334668726619292e-06, + "loss": 0.0238, + "step": 32708 + }, + { + "epoch": 4.28, + "learning_rate": 9.334139577575427e-06, + "loss": 0.0238, + "step": 32709 + }, + { + "epoch": 4.28, + "learning_rate": 9.333610430404258e-06, + "loss": 0.0217, + "step": 32710 + }, + { + "epoch": 4.28, + "learning_rate": 9.333081285107275e-06, + "loss": 0.0241, + "step": 32711 + }, + { + "epoch": 4.29, + "learning_rate": 9.332552141685967e-06, + "loss": 0.024, + "step": 32712 + }, + { + "epoch": 4.29, + "learning_rate": 9.332023000141818e-06, + "loss": 0.024, + "step": 32713 + }, + { + "epoch": 4.29, + "learning_rate": 9.331493860476318e-06, + "loss": 0.023, + "step": 32714 + }, + { + "epoch": 4.29, + "learning_rate": 9.330964722690958e-06, + "loss": 0.0191, + "step": 32715 + }, + { + "epoch": 4.29, + "learning_rate": 9.330435586787225e-06, + "loss": 0.0245, + "step": 32716 + }, + { + "epoch": 4.29, + "learning_rate": 9.329906452766603e-06, + "loss": 0.0215, + "step": 32717 + }, + { + "epoch": 4.29, + "learning_rate": 9.329377320630585e-06, + "loss": 0.0236, + "step": 32718 + }, + { + "epoch": 4.29, + "learning_rate": 9.32884819038066e-06, + "loss": 0.0283, + "step": 32719 + }, + { + "epoch": 4.29, + "learning_rate": 9.328319062018311e-06, + "loss": 0.0202, + "step": 32720 + }, + { + "epoch": 4.29, + "learning_rate": 9.327789935545031e-06, + "loss": 0.0267, + "step": 32721 + }, + { + "epoch": 4.29, + "learning_rate": 9.327260810962304e-06, + "loss": 0.0235, + "step": 32722 + }, + { + "epoch": 4.29, + "learning_rate": 9.32673168827162e-06, + "loss": 0.0179, + "step": 32723 + }, + { + "epoch": 4.29, + "learning_rate": 9.326202567474467e-06, + "loss": 0.0235, + "step": 32724 + }, + { + "epoch": 4.29, + "learning_rate": 9.325673448572332e-06, + "loss": 0.0194, + "step": 32725 + }, + { + "epoch": 4.29, + "learning_rate": 9.325144331566707e-06, + "loss": 0.0196, + "step": 32726 + }, + { + "epoch": 4.29, + "learning_rate": 9.324615216459078e-06, + "loss": 0.0223, + "step": 32727 + }, + { + "epoch": 4.29, + "learning_rate": 9.324086103250932e-06, + "loss": 0.0155, + "step": 32728 + }, + { + "epoch": 4.29, + "learning_rate": 9.32355699194376e-06, + "loss": 0.0267, + "step": 32729 + }, + { + "epoch": 4.29, + "learning_rate": 9.323027882539045e-06, + "loss": 0.0198, + "step": 32730 + }, + { + "epoch": 4.29, + "learning_rate": 9.32249877503828e-06, + "loss": 0.0246, + "step": 32731 + }, + { + "epoch": 4.29, + "learning_rate": 9.321969669442949e-06, + "loss": 0.0153, + "step": 32732 + }, + { + "epoch": 4.29, + "learning_rate": 9.321440565754544e-06, + "loss": 0.0193, + "step": 32733 + }, + { + "epoch": 4.29, + "learning_rate": 9.320911463974551e-06, + "loss": 0.0185, + "step": 32734 + }, + { + "epoch": 4.29, + "learning_rate": 9.320382364104458e-06, + "loss": 0.0262, + "step": 32735 + }, + { + "epoch": 4.29, + "learning_rate": 9.319853266145752e-06, + "loss": 0.0189, + "step": 32736 + }, + { + "epoch": 4.29, + "learning_rate": 9.319324170099925e-06, + "loss": 0.0264, + "step": 32737 + }, + { + "epoch": 4.29, + "learning_rate": 9.31879507596846e-06, + "loss": 0.0224, + "step": 32738 + }, + { + "epoch": 4.29, + "learning_rate": 9.318265983752848e-06, + "loss": 0.0216, + "step": 32739 + }, + { + "epoch": 4.29, + "learning_rate": 9.317736893454579e-06, + "loss": 0.0217, + "step": 32740 + }, + { + "epoch": 4.29, + "learning_rate": 9.317207805075135e-06, + "loss": 0.0251, + "step": 32741 + }, + { + "epoch": 4.29, + "learning_rate": 9.316678718616012e-06, + "loss": 0.025, + "step": 32742 + }, + { + "epoch": 4.29, + "learning_rate": 9.316149634078691e-06, + "loss": 0.023, + "step": 32743 + }, + { + "epoch": 4.29, + "learning_rate": 9.315620551464665e-06, + "loss": 0.0227, + "step": 32744 + }, + { + "epoch": 4.29, + "learning_rate": 9.315091470775417e-06, + "loss": 0.0179, + "step": 32745 + }, + { + "epoch": 4.29, + "learning_rate": 9.31456239201244e-06, + "loss": 0.0245, + "step": 32746 + }, + { + "epoch": 4.29, + "learning_rate": 9.314033315177214e-06, + "loss": 0.021, + "step": 32747 + }, + { + "epoch": 4.29, + "learning_rate": 9.31350424027124e-06, + "loss": 0.0293, + "step": 32748 + }, + { + "epoch": 4.29, + "learning_rate": 9.312975167295994e-06, + "loss": 0.0252, + "step": 32749 + }, + { + "epoch": 4.29, + "learning_rate": 9.312446096252973e-06, + "loss": 0.0187, + "step": 32750 + }, + { + "epoch": 4.29, + "learning_rate": 9.311917027143658e-06, + "loss": 0.0214, + "step": 32751 + }, + { + "epoch": 4.29, + "learning_rate": 9.311387959969542e-06, + "loss": 0.0232, + "step": 32752 + }, + { + "epoch": 4.29, + "learning_rate": 9.310858894732108e-06, + "loss": 0.0205, + "step": 32753 + }, + { + "epoch": 4.29, + "learning_rate": 9.310329831432847e-06, + "loss": 0.0193, + "step": 32754 + }, + { + "epoch": 4.29, + "learning_rate": 9.30980077007325e-06, + "loss": 0.0228, + "step": 32755 + }, + { + "epoch": 4.29, + "learning_rate": 9.309271710654796e-06, + "loss": 0.0244, + "step": 32756 + }, + { + "epoch": 4.29, + "learning_rate": 9.308742653178984e-06, + "loss": 0.0238, + "step": 32757 + }, + { + "epoch": 4.29, + "learning_rate": 9.308213597647292e-06, + "loss": 0.0231, + "step": 32758 + }, + { + "epoch": 4.29, + "learning_rate": 9.307684544061212e-06, + "loss": 0.0205, + "step": 32759 + }, + { + "epoch": 4.29, + "learning_rate": 9.307155492422234e-06, + "loss": 0.0186, + "step": 32760 + }, + { + "epoch": 4.29, + "learning_rate": 9.306626442731847e-06, + "loss": 0.0186, + "step": 32761 + }, + { + "epoch": 4.29, + "learning_rate": 9.306097394991534e-06, + "loss": 0.0222, + "step": 32762 + }, + { + "epoch": 4.29, + "learning_rate": 9.305568349202785e-06, + "loss": 0.0182, + "step": 32763 + }, + { + "epoch": 4.29, + "learning_rate": 9.30503930536709e-06, + "loss": 0.0208, + "step": 32764 + }, + { + "epoch": 4.29, + "learning_rate": 9.304510263485934e-06, + "loss": 0.0207, + "step": 32765 + }, + { + "epoch": 4.29, + "learning_rate": 9.303981223560805e-06, + "loss": 0.0244, + "step": 32766 + }, + { + "epoch": 4.29, + "learning_rate": 9.303452185593193e-06, + "loss": 0.0256, + "step": 32767 + }, + { + "epoch": 4.29, + "learning_rate": 9.302923149584584e-06, + "loss": 0.021, + "step": 32768 + }, + { + "epoch": 4.29, + "learning_rate": 9.302394115536462e-06, + "loss": 0.0331, + "step": 32769 + }, + { + "epoch": 4.29, + "learning_rate": 9.301865083450322e-06, + "loss": 0.0233, + "step": 32770 + }, + { + "epoch": 4.29, + "learning_rate": 9.301336053327651e-06, + "loss": 0.0232, + "step": 32771 + }, + { + "epoch": 4.29, + "learning_rate": 9.300807025169936e-06, + "loss": 0.0224, + "step": 32772 + }, + { + "epoch": 4.29, + "learning_rate": 9.300277998978664e-06, + "loss": 0.0222, + "step": 32773 + }, + { + "epoch": 4.29, + "learning_rate": 9.29974897475532e-06, + "loss": 0.0194, + "step": 32774 + }, + { + "epoch": 4.29, + "learning_rate": 9.299219952501396e-06, + "loss": 0.0275, + "step": 32775 + }, + { + "epoch": 4.29, + "learning_rate": 9.298690932218382e-06, + "loss": 0.0207, + "step": 32776 + }, + { + "epoch": 4.29, + "learning_rate": 9.298161913907759e-06, + "loss": 0.0171, + "step": 32777 + }, + { + "epoch": 4.29, + "learning_rate": 9.297632897571019e-06, + "loss": 0.0197, + "step": 32778 + }, + { + "epoch": 4.29, + "learning_rate": 9.297103883209646e-06, + "loss": 0.0224, + "step": 32779 + }, + { + "epoch": 4.29, + "learning_rate": 9.296574870825133e-06, + "loss": 0.0263, + "step": 32780 + }, + { + "epoch": 4.29, + "learning_rate": 9.296045860418967e-06, + "loss": 0.0177, + "step": 32781 + }, + { + "epoch": 4.29, + "learning_rate": 9.295516851992635e-06, + "loss": 0.0187, + "step": 32782 + }, + { + "epoch": 4.29, + "learning_rate": 9.294987845547622e-06, + "loss": 0.0161, + "step": 32783 + }, + { + "epoch": 4.29, + "learning_rate": 9.29445884108542e-06, + "loss": 0.0248, + "step": 32784 + }, + { + "epoch": 4.29, + "learning_rate": 9.293929838607514e-06, + "loss": 0.0239, + "step": 32785 + }, + { + "epoch": 4.29, + "learning_rate": 9.293400838115394e-06, + "loss": 0.0241, + "step": 32786 + }, + { + "epoch": 4.29, + "learning_rate": 9.292871839610544e-06, + "loss": 0.0224, + "step": 32787 + }, + { + "epoch": 4.29, + "learning_rate": 9.292342843094458e-06, + "loss": 0.0233, + "step": 32788 + }, + { + "epoch": 4.3, + "learning_rate": 9.291813848568617e-06, + "loss": 0.0203, + "step": 32789 + }, + { + "epoch": 4.3, + "learning_rate": 9.29128485603451e-06, + "loss": 0.0268, + "step": 32790 + }, + { + "epoch": 4.3, + "learning_rate": 9.29075586549363e-06, + "loss": 0.0272, + "step": 32791 + }, + { + "epoch": 4.3, + "learning_rate": 9.290226876947461e-06, + "loss": 0.0204, + "step": 32792 + }, + { + "epoch": 4.3, + "learning_rate": 9.28969789039749e-06, + "loss": 0.0256, + "step": 32793 + }, + { + "epoch": 4.3, + "learning_rate": 9.289168905845208e-06, + "loss": 0.025, + "step": 32794 + }, + { + "epoch": 4.3, + "learning_rate": 9.288639923292099e-06, + "loss": 0.0239, + "step": 32795 + }, + { + "epoch": 4.3, + "learning_rate": 9.288110942739655e-06, + "loss": 0.0213, + "step": 32796 + }, + { + "epoch": 4.3, + "learning_rate": 9.287581964189358e-06, + "loss": 0.0194, + "step": 32797 + }, + { + "epoch": 4.3, + "learning_rate": 9.287052987642697e-06, + "loss": 0.0188, + "step": 32798 + }, + { + "epoch": 4.3, + "learning_rate": 9.286524013101166e-06, + "loss": 0.0241, + "step": 32799 + }, + { + "epoch": 4.3, + "learning_rate": 9.285995040566246e-06, + "loss": 0.0234, + "step": 32800 + }, + { + "epoch": 4.3, + "learning_rate": 9.285466070039425e-06, + "loss": 0.0222, + "step": 32801 + }, + { + "epoch": 4.3, + "learning_rate": 9.284937101522194e-06, + "loss": 0.0225, + "step": 32802 + }, + { + "epoch": 4.3, + "learning_rate": 9.28440813501604e-06, + "loss": 0.0214, + "step": 32803 + }, + { + "epoch": 4.3, + "learning_rate": 9.283879170522449e-06, + "loss": 0.0238, + "step": 32804 + }, + { + "epoch": 4.3, + "learning_rate": 9.283350208042913e-06, + "loss": 0.0171, + "step": 32805 + }, + { + "epoch": 4.3, + "learning_rate": 9.282821247578912e-06, + "loss": 0.027, + "step": 32806 + }, + { + "epoch": 4.3, + "learning_rate": 9.282292289131943e-06, + "loss": 0.0235, + "step": 32807 + }, + { + "epoch": 4.3, + "learning_rate": 9.281763332703485e-06, + "loss": 0.0255, + "step": 32808 + }, + { + "epoch": 4.3, + "learning_rate": 9.281234378295033e-06, + "loss": 0.0266, + "step": 32809 + }, + { + "epoch": 4.3, + "learning_rate": 9.280705425908067e-06, + "loss": 0.0187, + "step": 32810 + }, + { + "epoch": 4.3, + "learning_rate": 9.280176475544079e-06, + "loss": 0.0208, + "step": 32811 + }, + { + "epoch": 4.3, + "learning_rate": 9.279647527204554e-06, + "loss": 0.0258, + "step": 32812 + }, + { + "epoch": 4.3, + "learning_rate": 9.279118580890987e-06, + "loss": 0.0178, + "step": 32813 + }, + { + "epoch": 4.3, + "learning_rate": 9.278589636604858e-06, + "loss": 0.0231, + "step": 32814 + }, + { + "epoch": 4.3, + "learning_rate": 9.27806069434766e-06, + "loss": 0.0247, + "step": 32815 + }, + { + "epoch": 4.3, + "learning_rate": 9.277531754120874e-06, + "loss": 0.0254, + "step": 32816 + }, + { + "epoch": 4.3, + "learning_rate": 9.277002815925996e-06, + "loss": 0.0174, + "step": 32817 + }, + { + "epoch": 4.3, + "learning_rate": 9.276473879764506e-06, + "loss": 0.0238, + "step": 32818 + }, + { + "epoch": 4.3, + "learning_rate": 9.275944945637893e-06, + "loss": 0.025, + "step": 32819 + }, + { + "epoch": 4.3, + "learning_rate": 9.275416013547651e-06, + "loss": 0.0231, + "step": 32820 + }, + { + "epoch": 4.3, + "learning_rate": 9.274887083495259e-06, + "loss": 0.0226, + "step": 32821 + }, + { + "epoch": 4.3, + "learning_rate": 9.274358155482212e-06, + "loss": 0.0193, + "step": 32822 + }, + { + "epoch": 4.3, + "learning_rate": 9.27382922950999e-06, + "loss": 0.0303, + "step": 32823 + }, + { + "epoch": 4.3, + "learning_rate": 9.273300305580083e-06, + "loss": 0.0195, + "step": 32824 + }, + { + "epoch": 4.3, + "learning_rate": 9.272771383693985e-06, + "loss": 0.0261, + "step": 32825 + }, + { + "epoch": 4.3, + "learning_rate": 9.272242463853178e-06, + "loss": 0.0204, + "step": 32826 + }, + { + "epoch": 4.3, + "learning_rate": 9.27171354605915e-06, + "loss": 0.0206, + "step": 32827 + }, + { + "epoch": 4.3, + "learning_rate": 9.27118463031339e-06, + "loss": 0.0255, + "step": 32828 + }, + { + "epoch": 4.3, + "learning_rate": 9.270655716617381e-06, + "loss": 0.0246, + "step": 32829 + }, + { + "epoch": 4.3, + "learning_rate": 9.270126804972619e-06, + "loss": 0.0273, + "step": 32830 + }, + { + "epoch": 4.3, + "learning_rate": 9.269597895380582e-06, + "loss": 0.0191, + "step": 32831 + }, + { + "epoch": 4.3, + "learning_rate": 9.269068987842766e-06, + "loss": 0.0202, + "step": 32832 + }, + { + "epoch": 4.3, + "learning_rate": 9.268540082360651e-06, + "loss": 0.022, + "step": 32833 + }, + { + "epoch": 4.3, + "learning_rate": 9.268011178935725e-06, + "loss": 0.0238, + "step": 32834 + }, + { + "epoch": 4.3, + "learning_rate": 9.267482277569483e-06, + "loss": 0.0205, + "step": 32835 + }, + { + "epoch": 4.3, + "learning_rate": 9.266953378263409e-06, + "loss": 0.0167, + "step": 32836 + }, + { + "epoch": 4.3, + "learning_rate": 9.266424481018988e-06, + "loss": 0.0247, + "step": 32837 + }, + { + "epoch": 4.3, + "learning_rate": 9.26589558583771e-06, + "loss": 0.0213, + "step": 32838 + }, + { + "epoch": 4.3, + "learning_rate": 9.26536669272106e-06, + "loss": 0.0168, + "step": 32839 + }, + { + "epoch": 4.3, + "learning_rate": 9.264837801670529e-06, + "loss": 0.0221, + "step": 32840 + }, + { + "epoch": 4.3, + "learning_rate": 9.264308912687603e-06, + "loss": 0.0201, + "step": 32841 + }, + { + "epoch": 4.3, + "learning_rate": 9.263780025773767e-06, + "loss": 0.0195, + "step": 32842 + }, + { + "epoch": 4.3, + "learning_rate": 9.263251140930512e-06, + "loss": 0.0198, + "step": 32843 + }, + { + "epoch": 4.3, + "learning_rate": 9.262722258159322e-06, + "loss": 0.0225, + "step": 32844 + }, + { + "epoch": 4.3, + "learning_rate": 9.262193377461685e-06, + "loss": 0.0197, + "step": 32845 + }, + { + "epoch": 4.3, + "learning_rate": 9.261664498839093e-06, + "loss": 0.0243, + "step": 32846 + }, + { + "epoch": 4.3, + "learning_rate": 9.261135622293028e-06, + "loss": 0.0228, + "step": 32847 + }, + { + "epoch": 4.3, + "learning_rate": 9.260606747824981e-06, + "loss": 0.0207, + "step": 32848 + }, + { + "epoch": 4.3, + "learning_rate": 9.26007787543644e-06, + "loss": 0.0182, + "step": 32849 + }, + { + "epoch": 4.3, + "learning_rate": 9.25954900512889e-06, + "loss": 0.021, + "step": 32850 + }, + { + "epoch": 4.3, + "learning_rate": 9.259020136903818e-06, + "loss": 0.0263, + "step": 32851 + }, + { + "epoch": 4.3, + "learning_rate": 9.258491270762711e-06, + "loss": 0.0198, + "step": 32852 + }, + { + "epoch": 4.3, + "learning_rate": 9.257962406707062e-06, + "loss": 0.0235, + "step": 32853 + }, + { + "epoch": 4.3, + "learning_rate": 9.25743354473835e-06, + "loss": 0.0208, + "step": 32854 + }, + { + "epoch": 4.3, + "learning_rate": 9.256904684858068e-06, + "loss": 0.0271, + "step": 32855 + }, + { + "epoch": 4.3, + "learning_rate": 9.256375827067698e-06, + "loss": 0.0223, + "step": 32856 + }, + { + "epoch": 4.3, + "learning_rate": 9.255846971368737e-06, + "loss": 0.0216, + "step": 32857 + }, + { + "epoch": 4.3, + "learning_rate": 9.255318117762666e-06, + "loss": 0.0254, + "step": 32858 + }, + { + "epoch": 4.3, + "learning_rate": 9.254789266250973e-06, + "loss": 0.0221, + "step": 32859 + }, + { + "epoch": 4.3, + "learning_rate": 9.254260416835143e-06, + "loss": 0.0196, + "step": 32860 + }, + { + "epoch": 4.3, + "learning_rate": 9.253731569516667e-06, + "loss": 0.0231, + "step": 32861 + }, + { + "epoch": 4.3, + "learning_rate": 9.253202724297035e-06, + "loss": 0.0188, + "step": 32862 + }, + { + "epoch": 4.3, + "learning_rate": 9.252673881177726e-06, + "loss": 0.0185, + "step": 32863 + }, + { + "epoch": 4.3, + "learning_rate": 9.252145040160235e-06, + "loss": 0.0229, + "step": 32864 + }, + { + "epoch": 4.31, + "learning_rate": 9.251616201246043e-06, + "loss": 0.0259, + "step": 32865 + }, + { + "epoch": 4.31, + "learning_rate": 9.251087364436644e-06, + "loss": 0.018, + "step": 32866 + }, + { + "epoch": 4.31, + "learning_rate": 9.250558529733518e-06, + "loss": 0.0198, + "step": 32867 + }, + { + "epoch": 4.31, + "learning_rate": 9.250029697138157e-06, + "loss": 0.0192, + "step": 32868 + }, + { + "epoch": 4.31, + "learning_rate": 9.249500866652048e-06, + "loss": 0.0188, + "step": 32869 + }, + { + "epoch": 4.31, + "learning_rate": 9.24897203827668e-06, + "loss": 0.0212, + "step": 32870 + }, + { + "epoch": 4.31, + "learning_rate": 9.248443212013536e-06, + "loss": 0.0189, + "step": 32871 + }, + { + "epoch": 4.31, + "learning_rate": 9.247914387864108e-06, + "loss": 0.0235, + "step": 32872 + }, + { + "epoch": 4.31, + "learning_rate": 9.247385565829879e-06, + "loss": 0.0239, + "step": 32873 + }, + { + "epoch": 4.31, + "learning_rate": 9.246856745912339e-06, + "loss": 0.0247, + "step": 32874 + }, + { + "epoch": 4.31, + "learning_rate": 9.246327928112973e-06, + "loss": 0.0193, + "step": 32875 + }, + { + "epoch": 4.31, + "learning_rate": 9.245799112433269e-06, + "loss": 0.0199, + "step": 32876 + }, + { + "epoch": 4.31, + "learning_rate": 9.245270298874717e-06, + "loss": 0.0239, + "step": 32877 + }, + { + "epoch": 4.31, + "learning_rate": 9.244741487438798e-06, + "loss": 0.0234, + "step": 32878 + }, + { + "epoch": 4.31, + "learning_rate": 9.244212678127007e-06, + "loss": 0.0274, + "step": 32879 + }, + { + "epoch": 4.31, + "learning_rate": 9.243683870940828e-06, + "loss": 0.0211, + "step": 32880 + }, + { + "epoch": 4.31, + "learning_rate": 9.243155065881745e-06, + "loss": 0.0188, + "step": 32881 + }, + { + "epoch": 4.31, + "learning_rate": 9.242626262951251e-06, + "loss": 0.0205, + "step": 32882 + }, + { + "epoch": 4.31, + "learning_rate": 9.242097462150828e-06, + "loss": 0.017, + "step": 32883 + }, + { + "epoch": 4.31, + "learning_rate": 9.241568663481965e-06, + "loss": 0.0218, + "step": 32884 + }, + { + "epoch": 4.31, + "learning_rate": 9.241039866946153e-06, + "loss": 0.018, + "step": 32885 + }, + { + "epoch": 4.31, + "learning_rate": 9.240511072544874e-06, + "loss": 0.0273, + "step": 32886 + }, + { + "epoch": 4.31, + "learning_rate": 9.23998228027962e-06, + "loss": 0.0217, + "step": 32887 + }, + { + "epoch": 4.31, + "learning_rate": 9.239453490151871e-06, + "loss": 0.0252, + "step": 32888 + }, + { + "epoch": 4.31, + "learning_rate": 9.238924702163116e-06, + "loss": 0.0185, + "step": 32889 + }, + { + "epoch": 4.31, + "learning_rate": 9.23839591631485e-06, + "loss": 0.0235, + "step": 32890 + }, + { + "epoch": 4.31, + "learning_rate": 9.237867132608556e-06, + "loss": 0.0198, + "step": 32891 + }, + { + "epoch": 4.31, + "learning_rate": 9.237338351045716e-06, + "loss": 0.0211, + "step": 32892 + }, + { + "epoch": 4.31, + "learning_rate": 9.236809571627825e-06, + "loss": 0.0204, + "step": 32893 + }, + { + "epoch": 4.31, + "learning_rate": 9.236280794356365e-06, + "loss": 0.0208, + "step": 32894 + }, + { + "epoch": 4.31, + "learning_rate": 9.235752019232826e-06, + "loss": 0.0184, + "step": 32895 + }, + { + "epoch": 4.31, + "learning_rate": 9.23522324625869e-06, + "loss": 0.018, + "step": 32896 + }, + { + "epoch": 4.31, + "learning_rate": 9.234694475435451e-06, + "loss": 0.02, + "step": 32897 + }, + { + "epoch": 4.31, + "learning_rate": 9.234165706764592e-06, + "loss": 0.0261, + "step": 32898 + }, + { + "epoch": 4.31, + "learning_rate": 9.233636940247601e-06, + "loss": 0.0213, + "step": 32899 + }, + { + "epoch": 4.31, + "learning_rate": 9.233108175885963e-06, + "loss": 0.0275, + "step": 32900 + }, + { + "epoch": 4.31, + "learning_rate": 9.23257941368117e-06, + "loss": 0.0175, + "step": 32901 + }, + { + "epoch": 4.31, + "learning_rate": 9.232050653634706e-06, + "loss": 0.0202, + "step": 32902 + }, + { + "epoch": 4.31, + "learning_rate": 9.23152189574806e-06, + "loss": 0.025, + "step": 32903 + }, + { + "epoch": 4.31, + "learning_rate": 9.230993140022718e-06, + "loss": 0.019, + "step": 32904 + }, + { + "epoch": 4.31, + "learning_rate": 9.230464386460164e-06, + "loss": 0.0177, + "step": 32905 + }, + { + "epoch": 4.31, + "learning_rate": 9.229935635061892e-06, + "loss": 0.0169, + "step": 32906 + }, + { + "epoch": 4.31, + "learning_rate": 9.229406885829382e-06, + "loss": 0.0224, + "step": 32907 + }, + { + "epoch": 4.31, + "learning_rate": 9.228878138764126e-06, + "loss": 0.0217, + "step": 32908 + }, + { + "epoch": 4.31, + "learning_rate": 9.228349393867607e-06, + "loss": 0.0194, + "step": 32909 + }, + { + "epoch": 4.31, + "learning_rate": 9.227820651141316e-06, + "loss": 0.0253, + "step": 32910 + }, + { + "epoch": 4.31, + "learning_rate": 9.227291910586733e-06, + "loss": 0.0242, + "step": 32911 + }, + { + "epoch": 4.31, + "learning_rate": 9.226763172205356e-06, + "loss": 0.0206, + "step": 32912 + }, + { + "epoch": 4.31, + "learning_rate": 9.226234435998664e-06, + "loss": 0.0219, + "step": 32913 + }, + { + "epoch": 4.31, + "learning_rate": 9.22570570196815e-06, + "loss": 0.013, + "step": 32914 + }, + { + "epoch": 4.31, + "learning_rate": 9.225176970115293e-06, + "loss": 0.0199, + "step": 32915 + }, + { + "epoch": 4.31, + "learning_rate": 9.224648240441589e-06, + "loss": 0.0238, + "step": 32916 + }, + { + "epoch": 4.31, + "learning_rate": 9.224119512948517e-06, + "loss": 0.0244, + "step": 32917 + }, + { + "epoch": 4.31, + "learning_rate": 9.22359078763757e-06, + "loss": 0.0222, + "step": 32918 + }, + { + "epoch": 4.31, + "learning_rate": 9.22306206451023e-06, + "loss": 0.0233, + "step": 32919 + }, + { + "epoch": 4.31, + "learning_rate": 9.222533343567989e-06, + "loss": 0.0216, + "step": 32920 + }, + { + "epoch": 4.31, + "learning_rate": 9.222004624812326e-06, + "loss": 0.0234, + "step": 32921 + }, + { + "epoch": 4.31, + "learning_rate": 9.22147590824474e-06, + "loss": 0.0189, + "step": 32922 + }, + { + "epoch": 4.31, + "learning_rate": 9.22094719386671e-06, + "loss": 0.0242, + "step": 32923 + }, + { + "epoch": 4.31, + "learning_rate": 9.220418481679726e-06, + "loss": 0.0195, + "step": 32924 + }, + { + "epoch": 4.31, + "learning_rate": 9.21988977168527e-06, + "loss": 0.0206, + "step": 32925 + }, + { + "epoch": 4.31, + "learning_rate": 9.219361063884833e-06, + "loss": 0.0196, + "step": 32926 + }, + { + "epoch": 4.31, + "learning_rate": 9.218832358279903e-06, + "loss": 0.0182, + "step": 32927 + }, + { + "epoch": 4.31, + "learning_rate": 9.218303654871966e-06, + "loss": 0.0219, + "step": 32928 + }, + { + "epoch": 4.31, + "learning_rate": 9.217774953662507e-06, + "loss": 0.0215, + "step": 32929 + }, + { + "epoch": 4.31, + "learning_rate": 9.217246254653013e-06, + "loss": 0.0238, + "step": 32930 + }, + { + "epoch": 4.31, + "learning_rate": 9.216717557844977e-06, + "loss": 0.0184, + "step": 32931 + }, + { + "epoch": 4.31, + "learning_rate": 9.216188863239875e-06, + "loss": 0.0193, + "step": 32932 + }, + { + "epoch": 4.31, + "learning_rate": 9.215660170839203e-06, + "loss": 0.0165, + "step": 32933 + }, + { + "epoch": 4.31, + "learning_rate": 9.215131480644444e-06, + "loss": 0.0164, + "step": 32934 + }, + { + "epoch": 4.31, + "learning_rate": 9.214602792657088e-06, + "loss": 0.0239, + "step": 32935 + }, + { + "epoch": 4.31, + "learning_rate": 9.214074106878618e-06, + "loss": 0.0253, + "step": 32936 + }, + { + "epoch": 4.31, + "learning_rate": 9.213545423310523e-06, + "loss": 0.0216, + "step": 32937 + }, + { + "epoch": 4.31, + "learning_rate": 9.21301674195429e-06, + "loss": 0.0211, + "step": 32938 + }, + { + "epoch": 4.31, + "learning_rate": 9.212488062811407e-06, + "loss": 0.0266, + "step": 32939 + }, + { + "epoch": 4.31, + "learning_rate": 9.211959385883358e-06, + "loss": 0.0243, + "step": 32940 + }, + { + "epoch": 4.32, + "learning_rate": 9.21143071117163e-06, + "loss": 0.0301, + "step": 32941 + }, + { + "epoch": 4.32, + "learning_rate": 9.210902038677714e-06, + "loss": 0.0174, + "step": 32942 + }, + { + "epoch": 4.32, + "learning_rate": 9.21037336840309e-06, + "loss": 0.0238, + "step": 32943 + }, + { + "epoch": 4.32, + "learning_rate": 9.20984470034925e-06, + "loss": 0.0201, + "step": 32944 + }, + { + "epoch": 4.32, + "learning_rate": 9.209316034517682e-06, + "loss": 0.0285, + "step": 32945 + }, + { + "epoch": 4.32, + "learning_rate": 9.208787370909868e-06, + "loss": 0.0213, + "step": 32946 + }, + { + "epoch": 4.32, + "learning_rate": 9.2082587095273e-06, + "loss": 0.0146, + "step": 32947 + }, + { + "epoch": 4.32, + "learning_rate": 9.207730050371462e-06, + "loss": 0.0211, + "step": 32948 + }, + { + "epoch": 4.32, + "learning_rate": 9.20720139344384e-06, + "loss": 0.0235, + "step": 32949 + }, + { + "epoch": 4.32, + "learning_rate": 9.206672738745922e-06, + "loss": 0.0199, + "step": 32950 + }, + { + "epoch": 4.32, + "learning_rate": 9.206144086279195e-06, + "loss": 0.0224, + "step": 32951 + }, + { + "epoch": 4.32, + "learning_rate": 9.205615436045146e-06, + "loss": 0.0174, + "step": 32952 + }, + { + "epoch": 4.32, + "learning_rate": 9.205086788045259e-06, + "loss": 0.0228, + "step": 32953 + }, + { + "epoch": 4.32, + "learning_rate": 9.204558142281022e-06, + "loss": 0.019, + "step": 32954 + }, + { + "epoch": 4.32, + "learning_rate": 9.204029498753924e-06, + "loss": 0.0191, + "step": 32955 + }, + { + "epoch": 4.32, + "learning_rate": 9.203500857465453e-06, + "loss": 0.0205, + "step": 32956 + }, + { + "epoch": 4.32, + "learning_rate": 9.202972218417093e-06, + "loss": 0.022, + "step": 32957 + }, + { + "epoch": 4.32, + "learning_rate": 9.202443581610333e-06, + "loss": 0.0194, + "step": 32958 + }, + { + "epoch": 4.32, + "learning_rate": 9.201914947046653e-06, + "loss": 0.0201, + "step": 32959 + }, + { + "epoch": 4.32, + "learning_rate": 9.20138631472755e-06, + "loss": 0.023, + "step": 32960 + }, + { + "epoch": 4.32, + "learning_rate": 9.200857684654501e-06, + "loss": 0.0155, + "step": 32961 + }, + { + "epoch": 4.32, + "learning_rate": 9.200329056829e-06, + "loss": 0.0177, + "step": 32962 + }, + { + "epoch": 4.32, + "learning_rate": 9.199800431252528e-06, + "loss": 0.023, + "step": 32963 + }, + { + "epoch": 4.32, + "learning_rate": 9.199271807926577e-06, + "loss": 0.0253, + "step": 32964 + }, + { + "epoch": 4.32, + "learning_rate": 9.198743186852627e-06, + "loss": 0.0199, + "step": 32965 + }, + { + "epoch": 4.32, + "learning_rate": 9.198214568032176e-06, + "loss": 0.0194, + "step": 32966 + }, + { + "epoch": 4.32, + "learning_rate": 9.1976859514667e-06, + "loss": 0.0206, + "step": 32967 + }, + { + "epoch": 4.32, + "learning_rate": 9.197157337157692e-06, + "loss": 0.024, + "step": 32968 + }, + { + "epoch": 4.32, + "learning_rate": 9.196628725106633e-06, + "loss": 0.0233, + "step": 32969 + }, + { + "epoch": 4.32, + "learning_rate": 9.196100115315015e-06, + "loss": 0.024, + "step": 32970 + }, + { + "epoch": 4.32, + "learning_rate": 9.195571507784323e-06, + "loss": 0.0185, + "step": 32971 + }, + { + "epoch": 4.32, + "learning_rate": 9.195042902516042e-06, + "loss": 0.0209, + "step": 32972 + }, + { + "epoch": 4.32, + "learning_rate": 9.194514299511661e-06, + "loss": 0.0209, + "step": 32973 + }, + { + "epoch": 4.32, + "learning_rate": 9.193985698772665e-06, + "loss": 0.0237, + "step": 32974 + }, + { + "epoch": 4.32, + "learning_rate": 9.193457100300543e-06, + "loss": 0.0224, + "step": 32975 + }, + { + "epoch": 4.32, + "learning_rate": 9.192928504096772e-06, + "loss": 0.0231, + "step": 32976 + }, + { + "epoch": 4.32, + "learning_rate": 9.192399910162856e-06, + "loss": 0.0231, + "step": 32977 + }, + { + "epoch": 4.32, + "learning_rate": 9.191871318500268e-06, + "loss": 0.0169, + "step": 32978 + }, + { + "epoch": 4.32, + "learning_rate": 9.191342729110502e-06, + "loss": 0.0226, + "step": 32979 + }, + { + "epoch": 4.32, + "learning_rate": 9.190814141995038e-06, + "loss": 0.0176, + "step": 32980 + }, + { + "epoch": 4.32, + "learning_rate": 9.19028555715537e-06, + "loss": 0.0234, + "step": 32981 + }, + { + "epoch": 4.32, + "learning_rate": 9.189756974592978e-06, + "loss": 0.0234, + "step": 32982 + }, + { + "epoch": 4.32, + "learning_rate": 9.189228394309353e-06, + "loss": 0.0219, + "step": 32983 + }, + { + "epoch": 4.32, + "learning_rate": 9.188699816305976e-06, + "loss": 0.0199, + "step": 32984 + }, + { + "epoch": 4.32, + "learning_rate": 9.188171240584342e-06, + "loss": 0.0261, + "step": 32985 + }, + { + "epoch": 4.32, + "learning_rate": 9.187642667145933e-06, + "loss": 0.025, + "step": 32986 + }, + { + "epoch": 4.32, + "learning_rate": 9.18711409599223e-06, + "loss": 0.0251, + "step": 32987 + }, + { + "epoch": 4.32, + "learning_rate": 9.186585527124729e-06, + "loss": 0.0216, + "step": 32988 + }, + { + "epoch": 4.32, + "learning_rate": 9.186056960544916e-06, + "loss": 0.0237, + "step": 32989 + }, + { + "epoch": 4.32, + "learning_rate": 9.185528396254271e-06, + "loss": 0.0184, + "step": 32990 + }, + { + "epoch": 4.32, + "learning_rate": 9.184999834254283e-06, + "loss": 0.0213, + "step": 32991 + }, + { + "epoch": 4.32, + "learning_rate": 9.184471274546443e-06, + "loss": 0.0198, + "step": 32992 + }, + { + "epoch": 4.32, + "learning_rate": 9.183942717132232e-06, + "loss": 0.0214, + "step": 32993 + }, + { + "epoch": 4.32, + "learning_rate": 9.183414162013141e-06, + "loss": 0.0145, + "step": 32994 + }, + { + "epoch": 4.32, + "learning_rate": 9.182885609190651e-06, + "loss": 0.0266, + "step": 32995 + }, + { + "epoch": 4.32, + "learning_rate": 9.182357058666255e-06, + "loss": 0.0205, + "step": 32996 + }, + { + "epoch": 4.32, + "learning_rate": 9.181828510441435e-06, + "loss": 0.0265, + "step": 32997 + }, + { + "epoch": 4.32, + "learning_rate": 9.181299964517674e-06, + "loss": 0.0191, + "step": 32998 + }, + { + "epoch": 4.32, + "learning_rate": 9.180771420896468e-06, + "loss": 0.0215, + "step": 32999 + }, + { + "epoch": 4.32, + "learning_rate": 9.1802428795793e-06, + "loss": 0.0234, + "step": 33000 + }, + { + "epoch": 4.32, + "learning_rate": 9.179714340567651e-06, + "loss": 0.0194, + "step": 33001 + }, + { + "epoch": 4.32, + "learning_rate": 9.179185803863016e-06, + "loss": 0.0242, + "step": 33002 + }, + { + "epoch": 4.32, + "learning_rate": 9.178657269466874e-06, + "loss": 0.0246, + "step": 33003 + }, + { + "epoch": 4.32, + "learning_rate": 9.178128737380717e-06, + "loss": 0.0265, + "step": 33004 + }, + { + "epoch": 4.32, + "learning_rate": 9.177600207606027e-06, + "loss": 0.0201, + "step": 33005 + }, + { + "epoch": 4.32, + "learning_rate": 9.177071680144293e-06, + "loss": 0.0202, + "step": 33006 + }, + { + "epoch": 4.32, + "learning_rate": 9.176543154997004e-06, + "loss": 0.0188, + "step": 33007 + }, + { + "epoch": 4.32, + "learning_rate": 9.17601463216564e-06, + "loss": 0.0237, + "step": 33008 + }, + { + "epoch": 4.32, + "learning_rate": 9.17548611165169e-06, + "loss": 0.0209, + "step": 33009 + }, + { + "epoch": 4.32, + "learning_rate": 9.174957593456647e-06, + "loss": 0.0297, + "step": 33010 + }, + { + "epoch": 4.32, + "learning_rate": 9.174429077581986e-06, + "loss": 0.0242, + "step": 33011 + }, + { + "epoch": 4.32, + "learning_rate": 9.173900564029203e-06, + "loss": 0.0203, + "step": 33012 + }, + { + "epoch": 4.32, + "learning_rate": 9.17337205279978e-06, + "loss": 0.0214, + "step": 33013 + }, + { + "epoch": 4.32, + "learning_rate": 9.172843543895203e-06, + "loss": 0.0263, + "step": 33014 + }, + { + "epoch": 4.32, + "learning_rate": 9.172315037316961e-06, + "loss": 0.0234, + "step": 33015 + }, + { + "epoch": 4.32, + "learning_rate": 9.171786533066537e-06, + "loss": 0.0199, + "step": 33016 + }, + { + "epoch": 4.32, + "learning_rate": 9.171258031145422e-06, + "loss": 0.0231, + "step": 33017 + }, + { + "epoch": 4.33, + "learning_rate": 9.170729531555097e-06, + "loss": 0.0228, + "step": 33018 + }, + { + "epoch": 4.33, + "learning_rate": 9.170201034297054e-06, + "loss": 0.0234, + "step": 33019 + }, + { + "epoch": 4.33, + "learning_rate": 9.169672539372767e-06, + "loss": 0.0215, + "step": 33020 + }, + { + "epoch": 4.33, + "learning_rate": 9.169144046783743e-06, + "loss": 0.0189, + "step": 33021 + }, + { + "epoch": 4.33, + "learning_rate": 9.168615556531451e-06, + "loss": 0.0214, + "step": 33022 + }, + { + "epoch": 4.33, + "learning_rate": 9.168087068617388e-06, + "loss": 0.0164, + "step": 33023 + }, + { + "epoch": 4.33, + "learning_rate": 9.167558583043032e-06, + "loss": 0.0244, + "step": 33024 + }, + { + "epoch": 4.33, + "learning_rate": 9.167030099809878e-06, + "loss": 0.0262, + "step": 33025 + }, + { + "epoch": 4.33, + "learning_rate": 9.166501618919403e-06, + "loss": 0.0188, + "step": 33026 + }, + { + "epoch": 4.33, + "learning_rate": 9.165973140373096e-06, + "loss": 0.0182, + "step": 33027 + }, + { + "epoch": 4.33, + "learning_rate": 9.16544466417245e-06, + "loss": 0.0209, + "step": 33028 + }, + { + "epoch": 4.33, + "learning_rate": 9.164916190318945e-06, + "loss": 0.0247, + "step": 33029 + }, + { + "epoch": 4.33, + "learning_rate": 9.16438771881407e-06, + "loss": 0.0195, + "step": 33030 + }, + { + "epoch": 4.33, + "learning_rate": 9.163859249659305e-06, + "loss": 0.0233, + "step": 33031 + }, + { + "epoch": 4.33, + "learning_rate": 9.163330782856144e-06, + "loss": 0.0252, + "step": 33032 + }, + { + "epoch": 4.33, + "learning_rate": 9.162802318406072e-06, + "loss": 0.0206, + "step": 33033 + }, + { + "epoch": 4.33, + "learning_rate": 9.162273856310573e-06, + "loss": 0.0258, + "step": 33034 + }, + { + "epoch": 4.33, + "learning_rate": 9.161745396571133e-06, + "loss": 0.0238, + "step": 33035 + }, + { + "epoch": 4.33, + "learning_rate": 9.161216939189243e-06, + "loss": 0.0234, + "step": 33036 + }, + { + "epoch": 4.33, + "learning_rate": 9.160688484166383e-06, + "loss": 0.024, + "step": 33037 + }, + { + "epoch": 4.33, + "learning_rate": 9.160160031504043e-06, + "loss": 0.0195, + "step": 33038 + }, + { + "epoch": 4.33, + "learning_rate": 9.159631581203708e-06, + "loss": 0.0197, + "step": 33039 + }, + { + "epoch": 4.33, + "learning_rate": 9.159103133266865e-06, + "loss": 0.02, + "step": 33040 + }, + { + "epoch": 4.33, + "learning_rate": 9.158574687694997e-06, + "loss": 0.025, + "step": 33041 + }, + { + "epoch": 4.33, + "learning_rate": 9.15804624448959e-06, + "loss": 0.0244, + "step": 33042 + }, + { + "epoch": 4.33, + "learning_rate": 9.157517803652138e-06, + "loss": 0.0178, + "step": 33043 + }, + { + "epoch": 4.33, + "learning_rate": 9.156989365184124e-06, + "loss": 0.0217, + "step": 33044 + }, + { + "epoch": 4.33, + "learning_rate": 9.15646092908703e-06, + "loss": 0.0189, + "step": 33045 + }, + { + "epoch": 4.33, + "learning_rate": 9.155932495362346e-06, + "loss": 0.0223, + "step": 33046 + }, + { + "epoch": 4.33, + "learning_rate": 9.155404064011555e-06, + "loss": 0.0204, + "step": 33047 + }, + { + "epoch": 4.33, + "learning_rate": 9.154875635036148e-06, + "loss": 0.0211, + "step": 33048 + }, + { + "epoch": 4.33, + "learning_rate": 9.154347208437607e-06, + "loss": 0.0206, + "step": 33049 + }, + { + "epoch": 4.33, + "learning_rate": 9.153818784217417e-06, + "loss": 0.0208, + "step": 33050 + }, + { + "epoch": 4.33, + "learning_rate": 9.15329036237707e-06, + "loss": 0.0247, + "step": 33051 + }, + { + "epoch": 4.33, + "learning_rate": 9.152761942918044e-06, + "loss": 0.0215, + "step": 33052 + }, + { + "epoch": 4.33, + "learning_rate": 9.152233525841832e-06, + "loss": 0.0191, + "step": 33053 + }, + { + "epoch": 4.33, + "learning_rate": 9.151705111149923e-06, + "loss": 0.0206, + "step": 33054 + }, + { + "epoch": 4.33, + "learning_rate": 9.151176698843792e-06, + "loss": 0.0222, + "step": 33055 + }, + { + "epoch": 4.33, + "learning_rate": 9.150648288924934e-06, + "loss": 0.0241, + "step": 33056 + }, + { + "epoch": 4.33, + "learning_rate": 9.150119881394834e-06, + "loss": 0.0195, + "step": 33057 + }, + { + "epoch": 4.33, + "learning_rate": 9.149591476254973e-06, + "loss": 0.0234, + "step": 33058 + }, + { + "epoch": 4.33, + "learning_rate": 9.149063073506845e-06, + "loss": 0.0242, + "step": 33059 + }, + { + "epoch": 4.33, + "learning_rate": 9.148534673151927e-06, + "loss": 0.0201, + "step": 33060 + }, + { + "epoch": 4.33, + "learning_rate": 9.148006275191714e-06, + "loss": 0.0225, + "step": 33061 + }, + { + "epoch": 4.33, + "learning_rate": 9.147477879627685e-06, + "loss": 0.0203, + "step": 33062 + }, + { + "epoch": 4.33, + "learning_rate": 9.146949486461328e-06, + "loss": 0.018, + "step": 33063 + }, + { + "epoch": 4.33, + "learning_rate": 9.146421095694132e-06, + "loss": 0.0214, + "step": 33064 + }, + { + "epoch": 4.33, + "learning_rate": 9.145892707327583e-06, + "loss": 0.0232, + "step": 33065 + }, + { + "epoch": 4.33, + "learning_rate": 9.145364321363162e-06, + "loss": 0.0179, + "step": 33066 + }, + { + "epoch": 4.33, + "learning_rate": 9.144835937802362e-06, + "loss": 0.0267, + "step": 33067 + }, + { + "epoch": 4.33, + "learning_rate": 9.144307556646664e-06, + "loss": 0.0217, + "step": 33068 + }, + { + "epoch": 4.33, + "learning_rate": 9.143779177897556e-06, + "loss": 0.0239, + "step": 33069 + }, + { + "epoch": 4.33, + "learning_rate": 9.143250801556522e-06, + "loss": 0.018, + "step": 33070 + }, + { + "epoch": 4.33, + "learning_rate": 9.142722427625048e-06, + "loss": 0.0183, + "step": 33071 + }, + { + "epoch": 4.33, + "learning_rate": 9.142194056104625e-06, + "loss": 0.0237, + "step": 33072 + }, + { + "epoch": 4.33, + "learning_rate": 9.141665686996733e-06, + "loss": 0.0218, + "step": 33073 + }, + { + "epoch": 4.33, + "learning_rate": 9.14113732030286e-06, + "loss": 0.0222, + "step": 33074 + }, + { + "epoch": 4.33, + "learning_rate": 9.140608956024495e-06, + "loss": 0.0239, + "step": 33075 + }, + { + "epoch": 4.33, + "learning_rate": 9.14008059416312e-06, + "loss": 0.0207, + "step": 33076 + }, + { + "epoch": 4.33, + "learning_rate": 9.139552234720223e-06, + "loss": 0.0187, + "step": 33077 + }, + { + "epoch": 4.33, + "learning_rate": 9.139023877697291e-06, + "loss": 0.0204, + "step": 33078 + }, + { + "epoch": 4.33, + "learning_rate": 9.138495523095807e-06, + "loss": 0.0187, + "step": 33079 + }, + { + "epoch": 4.33, + "learning_rate": 9.137967170917258e-06, + "loss": 0.023, + "step": 33080 + }, + { + "epoch": 4.33, + "learning_rate": 9.137438821163132e-06, + "loss": 0.0182, + "step": 33081 + }, + { + "epoch": 4.33, + "learning_rate": 9.136910473834913e-06, + "loss": 0.0201, + "step": 33082 + }, + { + "epoch": 4.33, + "learning_rate": 9.136382128934087e-06, + "loss": 0.0176, + "step": 33083 + }, + { + "epoch": 4.33, + "learning_rate": 9.135853786462141e-06, + "loss": 0.0197, + "step": 33084 + }, + { + "epoch": 4.33, + "learning_rate": 9.135325446420555e-06, + "loss": 0.0203, + "step": 33085 + }, + { + "epoch": 4.33, + "learning_rate": 9.134797108810826e-06, + "loss": 0.02, + "step": 33086 + }, + { + "epoch": 4.33, + "learning_rate": 9.134268773634431e-06, + "loss": 0.0169, + "step": 33087 + }, + { + "epoch": 4.33, + "learning_rate": 9.133740440892861e-06, + "loss": 0.0214, + "step": 33088 + }, + { + "epoch": 4.33, + "learning_rate": 9.133212110587599e-06, + "loss": 0.0245, + "step": 33089 + }, + { + "epoch": 4.33, + "learning_rate": 9.132683782720134e-06, + "loss": 0.0181, + "step": 33090 + }, + { + "epoch": 4.33, + "learning_rate": 9.132155457291947e-06, + "loss": 0.0158, + "step": 33091 + }, + { + "epoch": 4.33, + "learning_rate": 9.131627134304526e-06, + "loss": 0.021, + "step": 33092 + }, + { + "epoch": 4.33, + "learning_rate": 9.131098813759361e-06, + "loss": 0.0247, + "step": 33093 + }, + { + "epoch": 4.34, + "learning_rate": 9.13057049565793e-06, + "loss": 0.0233, + "step": 33094 + }, + { + "epoch": 4.34, + "learning_rate": 9.130042180001729e-06, + "loss": 0.0167, + "step": 33095 + }, + { + "epoch": 4.34, + "learning_rate": 9.12951386679223e-06, + "loss": 0.021, + "step": 33096 + }, + { + "epoch": 4.34, + "learning_rate": 9.12898555603093e-06, + "loss": 0.0186, + "step": 33097 + }, + { + "epoch": 4.34, + "learning_rate": 9.128457247719315e-06, + "loss": 0.0218, + "step": 33098 + }, + { + "epoch": 4.34, + "learning_rate": 9.127928941858864e-06, + "loss": 0.0249, + "step": 33099 + }, + { + "epoch": 4.34, + "learning_rate": 9.127400638451069e-06, + "loss": 0.0213, + "step": 33100 + }, + { + "epoch": 4.34, + "learning_rate": 9.126872337497414e-06, + "loss": 0.0206, + "step": 33101 + }, + { + "epoch": 4.34, + "learning_rate": 9.12634403899938e-06, + "loss": 0.0199, + "step": 33102 + }, + { + "epoch": 4.34, + "learning_rate": 9.125815742958462e-06, + "loss": 0.0209, + "step": 33103 + }, + { + "epoch": 4.34, + "learning_rate": 9.125287449376135e-06, + "loss": 0.0265, + "step": 33104 + }, + { + "epoch": 4.34, + "learning_rate": 9.124759158253895e-06, + "loss": 0.0224, + "step": 33105 + }, + { + "epoch": 4.34, + "learning_rate": 9.124230869593221e-06, + "loss": 0.0236, + "step": 33106 + }, + { + "epoch": 4.34, + "learning_rate": 9.1237025833956e-06, + "loss": 0.0205, + "step": 33107 + }, + { + "epoch": 4.34, + "learning_rate": 9.123174299662518e-06, + "loss": 0.0205, + "step": 33108 + }, + { + "epoch": 4.34, + "learning_rate": 9.122646018395466e-06, + "loss": 0.0177, + "step": 33109 + }, + { + "epoch": 4.34, + "learning_rate": 9.122117739595922e-06, + "loss": 0.0239, + "step": 33110 + }, + { + "epoch": 4.34, + "learning_rate": 9.12158946326538e-06, + "loss": 0.0171, + "step": 33111 + }, + { + "epoch": 4.34, + "learning_rate": 9.121061189405315e-06, + "loss": 0.0217, + "step": 33112 + }, + { + "epoch": 4.34, + "learning_rate": 9.120532918017221e-06, + "loss": 0.018, + "step": 33113 + }, + { + "epoch": 4.34, + "learning_rate": 9.120004649102581e-06, + "loss": 0.0225, + "step": 33114 + }, + { + "epoch": 4.34, + "learning_rate": 9.119476382662881e-06, + "loss": 0.0222, + "step": 33115 + }, + { + "epoch": 4.34, + "learning_rate": 9.118948118699609e-06, + "loss": 0.0213, + "step": 33116 + }, + { + "epoch": 4.34, + "learning_rate": 9.118419857214245e-06, + "loss": 0.0204, + "step": 33117 + }, + { + "epoch": 4.34, + "learning_rate": 9.117891598208278e-06, + "loss": 0.0261, + "step": 33118 + }, + { + "epoch": 4.34, + "learning_rate": 9.117363341683196e-06, + "loss": 0.0196, + "step": 33119 + }, + { + "epoch": 4.34, + "learning_rate": 9.116835087640483e-06, + "loss": 0.0277, + "step": 33120 + }, + { + "epoch": 4.34, + "learning_rate": 9.116306836081622e-06, + "loss": 0.0203, + "step": 33121 + }, + { + "epoch": 4.34, + "learning_rate": 9.115778587008103e-06, + "loss": 0.0206, + "step": 33122 + }, + { + "epoch": 4.34, + "learning_rate": 9.115250340421409e-06, + "loss": 0.0253, + "step": 33123 + }, + { + "epoch": 4.34, + "learning_rate": 9.114722096323028e-06, + "loss": 0.0181, + "step": 33124 + }, + { + "epoch": 4.34, + "learning_rate": 9.11419385471444e-06, + "loss": 0.0224, + "step": 33125 + }, + { + "epoch": 4.34, + "learning_rate": 9.11366561559714e-06, + "loss": 0.021, + "step": 33126 + }, + { + "epoch": 4.34, + "learning_rate": 9.113137378972604e-06, + "loss": 0.0208, + "step": 33127 + }, + { + "epoch": 4.34, + "learning_rate": 9.112609144842323e-06, + "loss": 0.0247, + "step": 33128 + }, + { + "epoch": 4.34, + "learning_rate": 9.11208091320778e-06, + "loss": 0.0232, + "step": 33129 + }, + { + "epoch": 4.34, + "learning_rate": 9.111552684070464e-06, + "loss": 0.0237, + "step": 33130 + }, + { + "epoch": 4.34, + "learning_rate": 9.111024457431858e-06, + "loss": 0.0208, + "step": 33131 + }, + { + "epoch": 4.34, + "learning_rate": 9.110496233293449e-06, + "loss": 0.02, + "step": 33132 + }, + { + "epoch": 4.34, + "learning_rate": 9.109968011656722e-06, + "loss": 0.0211, + "step": 33133 + }, + { + "epoch": 4.34, + "learning_rate": 9.109439792523165e-06, + "loss": 0.0176, + "step": 33134 + }, + { + "epoch": 4.34, + "learning_rate": 9.108911575894257e-06, + "loss": 0.0207, + "step": 33135 + }, + { + "epoch": 4.34, + "learning_rate": 9.10838336177149e-06, + "loss": 0.019, + "step": 33136 + }, + { + "epoch": 4.34, + "learning_rate": 9.107855150156348e-06, + "loss": 0.0229, + "step": 33137 + }, + { + "epoch": 4.34, + "learning_rate": 9.107326941050312e-06, + "loss": 0.0197, + "step": 33138 + }, + { + "epoch": 4.34, + "learning_rate": 9.106798734454878e-06, + "loss": 0.0238, + "step": 33139 + }, + { + "epoch": 4.34, + "learning_rate": 9.106270530371517e-06, + "loss": 0.0227, + "step": 33140 + }, + { + "epoch": 4.34, + "learning_rate": 9.105742328801725e-06, + "loss": 0.0235, + "step": 33141 + }, + { + "epoch": 4.34, + "learning_rate": 9.105214129746986e-06, + "loss": 0.0202, + "step": 33142 + }, + { + "epoch": 4.34, + "learning_rate": 9.104685933208787e-06, + "loss": 0.0223, + "step": 33143 + }, + { + "epoch": 4.34, + "learning_rate": 9.104157739188608e-06, + "loss": 0.0188, + "step": 33144 + }, + { + "epoch": 4.34, + "learning_rate": 9.10362954768794e-06, + "loss": 0.0252, + "step": 33145 + }, + { + "epoch": 4.34, + "learning_rate": 9.103101358708265e-06, + "loss": 0.0205, + "step": 33146 + }, + { + "epoch": 4.34, + "learning_rate": 9.102573172251073e-06, + "loss": 0.0217, + "step": 33147 + }, + { + "epoch": 4.34, + "learning_rate": 9.10204498831784e-06, + "loss": 0.0229, + "step": 33148 + }, + { + "epoch": 4.34, + "learning_rate": 9.101516806910063e-06, + "loss": 0.017, + "step": 33149 + }, + { + "epoch": 4.34, + "learning_rate": 9.10098862802922e-06, + "loss": 0.0205, + "step": 33150 + }, + { + "epoch": 4.34, + "learning_rate": 9.100460451676795e-06, + "loss": 0.0153, + "step": 33151 + }, + { + "epoch": 4.34, + "learning_rate": 9.09993227785428e-06, + "loss": 0.0242, + "step": 33152 + }, + { + "epoch": 4.34, + "learning_rate": 9.099404106563159e-06, + "loss": 0.0173, + "step": 33153 + }, + { + "epoch": 4.34, + "learning_rate": 9.098875937804915e-06, + "loss": 0.0258, + "step": 33154 + }, + { + "epoch": 4.34, + "learning_rate": 9.098347771581035e-06, + "loss": 0.0176, + "step": 33155 + }, + { + "epoch": 4.34, + "learning_rate": 9.097819607893003e-06, + "loss": 0.0229, + "step": 33156 + }, + { + "epoch": 4.34, + "learning_rate": 9.097291446742304e-06, + "loss": 0.0164, + "step": 33157 + }, + { + "epoch": 4.34, + "learning_rate": 9.096763288130428e-06, + "loss": 0.0203, + "step": 33158 + }, + { + "epoch": 4.34, + "learning_rate": 9.096235132058855e-06, + "loss": 0.0189, + "step": 33159 + }, + { + "epoch": 4.34, + "learning_rate": 9.095706978529076e-06, + "loss": 0.0168, + "step": 33160 + }, + { + "epoch": 4.34, + "learning_rate": 9.095178827542567e-06, + "loss": 0.0219, + "step": 33161 + }, + { + "epoch": 4.34, + "learning_rate": 9.09465067910082e-06, + "loss": 0.0205, + "step": 33162 + }, + { + "epoch": 4.34, + "learning_rate": 9.094122533205324e-06, + "loss": 0.0184, + "step": 33163 + }, + { + "epoch": 4.34, + "learning_rate": 9.093594389857559e-06, + "loss": 0.0224, + "step": 33164 + }, + { + "epoch": 4.34, + "learning_rate": 9.09306624905901e-06, + "loss": 0.0191, + "step": 33165 + }, + { + "epoch": 4.34, + "learning_rate": 9.092538110811165e-06, + "loss": 0.0204, + "step": 33166 + }, + { + "epoch": 4.34, + "learning_rate": 9.092009975115508e-06, + "loss": 0.0213, + "step": 33167 + }, + { + "epoch": 4.34, + "learning_rate": 9.091481841973527e-06, + "loss": 0.0179, + "step": 33168 + }, + { + "epoch": 4.34, + "learning_rate": 9.0909537113867e-06, + "loss": 0.0185, + "step": 33169 + }, + { + "epoch": 4.35, + "learning_rate": 9.09042558335652e-06, + "loss": 0.0227, + "step": 33170 + }, + { + "epoch": 4.35, + "learning_rate": 9.08989745788447e-06, + "loss": 0.0207, + "step": 33171 + }, + { + "epoch": 4.35, + "learning_rate": 9.089369334972034e-06, + "loss": 0.0204, + "step": 33172 + }, + { + "epoch": 4.35, + "learning_rate": 9.088841214620695e-06, + "loss": 0.024, + "step": 33173 + }, + { + "epoch": 4.35, + "learning_rate": 9.088313096831947e-06, + "loss": 0.0237, + "step": 33174 + }, + { + "epoch": 4.35, + "learning_rate": 9.087784981607268e-06, + "loss": 0.026, + "step": 33175 + }, + { + "epoch": 4.35, + "learning_rate": 9.087256868948146e-06, + "loss": 0.0256, + "step": 33176 + }, + { + "epoch": 4.35, + "learning_rate": 9.086728758856064e-06, + "loss": 0.021, + "step": 33177 + }, + { + "epoch": 4.35, + "learning_rate": 9.086200651332508e-06, + "loss": 0.0204, + "step": 33178 + }, + { + "epoch": 4.35, + "learning_rate": 9.085672546378967e-06, + "loss": 0.0189, + "step": 33179 + }, + { + "epoch": 4.35, + "learning_rate": 9.08514444399692e-06, + "loss": 0.0216, + "step": 33180 + }, + { + "epoch": 4.35, + "learning_rate": 9.08461634418786e-06, + "loss": 0.0223, + "step": 33181 + }, + { + "epoch": 4.35, + "learning_rate": 9.084088246953263e-06, + "loss": 0.0274, + "step": 33182 + }, + { + "epoch": 4.35, + "learning_rate": 9.083560152294618e-06, + "loss": 0.0214, + "step": 33183 + }, + { + "epoch": 4.35, + "learning_rate": 9.083032060213416e-06, + "loss": 0.0202, + "step": 33184 + }, + { + "epoch": 4.35, + "learning_rate": 9.082503970711135e-06, + "loss": 0.0238, + "step": 33185 + }, + { + "epoch": 4.35, + "learning_rate": 9.081975883789263e-06, + "loss": 0.0212, + "step": 33186 + }, + { + "epoch": 4.35, + "learning_rate": 9.081447799449287e-06, + "loss": 0.0207, + "step": 33187 + }, + { + "epoch": 4.35, + "learning_rate": 9.080919717692688e-06, + "loss": 0.0227, + "step": 33188 + }, + { + "epoch": 4.35, + "learning_rate": 9.080391638520955e-06, + "loss": 0.0258, + "step": 33189 + }, + { + "epoch": 4.35, + "learning_rate": 9.079863561935569e-06, + "loss": 0.0214, + "step": 33190 + }, + { + "epoch": 4.35, + "learning_rate": 9.07933548793802e-06, + "loss": 0.0191, + "step": 33191 + }, + { + "epoch": 4.35, + "learning_rate": 9.07880741652979e-06, + "loss": 0.0223, + "step": 33192 + }, + { + "epoch": 4.35, + "learning_rate": 9.078279347712365e-06, + "loss": 0.0185, + "step": 33193 + }, + { + "epoch": 4.35, + "learning_rate": 9.077751281487226e-06, + "loss": 0.0261, + "step": 33194 + }, + { + "epoch": 4.35, + "learning_rate": 9.077223217855869e-06, + "loss": 0.0264, + "step": 33195 + }, + { + "epoch": 4.35, + "learning_rate": 9.07669515681977e-06, + "loss": 0.0217, + "step": 33196 + }, + { + "epoch": 4.35, + "learning_rate": 9.076167098380418e-06, + "loss": 0.0233, + "step": 33197 + }, + { + "epoch": 4.35, + "learning_rate": 9.075639042539296e-06, + "loss": 0.022, + "step": 33198 + }, + { + "epoch": 4.35, + "learning_rate": 9.07511098929789e-06, + "loss": 0.0221, + "step": 33199 + }, + { + "epoch": 4.35, + "learning_rate": 9.074582938657686e-06, + "loss": 0.0183, + "step": 33200 + }, + { + "epoch": 4.35, + "learning_rate": 9.074054890620166e-06, + "loss": 0.0257, + "step": 33201 + }, + { + "epoch": 4.35, + "learning_rate": 9.07352684518682e-06, + "loss": 0.019, + "step": 33202 + }, + { + "epoch": 4.35, + "learning_rate": 9.072998802359127e-06, + "loss": 0.0233, + "step": 33203 + }, + { + "epoch": 4.35, + "learning_rate": 9.07247076213858e-06, + "loss": 0.0193, + "step": 33204 + }, + { + "epoch": 4.35, + "learning_rate": 9.071942724526654e-06, + "loss": 0.0183, + "step": 33205 + }, + { + "epoch": 4.35, + "learning_rate": 9.07141468952484e-06, + "loss": 0.0228, + "step": 33206 + }, + { + "epoch": 4.35, + "learning_rate": 9.070886657134627e-06, + "loss": 0.0212, + "step": 33207 + }, + { + "epoch": 4.35, + "learning_rate": 9.070358627357496e-06, + "loss": 0.0248, + "step": 33208 + }, + { + "epoch": 4.35, + "learning_rate": 9.069830600194929e-06, + "loss": 0.0235, + "step": 33209 + }, + { + "epoch": 4.35, + "learning_rate": 9.069302575648416e-06, + "loss": 0.0192, + "step": 33210 + }, + { + "epoch": 4.35, + "learning_rate": 9.068774553719437e-06, + "loss": 0.0226, + "step": 33211 + }, + { + "epoch": 4.35, + "learning_rate": 9.068246534409483e-06, + "loss": 0.0187, + "step": 33212 + }, + { + "epoch": 4.35, + "learning_rate": 9.067718517720036e-06, + "loss": 0.0234, + "step": 33213 + }, + { + "epoch": 4.35, + "learning_rate": 9.067190503652581e-06, + "loss": 0.0203, + "step": 33214 + }, + { + "epoch": 4.35, + "learning_rate": 9.066662492208601e-06, + "loss": 0.0212, + "step": 33215 + }, + { + "epoch": 4.35, + "learning_rate": 9.066134483389581e-06, + "loss": 0.0224, + "step": 33216 + }, + { + "epoch": 4.35, + "learning_rate": 9.06560647719701e-06, + "loss": 0.0266, + "step": 33217 + }, + { + "epoch": 4.35, + "learning_rate": 9.065078473632374e-06, + "loss": 0.0208, + "step": 33218 + }, + { + "epoch": 4.35, + "learning_rate": 9.06455047269715e-06, + "loss": 0.0191, + "step": 33219 + }, + { + "epoch": 4.35, + "learning_rate": 9.064022474392834e-06, + "loss": 0.0277, + "step": 33220 + }, + { + "epoch": 4.35, + "learning_rate": 9.063494478720901e-06, + "loss": 0.0216, + "step": 33221 + }, + { + "epoch": 4.35, + "learning_rate": 9.06296648568284e-06, + "loss": 0.0226, + "step": 33222 + }, + { + "epoch": 4.35, + "learning_rate": 9.062438495280138e-06, + "loss": 0.024, + "step": 33223 + }, + { + "epoch": 4.35, + "learning_rate": 9.061910507514276e-06, + "loss": 0.0212, + "step": 33224 + }, + { + "epoch": 4.35, + "learning_rate": 9.061382522386743e-06, + "loss": 0.0216, + "step": 33225 + }, + { + "epoch": 4.35, + "learning_rate": 9.060854539899018e-06, + "loss": 0.0185, + "step": 33226 + }, + { + "epoch": 4.35, + "learning_rate": 9.06032656005259e-06, + "loss": 0.0222, + "step": 33227 + }, + { + "epoch": 4.35, + "learning_rate": 9.059798582848945e-06, + "loss": 0.018, + "step": 33228 + }, + { + "epoch": 4.35, + "learning_rate": 9.059270608289567e-06, + "loss": 0.0192, + "step": 33229 + }, + { + "epoch": 4.35, + "learning_rate": 9.05874263637594e-06, + "loss": 0.0214, + "step": 33230 + }, + { + "epoch": 4.35, + "learning_rate": 9.05821466710955e-06, + "loss": 0.0191, + "step": 33231 + }, + { + "epoch": 4.35, + "learning_rate": 9.057686700491878e-06, + "loss": 0.0207, + "step": 33232 + }, + { + "epoch": 4.35, + "learning_rate": 9.057158736524416e-06, + "loss": 0.0183, + "step": 33233 + }, + { + "epoch": 4.35, + "learning_rate": 9.056630775208642e-06, + "loss": 0.0163, + "step": 33234 + }, + { + "epoch": 4.35, + "learning_rate": 9.056102816546046e-06, + "loss": 0.0263, + "step": 33235 + }, + { + "epoch": 4.35, + "learning_rate": 9.055574860538109e-06, + "loss": 0.0247, + "step": 33236 + }, + { + "epoch": 4.35, + "learning_rate": 9.055046907186316e-06, + "loss": 0.0153, + "step": 33237 + }, + { + "epoch": 4.35, + "learning_rate": 9.054518956492152e-06, + "loss": 0.0233, + "step": 33238 + }, + { + "epoch": 4.35, + "learning_rate": 9.053991008457108e-06, + "loss": 0.0212, + "step": 33239 + }, + { + "epoch": 4.35, + "learning_rate": 9.053463063082661e-06, + "loss": 0.0255, + "step": 33240 + }, + { + "epoch": 4.35, + "learning_rate": 9.0529351203703e-06, + "loss": 0.0275, + "step": 33241 + }, + { + "epoch": 4.35, + "learning_rate": 9.05240718032151e-06, + "loss": 0.0214, + "step": 33242 + }, + { + "epoch": 4.35, + "learning_rate": 9.05187924293777e-06, + "loss": 0.0213, + "step": 33243 + }, + { + "epoch": 4.35, + "learning_rate": 9.051351308220574e-06, + "loss": 0.021, + "step": 33244 + }, + { + "epoch": 4.35, + "learning_rate": 9.050823376171399e-06, + "loss": 0.0205, + "step": 33245 + }, + { + "epoch": 4.35, + "learning_rate": 9.050295446791733e-06, + "loss": 0.0208, + "step": 33246 + }, + { + "epoch": 4.36, + "learning_rate": 9.04976752008306e-06, + "loss": 0.0182, + "step": 33247 + }, + { + "epoch": 4.36, + "learning_rate": 9.049239596046868e-06, + "loss": 0.0171, + "step": 33248 + }, + { + "epoch": 4.36, + "learning_rate": 9.048711674684633e-06, + "loss": 0.0185, + "step": 33249 + }, + { + "epoch": 4.36, + "learning_rate": 9.048183755997849e-06, + "loss": 0.0234, + "step": 33250 + }, + { + "epoch": 4.36, + "learning_rate": 9.047655839987998e-06, + "loss": 0.0202, + "step": 33251 + }, + { + "epoch": 4.36, + "learning_rate": 9.047127926656566e-06, + "loss": 0.0253, + "step": 33252 + }, + { + "epoch": 4.36, + "learning_rate": 9.046600016005034e-06, + "loss": 0.0186, + "step": 33253 + }, + { + "epoch": 4.36, + "learning_rate": 9.04607210803489e-06, + "loss": 0.0274, + "step": 33254 + }, + { + "epoch": 4.36, + "learning_rate": 9.045544202747616e-06, + "loss": 0.0198, + "step": 33255 + }, + { + "epoch": 4.36, + "learning_rate": 9.0450163001447e-06, + "loss": 0.0211, + "step": 33256 + }, + { + "epoch": 4.36, + "learning_rate": 9.044488400227623e-06, + "loss": 0.0212, + "step": 33257 + }, + { + "epoch": 4.36, + "learning_rate": 9.04396050299787e-06, + "loss": 0.0187, + "step": 33258 + }, + { + "epoch": 4.36, + "learning_rate": 9.04343260845693e-06, + "loss": 0.026, + "step": 33259 + }, + { + "epoch": 4.36, + "learning_rate": 9.042904716606282e-06, + "loss": 0.0167, + "step": 33260 + }, + { + "epoch": 4.36, + "learning_rate": 9.042376827447414e-06, + "loss": 0.0235, + "step": 33261 + }, + { + "epoch": 4.36, + "learning_rate": 9.041848940981813e-06, + "loss": 0.0196, + "step": 33262 + }, + { + "epoch": 4.36, + "learning_rate": 9.041321057210958e-06, + "loss": 0.0217, + "step": 33263 + }, + { + "epoch": 4.36, + "learning_rate": 9.040793176136342e-06, + "loss": 0.0222, + "step": 33264 + }, + { + "epoch": 4.36, + "learning_rate": 9.040265297759438e-06, + "loss": 0.0205, + "step": 33265 + }, + { + "epoch": 4.36, + "learning_rate": 9.039737422081737e-06, + "loss": 0.0223, + "step": 33266 + }, + { + "epoch": 4.36, + "learning_rate": 9.039209549104727e-06, + "loss": 0.022, + "step": 33267 + }, + { + "epoch": 4.36, + "learning_rate": 9.038681678829887e-06, + "loss": 0.022, + "step": 33268 + }, + { + "epoch": 4.36, + "learning_rate": 9.038153811258705e-06, + "loss": 0.0199, + "step": 33269 + }, + { + "epoch": 4.36, + "learning_rate": 9.037625946392661e-06, + "loss": 0.0186, + "step": 33270 + }, + { + "epoch": 4.36, + "learning_rate": 9.037098084233242e-06, + "loss": 0.0183, + "step": 33271 + }, + { + "epoch": 4.36, + "learning_rate": 9.036570224781935e-06, + "loss": 0.0211, + "step": 33272 + }, + { + "epoch": 4.36, + "learning_rate": 9.036042368040226e-06, + "loss": 0.0168, + "step": 33273 + }, + { + "epoch": 4.36, + "learning_rate": 9.035514514009593e-06, + "loss": 0.0209, + "step": 33274 + }, + { + "epoch": 4.36, + "learning_rate": 9.034986662691528e-06, + "loss": 0.0189, + "step": 33275 + }, + { + "epoch": 4.36, + "learning_rate": 9.034458814087507e-06, + "loss": 0.0208, + "step": 33276 + }, + { + "epoch": 4.36, + "learning_rate": 9.033930968199024e-06, + "loss": 0.0184, + "step": 33277 + }, + { + "epoch": 4.36, + "learning_rate": 9.033403125027556e-06, + "loss": 0.0244, + "step": 33278 + }, + { + "epoch": 4.36, + "learning_rate": 9.03287528457459e-06, + "loss": 0.0259, + "step": 33279 + }, + { + "epoch": 4.36, + "learning_rate": 9.032347446841611e-06, + "loss": 0.0249, + "step": 33280 + }, + { + "epoch": 4.36, + "learning_rate": 9.031819611830103e-06, + "loss": 0.0235, + "step": 33281 + }, + { + "epoch": 4.36, + "learning_rate": 9.031291779541548e-06, + "loss": 0.0226, + "step": 33282 + }, + { + "epoch": 4.36, + "learning_rate": 9.030763949977437e-06, + "loss": 0.0209, + "step": 33283 + }, + { + "epoch": 4.36, + "learning_rate": 9.03023612313925e-06, + "loss": 0.0202, + "step": 33284 + }, + { + "epoch": 4.36, + "learning_rate": 9.029708299028473e-06, + "loss": 0.0184, + "step": 33285 + }, + { + "epoch": 4.36, + "learning_rate": 9.029180477646589e-06, + "loss": 0.0183, + "step": 33286 + }, + { + "epoch": 4.36, + "learning_rate": 9.028652658995083e-06, + "loss": 0.024, + "step": 33287 + }, + { + "epoch": 4.36, + "learning_rate": 9.028124843075441e-06, + "loss": 0.022, + "step": 33288 + }, + { + "epoch": 4.36, + "learning_rate": 9.027597029889143e-06, + "loss": 0.018, + "step": 33289 + }, + { + "epoch": 4.36, + "learning_rate": 9.02706921943768e-06, + "loss": 0.0231, + "step": 33290 + }, + { + "epoch": 4.36, + "learning_rate": 9.026541411722532e-06, + "loss": 0.0253, + "step": 33291 + }, + { + "epoch": 4.36, + "learning_rate": 9.026013606745186e-06, + "loss": 0.0219, + "step": 33292 + }, + { + "epoch": 4.36, + "learning_rate": 9.025485804507117e-06, + "loss": 0.0227, + "step": 33293 + }, + { + "epoch": 4.36, + "learning_rate": 9.024958005009825e-06, + "loss": 0.0263, + "step": 33294 + }, + { + "epoch": 4.36, + "learning_rate": 9.024430208254785e-06, + "loss": 0.0159, + "step": 33295 + }, + { + "epoch": 4.36, + "learning_rate": 9.023902414243484e-06, + "loss": 0.0213, + "step": 33296 + }, + { + "epoch": 4.36, + "learning_rate": 9.023374622977403e-06, + "loss": 0.0266, + "step": 33297 + }, + { + "epoch": 4.36, + "learning_rate": 9.022846834458032e-06, + "loss": 0.0264, + "step": 33298 + }, + { + "epoch": 4.36, + "learning_rate": 9.022319048686851e-06, + "loss": 0.0157, + "step": 33299 + }, + { + "epoch": 4.36, + "learning_rate": 9.021791265665346e-06, + "loss": 0.0211, + "step": 33300 + }, + { + "epoch": 4.36, + "learning_rate": 9.021263485395e-06, + "loss": 0.0227, + "step": 33301 + }, + { + "epoch": 4.36, + "learning_rate": 9.020735707877297e-06, + "loss": 0.0222, + "step": 33302 + }, + { + "epoch": 4.36, + "learning_rate": 9.020207933113726e-06, + "loss": 0.0221, + "step": 33303 + }, + { + "epoch": 4.36, + "learning_rate": 9.019680161105764e-06, + "loss": 0.0243, + "step": 33304 + }, + { + "epoch": 4.36, + "learning_rate": 9.019152391854899e-06, + "loss": 0.0219, + "step": 33305 + }, + { + "epoch": 4.36, + "learning_rate": 9.01862462536262e-06, + "loss": 0.0192, + "step": 33306 + }, + { + "epoch": 4.36, + "learning_rate": 9.018096861630405e-06, + "loss": 0.0204, + "step": 33307 + }, + { + "epoch": 4.36, + "learning_rate": 9.01756910065974e-06, + "loss": 0.0245, + "step": 33308 + }, + { + "epoch": 4.36, + "learning_rate": 9.017041342452111e-06, + "loss": 0.0248, + "step": 33309 + }, + { + "epoch": 4.36, + "learning_rate": 9.016513587008999e-06, + "loss": 0.0232, + "step": 33310 + }, + { + "epoch": 4.36, + "learning_rate": 9.015985834331894e-06, + "loss": 0.0203, + "step": 33311 + }, + { + "epoch": 4.36, + "learning_rate": 9.015458084422272e-06, + "loss": 0.0222, + "step": 33312 + }, + { + "epoch": 4.36, + "learning_rate": 9.014930337281624e-06, + "loss": 0.0253, + "step": 33313 + }, + { + "epoch": 4.36, + "learning_rate": 9.014402592911428e-06, + "loss": 0.0204, + "step": 33314 + }, + { + "epoch": 4.36, + "learning_rate": 9.013874851313175e-06, + "loss": 0.0203, + "step": 33315 + }, + { + "epoch": 4.36, + "learning_rate": 9.013347112488346e-06, + "loss": 0.0198, + "step": 33316 + }, + { + "epoch": 4.36, + "learning_rate": 9.012819376438428e-06, + "loss": 0.0226, + "step": 33317 + }, + { + "epoch": 4.36, + "learning_rate": 9.012291643164902e-06, + "loss": 0.0256, + "step": 33318 + }, + { + "epoch": 4.36, + "learning_rate": 9.011763912669254e-06, + "loss": 0.0216, + "step": 33319 + }, + { + "epoch": 4.36, + "learning_rate": 9.011236184952966e-06, + "loss": 0.0229, + "step": 33320 + }, + { + "epoch": 4.36, + "learning_rate": 9.010708460017524e-06, + "loss": 0.0185, + "step": 33321 + }, + { + "epoch": 4.36, + "learning_rate": 9.010180737864411e-06, + "loss": 0.0236, + "step": 33322 + }, + { + "epoch": 4.37, + "learning_rate": 9.009653018495111e-06, + "loss": 0.0192, + "step": 33323 + }, + { + "epoch": 4.37, + "learning_rate": 9.009125301911113e-06, + "loss": 0.024, + "step": 33324 + }, + { + "epoch": 4.37, + "learning_rate": 9.008597588113892e-06, + "loss": 0.0194, + "step": 33325 + }, + { + "epoch": 4.37, + "learning_rate": 9.008069877104939e-06, + "loss": 0.0217, + "step": 33326 + }, + { + "epoch": 4.37, + "learning_rate": 9.007542168885738e-06, + "loss": 0.0259, + "step": 33327 + }, + { + "epoch": 4.37, + "learning_rate": 9.007014463457772e-06, + "loss": 0.0209, + "step": 33328 + }, + { + "epoch": 4.37, + "learning_rate": 9.006486760822528e-06, + "loss": 0.0186, + "step": 33329 + }, + { + "epoch": 4.37, + "learning_rate": 9.005959060981482e-06, + "loss": 0.0181, + "step": 33330 + }, + { + "epoch": 4.37, + "learning_rate": 9.005431363936124e-06, + "loss": 0.0212, + "step": 33331 + }, + { + "epoch": 4.37, + "learning_rate": 9.00490366968794e-06, + "loss": 0.0208, + "step": 33332 + }, + { + "epoch": 4.37, + "learning_rate": 9.004375978238408e-06, + "loss": 0.022, + "step": 33333 + }, + { + "epoch": 4.37, + "learning_rate": 9.00384828958902e-06, + "loss": 0.0191, + "step": 33334 + }, + { + "epoch": 4.37, + "learning_rate": 9.003320603741253e-06, + "loss": 0.0241, + "step": 33335 + }, + { + "epoch": 4.37, + "learning_rate": 9.00279292069659e-06, + "loss": 0.0174, + "step": 33336 + }, + { + "epoch": 4.37, + "learning_rate": 9.00226524045652e-06, + "loss": 0.0172, + "step": 33337 + }, + { + "epoch": 4.37, + "learning_rate": 9.00173756302253e-06, + "loss": 0.0235, + "step": 33338 + }, + { + "epoch": 4.37, + "learning_rate": 9.001209888396098e-06, + "loss": 0.0242, + "step": 33339 + }, + { + "epoch": 4.37, + "learning_rate": 9.000682216578712e-06, + "loss": 0.0166, + "step": 33340 + }, + { + "epoch": 4.37, + "learning_rate": 9.000154547571852e-06, + "loss": 0.0243, + "step": 33341 + }, + { + "epoch": 4.37, + "learning_rate": 8.999626881377007e-06, + "loss": 0.0202, + "step": 33342 + }, + { + "epoch": 4.37, + "learning_rate": 8.999099217995654e-06, + "loss": 0.0248, + "step": 33343 + }, + { + "epoch": 4.37, + "learning_rate": 8.998571557429282e-06, + "loss": 0.0177, + "step": 33344 + }, + { + "epoch": 4.37, + "learning_rate": 8.998043899679376e-06, + "loss": 0.0219, + "step": 33345 + }, + { + "epoch": 4.37, + "learning_rate": 8.997516244747416e-06, + "loss": 0.0204, + "step": 33346 + }, + { + "epoch": 4.37, + "learning_rate": 8.996988592634886e-06, + "loss": 0.0216, + "step": 33347 + }, + { + "epoch": 4.37, + "learning_rate": 8.996460943343277e-06, + "loss": 0.0237, + "step": 33348 + }, + { + "epoch": 4.37, + "learning_rate": 8.995933296874067e-06, + "loss": 0.0209, + "step": 33349 + }, + { + "epoch": 4.37, + "learning_rate": 8.995405653228742e-06, + "loss": 0.0212, + "step": 33350 + }, + { + "epoch": 4.37, + "learning_rate": 8.994878012408783e-06, + "loss": 0.0164, + "step": 33351 + }, + { + "epoch": 4.37, + "learning_rate": 8.994350374415676e-06, + "loss": 0.0176, + "step": 33352 + }, + { + "epoch": 4.37, + "learning_rate": 8.993822739250907e-06, + "loss": 0.0253, + "step": 33353 + }, + { + "epoch": 4.37, + "learning_rate": 8.993295106915955e-06, + "loss": 0.0227, + "step": 33354 + }, + { + "epoch": 4.37, + "learning_rate": 8.99276747741231e-06, + "loss": 0.0216, + "step": 33355 + }, + { + "epoch": 4.37, + "learning_rate": 8.992239850741452e-06, + "loss": 0.0215, + "step": 33356 + }, + { + "epoch": 4.37, + "learning_rate": 8.991712226904867e-06, + "loss": 0.0195, + "step": 33357 + }, + { + "epoch": 4.37, + "learning_rate": 8.99118460590403e-06, + "loss": 0.0242, + "step": 33358 + }, + { + "epoch": 4.37, + "learning_rate": 8.99065698774044e-06, + "loss": 0.0202, + "step": 33359 + }, + { + "epoch": 4.37, + "learning_rate": 8.990129372415572e-06, + "loss": 0.0225, + "step": 33360 + }, + { + "epoch": 4.37, + "learning_rate": 8.989601759930913e-06, + "loss": 0.0237, + "step": 33361 + }, + { + "epoch": 4.37, + "learning_rate": 8.989074150287943e-06, + "loss": 0.0226, + "step": 33362 + }, + { + "epoch": 4.37, + "learning_rate": 8.98854654348815e-06, + "loss": 0.0226, + "step": 33363 + }, + { + "epoch": 4.37, + "learning_rate": 8.988018939533014e-06, + "loss": 0.0193, + "step": 33364 + }, + { + "epoch": 4.37, + "learning_rate": 8.987491338424023e-06, + "loss": 0.0171, + "step": 33365 + }, + { + "epoch": 4.37, + "learning_rate": 8.986963740162656e-06, + "loss": 0.017, + "step": 33366 + }, + { + "epoch": 4.37, + "learning_rate": 8.986436144750402e-06, + "loss": 0.0282, + "step": 33367 + }, + { + "epoch": 4.37, + "learning_rate": 8.98590855218874e-06, + "loss": 0.0244, + "step": 33368 + }, + { + "epoch": 4.37, + "learning_rate": 8.985380962479156e-06, + "loss": 0.0192, + "step": 33369 + }, + { + "epoch": 4.37, + "learning_rate": 8.984853375623134e-06, + "loss": 0.0181, + "step": 33370 + }, + { + "epoch": 4.37, + "learning_rate": 8.984325791622162e-06, + "loss": 0.0177, + "step": 33371 + }, + { + "epoch": 4.37, + "learning_rate": 8.983798210477716e-06, + "loss": 0.0242, + "step": 33372 + }, + { + "epoch": 4.37, + "learning_rate": 8.983270632191282e-06, + "loss": 0.0234, + "step": 33373 + }, + { + "epoch": 4.37, + "learning_rate": 8.982743056764349e-06, + "loss": 0.0191, + "step": 33374 + }, + { + "epoch": 4.37, + "learning_rate": 8.982215484198396e-06, + "loss": 0.0212, + "step": 33375 + }, + { + "epoch": 4.37, + "learning_rate": 8.981687914494907e-06, + "loss": 0.022, + "step": 33376 + }, + { + "epoch": 4.37, + "learning_rate": 8.981160347655366e-06, + "loss": 0.0253, + "step": 33377 + }, + { + "epoch": 4.37, + "learning_rate": 8.98063278368126e-06, + "loss": 0.0172, + "step": 33378 + }, + { + "epoch": 4.37, + "learning_rate": 8.980105222574068e-06, + "loss": 0.0185, + "step": 33379 + }, + { + "epoch": 4.37, + "learning_rate": 8.979577664335272e-06, + "loss": 0.0249, + "step": 33380 + }, + { + "epoch": 4.37, + "learning_rate": 8.979050108966362e-06, + "loss": 0.0233, + "step": 33381 + }, + { + "epoch": 4.37, + "learning_rate": 8.978522556468822e-06, + "loss": 0.0229, + "step": 33382 + }, + { + "epoch": 4.37, + "learning_rate": 8.977995006844132e-06, + "loss": 0.0212, + "step": 33383 + }, + { + "epoch": 4.37, + "learning_rate": 8.977467460093777e-06, + "loss": 0.0226, + "step": 33384 + }, + { + "epoch": 4.37, + "learning_rate": 8.97693991621924e-06, + "loss": 0.0198, + "step": 33385 + }, + { + "epoch": 4.37, + "learning_rate": 8.976412375222003e-06, + "loss": 0.0217, + "step": 33386 + }, + { + "epoch": 4.37, + "learning_rate": 8.975884837103552e-06, + "loss": 0.0271, + "step": 33387 + }, + { + "epoch": 4.37, + "learning_rate": 8.975357301865371e-06, + "loss": 0.0197, + "step": 33388 + }, + { + "epoch": 4.37, + "learning_rate": 8.974829769508945e-06, + "loss": 0.0265, + "step": 33389 + }, + { + "epoch": 4.37, + "learning_rate": 8.974302240035754e-06, + "loss": 0.0221, + "step": 33390 + }, + { + "epoch": 4.37, + "learning_rate": 8.97377471344728e-06, + "loss": 0.0223, + "step": 33391 + }, + { + "epoch": 4.37, + "learning_rate": 8.973247189745015e-06, + "loss": 0.025, + "step": 33392 + }, + { + "epoch": 4.37, + "learning_rate": 8.972719668930435e-06, + "loss": 0.018, + "step": 33393 + }, + { + "epoch": 4.37, + "learning_rate": 8.972192151005026e-06, + "loss": 0.0185, + "step": 33394 + }, + { + "epoch": 4.37, + "learning_rate": 8.971664635970274e-06, + "loss": 0.0189, + "step": 33395 + }, + { + "epoch": 4.37, + "learning_rate": 8.97113712382766e-06, + "loss": 0.0251, + "step": 33396 + }, + { + "epoch": 4.37, + "learning_rate": 8.970609614578668e-06, + "loss": 0.0223, + "step": 33397 + }, + { + "epoch": 4.37, + "learning_rate": 8.97008210822478e-06, + "loss": 0.0182, + "step": 33398 + }, + { + "epoch": 4.38, + "learning_rate": 8.969554604767484e-06, + "loss": 0.0218, + "step": 33399 + }, + { + "epoch": 4.38, + "learning_rate": 8.969027104208258e-06, + "loss": 0.0216, + "step": 33400 + }, + { + "epoch": 4.38, + "learning_rate": 8.96849960654859e-06, + "loss": 0.0191, + "step": 33401 + }, + { + "epoch": 4.38, + "learning_rate": 8.967972111789958e-06, + "loss": 0.0192, + "step": 33402 + }, + { + "epoch": 4.38, + "learning_rate": 8.967444619933853e-06, + "loss": 0.0235, + "step": 33403 + }, + { + "epoch": 4.38, + "learning_rate": 8.966917130981755e-06, + "loss": 0.0228, + "step": 33404 + }, + { + "epoch": 4.38, + "learning_rate": 8.96638964493515e-06, + "loss": 0.0234, + "step": 33405 + }, + { + "epoch": 4.38, + "learning_rate": 8.965862161795515e-06, + "loss": 0.0224, + "step": 33406 + }, + { + "epoch": 4.38, + "learning_rate": 8.965334681564341e-06, + "loss": 0.0213, + "step": 33407 + }, + { + "epoch": 4.38, + "learning_rate": 8.964807204243105e-06, + "loss": 0.0204, + "step": 33408 + }, + { + "epoch": 4.38, + "learning_rate": 8.964279729833294e-06, + "loss": 0.024, + "step": 33409 + }, + { + "epoch": 4.38, + "learning_rate": 8.963752258336393e-06, + "loss": 0.0225, + "step": 33410 + }, + { + "epoch": 4.38, + "learning_rate": 8.963224789753882e-06, + "loss": 0.0214, + "step": 33411 + }, + { + "epoch": 4.38, + "learning_rate": 8.962697324087248e-06, + "loss": 0.0173, + "step": 33412 + }, + { + "epoch": 4.38, + "learning_rate": 8.962169861337969e-06, + "loss": 0.0195, + "step": 33413 + }, + { + "epoch": 4.38, + "learning_rate": 8.961642401507533e-06, + "loss": 0.023, + "step": 33414 + }, + { + "epoch": 4.38, + "learning_rate": 8.961114944597426e-06, + "loss": 0.0211, + "step": 33415 + }, + { + "epoch": 4.38, + "learning_rate": 8.960587490609125e-06, + "loss": 0.0214, + "step": 33416 + }, + { + "epoch": 4.38, + "learning_rate": 8.960060039544116e-06, + "loss": 0.025, + "step": 33417 + }, + { + "epoch": 4.38, + "learning_rate": 8.959532591403883e-06, + "loss": 0.0232, + "step": 33418 + }, + { + "epoch": 4.38, + "learning_rate": 8.95900514618991e-06, + "loss": 0.0154, + "step": 33419 + }, + { + "epoch": 4.38, + "learning_rate": 8.95847770390368e-06, + "loss": 0.0247, + "step": 33420 + }, + { + "epoch": 4.38, + "learning_rate": 8.957950264546675e-06, + "loss": 0.0274, + "step": 33421 + }, + { + "epoch": 4.38, + "learning_rate": 8.95742282812038e-06, + "loss": 0.0274, + "step": 33422 + }, + { + "epoch": 4.38, + "learning_rate": 8.956895394626279e-06, + "loss": 0.0203, + "step": 33423 + }, + { + "epoch": 4.38, + "learning_rate": 8.956367964065849e-06, + "loss": 0.0244, + "step": 33424 + }, + { + "epoch": 4.38, + "learning_rate": 8.955840536440581e-06, + "loss": 0.0245, + "step": 33425 + }, + { + "epoch": 4.38, + "learning_rate": 8.955313111751958e-06, + "loss": 0.0234, + "step": 33426 + }, + { + "epoch": 4.38, + "learning_rate": 8.95478569000146e-06, + "loss": 0.0211, + "step": 33427 + }, + { + "epoch": 4.38, + "learning_rate": 8.954258271190575e-06, + "loss": 0.0216, + "step": 33428 + }, + { + "epoch": 4.38, + "learning_rate": 8.953730855320777e-06, + "loss": 0.0212, + "step": 33429 + }, + { + "epoch": 4.38, + "learning_rate": 8.95320344239356e-06, + "loss": 0.0221, + "step": 33430 + }, + { + "epoch": 4.38, + "learning_rate": 8.9526760324104e-06, + "loss": 0.0213, + "step": 33431 + }, + { + "epoch": 4.38, + "learning_rate": 8.952148625372783e-06, + "loss": 0.0168, + "step": 33432 + }, + { + "epoch": 4.38, + "learning_rate": 8.951621221282194e-06, + "loss": 0.0173, + "step": 33433 + }, + { + "epoch": 4.38, + "learning_rate": 8.951093820140113e-06, + "loss": 0.022, + "step": 33434 + }, + { + "epoch": 4.38, + "learning_rate": 8.950566421948022e-06, + "loss": 0.0169, + "step": 33435 + }, + { + "epoch": 4.38, + "learning_rate": 8.95003902670741e-06, + "loss": 0.0189, + "step": 33436 + }, + { + "epoch": 4.38, + "learning_rate": 8.949511634419759e-06, + "loss": 0.0219, + "step": 33437 + }, + { + "epoch": 4.38, + "learning_rate": 8.948984245086547e-06, + "loss": 0.0201, + "step": 33438 + }, + { + "epoch": 4.38, + "learning_rate": 8.948456858709265e-06, + "loss": 0.0197, + "step": 33439 + }, + { + "epoch": 4.38, + "learning_rate": 8.94792947528939e-06, + "loss": 0.0252, + "step": 33440 + }, + { + "epoch": 4.38, + "learning_rate": 8.947402094828409e-06, + "loss": 0.0235, + "step": 33441 + }, + { + "epoch": 4.38, + "learning_rate": 8.9468747173278e-06, + "loss": 0.0216, + "step": 33442 + }, + { + "epoch": 4.38, + "learning_rate": 8.946347342789053e-06, + "loss": 0.0207, + "step": 33443 + }, + { + "epoch": 4.38, + "learning_rate": 8.945819971213644e-06, + "loss": 0.024, + "step": 33444 + }, + { + "epoch": 4.38, + "learning_rate": 8.945292602603061e-06, + "loss": 0.0197, + "step": 33445 + }, + { + "epoch": 4.38, + "learning_rate": 8.944765236958787e-06, + "loss": 0.0233, + "step": 33446 + }, + { + "epoch": 4.38, + "learning_rate": 8.944237874282308e-06, + "loss": 0.0221, + "step": 33447 + }, + { + "epoch": 4.38, + "learning_rate": 8.9437105145751e-06, + "loss": 0.0252, + "step": 33448 + }, + { + "epoch": 4.38, + "learning_rate": 8.943183157838653e-06, + "loss": 0.0242, + "step": 33449 + }, + { + "epoch": 4.38, + "learning_rate": 8.942655804074445e-06, + "loss": 0.0201, + "step": 33450 + }, + { + "epoch": 4.38, + "learning_rate": 8.942128453283963e-06, + "loss": 0.0224, + "step": 33451 + }, + { + "epoch": 4.38, + "learning_rate": 8.941601105468686e-06, + "loss": 0.0233, + "step": 33452 + }, + { + "epoch": 4.38, + "learning_rate": 8.9410737606301e-06, + "loss": 0.021, + "step": 33453 + }, + { + "epoch": 4.38, + "learning_rate": 8.94054641876969e-06, + "loss": 0.021, + "step": 33454 + }, + { + "epoch": 4.38, + "learning_rate": 8.940019079888933e-06, + "loss": 0.0202, + "step": 33455 + }, + { + "epoch": 4.38, + "learning_rate": 8.939491743989316e-06, + "loss": 0.0231, + "step": 33456 + }, + { + "epoch": 4.38, + "learning_rate": 8.938964411072325e-06, + "loss": 0.0192, + "step": 33457 + }, + { + "epoch": 4.38, + "learning_rate": 8.938437081139438e-06, + "loss": 0.0241, + "step": 33458 + }, + { + "epoch": 4.38, + "learning_rate": 8.937909754192142e-06, + "loss": 0.0209, + "step": 33459 + }, + { + "epoch": 4.38, + "learning_rate": 8.937382430231918e-06, + "loss": 0.0248, + "step": 33460 + }, + { + "epoch": 4.38, + "learning_rate": 8.936855109260248e-06, + "loss": 0.0173, + "step": 33461 + }, + { + "epoch": 4.38, + "learning_rate": 8.936327791278618e-06, + "loss": 0.0273, + "step": 33462 + }, + { + "epoch": 4.38, + "learning_rate": 8.935800476288508e-06, + "loss": 0.0189, + "step": 33463 + }, + { + "epoch": 4.38, + "learning_rate": 8.935273164291405e-06, + "loss": 0.021, + "step": 33464 + }, + { + "epoch": 4.38, + "learning_rate": 8.934745855288787e-06, + "loss": 0.0193, + "step": 33465 + }, + { + "epoch": 4.38, + "learning_rate": 8.934218549282143e-06, + "loss": 0.0229, + "step": 33466 + }, + { + "epoch": 4.38, + "learning_rate": 8.933691246272943e-06, + "loss": 0.0202, + "step": 33467 + }, + { + "epoch": 4.38, + "learning_rate": 8.93316394626269e-06, + "loss": 0.0175, + "step": 33468 + }, + { + "epoch": 4.38, + "learning_rate": 8.932636649252851e-06, + "loss": 0.0183, + "step": 33469 + }, + { + "epoch": 4.38, + "learning_rate": 8.932109355244918e-06, + "loss": 0.0233, + "step": 33470 + }, + { + "epoch": 4.38, + "learning_rate": 8.931582064240368e-06, + "loss": 0.0251, + "step": 33471 + }, + { + "epoch": 4.38, + "learning_rate": 8.931054776240689e-06, + "loss": 0.0186, + "step": 33472 + }, + { + "epoch": 4.38, + "learning_rate": 8.93052749124736e-06, + "loss": 0.0199, + "step": 33473 + }, + { + "epoch": 4.38, + "learning_rate": 8.930000209261864e-06, + "loss": 0.023, + "step": 33474 + }, + { + "epoch": 4.38, + "learning_rate": 8.92947293028569e-06, + "loss": 0.0213, + "step": 33475 + }, + { + "epoch": 4.39, + "learning_rate": 8.92894565432031e-06, + "loss": 0.0214, + "step": 33476 + }, + { + "epoch": 4.39, + "learning_rate": 8.92841838136722e-06, + "loss": 0.0173, + "step": 33477 + }, + { + "epoch": 4.39, + "learning_rate": 8.92789111142789e-06, + "loss": 0.0176, + "step": 33478 + }, + { + "epoch": 4.39, + "learning_rate": 8.927363844503811e-06, + "loss": 0.0205, + "step": 33479 + }, + { + "epoch": 4.39, + "learning_rate": 8.926836580596466e-06, + "loss": 0.0257, + "step": 33480 + }, + { + "epoch": 4.39, + "learning_rate": 8.926309319707335e-06, + "loss": 0.0189, + "step": 33481 + }, + { + "epoch": 4.39, + "learning_rate": 8.9257820618379e-06, + "loss": 0.0283, + "step": 33482 + }, + { + "epoch": 4.39, + "learning_rate": 8.925254806989651e-06, + "loss": 0.0173, + "step": 33483 + }, + { + "epoch": 4.39, + "learning_rate": 8.92472755516406e-06, + "loss": 0.0219, + "step": 33484 + }, + { + "epoch": 4.39, + "learning_rate": 8.92420030636262e-06, + "loss": 0.023, + "step": 33485 + }, + { + "epoch": 4.39, + "learning_rate": 8.923673060586807e-06, + "loss": 0.0199, + "step": 33486 + }, + { + "epoch": 4.39, + "learning_rate": 8.923145817838108e-06, + "loss": 0.0199, + "step": 33487 + }, + { + "epoch": 4.39, + "learning_rate": 8.922618578118002e-06, + "loss": 0.0189, + "step": 33488 + }, + { + "epoch": 4.39, + "learning_rate": 8.92209134142797e-06, + "loss": 0.0278, + "step": 33489 + }, + { + "epoch": 4.39, + "learning_rate": 8.921564107769504e-06, + "loss": 0.0238, + "step": 33490 + }, + { + "epoch": 4.39, + "learning_rate": 8.92103687714408e-06, + "loss": 0.0199, + "step": 33491 + }, + { + "epoch": 4.39, + "learning_rate": 8.920509649553184e-06, + "loss": 0.0188, + "step": 33492 + }, + { + "epoch": 4.39, + "learning_rate": 8.919982424998297e-06, + "loss": 0.0221, + "step": 33493 + }, + { + "epoch": 4.39, + "learning_rate": 8.9194552034809e-06, + "loss": 0.023, + "step": 33494 + }, + { + "epoch": 4.39, + "learning_rate": 8.918927985002477e-06, + "loss": 0.0174, + "step": 33495 + }, + { + "epoch": 4.39, + "learning_rate": 8.918400769564516e-06, + "loss": 0.0177, + "step": 33496 + }, + { + "epoch": 4.39, + "learning_rate": 8.917873557168492e-06, + "loss": 0.0233, + "step": 33497 + }, + { + "epoch": 4.39, + "learning_rate": 8.917346347815892e-06, + "loss": 0.0222, + "step": 33498 + }, + { + "epoch": 4.39, + "learning_rate": 8.916819141508197e-06, + "loss": 0.0208, + "step": 33499 + }, + { + "epoch": 4.39, + "learning_rate": 8.916291938246888e-06, + "loss": 0.0178, + "step": 33500 + }, + { + "epoch": 4.39, + "learning_rate": 8.915764738033453e-06, + "loss": 0.0219, + "step": 33501 + }, + { + "epoch": 4.39, + "learning_rate": 8.915237540869372e-06, + "loss": 0.0208, + "step": 33502 + }, + { + "epoch": 4.39, + "learning_rate": 8.914710346756126e-06, + "loss": 0.0242, + "step": 33503 + }, + { + "epoch": 4.39, + "learning_rate": 8.914183155695202e-06, + "loss": 0.0184, + "step": 33504 + }, + { + "epoch": 4.39, + "learning_rate": 8.913655967688079e-06, + "loss": 0.0189, + "step": 33505 + }, + { + "epoch": 4.39, + "learning_rate": 8.913128782736241e-06, + "loss": 0.0234, + "step": 33506 + }, + { + "epoch": 4.39, + "learning_rate": 8.912601600841169e-06, + "loss": 0.0236, + "step": 33507 + }, + { + "epoch": 4.39, + "learning_rate": 8.91207442200435e-06, + "loss": 0.02, + "step": 33508 + }, + { + "epoch": 4.39, + "learning_rate": 8.911547246227261e-06, + "loss": 0.0254, + "step": 33509 + }, + { + "epoch": 4.39, + "learning_rate": 8.911020073511387e-06, + "loss": 0.0165, + "step": 33510 + }, + { + "epoch": 4.39, + "learning_rate": 8.91049290385821e-06, + "loss": 0.0159, + "step": 33511 + }, + { + "epoch": 4.39, + "learning_rate": 8.909965737269218e-06, + "loss": 0.0161, + "step": 33512 + }, + { + "epoch": 4.39, + "learning_rate": 8.909438573745885e-06, + "loss": 0.0267, + "step": 33513 + }, + { + "epoch": 4.39, + "learning_rate": 8.908911413289703e-06, + "loss": 0.0229, + "step": 33514 + }, + { + "epoch": 4.39, + "learning_rate": 8.908384255902145e-06, + "loss": 0.0181, + "step": 33515 + }, + { + "epoch": 4.39, + "learning_rate": 8.907857101584701e-06, + "loss": 0.019, + "step": 33516 + }, + { + "epoch": 4.39, + "learning_rate": 8.907329950338849e-06, + "loss": 0.0201, + "step": 33517 + }, + { + "epoch": 4.39, + "learning_rate": 8.906802802166074e-06, + "loss": 0.021, + "step": 33518 + }, + { + "epoch": 4.39, + "learning_rate": 8.906275657067858e-06, + "loss": 0.0207, + "step": 33519 + }, + { + "epoch": 4.39, + "learning_rate": 8.905748515045683e-06, + "loss": 0.0236, + "step": 33520 + }, + { + "epoch": 4.39, + "learning_rate": 8.905221376101034e-06, + "loss": 0.0207, + "step": 33521 + }, + { + "epoch": 4.39, + "learning_rate": 8.904694240235386e-06, + "loss": 0.0182, + "step": 33522 + }, + { + "epoch": 4.39, + "learning_rate": 8.90416710745023e-06, + "loss": 0.0227, + "step": 33523 + }, + { + "epoch": 4.39, + "learning_rate": 8.903639977747047e-06, + "loss": 0.022, + "step": 33524 + }, + { + "epoch": 4.39, + "learning_rate": 8.903112851127319e-06, + "loss": 0.0242, + "step": 33525 + }, + { + "epoch": 4.39, + "learning_rate": 8.902585727592527e-06, + "loss": 0.0262, + "step": 33526 + }, + { + "epoch": 4.39, + "learning_rate": 8.902058607144156e-06, + "loss": 0.0225, + "step": 33527 + }, + { + "epoch": 4.39, + "learning_rate": 8.901531489783682e-06, + "loss": 0.0199, + "step": 33528 + }, + { + "epoch": 4.39, + "learning_rate": 8.901004375512597e-06, + "loss": 0.0196, + "step": 33529 + }, + { + "epoch": 4.39, + "learning_rate": 8.900477264332375e-06, + "loss": 0.0182, + "step": 33530 + }, + { + "epoch": 4.39, + "learning_rate": 8.899950156244505e-06, + "loss": 0.0219, + "step": 33531 + }, + { + "epoch": 4.39, + "learning_rate": 8.899423051250465e-06, + "loss": 0.0198, + "step": 33532 + }, + { + "epoch": 4.39, + "learning_rate": 8.898895949351736e-06, + "loss": 0.0284, + "step": 33533 + }, + { + "epoch": 4.39, + "learning_rate": 8.898368850549806e-06, + "loss": 0.0177, + "step": 33534 + }, + { + "epoch": 4.39, + "learning_rate": 8.897841754846156e-06, + "loss": 0.0198, + "step": 33535 + }, + { + "epoch": 4.39, + "learning_rate": 8.897314662242266e-06, + "loss": 0.0185, + "step": 33536 + }, + { + "epoch": 4.39, + "learning_rate": 8.896787572739622e-06, + "loss": 0.0238, + "step": 33537 + }, + { + "epoch": 4.39, + "learning_rate": 8.896260486339703e-06, + "loss": 0.0196, + "step": 33538 + }, + { + "epoch": 4.39, + "learning_rate": 8.895733403043991e-06, + "loss": 0.0216, + "step": 33539 + }, + { + "epoch": 4.39, + "learning_rate": 8.895206322853973e-06, + "loss": 0.0162, + "step": 33540 + }, + { + "epoch": 4.39, + "learning_rate": 8.894679245771127e-06, + "loss": 0.0173, + "step": 33541 + }, + { + "epoch": 4.39, + "learning_rate": 8.894152171796937e-06, + "loss": 0.0206, + "step": 33542 + }, + { + "epoch": 4.39, + "learning_rate": 8.893625100932884e-06, + "loss": 0.0166, + "step": 33543 + }, + { + "epoch": 4.39, + "learning_rate": 8.89309803318045e-06, + "loss": 0.0197, + "step": 33544 + }, + { + "epoch": 4.39, + "learning_rate": 8.892570968541121e-06, + "loss": 0.0163, + "step": 33545 + }, + { + "epoch": 4.39, + "learning_rate": 8.892043907016378e-06, + "loss": 0.0195, + "step": 33546 + }, + { + "epoch": 4.39, + "learning_rate": 8.891516848607702e-06, + "loss": 0.0215, + "step": 33547 + }, + { + "epoch": 4.39, + "learning_rate": 8.890989793316578e-06, + "loss": 0.0209, + "step": 33548 + }, + { + "epoch": 4.39, + "learning_rate": 8.890462741144481e-06, + "loss": 0.0209, + "step": 33549 + }, + { + "epoch": 4.39, + "learning_rate": 8.889935692092903e-06, + "loss": 0.0202, + "step": 33550 + }, + { + "epoch": 4.39, + "learning_rate": 8.88940864616332e-06, + "loss": 0.0229, + "step": 33551 + }, + { + "epoch": 4.4, + "learning_rate": 8.888881603357217e-06, + "loss": 0.0207, + "step": 33552 + }, + { + "epoch": 4.4, + "learning_rate": 8.888354563676075e-06, + "loss": 0.0248, + "step": 33553 + }, + { + "epoch": 4.4, + "learning_rate": 8.887827527121373e-06, + "loss": 0.0186, + "step": 33554 + }, + { + "epoch": 4.4, + "learning_rate": 8.887300493694597e-06, + "loss": 0.0245, + "step": 33555 + }, + { + "epoch": 4.4, + "learning_rate": 8.886773463397233e-06, + "loss": 0.0224, + "step": 33556 + }, + { + "epoch": 4.4, + "learning_rate": 8.886246436230759e-06, + "loss": 0.0211, + "step": 33557 + }, + { + "epoch": 4.4, + "learning_rate": 8.885719412196658e-06, + "loss": 0.023, + "step": 33558 + }, + { + "epoch": 4.4, + "learning_rate": 8.88519239129641e-06, + "loss": 0.0203, + "step": 33559 + }, + { + "epoch": 4.4, + "learning_rate": 8.8846653735315e-06, + "loss": 0.0205, + "step": 33560 + }, + { + "epoch": 4.4, + "learning_rate": 8.884138358903409e-06, + "loss": 0.0245, + "step": 33561 + }, + { + "epoch": 4.4, + "learning_rate": 8.883611347413619e-06, + "loss": 0.019, + "step": 33562 + }, + { + "epoch": 4.4, + "learning_rate": 8.883084339063615e-06, + "loss": 0.0235, + "step": 33563 + }, + { + "epoch": 4.4, + "learning_rate": 8.882557333854873e-06, + "loss": 0.0195, + "step": 33564 + }, + { + "epoch": 4.4, + "learning_rate": 8.882030331788883e-06, + "loss": 0.022, + "step": 33565 + }, + { + "epoch": 4.4, + "learning_rate": 8.881503332867117e-06, + "loss": 0.0173, + "step": 33566 + }, + { + "epoch": 4.4, + "learning_rate": 8.880976337091067e-06, + "loss": 0.0221, + "step": 33567 + }, + { + "epoch": 4.4, + "learning_rate": 8.88044934446221e-06, + "loss": 0.0161, + "step": 33568 + }, + { + "epoch": 4.4, + "learning_rate": 8.879922354982033e-06, + "loss": 0.0258, + "step": 33569 + }, + { + "epoch": 4.4, + "learning_rate": 8.879395368652013e-06, + "loss": 0.0184, + "step": 33570 + }, + { + "epoch": 4.4, + "learning_rate": 8.878868385473635e-06, + "loss": 0.0203, + "step": 33571 + }, + { + "epoch": 4.4, + "learning_rate": 8.878341405448378e-06, + "loss": 0.0202, + "step": 33572 + }, + { + "epoch": 4.4, + "learning_rate": 8.87781442857773e-06, + "loss": 0.0224, + "step": 33573 + }, + { + "epoch": 4.4, + "learning_rate": 8.877287454863166e-06, + "loss": 0.0231, + "step": 33574 + }, + { + "epoch": 4.4, + "learning_rate": 8.87676048430617e-06, + "loss": 0.0222, + "step": 33575 + }, + { + "epoch": 4.4, + "learning_rate": 8.876233516908222e-06, + "loss": 0.0163, + "step": 33576 + }, + { + "epoch": 4.4, + "learning_rate": 8.875706552670814e-06, + "loss": 0.0194, + "step": 33577 + }, + { + "epoch": 4.4, + "learning_rate": 8.87517959159542e-06, + "loss": 0.0167, + "step": 33578 + }, + { + "epoch": 4.4, + "learning_rate": 8.874652633683524e-06, + "loss": 0.0164, + "step": 33579 + }, + { + "epoch": 4.4, + "learning_rate": 8.874125678936606e-06, + "loss": 0.025, + "step": 33580 + }, + { + "epoch": 4.4, + "learning_rate": 8.87359872735615e-06, + "loss": 0.0177, + "step": 33581 + }, + { + "epoch": 4.4, + "learning_rate": 8.873071778943638e-06, + "loss": 0.0173, + "step": 33582 + }, + { + "epoch": 4.4, + "learning_rate": 8.872544833700551e-06, + "loss": 0.0211, + "step": 33583 + }, + { + "epoch": 4.4, + "learning_rate": 8.872017891628372e-06, + "loss": 0.0229, + "step": 33584 + }, + { + "epoch": 4.4, + "learning_rate": 8.871490952728584e-06, + "loss": 0.0252, + "step": 33585 + }, + { + "epoch": 4.4, + "learning_rate": 8.870964017002667e-06, + "loss": 0.0237, + "step": 33586 + }, + { + "epoch": 4.4, + "learning_rate": 8.870437084452099e-06, + "loss": 0.0258, + "step": 33587 + }, + { + "epoch": 4.4, + "learning_rate": 8.86991015507837e-06, + "loss": 0.0205, + "step": 33588 + }, + { + "epoch": 4.4, + "learning_rate": 8.869383228882958e-06, + "loss": 0.022, + "step": 33589 + }, + { + "epoch": 4.4, + "learning_rate": 8.868856305867348e-06, + "loss": 0.0155, + "step": 33590 + }, + { + "epoch": 4.4, + "learning_rate": 8.868329386033016e-06, + "loss": 0.0222, + "step": 33591 + }, + { + "epoch": 4.4, + "learning_rate": 8.867802469381452e-06, + "loss": 0.0229, + "step": 33592 + }, + { + "epoch": 4.4, + "learning_rate": 8.867275555914129e-06, + "loss": 0.0163, + "step": 33593 + }, + { + "epoch": 4.4, + "learning_rate": 8.866748645632537e-06, + "loss": 0.015, + "step": 33594 + }, + { + "epoch": 4.4, + "learning_rate": 8.86622173853815e-06, + "loss": 0.0204, + "step": 33595 + }, + { + "epoch": 4.4, + "learning_rate": 8.865694834632457e-06, + "loss": 0.0217, + "step": 33596 + }, + { + "epoch": 4.4, + "learning_rate": 8.865167933916934e-06, + "loss": 0.023, + "step": 33597 + }, + { + "epoch": 4.4, + "learning_rate": 8.864641036393064e-06, + "loss": 0.026, + "step": 33598 + }, + { + "epoch": 4.4, + "learning_rate": 8.864114142062333e-06, + "loss": 0.0255, + "step": 33599 + }, + { + "epoch": 4.4, + "learning_rate": 8.863587250926221e-06, + "loss": 0.0223, + "step": 33600 + }, + { + "epoch": 4.4, + "learning_rate": 8.86306036298621e-06, + "loss": 0.0289, + "step": 33601 + }, + { + "epoch": 4.4, + "learning_rate": 8.862533478243783e-06, + "loss": 0.0207, + "step": 33602 + }, + { + "epoch": 4.4, + "learning_rate": 8.862006596700417e-06, + "loss": 0.0221, + "step": 33603 + }, + { + "epoch": 4.4, + "learning_rate": 8.861479718357596e-06, + "loss": 0.0191, + "step": 33604 + }, + { + "epoch": 4.4, + "learning_rate": 8.860952843216805e-06, + "loss": 0.0209, + "step": 33605 + }, + { + "epoch": 4.4, + "learning_rate": 8.860425971279524e-06, + "loss": 0.0206, + "step": 33606 + }, + { + "epoch": 4.4, + "learning_rate": 8.859899102547233e-06, + "loss": 0.0211, + "step": 33607 + }, + { + "epoch": 4.4, + "learning_rate": 8.859372237021415e-06, + "loss": 0.0152, + "step": 33608 + }, + { + "epoch": 4.4, + "learning_rate": 8.85884537470355e-06, + "loss": 0.0191, + "step": 33609 + }, + { + "epoch": 4.4, + "learning_rate": 8.858318515595122e-06, + "loss": 0.0227, + "step": 33610 + }, + { + "epoch": 4.4, + "learning_rate": 8.857791659697615e-06, + "loss": 0.0239, + "step": 33611 + }, + { + "epoch": 4.4, + "learning_rate": 8.857264807012506e-06, + "loss": 0.0223, + "step": 33612 + }, + { + "epoch": 4.4, + "learning_rate": 8.856737957541281e-06, + "loss": 0.0212, + "step": 33613 + }, + { + "epoch": 4.4, + "learning_rate": 8.856211111285418e-06, + "loss": 0.0197, + "step": 33614 + }, + { + "epoch": 4.4, + "learning_rate": 8.855684268246403e-06, + "loss": 0.0204, + "step": 33615 + }, + { + "epoch": 4.4, + "learning_rate": 8.855157428425712e-06, + "loss": 0.0179, + "step": 33616 + }, + { + "epoch": 4.4, + "learning_rate": 8.854630591824833e-06, + "loss": 0.0208, + "step": 33617 + }, + { + "epoch": 4.4, + "learning_rate": 8.854103758445241e-06, + "loss": 0.0213, + "step": 33618 + }, + { + "epoch": 4.4, + "learning_rate": 8.85357692828842e-06, + "loss": 0.0223, + "step": 33619 + }, + { + "epoch": 4.4, + "learning_rate": 8.853050101355854e-06, + "loss": 0.0246, + "step": 33620 + }, + { + "epoch": 4.4, + "learning_rate": 8.852523277649026e-06, + "loss": 0.0181, + "step": 33621 + }, + { + "epoch": 4.4, + "learning_rate": 8.851996457169413e-06, + "loss": 0.0231, + "step": 33622 + }, + { + "epoch": 4.4, + "learning_rate": 8.851469639918501e-06, + "loss": 0.0175, + "step": 33623 + }, + { + "epoch": 4.4, + "learning_rate": 8.850942825897768e-06, + "loss": 0.0225, + "step": 33624 + }, + { + "epoch": 4.4, + "learning_rate": 8.850416015108696e-06, + "loss": 0.0228, + "step": 33625 + }, + { + "epoch": 4.4, + "learning_rate": 8.84988920755277e-06, + "loss": 0.0265, + "step": 33626 + }, + { + "epoch": 4.4, + "learning_rate": 8.849362403231469e-06, + "loss": 0.0171, + "step": 33627 + }, + { + "epoch": 4.41, + "learning_rate": 8.848835602146276e-06, + "loss": 0.0214, + "step": 33628 + }, + { + "epoch": 4.41, + "learning_rate": 8.848308804298668e-06, + "loss": 0.021, + "step": 33629 + }, + { + "epoch": 4.41, + "learning_rate": 8.847782009690134e-06, + "loss": 0.0242, + "step": 33630 + }, + { + "epoch": 4.41, + "learning_rate": 8.847255218322147e-06, + "loss": 0.022, + "step": 33631 + }, + { + "epoch": 4.41, + "learning_rate": 8.846728430196194e-06, + "loss": 0.025, + "step": 33632 + }, + { + "epoch": 4.41, + "learning_rate": 8.846201645313758e-06, + "loss": 0.0207, + "step": 33633 + }, + { + "epoch": 4.41, + "learning_rate": 8.845674863676321e-06, + "loss": 0.0208, + "step": 33634 + }, + { + "epoch": 4.41, + "learning_rate": 8.845148085285358e-06, + "loss": 0.0214, + "step": 33635 + }, + { + "epoch": 4.41, + "learning_rate": 8.844621310142356e-06, + "loss": 0.0214, + "step": 33636 + }, + { + "epoch": 4.41, + "learning_rate": 8.844094538248795e-06, + "loss": 0.0228, + "step": 33637 + }, + { + "epoch": 4.41, + "learning_rate": 8.843567769606158e-06, + "loss": 0.0214, + "step": 33638 + }, + { + "epoch": 4.41, + "learning_rate": 8.843041004215922e-06, + "loss": 0.0193, + "step": 33639 + }, + { + "epoch": 4.41, + "learning_rate": 8.842514242079572e-06, + "loss": 0.023, + "step": 33640 + }, + { + "epoch": 4.41, + "learning_rate": 8.841987483198593e-06, + "loss": 0.0205, + "step": 33641 + }, + { + "epoch": 4.41, + "learning_rate": 8.841460727574456e-06, + "loss": 0.0249, + "step": 33642 + }, + { + "epoch": 4.41, + "learning_rate": 8.840933975208652e-06, + "loss": 0.022, + "step": 33643 + }, + { + "epoch": 4.41, + "learning_rate": 8.840407226102662e-06, + "loss": 0.0195, + "step": 33644 + }, + { + "epoch": 4.41, + "learning_rate": 8.839880480257962e-06, + "loss": 0.0186, + "step": 33645 + }, + { + "epoch": 4.41, + "learning_rate": 8.839353737676039e-06, + "loss": 0.0247, + "step": 33646 + }, + { + "epoch": 4.41, + "learning_rate": 8.83882699835837e-06, + "loss": 0.02, + "step": 33647 + }, + { + "epoch": 4.41, + "learning_rate": 8.838300262306437e-06, + "loss": 0.0221, + "step": 33648 + }, + { + "epoch": 4.41, + "learning_rate": 8.837773529521725e-06, + "loss": 0.021, + "step": 33649 + }, + { + "epoch": 4.41, + "learning_rate": 8.837246800005712e-06, + "loss": 0.0259, + "step": 33650 + }, + { + "epoch": 4.41, + "learning_rate": 8.836720073759882e-06, + "loss": 0.0253, + "step": 33651 + }, + { + "epoch": 4.41, + "learning_rate": 8.836193350785713e-06, + "loss": 0.0178, + "step": 33652 + }, + { + "epoch": 4.41, + "learning_rate": 8.835666631084687e-06, + "loss": 0.0278, + "step": 33653 + }, + { + "epoch": 4.41, + "learning_rate": 8.835139914658287e-06, + "loss": 0.0176, + "step": 33654 + }, + { + "epoch": 4.41, + "learning_rate": 8.834613201507997e-06, + "loss": 0.021, + "step": 33655 + }, + { + "epoch": 4.41, + "learning_rate": 8.834086491635293e-06, + "loss": 0.0221, + "step": 33656 + }, + { + "epoch": 4.41, + "learning_rate": 8.83355978504166e-06, + "loss": 0.0212, + "step": 33657 + }, + { + "epoch": 4.41, + "learning_rate": 8.833033081728576e-06, + "loss": 0.0239, + "step": 33658 + }, + { + "epoch": 4.41, + "learning_rate": 8.832506381697527e-06, + "loss": 0.0182, + "step": 33659 + }, + { + "epoch": 4.41, + "learning_rate": 8.83197968494999e-06, + "loss": 0.0231, + "step": 33660 + }, + { + "epoch": 4.41, + "learning_rate": 8.831452991487447e-06, + "loss": 0.0218, + "step": 33661 + }, + { + "epoch": 4.41, + "learning_rate": 8.830926301311383e-06, + "loss": 0.0189, + "step": 33662 + }, + { + "epoch": 4.41, + "learning_rate": 8.830399614423273e-06, + "loss": 0.0237, + "step": 33663 + }, + { + "epoch": 4.41, + "learning_rate": 8.829872930824602e-06, + "loss": 0.0223, + "step": 33664 + }, + { + "epoch": 4.41, + "learning_rate": 8.829346250516853e-06, + "loss": 0.0246, + "step": 33665 + }, + { + "epoch": 4.41, + "learning_rate": 8.828819573501505e-06, + "loss": 0.0184, + "step": 33666 + }, + { + "epoch": 4.41, + "learning_rate": 8.828292899780042e-06, + "loss": 0.0217, + "step": 33667 + }, + { + "epoch": 4.41, + "learning_rate": 8.82776622935394e-06, + "loss": 0.0156, + "step": 33668 + }, + { + "epoch": 4.41, + "learning_rate": 8.827239562224682e-06, + "loss": 0.0248, + "step": 33669 + }, + { + "epoch": 4.41, + "learning_rate": 8.826712898393753e-06, + "loss": 0.0205, + "step": 33670 + }, + { + "epoch": 4.41, + "learning_rate": 8.826186237862629e-06, + "loss": 0.0167, + "step": 33671 + }, + { + "epoch": 4.41, + "learning_rate": 8.825659580632797e-06, + "loss": 0.0228, + "step": 33672 + }, + { + "epoch": 4.41, + "learning_rate": 8.825132926705732e-06, + "loss": 0.026, + "step": 33673 + }, + { + "epoch": 4.41, + "learning_rate": 8.824606276082922e-06, + "loss": 0.0177, + "step": 33674 + }, + { + "epoch": 4.41, + "learning_rate": 8.824079628765835e-06, + "loss": 0.02, + "step": 33675 + }, + { + "epoch": 4.41, + "learning_rate": 8.823552984755968e-06, + "loss": 0.0193, + "step": 33676 + }, + { + "epoch": 4.41, + "learning_rate": 8.823026344054796e-06, + "loss": 0.0244, + "step": 33677 + }, + { + "epoch": 4.41, + "learning_rate": 8.8224997066638e-06, + "loss": 0.0163, + "step": 33678 + }, + { + "epoch": 4.41, + "learning_rate": 8.821973072584458e-06, + "loss": 0.0203, + "step": 33679 + }, + { + "epoch": 4.41, + "learning_rate": 8.821446441818258e-06, + "loss": 0.0185, + "step": 33680 + }, + { + "epoch": 4.41, + "learning_rate": 8.820919814366674e-06, + "loss": 0.0192, + "step": 33681 + }, + { + "epoch": 4.41, + "learning_rate": 8.820393190231193e-06, + "loss": 0.0195, + "step": 33682 + }, + { + "epoch": 4.41, + "learning_rate": 8.819866569413291e-06, + "loss": 0.0238, + "step": 33683 + }, + { + "epoch": 4.41, + "learning_rate": 8.819339951914451e-06, + "loss": 0.0219, + "step": 33684 + }, + { + "epoch": 4.41, + "learning_rate": 8.818813337736158e-06, + "loss": 0.0225, + "step": 33685 + }, + { + "epoch": 4.41, + "learning_rate": 8.818286726879884e-06, + "loss": 0.0183, + "step": 33686 + }, + { + "epoch": 4.41, + "learning_rate": 8.817760119347118e-06, + "loss": 0.0172, + "step": 33687 + }, + { + "epoch": 4.41, + "learning_rate": 8.81723351513934e-06, + "loss": 0.0303, + "step": 33688 + }, + { + "epoch": 4.41, + "learning_rate": 8.816706914258028e-06, + "loss": 0.0197, + "step": 33689 + }, + { + "epoch": 4.41, + "learning_rate": 8.816180316704666e-06, + "loss": 0.0248, + "step": 33690 + }, + { + "epoch": 4.41, + "learning_rate": 8.815653722480734e-06, + "loss": 0.0178, + "step": 33691 + }, + { + "epoch": 4.41, + "learning_rate": 8.815127131587714e-06, + "loss": 0.0222, + "step": 33692 + }, + { + "epoch": 4.41, + "learning_rate": 8.814600544027084e-06, + "loss": 0.0247, + "step": 33693 + }, + { + "epoch": 4.41, + "learning_rate": 8.814073959800326e-06, + "loss": 0.017, + "step": 33694 + }, + { + "epoch": 4.41, + "learning_rate": 8.813547378908925e-06, + "loss": 0.0219, + "step": 33695 + }, + { + "epoch": 4.41, + "learning_rate": 8.813020801354356e-06, + "loss": 0.0218, + "step": 33696 + }, + { + "epoch": 4.41, + "learning_rate": 8.812494227138101e-06, + "loss": 0.0218, + "step": 33697 + }, + { + "epoch": 4.41, + "learning_rate": 8.811967656261645e-06, + "loss": 0.0203, + "step": 33698 + }, + { + "epoch": 4.41, + "learning_rate": 8.81144108872647e-06, + "loss": 0.0169, + "step": 33699 + }, + { + "epoch": 4.41, + "learning_rate": 8.810914524534049e-06, + "loss": 0.0222, + "step": 33700 + }, + { + "epoch": 4.41, + "learning_rate": 8.81038796368587e-06, + "loss": 0.0213, + "step": 33701 + }, + { + "epoch": 4.41, + "learning_rate": 8.809861406183412e-06, + "loss": 0.0208, + "step": 33702 + }, + { + "epoch": 4.41, + "learning_rate": 8.809334852028155e-06, + "loss": 0.0202, + "step": 33703 + }, + { + "epoch": 4.41, + "learning_rate": 8.808808301221579e-06, + "loss": 0.0176, + "step": 33704 + }, + { + "epoch": 4.42, + "learning_rate": 8.808281753765165e-06, + "loss": 0.0177, + "step": 33705 + }, + { + "epoch": 4.42, + "learning_rate": 8.8077552096604e-06, + "loss": 0.0226, + "step": 33706 + }, + { + "epoch": 4.42, + "learning_rate": 8.807228668908753e-06, + "loss": 0.0245, + "step": 33707 + }, + { + "epoch": 4.42, + "learning_rate": 8.806702131511715e-06, + "loss": 0.0201, + "step": 33708 + }, + { + "epoch": 4.42, + "learning_rate": 8.806175597470765e-06, + "loss": 0.0183, + "step": 33709 + }, + { + "epoch": 4.42, + "learning_rate": 8.805649066787382e-06, + "loss": 0.0183, + "step": 33710 + }, + { + "epoch": 4.42, + "learning_rate": 8.805122539463046e-06, + "loss": 0.0206, + "step": 33711 + }, + { + "epoch": 4.42, + "learning_rate": 8.804596015499242e-06, + "loss": 0.0215, + "step": 33712 + }, + { + "epoch": 4.42, + "learning_rate": 8.804069494897447e-06, + "loss": 0.0197, + "step": 33713 + }, + { + "epoch": 4.42, + "learning_rate": 8.803542977659144e-06, + "loss": 0.0201, + "step": 33714 + }, + { + "epoch": 4.42, + "learning_rate": 8.803016463785811e-06, + "loss": 0.0238, + "step": 33715 + }, + { + "epoch": 4.42, + "learning_rate": 8.802489953278932e-06, + "loss": 0.02, + "step": 33716 + }, + { + "epoch": 4.42, + "learning_rate": 8.801963446139984e-06, + "loss": 0.0181, + "step": 33717 + }, + { + "epoch": 4.42, + "learning_rate": 8.801436942370448e-06, + "loss": 0.0251, + "step": 33718 + }, + { + "epoch": 4.42, + "learning_rate": 8.800910441971808e-06, + "loss": 0.0219, + "step": 33719 + }, + { + "epoch": 4.42, + "learning_rate": 8.800383944945548e-06, + "loss": 0.0262, + "step": 33720 + }, + { + "epoch": 4.42, + "learning_rate": 8.79985745129314e-06, + "loss": 0.019, + "step": 33721 + }, + { + "epoch": 4.42, + "learning_rate": 8.799330961016071e-06, + "loss": 0.0153, + "step": 33722 + }, + { + "epoch": 4.42, + "learning_rate": 8.79880447411582e-06, + "loss": 0.0248, + "step": 33723 + }, + { + "epoch": 4.42, + "learning_rate": 8.798277990593867e-06, + "loss": 0.0225, + "step": 33724 + }, + { + "epoch": 4.42, + "learning_rate": 8.797751510451693e-06, + "loss": 0.0227, + "step": 33725 + }, + { + "epoch": 4.42, + "learning_rate": 8.797225033690777e-06, + "loss": 0.0201, + "step": 33726 + }, + { + "epoch": 4.42, + "learning_rate": 8.796698560312606e-06, + "loss": 0.0213, + "step": 33727 + }, + { + "epoch": 4.42, + "learning_rate": 8.796172090318653e-06, + "loss": 0.022, + "step": 33728 + }, + { + "epoch": 4.42, + "learning_rate": 8.795645623710399e-06, + "loss": 0.0233, + "step": 33729 + }, + { + "epoch": 4.42, + "learning_rate": 8.795119160489332e-06, + "loss": 0.0249, + "step": 33730 + }, + { + "epoch": 4.42, + "learning_rate": 8.794592700656925e-06, + "loss": 0.0208, + "step": 33731 + }, + { + "epoch": 4.42, + "learning_rate": 8.794066244214665e-06, + "loss": 0.0222, + "step": 33732 + }, + { + "epoch": 4.42, + "learning_rate": 8.79353979116403e-06, + "loss": 0.0237, + "step": 33733 + }, + { + "epoch": 4.42, + "learning_rate": 8.793013341506497e-06, + "loss": 0.0234, + "step": 33734 + }, + { + "epoch": 4.42, + "learning_rate": 8.792486895243553e-06, + "loss": 0.0183, + "step": 33735 + }, + { + "epoch": 4.42, + "learning_rate": 8.791960452376672e-06, + "loss": 0.0234, + "step": 33736 + }, + { + "epoch": 4.42, + "learning_rate": 8.79143401290734e-06, + "loss": 0.0218, + "step": 33737 + }, + { + "epoch": 4.42, + "learning_rate": 8.790907576837035e-06, + "loss": 0.0191, + "step": 33738 + }, + { + "epoch": 4.42, + "learning_rate": 8.79038114416724e-06, + "loss": 0.0193, + "step": 33739 + }, + { + "epoch": 4.42, + "learning_rate": 8.789854714899425e-06, + "loss": 0.0204, + "step": 33740 + }, + { + "epoch": 4.42, + "learning_rate": 8.789328289035088e-06, + "loss": 0.0206, + "step": 33741 + }, + { + "epoch": 4.42, + "learning_rate": 8.788801866575697e-06, + "loss": 0.0279, + "step": 33742 + }, + { + "epoch": 4.42, + "learning_rate": 8.78827544752274e-06, + "loss": 0.0232, + "step": 33743 + }, + { + "epoch": 4.42, + "learning_rate": 8.78774903187769e-06, + "loss": 0.0236, + "step": 33744 + }, + { + "epoch": 4.42, + "learning_rate": 8.787222619642034e-06, + "loss": 0.0204, + "step": 33745 + }, + { + "epoch": 4.42, + "learning_rate": 8.786696210817248e-06, + "loss": 0.0211, + "step": 33746 + }, + { + "epoch": 4.42, + "learning_rate": 8.786169805404816e-06, + "loss": 0.0266, + "step": 33747 + }, + { + "epoch": 4.42, + "learning_rate": 8.785643403406216e-06, + "loss": 0.0193, + "step": 33748 + }, + { + "epoch": 4.42, + "learning_rate": 8.785117004822927e-06, + "loss": 0.0173, + "step": 33749 + }, + { + "epoch": 4.42, + "learning_rate": 8.784590609656436e-06, + "loss": 0.0194, + "step": 33750 + }, + { + "epoch": 4.42, + "learning_rate": 8.784064217908214e-06, + "loss": 0.0231, + "step": 33751 + }, + { + "epoch": 4.42, + "learning_rate": 8.783537829579748e-06, + "loss": 0.0264, + "step": 33752 + }, + { + "epoch": 4.42, + "learning_rate": 8.78301144467252e-06, + "loss": 0.0228, + "step": 33753 + }, + { + "epoch": 4.42, + "learning_rate": 8.782485063188006e-06, + "loss": 0.0203, + "step": 33754 + }, + { + "epoch": 4.42, + "learning_rate": 8.781958685127687e-06, + "loss": 0.0215, + "step": 33755 + }, + { + "epoch": 4.42, + "learning_rate": 8.781432310493047e-06, + "loss": 0.0196, + "step": 33756 + }, + { + "epoch": 4.42, + "learning_rate": 8.78090593928556e-06, + "loss": 0.0183, + "step": 33757 + }, + { + "epoch": 4.42, + "learning_rate": 8.780379571506715e-06, + "loss": 0.0203, + "step": 33758 + }, + { + "epoch": 4.42, + "learning_rate": 8.779853207157982e-06, + "loss": 0.0217, + "step": 33759 + }, + { + "epoch": 4.42, + "learning_rate": 8.779326846240851e-06, + "loss": 0.02, + "step": 33760 + }, + { + "epoch": 4.42, + "learning_rate": 8.778800488756797e-06, + "loss": 0.0238, + "step": 33761 + }, + { + "epoch": 4.42, + "learning_rate": 8.778274134707298e-06, + "loss": 0.019, + "step": 33762 + }, + { + "epoch": 4.42, + "learning_rate": 8.777747784093841e-06, + "loss": 0.0193, + "step": 33763 + }, + { + "epoch": 4.42, + "learning_rate": 8.777221436917904e-06, + "loss": 0.0181, + "step": 33764 + }, + { + "epoch": 4.42, + "learning_rate": 8.776695093180966e-06, + "loss": 0.0218, + "step": 33765 + }, + { + "epoch": 4.42, + "learning_rate": 8.77616875288451e-06, + "loss": 0.0217, + "step": 33766 + }, + { + "epoch": 4.42, + "learning_rate": 8.775642416030011e-06, + "loss": 0.0181, + "step": 33767 + }, + { + "epoch": 4.42, + "learning_rate": 8.775116082618956e-06, + "loss": 0.0206, + "step": 33768 + }, + { + "epoch": 4.42, + "learning_rate": 8.774589752652818e-06, + "loss": 0.0243, + "step": 33769 + }, + { + "epoch": 4.42, + "learning_rate": 8.774063426133082e-06, + "loss": 0.0216, + "step": 33770 + }, + { + "epoch": 4.42, + "learning_rate": 8.773537103061229e-06, + "loss": 0.0163, + "step": 33771 + }, + { + "epoch": 4.42, + "learning_rate": 8.773010783438736e-06, + "loss": 0.0201, + "step": 33772 + }, + { + "epoch": 4.42, + "learning_rate": 8.772484467267081e-06, + "loss": 0.0217, + "step": 33773 + }, + { + "epoch": 4.42, + "learning_rate": 8.771958154547755e-06, + "loss": 0.0171, + "step": 33774 + }, + { + "epoch": 4.42, + "learning_rate": 8.771431845282229e-06, + "loss": 0.0205, + "step": 33775 + }, + { + "epoch": 4.42, + "learning_rate": 8.770905539471984e-06, + "loss": 0.0175, + "step": 33776 + }, + { + "epoch": 4.42, + "learning_rate": 8.770379237118504e-06, + "loss": 0.0182, + "step": 33777 + }, + { + "epoch": 4.42, + "learning_rate": 8.769852938223265e-06, + "loss": 0.0199, + "step": 33778 + }, + { + "epoch": 4.42, + "learning_rate": 8.76932664278775e-06, + "loss": 0.0211, + "step": 33779 + }, + { + "epoch": 4.42, + "learning_rate": 8.768800350813438e-06, + "loss": 0.0249, + "step": 33780 + }, + { + "epoch": 4.43, + "learning_rate": 8.768274062301812e-06, + "loss": 0.0199, + "step": 33781 + }, + { + "epoch": 4.43, + "learning_rate": 8.767747777254347e-06, + "loss": 0.0222, + "step": 33782 + }, + { + "epoch": 4.43, + "learning_rate": 8.767221495672526e-06, + "loss": 0.0264, + "step": 33783 + }, + { + "epoch": 4.43, + "learning_rate": 8.766695217557824e-06, + "loss": 0.0197, + "step": 33784 + }, + { + "epoch": 4.43, + "learning_rate": 8.766168942911734e-06, + "loss": 0.0253, + "step": 33785 + }, + { + "epoch": 4.43, + "learning_rate": 8.765642671735724e-06, + "loss": 0.0228, + "step": 33786 + }, + { + "epoch": 4.43, + "learning_rate": 8.76511640403128e-06, + "loss": 0.017, + "step": 33787 + }, + { + "epoch": 4.43, + "learning_rate": 8.76459013979988e-06, + "loss": 0.018, + "step": 33788 + }, + { + "epoch": 4.43, + "learning_rate": 8.764063879043005e-06, + "loss": 0.0196, + "step": 33789 + }, + { + "epoch": 4.43, + "learning_rate": 8.763537621762132e-06, + "loss": 0.0211, + "step": 33790 + }, + { + "epoch": 4.43, + "learning_rate": 8.763011367958745e-06, + "loss": 0.0247, + "step": 33791 + }, + { + "epoch": 4.43, + "learning_rate": 8.762485117634325e-06, + "loss": 0.0169, + "step": 33792 + }, + { + "epoch": 4.43, + "learning_rate": 8.761958870790346e-06, + "loss": 0.0205, + "step": 33793 + }, + { + "epoch": 4.43, + "learning_rate": 8.761432627428295e-06, + "loss": 0.0202, + "step": 33794 + }, + { + "epoch": 4.43, + "learning_rate": 8.760906387549645e-06, + "loss": 0.0206, + "step": 33795 + }, + { + "epoch": 4.43, + "learning_rate": 8.760380151155881e-06, + "loss": 0.0233, + "step": 33796 + }, + { + "epoch": 4.43, + "learning_rate": 8.759853918248484e-06, + "loss": 0.0192, + "step": 33797 + }, + { + "epoch": 4.43, + "learning_rate": 8.759327688828929e-06, + "loss": 0.0208, + "step": 33798 + }, + { + "epoch": 4.43, + "learning_rate": 8.7588014628987e-06, + "loss": 0.0197, + "step": 33799 + }, + { + "epoch": 4.43, + "learning_rate": 8.758275240459278e-06, + "loss": 0.0214, + "step": 33800 + }, + { + "epoch": 4.43, + "learning_rate": 8.757749021512138e-06, + "loss": 0.0195, + "step": 33801 + }, + { + "epoch": 4.43, + "learning_rate": 8.757222806058765e-06, + "loss": 0.0198, + "step": 33802 + }, + { + "epoch": 4.43, + "learning_rate": 8.756696594100635e-06, + "loss": 0.0197, + "step": 33803 + }, + { + "epoch": 4.43, + "learning_rate": 8.756170385639231e-06, + "loss": 0.0238, + "step": 33804 + }, + { + "epoch": 4.43, + "learning_rate": 8.755644180676029e-06, + "loss": 0.0196, + "step": 33805 + }, + { + "epoch": 4.43, + "learning_rate": 8.75511797921251e-06, + "loss": 0.021, + "step": 33806 + }, + { + "epoch": 4.43, + "learning_rate": 8.754591781250159e-06, + "loss": 0.0247, + "step": 33807 + }, + { + "epoch": 4.43, + "learning_rate": 8.754065586790452e-06, + "loss": 0.0214, + "step": 33808 + }, + { + "epoch": 4.43, + "learning_rate": 8.75353939583487e-06, + "loss": 0.0211, + "step": 33809 + }, + { + "epoch": 4.43, + "learning_rate": 8.75301320838489e-06, + "loss": 0.0199, + "step": 33810 + }, + { + "epoch": 4.43, + "learning_rate": 8.752487024441995e-06, + "loss": 0.0205, + "step": 33811 + }, + { + "epoch": 4.43, + "learning_rate": 8.751960844007663e-06, + "loss": 0.0186, + "step": 33812 + }, + { + "epoch": 4.43, + "learning_rate": 8.751434667083378e-06, + "loss": 0.02, + "step": 33813 + }, + { + "epoch": 4.43, + "learning_rate": 8.750908493670612e-06, + "loss": 0.0195, + "step": 33814 + }, + { + "epoch": 4.43, + "learning_rate": 8.750382323770852e-06, + "loss": 0.02, + "step": 33815 + }, + { + "epoch": 4.43, + "learning_rate": 8.749856157385574e-06, + "loss": 0.0199, + "step": 33816 + }, + { + "epoch": 4.43, + "learning_rate": 8.749329994516258e-06, + "loss": 0.0213, + "step": 33817 + }, + { + "epoch": 4.43, + "learning_rate": 8.748803835164386e-06, + "loss": 0.0253, + "step": 33818 + }, + { + "epoch": 4.43, + "learning_rate": 8.748277679331436e-06, + "loss": 0.0192, + "step": 33819 + }, + { + "epoch": 4.43, + "learning_rate": 8.74775152701889e-06, + "loss": 0.0181, + "step": 33820 + }, + { + "epoch": 4.43, + "learning_rate": 8.747225378228224e-06, + "loss": 0.0242, + "step": 33821 + }, + { + "epoch": 4.43, + "learning_rate": 8.746699232960921e-06, + "loss": 0.0222, + "step": 33822 + }, + { + "epoch": 4.43, + "learning_rate": 8.746173091218461e-06, + "loss": 0.0254, + "step": 33823 + }, + { + "epoch": 4.43, + "learning_rate": 8.745646953002321e-06, + "loss": 0.0215, + "step": 33824 + }, + { + "epoch": 4.43, + "learning_rate": 8.745120818313984e-06, + "loss": 0.0239, + "step": 33825 + }, + { + "epoch": 4.43, + "learning_rate": 8.744594687154926e-06, + "loss": 0.0189, + "step": 33826 + }, + { + "epoch": 4.43, + "learning_rate": 8.744068559526626e-06, + "loss": 0.0228, + "step": 33827 + }, + { + "epoch": 4.43, + "learning_rate": 8.743542435430568e-06, + "loss": 0.0187, + "step": 33828 + }, + { + "epoch": 4.43, + "learning_rate": 8.743016314868233e-06, + "loss": 0.0205, + "step": 33829 + }, + { + "epoch": 4.43, + "learning_rate": 8.742490197841093e-06, + "loss": 0.0181, + "step": 33830 + }, + { + "epoch": 4.43, + "learning_rate": 8.741964084350638e-06, + "loss": 0.0227, + "step": 33831 + }, + { + "epoch": 4.43, + "learning_rate": 8.741437974398337e-06, + "loss": 0.0202, + "step": 33832 + }, + { + "epoch": 4.43, + "learning_rate": 8.740911867985679e-06, + "loss": 0.0182, + "step": 33833 + }, + { + "epoch": 4.43, + "learning_rate": 8.740385765114137e-06, + "loss": 0.0203, + "step": 33834 + }, + { + "epoch": 4.43, + "learning_rate": 8.739859665785192e-06, + "loss": 0.0223, + "step": 33835 + }, + { + "epoch": 4.43, + "learning_rate": 8.739333570000327e-06, + "loss": 0.0222, + "step": 33836 + }, + { + "epoch": 4.43, + "learning_rate": 8.738807477761018e-06, + "loss": 0.0201, + "step": 33837 + }, + { + "epoch": 4.43, + "learning_rate": 8.738281389068742e-06, + "loss": 0.0195, + "step": 33838 + }, + { + "epoch": 4.43, + "learning_rate": 8.737755303924988e-06, + "loss": 0.0195, + "step": 33839 + }, + { + "epoch": 4.43, + "learning_rate": 8.737229222331227e-06, + "loss": 0.0187, + "step": 33840 + }, + { + "epoch": 4.43, + "learning_rate": 8.73670314428894e-06, + "loss": 0.0174, + "step": 33841 + }, + { + "epoch": 4.43, + "learning_rate": 8.736177069799614e-06, + "loss": 0.024, + "step": 33842 + }, + { + "epoch": 4.43, + "learning_rate": 8.73565099886472e-06, + "loss": 0.0237, + "step": 33843 + }, + { + "epoch": 4.43, + "learning_rate": 8.73512493148574e-06, + "loss": 0.024, + "step": 33844 + }, + { + "epoch": 4.43, + "learning_rate": 8.734598867664153e-06, + "loss": 0.0214, + "step": 33845 + }, + { + "epoch": 4.43, + "learning_rate": 8.734072807401441e-06, + "loss": 0.0198, + "step": 33846 + }, + { + "epoch": 4.43, + "learning_rate": 8.73354675069908e-06, + "loss": 0.0207, + "step": 33847 + }, + { + "epoch": 4.43, + "learning_rate": 8.733020697558553e-06, + "loss": 0.023, + "step": 33848 + }, + { + "epoch": 4.43, + "learning_rate": 8.732494647981331e-06, + "loss": 0.0252, + "step": 33849 + }, + { + "epoch": 4.43, + "learning_rate": 8.731968601968907e-06, + "loss": 0.0197, + "step": 33850 + }, + { + "epoch": 4.43, + "learning_rate": 8.731442559522754e-06, + "loss": 0.0226, + "step": 33851 + }, + { + "epoch": 4.43, + "learning_rate": 8.730916520644351e-06, + "loss": 0.0214, + "step": 33852 + }, + { + "epoch": 4.43, + "learning_rate": 8.730390485335176e-06, + "loss": 0.0201, + "step": 33853 + }, + { + "epoch": 4.43, + "learning_rate": 8.729864453596712e-06, + "loss": 0.0296, + "step": 33854 + }, + { + "epoch": 4.43, + "learning_rate": 8.729338425430435e-06, + "loss": 0.0229, + "step": 33855 + }, + { + "epoch": 4.43, + "learning_rate": 8.728812400837825e-06, + "loss": 0.0212, + "step": 33856 + }, + { + "epoch": 4.44, + "learning_rate": 8.728286379820366e-06, + "loss": 0.0169, + "step": 33857 + }, + { + "epoch": 4.44, + "learning_rate": 8.72776036237953e-06, + "loss": 0.0183, + "step": 33858 + }, + { + "epoch": 4.44, + "learning_rate": 8.727234348516805e-06, + "loss": 0.0253, + "step": 33859 + }, + { + "epoch": 4.44, + "learning_rate": 8.726708338233658e-06, + "loss": 0.0233, + "step": 33860 + }, + { + "epoch": 4.44, + "learning_rate": 8.726182331531581e-06, + "loss": 0.0218, + "step": 33861 + }, + { + "epoch": 4.44, + "learning_rate": 8.725656328412046e-06, + "loss": 0.0223, + "step": 33862 + }, + { + "epoch": 4.44, + "learning_rate": 8.725130328876538e-06, + "loss": 0.0185, + "step": 33863 + }, + { + "epoch": 4.44, + "learning_rate": 8.724604332926532e-06, + "loss": 0.022, + "step": 33864 + }, + { + "epoch": 4.44, + "learning_rate": 8.72407834056351e-06, + "loss": 0.0207, + "step": 33865 + }, + { + "epoch": 4.44, + "learning_rate": 8.723552351788947e-06, + "loss": 0.0204, + "step": 33866 + }, + { + "epoch": 4.44, + "learning_rate": 8.723026366604327e-06, + "loss": 0.0177, + "step": 33867 + }, + { + "epoch": 4.44, + "learning_rate": 8.722500385011125e-06, + "loss": 0.0204, + "step": 33868 + }, + { + "epoch": 4.44, + "learning_rate": 8.721974407010826e-06, + "loss": 0.0175, + "step": 33869 + }, + { + "epoch": 4.44, + "learning_rate": 8.721448432604902e-06, + "loss": 0.025, + "step": 33870 + }, + { + "epoch": 4.44, + "learning_rate": 8.720922461794836e-06, + "loss": 0.0198, + "step": 33871 + }, + { + "epoch": 4.44, + "learning_rate": 8.720396494582109e-06, + "loss": 0.0214, + "step": 33872 + }, + { + "epoch": 4.44, + "learning_rate": 8.719870530968199e-06, + "loss": 0.0181, + "step": 33873 + }, + { + "epoch": 4.44, + "learning_rate": 8.719344570954586e-06, + "loss": 0.0182, + "step": 33874 + }, + { + "epoch": 4.44, + "learning_rate": 8.718818614542748e-06, + "loss": 0.0218, + "step": 33875 + }, + { + "epoch": 4.44, + "learning_rate": 8.718292661734164e-06, + "loss": 0.02, + "step": 33876 + }, + { + "epoch": 4.44, + "learning_rate": 8.71776671253031e-06, + "loss": 0.0205, + "step": 33877 + }, + { + "epoch": 4.44, + "learning_rate": 8.717240766932675e-06, + "loss": 0.0217, + "step": 33878 + }, + { + "epoch": 4.44, + "learning_rate": 8.716714824942729e-06, + "loss": 0.0216, + "step": 33879 + }, + { + "epoch": 4.44, + "learning_rate": 8.716188886561956e-06, + "loss": 0.0225, + "step": 33880 + }, + { + "epoch": 4.44, + "learning_rate": 8.715662951791832e-06, + "loss": 0.0173, + "step": 33881 + }, + { + "epoch": 4.44, + "learning_rate": 8.715137020633833e-06, + "loss": 0.0185, + "step": 33882 + }, + { + "epoch": 4.44, + "learning_rate": 8.714611093089449e-06, + "loss": 0.0247, + "step": 33883 + }, + { + "epoch": 4.44, + "learning_rate": 8.714085169160151e-06, + "loss": 0.0206, + "step": 33884 + }, + { + "epoch": 4.44, + "learning_rate": 8.713559248847418e-06, + "loss": 0.0244, + "step": 33885 + }, + { + "epoch": 4.44, + "learning_rate": 8.713033332152735e-06, + "loss": 0.0239, + "step": 33886 + }, + { + "epoch": 4.44, + "learning_rate": 8.712507419077575e-06, + "loss": 0.0235, + "step": 33887 + }, + { + "epoch": 4.44, + "learning_rate": 8.71198150962342e-06, + "loss": 0.018, + "step": 33888 + }, + { + "epoch": 4.44, + "learning_rate": 8.711455603791747e-06, + "loss": 0.0223, + "step": 33889 + }, + { + "epoch": 4.44, + "learning_rate": 8.710929701584038e-06, + "loss": 0.0211, + "step": 33890 + }, + { + "epoch": 4.44, + "learning_rate": 8.71040380300177e-06, + "loss": 0.02, + "step": 33891 + }, + { + "epoch": 4.44, + "learning_rate": 8.709877908046423e-06, + "loss": 0.024, + "step": 33892 + }, + { + "epoch": 4.44, + "learning_rate": 8.709352016719472e-06, + "loss": 0.0184, + "step": 33893 + }, + { + "epoch": 4.44, + "learning_rate": 8.708826129022405e-06, + "loss": 0.0191, + "step": 33894 + }, + { + "epoch": 4.44, + "learning_rate": 8.708300244956693e-06, + "loss": 0.0156, + "step": 33895 + }, + { + "epoch": 4.44, + "learning_rate": 8.70777436452382e-06, + "loss": 0.0255, + "step": 33896 + }, + { + "epoch": 4.44, + "learning_rate": 8.70724848772526e-06, + "loss": 0.028, + "step": 33897 + }, + { + "epoch": 4.44, + "learning_rate": 8.706722614562498e-06, + "loss": 0.0211, + "step": 33898 + }, + { + "epoch": 4.44, + "learning_rate": 8.706196745037008e-06, + "loss": 0.0205, + "step": 33899 + }, + { + "epoch": 4.44, + "learning_rate": 8.70567087915027e-06, + "loss": 0.0143, + "step": 33900 + }, + { + "epoch": 4.44, + "learning_rate": 8.705145016903764e-06, + "loss": 0.0213, + "step": 33901 + }, + { + "epoch": 4.44, + "learning_rate": 8.704619158298969e-06, + "loss": 0.0229, + "step": 33902 + }, + { + "epoch": 4.44, + "learning_rate": 8.704093303337365e-06, + "loss": 0.0235, + "step": 33903 + }, + { + "epoch": 4.44, + "learning_rate": 8.703567452020425e-06, + "loss": 0.0238, + "step": 33904 + }, + { + "epoch": 4.44, + "learning_rate": 8.703041604349636e-06, + "loss": 0.0188, + "step": 33905 + }, + { + "epoch": 4.44, + "learning_rate": 8.702515760326473e-06, + "loss": 0.0161, + "step": 33906 + }, + { + "epoch": 4.44, + "learning_rate": 8.701989919952417e-06, + "loss": 0.0192, + "step": 33907 + }, + { + "epoch": 4.44, + "learning_rate": 8.701464083228944e-06, + "loss": 0.0201, + "step": 33908 + }, + { + "epoch": 4.44, + "learning_rate": 8.700938250157534e-06, + "loss": 0.0194, + "step": 33909 + }, + { + "epoch": 4.44, + "learning_rate": 8.700412420739666e-06, + "loss": 0.0213, + "step": 33910 + }, + { + "epoch": 4.44, + "learning_rate": 8.69988659497682e-06, + "loss": 0.0234, + "step": 33911 + }, + { + "epoch": 4.44, + "learning_rate": 8.699360772870472e-06, + "loss": 0.02, + "step": 33912 + }, + { + "epoch": 4.44, + "learning_rate": 8.698834954422104e-06, + "loss": 0.0136, + "step": 33913 + }, + { + "epoch": 4.44, + "learning_rate": 8.69830913963319e-06, + "loss": 0.022, + "step": 33914 + }, + { + "epoch": 4.44, + "learning_rate": 8.697783328505213e-06, + "loss": 0.02, + "step": 33915 + }, + { + "epoch": 4.44, + "learning_rate": 8.697257521039652e-06, + "loss": 0.0179, + "step": 33916 + }, + { + "epoch": 4.44, + "learning_rate": 8.696731717237986e-06, + "loss": 0.0206, + "step": 33917 + }, + { + "epoch": 4.44, + "learning_rate": 8.696205917101691e-06, + "loss": 0.0203, + "step": 33918 + }, + { + "epoch": 4.44, + "learning_rate": 8.695680120632251e-06, + "loss": 0.0178, + "step": 33919 + }, + { + "epoch": 4.44, + "learning_rate": 8.69515432783114e-06, + "loss": 0.0178, + "step": 33920 + }, + { + "epoch": 4.44, + "learning_rate": 8.694628538699834e-06, + "loss": 0.0232, + "step": 33921 + }, + { + "epoch": 4.44, + "learning_rate": 8.694102753239821e-06, + "loss": 0.0235, + "step": 33922 + }, + { + "epoch": 4.44, + "learning_rate": 8.69357697145257e-06, + "loss": 0.0221, + "step": 33923 + }, + { + "epoch": 4.44, + "learning_rate": 8.693051193339567e-06, + "loss": 0.0191, + "step": 33924 + }, + { + "epoch": 4.44, + "learning_rate": 8.692525418902287e-06, + "loss": 0.0215, + "step": 33925 + }, + { + "epoch": 4.44, + "learning_rate": 8.691999648142208e-06, + "loss": 0.0187, + "step": 33926 + }, + { + "epoch": 4.44, + "learning_rate": 8.691473881060811e-06, + "loss": 0.0213, + "step": 33927 + }, + { + "epoch": 4.44, + "learning_rate": 8.690948117659576e-06, + "loss": 0.0184, + "step": 33928 + }, + { + "epoch": 4.44, + "learning_rate": 8.690422357939978e-06, + "loss": 0.0206, + "step": 33929 + }, + { + "epoch": 4.44, + "learning_rate": 8.6898966019035e-06, + "loss": 0.0249, + "step": 33930 + }, + { + "epoch": 4.44, + "learning_rate": 8.689370849551616e-06, + "loss": 0.0216, + "step": 33931 + }, + { + "epoch": 4.44, + "learning_rate": 8.688845100885807e-06, + "loss": 0.0201, + "step": 33932 + }, + { + "epoch": 4.44, + "learning_rate": 8.688319355907552e-06, + "loss": 0.0196, + "step": 33933 + }, + { + "epoch": 4.45, + "learning_rate": 8.687793614618329e-06, + "loss": 0.0213, + "step": 33934 + }, + { + "epoch": 4.45, + "learning_rate": 8.687267877019615e-06, + "loss": 0.024, + "step": 33935 + }, + { + "epoch": 4.45, + "learning_rate": 8.68674214311289e-06, + "loss": 0.0214, + "step": 33936 + }, + { + "epoch": 4.45, + "learning_rate": 8.686216412899631e-06, + "loss": 0.0158, + "step": 33937 + }, + { + "epoch": 4.45, + "learning_rate": 8.685690686381322e-06, + "loss": 0.0247, + "step": 33938 + }, + { + "epoch": 4.45, + "learning_rate": 8.685164963559437e-06, + "loss": 0.0219, + "step": 33939 + }, + { + "epoch": 4.45, + "learning_rate": 8.684639244435458e-06, + "loss": 0.0202, + "step": 33940 + }, + { + "epoch": 4.45, + "learning_rate": 8.684113529010857e-06, + "loss": 0.0231, + "step": 33941 + }, + { + "epoch": 4.45, + "learning_rate": 8.683587817287117e-06, + "loss": 0.0217, + "step": 33942 + }, + { + "epoch": 4.45, + "learning_rate": 8.68306210926572e-06, + "loss": 0.021, + "step": 33943 + }, + { + "epoch": 4.45, + "learning_rate": 8.682536404948136e-06, + "loss": 0.0197, + "step": 33944 + }, + { + "epoch": 4.45, + "learning_rate": 8.682010704335853e-06, + "loss": 0.0218, + "step": 33945 + }, + { + "epoch": 4.45, + "learning_rate": 8.681485007430341e-06, + "loss": 0.0197, + "step": 33946 + }, + { + "epoch": 4.45, + "learning_rate": 8.680959314233086e-06, + "loss": 0.0214, + "step": 33947 + }, + { + "epoch": 4.45, + "learning_rate": 8.680433624745555e-06, + "loss": 0.0207, + "step": 33948 + }, + { + "epoch": 4.45, + "learning_rate": 8.679907938969238e-06, + "loss": 0.0236, + "step": 33949 + }, + { + "epoch": 4.45, + "learning_rate": 8.67938225690561e-06, + "loss": 0.0188, + "step": 33950 + }, + { + "epoch": 4.45, + "learning_rate": 8.678856578556152e-06, + "loss": 0.0233, + "step": 33951 + }, + { + "epoch": 4.45, + "learning_rate": 8.678330903922338e-06, + "loss": 0.0222, + "step": 33952 + }, + { + "epoch": 4.45, + "learning_rate": 8.677805233005647e-06, + "loss": 0.0274, + "step": 33953 + }, + { + "epoch": 4.45, + "learning_rate": 8.67727956580756e-06, + "loss": 0.0191, + "step": 33954 + }, + { + "epoch": 4.45, + "learning_rate": 8.676753902329553e-06, + "loss": 0.0133, + "step": 33955 + }, + { + "epoch": 4.45, + "learning_rate": 8.676228242573104e-06, + "loss": 0.0207, + "step": 33956 + }, + { + "epoch": 4.45, + "learning_rate": 8.675702586539692e-06, + "loss": 0.0177, + "step": 33957 + }, + { + "epoch": 4.45, + "learning_rate": 8.675176934230799e-06, + "loss": 0.0187, + "step": 33958 + }, + { + "epoch": 4.45, + "learning_rate": 8.674651285647895e-06, + "loss": 0.018, + "step": 33959 + }, + { + "epoch": 4.45, + "learning_rate": 8.674125640792467e-06, + "loss": 0.0208, + "step": 33960 + }, + { + "epoch": 4.45, + "learning_rate": 8.67359999966599e-06, + "loss": 0.0197, + "step": 33961 + }, + { + "epoch": 4.45, + "learning_rate": 8.673074362269943e-06, + "loss": 0.0234, + "step": 33962 + }, + { + "epoch": 4.45, + "learning_rate": 8.672548728605805e-06, + "loss": 0.0179, + "step": 33963 + }, + { + "epoch": 4.45, + "learning_rate": 8.67202309867505e-06, + "loss": 0.0201, + "step": 33964 + }, + { + "epoch": 4.45, + "learning_rate": 8.67149747247916e-06, + "loss": 0.0166, + "step": 33965 + }, + { + "epoch": 4.45, + "learning_rate": 8.670971850019614e-06, + "loss": 0.0194, + "step": 33966 + }, + { + "epoch": 4.45, + "learning_rate": 8.670446231297888e-06, + "loss": 0.0237, + "step": 33967 + }, + { + "epoch": 4.45, + "learning_rate": 8.669920616315463e-06, + "loss": 0.0228, + "step": 33968 + }, + { + "epoch": 4.45, + "learning_rate": 8.669395005073812e-06, + "loss": 0.0215, + "step": 33969 + }, + { + "epoch": 4.45, + "learning_rate": 8.668869397574415e-06, + "loss": 0.024, + "step": 33970 + }, + { + "epoch": 4.45, + "learning_rate": 8.668343793818755e-06, + "loss": 0.0225, + "step": 33971 + }, + { + "epoch": 4.45, + "learning_rate": 8.667818193808309e-06, + "loss": 0.0166, + "step": 33972 + }, + { + "epoch": 4.45, + "learning_rate": 8.667292597544552e-06, + "loss": 0.0206, + "step": 33973 + }, + { + "epoch": 4.45, + "learning_rate": 8.666767005028965e-06, + "loss": 0.0249, + "step": 33974 + }, + { + "epoch": 4.45, + "learning_rate": 8.666241416263021e-06, + "loss": 0.0219, + "step": 33975 + }, + { + "epoch": 4.45, + "learning_rate": 8.665715831248205e-06, + "loss": 0.0224, + "step": 33976 + }, + { + "epoch": 4.45, + "learning_rate": 8.665190249985991e-06, + "loss": 0.021, + "step": 33977 + }, + { + "epoch": 4.45, + "learning_rate": 8.66466467247786e-06, + "loss": 0.0203, + "step": 33978 + }, + { + "epoch": 4.45, + "learning_rate": 8.664139098725289e-06, + "loss": 0.022, + "step": 33979 + }, + { + "epoch": 4.45, + "learning_rate": 8.66361352872975e-06, + "loss": 0.0269, + "step": 33980 + }, + { + "epoch": 4.45, + "learning_rate": 8.66308796249273e-06, + "loss": 0.0216, + "step": 33981 + }, + { + "epoch": 4.45, + "learning_rate": 8.662562400015706e-06, + "loss": 0.019, + "step": 33982 + }, + { + "epoch": 4.45, + "learning_rate": 8.662036841300152e-06, + "loss": 0.0201, + "step": 33983 + }, + { + "epoch": 4.45, + "learning_rate": 8.66151128634755e-06, + "loss": 0.0217, + "step": 33984 + }, + { + "epoch": 4.45, + "learning_rate": 8.660985735159375e-06, + "loss": 0.0205, + "step": 33985 + }, + { + "epoch": 4.45, + "learning_rate": 8.660460187737104e-06, + "loss": 0.0221, + "step": 33986 + }, + { + "epoch": 4.45, + "learning_rate": 8.659934644082221e-06, + "loss": 0.0178, + "step": 33987 + }, + { + "epoch": 4.45, + "learning_rate": 8.6594091041962e-06, + "loss": 0.0204, + "step": 33988 + }, + { + "epoch": 4.45, + "learning_rate": 8.658883568080518e-06, + "loss": 0.0194, + "step": 33989 + }, + { + "epoch": 4.45, + "learning_rate": 8.658358035736656e-06, + "loss": 0.0208, + "step": 33990 + }, + { + "epoch": 4.45, + "learning_rate": 8.657832507166086e-06, + "loss": 0.023, + "step": 33991 + }, + { + "epoch": 4.45, + "learning_rate": 8.657306982370293e-06, + "loss": 0.0169, + "step": 33992 + }, + { + "epoch": 4.45, + "learning_rate": 8.656781461350757e-06, + "loss": 0.0219, + "step": 33993 + }, + { + "epoch": 4.45, + "learning_rate": 8.656255944108946e-06, + "loss": 0.0226, + "step": 33994 + }, + { + "epoch": 4.45, + "learning_rate": 8.65573043064635e-06, + "loss": 0.0169, + "step": 33995 + }, + { + "epoch": 4.45, + "learning_rate": 8.655204920964435e-06, + "loss": 0.0165, + "step": 33996 + }, + { + "epoch": 4.45, + "learning_rate": 8.654679415064687e-06, + "loss": 0.0163, + "step": 33997 + }, + { + "epoch": 4.45, + "learning_rate": 8.65415391294858e-06, + "loss": 0.0198, + "step": 33998 + }, + { + "epoch": 4.45, + "learning_rate": 8.653628414617596e-06, + "loss": 0.018, + "step": 33999 + }, + { + "epoch": 4.45, + "learning_rate": 8.653102920073207e-06, + "loss": 0.0241, + "step": 34000 + }, + { + "epoch": 4.45, + "learning_rate": 8.652577429316895e-06, + "loss": 0.0197, + "step": 34001 + }, + { + "epoch": 4.45, + "learning_rate": 8.652051942350136e-06, + "loss": 0.0211, + "step": 34002 + }, + { + "epoch": 4.45, + "learning_rate": 8.651526459174414e-06, + "loss": 0.02, + "step": 34003 + }, + { + "epoch": 4.45, + "learning_rate": 8.651000979791198e-06, + "loss": 0.0189, + "step": 34004 + }, + { + "epoch": 4.45, + "learning_rate": 8.650475504201972e-06, + "loss": 0.0228, + "step": 34005 + }, + { + "epoch": 4.45, + "learning_rate": 8.64995003240821e-06, + "loss": 0.0267, + "step": 34006 + }, + { + "epoch": 4.45, + "learning_rate": 8.649424564411393e-06, + "loss": 0.0205, + "step": 34007 + }, + { + "epoch": 4.45, + "learning_rate": 8.648899100212997e-06, + "loss": 0.0179, + "step": 34008 + }, + { + "epoch": 4.45, + "learning_rate": 8.6483736398145e-06, + "loss": 0.0242, + "step": 34009 + }, + { + "epoch": 4.46, + "learning_rate": 8.647848183217382e-06, + "loss": 0.0188, + "step": 34010 + }, + { + "epoch": 4.46, + "learning_rate": 8.647322730423117e-06, + "loss": 0.0208, + "step": 34011 + }, + { + "epoch": 4.46, + "learning_rate": 8.646797281433188e-06, + "loss": 0.0192, + "step": 34012 + }, + { + "epoch": 4.46, + "learning_rate": 8.646271836249064e-06, + "loss": 0.0178, + "step": 34013 + }, + { + "epoch": 4.46, + "learning_rate": 8.64574639487223e-06, + "loss": 0.0225, + "step": 34014 + }, + { + "epoch": 4.46, + "learning_rate": 8.645220957304163e-06, + "loss": 0.0204, + "step": 34015 + }, + { + "epoch": 4.46, + "learning_rate": 8.644695523546342e-06, + "loss": 0.0209, + "step": 34016 + }, + { + "epoch": 4.46, + "learning_rate": 8.64417009360024e-06, + "loss": 0.0164, + "step": 34017 + }, + { + "epoch": 4.46, + "learning_rate": 8.64364466746734e-06, + "loss": 0.0217, + "step": 34018 + }, + { + "epoch": 4.46, + "learning_rate": 8.643119245149115e-06, + "loss": 0.0201, + "step": 34019 + }, + { + "epoch": 4.46, + "learning_rate": 8.642593826647046e-06, + "loss": 0.0196, + "step": 34020 + }, + { + "epoch": 4.46, + "learning_rate": 8.64206841196261e-06, + "loss": 0.0193, + "step": 34021 + }, + { + "epoch": 4.46, + "learning_rate": 8.641543001097282e-06, + "loss": 0.0235, + "step": 34022 + }, + { + "epoch": 4.46, + "learning_rate": 8.641017594052545e-06, + "loss": 0.0247, + "step": 34023 + }, + { + "epoch": 4.46, + "learning_rate": 8.640492190829868e-06, + "loss": 0.0218, + "step": 34024 + }, + { + "epoch": 4.46, + "learning_rate": 8.639966791430739e-06, + "loss": 0.0218, + "step": 34025 + }, + { + "epoch": 4.46, + "learning_rate": 8.63944139585663e-06, + "loss": 0.0192, + "step": 34026 + }, + { + "epoch": 4.46, + "learning_rate": 8.63891600410902e-06, + "loss": 0.0236, + "step": 34027 + }, + { + "epoch": 4.46, + "learning_rate": 8.638390616189386e-06, + "loss": 0.0214, + "step": 34028 + }, + { + "epoch": 4.46, + "learning_rate": 8.637865232099206e-06, + "loss": 0.0219, + "step": 34029 + }, + { + "epoch": 4.46, + "learning_rate": 8.637339851839957e-06, + "loss": 0.024, + "step": 34030 + }, + { + "epoch": 4.46, + "learning_rate": 8.636814475413118e-06, + "loss": 0.0185, + "step": 34031 + }, + { + "epoch": 4.46, + "learning_rate": 8.636289102820166e-06, + "loss": 0.0205, + "step": 34032 + }, + { + "epoch": 4.46, + "learning_rate": 8.635763734062579e-06, + "loss": 0.0239, + "step": 34033 + }, + { + "epoch": 4.46, + "learning_rate": 8.635238369141831e-06, + "loss": 0.0179, + "step": 34034 + }, + { + "epoch": 4.46, + "learning_rate": 8.634713008059401e-06, + "loss": 0.0225, + "step": 34035 + }, + { + "epoch": 4.46, + "learning_rate": 8.63418765081677e-06, + "loss": 0.0192, + "step": 34036 + }, + { + "epoch": 4.46, + "learning_rate": 8.633662297415415e-06, + "loss": 0.0265, + "step": 34037 + }, + { + "epoch": 4.46, + "learning_rate": 8.633136947856811e-06, + "loss": 0.0171, + "step": 34038 + }, + { + "epoch": 4.46, + "learning_rate": 8.632611602142438e-06, + "loss": 0.0205, + "step": 34039 + }, + { + "epoch": 4.46, + "learning_rate": 8.632086260273772e-06, + "loss": 0.0193, + "step": 34040 + }, + { + "epoch": 4.46, + "learning_rate": 8.63156092225229e-06, + "loss": 0.0217, + "step": 34041 + }, + { + "epoch": 4.46, + "learning_rate": 8.631035588079468e-06, + "loss": 0.0164, + "step": 34042 + }, + { + "epoch": 4.46, + "learning_rate": 8.630510257756787e-06, + "loss": 0.0249, + "step": 34043 + }, + { + "epoch": 4.46, + "learning_rate": 8.629984931285724e-06, + "loss": 0.0204, + "step": 34044 + }, + { + "epoch": 4.46, + "learning_rate": 8.629459608667754e-06, + "loss": 0.0179, + "step": 34045 + }, + { + "epoch": 4.46, + "learning_rate": 8.628934289904354e-06, + "loss": 0.0219, + "step": 34046 + }, + { + "epoch": 4.46, + "learning_rate": 8.628408974997005e-06, + "loss": 0.0173, + "step": 34047 + }, + { + "epoch": 4.46, + "learning_rate": 8.627883663947183e-06, + "loss": 0.0183, + "step": 34048 + }, + { + "epoch": 4.46, + "learning_rate": 8.627358356756367e-06, + "loss": 0.0199, + "step": 34049 + }, + { + "epoch": 4.46, + "learning_rate": 8.626833053426031e-06, + "loss": 0.0206, + "step": 34050 + }, + { + "epoch": 4.46, + "learning_rate": 8.626307753957652e-06, + "loss": 0.0178, + "step": 34051 + }, + { + "epoch": 4.46, + "learning_rate": 8.625782458352713e-06, + "loss": 0.0219, + "step": 34052 + }, + { + "epoch": 4.46, + "learning_rate": 8.625257166612684e-06, + "loss": 0.0268, + "step": 34053 + }, + { + "epoch": 4.46, + "learning_rate": 8.62473187873905e-06, + "loss": 0.0165, + "step": 34054 + }, + { + "epoch": 4.46, + "learning_rate": 8.624206594733282e-06, + "loss": 0.0232, + "step": 34055 + }, + { + "epoch": 4.46, + "learning_rate": 8.62368131459686e-06, + "loss": 0.0211, + "step": 34056 + }, + { + "epoch": 4.46, + "learning_rate": 8.623156038331257e-06, + "loss": 0.0222, + "step": 34057 + }, + { + "epoch": 4.46, + "learning_rate": 8.62263076593796e-06, + "loss": 0.0185, + "step": 34058 + }, + { + "epoch": 4.46, + "learning_rate": 8.622105497418437e-06, + "loss": 0.0268, + "step": 34059 + }, + { + "epoch": 4.46, + "learning_rate": 8.621580232774172e-06, + "loss": 0.0202, + "step": 34060 + }, + { + "epoch": 4.46, + "learning_rate": 8.621054972006637e-06, + "loss": 0.0238, + "step": 34061 + }, + { + "epoch": 4.46, + "learning_rate": 8.620529715117314e-06, + "loss": 0.022, + "step": 34062 + }, + { + "epoch": 4.46, + "learning_rate": 8.620004462107676e-06, + "loss": 0.023, + "step": 34063 + }, + { + "epoch": 4.46, + "learning_rate": 8.619479212979203e-06, + "loss": 0.0208, + "step": 34064 + }, + { + "epoch": 4.46, + "learning_rate": 8.61895396773337e-06, + "loss": 0.0172, + "step": 34065 + }, + { + "epoch": 4.46, + "learning_rate": 8.618428726371654e-06, + "loss": 0.0172, + "step": 34066 + }, + { + "epoch": 4.46, + "learning_rate": 8.617903488895537e-06, + "loss": 0.0218, + "step": 34067 + }, + { + "epoch": 4.46, + "learning_rate": 8.617378255306487e-06, + "loss": 0.022, + "step": 34068 + }, + { + "epoch": 4.46, + "learning_rate": 8.616853025605991e-06, + "loss": 0.0193, + "step": 34069 + }, + { + "epoch": 4.46, + "learning_rate": 8.616327799795524e-06, + "loss": 0.0196, + "step": 34070 + }, + { + "epoch": 4.46, + "learning_rate": 8.61580257787656e-06, + "loss": 0.0229, + "step": 34071 + }, + { + "epoch": 4.46, + "learning_rate": 8.615277359850576e-06, + "loss": 0.0189, + "step": 34072 + }, + { + "epoch": 4.46, + "learning_rate": 8.614752145719053e-06, + "loss": 0.0252, + "step": 34073 + }, + { + "epoch": 4.46, + "learning_rate": 8.614226935483464e-06, + "loss": 0.0201, + "step": 34074 + }, + { + "epoch": 4.46, + "learning_rate": 8.61370172914529e-06, + "loss": 0.0235, + "step": 34075 + }, + { + "epoch": 4.46, + "learning_rate": 8.613176526706003e-06, + "loss": 0.0176, + "step": 34076 + }, + { + "epoch": 4.46, + "learning_rate": 8.612651328167087e-06, + "loss": 0.0219, + "step": 34077 + }, + { + "epoch": 4.46, + "learning_rate": 8.612126133530014e-06, + "loss": 0.0226, + "step": 34078 + }, + { + "epoch": 4.46, + "learning_rate": 8.611600942796258e-06, + "loss": 0.0209, + "step": 34079 + }, + { + "epoch": 4.46, + "learning_rate": 8.611075755967303e-06, + "loss": 0.0187, + "step": 34080 + }, + { + "epoch": 4.46, + "learning_rate": 8.610550573044626e-06, + "loss": 0.0231, + "step": 34081 + }, + { + "epoch": 4.46, + "learning_rate": 8.610025394029699e-06, + "loss": 0.0217, + "step": 34082 + }, + { + "epoch": 4.46, + "learning_rate": 8.609500218924005e-06, + "loss": 0.0239, + "step": 34083 + }, + { + "epoch": 4.46, + "learning_rate": 8.608975047729014e-06, + "loss": 0.0214, + "step": 34084 + }, + { + "epoch": 4.46, + "learning_rate": 8.608449880446208e-06, + "loss": 0.0191, + "step": 34085 + }, + { + "epoch": 4.47, + "learning_rate": 8.607924717077061e-06, + "loss": 0.0235, + "step": 34086 + }, + { + "epoch": 4.47, + "learning_rate": 8.607399557623054e-06, + "loss": 0.0224, + "step": 34087 + }, + { + "epoch": 4.47, + "learning_rate": 8.60687440208566e-06, + "loss": 0.0198, + "step": 34088 + }, + { + "epoch": 4.47, + "learning_rate": 8.606349250466358e-06, + "loss": 0.0252, + "step": 34089 + }, + { + "epoch": 4.47, + "learning_rate": 8.605824102766622e-06, + "loss": 0.0191, + "step": 34090 + }, + { + "epoch": 4.47, + "learning_rate": 8.605298958987935e-06, + "loss": 0.0208, + "step": 34091 + }, + { + "epoch": 4.47, + "learning_rate": 8.604773819131769e-06, + "loss": 0.0224, + "step": 34092 + }, + { + "epoch": 4.47, + "learning_rate": 8.604248683199602e-06, + "loss": 0.0193, + "step": 34093 + }, + { + "epoch": 4.47, + "learning_rate": 8.603723551192912e-06, + "loss": 0.02, + "step": 34094 + }, + { + "epoch": 4.47, + "learning_rate": 8.603198423113175e-06, + "loss": 0.0182, + "step": 34095 + }, + { + "epoch": 4.47, + "learning_rate": 8.60267329896187e-06, + "loss": 0.0232, + "step": 34096 + }, + { + "epoch": 4.47, + "learning_rate": 8.602148178740469e-06, + "loss": 0.0161, + "step": 34097 + }, + { + "epoch": 4.47, + "learning_rate": 8.601623062450455e-06, + "loss": 0.0175, + "step": 34098 + }, + { + "epoch": 4.47, + "learning_rate": 8.601097950093299e-06, + "loss": 0.0233, + "step": 34099 + }, + { + "epoch": 4.47, + "learning_rate": 8.600572841670482e-06, + "loss": 0.0177, + "step": 34100 + }, + { + "epoch": 4.47, + "learning_rate": 8.600047737183473e-06, + "loss": 0.0177, + "step": 34101 + }, + { + "epoch": 4.47, + "learning_rate": 8.599522636633763e-06, + "loss": 0.0242, + "step": 34102 + }, + { + "epoch": 4.47, + "learning_rate": 8.598997540022818e-06, + "loss": 0.0206, + "step": 34103 + }, + { + "epoch": 4.47, + "learning_rate": 8.59847244735212e-06, + "loss": 0.018, + "step": 34104 + }, + { + "epoch": 4.47, + "learning_rate": 8.597947358623141e-06, + "loss": 0.018, + "step": 34105 + }, + { + "epoch": 4.47, + "learning_rate": 8.597422273837363e-06, + "loss": 0.0162, + "step": 34106 + }, + { + "epoch": 4.47, + "learning_rate": 8.596897192996258e-06, + "loss": 0.0165, + "step": 34107 + }, + { + "epoch": 4.47, + "learning_rate": 8.596372116101305e-06, + "loss": 0.0215, + "step": 34108 + }, + { + "epoch": 4.47, + "learning_rate": 8.595847043153983e-06, + "loss": 0.0187, + "step": 34109 + }, + { + "epoch": 4.47, + "learning_rate": 8.595321974155763e-06, + "loss": 0.0165, + "step": 34110 + }, + { + "epoch": 4.47, + "learning_rate": 8.594796909108125e-06, + "loss": 0.0226, + "step": 34111 + }, + { + "epoch": 4.47, + "learning_rate": 8.594271848012548e-06, + "loss": 0.023, + "step": 34112 + }, + { + "epoch": 4.47, + "learning_rate": 8.593746790870505e-06, + "loss": 0.0212, + "step": 34113 + }, + { + "epoch": 4.47, + "learning_rate": 8.593221737683477e-06, + "loss": 0.0228, + "step": 34114 + }, + { + "epoch": 4.47, + "learning_rate": 8.592696688452935e-06, + "loss": 0.0175, + "step": 34115 + }, + { + "epoch": 4.47, + "learning_rate": 8.59217164318036e-06, + "loss": 0.0221, + "step": 34116 + }, + { + "epoch": 4.47, + "learning_rate": 8.59164660186723e-06, + "loss": 0.0188, + "step": 34117 + }, + { + "epoch": 4.47, + "learning_rate": 8.591121564515015e-06, + "loss": 0.0204, + "step": 34118 + }, + { + "epoch": 4.47, + "learning_rate": 8.590596531125199e-06, + "loss": 0.02, + "step": 34119 + }, + { + "epoch": 4.47, + "learning_rate": 8.590071501699253e-06, + "loss": 0.0262, + "step": 34120 + }, + { + "epoch": 4.47, + "learning_rate": 8.589546476238656e-06, + "loss": 0.0209, + "step": 34121 + }, + { + "epoch": 4.47, + "learning_rate": 8.589021454744879e-06, + "loss": 0.0228, + "step": 34122 + }, + { + "epoch": 4.47, + "learning_rate": 8.588496437219412e-06, + "loss": 0.0232, + "step": 34123 + }, + { + "epoch": 4.47, + "learning_rate": 8.587971423663719e-06, + "loss": 0.0202, + "step": 34124 + }, + { + "epoch": 4.47, + "learning_rate": 8.587446414079284e-06, + "loss": 0.0193, + "step": 34125 + }, + { + "epoch": 4.47, + "learning_rate": 8.58692140846758e-06, + "loss": 0.0227, + "step": 34126 + }, + { + "epoch": 4.47, + "learning_rate": 8.586396406830086e-06, + "loss": 0.0272, + "step": 34127 + }, + { + "epoch": 4.47, + "learning_rate": 8.585871409168273e-06, + "loss": 0.0212, + "step": 34128 + }, + { + "epoch": 4.47, + "learning_rate": 8.585346415483623e-06, + "loss": 0.0237, + "step": 34129 + }, + { + "epoch": 4.47, + "learning_rate": 8.584821425777613e-06, + "loss": 0.0259, + "step": 34130 + }, + { + "epoch": 4.47, + "learning_rate": 8.584296440051715e-06, + "loss": 0.0206, + "step": 34131 + }, + { + "epoch": 4.47, + "learning_rate": 8.58377145830741e-06, + "loss": 0.0198, + "step": 34132 + }, + { + "epoch": 4.47, + "learning_rate": 8.583246480546167e-06, + "loss": 0.0178, + "step": 34133 + }, + { + "epoch": 4.47, + "learning_rate": 8.582721506769471e-06, + "loss": 0.0247, + "step": 34134 + }, + { + "epoch": 4.47, + "learning_rate": 8.582196536978798e-06, + "loss": 0.0204, + "step": 34135 + }, + { + "epoch": 4.47, + "learning_rate": 8.581671571175618e-06, + "loss": 0.0203, + "step": 34136 + }, + { + "epoch": 4.47, + "learning_rate": 8.581146609361412e-06, + "loss": 0.0215, + "step": 34137 + }, + { + "epoch": 4.47, + "learning_rate": 8.580621651537658e-06, + "loss": 0.0224, + "step": 34138 + }, + { + "epoch": 4.47, + "learning_rate": 8.580096697705828e-06, + "loss": 0.0262, + "step": 34139 + }, + { + "epoch": 4.47, + "learning_rate": 8.579571747867402e-06, + "loss": 0.0165, + "step": 34140 + }, + { + "epoch": 4.47, + "learning_rate": 8.579046802023854e-06, + "loss": 0.0234, + "step": 34141 + }, + { + "epoch": 4.47, + "learning_rate": 8.578521860176662e-06, + "loss": 0.018, + "step": 34142 + }, + { + "epoch": 4.47, + "learning_rate": 8.577996922327299e-06, + "loss": 0.0196, + "step": 34143 + }, + { + "epoch": 4.47, + "learning_rate": 8.577471988477242e-06, + "loss": 0.0224, + "step": 34144 + }, + { + "epoch": 4.47, + "learning_rate": 8.576947058627973e-06, + "loss": 0.0203, + "step": 34145 + }, + { + "epoch": 4.47, + "learning_rate": 8.576422132780964e-06, + "loss": 0.0182, + "step": 34146 + }, + { + "epoch": 4.47, + "learning_rate": 8.575897210937693e-06, + "loss": 0.019, + "step": 34147 + }, + { + "epoch": 4.47, + "learning_rate": 8.575372293099635e-06, + "loss": 0.016, + "step": 34148 + }, + { + "epoch": 4.47, + "learning_rate": 8.574847379268265e-06, + "loss": 0.0196, + "step": 34149 + }, + { + "epoch": 4.47, + "learning_rate": 8.574322469445064e-06, + "loss": 0.0205, + "step": 34150 + }, + { + "epoch": 4.47, + "learning_rate": 8.573797563631503e-06, + "loss": 0.0188, + "step": 34151 + }, + { + "epoch": 4.47, + "learning_rate": 8.57327266182906e-06, + "loss": 0.0243, + "step": 34152 + }, + { + "epoch": 4.47, + "learning_rate": 8.572747764039213e-06, + "loss": 0.0174, + "step": 34153 + }, + { + "epoch": 4.47, + "learning_rate": 8.572222870263435e-06, + "loss": 0.0215, + "step": 34154 + }, + { + "epoch": 4.47, + "learning_rate": 8.571697980503201e-06, + "loss": 0.023, + "step": 34155 + }, + { + "epoch": 4.47, + "learning_rate": 8.571173094759996e-06, + "loss": 0.0193, + "step": 34156 + }, + { + "epoch": 4.47, + "learning_rate": 8.570648213035288e-06, + "loss": 0.0215, + "step": 34157 + }, + { + "epoch": 4.47, + "learning_rate": 8.570123335330558e-06, + "loss": 0.0218, + "step": 34158 + }, + { + "epoch": 4.47, + "learning_rate": 8.56959846164728e-06, + "loss": 0.0224, + "step": 34159 + }, + { + "epoch": 4.47, + "learning_rate": 8.569073591986927e-06, + "loss": 0.0195, + "step": 34160 + }, + { + "epoch": 4.47, + "learning_rate": 8.568548726350982e-06, + "loss": 0.0184, + "step": 34161 + }, + { + "epoch": 4.47, + "learning_rate": 8.568023864740916e-06, + "loss": 0.0169, + "step": 34162 + }, + { + "epoch": 4.48, + "learning_rate": 8.567499007158207e-06, + "loss": 0.0217, + "step": 34163 + }, + { + "epoch": 4.48, + "learning_rate": 8.566974153604331e-06, + "loss": 0.0203, + "step": 34164 + }, + { + "epoch": 4.48, + "learning_rate": 8.566449304080765e-06, + "loss": 0.0159, + "step": 34165 + }, + { + "epoch": 4.48, + "learning_rate": 8.565924458588975e-06, + "loss": 0.0149, + "step": 34166 + }, + { + "epoch": 4.48, + "learning_rate": 8.565399617130455e-06, + "loss": 0.0225, + "step": 34167 + }, + { + "epoch": 4.48, + "learning_rate": 8.56487477970667e-06, + "loss": 0.0236, + "step": 34168 + }, + { + "epoch": 4.48, + "learning_rate": 8.5643499463191e-06, + "loss": 0.0182, + "step": 34169 + }, + { + "epoch": 4.48, + "learning_rate": 8.563825116969217e-06, + "loss": 0.0166, + "step": 34170 + }, + { + "epoch": 4.48, + "learning_rate": 8.5633002916585e-06, + "loss": 0.024, + "step": 34171 + }, + { + "epoch": 4.48, + "learning_rate": 8.562775470388424e-06, + "loss": 0.0211, + "step": 34172 + }, + { + "epoch": 4.48, + "learning_rate": 8.562250653160466e-06, + "loss": 0.0198, + "step": 34173 + }, + { + "epoch": 4.48, + "learning_rate": 8.561725839976104e-06, + "loss": 0.0154, + "step": 34174 + }, + { + "epoch": 4.48, + "learning_rate": 8.561201030836808e-06, + "loss": 0.018, + "step": 34175 + }, + { + "epoch": 4.48, + "learning_rate": 8.56067622574406e-06, + "loss": 0.0212, + "step": 34176 + }, + { + "epoch": 4.48, + "learning_rate": 8.560151424699329e-06, + "loss": 0.0234, + "step": 34177 + }, + { + "epoch": 4.48, + "learning_rate": 8.559626627704099e-06, + "loss": 0.0197, + "step": 34178 + }, + { + "epoch": 4.48, + "learning_rate": 8.55910183475984e-06, + "loss": 0.0191, + "step": 34179 + }, + { + "epoch": 4.48, + "learning_rate": 8.558577045868034e-06, + "loss": 0.0216, + "step": 34180 + }, + { + "epoch": 4.48, + "learning_rate": 8.55805226103015e-06, + "loss": 0.0197, + "step": 34181 + }, + { + "epoch": 4.48, + "learning_rate": 8.55752748024767e-06, + "loss": 0.0235, + "step": 34182 + }, + { + "epoch": 4.48, + "learning_rate": 8.557002703522066e-06, + "loss": 0.021, + "step": 34183 + }, + { + "epoch": 4.48, + "learning_rate": 8.556477930854817e-06, + "loss": 0.0204, + "step": 34184 + }, + { + "epoch": 4.48, + "learning_rate": 8.555953162247393e-06, + "loss": 0.0209, + "step": 34185 + }, + { + "epoch": 4.48, + "learning_rate": 8.555428397701277e-06, + "loss": 0.0188, + "step": 34186 + }, + { + "epoch": 4.48, + "learning_rate": 8.554903637217941e-06, + "loss": 0.02, + "step": 34187 + }, + { + "epoch": 4.48, + "learning_rate": 8.554378880798857e-06, + "loss": 0.0242, + "step": 34188 + }, + { + "epoch": 4.48, + "learning_rate": 8.55385412844551e-06, + "loss": 0.0167, + "step": 34189 + }, + { + "epoch": 4.48, + "learning_rate": 8.55332938015937e-06, + "loss": 0.0178, + "step": 34190 + }, + { + "epoch": 4.48, + "learning_rate": 8.552804635941915e-06, + "loss": 0.0195, + "step": 34191 + }, + { + "epoch": 4.48, + "learning_rate": 8.552279895794622e-06, + "loss": 0.0209, + "step": 34192 + }, + { + "epoch": 4.48, + "learning_rate": 8.55175515971896e-06, + "loss": 0.0237, + "step": 34193 + }, + { + "epoch": 4.48, + "learning_rate": 8.551230427716413e-06, + "loss": 0.0225, + "step": 34194 + }, + { + "epoch": 4.48, + "learning_rate": 8.550705699788453e-06, + "loss": 0.0201, + "step": 34195 + }, + { + "epoch": 4.48, + "learning_rate": 8.550180975936554e-06, + "loss": 0.0201, + "step": 34196 + }, + { + "epoch": 4.48, + "learning_rate": 8.549656256162196e-06, + "loss": 0.0229, + "step": 34197 + }, + { + "epoch": 4.48, + "learning_rate": 8.54913154046685e-06, + "loss": 0.0215, + "step": 34198 + }, + { + "epoch": 4.48, + "learning_rate": 8.548606828851994e-06, + "loss": 0.017, + "step": 34199 + }, + { + "epoch": 4.48, + "learning_rate": 8.548082121319107e-06, + "loss": 0.023, + "step": 34200 + }, + { + "epoch": 4.48, + "learning_rate": 8.547557417869659e-06, + "loss": 0.0236, + "step": 34201 + }, + { + "epoch": 4.48, + "learning_rate": 8.54703271850513e-06, + "loss": 0.0199, + "step": 34202 + }, + { + "epoch": 4.48, + "learning_rate": 8.546508023226996e-06, + "loss": 0.0194, + "step": 34203 + }, + { + "epoch": 4.48, + "learning_rate": 8.545983332036727e-06, + "loss": 0.021, + "step": 34204 + }, + { + "epoch": 4.48, + "learning_rate": 8.545458644935807e-06, + "loss": 0.016, + "step": 34205 + }, + { + "epoch": 4.48, + "learning_rate": 8.544933961925703e-06, + "loss": 0.0214, + "step": 34206 + }, + { + "epoch": 4.48, + "learning_rate": 8.544409283007898e-06, + "loss": 0.0184, + "step": 34207 + }, + { + "epoch": 4.48, + "learning_rate": 8.543884608183862e-06, + "loss": 0.0219, + "step": 34208 + }, + { + "epoch": 4.48, + "learning_rate": 8.543359937455071e-06, + "loss": 0.0202, + "step": 34209 + }, + { + "epoch": 4.48, + "learning_rate": 8.542835270823003e-06, + "loss": 0.0222, + "step": 34210 + }, + { + "epoch": 4.48, + "learning_rate": 8.542310608289137e-06, + "loss": 0.02, + "step": 34211 + }, + { + "epoch": 4.48, + "learning_rate": 8.541785949854942e-06, + "loss": 0.0201, + "step": 34212 + }, + { + "epoch": 4.48, + "learning_rate": 8.541261295521898e-06, + "loss": 0.0225, + "step": 34213 + }, + { + "epoch": 4.48, + "learning_rate": 8.54073664529148e-06, + "loss": 0.0149, + "step": 34214 + }, + { + "epoch": 4.48, + "learning_rate": 8.54021199916516e-06, + "loss": 0.0195, + "step": 34215 + }, + { + "epoch": 4.48, + "learning_rate": 8.539687357144417e-06, + "loss": 0.0193, + "step": 34216 + }, + { + "epoch": 4.48, + "learning_rate": 8.539162719230724e-06, + "loss": 0.0275, + "step": 34217 + }, + { + "epoch": 4.48, + "learning_rate": 8.538638085425562e-06, + "loss": 0.0239, + "step": 34218 + }, + { + "epoch": 4.48, + "learning_rate": 8.538113455730398e-06, + "loss": 0.0201, + "step": 34219 + }, + { + "epoch": 4.48, + "learning_rate": 8.537588830146715e-06, + "loss": 0.0201, + "step": 34220 + }, + { + "epoch": 4.48, + "learning_rate": 8.537064208675982e-06, + "loss": 0.0214, + "step": 34221 + }, + { + "epoch": 4.48, + "learning_rate": 8.536539591319681e-06, + "loss": 0.0202, + "step": 34222 + }, + { + "epoch": 4.48, + "learning_rate": 8.536014978079282e-06, + "loss": 0.0172, + "step": 34223 + }, + { + "epoch": 4.48, + "learning_rate": 8.535490368956267e-06, + "loss": 0.017, + "step": 34224 + }, + { + "epoch": 4.48, + "learning_rate": 8.534965763952105e-06, + "loss": 0.0211, + "step": 34225 + }, + { + "epoch": 4.48, + "learning_rate": 8.534441163068275e-06, + "loss": 0.0221, + "step": 34226 + }, + { + "epoch": 4.48, + "learning_rate": 8.533916566306249e-06, + "loss": 0.0199, + "step": 34227 + }, + { + "epoch": 4.48, + "learning_rate": 8.53339197366751e-06, + "loss": 0.0186, + "step": 34228 + }, + { + "epoch": 4.48, + "learning_rate": 8.53286738515352e-06, + "loss": 0.0211, + "step": 34229 + }, + { + "epoch": 4.48, + "learning_rate": 8.532342800765768e-06, + "loss": 0.0171, + "step": 34230 + }, + { + "epoch": 4.48, + "learning_rate": 8.531818220505722e-06, + "loss": 0.0206, + "step": 34231 + }, + { + "epoch": 4.48, + "learning_rate": 8.531293644374856e-06, + "loss": 0.0207, + "step": 34232 + }, + { + "epoch": 4.48, + "learning_rate": 8.53076907237465e-06, + "loss": 0.0229, + "step": 34233 + }, + { + "epoch": 4.48, + "learning_rate": 8.530244504506581e-06, + "loss": 0.0151, + "step": 34234 + }, + { + "epoch": 4.48, + "learning_rate": 8.52971994077212e-06, + "loss": 0.0211, + "step": 34235 + }, + { + "epoch": 4.48, + "learning_rate": 8.529195381172742e-06, + "loss": 0.0194, + "step": 34236 + }, + { + "epoch": 4.48, + "learning_rate": 8.528670825709924e-06, + "loss": 0.0165, + "step": 34237 + }, + { + "epoch": 4.48, + "learning_rate": 8.52814627438514e-06, + "loss": 0.021, + "step": 34238 + }, + { + "epoch": 4.49, + "learning_rate": 8.52762172719987e-06, + "loss": 0.0173, + "step": 34239 + }, + { + "epoch": 4.49, + "learning_rate": 8.527097184155581e-06, + "loss": 0.0218, + "step": 34240 + }, + { + "epoch": 4.49, + "learning_rate": 8.526572645253757e-06, + "loss": 0.0238, + "step": 34241 + }, + { + "epoch": 4.49, + "learning_rate": 8.526048110495862e-06, + "loss": 0.0223, + "step": 34242 + }, + { + "epoch": 4.49, + "learning_rate": 8.525523579883383e-06, + "loss": 0.0217, + "step": 34243 + }, + { + "epoch": 4.49, + "learning_rate": 8.52499905341779e-06, + "loss": 0.019, + "step": 34244 + }, + { + "epoch": 4.49, + "learning_rate": 8.52447453110056e-06, + "loss": 0.0207, + "step": 34245 + }, + { + "epoch": 4.49, + "learning_rate": 8.523950012933165e-06, + "loss": 0.0192, + "step": 34246 + }, + { + "epoch": 4.49, + "learning_rate": 8.523425498917084e-06, + "loss": 0.026, + "step": 34247 + }, + { + "epoch": 4.49, + "learning_rate": 8.522900989053787e-06, + "loss": 0.0183, + "step": 34248 + }, + { + "epoch": 4.49, + "learning_rate": 8.522376483344756e-06, + "loss": 0.0172, + "step": 34249 + }, + { + "epoch": 4.49, + "learning_rate": 8.52185198179146e-06, + "loss": 0.0252, + "step": 34250 + }, + { + "epoch": 4.49, + "learning_rate": 8.521327484395379e-06, + "loss": 0.0237, + "step": 34251 + }, + { + "epoch": 4.49, + "learning_rate": 8.520802991157982e-06, + "loss": 0.0189, + "step": 34252 + }, + { + "epoch": 4.49, + "learning_rate": 8.520278502080747e-06, + "loss": 0.021, + "step": 34253 + }, + { + "epoch": 4.49, + "learning_rate": 8.519754017165152e-06, + "loss": 0.0154, + "step": 34254 + }, + { + "epoch": 4.49, + "learning_rate": 8.519229536412673e-06, + "loss": 0.0221, + "step": 34255 + }, + { + "epoch": 4.49, + "learning_rate": 8.518705059824777e-06, + "loss": 0.0193, + "step": 34256 + }, + { + "epoch": 4.49, + "learning_rate": 8.518180587402949e-06, + "loss": 0.0188, + "step": 34257 + }, + { + "epoch": 4.49, + "learning_rate": 8.517656119148656e-06, + "loss": 0.0205, + "step": 34258 + }, + { + "epoch": 4.49, + "learning_rate": 8.517131655063377e-06, + "loss": 0.0206, + "step": 34259 + }, + { + "epoch": 4.49, + "learning_rate": 8.516607195148588e-06, + "loss": 0.0227, + "step": 34260 + }, + { + "epoch": 4.49, + "learning_rate": 8.516082739405758e-06, + "loss": 0.0232, + "step": 34261 + }, + { + "epoch": 4.49, + "learning_rate": 8.515558287836372e-06, + "loss": 0.0186, + "step": 34262 + }, + { + "epoch": 4.49, + "learning_rate": 8.515033840441895e-06, + "loss": 0.0208, + "step": 34263 + }, + { + "epoch": 4.49, + "learning_rate": 8.514509397223804e-06, + "loss": 0.0165, + "step": 34264 + }, + { + "epoch": 4.49, + "learning_rate": 8.513984958183582e-06, + "loss": 0.0211, + "step": 34265 + }, + { + "epoch": 4.49, + "learning_rate": 8.513460523322694e-06, + "loss": 0.0271, + "step": 34266 + }, + { + "epoch": 4.49, + "learning_rate": 8.51293609264262e-06, + "loss": 0.02, + "step": 34267 + }, + { + "epoch": 4.49, + "learning_rate": 8.512411666144838e-06, + "loss": 0.018, + "step": 34268 + }, + { + "epoch": 4.49, + "learning_rate": 8.511887243830814e-06, + "loss": 0.0251, + "step": 34269 + }, + { + "epoch": 4.49, + "learning_rate": 8.511362825702032e-06, + "loss": 0.0213, + "step": 34270 + }, + { + "epoch": 4.49, + "learning_rate": 8.51083841175996e-06, + "loss": 0.0226, + "step": 34271 + }, + { + "epoch": 4.49, + "learning_rate": 8.510314002006078e-06, + "loss": 0.0201, + "step": 34272 + }, + { + "epoch": 4.49, + "learning_rate": 8.509789596441856e-06, + "loss": 0.0238, + "step": 34273 + }, + { + "epoch": 4.49, + "learning_rate": 8.509265195068772e-06, + "loss": 0.0169, + "step": 34274 + }, + { + "epoch": 4.49, + "learning_rate": 8.508740797888297e-06, + "loss": 0.0182, + "step": 34275 + }, + { + "epoch": 4.49, + "learning_rate": 8.508216404901913e-06, + "loss": 0.0188, + "step": 34276 + }, + { + "epoch": 4.49, + "learning_rate": 8.50769201611109e-06, + "loss": 0.0183, + "step": 34277 + }, + { + "epoch": 4.49, + "learning_rate": 8.507167631517306e-06, + "loss": 0.0196, + "step": 34278 + }, + { + "epoch": 4.49, + "learning_rate": 8.506643251122031e-06, + "loss": 0.0204, + "step": 34279 + }, + { + "epoch": 4.49, + "learning_rate": 8.506118874926744e-06, + "loss": 0.0256, + "step": 34280 + }, + { + "epoch": 4.49, + "learning_rate": 8.505594502932916e-06, + "loss": 0.0184, + "step": 34281 + }, + { + "epoch": 4.49, + "learning_rate": 8.505070135142025e-06, + "loss": 0.0194, + "step": 34282 + }, + { + "epoch": 4.49, + "learning_rate": 8.504545771555545e-06, + "loss": 0.0224, + "step": 34283 + }, + { + "epoch": 4.49, + "learning_rate": 8.504021412174949e-06, + "loss": 0.0186, + "step": 34284 + }, + { + "epoch": 4.49, + "learning_rate": 8.503497057001717e-06, + "loss": 0.0206, + "step": 34285 + }, + { + "epoch": 4.49, + "learning_rate": 8.502972706037313e-06, + "loss": 0.0184, + "step": 34286 + }, + { + "epoch": 4.49, + "learning_rate": 8.502448359283222e-06, + "loss": 0.0188, + "step": 34287 + }, + { + "epoch": 4.49, + "learning_rate": 8.501924016740916e-06, + "loss": 0.0167, + "step": 34288 + }, + { + "epoch": 4.49, + "learning_rate": 8.501399678411871e-06, + "loss": 0.0179, + "step": 34289 + }, + { + "epoch": 4.49, + "learning_rate": 8.500875344297555e-06, + "loss": 0.0185, + "step": 34290 + }, + { + "epoch": 4.49, + "learning_rate": 8.500351014399452e-06, + "loss": 0.0212, + "step": 34291 + }, + { + "epoch": 4.49, + "learning_rate": 8.499826688719029e-06, + "loss": 0.0239, + "step": 34292 + }, + { + "epoch": 4.49, + "learning_rate": 8.499302367257764e-06, + "loss": 0.019, + "step": 34293 + }, + { + "epoch": 4.49, + "learning_rate": 8.498778050017132e-06, + "loss": 0.0244, + "step": 34294 + }, + { + "epoch": 4.49, + "learning_rate": 8.498253736998605e-06, + "loss": 0.0229, + "step": 34295 + }, + { + "epoch": 4.49, + "learning_rate": 8.497729428203662e-06, + "loss": 0.0165, + "step": 34296 + }, + { + "epoch": 4.49, + "learning_rate": 8.49720512363377e-06, + "loss": 0.0209, + "step": 34297 + }, + { + "epoch": 4.49, + "learning_rate": 8.49668082329041e-06, + "loss": 0.023, + "step": 34298 + }, + { + "epoch": 4.49, + "learning_rate": 8.49615652717506e-06, + "loss": 0.0225, + "step": 34299 + }, + { + "epoch": 4.49, + "learning_rate": 8.495632235289186e-06, + "loss": 0.0194, + "step": 34300 + }, + { + "epoch": 4.49, + "learning_rate": 8.49510794763427e-06, + "loss": 0.0184, + "step": 34301 + }, + { + "epoch": 4.49, + "learning_rate": 8.494583664211777e-06, + "loss": 0.0216, + "step": 34302 + }, + { + "epoch": 4.49, + "learning_rate": 8.49405938502319e-06, + "loss": 0.0273, + "step": 34303 + }, + { + "epoch": 4.49, + "learning_rate": 8.493535110069982e-06, + "loss": 0.0193, + "step": 34304 + }, + { + "epoch": 4.49, + "learning_rate": 8.493010839353623e-06, + "loss": 0.0165, + "step": 34305 + }, + { + "epoch": 4.49, + "learning_rate": 8.492486572875596e-06, + "loss": 0.0217, + "step": 34306 + }, + { + "epoch": 4.49, + "learning_rate": 8.491962310637367e-06, + "loss": 0.0194, + "step": 34307 + }, + { + "epoch": 4.49, + "learning_rate": 8.49143805264041e-06, + "loss": 0.022, + "step": 34308 + }, + { + "epoch": 4.49, + "learning_rate": 8.490913798886206e-06, + "loss": 0.0217, + "step": 34309 + }, + { + "epoch": 4.49, + "learning_rate": 8.49038954937623e-06, + "loss": 0.0214, + "step": 34310 + }, + { + "epoch": 4.49, + "learning_rate": 8.489865304111949e-06, + "loss": 0.02, + "step": 34311 + }, + { + "epoch": 4.49, + "learning_rate": 8.489341063094845e-06, + "loss": 0.0265, + "step": 34312 + }, + { + "epoch": 4.49, + "learning_rate": 8.488816826326384e-06, + "loss": 0.0206, + "step": 34313 + }, + { + "epoch": 4.49, + "learning_rate": 8.488292593808051e-06, + "loss": 0.0157, + "step": 34314 + }, + { + "epoch": 4.5, + "learning_rate": 8.48776836554131e-06, + "loss": 0.0215, + "step": 34315 + }, + { + "epoch": 4.5, + "learning_rate": 8.487244141527643e-06, + "loss": 0.0142, + "step": 34316 + }, + { + "epoch": 4.5, + "learning_rate": 8.486719921768519e-06, + "loss": 0.0224, + "step": 34317 + }, + { + "epoch": 4.5, + "learning_rate": 8.486195706265414e-06, + "loss": 0.0188, + "step": 34318 + }, + { + "epoch": 4.5, + "learning_rate": 8.485671495019803e-06, + "loss": 0.0225, + "step": 34319 + }, + { + "epoch": 4.5, + "learning_rate": 8.485147288033162e-06, + "loss": 0.022, + "step": 34320 + }, + { + "epoch": 4.5, + "learning_rate": 8.484623085306963e-06, + "loss": 0.0192, + "step": 34321 + }, + { + "epoch": 4.5, + "learning_rate": 8.484098886842683e-06, + "loss": 0.0214, + "step": 34322 + }, + { + "epoch": 4.5, + "learning_rate": 8.48357469264179e-06, + "loss": 0.0151, + "step": 34323 + }, + { + "epoch": 4.5, + "learning_rate": 8.483050502705763e-06, + "loss": 0.0265, + "step": 34324 + }, + { + "epoch": 4.5, + "learning_rate": 8.48252631703608e-06, + "loss": 0.0205, + "step": 34325 + }, + { + "epoch": 4.5, + "learning_rate": 8.482002135634207e-06, + "loss": 0.0175, + "step": 34326 + }, + { + "epoch": 4.5, + "learning_rate": 8.481477958501626e-06, + "loss": 0.0197, + "step": 34327 + }, + { + "epoch": 4.5, + "learning_rate": 8.480953785639803e-06, + "loss": 0.0178, + "step": 34328 + }, + { + "epoch": 4.5, + "learning_rate": 8.480429617050221e-06, + "loss": 0.0175, + "step": 34329 + }, + { + "epoch": 4.5, + "learning_rate": 8.479905452734343e-06, + "loss": 0.0214, + "step": 34330 + }, + { + "epoch": 4.5, + "learning_rate": 8.479381292693653e-06, + "loss": 0.0215, + "step": 34331 + }, + { + "epoch": 4.5, + "learning_rate": 8.478857136929624e-06, + "loss": 0.021, + "step": 34332 + }, + { + "epoch": 4.5, + "learning_rate": 8.478332985443729e-06, + "loss": 0.0211, + "step": 34333 + }, + { + "epoch": 4.5, + "learning_rate": 8.477808838237439e-06, + "loss": 0.017, + "step": 34334 + }, + { + "epoch": 4.5, + "learning_rate": 8.477284695312233e-06, + "loss": 0.0192, + "step": 34335 + }, + { + "epoch": 4.5, + "learning_rate": 8.47676055666958e-06, + "loss": 0.018, + "step": 34336 + }, + { + "epoch": 4.5, + "learning_rate": 8.47623642231096e-06, + "loss": 0.0213, + "step": 34337 + }, + { + "epoch": 4.5, + "learning_rate": 8.475712292237841e-06, + "loss": 0.0148, + "step": 34338 + }, + { + "epoch": 4.5, + "learning_rate": 8.4751881664517e-06, + "loss": 0.0219, + "step": 34339 + }, + { + "epoch": 4.5, + "learning_rate": 8.474664044954014e-06, + "loss": 0.0181, + "step": 34340 + }, + { + "epoch": 4.5, + "learning_rate": 8.474139927746248e-06, + "loss": 0.0185, + "step": 34341 + }, + { + "epoch": 4.5, + "learning_rate": 8.473615814829887e-06, + "loss": 0.0202, + "step": 34342 + }, + { + "epoch": 4.5, + "learning_rate": 8.4730917062064e-06, + "loss": 0.0257, + "step": 34343 + }, + { + "epoch": 4.5, + "learning_rate": 8.472567601877262e-06, + "loss": 0.0234, + "step": 34344 + }, + { + "epoch": 4.5, + "learning_rate": 8.472043501843943e-06, + "loss": 0.0196, + "step": 34345 + }, + { + "epoch": 4.5, + "learning_rate": 8.471519406107924e-06, + "loss": 0.0224, + "step": 34346 + }, + { + "epoch": 4.5, + "learning_rate": 8.470995314670674e-06, + "loss": 0.0148, + "step": 34347 + }, + { + "epoch": 4.5, + "learning_rate": 8.47047122753367e-06, + "loss": 0.0227, + "step": 34348 + }, + { + "epoch": 4.5, + "learning_rate": 8.469947144698381e-06, + "loss": 0.0208, + "step": 34349 + }, + { + "epoch": 4.5, + "learning_rate": 8.469423066166287e-06, + "loss": 0.0162, + "step": 34350 + }, + { + "epoch": 4.5, + "learning_rate": 8.468898991938857e-06, + "loss": 0.0211, + "step": 34351 + }, + { + "epoch": 4.5, + "learning_rate": 8.468374922017564e-06, + "loss": 0.017, + "step": 34352 + }, + { + "epoch": 4.5, + "learning_rate": 8.467850856403887e-06, + "loss": 0.0181, + "step": 34353 + }, + { + "epoch": 4.5, + "learning_rate": 8.467326795099303e-06, + "loss": 0.0199, + "step": 34354 + }, + { + "epoch": 4.5, + "learning_rate": 8.466802738105275e-06, + "loss": 0.0217, + "step": 34355 + }, + { + "epoch": 4.5, + "learning_rate": 8.466278685423286e-06, + "loss": 0.0179, + "step": 34356 + }, + { + "epoch": 4.5, + "learning_rate": 8.465754637054805e-06, + "loss": 0.0199, + "step": 34357 + }, + { + "epoch": 4.5, + "learning_rate": 8.465230593001309e-06, + "loss": 0.0198, + "step": 34358 + }, + { + "epoch": 4.5, + "learning_rate": 8.464706553264267e-06, + "loss": 0.0221, + "step": 34359 + }, + { + "epoch": 4.5, + "learning_rate": 8.464182517845158e-06, + "loss": 0.0192, + "step": 34360 + }, + { + "epoch": 4.5, + "learning_rate": 8.463658486745457e-06, + "loss": 0.0177, + "step": 34361 + }, + { + "epoch": 4.5, + "learning_rate": 8.46313445996663e-06, + "loss": 0.0224, + "step": 34362 + }, + { + "epoch": 4.5, + "learning_rate": 8.462610437510154e-06, + "loss": 0.0191, + "step": 34363 + }, + { + "epoch": 4.5, + "learning_rate": 8.462086419377508e-06, + "loss": 0.0209, + "step": 34364 + }, + { + "epoch": 4.5, + "learning_rate": 8.461562405570161e-06, + "loss": 0.0164, + "step": 34365 + }, + { + "epoch": 4.5, + "learning_rate": 8.46103839608959e-06, + "loss": 0.0235, + "step": 34366 + }, + { + "epoch": 4.5, + "learning_rate": 8.460514390937265e-06, + "loss": 0.023, + "step": 34367 + }, + { + "epoch": 4.5, + "learning_rate": 8.459990390114659e-06, + "loss": 0.0206, + "step": 34368 + }, + { + "epoch": 4.5, + "learning_rate": 8.459466393623252e-06, + "loss": 0.0162, + "step": 34369 + }, + { + "epoch": 4.5, + "learning_rate": 8.458942401464512e-06, + "loss": 0.0214, + "step": 34370 + }, + { + "epoch": 4.5, + "learning_rate": 8.458418413639915e-06, + "loss": 0.0172, + "step": 34371 + }, + { + "epoch": 4.5, + "learning_rate": 8.457894430150933e-06, + "loss": 0.0213, + "step": 34372 + }, + { + "epoch": 4.5, + "learning_rate": 8.457370450999038e-06, + "loss": 0.0252, + "step": 34373 + }, + { + "epoch": 4.5, + "learning_rate": 8.456846476185708e-06, + "loss": 0.0153, + "step": 34374 + }, + { + "epoch": 4.5, + "learning_rate": 8.45632250571242e-06, + "loss": 0.0236, + "step": 34375 + }, + { + "epoch": 4.5, + "learning_rate": 8.455798539580637e-06, + "loss": 0.0181, + "step": 34376 + }, + { + "epoch": 4.5, + "learning_rate": 8.455274577791843e-06, + "loss": 0.0178, + "step": 34377 + }, + { + "epoch": 4.5, + "learning_rate": 8.454750620347504e-06, + "loss": 0.0191, + "step": 34378 + }, + { + "epoch": 4.5, + "learning_rate": 8.454226667249099e-06, + "loss": 0.0228, + "step": 34379 + }, + { + "epoch": 4.5, + "learning_rate": 8.453702718498097e-06, + "loss": 0.0172, + "step": 34380 + }, + { + "epoch": 4.5, + "learning_rate": 8.453178774095976e-06, + "loss": 0.0184, + "step": 34381 + }, + { + "epoch": 4.5, + "learning_rate": 8.452654834044206e-06, + "loss": 0.018, + "step": 34382 + }, + { + "epoch": 4.5, + "learning_rate": 8.452130898344259e-06, + "loss": 0.0224, + "step": 34383 + }, + { + "epoch": 4.5, + "learning_rate": 8.451606966997612e-06, + "loss": 0.02, + "step": 34384 + }, + { + "epoch": 4.5, + "learning_rate": 8.45108304000574e-06, + "loss": 0.0181, + "step": 34385 + }, + { + "epoch": 4.5, + "learning_rate": 8.450559117370114e-06, + "loss": 0.0205, + "step": 34386 + }, + { + "epoch": 4.5, + "learning_rate": 8.45003519909221e-06, + "loss": 0.0181, + "step": 34387 + }, + { + "epoch": 4.5, + "learning_rate": 8.449511285173497e-06, + "loss": 0.0183, + "step": 34388 + }, + { + "epoch": 4.5, + "learning_rate": 8.448987375615452e-06, + "loss": 0.0172, + "step": 34389 + }, + { + "epoch": 4.5, + "learning_rate": 8.448463470419548e-06, + "loss": 0.0179, + "step": 34390 + }, + { + "epoch": 4.5, + "learning_rate": 8.447939569587255e-06, + "loss": 0.0218, + "step": 34391 + }, + { + "epoch": 4.51, + "learning_rate": 8.447415673120054e-06, + "loss": 0.0206, + "step": 34392 + }, + { + "epoch": 4.51, + "learning_rate": 8.44689178101941e-06, + "loss": 0.0191, + "step": 34393 + }, + { + "epoch": 4.51, + "learning_rate": 8.446367893286803e-06, + "loss": 0.025, + "step": 34394 + }, + { + "epoch": 4.51, + "learning_rate": 8.445844009923697e-06, + "loss": 0.0146, + "step": 34395 + }, + { + "epoch": 4.51, + "learning_rate": 8.445320130931578e-06, + "loss": 0.0215, + "step": 34396 + }, + { + "epoch": 4.51, + "learning_rate": 8.44479625631191e-06, + "loss": 0.0178, + "step": 34397 + }, + { + "epoch": 4.51, + "learning_rate": 8.444272386066174e-06, + "loss": 0.0135, + "step": 34398 + }, + { + "epoch": 4.51, + "learning_rate": 8.443748520195837e-06, + "loss": 0.0206, + "step": 34399 + }, + { + "epoch": 4.51, + "learning_rate": 8.443224658702376e-06, + "loss": 0.0201, + "step": 34400 + }, + { + "epoch": 4.51, + "learning_rate": 8.442700801587261e-06, + "loss": 0.0251, + "step": 34401 + }, + { + "epoch": 4.51, + "learning_rate": 8.442176948851967e-06, + "loss": 0.0225, + "step": 34402 + }, + { + "epoch": 4.51, + "learning_rate": 8.441653100497968e-06, + "loss": 0.0244, + "step": 34403 + }, + { + "epoch": 4.51, + "learning_rate": 8.441129256526734e-06, + "loss": 0.0234, + "step": 34404 + }, + { + "epoch": 4.51, + "learning_rate": 8.440605416939744e-06, + "loss": 0.0254, + "step": 34405 + }, + { + "epoch": 4.51, + "learning_rate": 8.440081581738464e-06, + "loss": 0.0194, + "step": 34406 + }, + { + "epoch": 4.51, + "learning_rate": 8.439557750924375e-06, + "loss": 0.0186, + "step": 34407 + }, + { + "epoch": 4.51, + "learning_rate": 8.439033924498949e-06, + "loss": 0.0207, + "step": 34408 + }, + { + "epoch": 4.51, + "learning_rate": 8.438510102463652e-06, + "loss": 0.0232, + "step": 34409 + }, + { + "epoch": 4.51, + "learning_rate": 8.437986284819966e-06, + "loss": 0.0162, + "step": 34410 + }, + { + "epoch": 4.51, + "learning_rate": 8.43746247156936e-06, + "loss": 0.0202, + "step": 34411 + }, + { + "epoch": 4.51, + "learning_rate": 8.436938662713305e-06, + "loss": 0.0167, + "step": 34412 + }, + { + "epoch": 4.51, + "learning_rate": 8.436414858253282e-06, + "loss": 0.0215, + "step": 34413 + }, + { + "epoch": 4.51, + "learning_rate": 8.435891058190752e-06, + "loss": 0.0201, + "step": 34414 + }, + { + "epoch": 4.51, + "learning_rate": 8.435367262527202e-06, + "loss": 0.0156, + "step": 34415 + }, + { + "epoch": 4.51, + "learning_rate": 8.434843471264094e-06, + "loss": 0.0192, + "step": 34416 + }, + { + "epoch": 4.51, + "learning_rate": 8.434319684402903e-06, + "loss": 0.0188, + "step": 34417 + }, + { + "epoch": 4.51, + "learning_rate": 8.433795901945107e-06, + "loss": 0.0254, + "step": 34418 + }, + { + "epoch": 4.51, + "learning_rate": 8.433272123892178e-06, + "loss": 0.0194, + "step": 34419 + }, + { + "epoch": 4.51, + "learning_rate": 8.432748350245587e-06, + "loss": 0.0181, + "step": 34420 + }, + { + "epoch": 4.51, + "learning_rate": 8.432224581006808e-06, + "loss": 0.0181, + "step": 34421 + }, + { + "epoch": 4.51, + "learning_rate": 8.431700816177313e-06, + "loss": 0.0217, + "step": 34422 + }, + { + "epoch": 4.51, + "learning_rate": 8.431177055758578e-06, + "loss": 0.0229, + "step": 34423 + }, + { + "epoch": 4.51, + "learning_rate": 8.430653299752072e-06, + "loss": 0.025, + "step": 34424 + }, + { + "epoch": 4.51, + "learning_rate": 8.430129548159271e-06, + "loss": 0.0203, + "step": 34425 + }, + { + "epoch": 4.51, + "learning_rate": 8.429605800981648e-06, + "loss": 0.0224, + "step": 34426 + }, + { + "epoch": 4.51, + "learning_rate": 8.429082058220672e-06, + "loss": 0.021, + "step": 34427 + }, + { + "epoch": 4.51, + "learning_rate": 8.428558319877818e-06, + "loss": 0.0223, + "step": 34428 + }, + { + "epoch": 4.51, + "learning_rate": 8.428034585954565e-06, + "loss": 0.0212, + "step": 34429 + }, + { + "epoch": 4.51, + "learning_rate": 8.427510856452376e-06, + "loss": 0.0225, + "step": 34430 + }, + { + "epoch": 4.51, + "learning_rate": 8.426987131372735e-06, + "loss": 0.023, + "step": 34431 + }, + { + "epoch": 4.51, + "learning_rate": 8.426463410717105e-06, + "loss": 0.0243, + "step": 34432 + }, + { + "epoch": 4.51, + "learning_rate": 8.425939694486962e-06, + "loss": 0.0163, + "step": 34433 + }, + { + "epoch": 4.51, + "learning_rate": 8.425415982683783e-06, + "loss": 0.0193, + "step": 34434 + }, + { + "epoch": 4.51, + "learning_rate": 8.424892275309036e-06, + "loss": 0.021, + "step": 34435 + }, + { + "epoch": 4.51, + "learning_rate": 8.424368572364197e-06, + "loss": 0.0221, + "step": 34436 + }, + { + "epoch": 4.51, + "learning_rate": 8.423844873850735e-06, + "loss": 0.0193, + "step": 34437 + }, + { + "epoch": 4.51, + "learning_rate": 8.423321179770128e-06, + "loss": 0.0262, + "step": 34438 + }, + { + "epoch": 4.51, + "learning_rate": 8.42279749012384e-06, + "loss": 0.0195, + "step": 34439 + }, + { + "epoch": 4.51, + "learning_rate": 8.422273804913355e-06, + "loss": 0.0193, + "step": 34440 + }, + { + "epoch": 4.51, + "learning_rate": 8.42175012414014e-06, + "loss": 0.0227, + "step": 34441 + }, + { + "epoch": 4.51, + "learning_rate": 8.421226447805671e-06, + "loss": 0.0149, + "step": 34442 + }, + { + "epoch": 4.51, + "learning_rate": 8.420702775911417e-06, + "loss": 0.0193, + "step": 34443 + }, + { + "epoch": 4.51, + "learning_rate": 8.420179108458854e-06, + "loss": 0.0208, + "step": 34444 + }, + { + "epoch": 4.51, + "learning_rate": 8.41965544544945e-06, + "loss": 0.0237, + "step": 34445 + }, + { + "epoch": 4.51, + "learning_rate": 8.419131786884682e-06, + "loss": 0.0219, + "step": 34446 + }, + { + "epoch": 4.51, + "learning_rate": 8.418608132766022e-06, + "loss": 0.0189, + "step": 34447 + }, + { + "epoch": 4.51, + "learning_rate": 8.418084483094943e-06, + "loss": 0.0196, + "step": 34448 + }, + { + "epoch": 4.51, + "learning_rate": 8.417560837872917e-06, + "loss": 0.0213, + "step": 34449 + }, + { + "epoch": 4.51, + "learning_rate": 8.417037197101415e-06, + "loss": 0.0192, + "step": 34450 + }, + { + "epoch": 4.51, + "learning_rate": 8.416513560781912e-06, + "loss": 0.0208, + "step": 34451 + }, + { + "epoch": 4.51, + "learning_rate": 8.415989928915882e-06, + "loss": 0.0201, + "step": 34452 + }, + { + "epoch": 4.51, + "learning_rate": 8.415466301504796e-06, + "loss": 0.0197, + "step": 34453 + }, + { + "epoch": 4.51, + "learning_rate": 8.414942678550125e-06, + "loss": 0.0205, + "step": 34454 + }, + { + "epoch": 4.51, + "learning_rate": 8.414419060053346e-06, + "loss": 0.02, + "step": 34455 + }, + { + "epoch": 4.51, + "learning_rate": 8.413895446015926e-06, + "loss": 0.0172, + "step": 34456 + }, + { + "epoch": 4.51, + "learning_rate": 8.413371836439342e-06, + "loss": 0.0202, + "step": 34457 + }, + { + "epoch": 4.51, + "learning_rate": 8.412848231325065e-06, + "loss": 0.0285, + "step": 34458 + }, + { + "epoch": 4.51, + "learning_rate": 8.412324630674569e-06, + "loss": 0.018, + "step": 34459 + }, + { + "epoch": 4.51, + "learning_rate": 8.411801034489322e-06, + "loss": 0.0245, + "step": 34460 + }, + { + "epoch": 4.51, + "learning_rate": 8.4112774427708e-06, + "loss": 0.018, + "step": 34461 + }, + { + "epoch": 4.51, + "learning_rate": 8.410753855520477e-06, + "loss": 0.0186, + "step": 34462 + }, + { + "epoch": 4.51, + "learning_rate": 8.410230272739828e-06, + "loss": 0.0214, + "step": 34463 + }, + { + "epoch": 4.51, + "learning_rate": 8.409706694430317e-06, + "loss": 0.0229, + "step": 34464 + }, + { + "epoch": 4.51, + "learning_rate": 8.409183120593425e-06, + "loss": 0.0135, + "step": 34465 + }, + { + "epoch": 4.51, + "learning_rate": 8.408659551230618e-06, + "loss": 0.0191, + "step": 34466 + }, + { + "epoch": 4.51, + "learning_rate": 8.408135986343373e-06, + "loss": 0.022, + "step": 34467 + }, + { + "epoch": 4.52, + "learning_rate": 8.407612425933159e-06, + "loss": 0.0208, + "step": 34468 + }, + { + "epoch": 4.52, + "learning_rate": 8.40708887000145e-06, + "loss": 0.0189, + "step": 34469 + }, + { + "epoch": 4.52, + "learning_rate": 8.406565318549721e-06, + "loss": 0.0243, + "step": 34470 + }, + { + "epoch": 4.52, + "learning_rate": 8.406041771579441e-06, + "loss": 0.0207, + "step": 34471 + }, + { + "epoch": 4.52, + "learning_rate": 8.40551822909208e-06, + "loss": 0.0201, + "step": 34472 + }, + { + "epoch": 4.52, + "learning_rate": 8.404994691089118e-06, + "loss": 0.0182, + "step": 34473 + }, + { + "epoch": 4.52, + "learning_rate": 8.404471157572023e-06, + "loss": 0.0215, + "step": 34474 + }, + { + "epoch": 4.52, + "learning_rate": 8.403947628542267e-06, + "loss": 0.0169, + "step": 34475 + }, + { + "epoch": 4.52, + "learning_rate": 8.403424104001327e-06, + "loss": 0.0247, + "step": 34476 + }, + { + "epoch": 4.52, + "learning_rate": 8.402900583950667e-06, + "loss": 0.0212, + "step": 34477 + }, + { + "epoch": 4.52, + "learning_rate": 8.402377068391768e-06, + "loss": 0.0238, + "step": 34478 + }, + { + "epoch": 4.52, + "learning_rate": 8.401853557326097e-06, + "loss": 0.0182, + "step": 34479 + }, + { + "epoch": 4.52, + "learning_rate": 8.401330050755128e-06, + "loss": 0.0196, + "step": 34480 + }, + { + "epoch": 4.52, + "learning_rate": 8.40080654868033e-06, + "loss": 0.0218, + "step": 34481 + }, + { + "epoch": 4.52, + "learning_rate": 8.400283051103183e-06, + "loss": 0.0171, + "step": 34482 + }, + { + "epoch": 4.52, + "learning_rate": 8.399759558025146e-06, + "loss": 0.0171, + "step": 34483 + }, + { + "epoch": 4.52, + "learning_rate": 8.399236069447709e-06, + "loss": 0.0162, + "step": 34484 + }, + { + "epoch": 4.52, + "learning_rate": 8.398712585372333e-06, + "loss": 0.0181, + "step": 34485 + }, + { + "epoch": 4.52, + "learning_rate": 8.398189105800492e-06, + "loss": 0.0201, + "step": 34486 + }, + { + "epoch": 4.52, + "learning_rate": 8.39766563073366e-06, + "loss": 0.0151, + "step": 34487 + }, + { + "epoch": 4.52, + "learning_rate": 8.397142160173308e-06, + "loss": 0.0198, + "step": 34488 + }, + { + "epoch": 4.52, + "learning_rate": 8.396618694120908e-06, + "loss": 0.0221, + "step": 34489 + }, + { + "epoch": 4.52, + "learning_rate": 8.39609523257793e-06, + "loss": 0.0245, + "step": 34490 + }, + { + "epoch": 4.52, + "learning_rate": 8.395571775545854e-06, + "loss": 0.0199, + "step": 34491 + }, + { + "epoch": 4.52, + "learning_rate": 8.395048323026143e-06, + "loss": 0.024, + "step": 34492 + }, + { + "epoch": 4.52, + "learning_rate": 8.394524875020274e-06, + "loss": 0.0176, + "step": 34493 + }, + { + "epoch": 4.52, + "learning_rate": 8.394001431529716e-06, + "loss": 0.0171, + "step": 34494 + }, + { + "epoch": 4.52, + "learning_rate": 8.393477992555946e-06, + "loss": 0.0194, + "step": 34495 + }, + { + "epoch": 4.52, + "learning_rate": 8.39295455810043e-06, + "loss": 0.0179, + "step": 34496 + }, + { + "epoch": 4.52, + "learning_rate": 8.39243112816465e-06, + "loss": 0.0192, + "step": 34497 + }, + { + "epoch": 4.52, + "learning_rate": 8.391907702750069e-06, + "loss": 0.0216, + "step": 34498 + }, + { + "epoch": 4.52, + "learning_rate": 8.391384281858162e-06, + "loss": 0.0212, + "step": 34499 + }, + { + "epoch": 4.52, + "learning_rate": 8.3908608654904e-06, + "loss": 0.021, + "step": 34500 + }, + { + "epoch": 4.52, + "learning_rate": 8.390337453648259e-06, + "loss": 0.024, + "step": 34501 + }, + { + "epoch": 4.52, + "learning_rate": 8.389814046333204e-06, + "loss": 0.0215, + "step": 34502 + }, + { + "epoch": 4.52, + "learning_rate": 8.389290643546716e-06, + "loss": 0.0215, + "step": 34503 + }, + { + "epoch": 4.52, + "learning_rate": 8.388767245290255e-06, + "loss": 0.0208, + "step": 34504 + }, + { + "epoch": 4.52, + "learning_rate": 8.388243851565305e-06, + "loss": 0.0202, + "step": 34505 + }, + { + "epoch": 4.52, + "learning_rate": 8.387720462373333e-06, + "loss": 0.0203, + "step": 34506 + }, + { + "epoch": 4.52, + "learning_rate": 8.387197077715813e-06, + "loss": 0.0192, + "step": 34507 + }, + { + "epoch": 4.52, + "learning_rate": 8.386673697594214e-06, + "loss": 0.0167, + "step": 34508 + }, + { + "epoch": 4.52, + "learning_rate": 8.38615032201001e-06, + "loss": 0.0176, + "step": 34509 + }, + { + "epoch": 4.52, + "learning_rate": 8.385626950964672e-06, + "loss": 0.0227, + "step": 34510 + }, + { + "epoch": 4.52, + "learning_rate": 8.385103584459671e-06, + "loss": 0.0192, + "step": 34511 + }, + { + "epoch": 4.52, + "learning_rate": 8.384580222496483e-06, + "loss": 0.0205, + "step": 34512 + }, + { + "epoch": 4.52, + "learning_rate": 8.384056865076573e-06, + "loss": 0.0183, + "step": 34513 + }, + { + "epoch": 4.52, + "learning_rate": 8.38353351220142e-06, + "loss": 0.0236, + "step": 34514 + }, + { + "epoch": 4.52, + "learning_rate": 8.383010163872491e-06, + "loss": 0.0228, + "step": 34515 + }, + { + "epoch": 4.52, + "learning_rate": 8.382486820091258e-06, + "loss": 0.0181, + "step": 34516 + }, + { + "epoch": 4.52, + "learning_rate": 8.3819634808592e-06, + "loss": 0.021, + "step": 34517 + }, + { + "epoch": 4.52, + "learning_rate": 8.38144014617778e-06, + "loss": 0.0206, + "step": 34518 + }, + { + "epoch": 4.52, + "learning_rate": 8.380916816048474e-06, + "loss": 0.0228, + "step": 34519 + }, + { + "epoch": 4.52, + "learning_rate": 8.380393490472756e-06, + "loss": 0.0202, + "step": 34520 + }, + { + "epoch": 4.52, + "learning_rate": 8.37987016945209e-06, + "loss": 0.0189, + "step": 34521 + }, + { + "epoch": 4.52, + "learning_rate": 8.379346852987959e-06, + "loss": 0.0208, + "step": 34522 + }, + { + "epoch": 4.52, + "learning_rate": 8.378823541081824e-06, + "loss": 0.0198, + "step": 34523 + }, + { + "epoch": 4.52, + "learning_rate": 8.378300233735163e-06, + "loss": 0.0192, + "step": 34524 + }, + { + "epoch": 4.52, + "learning_rate": 8.377776930949446e-06, + "loss": 0.0207, + "step": 34525 + }, + { + "epoch": 4.52, + "learning_rate": 8.377253632726142e-06, + "loss": 0.0196, + "step": 34526 + }, + { + "epoch": 4.52, + "learning_rate": 8.376730339066729e-06, + "loss": 0.0189, + "step": 34527 + }, + { + "epoch": 4.52, + "learning_rate": 8.376207049972676e-06, + "loss": 0.0183, + "step": 34528 + }, + { + "epoch": 4.52, + "learning_rate": 8.375683765445453e-06, + "loss": 0.0239, + "step": 34529 + }, + { + "epoch": 4.52, + "learning_rate": 8.375160485486534e-06, + "loss": 0.0215, + "step": 34530 + }, + { + "epoch": 4.52, + "learning_rate": 8.374637210097389e-06, + "loss": 0.0217, + "step": 34531 + }, + { + "epoch": 4.52, + "learning_rate": 8.374113939279491e-06, + "loss": 0.0234, + "step": 34532 + }, + { + "epoch": 4.52, + "learning_rate": 8.37359067303431e-06, + "loss": 0.0218, + "step": 34533 + }, + { + "epoch": 4.52, + "learning_rate": 8.373067411363319e-06, + "loss": 0.0206, + "step": 34534 + }, + { + "epoch": 4.52, + "learning_rate": 8.37254415426799e-06, + "loss": 0.0241, + "step": 34535 + }, + { + "epoch": 4.52, + "learning_rate": 8.372020901749792e-06, + "loss": 0.0201, + "step": 34536 + }, + { + "epoch": 4.52, + "learning_rate": 8.371497653810196e-06, + "loss": 0.0169, + "step": 34537 + }, + { + "epoch": 4.52, + "learning_rate": 8.370974410450681e-06, + "loss": 0.0212, + "step": 34538 + }, + { + "epoch": 4.52, + "learning_rate": 8.370451171672712e-06, + "loss": 0.0196, + "step": 34539 + }, + { + "epoch": 4.52, + "learning_rate": 8.369927937477761e-06, + "loss": 0.0247, + "step": 34540 + }, + { + "epoch": 4.52, + "learning_rate": 8.369404707867305e-06, + "loss": 0.0213, + "step": 34541 + }, + { + "epoch": 4.52, + "learning_rate": 8.368881482842807e-06, + "loss": 0.0233, + "step": 34542 + }, + { + "epoch": 4.52, + "learning_rate": 8.368358262405746e-06, + "loss": 0.0159, + "step": 34543 + }, + { + "epoch": 4.53, + "learning_rate": 8.36783504655759e-06, + "loss": 0.0168, + "step": 34544 + }, + { + "epoch": 4.53, + "learning_rate": 8.367311835299812e-06, + "loss": 0.0235, + "step": 34545 + }, + { + "epoch": 4.53, + "learning_rate": 8.366788628633879e-06, + "loss": 0.015, + "step": 34546 + }, + { + "epoch": 4.53, + "learning_rate": 8.366265426561272e-06, + "loss": 0.0238, + "step": 34547 + }, + { + "epoch": 4.53, + "learning_rate": 8.365742229083445e-06, + "loss": 0.0182, + "step": 34548 + }, + { + "epoch": 4.53, + "learning_rate": 8.365219036201891e-06, + "loss": 0.0222, + "step": 34549 + }, + { + "epoch": 4.53, + "learning_rate": 8.364695847918067e-06, + "loss": 0.0273, + "step": 34550 + }, + { + "epoch": 4.53, + "learning_rate": 8.364172664233453e-06, + "loss": 0.024, + "step": 34551 + }, + { + "epoch": 4.53, + "learning_rate": 8.363649485149513e-06, + "loss": 0.0173, + "step": 34552 + }, + { + "epoch": 4.53, + "learning_rate": 8.363126310667723e-06, + "loss": 0.0176, + "step": 34553 + }, + { + "epoch": 4.53, + "learning_rate": 8.362603140789553e-06, + "loss": 0.0209, + "step": 34554 + }, + { + "epoch": 4.53, + "learning_rate": 8.362079975516473e-06, + "loss": 0.0222, + "step": 34555 + }, + { + "epoch": 4.53, + "learning_rate": 8.361556814849959e-06, + "loss": 0.0182, + "step": 34556 + }, + { + "epoch": 4.53, + "learning_rate": 8.361033658791476e-06, + "loss": 0.0226, + "step": 34557 + }, + { + "epoch": 4.53, + "learning_rate": 8.360510507342501e-06, + "loss": 0.0165, + "step": 34558 + }, + { + "epoch": 4.53, + "learning_rate": 8.359987360504497e-06, + "loss": 0.0246, + "step": 34559 + }, + { + "epoch": 4.53, + "learning_rate": 8.359464218278947e-06, + "loss": 0.0182, + "step": 34560 + }, + { + "epoch": 4.53, + "learning_rate": 8.358941080667314e-06, + "loss": 0.0187, + "step": 34561 + }, + { + "epoch": 4.53, + "learning_rate": 8.358417947671074e-06, + "loss": 0.0228, + "step": 34562 + }, + { + "epoch": 4.53, + "learning_rate": 8.357894819291695e-06, + "loss": 0.0191, + "step": 34563 + }, + { + "epoch": 4.53, + "learning_rate": 8.357371695530652e-06, + "loss": 0.0229, + "step": 34564 + }, + { + "epoch": 4.53, + "learning_rate": 8.356848576389411e-06, + "loss": 0.0132, + "step": 34565 + }, + { + "epoch": 4.53, + "learning_rate": 8.356325461869446e-06, + "loss": 0.0243, + "step": 34566 + }, + { + "epoch": 4.53, + "learning_rate": 8.355802351972228e-06, + "loss": 0.0176, + "step": 34567 + }, + { + "epoch": 4.53, + "learning_rate": 8.35527924669923e-06, + "loss": 0.0201, + "step": 34568 + }, + { + "epoch": 4.53, + "learning_rate": 8.354756146051919e-06, + "loss": 0.0205, + "step": 34569 + }, + { + "epoch": 4.53, + "learning_rate": 8.354233050031767e-06, + "loss": 0.0184, + "step": 34570 + }, + { + "epoch": 4.53, + "learning_rate": 8.353709958640249e-06, + "loss": 0.0193, + "step": 34571 + }, + { + "epoch": 4.53, + "learning_rate": 8.353186871878836e-06, + "loss": 0.0221, + "step": 34572 + }, + { + "epoch": 4.53, + "learning_rate": 8.352663789748996e-06, + "loss": 0.0205, + "step": 34573 + }, + { + "epoch": 4.53, + "learning_rate": 8.352140712252203e-06, + "loss": 0.0211, + "step": 34574 + }, + { + "epoch": 4.53, + "learning_rate": 8.351617639389925e-06, + "loss": 0.0211, + "step": 34575 + }, + { + "epoch": 4.53, + "learning_rate": 8.351094571163633e-06, + "loss": 0.0204, + "step": 34576 + }, + { + "epoch": 4.53, + "learning_rate": 8.350571507574804e-06, + "loss": 0.0224, + "step": 34577 + }, + { + "epoch": 4.53, + "learning_rate": 8.3500484486249e-06, + "loss": 0.0187, + "step": 34578 + }, + { + "epoch": 4.53, + "learning_rate": 8.3495253943154e-06, + "loss": 0.0167, + "step": 34579 + }, + { + "epoch": 4.53, + "learning_rate": 8.34900234464777e-06, + "loss": 0.0221, + "step": 34580 + }, + { + "epoch": 4.53, + "learning_rate": 8.34847929962348e-06, + "loss": 0.0258, + "step": 34581 + }, + { + "epoch": 4.53, + "learning_rate": 8.34795625924401e-06, + "loss": 0.0208, + "step": 34582 + }, + { + "epoch": 4.53, + "learning_rate": 8.347433223510822e-06, + "loss": 0.0186, + "step": 34583 + }, + { + "epoch": 4.53, + "learning_rate": 8.346910192425391e-06, + "loss": 0.0204, + "step": 34584 + }, + { + "epoch": 4.53, + "learning_rate": 8.346387165989189e-06, + "loss": 0.0213, + "step": 34585 + }, + { + "epoch": 4.53, + "learning_rate": 8.345864144203682e-06, + "loss": 0.0252, + "step": 34586 + }, + { + "epoch": 4.53, + "learning_rate": 8.345341127070346e-06, + "loss": 0.0207, + "step": 34587 + }, + { + "epoch": 4.53, + "learning_rate": 8.344818114590648e-06, + "loss": 0.0231, + "step": 34588 + }, + { + "epoch": 4.53, + "learning_rate": 8.344295106766062e-06, + "loss": 0.0168, + "step": 34589 + }, + { + "epoch": 4.53, + "learning_rate": 8.343772103598058e-06, + "loss": 0.02, + "step": 34590 + }, + { + "epoch": 4.53, + "learning_rate": 8.343249105088106e-06, + "loss": 0.0161, + "step": 34591 + }, + { + "epoch": 4.53, + "learning_rate": 8.342726111237676e-06, + "loss": 0.0194, + "step": 34592 + }, + { + "epoch": 4.53, + "learning_rate": 8.342203122048243e-06, + "loss": 0.0209, + "step": 34593 + }, + { + "epoch": 4.53, + "learning_rate": 8.341680137521275e-06, + "loss": 0.0199, + "step": 34594 + }, + { + "epoch": 4.53, + "learning_rate": 8.341157157658244e-06, + "loss": 0.0202, + "step": 34595 + }, + { + "epoch": 4.53, + "learning_rate": 8.34063418246062e-06, + "loss": 0.0228, + "step": 34596 + }, + { + "epoch": 4.53, + "learning_rate": 8.340111211929875e-06, + "loss": 0.0231, + "step": 34597 + }, + { + "epoch": 4.53, + "learning_rate": 8.339588246067475e-06, + "loss": 0.0179, + "step": 34598 + }, + { + "epoch": 4.53, + "learning_rate": 8.339065284874897e-06, + "loss": 0.0255, + "step": 34599 + }, + { + "epoch": 4.53, + "learning_rate": 8.338542328353611e-06, + "loss": 0.0224, + "step": 34600 + }, + { + "epoch": 4.53, + "learning_rate": 8.338019376505084e-06, + "loss": 0.0187, + "step": 34601 + }, + { + "epoch": 4.53, + "learning_rate": 8.33749642933079e-06, + "loss": 0.0204, + "step": 34602 + }, + { + "epoch": 4.53, + "learning_rate": 8.336973486832198e-06, + "loss": 0.0181, + "step": 34603 + }, + { + "epoch": 4.53, + "learning_rate": 8.336450549010778e-06, + "loss": 0.0232, + "step": 34604 + }, + { + "epoch": 4.53, + "learning_rate": 8.335927615868003e-06, + "loss": 0.0163, + "step": 34605 + }, + { + "epoch": 4.53, + "learning_rate": 8.335404687405346e-06, + "loss": 0.023, + "step": 34606 + }, + { + "epoch": 4.53, + "learning_rate": 8.334881763624273e-06, + "loss": 0.0211, + "step": 34607 + }, + { + "epoch": 4.53, + "learning_rate": 8.334358844526256e-06, + "loss": 0.0219, + "step": 34608 + }, + { + "epoch": 4.53, + "learning_rate": 8.333835930112767e-06, + "loss": 0.0195, + "step": 34609 + }, + { + "epoch": 4.53, + "learning_rate": 8.333313020385277e-06, + "loss": 0.0193, + "step": 34610 + }, + { + "epoch": 4.53, + "learning_rate": 8.332790115345252e-06, + "loss": 0.0226, + "step": 34611 + }, + { + "epoch": 4.53, + "learning_rate": 8.332267214994166e-06, + "loss": 0.0211, + "step": 34612 + }, + { + "epoch": 4.53, + "learning_rate": 8.331744319333492e-06, + "loss": 0.0192, + "step": 34613 + }, + { + "epoch": 4.53, + "learning_rate": 8.331221428364695e-06, + "loss": 0.0194, + "step": 34614 + }, + { + "epoch": 4.53, + "learning_rate": 8.33069854208925e-06, + "loss": 0.0198, + "step": 34615 + }, + { + "epoch": 4.53, + "learning_rate": 8.33017566050863e-06, + "loss": 0.0194, + "step": 34616 + }, + { + "epoch": 4.53, + "learning_rate": 8.3296527836243e-06, + "loss": 0.0224, + "step": 34617 + }, + { + "epoch": 4.53, + "learning_rate": 8.329129911437732e-06, + "loss": 0.0214, + "step": 34618 + }, + { + "epoch": 4.53, + "learning_rate": 8.328607043950398e-06, + "loss": 0.0179, + "step": 34619 + }, + { + "epoch": 4.53, + "learning_rate": 8.328084181163765e-06, + "loss": 0.0205, + "step": 34620 + }, + { + "epoch": 4.54, + "learning_rate": 8.32756132307931e-06, + "loss": 0.0202, + "step": 34621 + }, + { + "epoch": 4.54, + "learning_rate": 8.327038469698498e-06, + "loss": 0.0186, + "step": 34622 + }, + { + "epoch": 4.54, + "learning_rate": 8.326515621022802e-06, + "loss": 0.0217, + "step": 34623 + }, + { + "epoch": 4.54, + "learning_rate": 8.32599277705369e-06, + "loss": 0.0235, + "step": 34624 + }, + { + "epoch": 4.54, + "learning_rate": 8.325469937792631e-06, + "loss": 0.0194, + "step": 34625 + }, + { + "epoch": 4.54, + "learning_rate": 8.324947103241103e-06, + "loss": 0.0173, + "step": 34626 + }, + { + "epoch": 4.54, + "learning_rate": 8.324424273400573e-06, + "loss": 0.0192, + "step": 34627 + }, + { + "epoch": 4.54, + "learning_rate": 8.323901448272509e-06, + "loss": 0.0189, + "step": 34628 + }, + { + "epoch": 4.54, + "learning_rate": 8.323378627858385e-06, + "loss": 0.0209, + "step": 34629 + }, + { + "epoch": 4.54, + "learning_rate": 8.322855812159667e-06, + "loss": 0.0187, + "step": 34630 + }, + { + "epoch": 4.54, + "learning_rate": 8.32233300117783e-06, + "loss": 0.023, + "step": 34631 + }, + { + "epoch": 4.54, + "learning_rate": 8.321810194914339e-06, + "loss": 0.0236, + "step": 34632 + }, + { + "epoch": 4.54, + "learning_rate": 8.32128739337067e-06, + "loss": 0.0178, + "step": 34633 + }, + { + "epoch": 4.54, + "learning_rate": 8.32076459654829e-06, + "loss": 0.0164, + "step": 34634 + }, + { + "epoch": 4.54, + "learning_rate": 8.320241804448665e-06, + "loss": 0.0171, + "step": 34635 + }, + { + "epoch": 4.54, + "learning_rate": 8.319719017073275e-06, + "loss": 0.0216, + "step": 34636 + }, + { + "epoch": 4.54, + "learning_rate": 8.319196234423587e-06, + "loss": 0.0199, + "step": 34637 + }, + { + "epoch": 4.54, + "learning_rate": 8.318673456501068e-06, + "loss": 0.023, + "step": 34638 + }, + { + "epoch": 4.54, + "learning_rate": 8.318150683307194e-06, + "loss": 0.0182, + "step": 34639 + }, + { + "epoch": 4.54, + "learning_rate": 8.317627914843429e-06, + "loss": 0.019, + "step": 34640 + }, + { + "epoch": 4.54, + "learning_rate": 8.317105151111244e-06, + "loss": 0.0125, + "step": 34641 + }, + { + "epoch": 4.54, + "learning_rate": 8.316582392112114e-06, + "loss": 0.0169, + "step": 34642 + }, + { + "epoch": 4.54, + "learning_rate": 8.316059637847506e-06, + "loss": 0.0158, + "step": 34643 + }, + { + "epoch": 4.54, + "learning_rate": 8.315536888318892e-06, + "loss": 0.0175, + "step": 34644 + }, + { + "epoch": 4.54, + "learning_rate": 8.315014143527738e-06, + "loss": 0.0182, + "step": 34645 + }, + { + "epoch": 4.54, + "learning_rate": 8.314491403475515e-06, + "loss": 0.0172, + "step": 34646 + }, + { + "epoch": 4.54, + "learning_rate": 8.313968668163698e-06, + "loss": 0.0221, + "step": 34647 + }, + { + "epoch": 4.54, + "learning_rate": 8.313445937593753e-06, + "loss": 0.0194, + "step": 34648 + }, + { + "epoch": 4.54, + "learning_rate": 8.312923211767154e-06, + "loss": 0.0189, + "step": 34649 + }, + { + "epoch": 4.54, + "learning_rate": 8.312400490685368e-06, + "loss": 0.0211, + "step": 34650 + }, + { + "epoch": 4.54, + "learning_rate": 8.311877774349864e-06, + "loss": 0.0201, + "step": 34651 + }, + { + "epoch": 4.54, + "learning_rate": 8.311355062762117e-06, + "loss": 0.0173, + "step": 34652 + }, + { + "epoch": 4.54, + "learning_rate": 8.310832355923591e-06, + "loss": 0.0201, + "step": 34653 + }, + { + "epoch": 4.54, + "learning_rate": 8.310309653835762e-06, + "loss": 0.0218, + "step": 34654 + }, + { + "epoch": 4.54, + "learning_rate": 8.309786956500095e-06, + "loss": 0.0167, + "step": 34655 + }, + { + "epoch": 4.54, + "learning_rate": 8.309264263918062e-06, + "loss": 0.0177, + "step": 34656 + }, + { + "epoch": 4.54, + "learning_rate": 8.30874157609113e-06, + "loss": 0.0194, + "step": 34657 + }, + { + "epoch": 4.54, + "learning_rate": 8.308218893020777e-06, + "loss": 0.0254, + "step": 34658 + }, + { + "epoch": 4.54, + "learning_rate": 8.307696214708467e-06, + "loss": 0.0213, + "step": 34659 + }, + { + "epoch": 4.54, + "learning_rate": 8.307173541155672e-06, + "loss": 0.0166, + "step": 34660 + }, + { + "epoch": 4.54, + "learning_rate": 8.30665087236386e-06, + "loss": 0.0189, + "step": 34661 + }, + { + "epoch": 4.54, + "learning_rate": 8.306128208334502e-06, + "loss": 0.0187, + "step": 34662 + }, + { + "epoch": 4.54, + "learning_rate": 8.305605549069072e-06, + "loss": 0.0223, + "step": 34663 + }, + { + "epoch": 4.54, + "learning_rate": 8.305082894569033e-06, + "loss": 0.0222, + "step": 34664 + }, + { + "epoch": 4.54, + "learning_rate": 8.304560244835858e-06, + "loss": 0.022, + "step": 34665 + }, + { + "epoch": 4.54, + "learning_rate": 8.304037599871017e-06, + "loss": 0.0218, + "step": 34666 + }, + { + "epoch": 4.54, + "learning_rate": 8.303514959675982e-06, + "loss": 0.0262, + "step": 34667 + }, + { + "epoch": 4.54, + "learning_rate": 8.302992324252216e-06, + "loss": 0.0187, + "step": 34668 + }, + { + "epoch": 4.54, + "learning_rate": 8.302469693601197e-06, + "loss": 0.0201, + "step": 34669 + }, + { + "epoch": 4.54, + "learning_rate": 8.30194706772439e-06, + "loss": 0.0201, + "step": 34670 + }, + { + "epoch": 4.54, + "learning_rate": 8.30142444662327e-06, + "loss": 0.0241, + "step": 34671 + }, + { + "epoch": 4.54, + "learning_rate": 8.300901830299301e-06, + "loss": 0.0221, + "step": 34672 + }, + { + "epoch": 4.54, + "learning_rate": 8.300379218753955e-06, + "loss": 0.0203, + "step": 34673 + }, + { + "epoch": 4.54, + "learning_rate": 8.299856611988702e-06, + "loss": 0.0169, + "step": 34674 + }, + { + "epoch": 4.54, + "learning_rate": 8.299334010005014e-06, + "loss": 0.0207, + "step": 34675 + }, + { + "epoch": 4.54, + "learning_rate": 8.298811412804356e-06, + "loss": 0.0164, + "step": 34676 + }, + { + "epoch": 4.54, + "learning_rate": 8.2982888203882e-06, + "loss": 0.0172, + "step": 34677 + }, + { + "epoch": 4.54, + "learning_rate": 8.297766232758019e-06, + "loss": 0.0168, + "step": 34678 + }, + { + "epoch": 4.54, + "learning_rate": 8.297243649915274e-06, + "loss": 0.0229, + "step": 34679 + }, + { + "epoch": 4.54, + "learning_rate": 8.296721071861444e-06, + "loss": 0.0254, + "step": 34680 + }, + { + "epoch": 4.54, + "learning_rate": 8.296198498597997e-06, + "loss": 0.0203, + "step": 34681 + }, + { + "epoch": 4.54, + "learning_rate": 8.295675930126399e-06, + "loss": 0.0204, + "step": 34682 + }, + { + "epoch": 4.54, + "learning_rate": 8.295153366448123e-06, + "loss": 0.0148, + "step": 34683 + }, + { + "epoch": 4.54, + "learning_rate": 8.294630807564636e-06, + "loss": 0.0167, + "step": 34684 + }, + { + "epoch": 4.54, + "learning_rate": 8.294108253477409e-06, + "loss": 0.0169, + "step": 34685 + }, + { + "epoch": 4.54, + "learning_rate": 8.293585704187914e-06, + "loss": 0.0231, + "step": 34686 + }, + { + "epoch": 4.54, + "learning_rate": 8.293063159697616e-06, + "loss": 0.0164, + "step": 34687 + }, + { + "epoch": 4.54, + "learning_rate": 8.29254062000799e-06, + "loss": 0.0184, + "step": 34688 + }, + { + "epoch": 4.54, + "learning_rate": 8.2920180851205e-06, + "loss": 0.0192, + "step": 34689 + }, + { + "epoch": 4.54, + "learning_rate": 8.291495555036616e-06, + "loss": 0.0186, + "step": 34690 + }, + { + "epoch": 4.54, + "learning_rate": 8.29097302975781e-06, + "loss": 0.0138, + "step": 34691 + }, + { + "epoch": 4.54, + "learning_rate": 8.290450509285556e-06, + "loss": 0.0211, + "step": 34692 + }, + { + "epoch": 4.54, + "learning_rate": 8.289927993621313e-06, + "loss": 0.0209, + "step": 34693 + }, + { + "epoch": 4.54, + "learning_rate": 8.289405482766563e-06, + "loss": 0.0177, + "step": 34694 + }, + { + "epoch": 4.54, + "learning_rate": 8.288882976722765e-06, + "loss": 0.0144, + "step": 34695 + }, + { + "epoch": 4.54, + "learning_rate": 8.288360475491395e-06, + "loss": 0.0228, + "step": 34696 + }, + { + "epoch": 4.55, + "learning_rate": 8.287837979073917e-06, + "loss": 0.0204, + "step": 34697 + }, + { + "epoch": 4.55, + "learning_rate": 8.287315487471807e-06, + "loss": 0.0209, + "step": 34698 + }, + { + "epoch": 4.55, + "learning_rate": 8.286793000686528e-06, + "loss": 0.0219, + "step": 34699 + }, + { + "epoch": 4.55, + "learning_rate": 8.286270518719552e-06, + "loss": 0.0196, + "step": 34700 + }, + { + "epoch": 4.55, + "learning_rate": 8.285748041572347e-06, + "loss": 0.0202, + "step": 34701 + }, + { + "epoch": 4.55, + "learning_rate": 8.28522556924639e-06, + "loss": 0.018, + "step": 34702 + }, + { + "epoch": 4.55, + "learning_rate": 8.284703101743142e-06, + "loss": 0.0252, + "step": 34703 + }, + { + "epoch": 4.55, + "learning_rate": 8.284180639064076e-06, + "loss": 0.0236, + "step": 34704 + }, + { + "epoch": 4.55, + "learning_rate": 8.28365818121066e-06, + "loss": 0.018, + "step": 34705 + }, + { + "epoch": 4.55, + "learning_rate": 8.283135728184363e-06, + "loss": 0.0163, + "step": 34706 + }, + { + "epoch": 4.55, + "learning_rate": 8.282613279986658e-06, + "loss": 0.0198, + "step": 34707 + }, + { + "epoch": 4.55, + "learning_rate": 8.28209083661901e-06, + "loss": 0.0199, + "step": 34708 + }, + { + "epoch": 4.55, + "learning_rate": 8.281568398082891e-06, + "loss": 0.0227, + "step": 34709 + }, + { + "epoch": 4.55, + "learning_rate": 8.281045964379768e-06, + "loss": 0.0173, + "step": 34710 + }, + { + "epoch": 4.55, + "learning_rate": 8.280523535511113e-06, + "loss": 0.0163, + "step": 34711 + }, + { + "epoch": 4.55, + "learning_rate": 8.280001111478389e-06, + "loss": 0.0236, + "step": 34712 + }, + { + "epoch": 4.55, + "learning_rate": 8.279478692283076e-06, + "loss": 0.0214, + "step": 34713 + }, + { + "epoch": 4.55, + "learning_rate": 8.278956277926636e-06, + "loss": 0.0212, + "step": 34714 + }, + { + "epoch": 4.55, + "learning_rate": 8.27843386841054e-06, + "loss": 0.0229, + "step": 34715 + }, + { + "epoch": 4.55, + "learning_rate": 8.277911463736257e-06, + "loss": 0.0278, + "step": 34716 + }, + { + "epoch": 4.55, + "learning_rate": 8.277389063905258e-06, + "loss": 0.0231, + "step": 34717 + }, + { + "epoch": 4.55, + "learning_rate": 8.276866668919007e-06, + "loss": 0.0238, + "step": 34718 + }, + { + "epoch": 4.55, + "learning_rate": 8.276344278778981e-06, + "loss": 0.0212, + "step": 34719 + }, + { + "epoch": 4.55, + "learning_rate": 8.27582189348664e-06, + "loss": 0.0218, + "step": 34720 + }, + { + "epoch": 4.55, + "learning_rate": 8.27529951304346e-06, + "loss": 0.0209, + "step": 34721 + }, + { + "epoch": 4.55, + "learning_rate": 8.27477713745091e-06, + "loss": 0.019, + "step": 34722 + }, + { + "epoch": 4.55, + "learning_rate": 8.274254766710455e-06, + "loss": 0.0171, + "step": 34723 + }, + { + "epoch": 4.55, + "learning_rate": 8.273732400823566e-06, + "loss": 0.0195, + "step": 34724 + }, + { + "epoch": 4.55, + "learning_rate": 8.273210039791715e-06, + "loss": 0.0211, + "step": 34725 + }, + { + "epoch": 4.55, + "learning_rate": 8.272687683616369e-06, + "loss": 0.0202, + "step": 34726 + }, + { + "epoch": 4.55, + "learning_rate": 8.272165332298995e-06, + "loss": 0.0173, + "step": 34727 + }, + { + "epoch": 4.55, + "learning_rate": 8.271642985841067e-06, + "loss": 0.0211, + "step": 34728 + }, + { + "epoch": 4.55, + "learning_rate": 8.271120644244048e-06, + "loss": 0.018, + "step": 34729 + }, + { + "epoch": 4.55, + "learning_rate": 8.270598307509412e-06, + "loss": 0.0217, + "step": 34730 + }, + { + "epoch": 4.55, + "learning_rate": 8.270075975638624e-06, + "loss": 0.0197, + "step": 34731 + }, + { + "epoch": 4.55, + "learning_rate": 8.26955364863316e-06, + "loss": 0.0206, + "step": 34732 + }, + { + "epoch": 4.55, + "learning_rate": 8.26903132649448e-06, + "loss": 0.0202, + "step": 34733 + }, + { + "epoch": 4.55, + "learning_rate": 8.268509009224055e-06, + "loss": 0.0204, + "step": 34734 + }, + { + "epoch": 4.55, + "learning_rate": 8.26798669682336e-06, + "loss": 0.0174, + "step": 34735 + }, + { + "epoch": 4.55, + "learning_rate": 8.26746438929386e-06, + "loss": 0.0218, + "step": 34736 + }, + { + "epoch": 4.55, + "learning_rate": 8.266942086637022e-06, + "loss": 0.0207, + "step": 34737 + }, + { + "epoch": 4.55, + "learning_rate": 8.266419788854321e-06, + "loss": 0.0192, + "step": 34738 + }, + { + "epoch": 4.55, + "learning_rate": 8.26589749594722e-06, + "loss": 0.0199, + "step": 34739 + }, + { + "epoch": 4.55, + "learning_rate": 8.26537520791719e-06, + "loss": 0.0242, + "step": 34740 + }, + { + "epoch": 4.55, + "learning_rate": 8.264852924765698e-06, + "loss": 0.0227, + "step": 34741 + }, + { + "epoch": 4.55, + "learning_rate": 8.264330646494216e-06, + "loss": 0.0265, + "step": 34742 + }, + { + "epoch": 4.55, + "learning_rate": 8.263808373104214e-06, + "loss": 0.019, + "step": 34743 + }, + { + "epoch": 4.55, + "learning_rate": 8.263286104597155e-06, + "loss": 0.0178, + "step": 34744 + }, + { + "epoch": 4.55, + "learning_rate": 8.26276384097451e-06, + "loss": 0.0191, + "step": 34745 + }, + { + "epoch": 4.55, + "learning_rate": 8.262241582237753e-06, + "loss": 0.0173, + "step": 34746 + }, + { + "epoch": 4.55, + "learning_rate": 8.261719328388348e-06, + "loss": 0.0177, + "step": 34747 + }, + { + "epoch": 4.55, + "learning_rate": 8.261197079427767e-06, + "loss": 0.0141, + "step": 34748 + }, + { + "epoch": 4.55, + "learning_rate": 8.260674835357473e-06, + "loss": 0.0207, + "step": 34749 + }, + { + "epoch": 4.55, + "learning_rate": 8.260152596178939e-06, + "loss": 0.0234, + "step": 34750 + }, + { + "epoch": 4.55, + "learning_rate": 8.259630361893636e-06, + "loss": 0.0206, + "step": 34751 + }, + { + "epoch": 4.55, + "learning_rate": 8.259108132503026e-06, + "loss": 0.0204, + "step": 34752 + }, + { + "epoch": 4.55, + "learning_rate": 8.258585908008586e-06, + "loss": 0.0184, + "step": 34753 + }, + { + "epoch": 4.55, + "learning_rate": 8.258063688411778e-06, + "loss": 0.0179, + "step": 34754 + }, + { + "epoch": 4.55, + "learning_rate": 8.257541473714074e-06, + "loss": 0.0192, + "step": 34755 + }, + { + "epoch": 4.55, + "learning_rate": 8.257019263916936e-06, + "loss": 0.0209, + "step": 34756 + }, + { + "epoch": 4.55, + "learning_rate": 8.256497059021843e-06, + "loss": 0.0175, + "step": 34757 + }, + { + "epoch": 4.55, + "learning_rate": 8.25597485903026e-06, + "loss": 0.0228, + "step": 34758 + }, + { + "epoch": 4.55, + "learning_rate": 8.255452663943656e-06, + "loss": 0.0191, + "step": 34759 + }, + { + "epoch": 4.55, + "learning_rate": 8.254930473763496e-06, + "loss": 0.019, + "step": 34760 + }, + { + "epoch": 4.55, + "learning_rate": 8.254408288491253e-06, + "loss": 0.0256, + "step": 34761 + }, + { + "epoch": 4.55, + "learning_rate": 8.253886108128393e-06, + "loss": 0.021, + "step": 34762 + }, + { + "epoch": 4.55, + "learning_rate": 8.253363932676384e-06, + "loss": 0.0219, + "step": 34763 + }, + { + "epoch": 4.55, + "learning_rate": 8.252841762136699e-06, + "loss": 0.02, + "step": 34764 + }, + { + "epoch": 4.55, + "learning_rate": 8.2523195965108e-06, + "loss": 0.019, + "step": 34765 + }, + { + "epoch": 4.55, + "learning_rate": 8.251797435800157e-06, + "loss": 0.0186, + "step": 34766 + }, + { + "epoch": 4.55, + "learning_rate": 8.251275280006246e-06, + "loss": 0.0173, + "step": 34767 + }, + { + "epoch": 4.55, + "learning_rate": 8.250753129130529e-06, + "loss": 0.0289, + "step": 34768 + }, + { + "epoch": 4.55, + "learning_rate": 8.250230983174476e-06, + "loss": 0.0207, + "step": 34769 + }, + { + "epoch": 4.55, + "learning_rate": 8.249708842139553e-06, + "loss": 0.0203, + "step": 34770 + }, + { + "epoch": 4.55, + "learning_rate": 8.24918670602723e-06, + "loss": 0.0225, + "step": 34771 + }, + { + "epoch": 4.55, + "learning_rate": 8.24866457483898e-06, + "loss": 0.015, + "step": 34772 + }, + { + "epoch": 4.56, + "learning_rate": 8.248142448576267e-06, + "loss": 0.0153, + "step": 34773 + }, + { + "epoch": 4.56, + "learning_rate": 8.24762032724056e-06, + "loss": 0.0239, + "step": 34774 + }, + { + "epoch": 4.56, + "learning_rate": 8.247098210833325e-06, + "loss": 0.0195, + "step": 34775 + }, + { + "epoch": 4.56, + "learning_rate": 8.246576099356036e-06, + "loss": 0.0194, + "step": 34776 + }, + { + "epoch": 4.56, + "learning_rate": 8.24605399281015e-06, + "loss": 0.0151, + "step": 34777 + }, + { + "epoch": 4.56, + "learning_rate": 8.245531891197151e-06, + "loss": 0.0201, + "step": 34778 + }, + { + "epoch": 4.56, + "learning_rate": 8.2450097945185e-06, + "loss": 0.0155, + "step": 34779 + }, + { + "epoch": 4.56, + "learning_rate": 8.244487702775665e-06, + "loss": 0.0213, + "step": 34780 + }, + { + "epoch": 4.56, + "learning_rate": 8.243965615970115e-06, + "loss": 0.0193, + "step": 34781 + }, + { + "epoch": 4.56, + "learning_rate": 8.243443534103319e-06, + "loss": 0.0174, + "step": 34782 + }, + { + "epoch": 4.56, + "learning_rate": 8.242921457176742e-06, + "loss": 0.0183, + "step": 34783 + }, + { + "epoch": 4.56, + "learning_rate": 8.242399385191858e-06, + "loss": 0.0163, + "step": 34784 + }, + { + "epoch": 4.56, + "learning_rate": 8.24187731815013e-06, + "loss": 0.0186, + "step": 34785 + }, + { + "epoch": 4.56, + "learning_rate": 8.24135525605303e-06, + "loss": 0.0184, + "step": 34786 + }, + { + "epoch": 4.56, + "learning_rate": 8.240833198902023e-06, + "loss": 0.018, + "step": 34787 + }, + { + "epoch": 4.56, + "learning_rate": 8.240311146698576e-06, + "loss": 0.0148, + "step": 34788 + }, + { + "epoch": 4.56, + "learning_rate": 8.239789099444164e-06, + "loss": 0.0199, + "step": 34789 + }, + { + "epoch": 4.56, + "learning_rate": 8.239267057140252e-06, + "loss": 0.0205, + "step": 34790 + }, + { + "epoch": 4.56, + "learning_rate": 8.238745019788305e-06, + "loss": 0.0217, + "step": 34791 + }, + { + "epoch": 4.56, + "learning_rate": 8.238222987389795e-06, + "loss": 0.0146, + "step": 34792 + }, + { + "epoch": 4.56, + "learning_rate": 8.237700959946192e-06, + "loss": 0.0167, + "step": 34793 + }, + { + "epoch": 4.56, + "learning_rate": 8.237178937458959e-06, + "loss": 0.0187, + "step": 34794 + }, + { + "epoch": 4.56, + "learning_rate": 8.236656919929567e-06, + "loss": 0.0179, + "step": 34795 + }, + { + "epoch": 4.56, + "learning_rate": 8.236134907359481e-06, + "loss": 0.0215, + "step": 34796 + }, + { + "epoch": 4.56, + "learning_rate": 8.235612899750175e-06, + "loss": 0.02, + "step": 34797 + }, + { + "epoch": 4.56, + "learning_rate": 8.23509089710311e-06, + "loss": 0.0175, + "step": 34798 + }, + { + "epoch": 4.56, + "learning_rate": 8.234568899419758e-06, + "loss": 0.0174, + "step": 34799 + }, + { + "epoch": 4.56, + "learning_rate": 8.234046906701589e-06, + "loss": 0.0191, + "step": 34800 + }, + { + "epoch": 4.56, + "learning_rate": 8.23352491895007e-06, + "loss": 0.0208, + "step": 34801 + }, + { + "epoch": 4.56, + "learning_rate": 8.233002936166666e-06, + "loss": 0.0188, + "step": 34802 + }, + { + "epoch": 4.56, + "learning_rate": 8.23248095835285e-06, + "loss": 0.0227, + "step": 34803 + }, + { + "epoch": 4.56, + "learning_rate": 8.231958985510085e-06, + "loss": 0.0214, + "step": 34804 + }, + { + "epoch": 4.56, + "learning_rate": 8.231437017639842e-06, + "loss": 0.0218, + "step": 34805 + }, + { + "epoch": 4.56, + "learning_rate": 8.230915054743587e-06, + "loss": 0.0173, + "step": 34806 + }, + { + "epoch": 4.56, + "learning_rate": 8.230393096822789e-06, + "loss": 0.021, + "step": 34807 + }, + { + "epoch": 4.56, + "learning_rate": 8.229871143878918e-06, + "loss": 0.0217, + "step": 34808 + }, + { + "epoch": 4.56, + "learning_rate": 8.229349195913439e-06, + "loss": 0.0214, + "step": 34809 + }, + { + "epoch": 4.56, + "learning_rate": 8.228827252927818e-06, + "loss": 0.0231, + "step": 34810 + }, + { + "epoch": 4.56, + "learning_rate": 8.228305314923532e-06, + "loss": 0.0207, + "step": 34811 + }, + { + "epoch": 4.56, + "learning_rate": 8.22778338190204e-06, + "loss": 0.016, + "step": 34812 + }, + { + "epoch": 4.56, + "learning_rate": 8.22726145386481e-06, + "loss": 0.0188, + "step": 34813 + }, + { + "epoch": 4.56, + "learning_rate": 8.226739530813319e-06, + "loss": 0.0202, + "step": 34814 + }, + { + "epoch": 4.56, + "learning_rate": 8.226217612749025e-06, + "loss": 0.0214, + "step": 34815 + }, + { + "epoch": 4.56, + "learning_rate": 8.2256956996734e-06, + "loss": 0.0221, + "step": 34816 + }, + { + "epoch": 4.56, + "learning_rate": 8.225173791587912e-06, + "loss": 0.0163, + "step": 34817 + }, + { + "epoch": 4.56, + "learning_rate": 8.224651888494028e-06, + "loss": 0.0214, + "step": 34818 + }, + { + "epoch": 4.56, + "learning_rate": 8.224129990393217e-06, + "loss": 0.0193, + "step": 34819 + }, + { + "epoch": 4.56, + "learning_rate": 8.223608097286946e-06, + "loss": 0.0193, + "step": 34820 + }, + { + "epoch": 4.56, + "learning_rate": 8.223086209176674e-06, + "loss": 0.0239, + "step": 34821 + }, + { + "epoch": 4.56, + "learning_rate": 8.222564326063886e-06, + "loss": 0.0187, + "step": 34822 + }, + { + "epoch": 4.56, + "learning_rate": 8.222042447950039e-06, + "loss": 0.0221, + "step": 34823 + }, + { + "epoch": 4.56, + "learning_rate": 8.221520574836606e-06, + "loss": 0.0206, + "step": 34824 + }, + { + "epoch": 4.56, + "learning_rate": 8.220998706725047e-06, + "loss": 0.0236, + "step": 34825 + }, + { + "epoch": 4.56, + "learning_rate": 8.22047684361684e-06, + "loss": 0.0203, + "step": 34826 + }, + { + "epoch": 4.56, + "learning_rate": 8.219954985513442e-06, + "loss": 0.0176, + "step": 34827 + }, + { + "epoch": 4.56, + "learning_rate": 8.219433132416327e-06, + "loss": 0.0183, + "step": 34828 + }, + { + "epoch": 4.56, + "learning_rate": 8.218911284326963e-06, + "loss": 0.0234, + "step": 34829 + }, + { + "epoch": 4.56, + "learning_rate": 8.218389441246815e-06, + "loss": 0.0258, + "step": 34830 + }, + { + "epoch": 4.56, + "learning_rate": 8.217867603177353e-06, + "loss": 0.0176, + "step": 34831 + }, + { + "epoch": 4.56, + "learning_rate": 8.21734577012004e-06, + "loss": 0.0159, + "step": 34832 + }, + { + "epoch": 4.56, + "learning_rate": 8.21682394207635e-06, + "loss": 0.0189, + "step": 34833 + }, + { + "epoch": 4.56, + "learning_rate": 8.216302119047748e-06, + "loss": 0.0194, + "step": 34834 + }, + { + "epoch": 4.56, + "learning_rate": 8.2157803010357e-06, + "loss": 0.0205, + "step": 34835 + }, + { + "epoch": 4.56, + "learning_rate": 8.215258488041676e-06, + "loss": 0.0203, + "step": 34836 + }, + { + "epoch": 4.56, + "learning_rate": 8.214736680067142e-06, + "loss": 0.0247, + "step": 34837 + }, + { + "epoch": 4.56, + "learning_rate": 8.214214877113566e-06, + "loss": 0.0242, + "step": 34838 + }, + { + "epoch": 4.56, + "learning_rate": 8.213693079182417e-06, + "loss": 0.0207, + "step": 34839 + }, + { + "epoch": 4.56, + "learning_rate": 8.21317128627516e-06, + "loss": 0.0236, + "step": 34840 + }, + { + "epoch": 4.56, + "learning_rate": 8.212649498393264e-06, + "loss": 0.0158, + "step": 34841 + }, + { + "epoch": 4.56, + "learning_rate": 8.212127715538193e-06, + "loss": 0.0183, + "step": 34842 + }, + { + "epoch": 4.56, + "learning_rate": 8.211605937711417e-06, + "loss": 0.0201, + "step": 34843 + }, + { + "epoch": 4.56, + "learning_rate": 8.211084164914407e-06, + "loss": 0.0176, + "step": 34844 + }, + { + "epoch": 4.56, + "learning_rate": 8.21056239714863e-06, + "loss": 0.0218, + "step": 34845 + }, + { + "epoch": 4.56, + "learning_rate": 8.210040634415547e-06, + "loss": 0.0205, + "step": 34846 + }, + { + "epoch": 4.56, + "learning_rate": 8.209518876716632e-06, + "loss": 0.0173, + "step": 34847 + }, + { + "epoch": 4.56, + "learning_rate": 8.208997124053348e-06, + "loss": 0.0188, + "step": 34848 + }, + { + "epoch": 4.56, + "learning_rate": 8.208475376427167e-06, + "loss": 0.021, + "step": 34849 + }, + { + "epoch": 4.57, + "learning_rate": 8.20795363383955e-06, + "loss": 0.0185, + "step": 34850 + }, + { + "epoch": 4.57, + "learning_rate": 8.207431896291969e-06, + "loss": 0.0158, + "step": 34851 + }, + { + "epoch": 4.57, + "learning_rate": 8.206910163785892e-06, + "loss": 0.0215, + "step": 34852 + }, + { + "epoch": 4.57, + "learning_rate": 8.206388436322784e-06, + "loss": 0.0188, + "step": 34853 + }, + { + "epoch": 4.57, + "learning_rate": 8.205866713904109e-06, + "loss": 0.0199, + "step": 34854 + }, + { + "epoch": 4.57, + "learning_rate": 8.205344996531343e-06, + "loss": 0.0171, + "step": 34855 + }, + { + "epoch": 4.57, + "learning_rate": 8.204823284205947e-06, + "loss": 0.0164, + "step": 34856 + }, + { + "epoch": 4.57, + "learning_rate": 8.20430157692939e-06, + "loss": 0.02, + "step": 34857 + }, + { + "epoch": 4.57, + "learning_rate": 8.20377987470314e-06, + "loss": 0.0169, + "step": 34858 + }, + { + "epoch": 4.57, + "learning_rate": 8.203258177528663e-06, + "loss": 0.0158, + "step": 34859 + }, + { + "epoch": 4.57, + "learning_rate": 8.202736485407428e-06, + "loss": 0.0167, + "step": 34860 + }, + { + "epoch": 4.57, + "learning_rate": 8.202214798340898e-06, + "loss": 0.0195, + "step": 34861 + }, + { + "epoch": 4.57, + "learning_rate": 8.201693116330547e-06, + "loss": 0.0169, + "step": 34862 + }, + { + "epoch": 4.57, + "learning_rate": 8.201171439377836e-06, + "loss": 0.0191, + "step": 34863 + }, + { + "epoch": 4.57, + "learning_rate": 8.200649767484236e-06, + "loss": 0.0191, + "step": 34864 + }, + { + "epoch": 4.57, + "learning_rate": 8.200128100651205e-06, + "loss": 0.0213, + "step": 34865 + }, + { + "epoch": 4.57, + "learning_rate": 8.199606438880227e-06, + "loss": 0.0204, + "step": 34866 + }, + { + "epoch": 4.57, + "learning_rate": 8.199084782172757e-06, + "loss": 0.0196, + "step": 34867 + }, + { + "epoch": 4.57, + "learning_rate": 8.198563130530266e-06, + "loss": 0.0172, + "step": 34868 + }, + { + "epoch": 4.57, + "learning_rate": 8.198041483954218e-06, + "loss": 0.0224, + "step": 34869 + }, + { + "epoch": 4.57, + "learning_rate": 8.197519842446087e-06, + "loss": 0.0169, + "step": 34870 + }, + { + "epoch": 4.57, + "learning_rate": 8.196998206007331e-06, + "loss": 0.0197, + "step": 34871 + }, + { + "epoch": 4.57, + "learning_rate": 8.196476574639423e-06, + "loss": 0.0247, + "step": 34872 + }, + { + "epoch": 4.57, + "learning_rate": 8.195954948343829e-06, + "loss": 0.0245, + "step": 34873 + }, + { + "epoch": 4.57, + "learning_rate": 8.195433327122015e-06, + "loss": 0.0215, + "step": 34874 + }, + { + "epoch": 4.57, + "learning_rate": 8.194911710975451e-06, + "loss": 0.0226, + "step": 34875 + }, + { + "epoch": 4.57, + "learning_rate": 8.194390099905596e-06, + "loss": 0.0169, + "step": 34876 + }, + { + "epoch": 4.57, + "learning_rate": 8.193868493913926e-06, + "loss": 0.0176, + "step": 34877 + }, + { + "epoch": 4.57, + "learning_rate": 8.193346893001904e-06, + "loss": 0.0245, + "step": 34878 + }, + { + "epoch": 4.57, + "learning_rate": 8.192825297170998e-06, + "loss": 0.0177, + "step": 34879 + }, + { + "epoch": 4.57, + "learning_rate": 8.192303706422677e-06, + "loss": 0.0188, + "step": 34880 + }, + { + "epoch": 4.57, + "learning_rate": 8.191782120758405e-06, + "loss": 0.0192, + "step": 34881 + }, + { + "epoch": 4.57, + "learning_rate": 8.191260540179646e-06, + "loss": 0.0163, + "step": 34882 + }, + { + "epoch": 4.57, + "learning_rate": 8.190738964687875e-06, + "loss": 0.0204, + "step": 34883 + }, + { + "epoch": 4.57, + "learning_rate": 8.190217394284552e-06, + "loss": 0.0213, + "step": 34884 + }, + { + "epoch": 4.57, + "learning_rate": 8.189695828971147e-06, + "loss": 0.0174, + "step": 34885 + }, + { + "epoch": 4.57, + "learning_rate": 8.189174268749124e-06, + "loss": 0.0161, + "step": 34886 + }, + { + "epoch": 4.57, + "learning_rate": 8.188652713619951e-06, + "loss": 0.0234, + "step": 34887 + }, + { + "epoch": 4.57, + "learning_rate": 8.188131163585097e-06, + "loss": 0.027, + "step": 34888 + }, + { + "epoch": 4.57, + "learning_rate": 8.18760961864603e-06, + "loss": 0.0199, + "step": 34889 + }, + { + "epoch": 4.57, + "learning_rate": 8.187088078804212e-06, + "loss": 0.0189, + "step": 34890 + }, + { + "epoch": 4.57, + "learning_rate": 8.186566544061115e-06, + "loss": 0.0193, + "step": 34891 + }, + { + "epoch": 4.57, + "learning_rate": 8.1860450144182e-06, + "loss": 0.0206, + "step": 34892 + }, + { + "epoch": 4.57, + "learning_rate": 8.185523489876937e-06, + "loss": 0.021, + "step": 34893 + }, + { + "epoch": 4.57, + "learning_rate": 8.185001970438797e-06, + "loss": 0.0219, + "step": 34894 + }, + { + "epoch": 4.57, + "learning_rate": 8.184480456105237e-06, + "loss": 0.0204, + "step": 34895 + }, + { + "epoch": 4.57, + "learning_rate": 8.183958946877732e-06, + "loss": 0.0163, + "step": 34896 + }, + { + "epoch": 4.57, + "learning_rate": 8.183437442757745e-06, + "loss": 0.0205, + "step": 34897 + }, + { + "epoch": 4.57, + "learning_rate": 8.182915943746739e-06, + "loss": 0.0234, + "step": 34898 + }, + { + "epoch": 4.57, + "learning_rate": 8.182394449846191e-06, + "loss": 0.0198, + "step": 34899 + }, + { + "epoch": 4.57, + "learning_rate": 8.18187296105756e-06, + "loss": 0.0205, + "step": 34900 + }, + { + "epoch": 4.57, + "learning_rate": 8.181351477382313e-06, + "loss": 0.0189, + "step": 34901 + }, + { + "epoch": 4.57, + "learning_rate": 8.180829998821921e-06, + "loss": 0.0184, + "step": 34902 + }, + { + "epoch": 4.57, + "learning_rate": 8.180308525377848e-06, + "loss": 0.0212, + "step": 34903 + }, + { + "epoch": 4.57, + "learning_rate": 8.179787057051559e-06, + "loss": 0.0226, + "step": 34904 + }, + { + "epoch": 4.57, + "learning_rate": 8.17926559384452e-06, + "loss": 0.0194, + "step": 34905 + }, + { + "epoch": 4.57, + "learning_rate": 8.178744135758204e-06, + "loss": 0.0206, + "step": 34906 + }, + { + "epoch": 4.57, + "learning_rate": 8.17822268279407e-06, + "loss": 0.0201, + "step": 34907 + }, + { + "epoch": 4.57, + "learning_rate": 8.177701234953586e-06, + "loss": 0.018, + "step": 34908 + }, + { + "epoch": 4.57, + "learning_rate": 8.177179792238221e-06, + "loss": 0.0225, + "step": 34909 + }, + { + "epoch": 4.57, + "learning_rate": 8.176658354649443e-06, + "loss": 0.02, + "step": 34910 + }, + { + "epoch": 4.57, + "learning_rate": 8.176136922188715e-06, + "loss": 0.0172, + "step": 34911 + }, + { + "epoch": 4.57, + "learning_rate": 8.175615494857508e-06, + "loss": 0.0183, + "step": 34912 + }, + { + "epoch": 4.57, + "learning_rate": 8.175094072657281e-06, + "loss": 0.0182, + "step": 34913 + }, + { + "epoch": 4.57, + "learning_rate": 8.174572655589508e-06, + "loss": 0.0165, + "step": 34914 + }, + { + "epoch": 4.57, + "learning_rate": 8.174051243655649e-06, + "loss": 0.0176, + "step": 34915 + }, + { + "epoch": 4.57, + "learning_rate": 8.173529836857174e-06, + "loss": 0.0217, + "step": 34916 + }, + { + "epoch": 4.57, + "learning_rate": 8.173008435195553e-06, + "loss": 0.0186, + "step": 34917 + }, + { + "epoch": 4.57, + "learning_rate": 8.172487038672244e-06, + "loss": 0.0204, + "step": 34918 + }, + { + "epoch": 4.57, + "learning_rate": 8.171965647288717e-06, + "loss": 0.023, + "step": 34919 + }, + { + "epoch": 4.57, + "learning_rate": 8.171444261046443e-06, + "loss": 0.0143, + "step": 34920 + }, + { + "epoch": 4.57, + "learning_rate": 8.170922879946883e-06, + "loss": 0.0176, + "step": 34921 + }, + { + "epoch": 4.57, + "learning_rate": 8.170401503991505e-06, + "loss": 0.0195, + "step": 34922 + }, + { + "epoch": 4.57, + "learning_rate": 8.169880133181777e-06, + "loss": 0.0181, + "step": 34923 + }, + { + "epoch": 4.57, + "learning_rate": 8.169358767519163e-06, + "loss": 0.0175, + "step": 34924 + }, + { + "epoch": 4.57, + "learning_rate": 8.16883740700513e-06, + "loss": 0.0165, + "step": 34925 + }, + { + "epoch": 4.58, + "learning_rate": 8.168316051641142e-06, + "loss": 0.0222, + "step": 34926 + }, + { + "epoch": 4.58, + "learning_rate": 8.167794701428672e-06, + "loss": 0.0161, + "step": 34927 + }, + { + "epoch": 4.58, + "learning_rate": 8.167273356369177e-06, + "loss": 0.0202, + "step": 34928 + }, + { + "epoch": 4.58, + "learning_rate": 8.16675201646413e-06, + "loss": 0.0191, + "step": 34929 + }, + { + "epoch": 4.58, + "learning_rate": 8.166230681714993e-06, + "loss": 0.0207, + "step": 34930 + }, + { + "epoch": 4.58, + "learning_rate": 8.16570935212324e-06, + "loss": 0.0184, + "step": 34931 + }, + { + "epoch": 4.58, + "learning_rate": 8.165188027690327e-06, + "loss": 0.0173, + "step": 34932 + }, + { + "epoch": 4.58, + "learning_rate": 8.164666708417729e-06, + "loss": 0.0215, + "step": 34933 + }, + { + "epoch": 4.58, + "learning_rate": 8.164145394306903e-06, + "loss": 0.0195, + "step": 34934 + }, + { + "epoch": 4.58, + "learning_rate": 8.163624085359325e-06, + "loss": 0.0191, + "step": 34935 + }, + { + "epoch": 4.58, + "learning_rate": 8.163102781576456e-06, + "loss": 0.018, + "step": 34936 + }, + { + "epoch": 4.58, + "learning_rate": 8.162581482959759e-06, + "loss": 0.0229, + "step": 34937 + }, + { + "epoch": 4.58, + "learning_rate": 8.162060189510708e-06, + "loss": 0.016, + "step": 34938 + }, + { + "epoch": 4.58, + "learning_rate": 8.161538901230762e-06, + "loss": 0.0186, + "step": 34939 + }, + { + "epoch": 4.58, + "learning_rate": 8.161017618121392e-06, + "loss": 0.0179, + "step": 34940 + }, + { + "epoch": 4.58, + "learning_rate": 8.16049634018406e-06, + "loss": 0.0182, + "step": 34941 + }, + { + "epoch": 4.58, + "learning_rate": 8.159975067420233e-06, + "loss": 0.0197, + "step": 34942 + }, + { + "epoch": 4.58, + "learning_rate": 8.159453799831379e-06, + "loss": 0.0157, + "step": 34943 + }, + { + "epoch": 4.58, + "learning_rate": 8.158932537418966e-06, + "loss": 0.0193, + "step": 34944 + }, + { + "epoch": 4.58, + "learning_rate": 8.158411280184453e-06, + "loss": 0.022, + "step": 34945 + }, + { + "epoch": 4.58, + "learning_rate": 8.157890028129317e-06, + "loss": 0.0178, + "step": 34946 + }, + { + "epoch": 4.58, + "learning_rate": 8.157368781255011e-06, + "loss": 0.0186, + "step": 34947 + }, + { + "epoch": 4.58, + "learning_rate": 8.15684753956301e-06, + "loss": 0.0184, + "step": 34948 + }, + { + "epoch": 4.58, + "learning_rate": 8.156326303054776e-06, + "loss": 0.023, + "step": 34949 + }, + { + "epoch": 4.58, + "learning_rate": 8.155805071731777e-06, + "loss": 0.0193, + "step": 34950 + }, + { + "epoch": 4.58, + "learning_rate": 8.155283845595476e-06, + "loss": 0.0209, + "step": 34951 + }, + { + "epoch": 4.58, + "learning_rate": 8.15476262464734e-06, + "loss": 0.0216, + "step": 34952 + }, + { + "epoch": 4.58, + "learning_rate": 8.154241408888837e-06, + "loss": 0.0219, + "step": 34953 + }, + { + "epoch": 4.58, + "learning_rate": 8.153720198321436e-06, + "loss": 0.0166, + "step": 34954 + }, + { + "epoch": 4.58, + "learning_rate": 8.153198992946593e-06, + "loss": 0.0214, + "step": 34955 + }, + { + "epoch": 4.58, + "learning_rate": 8.152677792765784e-06, + "loss": 0.0217, + "step": 34956 + }, + { + "epoch": 4.58, + "learning_rate": 8.152156597780468e-06, + "loss": 0.0184, + "step": 34957 + }, + { + "epoch": 4.58, + "learning_rate": 8.151635407992112e-06, + "loss": 0.0201, + "step": 34958 + }, + { + "epoch": 4.58, + "learning_rate": 8.151114223402186e-06, + "loss": 0.0205, + "step": 34959 + }, + { + "epoch": 4.58, + "learning_rate": 8.15059304401215e-06, + "loss": 0.0198, + "step": 34960 + }, + { + "epoch": 4.58, + "learning_rate": 8.150071869823476e-06, + "loss": 0.0195, + "step": 34961 + }, + { + "epoch": 4.58, + "learning_rate": 8.149550700837623e-06, + "loss": 0.02, + "step": 34962 + }, + { + "epoch": 4.58, + "learning_rate": 8.149029537056058e-06, + "loss": 0.0197, + "step": 34963 + }, + { + "epoch": 4.58, + "learning_rate": 8.148508378480254e-06, + "loss": 0.0177, + "step": 34964 + }, + { + "epoch": 4.58, + "learning_rate": 8.147987225111669e-06, + "loss": 0.0167, + "step": 34965 + }, + { + "epoch": 4.58, + "learning_rate": 8.14746607695177e-06, + "loss": 0.0186, + "step": 34966 + }, + { + "epoch": 4.58, + "learning_rate": 8.146944934002028e-06, + "loss": 0.018, + "step": 34967 + }, + { + "epoch": 4.58, + "learning_rate": 8.146423796263903e-06, + "loss": 0.0195, + "step": 34968 + }, + { + "epoch": 4.58, + "learning_rate": 8.145902663738863e-06, + "loss": 0.0228, + "step": 34969 + }, + { + "epoch": 4.58, + "learning_rate": 8.145381536428372e-06, + "loss": 0.0187, + "step": 34970 + }, + { + "epoch": 4.58, + "learning_rate": 8.144860414333897e-06, + "loss": 0.0265, + "step": 34971 + }, + { + "epoch": 4.58, + "learning_rate": 8.144339297456903e-06, + "loss": 0.0234, + "step": 34972 + }, + { + "epoch": 4.58, + "learning_rate": 8.143818185798855e-06, + "loss": 0.0237, + "step": 34973 + }, + { + "epoch": 4.58, + "learning_rate": 8.143297079361218e-06, + "loss": 0.0211, + "step": 34974 + }, + { + "epoch": 4.58, + "learning_rate": 8.142775978145463e-06, + "loss": 0.0177, + "step": 34975 + }, + { + "epoch": 4.58, + "learning_rate": 8.142254882153048e-06, + "loss": 0.0175, + "step": 34976 + }, + { + "epoch": 4.58, + "learning_rate": 8.141733791385446e-06, + "loss": 0.018, + "step": 34977 + }, + { + "epoch": 4.58, + "learning_rate": 8.141212705844117e-06, + "loss": 0.0217, + "step": 34978 + }, + { + "epoch": 4.58, + "learning_rate": 8.140691625530528e-06, + "loss": 0.0202, + "step": 34979 + }, + { + "epoch": 4.58, + "learning_rate": 8.140170550446146e-06, + "loss": 0.0201, + "step": 34980 + }, + { + "epoch": 4.58, + "learning_rate": 8.139649480592435e-06, + "loss": 0.0223, + "step": 34981 + }, + { + "epoch": 4.58, + "learning_rate": 8.139128415970862e-06, + "loss": 0.0186, + "step": 34982 + }, + { + "epoch": 4.58, + "learning_rate": 8.138607356582888e-06, + "loss": 0.0199, + "step": 34983 + }, + { + "epoch": 4.58, + "learning_rate": 8.138086302429986e-06, + "loss": 0.0196, + "step": 34984 + }, + { + "epoch": 4.58, + "learning_rate": 8.137565253513611e-06, + "loss": 0.0215, + "step": 34985 + }, + { + "epoch": 4.58, + "learning_rate": 8.137044209835237e-06, + "loss": 0.0205, + "step": 34986 + }, + { + "epoch": 4.58, + "learning_rate": 8.136523171396326e-06, + "loss": 0.0172, + "step": 34987 + }, + { + "epoch": 4.58, + "learning_rate": 8.136002138198348e-06, + "loss": 0.0237, + "step": 34988 + }, + { + "epoch": 4.58, + "learning_rate": 8.135481110242763e-06, + "loss": 0.0247, + "step": 34989 + }, + { + "epoch": 4.58, + "learning_rate": 8.134960087531038e-06, + "loss": 0.0154, + "step": 34990 + }, + { + "epoch": 4.58, + "learning_rate": 8.134439070064639e-06, + "loss": 0.0201, + "step": 34991 + }, + { + "epoch": 4.58, + "learning_rate": 8.133918057845031e-06, + "loss": 0.02, + "step": 34992 + }, + { + "epoch": 4.58, + "learning_rate": 8.133397050873677e-06, + "loss": 0.0185, + "step": 34993 + }, + { + "epoch": 4.58, + "learning_rate": 8.132876049152045e-06, + "loss": 0.0211, + "step": 34994 + }, + { + "epoch": 4.58, + "learning_rate": 8.132355052681603e-06, + "loss": 0.0165, + "step": 34995 + }, + { + "epoch": 4.58, + "learning_rate": 8.131834061463808e-06, + "loss": 0.0182, + "step": 34996 + }, + { + "epoch": 4.58, + "learning_rate": 8.13131307550013e-06, + "loss": 0.0224, + "step": 34997 + }, + { + "epoch": 4.58, + "learning_rate": 8.13079209479204e-06, + "loss": 0.0176, + "step": 34998 + }, + { + "epoch": 4.58, + "learning_rate": 8.130271119340994e-06, + "loss": 0.0204, + "step": 34999 + }, + { + "epoch": 4.58, + "learning_rate": 8.129750149148463e-06, + "loss": 0.011, + "step": 35000 + }, + { + "epoch": 4.58, + "learning_rate": 8.129229184215906e-06, + "loss": 0.0255, + "step": 35001 + }, + { + "epoch": 4.59, + "learning_rate": 8.128708224544795e-06, + "loss": 0.0191, + "step": 35002 + }, + { + "epoch": 4.59, + "learning_rate": 8.128187270136594e-06, + "loss": 0.018, + "step": 35003 + }, + { + "epoch": 4.59, + "learning_rate": 8.127666320992765e-06, + "loss": 0.019, + "step": 35004 + }, + { + "epoch": 4.59, + "learning_rate": 8.127145377114776e-06, + "loss": 0.0176, + "step": 35005 + }, + { + "epoch": 4.59, + "learning_rate": 8.126624438504088e-06, + "loss": 0.0191, + "step": 35006 + }, + { + "epoch": 4.59, + "learning_rate": 8.126103505162169e-06, + "loss": 0.0168, + "step": 35007 + }, + { + "epoch": 4.59, + "learning_rate": 8.125582577090486e-06, + "loss": 0.0228, + "step": 35008 + }, + { + "epoch": 4.59, + "learning_rate": 8.125061654290503e-06, + "loss": 0.0138, + "step": 35009 + }, + { + "epoch": 4.59, + "learning_rate": 8.124540736763682e-06, + "loss": 0.0164, + "step": 35010 + }, + { + "epoch": 4.59, + "learning_rate": 8.124019824511494e-06, + "loss": 0.0182, + "step": 35011 + }, + { + "epoch": 4.59, + "learning_rate": 8.123498917535395e-06, + "loss": 0.0191, + "step": 35012 + }, + { + "epoch": 4.59, + "learning_rate": 8.12297801583686e-06, + "loss": 0.0243, + "step": 35013 + }, + { + "epoch": 4.59, + "learning_rate": 8.122457119417347e-06, + "loss": 0.0174, + "step": 35014 + }, + { + "epoch": 4.59, + "learning_rate": 8.121936228278325e-06, + "loss": 0.0197, + "step": 35015 + }, + { + "epoch": 4.59, + "learning_rate": 8.121415342421256e-06, + "loss": 0.0161, + "step": 35016 + }, + { + "epoch": 4.59, + "learning_rate": 8.120894461847605e-06, + "loss": 0.0208, + "step": 35017 + }, + { + "epoch": 4.59, + "learning_rate": 8.12037358655884e-06, + "loss": 0.0195, + "step": 35018 + }, + { + "epoch": 4.59, + "learning_rate": 8.119852716556424e-06, + "loss": 0.0184, + "step": 35019 + }, + { + "epoch": 4.59, + "learning_rate": 8.119331851841821e-06, + "loss": 0.019, + "step": 35020 + }, + { + "epoch": 4.59, + "learning_rate": 8.1188109924165e-06, + "loss": 0.0139, + "step": 35021 + }, + { + "epoch": 4.59, + "learning_rate": 8.11829013828192e-06, + "loss": 0.0191, + "step": 35022 + }, + { + "epoch": 4.59, + "learning_rate": 8.11776928943955e-06, + "loss": 0.0232, + "step": 35023 + }, + { + "epoch": 4.59, + "learning_rate": 8.117248445890854e-06, + "loss": 0.0147, + "step": 35024 + }, + { + "epoch": 4.59, + "learning_rate": 8.116727607637296e-06, + "loss": 0.0181, + "step": 35025 + }, + { + "epoch": 4.59, + "learning_rate": 8.116206774680343e-06, + "loss": 0.021, + "step": 35026 + }, + { + "epoch": 4.59, + "learning_rate": 8.115685947021456e-06, + "loss": 0.0202, + "step": 35027 + }, + { + "epoch": 4.59, + "learning_rate": 8.115165124662104e-06, + "loss": 0.0195, + "step": 35028 + }, + { + "epoch": 4.59, + "learning_rate": 8.114644307603742e-06, + "loss": 0.0206, + "step": 35029 + }, + { + "epoch": 4.59, + "learning_rate": 8.114123495847851e-06, + "loss": 0.0179, + "step": 35030 + }, + { + "epoch": 4.59, + "learning_rate": 8.113602689395884e-06, + "loss": 0.0236, + "step": 35031 + }, + { + "epoch": 4.59, + "learning_rate": 8.113081888249311e-06, + "loss": 0.0161, + "step": 35032 + }, + { + "epoch": 4.59, + "learning_rate": 8.112561092409593e-06, + "loss": 0.0231, + "step": 35033 + }, + { + "epoch": 4.59, + "learning_rate": 8.112040301878198e-06, + "loss": 0.02, + "step": 35034 + }, + { + "epoch": 4.59, + "learning_rate": 8.111519516656588e-06, + "loss": 0.0164, + "step": 35035 + }, + { + "epoch": 4.59, + "learning_rate": 8.110998736746229e-06, + "loss": 0.0149, + "step": 35036 + }, + { + "epoch": 4.59, + "learning_rate": 8.110477962148586e-06, + "loss": 0.0238, + "step": 35037 + }, + { + "epoch": 4.59, + "learning_rate": 8.10995719286512e-06, + "loss": 0.0175, + "step": 35038 + }, + { + "epoch": 4.59, + "learning_rate": 8.109436428897299e-06, + "loss": 0.02, + "step": 35039 + }, + { + "epoch": 4.59, + "learning_rate": 8.10891567024659e-06, + "loss": 0.0137, + "step": 35040 + }, + { + "epoch": 4.59, + "learning_rate": 8.108394916914455e-06, + "loss": 0.0232, + "step": 35041 + }, + { + "epoch": 4.59, + "learning_rate": 8.10787416890236e-06, + "loss": 0.018, + "step": 35042 + }, + { + "epoch": 4.59, + "learning_rate": 8.107353426211764e-06, + "loss": 0.0177, + "step": 35043 + }, + { + "epoch": 4.59, + "learning_rate": 8.106832688844138e-06, + "loss": 0.0185, + "step": 35044 + }, + { + "epoch": 4.59, + "learning_rate": 8.106311956800947e-06, + "loss": 0.0129, + "step": 35045 + }, + { + "epoch": 4.59, + "learning_rate": 8.105791230083649e-06, + "loss": 0.0199, + "step": 35046 + }, + { + "epoch": 4.59, + "learning_rate": 8.105270508693715e-06, + "loss": 0.0178, + "step": 35047 + }, + { + "epoch": 4.59, + "learning_rate": 8.104749792632603e-06, + "loss": 0.022, + "step": 35048 + }, + { + "epoch": 4.59, + "learning_rate": 8.104229081901785e-06, + "loss": 0.0144, + "step": 35049 + }, + { + "epoch": 4.59, + "learning_rate": 8.103708376502717e-06, + "loss": 0.0195, + "step": 35050 + }, + { + "epoch": 4.59, + "learning_rate": 8.10318767643687e-06, + "loss": 0.0185, + "step": 35051 + }, + { + "epoch": 4.59, + "learning_rate": 8.102666981705709e-06, + "loss": 0.0219, + "step": 35052 + }, + { + "epoch": 4.59, + "learning_rate": 8.102146292310695e-06, + "loss": 0.0203, + "step": 35053 + }, + { + "epoch": 4.59, + "learning_rate": 8.101625608253293e-06, + "loss": 0.0185, + "step": 35054 + }, + { + "epoch": 4.59, + "learning_rate": 8.10110492953497e-06, + "loss": 0.0183, + "step": 35055 + }, + { + "epoch": 4.59, + "learning_rate": 8.100584256157186e-06, + "loss": 0.0215, + "step": 35056 + }, + { + "epoch": 4.59, + "learning_rate": 8.100063588121409e-06, + "loss": 0.0182, + "step": 35057 + }, + { + "epoch": 4.59, + "learning_rate": 8.0995429254291e-06, + "loss": 0.018, + "step": 35058 + }, + { + "epoch": 4.59, + "learning_rate": 8.099022268081727e-06, + "loss": 0.018, + "step": 35059 + }, + { + "epoch": 4.59, + "learning_rate": 8.098501616080752e-06, + "loss": 0.0142, + "step": 35060 + }, + { + "epoch": 4.59, + "learning_rate": 8.097980969427639e-06, + "loss": 0.0184, + "step": 35061 + }, + { + "epoch": 4.59, + "learning_rate": 8.097460328123852e-06, + "loss": 0.0228, + "step": 35062 + }, + { + "epoch": 4.59, + "learning_rate": 8.096939692170861e-06, + "loss": 0.0164, + "step": 35063 + }, + { + "epoch": 4.59, + "learning_rate": 8.096419061570123e-06, + "loss": 0.0214, + "step": 35064 + }, + { + "epoch": 4.59, + "learning_rate": 8.095898436323108e-06, + "loss": 0.0168, + "step": 35065 + }, + { + "epoch": 4.59, + "learning_rate": 8.095377816431274e-06, + "loss": 0.0252, + "step": 35066 + }, + { + "epoch": 4.59, + "learning_rate": 8.094857201896088e-06, + "loss": 0.0205, + "step": 35067 + }, + { + "epoch": 4.59, + "learning_rate": 8.094336592719018e-06, + "loss": 0.0215, + "step": 35068 + }, + { + "epoch": 4.59, + "learning_rate": 8.093815988901521e-06, + "loss": 0.0174, + "step": 35069 + }, + { + "epoch": 4.59, + "learning_rate": 8.093295390445069e-06, + "loss": 0.0208, + "step": 35070 + }, + { + "epoch": 4.59, + "learning_rate": 8.09277479735112e-06, + "loss": 0.0216, + "step": 35071 + }, + { + "epoch": 4.59, + "learning_rate": 8.092254209621137e-06, + "loss": 0.0156, + "step": 35072 + }, + { + "epoch": 4.59, + "learning_rate": 8.091733627256591e-06, + "loss": 0.0216, + "step": 35073 + }, + { + "epoch": 4.59, + "learning_rate": 8.091213050258942e-06, + "loss": 0.0121, + "step": 35074 + }, + { + "epoch": 4.59, + "learning_rate": 8.090692478629655e-06, + "loss": 0.0254, + "step": 35075 + }, + { + "epoch": 4.59, + "learning_rate": 8.090171912370196e-06, + "loss": 0.0203, + "step": 35076 + }, + { + "epoch": 4.59, + "learning_rate": 8.089651351482023e-06, + "loss": 0.0232, + "step": 35077 + }, + { + "epoch": 4.59, + "learning_rate": 8.089130795966607e-06, + "loss": 0.0158, + "step": 35078 + }, + { + "epoch": 4.6, + "learning_rate": 8.088610245825407e-06, + "loss": 0.0228, + "step": 35079 + }, + { + "epoch": 4.6, + "learning_rate": 8.08808970105989e-06, + "loss": 0.0222, + "step": 35080 + }, + { + "epoch": 4.6, + "learning_rate": 8.087569161671517e-06, + "loss": 0.0187, + "step": 35081 + }, + { + "epoch": 4.6, + "learning_rate": 8.087048627661755e-06, + "loss": 0.0195, + "step": 35082 + }, + { + "epoch": 4.6, + "learning_rate": 8.086528099032064e-06, + "loss": 0.0213, + "step": 35083 + }, + { + "epoch": 4.6, + "learning_rate": 8.086007575783915e-06, + "loss": 0.0132, + "step": 35084 + }, + { + "epoch": 4.6, + "learning_rate": 8.085487057918766e-06, + "loss": 0.0177, + "step": 35085 + }, + { + "epoch": 4.6, + "learning_rate": 8.084966545438084e-06, + "loss": 0.0176, + "step": 35086 + }, + { + "epoch": 4.6, + "learning_rate": 8.084446038343329e-06, + "loss": 0.0183, + "step": 35087 + }, + { + "epoch": 4.6, + "learning_rate": 8.08392553663597e-06, + "loss": 0.0171, + "step": 35088 + }, + { + "epoch": 4.6, + "learning_rate": 8.083405040317469e-06, + "loss": 0.0233, + "step": 35089 + }, + { + "epoch": 4.6, + "learning_rate": 8.082884549389288e-06, + "loss": 0.0179, + "step": 35090 + }, + { + "epoch": 4.6, + "learning_rate": 8.082364063852892e-06, + "loss": 0.0281, + "step": 35091 + }, + { + "epoch": 4.6, + "learning_rate": 8.081843583709744e-06, + "loss": 0.0247, + "step": 35092 + }, + { + "epoch": 4.6, + "learning_rate": 8.081323108961312e-06, + "loss": 0.0211, + "step": 35093 + }, + { + "epoch": 4.6, + "learning_rate": 8.080802639609049e-06, + "loss": 0.0152, + "step": 35094 + }, + { + "epoch": 4.6, + "learning_rate": 8.080282175654432e-06, + "loss": 0.0219, + "step": 35095 + }, + { + "epoch": 4.6, + "learning_rate": 8.07976171709892e-06, + "loss": 0.0182, + "step": 35096 + }, + { + "epoch": 4.6, + "learning_rate": 8.079241263943975e-06, + "loss": 0.0226, + "step": 35097 + }, + { + "epoch": 4.6, + "learning_rate": 8.07872081619106e-06, + "loss": 0.0187, + "step": 35098 + }, + { + "epoch": 4.6, + "learning_rate": 8.078200373841644e-06, + "loss": 0.0164, + "step": 35099 + }, + { + "epoch": 4.6, + "learning_rate": 8.077679936897185e-06, + "loss": 0.0233, + "step": 35100 + }, + { + "epoch": 4.6, + "learning_rate": 8.077159505359149e-06, + "loss": 0.0202, + "step": 35101 + }, + { + "epoch": 4.6, + "learning_rate": 8.076639079228999e-06, + "loss": 0.0217, + "step": 35102 + }, + { + "epoch": 4.6, + "learning_rate": 8.076118658508199e-06, + "loss": 0.0202, + "step": 35103 + }, + { + "epoch": 4.6, + "learning_rate": 8.075598243198213e-06, + "loss": 0.0215, + "step": 35104 + }, + { + "epoch": 4.6, + "learning_rate": 8.075077833300504e-06, + "loss": 0.0201, + "step": 35105 + }, + { + "epoch": 4.6, + "learning_rate": 8.074557428816535e-06, + "loss": 0.0196, + "step": 35106 + }, + { + "epoch": 4.6, + "learning_rate": 8.074037029747774e-06, + "loss": 0.0147, + "step": 35107 + }, + { + "epoch": 4.6, + "learning_rate": 8.073516636095679e-06, + "loss": 0.0184, + "step": 35108 + }, + { + "epoch": 4.6, + "learning_rate": 8.072996247861715e-06, + "loss": 0.0161, + "step": 35109 + }, + { + "epoch": 4.6, + "learning_rate": 8.07247586504735e-06, + "loss": 0.0218, + "step": 35110 + }, + { + "epoch": 4.6, + "learning_rate": 8.071955487654042e-06, + "loss": 0.0208, + "step": 35111 + }, + { + "epoch": 4.6, + "learning_rate": 8.071435115683257e-06, + "loss": 0.0183, + "step": 35112 + }, + { + "epoch": 4.6, + "learning_rate": 8.070914749136456e-06, + "loss": 0.0236, + "step": 35113 + }, + { + "epoch": 4.6, + "learning_rate": 8.070394388015109e-06, + "loss": 0.0248, + "step": 35114 + }, + { + "epoch": 4.6, + "learning_rate": 8.06987403232067e-06, + "loss": 0.0202, + "step": 35115 + }, + { + "epoch": 4.6, + "learning_rate": 8.069353682054608e-06, + "loss": 0.0192, + "step": 35116 + }, + { + "epoch": 4.6, + "learning_rate": 8.068833337218387e-06, + "loss": 0.0197, + "step": 35117 + }, + { + "epoch": 4.6, + "learning_rate": 8.068312997813472e-06, + "loss": 0.0221, + "step": 35118 + }, + { + "epoch": 4.6, + "learning_rate": 8.067792663841319e-06, + "loss": 0.0163, + "step": 35119 + }, + { + "epoch": 4.6, + "learning_rate": 8.067272335303401e-06, + "loss": 0.0199, + "step": 35120 + }, + { + "epoch": 4.6, + "learning_rate": 8.066752012201175e-06, + "loss": 0.0169, + "step": 35121 + }, + { + "epoch": 4.6, + "learning_rate": 8.066231694536106e-06, + "loss": 0.0214, + "step": 35122 + }, + { + "epoch": 4.6, + "learning_rate": 8.065711382309657e-06, + "loss": 0.0173, + "step": 35123 + }, + { + "epoch": 4.6, + "learning_rate": 8.065191075523291e-06, + "loss": 0.0234, + "step": 35124 + }, + { + "epoch": 4.6, + "learning_rate": 8.064670774178473e-06, + "loss": 0.0147, + "step": 35125 + }, + { + "epoch": 4.6, + "learning_rate": 8.064150478276665e-06, + "loss": 0.0148, + "step": 35126 + }, + { + "epoch": 4.6, + "learning_rate": 8.06363018781933e-06, + "loss": 0.0198, + "step": 35127 + }, + { + "epoch": 4.6, + "learning_rate": 8.063109902807933e-06, + "loss": 0.0211, + "step": 35128 + }, + { + "epoch": 4.6, + "learning_rate": 8.062589623243935e-06, + "loss": 0.0208, + "step": 35129 + }, + { + "epoch": 4.6, + "learning_rate": 8.062069349128804e-06, + "loss": 0.0171, + "step": 35130 + }, + { + "epoch": 4.6, + "learning_rate": 8.061549080463996e-06, + "loss": 0.0243, + "step": 35131 + }, + { + "epoch": 4.6, + "learning_rate": 8.061028817250979e-06, + "loss": 0.0232, + "step": 35132 + }, + { + "epoch": 4.6, + "learning_rate": 8.060508559491217e-06, + "loss": 0.0173, + "step": 35133 + }, + { + "epoch": 4.6, + "learning_rate": 8.059988307186168e-06, + "loss": 0.0199, + "step": 35134 + }, + { + "epoch": 4.6, + "learning_rate": 8.059468060337303e-06, + "loss": 0.0189, + "step": 35135 + }, + { + "epoch": 4.6, + "learning_rate": 8.058947818946079e-06, + "loss": 0.0196, + "step": 35136 + }, + { + "epoch": 4.6, + "learning_rate": 8.05842758301396e-06, + "loss": 0.0233, + "step": 35137 + }, + { + "epoch": 4.6, + "learning_rate": 8.057907352542406e-06, + "loss": 0.018, + "step": 35138 + }, + { + "epoch": 4.6, + "learning_rate": 8.05738712753289e-06, + "loss": 0.0154, + "step": 35139 + }, + { + "epoch": 4.6, + "learning_rate": 8.056866907986867e-06, + "loss": 0.0226, + "step": 35140 + }, + { + "epoch": 4.6, + "learning_rate": 8.056346693905803e-06, + "loss": 0.0239, + "step": 35141 + }, + { + "epoch": 4.6, + "learning_rate": 8.05582648529116e-06, + "loss": 0.0241, + "step": 35142 + }, + { + "epoch": 4.6, + "learning_rate": 8.055306282144403e-06, + "loss": 0.0248, + "step": 35143 + }, + { + "epoch": 4.6, + "learning_rate": 8.054786084466992e-06, + "loss": 0.0189, + "step": 35144 + }, + { + "epoch": 4.6, + "learning_rate": 8.054265892260391e-06, + "loss": 0.0202, + "step": 35145 + }, + { + "epoch": 4.6, + "learning_rate": 8.053745705526066e-06, + "loss": 0.0195, + "step": 35146 + }, + { + "epoch": 4.6, + "learning_rate": 8.053225524265475e-06, + "loss": 0.0182, + "step": 35147 + }, + { + "epoch": 4.6, + "learning_rate": 8.052705348480087e-06, + "loss": 0.0207, + "step": 35148 + }, + { + "epoch": 4.6, + "learning_rate": 8.052185178171356e-06, + "loss": 0.0215, + "step": 35149 + }, + { + "epoch": 4.6, + "learning_rate": 8.051665013340752e-06, + "loss": 0.0184, + "step": 35150 + }, + { + "epoch": 4.6, + "learning_rate": 8.051144853989741e-06, + "loss": 0.0173, + "step": 35151 + }, + { + "epoch": 4.6, + "learning_rate": 8.050624700119776e-06, + "loss": 0.0171, + "step": 35152 + }, + { + "epoch": 4.6, + "learning_rate": 8.050104551732327e-06, + "loss": 0.0196, + "step": 35153 + }, + { + "epoch": 4.6, + "learning_rate": 8.049584408828857e-06, + "loss": 0.0237, + "step": 35154 + }, + { + "epoch": 4.61, + "learning_rate": 8.049064271410826e-06, + "loss": 0.0212, + "step": 35155 + }, + { + "epoch": 4.61, + "learning_rate": 8.048544139479699e-06, + "loss": 0.0228, + "step": 35156 + }, + { + "epoch": 4.61, + "learning_rate": 8.048024013036935e-06, + "loss": 0.024, + "step": 35157 + }, + { + "epoch": 4.61, + "learning_rate": 8.047503892084003e-06, + "loss": 0.0166, + "step": 35158 + }, + { + "epoch": 4.61, + "learning_rate": 8.046983776622357e-06, + "loss": 0.0188, + "step": 35159 + }, + { + "epoch": 4.61, + "learning_rate": 8.046463666653467e-06, + "loss": 0.0194, + "step": 35160 + }, + { + "epoch": 4.61, + "learning_rate": 8.045943562178792e-06, + "loss": 0.0202, + "step": 35161 + }, + { + "epoch": 4.61, + "learning_rate": 8.045423463199803e-06, + "loss": 0.0228, + "step": 35162 + }, + { + "epoch": 4.61, + "learning_rate": 8.044903369717952e-06, + "loss": 0.0188, + "step": 35163 + }, + { + "epoch": 4.61, + "learning_rate": 8.044383281734709e-06, + "loss": 0.0234, + "step": 35164 + }, + { + "epoch": 4.61, + "learning_rate": 8.043863199251531e-06, + "loss": 0.0144, + "step": 35165 + }, + { + "epoch": 4.61, + "learning_rate": 8.043343122269886e-06, + "loss": 0.0158, + "step": 35166 + }, + { + "epoch": 4.61, + "learning_rate": 8.042823050791233e-06, + "loss": 0.0184, + "step": 35167 + }, + { + "epoch": 4.61, + "learning_rate": 8.042302984817033e-06, + "loss": 0.0204, + "step": 35168 + }, + { + "epoch": 4.61, + "learning_rate": 8.041782924348758e-06, + "loss": 0.0207, + "step": 35169 + }, + { + "epoch": 4.61, + "learning_rate": 8.041262869387858e-06, + "loss": 0.0215, + "step": 35170 + }, + { + "epoch": 4.61, + "learning_rate": 8.040742819935803e-06, + "loss": 0.0173, + "step": 35171 + }, + { + "epoch": 4.61, + "learning_rate": 8.040222775994057e-06, + "loss": 0.0189, + "step": 35172 + }, + { + "epoch": 4.61, + "learning_rate": 8.039702737564079e-06, + "loss": 0.0175, + "step": 35173 + }, + { + "epoch": 4.61, + "learning_rate": 8.039182704647332e-06, + "loss": 0.0175, + "step": 35174 + }, + { + "epoch": 4.61, + "learning_rate": 8.038662677245282e-06, + "loss": 0.0252, + "step": 35175 + }, + { + "epoch": 4.61, + "learning_rate": 8.038142655359386e-06, + "loss": 0.0199, + "step": 35176 + }, + { + "epoch": 4.61, + "learning_rate": 8.03762263899111e-06, + "loss": 0.0211, + "step": 35177 + }, + { + "epoch": 4.61, + "learning_rate": 8.037102628141917e-06, + "loss": 0.0214, + "step": 35178 + }, + { + "epoch": 4.61, + "learning_rate": 8.03658262281327e-06, + "loss": 0.0197, + "step": 35179 + }, + { + "epoch": 4.61, + "learning_rate": 8.036062623006626e-06, + "loss": 0.021, + "step": 35180 + }, + { + "epoch": 4.61, + "learning_rate": 8.035542628723449e-06, + "loss": 0.0168, + "step": 35181 + }, + { + "epoch": 4.61, + "learning_rate": 8.035022639965207e-06, + "loss": 0.0136, + "step": 35182 + }, + { + "epoch": 4.61, + "learning_rate": 8.03450265673336e-06, + "loss": 0.0195, + "step": 35183 + }, + { + "epoch": 4.61, + "learning_rate": 8.033982679029368e-06, + "loss": 0.0217, + "step": 35184 + }, + { + "epoch": 4.61, + "learning_rate": 8.033462706854698e-06, + "loss": 0.0157, + "step": 35185 + }, + { + "epoch": 4.61, + "learning_rate": 8.032942740210806e-06, + "loss": 0.0204, + "step": 35186 + }, + { + "epoch": 4.61, + "learning_rate": 8.03242277909916e-06, + "loss": 0.0196, + "step": 35187 + }, + { + "epoch": 4.61, + "learning_rate": 8.031902823521219e-06, + "loss": 0.0202, + "step": 35188 + }, + { + "epoch": 4.61, + "learning_rate": 8.031382873478446e-06, + "loss": 0.0226, + "step": 35189 + }, + { + "epoch": 4.61, + "learning_rate": 8.030862928972306e-06, + "loss": 0.0199, + "step": 35190 + }, + { + "epoch": 4.61, + "learning_rate": 8.030342990004257e-06, + "loss": 0.0215, + "step": 35191 + }, + { + "epoch": 4.61, + "learning_rate": 8.029823056575762e-06, + "loss": 0.0195, + "step": 35192 + }, + { + "epoch": 4.61, + "learning_rate": 8.029303128688289e-06, + "loss": 0.019, + "step": 35193 + }, + { + "epoch": 4.61, + "learning_rate": 8.028783206343293e-06, + "loss": 0.0214, + "step": 35194 + }, + { + "epoch": 4.61, + "learning_rate": 8.028263289542239e-06, + "loss": 0.0163, + "step": 35195 + }, + { + "epoch": 4.61, + "learning_rate": 8.027743378286592e-06, + "loss": 0.0168, + "step": 35196 + }, + { + "epoch": 4.61, + "learning_rate": 8.027223472577809e-06, + "loss": 0.0179, + "step": 35197 + }, + { + "epoch": 4.61, + "learning_rate": 8.02670357241736e-06, + "loss": 0.0203, + "step": 35198 + }, + { + "epoch": 4.61, + "learning_rate": 8.026183677806698e-06, + "loss": 0.0243, + "step": 35199 + }, + { + "epoch": 4.61, + "learning_rate": 8.02566378874729e-06, + "loss": 0.0251, + "step": 35200 + }, + { + "epoch": 4.61, + "learning_rate": 8.025143905240597e-06, + "loss": 0.0187, + "step": 35201 + }, + { + "epoch": 4.61, + "learning_rate": 8.024624027288083e-06, + "loss": 0.0186, + "step": 35202 + }, + { + "epoch": 4.61, + "learning_rate": 8.024104154891202e-06, + "loss": 0.0159, + "step": 35203 + }, + { + "epoch": 4.61, + "learning_rate": 8.02358428805143e-06, + "loss": 0.0233, + "step": 35204 + }, + { + "epoch": 4.61, + "learning_rate": 8.02306442677022e-06, + "loss": 0.0184, + "step": 35205 + }, + { + "epoch": 4.61, + "learning_rate": 8.022544571049037e-06, + "loss": 0.0196, + "step": 35206 + }, + { + "epoch": 4.61, + "learning_rate": 8.022024720889341e-06, + "loss": 0.0234, + "step": 35207 + }, + { + "epoch": 4.61, + "learning_rate": 8.021504876292598e-06, + "loss": 0.0157, + "step": 35208 + }, + { + "epoch": 4.61, + "learning_rate": 8.020985037260263e-06, + "loss": 0.0213, + "step": 35209 + }, + { + "epoch": 4.61, + "learning_rate": 8.020465203793804e-06, + "loss": 0.0169, + "step": 35210 + }, + { + "epoch": 4.61, + "learning_rate": 8.019945375894683e-06, + "loss": 0.0203, + "step": 35211 + }, + { + "epoch": 4.61, + "learning_rate": 8.019425553564357e-06, + "loss": 0.0189, + "step": 35212 + }, + { + "epoch": 4.61, + "learning_rate": 8.018905736804295e-06, + "loss": 0.0202, + "step": 35213 + }, + { + "epoch": 4.61, + "learning_rate": 8.01838592561595e-06, + "loss": 0.0182, + "step": 35214 + }, + { + "epoch": 4.61, + "learning_rate": 8.01786612000079e-06, + "loss": 0.0193, + "step": 35215 + }, + { + "epoch": 4.61, + "learning_rate": 8.01734631996028e-06, + "loss": 0.0195, + "step": 35216 + }, + { + "epoch": 4.61, + "learning_rate": 8.016826525495873e-06, + "loss": 0.0209, + "step": 35217 + }, + { + "epoch": 4.61, + "learning_rate": 8.01630673660904e-06, + "loss": 0.0197, + "step": 35218 + }, + { + "epoch": 4.61, + "learning_rate": 8.015786953301237e-06, + "loss": 0.0211, + "step": 35219 + }, + { + "epoch": 4.61, + "learning_rate": 8.015267175573927e-06, + "loss": 0.0208, + "step": 35220 + }, + { + "epoch": 4.61, + "learning_rate": 8.014747403428574e-06, + "loss": 0.0147, + "step": 35221 + }, + { + "epoch": 4.61, + "learning_rate": 8.014227636866637e-06, + "loss": 0.0171, + "step": 35222 + }, + { + "epoch": 4.61, + "learning_rate": 8.013707875889581e-06, + "loss": 0.0179, + "step": 35223 + }, + { + "epoch": 4.61, + "learning_rate": 8.013188120498863e-06, + "loss": 0.0201, + "step": 35224 + }, + { + "epoch": 4.61, + "learning_rate": 8.012668370695946e-06, + "loss": 0.0233, + "step": 35225 + }, + { + "epoch": 4.61, + "learning_rate": 8.012148626482294e-06, + "loss": 0.0232, + "step": 35226 + }, + { + "epoch": 4.61, + "learning_rate": 8.011628887859372e-06, + "loss": 0.0199, + "step": 35227 + }, + { + "epoch": 4.61, + "learning_rate": 8.011109154828636e-06, + "loss": 0.0229, + "step": 35228 + }, + { + "epoch": 4.61, + "learning_rate": 8.01058942739155e-06, + "loss": 0.0195, + "step": 35229 + }, + { + "epoch": 4.61, + "learning_rate": 8.010069705549575e-06, + "loss": 0.0167, + "step": 35230 + }, + { + "epoch": 4.62, + "learning_rate": 8.009549989304175e-06, + "loss": 0.0184, + "step": 35231 + }, + { + "epoch": 4.62, + "learning_rate": 8.009030278656805e-06, + "loss": 0.0198, + "step": 35232 + }, + { + "epoch": 4.62, + "learning_rate": 8.008510573608933e-06, + "loss": 0.0182, + "step": 35233 + }, + { + "epoch": 4.62, + "learning_rate": 8.007990874162021e-06, + "loss": 0.0174, + "step": 35234 + }, + { + "epoch": 4.62, + "learning_rate": 8.007471180317527e-06, + "loss": 0.0185, + "step": 35235 + }, + { + "epoch": 4.62, + "learning_rate": 8.006951492076912e-06, + "loss": 0.0178, + "step": 35236 + }, + { + "epoch": 4.62, + "learning_rate": 8.006431809441644e-06, + "loss": 0.0196, + "step": 35237 + }, + { + "epoch": 4.62, + "learning_rate": 8.005912132413178e-06, + "loss": 0.02, + "step": 35238 + }, + { + "epoch": 4.62, + "learning_rate": 8.005392460992978e-06, + "loss": 0.0173, + "step": 35239 + }, + { + "epoch": 4.62, + "learning_rate": 8.004872795182508e-06, + "loss": 0.0179, + "step": 35240 + }, + { + "epoch": 4.62, + "learning_rate": 8.004353134983224e-06, + "loss": 0.0192, + "step": 35241 + }, + { + "epoch": 4.62, + "learning_rate": 8.003833480396593e-06, + "loss": 0.0183, + "step": 35242 + }, + { + "epoch": 4.62, + "learning_rate": 8.003313831424072e-06, + "loss": 0.02, + "step": 35243 + }, + { + "epoch": 4.62, + "learning_rate": 8.002794188067126e-06, + "loss": 0.0211, + "step": 35244 + }, + { + "epoch": 4.62, + "learning_rate": 8.002274550327213e-06, + "loss": 0.0202, + "step": 35245 + }, + { + "epoch": 4.62, + "learning_rate": 8.001754918205798e-06, + "loss": 0.0193, + "step": 35246 + }, + { + "epoch": 4.62, + "learning_rate": 8.001235291704338e-06, + "loss": 0.0187, + "step": 35247 + }, + { + "epoch": 4.62, + "learning_rate": 8.000715670824301e-06, + "loss": 0.0233, + "step": 35248 + }, + { + "epoch": 4.62, + "learning_rate": 8.000196055567143e-06, + "loss": 0.0234, + "step": 35249 + }, + { + "epoch": 4.62, + "learning_rate": 7.999676445934329e-06, + "loss": 0.0208, + "step": 35250 + }, + { + "epoch": 4.62, + "learning_rate": 7.999156841927316e-06, + "loss": 0.0213, + "step": 35251 + }, + { + "epoch": 4.62, + "learning_rate": 7.998637243547571e-06, + "loss": 0.0169, + "step": 35252 + }, + { + "epoch": 4.62, + "learning_rate": 7.998117650796549e-06, + "loss": 0.0163, + "step": 35253 + }, + { + "epoch": 4.62, + "learning_rate": 7.997598063675716e-06, + "loss": 0.019, + "step": 35254 + }, + { + "epoch": 4.62, + "learning_rate": 7.997078482186532e-06, + "loss": 0.0166, + "step": 35255 + }, + { + "epoch": 4.62, + "learning_rate": 7.996558906330457e-06, + "loss": 0.0203, + "step": 35256 + }, + { + "epoch": 4.62, + "learning_rate": 7.996039336108954e-06, + "loss": 0.0237, + "step": 35257 + }, + { + "epoch": 4.62, + "learning_rate": 7.99551977152348e-06, + "loss": 0.0133, + "step": 35258 + }, + { + "epoch": 4.62, + "learning_rate": 7.995000212575504e-06, + "loss": 0.0154, + "step": 35259 + }, + { + "epoch": 4.62, + "learning_rate": 7.994480659266483e-06, + "loss": 0.0203, + "step": 35260 + }, + { + "epoch": 4.62, + "learning_rate": 7.993961111597878e-06, + "loss": 0.0207, + "step": 35261 + }, + { + "epoch": 4.62, + "learning_rate": 7.99344156957115e-06, + "loss": 0.0192, + "step": 35262 + }, + { + "epoch": 4.62, + "learning_rate": 7.992922033187763e-06, + "loss": 0.0218, + "step": 35263 + }, + { + "epoch": 4.62, + "learning_rate": 7.992402502449174e-06, + "loss": 0.0176, + "step": 35264 + }, + { + "epoch": 4.62, + "learning_rate": 7.991882977356847e-06, + "loss": 0.0213, + "step": 35265 + }, + { + "epoch": 4.62, + "learning_rate": 7.991363457912242e-06, + "loss": 0.0202, + "step": 35266 + }, + { + "epoch": 4.62, + "learning_rate": 7.99084394411682e-06, + "loss": 0.0173, + "step": 35267 + }, + { + "epoch": 4.62, + "learning_rate": 7.990324435972042e-06, + "loss": 0.0174, + "step": 35268 + }, + { + "epoch": 4.62, + "learning_rate": 7.989804933479367e-06, + "loss": 0.0143, + "step": 35269 + }, + { + "epoch": 4.62, + "learning_rate": 7.989285436640259e-06, + "loss": 0.0163, + "step": 35270 + }, + { + "epoch": 4.62, + "learning_rate": 7.988765945456183e-06, + "loss": 0.02, + "step": 35271 + }, + { + "epoch": 4.62, + "learning_rate": 7.988246459928592e-06, + "loss": 0.0197, + "step": 35272 + }, + { + "epoch": 4.62, + "learning_rate": 7.987726980058954e-06, + "loss": 0.0192, + "step": 35273 + }, + { + "epoch": 4.62, + "learning_rate": 7.987207505848724e-06, + "loss": 0.0176, + "step": 35274 + }, + { + "epoch": 4.62, + "learning_rate": 7.986688037299367e-06, + "loss": 0.0194, + "step": 35275 + }, + { + "epoch": 4.62, + "learning_rate": 7.986168574412343e-06, + "loss": 0.0198, + "step": 35276 + }, + { + "epoch": 4.62, + "learning_rate": 7.985649117189112e-06, + "loss": 0.0207, + "step": 35277 + }, + { + "epoch": 4.62, + "learning_rate": 7.985129665631136e-06, + "loss": 0.0134, + "step": 35278 + }, + { + "epoch": 4.62, + "learning_rate": 7.984610219739874e-06, + "loss": 0.018, + "step": 35279 + }, + { + "epoch": 4.62, + "learning_rate": 7.984090779516786e-06, + "loss": 0.0168, + "step": 35280 + }, + { + "epoch": 4.62, + "learning_rate": 7.983571344963341e-06, + "loss": 0.0202, + "step": 35281 + }, + { + "epoch": 4.62, + "learning_rate": 7.983051916080992e-06, + "loss": 0.0201, + "step": 35282 + }, + { + "epoch": 4.62, + "learning_rate": 7.9825324928712e-06, + "loss": 0.0208, + "step": 35283 + }, + { + "epoch": 4.62, + "learning_rate": 7.982013075335432e-06, + "loss": 0.0198, + "step": 35284 + }, + { + "epoch": 4.62, + "learning_rate": 7.981493663475142e-06, + "loss": 0.0187, + "step": 35285 + }, + { + "epoch": 4.62, + "learning_rate": 7.980974257291797e-06, + "loss": 0.0212, + "step": 35286 + }, + { + "epoch": 4.62, + "learning_rate": 7.98045485678685e-06, + "loss": 0.0223, + "step": 35287 + }, + { + "epoch": 4.62, + "learning_rate": 7.97993546196177e-06, + "loss": 0.0182, + "step": 35288 + }, + { + "epoch": 4.62, + "learning_rate": 7.97941607281801e-06, + "loss": 0.0153, + "step": 35289 + }, + { + "epoch": 4.62, + "learning_rate": 7.978896689357036e-06, + "loss": 0.0221, + "step": 35290 + }, + { + "epoch": 4.62, + "learning_rate": 7.978377311580306e-06, + "loss": 0.0184, + "step": 35291 + }, + { + "epoch": 4.62, + "learning_rate": 7.977857939489285e-06, + "loss": 0.0212, + "step": 35292 + }, + { + "epoch": 4.62, + "learning_rate": 7.977338573085429e-06, + "loss": 0.024, + "step": 35293 + }, + { + "epoch": 4.62, + "learning_rate": 7.976819212370203e-06, + "loss": 0.0246, + "step": 35294 + }, + { + "epoch": 4.62, + "learning_rate": 7.976299857345063e-06, + "loss": 0.0205, + "step": 35295 + }, + { + "epoch": 4.62, + "learning_rate": 7.975780508011472e-06, + "loss": 0.0196, + "step": 35296 + }, + { + "epoch": 4.62, + "learning_rate": 7.975261164370892e-06, + "loss": 0.0193, + "step": 35297 + }, + { + "epoch": 4.62, + "learning_rate": 7.974741826424782e-06, + "loss": 0.0182, + "step": 35298 + }, + { + "epoch": 4.62, + "learning_rate": 7.974222494174603e-06, + "loss": 0.0191, + "step": 35299 + }, + { + "epoch": 4.62, + "learning_rate": 7.973703167621814e-06, + "loss": 0.0231, + "step": 35300 + }, + { + "epoch": 4.62, + "learning_rate": 7.973183846767874e-06, + "loss": 0.0219, + "step": 35301 + }, + { + "epoch": 4.62, + "learning_rate": 7.972664531614251e-06, + "loss": 0.0206, + "step": 35302 + }, + { + "epoch": 4.62, + "learning_rate": 7.9721452221624e-06, + "loss": 0.0202, + "step": 35303 + }, + { + "epoch": 4.62, + "learning_rate": 7.971625918413781e-06, + "loss": 0.0187, + "step": 35304 + }, + { + "epoch": 4.62, + "learning_rate": 7.971106620369858e-06, + "loss": 0.0282, + "step": 35305 + }, + { + "epoch": 4.62, + "learning_rate": 7.970587328032088e-06, + "loss": 0.0219, + "step": 35306 + }, + { + "epoch": 4.62, + "learning_rate": 7.970068041401936e-06, + "loss": 0.0202, + "step": 35307 + }, + { + "epoch": 4.63, + "learning_rate": 7.969548760480858e-06, + "loss": 0.0171, + "step": 35308 + }, + { + "epoch": 4.63, + "learning_rate": 7.969029485270316e-06, + "loss": 0.0236, + "step": 35309 + }, + { + "epoch": 4.63, + "learning_rate": 7.96851021577177e-06, + "loss": 0.0177, + "step": 35310 + }, + { + "epoch": 4.63, + "learning_rate": 7.96799095198668e-06, + "loss": 0.0245, + "step": 35311 + }, + { + "epoch": 4.63, + "learning_rate": 7.967471693916506e-06, + "loss": 0.0193, + "step": 35312 + }, + { + "epoch": 4.63, + "learning_rate": 7.966952441562713e-06, + "loss": 0.0187, + "step": 35313 + }, + { + "epoch": 4.63, + "learning_rate": 7.966433194926756e-06, + "loss": 0.0186, + "step": 35314 + }, + { + "epoch": 4.63, + "learning_rate": 7.9659139540101e-06, + "loss": 0.0224, + "step": 35315 + }, + { + "epoch": 4.63, + "learning_rate": 7.9653947188142e-06, + "loss": 0.019, + "step": 35316 + }, + { + "epoch": 4.63, + "learning_rate": 7.964875489340521e-06, + "loss": 0.0233, + "step": 35317 + }, + { + "epoch": 4.63, + "learning_rate": 7.964356265590519e-06, + "loss": 0.0199, + "step": 35318 + }, + { + "epoch": 4.63, + "learning_rate": 7.963837047565658e-06, + "loss": 0.0235, + "step": 35319 + }, + { + "epoch": 4.63, + "learning_rate": 7.963317835267397e-06, + "loss": 0.0162, + "step": 35320 + }, + { + "epoch": 4.63, + "learning_rate": 7.962798628697195e-06, + "loss": 0.015, + "step": 35321 + }, + { + "epoch": 4.63, + "learning_rate": 7.962279427856515e-06, + "loss": 0.0229, + "step": 35322 + }, + { + "epoch": 4.63, + "learning_rate": 7.961760232746813e-06, + "loss": 0.0214, + "step": 35323 + }, + { + "epoch": 4.63, + "learning_rate": 7.961241043369554e-06, + "loss": 0.0208, + "step": 35324 + }, + { + "epoch": 4.63, + "learning_rate": 7.960721859726194e-06, + "loss": 0.0174, + "step": 35325 + }, + { + "epoch": 4.63, + "learning_rate": 7.960202681818198e-06, + "loss": 0.0159, + "step": 35326 + }, + { + "epoch": 4.63, + "learning_rate": 7.959683509647022e-06, + "loss": 0.02, + "step": 35327 + }, + { + "epoch": 4.63, + "learning_rate": 7.959164343214128e-06, + "loss": 0.0195, + "step": 35328 + }, + { + "epoch": 4.63, + "learning_rate": 7.958645182520974e-06, + "loss": 0.0202, + "step": 35329 + }, + { + "epoch": 4.63, + "learning_rate": 7.958126027569024e-06, + "loss": 0.025, + "step": 35330 + }, + { + "epoch": 4.63, + "learning_rate": 7.957606878359735e-06, + "loss": 0.0213, + "step": 35331 + }, + { + "epoch": 4.63, + "learning_rate": 7.957087734894569e-06, + "loss": 0.0211, + "step": 35332 + }, + { + "epoch": 4.63, + "learning_rate": 7.956568597174982e-06, + "loss": 0.0146, + "step": 35333 + }, + { + "epoch": 4.63, + "learning_rate": 7.956049465202435e-06, + "loss": 0.0187, + "step": 35334 + }, + { + "epoch": 4.63, + "learning_rate": 7.955530338978393e-06, + "loss": 0.0159, + "step": 35335 + }, + { + "epoch": 4.63, + "learning_rate": 7.955011218504317e-06, + "loss": 0.022, + "step": 35336 + }, + { + "epoch": 4.63, + "learning_rate": 7.954492103781658e-06, + "loss": 0.0182, + "step": 35337 + }, + { + "epoch": 4.63, + "learning_rate": 7.953972994811884e-06, + "loss": 0.0252, + "step": 35338 + }, + { + "epoch": 4.63, + "learning_rate": 7.953453891596449e-06, + "loss": 0.0208, + "step": 35339 + }, + { + "epoch": 4.63, + "learning_rate": 7.952934794136818e-06, + "loss": 0.0205, + "step": 35340 + }, + { + "epoch": 4.63, + "learning_rate": 7.95241570243445e-06, + "loss": 0.0211, + "step": 35341 + }, + { + "epoch": 4.63, + "learning_rate": 7.951896616490801e-06, + "loss": 0.0203, + "step": 35342 + }, + { + "epoch": 4.63, + "learning_rate": 7.951377536307337e-06, + "loss": 0.0176, + "step": 35343 + }, + { + "epoch": 4.63, + "learning_rate": 7.950858461885511e-06, + "loss": 0.0231, + "step": 35344 + }, + { + "epoch": 4.63, + "learning_rate": 7.950339393226785e-06, + "loss": 0.0182, + "step": 35345 + }, + { + "epoch": 4.63, + "learning_rate": 7.949820330332623e-06, + "loss": 0.0174, + "step": 35346 + }, + { + "epoch": 4.63, + "learning_rate": 7.949301273204483e-06, + "loss": 0.0186, + "step": 35347 + }, + { + "epoch": 4.63, + "learning_rate": 7.948782221843822e-06, + "loss": 0.0201, + "step": 35348 + }, + { + "epoch": 4.63, + "learning_rate": 7.948263176252105e-06, + "loss": 0.0186, + "step": 35349 + }, + { + "epoch": 4.63, + "learning_rate": 7.947744136430786e-06, + "loss": 0.0159, + "step": 35350 + }, + { + "epoch": 4.63, + "learning_rate": 7.94722510238133e-06, + "loss": 0.0226, + "step": 35351 + }, + { + "epoch": 4.63, + "learning_rate": 7.94670607410519e-06, + "loss": 0.0173, + "step": 35352 + }, + { + "epoch": 4.63, + "learning_rate": 7.946187051603835e-06, + "loss": 0.0157, + "step": 35353 + }, + { + "epoch": 4.63, + "learning_rate": 7.945668034878714e-06, + "loss": 0.019, + "step": 35354 + }, + { + "epoch": 4.63, + "learning_rate": 7.945149023931295e-06, + "loss": 0.019, + "step": 35355 + }, + { + "epoch": 4.63, + "learning_rate": 7.944630018763034e-06, + "loss": 0.0199, + "step": 35356 + }, + { + "epoch": 4.63, + "learning_rate": 7.944111019375392e-06, + "loss": 0.0238, + "step": 35357 + }, + { + "epoch": 4.63, + "learning_rate": 7.943592025769828e-06, + "loss": 0.0229, + "step": 35358 + }, + { + "epoch": 4.63, + "learning_rate": 7.943073037947804e-06, + "loss": 0.0219, + "step": 35359 + }, + { + "epoch": 4.63, + "learning_rate": 7.942554055910776e-06, + "loss": 0.0182, + "step": 35360 + }, + { + "epoch": 4.63, + "learning_rate": 7.942035079660205e-06, + "loss": 0.0175, + "step": 35361 + }, + { + "epoch": 4.63, + "learning_rate": 7.941516109197552e-06, + "loss": 0.0176, + "step": 35362 + }, + { + "epoch": 4.63, + "learning_rate": 7.940997144524272e-06, + "loss": 0.015, + "step": 35363 + }, + { + "epoch": 4.63, + "learning_rate": 7.940478185641832e-06, + "loss": 0.0183, + "step": 35364 + }, + { + "epoch": 4.63, + "learning_rate": 7.939959232551686e-06, + "loss": 0.0167, + "step": 35365 + }, + { + "epoch": 4.63, + "learning_rate": 7.939440285255296e-06, + "loss": 0.0231, + "step": 35366 + }, + { + "epoch": 4.63, + "learning_rate": 7.938921343754115e-06, + "loss": 0.0212, + "step": 35367 + }, + { + "epoch": 4.63, + "learning_rate": 7.938402408049612e-06, + "loss": 0.0165, + "step": 35368 + }, + { + "epoch": 4.63, + "learning_rate": 7.937883478143242e-06, + "loss": 0.0222, + "step": 35369 + }, + { + "epoch": 4.63, + "learning_rate": 7.937364554036466e-06, + "loss": 0.0192, + "step": 35370 + }, + { + "epoch": 4.63, + "learning_rate": 7.936845635730741e-06, + "loss": 0.0204, + "step": 35371 + }, + { + "epoch": 4.63, + "learning_rate": 7.93632672322753e-06, + "loss": 0.0156, + "step": 35372 + }, + { + "epoch": 4.63, + "learning_rate": 7.935807816528288e-06, + "loss": 0.0183, + "step": 35373 + }, + { + "epoch": 4.63, + "learning_rate": 7.935288915634478e-06, + "loss": 0.0212, + "step": 35374 + }, + { + "epoch": 4.63, + "learning_rate": 7.934770020547556e-06, + "loss": 0.0184, + "step": 35375 + }, + { + "epoch": 4.63, + "learning_rate": 7.934251131268984e-06, + "loss": 0.0211, + "step": 35376 + }, + { + "epoch": 4.63, + "learning_rate": 7.933732247800223e-06, + "loss": 0.0184, + "step": 35377 + }, + { + "epoch": 4.63, + "learning_rate": 7.933213370142725e-06, + "loss": 0.0173, + "step": 35378 + }, + { + "epoch": 4.63, + "learning_rate": 7.932694498297958e-06, + "loss": 0.0214, + "step": 35379 + }, + { + "epoch": 4.63, + "learning_rate": 7.932175632267378e-06, + "loss": 0.0266, + "step": 35380 + }, + { + "epoch": 4.63, + "learning_rate": 7.931656772052442e-06, + "loss": 0.0205, + "step": 35381 + }, + { + "epoch": 4.63, + "learning_rate": 7.931137917654615e-06, + "loss": 0.019, + "step": 35382 + }, + { + "epoch": 4.63, + "learning_rate": 7.930619069075347e-06, + "loss": 0.0211, + "step": 35383 + }, + { + "epoch": 4.64, + "learning_rate": 7.930100226316106e-06, + "loss": 0.0156, + "step": 35384 + }, + { + "epoch": 4.64, + "learning_rate": 7.92958138937835e-06, + "loss": 0.0189, + "step": 35385 + }, + { + "epoch": 4.64, + "learning_rate": 7.929062558263531e-06, + "loss": 0.0209, + "step": 35386 + }, + { + "epoch": 4.64, + "learning_rate": 7.928543732973117e-06, + "loss": 0.0196, + "step": 35387 + }, + { + "epoch": 4.64, + "learning_rate": 7.928024913508562e-06, + "loss": 0.0201, + "step": 35388 + }, + { + "epoch": 4.64, + "learning_rate": 7.927506099871323e-06, + "loss": 0.0184, + "step": 35389 + }, + { + "epoch": 4.64, + "learning_rate": 7.926987292062867e-06, + "loss": 0.019, + "step": 35390 + }, + { + "epoch": 4.64, + "learning_rate": 7.92646849008465e-06, + "loss": 0.0184, + "step": 35391 + }, + { + "epoch": 4.64, + "learning_rate": 7.925949693938126e-06, + "loss": 0.0182, + "step": 35392 + }, + { + "epoch": 4.64, + "learning_rate": 7.925430903624763e-06, + "loss": 0.0187, + "step": 35393 + }, + { + "epoch": 4.64, + "learning_rate": 7.924912119146011e-06, + "loss": 0.0182, + "step": 35394 + }, + { + "epoch": 4.64, + "learning_rate": 7.924393340503336e-06, + "loss": 0.0187, + "step": 35395 + }, + { + "epoch": 4.64, + "learning_rate": 7.923874567698191e-06, + "loss": 0.0198, + "step": 35396 + }, + { + "epoch": 4.64, + "learning_rate": 7.923355800732042e-06, + "loss": 0.0151, + "step": 35397 + }, + { + "epoch": 4.64, + "learning_rate": 7.922837039606342e-06, + "loss": 0.0183, + "step": 35398 + }, + { + "epoch": 4.64, + "learning_rate": 7.922318284322551e-06, + "loss": 0.0203, + "step": 35399 + }, + { + "epoch": 4.64, + "learning_rate": 7.921799534882128e-06, + "loss": 0.019, + "step": 35400 + }, + { + "epoch": 4.64, + "learning_rate": 7.921280791286536e-06, + "loss": 0.0172, + "step": 35401 + }, + { + "epoch": 4.64, + "learning_rate": 7.92076205353723e-06, + "loss": 0.021, + "step": 35402 + }, + { + "epoch": 4.64, + "learning_rate": 7.920243321635673e-06, + "loss": 0.0217, + "step": 35403 + }, + { + "epoch": 4.64, + "learning_rate": 7.919724595583318e-06, + "loss": 0.0226, + "step": 35404 + }, + { + "epoch": 4.64, + "learning_rate": 7.919205875381626e-06, + "loss": 0.0164, + "step": 35405 + }, + { + "epoch": 4.64, + "learning_rate": 7.91868716103206e-06, + "loss": 0.0169, + "step": 35406 + }, + { + "epoch": 4.64, + "learning_rate": 7.918168452536072e-06, + "loss": 0.0203, + "step": 35407 + }, + { + "epoch": 4.64, + "learning_rate": 7.917649749895128e-06, + "loss": 0.0187, + "step": 35408 + }, + { + "epoch": 4.64, + "learning_rate": 7.91713105311068e-06, + "loss": 0.0197, + "step": 35409 + }, + { + "epoch": 4.64, + "learning_rate": 7.916612362184193e-06, + "loss": 0.0231, + "step": 35410 + }, + { + "epoch": 4.64, + "learning_rate": 7.916093677117116e-06, + "loss": 0.0229, + "step": 35411 + }, + { + "epoch": 4.64, + "learning_rate": 7.915574997910921e-06, + "loss": 0.0189, + "step": 35412 + }, + { + "epoch": 4.64, + "learning_rate": 7.915056324567058e-06, + "loss": 0.0224, + "step": 35413 + }, + { + "epoch": 4.64, + "learning_rate": 7.914537657086991e-06, + "loss": 0.0161, + "step": 35414 + }, + { + "epoch": 4.64, + "learning_rate": 7.914018995472173e-06, + "loss": 0.0191, + "step": 35415 + }, + { + "epoch": 4.64, + "learning_rate": 7.913500339724068e-06, + "loss": 0.0213, + "step": 35416 + }, + { + "epoch": 4.64, + "learning_rate": 7.912981689844129e-06, + "loss": 0.0155, + "step": 35417 + }, + { + "epoch": 4.64, + "learning_rate": 7.91246304583382e-06, + "loss": 0.0159, + "step": 35418 + }, + { + "epoch": 4.64, + "learning_rate": 7.911944407694596e-06, + "loss": 0.0218, + "step": 35419 + }, + { + "epoch": 4.64, + "learning_rate": 7.911425775427916e-06, + "loss": 0.0267, + "step": 35420 + }, + { + "epoch": 4.64, + "learning_rate": 7.910907149035244e-06, + "loss": 0.014, + "step": 35421 + }, + { + "epoch": 4.64, + "learning_rate": 7.910388528518029e-06, + "loss": 0.0192, + "step": 35422 + }, + { + "epoch": 4.64, + "learning_rate": 7.909869913877739e-06, + "loss": 0.0185, + "step": 35423 + }, + { + "epoch": 4.64, + "learning_rate": 7.909351305115828e-06, + "loss": 0.0199, + "step": 35424 + }, + { + "epoch": 4.64, + "learning_rate": 7.908832702233755e-06, + "loss": 0.0191, + "step": 35425 + }, + { + "epoch": 4.64, + "learning_rate": 7.908314105232976e-06, + "loss": 0.0208, + "step": 35426 + }, + { + "epoch": 4.64, + "learning_rate": 7.907795514114957e-06, + "loss": 0.0174, + "step": 35427 + }, + { + "epoch": 4.64, + "learning_rate": 7.907276928881149e-06, + "loss": 0.0198, + "step": 35428 + }, + { + "epoch": 4.64, + "learning_rate": 7.906758349533013e-06, + "loss": 0.0184, + "step": 35429 + }, + { + "epoch": 4.64, + "learning_rate": 7.906239776072009e-06, + "loss": 0.0195, + "step": 35430 + }, + { + "epoch": 4.64, + "learning_rate": 7.905721208499593e-06, + "loss": 0.0237, + "step": 35431 + }, + { + "epoch": 4.64, + "learning_rate": 7.905202646817222e-06, + "loss": 0.0197, + "step": 35432 + }, + { + "epoch": 4.64, + "learning_rate": 7.90468409102636e-06, + "loss": 0.0214, + "step": 35433 + }, + { + "epoch": 4.64, + "learning_rate": 7.904165541128461e-06, + "loss": 0.0233, + "step": 35434 + }, + { + "epoch": 4.64, + "learning_rate": 7.903646997124987e-06, + "loss": 0.0183, + "step": 35435 + }, + { + "epoch": 4.64, + "learning_rate": 7.903128459017394e-06, + "loss": 0.0177, + "step": 35436 + }, + { + "epoch": 4.64, + "learning_rate": 7.902609926807141e-06, + "loss": 0.0185, + "step": 35437 + }, + { + "epoch": 4.64, + "learning_rate": 7.902091400495684e-06, + "loss": 0.0232, + "step": 35438 + }, + { + "epoch": 4.64, + "learning_rate": 7.901572880084485e-06, + "loss": 0.0193, + "step": 35439 + }, + { + "epoch": 4.64, + "learning_rate": 7.901054365574999e-06, + "loss": 0.0163, + "step": 35440 + }, + { + "epoch": 4.64, + "learning_rate": 7.900535856968685e-06, + "loss": 0.0182, + "step": 35441 + }, + { + "epoch": 4.64, + "learning_rate": 7.900017354267004e-06, + "loss": 0.0175, + "step": 35442 + }, + { + "epoch": 4.64, + "learning_rate": 7.899498857471411e-06, + "loss": 0.0177, + "step": 35443 + }, + { + "epoch": 4.64, + "learning_rate": 7.898980366583365e-06, + "loss": 0.0221, + "step": 35444 + }, + { + "epoch": 4.64, + "learning_rate": 7.898461881604327e-06, + "loss": 0.024, + "step": 35445 + }, + { + "epoch": 4.64, + "learning_rate": 7.89794340253575e-06, + "loss": 0.0203, + "step": 35446 + }, + { + "epoch": 4.64, + "learning_rate": 7.8974249293791e-06, + "loss": 0.0249, + "step": 35447 + }, + { + "epoch": 4.64, + "learning_rate": 7.896906462135826e-06, + "loss": 0.0203, + "step": 35448 + }, + { + "epoch": 4.64, + "learning_rate": 7.896388000807393e-06, + "loss": 0.0179, + "step": 35449 + }, + { + "epoch": 4.64, + "learning_rate": 7.895869545395257e-06, + "loss": 0.0208, + "step": 35450 + }, + { + "epoch": 4.64, + "learning_rate": 7.895351095900875e-06, + "loss": 0.021, + "step": 35451 + }, + { + "epoch": 4.64, + "learning_rate": 7.894832652325708e-06, + "loss": 0.0177, + "step": 35452 + }, + { + "epoch": 4.64, + "learning_rate": 7.894314214671208e-06, + "loss": 0.0198, + "step": 35453 + }, + { + "epoch": 4.64, + "learning_rate": 7.893795782938837e-06, + "loss": 0.0194, + "step": 35454 + }, + { + "epoch": 4.64, + "learning_rate": 7.893277357130055e-06, + "loss": 0.0189, + "step": 35455 + }, + { + "epoch": 4.64, + "learning_rate": 7.892758937246319e-06, + "loss": 0.0189, + "step": 35456 + }, + { + "epoch": 4.64, + "learning_rate": 7.892240523289086e-06, + "loss": 0.0154, + "step": 35457 + }, + { + "epoch": 4.64, + "learning_rate": 7.891722115259815e-06, + "loss": 0.017, + "step": 35458 + }, + { + "epoch": 4.64, + "learning_rate": 7.891203713159962e-06, + "loss": 0.0168, + "step": 35459 + }, + { + "epoch": 4.65, + "learning_rate": 7.890685316990987e-06, + "loss": 0.023, + "step": 35460 + }, + { + "epoch": 4.65, + "learning_rate": 7.890166926754346e-06, + "loss": 0.0216, + "step": 35461 + }, + { + "epoch": 4.65, + "learning_rate": 7.8896485424515e-06, + "loss": 0.0173, + "step": 35462 + }, + { + "epoch": 4.65, + "learning_rate": 7.889130164083905e-06, + "loss": 0.02, + "step": 35463 + }, + { + "epoch": 4.65, + "learning_rate": 7.888611791653018e-06, + "loss": 0.0158, + "step": 35464 + }, + { + "epoch": 4.65, + "learning_rate": 7.888093425160294e-06, + "loss": 0.0179, + "step": 35465 + }, + { + "epoch": 4.65, + "learning_rate": 7.8875750646072e-06, + "loss": 0.0178, + "step": 35466 + }, + { + "epoch": 4.65, + "learning_rate": 7.887056709995187e-06, + "loss": 0.0187, + "step": 35467 + }, + { + "epoch": 4.65, + "learning_rate": 7.886538361325718e-06, + "loss": 0.0166, + "step": 35468 + }, + { + "epoch": 4.65, + "learning_rate": 7.886020018600242e-06, + "loss": 0.0196, + "step": 35469 + }, + { + "epoch": 4.65, + "learning_rate": 7.885501681820225e-06, + "loss": 0.0189, + "step": 35470 + }, + { + "epoch": 4.65, + "learning_rate": 7.884983350987123e-06, + "loss": 0.0184, + "step": 35471 + }, + { + "epoch": 4.65, + "learning_rate": 7.884465026102389e-06, + "loss": 0.0186, + "step": 35472 + }, + { + "epoch": 4.65, + "learning_rate": 7.883946707167487e-06, + "loss": 0.0209, + "step": 35473 + }, + { + "epoch": 4.65, + "learning_rate": 7.883428394183871e-06, + "loss": 0.0213, + "step": 35474 + }, + { + "epoch": 4.65, + "learning_rate": 7.882910087153003e-06, + "loss": 0.018, + "step": 35475 + }, + { + "epoch": 4.65, + "learning_rate": 7.88239178607633e-06, + "loss": 0.0213, + "step": 35476 + }, + { + "epoch": 4.65, + "learning_rate": 7.881873490955324e-06, + "loss": 0.0162, + "step": 35477 + }, + { + "epoch": 4.65, + "learning_rate": 7.881355201791435e-06, + "loss": 0.0214, + "step": 35478 + }, + { + "epoch": 4.65, + "learning_rate": 7.88083691858612e-06, + "loss": 0.0194, + "step": 35479 + }, + { + "epoch": 4.65, + "learning_rate": 7.880318641340841e-06, + "loss": 0.0208, + "step": 35480 + }, + { + "epoch": 4.65, + "learning_rate": 7.879800370057052e-06, + "loss": 0.0222, + "step": 35481 + }, + { + "epoch": 4.65, + "learning_rate": 7.87928210473621e-06, + "loss": 0.0191, + "step": 35482 + }, + { + "epoch": 4.65, + "learning_rate": 7.878763845379778e-06, + "loss": 0.0228, + "step": 35483 + }, + { + "epoch": 4.65, + "learning_rate": 7.878245591989206e-06, + "loss": 0.0152, + "step": 35484 + }, + { + "epoch": 4.65, + "learning_rate": 7.877727344565955e-06, + "loss": 0.0182, + "step": 35485 + }, + { + "epoch": 4.65, + "learning_rate": 7.877209103111485e-06, + "loss": 0.0186, + "step": 35486 + }, + { + "epoch": 4.65, + "learning_rate": 7.876690867627249e-06, + "loss": 0.0152, + "step": 35487 + }, + { + "epoch": 4.65, + "learning_rate": 7.876172638114706e-06, + "loss": 0.0207, + "step": 35488 + }, + { + "epoch": 4.65, + "learning_rate": 7.875654414575318e-06, + "loss": 0.0221, + "step": 35489 + }, + { + "epoch": 4.65, + "learning_rate": 7.875136197010536e-06, + "loss": 0.0187, + "step": 35490 + }, + { + "epoch": 4.65, + "learning_rate": 7.874617985421823e-06, + "loss": 0.0202, + "step": 35491 + }, + { + "epoch": 4.65, + "learning_rate": 7.874099779810634e-06, + "loss": 0.0164, + "step": 35492 + }, + { + "epoch": 4.65, + "learning_rate": 7.873581580178422e-06, + "loss": 0.0195, + "step": 35493 + }, + { + "epoch": 4.65, + "learning_rate": 7.873063386526653e-06, + "loss": 0.0191, + "step": 35494 + }, + { + "epoch": 4.65, + "learning_rate": 7.872545198856777e-06, + "loss": 0.0168, + "step": 35495 + }, + { + "epoch": 4.65, + "learning_rate": 7.872027017170258e-06, + "loss": 0.023, + "step": 35496 + }, + { + "epoch": 4.65, + "learning_rate": 7.871508841468547e-06, + "loss": 0.0194, + "step": 35497 + }, + { + "epoch": 4.65, + "learning_rate": 7.8709906717531e-06, + "loss": 0.0204, + "step": 35498 + }, + { + "epoch": 4.65, + "learning_rate": 7.870472508025381e-06, + "loss": 0.019, + "step": 35499 + }, + { + "epoch": 4.65, + "learning_rate": 7.869954350286848e-06, + "loss": 0.0152, + "step": 35500 + }, + { + "epoch": 4.65, + "learning_rate": 7.869436198538954e-06, + "loss": 0.0149, + "step": 35501 + }, + { + "epoch": 4.65, + "learning_rate": 7.868918052783157e-06, + "loss": 0.0222, + "step": 35502 + }, + { + "epoch": 4.65, + "learning_rate": 7.868399913020913e-06, + "loss": 0.0215, + "step": 35503 + }, + { + "epoch": 4.65, + "learning_rate": 7.867881779253684e-06, + "loss": 0.0214, + "step": 35504 + }, + { + "epoch": 4.65, + "learning_rate": 7.86736365148292e-06, + "loss": 0.0199, + "step": 35505 + }, + { + "epoch": 4.65, + "learning_rate": 7.866845529710084e-06, + "loss": 0.02, + "step": 35506 + }, + { + "epoch": 4.65, + "learning_rate": 7.866327413936633e-06, + "loss": 0.0204, + "step": 35507 + }, + { + "epoch": 4.65, + "learning_rate": 7.865809304164019e-06, + "loss": 0.0192, + "step": 35508 + }, + { + "epoch": 4.65, + "learning_rate": 7.8652912003937e-06, + "loss": 0.0214, + "step": 35509 + }, + { + "epoch": 4.65, + "learning_rate": 7.864773102627143e-06, + "loss": 0.0178, + "step": 35510 + }, + { + "epoch": 4.65, + "learning_rate": 7.864255010865793e-06, + "loss": 0.0196, + "step": 35511 + }, + { + "epoch": 4.65, + "learning_rate": 7.863736925111114e-06, + "loss": 0.0166, + "step": 35512 + }, + { + "epoch": 4.65, + "learning_rate": 7.863218845364564e-06, + "loss": 0.0182, + "step": 35513 + }, + { + "epoch": 4.65, + "learning_rate": 7.862700771627593e-06, + "loss": 0.0195, + "step": 35514 + }, + { + "epoch": 4.65, + "learning_rate": 7.862182703901664e-06, + "loss": 0.0228, + "step": 35515 + }, + { + "epoch": 4.65, + "learning_rate": 7.861664642188232e-06, + "loss": 0.0193, + "step": 35516 + }, + { + "epoch": 4.65, + "learning_rate": 7.861146586488754e-06, + "loss": 0.0173, + "step": 35517 + }, + { + "epoch": 4.65, + "learning_rate": 7.860628536804688e-06, + "loss": 0.0238, + "step": 35518 + }, + { + "epoch": 4.65, + "learning_rate": 7.860110493137491e-06, + "loss": 0.0205, + "step": 35519 + }, + { + "epoch": 4.65, + "learning_rate": 7.859592455488613e-06, + "loss": 0.0179, + "step": 35520 + }, + { + "epoch": 4.65, + "learning_rate": 7.859074423859525e-06, + "loss": 0.0171, + "step": 35521 + }, + { + "epoch": 4.65, + "learning_rate": 7.858556398251674e-06, + "loss": 0.0202, + "step": 35522 + }, + { + "epoch": 4.65, + "learning_rate": 7.85803837866652e-06, + "loss": 0.0207, + "step": 35523 + }, + { + "epoch": 4.65, + "learning_rate": 7.857520365105516e-06, + "loss": 0.0239, + "step": 35524 + }, + { + "epoch": 4.65, + "learning_rate": 7.857002357570125e-06, + "loss": 0.0159, + "step": 35525 + }, + { + "epoch": 4.65, + "learning_rate": 7.856484356061798e-06, + "loss": 0.0175, + "step": 35526 + }, + { + "epoch": 4.65, + "learning_rate": 7.855966360581996e-06, + "loss": 0.0169, + "step": 35527 + }, + { + "epoch": 4.65, + "learning_rate": 7.855448371132176e-06, + "loss": 0.0191, + "step": 35528 + }, + { + "epoch": 4.65, + "learning_rate": 7.85493038771379e-06, + "loss": 0.0172, + "step": 35529 + }, + { + "epoch": 4.65, + "learning_rate": 7.854412410328301e-06, + "loss": 0.0203, + "step": 35530 + }, + { + "epoch": 4.65, + "learning_rate": 7.853894438977159e-06, + "loss": 0.0195, + "step": 35531 + }, + { + "epoch": 4.65, + "learning_rate": 7.853376473661826e-06, + "loss": 0.0173, + "step": 35532 + }, + { + "epoch": 4.65, + "learning_rate": 7.852858514383762e-06, + "loss": 0.0176, + "step": 35533 + }, + { + "epoch": 4.65, + "learning_rate": 7.852340561144413e-06, + "loss": 0.0195, + "step": 35534 + }, + { + "epoch": 4.65, + "learning_rate": 7.851822613945244e-06, + "loss": 0.0192, + "step": 35535 + }, + { + "epoch": 4.65, + "learning_rate": 7.851304672787712e-06, + "loss": 0.0191, + "step": 35536 + }, + { + "epoch": 4.66, + "learning_rate": 7.850786737673268e-06, + "loss": 0.0197, + "step": 35537 + }, + { + "epoch": 4.66, + "learning_rate": 7.850268808603373e-06, + "loss": 0.0191, + "step": 35538 + }, + { + "epoch": 4.66, + "learning_rate": 7.849750885579484e-06, + "loss": 0.0163, + "step": 35539 + }, + { + "epoch": 4.66, + "learning_rate": 7.849232968603055e-06, + "loss": 0.0182, + "step": 35540 + }, + { + "epoch": 4.66, + "learning_rate": 7.848715057675542e-06, + "loss": 0.016, + "step": 35541 + }, + { + "epoch": 4.66, + "learning_rate": 7.848197152798402e-06, + "loss": 0.016, + "step": 35542 + }, + { + "epoch": 4.66, + "learning_rate": 7.847679253973095e-06, + "loss": 0.0179, + "step": 35543 + }, + { + "epoch": 4.66, + "learning_rate": 7.847161361201078e-06, + "loss": 0.0143, + "step": 35544 + }, + { + "epoch": 4.66, + "learning_rate": 7.846643474483801e-06, + "loss": 0.0189, + "step": 35545 + }, + { + "epoch": 4.66, + "learning_rate": 7.846125593822728e-06, + "loss": 0.0197, + "step": 35546 + }, + { + "epoch": 4.66, + "learning_rate": 7.84560771921931e-06, + "loss": 0.0163, + "step": 35547 + }, + { + "epoch": 4.66, + "learning_rate": 7.845089850675008e-06, + "loss": 0.0201, + "step": 35548 + }, + { + "epoch": 4.66, + "learning_rate": 7.844571988191274e-06, + "loss": 0.0197, + "step": 35549 + }, + { + "epoch": 4.66, + "learning_rate": 7.844054131769566e-06, + "loss": 0.0158, + "step": 35550 + }, + { + "epoch": 4.66, + "learning_rate": 7.843536281411344e-06, + "loss": 0.0157, + "step": 35551 + }, + { + "epoch": 4.66, + "learning_rate": 7.843018437118058e-06, + "loss": 0.0196, + "step": 35552 + }, + { + "epoch": 4.66, + "learning_rate": 7.842500598891166e-06, + "loss": 0.021, + "step": 35553 + }, + { + "epoch": 4.66, + "learning_rate": 7.84198276673213e-06, + "loss": 0.0187, + "step": 35554 + }, + { + "epoch": 4.66, + "learning_rate": 7.8414649406424e-06, + "loss": 0.0182, + "step": 35555 + }, + { + "epoch": 4.66, + "learning_rate": 7.840947120623436e-06, + "loss": 0.0156, + "step": 35556 + }, + { + "epoch": 4.66, + "learning_rate": 7.840429306676696e-06, + "loss": 0.0159, + "step": 35557 + }, + { + "epoch": 4.66, + "learning_rate": 7.83991149880363e-06, + "loss": 0.0196, + "step": 35558 + }, + { + "epoch": 4.66, + "learning_rate": 7.8393936970057e-06, + "loss": 0.0184, + "step": 35559 + }, + { + "epoch": 4.66, + "learning_rate": 7.83887590128436e-06, + "loss": 0.0186, + "step": 35560 + }, + { + "epoch": 4.66, + "learning_rate": 7.838358111641066e-06, + "loss": 0.0179, + "step": 35561 + }, + { + "epoch": 4.66, + "learning_rate": 7.837840328077274e-06, + "loss": 0.0204, + "step": 35562 + }, + { + "epoch": 4.66, + "learning_rate": 7.837322550594439e-06, + "loss": 0.0209, + "step": 35563 + }, + { + "epoch": 4.66, + "learning_rate": 7.836804779194021e-06, + "loss": 0.0173, + "step": 35564 + }, + { + "epoch": 4.66, + "learning_rate": 7.836287013877475e-06, + "loss": 0.0173, + "step": 35565 + }, + { + "epoch": 4.66, + "learning_rate": 7.835769254646257e-06, + "loss": 0.0179, + "step": 35566 + }, + { + "epoch": 4.66, + "learning_rate": 7.835251501501823e-06, + "loss": 0.0199, + "step": 35567 + }, + { + "epoch": 4.66, + "learning_rate": 7.834733754445627e-06, + "loss": 0.0224, + "step": 35568 + }, + { + "epoch": 4.66, + "learning_rate": 7.83421601347913e-06, + "loss": 0.0178, + "step": 35569 + }, + { + "epoch": 4.66, + "learning_rate": 7.833698278603784e-06, + "loss": 0.0199, + "step": 35570 + }, + { + "epoch": 4.66, + "learning_rate": 7.833180549821043e-06, + "loss": 0.0186, + "step": 35571 + }, + { + "epoch": 4.66, + "learning_rate": 7.83266282713237e-06, + "loss": 0.0188, + "step": 35572 + }, + { + "epoch": 4.66, + "learning_rate": 7.832145110539215e-06, + "loss": 0.0194, + "step": 35573 + }, + { + "epoch": 4.66, + "learning_rate": 7.831627400043036e-06, + "loss": 0.0204, + "step": 35574 + }, + { + "epoch": 4.66, + "learning_rate": 7.831109695645291e-06, + "loss": 0.0184, + "step": 35575 + }, + { + "epoch": 4.66, + "learning_rate": 7.830591997347434e-06, + "loss": 0.0169, + "step": 35576 + }, + { + "epoch": 4.66, + "learning_rate": 7.830074305150923e-06, + "loss": 0.0179, + "step": 35577 + }, + { + "epoch": 4.66, + "learning_rate": 7.829556619057213e-06, + "loss": 0.0152, + "step": 35578 + }, + { + "epoch": 4.66, + "learning_rate": 7.829038939067757e-06, + "loss": 0.0202, + "step": 35579 + }, + { + "epoch": 4.66, + "learning_rate": 7.828521265184015e-06, + "loss": 0.0198, + "step": 35580 + }, + { + "epoch": 4.66, + "learning_rate": 7.82800359740744e-06, + "loss": 0.0162, + "step": 35581 + }, + { + "epoch": 4.66, + "learning_rate": 7.82748593573949e-06, + "loss": 0.0181, + "step": 35582 + }, + { + "epoch": 4.66, + "learning_rate": 7.82696828018162e-06, + "loss": 0.0204, + "step": 35583 + }, + { + "epoch": 4.66, + "learning_rate": 7.826450630735287e-06, + "loss": 0.0217, + "step": 35584 + }, + { + "epoch": 4.66, + "learning_rate": 7.82593298740194e-06, + "loss": 0.02, + "step": 35585 + }, + { + "epoch": 4.66, + "learning_rate": 7.825415350183047e-06, + "loss": 0.0159, + "step": 35586 + }, + { + "epoch": 4.66, + "learning_rate": 7.824897719080056e-06, + "loss": 0.0211, + "step": 35587 + }, + { + "epoch": 4.66, + "learning_rate": 7.824380094094425e-06, + "loss": 0.0222, + "step": 35588 + }, + { + "epoch": 4.66, + "learning_rate": 7.823862475227607e-06, + "loss": 0.0181, + "step": 35589 + }, + { + "epoch": 4.66, + "learning_rate": 7.823344862481063e-06, + "loss": 0.0144, + "step": 35590 + }, + { + "epoch": 4.66, + "learning_rate": 7.822827255856244e-06, + "loss": 0.0188, + "step": 35591 + }, + { + "epoch": 4.66, + "learning_rate": 7.822309655354607e-06, + "loss": 0.0167, + "step": 35592 + }, + { + "epoch": 4.66, + "learning_rate": 7.821792060977609e-06, + "loss": 0.0226, + "step": 35593 + }, + { + "epoch": 4.66, + "learning_rate": 7.821274472726704e-06, + "loss": 0.0197, + "step": 35594 + }, + { + "epoch": 4.66, + "learning_rate": 7.820756890603349e-06, + "loss": 0.0216, + "step": 35595 + }, + { + "epoch": 4.66, + "learning_rate": 7.820239314608997e-06, + "loss": 0.0231, + "step": 35596 + }, + { + "epoch": 4.66, + "learning_rate": 7.819721744745108e-06, + "loss": 0.0143, + "step": 35597 + }, + { + "epoch": 4.66, + "learning_rate": 7.819204181013135e-06, + "loss": 0.019, + "step": 35598 + }, + { + "epoch": 4.66, + "learning_rate": 7.818686623414535e-06, + "loss": 0.0161, + "step": 35599 + }, + { + "epoch": 4.66, + "learning_rate": 7.81816907195076e-06, + "loss": 0.0134, + "step": 35600 + }, + { + "epoch": 4.66, + "learning_rate": 7.817651526623273e-06, + "loss": 0.019, + "step": 35601 + }, + { + "epoch": 4.66, + "learning_rate": 7.817133987433522e-06, + "loss": 0.0211, + "step": 35602 + }, + { + "epoch": 4.66, + "learning_rate": 7.816616454382967e-06, + "loss": 0.0221, + "step": 35603 + }, + { + "epoch": 4.66, + "learning_rate": 7.81609892747306e-06, + "loss": 0.0152, + "step": 35604 + }, + { + "epoch": 4.66, + "learning_rate": 7.81558140670526e-06, + "loss": 0.0167, + "step": 35605 + }, + { + "epoch": 4.66, + "learning_rate": 7.81506389208102e-06, + "loss": 0.0136, + "step": 35606 + }, + { + "epoch": 4.66, + "learning_rate": 7.814546383601793e-06, + "loss": 0.0179, + "step": 35607 + }, + { + "epoch": 4.66, + "learning_rate": 7.814028881269041e-06, + "loss": 0.0168, + "step": 35608 + }, + { + "epoch": 4.66, + "learning_rate": 7.81351138508422e-06, + "loss": 0.0173, + "step": 35609 + }, + { + "epoch": 4.66, + "learning_rate": 7.812993895048777e-06, + "loss": 0.0176, + "step": 35610 + }, + { + "epoch": 4.66, + "learning_rate": 7.812476411164175e-06, + "loss": 0.0195, + "step": 35611 + }, + { + "epoch": 4.66, + "learning_rate": 7.811958933431864e-06, + "loss": 0.0191, + "step": 35612 + }, + { + "epoch": 4.67, + "learning_rate": 7.811441461853303e-06, + "loss": 0.0226, + "step": 35613 + }, + { + "epoch": 4.67, + "learning_rate": 7.810923996429948e-06, + "loss": 0.0215, + "step": 35614 + }, + { + "epoch": 4.67, + "learning_rate": 7.810406537163252e-06, + "loss": 0.0163, + "step": 35615 + }, + { + "epoch": 4.67, + "learning_rate": 7.809889084054672e-06, + "loss": 0.0195, + "step": 35616 + }, + { + "epoch": 4.67, + "learning_rate": 7.80937163710566e-06, + "loss": 0.018, + "step": 35617 + }, + { + "epoch": 4.67, + "learning_rate": 7.808854196317671e-06, + "loss": 0.0193, + "step": 35618 + }, + { + "epoch": 4.67, + "learning_rate": 7.808336761692168e-06, + "loss": 0.0194, + "step": 35619 + }, + { + "epoch": 4.67, + "learning_rate": 7.807819333230599e-06, + "loss": 0.0151, + "step": 35620 + }, + { + "epoch": 4.67, + "learning_rate": 7.80730191093442e-06, + "loss": 0.0238, + "step": 35621 + }, + { + "epoch": 4.67, + "learning_rate": 7.806784494805092e-06, + "loss": 0.0226, + "step": 35622 + }, + { + "epoch": 4.67, + "learning_rate": 7.806267084844063e-06, + "loss": 0.02, + "step": 35623 + }, + { + "epoch": 4.67, + "learning_rate": 7.805749681052792e-06, + "loss": 0.0263, + "step": 35624 + }, + { + "epoch": 4.67, + "learning_rate": 7.80523228343273e-06, + "loss": 0.0192, + "step": 35625 + }, + { + "epoch": 4.67, + "learning_rate": 7.804714891985341e-06, + "loss": 0.0165, + "step": 35626 + }, + { + "epoch": 4.67, + "learning_rate": 7.80419750671207e-06, + "loss": 0.0199, + "step": 35627 + }, + { + "epoch": 4.67, + "learning_rate": 7.803680127614377e-06, + "loss": 0.0228, + "step": 35628 + }, + { + "epoch": 4.67, + "learning_rate": 7.803162754693714e-06, + "loss": 0.0193, + "step": 35629 + }, + { + "epoch": 4.67, + "learning_rate": 7.802645387951546e-06, + "loss": 0.0185, + "step": 35630 + }, + { + "epoch": 4.67, + "learning_rate": 7.802128027389314e-06, + "loss": 0.0206, + "step": 35631 + }, + { + "epoch": 4.67, + "learning_rate": 7.801610673008486e-06, + "loss": 0.0221, + "step": 35632 + }, + { + "epoch": 4.67, + "learning_rate": 7.801093324810508e-06, + "loss": 0.018, + "step": 35633 + }, + { + "epoch": 4.67, + "learning_rate": 7.800575982796838e-06, + "loss": 0.0184, + "step": 35634 + }, + { + "epoch": 4.67, + "learning_rate": 7.80005864696893e-06, + "loss": 0.0163, + "step": 35635 + }, + { + "epoch": 4.67, + "learning_rate": 7.79954131732824e-06, + "loss": 0.0208, + "step": 35636 + }, + { + "epoch": 4.67, + "learning_rate": 7.799023993876225e-06, + "loss": 0.0182, + "step": 35637 + }, + { + "epoch": 4.67, + "learning_rate": 7.798506676614334e-06, + "loss": 0.0243, + "step": 35638 + }, + { + "epoch": 4.67, + "learning_rate": 7.797989365544031e-06, + "loss": 0.0184, + "step": 35639 + }, + { + "epoch": 4.67, + "learning_rate": 7.797472060666759e-06, + "loss": 0.0168, + "step": 35640 + }, + { + "epoch": 4.67, + "learning_rate": 7.796954761983984e-06, + "loss": 0.018, + "step": 35641 + }, + { + "epoch": 4.67, + "learning_rate": 7.796437469497154e-06, + "loss": 0.0217, + "step": 35642 + }, + { + "epoch": 4.67, + "learning_rate": 7.79592018320773e-06, + "loss": 0.0228, + "step": 35643 + }, + { + "epoch": 4.67, + "learning_rate": 7.795402903117159e-06, + "loss": 0.0188, + "step": 35644 + }, + { + "epoch": 4.67, + "learning_rate": 7.794885629226903e-06, + "loss": 0.0159, + "step": 35645 + }, + { + "epoch": 4.67, + "learning_rate": 7.794368361538412e-06, + "loss": 0.0169, + "step": 35646 + }, + { + "epoch": 4.67, + "learning_rate": 7.793851100053145e-06, + "loss": 0.0241, + "step": 35647 + }, + { + "epoch": 4.67, + "learning_rate": 7.793333844772552e-06, + "loss": 0.0183, + "step": 35648 + }, + { + "epoch": 4.67, + "learning_rate": 7.792816595698092e-06, + "loss": 0.0204, + "step": 35649 + }, + { + "epoch": 4.67, + "learning_rate": 7.792299352831216e-06, + "loss": 0.0141, + "step": 35650 + }, + { + "epoch": 4.67, + "learning_rate": 7.791782116173377e-06, + "loss": 0.0218, + "step": 35651 + }, + { + "epoch": 4.67, + "learning_rate": 7.791264885726038e-06, + "loss": 0.0164, + "step": 35652 + }, + { + "epoch": 4.67, + "learning_rate": 7.790747661490648e-06, + "loss": 0.0171, + "step": 35653 + }, + { + "epoch": 4.67, + "learning_rate": 7.790230443468662e-06, + "loss": 0.0196, + "step": 35654 + }, + { + "epoch": 4.67, + "learning_rate": 7.789713231661537e-06, + "loss": 0.0183, + "step": 35655 + }, + { + "epoch": 4.67, + "learning_rate": 7.789196026070724e-06, + "loss": 0.0227, + "step": 35656 + }, + { + "epoch": 4.67, + "learning_rate": 7.78867882669768e-06, + "loss": 0.0186, + "step": 35657 + }, + { + "epoch": 4.67, + "learning_rate": 7.788161633543859e-06, + "loss": 0.0197, + "step": 35658 + }, + { + "epoch": 4.67, + "learning_rate": 7.787644446610715e-06, + "loss": 0.0187, + "step": 35659 + }, + { + "epoch": 4.67, + "learning_rate": 7.787127265899706e-06, + "loss": 0.0212, + "step": 35660 + }, + { + "epoch": 4.67, + "learning_rate": 7.786610091412282e-06, + "loss": 0.018, + "step": 35661 + }, + { + "epoch": 4.67, + "learning_rate": 7.786092923149896e-06, + "loss": 0.0179, + "step": 35662 + }, + { + "epoch": 4.67, + "learning_rate": 7.785575761114007e-06, + "loss": 0.0164, + "step": 35663 + }, + { + "epoch": 4.67, + "learning_rate": 7.785058605306071e-06, + "loss": 0.013, + "step": 35664 + }, + { + "epoch": 4.67, + "learning_rate": 7.784541455727536e-06, + "loss": 0.0169, + "step": 35665 + }, + { + "epoch": 4.67, + "learning_rate": 7.784024312379865e-06, + "loss": 0.0192, + "step": 35666 + }, + { + "epoch": 4.67, + "learning_rate": 7.783507175264503e-06, + "loss": 0.0259, + "step": 35667 + }, + { + "epoch": 4.67, + "learning_rate": 7.782990044382915e-06, + "loss": 0.0191, + "step": 35668 + }, + { + "epoch": 4.67, + "learning_rate": 7.782472919736545e-06, + "loss": 0.0175, + "step": 35669 + }, + { + "epoch": 4.67, + "learning_rate": 7.781955801326854e-06, + "loss": 0.0224, + "step": 35670 + }, + { + "epoch": 4.67, + "learning_rate": 7.781438689155291e-06, + "loss": 0.017, + "step": 35671 + }, + { + "epoch": 4.67, + "learning_rate": 7.780921583223316e-06, + "loss": 0.0163, + "step": 35672 + }, + { + "epoch": 4.67, + "learning_rate": 7.780404483532377e-06, + "loss": 0.0122, + "step": 35673 + }, + { + "epoch": 4.67, + "learning_rate": 7.779887390083935e-06, + "loss": 0.023, + "step": 35674 + }, + { + "epoch": 4.67, + "learning_rate": 7.779370302879443e-06, + "loss": 0.0166, + "step": 35675 + }, + { + "epoch": 4.67, + "learning_rate": 7.778853221920353e-06, + "loss": 0.0225, + "step": 35676 + }, + { + "epoch": 4.67, + "learning_rate": 7.778336147208119e-06, + "loss": 0.0187, + "step": 35677 + }, + { + "epoch": 4.67, + "learning_rate": 7.777819078744197e-06, + "loss": 0.0186, + "step": 35678 + }, + { + "epoch": 4.67, + "learning_rate": 7.777302016530042e-06, + "loss": 0.0199, + "step": 35679 + }, + { + "epoch": 4.67, + "learning_rate": 7.776784960567105e-06, + "loss": 0.0192, + "step": 35680 + }, + { + "epoch": 4.67, + "learning_rate": 7.776267910856843e-06, + "loss": 0.0193, + "step": 35681 + }, + { + "epoch": 4.67, + "learning_rate": 7.775750867400706e-06, + "loss": 0.0151, + "step": 35682 + }, + { + "epoch": 4.67, + "learning_rate": 7.775233830200155e-06, + "loss": 0.0182, + "step": 35683 + }, + { + "epoch": 4.67, + "learning_rate": 7.774716799256636e-06, + "loss": 0.014, + "step": 35684 + }, + { + "epoch": 4.67, + "learning_rate": 7.77419977457161e-06, + "loss": 0.0183, + "step": 35685 + }, + { + "epoch": 4.67, + "learning_rate": 7.773682756146528e-06, + "loss": 0.0159, + "step": 35686 + }, + { + "epoch": 4.67, + "learning_rate": 7.773165743982847e-06, + "loss": 0.0164, + "step": 35687 + }, + { + "epoch": 4.67, + "learning_rate": 7.772648738082015e-06, + "loss": 0.0206, + "step": 35688 + }, + { + "epoch": 4.68, + "learning_rate": 7.772131738445495e-06, + "loss": 0.0194, + "step": 35689 + }, + { + "epoch": 4.68, + "learning_rate": 7.771614745074733e-06, + "loss": 0.0178, + "step": 35690 + }, + { + "epoch": 4.68, + "learning_rate": 7.771097757971186e-06, + "loss": 0.0166, + "step": 35691 + }, + { + "epoch": 4.68, + "learning_rate": 7.770580777136307e-06, + "loss": 0.0191, + "step": 35692 + }, + { + "epoch": 4.68, + "learning_rate": 7.770063802571552e-06, + "loss": 0.0146, + "step": 35693 + }, + { + "epoch": 4.68, + "learning_rate": 7.769546834278369e-06, + "loss": 0.0171, + "step": 35694 + }, + { + "epoch": 4.68, + "learning_rate": 7.769029872258224e-06, + "loss": 0.0192, + "step": 35695 + }, + { + "epoch": 4.68, + "learning_rate": 7.76851291651256e-06, + "loss": 0.0157, + "step": 35696 + }, + { + "epoch": 4.68, + "learning_rate": 7.767995967042836e-06, + "loss": 0.0263, + "step": 35697 + }, + { + "epoch": 4.68, + "learning_rate": 7.767479023850502e-06, + "loss": 0.0232, + "step": 35698 + }, + { + "epoch": 4.68, + "learning_rate": 7.766962086937019e-06, + "loss": 0.0134, + "step": 35699 + }, + { + "epoch": 4.68, + "learning_rate": 7.766445156303832e-06, + "loss": 0.0229, + "step": 35700 + }, + { + "epoch": 4.68, + "learning_rate": 7.7659282319524e-06, + "loss": 0.0189, + "step": 35701 + }, + { + "epoch": 4.68, + "learning_rate": 7.765411313884178e-06, + "loss": 0.0166, + "step": 35702 + }, + { + "epoch": 4.68, + "learning_rate": 7.764894402100615e-06, + "loss": 0.0213, + "step": 35703 + }, + { + "epoch": 4.68, + "learning_rate": 7.76437749660317e-06, + "loss": 0.0188, + "step": 35704 + }, + { + "epoch": 4.68, + "learning_rate": 7.76386059739329e-06, + "loss": 0.0225, + "step": 35705 + }, + { + "epoch": 4.68, + "learning_rate": 7.763343704472435e-06, + "loss": 0.0203, + "step": 35706 + }, + { + "epoch": 4.68, + "learning_rate": 7.762826817842056e-06, + "loss": 0.0207, + "step": 35707 + }, + { + "epoch": 4.68, + "learning_rate": 7.762309937503611e-06, + "loss": 0.0141, + "step": 35708 + }, + { + "epoch": 4.68, + "learning_rate": 7.761793063458546e-06, + "loss": 0.0194, + "step": 35709 + }, + { + "epoch": 4.68, + "learning_rate": 7.761276195708324e-06, + "loss": 0.02, + "step": 35710 + }, + { + "epoch": 4.68, + "learning_rate": 7.760759334254388e-06, + "loss": 0.0277, + "step": 35711 + }, + { + "epoch": 4.68, + "learning_rate": 7.760242479098202e-06, + "loss": 0.0227, + "step": 35712 + }, + { + "epoch": 4.68, + "learning_rate": 7.75972563024121e-06, + "loss": 0.0218, + "step": 35713 + }, + { + "epoch": 4.68, + "learning_rate": 7.759208787684876e-06, + "loss": 0.0176, + "step": 35714 + }, + { + "epoch": 4.68, + "learning_rate": 7.758691951430642e-06, + "loss": 0.0189, + "step": 35715 + }, + { + "epoch": 4.68, + "learning_rate": 7.758175121479967e-06, + "loss": 0.0174, + "step": 35716 + }, + { + "epoch": 4.68, + "learning_rate": 7.757658297834308e-06, + "loss": 0.0174, + "step": 35717 + }, + { + "epoch": 4.68, + "learning_rate": 7.757141480495117e-06, + "loss": 0.0154, + "step": 35718 + }, + { + "epoch": 4.68, + "learning_rate": 7.756624669463845e-06, + "loss": 0.023, + "step": 35719 + }, + { + "epoch": 4.68, + "learning_rate": 7.756107864741947e-06, + "loss": 0.0176, + "step": 35720 + }, + { + "epoch": 4.68, + "learning_rate": 7.755591066330875e-06, + "loss": 0.0226, + "step": 35721 + }, + { + "epoch": 4.68, + "learning_rate": 7.755074274232084e-06, + "loss": 0.0144, + "step": 35722 + }, + { + "epoch": 4.68, + "learning_rate": 7.754557488447029e-06, + "loss": 0.0183, + "step": 35723 + }, + { + "epoch": 4.68, + "learning_rate": 7.754040708977158e-06, + "loss": 0.0185, + "step": 35724 + }, + { + "epoch": 4.68, + "learning_rate": 7.753523935823932e-06, + "loss": 0.0144, + "step": 35725 + }, + { + "epoch": 4.68, + "learning_rate": 7.753007168988798e-06, + "loss": 0.0159, + "step": 35726 + }, + { + "epoch": 4.68, + "learning_rate": 7.752490408473207e-06, + "loss": 0.0146, + "step": 35727 + }, + { + "epoch": 4.68, + "learning_rate": 7.751973654278622e-06, + "loss": 0.014, + "step": 35728 + }, + { + "epoch": 4.68, + "learning_rate": 7.751456906406492e-06, + "loss": 0.0185, + "step": 35729 + }, + { + "epoch": 4.68, + "learning_rate": 7.750940164858268e-06, + "loss": 0.0248, + "step": 35730 + }, + { + "epoch": 4.68, + "learning_rate": 7.750423429635408e-06, + "loss": 0.0183, + "step": 35731 + }, + { + "epoch": 4.68, + "learning_rate": 7.74990670073936e-06, + "loss": 0.0198, + "step": 35732 + }, + { + "epoch": 4.68, + "learning_rate": 7.74938997817158e-06, + "loss": 0.0198, + "step": 35733 + }, + { + "epoch": 4.68, + "learning_rate": 7.74887326193352e-06, + "loss": 0.0217, + "step": 35734 + }, + { + "epoch": 4.68, + "learning_rate": 7.748356552026637e-06, + "loss": 0.0158, + "step": 35735 + }, + { + "epoch": 4.68, + "learning_rate": 7.747839848452377e-06, + "loss": 0.0211, + "step": 35736 + }, + { + "epoch": 4.68, + "learning_rate": 7.747323151212199e-06, + "loss": 0.0163, + "step": 35737 + }, + { + "epoch": 4.68, + "learning_rate": 7.746806460307553e-06, + "loss": 0.0174, + "step": 35738 + }, + { + "epoch": 4.68, + "learning_rate": 7.746289775739897e-06, + "loss": 0.0167, + "step": 35739 + }, + { + "epoch": 4.68, + "learning_rate": 7.745773097510678e-06, + "loss": 0.0197, + "step": 35740 + }, + { + "epoch": 4.68, + "learning_rate": 7.745256425621357e-06, + "loss": 0.0184, + "step": 35741 + }, + { + "epoch": 4.68, + "learning_rate": 7.744739760073378e-06, + "loss": 0.0187, + "step": 35742 + }, + { + "epoch": 4.68, + "learning_rate": 7.7442231008682e-06, + "loss": 0.0231, + "step": 35743 + }, + { + "epoch": 4.68, + "learning_rate": 7.743706448007274e-06, + "loss": 0.0178, + "step": 35744 + }, + { + "epoch": 4.68, + "learning_rate": 7.743189801492052e-06, + "loss": 0.027, + "step": 35745 + }, + { + "epoch": 4.68, + "learning_rate": 7.742673161323992e-06, + "loss": 0.0231, + "step": 35746 + }, + { + "epoch": 4.68, + "learning_rate": 7.74215652750454e-06, + "loss": 0.0201, + "step": 35747 + }, + { + "epoch": 4.68, + "learning_rate": 7.741639900035155e-06, + "loss": 0.0178, + "step": 35748 + }, + { + "epoch": 4.68, + "learning_rate": 7.741123278917284e-06, + "loss": 0.02, + "step": 35749 + }, + { + "epoch": 4.68, + "learning_rate": 7.740606664152386e-06, + "loss": 0.0192, + "step": 35750 + }, + { + "epoch": 4.68, + "learning_rate": 7.740090055741912e-06, + "loss": 0.0182, + "step": 35751 + }, + { + "epoch": 4.68, + "learning_rate": 7.739573453687315e-06, + "loss": 0.0204, + "step": 35752 + }, + { + "epoch": 4.68, + "learning_rate": 7.739056857990045e-06, + "loss": 0.0195, + "step": 35753 + }, + { + "epoch": 4.68, + "learning_rate": 7.73854026865156e-06, + "loss": 0.0199, + "step": 35754 + }, + { + "epoch": 4.68, + "learning_rate": 7.738023685673307e-06, + "loss": 0.0196, + "step": 35755 + }, + { + "epoch": 4.68, + "learning_rate": 7.737507109056745e-06, + "loss": 0.0177, + "step": 35756 + }, + { + "epoch": 4.68, + "learning_rate": 7.736990538803321e-06, + "loss": 0.0213, + "step": 35757 + }, + { + "epoch": 4.68, + "learning_rate": 7.736473974914491e-06, + "loss": 0.0218, + "step": 35758 + }, + { + "epoch": 4.68, + "learning_rate": 7.73595741739171e-06, + "loss": 0.0182, + "step": 35759 + }, + { + "epoch": 4.68, + "learning_rate": 7.735440866236424e-06, + "loss": 0.022, + "step": 35760 + }, + { + "epoch": 4.68, + "learning_rate": 7.73492432145009e-06, + "loss": 0.0202, + "step": 35761 + }, + { + "epoch": 4.68, + "learning_rate": 7.734407783034164e-06, + "loss": 0.022, + "step": 35762 + }, + { + "epoch": 4.68, + "learning_rate": 7.733891250990095e-06, + "loss": 0.0181, + "step": 35763 + }, + { + "epoch": 4.68, + "learning_rate": 7.733374725319336e-06, + "loss": 0.0168, + "step": 35764 + }, + { + "epoch": 4.68, + "learning_rate": 7.73285820602334e-06, + "loss": 0.0209, + "step": 35765 + }, + { + "epoch": 4.69, + "learning_rate": 7.732341693103558e-06, + "loss": 0.0226, + "step": 35766 + }, + { + "epoch": 4.69, + "learning_rate": 7.731825186561446e-06, + "loss": 0.0162, + "step": 35767 + }, + { + "epoch": 4.69, + "learning_rate": 7.731308686398453e-06, + "loss": 0.0163, + "step": 35768 + }, + { + "epoch": 4.69, + "learning_rate": 7.730792192616036e-06, + "loss": 0.0128, + "step": 35769 + }, + { + "epoch": 4.69, + "learning_rate": 7.730275705215643e-06, + "loss": 0.0185, + "step": 35770 + }, + { + "epoch": 4.69, + "learning_rate": 7.729759224198728e-06, + "loss": 0.0158, + "step": 35771 + }, + { + "epoch": 4.69, + "learning_rate": 7.729242749566743e-06, + "loss": 0.0202, + "step": 35772 + }, + { + "epoch": 4.69, + "learning_rate": 7.728726281321147e-06, + "loss": 0.0241, + "step": 35773 + }, + { + "epoch": 4.69, + "learning_rate": 7.728209819463383e-06, + "loss": 0.0184, + "step": 35774 + }, + { + "epoch": 4.69, + "learning_rate": 7.727693363994913e-06, + "loss": 0.017, + "step": 35775 + }, + { + "epoch": 4.69, + "learning_rate": 7.72717691491718e-06, + "loss": 0.0156, + "step": 35776 + }, + { + "epoch": 4.69, + "learning_rate": 7.726660472231645e-06, + "loss": 0.0174, + "step": 35777 + }, + { + "epoch": 4.69, + "learning_rate": 7.726144035939751e-06, + "loss": 0.0182, + "step": 35778 + }, + { + "epoch": 4.69, + "learning_rate": 7.725627606042958e-06, + "loss": 0.0234, + "step": 35779 + }, + { + "epoch": 4.69, + "learning_rate": 7.725111182542719e-06, + "loss": 0.0163, + "step": 35780 + }, + { + "epoch": 4.69, + "learning_rate": 7.72459476544048e-06, + "loss": 0.0202, + "step": 35781 + }, + { + "epoch": 4.69, + "learning_rate": 7.724078354737696e-06, + "loss": 0.02, + "step": 35782 + }, + { + "epoch": 4.69, + "learning_rate": 7.723561950435822e-06, + "loss": 0.0138, + "step": 35783 + }, + { + "epoch": 4.69, + "learning_rate": 7.72304555253631e-06, + "loss": 0.0233, + "step": 35784 + }, + { + "epoch": 4.69, + "learning_rate": 7.722529161040611e-06, + "loss": 0.0158, + "step": 35785 + }, + { + "epoch": 4.69, + "learning_rate": 7.722012775950177e-06, + "loss": 0.0165, + "step": 35786 + }, + { + "epoch": 4.69, + "learning_rate": 7.72149639726646e-06, + "loss": 0.0259, + "step": 35787 + }, + { + "epoch": 4.69, + "learning_rate": 7.720980024990916e-06, + "loss": 0.0156, + "step": 35788 + }, + { + "epoch": 4.69, + "learning_rate": 7.72046365912499e-06, + "loss": 0.016, + "step": 35789 + }, + { + "epoch": 4.69, + "learning_rate": 7.719947299670144e-06, + "loss": 0.0219, + "step": 35790 + }, + { + "epoch": 4.69, + "learning_rate": 7.71943094662782e-06, + "loss": 0.0152, + "step": 35791 + }, + { + "epoch": 4.69, + "learning_rate": 7.718914599999478e-06, + "loss": 0.0196, + "step": 35792 + }, + { + "epoch": 4.69, + "learning_rate": 7.71839825978656e-06, + "loss": 0.0195, + "step": 35793 + }, + { + "epoch": 4.69, + "learning_rate": 7.717881925990534e-06, + "loss": 0.0201, + "step": 35794 + }, + { + "epoch": 4.69, + "learning_rate": 7.71736559861284e-06, + "loss": 0.0181, + "step": 35795 + }, + { + "epoch": 4.69, + "learning_rate": 7.716849277654935e-06, + "loss": 0.0197, + "step": 35796 + }, + { + "epoch": 4.69, + "learning_rate": 7.716332963118269e-06, + "loss": 0.0206, + "step": 35797 + }, + { + "epoch": 4.69, + "learning_rate": 7.715816655004295e-06, + "loss": 0.0225, + "step": 35798 + }, + { + "epoch": 4.69, + "learning_rate": 7.715300353314465e-06, + "loss": 0.0208, + "step": 35799 + }, + { + "epoch": 4.69, + "learning_rate": 7.714784058050233e-06, + "loss": 0.0157, + "step": 35800 + }, + { + "epoch": 4.69, + "learning_rate": 7.714267769213047e-06, + "loss": 0.0176, + "step": 35801 + }, + { + "epoch": 4.69, + "learning_rate": 7.713751486804361e-06, + "loss": 0.0197, + "step": 35802 + }, + { + "epoch": 4.69, + "learning_rate": 7.71323521082563e-06, + "loss": 0.0196, + "step": 35803 + }, + { + "epoch": 4.69, + "learning_rate": 7.712718941278298e-06, + "loss": 0.0158, + "step": 35804 + }, + { + "epoch": 4.69, + "learning_rate": 7.712202678163823e-06, + "loss": 0.0179, + "step": 35805 + }, + { + "epoch": 4.69, + "learning_rate": 7.71168642148366e-06, + "loss": 0.0177, + "step": 35806 + }, + { + "epoch": 4.69, + "learning_rate": 7.711170171239255e-06, + "loss": 0.023, + "step": 35807 + }, + { + "epoch": 4.69, + "learning_rate": 7.710653927432061e-06, + "loss": 0.0207, + "step": 35808 + }, + { + "epoch": 4.69, + "learning_rate": 7.710137690063534e-06, + "loss": 0.0211, + "step": 35809 + }, + { + "epoch": 4.69, + "learning_rate": 7.709621459135121e-06, + "loss": 0.0173, + "step": 35810 + }, + { + "epoch": 4.69, + "learning_rate": 7.709105234648278e-06, + "loss": 0.0225, + "step": 35811 + }, + { + "epoch": 4.69, + "learning_rate": 7.70858901660445e-06, + "loss": 0.0162, + "step": 35812 + }, + { + "epoch": 4.69, + "learning_rate": 7.708072805005099e-06, + "loss": 0.0188, + "step": 35813 + }, + { + "epoch": 4.69, + "learning_rate": 7.707556599851666e-06, + "loss": 0.0206, + "step": 35814 + }, + { + "epoch": 4.69, + "learning_rate": 7.707040401145607e-06, + "loss": 0.0198, + "step": 35815 + }, + { + "epoch": 4.69, + "learning_rate": 7.706524208888376e-06, + "loss": 0.018, + "step": 35816 + }, + { + "epoch": 4.69, + "learning_rate": 7.706008023081427e-06, + "loss": 0.0203, + "step": 35817 + }, + { + "epoch": 4.69, + "learning_rate": 7.705491843726207e-06, + "loss": 0.0208, + "step": 35818 + }, + { + "epoch": 4.69, + "learning_rate": 7.704975670824168e-06, + "loss": 0.0208, + "step": 35819 + }, + { + "epoch": 4.69, + "learning_rate": 7.704459504376763e-06, + "loss": 0.0235, + "step": 35820 + }, + { + "epoch": 4.69, + "learning_rate": 7.703943344385445e-06, + "loss": 0.0174, + "step": 35821 + }, + { + "epoch": 4.69, + "learning_rate": 7.703427190851662e-06, + "loss": 0.0149, + "step": 35822 + }, + { + "epoch": 4.69, + "learning_rate": 7.702911043776867e-06, + "loss": 0.0166, + "step": 35823 + }, + { + "epoch": 4.69, + "learning_rate": 7.702394903162515e-06, + "loss": 0.018, + "step": 35824 + }, + { + "epoch": 4.69, + "learning_rate": 7.701878769010054e-06, + "loss": 0.022, + "step": 35825 + }, + { + "epoch": 4.69, + "learning_rate": 7.701362641320933e-06, + "loss": 0.0172, + "step": 35826 + }, + { + "epoch": 4.69, + "learning_rate": 7.700846520096612e-06, + "loss": 0.0175, + "step": 35827 + }, + { + "epoch": 4.69, + "learning_rate": 7.700330405338535e-06, + "loss": 0.0208, + "step": 35828 + }, + { + "epoch": 4.69, + "learning_rate": 7.699814297048156e-06, + "loss": 0.0168, + "step": 35829 + }, + { + "epoch": 4.69, + "learning_rate": 7.699298195226929e-06, + "loss": 0.0173, + "step": 35830 + }, + { + "epoch": 4.69, + "learning_rate": 7.698782099876303e-06, + "loss": 0.0171, + "step": 35831 + }, + { + "epoch": 4.69, + "learning_rate": 7.698266010997731e-06, + "loss": 0.0186, + "step": 35832 + }, + { + "epoch": 4.69, + "learning_rate": 7.69774992859266e-06, + "loss": 0.019, + "step": 35833 + }, + { + "epoch": 4.69, + "learning_rate": 7.697233852662548e-06, + "loss": 0.0205, + "step": 35834 + }, + { + "epoch": 4.69, + "learning_rate": 7.696717783208842e-06, + "loss": 0.0229, + "step": 35835 + }, + { + "epoch": 4.69, + "learning_rate": 7.69620172023299e-06, + "loss": 0.021, + "step": 35836 + }, + { + "epoch": 4.69, + "learning_rate": 7.695685663736451e-06, + "loss": 0.027, + "step": 35837 + }, + { + "epoch": 4.69, + "learning_rate": 7.695169613720677e-06, + "loss": 0.0188, + "step": 35838 + }, + { + "epoch": 4.69, + "learning_rate": 7.694653570187112e-06, + "loss": 0.0179, + "step": 35839 + }, + { + "epoch": 4.69, + "learning_rate": 7.694137533137214e-06, + "loss": 0.0151, + "step": 35840 + }, + { + "epoch": 4.69, + "learning_rate": 7.693621502572428e-06, + "loss": 0.0118, + "step": 35841 + }, + { + "epoch": 4.7, + "learning_rate": 7.693105478494212e-06, + "loss": 0.0141, + "step": 35842 + }, + { + "epoch": 4.7, + "learning_rate": 7.692589460904011e-06, + "loss": 0.0211, + "step": 35843 + }, + { + "epoch": 4.7, + "learning_rate": 7.69207344980328e-06, + "loss": 0.0167, + "step": 35844 + }, + { + "epoch": 4.7, + "learning_rate": 7.691557445193472e-06, + "loss": 0.0207, + "step": 35845 + }, + { + "epoch": 4.7, + "learning_rate": 7.691041447076031e-06, + "loss": 0.0141, + "step": 35846 + }, + { + "epoch": 4.7, + "learning_rate": 7.690525455452415e-06, + "loss": 0.0208, + "step": 35847 + }, + { + "epoch": 4.7, + "learning_rate": 7.690009470324074e-06, + "loss": 0.0199, + "step": 35848 + }, + { + "epoch": 4.7, + "learning_rate": 7.689493491692459e-06, + "loss": 0.0237, + "step": 35849 + }, + { + "epoch": 4.7, + "learning_rate": 7.68897751955902e-06, + "loss": 0.0147, + "step": 35850 + }, + { + "epoch": 4.7, + "learning_rate": 7.688461553925208e-06, + "loss": 0.0214, + "step": 35851 + }, + { + "epoch": 4.7, + "learning_rate": 7.687945594792474e-06, + "loss": 0.0198, + "step": 35852 + }, + { + "epoch": 4.7, + "learning_rate": 7.687429642162274e-06, + "loss": 0.0219, + "step": 35853 + }, + { + "epoch": 4.7, + "learning_rate": 7.68691369603605e-06, + "loss": 0.016, + "step": 35854 + }, + { + "epoch": 4.7, + "learning_rate": 7.686397756415262e-06, + "loss": 0.0218, + "step": 35855 + }, + { + "epoch": 4.7, + "learning_rate": 7.685881823301354e-06, + "loss": 0.0179, + "step": 35856 + }, + { + "epoch": 4.7, + "learning_rate": 7.685365896695783e-06, + "loss": 0.0213, + "step": 35857 + }, + { + "epoch": 4.7, + "learning_rate": 7.684849976599991e-06, + "loss": 0.0233, + "step": 35858 + }, + { + "epoch": 4.7, + "learning_rate": 7.684334063015441e-06, + "loss": 0.0188, + "step": 35859 + }, + { + "epoch": 4.7, + "learning_rate": 7.683818155943575e-06, + "loss": 0.0149, + "step": 35860 + }, + { + "epoch": 4.7, + "learning_rate": 7.683302255385851e-06, + "loss": 0.0202, + "step": 35861 + }, + { + "epoch": 4.7, + "learning_rate": 7.682786361343712e-06, + "loss": 0.0175, + "step": 35862 + }, + { + "epoch": 4.7, + "learning_rate": 7.682270473818618e-06, + "loss": 0.0193, + "step": 35863 + }, + { + "epoch": 4.7, + "learning_rate": 7.681754592812011e-06, + "loss": 0.0209, + "step": 35864 + }, + { + "epoch": 4.7, + "learning_rate": 7.681238718325348e-06, + "loss": 0.0185, + "step": 35865 + }, + { + "epoch": 4.7, + "learning_rate": 7.680722850360075e-06, + "loss": 0.0185, + "step": 35866 + }, + { + "epoch": 4.7, + "learning_rate": 7.680206988917645e-06, + "loss": 0.0217, + "step": 35867 + }, + { + "epoch": 4.7, + "learning_rate": 7.679691133999514e-06, + "loss": 0.0152, + "step": 35868 + }, + { + "epoch": 4.7, + "learning_rate": 7.67917528560712e-06, + "loss": 0.0193, + "step": 35869 + }, + { + "epoch": 4.7, + "learning_rate": 7.678659443741927e-06, + "loss": 0.0245, + "step": 35870 + }, + { + "epoch": 4.7, + "learning_rate": 7.678143608405383e-06, + "loss": 0.0173, + "step": 35871 + }, + { + "epoch": 4.7, + "learning_rate": 7.677627779598933e-06, + "loss": 0.0207, + "step": 35872 + }, + { + "epoch": 4.7, + "learning_rate": 7.677111957324031e-06, + "loss": 0.0265, + "step": 35873 + }, + { + "epoch": 4.7, + "learning_rate": 7.67659614158213e-06, + "loss": 0.0211, + "step": 35874 + }, + { + "epoch": 4.7, + "learning_rate": 7.676080332374678e-06, + "loss": 0.0195, + "step": 35875 + }, + { + "epoch": 4.7, + "learning_rate": 7.675564529703127e-06, + "loss": 0.0213, + "step": 35876 + }, + { + "epoch": 4.7, + "learning_rate": 7.675048733568926e-06, + "loss": 0.0204, + "step": 35877 + }, + { + "epoch": 4.7, + "learning_rate": 7.674532943973529e-06, + "loss": 0.0243, + "step": 35878 + }, + { + "epoch": 4.7, + "learning_rate": 7.67401716091838e-06, + "loss": 0.024, + "step": 35879 + }, + { + "epoch": 4.7, + "learning_rate": 7.673501384404934e-06, + "loss": 0.0186, + "step": 35880 + }, + { + "epoch": 4.7, + "learning_rate": 7.672985614434642e-06, + "loss": 0.0172, + "step": 35881 + }, + { + "epoch": 4.7, + "learning_rate": 7.672469851008957e-06, + "loss": 0.0251, + "step": 35882 + }, + { + "epoch": 4.7, + "learning_rate": 7.671954094129324e-06, + "loss": 0.0214, + "step": 35883 + }, + { + "epoch": 4.7, + "learning_rate": 7.6714383437972e-06, + "loss": 0.0217, + "step": 35884 + }, + { + "epoch": 4.7, + "learning_rate": 7.670922600014026e-06, + "loss": 0.0189, + "step": 35885 + }, + { + "epoch": 4.7, + "learning_rate": 7.670406862781263e-06, + "loss": 0.02, + "step": 35886 + }, + { + "epoch": 4.7, + "learning_rate": 7.669891132100354e-06, + "loss": 0.0201, + "step": 35887 + }, + { + "epoch": 4.7, + "learning_rate": 7.669375407972753e-06, + "loss": 0.019, + "step": 35888 + }, + { + "epoch": 4.7, + "learning_rate": 7.668859690399911e-06, + "loss": 0.0196, + "step": 35889 + }, + { + "epoch": 4.7, + "learning_rate": 7.668343979383273e-06, + "loss": 0.0234, + "step": 35890 + }, + { + "epoch": 4.7, + "learning_rate": 7.667828274924294e-06, + "loss": 0.0223, + "step": 35891 + }, + { + "epoch": 4.7, + "learning_rate": 7.667312577024426e-06, + "loss": 0.0214, + "step": 35892 + }, + { + "epoch": 4.7, + "learning_rate": 7.666796885685115e-06, + "loss": 0.0224, + "step": 35893 + }, + { + "epoch": 4.7, + "learning_rate": 7.666281200907815e-06, + "loss": 0.0124, + "step": 35894 + }, + { + "epoch": 4.7, + "learning_rate": 7.665765522693977e-06, + "loss": 0.0208, + "step": 35895 + }, + { + "epoch": 4.7, + "learning_rate": 7.665249851045046e-06, + "loss": 0.022, + "step": 35896 + }, + { + "epoch": 4.7, + "learning_rate": 7.664734185962478e-06, + "loss": 0.0205, + "step": 35897 + }, + { + "epoch": 4.7, + "learning_rate": 7.664218527447717e-06, + "loss": 0.0166, + "step": 35898 + }, + { + "epoch": 4.7, + "learning_rate": 7.66370287550222e-06, + "loss": 0.0179, + "step": 35899 + }, + { + "epoch": 4.7, + "learning_rate": 7.663187230127435e-06, + "loss": 0.017, + "step": 35900 + }, + { + "epoch": 4.7, + "learning_rate": 7.662671591324811e-06, + "loss": 0.0192, + "step": 35901 + }, + { + "epoch": 4.7, + "learning_rate": 7.662155959095794e-06, + "loss": 0.0233, + "step": 35902 + }, + { + "epoch": 4.7, + "learning_rate": 7.661640333441842e-06, + "loss": 0.0181, + "step": 35903 + }, + { + "epoch": 4.7, + "learning_rate": 7.661124714364404e-06, + "loss": 0.0196, + "step": 35904 + }, + { + "epoch": 4.7, + "learning_rate": 7.660609101864926e-06, + "loss": 0.0213, + "step": 35905 + }, + { + "epoch": 4.7, + "learning_rate": 7.660093495944862e-06, + "loss": 0.0139, + "step": 35906 + }, + { + "epoch": 4.7, + "learning_rate": 7.65957789660566e-06, + "loss": 0.0209, + "step": 35907 + }, + { + "epoch": 4.7, + "learning_rate": 7.65906230384877e-06, + "loss": 0.0251, + "step": 35908 + }, + { + "epoch": 4.7, + "learning_rate": 7.658546717675641e-06, + "loss": 0.0226, + "step": 35909 + }, + { + "epoch": 4.7, + "learning_rate": 7.658031138087727e-06, + "loss": 0.0146, + "step": 35910 + }, + { + "epoch": 4.7, + "learning_rate": 7.657515565086476e-06, + "loss": 0.0174, + "step": 35911 + }, + { + "epoch": 4.7, + "learning_rate": 7.656999998673338e-06, + "loss": 0.0192, + "step": 35912 + }, + { + "epoch": 4.7, + "learning_rate": 7.656484438849759e-06, + "loss": 0.0218, + "step": 35913 + }, + { + "epoch": 4.7, + "learning_rate": 7.655968885617194e-06, + "loss": 0.0171, + "step": 35914 + }, + { + "epoch": 4.7, + "learning_rate": 7.655453338977093e-06, + "loss": 0.0218, + "step": 35915 + }, + { + "epoch": 4.7, + "learning_rate": 7.654937798930905e-06, + "loss": 0.0212, + "step": 35916 + }, + { + "epoch": 4.7, + "learning_rate": 7.654422265480078e-06, + "loss": 0.0204, + "step": 35917 + }, + { + "epoch": 4.71, + "learning_rate": 7.653906738626065e-06, + "loss": 0.0178, + "step": 35918 + }, + { + "epoch": 4.71, + "learning_rate": 7.653391218370314e-06, + "loss": 0.0188, + "step": 35919 + }, + { + "epoch": 4.71, + "learning_rate": 7.652875704714276e-06, + "loss": 0.0198, + "step": 35920 + }, + { + "epoch": 4.71, + "learning_rate": 7.6523601976594e-06, + "loss": 0.0187, + "step": 35921 + }, + { + "epoch": 4.71, + "learning_rate": 7.651844697207136e-06, + "loss": 0.0186, + "step": 35922 + }, + { + "epoch": 4.71, + "learning_rate": 7.651329203358931e-06, + "loss": 0.0118, + "step": 35923 + }, + { + "epoch": 4.71, + "learning_rate": 7.650813716116238e-06, + "loss": 0.0169, + "step": 35924 + }, + { + "epoch": 4.71, + "learning_rate": 7.650298235480507e-06, + "loss": 0.0204, + "step": 35925 + }, + { + "epoch": 4.71, + "learning_rate": 7.649782761453189e-06, + "loss": 0.0174, + "step": 35926 + }, + { + "epoch": 4.71, + "learning_rate": 7.649267294035729e-06, + "loss": 0.0119, + "step": 35927 + }, + { + "epoch": 4.71, + "learning_rate": 7.648751833229583e-06, + "loss": 0.0188, + "step": 35928 + }, + { + "epoch": 4.71, + "learning_rate": 7.648236379036194e-06, + "loss": 0.0179, + "step": 35929 + }, + { + "epoch": 4.71, + "learning_rate": 7.647720931457017e-06, + "loss": 0.0196, + "step": 35930 + }, + { + "epoch": 4.71, + "learning_rate": 7.6472054904935e-06, + "loss": 0.0188, + "step": 35931 + }, + { + "epoch": 4.71, + "learning_rate": 7.64669005614709e-06, + "loss": 0.017, + "step": 35932 + }, + { + "epoch": 4.71, + "learning_rate": 7.646174628419243e-06, + "loss": 0.0222, + "step": 35933 + }, + { + "epoch": 4.71, + "learning_rate": 7.6456592073114e-06, + "loss": 0.0174, + "step": 35934 + }, + { + "epoch": 4.71, + "learning_rate": 7.645143792825014e-06, + "loss": 0.0185, + "step": 35935 + }, + { + "epoch": 4.71, + "learning_rate": 7.64462838496154e-06, + "loss": 0.0164, + "step": 35936 + }, + { + "epoch": 4.71, + "learning_rate": 7.644112983722421e-06, + "loss": 0.0204, + "step": 35937 + }, + { + "epoch": 4.71, + "learning_rate": 7.643597589109108e-06, + "loss": 0.0255, + "step": 35938 + }, + { + "epoch": 4.71, + "learning_rate": 7.643082201123056e-06, + "loss": 0.016, + "step": 35939 + }, + { + "epoch": 4.71, + "learning_rate": 7.642566819765704e-06, + "loss": 0.0231, + "step": 35940 + }, + { + "epoch": 4.71, + "learning_rate": 7.642051445038512e-06, + "loss": 0.0134, + "step": 35941 + }, + { + "epoch": 4.71, + "learning_rate": 7.641536076942921e-06, + "loss": 0.0164, + "step": 35942 + }, + { + "epoch": 4.71, + "learning_rate": 7.641020715480389e-06, + "loss": 0.0179, + "step": 35943 + }, + { + "epoch": 4.71, + "learning_rate": 7.640505360652356e-06, + "loss": 0.0179, + "step": 35944 + }, + { + "epoch": 4.71, + "learning_rate": 7.639990012460277e-06, + "loss": 0.0237, + "step": 35945 + }, + { + "epoch": 4.71, + "learning_rate": 7.639474670905598e-06, + "loss": 0.0168, + "step": 35946 + }, + { + "epoch": 4.71, + "learning_rate": 7.638959335989775e-06, + "loss": 0.0158, + "step": 35947 + }, + { + "epoch": 4.71, + "learning_rate": 7.638444007714251e-06, + "loss": 0.0181, + "step": 35948 + }, + { + "epoch": 4.71, + "learning_rate": 7.637928686080479e-06, + "loss": 0.0166, + "step": 35949 + }, + { + "epoch": 4.71, + "learning_rate": 7.637413371089906e-06, + "loss": 0.018, + "step": 35950 + }, + { + "epoch": 4.71, + "learning_rate": 7.636898062743983e-06, + "loss": 0.0137, + "step": 35951 + }, + { + "epoch": 4.71, + "learning_rate": 7.636382761044156e-06, + "loss": 0.0191, + "step": 35952 + }, + { + "epoch": 4.71, + "learning_rate": 7.635867465991878e-06, + "loss": 0.0204, + "step": 35953 + }, + { + "epoch": 4.71, + "learning_rate": 7.6353521775886e-06, + "loss": 0.0181, + "step": 35954 + }, + { + "epoch": 4.71, + "learning_rate": 7.634836895835764e-06, + "loss": 0.0194, + "step": 35955 + }, + { + "epoch": 4.71, + "learning_rate": 7.634321620734824e-06, + "loss": 0.0174, + "step": 35956 + }, + { + "epoch": 4.71, + "learning_rate": 7.633806352287226e-06, + "loss": 0.0167, + "step": 35957 + }, + { + "epoch": 4.71, + "learning_rate": 7.633291090494424e-06, + "loss": 0.018, + "step": 35958 + }, + { + "epoch": 4.71, + "learning_rate": 7.632775835357864e-06, + "loss": 0.0193, + "step": 35959 + }, + { + "epoch": 4.71, + "learning_rate": 7.632260586878999e-06, + "loss": 0.0157, + "step": 35960 + }, + { + "epoch": 4.71, + "learning_rate": 7.631745345059271e-06, + "loss": 0.0215, + "step": 35961 + }, + { + "epoch": 4.71, + "learning_rate": 7.631230109900139e-06, + "loss": 0.0168, + "step": 35962 + }, + { + "epoch": 4.71, + "learning_rate": 7.63071488140304e-06, + "loss": 0.0208, + "step": 35963 + }, + { + "epoch": 4.71, + "learning_rate": 7.630199659569432e-06, + "loss": 0.0184, + "step": 35964 + }, + { + "epoch": 4.71, + "learning_rate": 7.62968444440076e-06, + "loss": 0.0203, + "step": 35965 + }, + { + "epoch": 4.71, + "learning_rate": 7.629169235898478e-06, + "loss": 0.0203, + "step": 35966 + }, + { + "epoch": 4.71, + "learning_rate": 7.628654034064023e-06, + "loss": 0.0205, + "step": 35967 + }, + { + "epoch": 4.71, + "learning_rate": 7.62813883889886e-06, + "loss": 0.0203, + "step": 35968 + }, + { + "epoch": 4.71, + "learning_rate": 7.6276236504044265e-06, + "loss": 0.017, + "step": 35969 + }, + { + "epoch": 4.71, + "learning_rate": 7.627108468582178e-06, + "loss": 0.0194, + "step": 35970 + }, + { + "epoch": 4.71, + "learning_rate": 7.626593293433557e-06, + "loss": 0.0205, + "step": 35971 + }, + { + "epoch": 4.71, + "learning_rate": 7.62607812496002e-06, + "loss": 0.0148, + "step": 35972 + }, + { + "epoch": 4.71, + "learning_rate": 7.625562963163009e-06, + "loss": 0.0187, + "step": 35973 + }, + { + "epoch": 4.71, + "learning_rate": 7.625047808043976e-06, + "loss": 0.0199, + "step": 35974 + }, + { + "epoch": 4.71, + "learning_rate": 7.624532659604371e-06, + "loss": 0.019, + "step": 35975 + }, + { + "epoch": 4.71, + "learning_rate": 7.624017517845641e-06, + "loss": 0.0234, + "step": 35976 + }, + { + "epoch": 4.71, + "learning_rate": 7.623502382769236e-06, + "loss": 0.0189, + "step": 35977 + }, + { + "epoch": 4.71, + "learning_rate": 7.6229872543765996e-06, + "loss": 0.0182, + "step": 35978 + }, + { + "epoch": 4.71, + "learning_rate": 7.622472132669188e-06, + "loss": 0.0183, + "step": 35979 + }, + { + "epoch": 4.71, + "learning_rate": 7.6219570176484455e-06, + "loss": 0.0195, + "step": 35980 + }, + { + "epoch": 4.71, + "learning_rate": 7.621441909315825e-06, + "loss": 0.0152, + "step": 35981 + }, + { + "epoch": 4.71, + "learning_rate": 7.620926807672771e-06, + "loss": 0.0169, + "step": 35982 + }, + { + "epoch": 4.71, + "learning_rate": 7.6204117127207345e-06, + "loss": 0.0196, + "step": 35983 + }, + { + "epoch": 4.71, + "learning_rate": 7.619896624461163e-06, + "loss": 0.0197, + "step": 35984 + }, + { + "epoch": 4.71, + "learning_rate": 7.6193815428955065e-06, + "loss": 0.0143, + "step": 35985 + }, + { + "epoch": 4.71, + "learning_rate": 7.61886646802521e-06, + "loss": 0.0177, + "step": 35986 + }, + { + "epoch": 4.71, + "learning_rate": 7.618351399851727e-06, + "loss": 0.0161, + "step": 35987 + }, + { + "epoch": 4.71, + "learning_rate": 7.617836338376503e-06, + "loss": 0.022, + "step": 35988 + }, + { + "epoch": 4.71, + "learning_rate": 7.617321283600984e-06, + "loss": 0.0155, + "step": 35989 + }, + { + "epoch": 4.71, + "learning_rate": 7.6168062355266244e-06, + "loss": 0.0204, + "step": 35990 + }, + { + "epoch": 4.71, + "learning_rate": 7.6162911941548735e-06, + "loss": 0.0207, + "step": 35991 + }, + { + "epoch": 4.71, + "learning_rate": 7.615776159487174e-06, + "loss": 0.022, + "step": 35992 + }, + { + "epoch": 4.71, + "learning_rate": 7.6152611315249795e-06, + "loss": 0.0181, + "step": 35993 + }, + { + "epoch": 4.71, + "learning_rate": 7.614746110269733e-06, + "loss": 0.0174, + "step": 35994 + }, + { + "epoch": 4.72, + "learning_rate": 7.6142310957228856e-06, + "loss": 0.0182, + "step": 35995 + }, + { + "epoch": 4.72, + "learning_rate": 7.61371608788589e-06, + "loss": 0.0138, + "step": 35996 + }, + { + "epoch": 4.72, + "learning_rate": 7.613201086760188e-06, + "loss": 0.0207, + "step": 35997 + }, + { + "epoch": 4.72, + "learning_rate": 7.612686092347233e-06, + "loss": 0.0266, + "step": 35998 + }, + { + "epoch": 4.72, + "learning_rate": 7.612171104648469e-06, + "loss": 0.0231, + "step": 35999 + }, + { + "epoch": 4.72, + "learning_rate": 7.611656123665344e-06, + "loss": 0.0177, + "step": 36000 + }, + { + "epoch": 4.72, + "learning_rate": 7.611141149399313e-06, + "loss": 0.0206, + "step": 36001 + }, + { + "epoch": 4.72, + "learning_rate": 7.610626181851819e-06, + "loss": 0.0191, + "step": 36002 + }, + { + "epoch": 4.72, + "learning_rate": 7.6101112210243125e-06, + "loss": 0.0217, + "step": 36003 + }, + { + "epoch": 4.72, + "learning_rate": 7.609596266918242e-06, + "loss": 0.017, + "step": 36004 + }, + { + "epoch": 4.72, + "learning_rate": 7.609081319535052e-06, + "loss": 0.0197, + "step": 36005 + }, + { + "epoch": 4.72, + "learning_rate": 7.608566378876196e-06, + "loss": 0.0172, + "step": 36006 + }, + { + "epoch": 4.72, + "learning_rate": 7.608051444943118e-06, + "loss": 0.0233, + "step": 36007 + }, + { + "epoch": 4.72, + "learning_rate": 7.607536517737271e-06, + "loss": 0.0185, + "step": 36008 + }, + { + "epoch": 4.72, + "learning_rate": 7.6070215972600955e-06, + "loss": 0.0165, + "step": 36009 + }, + { + "epoch": 4.72, + "learning_rate": 7.606506683513047e-06, + "loss": 0.0196, + "step": 36010 + }, + { + "epoch": 4.72, + "learning_rate": 7.605991776497567e-06, + "loss": 0.0195, + "step": 36011 + }, + { + "epoch": 4.72, + "learning_rate": 7.605476876215112e-06, + "loss": 0.0186, + "step": 36012 + }, + { + "epoch": 4.72, + "learning_rate": 7.6049619826671235e-06, + "loss": 0.0197, + "step": 36013 + }, + { + "epoch": 4.72, + "learning_rate": 7.604447095855055e-06, + "loss": 0.0161, + "step": 36014 + }, + { + "epoch": 4.72, + "learning_rate": 7.60393221578035e-06, + "loss": 0.0148, + "step": 36015 + }, + { + "epoch": 4.72, + "learning_rate": 7.603417342444458e-06, + "loss": 0.0199, + "step": 36016 + }, + { + "epoch": 4.72, + "learning_rate": 7.602902475848826e-06, + "loss": 0.0194, + "step": 36017 + }, + { + "epoch": 4.72, + "learning_rate": 7.602387615994904e-06, + "loss": 0.0189, + "step": 36018 + }, + { + "epoch": 4.72, + "learning_rate": 7.601872762884141e-06, + "loss": 0.0209, + "step": 36019 + }, + { + "epoch": 4.72, + "learning_rate": 7.601357916517979e-06, + "loss": 0.0171, + "step": 36020 + }, + { + "epoch": 4.72, + "learning_rate": 7.6008430768978745e-06, + "loss": 0.0161, + "step": 36021 + }, + { + "epoch": 4.72, + "learning_rate": 7.600328244025266e-06, + "loss": 0.015, + "step": 36022 + }, + { + "epoch": 4.72, + "learning_rate": 7.599813417901609e-06, + "loss": 0.019, + "step": 36023 + }, + { + "epoch": 4.72, + "learning_rate": 7.599298598528348e-06, + "loss": 0.0219, + "step": 36024 + }, + { + "epoch": 4.72, + "learning_rate": 7.598783785906935e-06, + "loss": 0.0204, + "step": 36025 + }, + { + "epoch": 4.72, + "learning_rate": 7.598268980038812e-06, + "loss": 0.0162, + "step": 36026 + }, + { + "epoch": 4.72, + "learning_rate": 7.5977541809254315e-06, + "loss": 0.0171, + "step": 36027 + }, + { + "epoch": 4.72, + "learning_rate": 7.597239388568238e-06, + "loss": 0.022, + "step": 36028 + }, + { + "epoch": 4.72, + "learning_rate": 7.5967246029686835e-06, + "loss": 0.0189, + "step": 36029 + }, + { + "epoch": 4.72, + "learning_rate": 7.59620982412821e-06, + "loss": 0.0144, + "step": 36030 + }, + { + "epoch": 4.72, + "learning_rate": 7.595695052048271e-06, + "loss": 0.0224, + "step": 36031 + }, + { + "epoch": 4.72, + "learning_rate": 7.595180286730308e-06, + "loss": 0.0213, + "step": 36032 + }, + { + "epoch": 4.72, + "learning_rate": 7.594665528175772e-06, + "loss": 0.0216, + "step": 36033 + }, + { + "epoch": 4.72, + "learning_rate": 7.594150776386113e-06, + "loss": 0.0166, + "step": 36034 + }, + { + "epoch": 4.72, + "learning_rate": 7.593636031362779e-06, + "loss": 0.0201, + "step": 36035 + }, + { + "epoch": 4.72, + "learning_rate": 7.593121293107212e-06, + "loss": 0.0188, + "step": 36036 + }, + { + "epoch": 4.72, + "learning_rate": 7.592606561620867e-06, + "loss": 0.0203, + "step": 36037 + }, + { + "epoch": 4.72, + "learning_rate": 7.5920918369051845e-06, + "loss": 0.0162, + "step": 36038 + }, + { + "epoch": 4.72, + "learning_rate": 7.591577118961616e-06, + "loss": 0.0192, + "step": 36039 + }, + { + "epoch": 4.72, + "learning_rate": 7.5910624077916105e-06, + "loss": 0.0202, + "step": 36040 + }, + { + "epoch": 4.72, + "learning_rate": 7.590547703396612e-06, + "loss": 0.016, + "step": 36041 + }, + { + "epoch": 4.72, + "learning_rate": 7.590033005778071e-06, + "loss": 0.0185, + "step": 36042 + }, + { + "epoch": 4.72, + "learning_rate": 7.589518314937433e-06, + "loss": 0.0199, + "step": 36043 + }, + { + "epoch": 4.72, + "learning_rate": 7.589003630876144e-06, + "loss": 0.0164, + "step": 36044 + }, + { + "epoch": 4.72, + "learning_rate": 7.588488953595655e-06, + "loss": 0.0191, + "step": 36045 + }, + { + "epoch": 4.72, + "learning_rate": 7.587974283097415e-06, + "loss": 0.0151, + "step": 36046 + }, + { + "epoch": 4.72, + "learning_rate": 7.5874596193828665e-06, + "loss": 0.0185, + "step": 36047 + }, + { + "epoch": 4.72, + "learning_rate": 7.586944962453462e-06, + "loss": 0.0149, + "step": 36048 + }, + { + "epoch": 4.72, + "learning_rate": 7.586430312310644e-06, + "loss": 0.0194, + "step": 36049 + }, + { + "epoch": 4.72, + "learning_rate": 7.585915668955866e-06, + "loss": 0.0175, + "step": 36050 + }, + { + "epoch": 4.72, + "learning_rate": 7.585401032390567e-06, + "loss": 0.0202, + "step": 36051 + }, + { + "epoch": 4.72, + "learning_rate": 7.584886402616202e-06, + "loss": 0.0191, + "step": 36052 + }, + { + "epoch": 4.72, + "learning_rate": 7.584371779634213e-06, + "loss": 0.0134, + "step": 36053 + }, + { + "epoch": 4.72, + "learning_rate": 7.58385716344605e-06, + "loss": 0.0188, + "step": 36054 + }, + { + "epoch": 4.72, + "learning_rate": 7.583342554053158e-06, + "loss": 0.0192, + "step": 36055 + }, + { + "epoch": 4.72, + "learning_rate": 7.58282795145699e-06, + "loss": 0.0153, + "step": 36056 + }, + { + "epoch": 4.72, + "learning_rate": 7.582313355658989e-06, + "loss": 0.0193, + "step": 36057 + }, + { + "epoch": 4.72, + "learning_rate": 7.581798766660604e-06, + "loss": 0.0209, + "step": 36058 + }, + { + "epoch": 4.72, + "learning_rate": 7.581284184463279e-06, + "loss": 0.0248, + "step": 36059 + }, + { + "epoch": 4.72, + "learning_rate": 7.580769609068464e-06, + "loss": 0.0186, + "step": 36060 + }, + { + "epoch": 4.72, + "learning_rate": 7.580255040477607e-06, + "loss": 0.0166, + "step": 36061 + }, + { + "epoch": 4.72, + "learning_rate": 7.579740478692152e-06, + "loss": 0.0146, + "step": 36062 + }, + { + "epoch": 4.72, + "learning_rate": 7.57922592371355e-06, + "loss": 0.0207, + "step": 36063 + }, + { + "epoch": 4.72, + "learning_rate": 7.578711375543245e-06, + "loss": 0.0188, + "step": 36064 + }, + { + "epoch": 4.72, + "learning_rate": 7.578196834182687e-06, + "loss": 0.0185, + "step": 36065 + }, + { + "epoch": 4.72, + "learning_rate": 7.577682299633317e-06, + "loss": 0.0165, + "step": 36066 + }, + { + "epoch": 4.72, + "learning_rate": 7.577167771896589e-06, + "loss": 0.0193, + "step": 36067 + }, + { + "epoch": 4.72, + "learning_rate": 7.576653250973947e-06, + "loss": 0.0198, + "step": 36068 + }, + { + "epoch": 4.72, + "learning_rate": 7.57613873686684e-06, + "loss": 0.0205, + "step": 36069 + }, + { + "epoch": 4.72, + "learning_rate": 7.575624229576714e-06, + "loss": 0.0203, + "step": 36070 + }, + { + "epoch": 4.73, + "learning_rate": 7.575109729105016e-06, + "loss": 0.0223, + "step": 36071 + }, + { + "epoch": 4.73, + "learning_rate": 7.574595235453191e-06, + "loss": 0.0182, + "step": 36072 + }, + { + "epoch": 4.73, + "learning_rate": 7.57408074862269e-06, + "loss": 0.0186, + "step": 36073 + }, + { + "epoch": 4.73, + "learning_rate": 7.573566268614956e-06, + "loss": 0.0201, + "step": 36074 + }, + { + "epoch": 4.73, + "learning_rate": 7.573051795431436e-06, + "loss": 0.0199, + "step": 36075 + }, + { + "epoch": 4.73, + "learning_rate": 7.572537329073583e-06, + "loss": 0.0164, + "step": 36076 + }, + { + "epoch": 4.73, + "learning_rate": 7.572022869542834e-06, + "loss": 0.0169, + "step": 36077 + }, + { + "epoch": 4.73, + "learning_rate": 7.571508416840644e-06, + "loss": 0.0144, + "step": 36078 + }, + { + "epoch": 4.73, + "learning_rate": 7.5709939709684585e-06, + "loss": 0.0202, + "step": 36079 + }, + { + "epoch": 4.73, + "learning_rate": 7.5704795319277215e-06, + "loss": 0.0153, + "step": 36080 + }, + { + "epoch": 4.73, + "learning_rate": 7.569965099719884e-06, + "loss": 0.0175, + "step": 36081 + }, + { + "epoch": 4.73, + "learning_rate": 7.569450674346388e-06, + "loss": 0.0173, + "step": 36082 + }, + { + "epoch": 4.73, + "learning_rate": 7.568936255808682e-06, + "loss": 0.0163, + "step": 36083 + }, + { + "epoch": 4.73, + "learning_rate": 7.5684218441082155e-06, + "loss": 0.0219, + "step": 36084 + }, + { + "epoch": 4.73, + "learning_rate": 7.567907439246431e-06, + "loss": 0.0147, + "step": 36085 + }, + { + "epoch": 4.73, + "learning_rate": 7.56739304122478e-06, + "loss": 0.0158, + "step": 36086 + }, + { + "epoch": 4.73, + "learning_rate": 7.5668786500447045e-06, + "loss": 0.0164, + "step": 36087 + }, + { + "epoch": 4.73, + "learning_rate": 7.56636426570765e-06, + "loss": 0.0215, + "step": 36088 + }, + { + "epoch": 4.73, + "learning_rate": 7.565849888215069e-06, + "loss": 0.0204, + "step": 36089 + }, + { + "epoch": 4.73, + "learning_rate": 7.5653355175684084e-06, + "loss": 0.0183, + "step": 36090 + }, + { + "epoch": 4.73, + "learning_rate": 7.564821153769108e-06, + "loss": 0.0216, + "step": 36091 + }, + { + "epoch": 4.73, + "learning_rate": 7.564306796818624e-06, + "loss": 0.019, + "step": 36092 + }, + { + "epoch": 4.73, + "learning_rate": 7.563792446718392e-06, + "loss": 0.0217, + "step": 36093 + }, + { + "epoch": 4.73, + "learning_rate": 7.563278103469868e-06, + "loss": 0.0182, + "step": 36094 + }, + { + "epoch": 4.73, + "learning_rate": 7.562763767074492e-06, + "loss": 0.0133, + "step": 36095 + }, + { + "epoch": 4.73, + "learning_rate": 7.562249437533712e-06, + "loss": 0.0252, + "step": 36096 + }, + { + "epoch": 4.73, + "learning_rate": 7.561735114848978e-06, + "loss": 0.0224, + "step": 36097 + }, + { + "epoch": 4.73, + "learning_rate": 7.561220799021731e-06, + "loss": 0.0223, + "step": 36098 + }, + { + "epoch": 4.73, + "learning_rate": 7.560706490053422e-06, + "loss": 0.0178, + "step": 36099 + }, + { + "epoch": 4.73, + "learning_rate": 7.560192187945499e-06, + "loss": 0.0211, + "step": 36100 + }, + { + "epoch": 4.73, + "learning_rate": 7.559677892699403e-06, + "loss": 0.018, + "step": 36101 + }, + { + "epoch": 4.73, + "learning_rate": 7.5591636043165854e-06, + "loss": 0.0166, + "step": 36102 + }, + { + "epoch": 4.73, + "learning_rate": 7.558649322798487e-06, + "loss": 0.0164, + "step": 36103 + }, + { + "epoch": 4.73, + "learning_rate": 7.558135048146558e-06, + "loss": 0.0194, + "step": 36104 + }, + { + "epoch": 4.73, + "learning_rate": 7.557620780362247e-06, + "loss": 0.0182, + "step": 36105 + }, + { + "epoch": 4.73, + "learning_rate": 7.557106519446994e-06, + "loss": 0.0198, + "step": 36106 + }, + { + "epoch": 4.73, + "learning_rate": 7.556592265402252e-06, + "loss": 0.0141, + "step": 36107 + }, + { + "epoch": 4.73, + "learning_rate": 7.556078018229461e-06, + "loss": 0.0147, + "step": 36108 + }, + { + "epoch": 4.73, + "learning_rate": 7.55556377793007e-06, + "loss": 0.0147, + "step": 36109 + }, + { + "epoch": 4.73, + "learning_rate": 7.555049544505526e-06, + "loss": 0.0211, + "step": 36110 + }, + { + "epoch": 4.73, + "learning_rate": 7.554535317957278e-06, + "loss": 0.0187, + "step": 36111 + }, + { + "epoch": 4.73, + "learning_rate": 7.5540210982867665e-06, + "loss": 0.0191, + "step": 36112 + }, + { + "epoch": 4.73, + "learning_rate": 7.553506885495443e-06, + "loss": 0.0169, + "step": 36113 + }, + { + "epoch": 4.73, + "learning_rate": 7.552992679584749e-06, + "loss": 0.0195, + "step": 36114 + }, + { + "epoch": 4.73, + "learning_rate": 7.552478480556135e-06, + "loss": 0.0186, + "step": 36115 + }, + { + "epoch": 4.73, + "learning_rate": 7.551964288411042e-06, + "loss": 0.0211, + "step": 36116 + }, + { + "epoch": 4.73, + "learning_rate": 7.551450103150922e-06, + "loss": 0.0195, + "step": 36117 + }, + { + "epoch": 4.73, + "learning_rate": 7.5509359247772165e-06, + "loss": 0.0233, + "step": 36118 + }, + { + "epoch": 4.73, + "learning_rate": 7.550421753291373e-06, + "loss": 0.0165, + "step": 36119 + }, + { + "epoch": 4.73, + "learning_rate": 7.549907588694835e-06, + "loss": 0.0184, + "step": 36120 + }, + { + "epoch": 4.73, + "learning_rate": 7.549393430989056e-06, + "loss": 0.0206, + "step": 36121 + }, + { + "epoch": 4.73, + "learning_rate": 7.548879280175475e-06, + "loss": 0.0167, + "step": 36122 + }, + { + "epoch": 4.73, + "learning_rate": 7.548365136255543e-06, + "loss": 0.0206, + "step": 36123 + }, + { + "epoch": 4.73, + "learning_rate": 7.547850999230701e-06, + "loss": 0.0219, + "step": 36124 + }, + { + "epoch": 4.73, + "learning_rate": 7.547336869102398e-06, + "loss": 0.0172, + "step": 36125 + }, + { + "epoch": 4.73, + "learning_rate": 7.546822745872082e-06, + "loss": 0.0242, + "step": 36126 + }, + { + "epoch": 4.73, + "learning_rate": 7.546308629541194e-06, + "loss": 0.0172, + "step": 36127 + }, + { + "epoch": 4.73, + "learning_rate": 7.545794520111185e-06, + "loss": 0.017, + "step": 36128 + }, + { + "epoch": 4.73, + "learning_rate": 7.545280417583495e-06, + "loss": 0.0156, + "step": 36129 + }, + { + "epoch": 4.73, + "learning_rate": 7.544766321959574e-06, + "loss": 0.0212, + "step": 36130 + }, + { + "epoch": 4.73, + "learning_rate": 7.5442522332408654e-06, + "loss": 0.0181, + "step": 36131 + }, + { + "epoch": 4.73, + "learning_rate": 7.5437381514288175e-06, + "loss": 0.0122, + "step": 36132 + }, + { + "epoch": 4.73, + "learning_rate": 7.5432240765248755e-06, + "loss": 0.0186, + "step": 36133 + }, + { + "epoch": 4.73, + "learning_rate": 7.542710008530487e-06, + "loss": 0.0207, + "step": 36134 + }, + { + "epoch": 4.73, + "learning_rate": 7.542195947447093e-06, + "loss": 0.017, + "step": 36135 + }, + { + "epoch": 4.73, + "learning_rate": 7.541681893276145e-06, + "loss": 0.0177, + "step": 36136 + }, + { + "epoch": 4.73, + "learning_rate": 7.541167846019083e-06, + "loss": 0.0213, + "step": 36137 + }, + { + "epoch": 4.73, + "learning_rate": 7.540653805677358e-06, + "loss": 0.0189, + "step": 36138 + }, + { + "epoch": 4.73, + "learning_rate": 7.540139772252411e-06, + "loss": 0.0173, + "step": 36139 + }, + { + "epoch": 4.73, + "learning_rate": 7.53962574574569e-06, + "loss": 0.0201, + "step": 36140 + }, + { + "epoch": 4.73, + "learning_rate": 7.539111726158642e-06, + "loss": 0.0197, + "step": 36141 + }, + { + "epoch": 4.73, + "learning_rate": 7.538597713492708e-06, + "loss": 0.0159, + "step": 36142 + }, + { + "epoch": 4.73, + "learning_rate": 7.538083707749339e-06, + "loss": 0.0189, + "step": 36143 + }, + { + "epoch": 4.73, + "learning_rate": 7.537569708929979e-06, + "loss": 0.0155, + "step": 36144 + }, + { + "epoch": 4.73, + "learning_rate": 7.537055717036072e-06, + "loss": 0.0198, + "step": 36145 + }, + { + "epoch": 4.73, + "learning_rate": 7.5365417320690645e-06, + "loss": 0.0182, + "step": 36146 + }, + { + "epoch": 4.74, + "learning_rate": 7.536027754030405e-06, + "loss": 0.0165, + "step": 36147 + }, + { + "epoch": 4.74, + "learning_rate": 7.535513782921534e-06, + "loss": 0.0172, + "step": 36148 + }, + { + "epoch": 4.74, + "learning_rate": 7.5349998187439014e-06, + "loss": 0.0166, + "step": 36149 + }, + { + "epoch": 4.74, + "learning_rate": 7.534485861498948e-06, + "loss": 0.0201, + "step": 36150 + }, + { + "epoch": 4.74, + "learning_rate": 7.533971911188124e-06, + "loss": 0.018, + "step": 36151 + }, + { + "epoch": 4.74, + "learning_rate": 7.533457967812869e-06, + "loss": 0.0188, + "step": 36152 + }, + { + "epoch": 4.74, + "learning_rate": 7.532944031374633e-06, + "loss": 0.0179, + "step": 36153 + }, + { + "epoch": 4.74, + "learning_rate": 7.53243010187486e-06, + "loss": 0.0205, + "step": 36154 + }, + { + "epoch": 4.74, + "learning_rate": 7.531916179315e-06, + "loss": 0.0163, + "step": 36155 + }, + { + "epoch": 4.74, + "learning_rate": 7.531402263696491e-06, + "loss": 0.0151, + "step": 36156 + }, + { + "epoch": 4.74, + "learning_rate": 7.530888355020785e-06, + "loss": 0.0209, + "step": 36157 + }, + { + "epoch": 4.74, + "learning_rate": 7.53037445328932e-06, + "loss": 0.0188, + "step": 36158 + }, + { + "epoch": 4.74, + "learning_rate": 7.529860558503549e-06, + "loss": 0.0141, + "step": 36159 + }, + { + "epoch": 4.74, + "learning_rate": 7.529346670664911e-06, + "loss": 0.0139, + "step": 36160 + }, + { + "epoch": 4.74, + "learning_rate": 7.528832789774854e-06, + "loss": 0.0248, + "step": 36161 + }, + { + "epoch": 4.74, + "learning_rate": 7.528318915834824e-06, + "loss": 0.0155, + "step": 36162 + }, + { + "epoch": 4.74, + "learning_rate": 7.527805048846266e-06, + "loss": 0.0216, + "step": 36163 + }, + { + "epoch": 4.74, + "learning_rate": 7.527291188810621e-06, + "loss": 0.0224, + "step": 36164 + }, + { + "epoch": 4.74, + "learning_rate": 7.526777335729341e-06, + "loss": 0.0194, + "step": 36165 + }, + { + "epoch": 4.74, + "learning_rate": 7.526263489603868e-06, + "loss": 0.0229, + "step": 36166 + }, + { + "epoch": 4.74, + "learning_rate": 7.525749650435649e-06, + "loss": 0.0193, + "step": 36167 + }, + { + "epoch": 4.74, + "learning_rate": 7.5252358182261244e-06, + "loss": 0.0202, + "step": 36168 + }, + { + "epoch": 4.74, + "learning_rate": 7.524721992976742e-06, + "loss": 0.0186, + "step": 36169 + }, + { + "epoch": 4.74, + "learning_rate": 7.52420817468895e-06, + "loss": 0.0182, + "step": 36170 + }, + { + "epoch": 4.74, + "learning_rate": 7.523694363364189e-06, + "loss": 0.0213, + "step": 36171 + }, + { + "epoch": 4.74, + "learning_rate": 7.523180559003906e-06, + "loss": 0.0203, + "step": 36172 + }, + { + "epoch": 4.74, + "learning_rate": 7.522666761609545e-06, + "loss": 0.0146, + "step": 36173 + }, + { + "epoch": 4.74, + "learning_rate": 7.522152971182555e-06, + "loss": 0.0178, + "step": 36174 + }, + { + "epoch": 4.74, + "learning_rate": 7.52163918772437e-06, + "loss": 0.0221, + "step": 36175 + }, + { + "epoch": 4.74, + "learning_rate": 7.52112541123645e-06, + "loss": 0.02, + "step": 36176 + }, + { + "epoch": 4.74, + "learning_rate": 7.520611641720229e-06, + "loss": 0.0183, + "step": 36177 + }, + { + "epoch": 4.74, + "learning_rate": 7.520097879177159e-06, + "loss": 0.0184, + "step": 36178 + }, + { + "epoch": 4.74, + "learning_rate": 7.51958412360868e-06, + "loss": 0.0153, + "step": 36179 + }, + { + "epoch": 4.74, + "learning_rate": 7.51907037501624e-06, + "loss": 0.0213, + "step": 36180 + }, + { + "epoch": 4.74, + "learning_rate": 7.51855663340128e-06, + "loss": 0.0168, + "step": 36181 + }, + { + "epoch": 4.74, + "learning_rate": 7.51804289876525e-06, + "loss": 0.016, + "step": 36182 + }, + { + "epoch": 4.74, + "learning_rate": 7.51752917110959e-06, + "loss": 0.0204, + "step": 36183 + }, + { + "epoch": 4.74, + "learning_rate": 7.517015450435747e-06, + "loss": 0.0202, + "step": 36184 + }, + { + "epoch": 4.74, + "learning_rate": 7.516501736745167e-06, + "loss": 0.0165, + "step": 36185 + }, + { + "epoch": 4.74, + "learning_rate": 7.515988030039292e-06, + "loss": 0.0214, + "step": 36186 + }, + { + "epoch": 4.74, + "learning_rate": 7.515474330319568e-06, + "loss": 0.0236, + "step": 36187 + }, + { + "epoch": 4.74, + "learning_rate": 7.514960637587443e-06, + "loss": 0.0202, + "step": 36188 + }, + { + "epoch": 4.74, + "learning_rate": 7.5144469518443565e-06, + "loss": 0.0258, + "step": 36189 + }, + { + "epoch": 4.74, + "learning_rate": 7.513933273091756e-06, + "loss": 0.0148, + "step": 36190 + }, + { + "epoch": 4.74, + "learning_rate": 7.513419601331088e-06, + "loss": 0.0219, + "step": 36191 + }, + { + "epoch": 4.74, + "learning_rate": 7.512905936563793e-06, + "loss": 0.0212, + "step": 36192 + }, + { + "epoch": 4.74, + "learning_rate": 7.512392278791318e-06, + "loss": 0.02, + "step": 36193 + }, + { + "epoch": 4.74, + "learning_rate": 7.511878628015108e-06, + "loss": 0.0153, + "step": 36194 + }, + { + "epoch": 4.74, + "learning_rate": 7.511364984236607e-06, + "loss": 0.0227, + "step": 36195 + }, + { + "epoch": 4.74, + "learning_rate": 7.510851347457257e-06, + "loss": 0.0153, + "step": 36196 + }, + { + "epoch": 4.74, + "learning_rate": 7.510337717678504e-06, + "loss": 0.0186, + "step": 36197 + }, + { + "epoch": 4.74, + "learning_rate": 7.509824094901795e-06, + "loss": 0.0133, + "step": 36198 + }, + { + "epoch": 4.74, + "learning_rate": 7.509310479128577e-06, + "loss": 0.0137, + "step": 36199 + }, + { + "epoch": 4.74, + "learning_rate": 7.508796870360286e-06, + "loss": 0.0189, + "step": 36200 + }, + { + "epoch": 4.74, + "learning_rate": 7.508283268598375e-06, + "loss": 0.0186, + "step": 36201 + }, + { + "epoch": 4.74, + "learning_rate": 7.507769673844282e-06, + "loss": 0.0222, + "step": 36202 + }, + { + "epoch": 4.74, + "learning_rate": 7.5072560860994555e-06, + "loss": 0.0211, + "step": 36203 + }, + { + "epoch": 4.74, + "learning_rate": 7.506742505365337e-06, + "loss": 0.0161, + "step": 36204 + }, + { + "epoch": 4.74, + "learning_rate": 7.5062289316433734e-06, + "loss": 0.0189, + "step": 36205 + }, + { + "epoch": 4.74, + "learning_rate": 7.5057153649350086e-06, + "loss": 0.0198, + "step": 36206 + }, + { + "epoch": 4.74, + "learning_rate": 7.505201805241685e-06, + "loss": 0.0185, + "step": 36207 + }, + { + "epoch": 4.74, + "learning_rate": 7.504688252564847e-06, + "loss": 0.015, + "step": 36208 + }, + { + "epoch": 4.74, + "learning_rate": 7.504174706905944e-06, + "loss": 0.0155, + "step": 36209 + }, + { + "epoch": 4.74, + "learning_rate": 7.503661168266414e-06, + "loss": 0.0195, + "step": 36210 + }, + { + "epoch": 4.74, + "learning_rate": 7.5031476366477056e-06, + "loss": 0.0192, + "step": 36211 + }, + { + "epoch": 4.74, + "learning_rate": 7.5026341120512625e-06, + "loss": 0.0219, + "step": 36212 + }, + { + "epoch": 4.74, + "learning_rate": 7.502120594478526e-06, + "loss": 0.0205, + "step": 36213 + }, + { + "epoch": 4.74, + "learning_rate": 7.501607083930944e-06, + "loss": 0.0197, + "step": 36214 + }, + { + "epoch": 4.74, + "learning_rate": 7.501093580409957e-06, + "loss": 0.0166, + "step": 36215 + }, + { + "epoch": 4.74, + "learning_rate": 7.500580083917014e-06, + "loss": 0.0194, + "step": 36216 + }, + { + "epoch": 4.74, + "learning_rate": 7.500066594453554e-06, + "loss": 0.0179, + "step": 36217 + }, + { + "epoch": 4.74, + "learning_rate": 7.499553112021026e-06, + "loss": 0.0233, + "step": 36218 + }, + { + "epoch": 4.74, + "learning_rate": 7.499039636620865e-06, + "loss": 0.0209, + "step": 36219 + }, + { + "epoch": 4.74, + "learning_rate": 7.498526168254529e-06, + "loss": 0.0197, + "step": 36220 + }, + { + "epoch": 4.74, + "learning_rate": 7.4980127069234524e-06, + "loss": 0.0152, + "step": 36221 + }, + { + "epoch": 4.74, + "learning_rate": 7.497499252629084e-06, + "loss": 0.0193, + "step": 36222 + }, + { + "epoch": 4.74, + "learning_rate": 7.496985805372863e-06, + "loss": 0.0195, + "step": 36223 + }, + { + "epoch": 4.75, + "learning_rate": 7.4964723651562375e-06, + "loss": 0.0164, + "step": 36224 + }, + { + "epoch": 4.75, + "learning_rate": 7.495958931980649e-06, + "loss": 0.0193, + "step": 36225 + }, + { + "epoch": 4.75, + "learning_rate": 7.4954455058475425e-06, + "loss": 0.0278, + "step": 36226 + }, + { + "epoch": 4.75, + "learning_rate": 7.494932086758364e-06, + "loss": 0.019, + "step": 36227 + }, + { + "epoch": 4.75, + "learning_rate": 7.494418674714554e-06, + "loss": 0.015, + "step": 36228 + }, + { + "epoch": 4.75, + "learning_rate": 7.493905269717556e-06, + "loss": 0.0228, + "step": 36229 + }, + { + "epoch": 4.75, + "learning_rate": 7.4933918717688204e-06, + "loss": 0.018, + "step": 36230 + }, + { + "epoch": 4.75, + "learning_rate": 7.492878480869784e-06, + "loss": 0.0171, + "step": 36231 + }, + { + "epoch": 4.75, + "learning_rate": 7.492365097021896e-06, + "loss": 0.0219, + "step": 36232 + }, + { + "epoch": 4.75, + "learning_rate": 7.491851720226595e-06, + "loss": 0.0196, + "step": 36233 + }, + { + "epoch": 4.75, + "learning_rate": 7.491338350485328e-06, + "loss": 0.0237, + "step": 36234 + }, + { + "epoch": 4.75, + "learning_rate": 7.490824987799539e-06, + "loss": 0.0153, + "step": 36235 + }, + { + "epoch": 4.75, + "learning_rate": 7.49031163217067e-06, + "loss": 0.0203, + "step": 36236 + }, + { + "epoch": 4.75, + "learning_rate": 7.489798283600169e-06, + "loss": 0.016, + "step": 36237 + }, + { + "epoch": 4.75, + "learning_rate": 7.489284942089473e-06, + "loss": 0.0194, + "step": 36238 + }, + { + "epoch": 4.75, + "learning_rate": 7.488771607640033e-06, + "loss": 0.0197, + "step": 36239 + }, + { + "epoch": 4.75, + "learning_rate": 7.488258280253281e-06, + "loss": 0.0196, + "step": 36240 + }, + { + "epoch": 4.75, + "learning_rate": 7.487744959930676e-06, + "loss": 0.0172, + "step": 36241 + }, + { + "epoch": 4.75, + "learning_rate": 7.487231646673653e-06, + "loss": 0.0161, + "step": 36242 + }, + { + "epoch": 4.75, + "learning_rate": 7.486718340483658e-06, + "loss": 0.0202, + "step": 36243 + }, + { + "epoch": 4.75, + "learning_rate": 7.486205041362133e-06, + "loss": 0.0194, + "step": 36244 + }, + { + "epoch": 4.75, + "learning_rate": 7.485691749310524e-06, + "loss": 0.017, + "step": 36245 + }, + { + "epoch": 4.75, + "learning_rate": 7.4851784643302714e-06, + "loss": 0.0179, + "step": 36246 + }, + { + "epoch": 4.75, + "learning_rate": 7.48466518642282e-06, + "loss": 0.0189, + "step": 36247 + }, + { + "epoch": 4.75, + "learning_rate": 7.484151915589615e-06, + "loss": 0.0195, + "step": 36248 + }, + { + "epoch": 4.75, + "learning_rate": 7.483638651832098e-06, + "loss": 0.0235, + "step": 36249 + }, + { + "epoch": 4.75, + "learning_rate": 7.483125395151716e-06, + "loss": 0.0186, + "step": 36250 + }, + { + "epoch": 4.75, + "learning_rate": 7.482612145549904e-06, + "loss": 0.0213, + "step": 36251 + }, + { + "epoch": 4.75, + "learning_rate": 7.4820989030281146e-06, + "loss": 0.0217, + "step": 36252 + }, + { + "epoch": 4.75, + "learning_rate": 7.481585667587789e-06, + "loss": 0.0189, + "step": 36253 + }, + { + "epoch": 4.75, + "learning_rate": 7.481072439230368e-06, + "loss": 0.0208, + "step": 36254 + }, + { + "epoch": 4.75, + "learning_rate": 7.480559217957295e-06, + "loss": 0.0135, + "step": 36255 + }, + { + "epoch": 4.75, + "learning_rate": 7.4800460037700195e-06, + "loss": 0.0157, + "step": 36256 + }, + { + "epoch": 4.75, + "learning_rate": 7.479532796669977e-06, + "loss": 0.0202, + "step": 36257 + }, + { + "epoch": 4.75, + "learning_rate": 7.479019596658617e-06, + "loss": 0.0133, + "step": 36258 + }, + { + "epoch": 4.75, + "learning_rate": 7.478506403737378e-06, + "loss": 0.0206, + "step": 36259 + }, + { + "epoch": 4.75, + "learning_rate": 7.477993217907707e-06, + "loss": 0.0186, + "step": 36260 + }, + { + "epoch": 4.75, + "learning_rate": 7.477480039171046e-06, + "loss": 0.0161, + "step": 36261 + }, + { + "epoch": 4.75, + "learning_rate": 7.476966867528832e-06, + "loss": 0.0195, + "step": 36262 + }, + { + "epoch": 4.75, + "learning_rate": 7.47645370298252e-06, + "loss": 0.0214, + "step": 36263 + }, + { + "epoch": 4.75, + "learning_rate": 7.475940545533548e-06, + "loss": 0.0197, + "step": 36264 + }, + { + "epoch": 4.75, + "learning_rate": 7.475427395183356e-06, + "loss": 0.0161, + "step": 36265 + }, + { + "epoch": 4.75, + "learning_rate": 7.474914251933394e-06, + "loss": 0.0269, + "step": 36266 + }, + { + "epoch": 4.75, + "learning_rate": 7.474401115785097e-06, + "loss": 0.021, + "step": 36267 + }, + { + "epoch": 4.75, + "learning_rate": 7.473887986739916e-06, + "loss": 0.0199, + "step": 36268 + }, + { + "epoch": 4.75, + "learning_rate": 7.473374864799288e-06, + "loss": 0.0194, + "step": 36269 + }, + { + "epoch": 4.75, + "learning_rate": 7.47286174996466e-06, + "loss": 0.0201, + "step": 36270 + }, + { + "epoch": 4.75, + "learning_rate": 7.472348642237472e-06, + "loss": 0.0163, + "step": 36271 + }, + { + "epoch": 4.75, + "learning_rate": 7.47183554161917e-06, + "loss": 0.0191, + "step": 36272 + }, + { + "epoch": 4.75, + "learning_rate": 7.4713224481111935e-06, + "loss": 0.0226, + "step": 36273 + }, + { + "epoch": 4.75, + "learning_rate": 7.4708093617149925e-06, + "loss": 0.0185, + "step": 36274 + }, + { + "epoch": 4.75, + "learning_rate": 7.470296282432001e-06, + "loss": 0.0192, + "step": 36275 + }, + { + "epoch": 4.75, + "learning_rate": 7.469783210263668e-06, + "loss": 0.0156, + "step": 36276 + }, + { + "epoch": 4.75, + "learning_rate": 7.469270145211438e-06, + "loss": 0.0179, + "step": 36277 + }, + { + "epoch": 4.75, + "learning_rate": 7.468757087276748e-06, + "loss": 0.0151, + "step": 36278 + }, + { + "epoch": 4.75, + "learning_rate": 7.468244036461045e-06, + "loss": 0.0185, + "step": 36279 + }, + { + "epoch": 4.75, + "learning_rate": 7.467730992765771e-06, + "loss": 0.0194, + "step": 36280 + }, + { + "epoch": 4.75, + "learning_rate": 7.467217956192369e-06, + "loss": 0.0216, + "step": 36281 + }, + { + "epoch": 4.75, + "learning_rate": 7.46670492674228e-06, + "loss": 0.017, + "step": 36282 + }, + { + "epoch": 4.75, + "learning_rate": 7.46619190441695e-06, + "loss": 0.0174, + "step": 36283 + }, + { + "epoch": 4.75, + "learning_rate": 7.465678889217816e-06, + "loss": 0.0177, + "step": 36284 + }, + { + "epoch": 4.75, + "learning_rate": 7.46516588114633e-06, + "loss": 0.0232, + "step": 36285 + }, + { + "epoch": 4.75, + "learning_rate": 7.4646528802039275e-06, + "loss": 0.0204, + "step": 36286 + }, + { + "epoch": 4.75, + "learning_rate": 7.464139886392056e-06, + "loss": 0.0201, + "step": 36287 + }, + { + "epoch": 4.75, + "learning_rate": 7.463626899712155e-06, + "loss": 0.0188, + "step": 36288 + }, + { + "epoch": 4.75, + "learning_rate": 7.46311392016567e-06, + "loss": 0.0219, + "step": 36289 + }, + { + "epoch": 4.75, + "learning_rate": 7.462600947754039e-06, + "loss": 0.0215, + "step": 36290 + }, + { + "epoch": 4.75, + "learning_rate": 7.462087982478707e-06, + "loss": 0.0166, + "step": 36291 + }, + { + "epoch": 4.75, + "learning_rate": 7.461575024341121e-06, + "loss": 0.018, + "step": 36292 + }, + { + "epoch": 4.75, + "learning_rate": 7.461062073342718e-06, + "loss": 0.0168, + "step": 36293 + }, + { + "epoch": 4.75, + "learning_rate": 7.460549129484946e-06, + "loss": 0.0179, + "step": 36294 + }, + { + "epoch": 4.75, + "learning_rate": 7.460036192769238e-06, + "loss": 0.0194, + "step": 36295 + }, + { + "epoch": 4.75, + "learning_rate": 7.459523263197046e-06, + "loss": 0.0235, + "step": 36296 + }, + { + "epoch": 4.75, + "learning_rate": 7.459010340769809e-06, + "loss": 0.013, + "step": 36297 + }, + { + "epoch": 4.75, + "learning_rate": 7.458497425488972e-06, + "loss": 0.0202, + "step": 36298 + }, + { + "epoch": 4.75, + "learning_rate": 7.457984517355974e-06, + "loss": 0.0206, + "step": 36299 + }, + { + "epoch": 4.76, + "learning_rate": 7.457471616372261e-06, + "loss": 0.0161, + "step": 36300 + }, + { + "epoch": 4.76, + "learning_rate": 7.456958722539273e-06, + "loss": 0.0203, + "step": 36301 + }, + { + "epoch": 4.76, + "learning_rate": 7.456445835858453e-06, + "loss": 0.0157, + "step": 36302 + }, + { + "epoch": 4.76, + "learning_rate": 7.455932956331243e-06, + "loss": 0.0176, + "step": 36303 + }, + { + "epoch": 4.76, + "learning_rate": 7.4554200839590885e-06, + "loss": 0.0175, + "step": 36304 + }, + { + "epoch": 4.76, + "learning_rate": 7.454907218743426e-06, + "loss": 0.0173, + "step": 36305 + }, + { + "epoch": 4.76, + "learning_rate": 7.4543943606857e-06, + "loss": 0.0186, + "step": 36306 + }, + { + "epoch": 4.76, + "learning_rate": 7.453881509787357e-06, + "loss": 0.0204, + "step": 36307 + }, + { + "epoch": 4.76, + "learning_rate": 7.453368666049839e-06, + "loss": 0.0187, + "step": 36308 + }, + { + "epoch": 4.76, + "learning_rate": 7.452855829474582e-06, + "loss": 0.0173, + "step": 36309 + }, + { + "epoch": 4.76, + "learning_rate": 7.452343000063036e-06, + "loss": 0.0183, + "step": 36310 + }, + { + "epoch": 4.76, + "learning_rate": 7.451830177816637e-06, + "loss": 0.0196, + "step": 36311 + }, + { + "epoch": 4.76, + "learning_rate": 7.45131736273683e-06, + "loss": 0.0181, + "step": 36312 + }, + { + "epoch": 4.76, + "learning_rate": 7.450804554825059e-06, + "loss": 0.018, + "step": 36313 + }, + { + "epoch": 4.76, + "learning_rate": 7.450291754082764e-06, + "loss": 0.019, + "step": 36314 + }, + { + "epoch": 4.76, + "learning_rate": 7.449778960511387e-06, + "loss": 0.0161, + "step": 36315 + }, + { + "epoch": 4.76, + "learning_rate": 7.44926617411237e-06, + "loss": 0.0194, + "step": 36316 + }, + { + "epoch": 4.76, + "learning_rate": 7.448753394887155e-06, + "loss": 0.0191, + "step": 36317 + }, + { + "epoch": 4.76, + "learning_rate": 7.448240622837188e-06, + "loss": 0.0204, + "step": 36318 + }, + { + "epoch": 4.76, + "learning_rate": 7.447727857963907e-06, + "loss": 0.019, + "step": 36319 + }, + { + "epoch": 4.76, + "learning_rate": 7.447215100268756e-06, + "loss": 0.0224, + "step": 36320 + }, + { + "epoch": 4.76, + "learning_rate": 7.446702349753178e-06, + "loss": 0.023, + "step": 36321 + }, + { + "epoch": 4.76, + "learning_rate": 7.446189606418611e-06, + "loss": 0.0163, + "step": 36322 + }, + { + "epoch": 4.76, + "learning_rate": 7.445676870266503e-06, + "loss": 0.0144, + "step": 36323 + }, + { + "epoch": 4.76, + "learning_rate": 7.445164141298289e-06, + "loss": 0.0197, + "step": 36324 + }, + { + "epoch": 4.76, + "learning_rate": 7.444651419515419e-06, + "loss": 0.016, + "step": 36325 + }, + { + "epoch": 4.76, + "learning_rate": 7.444138704919328e-06, + "loss": 0.0149, + "step": 36326 + }, + { + "epoch": 4.76, + "learning_rate": 7.44362599751146e-06, + "loss": 0.0177, + "step": 36327 + }, + { + "epoch": 4.76, + "learning_rate": 7.4431132972932566e-06, + "loss": 0.0181, + "step": 36328 + }, + { + "epoch": 4.76, + "learning_rate": 7.442600604266164e-06, + "loss": 0.0164, + "step": 36329 + }, + { + "epoch": 4.76, + "learning_rate": 7.44208791843162e-06, + "loss": 0.0179, + "step": 36330 + }, + { + "epoch": 4.76, + "learning_rate": 7.4415752397910675e-06, + "loss": 0.0145, + "step": 36331 + }, + { + "epoch": 4.76, + "learning_rate": 7.441062568345949e-06, + "loss": 0.015, + "step": 36332 + }, + { + "epoch": 4.76, + "learning_rate": 7.440549904097707e-06, + "loss": 0.0165, + "step": 36333 + }, + { + "epoch": 4.76, + "learning_rate": 7.44003724704778e-06, + "loss": 0.0169, + "step": 36334 + }, + { + "epoch": 4.76, + "learning_rate": 7.439524597197612e-06, + "loss": 0.0165, + "step": 36335 + }, + { + "epoch": 4.76, + "learning_rate": 7.439011954548645e-06, + "loss": 0.0161, + "step": 36336 + }, + { + "epoch": 4.76, + "learning_rate": 7.438499319102321e-06, + "loss": 0.0149, + "step": 36337 + }, + { + "epoch": 4.76, + "learning_rate": 7.437986690860082e-06, + "loss": 0.0166, + "step": 36338 + }, + { + "epoch": 4.76, + "learning_rate": 7.437474069823365e-06, + "loss": 0.012, + "step": 36339 + }, + { + "epoch": 4.76, + "learning_rate": 7.436961455993618e-06, + "loss": 0.0205, + "step": 36340 + }, + { + "epoch": 4.76, + "learning_rate": 7.43644884937228e-06, + "loss": 0.0145, + "step": 36341 + }, + { + "epoch": 4.76, + "learning_rate": 7.435936249960795e-06, + "loss": 0.0192, + "step": 36342 + }, + { + "epoch": 4.76, + "learning_rate": 7.435423657760601e-06, + "loss": 0.0175, + "step": 36343 + }, + { + "epoch": 4.76, + "learning_rate": 7.434911072773145e-06, + "loss": 0.0203, + "step": 36344 + }, + { + "epoch": 4.76, + "learning_rate": 7.434398494999862e-06, + "loss": 0.0219, + "step": 36345 + }, + { + "epoch": 4.76, + "learning_rate": 7.433885924442197e-06, + "loss": 0.0215, + "step": 36346 + }, + { + "epoch": 4.76, + "learning_rate": 7.43337336110159e-06, + "loss": 0.0238, + "step": 36347 + }, + { + "epoch": 4.76, + "learning_rate": 7.432860804979487e-06, + "loss": 0.0173, + "step": 36348 + }, + { + "epoch": 4.76, + "learning_rate": 7.432348256077323e-06, + "loss": 0.0183, + "step": 36349 + }, + { + "epoch": 4.76, + "learning_rate": 7.4318357143965405e-06, + "loss": 0.0167, + "step": 36350 + }, + { + "epoch": 4.76, + "learning_rate": 7.431323179938587e-06, + "loss": 0.0188, + "step": 36351 + }, + { + "epoch": 4.76, + "learning_rate": 7.430810652704901e-06, + "loss": 0.0186, + "step": 36352 + }, + { + "epoch": 4.76, + "learning_rate": 7.43029813269692e-06, + "loss": 0.0176, + "step": 36353 + }, + { + "epoch": 4.76, + "learning_rate": 7.429785619916093e-06, + "loss": 0.0144, + "step": 36354 + }, + { + "epoch": 4.76, + "learning_rate": 7.4292731143638555e-06, + "loss": 0.02, + "step": 36355 + }, + { + "epoch": 4.76, + "learning_rate": 7.428760616041648e-06, + "loss": 0.021, + "step": 36356 + }, + { + "epoch": 4.76, + "learning_rate": 7.428248124950917e-06, + "loss": 0.0162, + "step": 36357 + }, + { + "epoch": 4.76, + "learning_rate": 7.4277356410931e-06, + "loss": 0.0224, + "step": 36358 + }, + { + "epoch": 4.76, + "learning_rate": 7.427223164469641e-06, + "loss": 0.0172, + "step": 36359 + }, + { + "epoch": 4.76, + "learning_rate": 7.426710695081977e-06, + "loss": 0.0192, + "step": 36360 + }, + { + "epoch": 4.76, + "learning_rate": 7.426198232931552e-06, + "loss": 0.02, + "step": 36361 + }, + { + "epoch": 4.76, + "learning_rate": 7.4256857780198085e-06, + "loss": 0.0167, + "step": 36362 + }, + { + "epoch": 4.76, + "learning_rate": 7.42517333034819e-06, + "loss": 0.0186, + "step": 36363 + }, + { + "epoch": 4.76, + "learning_rate": 7.4246608899181305e-06, + "loss": 0.0195, + "step": 36364 + }, + { + "epoch": 4.76, + "learning_rate": 7.424148456731077e-06, + "loss": 0.0183, + "step": 36365 + }, + { + "epoch": 4.76, + "learning_rate": 7.423636030788468e-06, + "loss": 0.0156, + "step": 36366 + }, + { + "epoch": 4.76, + "learning_rate": 7.423123612091747e-06, + "loss": 0.0157, + "step": 36367 + }, + { + "epoch": 4.76, + "learning_rate": 7.422611200642352e-06, + "loss": 0.0211, + "step": 36368 + }, + { + "epoch": 4.76, + "learning_rate": 7.422098796441728e-06, + "loss": 0.0199, + "step": 36369 + }, + { + "epoch": 4.76, + "learning_rate": 7.421586399491311e-06, + "loss": 0.0187, + "step": 36370 + }, + { + "epoch": 4.76, + "learning_rate": 7.4210740097925425e-06, + "loss": 0.0213, + "step": 36371 + }, + { + "epoch": 4.76, + "learning_rate": 7.42056162734687e-06, + "loss": 0.0206, + "step": 36372 + }, + { + "epoch": 4.76, + "learning_rate": 7.42004925215573e-06, + "loss": 0.0184, + "step": 36373 + }, + { + "epoch": 4.76, + "learning_rate": 7.419536884220564e-06, + "loss": 0.0146, + "step": 36374 + }, + { + "epoch": 4.76, + "learning_rate": 7.419024523542814e-06, + "loss": 0.0216, + "step": 36375 + }, + { + "epoch": 4.76, + "learning_rate": 7.4185121701239185e-06, + "loss": 0.0195, + "step": 36376 + }, + { + "epoch": 4.77, + "learning_rate": 7.417999823965321e-06, + "loss": 0.0197, + "step": 36377 + }, + { + "epoch": 4.77, + "learning_rate": 7.417487485068462e-06, + "loss": 0.0198, + "step": 36378 + }, + { + "epoch": 4.77, + "learning_rate": 7.41697515343478e-06, + "loss": 0.0208, + "step": 36379 + }, + { + "epoch": 4.77, + "learning_rate": 7.41646282906572e-06, + "loss": 0.0175, + "step": 36380 + }, + { + "epoch": 4.77, + "learning_rate": 7.415950511962719e-06, + "loss": 0.0181, + "step": 36381 + }, + { + "epoch": 4.77, + "learning_rate": 7.415438202127217e-06, + "loss": 0.0222, + "step": 36382 + }, + { + "epoch": 4.77, + "learning_rate": 7.414925899560663e-06, + "loss": 0.0149, + "step": 36383 + }, + { + "epoch": 4.77, + "learning_rate": 7.414413604264489e-06, + "loss": 0.019, + "step": 36384 + }, + { + "epoch": 4.77, + "learning_rate": 7.413901316240139e-06, + "loss": 0.0196, + "step": 36385 + }, + { + "epoch": 4.77, + "learning_rate": 7.413389035489056e-06, + "loss": 0.0189, + "step": 36386 + }, + { + "epoch": 4.77, + "learning_rate": 7.4128767620126775e-06, + "loss": 0.0229, + "step": 36387 + }, + { + "epoch": 4.77, + "learning_rate": 7.412364495812446e-06, + "loss": 0.0178, + "step": 36388 + }, + { + "epoch": 4.77, + "learning_rate": 7.411852236889801e-06, + "loss": 0.0179, + "step": 36389 + }, + { + "epoch": 4.77, + "learning_rate": 7.411339985246184e-06, + "loss": 0.015, + "step": 36390 + }, + { + "epoch": 4.77, + "learning_rate": 7.410827740883036e-06, + "loss": 0.0199, + "step": 36391 + }, + { + "epoch": 4.77, + "learning_rate": 7.410315503801794e-06, + "loss": 0.0175, + "step": 36392 + }, + { + "epoch": 4.77, + "learning_rate": 7.409803274003902e-06, + "loss": 0.0161, + "step": 36393 + }, + { + "epoch": 4.77, + "learning_rate": 7.4092910514908024e-06, + "loss": 0.0189, + "step": 36394 + }, + { + "epoch": 4.77, + "learning_rate": 7.408778836263932e-06, + "loss": 0.0206, + "step": 36395 + }, + { + "epoch": 4.77, + "learning_rate": 7.408266628324736e-06, + "loss": 0.0156, + "step": 36396 + }, + { + "epoch": 4.77, + "learning_rate": 7.407754427674652e-06, + "loss": 0.0208, + "step": 36397 + }, + { + "epoch": 4.77, + "learning_rate": 7.40724223431512e-06, + "loss": 0.017, + "step": 36398 + }, + { + "epoch": 4.77, + "learning_rate": 7.406730048247579e-06, + "loss": 0.017, + "step": 36399 + }, + { + "epoch": 4.77, + "learning_rate": 7.4062178694734735e-06, + "loss": 0.0181, + "step": 36400 + }, + { + "epoch": 4.77, + "learning_rate": 7.405705697994243e-06, + "loss": 0.0205, + "step": 36401 + }, + { + "epoch": 4.77, + "learning_rate": 7.4051935338113246e-06, + "loss": 0.0182, + "step": 36402 + }, + { + "epoch": 4.77, + "learning_rate": 7.404681376926164e-06, + "loss": 0.0168, + "step": 36403 + }, + { + "epoch": 4.77, + "learning_rate": 7.404169227340194e-06, + "loss": 0.0165, + "step": 36404 + }, + { + "epoch": 4.77, + "learning_rate": 7.403657085054863e-06, + "loss": 0.0196, + "step": 36405 + }, + { + "epoch": 4.77, + "learning_rate": 7.403144950071608e-06, + "loss": 0.0189, + "step": 36406 + }, + { + "epoch": 4.77, + "learning_rate": 7.402632822391871e-06, + "loss": 0.0178, + "step": 36407 + }, + { + "epoch": 4.77, + "learning_rate": 7.402120702017088e-06, + "loss": 0.0198, + "step": 36408 + }, + { + "epoch": 4.77, + "learning_rate": 7.4016085889487055e-06, + "loss": 0.0158, + "step": 36409 + }, + { + "epoch": 4.77, + "learning_rate": 7.4010964831881584e-06, + "loss": 0.0205, + "step": 36410 + }, + { + "epoch": 4.77, + "learning_rate": 7.400584384736891e-06, + "loss": 0.0192, + "step": 36411 + }, + { + "epoch": 4.77, + "learning_rate": 7.400072293596339e-06, + "loss": 0.0202, + "step": 36412 + }, + { + "epoch": 4.77, + "learning_rate": 7.399560209767946e-06, + "loss": 0.021, + "step": 36413 + }, + { + "epoch": 4.77, + "learning_rate": 7.3990481332531525e-06, + "loss": 0.0159, + "step": 36414 + }, + { + "epoch": 4.77, + "learning_rate": 7.398536064053394e-06, + "loss": 0.0182, + "step": 36415 + }, + { + "epoch": 4.77, + "learning_rate": 7.398024002170117e-06, + "loss": 0.013, + "step": 36416 + }, + { + "epoch": 4.77, + "learning_rate": 7.3975119476047605e-06, + "loss": 0.0207, + "step": 36417 + }, + { + "epoch": 4.77, + "learning_rate": 7.39699990035876e-06, + "loss": 0.0212, + "step": 36418 + }, + { + "epoch": 4.77, + "learning_rate": 7.396487860433562e-06, + "loss": 0.0175, + "step": 36419 + }, + { + "epoch": 4.77, + "learning_rate": 7.395975827830602e-06, + "loss": 0.0162, + "step": 36420 + }, + { + "epoch": 4.77, + "learning_rate": 7.395463802551319e-06, + "loss": 0.0154, + "step": 36421 + }, + { + "epoch": 4.77, + "learning_rate": 7.394951784597158e-06, + "loss": 0.0221, + "step": 36422 + }, + { + "epoch": 4.77, + "learning_rate": 7.394439773969555e-06, + "loss": 0.0177, + "step": 36423 + }, + { + "epoch": 4.77, + "learning_rate": 7.393927770669953e-06, + "loss": 0.0161, + "step": 36424 + }, + { + "epoch": 4.77, + "learning_rate": 7.393415774699789e-06, + "loss": 0.0219, + "step": 36425 + }, + { + "epoch": 4.77, + "learning_rate": 7.392903786060501e-06, + "loss": 0.0201, + "step": 36426 + }, + { + "epoch": 4.77, + "learning_rate": 7.392391804753536e-06, + "loss": 0.0195, + "step": 36427 + }, + { + "epoch": 4.77, + "learning_rate": 7.391879830780331e-06, + "loss": 0.0224, + "step": 36428 + }, + { + "epoch": 4.77, + "learning_rate": 7.391367864142324e-06, + "loss": 0.0223, + "step": 36429 + }, + { + "epoch": 4.77, + "learning_rate": 7.390855904840957e-06, + "loss": 0.0159, + "step": 36430 + }, + { + "epoch": 4.77, + "learning_rate": 7.390343952877668e-06, + "loss": 0.0201, + "step": 36431 + }, + { + "epoch": 4.77, + "learning_rate": 7.389832008253899e-06, + "loss": 0.0187, + "step": 36432 + }, + { + "epoch": 4.77, + "learning_rate": 7.389320070971087e-06, + "loss": 0.0179, + "step": 36433 + }, + { + "epoch": 4.77, + "learning_rate": 7.3888081410306745e-06, + "loss": 0.0182, + "step": 36434 + }, + { + "epoch": 4.77, + "learning_rate": 7.3882962184340964e-06, + "loss": 0.0203, + "step": 36435 + }, + { + "epoch": 4.77, + "learning_rate": 7.387784303182799e-06, + "loss": 0.0138, + "step": 36436 + }, + { + "epoch": 4.77, + "learning_rate": 7.387272395278215e-06, + "loss": 0.0201, + "step": 36437 + }, + { + "epoch": 4.77, + "learning_rate": 7.386760494721794e-06, + "loss": 0.0175, + "step": 36438 + }, + { + "epoch": 4.77, + "learning_rate": 7.386248601514967e-06, + "loss": 0.0217, + "step": 36439 + }, + { + "epoch": 4.77, + "learning_rate": 7.385736715659179e-06, + "loss": 0.0166, + "step": 36440 + }, + { + "epoch": 4.77, + "learning_rate": 7.385224837155865e-06, + "loss": 0.0164, + "step": 36441 + }, + { + "epoch": 4.77, + "learning_rate": 7.3847129660064675e-06, + "loss": 0.0185, + "step": 36442 + }, + { + "epoch": 4.77, + "learning_rate": 7.384201102212427e-06, + "loss": 0.0175, + "step": 36443 + }, + { + "epoch": 4.77, + "learning_rate": 7.38368924577518e-06, + "loss": 0.0231, + "step": 36444 + }, + { + "epoch": 4.77, + "learning_rate": 7.383177396696168e-06, + "loss": 0.0241, + "step": 36445 + }, + { + "epoch": 4.77, + "learning_rate": 7.382665554976831e-06, + "loss": 0.0177, + "step": 36446 + }, + { + "epoch": 4.77, + "learning_rate": 7.382153720618608e-06, + "loss": 0.0152, + "step": 36447 + }, + { + "epoch": 4.77, + "learning_rate": 7.381641893622935e-06, + "loss": 0.0164, + "step": 36448 + }, + { + "epoch": 4.77, + "learning_rate": 7.381130073991256e-06, + "loss": 0.0198, + "step": 36449 + }, + { + "epoch": 4.77, + "learning_rate": 7.380618261725009e-06, + "loss": 0.019, + "step": 36450 + }, + { + "epoch": 4.77, + "learning_rate": 7.3801064568256365e-06, + "loss": 0.019, + "step": 36451 + }, + { + "epoch": 4.77, + "learning_rate": 7.3795946592945725e-06, + "loss": 0.0204, + "step": 36452 + }, + { + "epoch": 4.78, + "learning_rate": 7.379082869133261e-06, + "loss": 0.0111, + "step": 36453 + }, + { + "epoch": 4.78, + "learning_rate": 7.3785710863431385e-06, + "loss": 0.0193, + "step": 36454 + }, + { + "epoch": 4.78, + "learning_rate": 7.378059310925645e-06, + "loss": 0.0139, + "step": 36455 + }, + { + "epoch": 4.78, + "learning_rate": 7.3775475428822195e-06, + "loss": 0.0195, + "step": 36456 + }, + { + "epoch": 4.78, + "learning_rate": 7.377035782214301e-06, + "loss": 0.0198, + "step": 36457 + }, + { + "epoch": 4.78, + "learning_rate": 7.376524028923333e-06, + "loss": 0.0177, + "step": 36458 + }, + { + "epoch": 4.78, + "learning_rate": 7.376012283010747e-06, + "loss": 0.0182, + "step": 36459 + }, + { + "epoch": 4.78, + "learning_rate": 7.375500544477989e-06, + "loss": 0.0182, + "step": 36460 + }, + { + "epoch": 4.78, + "learning_rate": 7.374988813326498e-06, + "loss": 0.0259, + "step": 36461 + }, + { + "epoch": 4.78, + "learning_rate": 7.374477089557708e-06, + "loss": 0.0214, + "step": 36462 + }, + { + "epoch": 4.78, + "learning_rate": 7.373965373173062e-06, + "loss": 0.0185, + "step": 36463 + }, + { + "epoch": 4.78, + "learning_rate": 7.373453664174e-06, + "loss": 0.0195, + "step": 36464 + }, + { + "epoch": 4.78, + "learning_rate": 7.3729419625619595e-06, + "loss": 0.016, + "step": 36465 + }, + { + "epoch": 4.78, + "learning_rate": 7.3724302683383795e-06, + "loss": 0.0165, + "step": 36466 + }, + { + "epoch": 4.78, + "learning_rate": 7.371918581504699e-06, + "loss": 0.0178, + "step": 36467 + }, + { + "epoch": 4.78, + "learning_rate": 7.371406902062359e-06, + "loss": 0.0223, + "step": 36468 + }, + { + "epoch": 4.78, + "learning_rate": 7.370895230012796e-06, + "loss": 0.0151, + "step": 36469 + }, + { + "epoch": 4.78, + "learning_rate": 7.370383565357446e-06, + "loss": 0.0177, + "step": 36470 + }, + { + "epoch": 4.78, + "learning_rate": 7.369871908097757e-06, + "loss": 0.017, + "step": 36471 + }, + { + "epoch": 4.78, + "learning_rate": 7.369360258235163e-06, + "loss": 0.0182, + "step": 36472 + }, + { + "epoch": 4.78, + "learning_rate": 7.368848615771103e-06, + "loss": 0.0158, + "step": 36473 + }, + { + "epoch": 4.78, + "learning_rate": 7.368336980707016e-06, + "loss": 0.0147, + "step": 36474 + }, + { + "epoch": 4.78, + "learning_rate": 7.36782535304434e-06, + "loss": 0.0191, + "step": 36475 + }, + { + "epoch": 4.78, + "learning_rate": 7.367313732784518e-06, + "loss": 0.0172, + "step": 36476 + }, + { + "epoch": 4.78, + "learning_rate": 7.366802119928982e-06, + "loss": 0.0239, + "step": 36477 + }, + { + "epoch": 4.78, + "learning_rate": 7.366290514479176e-06, + "loss": 0.014, + "step": 36478 + }, + { + "epoch": 4.78, + "learning_rate": 7.3657789164365405e-06, + "loss": 0.021, + "step": 36479 + }, + { + "epoch": 4.78, + "learning_rate": 7.365267325802507e-06, + "loss": 0.0174, + "step": 36480 + }, + { + "epoch": 4.78, + "learning_rate": 7.36475574257852e-06, + "loss": 0.016, + "step": 36481 + }, + { + "epoch": 4.78, + "learning_rate": 7.364244166766019e-06, + "loss": 0.0171, + "step": 36482 + }, + { + "epoch": 4.78, + "learning_rate": 7.36373259836644e-06, + "loss": 0.0148, + "step": 36483 + }, + { + "epoch": 4.78, + "learning_rate": 7.363221037381224e-06, + "loss": 0.0173, + "step": 36484 + }, + { + "epoch": 4.78, + "learning_rate": 7.362709483811807e-06, + "loss": 0.0187, + "step": 36485 + }, + { + "epoch": 4.78, + "learning_rate": 7.362197937659629e-06, + "loss": 0.0162, + "step": 36486 + }, + { + "epoch": 4.78, + "learning_rate": 7.36168639892613e-06, + "loss": 0.0221, + "step": 36487 + }, + { + "epoch": 4.78, + "learning_rate": 7.361174867612746e-06, + "loss": 0.0154, + "step": 36488 + }, + { + "epoch": 4.78, + "learning_rate": 7.360663343720919e-06, + "loss": 0.0197, + "step": 36489 + }, + { + "epoch": 4.78, + "learning_rate": 7.360151827252085e-06, + "loss": 0.0159, + "step": 36490 + }, + { + "epoch": 4.78, + "learning_rate": 7.3596403182076795e-06, + "loss": 0.0184, + "step": 36491 + }, + { + "epoch": 4.78, + "learning_rate": 7.359128816589149e-06, + "loss": 0.0187, + "step": 36492 + }, + { + "epoch": 4.78, + "learning_rate": 7.3586173223979286e-06, + "loss": 0.019, + "step": 36493 + }, + { + "epoch": 4.78, + "learning_rate": 7.358105835635454e-06, + "loss": 0.0174, + "step": 36494 + }, + { + "epoch": 4.78, + "learning_rate": 7.357594356303169e-06, + "loss": 0.0187, + "step": 36495 + }, + { + "epoch": 4.78, + "learning_rate": 7.357082884402506e-06, + "loss": 0.0176, + "step": 36496 + }, + { + "epoch": 4.78, + "learning_rate": 7.356571419934911e-06, + "loss": 0.0161, + "step": 36497 + }, + { + "epoch": 4.78, + "learning_rate": 7.356059962901815e-06, + "loss": 0.0149, + "step": 36498 + }, + { + "epoch": 4.78, + "learning_rate": 7.355548513304661e-06, + "loss": 0.017, + "step": 36499 + }, + { + "epoch": 4.78, + "learning_rate": 7.3550370711448835e-06, + "loss": 0.0148, + "step": 36500 + }, + { + "epoch": 4.78, + "learning_rate": 7.354525636423925e-06, + "loss": 0.0181, + "step": 36501 + }, + { + "epoch": 4.78, + "learning_rate": 7.3540142091432184e-06, + "loss": 0.0198, + "step": 36502 + }, + { + "epoch": 4.78, + "learning_rate": 7.353502789304212e-06, + "loss": 0.0201, + "step": 36503 + }, + { + "epoch": 4.78, + "learning_rate": 7.352991376908334e-06, + "loss": 0.0179, + "step": 36504 + }, + { + "epoch": 4.78, + "learning_rate": 7.352479971957031e-06, + "loss": 0.0172, + "step": 36505 + }, + { + "epoch": 4.78, + "learning_rate": 7.3519685744517335e-06, + "loss": 0.0184, + "step": 36506 + }, + { + "epoch": 4.78, + "learning_rate": 7.351457184393884e-06, + "loss": 0.0165, + "step": 36507 + }, + { + "epoch": 4.78, + "learning_rate": 7.3509458017849215e-06, + "loss": 0.0174, + "step": 36508 + }, + { + "epoch": 4.78, + "learning_rate": 7.35043442662628e-06, + "loss": 0.0173, + "step": 36509 + }, + { + "epoch": 4.78, + "learning_rate": 7.3499230589194044e-06, + "loss": 0.0182, + "step": 36510 + }, + { + "epoch": 4.78, + "learning_rate": 7.349411698665727e-06, + "loss": 0.0156, + "step": 36511 + }, + { + "epoch": 4.78, + "learning_rate": 7.34890034586669e-06, + "loss": 0.0171, + "step": 36512 + }, + { + "epoch": 4.78, + "learning_rate": 7.348389000523723e-06, + "loss": 0.0187, + "step": 36513 + }, + { + "epoch": 4.78, + "learning_rate": 7.347877662638277e-06, + "loss": 0.0206, + "step": 36514 + }, + { + "epoch": 4.78, + "learning_rate": 7.347366332211781e-06, + "loss": 0.018, + "step": 36515 + }, + { + "epoch": 4.78, + "learning_rate": 7.346855009245679e-06, + "loss": 0.0167, + "step": 36516 + }, + { + "epoch": 4.78, + "learning_rate": 7.346343693741403e-06, + "loss": 0.0149, + "step": 36517 + }, + { + "epoch": 4.78, + "learning_rate": 7.345832385700398e-06, + "loss": 0.0187, + "step": 36518 + }, + { + "epoch": 4.78, + "learning_rate": 7.345321085124093e-06, + "loss": 0.019, + "step": 36519 + }, + { + "epoch": 4.78, + "learning_rate": 7.344809792013936e-06, + "loss": 0.0182, + "step": 36520 + }, + { + "epoch": 4.78, + "learning_rate": 7.344298506371357e-06, + "loss": 0.014, + "step": 36521 + }, + { + "epoch": 4.78, + "learning_rate": 7.343787228197796e-06, + "loss": 0.0187, + "step": 36522 + }, + { + "epoch": 4.78, + "learning_rate": 7.343275957494696e-06, + "loss": 0.0176, + "step": 36523 + }, + { + "epoch": 4.78, + "learning_rate": 7.342764694263484e-06, + "loss": 0.0153, + "step": 36524 + }, + { + "epoch": 4.78, + "learning_rate": 7.342253438505609e-06, + "loss": 0.0137, + "step": 36525 + }, + { + "epoch": 4.78, + "learning_rate": 7.341742190222507e-06, + "loss": 0.0202, + "step": 36526 + }, + { + "epoch": 4.78, + "learning_rate": 7.34123094941561e-06, + "loss": 0.0223, + "step": 36527 + }, + { + "epoch": 4.78, + "learning_rate": 7.340719716086361e-06, + "loss": 0.021, + "step": 36528 + }, + { + "epoch": 4.79, + "learning_rate": 7.340208490236197e-06, + "loss": 0.0125, + "step": 36529 + }, + { + "epoch": 4.79, + "learning_rate": 7.339697271866553e-06, + "loss": 0.0161, + "step": 36530 + }, + { + "epoch": 4.79, + "learning_rate": 7.339186060978871e-06, + "loss": 0.0167, + "step": 36531 + }, + { + "epoch": 4.79, + "learning_rate": 7.338674857574585e-06, + "loss": 0.0172, + "step": 36532 + }, + { + "epoch": 4.79, + "learning_rate": 7.338163661655136e-06, + "loss": 0.0167, + "step": 36533 + }, + { + "epoch": 4.79, + "learning_rate": 7.3376524732219564e-06, + "loss": 0.0189, + "step": 36534 + }, + { + "epoch": 4.79, + "learning_rate": 7.3371412922764875e-06, + "loss": 0.022, + "step": 36535 + }, + { + "epoch": 4.79, + "learning_rate": 7.336630118820168e-06, + "loss": 0.0169, + "step": 36536 + }, + { + "epoch": 4.79, + "learning_rate": 7.336118952854436e-06, + "loss": 0.0163, + "step": 36537 + }, + { + "epoch": 4.79, + "learning_rate": 7.335607794380727e-06, + "loss": 0.019, + "step": 36538 + }, + { + "epoch": 4.79, + "learning_rate": 7.3350966434004815e-06, + "loss": 0.0172, + "step": 36539 + }, + { + "epoch": 4.79, + "learning_rate": 7.33458549991513e-06, + "loss": 0.0181, + "step": 36540 + }, + { + "epoch": 4.79, + "learning_rate": 7.3340743639261205e-06, + "loss": 0.0219, + "step": 36541 + }, + { + "epoch": 4.79, + "learning_rate": 7.3335632354348795e-06, + "loss": 0.0175, + "step": 36542 + }, + { + "epoch": 4.79, + "learning_rate": 7.3330521144428516e-06, + "loss": 0.015, + "step": 36543 + }, + { + "epoch": 4.79, + "learning_rate": 7.332541000951474e-06, + "loss": 0.0172, + "step": 36544 + }, + { + "epoch": 4.79, + "learning_rate": 7.332029894962182e-06, + "loss": 0.0225, + "step": 36545 + }, + { + "epoch": 4.79, + "learning_rate": 7.33151879647641e-06, + "loss": 0.0211, + "step": 36546 + }, + { + "epoch": 4.79, + "learning_rate": 7.331007705495605e-06, + "loss": 0.0201, + "step": 36547 + }, + { + "epoch": 4.79, + "learning_rate": 7.330496622021195e-06, + "loss": 0.0169, + "step": 36548 + }, + { + "epoch": 4.79, + "learning_rate": 7.329985546054624e-06, + "loss": 0.0156, + "step": 36549 + }, + { + "epoch": 4.79, + "learning_rate": 7.329474477597325e-06, + "loss": 0.0191, + "step": 36550 + }, + { + "epoch": 4.79, + "learning_rate": 7.328963416650736e-06, + "loss": 0.0176, + "step": 36551 + }, + { + "epoch": 4.79, + "learning_rate": 7.328452363216299e-06, + "loss": 0.0186, + "step": 36552 + }, + { + "epoch": 4.79, + "learning_rate": 7.327941317295443e-06, + "loss": 0.0181, + "step": 36553 + }, + { + "epoch": 4.79, + "learning_rate": 7.327430278889613e-06, + "loss": 0.0168, + "step": 36554 + }, + { + "epoch": 4.79, + "learning_rate": 7.326919248000239e-06, + "loss": 0.0224, + "step": 36555 + }, + { + "epoch": 4.79, + "learning_rate": 7.326408224628767e-06, + "loss": 0.0263, + "step": 36556 + }, + { + "epoch": 4.79, + "learning_rate": 7.325897208776622e-06, + "loss": 0.0174, + "step": 36557 + }, + { + "epoch": 4.79, + "learning_rate": 7.325386200445255e-06, + "loss": 0.014, + "step": 36558 + }, + { + "epoch": 4.79, + "learning_rate": 7.324875199636096e-06, + "loss": 0.0189, + "step": 36559 + }, + { + "epoch": 4.79, + "learning_rate": 7.324364206350584e-06, + "loss": 0.0133, + "step": 36560 + }, + { + "epoch": 4.79, + "learning_rate": 7.323853220590154e-06, + "loss": 0.0194, + "step": 36561 + }, + { + "epoch": 4.79, + "learning_rate": 7.323342242356244e-06, + "loss": 0.0186, + "step": 36562 + }, + { + "epoch": 4.79, + "learning_rate": 7.322831271650292e-06, + "loss": 0.0172, + "step": 36563 + }, + { + "epoch": 4.79, + "learning_rate": 7.3223203084737326e-06, + "loss": 0.0195, + "step": 36564 + }, + { + "epoch": 4.79, + "learning_rate": 7.321809352828008e-06, + "loss": 0.0209, + "step": 36565 + }, + { + "epoch": 4.79, + "learning_rate": 7.32129840471455e-06, + "loss": 0.0181, + "step": 36566 + }, + { + "epoch": 4.79, + "learning_rate": 7.320787464134799e-06, + "loss": 0.0177, + "step": 36567 + }, + { + "epoch": 4.79, + "learning_rate": 7.320276531090186e-06, + "loss": 0.0177, + "step": 36568 + }, + { + "epoch": 4.79, + "learning_rate": 7.319765605582155e-06, + "loss": 0.0167, + "step": 36569 + }, + { + "epoch": 4.79, + "learning_rate": 7.319254687612143e-06, + "loss": 0.0217, + "step": 36570 + }, + { + "epoch": 4.79, + "learning_rate": 7.3187437771815825e-06, + "loss": 0.0172, + "step": 36571 + }, + { + "epoch": 4.79, + "learning_rate": 7.318232874291911e-06, + "loss": 0.0114, + "step": 36572 + }, + { + "epoch": 4.79, + "learning_rate": 7.317721978944571e-06, + "loss": 0.0161, + "step": 36573 + }, + { + "epoch": 4.79, + "learning_rate": 7.3172110911409926e-06, + "loss": 0.0165, + "step": 36574 + }, + { + "epoch": 4.79, + "learning_rate": 7.3167002108826146e-06, + "loss": 0.0177, + "step": 36575 + }, + { + "epoch": 4.79, + "learning_rate": 7.316189338170875e-06, + "loss": 0.0206, + "step": 36576 + }, + { + "epoch": 4.79, + "learning_rate": 7.31567847300721e-06, + "loss": 0.0198, + "step": 36577 + }, + { + "epoch": 4.79, + "learning_rate": 7.315167615393057e-06, + "loss": 0.0178, + "step": 36578 + }, + { + "epoch": 4.79, + "learning_rate": 7.314656765329848e-06, + "loss": 0.0202, + "step": 36579 + }, + { + "epoch": 4.79, + "learning_rate": 7.314145922819027e-06, + "loss": 0.02, + "step": 36580 + }, + { + "epoch": 4.79, + "learning_rate": 7.313635087862028e-06, + "loss": 0.0182, + "step": 36581 + }, + { + "epoch": 4.79, + "learning_rate": 7.313124260460287e-06, + "loss": 0.0153, + "step": 36582 + }, + { + "epoch": 4.79, + "learning_rate": 7.312613440615241e-06, + "loss": 0.0136, + "step": 36583 + }, + { + "epoch": 4.79, + "learning_rate": 7.312102628328326e-06, + "loss": 0.016, + "step": 36584 + }, + { + "epoch": 4.79, + "learning_rate": 7.311591823600981e-06, + "loss": 0.0173, + "step": 36585 + }, + { + "epoch": 4.79, + "learning_rate": 7.311081026434639e-06, + "loss": 0.0205, + "step": 36586 + }, + { + "epoch": 4.79, + "learning_rate": 7.3105702368307385e-06, + "loss": 0.0217, + "step": 36587 + }, + { + "epoch": 4.79, + "learning_rate": 7.310059454790718e-06, + "loss": 0.0173, + "step": 36588 + }, + { + "epoch": 4.79, + "learning_rate": 7.30954868031601e-06, + "loss": 0.0155, + "step": 36589 + }, + { + "epoch": 4.79, + "learning_rate": 7.30903791340805e-06, + "loss": 0.016, + "step": 36590 + }, + { + "epoch": 4.79, + "learning_rate": 7.308527154068283e-06, + "loss": 0.0173, + "step": 36591 + }, + { + "epoch": 4.79, + "learning_rate": 7.308016402298137e-06, + "loss": 0.0182, + "step": 36592 + }, + { + "epoch": 4.79, + "learning_rate": 7.307505658099052e-06, + "loss": 0.0163, + "step": 36593 + }, + { + "epoch": 4.79, + "learning_rate": 7.306994921472466e-06, + "loss": 0.0187, + "step": 36594 + }, + { + "epoch": 4.79, + "learning_rate": 7.306484192419811e-06, + "loss": 0.0228, + "step": 36595 + }, + { + "epoch": 4.79, + "learning_rate": 7.305973470942529e-06, + "loss": 0.0226, + "step": 36596 + }, + { + "epoch": 4.79, + "learning_rate": 7.30546275704205e-06, + "loss": 0.019, + "step": 36597 + }, + { + "epoch": 4.79, + "learning_rate": 7.304952050719816e-06, + "loss": 0.0211, + "step": 36598 + }, + { + "epoch": 4.79, + "learning_rate": 7.304441351977258e-06, + "loss": 0.0151, + "step": 36599 + }, + { + "epoch": 4.79, + "learning_rate": 7.3039306608158176e-06, + "loss": 0.0172, + "step": 36600 + }, + { + "epoch": 4.79, + "learning_rate": 7.303419977236923e-06, + "loss": 0.0173, + "step": 36601 + }, + { + "epoch": 4.79, + "learning_rate": 7.302909301242021e-06, + "loss": 0.0233, + "step": 36602 + }, + { + "epoch": 4.79, + "learning_rate": 7.302398632832542e-06, + "loss": 0.0161, + "step": 36603 + }, + { + "epoch": 4.79, + "learning_rate": 7.301887972009924e-06, + "loss": 0.0175, + "step": 36604 + }, + { + "epoch": 4.79, + "learning_rate": 7.301377318775602e-06, + "loss": 0.0195, + "step": 36605 + }, + { + "epoch": 4.8, + "learning_rate": 7.300866673131015e-06, + "loss": 0.016, + "step": 36606 + }, + { + "epoch": 4.8, + "learning_rate": 7.300356035077594e-06, + "loss": 0.0153, + "step": 36607 + }, + { + "epoch": 4.8, + "learning_rate": 7.299845404616776e-06, + "loss": 0.0173, + "step": 36608 + }, + { + "epoch": 4.8, + "learning_rate": 7.299334781750002e-06, + "loss": 0.0169, + "step": 36609 + }, + { + "epoch": 4.8, + "learning_rate": 7.298824166478705e-06, + "loss": 0.014, + "step": 36610 + }, + { + "epoch": 4.8, + "learning_rate": 7.298313558804322e-06, + "loss": 0.02, + "step": 36611 + }, + { + "epoch": 4.8, + "learning_rate": 7.297802958728284e-06, + "loss": 0.0209, + "step": 36612 + }, + { + "epoch": 4.8, + "learning_rate": 7.297292366252033e-06, + "loss": 0.0188, + "step": 36613 + }, + { + "epoch": 4.8, + "learning_rate": 7.296781781377005e-06, + "loss": 0.0175, + "step": 36614 + }, + { + "epoch": 4.8, + "learning_rate": 7.296271204104635e-06, + "loss": 0.0163, + "step": 36615 + }, + { + "epoch": 4.8, + "learning_rate": 7.295760634436355e-06, + "loss": 0.0197, + "step": 36616 + }, + { + "epoch": 4.8, + "learning_rate": 7.295250072373607e-06, + "loss": 0.0155, + "step": 36617 + }, + { + "epoch": 4.8, + "learning_rate": 7.294739517917824e-06, + "loss": 0.0135, + "step": 36618 + }, + { + "epoch": 4.8, + "learning_rate": 7.294228971070443e-06, + "loss": 0.0194, + "step": 36619 + }, + { + "epoch": 4.8, + "learning_rate": 7.293718431832896e-06, + "loss": 0.0202, + "step": 36620 + }, + { + "epoch": 4.8, + "learning_rate": 7.2932079002066246e-06, + "loss": 0.0146, + "step": 36621 + }, + { + "epoch": 4.8, + "learning_rate": 7.292697376193054e-06, + "loss": 0.0174, + "step": 36622 + }, + { + "epoch": 4.8, + "learning_rate": 7.292186859793636e-06, + "loss": 0.0207, + "step": 36623 + }, + { + "epoch": 4.8, + "learning_rate": 7.291676351009795e-06, + "loss": 0.0194, + "step": 36624 + }, + { + "epoch": 4.8, + "learning_rate": 7.291165849842974e-06, + "loss": 0.0178, + "step": 36625 + }, + { + "epoch": 4.8, + "learning_rate": 7.290655356294601e-06, + "loss": 0.0171, + "step": 36626 + }, + { + "epoch": 4.8, + "learning_rate": 7.290144870366118e-06, + "loss": 0.018, + "step": 36627 + }, + { + "epoch": 4.8, + "learning_rate": 7.289634392058958e-06, + "loss": 0.0197, + "step": 36628 + }, + { + "epoch": 4.8, + "learning_rate": 7.289123921374555e-06, + "loss": 0.0199, + "step": 36629 + }, + { + "epoch": 4.8, + "learning_rate": 7.288613458314349e-06, + "loss": 0.0198, + "step": 36630 + }, + { + "epoch": 4.8, + "learning_rate": 7.288103002879769e-06, + "loss": 0.0191, + "step": 36631 + }, + { + "epoch": 4.8, + "learning_rate": 7.28759255507226e-06, + "loss": 0.0205, + "step": 36632 + }, + { + "epoch": 4.8, + "learning_rate": 7.287082114893247e-06, + "loss": 0.0134, + "step": 36633 + }, + { + "epoch": 4.8, + "learning_rate": 7.286571682344174e-06, + "loss": 0.0157, + "step": 36634 + }, + { + "epoch": 4.8, + "learning_rate": 7.286061257426475e-06, + "loss": 0.0169, + "step": 36635 + }, + { + "epoch": 4.8, + "learning_rate": 7.285550840141582e-06, + "loss": 0.0208, + "step": 36636 + }, + { + "epoch": 4.8, + "learning_rate": 7.2850404304909336e-06, + "loss": 0.0167, + "step": 36637 + }, + { + "epoch": 4.8, + "learning_rate": 7.284530028475967e-06, + "loss": 0.0174, + "step": 36638 + }, + { + "epoch": 4.8, + "learning_rate": 7.284019634098112e-06, + "loss": 0.0173, + "step": 36639 + }, + { + "epoch": 4.8, + "learning_rate": 7.283509247358809e-06, + "loss": 0.0176, + "step": 36640 + }, + { + "epoch": 4.8, + "learning_rate": 7.282998868259489e-06, + "loss": 0.0193, + "step": 36641 + }, + { + "epoch": 4.8, + "learning_rate": 7.282488496801593e-06, + "loss": 0.0221, + "step": 36642 + }, + { + "epoch": 4.8, + "learning_rate": 7.281978132986551e-06, + "loss": 0.0179, + "step": 36643 + }, + { + "epoch": 4.8, + "learning_rate": 7.281467776815798e-06, + "loss": 0.0136, + "step": 36644 + }, + { + "epoch": 4.8, + "learning_rate": 7.280957428290776e-06, + "loss": 0.0198, + "step": 36645 + }, + { + "epoch": 4.8, + "learning_rate": 7.280447087412916e-06, + "loss": 0.0189, + "step": 36646 + }, + { + "epoch": 4.8, + "learning_rate": 7.279936754183654e-06, + "loss": 0.0147, + "step": 36647 + }, + { + "epoch": 4.8, + "learning_rate": 7.279426428604426e-06, + "loss": 0.0204, + "step": 36648 + }, + { + "epoch": 4.8, + "learning_rate": 7.278916110676665e-06, + "loss": 0.017, + "step": 36649 + }, + { + "epoch": 4.8, + "learning_rate": 7.278405800401809e-06, + "loss": 0.0199, + "step": 36650 + }, + { + "epoch": 4.8, + "learning_rate": 7.27789549778129e-06, + "loss": 0.0159, + "step": 36651 + }, + { + "epoch": 4.8, + "learning_rate": 7.277385202816545e-06, + "loss": 0.0149, + "step": 36652 + }, + { + "epoch": 4.8, + "learning_rate": 7.27687491550901e-06, + "loss": 0.0222, + "step": 36653 + }, + { + "epoch": 4.8, + "learning_rate": 7.276364635860118e-06, + "loss": 0.015, + "step": 36654 + }, + { + "epoch": 4.8, + "learning_rate": 7.2758543638713045e-06, + "loss": 0.0164, + "step": 36655 + }, + { + "epoch": 4.8, + "learning_rate": 7.275344099544008e-06, + "loss": 0.0187, + "step": 36656 + }, + { + "epoch": 4.8, + "learning_rate": 7.27483384287966e-06, + "loss": 0.0154, + "step": 36657 + }, + { + "epoch": 4.8, + "learning_rate": 7.274323593879695e-06, + "loss": 0.0181, + "step": 36658 + }, + { + "epoch": 4.8, + "learning_rate": 7.2738133525455535e-06, + "loss": 0.0174, + "step": 36659 + }, + { + "epoch": 4.8, + "learning_rate": 7.273303118878665e-06, + "loss": 0.0153, + "step": 36660 + }, + { + "epoch": 4.8, + "learning_rate": 7.272792892880468e-06, + "loss": 0.0236, + "step": 36661 + }, + { + "epoch": 4.8, + "learning_rate": 7.272282674552394e-06, + "loss": 0.0193, + "step": 36662 + }, + { + "epoch": 4.8, + "learning_rate": 7.271772463895881e-06, + "loss": 0.0143, + "step": 36663 + }, + { + "epoch": 4.8, + "learning_rate": 7.2712622609123596e-06, + "loss": 0.0118, + "step": 36664 + }, + { + "epoch": 4.8, + "learning_rate": 7.270752065603271e-06, + "loss": 0.0199, + "step": 36665 + }, + { + "epoch": 4.8, + "learning_rate": 7.270241877970041e-06, + "loss": 0.017, + "step": 36666 + }, + { + "epoch": 4.8, + "learning_rate": 7.269731698014116e-06, + "loss": 0.0148, + "step": 36667 + }, + { + "epoch": 4.8, + "learning_rate": 7.2692215257369244e-06, + "loss": 0.0177, + "step": 36668 + }, + { + "epoch": 4.8, + "learning_rate": 7.2687113611399015e-06, + "loss": 0.0185, + "step": 36669 + }, + { + "epoch": 4.8, + "learning_rate": 7.268201204224483e-06, + "loss": 0.0202, + "step": 36670 + }, + { + "epoch": 4.8, + "learning_rate": 7.267691054992103e-06, + "loss": 0.0162, + "step": 36671 + }, + { + "epoch": 4.8, + "learning_rate": 7.267180913444196e-06, + "loss": 0.0195, + "step": 36672 + }, + { + "epoch": 4.8, + "learning_rate": 7.2666707795821965e-06, + "loss": 0.0173, + "step": 36673 + }, + { + "epoch": 4.8, + "learning_rate": 7.266160653407541e-06, + "loss": 0.0171, + "step": 36674 + }, + { + "epoch": 4.8, + "learning_rate": 7.2656505349216626e-06, + "loss": 0.0147, + "step": 36675 + }, + { + "epoch": 4.8, + "learning_rate": 7.265140424125998e-06, + "loss": 0.0217, + "step": 36676 + }, + { + "epoch": 4.8, + "learning_rate": 7.264630321021976e-06, + "loss": 0.023, + "step": 36677 + }, + { + "epoch": 4.8, + "learning_rate": 7.264120225611037e-06, + "loss": 0.0195, + "step": 36678 + }, + { + "epoch": 4.8, + "learning_rate": 7.263610137894616e-06, + "loss": 0.0186, + "step": 36679 + }, + { + "epoch": 4.8, + "learning_rate": 7.263100057874146e-06, + "loss": 0.0175, + "step": 36680 + }, + { + "epoch": 4.8, + "learning_rate": 7.26258998555106e-06, + "loss": 0.0146, + "step": 36681 + }, + { + "epoch": 4.81, + "learning_rate": 7.262079920926797e-06, + "loss": 0.0214, + "step": 36682 + }, + { + "epoch": 4.81, + "learning_rate": 7.261569864002784e-06, + "loss": 0.0189, + "step": 36683 + }, + { + "epoch": 4.81, + "learning_rate": 7.261059814780464e-06, + "loss": 0.0186, + "step": 36684 + }, + { + "epoch": 4.81, + "learning_rate": 7.260549773261264e-06, + "loss": 0.0161, + "step": 36685 + }, + { + "epoch": 4.81, + "learning_rate": 7.260039739446625e-06, + "loss": 0.0145, + "step": 36686 + }, + { + "epoch": 4.81, + "learning_rate": 7.259529713337976e-06, + "loss": 0.0174, + "step": 36687 + }, + { + "epoch": 4.81, + "learning_rate": 7.259019694936751e-06, + "loss": 0.0187, + "step": 36688 + }, + { + "epoch": 4.81, + "learning_rate": 7.25850968424439e-06, + "loss": 0.0202, + "step": 36689 + }, + { + "epoch": 4.81, + "learning_rate": 7.257999681262327e-06, + "loss": 0.0158, + "step": 36690 + }, + { + "epoch": 4.81, + "learning_rate": 7.25748968599199e-06, + "loss": 0.0159, + "step": 36691 + }, + { + "epoch": 4.81, + "learning_rate": 7.256979698434821e-06, + "loss": 0.0198, + "step": 36692 + }, + { + "epoch": 4.81, + "learning_rate": 7.2564697185922475e-06, + "loss": 0.0201, + "step": 36693 + }, + { + "epoch": 4.81, + "learning_rate": 7.255959746465706e-06, + "loss": 0.0172, + "step": 36694 + }, + { + "epoch": 4.81, + "learning_rate": 7.255449782056636e-06, + "loss": 0.0181, + "step": 36695 + }, + { + "epoch": 4.81, + "learning_rate": 7.254939825366463e-06, + "loss": 0.0163, + "step": 36696 + }, + { + "epoch": 4.81, + "learning_rate": 7.254429876396629e-06, + "loss": 0.0138, + "step": 36697 + }, + { + "epoch": 4.81, + "learning_rate": 7.253919935148562e-06, + "loss": 0.0171, + "step": 36698 + }, + { + "epoch": 4.81, + "learning_rate": 7.2534100016236975e-06, + "loss": 0.0176, + "step": 36699 + }, + { + "epoch": 4.81, + "learning_rate": 7.252900075823474e-06, + "loss": 0.0157, + "step": 36700 + }, + { + "epoch": 4.81, + "learning_rate": 7.252390157749323e-06, + "loss": 0.019, + "step": 36701 + }, + { + "epoch": 4.81, + "learning_rate": 7.251880247402676e-06, + "loss": 0.0193, + "step": 36702 + }, + { + "epoch": 4.81, + "learning_rate": 7.251370344784974e-06, + "loss": 0.0197, + "step": 36703 + }, + { + "epoch": 4.81, + "learning_rate": 7.250860449897642e-06, + "loss": 0.0195, + "step": 36704 + }, + { + "epoch": 4.81, + "learning_rate": 7.250350562742121e-06, + "loss": 0.0167, + "step": 36705 + }, + { + "epoch": 4.81, + "learning_rate": 7.249840683319842e-06, + "loss": 0.0181, + "step": 36706 + }, + { + "epoch": 4.81, + "learning_rate": 7.24933081163224e-06, + "loss": 0.0186, + "step": 36707 + }, + { + "epoch": 4.81, + "learning_rate": 7.248820947680748e-06, + "loss": 0.0172, + "step": 36708 + }, + { + "epoch": 4.81, + "learning_rate": 7.2483110914668e-06, + "loss": 0.0174, + "step": 36709 + }, + { + "epoch": 4.81, + "learning_rate": 7.2478012429918285e-06, + "loss": 0.0151, + "step": 36710 + }, + { + "epoch": 4.81, + "learning_rate": 7.2472914022572735e-06, + "loss": 0.0116, + "step": 36711 + }, + { + "epoch": 4.81, + "learning_rate": 7.2467815692645624e-06, + "loss": 0.0136, + "step": 36712 + }, + { + "epoch": 4.81, + "learning_rate": 7.246271744015134e-06, + "loss": 0.0124, + "step": 36713 + }, + { + "epoch": 4.81, + "learning_rate": 7.245761926510418e-06, + "loss": 0.0161, + "step": 36714 + }, + { + "epoch": 4.81, + "learning_rate": 7.245252116751852e-06, + "loss": 0.0153, + "step": 36715 + }, + { + "epoch": 4.81, + "learning_rate": 7.244742314740864e-06, + "loss": 0.0165, + "step": 36716 + }, + { + "epoch": 4.81, + "learning_rate": 7.244232520478893e-06, + "loss": 0.0148, + "step": 36717 + }, + { + "epoch": 4.81, + "learning_rate": 7.243722733967374e-06, + "loss": 0.0204, + "step": 36718 + }, + { + "epoch": 4.81, + "learning_rate": 7.243212955207735e-06, + "loss": 0.0228, + "step": 36719 + }, + { + "epoch": 4.81, + "learning_rate": 7.242703184201415e-06, + "loss": 0.0187, + "step": 36720 + }, + { + "epoch": 4.81, + "learning_rate": 7.242193420949841e-06, + "loss": 0.0238, + "step": 36721 + }, + { + "epoch": 4.81, + "learning_rate": 7.2416836654544535e-06, + "loss": 0.014, + "step": 36722 + }, + { + "epoch": 4.81, + "learning_rate": 7.241173917716684e-06, + "loss": 0.0176, + "step": 36723 + }, + { + "epoch": 4.81, + "learning_rate": 7.240664177737968e-06, + "loss": 0.0163, + "step": 36724 + }, + { + "epoch": 4.81, + "learning_rate": 7.240154445519734e-06, + "loss": 0.0159, + "step": 36725 + }, + { + "epoch": 4.81, + "learning_rate": 7.239644721063422e-06, + "loss": 0.0154, + "step": 36726 + }, + { + "epoch": 4.81, + "learning_rate": 7.2391350043704585e-06, + "loss": 0.0147, + "step": 36727 + }, + { + "epoch": 4.81, + "learning_rate": 7.2386252954422854e-06, + "loss": 0.0202, + "step": 36728 + }, + { + "epoch": 4.81, + "learning_rate": 7.238115594280327e-06, + "loss": 0.0184, + "step": 36729 + }, + { + "epoch": 4.81, + "learning_rate": 7.2376059008860225e-06, + "loss": 0.0145, + "step": 36730 + }, + { + "epoch": 4.81, + "learning_rate": 7.237096215260806e-06, + "loss": 0.0162, + "step": 36731 + }, + { + "epoch": 4.81, + "learning_rate": 7.236586537406106e-06, + "loss": 0.016, + "step": 36732 + }, + { + "epoch": 4.81, + "learning_rate": 7.236076867323361e-06, + "loss": 0.0157, + "step": 36733 + }, + { + "epoch": 4.81, + "learning_rate": 7.235567205014005e-06, + "loss": 0.0183, + "step": 36734 + }, + { + "epoch": 4.81, + "learning_rate": 7.2350575504794655e-06, + "loss": 0.0172, + "step": 36735 + }, + { + "epoch": 4.81, + "learning_rate": 7.234547903721183e-06, + "loss": 0.0168, + "step": 36736 + }, + { + "epoch": 4.81, + "learning_rate": 7.234038264740584e-06, + "loss": 0.0248, + "step": 36737 + }, + { + "epoch": 4.81, + "learning_rate": 7.233528633539106e-06, + "loss": 0.0181, + "step": 36738 + }, + { + "epoch": 4.81, + "learning_rate": 7.233019010118183e-06, + "loss": 0.0179, + "step": 36739 + }, + { + "epoch": 4.81, + "learning_rate": 7.232509394479244e-06, + "loss": 0.0193, + "step": 36740 + }, + { + "epoch": 4.81, + "learning_rate": 7.231999786623729e-06, + "loss": 0.0195, + "step": 36741 + }, + { + "epoch": 4.81, + "learning_rate": 7.231490186553063e-06, + "loss": 0.02, + "step": 36742 + }, + { + "epoch": 4.81, + "learning_rate": 7.230980594268682e-06, + "loss": 0.0194, + "step": 36743 + }, + { + "epoch": 4.81, + "learning_rate": 7.230471009772023e-06, + "loss": 0.0176, + "step": 36744 + }, + { + "epoch": 4.81, + "learning_rate": 7.229961433064518e-06, + "loss": 0.0211, + "step": 36745 + }, + { + "epoch": 4.81, + "learning_rate": 7.229451864147599e-06, + "loss": 0.0183, + "step": 36746 + }, + { + "epoch": 4.81, + "learning_rate": 7.228942303022699e-06, + "loss": 0.0176, + "step": 36747 + }, + { + "epoch": 4.81, + "learning_rate": 7.22843274969125e-06, + "loss": 0.0173, + "step": 36748 + }, + { + "epoch": 4.81, + "learning_rate": 7.227923204154689e-06, + "loss": 0.0197, + "step": 36749 + }, + { + "epoch": 4.81, + "learning_rate": 7.227413666414444e-06, + "loss": 0.0165, + "step": 36750 + }, + { + "epoch": 4.81, + "learning_rate": 7.2269041364719515e-06, + "loss": 0.0178, + "step": 36751 + }, + { + "epoch": 4.81, + "learning_rate": 7.226394614328642e-06, + "loss": 0.0213, + "step": 36752 + }, + { + "epoch": 4.81, + "learning_rate": 7.22588509998595e-06, + "loss": 0.0187, + "step": 36753 + }, + { + "epoch": 4.81, + "learning_rate": 7.225375593445305e-06, + "loss": 0.0169, + "step": 36754 + }, + { + "epoch": 4.81, + "learning_rate": 7.224866094708149e-06, + "loss": 0.0212, + "step": 36755 + }, + { + "epoch": 4.81, + "learning_rate": 7.2243566037759085e-06, + "loss": 0.0177, + "step": 36756 + }, + { + "epoch": 4.81, + "learning_rate": 7.223847120650017e-06, + "loss": 0.0175, + "step": 36757 + }, + { + "epoch": 4.82, + "learning_rate": 7.223337645331907e-06, + "loss": 0.0168, + "step": 36758 + }, + { + "epoch": 4.82, + "learning_rate": 7.222828177823012e-06, + "loss": 0.019, + "step": 36759 + }, + { + "epoch": 4.82, + "learning_rate": 7.222318718124765e-06, + "loss": 0.0191, + "step": 36760 + }, + { + "epoch": 4.82, + "learning_rate": 7.221809266238598e-06, + "loss": 0.0162, + "step": 36761 + }, + { + "epoch": 4.82, + "learning_rate": 7.221299822165947e-06, + "loss": 0.0201, + "step": 36762 + }, + { + "epoch": 4.82, + "learning_rate": 7.220790385908239e-06, + "loss": 0.0248, + "step": 36763 + }, + { + "epoch": 4.82, + "learning_rate": 7.2202809574669095e-06, + "loss": 0.0162, + "step": 36764 + }, + { + "epoch": 4.82, + "learning_rate": 7.219771536843394e-06, + "loss": 0.021, + "step": 36765 + }, + { + "epoch": 4.82, + "learning_rate": 7.2192621240391235e-06, + "loss": 0.0164, + "step": 36766 + }, + { + "epoch": 4.82, + "learning_rate": 7.2187527190555284e-06, + "loss": 0.0205, + "step": 36767 + }, + { + "epoch": 4.82, + "learning_rate": 7.2182433218940455e-06, + "loss": 0.0188, + "step": 36768 + }, + { + "epoch": 4.82, + "learning_rate": 7.217733932556104e-06, + "loss": 0.0213, + "step": 36769 + }, + { + "epoch": 4.82, + "learning_rate": 7.2172245510431385e-06, + "loss": 0.018, + "step": 36770 + }, + { + "epoch": 4.82, + "learning_rate": 7.216715177356578e-06, + "loss": 0.0139, + "step": 36771 + }, + { + "epoch": 4.82, + "learning_rate": 7.21620581149786e-06, + "loss": 0.0177, + "step": 36772 + }, + { + "epoch": 4.82, + "learning_rate": 7.215696453468415e-06, + "loss": 0.012, + "step": 36773 + }, + { + "epoch": 4.82, + "learning_rate": 7.215187103269674e-06, + "loss": 0.0203, + "step": 36774 + }, + { + "epoch": 4.82, + "learning_rate": 7.21467776090307e-06, + "loss": 0.0238, + "step": 36775 + }, + { + "epoch": 4.82, + "learning_rate": 7.214168426370038e-06, + "loss": 0.0176, + "step": 36776 + }, + { + "epoch": 4.82, + "learning_rate": 7.2136590996720106e-06, + "loss": 0.0184, + "step": 36777 + }, + { + "epoch": 4.82, + "learning_rate": 7.213149780810418e-06, + "loss": 0.0191, + "step": 36778 + }, + { + "epoch": 4.82, + "learning_rate": 7.2126404697866915e-06, + "loss": 0.0218, + "step": 36779 + }, + { + "epoch": 4.82, + "learning_rate": 7.2121311666022655e-06, + "loss": 0.0189, + "step": 36780 + }, + { + "epoch": 4.82, + "learning_rate": 7.211621871258574e-06, + "loss": 0.0148, + "step": 36781 + }, + { + "epoch": 4.82, + "learning_rate": 7.211112583757046e-06, + "loss": 0.0219, + "step": 36782 + }, + { + "epoch": 4.82, + "learning_rate": 7.2106033040991174e-06, + "loss": 0.0175, + "step": 36783 + }, + { + "epoch": 4.82, + "learning_rate": 7.210094032286217e-06, + "loss": 0.0177, + "step": 36784 + }, + { + "epoch": 4.82, + "learning_rate": 7.20958476831978e-06, + "loss": 0.0166, + "step": 36785 + }, + { + "epoch": 4.82, + "learning_rate": 7.2090755122012316e-06, + "loss": 0.0183, + "step": 36786 + }, + { + "epoch": 4.82, + "learning_rate": 7.208566263932013e-06, + "loss": 0.0211, + "step": 36787 + }, + { + "epoch": 4.82, + "learning_rate": 7.208057023513553e-06, + "loss": 0.0161, + "step": 36788 + }, + { + "epoch": 4.82, + "learning_rate": 7.207547790947287e-06, + "loss": 0.0174, + "step": 36789 + }, + { + "epoch": 4.82, + "learning_rate": 7.207038566234641e-06, + "loss": 0.0172, + "step": 36790 + }, + { + "epoch": 4.82, + "learning_rate": 7.206529349377052e-06, + "loss": 0.0139, + "step": 36791 + }, + { + "epoch": 4.82, + "learning_rate": 7.206020140375949e-06, + "loss": 0.0184, + "step": 36792 + }, + { + "epoch": 4.82, + "learning_rate": 7.205510939232768e-06, + "loss": 0.0189, + "step": 36793 + }, + { + "epoch": 4.82, + "learning_rate": 7.2050017459489364e-06, + "loss": 0.02, + "step": 36794 + }, + { + "epoch": 4.82, + "learning_rate": 7.204492560525888e-06, + "loss": 0.0151, + "step": 36795 + }, + { + "epoch": 4.82, + "learning_rate": 7.2039833829650586e-06, + "loss": 0.0201, + "step": 36796 + }, + { + "epoch": 4.82, + "learning_rate": 7.203474213267872e-06, + "loss": 0.0233, + "step": 36797 + }, + { + "epoch": 4.82, + "learning_rate": 7.202965051435766e-06, + "loss": 0.0234, + "step": 36798 + }, + { + "epoch": 4.82, + "learning_rate": 7.202455897470177e-06, + "loss": 0.0178, + "step": 36799 + }, + { + "epoch": 4.82, + "learning_rate": 7.201946751372527e-06, + "loss": 0.0156, + "step": 36800 + }, + { + "epoch": 4.82, + "learning_rate": 7.201437613144257e-06, + "loss": 0.0202, + "step": 36801 + }, + { + "epoch": 4.82, + "learning_rate": 7.200928482786791e-06, + "loss": 0.0237, + "step": 36802 + }, + { + "epoch": 4.82, + "learning_rate": 7.2004193603015645e-06, + "loss": 0.0174, + "step": 36803 + }, + { + "epoch": 4.82, + "learning_rate": 7.199910245690013e-06, + "loss": 0.0196, + "step": 36804 + }, + { + "epoch": 4.82, + "learning_rate": 7.1994011389535615e-06, + "loss": 0.0148, + "step": 36805 + }, + { + "epoch": 4.82, + "learning_rate": 7.198892040093648e-06, + "loss": 0.021, + "step": 36806 + }, + { + "epoch": 4.82, + "learning_rate": 7.1983829491116985e-06, + "loss": 0.0162, + "step": 36807 + }, + { + "epoch": 4.82, + "learning_rate": 7.197873866009146e-06, + "loss": 0.0167, + "step": 36808 + }, + { + "epoch": 4.82, + "learning_rate": 7.197364790787427e-06, + "loss": 0.0227, + "step": 36809 + }, + { + "epoch": 4.82, + "learning_rate": 7.19685572344797e-06, + "loss": 0.0208, + "step": 36810 + }, + { + "epoch": 4.82, + "learning_rate": 7.196346663992209e-06, + "loss": 0.0161, + "step": 36811 + }, + { + "epoch": 4.82, + "learning_rate": 7.195837612421572e-06, + "loss": 0.0162, + "step": 36812 + }, + { + "epoch": 4.82, + "learning_rate": 7.195328568737492e-06, + "loss": 0.0214, + "step": 36813 + }, + { + "epoch": 4.82, + "learning_rate": 7.194819532941403e-06, + "loss": 0.0192, + "step": 36814 + }, + { + "epoch": 4.82, + "learning_rate": 7.194310505034732e-06, + "loss": 0.0215, + "step": 36815 + }, + { + "epoch": 4.82, + "learning_rate": 7.193801485018917e-06, + "loss": 0.0143, + "step": 36816 + }, + { + "epoch": 4.82, + "learning_rate": 7.193292472895382e-06, + "loss": 0.0161, + "step": 36817 + }, + { + "epoch": 4.82, + "learning_rate": 7.192783468665563e-06, + "loss": 0.0162, + "step": 36818 + }, + { + "epoch": 4.82, + "learning_rate": 7.192274472330889e-06, + "loss": 0.0215, + "step": 36819 + }, + { + "epoch": 4.82, + "learning_rate": 7.191765483892798e-06, + "loss": 0.0171, + "step": 36820 + }, + { + "epoch": 4.82, + "learning_rate": 7.191256503352713e-06, + "loss": 0.015, + "step": 36821 + }, + { + "epoch": 4.82, + "learning_rate": 7.190747530712074e-06, + "loss": 0.0185, + "step": 36822 + }, + { + "epoch": 4.82, + "learning_rate": 7.190238565972305e-06, + "loss": 0.0172, + "step": 36823 + }, + { + "epoch": 4.82, + "learning_rate": 7.189729609134841e-06, + "loss": 0.0155, + "step": 36824 + }, + { + "epoch": 4.82, + "learning_rate": 7.189220660201115e-06, + "loss": 0.0227, + "step": 36825 + }, + { + "epoch": 4.82, + "learning_rate": 7.188711719172552e-06, + "loss": 0.011, + "step": 36826 + }, + { + "epoch": 4.82, + "learning_rate": 7.188202786050592e-06, + "loss": 0.0182, + "step": 36827 + }, + { + "epoch": 4.82, + "learning_rate": 7.1876938608366585e-06, + "loss": 0.0169, + "step": 36828 + }, + { + "epoch": 4.82, + "learning_rate": 7.18718494353219e-06, + "loss": 0.0163, + "step": 36829 + }, + { + "epoch": 4.82, + "learning_rate": 7.186676034138606e-06, + "loss": 0.0177, + "step": 36830 + }, + { + "epoch": 4.82, + "learning_rate": 7.186167132657353e-06, + "loss": 0.0198, + "step": 36831 + }, + { + "epoch": 4.82, + "learning_rate": 7.185658239089852e-06, + "loss": 0.0159, + "step": 36832 + }, + { + "epoch": 4.82, + "learning_rate": 7.185149353437541e-06, + "loss": 0.0179, + "step": 36833 + }, + { + "epoch": 4.82, + "learning_rate": 7.184640475701844e-06, + "loss": 0.0226, + "step": 36834 + }, + { + "epoch": 4.83, + "learning_rate": 7.1841316058842e-06, + "loss": 0.0154, + "step": 36835 + }, + { + "epoch": 4.83, + "learning_rate": 7.18362274398603e-06, + "loss": 0.018, + "step": 36836 + }, + { + "epoch": 4.83, + "learning_rate": 7.183113890008777e-06, + "loss": 0.0149, + "step": 36837 + }, + { + "epoch": 4.83, + "learning_rate": 7.182605043953862e-06, + "loss": 0.0165, + "step": 36838 + }, + { + "epoch": 4.83, + "learning_rate": 7.1820962058227205e-06, + "loss": 0.0206, + "step": 36839 + }, + { + "epoch": 4.83, + "learning_rate": 7.181587375616785e-06, + "loss": 0.015, + "step": 36840 + }, + { + "epoch": 4.83, + "learning_rate": 7.181078553337482e-06, + "loss": 0.0185, + "step": 36841 + }, + { + "epoch": 4.83, + "learning_rate": 7.1805697389862475e-06, + "loss": 0.0208, + "step": 36842 + }, + { + "epoch": 4.83, + "learning_rate": 7.180060932564512e-06, + "loss": 0.0186, + "step": 36843 + }, + { + "epoch": 4.83, + "learning_rate": 7.179552134073703e-06, + "loss": 0.0178, + "step": 36844 + }, + { + "epoch": 4.83, + "learning_rate": 7.179043343515253e-06, + "loss": 0.0208, + "step": 36845 + }, + { + "epoch": 4.83, + "learning_rate": 7.178534560890596e-06, + "loss": 0.0214, + "step": 36846 + }, + { + "epoch": 4.83, + "learning_rate": 7.178025786201158e-06, + "loss": 0.0227, + "step": 36847 + }, + { + "epoch": 4.83, + "learning_rate": 7.177517019448375e-06, + "loss": 0.0211, + "step": 36848 + }, + { + "epoch": 4.83, + "learning_rate": 7.177008260633671e-06, + "loss": 0.0156, + "step": 36849 + }, + { + "epoch": 4.83, + "learning_rate": 7.1764995097584854e-06, + "loss": 0.0149, + "step": 36850 + }, + { + "epoch": 4.83, + "learning_rate": 7.175990766824241e-06, + "loss": 0.0137, + "step": 36851 + }, + { + "epoch": 4.83, + "learning_rate": 7.175482031832371e-06, + "loss": 0.0183, + "step": 36852 + }, + { + "epoch": 4.83, + "learning_rate": 7.174973304784308e-06, + "loss": 0.0215, + "step": 36853 + }, + { + "epoch": 4.83, + "learning_rate": 7.174464585681486e-06, + "loss": 0.014, + "step": 36854 + }, + { + "epoch": 4.83, + "learning_rate": 7.173955874525329e-06, + "loss": 0.0151, + "step": 36855 + }, + { + "epoch": 4.83, + "learning_rate": 7.173447171317271e-06, + "loss": 0.0137, + "step": 36856 + }, + { + "epoch": 4.83, + "learning_rate": 7.172938476058743e-06, + "loss": 0.0164, + "step": 36857 + }, + { + "epoch": 4.83, + "learning_rate": 7.172429788751176e-06, + "loss": 0.0143, + "step": 36858 + }, + { + "epoch": 4.83, + "learning_rate": 7.171921109395997e-06, + "loss": 0.0217, + "step": 36859 + }, + { + "epoch": 4.83, + "learning_rate": 7.17141243799464e-06, + "loss": 0.0146, + "step": 36860 + }, + { + "epoch": 4.83, + "learning_rate": 7.170903774548537e-06, + "loss": 0.0172, + "step": 36861 + }, + { + "epoch": 4.83, + "learning_rate": 7.170395119059115e-06, + "loss": 0.0133, + "step": 36862 + }, + { + "epoch": 4.83, + "learning_rate": 7.169886471527803e-06, + "loss": 0.0186, + "step": 36863 + }, + { + "epoch": 4.83, + "learning_rate": 7.169377831956038e-06, + "loss": 0.0167, + "step": 36864 + }, + { + "epoch": 4.83, + "learning_rate": 7.168869200345246e-06, + "loss": 0.02, + "step": 36865 + }, + { + "epoch": 4.83, + "learning_rate": 7.16836057669686e-06, + "loss": 0.0158, + "step": 36866 + }, + { + "epoch": 4.83, + "learning_rate": 7.167851961012308e-06, + "loss": 0.0204, + "step": 36867 + }, + { + "epoch": 4.83, + "learning_rate": 7.167343353293022e-06, + "loss": 0.0194, + "step": 36868 + }, + { + "epoch": 4.83, + "learning_rate": 7.166834753540432e-06, + "loss": 0.0199, + "step": 36869 + }, + { + "epoch": 4.83, + "learning_rate": 7.166326161755967e-06, + "loss": 0.0178, + "step": 36870 + }, + { + "epoch": 4.83, + "learning_rate": 7.165817577941062e-06, + "loss": 0.0205, + "step": 36871 + }, + { + "epoch": 4.83, + "learning_rate": 7.165309002097142e-06, + "loss": 0.018, + "step": 36872 + }, + { + "epoch": 4.83, + "learning_rate": 7.16480043422564e-06, + "loss": 0.0154, + "step": 36873 + }, + { + "epoch": 4.83, + "learning_rate": 7.164291874327981e-06, + "loss": 0.0121, + "step": 36874 + }, + { + "epoch": 4.83, + "learning_rate": 7.163783322405606e-06, + "loss": 0.0208, + "step": 36875 + }, + { + "epoch": 4.83, + "learning_rate": 7.163274778459937e-06, + "loss": 0.0147, + "step": 36876 + }, + { + "epoch": 4.83, + "learning_rate": 7.1627662424924095e-06, + "loss": 0.0197, + "step": 36877 + }, + { + "epoch": 4.83, + "learning_rate": 7.162257714504448e-06, + "loss": 0.0183, + "step": 36878 + }, + { + "epoch": 4.83, + "learning_rate": 7.161749194497487e-06, + "loss": 0.0154, + "step": 36879 + }, + { + "epoch": 4.83, + "learning_rate": 7.1612406824729545e-06, + "loss": 0.0175, + "step": 36880 + }, + { + "epoch": 4.83, + "learning_rate": 7.16073217843228e-06, + "loss": 0.0166, + "step": 36881 + }, + { + "epoch": 4.83, + "learning_rate": 7.160223682376899e-06, + "loss": 0.0148, + "step": 36882 + }, + { + "epoch": 4.83, + "learning_rate": 7.1597151943082345e-06, + "loss": 0.0147, + "step": 36883 + }, + { + "epoch": 4.83, + "learning_rate": 7.159206714227721e-06, + "loss": 0.0172, + "step": 36884 + }, + { + "epoch": 4.83, + "learning_rate": 7.158698242136783e-06, + "loss": 0.0199, + "step": 36885 + }, + { + "epoch": 4.83, + "learning_rate": 7.158189778036859e-06, + "loss": 0.0132, + "step": 36886 + }, + { + "epoch": 4.83, + "learning_rate": 7.157681321929376e-06, + "loss": 0.0168, + "step": 36887 + }, + { + "epoch": 4.83, + "learning_rate": 7.1571728738157605e-06, + "loss": 0.0166, + "step": 36888 + }, + { + "epoch": 4.83, + "learning_rate": 7.156664433697446e-06, + "loss": 0.019, + "step": 36889 + }, + { + "epoch": 4.83, + "learning_rate": 7.156156001575862e-06, + "loss": 0.0166, + "step": 36890 + }, + { + "epoch": 4.83, + "learning_rate": 7.155647577452436e-06, + "loss": 0.0167, + "step": 36891 + }, + { + "epoch": 4.83, + "learning_rate": 7.1551391613286034e-06, + "loss": 0.0129, + "step": 36892 + }, + { + "epoch": 4.83, + "learning_rate": 7.154630753205787e-06, + "loss": 0.0162, + "step": 36893 + }, + { + "epoch": 4.83, + "learning_rate": 7.154122353085422e-06, + "loss": 0.0178, + "step": 36894 + }, + { + "epoch": 4.83, + "learning_rate": 7.1536139609689305e-06, + "loss": 0.019, + "step": 36895 + }, + { + "epoch": 4.83, + "learning_rate": 7.153105576857754e-06, + "loss": 0.0222, + "step": 36896 + }, + { + "epoch": 4.83, + "learning_rate": 7.1525972007533155e-06, + "loss": 0.0124, + "step": 36897 + }, + { + "epoch": 4.83, + "learning_rate": 7.152088832657046e-06, + "loss": 0.0148, + "step": 36898 + }, + { + "epoch": 4.83, + "learning_rate": 7.151580472570374e-06, + "loss": 0.0182, + "step": 36899 + }, + { + "epoch": 4.83, + "learning_rate": 7.151072120494732e-06, + "loss": 0.0202, + "step": 36900 + }, + { + "epoch": 4.83, + "learning_rate": 7.150563776431546e-06, + "loss": 0.0146, + "step": 36901 + }, + { + "epoch": 4.83, + "learning_rate": 7.1500554403822495e-06, + "loss": 0.0188, + "step": 36902 + }, + { + "epoch": 4.83, + "learning_rate": 7.14954711234827e-06, + "loss": 0.0222, + "step": 36903 + }, + { + "epoch": 4.83, + "learning_rate": 7.1490387923310355e-06, + "loss": 0.0161, + "step": 36904 + }, + { + "epoch": 4.83, + "learning_rate": 7.148530480331979e-06, + "loss": 0.0159, + "step": 36905 + }, + { + "epoch": 4.83, + "learning_rate": 7.148022176352526e-06, + "loss": 0.0176, + "step": 36906 + }, + { + "epoch": 4.83, + "learning_rate": 7.14751388039411e-06, + "loss": 0.0197, + "step": 36907 + }, + { + "epoch": 4.83, + "learning_rate": 7.147005592458163e-06, + "loss": 0.02, + "step": 36908 + }, + { + "epoch": 4.83, + "learning_rate": 7.146497312546108e-06, + "loss": 0.0102, + "step": 36909 + }, + { + "epoch": 4.83, + "learning_rate": 7.145989040659376e-06, + "loss": 0.0177, + "step": 36910 + }, + { + "epoch": 4.84, + "learning_rate": 7.1454807767994e-06, + "loss": 0.0158, + "step": 36911 + }, + { + "epoch": 4.84, + "learning_rate": 7.144972520967607e-06, + "loss": 0.0159, + "step": 36912 + }, + { + "epoch": 4.84, + "learning_rate": 7.1444642731654275e-06, + "loss": 0.0195, + "step": 36913 + }, + { + "epoch": 4.84, + "learning_rate": 7.143956033394289e-06, + "loss": 0.0145, + "step": 36914 + }, + { + "epoch": 4.84, + "learning_rate": 7.143447801655623e-06, + "loss": 0.0183, + "step": 36915 + }, + { + "epoch": 4.84, + "learning_rate": 7.142939577950857e-06, + "loss": 0.0167, + "step": 36916 + }, + { + "epoch": 4.84, + "learning_rate": 7.142431362281419e-06, + "loss": 0.0176, + "step": 36917 + }, + { + "epoch": 4.84, + "learning_rate": 7.141923154648742e-06, + "loss": 0.021, + "step": 36918 + }, + { + "epoch": 4.84, + "learning_rate": 7.141414955054257e-06, + "loss": 0.019, + "step": 36919 + }, + { + "epoch": 4.84, + "learning_rate": 7.140906763499387e-06, + "loss": 0.0141, + "step": 36920 + }, + { + "epoch": 4.84, + "learning_rate": 7.140398579985568e-06, + "loss": 0.0162, + "step": 36921 + }, + { + "epoch": 4.84, + "learning_rate": 7.139890404514222e-06, + "loss": 0.0175, + "step": 36922 + }, + { + "epoch": 4.84, + "learning_rate": 7.139382237086785e-06, + "loss": 0.0207, + "step": 36923 + }, + { + "epoch": 4.84, + "learning_rate": 7.138874077704681e-06, + "loss": 0.0198, + "step": 36924 + }, + { + "epoch": 4.84, + "learning_rate": 7.138365926369341e-06, + "loss": 0.0155, + "step": 36925 + }, + { + "epoch": 4.84, + "learning_rate": 7.137857783082196e-06, + "loss": 0.023, + "step": 36926 + }, + { + "epoch": 4.84, + "learning_rate": 7.137349647844673e-06, + "loss": 0.0141, + "step": 36927 + }, + { + "epoch": 4.84, + "learning_rate": 7.136841520658198e-06, + "loss": 0.0208, + "step": 36928 + }, + { + "epoch": 4.84, + "learning_rate": 7.136333401524209e-06, + "loss": 0.0175, + "step": 36929 + }, + { + "epoch": 4.84, + "learning_rate": 7.135825290444127e-06, + "loss": 0.0175, + "step": 36930 + }, + { + "epoch": 4.84, + "learning_rate": 7.135317187419384e-06, + "loss": 0.0169, + "step": 36931 + }, + { + "epoch": 4.84, + "learning_rate": 7.134809092451411e-06, + "loss": 0.0157, + "step": 36932 + }, + { + "epoch": 4.84, + "learning_rate": 7.1343010055416324e-06, + "loss": 0.0211, + "step": 36933 + }, + { + "epoch": 4.84, + "learning_rate": 7.133792926691483e-06, + "loss": 0.0194, + "step": 36934 + }, + { + "epoch": 4.84, + "learning_rate": 7.133284855902385e-06, + "loss": 0.0172, + "step": 36935 + }, + { + "epoch": 4.84, + "learning_rate": 7.132776793175773e-06, + "loss": 0.0175, + "step": 36936 + }, + { + "epoch": 4.84, + "learning_rate": 7.132268738513073e-06, + "loss": 0.0152, + "step": 36937 + }, + { + "epoch": 4.84, + "learning_rate": 7.1317606919157145e-06, + "loss": 0.0183, + "step": 36938 + }, + { + "epoch": 4.84, + "learning_rate": 7.131252653385121e-06, + "loss": 0.0176, + "step": 36939 + }, + { + "epoch": 4.84, + "learning_rate": 7.130744622922732e-06, + "loss": 0.0158, + "step": 36940 + }, + { + "epoch": 4.84, + "learning_rate": 7.13023660052997e-06, + "loss": 0.0139, + "step": 36941 + }, + { + "epoch": 4.84, + "learning_rate": 7.129728586208267e-06, + "loss": 0.0168, + "step": 36942 + }, + { + "epoch": 4.84, + "learning_rate": 7.129220579959047e-06, + "loss": 0.0185, + "step": 36943 + }, + { + "epoch": 4.84, + "learning_rate": 7.1287125817837425e-06, + "loss": 0.018, + "step": 36944 + }, + { + "epoch": 4.84, + "learning_rate": 7.128204591683779e-06, + "loss": 0.0163, + "step": 36945 + }, + { + "epoch": 4.84, + "learning_rate": 7.1276966096605885e-06, + "loss": 0.0156, + "step": 36946 + }, + { + "epoch": 4.84, + "learning_rate": 7.1271886357155985e-06, + "loss": 0.0162, + "step": 36947 + }, + { + "epoch": 4.84, + "learning_rate": 7.126680669850236e-06, + "loss": 0.0181, + "step": 36948 + }, + { + "epoch": 4.84, + "learning_rate": 7.126172712065935e-06, + "loss": 0.0205, + "step": 36949 + }, + { + "epoch": 4.84, + "learning_rate": 7.125664762364115e-06, + "loss": 0.0131, + "step": 36950 + }, + { + "epoch": 4.84, + "learning_rate": 7.125156820746211e-06, + "loss": 0.0138, + "step": 36951 + }, + { + "epoch": 4.84, + "learning_rate": 7.124648887213652e-06, + "loss": 0.0164, + "step": 36952 + }, + { + "epoch": 4.84, + "learning_rate": 7.124140961767865e-06, + "loss": 0.0169, + "step": 36953 + }, + { + "epoch": 4.84, + "learning_rate": 7.123633044410276e-06, + "loss": 0.0155, + "step": 36954 + }, + { + "epoch": 4.84, + "learning_rate": 7.123125135142319e-06, + "loss": 0.0145, + "step": 36955 + }, + { + "epoch": 4.84, + "learning_rate": 7.122617233965419e-06, + "loss": 0.0179, + "step": 36956 + }, + { + "epoch": 4.84, + "learning_rate": 7.122109340881005e-06, + "loss": 0.0184, + "step": 36957 + }, + { + "epoch": 4.84, + "learning_rate": 7.121601455890503e-06, + "loss": 0.0162, + "step": 36958 + }, + { + "epoch": 4.84, + "learning_rate": 7.121093578995346e-06, + "loss": 0.0172, + "step": 36959 + }, + { + "epoch": 4.84, + "learning_rate": 7.120585710196959e-06, + "loss": 0.0158, + "step": 36960 + }, + { + "epoch": 4.84, + "learning_rate": 7.120077849496767e-06, + "loss": 0.018, + "step": 36961 + }, + { + "epoch": 4.84, + "learning_rate": 7.119569996896207e-06, + "loss": 0.0182, + "step": 36962 + }, + { + "epoch": 4.84, + "learning_rate": 7.119062152396704e-06, + "loss": 0.0186, + "step": 36963 + }, + { + "epoch": 4.84, + "learning_rate": 7.118554315999684e-06, + "loss": 0.0171, + "step": 36964 + }, + { + "epoch": 4.84, + "learning_rate": 7.118046487706578e-06, + "loss": 0.0153, + "step": 36965 + }, + { + "epoch": 4.84, + "learning_rate": 7.117538667518811e-06, + "loss": 0.0206, + "step": 36966 + }, + { + "epoch": 4.84, + "learning_rate": 7.117030855437814e-06, + "loss": 0.0154, + "step": 36967 + }, + { + "epoch": 4.84, + "learning_rate": 7.116523051465012e-06, + "loss": 0.0162, + "step": 36968 + }, + { + "epoch": 4.84, + "learning_rate": 7.116015255601837e-06, + "loss": 0.0242, + "step": 36969 + }, + { + "epoch": 4.84, + "learning_rate": 7.115507467849716e-06, + "loss": 0.0174, + "step": 36970 + }, + { + "epoch": 4.84, + "learning_rate": 7.114999688210075e-06, + "loss": 0.0203, + "step": 36971 + }, + { + "epoch": 4.84, + "learning_rate": 7.114491916684343e-06, + "loss": 0.0112, + "step": 36972 + }, + { + "epoch": 4.84, + "learning_rate": 7.113984153273952e-06, + "loss": 0.0188, + "step": 36973 + }, + { + "epoch": 4.84, + "learning_rate": 7.113476397980326e-06, + "loss": 0.0176, + "step": 36974 + }, + { + "epoch": 4.84, + "learning_rate": 7.112968650804892e-06, + "loss": 0.0185, + "step": 36975 + }, + { + "epoch": 4.84, + "learning_rate": 7.112460911749083e-06, + "loss": 0.0142, + "step": 36976 + }, + { + "epoch": 4.84, + "learning_rate": 7.111953180814321e-06, + "loss": 0.0168, + "step": 36977 + }, + { + "epoch": 4.84, + "learning_rate": 7.111445458002039e-06, + "loss": 0.0217, + "step": 36978 + }, + { + "epoch": 4.84, + "learning_rate": 7.1109377433136616e-06, + "loss": 0.0193, + "step": 36979 + }, + { + "epoch": 4.84, + "learning_rate": 7.110430036750619e-06, + "loss": 0.0186, + "step": 36980 + }, + { + "epoch": 4.84, + "learning_rate": 7.109922338314338e-06, + "loss": 0.0134, + "step": 36981 + }, + { + "epoch": 4.84, + "learning_rate": 7.1094146480062476e-06, + "loss": 0.0122, + "step": 36982 + }, + { + "epoch": 4.84, + "learning_rate": 7.108906965827768e-06, + "loss": 0.014, + "step": 36983 + }, + { + "epoch": 4.84, + "learning_rate": 7.108399291780339e-06, + "loss": 0.0172, + "step": 36984 + }, + { + "epoch": 4.84, + "learning_rate": 7.107891625865382e-06, + "loss": 0.0135, + "step": 36985 + }, + { + "epoch": 4.84, + "learning_rate": 7.107383968084327e-06, + "loss": 0.0179, + "step": 36986 + }, + { + "epoch": 4.85, + "learning_rate": 7.1068763184386e-06, + "loss": 0.0197, + "step": 36987 + }, + { + "epoch": 4.85, + "learning_rate": 7.10636867692963e-06, + "loss": 0.0152, + "step": 36988 + }, + { + "epoch": 4.85, + "learning_rate": 7.105861043558843e-06, + "loss": 0.0184, + "step": 36989 + }, + { + "epoch": 4.85, + "learning_rate": 7.105353418327667e-06, + "loss": 0.0175, + "step": 36990 + }, + { + "epoch": 4.85, + "learning_rate": 7.104845801237532e-06, + "loss": 0.0165, + "step": 36991 + }, + { + "epoch": 4.85, + "learning_rate": 7.1043381922898635e-06, + "loss": 0.0179, + "step": 36992 + }, + { + "epoch": 4.85, + "learning_rate": 7.1038305914860895e-06, + "loss": 0.0145, + "step": 36993 + }, + { + "epoch": 4.85, + "learning_rate": 7.103322998827635e-06, + "loss": 0.0142, + "step": 36994 + }, + { + "epoch": 4.85, + "learning_rate": 7.1028154143159325e-06, + "loss": 0.0186, + "step": 36995 + }, + { + "epoch": 4.85, + "learning_rate": 7.1023078379524065e-06, + "loss": 0.019, + "step": 36996 + }, + { + "epoch": 4.85, + "learning_rate": 7.101800269738488e-06, + "loss": 0.0184, + "step": 36997 + }, + { + "epoch": 4.85, + "learning_rate": 7.1012927096756e-06, + "loss": 0.0173, + "step": 36998 + }, + { + "epoch": 4.85, + "learning_rate": 7.100785157765174e-06, + "loss": 0.019, + "step": 36999 + }, + { + "epoch": 4.85, + "learning_rate": 7.100277614008633e-06, + "loss": 0.0178, + "step": 37000 + }, + { + "epoch": 4.85, + "learning_rate": 7.099770078407409e-06, + "loss": 0.0148, + "step": 37001 + }, + { + "epoch": 4.85, + "learning_rate": 7.099262550962925e-06, + "loss": 0.016, + "step": 37002 + }, + { + "epoch": 4.85, + "learning_rate": 7.098755031676613e-06, + "loss": 0.0163, + "step": 37003 + }, + { + "epoch": 4.85, + "learning_rate": 7.098247520549894e-06, + "loss": 0.0158, + "step": 37004 + }, + { + "epoch": 4.85, + "learning_rate": 7.097740017584201e-06, + "loss": 0.0207, + "step": 37005 + }, + { + "epoch": 4.85, + "learning_rate": 7.097232522780959e-06, + "loss": 0.0146, + "step": 37006 + }, + { + "epoch": 4.85, + "learning_rate": 7.0967250361415984e-06, + "loss": 0.0177, + "step": 37007 + }, + { + "epoch": 4.85, + "learning_rate": 7.0962175576675416e-06, + "loss": 0.0147, + "step": 37008 + }, + { + "epoch": 4.85, + "learning_rate": 7.0957100873602215e-06, + "loss": 0.018, + "step": 37009 + }, + { + "epoch": 4.85, + "learning_rate": 7.0952026252210595e-06, + "loss": 0.0169, + "step": 37010 + }, + { + "epoch": 4.85, + "learning_rate": 7.094695171251485e-06, + "loss": 0.0182, + "step": 37011 + }, + { + "epoch": 4.85, + "learning_rate": 7.094187725452928e-06, + "loss": 0.0205, + "step": 37012 + }, + { + "epoch": 4.85, + "learning_rate": 7.0936802878268114e-06, + "loss": 0.0157, + "step": 37013 + }, + { + "epoch": 4.85, + "learning_rate": 7.093172858374566e-06, + "loss": 0.0156, + "step": 37014 + }, + { + "epoch": 4.85, + "learning_rate": 7.092665437097615e-06, + "loss": 0.0153, + "step": 37015 + }, + { + "epoch": 4.85, + "learning_rate": 7.092158023997387e-06, + "loss": 0.0177, + "step": 37016 + }, + { + "epoch": 4.85, + "learning_rate": 7.091650619075312e-06, + "loss": 0.0204, + "step": 37017 + }, + { + "epoch": 4.85, + "learning_rate": 7.091143222332814e-06, + "loss": 0.0158, + "step": 37018 + }, + { + "epoch": 4.85, + "learning_rate": 7.09063583377132e-06, + "loss": 0.0147, + "step": 37019 + }, + { + "epoch": 4.85, + "learning_rate": 7.09012845339226e-06, + "loss": 0.0142, + "step": 37020 + }, + { + "epoch": 4.85, + "learning_rate": 7.089621081197056e-06, + "loss": 0.0165, + "step": 37021 + }, + { + "epoch": 4.85, + "learning_rate": 7.08911371718714e-06, + "loss": 0.0135, + "step": 37022 + }, + { + "epoch": 4.85, + "learning_rate": 7.088606361363935e-06, + "loss": 0.0185, + "step": 37023 + }, + { + "epoch": 4.85, + "learning_rate": 7.088099013728873e-06, + "loss": 0.0148, + "step": 37024 + }, + { + "epoch": 4.85, + "learning_rate": 7.087591674283374e-06, + "loss": 0.0157, + "step": 37025 + }, + { + "epoch": 4.85, + "learning_rate": 7.087084343028865e-06, + "loss": 0.0219, + "step": 37026 + }, + { + "epoch": 4.85, + "learning_rate": 7.08657701996678e-06, + "loss": 0.0217, + "step": 37027 + }, + { + "epoch": 4.85, + "learning_rate": 7.086069705098544e-06, + "loss": 0.0191, + "step": 37028 + }, + { + "epoch": 4.85, + "learning_rate": 7.08556239842558e-06, + "loss": 0.0208, + "step": 37029 + }, + { + "epoch": 4.85, + "learning_rate": 7.085055099949317e-06, + "loss": 0.023, + "step": 37030 + }, + { + "epoch": 4.85, + "learning_rate": 7.08454780967118e-06, + "loss": 0.0172, + "step": 37031 + }, + { + "epoch": 4.85, + "learning_rate": 7.084040527592599e-06, + "loss": 0.0222, + "step": 37032 + }, + { + "epoch": 4.85, + "learning_rate": 7.083533253714998e-06, + "loss": 0.0204, + "step": 37033 + }, + { + "epoch": 4.85, + "learning_rate": 7.0830259880398025e-06, + "loss": 0.02, + "step": 37034 + }, + { + "epoch": 4.85, + "learning_rate": 7.082518730568444e-06, + "loss": 0.0176, + "step": 37035 + }, + { + "epoch": 4.85, + "learning_rate": 7.082011481302345e-06, + "loss": 0.0143, + "step": 37036 + }, + { + "epoch": 4.85, + "learning_rate": 7.081504240242929e-06, + "loss": 0.0208, + "step": 37037 + }, + { + "epoch": 4.85, + "learning_rate": 7.080997007391634e-06, + "loss": 0.0215, + "step": 37038 + }, + { + "epoch": 4.85, + "learning_rate": 7.0804897827498755e-06, + "loss": 0.0183, + "step": 37039 + }, + { + "epoch": 4.85, + "learning_rate": 7.079982566319084e-06, + "loss": 0.0131, + "step": 37040 + }, + { + "epoch": 4.85, + "learning_rate": 7.079475358100688e-06, + "loss": 0.0184, + "step": 37041 + }, + { + "epoch": 4.85, + "learning_rate": 7.078968158096111e-06, + "loss": 0.0183, + "step": 37042 + }, + { + "epoch": 4.85, + "learning_rate": 7.078460966306782e-06, + "loss": 0.0196, + "step": 37043 + }, + { + "epoch": 4.85, + "learning_rate": 7.0779537827341235e-06, + "loss": 0.0214, + "step": 37044 + }, + { + "epoch": 4.85, + "learning_rate": 7.077446607379567e-06, + "loss": 0.0175, + "step": 37045 + }, + { + "epoch": 4.85, + "learning_rate": 7.076939440244533e-06, + "loss": 0.0215, + "step": 37046 + }, + { + "epoch": 4.85, + "learning_rate": 7.076432281330453e-06, + "loss": 0.0182, + "step": 37047 + }, + { + "epoch": 4.85, + "learning_rate": 7.075925130638748e-06, + "loss": 0.0183, + "step": 37048 + }, + { + "epoch": 4.85, + "learning_rate": 7.075417988170853e-06, + "loss": 0.0154, + "step": 37049 + }, + { + "epoch": 4.85, + "learning_rate": 7.074910853928187e-06, + "loss": 0.02, + "step": 37050 + }, + { + "epoch": 4.85, + "learning_rate": 7.07440372791218e-06, + "loss": 0.0176, + "step": 37051 + }, + { + "epoch": 4.85, + "learning_rate": 7.073896610124256e-06, + "loss": 0.0172, + "step": 37052 + }, + { + "epoch": 4.85, + "learning_rate": 7.073389500565841e-06, + "loss": 0.0167, + "step": 37053 + }, + { + "epoch": 4.85, + "learning_rate": 7.072882399238363e-06, + "loss": 0.0197, + "step": 37054 + }, + { + "epoch": 4.85, + "learning_rate": 7.072375306143246e-06, + "loss": 0.0145, + "step": 37055 + }, + { + "epoch": 4.85, + "learning_rate": 7.07186822128192e-06, + "loss": 0.0165, + "step": 37056 + }, + { + "epoch": 4.85, + "learning_rate": 7.071361144655806e-06, + "loss": 0.0171, + "step": 37057 + }, + { + "epoch": 4.85, + "learning_rate": 7.070854076266336e-06, + "loss": 0.016, + "step": 37058 + }, + { + "epoch": 4.85, + "learning_rate": 7.070347016114929e-06, + "loss": 0.0153, + "step": 37059 + }, + { + "epoch": 4.85, + "learning_rate": 7.069839964203016e-06, + "loss": 0.0181, + "step": 37060 + }, + { + "epoch": 4.85, + "learning_rate": 7.069332920532022e-06, + "loss": 0.0173, + "step": 37061 + }, + { + "epoch": 4.85, + "learning_rate": 7.068825885103377e-06, + "loss": 0.0145, + "step": 37062 + }, + { + "epoch": 4.85, + "learning_rate": 7.0683188579184995e-06, + "loss": 0.0167, + "step": 37063 + }, + { + "epoch": 4.86, + "learning_rate": 7.06781183897882e-06, + "loss": 0.0214, + "step": 37064 + }, + { + "epoch": 4.86, + "learning_rate": 7.067304828285764e-06, + "loss": 0.0194, + "step": 37065 + }, + { + "epoch": 4.86, + "learning_rate": 7.066797825840756e-06, + "loss": 0.0155, + "step": 37066 + }, + { + "epoch": 4.86, + "learning_rate": 7.066290831645224e-06, + "loss": 0.0231, + "step": 37067 + }, + { + "epoch": 4.86, + "learning_rate": 7.0657838457005936e-06, + "loss": 0.0177, + "step": 37068 + }, + { + "epoch": 4.86, + "learning_rate": 7.065276868008288e-06, + "loss": 0.0169, + "step": 37069 + }, + { + "epoch": 4.86, + "learning_rate": 7.0647698985697325e-06, + "loss": 0.0201, + "step": 37070 + }, + { + "epoch": 4.86, + "learning_rate": 7.064262937386357e-06, + "loss": 0.0177, + "step": 37071 + }, + { + "epoch": 4.86, + "learning_rate": 7.063755984459588e-06, + "loss": 0.02, + "step": 37072 + }, + { + "epoch": 4.86, + "learning_rate": 7.063249039790847e-06, + "loss": 0.0158, + "step": 37073 + }, + { + "epoch": 4.86, + "learning_rate": 7.062742103381564e-06, + "loss": 0.0183, + "step": 37074 + }, + { + "epoch": 4.86, + "learning_rate": 7.062235175233161e-06, + "loss": 0.0168, + "step": 37075 + }, + { + "epoch": 4.86, + "learning_rate": 7.061728255347064e-06, + "loss": 0.0172, + "step": 37076 + }, + { + "epoch": 4.86, + "learning_rate": 7.061221343724702e-06, + "loss": 0.0163, + "step": 37077 + }, + { + "epoch": 4.86, + "learning_rate": 7.060714440367496e-06, + "loss": 0.0197, + "step": 37078 + }, + { + "epoch": 4.86, + "learning_rate": 7.0602075452768785e-06, + "loss": 0.0146, + "step": 37079 + }, + { + "epoch": 4.86, + "learning_rate": 7.059700658454266e-06, + "loss": 0.0149, + "step": 37080 + }, + { + "epoch": 4.86, + "learning_rate": 7.059193779901088e-06, + "loss": 0.0184, + "step": 37081 + }, + { + "epoch": 4.86, + "learning_rate": 7.058686909618775e-06, + "loss": 0.0138, + "step": 37082 + }, + { + "epoch": 4.86, + "learning_rate": 7.0581800476087445e-06, + "loss": 0.0162, + "step": 37083 + }, + { + "epoch": 4.86, + "learning_rate": 7.0576731938724285e-06, + "loss": 0.0207, + "step": 37084 + }, + { + "epoch": 4.86, + "learning_rate": 7.057166348411251e-06, + "loss": 0.0174, + "step": 37085 + }, + { + "epoch": 4.86, + "learning_rate": 7.056659511226635e-06, + "loss": 0.0152, + "step": 37086 + }, + { + "epoch": 4.86, + "learning_rate": 7.056152682320009e-06, + "loss": 0.0185, + "step": 37087 + }, + { + "epoch": 4.86, + "learning_rate": 7.055645861692795e-06, + "loss": 0.0215, + "step": 37088 + }, + { + "epoch": 4.86, + "learning_rate": 7.055139049346421e-06, + "loss": 0.0173, + "step": 37089 + }, + { + "epoch": 4.86, + "learning_rate": 7.054632245282311e-06, + "loss": 0.0199, + "step": 37090 + }, + { + "epoch": 4.86, + "learning_rate": 7.054125449501889e-06, + "loss": 0.0234, + "step": 37091 + }, + { + "epoch": 4.86, + "learning_rate": 7.053618662006583e-06, + "loss": 0.0159, + "step": 37092 + }, + { + "epoch": 4.86, + "learning_rate": 7.05311188279782e-06, + "loss": 0.0175, + "step": 37093 + }, + { + "epoch": 4.86, + "learning_rate": 7.052605111877021e-06, + "loss": 0.0185, + "step": 37094 + }, + { + "epoch": 4.86, + "learning_rate": 7.0520983492456155e-06, + "loss": 0.017, + "step": 37095 + }, + { + "epoch": 4.86, + "learning_rate": 7.051591594905025e-06, + "loss": 0.0184, + "step": 37096 + }, + { + "epoch": 4.86, + "learning_rate": 7.051084848856675e-06, + "loss": 0.0193, + "step": 37097 + }, + { + "epoch": 4.86, + "learning_rate": 7.050578111101994e-06, + "loss": 0.0206, + "step": 37098 + }, + { + "epoch": 4.86, + "learning_rate": 7.050071381642402e-06, + "loss": 0.022, + "step": 37099 + }, + { + "epoch": 4.86, + "learning_rate": 7.049564660479331e-06, + "loss": 0.0181, + "step": 37100 + }, + { + "epoch": 4.86, + "learning_rate": 7.0490579476142005e-06, + "loss": 0.0198, + "step": 37101 + }, + { + "epoch": 4.86, + "learning_rate": 7.0485512430484384e-06, + "loss": 0.0203, + "step": 37102 + }, + { + "epoch": 4.86, + "learning_rate": 7.048044546783465e-06, + "loss": 0.0202, + "step": 37103 + }, + { + "epoch": 4.86, + "learning_rate": 7.047537858820711e-06, + "loss": 0.0171, + "step": 37104 + }, + { + "epoch": 4.86, + "learning_rate": 7.0470311791616e-06, + "loss": 0.0152, + "step": 37105 + }, + { + "epoch": 4.86, + "learning_rate": 7.046524507807557e-06, + "loss": 0.0174, + "step": 37106 + }, + { + "epoch": 4.86, + "learning_rate": 7.046017844760007e-06, + "loss": 0.0157, + "step": 37107 + }, + { + "epoch": 4.86, + "learning_rate": 7.045511190020375e-06, + "loss": 0.0185, + "step": 37108 + }, + { + "epoch": 4.86, + "learning_rate": 7.045004543590084e-06, + "loss": 0.0179, + "step": 37109 + }, + { + "epoch": 4.86, + "learning_rate": 7.044497905470562e-06, + "loss": 0.0141, + "step": 37110 + }, + { + "epoch": 4.86, + "learning_rate": 7.043991275663229e-06, + "loss": 0.0173, + "step": 37111 + }, + { + "epoch": 4.86, + "learning_rate": 7.043484654169514e-06, + "loss": 0.0171, + "step": 37112 + }, + { + "epoch": 4.86, + "learning_rate": 7.042978040990844e-06, + "loss": 0.0173, + "step": 37113 + }, + { + "epoch": 4.86, + "learning_rate": 7.042471436128636e-06, + "loss": 0.0153, + "step": 37114 + }, + { + "epoch": 4.86, + "learning_rate": 7.041964839584321e-06, + "loss": 0.0171, + "step": 37115 + }, + { + "epoch": 4.86, + "learning_rate": 7.0414582513593255e-06, + "loss": 0.015, + "step": 37116 + }, + { + "epoch": 4.86, + "learning_rate": 7.04095167145507e-06, + "loss": 0.0157, + "step": 37117 + }, + { + "epoch": 4.86, + "learning_rate": 7.04044509987298e-06, + "loss": 0.0182, + "step": 37118 + }, + { + "epoch": 4.86, + "learning_rate": 7.039938536614479e-06, + "loss": 0.0189, + "step": 37119 + }, + { + "epoch": 4.86, + "learning_rate": 7.039431981680994e-06, + "loss": 0.0237, + "step": 37120 + }, + { + "epoch": 4.86, + "learning_rate": 7.038925435073951e-06, + "loss": 0.0163, + "step": 37121 + }, + { + "epoch": 4.86, + "learning_rate": 7.0384188967947695e-06, + "loss": 0.0181, + "step": 37122 + }, + { + "epoch": 4.86, + "learning_rate": 7.037912366844881e-06, + "loss": 0.0181, + "step": 37123 + }, + { + "epoch": 4.86, + "learning_rate": 7.037405845225702e-06, + "loss": 0.0172, + "step": 37124 + }, + { + "epoch": 4.86, + "learning_rate": 7.0368993319386604e-06, + "loss": 0.023, + "step": 37125 + }, + { + "epoch": 4.86, + "learning_rate": 7.036392826985183e-06, + "loss": 0.021, + "step": 37126 + }, + { + "epoch": 4.86, + "learning_rate": 7.035886330366697e-06, + "loss": 0.0212, + "step": 37127 + }, + { + "epoch": 4.86, + "learning_rate": 7.035379842084619e-06, + "loss": 0.0183, + "step": 37128 + }, + { + "epoch": 4.86, + "learning_rate": 7.0348733621403795e-06, + "loss": 0.0136, + "step": 37129 + }, + { + "epoch": 4.86, + "learning_rate": 7.034366890535399e-06, + "loss": 0.0149, + "step": 37130 + }, + { + "epoch": 4.86, + "learning_rate": 7.033860427271107e-06, + "loss": 0.02, + "step": 37131 + }, + { + "epoch": 4.86, + "learning_rate": 7.03335397234892e-06, + "loss": 0.0166, + "step": 37132 + }, + { + "epoch": 4.86, + "learning_rate": 7.03284752577027e-06, + "loss": 0.019, + "step": 37133 + }, + { + "epoch": 4.86, + "learning_rate": 7.032341087536577e-06, + "loss": 0.022, + "step": 37134 + }, + { + "epoch": 4.86, + "learning_rate": 7.031834657649266e-06, + "loss": 0.0172, + "step": 37135 + }, + { + "epoch": 4.86, + "learning_rate": 7.0313282361097605e-06, + "loss": 0.021, + "step": 37136 + }, + { + "epoch": 4.86, + "learning_rate": 7.030821822919489e-06, + "loss": 0.0169, + "step": 37137 + }, + { + "epoch": 4.86, + "learning_rate": 7.030315418079873e-06, + "loss": 0.0206, + "step": 37138 + }, + { + "epoch": 4.86, + "learning_rate": 7.029809021592337e-06, + "loss": 0.0225, + "step": 37139 + }, + { + "epoch": 4.87, + "learning_rate": 7.029302633458304e-06, + "loss": 0.0124, + "step": 37140 + }, + { + "epoch": 4.87, + "learning_rate": 7.028796253679199e-06, + "loss": 0.0168, + "step": 37141 + }, + { + "epoch": 4.87, + "learning_rate": 7.028289882256447e-06, + "loss": 0.0186, + "step": 37142 + }, + { + "epoch": 4.87, + "learning_rate": 7.027783519191471e-06, + "loss": 0.0181, + "step": 37143 + }, + { + "epoch": 4.87, + "learning_rate": 7.027277164485697e-06, + "loss": 0.0177, + "step": 37144 + }, + { + "epoch": 4.87, + "learning_rate": 7.026770818140547e-06, + "loss": 0.0179, + "step": 37145 + }, + { + "epoch": 4.87, + "learning_rate": 7.0262644801574455e-06, + "loss": 0.017, + "step": 37146 + }, + { + "epoch": 4.87, + "learning_rate": 7.0257581505378114e-06, + "loss": 0.019, + "step": 37147 + }, + { + "epoch": 4.87, + "learning_rate": 7.025251829283082e-06, + "loss": 0.0183, + "step": 37148 + }, + { + "epoch": 4.87, + "learning_rate": 7.0247455163946684e-06, + "loss": 0.014, + "step": 37149 + }, + { + "epoch": 4.87, + "learning_rate": 7.024239211874005e-06, + "loss": 0.0247, + "step": 37150 + }, + { + "epoch": 4.87, + "learning_rate": 7.023732915722505e-06, + "loss": 0.015, + "step": 37151 + }, + { + "epoch": 4.87, + "learning_rate": 7.023226627941602e-06, + "loss": 0.0144, + "step": 37152 + }, + { + "epoch": 4.87, + "learning_rate": 7.022720348532712e-06, + "loss": 0.0207, + "step": 37153 + }, + { + "epoch": 4.87, + "learning_rate": 7.0222140774972655e-06, + "loss": 0.0179, + "step": 37154 + }, + { + "epoch": 4.87, + "learning_rate": 7.021707814836681e-06, + "loss": 0.0222, + "step": 37155 + }, + { + "epoch": 4.87, + "learning_rate": 7.0212015605523844e-06, + "loss": 0.0165, + "step": 37156 + }, + { + "epoch": 4.87, + "learning_rate": 7.0206953146457975e-06, + "loss": 0.0175, + "step": 37157 + }, + { + "epoch": 4.87, + "learning_rate": 7.020189077118351e-06, + "loss": 0.0226, + "step": 37158 + }, + { + "epoch": 4.87, + "learning_rate": 7.019682847971463e-06, + "loss": 0.0159, + "step": 37159 + }, + { + "epoch": 4.87, + "learning_rate": 7.019176627206559e-06, + "loss": 0.0172, + "step": 37160 + }, + { + "epoch": 4.87, + "learning_rate": 7.018670414825061e-06, + "loss": 0.0167, + "step": 37161 + }, + { + "epoch": 4.87, + "learning_rate": 7.018164210828392e-06, + "loss": 0.0153, + "step": 37162 + }, + { + "epoch": 4.87, + "learning_rate": 7.017658015217981e-06, + "loss": 0.016, + "step": 37163 + }, + { + "epoch": 4.87, + "learning_rate": 7.017151827995245e-06, + "loss": 0.0184, + "step": 37164 + }, + { + "epoch": 4.87, + "learning_rate": 7.016645649161613e-06, + "loss": 0.0153, + "step": 37165 + }, + { + "epoch": 4.87, + "learning_rate": 7.016139478718504e-06, + "loss": 0.0137, + "step": 37166 + }, + { + "epoch": 4.87, + "learning_rate": 7.015633316667346e-06, + "loss": 0.0155, + "step": 37167 + }, + { + "epoch": 4.87, + "learning_rate": 7.015127163009557e-06, + "loss": 0.0168, + "step": 37168 + }, + { + "epoch": 4.87, + "learning_rate": 7.0146210177465656e-06, + "loss": 0.016, + "step": 37169 + }, + { + "epoch": 4.87, + "learning_rate": 7.014114880879792e-06, + "loss": 0.0199, + "step": 37170 + }, + { + "epoch": 4.87, + "learning_rate": 7.013608752410666e-06, + "loss": 0.017, + "step": 37171 + }, + { + "epoch": 4.87, + "learning_rate": 7.013102632340603e-06, + "loss": 0.0162, + "step": 37172 + }, + { + "epoch": 4.87, + "learning_rate": 7.012596520671031e-06, + "loss": 0.0159, + "step": 37173 + }, + { + "epoch": 4.87, + "learning_rate": 7.012090417403371e-06, + "loss": 0.0173, + "step": 37174 + }, + { + "epoch": 4.87, + "learning_rate": 7.011584322539049e-06, + "loss": 0.0196, + "step": 37175 + }, + { + "epoch": 4.87, + "learning_rate": 7.011078236079486e-06, + "loss": 0.0191, + "step": 37176 + }, + { + "epoch": 4.87, + "learning_rate": 7.010572158026106e-06, + "loss": 0.0181, + "step": 37177 + }, + { + "epoch": 4.87, + "learning_rate": 7.010066088380334e-06, + "loss": 0.0168, + "step": 37178 + }, + { + "epoch": 4.87, + "learning_rate": 7.009560027143588e-06, + "loss": 0.0199, + "step": 37179 + }, + { + "epoch": 4.87, + "learning_rate": 7.009053974317297e-06, + "loss": 0.0152, + "step": 37180 + }, + { + "epoch": 4.87, + "learning_rate": 7.008547929902886e-06, + "loss": 0.0152, + "step": 37181 + }, + { + "epoch": 4.87, + "learning_rate": 7.0080418939017715e-06, + "loss": 0.0226, + "step": 37182 + }, + { + "epoch": 4.87, + "learning_rate": 7.007535866315381e-06, + "loss": 0.0139, + "step": 37183 + }, + { + "epoch": 4.87, + "learning_rate": 7.007029847145136e-06, + "loss": 0.0167, + "step": 37184 + }, + { + "epoch": 4.87, + "learning_rate": 7.006523836392459e-06, + "loss": 0.015, + "step": 37185 + }, + { + "epoch": 4.87, + "learning_rate": 7.006017834058777e-06, + "loss": 0.0193, + "step": 37186 + }, + { + "epoch": 4.87, + "learning_rate": 7.005511840145509e-06, + "loss": 0.0165, + "step": 37187 + }, + { + "epoch": 4.87, + "learning_rate": 7.00500585465408e-06, + "loss": 0.0177, + "step": 37188 + }, + { + "epoch": 4.87, + "learning_rate": 7.00449987758591e-06, + "loss": 0.0198, + "step": 37189 + }, + { + "epoch": 4.87, + "learning_rate": 7.0039939089424255e-06, + "loss": 0.0151, + "step": 37190 + }, + { + "epoch": 4.87, + "learning_rate": 7.003487948725048e-06, + "loss": 0.0168, + "step": 37191 + }, + { + "epoch": 4.87, + "learning_rate": 7.002981996935206e-06, + "loss": 0.0158, + "step": 37192 + }, + { + "epoch": 4.87, + "learning_rate": 7.002476053574312e-06, + "loss": 0.0117, + "step": 37193 + }, + { + "epoch": 4.87, + "learning_rate": 7.001970118643799e-06, + "loss": 0.0206, + "step": 37194 + }, + { + "epoch": 4.87, + "learning_rate": 7.001464192145082e-06, + "loss": 0.021, + "step": 37195 + }, + { + "epoch": 4.87, + "learning_rate": 7.00095827407959e-06, + "loss": 0.0172, + "step": 37196 + }, + { + "epoch": 4.87, + "learning_rate": 7.00045236444874e-06, + "loss": 0.0159, + "step": 37197 + }, + { + "epoch": 4.87, + "learning_rate": 6.999946463253962e-06, + "loss": 0.0158, + "step": 37198 + }, + { + "epoch": 4.87, + "learning_rate": 6.99944057049667e-06, + "loss": 0.0175, + "step": 37199 + }, + { + "epoch": 4.87, + "learning_rate": 6.9989346861782935e-06, + "loss": 0.0186, + "step": 37200 + }, + { + "epoch": 4.87, + "learning_rate": 6.998428810300251e-06, + "loss": 0.0141, + "step": 37201 + }, + { + "epoch": 4.87, + "learning_rate": 6.997922942863973e-06, + "loss": 0.0164, + "step": 37202 + }, + { + "epoch": 4.87, + "learning_rate": 6.997417083870872e-06, + "loss": 0.0172, + "step": 37203 + }, + { + "epoch": 4.87, + "learning_rate": 6.99691123332238e-06, + "loss": 0.0184, + "step": 37204 + }, + { + "epoch": 4.87, + "learning_rate": 6.996405391219911e-06, + "loss": 0.0165, + "step": 37205 + }, + { + "epoch": 4.87, + "learning_rate": 6.995899557564893e-06, + "loss": 0.0211, + "step": 37206 + }, + { + "epoch": 4.87, + "learning_rate": 6.99539373235875e-06, + "loss": 0.0148, + "step": 37207 + }, + { + "epoch": 4.87, + "learning_rate": 6.994887915602901e-06, + "loss": 0.019, + "step": 37208 + }, + { + "epoch": 4.87, + "learning_rate": 6.9943821072987704e-06, + "loss": 0.0186, + "step": 37209 + }, + { + "epoch": 4.87, + "learning_rate": 6.993876307447778e-06, + "loss": 0.0141, + "step": 37210 + }, + { + "epoch": 4.87, + "learning_rate": 6.993370516051351e-06, + "loss": 0.0201, + "step": 37211 + }, + { + "epoch": 4.87, + "learning_rate": 6.992864733110902e-06, + "loss": 0.0157, + "step": 37212 + }, + { + "epoch": 4.87, + "learning_rate": 6.99235895862787e-06, + "loss": 0.0166, + "step": 37213 + }, + { + "epoch": 4.87, + "learning_rate": 6.991853192603663e-06, + "loss": 0.0173, + "step": 37214 + }, + { + "epoch": 4.87, + "learning_rate": 6.991347435039711e-06, + "loss": 0.0145, + "step": 37215 + }, + { + "epoch": 4.88, + "learning_rate": 6.990841685937433e-06, + "loss": 0.0173, + "step": 37216 + }, + { + "epoch": 4.88, + "learning_rate": 6.990335945298255e-06, + "loss": 0.0141, + "step": 37217 + }, + { + "epoch": 4.88, + "learning_rate": 6.989830213123594e-06, + "loss": 0.0146, + "step": 37218 + }, + { + "epoch": 4.88, + "learning_rate": 6.989324489414877e-06, + "loss": 0.0186, + "step": 37219 + }, + { + "epoch": 4.88, + "learning_rate": 6.988818774173524e-06, + "loss": 0.0173, + "step": 37220 + }, + { + "epoch": 4.88, + "learning_rate": 6.9883130674009555e-06, + "loss": 0.0162, + "step": 37221 + }, + { + "epoch": 4.88, + "learning_rate": 6.987807369098601e-06, + "loss": 0.0197, + "step": 37222 + }, + { + "epoch": 4.88, + "learning_rate": 6.9873016792678715e-06, + "loss": 0.0132, + "step": 37223 + }, + { + "epoch": 4.88, + "learning_rate": 6.986795997910198e-06, + "loss": 0.0171, + "step": 37224 + }, + { + "epoch": 4.88, + "learning_rate": 6.986290325027003e-06, + "loss": 0.018, + "step": 37225 + }, + { + "epoch": 4.88, + "learning_rate": 6.985784660619704e-06, + "loss": 0.0153, + "step": 37226 + }, + { + "epoch": 4.88, + "learning_rate": 6.985279004689725e-06, + "loss": 0.0169, + "step": 37227 + }, + { + "epoch": 4.88, + "learning_rate": 6.98477335723849e-06, + "loss": 0.0147, + "step": 37228 + }, + { + "epoch": 4.88, + "learning_rate": 6.984267718267417e-06, + "loss": 0.0183, + "step": 37229 + }, + { + "epoch": 4.88, + "learning_rate": 6.983762087777933e-06, + "loss": 0.0126, + "step": 37230 + }, + { + "epoch": 4.88, + "learning_rate": 6.983256465771455e-06, + "loss": 0.0175, + "step": 37231 + }, + { + "epoch": 4.88, + "learning_rate": 6.982750852249412e-06, + "loss": 0.0151, + "step": 37232 + }, + { + "epoch": 4.88, + "learning_rate": 6.982245247213216e-06, + "loss": 0.0194, + "step": 37233 + }, + { + "epoch": 4.88, + "learning_rate": 6.981739650664294e-06, + "loss": 0.0231, + "step": 37234 + }, + { + "epoch": 4.88, + "learning_rate": 6.9812340626040695e-06, + "loss": 0.0148, + "step": 37235 + }, + { + "epoch": 4.88, + "learning_rate": 6.980728483033967e-06, + "loss": 0.0157, + "step": 37236 + }, + { + "epoch": 4.88, + "learning_rate": 6.9802229119554035e-06, + "loss": 0.0177, + "step": 37237 + }, + { + "epoch": 4.88, + "learning_rate": 6.979717349369802e-06, + "loss": 0.0144, + "step": 37238 + }, + { + "epoch": 4.88, + "learning_rate": 6.979211795278585e-06, + "loss": 0.0203, + "step": 37239 + }, + { + "epoch": 4.88, + "learning_rate": 6.9787062496831735e-06, + "loss": 0.0186, + "step": 37240 + }, + { + "epoch": 4.88, + "learning_rate": 6.978200712584989e-06, + "loss": 0.014, + "step": 37241 + }, + { + "epoch": 4.88, + "learning_rate": 6.9776951839854535e-06, + "loss": 0.0159, + "step": 37242 + }, + { + "epoch": 4.88, + "learning_rate": 6.977189663885992e-06, + "loss": 0.0161, + "step": 37243 + }, + { + "epoch": 4.88, + "learning_rate": 6.976684152288022e-06, + "loss": 0.0254, + "step": 37244 + }, + { + "epoch": 4.88, + "learning_rate": 6.976178649192964e-06, + "loss": 0.016, + "step": 37245 + }, + { + "epoch": 4.88, + "learning_rate": 6.975673154602247e-06, + "loss": 0.0165, + "step": 37246 + }, + { + "epoch": 4.88, + "learning_rate": 6.975167668517286e-06, + "loss": 0.0193, + "step": 37247 + }, + { + "epoch": 4.88, + "learning_rate": 6.9746621909395075e-06, + "loss": 0.0145, + "step": 37248 + }, + { + "epoch": 4.88, + "learning_rate": 6.9741567218703265e-06, + "loss": 0.0177, + "step": 37249 + }, + { + "epoch": 4.88, + "learning_rate": 6.973651261311169e-06, + "loss": 0.0173, + "step": 37250 + }, + { + "epoch": 4.88, + "learning_rate": 6.973145809263459e-06, + "loss": 0.0194, + "step": 37251 + }, + { + "epoch": 4.88, + "learning_rate": 6.972640365728615e-06, + "loss": 0.0225, + "step": 37252 + }, + { + "epoch": 4.88, + "learning_rate": 6.9721349307080586e-06, + "loss": 0.0149, + "step": 37253 + }, + { + "epoch": 4.88, + "learning_rate": 6.97162950420321e-06, + "loss": 0.0177, + "step": 37254 + }, + { + "epoch": 4.88, + "learning_rate": 6.971124086215495e-06, + "loss": 0.0184, + "step": 37255 + }, + { + "epoch": 4.88, + "learning_rate": 6.9706186767463235e-06, + "loss": 0.016, + "step": 37256 + }, + { + "epoch": 4.88, + "learning_rate": 6.970113275797135e-06, + "loss": 0.0176, + "step": 37257 + }, + { + "epoch": 4.88, + "learning_rate": 6.969607883369337e-06, + "loss": 0.0151, + "step": 37258 + }, + { + "epoch": 4.88, + "learning_rate": 6.9691024994643585e-06, + "loss": 0.0168, + "step": 37259 + }, + { + "epoch": 4.88, + "learning_rate": 6.9685971240836155e-06, + "loss": 0.0166, + "step": 37260 + }, + { + "epoch": 4.88, + "learning_rate": 6.9680917572285345e-06, + "loss": 0.0138, + "step": 37261 + }, + { + "epoch": 4.88, + "learning_rate": 6.967586398900531e-06, + "loss": 0.0193, + "step": 37262 + }, + { + "epoch": 4.88, + "learning_rate": 6.96708104910103e-06, + "loss": 0.0193, + "step": 37263 + }, + { + "epoch": 4.88, + "learning_rate": 6.966575707831453e-06, + "loss": 0.0188, + "step": 37264 + }, + { + "epoch": 4.88, + "learning_rate": 6.966070375093219e-06, + "loss": 0.0127, + "step": 37265 + }, + { + "epoch": 4.88, + "learning_rate": 6.965565050887752e-06, + "loss": 0.0239, + "step": 37266 + }, + { + "epoch": 4.88, + "learning_rate": 6.965059735216468e-06, + "loss": 0.0141, + "step": 37267 + }, + { + "epoch": 4.88, + "learning_rate": 6.964554428080794e-06, + "loss": 0.0166, + "step": 37268 + }, + { + "epoch": 4.88, + "learning_rate": 6.964049129482151e-06, + "loss": 0.014, + "step": 37269 + }, + { + "epoch": 4.88, + "learning_rate": 6.963543839421957e-06, + "loss": 0.0168, + "step": 37270 + }, + { + "epoch": 4.88, + "learning_rate": 6.963038557901634e-06, + "loss": 0.0202, + "step": 37271 + }, + { + "epoch": 4.88, + "learning_rate": 6.962533284922603e-06, + "loss": 0.0215, + "step": 37272 + }, + { + "epoch": 4.88, + "learning_rate": 6.962028020486287e-06, + "loss": 0.0188, + "step": 37273 + }, + { + "epoch": 4.88, + "learning_rate": 6.961522764594105e-06, + "loss": 0.0228, + "step": 37274 + }, + { + "epoch": 4.88, + "learning_rate": 6.961017517247477e-06, + "loss": 0.0195, + "step": 37275 + }, + { + "epoch": 4.88, + "learning_rate": 6.9605122784478276e-06, + "loss": 0.0161, + "step": 37276 + }, + { + "epoch": 4.88, + "learning_rate": 6.960007048196573e-06, + "loss": 0.0186, + "step": 37277 + }, + { + "epoch": 4.88, + "learning_rate": 6.959501826495135e-06, + "loss": 0.0166, + "step": 37278 + }, + { + "epoch": 4.88, + "learning_rate": 6.958996613344938e-06, + "loss": 0.0188, + "step": 37279 + }, + { + "epoch": 4.88, + "learning_rate": 6.958491408747403e-06, + "loss": 0.0173, + "step": 37280 + }, + { + "epoch": 4.88, + "learning_rate": 6.9579862127039476e-06, + "loss": 0.0163, + "step": 37281 + }, + { + "epoch": 4.88, + "learning_rate": 6.957481025215995e-06, + "loss": 0.0165, + "step": 37282 + }, + { + "epoch": 4.88, + "learning_rate": 6.956975846284964e-06, + "loss": 0.018, + "step": 37283 + }, + { + "epoch": 4.88, + "learning_rate": 6.956470675912279e-06, + "loss": 0.0198, + "step": 37284 + }, + { + "epoch": 4.88, + "learning_rate": 6.9559655140993545e-06, + "loss": 0.0146, + "step": 37285 + }, + { + "epoch": 4.88, + "learning_rate": 6.955460360847616e-06, + "loss": 0.0154, + "step": 37286 + }, + { + "epoch": 4.88, + "learning_rate": 6.954955216158485e-06, + "loss": 0.0172, + "step": 37287 + }, + { + "epoch": 4.88, + "learning_rate": 6.954450080033375e-06, + "loss": 0.0152, + "step": 37288 + }, + { + "epoch": 4.88, + "learning_rate": 6.953944952473714e-06, + "loss": 0.0215, + "step": 37289 + }, + { + "epoch": 4.88, + "learning_rate": 6.953439833480924e-06, + "loss": 0.0181, + "step": 37290 + }, + { + "epoch": 4.88, + "learning_rate": 6.952934723056421e-06, + "loss": 0.0163, + "step": 37291 + }, + { + "epoch": 4.88, + "learning_rate": 6.9524296212016255e-06, + "loss": 0.0208, + "step": 37292 + }, + { + "epoch": 4.89, + "learning_rate": 6.951924527917962e-06, + "loss": 0.0221, + "step": 37293 + }, + { + "epoch": 4.89, + "learning_rate": 6.951419443206845e-06, + "loss": 0.0173, + "step": 37294 + }, + { + "epoch": 4.89, + "learning_rate": 6.9509143670697014e-06, + "loss": 0.0195, + "step": 37295 + }, + { + "epoch": 4.89, + "learning_rate": 6.9504092995079465e-06, + "loss": 0.0185, + "step": 37296 + }, + { + "epoch": 4.89, + "learning_rate": 6.949904240523006e-06, + "loss": 0.0135, + "step": 37297 + }, + { + "epoch": 4.89, + "learning_rate": 6.949399190116295e-06, + "loss": 0.0242, + "step": 37298 + }, + { + "epoch": 4.89, + "learning_rate": 6.948894148289234e-06, + "loss": 0.0188, + "step": 37299 + }, + { + "epoch": 4.89, + "learning_rate": 6.948389115043247e-06, + "loss": 0.0201, + "step": 37300 + }, + { + "epoch": 4.89, + "learning_rate": 6.947884090379755e-06, + "loss": 0.0123, + "step": 37301 + }, + { + "epoch": 4.89, + "learning_rate": 6.947379074300176e-06, + "loss": 0.0188, + "step": 37302 + }, + { + "epoch": 4.89, + "learning_rate": 6.946874066805932e-06, + "loss": 0.0173, + "step": 37303 + }, + { + "epoch": 4.89, + "learning_rate": 6.94636906789844e-06, + "loss": 0.0236, + "step": 37304 + }, + { + "epoch": 4.89, + "learning_rate": 6.945864077579126e-06, + "loss": 0.0175, + "step": 37305 + }, + { + "epoch": 4.89, + "learning_rate": 6.945359095849402e-06, + "loss": 0.0187, + "step": 37306 + }, + { + "epoch": 4.89, + "learning_rate": 6.944854122710693e-06, + "loss": 0.0164, + "step": 37307 + }, + { + "epoch": 4.89, + "learning_rate": 6.944349158164423e-06, + "loss": 0.0156, + "step": 37308 + }, + { + "epoch": 4.89, + "learning_rate": 6.943844202212006e-06, + "loss": 0.0179, + "step": 37309 + }, + { + "epoch": 4.89, + "learning_rate": 6.943339254854862e-06, + "loss": 0.0165, + "step": 37310 + }, + { + "epoch": 4.89, + "learning_rate": 6.942834316094418e-06, + "loss": 0.0149, + "step": 37311 + }, + { + "epoch": 4.89, + "learning_rate": 6.942329385932086e-06, + "loss": 0.02, + "step": 37312 + }, + { + "epoch": 4.89, + "learning_rate": 6.941824464369291e-06, + "loss": 0.0154, + "step": 37313 + }, + { + "epoch": 4.89, + "learning_rate": 6.9413195514074536e-06, + "loss": 0.0163, + "step": 37314 + }, + { + "epoch": 4.89, + "learning_rate": 6.940814647047991e-06, + "loss": 0.02, + "step": 37315 + }, + { + "epoch": 4.89, + "learning_rate": 6.940309751292324e-06, + "loss": 0.016, + "step": 37316 + }, + { + "epoch": 4.89, + "learning_rate": 6.939804864141874e-06, + "loss": 0.0157, + "step": 37317 + }, + { + "epoch": 4.89, + "learning_rate": 6.93929998559806e-06, + "loss": 0.0154, + "step": 37318 + }, + { + "epoch": 4.89, + "learning_rate": 6.9387951156623e-06, + "loss": 0.0183, + "step": 37319 + }, + { + "epoch": 4.89, + "learning_rate": 6.938290254336018e-06, + "loss": 0.0191, + "step": 37320 + }, + { + "epoch": 4.89, + "learning_rate": 6.937785401620625e-06, + "loss": 0.0172, + "step": 37321 + }, + { + "epoch": 4.89, + "learning_rate": 6.937280557517554e-06, + "loss": 0.0153, + "step": 37322 + }, + { + "epoch": 4.89, + "learning_rate": 6.936775722028215e-06, + "loss": 0.0145, + "step": 37323 + }, + { + "epoch": 4.89, + "learning_rate": 6.9362708951540354e-06, + "loss": 0.0179, + "step": 37324 + }, + { + "epoch": 4.89, + "learning_rate": 6.935766076896428e-06, + "loss": 0.0154, + "step": 37325 + }, + { + "epoch": 4.89, + "learning_rate": 6.935261267256817e-06, + "loss": 0.0168, + "step": 37326 + }, + { + "epoch": 4.89, + "learning_rate": 6.934756466236618e-06, + "loss": 0.016, + "step": 37327 + }, + { + "epoch": 4.89, + "learning_rate": 6.934251673837253e-06, + "loss": 0.0139, + "step": 37328 + }, + { + "epoch": 4.89, + "learning_rate": 6.933746890060146e-06, + "loss": 0.0196, + "step": 37329 + }, + { + "epoch": 4.89, + "learning_rate": 6.933242114906709e-06, + "loss": 0.016, + "step": 37330 + }, + { + "epoch": 4.89, + "learning_rate": 6.932737348378366e-06, + "loss": 0.0171, + "step": 37331 + }, + { + "epoch": 4.89, + "learning_rate": 6.932232590476534e-06, + "loss": 0.0122, + "step": 37332 + }, + { + "epoch": 4.89, + "learning_rate": 6.931727841202635e-06, + "loss": 0.0202, + "step": 37333 + }, + { + "epoch": 4.89, + "learning_rate": 6.931223100558092e-06, + "loss": 0.0155, + "step": 37334 + }, + { + "epoch": 4.89, + "learning_rate": 6.930718368544318e-06, + "loss": 0.0224, + "step": 37335 + }, + { + "epoch": 4.89, + "learning_rate": 6.930213645162733e-06, + "loss": 0.0127, + "step": 37336 + }, + { + "epoch": 4.89, + "learning_rate": 6.929708930414762e-06, + "loss": 0.0183, + "step": 37337 + }, + { + "epoch": 4.89, + "learning_rate": 6.92920422430182e-06, + "loss": 0.0168, + "step": 37338 + }, + { + "epoch": 4.89, + "learning_rate": 6.928699526825329e-06, + "loss": 0.0208, + "step": 37339 + }, + { + "epoch": 4.89, + "learning_rate": 6.928194837986706e-06, + "loss": 0.0143, + "step": 37340 + }, + { + "epoch": 4.89, + "learning_rate": 6.9276901577873725e-06, + "loss": 0.0161, + "step": 37341 + }, + { + "epoch": 4.89, + "learning_rate": 6.927185486228745e-06, + "loss": 0.0209, + "step": 37342 + }, + { + "epoch": 4.89, + "learning_rate": 6.9266808233122414e-06, + "loss": 0.0167, + "step": 37343 + }, + { + "epoch": 4.89, + "learning_rate": 6.926176169039287e-06, + "loss": 0.0159, + "step": 37344 + }, + { + "epoch": 4.89, + "learning_rate": 6.925671523411301e-06, + "loss": 0.0138, + "step": 37345 + }, + { + "epoch": 4.89, + "learning_rate": 6.925166886429697e-06, + "loss": 0.0147, + "step": 37346 + }, + { + "epoch": 4.89, + "learning_rate": 6.924662258095902e-06, + "loss": 0.0187, + "step": 37347 + }, + { + "epoch": 4.89, + "learning_rate": 6.924157638411326e-06, + "loss": 0.0173, + "step": 37348 + }, + { + "epoch": 4.89, + "learning_rate": 6.923653027377396e-06, + "loss": 0.0173, + "step": 37349 + }, + { + "epoch": 4.89, + "learning_rate": 6.9231484249955265e-06, + "loss": 0.0152, + "step": 37350 + }, + { + "epoch": 4.89, + "learning_rate": 6.922643831267137e-06, + "loss": 0.0152, + "step": 37351 + }, + { + "epoch": 4.89, + "learning_rate": 6.922139246193651e-06, + "loss": 0.0178, + "step": 37352 + }, + { + "epoch": 4.89, + "learning_rate": 6.92163466977648e-06, + "loss": 0.0164, + "step": 37353 + }, + { + "epoch": 4.89, + "learning_rate": 6.921130102017049e-06, + "loss": 0.0198, + "step": 37354 + }, + { + "epoch": 4.89, + "learning_rate": 6.920625542916777e-06, + "loss": 0.0134, + "step": 37355 + }, + { + "epoch": 4.89, + "learning_rate": 6.920120992477081e-06, + "loss": 0.0142, + "step": 37356 + }, + { + "epoch": 4.89, + "learning_rate": 6.91961645069938e-06, + "loss": 0.0193, + "step": 37357 + }, + { + "epoch": 4.89, + "learning_rate": 6.919111917585096e-06, + "loss": 0.0156, + "step": 37358 + }, + { + "epoch": 4.89, + "learning_rate": 6.918607393135643e-06, + "loss": 0.0148, + "step": 37359 + }, + { + "epoch": 4.89, + "learning_rate": 6.918102877352446e-06, + "loss": 0.0166, + "step": 37360 + }, + { + "epoch": 4.89, + "learning_rate": 6.917598370236918e-06, + "loss": 0.0229, + "step": 37361 + }, + { + "epoch": 4.89, + "learning_rate": 6.917093871790483e-06, + "loss": 0.018, + "step": 37362 + }, + { + "epoch": 4.89, + "learning_rate": 6.9165893820145535e-06, + "loss": 0.0168, + "step": 37363 + }, + { + "epoch": 4.89, + "learning_rate": 6.916084900910553e-06, + "loss": 0.0117, + "step": 37364 + }, + { + "epoch": 4.89, + "learning_rate": 6.915580428479898e-06, + "loss": 0.0159, + "step": 37365 + }, + { + "epoch": 4.89, + "learning_rate": 6.915075964724014e-06, + "loss": 0.0171, + "step": 37366 + }, + { + "epoch": 4.89, + "learning_rate": 6.91457150964431e-06, + "loss": 0.0156, + "step": 37367 + }, + { + "epoch": 4.89, + "learning_rate": 6.914067063242213e-06, + "loss": 0.0143, + "step": 37368 + }, + { + "epoch": 4.9, + "learning_rate": 6.913562625519134e-06, + "loss": 0.0151, + "step": 37369 + }, + { + "epoch": 4.9, + "learning_rate": 6.913058196476501e-06, + "loss": 0.0152, + "step": 37370 + }, + { + "epoch": 4.9, + "learning_rate": 6.912553776115722e-06, + "loss": 0.0193, + "step": 37371 + }, + { + "epoch": 4.9, + "learning_rate": 6.912049364438224e-06, + "loss": 0.0189, + "step": 37372 + }, + { + "epoch": 4.9, + "learning_rate": 6.911544961445423e-06, + "loss": 0.018, + "step": 37373 + }, + { + "epoch": 4.9, + "learning_rate": 6.911040567138736e-06, + "loss": 0.0166, + "step": 37374 + }, + { + "epoch": 4.9, + "learning_rate": 6.910536181519584e-06, + "loss": 0.0163, + "step": 37375 + }, + { + "epoch": 4.9, + "learning_rate": 6.910031804589382e-06, + "loss": 0.0191, + "step": 37376 + }, + { + "epoch": 4.9, + "learning_rate": 6.909527436349552e-06, + "loss": 0.0148, + "step": 37377 + }, + { + "epoch": 4.9, + "learning_rate": 6.909023076801512e-06, + "loss": 0.0188, + "step": 37378 + }, + { + "epoch": 4.9, + "learning_rate": 6.908518725946682e-06, + "loss": 0.0175, + "step": 37379 + }, + { + "epoch": 4.9, + "learning_rate": 6.908014383786475e-06, + "loss": 0.0161, + "step": 37380 + }, + { + "epoch": 4.9, + "learning_rate": 6.9075100503223165e-06, + "loss": 0.0157, + "step": 37381 + }, + { + "epoch": 4.9, + "learning_rate": 6.9070057255556185e-06, + "loss": 0.0142, + "step": 37382 + }, + { + "epoch": 4.9, + "learning_rate": 6.906501409487805e-06, + "loss": 0.0168, + "step": 37383 + }, + { + "epoch": 4.9, + "learning_rate": 6.90599710212029e-06, + "loss": 0.0141, + "step": 37384 + }, + { + "epoch": 4.9, + "learning_rate": 6.905492803454495e-06, + "loss": 0.0149, + "step": 37385 + }, + { + "epoch": 4.9, + "learning_rate": 6.904988513491832e-06, + "loss": 0.0132, + "step": 37386 + }, + { + "epoch": 4.9, + "learning_rate": 6.9044842322337245e-06, + "loss": 0.0168, + "step": 37387 + }, + { + "epoch": 4.9, + "learning_rate": 6.903979959681592e-06, + "loss": 0.0185, + "step": 37388 + }, + { + "epoch": 4.9, + "learning_rate": 6.903475695836853e-06, + "loss": 0.0166, + "step": 37389 + }, + { + "epoch": 4.9, + "learning_rate": 6.902971440700922e-06, + "loss": 0.0172, + "step": 37390 + }, + { + "epoch": 4.9, + "learning_rate": 6.902467194275219e-06, + "loss": 0.0157, + "step": 37391 + }, + { + "epoch": 4.9, + "learning_rate": 6.901962956561164e-06, + "loss": 0.0162, + "step": 37392 + }, + { + "epoch": 4.9, + "learning_rate": 6.901458727560169e-06, + "loss": 0.0155, + "step": 37393 + }, + { + "epoch": 4.9, + "learning_rate": 6.90095450727366e-06, + "loss": 0.0165, + "step": 37394 + }, + { + "epoch": 4.9, + "learning_rate": 6.900450295703049e-06, + "loss": 0.0188, + "step": 37395 + }, + { + "epoch": 4.9, + "learning_rate": 6.89994609284976e-06, + "loss": 0.0178, + "step": 37396 + }, + { + "epoch": 4.9, + "learning_rate": 6.899441898715205e-06, + "loss": 0.0178, + "step": 37397 + }, + { + "epoch": 4.9, + "learning_rate": 6.898937713300802e-06, + "loss": 0.0211, + "step": 37398 + }, + { + "epoch": 4.9, + "learning_rate": 6.898433536607974e-06, + "loss": 0.0172, + "step": 37399 + }, + { + "epoch": 4.9, + "learning_rate": 6.897929368638136e-06, + "loss": 0.0163, + "step": 37400 + }, + { + "epoch": 4.9, + "learning_rate": 6.897425209392707e-06, + "loss": 0.019, + "step": 37401 + }, + { + "epoch": 4.9, + "learning_rate": 6.896921058873106e-06, + "loss": 0.0188, + "step": 37402 + }, + { + "epoch": 4.9, + "learning_rate": 6.8964169170807485e-06, + "loss": 0.0208, + "step": 37403 + }, + { + "epoch": 4.9, + "learning_rate": 6.8959127840170534e-06, + "loss": 0.0201, + "step": 37404 + }, + { + "epoch": 4.9, + "learning_rate": 6.8954086596834356e-06, + "loss": 0.017, + "step": 37405 + }, + { + "epoch": 4.9, + "learning_rate": 6.89490454408132e-06, + "loss": 0.0154, + "step": 37406 + }, + { + "epoch": 4.9, + "learning_rate": 6.894400437212117e-06, + "loss": 0.0179, + "step": 37407 + }, + { + "epoch": 4.9, + "learning_rate": 6.893896339077247e-06, + "loss": 0.0194, + "step": 37408 + }, + { + "epoch": 4.9, + "learning_rate": 6.893392249678126e-06, + "loss": 0.0177, + "step": 37409 + }, + { + "epoch": 4.9, + "learning_rate": 6.892888169016178e-06, + "loss": 0.0174, + "step": 37410 + }, + { + "epoch": 4.9, + "learning_rate": 6.892384097092815e-06, + "loss": 0.0157, + "step": 37411 + }, + { + "epoch": 4.9, + "learning_rate": 6.8918800339094575e-06, + "loss": 0.0238, + "step": 37412 + }, + { + "epoch": 4.9, + "learning_rate": 6.891375979467521e-06, + "loss": 0.0179, + "step": 37413 + }, + { + "epoch": 4.9, + "learning_rate": 6.890871933768422e-06, + "loss": 0.0171, + "step": 37414 + }, + { + "epoch": 4.9, + "learning_rate": 6.890367896813583e-06, + "loss": 0.0236, + "step": 37415 + }, + { + "epoch": 4.9, + "learning_rate": 6.889863868604419e-06, + "loss": 0.0189, + "step": 37416 + }, + { + "epoch": 4.9, + "learning_rate": 6.889359849142346e-06, + "loss": 0.015, + "step": 37417 + }, + { + "epoch": 4.9, + "learning_rate": 6.888855838428783e-06, + "loss": 0.0127, + "step": 37418 + }, + { + "epoch": 4.9, + "learning_rate": 6.888351836465143e-06, + "loss": 0.0193, + "step": 37419 + }, + { + "epoch": 4.9, + "learning_rate": 6.887847843252854e-06, + "loss": 0.0158, + "step": 37420 + }, + { + "epoch": 4.9, + "learning_rate": 6.887343858793325e-06, + "loss": 0.0157, + "step": 37421 + }, + { + "epoch": 4.9, + "learning_rate": 6.886839883087974e-06, + "loss": 0.0182, + "step": 37422 + }, + { + "epoch": 4.9, + "learning_rate": 6.886335916138224e-06, + "loss": 0.0194, + "step": 37423 + }, + { + "epoch": 4.9, + "learning_rate": 6.885831957945486e-06, + "loss": 0.0203, + "step": 37424 + }, + { + "epoch": 4.9, + "learning_rate": 6.885328008511181e-06, + "loss": 0.025, + "step": 37425 + }, + { + "epoch": 4.9, + "learning_rate": 6.884824067836723e-06, + "loss": 0.0173, + "step": 37426 + }, + { + "epoch": 4.9, + "learning_rate": 6.884320135923534e-06, + "loss": 0.0176, + "step": 37427 + }, + { + "epoch": 4.9, + "learning_rate": 6.883816212773027e-06, + "loss": 0.0185, + "step": 37428 + }, + { + "epoch": 4.9, + "learning_rate": 6.88331229838662e-06, + "loss": 0.0166, + "step": 37429 + }, + { + "epoch": 4.9, + "learning_rate": 6.88280839276573e-06, + "loss": 0.0171, + "step": 37430 + }, + { + "epoch": 4.9, + "learning_rate": 6.882304495911778e-06, + "loss": 0.0172, + "step": 37431 + }, + { + "epoch": 4.9, + "learning_rate": 6.881800607826178e-06, + "loss": 0.0145, + "step": 37432 + }, + { + "epoch": 4.9, + "learning_rate": 6.88129672851035e-06, + "loss": 0.0154, + "step": 37433 + }, + { + "epoch": 4.9, + "learning_rate": 6.880792857965706e-06, + "loss": 0.0149, + "step": 37434 + }, + { + "epoch": 4.9, + "learning_rate": 6.880288996193667e-06, + "loss": 0.015, + "step": 37435 + }, + { + "epoch": 4.9, + "learning_rate": 6.87978514319565e-06, + "loss": 0.0162, + "step": 37436 + }, + { + "epoch": 4.9, + "learning_rate": 6.879281298973068e-06, + "loss": 0.0188, + "step": 37437 + }, + { + "epoch": 4.9, + "learning_rate": 6.878777463527344e-06, + "loss": 0.0189, + "step": 37438 + }, + { + "epoch": 4.9, + "learning_rate": 6.878273636859891e-06, + "loss": 0.0153, + "step": 37439 + }, + { + "epoch": 4.9, + "learning_rate": 6.8777698189721296e-06, + "loss": 0.0179, + "step": 37440 + }, + { + "epoch": 4.9, + "learning_rate": 6.877266009865469e-06, + "loss": 0.0139, + "step": 37441 + }, + { + "epoch": 4.9, + "learning_rate": 6.876762209541334e-06, + "loss": 0.0167, + "step": 37442 + }, + { + "epoch": 4.9, + "learning_rate": 6.8762584180011395e-06, + "loss": 0.0156, + "step": 37443 + }, + { + "epoch": 4.9, + "learning_rate": 6.875754635246303e-06, + "loss": 0.0176, + "step": 37444 + }, + { + "epoch": 4.91, + "learning_rate": 6.875250861278239e-06, + "loss": 0.0189, + "step": 37445 + }, + { + "epoch": 4.91, + "learning_rate": 6.874747096098366e-06, + "loss": 0.0205, + "step": 37446 + }, + { + "epoch": 4.91, + "learning_rate": 6.8742433397081e-06, + "loss": 0.0201, + "step": 37447 + }, + { + "epoch": 4.91, + "learning_rate": 6.87373959210886e-06, + "loss": 0.0197, + "step": 37448 + }, + { + "epoch": 4.91, + "learning_rate": 6.873235853302058e-06, + "loss": 0.0145, + "step": 37449 + }, + { + "epoch": 4.91, + "learning_rate": 6.872732123289116e-06, + "loss": 0.0151, + "step": 37450 + }, + { + "epoch": 4.91, + "learning_rate": 6.8722284020714484e-06, + "loss": 0.0156, + "step": 37451 + }, + { + "epoch": 4.91, + "learning_rate": 6.871724689650468e-06, + "loss": 0.0204, + "step": 37452 + }, + { + "epoch": 4.91, + "learning_rate": 6.8712209860275956e-06, + "loss": 0.0211, + "step": 37453 + }, + { + "epoch": 4.91, + "learning_rate": 6.870717291204252e-06, + "loss": 0.0207, + "step": 37454 + }, + { + "epoch": 4.91, + "learning_rate": 6.870213605181847e-06, + "loss": 0.0184, + "step": 37455 + }, + { + "epoch": 4.91, + "learning_rate": 6.869709927961801e-06, + "loss": 0.018, + "step": 37456 + }, + { + "epoch": 4.91, + "learning_rate": 6.869206259545527e-06, + "loss": 0.0192, + "step": 37457 + }, + { + "epoch": 4.91, + "learning_rate": 6.868702599934445e-06, + "loss": 0.0217, + "step": 37458 + }, + { + "epoch": 4.91, + "learning_rate": 6.8681989491299705e-06, + "loss": 0.0201, + "step": 37459 + }, + { + "epoch": 4.91, + "learning_rate": 6.867695307133518e-06, + "loss": 0.0198, + "step": 37460 + }, + { + "epoch": 4.91, + "learning_rate": 6.867191673946509e-06, + "loss": 0.0167, + "step": 37461 + }, + { + "epoch": 4.91, + "learning_rate": 6.866688049570354e-06, + "loss": 0.0156, + "step": 37462 + }, + { + "epoch": 4.91, + "learning_rate": 6.8661844340064684e-06, + "loss": 0.016, + "step": 37463 + }, + { + "epoch": 4.91, + "learning_rate": 6.865680827256276e-06, + "loss": 0.0183, + "step": 37464 + }, + { + "epoch": 4.91, + "learning_rate": 6.86517722932119e-06, + "loss": 0.0216, + "step": 37465 + }, + { + "epoch": 4.91, + "learning_rate": 6.864673640202626e-06, + "loss": 0.0182, + "step": 37466 + }, + { + "epoch": 4.91, + "learning_rate": 6.864170059902001e-06, + "loss": 0.0172, + "step": 37467 + }, + { + "epoch": 4.91, + "learning_rate": 6.8636664884207285e-06, + "loss": 0.0186, + "step": 37468 + }, + { + "epoch": 4.91, + "learning_rate": 6.86316292576023e-06, + "loss": 0.018, + "step": 37469 + }, + { + "epoch": 4.91, + "learning_rate": 6.862659371921916e-06, + "loss": 0.017, + "step": 37470 + }, + { + "epoch": 4.91, + "learning_rate": 6.8621558269072085e-06, + "loss": 0.0184, + "step": 37471 + }, + { + "epoch": 4.91, + "learning_rate": 6.861652290717518e-06, + "loss": 0.013, + "step": 37472 + }, + { + "epoch": 4.91, + "learning_rate": 6.861148763354262e-06, + "loss": 0.0196, + "step": 37473 + }, + { + "epoch": 4.91, + "learning_rate": 6.8606452448188575e-06, + "loss": 0.0137, + "step": 37474 + }, + { + "epoch": 4.91, + "learning_rate": 6.860141735112724e-06, + "loss": 0.0164, + "step": 37475 + }, + { + "epoch": 4.91, + "learning_rate": 6.859638234237273e-06, + "loss": 0.02, + "step": 37476 + }, + { + "epoch": 4.91, + "learning_rate": 6.859134742193925e-06, + "loss": 0.0204, + "step": 37477 + }, + { + "epoch": 4.91, + "learning_rate": 6.858631258984091e-06, + "loss": 0.0157, + "step": 37478 + }, + { + "epoch": 4.91, + "learning_rate": 6.85812778460919e-06, + "loss": 0.0148, + "step": 37479 + }, + { + "epoch": 4.91, + "learning_rate": 6.8576243190706375e-06, + "loss": 0.015, + "step": 37480 + }, + { + "epoch": 4.91, + "learning_rate": 6.857120862369848e-06, + "loss": 0.0136, + "step": 37481 + }, + { + "epoch": 4.91, + "learning_rate": 6.856617414508242e-06, + "loss": 0.0144, + "step": 37482 + }, + { + "epoch": 4.91, + "learning_rate": 6.8561139754872285e-06, + "loss": 0.0142, + "step": 37483 + }, + { + "epoch": 4.91, + "learning_rate": 6.855610545308229e-06, + "loss": 0.0215, + "step": 37484 + }, + { + "epoch": 4.91, + "learning_rate": 6.855107123972654e-06, + "loss": 0.0177, + "step": 37485 + }, + { + "epoch": 4.91, + "learning_rate": 6.8546037114819245e-06, + "loss": 0.0144, + "step": 37486 + }, + { + "epoch": 4.91, + "learning_rate": 6.854100307837454e-06, + "loss": 0.0262, + "step": 37487 + }, + { + "epoch": 4.91, + "learning_rate": 6.8535969130406635e-06, + "loss": 0.0173, + "step": 37488 + }, + { + "epoch": 4.91, + "learning_rate": 6.853093527092959e-06, + "loss": 0.0135, + "step": 37489 + }, + { + "epoch": 4.91, + "learning_rate": 6.852590149995765e-06, + "loss": 0.0168, + "step": 37490 + }, + { + "epoch": 4.91, + "learning_rate": 6.852086781750491e-06, + "loss": 0.0178, + "step": 37491 + }, + { + "epoch": 4.91, + "learning_rate": 6.8515834223585575e-06, + "loss": 0.0186, + "step": 37492 + }, + { + "epoch": 4.91, + "learning_rate": 6.851080071821377e-06, + "loss": 0.0146, + "step": 37493 + }, + { + "epoch": 4.91, + "learning_rate": 6.850576730140364e-06, + "loss": 0.0159, + "step": 37494 + }, + { + "epoch": 4.91, + "learning_rate": 6.850073397316938e-06, + "loss": 0.0156, + "step": 37495 + }, + { + "epoch": 4.91, + "learning_rate": 6.849570073352512e-06, + "loss": 0.0159, + "step": 37496 + }, + { + "epoch": 4.91, + "learning_rate": 6.849066758248502e-06, + "loss": 0.0179, + "step": 37497 + }, + { + "epoch": 4.91, + "learning_rate": 6.848563452006326e-06, + "loss": 0.018, + "step": 37498 + }, + { + "epoch": 4.91, + "learning_rate": 6.848060154627396e-06, + "loss": 0.0165, + "step": 37499 + }, + { + "epoch": 4.91, + "learning_rate": 6.847556866113131e-06, + "loss": 0.0168, + "step": 37500 + }, + { + "epoch": 4.91, + "learning_rate": 6.847053586464942e-06, + "loss": 0.0187, + "step": 37501 + }, + { + "epoch": 4.91, + "learning_rate": 6.846550315684247e-06, + "loss": 0.0142, + "step": 37502 + }, + { + "epoch": 4.91, + "learning_rate": 6.846047053772464e-06, + "loss": 0.0183, + "step": 37503 + }, + { + "epoch": 4.91, + "learning_rate": 6.8455438007310026e-06, + "loss": 0.0138, + "step": 37504 + }, + { + "epoch": 4.91, + "learning_rate": 6.8450405565612845e-06, + "loss": 0.0141, + "step": 37505 + }, + { + "epoch": 4.91, + "learning_rate": 6.8445373212647194e-06, + "loss": 0.0177, + "step": 37506 + }, + { + "epoch": 4.91, + "learning_rate": 6.844034094842722e-06, + "loss": 0.0176, + "step": 37507 + }, + { + "epoch": 4.91, + "learning_rate": 6.843530877296714e-06, + "loss": 0.0196, + "step": 37508 + }, + { + "epoch": 4.91, + "learning_rate": 6.843027668628108e-06, + "loss": 0.0173, + "step": 37509 + }, + { + "epoch": 4.91, + "learning_rate": 6.842524468838317e-06, + "loss": 0.019, + "step": 37510 + }, + { + "epoch": 4.91, + "learning_rate": 6.842021277928761e-06, + "loss": 0.0185, + "step": 37511 + }, + { + "epoch": 4.91, + "learning_rate": 6.841518095900849e-06, + "loss": 0.0204, + "step": 37512 + }, + { + "epoch": 4.91, + "learning_rate": 6.841014922756002e-06, + "loss": 0.0228, + "step": 37513 + }, + { + "epoch": 4.91, + "learning_rate": 6.840511758495629e-06, + "loss": 0.0182, + "step": 37514 + }, + { + "epoch": 4.91, + "learning_rate": 6.840008603121151e-06, + "loss": 0.0211, + "step": 37515 + }, + { + "epoch": 4.91, + "learning_rate": 6.839505456633978e-06, + "loss": 0.0173, + "step": 37516 + }, + { + "epoch": 4.91, + "learning_rate": 6.8390023190355285e-06, + "loss": 0.0165, + "step": 37517 + }, + { + "epoch": 4.91, + "learning_rate": 6.838499190327214e-06, + "loss": 0.0185, + "step": 37518 + }, + { + "epoch": 4.91, + "learning_rate": 6.837996070510457e-06, + "loss": 0.0124, + "step": 37519 + }, + { + "epoch": 4.91, + "learning_rate": 6.837492959586665e-06, + "loss": 0.0186, + "step": 37520 + }, + { + "epoch": 4.91, + "learning_rate": 6.836989857557257e-06, + "loss": 0.0199, + "step": 37521 + }, + { + "epoch": 4.92, + "learning_rate": 6.836486764423646e-06, + "loss": 0.0162, + "step": 37522 + }, + { + "epoch": 4.92, + "learning_rate": 6.8359836801872455e-06, + "loss": 0.014, + "step": 37523 + }, + { + "epoch": 4.92, + "learning_rate": 6.835480604849476e-06, + "loss": 0.0146, + "step": 37524 + }, + { + "epoch": 4.92, + "learning_rate": 6.834977538411746e-06, + "loss": 0.0167, + "step": 37525 + }, + { + "epoch": 4.92, + "learning_rate": 6.834474480875477e-06, + "loss": 0.0156, + "step": 37526 + }, + { + "epoch": 4.92, + "learning_rate": 6.833971432242077e-06, + "loss": 0.0188, + "step": 37527 + }, + { + "epoch": 4.92, + "learning_rate": 6.833468392512965e-06, + "loss": 0.0167, + "step": 37528 + }, + { + "epoch": 4.92, + "learning_rate": 6.832965361689549e-06, + "loss": 0.0209, + "step": 37529 + }, + { + "epoch": 4.92, + "learning_rate": 6.832462339773255e-06, + "loss": 0.0202, + "step": 37530 + }, + { + "epoch": 4.92, + "learning_rate": 6.831959326765489e-06, + "loss": 0.0186, + "step": 37531 + }, + { + "epoch": 4.92, + "learning_rate": 6.831456322667673e-06, + "loss": 0.0163, + "step": 37532 + }, + { + "epoch": 4.92, + "learning_rate": 6.830953327481213e-06, + "loss": 0.0182, + "step": 37533 + }, + { + "epoch": 4.92, + "learning_rate": 6.830450341207532e-06, + "loss": 0.0223, + "step": 37534 + }, + { + "epoch": 4.92, + "learning_rate": 6.829947363848037e-06, + "loss": 0.0188, + "step": 37535 + }, + { + "epoch": 4.92, + "learning_rate": 6.82944439540415e-06, + "loss": 0.0134, + "step": 37536 + }, + { + "epoch": 4.92, + "learning_rate": 6.828941435877278e-06, + "loss": 0.0194, + "step": 37537 + }, + { + "epoch": 4.92, + "learning_rate": 6.82843848526884e-06, + "loss": 0.0171, + "step": 37538 + }, + { + "epoch": 4.92, + "learning_rate": 6.827935543580253e-06, + "loss": 0.016, + "step": 37539 + }, + { + "epoch": 4.92, + "learning_rate": 6.827432610812923e-06, + "loss": 0.0203, + "step": 37540 + }, + { + "epoch": 4.92, + "learning_rate": 6.826929686968272e-06, + "loss": 0.0184, + "step": 37541 + }, + { + "epoch": 4.92, + "learning_rate": 6.826426772047714e-06, + "loss": 0.017, + "step": 37542 + }, + { + "epoch": 4.92, + "learning_rate": 6.82592386605266e-06, + "loss": 0.0175, + "step": 37543 + }, + { + "epoch": 4.92, + "learning_rate": 6.8254209689845265e-06, + "loss": 0.0148, + "step": 37544 + }, + { + "epoch": 4.92, + "learning_rate": 6.824918080844729e-06, + "loss": 0.0159, + "step": 37545 + }, + { + "epoch": 4.92, + "learning_rate": 6.82441520163468e-06, + "loss": 0.0188, + "step": 37546 + }, + { + "epoch": 4.92, + "learning_rate": 6.8239123313557944e-06, + "loss": 0.0127, + "step": 37547 + }, + { + "epoch": 4.92, + "learning_rate": 6.823409470009485e-06, + "loss": 0.0207, + "step": 37548 + }, + { + "epoch": 4.92, + "learning_rate": 6.8229066175971685e-06, + "loss": 0.0197, + "step": 37549 + }, + { + "epoch": 4.92, + "learning_rate": 6.8224037741202546e-06, + "loss": 0.0146, + "step": 37550 + }, + { + "epoch": 4.92, + "learning_rate": 6.821900939580162e-06, + "loss": 0.0181, + "step": 37551 + }, + { + "epoch": 4.92, + "learning_rate": 6.8213981139783034e-06, + "loss": 0.0164, + "step": 37552 + }, + { + "epoch": 4.92, + "learning_rate": 6.820895297316097e-06, + "loss": 0.0168, + "step": 37553 + }, + { + "epoch": 4.92, + "learning_rate": 6.82039248959495e-06, + "loss": 0.015, + "step": 37554 + }, + { + "epoch": 4.92, + "learning_rate": 6.8198896908162825e-06, + "loss": 0.0174, + "step": 37555 + }, + { + "epoch": 4.92, + "learning_rate": 6.819386900981504e-06, + "loss": 0.0208, + "step": 37556 + }, + { + "epoch": 4.92, + "learning_rate": 6.818884120092033e-06, + "loss": 0.0169, + "step": 37557 + }, + { + "epoch": 4.92, + "learning_rate": 6.818381348149276e-06, + "loss": 0.0149, + "step": 37558 + }, + { + "epoch": 4.92, + "learning_rate": 6.817878585154656e-06, + "loss": 0.0178, + "step": 37559 + }, + { + "epoch": 4.92, + "learning_rate": 6.817375831109583e-06, + "loss": 0.0154, + "step": 37560 + }, + { + "epoch": 4.92, + "learning_rate": 6.8168730860154676e-06, + "loss": 0.018, + "step": 37561 + }, + { + "epoch": 4.92, + "learning_rate": 6.8163703498737284e-06, + "loss": 0.0171, + "step": 37562 + }, + { + "epoch": 4.92, + "learning_rate": 6.81586762268578e-06, + "loss": 0.0153, + "step": 37563 + }, + { + "epoch": 4.92, + "learning_rate": 6.815364904453032e-06, + "loss": 0.0207, + "step": 37564 + }, + { + "epoch": 4.92, + "learning_rate": 6.8148621951769045e-06, + "loss": 0.019, + "step": 37565 + }, + { + "epoch": 4.92, + "learning_rate": 6.814359494858804e-06, + "loss": 0.0189, + "step": 37566 + }, + { + "epoch": 4.92, + "learning_rate": 6.813856803500149e-06, + "loss": 0.018, + "step": 37567 + }, + { + "epoch": 4.92, + "learning_rate": 6.813354121102353e-06, + "loss": 0.0173, + "step": 37568 + }, + { + "epoch": 4.92, + "learning_rate": 6.812851447666826e-06, + "loss": 0.0209, + "step": 37569 + }, + { + "epoch": 4.92, + "learning_rate": 6.812348783194988e-06, + "loss": 0.0195, + "step": 37570 + }, + { + "epoch": 4.92, + "learning_rate": 6.811846127688246e-06, + "loss": 0.0162, + "step": 37571 + }, + { + "epoch": 4.92, + "learning_rate": 6.8113434811480164e-06, + "loss": 0.0173, + "step": 37572 + }, + { + "epoch": 4.92, + "learning_rate": 6.810840843575715e-06, + "loss": 0.0126, + "step": 37573 + }, + { + "epoch": 4.92, + "learning_rate": 6.810338214972756e-06, + "loss": 0.0144, + "step": 37574 + }, + { + "epoch": 4.92, + "learning_rate": 6.809835595340548e-06, + "loss": 0.0159, + "step": 37575 + }, + { + "epoch": 4.92, + "learning_rate": 6.809332984680511e-06, + "loss": 0.0166, + "step": 37576 + }, + { + "epoch": 4.92, + "learning_rate": 6.808830382994052e-06, + "loss": 0.0132, + "step": 37577 + }, + { + "epoch": 4.92, + "learning_rate": 6.8083277902825896e-06, + "loss": 0.0154, + "step": 37578 + }, + { + "epoch": 4.92, + "learning_rate": 6.8078252065475336e-06, + "loss": 0.0147, + "step": 37579 + }, + { + "epoch": 4.92, + "learning_rate": 6.8073226317903e-06, + "loss": 0.0189, + "step": 37580 + }, + { + "epoch": 4.92, + "learning_rate": 6.806820066012303e-06, + "loss": 0.0175, + "step": 37581 + }, + { + "epoch": 4.92, + "learning_rate": 6.806317509214953e-06, + "loss": 0.0184, + "step": 37582 + }, + { + "epoch": 4.92, + "learning_rate": 6.805814961399662e-06, + "loss": 0.0171, + "step": 37583 + }, + { + "epoch": 4.92, + "learning_rate": 6.8053124225678495e-06, + "loss": 0.0144, + "step": 37584 + }, + { + "epoch": 4.92, + "learning_rate": 6.804809892720925e-06, + "loss": 0.0184, + "step": 37585 + }, + { + "epoch": 4.92, + "learning_rate": 6.804307371860303e-06, + "loss": 0.0157, + "step": 37586 + }, + { + "epoch": 4.92, + "learning_rate": 6.803804859987396e-06, + "loss": 0.0144, + "step": 37587 + }, + { + "epoch": 4.92, + "learning_rate": 6.803302357103618e-06, + "loss": 0.0158, + "step": 37588 + }, + { + "epoch": 4.92, + "learning_rate": 6.802799863210383e-06, + "loss": 0.0168, + "step": 37589 + }, + { + "epoch": 4.92, + "learning_rate": 6.802297378309101e-06, + "loss": 0.0113, + "step": 37590 + }, + { + "epoch": 4.92, + "learning_rate": 6.801794902401189e-06, + "loss": 0.0191, + "step": 37591 + }, + { + "epoch": 4.92, + "learning_rate": 6.801292435488057e-06, + "loss": 0.0155, + "step": 37592 + }, + { + "epoch": 4.92, + "learning_rate": 6.800789977571123e-06, + "loss": 0.0167, + "step": 37593 + }, + { + "epoch": 4.92, + "learning_rate": 6.800287528651788e-06, + "loss": 0.0135, + "step": 37594 + }, + { + "epoch": 4.92, + "learning_rate": 6.799785088731482e-06, + "loss": 0.0149, + "step": 37595 + }, + { + "epoch": 4.92, + "learning_rate": 6.799282657811608e-06, + "loss": 0.0157, + "step": 37596 + }, + { + "epoch": 4.92, + "learning_rate": 6.798780235893583e-06, + "loss": 0.0168, + "step": 37597 + }, + { + "epoch": 4.93, + "learning_rate": 6.798277822978815e-06, + "loss": 0.0161, + "step": 37598 + }, + { + "epoch": 4.93, + "learning_rate": 6.797775419068722e-06, + "loss": 0.0169, + "step": 37599 + }, + { + "epoch": 4.93, + "learning_rate": 6.7972730241647146e-06, + "loss": 0.0197, + "step": 37600 + }, + { + "epoch": 4.93, + "learning_rate": 6.796770638268207e-06, + "loss": 0.0145, + "step": 37601 + }, + { + "epoch": 4.93, + "learning_rate": 6.79626826138061e-06, + "loss": 0.0155, + "step": 37602 + }, + { + "epoch": 4.93, + "learning_rate": 6.795765893503339e-06, + "loss": 0.0192, + "step": 37603 + }, + { + "epoch": 4.93, + "learning_rate": 6.795263534637806e-06, + "loss": 0.0183, + "step": 37604 + }, + { + "epoch": 4.93, + "learning_rate": 6.79476118478542e-06, + "loss": 0.0221, + "step": 37605 + }, + { + "epoch": 4.93, + "learning_rate": 6.7942588439476e-06, + "loss": 0.0172, + "step": 37606 + }, + { + "epoch": 4.93, + "learning_rate": 6.793756512125758e-06, + "loss": 0.0169, + "step": 37607 + }, + { + "epoch": 4.93, + "learning_rate": 6.793254189321304e-06, + "loss": 0.0157, + "step": 37608 + }, + { + "epoch": 4.93, + "learning_rate": 6.7927518755356505e-06, + "loss": 0.0168, + "step": 37609 + }, + { + "epoch": 4.93, + "learning_rate": 6.792249570770213e-06, + "loss": 0.0162, + "step": 37610 + }, + { + "epoch": 4.93, + "learning_rate": 6.791747275026402e-06, + "loss": 0.0193, + "step": 37611 + }, + { + "epoch": 4.93, + "learning_rate": 6.791244988305633e-06, + "loss": 0.012, + "step": 37612 + }, + { + "epoch": 4.93, + "learning_rate": 6.790742710609315e-06, + "loss": 0.0151, + "step": 37613 + }, + { + "epoch": 4.93, + "learning_rate": 6.790240441938863e-06, + "loss": 0.0157, + "step": 37614 + }, + { + "epoch": 4.93, + "learning_rate": 6.789738182295687e-06, + "loss": 0.0153, + "step": 37615 + }, + { + "epoch": 4.93, + "learning_rate": 6.789235931681199e-06, + "loss": 0.014, + "step": 37616 + }, + { + "epoch": 4.93, + "learning_rate": 6.788733690096818e-06, + "loss": 0.0131, + "step": 37617 + }, + { + "epoch": 4.93, + "learning_rate": 6.7882314575439524e-06, + "loss": 0.0119, + "step": 37618 + }, + { + "epoch": 4.93, + "learning_rate": 6.7877292340240144e-06, + "loss": 0.0174, + "step": 37619 + }, + { + "epoch": 4.93, + "learning_rate": 6.787227019538418e-06, + "loss": 0.0231, + "step": 37620 + }, + { + "epoch": 4.93, + "learning_rate": 6.786724814088572e-06, + "loss": 0.0185, + "step": 37621 + }, + { + "epoch": 4.93, + "learning_rate": 6.786222617675894e-06, + "loss": 0.016, + "step": 37622 + }, + { + "epoch": 4.93, + "learning_rate": 6.785720430301792e-06, + "loss": 0.0161, + "step": 37623 + }, + { + "epoch": 4.93, + "learning_rate": 6.78521825196768e-06, + "loss": 0.0136, + "step": 37624 + }, + { + "epoch": 4.93, + "learning_rate": 6.784716082674971e-06, + "loss": 0.0212, + "step": 37625 + }, + { + "epoch": 4.93, + "learning_rate": 6.784213922425077e-06, + "loss": 0.0156, + "step": 37626 + }, + { + "epoch": 4.93, + "learning_rate": 6.7837117712194065e-06, + "loss": 0.0191, + "step": 37627 + }, + { + "epoch": 4.93, + "learning_rate": 6.783209629059379e-06, + "loss": 0.0185, + "step": 37628 + }, + { + "epoch": 4.93, + "learning_rate": 6.7827074959464035e-06, + "loss": 0.0165, + "step": 37629 + }, + { + "epoch": 4.93, + "learning_rate": 6.78220537188189e-06, + "loss": 0.0203, + "step": 37630 + }, + { + "epoch": 4.93, + "learning_rate": 6.781703256867256e-06, + "loss": 0.0154, + "step": 37631 + }, + { + "epoch": 4.93, + "learning_rate": 6.781201150903907e-06, + "loss": 0.0146, + "step": 37632 + }, + { + "epoch": 4.93, + "learning_rate": 6.7806990539932605e-06, + "loss": 0.0195, + "step": 37633 + }, + { + "epoch": 4.93, + "learning_rate": 6.780196966136724e-06, + "loss": 0.0152, + "step": 37634 + }, + { + "epoch": 4.93, + "learning_rate": 6.779694887335715e-06, + "loss": 0.0183, + "step": 37635 + }, + { + "epoch": 4.93, + "learning_rate": 6.77919281759164e-06, + "loss": 0.0134, + "step": 37636 + }, + { + "epoch": 4.93, + "learning_rate": 6.778690756905917e-06, + "loss": 0.0147, + "step": 37637 + }, + { + "epoch": 4.93, + "learning_rate": 6.778188705279947e-06, + "loss": 0.0215, + "step": 37638 + }, + { + "epoch": 4.93, + "learning_rate": 6.777686662715155e-06, + "loss": 0.0206, + "step": 37639 + }, + { + "epoch": 4.93, + "learning_rate": 6.777184629212948e-06, + "loss": 0.017, + "step": 37640 + }, + { + "epoch": 4.93, + "learning_rate": 6.776682604774738e-06, + "loss": 0.0166, + "step": 37641 + }, + { + "epoch": 4.93, + "learning_rate": 6.776180589401935e-06, + "loss": 0.0146, + "step": 37642 + }, + { + "epoch": 4.93, + "learning_rate": 6.7756785830959545e-06, + "loss": 0.0155, + "step": 37643 + }, + { + "epoch": 4.93, + "learning_rate": 6.775176585858205e-06, + "loss": 0.0176, + "step": 37644 + }, + { + "epoch": 4.93, + "learning_rate": 6.774674597690099e-06, + "loss": 0.0173, + "step": 37645 + }, + { + "epoch": 4.93, + "learning_rate": 6.774172618593051e-06, + "loss": 0.0197, + "step": 37646 + }, + { + "epoch": 4.93, + "learning_rate": 6.773670648568468e-06, + "loss": 0.0205, + "step": 37647 + }, + { + "epoch": 4.93, + "learning_rate": 6.773168687617767e-06, + "loss": 0.0205, + "step": 37648 + }, + { + "epoch": 4.93, + "learning_rate": 6.772666735742353e-06, + "loss": 0.0139, + "step": 37649 + }, + { + "epoch": 4.93, + "learning_rate": 6.772164792943645e-06, + "loss": 0.0171, + "step": 37650 + }, + { + "epoch": 4.93, + "learning_rate": 6.7716628592230525e-06, + "loss": 0.0097, + "step": 37651 + }, + { + "epoch": 4.93, + "learning_rate": 6.771160934581984e-06, + "loss": 0.0159, + "step": 37652 + }, + { + "epoch": 4.93, + "learning_rate": 6.770659019021854e-06, + "loss": 0.0168, + "step": 37653 + }, + { + "epoch": 4.93, + "learning_rate": 6.770157112544076e-06, + "loss": 0.0184, + "step": 37654 + }, + { + "epoch": 4.93, + "learning_rate": 6.769655215150057e-06, + "loss": 0.0191, + "step": 37655 + }, + { + "epoch": 4.93, + "learning_rate": 6.769153326841213e-06, + "loss": 0.0219, + "step": 37656 + }, + { + "epoch": 4.93, + "learning_rate": 6.7686514476189505e-06, + "loss": 0.0147, + "step": 37657 + }, + { + "epoch": 4.93, + "learning_rate": 6.768149577484686e-06, + "loss": 0.0159, + "step": 37658 + }, + { + "epoch": 4.93, + "learning_rate": 6.767647716439826e-06, + "loss": 0.0157, + "step": 37659 + }, + { + "epoch": 4.93, + "learning_rate": 6.767145864485782e-06, + "loss": 0.0159, + "step": 37660 + }, + { + "epoch": 4.93, + "learning_rate": 6.766644021623971e-06, + "loss": 0.0176, + "step": 37661 + }, + { + "epoch": 4.93, + "learning_rate": 6.766142187855805e-06, + "loss": 0.0184, + "step": 37662 + }, + { + "epoch": 4.93, + "learning_rate": 6.765640363182688e-06, + "loss": 0.0162, + "step": 37663 + }, + { + "epoch": 4.93, + "learning_rate": 6.7651385476060385e-06, + "loss": 0.0172, + "step": 37664 + }, + { + "epoch": 4.93, + "learning_rate": 6.764636741127261e-06, + "loss": 0.0197, + "step": 37665 + }, + { + "epoch": 4.93, + "learning_rate": 6.7641349437477755e-06, + "loss": 0.0159, + "step": 37666 + }, + { + "epoch": 4.93, + "learning_rate": 6.763633155468983e-06, + "loss": 0.02, + "step": 37667 + }, + { + "epoch": 4.93, + "learning_rate": 6.763131376292301e-06, + "loss": 0.0173, + "step": 37668 + }, + { + "epoch": 4.93, + "learning_rate": 6.762629606219142e-06, + "loss": 0.0182, + "step": 37669 + }, + { + "epoch": 4.93, + "learning_rate": 6.7621278452509125e-06, + "loss": 0.0164, + "step": 37670 + }, + { + "epoch": 4.93, + "learning_rate": 6.761626093389023e-06, + "loss": 0.0179, + "step": 37671 + }, + { + "epoch": 4.93, + "learning_rate": 6.7611243506348935e-06, + "loss": 0.0164, + "step": 37672 + }, + { + "epoch": 4.93, + "learning_rate": 6.760622616989926e-06, + "loss": 0.0196, + "step": 37673 + }, + { + "epoch": 4.94, + "learning_rate": 6.760120892455535e-06, + "loss": 0.021, + "step": 37674 + }, + { + "epoch": 4.94, + "learning_rate": 6.759619177033134e-06, + "loss": 0.0233, + "step": 37675 + }, + { + "epoch": 4.94, + "learning_rate": 6.759117470724129e-06, + "loss": 0.0156, + "step": 37676 + }, + { + "epoch": 4.94, + "learning_rate": 6.7586157735299355e-06, + "loss": 0.0183, + "step": 37677 + }, + { + "epoch": 4.94, + "learning_rate": 6.758114085451961e-06, + "loss": 0.0182, + "step": 37678 + }, + { + "epoch": 4.94, + "learning_rate": 6.75761240649162e-06, + "loss": 0.0136, + "step": 37679 + }, + { + "epoch": 4.94, + "learning_rate": 6.757110736650319e-06, + "loss": 0.0181, + "step": 37680 + }, + { + "epoch": 4.94, + "learning_rate": 6.756609075929471e-06, + "loss": 0.0155, + "step": 37681 + }, + { + "epoch": 4.94, + "learning_rate": 6.756107424330486e-06, + "loss": 0.0146, + "step": 37682 + }, + { + "epoch": 4.94, + "learning_rate": 6.755605781854779e-06, + "loss": 0.0192, + "step": 37683 + }, + { + "epoch": 4.94, + "learning_rate": 6.755104148503757e-06, + "loss": 0.0179, + "step": 37684 + }, + { + "epoch": 4.94, + "learning_rate": 6.754602524278835e-06, + "loss": 0.0205, + "step": 37685 + }, + { + "epoch": 4.94, + "learning_rate": 6.754100909181417e-06, + "loss": 0.0206, + "step": 37686 + }, + { + "epoch": 4.94, + "learning_rate": 6.753599303212918e-06, + "loss": 0.0172, + "step": 37687 + }, + { + "epoch": 4.94, + "learning_rate": 6.753097706374749e-06, + "loss": 0.0172, + "step": 37688 + }, + { + "epoch": 4.94, + "learning_rate": 6.752596118668318e-06, + "loss": 0.0181, + "step": 37689 + }, + { + "epoch": 4.94, + "learning_rate": 6.75209454009504e-06, + "loss": 0.0181, + "step": 37690 + }, + { + "epoch": 4.94, + "learning_rate": 6.751592970656322e-06, + "loss": 0.0162, + "step": 37691 + }, + { + "epoch": 4.94, + "learning_rate": 6.7510914103535715e-06, + "loss": 0.0175, + "step": 37692 + }, + { + "epoch": 4.94, + "learning_rate": 6.75058985918821e-06, + "loss": 0.0122, + "step": 37693 + }, + { + "epoch": 4.94, + "learning_rate": 6.750088317161638e-06, + "loss": 0.0206, + "step": 37694 + }, + { + "epoch": 4.94, + "learning_rate": 6.74958678427527e-06, + "loss": 0.0142, + "step": 37695 + }, + { + "epoch": 4.94, + "learning_rate": 6.749085260530517e-06, + "loss": 0.0163, + "step": 37696 + }, + { + "epoch": 4.94, + "learning_rate": 6.748583745928788e-06, + "loss": 0.0157, + "step": 37697 + }, + { + "epoch": 4.94, + "learning_rate": 6.7480822404714956e-06, + "loss": 0.0131, + "step": 37698 + }, + { + "epoch": 4.94, + "learning_rate": 6.747580744160046e-06, + "loss": 0.0168, + "step": 37699 + }, + { + "epoch": 4.94, + "learning_rate": 6.747079256995855e-06, + "loss": 0.0185, + "step": 37700 + }, + { + "epoch": 4.94, + "learning_rate": 6.7465777789803275e-06, + "loss": 0.0168, + "step": 37701 + }, + { + "epoch": 4.94, + "learning_rate": 6.746076310114881e-06, + "loss": 0.0209, + "step": 37702 + }, + { + "epoch": 4.94, + "learning_rate": 6.745574850400914e-06, + "loss": 0.0162, + "step": 37703 + }, + { + "epoch": 4.94, + "learning_rate": 6.74507339983985e-06, + "loss": 0.0184, + "step": 37704 + }, + { + "epoch": 4.94, + "learning_rate": 6.744571958433093e-06, + "loss": 0.019, + "step": 37705 + }, + { + "epoch": 4.94, + "learning_rate": 6.744070526182055e-06, + "loss": 0.019, + "step": 37706 + }, + { + "epoch": 4.94, + "learning_rate": 6.7435691030881435e-06, + "loss": 0.0165, + "step": 37707 + }, + { + "epoch": 4.94, + "learning_rate": 6.743067689152772e-06, + "loss": 0.0154, + "step": 37708 + }, + { + "epoch": 4.94, + "learning_rate": 6.742566284377348e-06, + "loss": 0.0127, + "step": 37709 + }, + { + "epoch": 4.94, + "learning_rate": 6.7420648887632825e-06, + "loss": 0.0116, + "step": 37710 + }, + { + "epoch": 4.94, + "learning_rate": 6.741563502311988e-06, + "loss": 0.0172, + "step": 37711 + }, + { + "epoch": 4.94, + "learning_rate": 6.741062125024871e-06, + "loss": 0.0159, + "step": 37712 + }, + { + "epoch": 4.94, + "learning_rate": 6.740560756903345e-06, + "loss": 0.0132, + "step": 37713 + }, + { + "epoch": 4.94, + "learning_rate": 6.740059397948815e-06, + "loss": 0.0161, + "step": 37714 + }, + { + "epoch": 4.94, + "learning_rate": 6.739558048162696e-06, + "loss": 0.0144, + "step": 37715 + }, + { + "epoch": 4.94, + "learning_rate": 6.7390567075463966e-06, + "loss": 0.0157, + "step": 37716 + }, + { + "epoch": 4.94, + "learning_rate": 6.7385553761013255e-06, + "loss": 0.0146, + "step": 37717 + }, + { + "epoch": 4.94, + "learning_rate": 6.7380540538288954e-06, + "loss": 0.0162, + "step": 37718 + }, + { + "epoch": 4.94, + "learning_rate": 6.737552740730515e-06, + "loss": 0.0211, + "step": 37719 + }, + { + "epoch": 4.94, + "learning_rate": 6.7370514368075916e-06, + "loss": 0.0169, + "step": 37720 + }, + { + "epoch": 4.94, + "learning_rate": 6.73655014206154e-06, + "loss": 0.0152, + "step": 37721 + }, + { + "epoch": 4.94, + "learning_rate": 6.736048856493764e-06, + "loss": 0.0151, + "step": 37722 + }, + { + "epoch": 4.94, + "learning_rate": 6.735547580105681e-06, + "loss": 0.0158, + "step": 37723 + }, + { + "epoch": 4.94, + "learning_rate": 6.735046312898693e-06, + "loss": 0.0145, + "step": 37724 + }, + { + "epoch": 4.94, + "learning_rate": 6.734545054874212e-06, + "loss": 0.0156, + "step": 37725 + }, + { + "epoch": 4.94, + "learning_rate": 6.734043806033651e-06, + "loss": 0.0157, + "step": 37726 + }, + { + "epoch": 4.94, + "learning_rate": 6.733542566378419e-06, + "loss": 0.0174, + "step": 37727 + }, + { + "epoch": 4.94, + "learning_rate": 6.733041335909925e-06, + "loss": 0.0144, + "step": 37728 + }, + { + "epoch": 4.94, + "learning_rate": 6.7325401146295775e-06, + "loss": 0.0108, + "step": 37729 + }, + { + "epoch": 4.94, + "learning_rate": 6.732038902538787e-06, + "loss": 0.0239, + "step": 37730 + }, + { + "epoch": 4.94, + "learning_rate": 6.731537699638961e-06, + "loss": 0.0163, + "step": 37731 + }, + { + "epoch": 4.94, + "learning_rate": 6.731036505931515e-06, + "loss": 0.0139, + "step": 37732 + }, + { + "epoch": 4.94, + "learning_rate": 6.730535321417852e-06, + "loss": 0.0183, + "step": 37733 + }, + { + "epoch": 4.94, + "learning_rate": 6.7300341460993865e-06, + "loss": 0.0172, + "step": 37734 + }, + { + "epoch": 4.94, + "learning_rate": 6.729532979977524e-06, + "loss": 0.0168, + "step": 37735 + }, + { + "epoch": 4.94, + "learning_rate": 6.729031823053673e-06, + "loss": 0.0183, + "step": 37736 + }, + { + "epoch": 4.94, + "learning_rate": 6.72853067532925e-06, + "loss": 0.0147, + "step": 37737 + }, + { + "epoch": 4.94, + "learning_rate": 6.728029536805659e-06, + "loss": 0.0172, + "step": 37738 + }, + { + "epoch": 4.94, + "learning_rate": 6.72752840748431e-06, + "loss": 0.0206, + "step": 37739 + }, + { + "epoch": 4.94, + "learning_rate": 6.727027287366617e-06, + "loss": 0.0136, + "step": 37740 + }, + { + "epoch": 4.94, + "learning_rate": 6.72652617645398e-06, + "loss": 0.0155, + "step": 37741 + }, + { + "epoch": 4.94, + "learning_rate": 6.726025074747818e-06, + "loss": 0.013, + "step": 37742 + }, + { + "epoch": 4.94, + "learning_rate": 6.7255239822495336e-06, + "loss": 0.0159, + "step": 37743 + }, + { + "epoch": 4.94, + "learning_rate": 6.725022898960541e-06, + "loss": 0.0134, + "step": 37744 + }, + { + "epoch": 4.94, + "learning_rate": 6.724521824882245e-06, + "loss": 0.0153, + "step": 37745 + }, + { + "epoch": 4.94, + "learning_rate": 6.724020760016057e-06, + "loss": 0.0171, + "step": 37746 + }, + { + "epoch": 4.94, + "learning_rate": 6.7235197043633835e-06, + "loss": 0.0151, + "step": 37747 + }, + { + "epoch": 4.94, + "learning_rate": 6.723018657925642e-06, + "loss": 0.0163, + "step": 37748 + }, + { + "epoch": 4.94, + "learning_rate": 6.722517620704232e-06, + "loss": 0.0166, + "step": 37749 + }, + { + "epoch": 4.94, + "learning_rate": 6.72201659270057e-06, + "loss": 0.0152, + "step": 37750 + }, + { + "epoch": 4.95, + "learning_rate": 6.721515573916059e-06, + "loss": 0.0184, + "step": 37751 + }, + { + "epoch": 4.95, + "learning_rate": 6.721014564352113e-06, + "loss": 0.0217, + "step": 37752 + }, + { + "epoch": 4.95, + "learning_rate": 6.720513564010137e-06, + "loss": 0.0206, + "step": 37753 + }, + { + "epoch": 4.95, + "learning_rate": 6.720012572891542e-06, + "loss": 0.0168, + "step": 37754 + }, + { + "epoch": 4.95, + "learning_rate": 6.719511590997739e-06, + "loss": 0.0194, + "step": 37755 + }, + { + "epoch": 4.95, + "learning_rate": 6.7190106183301325e-06, + "loss": 0.0157, + "step": 37756 + }, + { + "epoch": 4.95, + "learning_rate": 6.718509654890137e-06, + "loss": 0.0126, + "step": 37757 + }, + { + "epoch": 4.95, + "learning_rate": 6.718008700679152e-06, + "loss": 0.0201, + "step": 37758 + }, + { + "epoch": 4.95, + "learning_rate": 6.717507755698598e-06, + "loss": 0.0188, + "step": 37759 + }, + { + "epoch": 4.95, + "learning_rate": 6.717006819949876e-06, + "loss": 0.0185, + "step": 37760 + }, + { + "epoch": 4.95, + "learning_rate": 6.716505893434402e-06, + "loss": 0.0171, + "step": 37761 + }, + { + "epoch": 4.95, + "learning_rate": 6.716004976153576e-06, + "loss": 0.0182, + "step": 37762 + }, + { + "epoch": 4.95, + "learning_rate": 6.715504068108814e-06, + "loss": 0.0141, + "step": 37763 + }, + { + "epoch": 4.95, + "learning_rate": 6.715003169301521e-06, + "loss": 0.0162, + "step": 37764 + }, + { + "epoch": 4.95, + "learning_rate": 6.714502279733106e-06, + "loss": 0.0204, + "step": 37765 + }, + { + "epoch": 4.95, + "learning_rate": 6.714001399404979e-06, + "loss": 0.0203, + "step": 37766 + }, + { + "epoch": 4.95, + "learning_rate": 6.713500528318548e-06, + "loss": 0.015, + "step": 37767 + }, + { + "epoch": 4.95, + "learning_rate": 6.712999666475222e-06, + "loss": 0.0149, + "step": 37768 + }, + { + "epoch": 4.95, + "learning_rate": 6.712498813876406e-06, + "loss": 0.017, + "step": 37769 + }, + { + "epoch": 4.95, + "learning_rate": 6.711997970523514e-06, + "loss": 0.0139, + "step": 37770 + }, + { + "epoch": 4.95, + "learning_rate": 6.711497136417955e-06, + "loss": 0.0177, + "step": 37771 + }, + { + "epoch": 4.95, + "learning_rate": 6.7109963115611336e-06, + "loss": 0.0131, + "step": 37772 + }, + { + "epoch": 4.95, + "learning_rate": 6.710495495954461e-06, + "loss": 0.0168, + "step": 37773 + }, + { + "epoch": 4.95, + "learning_rate": 6.709994689599342e-06, + "loss": 0.0209, + "step": 37774 + }, + { + "epoch": 4.95, + "learning_rate": 6.709493892497189e-06, + "loss": 0.0133, + "step": 37775 + }, + { + "epoch": 4.95, + "learning_rate": 6.708993104649411e-06, + "loss": 0.0181, + "step": 37776 + }, + { + "epoch": 4.95, + "learning_rate": 6.708492326057412e-06, + "loss": 0.0158, + "step": 37777 + }, + { + "epoch": 4.95, + "learning_rate": 6.707991556722606e-06, + "loss": 0.0183, + "step": 37778 + }, + { + "epoch": 4.95, + "learning_rate": 6.707490796646396e-06, + "loss": 0.0186, + "step": 37779 + }, + { + "epoch": 4.95, + "learning_rate": 6.70699004583019e-06, + "loss": 0.0166, + "step": 37780 + }, + { + "epoch": 4.95, + "learning_rate": 6.706489304275402e-06, + "loss": 0.0173, + "step": 37781 + }, + { + "epoch": 4.95, + "learning_rate": 6.705988571983439e-06, + "loss": 0.0168, + "step": 37782 + }, + { + "epoch": 4.95, + "learning_rate": 6.705487848955708e-06, + "loss": 0.0148, + "step": 37783 + }, + { + "epoch": 4.95, + "learning_rate": 6.704987135193615e-06, + "loss": 0.0188, + "step": 37784 + }, + { + "epoch": 4.95, + "learning_rate": 6.704486430698571e-06, + "loss": 0.0216, + "step": 37785 + }, + { + "epoch": 4.95, + "learning_rate": 6.703985735471985e-06, + "loss": 0.0143, + "step": 37786 + }, + { + "epoch": 4.95, + "learning_rate": 6.703485049515262e-06, + "loss": 0.0139, + "step": 37787 + }, + { + "epoch": 4.95, + "learning_rate": 6.7029843728298124e-06, + "loss": 0.0173, + "step": 37788 + }, + { + "epoch": 4.95, + "learning_rate": 6.702483705417042e-06, + "loss": 0.0172, + "step": 37789 + }, + { + "epoch": 4.95, + "learning_rate": 6.701983047278361e-06, + "loss": 0.0152, + "step": 37790 + }, + { + "epoch": 4.95, + "learning_rate": 6.701482398415174e-06, + "loss": 0.0168, + "step": 37791 + }, + { + "epoch": 4.95, + "learning_rate": 6.700981758828897e-06, + "loss": 0.0158, + "step": 37792 + }, + { + "epoch": 4.95, + "learning_rate": 6.700481128520932e-06, + "loss": 0.0137, + "step": 37793 + }, + { + "epoch": 4.95, + "learning_rate": 6.699980507492689e-06, + "loss": 0.0168, + "step": 37794 + }, + { + "epoch": 4.95, + "learning_rate": 6.699479895745573e-06, + "loss": 0.0174, + "step": 37795 + }, + { + "epoch": 4.95, + "learning_rate": 6.698979293280995e-06, + "loss": 0.0192, + "step": 37796 + }, + { + "epoch": 4.95, + "learning_rate": 6.698478700100364e-06, + "loss": 0.0142, + "step": 37797 + }, + { + "epoch": 4.95, + "learning_rate": 6.697978116205082e-06, + "loss": 0.0189, + "step": 37798 + }, + { + "epoch": 4.95, + "learning_rate": 6.697477541596564e-06, + "loss": 0.0196, + "step": 37799 + }, + { + "epoch": 4.95, + "learning_rate": 6.696976976276212e-06, + "loss": 0.0203, + "step": 37800 + }, + { + "epoch": 4.95, + "learning_rate": 6.696476420245439e-06, + "loss": 0.0188, + "step": 37801 + }, + { + "epoch": 4.95, + "learning_rate": 6.6959758735056464e-06, + "loss": 0.018, + "step": 37802 + }, + { + "epoch": 4.95, + "learning_rate": 6.695475336058247e-06, + "loss": 0.0165, + "step": 37803 + }, + { + "epoch": 4.95, + "learning_rate": 6.694974807904647e-06, + "loss": 0.0196, + "step": 37804 + }, + { + "epoch": 4.95, + "learning_rate": 6.6944742890462576e-06, + "loss": 0.0156, + "step": 37805 + }, + { + "epoch": 4.95, + "learning_rate": 6.69397377948448e-06, + "loss": 0.0182, + "step": 37806 + }, + { + "epoch": 4.95, + "learning_rate": 6.693473279220728e-06, + "loss": 0.0152, + "step": 37807 + }, + { + "epoch": 4.95, + "learning_rate": 6.692972788256403e-06, + "loss": 0.0166, + "step": 37808 + }, + { + "epoch": 4.95, + "learning_rate": 6.692472306592919e-06, + "loss": 0.0177, + "step": 37809 + }, + { + "epoch": 4.95, + "learning_rate": 6.6919718342316785e-06, + "loss": 0.0183, + "step": 37810 + }, + { + "epoch": 4.95, + "learning_rate": 6.6914713711740905e-06, + "loss": 0.0163, + "step": 37811 + }, + { + "epoch": 4.95, + "learning_rate": 6.690970917421567e-06, + "loss": 0.0229, + "step": 37812 + }, + { + "epoch": 4.95, + "learning_rate": 6.690470472975505e-06, + "loss": 0.0216, + "step": 37813 + }, + { + "epoch": 4.95, + "learning_rate": 6.689970037837321e-06, + "loss": 0.017, + "step": 37814 + }, + { + "epoch": 4.95, + "learning_rate": 6.689469612008423e-06, + "loss": 0.0123, + "step": 37815 + }, + { + "epoch": 4.95, + "learning_rate": 6.6889691954902135e-06, + "loss": 0.0178, + "step": 37816 + }, + { + "epoch": 4.95, + "learning_rate": 6.688468788284103e-06, + "loss": 0.0168, + "step": 37817 + }, + { + "epoch": 4.95, + "learning_rate": 6.687968390391495e-06, + "loss": 0.0141, + "step": 37818 + }, + { + "epoch": 4.95, + "learning_rate": 6.6874680018138015e-06, + "loss": 0.0133, + "step": 37819 + }, + { + "epoch": 4.95, + "learning_rate": 6.686967622552428e-06, + "loss": 0.0131, + "step": 37820 + }, + { + "epoch": 4.95, + "learning_rate": 6.686467252608782e-06, + "loss": 0.0133, + "step": 37821 + }, + { + "epoch": 4.95, + "learning_rate": 6.68596689198427e-06, + "loss": 0.0112, + "step": 37822 + }, + { + "epoch": 4.95, + "learning_rate": 6.685466540680296e-06, + "loss": 0.0166, + "step": 37823 + }, + { + "epoch": 4.95, + "learning_rate": 6.684966198698273e-06, + "loss": 0.0149, + "step": 37824 + }, + { + "epoch": 4.95, + "learning_rate": 6.684465866039607e-06, + "loss": 0.0169, + "step": 37825 + }, + { + "epoch": 4.95, + "learning_rate": 6.683965542705706e-06, + "loss": 0.0138, + "step": 37826 + }, + { + "epoch": 4.96, + "learning_rate": 6.683465228697971e-06, + "loss": 0.0172, + "step": 37827 + }, + { + "epoch": 4.96, + "learning_rate": 6.682964924017819e-06, + "loss": 0.0186, + "step": 37828 + }, + { + "epoch": 4.96, + "learning_rate": 6.682464628666647e-06, + "loss": 0.0169, + "step": 37829 + }, + { + "epoch": 4.96, + "learning_rate": 6.6819643426458704e-06, + "loss": 0.0166, + "step": 37830 + }, + { + "epoch": 4.96, + "learning_rate": 6.681464065956889e-06, + "loss": 0.0162, + "step": 37831 + }, + { + "epoch": 4.96, + "learning_rate": 6.680963798601116e-06, + "loss": 0.0178, + "step": 37832 + }, + { + "epoch": 4.96, + "learning_rate": 6.6804635405799526e-06, + "loss": 0.0134, + "step": 37833 + }, + { + "epoch": 4.96, + "learning_rate": 6.679963291894807e-06, + "loss": 0.0146, + "step": 37834 + }, + { + "epoch": 4.96, + "learning_rate": 6.6794630525470905e-06, + "loss": 0.0181, + "step": 37835 + }, + { + "epoch": 4.96, + "learning_rate": 6.678962822538209e-06, + "loss": 0.0152, + "step": 37836 + }, + { + "epoch": 4.96, + "learning_rate": 6.6784626018695656e-06, + "loss": 0.0156, + "step": 37837 + }, + { + "epoch": 4.96, + "learning_rate": 6.677962390542572e-06, + "loss": 0.0155, + "step": 37838 + }, + { + "epoch": 4.96, + "learning_rate": 6.677462188558628e-06, + "loss": 0.0201, + "step": 37839 + }, + { + "epoch": 4.96, + "learning_rate": 6.676961995919147e-06, + "loss": 0.0168, + "step": 37840 + }, + { + "epoch": 4.96, + "learning_rate": 6.676461812625534e-06, + "loss": 0.0193, + "step": 37841 + }, + { + "epoch": 4.96, + "learning_rate": 6.6759616386791935e-06, + "loss": 0.0185, + "step": 37842 + }, + { + "epoch": 4.96, + "learning_rate": 6.675461474081537e-06, + "loss": 0.016, + "step": 37843 + }, + { + "epoch": 4.96, + "learning_rate": 6.674961318833966e-06, + "loss": 0.0118, + "step": 37844 + }, + { + "epoch": 4.96, + "learning_rate": 6.674461172937885e-06, + "loss": 0.0174, + "step": 37845 + }, + { + "epoch": 4.96, + "learning_rate": 6.6739610363947085e-06, + "loss": 0.0188, + "step": 37846 + }, + { + "epoch": 4.96, + "learning_rate": 6.673460909205841e-06, + "loss": 0.0179, + "step": 37847 + }, + { + "epoch": 4.96, + "learning_rate": 6.672960791372687e-06, + "loss": 0.0165, + "step": 37848 + }, + { + "epoch": 4.96, + "learning_rate": 6.672460682896653e-06, + "loss": 0.0149, + "step": 37849 + }, + { + "epoch": 4.96, + "learning_rate": 6.671960583779147e-06, + "loss": 0.016, + "step": 37850 + }, + { + "epoch": 4.96, + "learning_rate": 6.671460494021574e-06, + "loss": 0.0177, + "step": 37851 + }, + { + "epoch": 4.96, + "learning_rate": 6.67096041362534e-06, + "loss": 0.0177, + "step": 37852 + }, + { + "epoch": 4.96, + "learning_rate": 6.670460342591856e-06, + "loss": 0.0184, + "step": 37853 + }, + { + "epoch": 4.96, + "learning_rate": 6.669960280922521e-06, + "loss": 0.0174, + "step": 37854 + }, + { + "epoch": 4.96, + "learning_rate": 6.669460228618747e-06, + "loss": 0.0205, + "step": 37855 + }, + { + "epoch": 4.96, + "learning_rate": 6.668960185681935e-06, + "loss": 0.0197, + "step": 37856 + }, + { + "epoch": 4.96, + "learning_rate": 6.668460152113501e-06, + "loss": 0.0156, + "step": 37857 + }, + { + "epoch": 4.96, + "learning_rate": 6.667960127914841e-06, + "loss": 0.0174, + "step": 37858 + }, + { + "epoch": 4.96, + "learning_rate": 6.6674601130873695e-06, + "loss": 0.0198, + "step": 37859 + }, + { + "epoch": 4.96, + "learning_rate": 6.666960107632487e-06, + "loss": 0.017, + "step": 37860 + }, + { + "epoch": 4.96, + "learning_rate": 6.6664601115516e-06, + "loss": 0.0199, + "step": 37861 + }, + { + "epoch": 4.96, + "learning_rate": 6.665960124846121e-06, + "loss": 0.0162, + "step": 37862 + }, + { + "epoch": 4.96, + "learning_rate": 6.6654601475174485e-06, + "loss": 0.0177, + "step": 37863 + }, + { + "epoch": 4.96, + "learning_rate": 6.664960179566993e-06, + "loss": 0.0122, + "step": 37864 + }, + { + "epoch": 4.96, + "learning_rate": 6.664460220996159e-06, + "loss": 0.015, + "step": 37865 + }, + { + "epoch": 4.96, + "learning_rate": 6.663960271806354e-06, + "loss": 0.021, + "step": 37866 + }, + { + "epoch": 4.96, + "learning_rate": 6.663460331998979e-06, + "loss": 0.0145, + "step": 37867 + }, + { + "epoch": 4.96, + "learning_rate": 6.662960401575445e-06, + "loss": 0.0173, + "step": 37868 + }, + { + "epoch": 4.96, + "learning_rate": 6.662460480537159e-06, + "loss": 0.0159, + "step": 37869 + }, + { + "epoch": 4.96, + "learning_rate": 6.661960568885526e-06, + "loss": 0.0123, + "step": 37870 + }, + { + "epoch": 4.96, + "learning_rate": 6.661460666621951e-06, + "loss": 0.0165, + "step": 37871 + }, + { + "epoch": 4.96, + "learning_rate": 6.66096077374784e-06, + "loss": 0.0202, + "step": 37872 + }, + { + "epoch": 4.96, + "learning_rate": 6.660460890264597e-06, + "loss": 0.0172, + "step": 37873 + }, + { + "epoch": 4.96, + "learning_rate": 6.659961016173633e-06, + "loss": 0.0159, + "step": 37874 + }, + { + "epoch": 4.96, + "learning_rate": 6.659461151476348e-06, + "loss": 0.0196, + "step": 37875 + }, + { + "epoch": 4.96, + "learning_rate": 6.658961296174151e-06, + "loss": 0.0164, + "step": 37876 + }, + { + "epoch": 4.96, + "learning_rate": 6.65846145026845e-06, + "loss": 0.0148, + "step": 37877 + }, + { + "epoch": 4.96, + "learning_rate": 6.657961613760644e-06, + "loss": 0.0147, + "step": 37878 + }, + { + "epoch": 4.96, + "learning_rate": 6.657461786652144e-06, + "loss": 0.0154, + "step": 37879 + }, + { + "epoch": 4.96, + "learning_rate": 6.656961968944358e-06, + "loss": 0.016, + "step": 37880 + }, + { + "epoch": 4.96, + "learning_rate": 6.656462160638686e-06, + "loss": 0.0164, + "step": 37881 + }, + { + "epoch": 4.96, + "learning_rate": 6.655962361736538e-06, + "loss": 0.0146, + "step": 37882 + }, + { + "epoch": 4.96, + "learning_rate": 6.655462572239315e-06, + "loss": 0.0146, + "step": 37883 + }, + { + "epoch": 4.96, + "learning_rate": 6.654962792148426e-06, + "loss": 0.012, + "step": 37884 + }, + { + "epoch": 4.96, + "learning_rate": 6.654463021465279e-06, + "loss": 0.0158, + "step": 37885 + }, + { + "epoch": 4.96, + "learning_rate": 6.653963260191273e-06, + "loss": 0.0192, + "step": 37886 + }, + { + "epoch": 4.96, + "learning_rate": 6.65346350832782e-06, + "loss": 0.0197, + "step": 37887 + }, + { + "epoch": 4.96, + "learning_rate": 6.652963765876321e-06, + "loss": 0.0163, + "step": 37888 + }, + { + "epoch": 4.96, + "learning_rate": 6.652464032838179e-06, + "loss": 0.0176, + "step": 37889 + }, + { + "epoch": 4.96, + "learning_rate": 6.651964309214809e-06, + "loss": 0.0172, + "step": 37890 + }, + { + "epoch": 4.96, + "learning_rate": 6.651464595007611e-06, + "loss": 0.014, + "step": 37891 + }, + { + "epoch": 4.96, + "learning_rate": 6.650964890217989e-06, + "loss": 0.0196, + "step": 37892 + }, + { + "epoch": 4.96, + "learning_rate": 6.650465194847352e-06, + "loss": 0.0194, + "step": 37893 + }, + { + "epoch": 4.96, + "learning_rate": 6.649965508897101e-06, + "loss": 0.017, + "step": 37894 + }, + { + "epoch": 4.96, + "learning_rate": 6.649465832368646e-06, + "loss": 0.0168, + "step": 37895 + }, + { + "epoch": 4.96, + "learning_rate": 6.648966165263388e-06, + "loss": 0.0189, + "step": 37896 + }, + { + "epoch": 4.96, + "learning_rate": 6.648466507582734e-06, + "loss": 0.018, + "step": 37897 + }, + { + "epoch": 4.96, + "learning_rate": 6.64796685932809e-06, + "loss": 0.015, + "step": 37898 + }, + { + "epoch": 4.96, + "learning_rate": 6.64746722050086e-06, + "loss": 0.0147, + "step": 37899 + }, + { + "epoch": 4.96, + "learning_rate": 6.6469675911024485e-06, + "loss": 0.0187, + "step": 37900 + }, + { + "epoch": 4.96, + "learning_rate": 6.646467971134265e-06, + "loss": 0.0163, + "step": 37901 + }, + { + "epoch": 4.96, + "learning_rate": 6.645968360597711e-06, + "loss": 0.0141, + "step": 37902 + }, + { + "epoch": 4.97, + "learning_rate": 6.645468759494194e-06, + "loss": 0.0185, + "step": 37903 + }, + { + "epoch": 4.97, + "learning_rate": 6.644969167825115e-06, + "loss": 0.0171, + "step": 37904 + }, + { + "epoch": 4.97, + "learning_rate": 6.644469585591882e-06, + "loss": 0.0155, + "step": 37905 + }, + { + "epoch": 4.97, + "learning_rate": 6.6439700127959004e-06, + "loss": 0.0156, + "step": 37906 + }, + { + "epoch": 4.97, + "learning_rate": 6.643470449438574e-06, + "loss": 0.0206, + "step": 37907 + }, + { + "epoch": 4.97, + "learning_rate": 6.642970895521309e-06, + "loss": 0.0148, + "step": 37908 + }, + { + "epoch": 4.97, + "learning_rate": 6.642471351045507e-06, + "loss": 0.0154, + "step": 37909 + }, + { + "epoch": 4.97, + "learning_rate": 6.6419718160125805e-06, + "loss": 0.0182, + "step": 37910 + }, + { + "epoch": 4.97, + "learning_rate": 6.6414722904239204e-06, + "loss": 0.0159, + "step": 37911 + }, + { + "epoch": 4.97, + "learning_rate": 6.640972774280948e-06, + "loss": 0.0158, + "step": 37912 + }, + { + "epoch": 4.97, + "learning_rate": 6.640473267585059e-06, + "loss": 0.0208, + "step": 37913 + }, + { + "epoch": 4.97, + "learning_rate": 6.639973770337662e-06, + "loss": 0.0162, + "step": 37914 + }, + { + "epoch": 4.97, + "learning_rate": 6.639474282540157e-06, + "loss": 0.0152, + "step": 37915 + }, + { + "epoch": 4.97, + "learning_rate": 6.6389748041939536e-06, + "loss": 0.014, + "step": 37916 + }, + { + "epoch": 4.97, + "learning_rate": 6.638475335300454e-06, + "loss": 0.018, + "step": 37917 + }, + { + "epoch": 4.97, + "learning_rate": 6.6379758758610645e-06, + "loss": 0.0159, + "step": 37918 + }, + { + "epoch": 4.97, + "learning_rate": 6.637476425877186e-06, + "loss": 0.0175, + "step": 37919 + }, + { + "epoch": 4.97, + "learning_rate": 6.636976985350228e-06, + "loss": 0.0135, + "step": 37920 + }, + { + "epoch": 4.97, + "learning_rate": 6.636477554281593e-06, + "loss": 0.0177, + "step": 37921 + }, + { + "epoch": 4.97, + "learning_rate": 6.635978132672683e-06, + "loss": 0.0165, + "step": 37922 + }, + { + "epoch": 4.97, + "learning_rate": 6.635478720524907e-06, + "loss": 0.0215, + "step": 37923 + }, + { + "epoch": 4.97, + "learning_rate": 6.634979317839669e-06, + "loss": 0.0178, + "step": 37924 + }, + { + "epoch": 4.97, + "learning_rate": 6.634479924618373e-06, + "loss": 0.0147, + "step": 37925 + }, + { + "epoch": 4.97, + "learning_rate": 6.6339805408624215e-06, + "loss": 0.0152, + "step": 37926 + }, + { + "epoch": 4.97, + "learning_rate": 6.633481166573222e-06, + "loss": 0.0173, + "step": 37927 + }, + { + "epoch": 4.97, + "learning_rate": 6.6329818017521765e-06, + "loss": 0.0212, + "step": 37928 + }, + { + "epoch": 4.97, + "learning_rate": 6.632482446400692e-06, + "loss": 0.0176, + "step": 37929 + }, + { + "epoch": 4.97, + "learning_rate": 6.631983100520168e-06, + "loss": 0.0143, + "step": 37930 + }, + { + "epoch": 4.97, + "learning_rate": 6.631483764112016e-06, + "loss": 0.0168, + "step": 37931 + }, + { + "epoch": 4.97, + "learning_rate": 6.630984437177634e-06, + "loss": 0.0175, + "step": 37932 + }, + { + "epoch": 4.97, + "learning_rate": 6.6304851197184265e-06, + "loss": 0.0169, + "step": 37933 + }, + { + "epoch": 4.97, + "learning_rate": 6.629985811735802e-06, + "loss": 0.0149, + "step": 37934 + }, + { + "epoch": 4.97, + "learning_rate": 6.629486513231165e-06, + "loss": 0.0142, + "step": 37935 + }, + { + "epoch": 4.97, + "learning_rate": 6.628987224205916e-06, + "loss": 0.018, + "step": 37936 + }, + { + "epoch": 4.97, + "learning_rate": 6.628487944661464e-06, + "loss": 0.0176, + "step": 37937 + }, + { + "epoch": 4.97, + "learning_rate": 6.6279886745992065e-06, + "loss": 0.0163, + "step": 37938 + }, + { + "epoch": 4.97, + "learning_rate": 6.627489414020554e-06, + "loss": 0.0144, + "step": 37939 + }, + { + "epoch": 4.97, + "learning_rate": 6.626990162926905e-06, + "loss": 0.0145, + "step": 37940 + }, + { + "epoch": 4.97, + "learning_rate": 6.626490921319667e-06, + "loss": 0.0155, + "step": 37941 + }, + { + "epoch": 4.97, + "learning_rate": 6.625991689200247e-06, + "loss": 0.019, + "step": 37942 + }, + { + "epoch": 4.97, + "learning_rate": 6.625492466570041e-06, + "loss": 0.0144, + "step": 37943 + }, + { + "epoch": 4.97, + "learning_rate": 6.624993253430458e-06, + "loss": 0.0169, + "step": 37944 + }, + { + "epoch": 4.97, + "learning_rate": 6.624494049782905e-06, + "loss": 0.0149, + "step": 37945 + }, + { + "epoch": 4.97, + "learning_rate": 6.623994855628781e-06, + "loss": 0.0168, + "step": 37946 + }, + { + "epoch": 4.97, + "learning_rate": 6.623495670969491e-06, + "loss": 0.0154, + "step": 37947 + }, + { + "epoch": 4.97, + "learning_rate": 6.622996495806442e-06, + "loss": 0.0196, + "step": 37948 + }, + { + "epoch": 4.97, + "learning_rate": 6.6224973301410335e-06, + "loss": 0.0233, + "step": 37949 + }, + { + "epoch": 4.97, + "learning_rate": 6.621998173974673e-06, + "loss": 0.0165, + "step": 37950 + }, + { + "epoch": 4.97, + "learning_rate": 6.621499027308762e-06, + "loss": 0.0135, + "step": 37951 + }, + { + "epoch": 4.97, + "learning_rate": 6.620999890144706e-06, + "loss": 0.0155, + "step": 37952 + }, + { + "epoch": 4.97, + "learning_rate": 6.620500762483906e-06, + "loss": 0.0195, + "step": 37953 + }, + { + "epoch": 4.97, + "learning_rate": 6.620001644327766e-06, + "loss": 0.0173, + "step": 37954 + }, + { + "epoch": 4.97, + "learning_rate": 6.619502535677694e-06, + "loss": 0.0129, + "step": 37955 + }, + { + "epoch": 4.97, + "learning_rate": 6.619003436535092e-06, + "loss": 0.0151, + "step": 37956 + }, + { + "epoch": 4.97, + "learning_rate": 6.61850434690136e-06, + "loss": 0.012, + "step": 37957 + }, + { + "epoch": 4.97, + "learning_rate": 6.6180052667779085e-06, + "loss": 0.0135, + "step": 37958 + }, + { + "epoch": 4.97, + "learning_rate": 6.6175061961661334e-06, + "loss": 0.022, + "step": 37959 + }, + { + "epoch": 4.97, + "learning_rate": 6.617007135067445e-06, + "loss": 0.0146, + "step": 37960 + }, + { + "epoch": 4.97, + "learning_rate": 6.6165080834832415e-06, + "loss": 0.0168, + "step": 37961 + }, + { + "epoch": 4.97, + "learning_rate": 6.616009041414928e-06, + "loss": 0.0157, + "step": 37962 + }, + { + "epoch": 4.97, + "learning_rate": 6.615510008863912e-06, + "loss": 0.0183, + "step": 37963 + }, + { + "epoch": 4.97, + "learning_rate": 6.615010985831593e-06, + "loss": 0.0198, + "step": 37964 + }, + { + "epoch": 4.97, + "learning_rate": 6.614511972319371e-06, + "loss": 0.0151, + "step": 37965 + }, + { + "epoch": 4.97, + "learning_rate": 6.614012968328659e-06, + "loss": 0.0226, + "step": 37966 + }, + { + "epoch": 4.97, + "learning_rate": 6.613513973860852e-06, + "loss": 0.0219, + "step": 37967 + }, + { + "epoch": 4.97, + "learning_rate": 6.613014988917361e-06, + "loss": 0.0167, + "step": 37968 + }, + { + "epoch": 4.97, + "learning_rate": 6.612516013499581e-06, + "loss": 0.0172, + "step": 37969 + }, + { + "epoch": 4.97, + "learning_rate": 6.612017047608921e-06, + "loss": 0.0143, + "step": 37970 + }, + { + "epoch": 4.97, + "learning_rate": 6.611518091246783e-06, + "loss": 0.0153, + "step": 37971 + }, + { + "epoch": 4.97, + "learning_rate": 6.6110191444145675e-06, + "loss": 0.0144, + "step": 37972 + }, + { + "epoch": 4.97, + "learning_rate": 6.610520207113684e-06, + "loss": 0.0175, + "step": 37973 + }, + { + "epoch": 4.97, + "learning_rate": 6.6100212793455285e-06, + "loss": 0.0156, + "step": 37974 + }, + { + "epoch": 4.97, + "learning_rate": 6.60952236111151e-06, + "loss": 0.0173, + "step": 37975 + }, + { + "epoch": 4.97, + "learning_rate": 6.609023452413024e-06, + "loss": 0.0152, + "step": 37976 + }, + { + "epoch": 4.97, + "learning_rate": 6.608524553251484e-06, + "loss": 0.0152, + "step": 37977 + }, + { + "epoch": 4.97, + "learning_rate": 6.6080256636282855e-06, + "loss": 0.0174, + "step": 37978 + }, + { + "epoch": 4.97, + "learning_rate": 6.607526783544838e-06, + "loss": 0.0156, + "step": 37979 + }, + { + "epoch": 4.98, + "learning_rate": 6.607027913002538e-06, + "loss": 0.0189, + "step": 37980 + }, + { + "epoch": 4.98, + "learning_rate": 6.606529052002794e-06, + "loss": 0.0164, + "step": 37981 + }, + { + "epoch": 4.98, + "learning_rate": 6.606030200547002e-06, + "loss": 0.0171, + "step": 37982 + }, + { + "epoch": 4.98, + "learning_rate": 6.605531358636574e-06, + "loss": 0.0198, + "step": 37983 + }, + { + "epoch": 4.98, + "learning_rate": 6.6050325262729054e-06, + "loss": 0.0209, + "step": 37984 + }, + { + "epoch": 4.98, + "learning_rate": 6.604533703457401e-06, + "loss": 0.0131, + "step": 37985 + }, + { + "epoch": 4.98, + "learning_rate": 6.604034890191468e-06, + "loss": 0.0133, + "step": 37986 + }, + { + "epoch": 4.98, + "learning_rate": 6.603536086476502e-06, + "loss": 0.0201, + "step": 37987 + }, + { + "epoch": 4.98, + "learning_rate": 6.6030372923139116e-06, + "loss": 0.0144, + "step": 37988 + }, + { + "epoch": 4.98, + "learning_rate": 6.6025385077051e-06, + "loss": 0.0133, + "step": 37989 + }, + { + "epoch": 4.98, + "learning_rate": 6.602039732651465e-06, + "loss": 0.0207, + "step": 37990 + }, + { + "epoch": 4.98, + "learning_rate": 6.601540967154413e-06, + "loss": 0.017, + "step": 37991 + }, + { + "epoch": 4.98, + "learning_rate": 6.601042211215348e-06, + "loss": 0.0172, + "step": 37992 + }, + { + "epoch": 4.98, + "learning_rate": 6.600543464835669e-06, + "loss": 0.0172, + "step": 37993 + }, + { + "epoch": 4.98, + "learning_rate": 6.600044728016782e-06, + "loss": 0.0169, + "step": 37994 + }, + { + "epoch": 4.98, + "learning_rate": 6.599546000760087e-06, + "loss": 0.0219, + "step": 37995 + }, + { + "epoch": 4.98, + "learning_rate": 6.59904728306699e-06, + "loss": 0.0185, + "step": 37996 + }, + { + "epoch": 4.98, + "learning_rate": 6.5985485749388876e-06, + "loss": 0.0208, + "step": 37997 + }, + { + "epoch": 4.98, + "learning_rate": 6.598049876377186e-06, + "loss": 0.0179, + "step": 37998 + }, + { + "epoch": 4.98, + "learning_rate": 6.5975511873832886e-06, + "loss": 0.0175, + "step": 37999 + }, + { + "epoch": 4.98, + "learning_rate": 6.5970525079586e-06, + "loss": 0.0158, + "step": 38000 + }, + { + "epoch": 4.98, + "learning_rate": 6.596553838104518e-06, + "loss": 0.0147, + "step": 38001 + }, + { + "epoch": 4.98, + "learning_rate": 6.596055177822449e-06, + "loss": 0.0185, + "step": 38002 + }, + { + "epoch": 4.98, + "learning_rate": 6.595556527113791e-06, + "loss": 0.0198, + "step": 38003 + }, + { + "epoch": 4.98, + "learning_rate": 6.595057885979952e-06, + "loss": 0.0177, + "step": 38004 + }, + { + "epoch": 4.98, + "learning_rate": 6.594559254422328e-06, + "loss": 0.0168, + "step": 38005 + }, + { + "epoch": 4.98, + "learning_rate": 6.594060632442326e-06, + "loss": 0.0159, + "step": 38006 + }, + { + "epoch": 4.98, + "learning_rate": 6.5935620200413485e-06, + "loss": 0.0139, + "step": 38007 + }, + { + "epoch": 4.98, + "learning_rate": 6.5930634172207954e-06, + "loss": 0.0116, + "step": 38008 + }, + { + "epoch": 4.98, + "learning_rate": 6.592564823982066e-06, + "loss": 0.019, + "step": 38009 + }, + { + "epoch": 4.98, + "learning_rate": 6.592066240326573e-06, + "loss": 0.0149, + "step": 38010 + }, + { + "epoch": 4.98, + "learning_rate": 6.591567666255709e-06, + "loss": 0.0174, + "step": 38011 + }, + { + "epoch": 4.98, + "learning_rate": 6.5910691017708795e-06, + "loss": 0.0162, + "step": 38012 + }, + { + "epoch": 4.98, + "learning_rate": 6.590570546873488e-06, + "loss": 0.0196, + "step": 38013 + }, + { + "epoch": 4.98, + "learning_rate": 6.590072001564935e-06, + "loss": 0.0171, + "step": 38014 + }, + { + "epoch": 4.98, + "learning_rate": 6.589573465846624e-06, + "loss": 0.0174, + "step": 38015 + }, + { + "epoch": 4.98, + "learning_rate": 6.589074939719953e-06, + "loss": 0.016, + "step": 38016 + }, + { + "epoch": 4.98, + "learning_rate": 6.58857642318633e-06, + "loss": 0.0199, + "step": 38017 + }, + { + "epoch": 4.98, + "learning_rate": 6.588077916247151e-06, + "loss": 0.018, + "step": 38018 + }, + { + "epoch": 4.98, + "learning_rate": 6.587579418903825e-06, + "loss": 0.0165, + "step": 38019 + }, + { + "epoch": 4.98, + "learning_rate": 6.587080931157743e-06, + "loss": 0.0179, + "step": 38020 + }, + { + "epoch": 4.98, + "learning_rate": 6.586582453010321e-06, + "loss": 0.0144, + "step": 38021 + }, + { + "epoch": 4.98, + "learning_rate": 6.586083984462951e-06, + "loss": 0.0161, + "step": 38022 + }, + { + "epoch": 4.98, + "learning_rate": 6.58558552551704e-06, + "loss": 0.0174, + "step": 38023 + }, + { + "epoch": 4.98, + "learning_rate": 6.585087076173987e-06, + "loss": 0.0145, + "step": 38024 + }, + { + "epoch": 4.98, + "learning_rate": 6.5845886364351965e-06, + "loss": 0.0165, + "step": 38025 + }, + { + "epoch": 4.98, + "learning_rate": 6.584090206302065e-06, + "loss": 0.0159, + "step": 38026 + }, + { + "epoch": 4.98, + "learning_rate": 6.583591785776e-06, + "loss": 0.0145, + "step": 38027 + }, + { + "epoch": 4.98, + "learning_rate": 6.583093374858402e-06, + "loss": 0.0169, + "step": 38028 + }, + { + "epoch": 4.98, + "learning_rate": 6.582594973550671e-06, + "loss": 0.0164, + "step": 38029 + }, + { + "epoch": 4.98, + "learning_rate": 6.58209658185421e-06, + "loss": 0.0168, + "step": 38030 + }, + { + "epoch": 4.98, + "learning_rate": 6.5815981997704185e-06, + "loss": 0.0148, + "step": 38031 + }, + { + "epoch": 4.98, + "learning_rate": 6.5810998273007e-06, + "loss": 0.0172, + "step": 38032 + }, + { + "epoch": 4.98, + "learning_rate": 6.580601464446459e-06, + "loss": 0.0156, + "step": 38033 + }, + { + "epoch": 4.98, + "learning_rate": 6.580103111209092e-06, + "loss": 0.0192, + "step": 38034 + }, + { + "epoch": 4.98, + "learning_rate": 6.579604767590002e-06, + "loss": 0.0147, + "step": 38035 + }, + { + "epoch": 4.98, + "learning_rate": 6.579106433590596e-06, + "loss": 0.0142, + "step": 38036 + }, + { + "epoch": 4.98, + "learning_rate": 6.578608109212267e-06, + "loss": 0.0147, + "step": 38037 + }, + { + "epoch": 4.98, + "learning_rate": 6.578109794456422e-06, + "loss": 0.0156, + "step": 38038 + }, + { + "epoch": 4.98, + "learning_rate": 6.57761148932446e-06, + "loss": 0.0131, + "step": 38039 + }, + { + "epoch": 4.98, + "learning_rate": 6.577113193817786e-06, + "loss": 0.0156, + "step": 38040 + }, + { + "epoch": 4.98, + "learning_rate": 6.5766149079377974e-06, + "loss": 0.0178, + "step": 38041 + }, + { + "epoch": 4.98, + "learning_rate": 6.576116631685893e-06, + "loss": 0.019, + "step": 38042 + }, + { + "epoch": 4.98, + "learning_rate": 6.575618365063481e-06, + "loss": 0.0147, + "step": 38043 + }, + { + "epoch": 4.98, + "learning_rate": 6.575120108071962e-06, + "loss": 0.0165, + "step": 38044 + }, + { + "epoch": 4.98, + "learning_rate": 6.574621860712733e-06, + "loss": 0.0211, + "step": 38045 + }, + { + "epoch": 4.98, + "learning_rate": 6.574123622987201e-06, + "loss": 0.0174, + "step": 38046 + }, + { + "epoch": 4.98, + "learning_rate": 6.573625394896761e-06, + "loss": 0.0186, + "step": 38047 + }, + { + "epoch": 4.98, + "learning_rate": 6.573127176442817e-06, + "loss": 0.0187, + "step": 38048 + }, + { + "epoch": 4.98, + "learning_rate": 6.572628967626773e-06, + "loss": 0.0196, + "step": 38049 + }, + { + "epoch": 4.98, + "learning_rate": 6.572130768450024e-06, + "loss": 0.016, + "step": 38050 + }, + { + "epoch": 4.98, + "learning_rate": 6.5716325789139775e-06, + "loss": 0.0182, + "step": 38051 + }, + { + "epoch": 4.98, + "learning_rate": 6.571134399020029e-06, + "loss": 0.017, + "step": 38052 + }, + { + "epoch": 4.98, + "learning_rate": 6.570636228769583e-06, + "loss": 0.0186, + "step": 38053 + }, + { + "epoch": 4.98, + "learning_rate": 6.570138068164041e-06, + "loss": 0.0167, + "step": 38054 + }, + { + "epoch": 4.98, + "learning_rate": 6.569639917204802e-06, + "loss": 0.0166, + "step": 38055 + }, + { + "epoch": 4.99, + "learning_rate": 6.569141775893268e-06, + "loss": 0.0225, + "step": 38056 + }, + { + "epoch": 4.99, + "learning_rate": 6.568643644230843e-06, + "loss": 0.0153, + "step": 38057 + }, + { + "epoch": 4.99, + "learning_rate": 6.568145522218923e-06, + "loss": 0.0155, + "step": 38058 + }, + { + "epoch": 4.99, + "learning_rate": 6.5676474098589125e-06, + "loss": 0.0159, + "step": 38059 + }, + { + "epoch": 4.99, + "learning_rate": 6.567149307152209e-06, + "loss": 0.0178, + "step": 38060 + }, + { + "epoch": 4.99, + "learning_rate": 6.566651214100217e-06, + "loss": 0.0158, + "step": 38061 + }, + { + "epoch": 4.99, + "learning_rate": 6.566153130704333e-06, + "loss": 0.0183, + "step": 38062 + }, + { + "epoch": 4.99, + "learning_rate": 6.565655056965961e-06, + "loss": 0.0139, + "step": 38063 + }, + { + "epoch": 4.99, + "learning_rate": 6.565156992886499e-06, + "loss": 0.0159, + "step": 38064 + }, + { + "epoch": 4.99, + "learning_rate": 6.564658938467355e-06, + "loss": 0.0192, + "step": 38065 + }, + { + "epoch": 4.99, + "learning_rate": 6.564160893709922e-06, + "loss": 0.0166, + "step": 38066 + }, + { + "epoch": 4.99, + "learning_rate": 6.563662858615606e-06, + "loss": 0.0155, + "step": 38067 + }, + { + "epoch": 4.99, + "learning_rate": 6.5631648331858025e-06, + "loss": 0.0204, + "step": 38068 + }, + { + "epoch": 4.99, + "learning_rate": 6.562666817421917e-06, + "loss": 0.0156, + "step": 38069 + }, + { + "epoch": 4.99, + "learning_rate": 6.562168811325347e-06, + "loss": 0.0165, + "step": 38070 + }, + { + "epoch": 4.99, + "learning_rate": 6.561670814897494e-06, + "loss": 0.0134, + "step": 38071 + }, + { + "epoch": 4.99, + "learning_rate": 6.561172828139759e-06, + "loss": 0.0176, + "step": 38072 + }, + { + "epoch": 4.99, + "learning_rate": 6.560674851053541e-06, + "loss": 0.0155, + "step": 38073 + }, + { + "epoch": 4.99, + "learning_rate": 6.5601768836402455e-06, + "loss": 0.0209, + "step": 38074 + }, + { + "epoch": 4.99, + "learning_rate": 6.559678925901264e-06, + "loss": 0.0103, + "step": 38075 + }, + { + "epoch": 4.99, + "learning_rate": 6.559180977838004e-06, + "loss": 0.0167, + "step": 38076 + }, + { + "epoch": 4.99, + "learning_rate": 6.558683039451865e-06, + "loss": 0.0154, + "step": 38077 + }, + { + "epoch": 4.99, + "learning_rate": 6.558185110744249e-06, + "loss": 0.0179, + "step": 38078 + }, + { + "epoch": 4.99, + "learning_rate": 6.557687191716551e-06, + "loss": 0.0204, + "step": 38079 + }, + { + "epoch": 4.99, + "learning_rate": 6.5571892823701775e-06, + "loss": 0.0154, + "step": 38080 + }, + { + "epoch": 4.99, + "learning_rate": 6.556691382706523e-06, + "loss": 0.0172, + "step": 38081 + }, + { + "epoch": 4.99, + "learning_rate": 6.556193492726992e-06, + "loss": 0.0179, + "step": 38082 + }, + { + "epoch": 4.99, + "learning_rate": 6.555695612432983e-06, + "loss": 0.0157, + "step": 38083 + }, + { + "epoch": 4.99, + "learning_rate": 6.555197741825898e-06, + "loss": 0.0234, + "step": 38084 + }, + { + "epoch": 4.99, + "learning_rate": 6.55469988090713e-06, + "loss": 0.0199, + "step": 38085 + }, + { + "epoch": 4.99, + "learning_rate": 6.554202029678091e-06, + "loss": 0.0161, + "step": 38086 + }, + { + "epoch": 4.99, + "learning_rate": 6.553704188140173e-06, + "loss": 0.0143, + "step": 38087 + }, + { + "epoch": 4.99, + "learning_rate": 6.553206356294781e-06, + "loss": 0.0144, + "step": 38088 + }, + { + "epoch": 4.99, + "learning_rate": 6.55270853414331e-06, + "loss": 0.0155, + "step": 38089 + }, + { + "epoch": 4.99, + "learning_rate": 6.552210721687165e-06, + "loss": 0.0164, + "step": 38090 + }, + { + "epoch": 4.99, + "learning_rate": 6.551712918927742e-06, + "loss": 0.0177, + "step": 38091 + }, + { + "epoch": 4.99, + "learning_rate": 6.551215125866442e-06, + "loss": 0.0146, + "step": 38092 + }, + { + "epoch": 4.99, + "learning_rate": 6.550717342504669e-06, + "loss": 0.014, + "step": 38093 + }, + { + "epoch": 4.99, + "learning_rate": 6.550219568843816e-06, + "loss": 0.0139, + "step": 38094 + }, + { + "epoch": 4.99, + "learning_rate": 6.549721804885289e-06, + "loss": 0.0116, + "step": 38095 + }, + { + "epoch": 4.99, + "learning_rate": 6.549224050630484e-06, + "loss": 0.0143, + "step": 38096 + }, + { + "epoch": 4.99, + "learning_rate": 6.548726306080803e-06, + "loss": 0.0184, + "step": 38097 + }, + { + "epoch": 4.99, + "learning_rate": 6.548228571237645e-06, + "loss": 0.018, + "step": 38098 + }, + { + "epoch": 4.99, + "learning_rate": 6.547730846102413e-06, + "loss": 0.0159, + "step": 38099 + }, + { + "epoch": 4.99, + "learning_rate": 6.547233130676503e-06, + "loss": 0.0165, + "step": 38100 + }, + { + "epoch": 4.99, + "learning_rate": 6.546735424961317e-06, + "loss": 0.0154, + "step": 38101 + }, + { + "epoch": 4.99, + "learning_rate": 6.5462377289582515e-06, + "loss": 0.019, + "step": 38102 + }, + { + "epoch": 4.99, + "learning_rate": 6.54574004266871e-06, + "loss": 0.0134, + "step": 38103 + }, + { + "epoch": 4.99, + "learning_rate": 6.545242366094089e-06, + "loss": 0.018, + "step": 38104 + }, + { + "epoch": 4.99, + "learning_rate": 6.5447446992357935e-06, + "loss": 0.0147, + "step": 38105 + }, + { + "epoch": 4.99, + "learning_rate": 6.544247042095215e-06, + "loss": 0.0176, + "step": 38106 + }, + { + "epoch": 4.99, + "learning_rate": 6.543749394673757e-06, + "loss": 0.0186, + "step": 38107 + }, + { + "epoch": 4.99, + "learning_rate": 6.543251756972821e-06, + "loss": 0.0158, + "step": 38108 + }, + { + "epoch": 4.99, + "learning_rate": 6.542754128993809e-06, + "loss": 0.013, + "step": 38109 + }, + { + "epoch": 4.99, + "learning_rate": 6.542256510738114e-06, + "loss": 0.012, + "step": 38110 + }, + { + "epoch": 4.99, + "learning_rate": 6.54175890220714e-06, + "loss": 0.0154, + "step": 38111 + }, + { + "epoch": 4.99, + "learning_rate": 6.541261303402284e-06, + "loss": 0.0134, + "step": 38112 + }, + { + "epoch": 4.99, + "learning_rate": 6.540763714324946e-06, + "loss": 0.0169, + "step": 38113 + }, + { + "epoch": 4.99, + "learning_rate": 6.540266134976527e-06, + "loss": 0.0165, + "step": 38114 + }, + { + "epoch": 4.99, + "learning_rate": 6.539768565358424e-06, + "loss": 0.0138, + "step": 38115 + }, + { + "epoch": 4.99, + "learning_rate": 6.539271005472039e-06, + "loss": 0.0161, + "step": 38116 + }, + { + "epoch": 4.99, + "learning_rate": 6.538773455318769e-06, + "loss": 0.0138, + "step": 38117 + }, + { + "epoch": 4.99, + "learning_rate": 6.538275914900012e-06, + "loss": 0.0183, + "step": 38118 + }, + { + "epoch": 4.99, + "learning_rate": 6.537778384217172e-06, + "loss": 0.0124, + "step": 38119 + }, + { + "epoch": 4.99, + "learning_rate": 6.537280863271647e-06, + "loss": 0.0134, + "step": 38120 + }, + { + "epoch": 4.99, + "learning_rate": 6.536783352064832e-06, + "loss": 0.0146, + "step": 38121 + }, + { + "epoch": 4.99, + "learning_rate": 6.5362858505981346e-06, + "loss": 0.0138, + "step": 38122 + }, + { + "epoch": 4.99, + "learning_rate": 6.535788358872944e-06, + "loss": 0.0169, + "step": 38123 + }, + { + "epoch": 4.99, + "learning_rate": 6.535290876890667e-06, + "loss": 0.0164, + "step": 38124 + }, + { + "epoch": 4.99, + "learning_rate": 6.5347934046527e-06, + "loss": 0.0175, + "step": 38125 + }, + { + "epoch": 4.99, + "learning_rate": 6.53429594216044e-06, + "loss": 0.0186, + "step": 38126 + }, + { + "epoch": 4.99, + "learning_rate": 6.533798489415288e-06, + "loss": 0.0162, + "step": 38127 + }, + { + "epoch": 4.99, + "learning_rate": 6.533301046418646e-06, + "loss": 0.0221, + "step": 38128 + }, + { + "epoch": 4.99, + "learning_rate": 6.532803613171904e-06, + "loss": 0.0158, + "step": 38129 + }, + { + "epoch": 4.99, + "learning_rate": 6.532306189676473e-06, + "loss": 0.0156, + "step": 38130 + }, + { + "epoch": 4.99, + "learning_rate": 6.531808775933744e-06, + "loss": 0.0132, + "step": 38131 + }, + { + "epoch": 5.0, + "learning_rate": 6.5313113719451205e-06, + "loss": 0.0151, + "step": 38132 + }, + { + "epoch": 5.0, + "learning_rate": 6.530813977711996e-06, + "loss": 0.0151, + "step": 38133 + }, + { + "epoch": 5.0, + "learning_rate": 6.5303165932357745e-06, + "loss": 0.0177, + "step": 38134 + }, + { + "epoch": 5.0, + "learning_rate": 6.529819218517852e-06, + "loss": 0.0111, + "step": 38135 + }, + { + "epoch": 5.0, + "learning_rate": 6.529321853559628e-06, + "loss": 0.0173, + "step": 38136 + }, + { + "epoch": 5.0, + "learning_rate": 6.528824498362503e-06, + "loss": 0.0191, + "step": 38137 + }, + { + "epoch": 5.0, + "learning_rate": 6.528327152927872e-06, + "loss": 0.0138, + "step": 38138 + }, + { + "epoch": 5.0, + "learning_rate": 6.5278298172571385e-06, + "loss": 0.0192, + "step": 38139 + }, + { + "epoch": 5.0, + "learning_rate": 6.527332491351694e-06, + "loss": 0.0215, + "step": 38140 + }, + { + "epoch": 5.0, + "learning_rate": 6.526835175212944e-06, + "loss": 0.0137, + "step": 38141 + }, + { + "epoch": 5.0, + "learning_rate": 6.526337868842287e-06, + "loss": 0.0181, + "step": 38142 + }, + { + "epoch": 5.0, + "learning_rate": 6.52584057224112e-06, + "loss": 0.0187, + "step": 38143 + }, + { + "epoch": 5.0, + "learning_rate": 6.525343285410839e-06, + "loss": 0.0166, + "step": 38144 + }, + { + "epoch": 5.0, + "learning_rate": 6.524846008352848e-06, + "loss": 0.0151, + "step": 38145 + }, + { + "epoch": 5.0, + "learning_rate": 6.52434874106854e-06, + "loss": 0.0181, + "step": 38146 + }, + { + "epoch": 5.0, + "learning_rate": 6.523851483559319e-06, + "loss": 0.0191, + "step": 38147 + }, + { + "epoch": 5.0, + "learning_rate": 6.523354235826578e-06, + "loss": 0.0147, + "step": 38148 + }, + { + "epoch": 5.0, + "learning_rate": 6.52285699787172e-06, + "loss": 0.0185, + "step": 38149 + }, + { + "epoch": 5.0, + "learning_rate": 6.52235976969614e-06, + "loss": 0.0129, + "step": 38150 + }, + { + "epoch": 5.0, + "learning_rate": 6.521862551301235e-06, + "loss": 0.017, + "step": 38151 + }, + { + "epoch": 5.0, + "learning_rate": 6.521365342688408e-06, + "loss": 0.0138, + "step": 38152 + }, + { + "epoch": 5.0, + "learning_rate": 6.520868143859059e-06, + "loss": 0.0146, + "step": 38153 + }, + { + "epoch": 5.0, + "learning_rate": 6.520370954814581e-06, + "loss": 0.0139, + "step": 38154 + }, + { + "epoch": 5.0, + "learning_rate": 6.519873775556376e-06, + "loss": 0.0176, + "step": 38155 + }, + { + "epoch": 5.0, + "learning_rate": 6.519376606085839e-06, + "loss": 0.0169, + "step": 38156 + }, + { + "epoch": 5.0, + "learning_rate": 6.518879446404369e-06, + "loss": 0.018, + "step": 38157 + }, + { + "epoch": 5.0, + "learning_rate": 6.518382296513368e-06, + "loss": 0.0171, + "step": 38158 + }, + { + "epoch": 5.0, + "learning_rate": 6.517885156414229e-06, + "loss": 0.019, + "step": 38159 + }, + { + "epoch": 5.0, + "learning_rate": 6.517388026108356e-06, + "loss": 0.015, + "step": 38160 + }, + { + "epoch": 5.0, + "learning_rate": 6.516890905597139e-06, + "loss": 0.014, + "step": 38161 + }, + { + "epoch": 5.0, + "learning_rate": 6.516393794881981e-06, + "loss": 0.0194, + "step": 38162 + }, + { + "epoch": 5.0, + "learning_rate": 6.515896693964281e-06, + "loss": 0.0165, + "step": 38163 + }, + { + "epoch": 5.0, + "learning_rate": 6.515399602845438e-06, + "loss": 0.0165, + "step": 38164 + }, + { + "epoch": 5.0, + "learning_rate": 6.514902521526847e-06, + "loss": 0.0191, + "step": 38165 + }, + { + "epoch": 5.0, + "learning_rate": 6.514405450009908e-06, + "loss": 0.017, + "step": 38166 + }, + { + "epoch": 5.0, + "learning_rate": 6.513908388296015e-06, + "loss": 0.0219, + "step": 38167 + }, + { + "epoch": 5.0, + "learning_rate": 6.513411336386573e-06, + "loss": 0.0147, + "step": 38168 + }, + { + "epoch": 5.0, + "learning_rate": 6.512914294282974e-06, + "loss": 0.0169, + "step": 38169 + }, + { + "epoch": 5.0, + "learning_rate": 6.512417261986619e-06, + "loss": 0.0098, + "step": 38170 + }, + { + "epoch": 5.0, + "learning_rate": 6.511920239498904e-06, + "loss": 0.0095, + "step": 38171 + }, + { + "epoch": 5.0, + "learning_rate": 6.511423226821225e-06, + "loss": 0.0067, + "step": 38172 + }, + { + "epoch": 5.0, + "learning_rate": 6.510926223954982e-06, + "loss": 0.0075, + "step": 38173 + }, + { + "epoch": 5.0, + "learning_rate": 6.510429230901577e-06, + "loss": 0.0094, + "step": 38174 + }, + { + "epoch": 5.0, + "learning_rate": 6.5099322476624025e-06, + "loss": 0.0081, + "step": 38175 + }, + { + "epoch": 5.0, + "learning_rate": 6.509435274238859e-06, + "loss": 0.0061, + "step": 38176 + }, + { + "epoch": 5.0, + "learning_rate": 6.508938310632341e-06, + "loss": 0.0119, + "step": 38177 + }, + { + "epoch": 5.0, + "learning_rate": 6.508441356844249e-06, + "loss": 0.008, + "step": 38178 + }, + { + "epoch": 5.0, + "learning_rate": 6.507944412875981e-06, + "loss": 0.0079, + "step": 38179 + }, + { + "epoch": 5.0, + "learning_rate": 6.507447478728931e-06, + "loss": 0.0088, + "step": 38180 + }, + { + "epoch": 5.0, + "learning_rate": 6.506950554404501e-06, + "loss": 0.0096, + "step": 38181 + }, + { + "epoch": 5.0, + "learning_rate": 6.506453639904084e-06, + "loss": 0.006, + "step": 38182 + }, + { + "epoch": 5.0, + "learning_rate": 6.505956735229085e-06, + "loss": 0.0106, + "step": 38183 + }, + { + "epoch": 5.0, + "learning_rate": 6.50545984038089e-06, + "loss": 0.0071, + "step": 38184 + }, + { + "epoch": 5.0, + "learning_rate": 6.5049629553609054e-06, + "loss": 0.0071, + "step": 38185 + }, + { + "epoch": 5.0, + "learning_rate": 6.504466080170526e-06, + "loss": 0.0057, + "step": 38186 + }, + { + "epoch": 5.0, + "learning_rate": 6.503969214811153e-06, + "loss": 0.0071, + "step": 38187 + }, + { + "epoch": 5.0, + "learning_rate": 6.503472359284178e-06, + "loss": 0.0104, + "step": 38188 + }, + { + "epoch": 5.0, + "learning_rate": 6.502975513591002e-06, + "loss": 0.0092, + "step": 38189 + }, + { + "epoch": 5.0, + "learning_rate": 6.502478677733019e-06, + "loss": 0.0073, + "step": 38190 + }, + { + "epoch": 5.0, + "learning_rate": 6.5019818517116314e-06, + "loss": 0.0064, + "step": 38191 + }, + { + "epoch": 5.0, + "learning_rate": 6.501485035528232e-06, + "loss": 0.0072, + "step": 38192 + }, + { + "epoch": 5.0, + "learning_rate": 6.500988229184218e-06, + "loss": 0.0076, + "step": 38193 + }, + { + "epoch": 5.0, + "learning_rate": 6.500491432680992e-06, + "loss": 0.0067, + "step": 38194 + }, + { + "epoch": 5.0, + "learning_rate": 6.499994646019942e-06, + "loss": 0.0112, + "step": 38195 + }, + { + "epoch": 5.0, + "learning_rate": 6.499497869202474e-06, + "loss": 0.0103, + "step": 38196 + }, + { + "epoch": 5.0, + "learning_rate": 6.499001102229983e-06, + "loss": 0.0093, + "step": 38197 + }, + { + "epoch": 5.0, + "learning_rate": 6.498504345103863e-06, + "loss": 0.0072, + "step": 38198 + }, + { + "epoch": 5.0, + "learning_rate": 6.4980075978255156e-06, + "loss": 0.008, + "step": 38199 + }, + { + "epoch": 5.0, + "learning_rate": 6.497510860396332e-06, + "loss": 0.0094, + "step": 38200 + }, + { + "epoch": 5.0, + "learning_rate": 6.4970141328177136e-06, + "loss": 0.0104, + "step": 38201 + }, + { + "epoch": 5.0, + "learning_rate": 6.496517415091059e-06, + "loss": 0.0108, + "step": 38202 + }, + { + "epoch": 5.0, + "learning_rate": 6.4960207072177594e-06, + "loss": 0.0128, + "step": 38203 + }, + { + "epoch": 5.0, + "learning_rate": 6.495524009199218e-06, + "loss": 0.0087, + "step": 38204 + }, + { + "epoch": 5.0, + "learning_rate": 6.495027321036826e-06, + "loss": 0.0097, + "step": 38205 + }, + { + "epoch": 5.0, + "learning_rate": 6.494530642731981e-06, + "loss": 0.0087, + "step": 38206 + }, + { + "epoch": 5.0, + "learning_rate": 6.4940339742860835e-06, + "loss": 0.0084, + "step": 38207 + }, + { + "epoch": 5.0, + "learning_rate": 6.493537315700531e-06, + "loss": 0.0087, + "step": 38208 + }, + { + "epoch": 5.01, + "learning_rate": 6.493040666976717e-06, + "loss": 0.009, + "step": 38209 + }, + { + "epoch": 5.01, + "learning_rate": 6.492544028116041e-06, + "loss": 0.0125, + "step": 38210 + }, + { + "epoch": 5.01, + "learning_rate": 6.492047399119896e-06, + "loss": 0.01, + "step": 38211 + }, + { + "epoch": 5.01, + "learning_rate": 6.491550779989683e-06, + "loss": 0.0079, + "step": 38212 + }, + { + "epoch": 5.01, + "learning_rate": 6.491054170726794e-06, + "loss": 0.0073, + "step": 38213 + }, + { + "epoch": 5.01, + "learning_rate": 6.490557571332629e-06, + "loss": 0.0076, + "step": 38214 + }, + { + "epoch": 5.01, + "learning_rate": 6.490060981808586e-06, + "loss": 0.0079, + "step": 38215 + }, + { + "epoch": 5.01, + "learning_rate": 6.489564402156055e-06, + "loss": 0.0076, + "step": 38216 + }, + { + "epoch": 5.01, + "learning_rate": 6.489067832376439e-06, + "loss": 0.0073, + "step": 38217 + }, + { + "epoch": 5.01, + "learning_rate": 6.488571272471135e-06, + "loss": 0.0075, + "step": 38218 + }, + { + "epoch": 5.01, + "learning_rate": 6.488074722441536e-06, + "loss": 0.0063, + "step": 38219 + }, + { + "epoch": 5.01, + "learning_rate": 6.487578182289041e-06, + "loss": 0.0077, + "step": 38220 + }, + { + "epoch": 5.01, + "learning_rate": 6.4870816520150425e-06, + "loss": 0.0081, + "step": 38221 + }, + { + "epoch": 5.01, + "learning_rate": 6.4865851316209415e-06, + "loss": 0.0075, + "step": 38222 + }, + { + "epoch": 5.01, + "learning_rate": 6.486088621108133e-06, + "loss": 0.0083, + "step": 38223 + }, + { + "epoch": 5.01, + "learning_rate": 6.4855921204780125e-06, + "loss": 0.0062, + "step": 38224 + }, + { + "epoch": 5.01, + "learning_rate": 6.485095629731979e-06, + "loss": 0.0077, + "step": 38225 + }, + { + "epoch": 5.01, + "learning_rate": 6.4845991488714245e-06, + "loss": 0.0088, + "step": 38226 + }, + { + "epoch": 5.01, + "learning_rate": 6.484102677897745e-06, + "loss": 0.0088, + "step": 38227 + }, + { + "epoch": 5.01, + "learning_rate": 6.4836062168123425e-06, + "loss": 0.0067, + "step": 38228 + }, + { + "epoch": 5.01, + "learning_rate": 6.483109765616612e-06, + "loss": 0.0086, + "step": 38229 + }, + { + "epoch": 5.01, + "learning_rate": 6.482613324311946e-06, + "loss": 0.0098, + "step": 38230 + }, + { + "epoch": 5.01, + "learning_rate": 6.482116892899745e-06, + "loss": 0.0077, + "step": 38231 + }, + { + "epoch": 5.01, + "learning_rate": 6.4816204713814005e-06, + "loss": 0.0081, + "step": 38232 + }, + { + "epoch": 5.01, + "learning_rate": 6.481124059758313e-06, + "loss": 0.0085, + "step": 38233 + }, + { + "epoch": 5.01, + "learning_rate": 6.480627658031875e-06, + "loss": 0.0082, + "step": 38234 + }, + { + "epoch": 5.01, + "learning_rate": 6.480131266203487e-06, + "loss": 0.0073, + "step": 38235 + }, + { + "epoch": 5.01, + "learning_rate": 6.479634884274538e-06, + "loss": 0.0102, + "step": 38236 + }, + { + "epoch": 5.01, + "learning_rate": 6.47913851224643e-06, + "loss": 0.0072, + "step": 38237 + }, + { + "epoch": 5.01, + "learning_rate": 6.4786421501205556e-06, + "loss": 0.0092, + "step": 38238 + }, + { + "epoch": 5.01, + "learning_rate": 6.478145797898316e-06, + "loss": 0.0087, + "step": 38239 + }, + { + "epoch": 5.01, + "learning_rate": 6.477649455581102e-06, + "loss": 0.0078, + "step": 38240 + }, + { + "epoch": 5.01, + "learning_rate": 6.4771531231703145e-06, + "loss": 0.0087, + "step": 38241 + }, + { + "epoch": 5.01, + "learning_rate": 6.476656800667343e-06, + "loss": 0.0082, + "step": 38242 + }, + { + "epoch": 5.01, + "learning_rate": 6.476160488073587e-06, + "loss": 0.0093, + "step": 38243 + }, + { + "epoch": 5.01, + "learning_rate": 6.475664185390442e-06, + "loss": 0.0064, + "step": 38244 + }, + { + "epoch": 5.01, + "learning_rate": 6.475167892619305e-06, + "loss": 0.0089, + "step": 38245 + }, + { + "epoch": 5.01, + "learning_rate": 6.474671609761571e-06, + "loss": 0.0067, + "step": 38246 + }, + { + "epoch": 5.01, + "learning_rate": 6.474175336818634e-06, + "loss": 0.0095, + "step": 38247 + }, + { + "epoch": 5.01, + "learning_rate": 6.473679073791892e-06, + "loss": 0.0088, + "step": 38248 + }, + { + "epoch": 5.01, + "learning_rate": 6.4731828206827366e-06, + "loss": 0.0077, + "step": 38249 + }, + { + "epoch": 5.01, + "learning_rate": 6.472686577492568e-06, + "loss": 0.0075, + "step": 38250 + }, + { + "epoch": 5.01, + "learning_rate": 6.472190344222782e-06, + "loss": 0.0076, + "step": 38251 + }, + { + "epoch": 5.01, + "learning_rate": 6.471694120874774e-06, + "loss": 0.0068, + "step": 38252 + }, + { + "epoch": 5.01, + "learning_rate": 6.4711979074499355e-06, + "loss": 0.0074, + "step": 38253 + }, + { + "epoch": 5.01, + "learning_rate": 6.470701703949669e-06, + "loss": 0.0072, + "step": 38254 + }, + { + "epoch": 5.01, + "learning_rate": 6.470205510375363e-06, + "loss": 0.0064, + "step": 38255 + }, + { + "epoch": 5.01, + "learning_rate": 6.469709326728417e-06, + "loss": 0.0099, + "step": 38256 + }, + { + "epoch": 5.01, + "learning_rate": 6.469213153010224e-06, + "loss": 0.0106, + "step": 38257 + }, + { + "epoch": 5.01, + "learning_rate": 6.468716989222181e-06, + "loss": 0.0095, + "step": 38258 + }, + { + "epoch": 5.01, + "learning_rate": 6.468220835365686e-06, + "loss": 0.0076, + "step": 38259 + }, + { + "epoch": 5.01, + "learning_rate": 6.467724691442127e-06, + "loss": 0.0066, + "step": 38260 + }, + { + "epoch": 5.01, + "learning_rate": 6.467228557452907e-06, + "loss": 0.0095, + "step": 38261 + }, + { + "epoch": 5.01, + "learning_rate": 6.466732433399421e-06, + "loss": 0.009, + "step": 38262 + }, + { + "epoch": 5.01, + "learning_rate": 6.466236319283059e-06, + "loss": 0.0079, + "step": 38263 + }, + { + "epoch": 5.01, + "learning_rate": 6.46574021510522e-06, + "loss": 0.0074, + "step": 38264 + }, + { + "epoch": 5.01, + "learning_rate": 6.4652441208673e-06, + "loss": 0.0076, + "step": 38265 + }, + { + "epoch": 5.01, + "learning_rate": 6.46474803657069e-06, + "loss": 0.0075, + "step": 38266 + }, + { + "epoch": 5.01, + "learning_rate": 6.46425196221679e-06, + "loss": 0.011, + "step": 38267 + }, + { + "epoch": 5.01, + "learning_rate": 6.4637558978069914e-06, + "loss": 0.0089, + "step": 38268 + }, + { + "epoch": 5.01, + "learning_rate": 6.463259843342693e-06, + "loss": 0.0092, + "step": 38269 + }, + { + "epoch": 5.01, + "learning_rate": 6.462763798825287e-06, + "loss": 0.0092, + "step": 38270 + }, + { + "epoch": 5.01, + "learning_rate": 6.462267764256165e-06, + "loss": 0.0075, + "step": 38271 + }, + { + "epoch": 5.01, + "learning_rate": 6.4617717396367305e-06, + "loss": 0.0099, + "step": 38272 + }, + { + "epoch": 5.01, + "learning_rate": 6.461275724968376e-06, + "loss": 0.0088, + "step": 38273 + }, + { + "epoch": 5.01, + "learning_rate": 6.460779720252492e-06, + "loss": 0.0076, + "step": 38274 + }, + { + "epoch": 5.01, + "learning_rate": 6.460283725490481e-06, + "loss": 0.0057, + "step": 38275 + }, + { + "epoch": 5.01, + "learning_rate": 6.45978774068373e-06, + "loss": 0.0078, + "step": 38276 + }, + { + "epoch": 5.01, + "learning_rate": 6.4592917658336395e-06, + "loss": 0.0112, + "step": 38277 + }, + { + "epoch": 5.01, + "learning_rate": 6.458795800941601e-06, + "loss": 0.0101, + "step": 38278 + }, + { + "epoch": 5.01, + "learning_rate": 6.45829984600901e-06, + "loss": 0.0074, + "step": 38279 + }, + { + "epoch": 5.01, + "learning_rate": 6.457803901037264e-06, + "loss": 0.0078, + "step": 38280 + }, + { + "epoch": 5.01, + "learning_rate": 6.457307966027755e-06, + "loss": 0.009, + "step": 38281 + }, + { + "epoch": 5.01, + "learning_rate": 6.456812040981875e-06, + "loss": 0.0073, + "step": 38282 + }, + { + "epoch": 5.01, + "learning_rate": 6.456316125901027e-06, + "loss": 0.0075, + "step": 38283 + }, + { + "epoch": 5.01, + "learning_rate": 6.455820220786599e-06, + "loss": 0.0076, + "step": 38284 + }, + { + "epoch": 5.02, + "learning_rate": 6.455324325639991e-06, + "loss": 0.0073, + "step": 38285 + }, + { + "epoch": 5.02, + "learning_rate": 6.454828440462593e-06, + "loss": 0.0087, + "step": 38286 + }, + { + "epoch": 5.02, + "learning_rate": 6.4543325652557985e-06, + "loss": 0.007, + "step": 38287 + }, + { + "epoch": 5.02, + "learning_rate": 6.45383670002101e-06, + "loss": 0.008, + "step": 38288 + }, + { + "epoch": 5.02, + "learning_rate": 6.453340844759612e-06, + "loss": 0.0067, + "step": 38289 + }, + { + "epoch": 5.02, + "learning_rate": 6.452844999473009e-06, + "loss": 0.0083, + "step": 38290 + }, + { + "epoch": 5.02, + "learning_rate": 6.4523491641625855e-06, + "loss": 0.0083, + "step": 38291 + }, + { + "epoch": 5.02, + "learning_rate": 6.451853338829746e-06, + "loss": 0.0092, + "step": 38292 + }, + { + "epoch": 5.02, + "learning_rate": 6.4513575234758715e-06, + "loss": 0.0083, + "step": 38293 + }, + { + "epoch": 5.02, + "learning_rate": 6.450861718102372e-06, + "loss": 0.0061, + "step": 38294 + }, + { + "epoch": 5.02, + "learning_rate": 6.450365922710633e-06, + "loss": 0.0077, + "step": 38295 + }, + { + "epoch": 5.02, + "learning_rate": 6.449870137302053e-06, + "loss": 0.0114, + "step": 38296 + }, + { + "epoch": 5.02, + "learning_rate": 6.449374361878023e-06, + "loss": 0.0054, + "step": 38297 + }, + { + "epoch": 5.02, + "learning_rate": 6.448878596439938e-06, + "loss": 0.0059, + "step": 38298 + }, + { + "epoch": 5.02, + "learning_rate": 6.448382840989193e-06, + "loss": 0.0086, + "step": 38299 + }, + { + "epoch": 5.02, + "learning_rate": 6.447887095527183e-06, + "loss": 0.008, + "step": 38300 + }, + { + "epoch": 5.02, + "learning_rate": 6.4473913600553005e-06, + "loss": 0.0093, + "step": 38301 + }, + { + "epoch": 5.02, + "learning_rate": 6.44689563457494e-06, + "loss": 0.0061, + "step": 38302 + }, + { + "epoch": 5.02, + "learning_rate": 6.4463999190874985e-06, + "loss": 0.0075, + "step": 38303 + }, + { + "epoch": 5.02, + "learning_rate": 6.4459042135943625e-06, + "loss": 0.0072, + "step": 38304 + }, + { + "epoch": 5.02, + "learning_rate": 6.445408518096936e-06, + "loss": 0.006, + "step": 38305 + }, + { + "epoch": 5.02, + "learning_rate": 6.44491283259661e-06, + "loss": 0.0083, + "step": 38306 + }, + { + "epoch": 5.02, + "learning_rate": 6.4444171570947745e-06, + "loss": 0.0094, + "step": 38307 + }, + { + "epoch": 5.02, + "learning_rate": 6.443921491592827e-06, + "loss": 0.0089, + "step": 38308 + }, + { + "epoch": 5.02, + "learning_rate": 6.443425836092163e-06, + "loss": 0.0087, + "step": 38309 + }, + { + "epoch": 5.02, + "learning_rate": 6.442930190594172e-06, + "loss": 0.0077, + "step": 38310 + }, + { + "epoch": 5.02, + "learning_rate": 6.442434555100252e-06, + "loss": 0.0113, + "step": 38311 + }, + { + "epoch": 5.02, + "learning_rate": 6.441938929611795e-06, + "loss": 0.0098, + "step": 38312 + }, + { + "epoch": 5.02, + "learning_rate": 6.441443314130196e-06, + "loss": 0.0075, + "step": 38313 + }, + { + "epoch": 5.02, + "learning_rate": 6.440947708656846e-06, + "loss": 0.0102, + "step": 38314 + }, + { + "epoch": 5.02, + "learning_rate": 6.440452113193141e-06, + "loss": 0.0094, + "step": 38315 + }, + { + "epoch": 5.02, + "learning_rate": 6.439956527740474e-06, + "loss": 0.0067, + "step": 38316 + }, + { + "epoch": 5.02, + "learning_rate": 6.439460952300244e-06, + "loss": 0.0072, + "step": 38317 + }, + { + "epoch": 5.02, + "learning_rate": 6.4389653868738366e-06, + "loss": 0.0087, + "step": 38318 + }, + { + "epoch": 5.02, + "learning_rate": 6.4384698314626526e-06, + "loss": 0.0065, + "step": 38319 + }, + { + "epoch": 5.02, + "learning_rate": 6.43797428606808e-06, + "loss": 0.0108, + "step": 38320 + }, + { + "epoch": 5.02, + "learning_rate": 6.4374787506915185e-06, + "loss": 0.0079, + "step": 38321 + }, + { + "epoch": 5.02, + "learning_rate": 6.436983225334354e-06, + "loss": 0.0071, + "step": 38322 + }, + { + "epoch": 5.02, + "learning_rate": 6.436487709997986e-06, + "loss": 0.0071, + "step": 38323 + }, + { + "epoch": 5.02, + "learning_rate": 6.435992204683809e-06, + "loss": 0.0068, + "step": 38324 + }, + { + "epoch": 5.02, + "learning_rate": 6.435496709393211e-06, + "loss": 0.0095, + "step": 38325 + }, + { + "epoch": 5.02, + "learning_rate": 6.4350012241275865e-06, + "loss": 0.008, + "step": 38326 + }, + { + "epoch": 5.02, + "learning_rate": 6.434505748888335e-06, + "loss": 0.0085, + "step": 38327 + }, + { + "epoch": 5.02, + "learning_rate": 6.434010283676846e-06, + "loss": 0.0075, + "step": 38328 + }, + { + "epoch": 5.02, + "learning_rate": 6.433514828494512e-06, + "loss": 0.0062, + "step": 38329 + }, + { + "epoch": 5.02, + "learning_rate": 6.433019383342729e-06, + "loss": 0.0098, + "step": 38330 + }, + { + "epoch": 5.02, + "learning_rate": 6.432523948222888e-06, + "loss": 0.0072, + "step": 38331 + }, + { + "epoch": 5.02, + "learning_rate": 6.432028523136384e-06, + "loss": 0.0071, + "step": 38332 + }, + { + "epoch": 5.02, + "learning_rate": 6.43153310808461e-06, + "loss": 0.0062, + "step": 38333 + }, + { + "epoch": 5.02, + "learning_rate": 6.431037703068961e-06, + "loss": 0.0075, + "step": 38334 + }, + { + "epoch": 5.02, + "learning_rate": 6.430542308090823e-06, + "loss": 0.0061, + "step": 38335 + }, + { + "epoch": 5.02, + "learning_rate": 6.430046923151601e-06, + "loss": 0.0087, + "step": 38336 + }, + { + "epoch": 5.02, + "learning_rate": 6.429551548252673e-06, + "loss": 0.0075, + "step": 38337 + }, + { + "epoch": 5.02, + "learning_rate": 6.429056183395448e-06, + "loss": 0.0118, + "step": 38338 + }, + { + "epoch": 5.02, + "learning_rate": 6.428560828581311e-06, + "loss": 0.0071, + "step": 38339 + }, + { + "epoch": 5.02, + "learning_rate": 6.428065483811659e-06, + "loss": 0.007, + "step": 38340 + }, + { + "epoch": 5.02, + "learning_rate": 6.427570149087878e-06, + "loss": 0.0088, + "step": 38341 + }, + { + "epoch": 5.02, + "learning_rate": 6.42707482441137e-06, + "loss": 0.0085, + "step": 38342 + }, + { + "epoch": 5.02, + "learning_rate": 6.42657950978352e-06, + "loss": 0.0077, + "step": 38343 + }, + { + "epoch": 5.02, + "learning_rate": 6.426084205205727e-06, + "loss": 0.008, + "step": 38344 + }, + { + "epoch": 5.02, + "learning_rate": 6.425588910679381e-06, + "loss": 0.0101, + "step": 38345 + }, + { + "epoch": 5.02, + "learning_rate": 6.425093626205876e-06, + "loss": 0.0079, + "step": 38346 + }, + { + "epoch": 5.02, + "learning_rate": 6.424598351786603e-06, + "loss": 0.0064, + "step": 38347 + }, + { + "epoch": 5.02, + "learning_rate": 6.424103087422959e-06, + "loss": 0.0077, + "step": 38348 + }, + { + "epoch": 5.02, + "learning_rate": 6.423607833116335e-06, + "loss": 0.0069, + "step": 38349 + }, + { + "epoch": 5.02, + "learning_rate": 6.423112588868123e-06, + "loss": 0.006, + "step": 38350 + }, + { + "epoch": 5.02, + "learning_rate": 6.422617354679716e-06, + "loss": 0.008, + "step": 38351 + }, + { + "epoch": 5.02, + "learning_rate": 6.422122130552507e-06, + "loss": 0.0079, + "step": 38352 + }, + { + "epoch": 5.02, + "learning_rate": 6.42162691648789e-06, + "loss": 0.0071, + "step": 38353 + }, + { + "epoch": 5.02, + "learning_rate": 6.421131712487257e-06, + "loss": 0.0083, + "step": 38354 + }, + { + "epoch": 5.02, + "learning_rate": 6.420636518552e-06, + "loss": 0.0094, + "step": 38355 + }, + { + "epoch": 5.02, + "learning_rate": 6.4201413346835115e-06, + "loss": 0.0085, + "step": 38356 + }, + { + "epoch": 5.02, + "learning_rate": 6.419646160883186e-06, + "loss": 0.0079, + "step": 38357 + }, + { + "epoch": 5.02, + "learning_rate": 6.41915099715241e-06, + "loss": 0.0078, + "step": 38358 + }, + { + "epoch": 5.02, + "learning_rate": 6.418655843492587e-06, + "loss": 0.0073, + "step": 38359 + }, + { + "epoch": 5.02, + "learning_rate": 6.4181606999051e-06, + "loss": 0.0096, + "step": 38360 + }, + { + "epoch": 5.03, + "learning_rate": 6.417665566391349e-06, + "loss": 0.0105, + "step": 38361 + }, + { + "epoch": 5.03, + "learning_rate": 6.41717044295272e-06, + "loss": 0.0078, + "step": 38362 + }, + { + "epoch": 5.03, + "learning_rate": 6.416675329590611e-06, + "loss": 0.0095, + "step": 38363 + }, + { + "epoch": 5.03, + "learning_rate": 6.41618022630641e-06, + "loss": 0.0097, + "step": 38364 + }, + { + "epoch": 5.03, + "learning_rate": 6.415685133101511e-06, + "loss": 0.0084, + "step": 38365 + }, + { + "epoch": 5.03, + "learning_rate": 6.415190049977309e-06, + "loss": 0.0086, + "step": 38366 + }, + { + "epoch": 5.03, + "learning_rate": 6.4146949769351915e-06, + "loss": 0.0081, + "step": 38367 + }, + { + "epoch": 5.03, + "learning_rate": 6.414199913976556e-06, + "loss": 0.0086, + "step": 38368 + }, + { + "epoch": 5.03, + "learning_rate": 6.413704861102787e-06, + "loss": 0.0084, + "step": 38369 + }, + { + "epoch": 5.03, + "learning_rate": 6.413209818315285e-06, + "loss": 0.0092, + "step": 38370 + }, + { + "epoch": 5.03, + "learning_rate": 6.412714785615442e-06, + "loss": 0.0087, + "step": 38371 + }, + { + "epoch": 5.03, + "learning_rate": 6.412219763004644e-06, + "loss": 0.006, + "step": 38372 + }, + { + "epoch": 5.03, + "learning_rate": 6.411724750484289e-06, + "loss": 0.0078, + "step": 38373 + }, + { + "epoch": 5.03, + "learning_rate": 6.411229748055767e-06, + "loss": 0.0075, + "step": 38374 + }, + { + "epoch": 5.03, + "learning_rate": 6.410734755720469e-06, + "loss": 0.0103, + "step": 38375 + }, + { + "epoch": 5.03, + "learning_rate": 6.410239773479792e-06, + "loss": 0.009, + "step": 38376 + }, + { + "epoch": 5.03, + "learning_rate": 6.40974480133512e-06, + "loss": 0.0089, + "step": 38377 + }, + { + "epoch": 5.03, + "learning_rate": 6.409249839287852e-06, + "loss": 0.0074, + "step": 38378 + }, + { + "epoch": 5.03, + "learning_rate": 6.408754887339377e-06, + "loss": 0.0108, + "step": 38379 + }, + { + "epoch": 5.03, + "learning_rate": 6.408259945491084e-06, + "loss": 0.007, + "step": 38380 + }, + { + "epoch": 5.03, + "learning_rate": 6.407765013744372e-06, + "loss": 0.0095, + "step": 38381 + }, + { + "epoch": 5.03, + "learning_rate": 6.40727009210063e-06, + "loss": 0.0063, + "step": 38382 + }, + { + "epoch": 5.03, + "learning_rate": 6.406775180561249e-06, + "loss": 0.0077, + "step": 38383 + }, + { + "epoch": 5.03, + "learning_rate": 6.406280279127623e-06, + "loss": 0.0102, + "step": 38384 + }, + { + "epoch": 5.03, + "learning_rate": 6.40578538780114e-06, + "loss": 0.0095, + "step": 38385 + }, + { + "epoch": 5.03, + "learning_rate": 6.405290506583197e-06, + "loss": 0.0069, + "step": 38386 + }, + { + "epoch": 5.03, + "learning_rate": 6.404795635475181e-06, + "loss": 0.007, + "step": 38387 + }, + { + "epoch": 5.03, + "learning_rate": 6.404300774478485e-06, + "loss": 0.0095, + "step": 38388 + }, + { + "epoch": 5.03, + "learning_rate": 6.403805923594504e-06, + "loss": 0.0085, + "step": 38389 + }, + { + "epoch": 5.03, + "learning_rate": 6.403311082824625e-06, + "loss": 0.0081, + "step": 38390 + }, + { + "epoch": 5.03, + "learning_rate": 6.402816252170241e-06, + "loss": 0.0075, + "step": 38391 + }, + { + "epoch": 5.03, + "learning_rate": 6.402321431632747e-06, + "loss": 0.0081, + "step": 38392 + }, + { + "epoch": 5.03, + "learning_rate": 6.401826621213531e-06, + "loss": 0.0072, + "step": 38393 + }, + { + "epoch": 5.03, + "learning_rate": 6.401331820913987e-06, + "loss": 0.0095, + "step": 38394 + }, + { + "epoch": 5.03, + "learning_rate": 6.400837030735507e-06, + "loss": 0.007, + "step": 38395 + }, + { + "epoch": 5.03, + "learning_rate": 6.400342250679479e-06, + "loss": 0.0074, + "step": 38396 + }, + { + "epoch": 5.03, + "learning_rate": 6.399847480747299e-06, + "loss": 0.0066, + "step": 38397 + }, + { + "epoch": 5.03, + "learning_rate": 6.399352720940355e-06, + "loss": 0.0096, + "step": 38398 + }, + { + "epoch": 5.03, + "learning_rate": 6.39885797126004e-06, + "loss": 0.0083, + "step": 38399 + }, + { + "epoch": 5.03, + "learning_rate": 6.398363231707745e-06, + "loss": 0.0087, + "step": 38400 + }, + { + "epoch": 5.03, + "learning_rate": 6.397868502284863e-06, + "loss": 0.0079, + "step": 38401 + }, + { + "epoch": 5.03, + "learning_rate": 6.397373782992778e-06, + "loss": 0.0068, + "step": 38402 + }, + { + "epoch": 5.03, + "learning_rate": 6.396879073832891e-06, + "loss": 0.0096, + "step": 38403 + }, + { + "epoch": 5.03, + "learning_rate": 6.39638437480659e-06, + "loss": 0.0094, + "step": 38404 + }, + { + "epoch": 5.03, + "learning_rate": 6.395889685915269e-06, + "loss": 0.0086, + "step": 38405 + }, + { + "epoch": 5.03, + "learning_rate": 6.395395007160312e-06, + "loss": 0.0067, + "step": 38406 + }, + { + "epoch": 5.03, + "learning_rate": 6.394900338543119e-06, + "loss": 0.008, + "step": 38407 + }, + { + "epoch": 5.03, + "learning_rate": 6.394405680065074e-06, + "loss": 0.0086, + "step": 38408 + }, + { + "epoch": 5.03, + "learning_rate": 6.39391103172757e-06, + "loss": 0.0088, + "step": 38409 + }, + { + "epoch": 5.03, + "learning_rate": 6.393416393532001e-06, + "loss": 0.0094, + "step": 38410 + }, + { + "epoch": 5.03, + "learning_rate": 6.392921765479756e-06, + "loss": 0.0084, + "step": 38411 + }, + { + "epoch": 5.03, + "learning_rate": 6.392427147572228e-06, + "loss": 0.0081, + "step": 38412 + }, + { + "epoch": 5.03, + "learning_rate": 6.391932539810801e-06, + "loss": 0.0087, + "step": 38413 + }, + { + "epoch": 5.03, + "learning_rate": 6.391437942196875e-06, + "loss": 0.0089, + "step": 38414 + }, + { + "epoch": 5.03, + "learning_rate": 6.390943354731838e-06, + "loss": 0.0097, + "step": 38415 + }, + { + "epoch": 5.03, + "learning_rate": 6.390448777417082e-06, + "loss": 0.0075, + "step": 38416 + }, + { + "epoch": 5.03, + "learning_rate": 6.389954210253995e-06, + "loss": 0.0053, + "step": 38417 + }, + { + "epoch": 5.03, + "learning_rate": 6.3894596532439705e-06, + "loss": 0.0084, + "step": 38418 + }, + { + "epoch": 5.03, + "learning_rate": 6.388965106388396e-06, + "loss": 0.0093, + "step": 38419 + }, + { + "epoch": 5.03, + "learning_rate": 6.388470569688669e-06, + "loss": 0.0085, + "step": 38420 + }, + { + "epoch": 5.03, + "learning_rate": 6.3879760431461734e-06, + "loss": 0.0073, + "step": 38421 + }, + { + "epoch": 5.03, + "learning_rate": 6.387481526762305e-06, + "loss": 0.0061, + "step": 38422 + }, + { + "epoch": 5.03, + "learning_rate": 6.386987020538449e-06, + "loss": 0.0095, + "step": 38423 + }, + { + "epoch": 5.03, + "learning_rate": 6.386492524475998e-06, + "loss": 0.0063, + "step": 38424 + }, + { + "epoch": 5.03, + "learning_rate": 6.385998038576347e-06, + "loss": 0.0065, + "step": 38425 + }, + { + "epoch": 5.03, + "learning_rate": 6.385503562840887e-06, + "loss": 0.0056, + "step": 38426 + }, + { + "epoch": 5.03, + "learning_rate": 6.385009097271002e-06, + "loss": 0.0087, + "step": 38427 + }, + { + "epoch": 5.03, + "learning_rate": 6.384514641868089e-06, + "loss": 0.0068, + "step": 38428 + }, + { + "epoch": 5.03, + "learning_rate": 6.384020196633533e-06, + "loss": 0.0106, + "step": 38429 + }, + { + "epoch": 5.03, + "learning_rate": 6.38352576156873e-06, + "loss": 0.0073, + "step": 38430 + }, + { + "epoch": 5.03, + "learning_rate": 6.38303133667507e-06, + "loss": 0.0081, + "step": 38431 + }, + { + "epoch": 5.03, + "learning_rate": 6.382536921953937e-06, + "loss": 0.009, + "step": 38432 + }, + { + "epoch": 5.03, + "learning_rate": 6.38204251740673e-06, + "loss": 0.0075, + "step": 38433 + }, + { + "epoch": 5.03, + "learning_rate": 6.381548123034834e-06, + "loss": 0.0088, + "step": 38434 + }, + { + "epoch": 5.03, + "learning_rate": 6.3810537388396395e-06, + "loss": 0.0082, + "step": 38435 + }, + { + "epoch": 5.03, + "learning_rate": 6.380559364822542e-06, + "loss": 0.007, + "step": 38436 + }, + { + "epoch": 5.03, + "learning_rate": 6.380065000984927e-06, + "loss": 0.0071, + "step": 38437 + }, + { + "epoch": 5.04, + "learning_rate": 6.379570647328186e-06, + "loss": 0.009, + "step": 38438 + }, + { + "epoch": 5.04, + "learning_rate": 6.379076303853711e-06, + "loss": 0.0073, + "step": 38439 + }, + { + "epoch": 5.04, + "learning_rate": 6.37858197056289e-06, + "loss": 0.0085, + "step": 38440 + }, + { + "epoch": 5.04, + "learning_rate": 6.378087647457116e-06, + "loss": 0.0054, + "step": 38441 + }, + { + "epoch": 5.04, + "learning_rate": 6.3775933345377765e-06, + "loss": 0.0074, + "step": 38442 + }, + { + "epoch": 5.04, + "learning_rate": 6.377099031806264e-06, + "loss": 0.0078, + "step": 38443 + }, + { + "epoch": 5.04, + "learning_rate": 6.3766047392639655e-06, + "loss": 0.0091, + "step": 38444 + }, + { + "epoch": 5.04, + "learning_rate": 6.376110456912274e-06, + "loss": 0.0077, + "step": 38445 + }, + { + "epoch": 5.04, + "learning_rate": 6.375616184752576e-06, + "loss": 0.0077, + "step": 38446 + }, + { + "epoch": 5.04, + "learning_rate": 6.375121922786268e-06, + "loss": 0.0074, + "step": 38447 + }, + { + "epoch": 5.04, + "learning_rate": 6.374627671014736e-06, + "loss": 0.0096, + "step": 38448 + }, + { + "epoch": 5.04, + "learning_rate": 6.374133429439371e-06, + "loss": 0.0088, + "step": 38449 + }, + { + "epoch": 5.04, + "learning_rate": 6.373639198061562e-06, + "loss": 0.0093, + "step": 38450 + }, + { + "epoch": 5.04, + "learning_rate": 6.373144976882701e-06, + "loss": 0.0072, + "step": 38451 + }, + { + "epoch": 5.04, + "learning_rate": 6.3726507659041755e-06, + "loss": 0.0067, + "step": 38452 + }, + { + "epoch": 5.04, + "learning_rate": 6.372156565127376e-06, + "loss": 0.0073, + "step": 38453 + }, + { + "epoch": 5.04, + "learning_rate": 6.371662374553694e-06, + "loss": 0.0078, + "step": 38454 + }, + { + "epoch": 5.04, + "learning_rate": 6.371168194184518e-06, + "loss": 0.0074, + "step": 38455 + }, + { + "epoch": 5.04, + "learning_rate": 6.37067402402124e-06, + "loss": 0.0068, + "step": 38456 + }, + { + "epoch": 5.04, + "learning_rate": 6.370179864065242e-06, + "loss": 0.0079, + "step": 38457 + }, + { + "epoch": 5.04, + "learning_rate": 6.3696857143179235e-06, + "loss": 0.0109, + "step": 38458 + }, + { + "epoch": 5.04, + "learning_rate": 6.369191574780671e-06, + "loss": 0.0089, + "step": 38459 + }, + { + "epoch": 5.04, + "learning_rate": 6.368697445454874e-06, + "loss": 0.0071, + "step": 38460 + }, + { + "epoch": 5.04, + "learning_rate": 6.368203326341922e-06, + "loss": 0.0099, + "step": 38461 + }, + { + "epoch": 5.04, + "learning_rate": 6.367709217443206e-06, + "loss": 0.01, + "step": 38462 + }, + { + "epoch": 5.04, + "learning_rate": 6.367215118760113e-06, + "loss": 0.0093, + "step": 38463 + }, + { + "epoch": 5.04, + "learning_rate": 6.366721030294034e-06, + "loss": 0.0081, + "step": 38464 + }, + { + "epoch": 5.04, + "learning_rate": 6.3662269520463595e-06, + "loss": 0.0061, + "step": 38465 + }, + { + "epoch": 5.04, + "learning_rate": 6.365732884018478e-06, + "loss": 0.0083, + "step": 38466 + }, + { + "epoch": 5.04, + "learning_rate": 6.365238826211778e-06, + "loss": 0.0072, + "step": 38467 + }, + { + "epoch": 5.04, + "learning_rate": 6.364744778627649e-06, + "loss": 0.0063, + "step": 38468 + }, + { + "epoch": 5.04, + "learning_rate": 6.364250741267481e-06, + "loss": 0.0113, + "step": 38469 + }, + { + "epoch": 5.04, + "learning_rate": 6.363756714132669e-06, + "loss": 0.0064, + "step": 38470 + }, + { + "epoch": 5.04, + "learning_rate": 6.363262697224594e-06, + "loss": 0.0082, + "step": 38471 + }, + { + "epoch": 5.04, + "learning_rate": 6.362768690544653e-06, + "loss": 0.0065, + "step": 38472 + }, + { + "epoch": 5.04, + "learning_rate": 6.362274694094228e-06, + "loss": 0.0118, + "step": 38473 + }, + { + "epoch": 5.04, + "learning_rate": 6.361780707874712e-06, + "loss": 0.0083, + "step": 38474 + }, + { + "epoch": 5.04, + "learning_rate": 6.361286731887495e-06, + "loss": 0.0108, + "step": 38475 + }, + { + "epoch": 5.04, + "learning_rate": 6.360792766133964e-06, + "loss": 0.0077, + "step": 38476 + }, + { + "epoch": 5.04, + "learning_rate": 6.360298810615511e-06, + "loss": 0.0114, + "step": 38477 + }, + { + "epoch": 5.04, + "learning_rate": 6.359804865333521e-06, + "loss": 0.0084, + "step": 38478 + }, + { + "epoch": 5.04, + "learning_rate": 6.359310930289387e-06, + "loss": 0.0075, + "step": 38479 + }, + { + "epoch": 5.04, + "learning_rate": 6.3588170054844975e-06, + "loss": 0.0063, + "step": 38480 + }, + { + "epoch": 5.04, + "learning_rate": 6.358323090920244e-06, + "loss": 0.0065, + "step": 38481 + }, + { + "epoch": 5.04, + "learning_rate": 6.3578291865980105e-06, + "loss": 0.0088, + "step": 38482 + }, + { + "epoch": 5.04, + "learning_rate": 6.357335292519189e-06, + "loss": 0.0078, + "step": 38483 + }, + { + "epoch": 5.04, + "learning_rate": 6.3568414086851685e-06, + "loss": 0.0088, + "step": 38484 + }, + { + "epoch": 5.04, + "learning_rate": 6.3563475350973366e-06, + "loss": 0.0088, + "step": 38485 + }, + { + "epoch": 5.04, + "learning_rate": 6.355853671757084e-06, + "loss": 0.0063, + "step": 38486 + }, + { + "epoch": 5.04, + "learning_rate": 6.355359818665798e-06, + "loss": 0.0061, + "step": 38487 + }, + { + "epoch": 5.04, + "learning_rate": 6.3548659758248685e-06, + "loss": 0.0059, + "step": 38488 + }, + { + "epoch": 5.04, + "learning_rate": 6.354372143235682e-06, + "loss": 0.0102, + "step": 38489 + }, + { + "epoch": 5.04, + "learning_rate": 6.3538783208996316e-06, + "loss": 0.0081, + "step": 38490 + }, + { + "epoch": 5.04, + "learning_rate": 6.353384508818105e-06, + "loss": 0.0062, + "step": 38491 + }, + { + "epoch": 5.04, + "learning_rate": 6.352890706992492e-06, + "loss": 0.01, + "step": 38492 + }, + { + "epoch": 5.04, + "learning_rate": 6.352396915424178e-06, + "loss": 0.0114, + "step": 38493 + }, + { + "epoch": 5.04, + "learning_rate": 6.351903134114552e-06, + "loss": 0.0057, + "step": 38494 + }, + { + "epoch": 5.04, + "learning_rate": 6.351409363065006e-06, + "loss": 0.009, + "step": 38495 + }, + { + "epoch": 5.04, + "learning_rate": 6.350915602276927e-06, + "loss": 0.0068, + "step": 38496 + }, + { + "epoch": 5.04, + "learning_rate": 6.350421851751703e-06, + "loss": 0.008, + "step": 38497 + }, + { + "epoch": 5.04, + "learning_rate": 6.349928111490724e-06, + "loss": 0.006, + "step": 38498 + }, + { + "epoch": 5.04, + "learning_rate": 6.349434381495376e-06, + "loss": 0.0096, + "step": 38499 + }, + { + "epoch": 5.04, + "learning_rate": 6.348940661767047e-06, + "loss": 0.006, + "step": 38500 + }, + { + "epoch": 5.04, + "learning_rate": 6.348446952307133e-06, + "loss": 0.0092, + "step": 38501 + }, + { + "epoch": 5.04, + "learning_rate": 6.347953253117014e-06, + "loss": 0.0077, + "step": 38502 + }, + { + "epoch": 5.04, + "learning_rate": 6.347459564198084e-06, + "loss": 0.0083, + "step": 38503 + }, + { + "epoch": 5.04, + "learning_rate": 6.34696588555173e-06, + "loss": 0.0056, + "step": 38504 + }, + { + "epoch": 5.04, + "learning_rate": 6.346472217179338e-06, + "loss": 0.009, + "step": 38505 + }, + { + "epoch": 5.04, + "learning_rate": 6.345978559082302e-06, + "loss": 0.0075, + "step": 38506 + }, + { + "epoch": 5.04, + "learning_rate": 6.345484911262003e-06, + "loss": 0.0107, + "step": 38507 + }, + { + "epoch": 5.04, + "learning_rate": 6.344991273719836e-06, + "loss": 0.0078, + "step": 38508 + }, + { + "epoch": 5.04, + "learning_rate": 6.344497646457184e-06, + "loss": 0.0081, + "step": 38509 + }, + { + "epoch": 5.04, + "learning_rate": 6.344004029475437e-06, + "loss": 0.007, + "step": 38510 + }, + { + "epoch": 5.04, + "learning_rate": 6.343510422775982e-06, + "loss": 0.0085, + "step": 38511 + }, + { + "epoch": 5.04, + "learning_rate": 6.343016826360214e-06, + "loss": 0.0091, + "step": 38512 + }, + { + "epoch": 5.04, + "learning_rate": 6.342523240229515e-06, + "loss": 0.0077, + "step": 38513 + }, + { + "epoch": 5.05, + "learning_rate": 6.342029664385275e-06, + "loss": 0.0051, + "step": 38514 + }, + { + "epoch": 5.05, + "learning_rate": 6.341536098828881e-06, + "loss": 0.0085, + "step": 38515 + }, + { + "epoch": 5.05, + "learning_rate": 6.341042543561724e-06, + "loss": 0.007, + "step": 38516 + }, + { + "epoch": 5.05, + "learning_rate": 6.340548998585188e-06, + "loss": 0.0071, + "step": 38517 + }, + { + "epoch": 5.05, + "learning_rate": 6.340055463900661e-06, + "loss": 0.0088, + "step": 38518 + }, + { + "epoch": 5.05, + "learning_rate": 6.339561939509537e-06, + "loss": 0.0097, + "step": 38519 + }, + { + "epoch": 5.05, + "learning_rate": 6.339068425413197e-06, + "loss": 0.0086, + "step": 38520 + }, + { + "epoch": 5.05, + "learning_rate": 6.3385749216130364e-06, + "loss": 0.0069, + "step": 38521 + }, + { + "epoch": 5.05, + "learning_rate": 6.338081428110432e-06, + "loss": 0.0118, + "step": 38522 + }, + { + "epoch": 5.05, + "learning_rate": 6.337587944906782e-06, + "loss": 0.0058, + "step": 38523 + }, + { + "epoch": 5.05, + "learning_rate": 6.337094472003471e-06, + "loss": 0.008, + "step": 38524 + }, + { + "epoch": 5.05, + "learning_rate": 6.3366010094018885e-06, + "loss": 0.0081, + "step": 38525 + }, + { + "epoch": 5.05, + "learning_rate": 6.336107557103419e-06, + "loss": 0.0088, + "step": 38526 + }, + { + "epoch": 5.05, + "learning_rate": 6.335614115109454e-06, + "loss": 0.0087, + "step": 38527 + }, + { + "epoch": 5.05, + "learning_rate": 6.335120683421376e-06, + "loss": 0.0086, + "step": 38528 + }, + { + "epoch": 5.05, + "learning_rate": 6.334627262040578e-06, + "loss": 0.0079, + "step": 38529 + }, + { + "epoch": 5.05, + "learning_rate": 6.334133850968445e-06, + "loss": 0.0091, + "step": 38530 + }, + { + "epoch": 5.05, + "learning_rate": 6.333640450206364e-06, + "loss": 0.0103, + "step": 38531 + }, + { + "epoch": 5.05, + "learning_rate": 6.333147059755726e-06, + "loss": 0.0058, + "step": 38532 + }, + { + "epoch": 5.05, + "learning_rate": 6.332653679617914e-06, + "loss": 0.0077, + "step": 38533 + }, + { + "epoch": 5.05, + "learning_rate": 6.33216030979432e-06, + "loss": 0.0074, + "step": 38534 + }, + { + "epoch": 5.05, + "learning_rate": 6.331666950286331e-06, + "loss": 0.0069, + "step": 38535 + }, + { + "epoch": 5.05, + "learning_rate": 6.331173601095331e-06, + "loss": 0.0085, + "step": 38536 + }, + { + "epoch": 5.05, + "learning_rate": 6.330680262222713e-06, + "loss": 0.0065, + "step": 38537 + }, + { + "epoch": 5.05, + "learning_rate": 6.330186933669857e-06, + "loss": 0.0089, + "step": 38538 + }, + { + "epoch": 5.05, + "learning_rate": 6.329693615438157e-06, + "loss": 0.0069, + "step": 38539 + }, + { + "epoch": 5.05, + "learning_rate": 6.329200307529e-06, + "loss": 0.008, + "step": 38540 + }, + { + "epoch": 5.05, + "learning_rate": 6.32870700994377e-06, + "loss": 0.005, + "step": 38541 + }, + { + "epoch": 5.05, + "learning_rate": 6.328213722683858e-06, + "loss": 0.008, + "step": 38542 + }, + { + "epoch": 5.05, + "learning_rate": 6.327720445750647e-06, + "loss": 0.0074, + "step": 38543 + }, + { + "epoch": 5.05, + "learning_rate": 6.327227179145523e-06, + "loss": 0.0085, + "step": 38544 + }, + { + "epoch": 5.05, + "learning_rate": 6.326733922869881e-06, + "loss": 0.0063, + "step": 38545 + }, + { + "epoch": 5.05, + "learning_rate": 6.326240676925106e-06, + "loss": 0.0085, + "step": 38546 + }, + { + "epoch": 5.05, + "learning_rate": 6.3257474413125805e-06, + "loss": 0.0108, + "step": 38547 + }, + { + "epoch": 5.05, + "learning_rate": 6.3252542160336985e-06, + "loss": 0.0072, + "step": 38548 + }, + { + "epoch": 5.05, + "learning_rate": 6.32476100108984e-06, + "loss": 0.0069, + "step": 38549 + }, + { + "epoch": 5.05, + "learning_rate": 6.324267796482397e-06, + "loss": 0.0083, + "step": 38550 + }, + { + "epoch": 5.05, + "learning_rate": 6.323774602212753e-06, + "loss": 0.0066, + "step": 38551 + }, + { + "epoch": 5.05, + "learning_rate": 6.323281418282301e-06, + "loss": 0.0086, + "step": 38552 + }, + { + "epoch": 5.05, + "learning_rate": 6.3227882446924215e-06, + "loss": 0.0074, + "step": 38553 + }, + { + "epoch": 5.05, + "learning_rate": 6.322295081444502e-06, + "loss": 0.0086, + "step": 38554 + }, + { + "epoch": 5.05, + "learning_rate": 6.321801928539932e-06, + "loss": 0.0103, + "step": 38555 + }, + { + "epoch": 5.05, + "learning_rate": 6.321308785980103e-06, + "loss": 0.0082, + "step": 38556 + }, + { + "epoch": 5.05, + "learning_rate": 6.320815653766393e-06, + "loss": 0.007, + "step": 38557 + }, + { + "epoch": 5.05, + "learning_rate": 6.320322531900196e-06, + "loss": 0.0073, + "step": 38558 + }, + { + "epoch": 5.05, + "learning_rate": 6.319829420382895e-06, + "loss": 0.0087, + "step": 38559 + }, + { + "epoch": 5.05, + "learning_rate": 6.3193363192158764e-06, + "loss": 0.0092, + "step": 38560 + }, + { + "epoch": 5.05, + "learning_rate": 6.318843228400531e-06, + "loss": 0.0073, + "step": 38561 + }, + { + "epoch": 5.05, + "learning_rate": 6.318350147938239e-06, + "loss": 0.008, + "step": 38562 + }, + { + "epoch": 5.05, + "learning_rate": 6.317857077830396e-06, + "loss": 0.0066, + "step": 38563 + }, + { + "epoch": 5.05, + "learning_rate": 6.31736401807838e-06, + "loss": 0.0068, + "step": 38564 + }, + { + "epoch": 5.05, + "learning_rate": 6.316870968683584e-06, + "loss": 0.0089, + "step": 38565 + }, + { + "epoch": 5.05, + "learning_rate": 6.316377929647389e-06, + "loss": 0.0084, + "step": 38566 + }, + { + "epoch": 5.05, + "learning_rate": 6.315884900971187e-06, + "loss": 0.0061, + "step": 38567 + }, + { + "epoch": 5.05, + "learning_rate": 6.315391882656361e-06, + "loss": 0.0073, + "step": 38568 + }, + { + "epoch": 5.05, + "learning_rate": 6.314898874704303e-06, + "loss": 0.0082, + "step": 38569 + }, + { + "epoch": 5.05, + "learning_rate": 6.314405877116392e-06, + "loss": 0.0056, + "step": 38570 + }, + { + "epoch": 5.05, + "learning_rate": 6.313912889894021e-06, + "loss": 0.0071, + "step": 38571 + }, + { + "epoch": 5.05, + "learning_rate": 6.313419913038572e-06, + "loss": 0.0096, + "step": 38572 + }, + { + "epoch": 5.05, + "learning_rate": 6.312926946551434e-06, + "loss": 0.0087, + "step": 38573 + }, + { + "epoch": 5.05, + "learning_rate": 6.312433990433992e-06, + "loss": 0.0055, + "step": 38574 + }, + { + "epoch": 5.05, + "learning_rate": 6.311941044687633e-06, + "loss": 0.0065, + "step": 38575 + }, + { + "epoch": 5.05, + "learning_rate": 6.3114481093137446e-06, + "loss": 0.0095, + "step": 38576 + }, + { + "epoch": 5.05, + "learning_rate": 6.310955184313707e-06, + "loss": 0.0072, + "step": 38577 + }, + { + "epoch": 5.05, + "learning_rate": 6.310462269688915e-06, + "loss": 0.0091, + "step": 38578 + }, + { + "epoch": 5.05, + "learning_rate": 6.309969365440753e-06, + "loss": 0.0101, + "step": 38579 + }, + { + "epoch": 5.05, + "learning_rate": 6.309476471570605e-06, + "loss": 0.0081, + "step": 38580 + }, + { + "epoch": 5.05, + "learning_rate": 6.308983588079856e-06, + "loss": 0.0086, + "step": 38581 + }, + { + "epoch": 5.05, + "learning_rate": 6.3084907149698974e-06, + "loss": 0.009, + "step": 38582 + }, + { + "epoch": 5.05, + "learning_rate": 6.307997852242109e-06, + "loss": 0.0072, + "step": 38583 + }, + { + "epoch": 5.05, + "learning_rate": 6.307504999897882e-06, + "loss": 0.01, + "step": 38584 + }, + { + "epoch": 5.05, + "learning_rate": 6.307012157938599e-06, + "loss": 0.0067, + "step": 38585 + }, + { + "epoch": 5.05, + "learning_rate": 6.30651932636565e-06, + "loss": 0.0056, + "step": 38586 + }, + { + "epoch": 5.05, + "learning_rate": 6.306026505180416e-06, + "loss": 0.0087, + "step": 38587 + }, + { + "epoch": 5.05, + "learning_rate": 6.305533694384285e-06, + "loss": 0.0096, + "step": 38588 + }, + { + "epoch": 5.05, + "learning_rate": 6.305040893978645e-06, + "loss": 0.0071, + "step": 38589 + }, + { + "epoch": 5.06, + "learning_rate": 6.304548103964882e-06, + "loss": 0.0075, + "step": 38590 + }, + { + "epoch": 5.06, + "learning_rate": 6.3040553243443795e-06, + "loss": 0.0056, + "step": 38591 + }, + { + "epoch": 5.06, + "learning_rate": 6.303562555118525e-06, + "loss": 0.0076, + "step": 38592 + }, + { + "epoch": 5.06, + "learning_rate": 6.303069796288704e-06, + "loss": 0.0066, + "step": 38593 + }, + { + "epoch": 5.06, + "learning_rate": 6.3025770478563044e-06, + "loss": 0.0066, + "step": 38594 + }, + { + "epoch": 5.06, + "learning_rate": 6.302084309822707e-06, + "loss": 0.0104, + "step": 38595 + }, + { + "epoch": 5.06, + "learning_rate": 6.3015915821893e-06, + "loss": 0.0086, + "step": 38596 + }, + { + "epoch": 5.06, + "learning_rate": 6.301098864957473e-06, + "loss": 0.0089, + "step": 38597 + }, + { + "epoch": 5.06, + "learning_rate": 6.300606158128606e-06, + "loss": 0.009, + "step": 38598 + }, + { + "epoch": 5.06, + "learning_rate": 6.300113461704085e-06, + "loss": 0.0069, + "step": 38599 + }, + { + "epoch": 5.06, + "learning_rate": 6.2996207756853025e-06, + "loss": 0.0068, + "step": 38600 + }, + { + "epoch": 5.06, + "learning_rate": 6.299128100073636e-06, + "loss": 0.0074, + "step": 38601 + }, + { + "epoch": 5.06, + "learning_rate": 6.298635434870477e-06, + "loss": 0.0056, + "step": 38602 + }, + { + "epoch": 5.06, + "learning_rate": 6.2981427800772075e-06, + "loss": 0.0079, + "step": 38603 + }, + { + "epoch": 5.06, + "learning_rate": 6.297650135695215e-06, + "loss": 0.0077, + "step": 38604 + }, + { + "epoch": 5.06, + "learning_rate": 6.2971575017258856e-06, + "loss": 0.0075, + "step": 38605 + }, + { + "epoch": 5.06, + "learning_rate": 6.296664878170601e-06, + "loss": 0.0089, + "step": 38606 + }, + { + "epoch": 5.06, + "learning_rate": 6.296172265030751e-06, + "loss": 0.0075, + "step": 38607 + }, + { + "epoch": 5.06, + "learning_rate": 6.2956796623077184e-06, + "loss": 0.01, + "step": 38608 + }, + { + "epoch": 5.06, + "learning_rate": 6.295187070002886e-06, + "loss": 0.009, + "step": 38609 + }, + { + "epoch": 5.06, + "learning_rate": 6.2946944881176465e-06, + "loss": 0.007, + "step": 38610 + }, + { + "epoch": 5.06, + "learning_rate": 6.294201916653382e-06, + "loss": 0.0075, + "step": 38611 + }, + { + "epoch": 5.06, + "learning_rate": 6.293709355611476e-06, + "loss": 0.0075, + "step": 38612 + }, + { + "epoch": 5.06, + "learning_rate": 6.293216804993316e-06, + "loss": 0.0072, + "step": 38613 + }, + { + "epoch": 5.06, + "learning_rate": 6.292724264800286e-06, + "loss": 0.0076, + "step": 38614 + }, + { + "epoch": 5.06, + "learning_rate": 6.292231735033774e-06, + "loss": 0.0084, + "step": 38615 + }, + { + "epoch": 5.06, + "learning_rate": 6.291739215695158e-06, + "loss": 0.0075, + "step": 38616 + }, + { + "epoch": 5.06, + "learning_rate": 6.291246706785832e-06, + "loss": 0.0077, + "step": 38617 + }, + { + "epoch": 5.06, + "learning_rate": 6.290754208307175e-06, + "loss": 0.0079, + "step": 38618 + }, + { + "epoch": 5.06, + "learning_rate": 6.290261720260573e-06, + "loss": 0.0097, + "step": 38619 + }, + { + "epoch": 5.06, + "learning_rate": 6.289769242647412e-06, + "loss": 0.0085, + "step": 38620 + }, + { + "epoch": 5.06, + "learning_rate": 6.289276775469081e-06, + "loss": 0.0062, + "step": 38621 + }, + { + "epoch": 5.06, + "learning_rate": 6.288784318726959e-06, + "loss": 0.0067, + "step": 38622 + }, + { + "epoch": 5.06, + "learning_rate": 6.288291872422436e-06, + "loss": 0.0068, + "step": 38623 + }, + { + "epoch": 5.06, + "learning_rate": 6.287799436556891e-06, + "loss": 0.0084, + "step": 38624 + }, + { + "epoch": 5.06, + "learning_rate": 6.287307011131714e-06, + "loss": 0.006, + "step": 38625 + }, + { + "epoch": 5.06, + "learning_rate": 6.28681459614829e-06, + "loss": 0.0059, + "step": 38626 + }, + { + "epoch": 5.06, + "learning_rate": 6.286322191608001e-06, + "loss": 0.0073, + "step": 38627 + }, + { + "epoch": 5.06, + "learning_rate": 6.285829797512233e-06, + "loss": 0.0089, + "step": 38628 + }, + { + "epoch": 5.06, + "learning_rate": 6.28533741386237e-06, + "loss": 0.0066, + "step": 38629 + }, + { + "epoch": 5.06, + "learning_rate": 6.284845040659799e-06, + "loss": 0.0076, + "step": 38630 + }, + { + "epoch": 5.06, + "learning_rate": 6.284352677905898e-06, + "loss": 0.0098, + "step": 38631 + }, + { + "epoch": 5.06, + "learning_rate": 6.283860325602064e-06, + "loss": 0.0073, + "step": 38632 + }, + { + "epoch": 5.06, + "learning_rate": 6.283367983749673e-06, + "loss": 0.0065, + "step": 38633 + }, + { + "epoch": 5.06, + "learning_rate": 6.282875652350113e-06, + "loss": 0.0082, + "step": 38634 + }, + { + "epoch": 5.06, + "learning_rate": 6.282383331404765e-06, + "loss": 0.009, + "step": 38635 + }, + { + "epoch": 5.06, + "learning_rate": 6.281891020915018e-06, + "loss": 0.0048, + "step": 38636 + }, + { + "epoch": 5.06, + "learning_rate": 6.281398720882253e-06, + "loss": 0.0106, + "step": 38637 + }, + { + "epoch": 5.06, + "learning_rate": 6.280906431307858e-06, + "loss": 0.0061, + "step": 38638 + }, + { + "epoch": 5.06, + "learning_rate": 6.280414152193213e-06, + "loss": 0.0065, + "step": 38639 + }, + { + "epoch": 5.06, + "learning_rate": 6.279921883539706e-06, + "loss": 0.0078, + "step": 38640 + }, + { + "epoch": 5.06, + "learning_rate": 6.2794296253487215e-06, + "loss": 0.0075, + "step": 38641 + }, + { + "epoch": 5.06, + "learning_rate": 6.278937377621639e-06, + "loss": 0.0074, + "step": 38642 + }, + { + "epoch": 5.06, + "learning_rate": 6.27844514035985e-06, + "loss": 0.0084, + "step": 38643 + }, + { + "epoch": 5.06, + "learning_rate": 6.277952913564738e-06, + "loss": 0.007, + "step": 38644 + }, + { + "epoch": 5.06, + "learning_rate": 6.277460697237683e-06, + "loss": 0.0104, + "step": 38645 + }, + { + "epoch": 5.06, + "learning_rate": 6.276968491380071e-06, + "loss": 0.0069, + "step": 38646 + }, + { + "epoch": 5.06, + "learning_rate": 6.276476295993288e-06, + "loss": 0.0094, + "step": 38647 + }, + { + "epoch": 5.06, + "learning_rate": 6.275984111078717e-06, + "loss": 0.0059, + "step": 38648 + }, + { + "epoch": 5.06, + "learning_rate": 6.275491936637743e-06, + "loss": 0.009, + "step": 38649 + }, + { + "epoch": 5.06, + "learning_rate": 6.2749997726717485e-06, + "loss": 0.0067, + "step": 38650 + }, + { + "epoch": 5.06, + "learning_rate": 6.274507619182121e-06, + "loss": 0.0095, + "step": 38651 + }, + { + "epoch": 5.06, + "learning_rate": 6.274015476170239e-06, + "loss": 0.0081, + "step": 38652 + }, + { + "epoch": 5.06, + "learning_rate": 6.273523343637489e-06, + "loss": 0.0077, + "step": 38653 + }, + { + "epoch": 5.06, + "learning_rate": 6.273031221585259e-06, + "loss": 0.0064, + "step": 38654 + }, + { + "epoch": 5.06, + "learning_rate": 6.27253911001493e-06, + "loss": 0.0071, + "step": 38655 + }, + { + "epoch": 5.06, + "learning_rate": 6.272047008927886e-06, + "loss": 0.0084, + "step": 38656 + }, + { + "epoch": 5.06, + "learning_rate": 6.2715549183255135e-06, + "loss": 0.0068, + "step": 38657 + }, + { + "epoch": 5.06, + "learning_rate": 6.271062838209191e-06, + "loss": 0.0076, + "step": 38658 + }, + { + "epoch": 5.06, + "learning_rate": 6.270570768580309e-06, + "loss": 0.0075, + "step": 38659 + }, + { + "epoch": 5.06, + "learning_rate": 6.270078709440244e-06, + "loss": 0.0078, + "step": 38660 + }, + { + "epoch": 5.06, + "learning_rate": 6.2695866607903855e-06, + "loss": 0.0081, + "step": 38661 + }, + { + "epoch": 5.06, + "learning_rate": 6.269094622632118e-06, + "loss": 0.0084, + "step": 38662 + }, + { + "epoch": 5.06, + "learning_rate": 6.2686025949668215e-06, + "loss": 0.0079, + "step": 38663 + }, + { + "epoch": 5.06, + "learning_rate": 6.268110577795878e-06, + "loss": 0.0099, + "step": 38664 + }, + { + "epoch": 5.06, + "learning_rate": 6.26761857112068e-06, + "loss": 0.01, + "step": 38665 + }, + { + "epoch": 5.06, + "learning_rate": 6.267126574942603e-06, + "loss": 0.0068, + "step": 38666 + }, + { + "epoch": 5.07, + "learning_rate": 6.266634589263036e-06, + "loss": 0.008, + "step": 38667 + }, + { + "epoch": 5.07, + "learning_rate": 6.266142614083359e-06, + "loss": 0.0085, + "step": 38668 + }, + { + "epoch": 5.07, + "learning_rate": 6.265650649404957e-06, + "loss": 0.0069, + "step": 38669 + }, + { + "epoch": 5.07, + "learning_rate": 6.265158695229217e-06, + "loss": 0.0065, + "step": 38670 + }, + { + "epoch": 5.07, + "learning_rate": 6.264666751557515e-06, + "loss": 0.0066, + "step": 38671 + }, + { + "epoch": 5.07, + "learning_rate": 6.264174818391242e-06, + "loss": 0.0091, + "step": 38672 + }, + { + "epoch": 5.07, + "learning_rate": 6.263682895731776e-06, + "loss": 0.0076, + "step": 38673 + }, + { + "epoch": 5.07, + "learning_rate": 6.263190983580505e-06, + "loss": 0.0076, + "step": 38674 + }, + { + "epoch": 5.07, + "learning_rate": 6.262699081938804e-06, + "loss": 0.0076, + "step": 38675 + }, + { + "epoch": 5.07, + "learning_rate": 6.2622071908080686e-06, + "loss": 0.0098, + "step": 38676 + }, + { + "epoch": 5.07, + "learning_rate": 6.261715310189676e-06, + "loss": 0.0078, + "step": 38677 + }, + { + "epoch": 5.07, + "learning_rate": 6.261223440085009e-06, + "loss": 0.0055, + "step": 38678 + }, + { + "epoch": 5.07, + "learning_rate": 6.260731580495453e-06, + "loss": 0.0089, + "step": 38679 + }, + { + "epoch": 5.07, + "learning_rate": 6.260239731422391e-06, + "loss": 0.0082, + "step": 38680 + }, + { + "epoch": 5.07, + "learning_rate": 6.2597478928672026e-06, + "loss": 0.0106, + "step": 38681 + }, + { + "epoch": 5.07, + "learning_rate": 6.259256064831276e-06, + "loss": 0.0101, + "step": 38682 + }, + { + "epoch": 5.07, + "learning_rate": 6.258764247315994e-06, + "loss": 0.0088, + "step": 38683 + }, + { + "epoch": 5.07, + "learning_rate": 6.258272440322737e-06, + "loss": 0.0077, + "step": 38684 + }, + { + "epoch": 5.07, + "learning_rate": 6.25778064385289e-06, + "loss": 0.0105, + "step": 38685 + }, + { + "epoch": 5.07, + "learning_rate": 6.257288857907832e-06, + "loss": 0.0078, + "step": 38686 + }, + { + "epoch": 5.07, + "learning_rate": 6.2567970824889525e-06, + "loss": 0.0094, + "step": 38687 + }, + { + "epoch": 5.07, + "learning_rate": 6.256305317597633e-06, + "loss": 0.0075, + "step": 38688 + }, + { + "epoch": 5.07, + "learning_rate": 6.255813563235254e-06, + "loss": 0.0089, + "step": 38689 + }, + { + "epoch": 5.07, + "learning_rate": 6.255321819403199e-06, + "loss": 0.0094, + "step": 38690 + }, + { + "epoch": 5.07, + "learning_rate": 6.254830086102855e-06, + "loss": 0.0068, + "step": 38691 + }, + { + "epoch": 5.07, + "learning_rate": 6.2543383633355995e-06, + "loss": 0.0081, + "step": 38692 + }, + { + "epoch": 5.07, + "learning_rate": 6.2538466511028206e-06, + "loss": 0.0078, + "step": 38693 + }, + { + "epoch": 5.07, + "learning_rate": 6.2533549494058966e-06, + "loss": 0.008, + "step": 38694 + }, + { + "epoch": 5.07, + "learning_rate": 6.252863258246213e-06, + "loss": 0.0084, + "step": 38695 + }, + { + "epoch": 5.07, + "learning_rate": 6.25237157762515e-06, + "loss": 0.0066, + "step": 38696 + }, + { + "epoch": 5.07, + "learning_rate": 6.251879907544092e-06, + "loss": 0.0082, + "step": 38697 + }, + { + "epoch": 5.07, + "learning_rate": 6.251388248004422e-06, + "loss": 0.0059, + "step": 38698 + }, + { + "epoch": 5.07, + "learning_rate": 6.2508965990075265e-06, + "loss": 0.0084, + "step": 38699 + }, + { + "epoch": 5.07, + "learning_rate": 6.2504049605547826e-06, + "loss": 0.0074, + "step": 38700 + }, + { + "epoch": 5.07, + "learning_rate": 6.249913332647577e-06, + "loss": 0.0068, + "step": 38701 + }, + { + "epoch": 5.07, + "learning_rate": 6.249421715287287e-06, + "loss": 0.0059, + "step": 38702 + }, + { + "epoch": 5.07, + "learning_rate": 6.2489301084753015e-06, + "loss": 0.0053, + "step": 38703 + }, + { + "epoch": 5.07, + "learning_rate": 6.248438512213e-06, + "loss": 0.0125, + "step": 38704 + }, + { + "epoch": 5.07, + "learning_rate": 6.247946926501764e-06, + "loss": 0.0065, + "step": 38705 + }, + { + "epoch": 5.07, + "learning_rate": 6.247455351342979e-06, + "loss": 0.0065, + "step": 38706 + }, + { + "epoch": 5.07, + "learning_rate": 6.246963786738025e-06, + "loss": 0.0089, + "step": 38707 + }, + { + "epoch": 5.07, + "learning_rate": 6.246472232688283e-06, + "loss": 0.0089, + "step": 38708 + }, + { + "epoch": 5.07, + "learning_rate": 6.245980689195142e-06, + "loss": 0.0074, + "step": 38709 + }, + { + "epoch": 5.07, + "learning_rate": 6.245489156259979e-06, + "loss": 0.0092, + "step": 38710 + }, + { + "epoch": 5.07, + "learning_rate": 6.244997633884176e-06, + "loss": 0.0092, + "step": 38711 + }, + { + "epoch": 5.07, + "learning_rate": 6.2445061220691205e-06, + "loss": 0.0082, + "step": 38712 + }, + { + "epoch": 5.07, + "learning_rate": 6.244014620816189e-06, + "loss": 0.0085, + "step": 38713 + }, + { + "epoch": 5.07, + "learning_rate": 6.24352313012677e-06, + "loss": 0.0057, + "step": 38714 + }, + { + "epoch": 5.07, + "learning_rate": 6.243031650002238e-06, + "loss": 0.0078, + "step": 38715 + }, + { + "epoch": 5.07, + "learning_rate": 6.242540180443982e-06, + "loss": 0.0053, + "step": 38716 + }, + { + "epoch": 5.07, + "learning_rate": 6.24204872145338e-06, + "loss": 0.0083, + "step": 38717 + }, + { + "epoch": 5.07, + "learning_rate": 6.241557273031817e-06, + "loss": 0.006, + "step": 38718 + }, + { + "epoch": 5.07, + "learning_rate": 6.24106583518067e-06, + "loss": 0.0059, + "step": 38719 + }, + { + "epoch": 5.07, + "learning_rate": 6.2405744079013295e-06, + "loss": 0.0067, + "step": 38720 + }, + { + "epoch": 5.07, + "learning_rate": 6.240082991195172e-06, + "loss": 0.0063, + "step": 38721 + }, + { + "epoch": 5.07, + "learning_rate": 6.239591585063582e-06, + "loss": 0.0083, + "step": 38722 + }, + { + "epoch": 5.07, + "learning_rate": 6.239100189507939e-06, + "loss": 0.0097, + "step": 38723 + }, + { + "epoch": 5.07, + "learning_rate": 6.2386088045296286e-06, + "loss": 0.0094, + "step": 38724 + }, + { + "epoch": 5.07, + "learning_rate": 6.238117430130028e-06, + "loss": 0.0078, + "step": 38725 + }, + { + "epoch": 5.07, + "learning_rate": 6.237626066310523e-06, + "loss": 0.007, + "step": 38726 + }, + { + "epoch": 5.07, + "learning_rate": 6.237134713072496e-06, + "loss": 0.0079, + "step": 38727 + }, + { + "epoch": 5.07, + "learning_rate": 6.236643370417325e-06, + "loss": 0.0099, + "step": 38728 + }, + { + "epoch": 5.07, + "learning_rate": 6.236152038346396e-06, + "loss": 0.0087, + "step": 38729 + }, + { + "epoch": 5.07, + "learning_rate": 6.235660716861085e-06, + "loss": 0.006, + "step": 38730 + }, + { + "epoch": 5.07, + "learning_rate": 6.23516940596278e-06, + "loss": 0.0066, + "step": 38731 + }, + { + "epoch": 5.07, + "learning_rate": 6.234678105652859e-06, + "loss": 0.0072, + "step": 38732 + }, + { + "epoch": 5.07, + "learning_rate": 6.234186815932709e-06, + "loss": 0.0059, + "step": 38733 + }, + { + "epoch": 5.07, + "learning_rate": 6.2336955368037056e-06, + "loss": 0.0065, + "step": 38734 + }, + { + "epoch": 5.07, + "learning_rate": 6.233204268267234e-06, + "loss": 0.0082, + "step": 38735 + }, + { + "epoch": 5.07, + "learning_rate": 6.2327130103246745e-06, + "loss": 0.0081, + "step": 38736 + }, + { + "epoch": 5.07, + "learning_rate": 6.232221762977411e-06, + "loss": 0.0078, + "step": 38737 + }, + { + "epoch": 5.07, + "learning_rate": 6.23173052622682e-06, + "loss": 0.0094, + "step": 38738 + }, + { + "epoch": 5.07, + "learning_rate": 6.231239300074289e-06, + "loss": 0.0068, + "step": 38739 + }, + { + "epoch": 5.07, + "learning_rate": 6.2307480845211945e-06, + "loss": 0.0082, + "step": 38740 + }, + { + "epoch": 5.07, + "learning_rate": 6.2302568795689176e-06, + "loss": 0.0064, + "step": 38741 + }, + { + "epoch": 5.07, + "learning_rate": 6.2297656852188445e-06, + "loss": 0.0106, + "step": 38742 + }, + { + "epoch": 5.08, + "learning_rate": 6.229274501472358e-06, + "loss": 0.0073, + "step": 38743 + }, + { + "epoch": 5.08, + "learning_rate": 6.228783328330834e-06, + "loss": 0.0092, + "step": 38744 + }, + { + "epoch": 5.08, + "learning_rate": 6.2282921657956574e-06, + "loss": 0.0085, + "step": 38745 + }, + { + "epoch": 5.08, + "learning_rate": 6.227801013868206e-06, + "loss": 0.009, + "step": 38746 + }, + { + "epoch": 5.08, + "learning_rate": 6.227309872549862e-06, + "loss": 0.0086, + "step": 38747 + }, + { + "epoch": 5.08, + "learning_rate": 6.226818741842014e-06, + "loss": 0.0076, + "step": 38748 + }, + { + "epoch": 5.08, + "learning_rate": 6.226327621746032e-06, + "loss": 0.0087, + "step": 38749 + }, + { + "epoch": 5.08, + "learning_rate": 6.225836512263306e-06, + "loss": 0.0097, + "step": 38750 + }, + { + "epoch": 5.08, + "learning_rate": 6.225345413395209e-06, + "loss": 0.0078, + "step": 38751 + }, + { + "epoch": 5.08, + "learning_rate": 6.2248543251431305e-06, + "loss": 0.0085, + "step": 38752 + }, + { + "epoch": 5.08, + "learning_rate": 6.22436324750845e-06, + "loss": 0.0071, + "step": 38753 + }, + { + "epoch": 5.08, + "learning_rate": 6.223872180492543e-06, + "loss": 0.0089, + "step": 38754 + }, + { + "epoch": 5.08, + "learning_rate": 6.223381124096795e-06, + "loss": 0.0071, + "step": 38755 + }, + { + "epoch": 5.08, + "learning_rate": 6.222890078322589e-06, + "loss": 0.009, + "step": 38756 + }, + { + "epoch": 5.08, + "learning_rate": 6.222399043171301e-06, + "loss": 0.0068, + "step": 38757 + }, + { + "epoch": 5.08, + "learning_rate": 6.221908018644317e-06, + "loss": 0.0073, + "step": 38758 + }, + { + "epoch": 5.08, + "learning_rate": 6.221417004743014e-06, + "loss": 0.0076, + "step": 38759 + }, + { + "epoch": 5.08, + "learning_rate": 6.220926001468776e-06, + "loss": 0.0079, + "step": 38760 + }, + { + "epoch": 5.08, + "learning_rate": 6.22043500882298e-06, + "loss": 0.0063, + "step": 38761 + }, + { + "epoch": 5.08, + "learning_rate": 6.219944026807008e-06, + "loss": 0.0098, + "step": 38762 + }, + { + "epoch": 5.08, + "learning_rate": 6.219453055422242e-06, + "loss": 0.0044, + "step": 38763 + }, + { + "epoch": 5.08, + "learning_rate": 6.218962094670067e-06, + "loss": 0.0075, + "step": 38764 + }, + { + "epoch": 5.08, + "learning_rate": 6.218471144551857e-06, + "loss": 0.009, + "step": 38765 + }, + { + "epoch": 5.08, + "learning_rate": 6.217980205068999e-06, + "loss": 0.0064, + "step": 38766 + }, + { + "epoch": 5.08, + "learning_rate": 6.217489276222867e-06, + "loss": 0.0086, + "step": 38767 + }, + { + "epoch": 5.08, + "learning_rate": 6.216998358014846e-06, + "loss": 0.006, + "step": 38768 + }, + { + "epoch": 5.08, + "learning_rate": 6.2165074504463155e-06, + "loss": 0.0066, + "step": 38769 + }, + { + "epoch": 5.08, + "learning_rate": 6.216016553518656e-06, + "loss": 0.0058, + "step": 38770 + }, + { + "epoch": 5.08, + "learning_rate": 6.215525667233248e-06, + "loss": 0.0066, + "step": 38771 + }, + { + "epoch": 5.08, + "learning_rate": 6.215034791591474e-06, + "loss": 0.0064, + "step": 38772 + }, + { + "epoch": 5.08, + "learning_rate": 6.214543926594709e-06, + "loss": 0.0082, + "step": 38773 + }, + { + "epoch": 5.08, + "learning_rate": 6.214053072244342e-06, + "loss": 0.0065, + "step": 38774 + }, + { + "epoch": 5.08, + "learning_rate": 6.213562228541748e-06, + "loss": 0.0064, + "step": 38775 + }, + { + "epoch": 5.08, + "learning_rate": 6.213071395488308e-06, + "loss": 0.0065, + "step": 38776 + }, + { + "epoch": 5.08, + "learning_rate": 6.212580573085405e-06, + "loss": 0.0076, + "step": 38777 + }, + { + "epoch": 5.08, + "learning_rate": 6.212089761334415e-06, + "loss": 0.0068, + "step": 38778 + }, + { + "epoch": 5.08, + "learning_rate": 6.211598960236724e-06, + "loss": 0.0066, + "step": 38779 + }, + { + "epoch": 5.08, + "learning_rate": 6.211108169793707e-06, + "loss": 0.005, + "step": 38780 + }, + { + "epoch": 5.08, + "learning_rate": 6.2106173900067475e-06, + "loss": 0.0073, + "step": 38781 + }, + { + "epoch": 5.08, + "learning_rate": 6.2101266208772225e-06, + "loss": 0.007, + "step": 38782 + }, + { + "epoch": 5.08, + "learning_rate": 6.209635862406518e-06, + "loss": 0.0092, + "step": 38783 + }, + { + "epoch": 5.08, + "learning_rate": 6.209145114596003e-06, + "loss": 0.0065, + "step": 38784 + }, + { + "epoch": 5.08, + "learning_rate": 6.208654377447073e-06, + "loss": 0.0055, + "step": 38785 + }, + { + "epoch": 5.08, + "learning_rate": 6.208163650961099e-06, + "loss": 0.0078, + "step": 38786 + }, + { + "epoch": 5.08, + "learning_rate": 6.207672935139464e-06, + "loss": 0.007, + "step": 38787 + }, + { + "epoch": 5.08, + "learning_rate": 6.207182229983544e-06, + "loss": 0.0082, + "step": 38788 + }, + { + "epoch": 5.08, + "learning_rate": 6.206691535494725e-06, + "loss": 0.0071, + "step": 38789 + }, + { + "epoch": 5.08, + "learning_rate": 6.206200851674382e-06, + "loss": 0.0081, + "step": 38790 + }, + { + "epoch": 5.08, + "learning_rate": 6.205710178523898e-06, + "loss": 0.0103, + "step": 38791 + }, + { + "epoch": 5.08, + "learning_rate": 6.205219516044653e-06, + "loss": 0.0081, + "step": 38792 + }, + { + "epoch": 5.08, + "learning_rate": 6.204728864238024e-06, + "loss": 0.0085, + "step": 38793 + }, + { + "epoch": 5.08, + "learning_rate": 6.204238223105395e-06, + "loss": 0.005, + "step": 38794 + }, + { + "epoch": 5.08, + "learning_rate": 6.20374759264814e-06, + "loss": 0.0053, + "step": 38795 + }, + { + "epoch": 5.08, + "learning_rate": 6.2032569728676435e-06, + "loss": 0.0114, + "step": 38796 + }, + { + "epoch": 5.08, + "learning_rate": 6.202766363765287e-06, + "loss": 0.0068, + "step": 38797 + }, + { + "epoch": 5.08, + "learning_rate": 6.2022757653424495e-06, + "loss": 0.0118, + "step": 38798 + }, + { + "epoch": 5.08, + "learning_rate": 6.201785177600505e-06, + "loss": 0.0066, + "step": 38799 + }, + { + "epoch": 5.08, + "learning_rate": 6.201294600540842e-06, + "loss": 0.0098, + "step": 38800 + }, + { + "epoch": 5.08, + "learning_rate": 6.200804034164831e-06, + "loss": 0.0062, + "step": 38801 + }, + { + "epoch": 5.08, + "learning_rate": 6.2003134784738604e-06, + "loss": 0.0082, + "step": 38802 + }, + { + "epoch": 5.08, + "learning_rate": 6.199822933469304e-06, + "loss": 0.0086, + "step": 38803 + }, + { + "epoch": 5.08, + "learning_rate": 6.199332399152544e-06, + "loss": 0.0095, + "step": 38804 + }, + { + "epoch": 5.08, + "learning_rate": 6.1988418755249565e-06, + "loss": 0.006, + "step": 38805 + }, + { + "epoch": 5.08, + "learning_rate": 6.198351362587924e-06, + "loss": 0.0072, + "step": 38806 + }, + { + "epoch": 5.08, + "learning_rate": 6.197860860342826e-06, + "loss": 0.0076, + "step": 38807 + }, + { + "epoch": 5.08, + "learning_rate": 6.197370368791044e-06, + "loss": 0.006, + "step": 38808 + }, + { + "epoch": 5.08, + "learning_rate": 6.196879887933955e-06, + "loss": 0.0066, + "step": 38809 + }, + { + "epoch": 5.08, + "learning_rate": 6.196389417772939e-06, + "loss": 0.005, + "step": 38810 + }, + { + "epoch": 5.08, + "learning_rate": 6.195898958309375e-06, + "loss": 0.0078, + "step": 38811 + }, + { + "epoch": 5.08, + "learning_rate": 6.195408509544641e-06, + "loss": 0.0082, + "step": 38812 + }, + { + "epoch": 5.08, + "learning_rate": 6.19491807148012e-06, + "loss": 0.0075, + "step": 38813 + }, + { + "epoch": 5.08, + "learning_rate": 6.194427644117189e-06, + "loss": 0.0088, + "step": 38814 + }, + { + "epoch": 5.08, + "learning_rate": 6.193937227457227e-06, + "loss": 0.0086, + "step": 38815 + }, + { + "epoch": 5.08, + "learning_rate": 6.193446821501614e-06, + "loss": 0.0065, + "step": 38816 + }, + { + "epoch": 5.08, + "learning_rate": 6.192956426251726e-06, + "loss": 0.008, + "step": 38817 + }, + { + "epoch": 5.08, + "learning_rate": 6.19246604170895e-06, + "loss": 0.0074, + "step": 38818 + }, + { + "epoch": 5.09, + "learning_rate": 6.191975667874659e-06, + "loss": 0.0104, + "step": 38819 + }, + { + "epoch": 5.09, + "learning_rate": 6.191485304750232e-06, + "loss": 0.0071, + "step": 38820 + }, + { + "epoch": 5.09, + "learning_rate": 6.190994952337053e-06, + "loss": 0.0067, + "step": 38821 + }, + { + "epoch": 5.09, + "learning_rate": 6.190504610636495e-06, + "loss": 0.0071, + "step": 38822 + }, + { + "epoch": 5.09, + "learning_rate": 6.190014279649943e-06, + "loss": 0.0081, + "step": 38823 + }, + { + "epoch": 5.09, + "learning_rate": 6.189523959378769e-06, + "loss": 0.0071, + "step": 38824 + }, + { + "epoch": 5.09, + "learning_rate": 6.1890336498243595e-06, + "loss": 0.0059, + "step": 38825 + }, + { + "epoch": 5.09, + "learning_rate": 6.188543350988088e-06, + "loss": 0.0064, + "step": 38826 + }, + { + "epoch": 5.09, + "learning_rate": 6.1880530628713355e-06, + "loss": 0.0087, + "step": 38827 + }, + { + "epoch": 5.09, + "learning_rate": 6.187562785475478e-06, + "loss": 0.0075, + "step": 38828 + }, + { + "epoch": 5.09, + "learning_rate": 6.1870725188019006e-06, + "loss": 0.0071, + "step": 38829 + }, + { + "epoch": 5.09, + "learning_rate": 6.1865822628519775e-06, + "loss": 0.0084, + "step": 38830 + }, + { + "epoch": 5.09, + "learning_rate": 6.186092017627091e-06, + "loss": 0.0063, + "step": 38831 + }, + { + "epoch": 5.09, + "learning_rate": 6.185601783128615e-06, + "loss": 0.0053, + "step": 38832 + }, + { + "epoch": 5.09, + "learning_rate": 6.185111559357934e-06, + "loss": 0.0108, + "step": 38833 + }, + { + "epoch": 5.09, + "learning_rate": 6.18462134631642e-06, + "loss": 0.0068, + "step": 38834 + }, + { + "epoch": 5.09, + "learning_rate": 6.184131144005457e-06, + "loss": 0.0064, + "step": 38835 + }, + { + "epoch": 5.09, + "learning_rate": 6.183640952426422e-06, + "loss": 0.0082, + "step": 38836 + }, + { + "epoch": 5.09, + "learning_rate": 6.1831507715806925e-06, + "loss": 0.0073, + "step": 38837 + }, + { + "epoch": 5.09, + "learning_rate": 6.18266060146965e-06, + "loss": 0.0065, + "step": 38838 + }, + { + "epoch": 5.09, + "learning_rate": 6.182170442094668e-06, + "loss": 0.0082, + "step": 38839 + }, + { + "epoch": 5.09, + "learning_rate": 6.181680293457131e-06, + "loss": 0.0076, + "step": 38840 + }, + { + "epoch": 5.09, + "learning_rate": 6.181190155558412e-06, + "loss": 0.0078, + "step": 38841 + }, + { + "epoch": 5.09, + "learning_rate": 6.1807000283998974e-06, + "loss": 0.0056, + "step": 38842 + }, + { + "epoch": 5.09, + "learning_rate": 6.180209911982957e-06, + "loss": 0.0076, + "step": 38843 + }, + { + "epoch": 5.09, + "learning_rate": 6.179719806308975e-06, + "loss": 0.0068, + "step": 38844 + }, + { + "epoch": 5.09, + "learning_rate": 6.179229711379325e-06, + "loss": 0.0092, + "step": 38845 + }, + { + "epoch": 5.09, + "learning_rate": 6.1787396271953915e-06, + "loss": 0.0078, + "step": 38846 + }, + { + "epoch": 5.09, + "learning_rate": 6.178249553758546e-06, + "loss": 0.0088, + "step": 38847 + }, + { + "epoch": 5.09, + "learning_rate": 6.17775949107017e-06, + "loss": 0.0064, + "step": 38848 + }, + { + "epoch": 5.09, + "learning_rate": 6.177269439131645e-06, + "loss": 0.0089, + "step": 38849 + }, + { + "epoch": 5.09, + "learning_rate": 6.176779397944341e-06, + "loss": 0.0094, + "step": 38850 + }, + { + "epoch": 5.09, + "learning_rate": 6.176289367509643e-06, + "loss": 0.0063, + "step": 38851 + }, + { + "epoch": 5.09, + "learning_rate": 6.1757993478289315e-06, + "loss": 0.0056, + "step": 38852 + }, + { + "epoch": 5.09, + "learning_rate": 6.175309338903576e-06, + "loss": 0.0073, + "step": 38853 + }, + { + "epoch": 5.09, + "learning_rate": 6.174819340734963e-06, + "loss": 0.0114, + "step": 38854 + }, + { + "epoch": 5.09, + "learning_rate": 6.1743293533244665e-06, + "loss": 0.0065, + "step": 38855 + }, + { + "epoch": 5.09, + "learning_rate": 6.173839376673463e-06, + "loss": 0.0079, + "step": 38856 + }, + { + "epoch": 5.09, + "learning_rate": 6.173349410783334e-06, + "loss": 0.0088, + "step": 38857 + }, + { + "epoch": 5.09, + "learning_rate": 6.172859455655456e-06, + "loss": 0.0097, + "step": 38858 + }, + { + "epoch": 5.09, + "learning_rate": 6.172369511291208e-06, + "loss": 0.0091, + "step": 38859 + }, + { + "epoch": 5.09, + "learning_rate": 6.171879577691964e-06, + "loss": 0.0067, + "step": 38860 + }, + { + "epoch": 5.09, + "learning_rate": 6.171389654859104e-06, + "loss": 0.0082, + "step": 38861 + }, + { + "epoch": 5.09, + "learning_rate": 6.170899742794008e-06, + "loss": 0.0083, + "step": 38862 + }, + { + "epoch": 5.09, + "learning_rate": 6.170409841498053e-06, + "loss": 0.0075, + "step": 38863 + }, + { + "epoch": 5.09, + "learning_rate": 6.169919950972616e-06, + "loss": 0.0073, + "step": 38864 + }, + { + "epoch": 5.09, + "learning_rate": 6.169430071219078e-06, + "loss": 0.0082, + "step": 38865 + }, + { + "epoch": 5.09, + "learning_rate": 6.168940202238811e-06, + "loss": 0.0062, + "step": 38866 + }, + { + "epoch": 5.09, + "learning_rate": 6.168450344033199e-06, + "loss": 0.0079, + "step": 38867 + }, + { + "epoch": 5.09, + "learning_rate": 6.1679604966036125e-06, + "loss": 0.008, + "step": 38868 + }, + { + "epoch": 5.09, + "learning_rate": 6.167470659951436e-06, + "loss": 0.0065, + "step": 38869 + }, + { + "epoch": 5.09, + "learning_rate": 6.166980834078042e-06, + "loss": 0.0065, + "step": 38870 + }, + { + "epoch": 5.09, + "learning_rate": 6.16649101898481e-06, + "loss": 0.0064, + "step": 38871 + }, + { + "epoch": 5.09, + "learning_rate": 6.166001214673116e-06, + "loss": 0.009, + "step": 38872 + }, + { + "epoch": 5.09, + "learning_rate": 6.165511421144343e-06, + "loss": 0.0084, + "step": 38873 + }, + { + "epoch": 5.09, + "learning_rate": 6.165021638399864e-06, + "loss": 0.0081, + "step": 38874 + }, + { + "epoch": 5.09, + "learning_rate": 6.1645318664410596e-06, + "loss": 0.0108, + "step": 38875 + }, + { + "epoch": 5.09, + "learning_rate": 6.164042105269301e-06, + "loss": 0.0078, + "step": 38876 + }, + { + "epoch": 5.09, + "learning_rate": 6.163552354885971e-06, + "loss": 0.0055, + "step": 38877 + }, + { + "epoch": 5.09, + "learning_rate": 6.163062615292447e-06, + "loss": 0.0085, + "step": 38878 + }, + { + "epoch": 5.09, + "learning_rate": 6.162572886490106e-06, + "loss": 0.0078, + "step": 38879 + }, + { + "epoch": 5.09, + "learning_rate": 6.162083168480324e-06, + "loss": 0.0088, + "step": 38880 + }, + { + "epoch": 5.09, + "learning_rate": 6.161593461264477e-06, + "loss": 0.0097, + "step": 38881 + }, + { + "epoch": 5.09, + "learning_rate": 6.161103764843941e-06, + "loss": 0.0078, + "step": 38882 + }, + { + "epoch": 5.09, + "learning_rate": 6.160614079220102e-06, + "loss": 0.0083, + "step": 38883 + }, + { + "epoch": 5.09, + "learning_rate": 6.160124404394327e-06, + "loss": 0.0079, + "step": 38884 + }, + { + "epoch": 5.09, + "learning_rate": 6.159634740367999e-06, + "loss": 0.0089, + "step": 38885 + }, + { + "epoch": 5.09, + "learning_rate": 6.159145087142497e-06, + "loss": 0.0081, + "step": 38886 + }, + { + "epoch": 5.09, + "learning_rate": 6.158655444719192e-06, + "loss": 0.0067, + "step": 38887 + }, + { + "epoch": 5.09, + "learning_rate": 6.1581658130994655e-06, + "loss": 0.0091, + "step": 38888 + }, + { + "epoch": 5.09, + "learning_rate": 6.1576761922846915e-06, + "loss": 0.0098, + "step": 38889 + }, + { + "epoch": 5.09, + "learning_rate": 6.157186582276251e-06, + "loss": 0.0106, + "step": 38890 + }, + { + "epoch": 5.09, + "learning_rate": 6.156696983075516e-06, + "loss": 0.0067, + "step": 38891 + }, + { + "epoch": 5.09, + "learning_rate": 6.156207394683866e-06, + "loss": 0.0071, + "step": 38892 + }, + { + "epoch": 5.09, + "learning_rate": 6.1557178171026756e-06, + "loss": 0.0104, + "step": 38893 + }, + { + "epoch": 5.09, + "learning_rate": 6.155228250333329e-06, + "loss": 0.0079, + "step": 38894 + }, + { + "epoch": 5.09, + "learning_rate": 6.154738694377196e-06, + "loss": 0.0092, + "step": 38895 + }, + { + "epoch": 5.1, + "learning_rate": 6.154249149235658e-06, + "loss": 0.0104, + "step": 38896 + }, + { + "epoch": 5.1, + "learning_rate": 6.153759614910087e-06, + "loss": 0.0078, + "step": 38897 + }, + { + "epoch": 5.1, + "learning_rate": 6.153270091401862e-06, + "loss": 0.0095, + "step": 38898 + }, + { + "epoch": 5.1, + "learning_rate": 6.152780578712363e-06, + "loss": 0.0087, + "step": 38899 + }, + { + "epoch": 5.1, + "learning_rate": 6.15229107684296e-06, + "loss": 0.0082, + "step": 38900 + }, + { + "epoch": 5.1, + "learning_rate": 6.151801585795036e-06, + "loss": 0.0077, + "step": 38901 + }, + { + "epoch": 5.1, + "learning_rate": 6.151312105569964e-06, + "loss": 0.0073, + "step": 38902 + }, + { + "epoch": 5.1, + "learning_rate": 6.150822636169124e-06, + "loss": 0.0094, + "step": 38903 + }, + { + "epoch": 5.1, + "learning_rate": 6.150333177593886e-06, + "loss": 0.0076, + "step": 38904 + }, + { + "epoch": 5.1, + "learning_rate": 6.1498437298456325e-06, + "loss": 0.0097, + "step": 38905 + }, + { + "epoch": 5.1, + "learning_rate": 6.149354292925738e-06, + "loss": 0.0078, + "step": 38906 + }, + { + "epoch": 5.1, + "learning_rate": 6.148864866835582e-06, + "loss": 0.0057, + "step": 38907 + }, + { + "epoch": 5.1, + "learning_rate": 6.148375451576537e-06, + "loss": 0.0058, + "step": 38908 + }, + { + "epoch": 5.1, + "learning_rate": 6.147886047149982e-06, + "loss": 0.0057, + "step": 38909 + }, + { + "epoch": 5.1, + "learning_rate": 6.147396653557291e-06, + "loss": 0.0082, + "step": 38910 + }, + { + "epoch": 5.1, + "learning_rate": 6.146907270799845e-06, + "loss": 0.0068, + "step": 38911 + }, + { + "epoch": 5.1, + "learning_rate": 6.146417898879013e-06, + "loss": 0.0088, + "step": 38912 + }, + { + "epoch": 5.1, + "learning_rate": 6.145928537796175e-06, + "loss": 0.0069, + "step": 38913 + }, + { + "epoch": 5.1, + "learning_rate": 6.145439187552712e-06, + "loss": 0.0072, + "step": 38914 + }, + { + "epoch": 5.1, + "learning_rate": 6.144949848149991e-06, + "loss": 0.0076, + "step": 38915 + }, + { + "epoch": 5.1, + "learning_rate": 6.144460519589395e-06, + "loss": 0.0093, + "step": 38916 + }, + { + "epoch": 5.1, + "learning_rate": 6.143971201872301e-06, + "loss": 0.0119, + "step": 38917 + }, + { + "epoch": 5.1, + "learning_rate": 6.143481895000081e-06, + "loss": 0.0082, + "step": 38918 + }, + { + "epoch": 5.1, + "learning_rate": 6.142992598974113e-06, + "loss": 0.0089, + "step": 38919 + }, + { + "epoch": 5.1, + "learning_rate": 6.142503313795773e-06, + "loss": 0.0084, + "step": 38920 + }, + { + "epoch": 5.1, + "learning_rate": 6.142014039466435e-06, + "loss": 0.0068, + "step": 38921 + }, + { + "epoch": 5.1, + "learning_rate": 6.14152477598748e-06, + "loss": 0.0082, + "step": 38922 + }, + { + "epoch": 5.1, + "learning_rate": 6.141035523360279e-06, + "loss": 0.0076, + "step": 38923 + }, + { + "epoch": 5.1, + "learning_rate": 6.140546281586212e-06, + "loss": 0.0091, + "step": 38924 + }, + { + "epoch": 5.1, + "learning_rate": 6.140057050666653e-06, + "loss": 0.0061, + "step": 38925 + }, + { + "epoch": 5.1, + "learning_rate": 6.139567830602972e-06, + "loss": 0.0068, + "step": 38926 + }, + { + "epoch": 5.1, + "learning_rate": 6.139078621396557e-06, + "loss": 0.0092, + "step": 38927 + }, + { + "epoch": 5.1, + "learning_rate": 6.1385894230487775e-06, + "loss": 0.008, + "step": 38928 + }, + { + "epoch": 5.1, + "learning_rate": 6.138100235561008e-06, + "loss": 0.0077, + "step": 38929 + }, + { + "epoch": 5.1, + "learning_rate": 6.1376110589346284e-06, + "loss": 0.0065, + "step": 38930 + }, + { + "epoch": 5.1, + "learning_rate": 6.13712189317101e-06, + "loss": 0.0074, + "step": 38931 + }, + { + "epoch": 5.1, + "learning_rate": 6.1366327382715314e-06, + "loss": 0.0077, + "step": 38932 + }, + { + "epoch": 5.1, + "learning_rate": 6.136143594237567e-06, + "loss": 0.006, + "step": 38933 + }, + { + "epoch": 5.1, + "learning_rate": 6.135654461070495e-06, + "loss": 0.0081, + "step": 38934 + }, + { + "epoch": 5.1, + "learning_rate": 6.135165338771685e-06, + "loss": 0.0084, + "step": 38935 + }, + { + "epoch": 5.1, + "learning_rate": 6.134676227342518e-06, + "loss": 0.005, + "step": 38936 + }, + { + "epoch": 5.1, + "learning_rate": 6.134187126784367e-06, + "loss": 0.0083, + "step": 38937 + }, + { + "epoch": 5.1, + "learning_rate": 6.133698037098612e-06, + "loss": 0.0071, + "step": 38938 + }, + { + "epoch": 5.1, + "learning_rate": 6.133208958286623e-06, + "loss": 0.0062, + "step": 38939 + }, + { + "epoch": 5.1, + "learning_rate": 6.132719890349781e-06, + "loss": 0.0066, + "step": 38940 + }, + { + "epoch": 5.1, + "learning_rate": 6.1322308332894556e-06, + "loss": 0.0097, + "step": 38941 + }, + { + "epoch": 5.1, + "learning_rate": 6.131741787107025e-06, + "loss": 0.009, + "step": 38942 + }, + { + "epoch": 5.1, + "learning_rate": 6.131252751803867e-06, + "loss": 0.0093, + "step": 38943 + }, + { + "epoch": 5.1, + "learning_rate": 6.130763727381353e-06, + "loss": 0.0084, + "step": 38944 + }, + { + "epoch": 5.1, + "learning_rate": 6.13027471384086e-06, + "loss": 0.0082, + "step": 38945 + }, + { + "epoch": 5.1, + "learning_rate": 6.1297857111837635e-06, + "loss": 0.007, + "step": 38946 + }, + { + "epoch": 5.1, + "learning_rate": 6.129296719411439e-06, + "loss": 0.0091, + "step": 38947 + }, + { + "epoch": 5.1, + "learning_rate": 6.1288077385252566e-06, + "loss": 0.0068, + "step": 38948 + }, + { + "epoch": 5.1, + "learning_rate": 6.128318768526602e-06, + "loss": 0.0064, + "step": 38949 + }, + { + "epoch": 5.1, + "learning_rate": 6.127829809416842e-06, + "loss": 0.0082, + "step": 38950 + }, + { + "epoch": 5.1, + "learning_rate": 6.1273408611973575e-06, + "loss": 0.0072, + "step": 38951 + }, + { + "epoch": 5.1, + "learning_rate": 6.1268519238695165e-06, + "loss": 0.0064, + "step": 38952 + }, + { + "epoch": 5.1, + "learning_rate": 6.126362997434703e-06, + "loss": 0.0107, + "step": 38953 + }, + { + "epoch": 5.1, + "learning_rate": 6.125874081894283e-06, + "loss": 0.0084, + "step": 38954 + }, + { + "epoch": 5.1, + "learning_rate": 6.12538517724964e-06, + "loss": 0.0077, + "step": 38955 + }, + { + "epoch": 5.1, + "learning_rate": 6.124896283502142e-06, + "loss": 0.009, + "step": 38956 + }, + { + "epoch": 5.1, + "learning_rate": 6.124407400653165e-06, + "loss": 0.0078, + "step": 38957 + }, + { + "epoch": 5.1, + "learning_rate": 6.1239185287040904e-06, + "loss": 0.0107, + "step": 38958 + }, + { + "epoch": 5.1, + "learning_rate": 6.1234296676562835e-06, + "loss": 0.0107, + "step": 38959 + }, + { + "epoch": 5.1, + "learning_rate": 6.122940817511126e-06, + "loss": 0.0066, + "step": 38960 + }, + { + "epoch": 5.1, + "learning_rate": 6.122451978269993e-06, + "loss": 0.01, + "step": 38961 + }, + { + "epoch": 5.1, + "learning_rate": 6.121963149934255e-06, + "loss": 0.0091, + "step": 38962 + }, + { + "epoch": 5.1, + "learning_rate": 6.121474332505291e-06, + "loss": 0.0073, + "step": 38963 + }, + { + "epoch": 5.1, + "learning_rate": 6.120985525984473e-06, + "loss": 0.0083, + "step": 38964 + }, + { + "epoch": 5.1, + "learning_rate": 6.120496730373176e-06, + "loss": 0.0063, + "step": 38965 + }, + { + "epoch": 5.1, + "learning_rate": 6.120007945672779e-06, + "loss": 0.0067, + "step": 38966 + }, + { + "epoch": 5.1, + "learning_rate": 6.1195191718846495e-06, + "loss": 0.0102, + "step": 38967 + }, + { + "epoch": 5.1, + "learning_rate": 6.119030409010167e-06, + "loss": 0.0095, + "step": 38968 + }, + { + "epoch": 5.1, + "learning_rate": 6.118541657050703e-06, + "loss": 0.0066, + "step": 38969 + }, + { + "epoch": 5.1, + "learning_rate": 6.118052916007633e-06, + "loss": 0.0101, + "step": 38970 + }, + { + "epoch": 5.1, + "learning_rate": 6.117564185882334e-06, + "loss": 0.0098, + "step": 38971 + }, + { + "epoch": 5.11, + "learning_rate": 6.117075466676181e-06, + "loss": 0.0092, + "step": 38972 + }, + { + "epoch": 5.11, + "learning_rate": 6.116586758390544e-06, + "loss": 0.009, + "step": 38973 + }, + { + "epoch": 5.11, + "learning_rate": 6.116098061026802e-06, + "loss": 0.0092, + "step": 38974 + }, + { + "epoch": 5.11, + "learning_rate": 6.115609374586325e-06, + "loss": 0.0089, + "step": 38975 + }, + { + "epoch": 5.11, + "learning_rate": 6.115120699070493e-06, + "loss": 0.0055, + "step": 38976 + }, + { + "epoch": 5.11, + "learning_rate": 6.114632034480674e-06, + "loss": 0.0085, + "step": 38977 + }, + { + "epoch": 5.11, + "learning_rate": 6.114143380818247e-06, + "loss": 0.0078, + "step": 38978 + }, + { + "epoch": 5.11, + "learning_rate": 6.113654738084584e-06, + "loss": 0.0076, + "step": 38979 + }, + { + "epoch": 5.11, + "learning_rate": 6.113166106281061e-06, + "loss": 0.0067, + "step": 38980 + }, + { + "epoch": 5.11, + "learning_rate": 6.112677485409049e-06, + "loss": 0.0075, + "step": 38981 + }, + { + "epoch": 5.11, + "learning_rate": 6.112188875469926e-06, + "loss": 0.0079, + "step": 38982 + }, + { + "epoch": 5.11, + "learning_rate": 6.111700276465066e-06, + "loss": 0.0082, + "step": 38983 + }, + { + "epoch": 5.11, + "learning_rate": 6.1112116883958414e-06, + "loss": 0.0068, + "step": 38984 + }, + { + "epoch": 5.11, + "learning_rate": 6.110723111263627e-06, + "loss": 0.0083, + "step": 38985 + }, + { + "epoch": 5.11, + "learning_rate": 6.110234545069795e-06, + "loss": 0.0056, + "step": 38986 + }, + { + "epoch": 5.11, + "learning_rate": 6.109745989815724e-06, + "loss": 0.0103, + "step": 38987 + }, + { + "epoch": 5.11, + "learning_rate": 6.109257445502783e-06, + "loss": 0.0091, + "step": 38988 + }, + { + "epoch": 5.11, + "learning_rate": 6.10876891213235e-06, + "loss": 0.01, + "step": 38989 + }, + { + "epoch": 5.11, + "learning_rate": 6.1082803897057965e-06, + "loss": 0.0097, + "step": 38990 + }, + { + "epoch": 5.11, + "learning_rate": 6.107791878224498e-06, + "loss": 0.0066, + "step": 38991 + }, + { + "epoch": 5.11, + "learning_rate": 6.107303377689822e-06, + "loss": 0.0078, + "step": 38992 + }, + { + "epoch": 5.11, + "learning_rate": 6.1068148881031546e-06, + "loss": 0.0067, + "step": 38993 + }, + { + "epoch": 5.11, + "learning_rate": 6.1063264094658595e-06, + "loss": 0.0069, + "step": 38994 + }, + { + "epoch": 5.11, + "learning_rate": 6.1058379417793156e-06, + "loss": 0.0045, + "step": 38995 + }, + { + "epoch": 5.11, + "learning_rate": 6.105349485044895e-06, + "loss": 0.0082, + "step": 38996 + }, + { + "epoch": 5.11, + "learning_rate": 6.1048610392639715e-06, + "loss": 0.0072, + "step": 38997 + }, + { + "epoch": 5.11, + "learning_rate": 6.104372604437919e-06, + "loss": 0.0089, + "step": 38998 + }, + { + "epoch": 5.11, + "learning_rate": 6.10388418056811e-06, + "loss": 0.0102, + "step": 38999 + }, + { + "epoch": 5.11, + "learning_rate": 6.103395767655922e-06, + "loss": 0.0084, + "step": 39000 + }, + { + "epoch": 5.11, + "learning_rate": 6.102907365702724e-06, + "loss": 0.0093, + "step": 39001 + }, + { + "epoch": 5.11, + "learning_rate": 6.102418974709891e-06, + "loss": 0.0082, + "step": 39002 + }, + { + "epoch": 5.11, + "learning_rate": 6.101930594678795e-06, + "loss": 0.0081, + "step": 39003 + }, + { + "epoch": 5.11, + "learning_rate": 6.101442225610814e-06, + "loss": 0.0078, + "step": 39004 + }, + { + "epoch": 5.11, + "learning_rate": 6.100953867507321e-06, + "loss": 0.0084, + "step": 39005 + }, + { + "epoch": 5.11, + "learning_rate": 6.100465520369685e-06, + "loss": 0.008, + "step": 39006 + }, + { + "epoch": 5.11, + "learning_rate": 6.0999771841992826e-06, + "loss": 0.0072, + "step": 39007 + }, + { + "epoch": 5.11, + "learning_rate": 6.099488858997489e-06, + "loss": 0.0075, + "step": 39008 + }, + { + "epoch": 5.11, + "learning_rate": 6.099000544765672e-06, + "loss": 0.0077, + "step": 39009 + }, + { + "epoch": 5.11, + "learning_rate": 6.098512241505211e-06, + "loss": 0.0075, + "step": 39010 + }, + { + "epoch": 5.11, + "learning_rate": 6.098023949217475e-06, + "loss": 0.0076, + "step": 39011 + }, + { + "epoch": 5.11, + "learning_rate": 6.09753566790384e-06, + "loss": 0.008, + "step": 39012 + }, + { + "epoch": 5.11, + "learning_rate": 6.097047397565672e-06, + "loss": 0.0076, + "step": 39013 + }, + { + "epoch": 5.11, + "learning_rate": 6.096559138204358e-06, + "loss": 0.0068, + "step": 39014 + }, + { + "epoch": 5.11, + "learning_rate": 6.096070889821259e-06, + "loss": 0.0116, + "step": 39015 + }, + { + "epoch": 5.11, + "learning_rate": 6.095582652417757e-06, + "loss": 0.0062, + "step": 39016 + }, + { + "epoch": 5.11, + "learning_rate": 6.095094425995218e-06, + "loss": 0.0073, + "step": 39017 + }, + { + "epoch": 5.11, + "learning_rate": 6.0946062105550195e-06, + "loss": 0.0072, + "step": 39018 + }, + { + "epoch": 5.11, + "learning_rate": 6.094118006098532e-06, + "loss": 0.0082, + "step": 39019 + }, + { + "epoch": 5.11, + "learning_rate": 6.093629812627131e-06, + "loss": 0.01, + "step": 39020 + }, + { + "epoch": 5.11, + "learning_rate": 6.093141630142186e-06, + "loss": 0.0062, + "step": 39021 + }, + { + "epoch": 5.11, + "learning_rate": 6.092653458645073e-06, + "loss": 0.0067, + "step": 39022 + }, + { + "epoch": 5.11, + "learning_rate": 6.092165298137165e-06, + "loss": 0.0056, + "step": 39023 + }, + { + "epoch": 5.11, + "learning_rate": 6.09167714861983e-06, + "loss": 0.0074, + "step": 39024 + }, + { + "epoch": 5.11, + "learning_rate": 6.091189010094447e-06, + "loss": 0.01, + "step": 39025 + }, + { + "epoch": 5.11, + "learning_rate": 6.0907008825623905e-06, + "loss": 0.0075, + "step": 39026 + }, + { + "epoch": 5.11, + "learning_rate": 6.090212766025025e-06, + "loss": 0.0086, + "step": 39027 + }, + { + "epoch": 5.11, + "learning_rate": 6.089724660483729e-06, + "loss": 0.0081, + "step": 39028 + }, + { + "epoch": 5.11, + "learning_rate": 6.089236565939877e-06, + "loss": 0.0084, + "step": 39029 + }, + { + "epoch": 5.11, + "learning_rate": 6.088748482394835e-06, + "loss": 0.0085, + "step": 39030 + }, + { + "epoch": 5.11, + "learning_rate": 6.0882604098499845e-06, + "loss": 0.0086, + "step": 39031 + }, + { + "epoch": 5.11, + "learning_rate": 6.087772348306688e-06, + "loss": 0.0109, + "step": 39032 + }, + { + "epoch": 5.11, + "learning_rate": 6.087284297766327e-06, + "loss": 0.0069, + "step": 39033 + }, + { + "epoch": 5.11, + "learning_rate": 6.086796258230269e-06, + "loss": 0.008, + "step": 39034 + }, + { + "epoch": 5.11, + "learning_rate": 6.086308229699885e-06, + "loss": 0.0066, + "step": 39035 + }, + { + "epoch": 5.11, + "learning_rate": 6.085820212176554e-06, + "loss": 0.0069, + "step": 39036 + }, + { + "epoch": 5.11, + "learning_rate": 6.085332205661646e-06, + "loss": 0.0101, + "step": 39037 + }, + { + "epoch": 5.11, + "learning_rate": 6.084844210156531e-06, + "loss": 0.0066, + "step": 39038 + }, + { + "epoch": 5.11, + "learning_rate": 6.084356225662587e-06, + "loss": 0.0056, + "step": 39039 + }, + { + "epoch": 5.11, + "learning_rate": 6.083868252181178e-06, + "loss": 0.0077, + "step": 39040 + }, + { + "epoch": 5.11, + "learning_rate": 6.083380289713683e-06, + "loss": 0.0074, + "step": 39041 + }, + { + "epoch": 5.11, + "learning_rate": 6.082892338261472e-06, + "loss": 0.0082, + "step": 39042 + }, + { + "epoch": 5.11, + "learning_rate": 6.082404397825917e-06, + "loss": 0.0066, + "step": 39043 + }, + { + "epoch": 5.11, + "learning_rate": 6.081916468408393e-06, + "loss": 0.0062, + "step": 39044 + }, + { + "epoch": 5.11, + "learning_rate": 6.081428550010269e-06, + "loss": 0.0044, + "step": 39045 + }, + { + "epoch": 5.11, + "learning_rate": 6.080940642632915e-06, + "loss": 0.0103, + "step": 39046 + }, + { + "epoch": 5.11, + "learning_rate": 6.0804527462777116e-06, + "loss": 0.0077, + "step": 39047 + }, + { + "epoch": 5.12, + "learning_rate": 6.079964860946023e-06, + "loss": 0.0065, + "step": 39048 + }, + { + "epoch": 5.12, + "learning_rate": 6.079476986639227e-06, + "loss": 0.0074, + "step": 39049 + }, + { + "epoch": 5.12, + "learning_rate": 6.078989123358694e-06, + "loss": 0.0078, + "step": 39050 + }, + { + "epoch": 5.12, + "learning_rate": 6.078501271105792e-06, + "loss": 0.0084, + "step": 39051 + }, + { + "epoch": 5.12, + "learning_rate": 6.0780134298819e-06, + "loss": 0.0055, + "step": 39052 + }, + { + "epoch": 5.12, + "learning_rate": 6.077525599688383e-06, + "loss": 0.0061, + "step": 39053 + }, + { + "epoch": 5.12, + "learning_rate": 6.07703778052662e-06, + "loss": 0.0093, + "step": 39054 + }, + { + "epoch": 5.12, + "learning_rate": 6.076549972397976e-06, + "loss": 0.0088, + "step": 39055 + }, + { + "epoch": 5.12, + "learning_rate": 6.07606217530383e-06, + "loss": 0.006, + "step": 39056 + }, + { + "epoch": 5.12, + "learning_rate": 6.075574389245542e-06, + "loss": 0.0097, + "step": 39057 + }, + { + "epoch": 5.12, + "learning_rate": 6.075086614224499e-06, + "loss": 0.0078, + "step": 39058 + }, + { + "epoch": 5.12, + "learning_rate": 6.0745988502420635e-06, + "loss": 0.0082, + "step": 39059 + }, + { + "epoch": 5.12, + "learning_rate": 6.074111097299612e-06, + "loss": 0.0078, + "step": 39060 + }, + { + "epoch": 5.12, + "learning_rate": 6.073623355398513e-06, + "loss": 0.0063, + "step": 39061 + }, + { + "epoch": 5.12, + "learning_rate": 6.0731356245401385e-06, + "loss": 0.0056, + "step": 39062 + }, + { + "epoch": 5.12, + "learning_rate": 6.072647904725861e-06, + "loss": 0.0075, + "step": 39063 + }, + { + "epoch": 5.12, + "learning_rate": 6.072160195957052e-06, + "loss": 0.0114, + "step": 39064 + }, + { + "epoch": 5.12, + "learning_rate": 6.071672498235086e-06, + "loss": 0.008, + "step": 39065 + }, + { + "epoch": 5.12, + "learning_rate": 6.071184811561328e-06, + "loss": 0.0064, + "step": 39066 + }, + { + "epoch": 5.12, + "learning_rate": 6.070697135937157e-06, + "loss": 0.0082, + "step": 39067 + }, + { + "epoch": 5.12, + "learning_rate": 6.070209471363936e-06, + "loss": 0.007, + "step": 39068 + }, + { + "epoch": 5.12, + "learning_rate": 6.069721817843044e-06, + "loss": 0.0089, + "step": 39069 + }, + { + "epoch": 5.12, + "learning_rate": 6.069234175375853e-06, + "loss": 0.0074, + "step": 39070 + }, + { + "epoch": 5.12, + "learning_rate": 6.068746543963729e-06, + "loss": 0.0064, + "step": 39071 + }, + { + "epoch": 5.12, + "learning_rate": 6.0682589236080445e-06, + "loss": 0.0089, + "step": 39072 + }, + { + "epoch": 5.12, + "learning_rate": 6.0677713143101755e-06, + "loss": 0.008, + "step": 39073 + }, + { + "epoch": 5.12, + "learning_rate": 6.067283716071488e-06, + "loss": 0.0073, + "step": 39074 + }, + { + "epoch": 5.12, + "learning_rate": 6.066796128893359e-06, + "loss": 0.0061, + "step": 39075 + }, + { + "epoch": 5.12, + "learning_rate": 6.066308552777154e-06, + "loss": 0.0064, + "step": 39076 + }, + { + "epoch": 5.12, + "learning_rate": 6.065820987724247e-06, + "loss": 0.008, + "step": 39077 + }, + { + "epoch": 5.12, + "learning_rate": 6.06533343373601e-06, + "loss": 0.0075, + "step": 39078 + }, + { + "epoch": 5.12, + "learning_rate": 6.064845890813808e-06, + "loss": 0.0057, + "step": 39079 + }, + { + "epoch": 5.12, + "learning_rate": 6.064358358959021e-06, + "loss": 0.0136, + "step": 39080 + }, + { + "epoch": 5.12, + "learning_rate": 6.063870838173018e-06, + "loss": 0.007, + "step": 39081 + }, + { + "epoch": 5.12, + "learning_rate": 6.0633833284571665e-06, + "loss": 0.008, + "step": 39082 + }, + { + "epoch": 5.12, + "learning_rate": 6.0628958298128425e-06, + "loss": 0.0081, + "step": 39083 + }, + { + "epoch": 5.12, + "learning_rate": 6.062408342241413e-06, + "loss": 0.0093, + "step": 39084 + }, + { + "epoch": 5.12, + "learning_rate": 6.061920865744251e-06, + "loss": 0.0098, + "step": 39085 + }, + { + "epoch": 5.12, + "learning_rate": 6.061433400322725e-06, + "loss": 0.0079, + "step": 39086 + }, + { + "epoch": 5.12, + "learning_rate": 6.060945945978207e-06, + "loss": 0.0058, + "step": 39087 + }, + { + "epoch": 5.12, + "learning_rate": 6.060458502712072e-06, + "loss": 0.0067, + "step": 39088 + }, + { + "epoch": 5.12, + "learning_rate": 6.0599710705256855e-06, + "loss": 0.0094, + "step": 39089 + }, + { + "epoch": 5.12, + "learning_rate": 6.059483649420419e-06, + "loss": 0.0103, + "step": 39090 + }, + { + "epoch": 5.12, + "learning_rate": 6.058996239397648e-06, + "loss": 0.0072, + "step": 39091 + }, + { + "epoch": 5.12, + "learning_rate": 6.058508840458739e-06, + "loss": 0.0059, + "step": 39092 + }, + { + "epoch": 5.12, + "learning_rate": 6.058021452605063e-06, + "loss": 0.0066, + "step": 39093 + }, + { + "epoch": 5.12, + "learning_rate": 6.057534075837994e-06, + "loss": 0.0082, + "step": 39094 + }, + { + "epoch": 5.12, + "learning_rate": 6.057046710158898e-06, + "loss": 0.0075, + "step": 39095 + }, + { + "epoch": 5.12, + "learning_rate": 6.056559355569152e-06, + "loss": 0.007, + "step": 39096 + }, + { + "epoch": 5.12, + "learning_rate": 6.05607201207012e-06, + "loss": 0.0102, + "step": 39097 + }, + { + "epoch": 5.12, + "learning_rate": 6.055584679663175e-06, + "loss": 0.0084, + "step": 39098 + }, + { + "epoch": 5.12, + "learning_rate": 6.05509735834969e-06, + "loss": 0.0076, + "step": 39099 + }, + { + "epoch": 5.12, + "learning_rate": 6.054610048131033e-06, + "loss": 0.0074, + "step": 39100 + }, + { + "epoch": 5.12, + "learning_rate": 6.054122749008571e-06, + "loss": 0.0069, + "step": 39101 + }, + { + "epoch": 5.12, + "learning_rate": 6.0536354609836825e-06, + "loss": 0.006, + "step": 39102 + }, + { + "epoch": 5.12, + "learning_rate": 6.053148184057733e-06, + "loss": 0.0055, + "step": 39103 + }, + { + "epoch": 5.12, + "learning_rate": 6.052660918232096e-06, + "loss": 0.0066, + "step": 39104 + }, + { + "epoch": 5.12, + "learning_rate": 6.0521736635081386e-06, + "loss": 0.0087, + "step": 39105 + }, + { + "epoch": 5.12, + "learning_rate": 6.051686419887235e-06, + "loss": 0.0068, + "step": 39106 + }, + { + "epoch": 5.12, + "learning_rate": 6.051199187370748e-06, + "loss": 0.0088, + "step": 39107 + }, + { + "epoch": 5.12, + "learning_rate": 6.050711965960056e-06, + "loss": 0.0077, + "step": 39108 + }, + { + "epoch": 5.12, + "learning_rate": 6.050224755656529e-06, + "loss": 0.0065, + "step": 39109 + }, + { + "epoch": 5.12, + "learning_rate": 6.04973755646153e-06, + "loss": 0.0101, + "step": 39110 + }, + { + "epoch": 5.12, + "learning_rate": 6.049250368376437e-06, + "loss": 0.0064, + "step": 39111 + }, + { + "epoch": 5.12, + "learning_rate": 6.048763191402612e-06, + "loss": 0.0092, + "step": 39112 + }, + { + "epoch": 5.12, + "learning_rate": 6.048276025541433e-06, + "loss": 0.0088, + "step": 39113 + }, + { + "epoch": 5.12, + "learning_rate": 6.047788870794267e-06, + "loss": 0.006, + "step": 39114 + }, + { + "epoch": 5.12, + "learning_rate": 6.047301727162487e-06, + "loss": 0.0083, + "step": 39115 + }, + { + "epoch": 5.12, + "learning_rate": 6.046814594647458e-06, + "loss": 0.0076, + "step": 39116 + }, + { + "epoch": 5.12, + "learning_rate": 6.046327473250554e-06, + "loss": 0.0086, + "step": 39117 + }, + { + "epoch": 5.12, + "learning_rate": 6.045840362973141e-06, + "loss": 0.0086, + "step": 39118 + }, + { + "epoch": 5.12, + "learning_rate": 6.045353263816595e-06, + "loss": 0.0093, + "step": 39119 + }, + { + "epoch": 5.12, + "learning_rate": 6.044866175782279e-06, + "loss": 0.0098, + "step": 39120 + }, + { + "epoch": 5.12, + "learning_rate": 6.044379098871569e-06, + "loss": 0.0078, + "step": 39121 + }, + { + "epoch": 5.12, + "learning_rate": 6.04389203308583e-06, + "loss": 0.0069, + "step": 39122 + }, + { + "epoch": 5.12, + "learning_rate": 6.043404978426432e-06, + "loss": 0.0087, + "step": 39123 + }, + { + "epoch": 5.12, + "learning_rate": 6.042917934894747e-06, + "loss": 0.0122, + "step": 39124 + }, + { + "epoch": 5.13, + "learning_rate": 6.042430902492149e-06, + "loss": 0.0082, + "step": 39125 + }, + { + "epoch": 5.13, + "learning_rate": 6.041943881220001e-06, + "loss": 0.008, + "step": 39126 + }, + { + "epoch": 5.13, + "learning_rate": 6.041456871079676e-06, + "loss": 0.0047, + "step": 39127 + }, + { + "epoch": 5.13, + "learning_rate": 6.040969872072542e-06, + "loss": 0.009, + "step": 39128 + }, + { + "epoch": 5.13, + "learning_rate": 6.040482884199968e-06, + "loss": 0.0083, + "step": 39129 + }, + { + "epoch": 5.13, + "learning_rate": 6.039995907463327e-06, + "loss": 0.0078, + "step": 39130 + }, + { + "epoch": 5.13, + "learning_rate": 6.039508941863986e-06, + "loss": 0.0057, + "step": 39131 + }, + { + "epoch": 5.13, + "learning_rate": 6.039021987403318e-06, + "loss": 0.0102, + "step": 39132 + }, + { + "epoch": 5.13, + "learning_rate": 6.038535044082685e-06, + "loss": 0.0069, + "step": 39133 + }, + { + "epoch": 5.13, + "learning_rate": 6.038048111903461e-06, + "loss": 0.0097, + "step": 39134 + }, + { + "epoch": 5.13, + "learning_rate": 6.037561190867018e-06, + "loss": 0.0072, + "step": 39135 + }, + { + "epoch": 5.13, + "learning_rate": 6.037074280974723e-06, + "loss": 0.0086, + "step": 39136 + }, + { + "epoch": 5.13, + "learning_rate": 6.036587382227945e-06, + "loss": 0.0066, + "step": 39137 + }, + { + "epoch": 5.13, + "learning_rate": 6.036100494628056e-06, + "loss": 0.0076, + "step": 39138 + }, + { + "epoch": 5.13, + "learning_rate": 6.035613618176421e-06, + "loss": 0.0079, + "step": 39139 + }, + { + "epoch": 5.13, + "learning_rate": 6.035126752874414e-06, + "loss": 0.0081, + "step": 39140 + }, + { + "epoch": 5.13, + "learning_rate": 6.034639898723399e-06, + "loss": 0.0086, + "step": 39141 + }, + { + "epoch": 5.13, + "learning_rate": 6.0341530557247504e-06, + "loss": 0.0102, + "step": 39142 + }, + { + "epoch": 5.13, + "learning_rate": 6.033666223879834e-06, + "loss": 0.0093, + "step": 39143 + }, + { + "epoch": 5.13, + "learning_rate": 6.033179403190017e-06, + "loss": 0.008, + "step": 39144 + }, + { + "epoch": 5.13, + "learning_rate": 6.0326925936566746e-06, + "loss": 0.0089, + "step": 39145 + }, + { + "epoch": 5.13, + "learning_rate": 6.032205795281174e-06, + "loss": 0.0097, + "step": 39146 + }, + { + "epoch": 5.13, + "learning_rate": 6.031719008064882e-06, + "loss": 0.007, + "step": 39147 + }, + { + "epoch": 5.13, + "learning_rate": 6.031232232009172e-06, + "loss": 0.0092, + "step": 39148 + }, + { + "epoch": 5.13, + "learning_rate": 6.030745467115406e-06, + "loss": 0.0085, + "step": 39149 + }, + { + "epoch": 5.13, + "learning_rate": 6.03025871338496e-06, + "loss": 0.0048, + "step": 39150 + }, + { + "epoch": 5.13, + "learning_rate": 6.029771970819198e-06, + "loss": 0.0078, + "step": 39151 + }, + { + "epoch": 5.13, + "learning_rate": 6.029285239419491e-06, + "loss": 0.0057, + "step": 39152 + }, + { + "epoch": 5.13, + "learning_rate": 6.028798519187209e-06, + "loss": 0.0101, + "step": 39153 + }, + { + "epoch": 5.13, + "learning_rate": 6.028311810123719e-06, + "loss": 0.0072, + "step": 39154 + }, + { + "epoch": 5.13, + "learning_rate": 6.027825112230388e-06, + "loss": 0.0077, + "step": 39155 + }, + { + "epoch": 5.13, + "learning_rate": 6.027338425508591e-06, + "loss": 0.0072, + "step": 39156 + }, + { + "epoch": 5.13, + "learning_rate": 6.02685174995969e-06, + "loss": 0.0093, + "step": 39157 + }, + { + "epoch": 5.13, + "learning_rate": 6.026365085585058e-06, + "loss": 0.0091, + "step": 39158 + }, + { + "epoch": 5.13, + "learning_rate": 6.025878432386066e-06, + "loss": 0.0091, + "step": 39159 + }, + { + "epoch": 5.13, + "learning_rate": 6.025391790364074e-06, + "loss": 0.0072, + "step": 39160 + }, + { + "epoch": 5.13, + "learning_rate": 6.024905159520461e-06, + "loss": 0.0085, + "step": 39161 + }, + { + "epoch": 5.13, + "learning_rate": 6.0244185398565865e-06, + "loss": 0.0086, + "step": 39162 + }, + { + "epoch": 5.13, + "learning_rate": 6.023931931373828e-06, + "loss": 0.0076, + "step": 39163 + }, + { + "epoch": 5.13, + "learning_rate": 6.023445334073545e-06, + "loss": 0.0096, + "step": 39164 + }, + { + "epoch": 5.13, + "learning_rate": 6.022958747957111e-06, + "loss": 0.0079, + "step": 39165 + }, + { + "epoch": 5.13, + "learning_rate": 6.0224721730258905e-06, + "loss": 0.0108, + "step": 39166 + }, + { + "epoch": 5.13, + "learning_rate": 6.02198560928126e-06, + "loss": 0.0075, + "step": 39167 + }, + { + "epoch": 5.13, + "learning_rate": 6.02149905672458e-06, + "loss": 0.0073, + "step": 39168 + }, + { + "epoch": 5.13, + "learning_rate": 6.021012515357226e-06, + "loss": 0.0065, + "step": 39169 + }, + { + "epoch": 5.13, + "learning_rate": 6.020525985180559e-06, + "loss": 0.0073, + "step": 39170 + }, + { + "epoch": 5.13, + "learning_rate": 6.020039466195954e-06, + "loss": 0.0081, + "step": 39171 + }, + { + "epoch": 5.13, + "learning_rate": 6.019552958404773e-06, + "loss": 0.0083, + "step": 39172 + }, + { + "epoch": 5.13, + "learning_rate": 6.019066461808387e-06, + "loss": 0.0085, + "step": 39173 + }, + { + "epoch": 5.13, + "learning_rate": 6.018579976408168e-06, + "loss": 0.0078, + "step": 39174 + }, + { + "epoch": 5.13, + "learning_rate": 6.018093502205478e-06, + "loss": 0.0093, + "step": 39175 + }, + { + "epoch": 5.13, + "learning_rate": 6.017607039201689e-06, + "loss": 0.0066, + "step": 39176 + }, + { + "epoch": 5.13, + "learning_rate": 6.017120587398166e-06, + "loss": 0.0085, + "step": 39177 + }, + { + "epoch": 5.13, + "learning_rate": 6.0166341467962805e-06, + "loss": 0.0095, + "step": 39178 + }, + { + "epoch": 5.13, + "learning_rate": 6.016147717397399e-06, + "loss": 0.006, + "step": 39179 + }, + { + "epoch": 5.13, + "learning_rate": 6.0156612992028925e-06, + "loss": 0.0065, + "step": 39180 + }, + { + "epoch": 5.13, + "learning_rate": 6.015174892214124e-06, + "loss": 0.0052, + "step": 39181 + }, + { + "epoch": 5.13, + "learning_rate": 6.014688496432466e-06, + "loss": 0.0056, + "step": 39182 + }, + { + "epoch": 5.13, + "learning_rate": 6.0142021118592816e-06, + "loss": 0.0087, + "step": 39183 + }, + { + "epoch": 5.13, + "learning_rate": 6.013715738495944e-06, + "loss": 0.0072, + "step": 39184 + }, + { + "epoch": 5.13, + "learning_rate": 6.013229376343818e-06, + "loss": 0.006, + "step": 39185 + }, + { + "epoch": 5.13, + "learning_rate": 6.012743025404273e-06, + "loss": 0.0078, + "step": 39186 + }, + { + "epoch": 5.13, + "learning_rate": 6.012256685678674e-06, + "loss": 0.008, + "step": 39187 + }, + { + "epoch": 5.13, + "learning_rate": 6.011770357168388e-06, + "loss": 0.008, + "step": 39188 + }, + { + "epoch": 5.13, + "learning_rate": 6.0112840398747874e-06, + "loss": 0.0085, + "step": 39189 + }, + { + "epoch": 5.13, + "learning_rate": 6.010797733799241e-06, + "loss": 0.007, + "step": 39190 + }, + { + "epoch": 5.13, + "learning_rate": 6.010311438943112e-06, + "loss": 0.0086, + "step": 39191 + }, + { + "epoch": 5.13, + "learning_rate": 6.009825155307771e-06, + "loss": 0.007, + "step": 39192 + }, + { + "epoch": 5.13, + "learning_rate": 6.009338882894582e-06, + "loss": 0.0079, + "step": 39193 + }, + { + "epoch": 5.13, + "learning_rate": 6.0088526217049145e-06, + "loss": 0.0063, + "step": 39194 + }, + { + "epoch": 5.13, + "learning_rate": 6.008366371740138e-06, + "loss": 0.0068, + "step": 39195 + }, + { + "epoch": 5.13, + "learning_rate": 6.007880133001618e-06, + "loss": 0.0079, + "step": 39196 + }, + { + "epoch": 5.13, + "learning_rate": 6.007393905490724e-06, + "loss": 0.0084, + "step": 39197 + }, + { + "epoch": 5.13, + "learning_rate": 6.006907689208821e-06, + "loss": 0.0079, + "step": 39198 + }, + { + "epoch": 5.13, + "learning_rate": 6.006421484157273e-06, + "loss": 0.0093, + "step": 39199 + }, + { + "epoch": 5.13, + "learning_rate": 6.005935290337458e-06, + "loss": 0.0081, + "step": 39200 + }, + { + "epoch": 5.14, + "learning_rate": 6.005449107750733e-06, + "loss": 0.0091, + "step": 39201 + }, + { + "epoch": 5.14, + "learning_rate": 6.004962936398473e-06, + "loss": 0.0088, + "step": 39202 + }, + { + "epoch": 5.14, + "learning_rate": 6.004476776282042e-06, + "loss": 0.0061, + "step": 39203 + }, + { + "epoch": 5.14, + "learning_rate": 6.003990627402804e-06, + "loss": 0.0069, + "step": 39204 + }, + { + "epoch": 5.14, + "learning_rate": 6.0035044897621355e-06, + "loss": 0.0074, + "step": 39205 + }, + { + "epoch": 5.14, + "learning_rate": 6.003018363361392e-06, + "loss": 0.0083, + "step": 39206 + }, + { + "epoch": 5.14, + "learning_rate": 6.002532248201951e-06, + "loss": 0.009, + "step": 39207 + }, + { + "epoch": 5.14, + "learning_rate": 6.002046144285172e-06, + "loss": 0.0081, + "step": 39208 + }, + { + "epoch": 5.14, + "learning_rate": 6.0015600516124255e-06, + "loss": 0.0072, + "step": 39209 + }, + { + "epoch": 5.14, + "learning_rate": 6.001073970185076e-06, + "loss": 0.0071, + "step": 39210 + }, + { + "epoch": 5.14, + "learning_rate": 6.000587900004497e-06, + "loss": 0.0067, + "step": 39211 + }, + { + "epoch": 5.14, + "learning_rate": 6.00010184107205e-06, + "loss": 0.0084, + "step": 39212 + }, + { + "epoch": 5.14, + "learning_rate": 5.999615793389107e-06, + "loss": 0.0071, + "step": 39213 + }, + { + "epoch": 5.14, + "learning_rate": 5.999129756957028e-06, + "loss": 0.0097, + "step": 39214 + }, + { + "epoch": 5.14, + "learning_rate": 5.998643731777183e-06, + "loss": 0.0058, + "step": 39215 + }, + { + "epoch": 5.14, + "learning_rate": 5.9981577178509406e-06, + "loss": 0.0041, + "step": 39216 + }, + { + "epoch": 5.14, + "learning_rate": 5.997671715179667e-06, + "loss": 0.0084, + "step": 39217 + }, + { + "epoch": 5.14, + "learning_rate": 5.997185723764729e-06, + "loss": 0.0081, + "step": 39218 + }, + { + "epoch": 5.14, + "learning_rate": 5.996699743607491e-06, + "loss": 0.005, + "step": 39219 + }, + { + "epoch": 5.14, + "learning_rate": 5.996213774709325e-06, + "loss": 0.0096, + "step": 39220 + }, + { + "epoch": 5.14, + "learning_rate": 5.995727817071589e-06, + "loss": 0.0091, + "step": 39221 + }, + { + "epoch": 5.14, + "learning_rate": 5.995241870695658e-06, + "loss": 0.0077, + "step": 39222 + }, + { + "epoch": 5.14, + "learning_rate": 5.994755935582896e-06, + "loss": 0.0091, + "step": 39223 + }, + { + "epoch": 5.14, + "learning_rate": 5.994270011734672e-06, + "loss": 0.0071, + "step": 39224 + }, + { + "epoch": 5.14, + "learning_rate": 5.9937840991523475e-06, + "loss": 0.0082, + "step": 39225 + }, + { + "epoch": 5.14, + "learning_rate": 5.993298197837294e-06, + "loss": 0.0058, + "step": 39226 + }, + { + "epoch": 5.14, + "learning_rate": 5.992812307790874e-06, + "loss": 0.0095, + "step": 39227 + }, + { + "epoch": 5.14, + "learning_rate": 5.992326429014459e-06, + "loss": 0.0091, + "step": 39228 + }, + { + "epoch": 5.14, + "learning_rate": 5.99184056150941e-06, + "loss": 0.0091, + "step": 39229 + }, + { + "epoch": 5.14, + "learning_rate": 5.991354705277095e-06, + "loss": 0.0061, + "step": 39230 + }, + { + "epoch": 5.14, + "learning_rate": 5.990868860318884e-06, + "loss": 0.0073, + "step": 39231 + }, + { + "epoch": 5.14, + "learning_rate": 5.990383026636137e-06, + "loss": 0.0072, + "step": 39232 + }, + { + "epoch": 5.14, + "learning_rate": 5.989897204230226e-06, + "loss": 0.008, + "step": 39233 + }, + { + "epoch": 5.14, + "learning_rate": 5.989411393102518e-06, + "loss": 0.0095, + "step": 39234 + }, + { + "epoch": 5.14, + "learning_rate": 5.9889255932543745e-06, + "loss": 0.0079, + "step": 39235 + }, + { + "epoch": 5.14, + "learning_rate": 5.9884398046871674e-06, + "loss": 0.0096, + "step": 39236 + }, + { + "epoch": 5.14, + "learning_rate": 5.987954027402256e-06, + "loss": 0.0082, + "step": 39237 + }, + { + "epoch": 5.14, + "learning_rate": 5.987468261401011e-06, + "loss": 0.0072, + "step": 39238 + }, + { + "epoch": 5.14, + "learning_rate": 5.986982506684799e-06, + "loss": 0.006, + "step": 39239 + }, + { + "epoch": 5.14, + "learning_rate": 5.986496763254983e-06, + "loss": 0.0056, + "step": 39240 + }, + { + "epoch": 5.14, + "learning_rate": 5.986011031112935e-06, + "loss": 0.0094, + "step": 39241 + }, + { + "epoch": 5.14, + "learning_rate": 5.985525310260014e-06, + "loss": 0.0085, + "step": 39242 + }, + { + "epoch": 5.14, + "learning_rate": 5.985039600697587e-06, + "loss": 0.0076, + "step": 39243 + }, + { + "epoch": 5.14, + "learning_rate": 5.984553902427024e-06, + "loss": 0.0068, + "step": 39244 + }, + { + "epoch": 5.14, + "learning_rate": 5.984068215449692e-06, + "loss": 0.0084, + "step": 39245 + }, + { + "epoch": 5.14, + "learning_rate": 5.983582539766952e-06, + "loss": 0.0094, + "step": 39246 + }, + { + "epoch": 5.14, + "learning_rate": 5.983096875380174e-06, + "loss": 0.0063, + "step": 39247 + }, + { + "epoch": 5.14, + "learning_rate": 5.9826112222907216e-06, + "loss": 0.0071, + "step": 39248 + }, + { + "epoch": 5.14, + "learning_rate": 5.982125580499962e-06, + "loss": 0.0069, + "step": 39249 + }, + { + "epoch": 5.14, + "learning_rate": 5.981639950009258e-06, + "loss": 0.0069, + "step": 39250 + }, + { + "epoch": 5.14, + "learning_rate": 5.98115433081998e-06, + "loss": 0.0055, + "step": 39251 + }, + { + "epoch": 5.14, + "learning_rate": 5.98066872293349e-06, + "loss": 0.0117, + "step": 39252 + }, + { + "epoch": 5.14, + "learning_rate": 5.980183126351154e-06, + "loss": 0.0084, + "step": 39253 + }, + { + "epoch": 5.14, + "learning_rate": 5.979697541074338e-06, + "loss": 0.006, + "step": 39254 + }, + { + "epoch": 5.14, + "learning_rate": 5.979211967104412e-06, + "loss": 0.0077, + "step": 39255 + }, + { + "epoch": 5.14, + "learning_rate": 5.978726404442736e-06, + "loss": 0.0056, + "step": 39256 + }, + { + "epoch": 5.14, + "learning_rate": 5.97824085309068e-06, + "loss": 0.0057, + "step": 39257 + }, + { + "epoch": 5.14, + "learning_rate": 5.977755313049605e-06, + "loss": 0.0053, + "step": 39258 + }, + { + "epoch": 5.14, + "learning_rate": 5.977269784320881e-06, + "loss": 0.0085, + "step": 39259 + }, + { + "epoch": 5.14, + "learning_rate": 5.976784266905872e-06, + "loss": 0.0057, + "step": 39260 + }, + { + "epoch": 5.14, + "learning_rate": 5.9762987608059404e-06, + "loss": 0.0121, + "step": 39261 + }, + { + "epoch": 5.14, + "learning_rate": 5.975813266022457e-06, + "loss": 0.0066, + "step": 39262 + }, + { + "epoch": 5.14, + "learning_rate": 5.975327782556782e-06, + "loss": 0.0085, + "step": 39263 + }, + { + "epoch": 5.14, + "learning_rate": 5.974842310410287e-06, + "loss": 0.0086, + "step": 39264 + }, + { + "epoch": 5.14, + "learning_rate": 5.974356849584326e-06, + "loss": 0.0093, + "step": 39265 + }, + { + "epoch": 5.14, + "learning_rate": 5.973871400080278e-06, + "loss": 0.0063, + "step": 39266 + }, + { + "epoch": 5.14, + "learning_rate": 5.9733859618995e-06, + "loss": 0.0053, + "step": 39267 + }, + { + "epoch": 5.14, + "learning_rate": 5.972900535043362e-06, + "loss": 0.0068, + "step": 39268 + }, + { + "epoch": 5.14, + "learning_rate": 5.972415119513227e-06, + "loss": 0.0088, + "step": 39269 + }, + { + "epoch": 5.14, + "learning_rate": 5.971929715310459e-06, + "loss": 0.0089, + "step": 39270 + }, + { + "epoch": 5.14, + "learning_rate": 5.9714443224364225e-06, + "loss": 0.0074, + "step": 39271 + }, + { + "epoch": 5.14, + "learning_rate": 5.970958940892487e-06, + "loss": 0.0064, + "step": 39272 + }, + { + "epoch": 5.14, + "learning_rate": 5.970473570680014e-06, + "loss": 0.0065, + "step": 39273 + }, + { + "epoch": 5.14, + "learning_rate": 5.9699882118003675e-06, + "loss": 0.0096, + "step": 39274 + }, + { + "epoch": 5.14, + "learning_rate": 5.9695028642549124e-06, + "loss": 0.0082, + "step": 39275 + }, + { + "epoch": 5.14, + "learning_rate": 5.969017528045022e-06, + "loss": 0.0074, + "step": 39276 + }, + { + "epoch": 5.15, + "learning_rate": 5.968532203172052e-06, + "loss": 0.0065, + "step": 39277 + }, + { + "epoch": 5.15, + "learning_rate": 5.968046889637372e-06, + "loss": 0.0084, + "step": 39278 + }, + { + "epoch": 5.15, + "learning_rate": 5.967561587442344e-06, + "loss": 0.0045, + "step": 39279 + }, + { + "epoch": 5.15, + "learning_rate": 5.967076296588334e-06, + "loss": 0.0091, + "step": 39280 + }, + { + "epoch": 5.15, + "learning_rate": 5.9665910170767095e-06, + "loss": 0.0066, + "step": 39281 + }, + { + "epoch": 5.15, + "learning_rate": 5.96610574890883e-06, + "loss": 0.0092, + "step": 39282 + }, + { + "epoch": 5.15, + "learning_rate": 5.965620492086067e-06, + "loss": 0.0052, + "step": 39283 + }, + { + "epoch": 5.15, + "learning_rate": 5.965135246609779e-06, + "loss": 0.0118, + "step": 39284 + }, + { + "epoch": 5.15, + "learning_rate": 5.964650012481334e-06, + "loss": 0.008, + "step": 39285 + }, + { + "epoch": 5.15, + "learning_rate": 5.964164789702093e-06, + "loss": 0.0055, + "step": 39286 + }, + { + "epoch": 5.15, + "learning_rate": 5.963679578273424e-06, + "loss": 0.0072, + "step": 39287 + }, + { + "epoch": 5.15, + "learning_rate": 5.963194378196694e-06, + "loss": 0.007, + "step": 39288 + }, + { + "epoch": 5.15, + "learning_rate": 5.962709189473266e-06, + "loss": 0.0067, + "step": 39289 + }, + { + "epoch": 5.15, + "learning_rate": 5.962224012104501e-06, + "loss": 0.0082, + "step": 39290 + }, + { + "epoch": 5.15, + "learning_rate": 5.961738846091767e-06, + "loss": 0.0069, + "step": 39291 + }, + { + "epoch": 5.15, + "learning_rate": 5.961253691436426e-06, + "loss": 0.0069, + "step": 39292 + }, + { + "epoch": 5.15, + "learning_rate": 5.960768548139847e-06, + "loss": 0.0081, + "step": 39293 + }, + { + "epoch": 5.15, + "learning_rate": 5.960283416203388e-06, + "loss": 0.0084, + "step": 39294 + }, + { + "epoch": 5.15, + "learning_rate": 5.959798295628418e-06, + "loss": 0.0067, + "step": 39295 + }, + { + "epoch": 5.15, + "learning_rate": 5.9593131864163015e-06, + "loss": 0.0084, + "step": 39296 + }, + { + "epoch": 5.15, + "learning_rate": 5.958828088568397e-06, + "loss": 0.0109, + "step": 39297 + }, + { + "epoch": 5.15, + "learning_rate": 5.958343002086076e-06, + "loss": 0.0102, + "step": 39298 + }, + { + "epoch": 5.15, + "learning_rate": 5.957857926970702e-06, + "loss": 0.0062, + "step": 39299 + }, + { + "epoch": 5.15, + "learning_rate": 5.957372863223634e-06, + "loss": 0.0081, + "step": 39300 + }, + { + "epoch": 5.15, + "learning_rate": 5.956887810846244e-06, + "loss": 0.0075, + "step": 39301 + }, + { + "epoch": 5.15, + "learning_rate": 5.956402769839889e-06, + "loss": 0.0093, + "step": 39302 + }, + { + "epoch": 5.15, + "learning_rate": 5.9559177402059345e-06, + "loss": 0.0114, + "step": 39303 + }, + { + "epoch": 5.15, + "learning_rate": 5.955432721945749e-06, + "loss": 0.0068, + "step": 39304 + }, + { + "epoch": 5.15, + "learning_rate": 5.954947715060691e-06, + "loss": 0.0085, + "step": 39305 + }, + { + "epoch": 5.15, + "learning_rate": 5.95446271955213e-06, + "loss": 0.0081, + "step": 39306 + }, + { + "epoch": 5.15, + "learning_rate": 5.953977735421424e-06, + "loss": 0.0057, + "step": 39307 + }, + { + "epoch": 5.15, + "learning_rate": 5.953492762669939e-06, + "loss": 0.0068, + "step": 39308 + }, + { + "epoch": 5.15, + "learning_rate": 5.953007801299042e-06, + "loss": 0.0101, + "step": 39309 + }, + { + "epoch": 5.15, + "learning_rate": 5.952522851310096e-06, + "loss": 0.0063, + "step": 39310 + }, + { + "epoch": 5.15, + "learning_rate": 5.952037912704462e-06, + "loss": 0.0087, + "step": 39311 + }, + { + "epoch": 5.15, + "learning_rate": 5.951552985483511e-06, + "loss": 0.0071, + "step": 39312 + }, + { + "epoch": 5.15, + "learning_rate": 5.951068069648596e-06, + "loss": 0.0082, + "step": 39313 + }, + { + "epoch": 5.15, + "learning_rate": 5.950583165201091e-06, + "loss": 0.0098, + "step": 39314 + }, + { + "epoch": 5.15, + "learning_rate": 5.95009827214235e-06, + "loss": 0.0069, + "step": 39315 + }, + { + "epoch": 5.15, + "learning_rate": 5.949613390473745e-06, + "loss": 0.0073, + "step": 39316 + }, + { + "epoch": 5.15, + "learning_rate": 5.9491285201966375e-06, + "loss": 0.0093, + "step": 39317 + }, + { + "epoch": 5.15, + "learning_rate": 5.9486436613123874e-06, + "loss": 0.0094, + "step": 39318 + }, + { + "epoch": 5.15, + "learning_rate": 5.948158813822361e-06, + "loss": 0.0052, + "step": 39319 + }, + { + "epoch": 5.15, + "learning_rate": 5.9476739777279255e-06, + "loss": 0.0069, + "step": 39320 + }, + { + "epoch": 5.15, + "learning_rate": 5.947189153030439e-06, + "loss": 0.0072, + "step": 39321 + }, + { + "epoch": 5.15, + "learning_rate": 5.94670433973127e-06, + "loss": 0.007, + "step": 39322 + }, + { + "epoch": 5.15, + "learning_rate": 5.946219537831777e-06, + "loss": 0.0104, + "step": 39323 + }, + { + "epoch": 5.15, + "learning_rate": 5.945734747333324e-06, + "loss": 0.0051, + "step": 39324 + }, + { + "epoch": 5.15, + "learning_rate": 5.945249968237279e-06, + "loss": 0.0046, + "step": 39325 + }, + { + "epoch": 5.15, + "learning_rate": 5.944765200545001e-06, + "loss": 0.0067, + "step": 39326 + }, + { + "epoch": 5.15, + "learning_rate": 5.944280444257857e-06, + "loss": 0.0076, + "step": 39327 + }, + { + "epoch": 5.15, + "learning_rate": 5.943795699377206e-06, + "loss": 0.0053, + "step": 39328 + }, + { + "epoch": 5.15, + "learning_rate": 5.943310965904415e-06, + "loss": 0.0099, + "step": 39329 + }, + { + "epoch": 5.15, + "learning_rate": 5.94282624384084e-06, + "loss": 0.0084, + "step": 39330 + }, + { + "epoch": 5.15, + "learning_rate": 5.942341533187856e-06, + "loss": 0.0093, + "step": 39331 + }, + { + "epoch": 5.15, + "learning_rate": 5.94185683394682e-06, + "loss": 0.0061, + "step": 39332 + }, + { + "epoch": 5.15, + "learning_rate": 5.941372146119096e-06, + "loss": 0.0101, + "step": 39333 + }, + { + "epoch": 5.15, + "learning_rate": 5.940887469706045e-06, + "loss": 0.0075, + "step": 39334 + }, + { + "epoch": 5.15, + "learning_rate": 5.940402804709035e-06, + "loss": 0.0104, + "step": 39335 + }, + { + "epoch": 5.15, + "learning_rate": 5.939918151129422e-06, + "loss": 0.0103, + "step": 39336 + }, + { + "epoch": 5.15, + "learning_rate": 5.9394335089685765e-06, + "loss": 0.0102, + "step": 39337 + }, + { + "epoch": 5.15, + "learning_rate": 5.938948878227855e-06, + "loss": 0.0098, + "step": 39338 + }, + { + "epoch": 5.15, + "learning_rate": 5.938464258908624e-06, + "loss": 0.0088, + "step": 39339 + }, + { + "epoch": 5.15, + "learning_rate": 5.937979651012248e-06, + "loss": 0.0076, + "step": 39340 + }, + { + "epoch": 5.15, + "learning_rate": 5.937495054540083e-06, + "loss": 0.0085, + "step": 39341 + }, + { + "epoch": 5.15, + "learning_rate": 5.937010469493501e-06, + "loss": 0.009, + "step": 39342 + }, + { + "epoch": 5.15, + "learning_rate": 5.936525895873861e-06, + "loss": 0.0079, + "step": 39343 + }, + { + "epoch": 5.15, + "learning_rate": 5.9360413336825225e-06, + "loss": 0.0082, + "step": 39344 + }, + { + "epoch": 5.15, + "learning_rate": 5.935556782920854e-06, + "loss": 0.0074, + "step": 39345 + }, + { + "epoch": 5.15, + "learning_rate": 5.935072243590216e-06, + "loss": 0.0081, + "step": 39346 + }, + { + "epoch": 5.15, + "learning_rate": 5.934587715691969e-06, + "loss": 0.0094, + "step": 39347 + }, + { + "epoch": 5.15, + "learning_rate": 5.93410319922748e-06, + "loss": 0.0089, + "step": 39348 + }, + { + "epoch": 5.15, + "learning_rate": 5.933618694198106e-06, + "loss": 0.0125, + "step": 39349 + }, + { + "epoch": 5.15, + "learning_rate": 5.9331342006052164e-06, + "loss": 0.0054, + "step": 39350 + }, + { + "epoch": 5.15, + "learning_rate": 5.932649718450167e-06, + "loss": 0.0098, + "step": 39351 + }, + { + "epoch": 5.15, + "learning_rate": 5.932165247734323e-06, + "loss": 0.0102, + "step": 39352 + }, + { + "epoch": 5.15, + "learning_rate": 5.931680788459049e-06, + "loss": 0.0063, + "step": 39353 + }, + { + "epoch": 5.16, + "learning_rate": 5.931196340625708e-06, + "loss": 0.0067, + "step": 39354 + }, + { + "epoch": 5.16, + "learning_rate": 5.930711904235659e-06, + "loss": 0.0081, + "step": 39355 + }, + { + "epoch": 5.16, + "learning_rate": 5.930227479290267e-06, + "loss": 0.0083, + "step": 39356 + }, + { + "epoch": 5.16, + "learning_rate": 5.9297430657908924e-06, + "loss": 0.0066, + "step": 39357 + }, + { + "epoch": 5.16, + "learning_rate": 5.929258663738901e-06, + "loss": 0.006, + "step": 39358 + }, + { + "epoch": 5.16, + "learning_rate": 5.92877427313565e-06, + "loss": 0.0083, + "step": 39359 + }, + { + "epoch": 5.16, + "learning_rate": 5.928289893982505e-06, + "loss": 0.0078, + "step": 39360 + }, + { + "epoch": 5.16, + "learning_rate": 5.92780552628083e-06, + "loss": 0.0068, + "step": 39361 + }, + { + "epoch": 5.16, + "learning_rate": 5.9273211700319835e-06, + "loss": 0.0097, + "step": 39362 + }, + { + "epoch": 5.16, + "learning_rate": 5.926836825237326e-06, + "loss": 0.0074, + "step": 39363 + }, + { + "epoch": 5.16, + "learning_rate": 5.926352491898229e-06, + "loss": 0.0079, + "step": 39364 + }, + { + "epoch": 5.16, + "learning_rate": 5.925868170016045e-06, + "loss": 0.0072, + "step": 39365 + }, + { + "epoch": 5.16, + "learning_rate": 5.92538385959214e-06, + "loss": 0.0068, + "step": 39366 + }, + { + "epoch": 5.16, + "learning_rate": 5.924899560627877e-06, + "loss": 0.0086, + "step": 39367 + }, + { + "epoch": 5.16, + "learning_rate": 5.9244152731246166e-06, + "loss": 0.0082, + "step": 39368 + }, + { + "epoch": 5.16, + "learning_rate": 5.923930997083722e-06, + "loss": 0.0096, + "step": 39369 + }, + { + "epoch": 5.16, + "learning_rate": 5.923446732506553e-06, + "loss": 0.0053, + "step": 39370 + }, + { + "epoch": 5.16, + "learning_rate": 5.922962479394477e-06, + "loss": 0.0069, + "step": 39371 + }, + { + "epoch": 5.16, + "learning_rate": 5.922478237748847e-06, + "loss": 0.009, + "step": 39372 + }, + { + "epoch": 5.16, + "learning_rate": 5.9219940075710325e-06, + "loss": 0.0087, + "step": 39373 + }, + { + "epoch": 5.16, + "learning_rate": 5.9215097888623874e-06, + "loss": 0.0054, + "step": 39374 + }, + { + "epoch": 5.16, + "learning_rate": 5.9210255816242845e-06, + "loss": 0.0054, + "step": 39375 + }, + { + "epoch": 5.16, + "learning_rate": 5.920541385858077e-06, + "loss": 0.0061, + "step": 39376 + }, + { + "epoch": 5.16, + "learning_rate": 5.920057201565132e-06, + "loss": 0.007, + "step": 39377 + }, + { + "epoch": 5.16, + "learning_rate": 5.919573028746808e-06, + "loss": 0.009, + "step": 39378 + }, + { + "epoch": 5.16, + "learning_rate": 5.9190888674044686e-06, + "loss": 0.0061, + "step": 39379 + }, + { + "epoch": 5.16, + "learning_rate": 5.918604717539473e-06, + "loss": 0.007, + "step": 39380 + }, + { + "epoch": 5.16, + "learning_rate": 5.918120579153184e-06, + "loss": 0.0054, + "step": 39381 + }, + { + "epoch": 5.16, + "learning_rate": 5.9176364522469656e-06, + "loss": 0.0078, + "step": 39382 + }, + { + "epoch": 5.16, + "learning_rate": 5.917152336822175e-06, + "loss": 0.0076, + "step": 39383 + }, + { + "epoch": 5.16, + "learning_rate": 5.916668232880179e-06, + "loss": 0.009, + "step": 39384 + }, + { + "epoch": 5.16, + "learning_rate": 5.91618414042233e-06, + "loss": 0.0081, + "step": 39385 + }, + { + "epoch": 5.16, + "learning_rate": 5.915700059449999e-06, + "loss": 0.0081, + "step": 39386 + }, + { + "epoch": 5.16, + "learning_rate": 5.915215989964547e-06, + "loss": 0.0077, + "step": 39387 + }, + { + "epoch": 5.16, + "learning_rate": 5.914731931967329e-06, + "loss": 0.0091, + "step": 39388 + }, + { + "epoch": 5.16, + "learning_rate": 5.914247885459709e-06, + "loss": 0.0099, + "step": 39389 + }, + { + "epoch": 5.16, + "learning_rate": 5.913763850443054e-06, + "loss": 0.0123, + "step": 39390 + }, + { + "epoch": 5.16, + "learning_rate": 5.9132798269187174e-06, + "loss": 0.0057, + "step": 39391 + }, + { + "epoch": 5.16, + "learning_rate": 5.9127958148880656e-06, + "loss": 0.0072, + "step": 39392 + }, + { + "epoch": 5.16, + "learning_rate": 5.912311814352456e-06, + "loss": 0.008, + "step": 39393 + }, + { + "epoch": 5.16, + "learning_rate": 5.9118278253132534e-06, + "loss": 0.008, + "step": 39394 + }, + { + "epoch": 5.16, + "learning_rate": 5.9113438477718155e-06, + "loss": 0.0079, + "step": 39395 + }, + { + "epoch": 5.16, + "learning_rate": 5.910859881729503e-06, + "loss": 0.0134, + "step": 39396 + }, + { + "epoch": 5.16, + "learning_rate": 5.910375927187681e-06, + "loss": 0.0072, + "step": 39397 + }, + { + "epoch": 5.16, + "learning_rate": 5.909891984147712e-06, + "loss": 0.0078, + "step": 39398 + }, + { + "epoch": 5.16, + "learning_rate": 5.909408052610952e-06, + "loss": 0.0081, + "step": 39399 + }, + { + "epoch": 5.16, + "learning_rate": 5.908924132578765e-06, + "loss": 0.0073, + "step": 39400 + }, + { + "epoch": 5.16, + "learning_rate": 5.9084402240525095e-06, + "loss": 0.0085, + "step": 39401 + }, + { + "epoch": 5.16, + "learning_rate": 5.907956327033551e-06, + "loss": 0.0057, + "step": 39402 + }, + { + "epoch": 5.16, + "learning_rate": 5.907472441523243e-06, + "loss": 0.0105, + "step": 39403 + }, + { + "epoch": 5.16, + "learning_rate": 5.906988567522953e-06, + "loss": 0.0096, + "step": 39404 + }, + { + "epoch": 5.16, + "learning_rate": 5.906504705034041e-06, + "loss": 0.0095, + "step": 39405 + }, + { + "epoch": 5.16, + "learning_rate": 5.906020854057862e-06, + "loss": 0.0076, + "step": 39406 + }, + { + "epoch": 5.16, + "learning_rate": 5.905537014595784e-06, + "loss": 0.0059, + "step": 39407 + }, + { + "epoch": 5.16, + "learning_rate": 5.9050531866491676e-06, + "loss": 0.0096, + "step": 39408 + }, + { + "epoch": 5.16, + "learning_rate": 5.904569370219369e-06, + "loss": 0.0077, + "step": 39409 + }, + { + "epoch": 5.16, + "learning_rate": 5.904085565307749e-06, + "loss": 0.0097, + "step": 39410 + }, + { + "epoch": 5.16, + "learning_rate": 5.903601771915673e-06, + "loss": 0.0093, + "step": 39411 + }, + { + "epoch": 5.16, + "learning_rate": 5.903117990044498e-06, + "loss": 0.0076, + "step": 39412 + }, + { + "epoch": 5.16, + "learning_rate": 5.902634219695587e-06, + "loss": 0.0109, + "step": 39413 + }, + { + "epoch": 5.16, + "learning_rate": 5.902150460870297e-06, + "loss": 0.0069, + "step": 39414 + }, + { + "epoch": 5.16, + "learning_rate": 5.901666713569993e-06, + "loss": 0.0098, + "step": 39415 + }, + { + "epoch": 5.16, + "learning_rate": 5.90118297779603e-06, + "loss": 0.0081, + "step": 39416 + }, + { + "epoch": 5.16, + "learning_rate": 5.900699253549771e-06, + "loss": 0.01, + "step": 39417 + }, + { + "epoch": 5.16, + "learning_rate": 5.9002155408325765e-06, + "loss": 0.0091, + "step": 39418 + }, + { + "epoch": 5.16, + "learning_rate": 5.899731839645812e-06, + "loss": 0.008, + "step": 39419 + }, + { + "epoch": 5.16, + "learning_rate": 5.899248149990831e-06, + "loss": 0.0082, + "step": 39420 + }, + { + "epoch": 5.16, + "learning_rate": 5.898764471868999e-06, + "loss": 0.0098, + "step": 39421 + }, + { + "epoch": 5.16, + "learning_rate": 5.898280805281669e-06, + "loss": 0.0086, + "step": 39422 + }, + { + "epoch": 5.16, + "learning_rate": 5.897797150230209e-06, + "loss": 0.0084, + "step": 39423 + }, + { + "epoch": 5.16, + "learning_rate": 5.897313506715974e-06, + "loss": 0.012, + "step": 39424 + }, + { + "epoch": 5.16, + "learning_rate": 5.896829874740326e-06, + "loss": 0.0054, + "step": 39425 + }, + { + "epoch": 5.16, + "learning_rate": 5.896346254304629e-06, + "loss": 0.0068, + "step": 39426 + }, + { + "epoch": 5.16, + "learning_rate": 5.895862645410236e-06, + "loss": 0.0081, + "step": 39427 + }, + { + "epoch": 5.16, + "learning_rate": 5.895379048058509e-06, + "loss": 0.0119, + "step": 39428 + }, + { + "epoch": 5.16, + "learning_rate": 5.894895462250815e-06, + "loss": 0.0068, + "step": 39429 + }, + { + "epoch": 5.17, + "learning_rate": 5.8944118879885055e-06, + "loss": 0.0058, + "step": 39430 + }, + { + "epoch": 5.17, + "learning_rate": 5.893928325272945e-06, + "loss": 0.0075, + "step": 39431 + }, + { + "epoch": 5.17, + "learning_rate": 5.893444774105495e-06, + "loss": 0.0081, + "step": 39432 + }, + { + "epoch": 5.17, + "learning_rate": 5.892961234487509e-06, + "loss": 0.0113, + "step": 39433 + }, + { + "epoch": 5.17, + "learning_rate": 5.892477706420354e-06, + "loss": 0.0056, + "step": 39434 + }, + { + "epoch": 5.17, + "learning_rate": 5.891994189905383e-06, + "loss": 0.0071, + "step": 39435 + }, + { + "epoch": 5.17, + "learning_rate": 5.891510684943964e-06, + "loss": 0.0092, + "step": 39436 + }, + { + "epoch": 5.17, + "learning_rate": 5.891027191537448e-06, + "loss": 0.0084, + "step": 39437 + }, + { + "epoch": 5.17, + "learning_rate": 5.890543709687204e-06, + "loss": 0.0082, + "step": 39438 + }, + { + "epoch": 5.17, + "learning_rate": 5.890060239394579e-06, + "loss": 0.0083, + "step": 39439 + }, + { + "epoch": 5.17, + "learning_rate": 5.889576780660947e-06, + "loss": 0.0063, + "step": 39440 + }, + { + "epoch": 5.17, + "learning_rate": 5.88909333348766e-06, + "loss": 0.0088, + "step": 39441 + }, + { + "epoch": 5.17, + "learning_rate": 5.888609897876081e-06, + "loss": 0.0073, + "step": 39442 + }, + { + "epoch": 5.17, + "learning_rate": 5.888126473827564e-06, + "loss": 0.0062, + "step": 39443 + }, + { + "epoch": 5.17, + "learning_rate": 5.887643061343477e-06, + "loss": 0.009, + "step": 39444 + }, + { + "epoch": 5.17, + "learning_rate": 5.8871596604251715e-06, + "loss": 0.0078, + "step": 39445 + }, + { + "epoch": 5.17, + "learning_rate": 5.88667627107401e-06, + "loss": 0.0071, + "step": 39446 + }, + { + "epoch": 5.17, + "learning_rate": 5.886192893291356e-06, + "loss": 0.0068, + "step": 39447 + }, + { + "epoch": 5.17, + "learning_rate": 5.885709527078562e-06, + "loss": 0.0087, + "step": 39448 + }, + { + "epoch": 5.17, + "learning_rate": 5.885226172436993e-06, + "loss": 0.0074, + "step": 39449 + }, + { + "epoch": 5.17, + "learning_rate": 5.884742829368003e-06, + "loss": 0.01, + "step": 39450 + }, + { + "epoch": 5.17, + "learning_rate": 5.884259497872956e-06, + "loss": 0.0071, + "step": 39451 + }, + { + "epoch": 5.17, + "learning_rate": 5.883776177953212e-06, + "loss": 0.0095, + "step": 39452 + }, + { + "epoch": 5.17, + "learning_rate": 5.883292869610127e-06, + "loss": 0.0074, + "step": 39453 + }, + { + "epoch": 5.17, + "learning_rate": 5.8828095728450604e-06, + "loss": 0.0095, + "step": 39454 + }, + { + "epoch": 5.17, + "learning_rate": 5.882326287659376e-06, + "loss": 0.0059, + "step": 39455 + }, + { + "epoch": 5.17, + "learning_rate": 5.881843014054427e-06, + "loss": 0.0078, + "step": 39456 + }, + { + "epoch": 5.17, + "learning_rate": 5.881359752031577e-06, + "loss": 0.0098, + "step": 39457 + }, + { + "epoch": 5.17, + "learning_rate": 5.880876501592181e-06, + "loss": 0.0081, + "step": 39458 + }, + { + "epoch": 5.17, + "learning_rate": 5.880393262737602e-06, + "loss": 0.01, + "step": 39459 + }, + { + "epoch": 5.17, + "learning_rate": 5.879910035469198e-06, + "loss": 0.0078, + "step": 39460 + }, + { + "epoch": 5.17, + "learning_rate": 5.879426819788324e-06, + "loss": 0.0081, + "step": 39461 + }, + { + "epoch": 5.17, + "learning_rate": 5.878943615696344e-06, + "loss": 0.0102, + "step": 39462 + }, + { + "epoch": 5.17, + "learning_rate": 5.878460423194618e-06, + "loss": 0.0073, + "step": 39463 + }, + { + "epoch": 5.17, + "learning_rate": 5.877977242284503e-06, + "loss": 0.0088, + "step": 39464 + }, + { + "epoch": 5.17, + "learning_rate": 5.877494072967357e-06, + "loss": 0.0068, + "step": 39465 + }, + { + "epoch": 5.17, + "learning_rate": 5.877010915244537e-06, + "loss": 0.0078, + "step": 39466 + }, + { + "epoch": 5.17, + "learning_rate": 5.876527769117408e-06, + "loss": 0.0076, + "step": 39467 + }, + { + "epoch": 5.17, + "learning_rate": 5.876044634587321e-06, + "loss": 0.007, + "step": 39468 + }, + { + "epoch": 5.17, + "learning_rate": 5.875561511655641e-06, + "loss": 0.0086, + "step": 39469 + }, + { + "epoch": 5.17, + "learning_rate": 5.875078400323726e-06, + "loss": 0.0094, + "step": 39470 + }, + { + "epoch": 5.17, + "learning_rate": 5.8745953005929315e-06, + "loss": 0.0076, + "step": 39471 + }, + { + "epoch": 5.17, + "learning_rate": 5.874112212464615e-06, + "loss": 0.0086, + "step": 39472 + }, + { + "epoch": 5.17, + "learning_rate": 5.873629135940142e-06, + "loss": 0.0103, + "step": 39473 + }, + { + "epoch": 5.17, + "learning_rate": 5.873146071020866e-06, + "loss": 0.0076, + "step": 39474 + }, + { + "epoch": 5.17, + "learning_rate": 5.872663017708146e-06, + "loss": 0.0054, + "step": 39475 + }, + { + "epoch": 5.17, + "learning_rate": 5.872179976003346e-06, + "loss": 0.0078, + "step": 39476 + }, + { + "epoch": 5.17, + "learning_rate": 5.871696945907816e-06, + "loss": 0.0077, + "step": 39477 + }, + { + "epoch": 5.17, + "learning_rate": 5.871213927422921e-06, + "loss": 0.0064, + "step": 39478 + }, + { + "epoch": 5.17, + "learning_rate": 5.870730920550014e-06, + "loss": 0.0082, + "step": 39479 + }, + { + "epoch": 5.17, + "learning_rate": 5.87024792529046e-06, + "loss": 0.007, + "step": 39480 + }, + { + "epoch": 5.17, + "learning_rate": 5.86976494164561e-06, + "loss": 0.0073, + "step": 39481 + }, + { + "epoch": 5.17, + "learning_rate": 5.869281969616829e-06, + "loss": 0.0066, + "step": 39482 + }, + { + "epoch": 5.17, + "learning_rate": 5.868799009205468e-06, + "loss": 0.0059, + "step": 39483 + }, + { + "epoch": 5.17, + "learning_rate": 5.868316060412894e-06, + "loss": 0.0059, + "step": 39484 + }, + { + "epoch": 5.17, + "learning_rate": 5.867833123240461e-06, + "loss": 0.0076, + "step": 39485 + }, + { + "epoch": 5.17, + "learning_rate": 5.867350197689527e-06, + "loss": 0.0065, + "step": 39486 + }, + { + "epoch": 5.17, + "learning_rate": 5.8668672837614484e-06, + "loss": 0.0082, + "step": 39487 + }, + { + "epoch": 5.17, + "learning_rate": 5.866384381457589e-06, + "loss": 0.0077, + "step": 39488 + }, + { + "epoch": 5.17, + "learning_rate": 5.8659014907793e-06, + "loss": 0.0062, + "step": 39489 + }, + { + "epoch": 5.17, + "learning_rate": 5.865418611727944e-06, + "loss": 0.0103, + "step": 39490 + }, + { + "epoch": 5.17, + "learning_rate": 5.864935744304881e-06, + "loss": 0.0078, + "step": 39491 + }, + { + "epoch": 5.17, + "learning_rate": 5.864452888511462e-06, + "loss": 0.0093, + "step": 39492 + }, + { + "epoch": 5.17, + "learning_rate": 5.863970044349051e-06, + "loss": 0.0098, + "step": 39493 + }, + { + "epoch": 5.17, + "learning_rate": 5.863487211819001e-06, + "loss": 0.0067, + "step": 39494 + }, + { + "epoch": 5.17, + "learning_rate": 5.863004390922674e-06, + "loss": 0.0072, + "step": 39495 + }, + { + "epoch": 5.17, + "learning_rate": 5.862521581661428e-06, + "loss": 0.0086, + "step": 39496 + }, + { + "epoch": 5.17, + "learning_rate": 5.8620387840366215e-06, + "loss": 0.0085, + "step": 39497 + }, + { + "epoch": 5.17, + "learning_rate": 5.861555998049609e-06, + "loss": 0.0073, + "step": 39498 + }, + { + "epoch": 5.17, + "learning_rate": 5.8610732237017495e-06, + "loss": 0.0106, + "step": 39499 + }, + { + "epoch": 5.17, + "learning_rate": 5.8605904609944006e-06, + "loss": 0.0082, + "step": 39500 + }, + { + "epoch": 5.17, + "learning_rate": 5.860107709928924e-06, + "loss": 0.0084, + "step": 39501 + }, + { + "epoch": 5.17, + "learning_rate": 5.85962497050667e-06, + "loss": 0.0078, + "step": 39502 + }, + { + "epoch": 5.17, + "learning_rate": 5.859142242729003e-06, + "loss": 0.0095, + "step": 39503 + }, + { + "epoch": 5.17, + "learning_rate": 5.858659526597275e-06, + "loss": 0.0099, + "step": 39504 + }, + { + "epoch": 5.17, + "learning_rate": 5.858176822112847e-06, + "loss": 0.0066, + "step": 39505 + }, + { + "epoch": 5.18, + "learning_rate": 5.857694129277076e-06, + "loss": 0.0059, + "step": 39506 + }, + { + "epoch": 5.18, + "learning_rate": 5.8572114480913225e-06, + "loss": 0.0076, + "step": 39507 + }, + { + "epoch": 5.18, + "learning_rate": 5.856728778556938e-06, + "loss": 0.0051, + "step": 39508 + }, + { + "epoch": 5.18, + "learning_rate": 5.856246120675286e-06, + "loss": 0.0074, + "step": 39509 + }, + { + "epoch": 5.18, + "learning_rate": 5.85576347444772e-06, + "loss": 0.0067, + "step": 39510 + }, + { + "epoch": 5.18, + "learning_rate": 5.8552808398755976e-06, + "loss": 0.0063, + "step": 39511 + }, + { + "epoch": 5.18, + "learning_rate": 5.854798216960279e-06, + "loss": 0.0083, + "step": 39512 + }, + { + "epoch": 5.18, + "learning_rate": 5.854315605703118e-06, + "loss": 0.0102, + "step": 39513 + }, + { + "epoch": 5.18, + "learning_rate": 5.853833006105476e-06, + "loss": 0.0063, + "step": 39514 + }, + { + "epoch": 5.18, + "learning_rate": 5.853350418168705e-06, + "loss": 0.0071, + "step": 39515 + }, + { + "epoch": 5.18, + "learning_rate": 5.852867841894163e-06, + "loss": 0.0083, + "step": 39516 + }, + { + "epoch": 5.18, + "learning_rate": 5.852385277283215e-06, + "loss": 0.0072, + "step": 39517 + }, + { + "epoch": 5.18, + "learning_rate": 5.851902724337208e-06, + "loss": 0.0079, + "step": 39518 + }, + { + "epoch": 5.18, + "learning_rate": 5.8514201830575055e-06, + "loss": 0.0071, + "step": 39519 + }, + { + "epoch": 5.18, + "learning_rate": 5.850937653445465e-06, + "loss": 0.0076, + "step": 39520 + }, + { + "epoch": 5.18, + "learning_rate": 5.850455135502439e-06, + "loss": 0.007, + "step": 39521 + }, + { + "epoch": 5.18, + "learning_rate": 5.849972629229789e-06, + "loss": 0.0111, + "step": 39522 + }, + { + "epoch": 5.18, + "learning_rate": 5.849490134628868e-06, + "loss": 0.0066, + "step": 39523 + }, + { + "epoch": 5.18, + "learning_rate": 5.849007651701036e-06, + "loss": 0.0071, + "step": 39524 + }, + { + "epoch": 5.18, + "learning_rate": 5.848525180447649e-06, + "loss": 0.01, + "step": 39525 + }, + { + "epoch": 5.18, + "learning_rate": 5.848042720870062e-06, + "loss": 0.0088, + "step": 39526 + }, + { + "epoch": 5.18, + "learning_rate": 5.847560272969632e-06, + "loss": 0.0063, + "step": 39527 + }, + { + "epoch": 5.18, + "learning_rate": 5.847077836747721e-06, + "loss": 0.0061, + "step": 39528 + }, + { + "epoch": 5.18, + "learning_rate": 5.846595412205682e-06, + "loss": 0.0064, + "step": 39529 + }, + { + "epoch": 5.18, + "learning_rate": 5.846112999344873e-06, + "loss": 0.0074, + "step": 39530 + }, + { + "epoch": 5.18, + "learning_rate": 5.845630598166649e-06, + "loss": 0.0052, + "step": 39531 + }, + { + "epoch": 5.18, + "learning_rate": 5.845148208672367e-06, + "loss": 0.0108, + "step": 39532 + }, + { + "epoch": 5.18, + "learning_rate": 5.844665830863387e-06, + "loss": 0.0098, + "step": 39533 + }, + { + "epoch": 5.18, + "learning_rate": 5.84418346474106e-06, + "loss": 0.0083, + "step": 39534 + }, + { + "epoch": 5.18, + "learning_rate": 5.843701110306748e-06, + "loss": 0.0089, + "step": 39535 + }, + { + "epoch": 5.18, + "learning_rate": 5.843218767561805e-06, + "loss": 0.0075, + "step": 39536 + }, + { + "epoch": 5.18, + "learning_rate": 5.842736436507583e-06, + "loss": 0.0094, + "step": 39537 + }, + { + "epoch": 5.18, + "learning_rate": 5.842254117145449e-06, + "loss": 0.0079, + "step": 39538 + }, + { + "epoch": 5.18, + "learning_rate": 5.841771809476753e-06, + "loss": 0.0083, + "step": 39539 + }, + { + "epoch": 5.18, + "learning_rate": 5.84128951350285e-06, + "loss": 0.007, + "step": 39540 + }, + { + "epoch": 5.18, + "learning_rate": 5.840807229225103e-06, + "loss": 0.0072, + "step": 39541 + }, + { + "epoch": 5.18, + "learning_rate": 5.840324956644861e-06, + "loss": 0.0098, + "step": 39542 + }, + { + "epoch": 5.18, + "learning_rate": 5.839842695763485e-06, + "loss": 0.0076, + "step": 39543 + }, + { + "epoch": 5.18, + "learning_rate": 5.839360446582329e-06, + "loss": 0.008, + "step": 39544 + }, + { + "epoch": 5.18, + "learning_rate": 5.838878209102752e-06, + "loss": 0.0075, + "step": 39545 + }, + { + "epoch": 5.18, + "learning_rate": 5.838395983326107e-06, + "loss": 0.0093, + "step": 39546 + }, + { + "epoch": 5.18, + "learning_rate": 5.837913769253749e-06, + "loss": 0.0101, + "step": 39547 + }, + { + "epoch": 5.18, + "learning_rate": 5.837431566887037e-06, + "loss": 0.0078, + "step": 39548 + }, + { + "epoch": 5.18, + "learning_rate": 5.83694937622733e-06, + "loss": 0.0052, + "step": 39549 + }, + { + "epoch": 5.18, + "learning_rate": 5.836467197275982e-06, + "loss": 0.0095, + "step": 39550 + }, + { + "epoch": 5.18, + "learning_rate": 5.8359850300343445e-06, + "loss": 0.0056, + "step": 39551 + }, + { + "epoch": 5.18, + "learning_rate": 5.835502874503781e-06, + "loss": 0.0072, + "step": 39552 + }, + { + "epoch": 5.18, + "learning_rate": 5.835020730685644e-06, + "loss": 0.0079, + "step": 39553 + }, + { + "epoch": 5.18, + "learning_rate": 5.834538598581285e-06, + "loss": 0.0113, + "step": 39554 + }, + { + "epoch": 5.18, + "learning_rate": 5.8340564781920694e-06, + "loss": 0.0059, + "step": 39555 + }, + { + "epoch": 5.18, + "learning_rate": 5.833574369519347e-06, + "loss": 0.0056, + "step": 39556 + }, + { + "epoch": 5.18, + "learning_rate": 5.8330922725644755e-06, + "loss": 0.0073, + "step": 39557 + }, + { + "epoch": 5.18, + "learning_rate": 5.832610187328805e-06, + "loss": 0.009, + "step": 39558 + }, + { + "epoch": 5.18, + "learning_rate": 5.832128113813697e-06, + "loss": 0.0084, + "step": 39559 + }, + { + "epoch": 5.18, + "learning_rate": 5.831646052020512e-06, + "loss": 0.0066, + "step": 39560 + }, + { + "epoch": 5.18, + "learning_rate": 5.8311640019506e-06, + "loss": 0.0066, + "step": 39561 + }, + { + "epoch": 5.18, + "learning_rate": 5.830681963605312e-06, + "loss": 0.0085, + "step": 39562 + }, + { + "epoch": 5.18, + "learning_rate": 5.830199936986014e-06, + "loss": 0.0067, + "step": 39563 + }, + { + "epoch": 5.18, + "learning_rate": 5.829717922094056e-06, + "loss": 0.0059, + "step": 39564 + }, + { + "epoch": 5.18, + "learning_rate": 5.8292359189307944e-06, + "loss": 0.007, + "step": 39565 + }, + { + "epoch": 5.18, + "learning_rate": 5.828753927497581e-06, + "loss": 0.0085, + "step": 39566 + }, + { + "epoch": 5.18, + "learning_rate": 5.828271947795779e-06, + "loss": 0.0067, + "step": 39567 + }, + { + "epoch": 5.18, + "learning_rate": 5.827789979826739e-06, + "loss": 0.0074, + "step": 39568 + }, + { + "epoch": 5.18, + "learning_rate": 5.827308023591815e-06, + "loss": 0.0073, + "step": 39569 + }, + { + "epoch": 5.18, + "learning_rate": 5.826826079092365e-06, + "loss": 0.0077, + "step": 39570 + }, + { + "epoch": 5.18, + "learning_rate": 5.826344146329748e-06, + "loss": 0.0088, + "step": 39571 + }, + { + "epoch": 5.18, + "learning_rate": 5.825862225305314e-06, + "loss": 0.0138, + "step": 39572 + }, + { + "epoch": 5.18, + "learning_rate": 5.8253803160204215e-06, + "loss": 0.0111, + "step": 39573 + }, + { + "epoch": 5.18, + "learning_rate": 5.824898418476421e-06, + "loss": 0.0048, + "step": 39574 + }, + { + "epoch": 5.18, + "learning_rate": 5.824416532674675e-06, + "loss": 0.0077, + "step": 39575 + }, + { + "epoch": 5.18, + "learning_rate": 5.823934658616537e-06, + "loss": 0.0075, + "step": 39576 + }, + { + "epoch": 5.18, + "learning_rate": 5.8234527963033535e-06, + "loss": 0.0081, + "step": 39577 + }, + { + "epoch": 5.18, + "learning_rate": 5.8229709457364925e-06, + "loss": 0.0072, + "step": 39578 + }, + { + "epoch": 5.18, + "learning_rate": 5.822489106917302e-06, + "loss": 0.0065, + "step": 39579 + }, + { + "epoch": 5.18, + "learning_rate": 5.82200727984714e-06, + "loss": 0.0065, + "step": 39580 + }, + { + "epoch": 5.18, + "learning_rate": 5.8215254645273535e-06, + "loss": 0.0105, + "step": 39581 + }, + { + "epoch": 5.18, + "learning_rate": 5.8210436609593066e-06, + "loss": 0.0091, + "step": 39582 + }, + { + "epoch": 5.19, + "learning_rate": 5.820561869144354e-06, + "loss": 0.0071, + "step": 39583 + }, + { + "epoch": 5.19, + "learning_rate": 5.820080089083851e-06, + "loss": 0.0078, + "step": 39584 + }, + { + "epoch": 5.19, + "learning_rate": 5.819598320779143e-06, + "loss": 0.0063, + "step": 39585 + }, + { + "epoch": 5.19, + "learning_rate": 5.8191165642315986e-06, + "loss": 0.0074, + "step": 39586 + }, + { + "epoch": 5.19, + "learning_rate": 5.818634819442566e-06, + "loss": 0.0078, + "step": 39587 + }, + { + "epoch": 5.19, + "learning_rate": 5.8181530864133986e-06, + "loss": 0.01, + "step": 39588 + }, + { + "epoch": 5.19, + "learning_rate": 5.8176713651454495e-06, + "loss": 0.0096, + "step": 39589 + }, + { + "epoch": 5.19, + "learning_rate": 5.817189655640082e-06, + "loss": 0.0069, + "step": 39590 + }, + { + "epoch": 5.19, + "learning_rate": 5.816707957898645e-06, + "loss": 0.0067, + "step": 39591 + }, + { + "epoch": 5.19, + "learning_rate": 5.816226271922491e-06, + "loss": 0.0087, + "step": 39592 + }, + { + "epoch": 5.19, + "learning_rate": 5.815744597712976e-06, + "loss": 0.0059, + "step": 39593 + }, + { + "epoch": 5.19, + "learning_rate": 5.815262935271463e-06, + "loss": 0.0092, + "step": 39594 + }, + { + "epoch": 5.19, + "learning_rate": 5.814781284599298e-06, + "loss": 0.0099, + "step": 39595 + }, + { + "epoch": 5.19, + "learning_rate": 5.814299645697838e-06, + "loss": 0.0086, + "step": 39596 + }, + { + "epoch": 5.19, + "learning_rate": 5.813818018568435e-06, + "loss": 0.0085, + "step": 39597 + }, + { + "epoch": 5.19, + "learning_rate": 5.8133364032124485e-06, + "loss": 0.0061, + "step": 39598 + }, + { + "epoch": 5.19, + "learning_rate": 5.812854799631231e-06, + "loss": 0.0092, + "step": 39599 + }, + { + "epoch": 5.19, + "learning_rate": 5.812373207826132e-06, + "loss": 0.0053, + "step": 39600 + }, + { + "epoch": 5.19, + "learning_rate": 5.8118916277985145e-06, + "loss": 0.0088, + "step": 39601 + }, + { + "epoch": 5.19, + "learning_rate": 5.811410059549729e-06, + "loss": 0.0055, + "step": 39602 + }, + { + "epoch": 5.19, + "learning_rate": 5.810928503081126e-06, + "loss": 0.0063, + "step": 39603 + }, + { + "epoch": 5.19, + "learning_rate": 5.810446958394062e-06, + "loss": 0.0056, + "step": 39604 + }, + { + "epoch": 5.19, + "learning_rate": 5.809965425489898e-06, + "loss": 0.0057, + "step": 39605 + }, + { + "epoch": 5.19, + "learning_rate": 5.809483904369983e-06, + "loss": 0.0071, + "step": 39606 + }, + { + "epoch": 5.19, + "learning_rate": 5.809002395035672e-06, + "loss": 0.0082, + "step": 39607 + }, + { + "epoch": 5.19, + "learning_rate": 5.808520897488313e-06, + "loss": 0.0067, + "step": 39608 + }, + { + "epoch": 5.19, + "learning_rate": 5.808039411729272e-06, + "loss": 0.007, + "step": 39609 + }, + { + "epoch": 5.19, + "learning_rate": 5.807557937759895e-06, + "loss": 0.0096, + "step": 39610 + }, + { + "epoch": 5.19, + "learning_rate": 5.807076475581538e-06, + "loss": 0.008, + "step": 39611 + }, + { + "epoch": 5.19, + "learning_rate": 5.806595025195552e-06, + "loss": 0.0054, + "step": 39612 + }, + { + "epoch": 5.19, + "learning_rate": 5.806113586603297e-06, + "loss": 0.0087, + "step": 39613 + }, + { + "epoch": 5.19, + "learning_rate": 5.8056321598061215e-06, + "loss": 0.0074, + "step": 39614 + }, + { + "epoch": 5.19, + "learning_rate": 5.805150744805386e-06, + "loss": 0.007, + "step": 39615 + }, + { + "epoch": 5.19, + "learning_rate": 5.804669341602435e-06, + "loss": 0.0089, + "step": 39616 + }, + { + "epoch": 5.19, + "learning_rate": 5.804187950198634e-06, + "loss": 0.0069, + "step": 39617 + }, + { + "epoch": 5.19, + "learning_rate": 5.803706570595331e-06, + "loss": 0.0097, + "step": 39618 + }, + { + "epoch": 5.19, + "learning_rate": 5.803225202793875e-06, + "loss": 0.0072, + "step": 39619 + }, + { + "epoch": 5.19, + "learning_rate": 5.802743846795629e-06, + "loss": 0.0075, + "step": 39620 + }, + { + "epoch": 5.19, + "learning_rate": 5.802262502601942e-06, + "loss": 0.008, + "step": 39621 + }, + { + "epoch": 5.19, + "learning_rate": 5.801781170214169e-06, + "loss": 0.0073, + "step": 39622 + }, + { + "epoch": 5.19, + "learning_rate": 5.801299849633657e-06, + "loss": 0.0068, + "step": 39623 + }, + { + "epoch": 5.19, + "learning_rate": 5.8008185408617725e-06, + "loss": 0.0088, + "step": 39624 + }, + { + "epoch": 5.19, + "learning_rate": 5.800337243899856e-06, + "loss": 0.0071, + "step": 39625 + }, + { + "epoch": 5.19, + "learning_rate": 5.799855958749272e-06, + "loss": 0.0069, + "step": 39626 + }, + { + "epoch": 5.19, + "learning_rate": 5.799374685411365e-06, + "loss": 0.0068, + "step": 39627 + }, + { + "epoch": 5.19, + "learning_rate": 5.7988934238874974e-06, + "loss": 0.0075, + "step": 39628 + }, + { + "epoch": 5.19, + "learning_rate": 5.798412174179018e-06, + "loss": 0.0071, + "step": 39629 + }, + { + "epoch": 5.19, + "learning_rate": 5.797930936287282e-06, + "loss": 0.006, + "step": 39630 + }, + { + "epoch": 5.19, + "learning_rate": 5.797449710213636e-06, + "loss": 0.0075, + "step": 39631 + }, + { + "epoch": 5.19, + "learning_rate": 5.796968495959443e-06, + "loss": 0.0071, + "step": 39632 + }, + { + "epoch": 5.19, + "learning_rate": 5.796487293526052e-06, + "loss": 0.01, + "step": 39633 + }, + { + "epoch": 5.19, + "learning_rate": 5.796006102914813e-06, + "loss": 0.0058, + "step": 39634 + }, + { + "epoch": 5.19, + "learning_rate": 5.795524924127087e-06, + "loss": 0.006, + "step": 39635 + }, + { + "epoch": 5.19, + "learning_rate": 5.79504375716422e-06, + "loss": 0.0059, + "step": 39636 + }, + { + "epoch": 5.19, + "learning_rate": 5.794562602027571e-06, + "loss": 0.0056, + "step": 39637 + }, + { + "epoch": 5.19, + "learning_rate": 5.794081458718493e-06, + "loss": 0.0089, + "step": 39638 + }, + { + "epoch": 5.19, + "learning_rate": 5.793600327238331e-06, + "loss": 0.0063, + "step": 39639 + }, + { + "epoch": 5.19, + "learning_rate": 5.7931192075884494e-06, + "loss": 0.0074, + "step": 39640 + }, + { + "epoch": 5.19, + "learning_rate": 5.792638099770195e-06, + "loss": 0.008, + "step": 39641 + }, + { + "epoch": 5.19, + "learning_rate": 5.792157003784918e-06, + "loss": 0.0068, + "step": 39642 + }, + { + "epoch": 5.19, + "learning_rate": 5.791675919633979e-06, + "loss": 0.008, + "step": 39643 + }, + { + "epoch": 5.19, + "learning_rate": 5.7911948473187285e-06, + "loss": 0.0075, + "step": 39644 + }, + { + "epoch": 5.19, + "learning_rate": 5.7907137868405185e-06, + "loss": 0.0097, + "step": 39645 + }, + { + "epoch": 5.19, + "learning_rate": 5.790232738200696e-06, + "loss": 0.0106, + "step": 39646 + }, + { + "epoch": 5.19, + "learning_rate": 5.789751701400621e-06, + "loss": 0.0067, + "step": 39647 + }, + { + "epoch": 5.19, + "learning_rate": 5.7892706764416495e-06, + "loss": 0.0051, + "step": 39648 + }, + { + "epoch": 5.19, + "learning_rate": 5.788789663325129e-06, + "loss": 0.0077, + "step": 39649 + }, + { + "epoch": 5.19, + "learning_rate": 5.788308662052409e-06, + "loss": 0.007, + "step": 39650 + }, + { + "epoch": 5.19, + "learning_rate": 5.7878276726248514e-06, + "loss": 0.0059, + "step": 39651 + }, + { + "epoch": 5.19, + "learning_rate": 5.787346695043805e-06, + "loss": 0.0088, + "step": 39652 + }, + { + "epoch": 5.19, + "learning_rate": 5.78686572931062e-06, + "loss": 0.0097, + "step": 39653 + }, + { + "epoch": 5.19, + "learning_rate": 5.786384775426648e-06, + "loss": 0.008, + "step": 39654 + }, + { + "epoch": 5.19, + "learning_rate": 5.7859038333932475e-06, + "loss": 0.0078, + "step": 39655 + }, + { + "epoch": 5.19, + "learning_rate": 5.785422903211768e-06, + "loss": 0.0081, + "step": 39656 + }, + { + "epoch": 5.19, + "learning_rate": 5.784941984883559e-06, + "loss": 0.0073, + "step": 39657 + }, + { + "epoch": 5.19, + "learning_rate": 5.784461078409976e-06, + "loss": 0.0078, + "step": 39658 + }, + { + "epoch": 5.2, + "learning_rate": 5.783980183792376e-06, + "loss": 0.0073, + "step": 39659 + }, + { + "epoch": 5.2, + "learning_rate": 5.783499301032107e-06, + "loss": 0.0076, + "step": 39660 + }, + { + "epoch": 5.2, + "learning_rate": 5.783018430130521e-06, + "loss": 0.0089, + "step": 39661 + }, + { + "epoch": 5.2, + "learning_rate": 5.782537571088967e-06, + "loss": 0.0061, + "step": 39662 + }, + { + "epoch": 5.2, + "learning_rate": 5.782056723908807e-06, + "loss": 0.0087, + "step": 39663 + }, + { + "epoch": 5.2, + "learning_rate": 5.781575888591386e-06, + "loss": 0.0085, + "step": 39664 + }, + { + "epoch": 5.2, + "learning_rate": 5.781095065138056e-06, + "loss": 0.0064, + "step": 39665 + }, + { + "epoch": 5.2, + "learning_rate": 5.780614253550174e-06, + "loss": 0.0064, + "step": 39666 + }, + { + "epoch": 5.2, + "learning_rate": 5.780133453829091e-06, + "loss": 0.0076, + "step": 39667 + }, + { + "epoch": 5.2, + "learning_rate": 5.779652665976157e-06, + "loss": 0.0085, + "step": 39668 + }, + { + "epoch": 5.2, + "learning_rate": 5.779171889992716e-06, + "loss": 0.0073, + "step": 39669 + }, + { + "epoch": 5.2, + "learning_rate": 5.778691125880139e-06, + "loss": 0.0082, + "step": 39670 + }, + { + "epoch": 5.2, + "learning_rate": 5.778210373639768e-06, + "loss": 0.0093, + "step": 39671 + }, + { + "epoch": 5.2, + "learning_rate": 5.777729633272956e-06, + "loss": 0.0083, + "step": 39672 + }, + { + "epoch": 5.2, + "learning_rate": 5.77724890478105e-06, + "loss": 0.0076, + "step": 39673 + }, + { + "epoch": 5.2, + "learning_rate": 5.776768188165412e-06, + "loss": 0.0068, + "step": 39674 + }, + { + "epoch": 5.2, + "learning_rate": 5.776287483427385e-06, + "loss": 0.0089, + "step": 39675 + }, + { + "epoch": 5.2, + "learning_rate": 5.775806790568328e-06, + "loss": 0.0088, + "step": 39676 + }, + { + "epoch": 5.2, + "learning_rate": 5.775326109589584e-06, + "loss": 0.0059, + "step": 39677 + }, + { + "epoch": 5.2, + "learning_rate": 5.774845440492514e-06, + "loss": 0.0111, + "step": 39678 + }, + { + "epoch": 5.2, + "learning_rate": 5.774364783278463e-06, + "loss": 0.0083, + "step": 39679 + }, + { + "epoch": 5.2, + "learning_rate": 5.77388413794879e-06, + "loss": 0.0112, + "step": 39680 + }, + { + "epoch": 5.2, + "learning_rate": 5.773403504504837e-06, + "loss": 0.009, + "step": 39681 + }, + { + "epoch": 5.2, + "learning_rate": 5.772922882947969e-06, + "loss": 0.0061, + "step": 39682 + }, + { + "epoch": 5.2, + "learning_rate": 5.772442273279528e-06, + "loss": 0.0084, + "step": 39683 + }, + { + "epoch": 5.2, + "learning_rate": 5.771961675500864e-06, + "loss": 0.0058, + "step": 39684 + }, + { + "epoch": 5.2, + "learning_rate": 5.7714810896133374e-06, + "loss": 0.0075, + "step": 39685 + }, + { + "epoch": 5.2, + "learning_rate": 5.771000515618295e-06, + "loss": 0.0061, + "step": 39686 + }, + { + "epoch": 5.2, + "learning_rate": 5.770519953517087e-06, + "loss": 0.0092, + "step": 39687 + }, + { + "epoch": 5.2, + "learning_rate": 5.770039403311064e-06, + "loss": 0.0067, + "step": 39688 + }, + { + "epoch": 5.2, + "learning_rate": 5.769558865001583e-06, + "loss": 0.0061, + "step": 39689 + }, + { + "epoch": 5.2, + "learning_rate": 5.769078338589989e-06, + "loss": 0.0079, + "step": 39690 + }, + { + "epoch": 5.2, + "learning_rate": 5.768597824077642e-06, + "loss": 0.0093, + "step": 39691 + }, + { + "epoch": 5.2, + "learning_rate": 5.768117321465883e-06, + "loss": 0.0052, + "step": 39692 + }, + { + "epoch": 5.2, + "learning_rate": 5.767636830756071e-06, + "loss": 0.0063, + "step": 39693 + }, + { + "epoch": 5.2, + "learning_rate": 5.767156351949558e-06, + "loss": 0.0096, + "step": 39694 + }, + { + "epoch": 5.2, + "learning_rate": 5.76667588504769e-06, + "loss": 0.0071, + "step": 39695 + }, + { + "epoch": 5.2, + "learning_rate": 5.766195430051818e-06, + "loss": 0.0097, + "step": 39696 + }, + { + "epoch": 5.2, + "learning_rate": 5.7657149869633e-06, + "loss": 0.0086, + "step": 39697 + }, + { + "epoch": 5.2, + "learning_rate": 5.765234555783483e-06, + "loss": 0.0081, + "step": 39698 + }, + { + "epoch": 5.2, + "learning_rate": 5.764754136513713e-06, + "loss": 0.0077, + "step": 39699 + }, + { + "epoch": 5.2, + "learning_rate": 5.764273729155352e-06, + "loss": 0.006, + "step": 39700 + }, + { + "epoch": 5.2, + "learning_rate": 5.763793333709741e-06, + "loss": 0.0076, + "step": 39701 + }, + { + "epoch": 5.2, + "learning_rate": 5.7633129501782396e-06, + "loss": 0.0083, + "step": 39702 + }, + { + "epoch": 5.2, + "learning_rate": 5.762832578562194e-06, + "loss": 0.0088, + "step": 39703 + }, + { + "epoch": 5.2, + "learning_rate": 5.762352218862954e-06, + "loss": 0.0063, + "step": 39704 + }, + { + "epoch": 5.2, + "learning_rate": 5.761871871081876e-06, + "loss": 0.0094, + "step": 39705 + }, + { + "epoch": 5.2, + "learning_rate": 5.761391535220306e-06, + "loss": 0.0067, + "step": 39706 + }, + { + "epoch": 5.2, + "learning_rate": 5.760911211279595e-06, + "loss": 0.0083, + "step": 39707 + }, + { + "epoch": 5.2, + "learning_rate": 5.760430899261099e-06, + "loss": 0.0077, + "step": 39708 + }, + { + "epoch": 5.2, + "learning_rate": 5.759950599166164e-06, + "loss": 0.0076, + "step": 39709 + }, + { + "epoch": 5.2, + "learning_rate": 5.7594703109961435e-06, + "loss": 0.0085, + "step": 39710 + }, + { + "epoch": 5.2, + "learning_rate": 5.758990034752383e-06, + "loss": 0.0081, + "step": 39711 + }, + { + "epoch": 5.2, + "learning_rate": 5.758509770436236e-06, + "loss": 0.0046, + "step": 39712 + }, + { + "epoch": 5.2, + "learning_rate": 5.75802951804906e-06, + "loss": 0.0085, + "step": 39713 + }, + { + "epoch": 5.2, + "learning_rate": 5.757549277592198e-06, + "loss": 0.0087, + "step": 39714 + }, + { + "epoch": 5.2, + "learning_rate": 5.757069049067e-06, + "loss": 0.011, + "step": 39715 + }, + { + "epoch": 5.2, + "learning_rate": 5.756588832474823e-06, + "loss": 0.0091, + "step": 39716 + }, + { + "epoch": 5.2, + "learning_rate": 5.756108627817014e-06, + "loss": 0.0075, + "step": 39717 + }, + { + "epoch": 5.2, + "learning_rate": 5.755628435094924e-06, + "loss": 0.0077, + "step": 39718 + }, + { + "epoch": 5.2, + "learning_rate": 5.755148254309899e-06, + "loss": 0.0072, + "step": 39719 + }, + { + "epoch": 5.2, + "learning_rate": 5.754668085463298e-06, + "loss": 0.006, + "step": 39720 + }, + { + "epoch": 5.2, + "learning_rate": 5.754187928556465e-06, + "loss": 0.0078, + "step": 39721 + }, + { + "epoch": 5.2, + "learning_rate": 5.753707783590749e-06, + "loss": 0.0075, + "step": 39722 + }, + { + "epoch": 5.2, + "learning_rate": 5.753227650567505e-06, + "loss": 0.0065, + "step": 39723 + }, + { + "epoch": 5.2, + "learning_rate": 5.752747529488084e-06, + "loss": 0.0075, + "step": 39724 + }, + { + "epoch": 5.2, + "learning_rate": 5.752267420353836e-06, + "loss": 0.0094, + "step": 39725 + }, + { + "epoch": 5.2, + "learning_rate": 5.7517873231661095e-06, + "loss": 0.0088, + "step": 39726 + }, + { + "epoch": 5.2, + "learning_rate": 5.75130723792625e-06, + "loss": 0.0078, + "step": 39727 + }, + { + "epoch": 5.2, + "learning_rate": 5.750827164635618e-06, + "loss": 0.0059, + "step": 39728 + }, + { + "epoch": 5.2, + "learning_rate": 5.750347103295557e-06, + "loss": 0.009, + "step": 39729 + }, + { + "epoch": 5.2, + "learning_rate": 5.749867053907415e-06, + "loss": 0.0078, + "step": 39730 + }, + { + "epoch": 5.2, + "learning_rate": 5.749387016472549e-06, + "loss": 0.0079, + "step": 39731 + }, + { + "epoch": 5.2, + "learning_rate": 5.748906990992307e-06, + "loss": 0.0076, + "step": 39732 + }, + { + "epoch": 5.2, + "learning_rate": 5.748426977468038e-06, + "loss": 0.0069, + "step": 39733 + }, + { + "epoch": 5.2, + "learning_rate": 5.747946975901082e-06, + "loss": 0.0075, + "step": 39734 + }, + { + "epoch": 5.21, + "learning_rate": 5.7474669862928065e-06, + "loss": 0.0075, + "step": 39735 + }, + { + "epoch": 5.21, + "learning_rate": 5.746987008644555e-06, + "loss": 0.0081, + "step": 39736 + }, + { + "epoch": 5.21, + "learning_rate": 5.7465070429576765e-06, + "loss": 0.0102, + "step": 39737 + }, + { + "epoch": 5.21, + "learning_rate": 5.746027089233514e-06, + "loss": 0.0066, + "step": 39738 + }, + { + "epoch": 5.21, + "learning_rate": 5.745547147473431e-06, + "loss": 0.0082, + "step": 39739 + }, + { + "epoch": 5.21, + "learning_rate": 5.745067217678768e-06, + "loss": 0.0089, + "step": 39740 + }, + { + "epoch": 5.21, + "learning_rate": 5.744587299850873e-06, + "loss": 0.0103, + "step": 39741 + }, + { + "epoch": 5.21, + "learning_rate": 5.744107393991104e-06, + "loss": 0.0094, + "step": 39742 + }, + { + "epoch": 5.21, + "learning_rate": 5.743627500100806e-06, + "loss": 0.0072, + "step": 39743 + }, + { + "epoch": 5.21, + "learning_rate": 5.743147618181329e-06, + "loss": 0.0075, + "step": 39744 + }, + { + "epoch": 5.21, + "learning_rate": 5.742667748234019e-06, + "loss": 0.0065, + "step": 39745 + }, + { + "epoch": 5.21, + "learning_rate": 5.7421878902602294e-06, + "loss": 0.0078, + "step": 39746 + }, + { + "epoch": 5.21, + "learning_rate": 5.7417080442613134e-06, + "loss": 0.0079, + "step": 39747 + }, + { + "epoch": 5.21, + "learning_rate": 5.741228210238617e-06, + "loss": 0.008, + "step": 39748 + }, + { + "epoch": 5.21, + "learning_rate": 5.740748388193484e-06, + "loss": 0.0102, + "step": 39749 + }, + { + "epoch": 5.21, + "learning_rate": 5.740268578127275e-06, + "loss": 0.0092, + "step": 39750 + }, + { + "epoch": 5.21, + "learning_rate": 5.7397887800413336e-06, + "loss": 0.0087, + "step": 39751 + }, + { + "epoch": 5.21, + "learning_rate": 5.739308993937009e-06, + "loss": 0.008, + "step": 39752 + }, + { + "epoch": 5.21, + "learning_rate": 5.738829219815647e-06, + "loss": 0.0077, + "step": 39753 + }, + { + "epoch": 5.21, + "learning_rate": 5.738349457678605e-06, + "loss": 0.0074, + "step": 39754 + }, + { + "epoch": 5.21, + "learning_rate": 5.737869707527228e-06, + "loss": 0.0065, + "step": 39755 + }, + { + "epoch": 5.21, + "learning_rate": 5.737389969362861e-06, + "loss": 0.0084, + "step": 39756 + }, + { + "epoch": 5.21, + "learning_rate": 5.736910243186858e-06, + "loss": 0.0116, + "step": 39757 + }, + { + "epoch": 5.21, + "learning_rate": 5.736430529000572e-06, + "loss": 0.0065, + "step": 39758 + }, + { + "epoch": 5.21, + "learning_rate": 5.735950826805349e-06, + "loss": 0.0084, + "step": 39759 + }, + { + "epoch": 5.21, + "learning_rate": 5.735471136602536e-06, + "loss": 0.0086, + "step": 39760 + }, + { + "epoch": 5.21, + "learning_rate": 5.734991458393479e-06, + "loss": 0.0085, + "step": 39761 + }, + { + "epoch": 5.21, + "learning_rate": 5.734511792179538e-06, + "loss": 0.0086, + "step": 39762 + }, + { + "epoch": 5.21, + "learning_rate": 5.734032137962054e-06, + "loss": 0.0075, + "step": 39763 + }, + { + "epoch": 5.21, + "learning_rate": 5.733552495742372e-06, + "loss": 0.0086, + "step": 39764 + }, + { + "epoch": 5.21, + "learning_rate": 5.7330728655218514e-06, + "loss": 0.0081, + "step": 39765 + }, + { + "epoch": 5.21, + "learning_rate": 5.7325932473018365e-06, + "loss": 0.0075, + "step": 39766 + }, + { + "epoch": 5.21, + "learning_rate": 5.732113641083672e-06, + "loss": 0.0088, + "step": 39767 + }, + { + "epoch": 5.21, + "learning_rate": 5.731634046868714e-06, + "loss": 0.009, + "step": 39768 + }, + { + "epoch": 5.21, + "learning_rate": 5.731154464658304e-06, + "loss": 0.0109, + "step": 39769 + }, + { + "epoch": 5.21, + "learning_rate": 5.7306748944537986e-06, + "loss": 0.0072, + "step": 39770 + }, + { + "epoch": 5.21, + "learning_rate": 5.730195336256543e-06, + "loss": 0.0086, + "step": 39771 + }, + { + "epoch": 5.21, + "learning_rate": 5.7297157900678815e-06, + "loss": 0.0068, + "step": 39772 + }, + { + "epoch": 5.21, + "learning_rate": 5.729236255889171e-06, + "loss": 0.0087, + "step": 39773 + }, + { + "epoch": 5.21, + "learning_rate": 5.7287567337217565e-06, + "loss": 0.0058, + "step": 39774 + }, + { + "epoch": 5.21, + "learning_rate": 5.728277223566985e-06, + "loss": 0.0053, + "step": 39775 + }, + { + "epoch": 5.21, + "learning_rate": 5.727797725426203e-06, + "loss": 0.0067, + "step": 39776 + }, + { + "epoch": 5.21, + "learning_rate": 5.727318239300768e-06, + "loss": 0.0097, + "step": 39777 + }, + { + "epoch": 5.21, + "learning_rate": 5.726838765192016e-06, + "loss": 0.0074, + "step": 39778 + }, + { + "epoch": 5.21, + "learning_rate": 5.726359303101309e-06, + "loss": 0.0066, + "step": 39779 + }, + { + "epoch": 5.21, + "learning_rate": 5.7258798530299835e-06, + "loss": 0.0049, + "step": 39780 + }, + { + "epoch": 5.21, + "learning_rate": 5.725400414979397e-06, + "loss": 0.0075, + "step": 39781 + }, + { + "epoch": 5.21, + "learning_rate": 5.724920988950895e-06, + "loss": 0.0077, + "step": 39782 + }, + { + "epoch": 5.21, + "learning_rate": 5.724441574945825e-06, + "loss": 0.0051, + "step": 39783 + }, + { + "epoch": 5.21, + "learning_rate": 5.723962172965531e-06, + "loss": 0.0094, + "step": 39784 + }, + { + "epoch": 5.21, + "learning_rate": 5.72348278301137e-06, + "loss": 0.0061, + "step": 39785 + }, + { + "epoch": 5.21, + "learning_rate": 5.723003405084686e-06, + "loss": 0.0076, + "step": 39786 + }, + { + "epoch": 5.21, + "learning_rate": 5.722524039186823e-06, + "loss": 0.0086, + "step": 39787 + }, + { + "epoch": 5.21, + "learning_rate": 5.722044685319137e-06, + "loss": 0.0091, + "step": 39788 + }, + { + "epoch": 5.21, + "learning_rate": 5.721565343482969e-06, + "loss": 0.0066, + "step": 39789 + }, + { + "epoch": 5.21, + "learning_rate": 5.7210860136796755e-06, + "loss": 0.0067, + "step": 39790 + }, + { + "epoch": 5.21, + "learning_rate": 5.720606695910594e-06, + "loss": 0.0091, + "step": 39791 + }, + { + "epoch": 5.21, + "learning_rate": 5.720127390177083e-06, + "loss": 0.0065, + "step": 39792 + }, + { + "epoch": 5.21, + "learning_rate": 5.719648096480486e-06, + "loss": 0.0059, + "step": 39793 + }, + { + "epoch": 5.21, + "learning_rate": 5.7191688148221505e-06, + "loss": 0.0063, + "step": 39794 + }, + { + "epoch": 5.21, + "learning_rate": 5.7186895452034205e-06, + "loss": 0.0074, + "step": 39795 + }, + { + "epoch": 5.21, + "learning_rate": 5.718210287625652e-06, + "loss": 0.0061, + "step": 39796 + }, + { + "epoch": 5.21, + "learning_rate": 5.71773104209019e-06, + "loss": 0.0091, + "step": 39797 + }, + { + "epoch": 5.21, + "learning_rate": 5.7172518085983805e-06, + "loss": 0.009, + "step": 39798 + }, + { + "epoch": 5.21, + "learning_rate": 5.716772587151569e-06, + "loss": 0.0073, + "step": 39799 + }, + { + "epoch": 5.21, + "learning_rate": 5.716293377751106e-06, + "loss": 0.0065, + "step": 39800 + }, + { + "epoch": 5.21, + "learning_rate": 5.715814180398345e-06, + "loss": 0.0077, + "step": 39801 + }, + { + "epoch": 5.21, + "learning_rate": 5.715334995094626e-06, + "loss": 0.0071, + "step": 39802 + }, + { + "epoch": 5.21, + "learning_rate": 5.714855821841297e-06, + "loss": 0.0083, + "step": 39803 + }, + { + "epoch": 5.21, + "learning_rate": 5.714376660639711e-06, + "loss": 0.0098, + "step": 39804 + }, + { + "epoch": 5.21, + "learning_rate": 5.7138975114912135e-06, + "loss": 0.0092, + "step": 39805 + }, + { + "epoch": 5.21, + "learning_rate": 5.713418374397146e-06, + "loss": 0.0113, + "step": 39806 + }, + { + "epoch": 5.21, + "learning_rate": 5.712939249358865e-06, + "loss": 0.0051, + "step": 39807 + }, + { + "epoch": 5.21, + "learning_rate": 5.712460136377714e-06, + "loss": 0.0096, + "step": 39808 + }, + { + "epoch": 5.21, + "learning_rate": 5.711981035455041e-06, + "loss": 0.0083, + "step": 39809 + }, + { + "epoch": 5.21, + "learning_rate": 5.71150194659219e-06, + "loss": 0.0085, + "step": 39810 + }, + { + "epoch": 5.21, + "learning_rate": 5.71102286979051e-06, + "loss": 0.0076, + "step": 39811 + }, + { + "epoch": 5.22, + "learning_rate": 5.710543805051354e-06, + "loss": 0.0074, + "step": 39812 + }, + { + "epoch": 5.22, + "learning_rate": 5.7100647523760655e-06, + "loss": 0.0102, + "step": 39813 + }, + { + "epoch": 5.22, + "learning_rate": 5.709585711765986e-06, + "loss": 0.0068, + "step": 39814 + }, + { + "epoch": 5.22, + "learning_rate": 5.709106683222474e-06, + "loss": 0.0099, + "step": 39815 + }, + { + "epoch": 5.22, + "learning_rate": 5.70862766674687e-06, + "loss": 0.0079, + "step": 39816 + }, + { + "epoch": 5.22, + "learning_rate": 5.708148662340523e-06, + "loss": 0.0062, + "step": 39817 + }, + { + "epoch": 5.22, + "learning_rate": 5.707669670004774e-06, + "loss": 0.0094, + "step": 39818 + }, + { + "epoch": 5.22, + "learning_rate": 5.70719068974098e-06, + "loss": 0.0071, + "step": 39819 + }, + { + "epoch": 5.22, + "learning_rate": 5.706711721550484e-06, + "loss": 0.0074, + "step": 39820 + }, + { + "epoch": 5.22, + "learning_rate": 5.7062327654346285e-06, + "loss": 0.0071, + "step": 39821 + }, + { + "epoch": 5.22, + "learning_rate": 5.705753821394765e-06, + "loss": 0.0075, + "step": 39822 + }, + { + "epoch": 5.22, + "learning_rate": 5.705274889432244e-06, + "loss": 0.0065, + "step": 39823 + }, + { + "epoch": 5.22, + "learning_rate": 5.704795969548408e-06, + "loss": 0.0095, + "step": 39824 + }, + { + "epoch": 5.22, + "learning_rate": 5.704317061744605e-06, + "loss": 0.0076, + "step": 39825 + }, + { + "epoch": 5.22, + "learning_rate": 5.703838166022178e-06, + "loss": 0.0084, + "step": 39826 + }, + { + "epoch": 5.22, + "learning_rate": 5.703359282382481e-06, + "loss": 0.0094, + "step": 39827 + }, + { + "epoch": 5.22, + "learning_rate": 5.7028804108268566e-06, + "loss": 0.0089, + "step": 39828 + }, + { + "epoch": 5.22, + "learning_rate": 5.702401551356649e-06, + "loss": 0.0087, + "step": 39829 + }, + { + "epoch": 5.22, + "learning_rate": 5.701922703973213e-06, + "loss": 0.0098, + "step": 39830 + }, + { + "epoch": 5.22, + "learning_rate": 5.701443868677889e-06, + "loss": 0.0054, + "step": 39831 + }, + { + "epoch": 5.22, + "learning_rate": 5.700965045472021e-06, + "loss": 0.0075, + "step": 39832 + }, + { + "epoch": 5.22, + "learning_rate": 5.700486234356966e-06, + "loss": 0.0086, + "step": 39833 + }, + { + "epoch": 5.22, + "learning_rate": 5.700007435334058e-06, + "loss": 0.0067, + "step": 39834 + }, + { + "epoch": 5.22, + "learning_rate": 5.6995286484046556e-06, + "loss": 0.0092, + "step": 39835 + }, + { + "epoch": 5.22, + "learning_rate": 5.6990498735701e-06, + "loss": 0.0065, + "step": 39836 + }, + { + "epoch": 5.22, + "learning_rate": 5.698571110831732e-06, + "loss": 0.0088, + "step": 39837 + }, + { + "epoch": 5.22, + "learning_rate": 5.6980923601909085e-06, + "loss": 0.0122, + "step": 39838 + }, + { + "epoch": 5.22, + "learning_rate": 5.697613621648972e-06, + "loss": 0.0083, + "step": 39839 + }, + { + "epoch": 5.22, + "learning_rate": 5.6971348952072665e-06, + "loss": 0.0076, + "step": 39840 + }, + { + "epoch": 5.22, + "learning_rate": 5.696656180867137e-06, + "loss": 0.007, + "step": 39841 + }, + { + "epoch": 5.22, + "learning_rate": 5.696177478629937e-06, + "loss": 0.0071, + "step": 39842 + }, + { + "epoch": 5.22, + "learning_rate": 5.695698788497004e-06, + "loss": 0.0062, + "step": 39843 + }, + { + "epoch": 5.22, + "learning_rate": 5.695220110469695e-06, + "loss": 0.0086, + "step": 39844 + }, + { + "epoch": 5.22, + "learning_rate": 5.694741444549343e-06, + "loss": 0.0068, + "step": 39845 + }, + { + "epoch": 5.22, + "learning_rate": 5.694262790737307e-06, + "loss": 0.0093, + "step": 39846 + }, + { + "epoch": 5.22, + "learning_rate": 5.693784149034928e-06, + "loss": 0.0074, + "step": 39847 + }, + { + "epoch": 5.22, + "learning_rate": 5.69330551944355e-06, + "loss": 0.0061, + "step": 39848 + }, + { + "epoch": 5.22, + "learning_rate": 5.692826901964519e-06, + "loss": 0.007, + "step": 39849 + }, + { + "epoch": 5.22, + "learning_rate": 5.692348296599186e-06, + "loss": 0.0087, + "step": 39850 + }, + { + "epoch": 5.22, + "learning_rate": 5.691869703348895e-06, + "loss": 0.0084, + "step": 39851 + }, + { + "epoch": 5.22, + "learning_rate": 5.691391122214986e-06, + "loss": 0.0066, + "step": 39852 + }, + { + "epoch": 5.22, + "learning_rate": 5.690912553198814e-06, + "loss": 0.0084, + "step": 39853 + }, + { + "epoch": 5.22, + "learning_rate": 5.690433996301717e-06, + "loss": 0.01, + "step": 39854 + }, + { + "epoch": 5.22, + "learning_rate": 5.689955451525049e-06, + "loss": 0.0098, + "step": 39855 + }, + { + "epoch": 5.22, + "learning_rate": 5.689476918870148e-06, + "loss": 0.0093, + "step": 39856 + }, + { + "epoch": 5.22, + "learning_rate": 5.688998398338368e-06, + "loss": 0.0079, + "step": 39857 + }, + { + "epoch": 5.22, + "learning_rate": 5.6885198899310504e-06, + "loss": 0.0053, + "step": 39858 + }, + { + "epoch": 5.22, + "learning_rate": 5.68804139364954e-06, + "loss": 0.007, + "step": 39859 + }, + { + "epoch": 5.22, + "learning_rate": 5.68756290949518e-06, + "loss": 0.0095, + "step": 39860 + }, + { + "epoch": 5.22, + "learning_rate": 5.6870844374693235e-06, + "loss": 0.0068, + "step": 39861 + }, + { + "epoch": 5.22, + "learning_rate": 5.686605977573313e-06, + "loss": 0.0055, + "step": 39862 + }, + { + "epoch": 5.22, + "learning_rate": 5.686127529808493e-06, + "loss": 0.0097, + "step": 39863 + }, + { + "epoch": 5.22, + "learning_rate": 5.6856490941762055e-06, + "loss": 0.0077, + "step": 39864 + }, + { + "epoch": 5.22, + "learning_rate": 5.6851706706778e-06, + "loss": 0.0122, + "step": 39865 + }, + { + "epoch": 5.22, + "learning_rate": 5.684692259314627e-06, + "loss": 0.0085, + "step": 39866 + }, + { + "epoch": 5.22, + "learning_rate": 5.684213860088026e-06, + "loss": 0.0094, + "step": 39867 + }, + { + "epoch": 5.22, + "learning_rate": 5.683735472999341e-06, + "loss": 0.0066, + "step": 39868 + }, + { + "epoch": 5.22, + "learning_rate": 5.683257098049924e-06, + "loss": 0.0084, + "step": 39869 + }, + { + "epoch": 5.22, + "learning_rate": 5.682778735241116e-06, + "loss": 0.0068, + "step": 39870 + }, + { + "epoch": 5.22, + "learning_rate": 5.682300384574258e-06, + "loss": 0.0113, + "step": 39871 + }, + { + "epoch": 5.22, + "learning_rate": 5.681822046050707e-06, + "loss": 0.0066, + "step": 39872 + }, + { + "epoch": 5.22, + "learning_rate": 5.681343719671799e-06, + "loss": 0.0074, + "step": 39873 + }, + { + "epoch": 5.22, + "learning_rate": 5.680865405438882e-06, + "loss": 0.0064, + "step": 39874 + }, + { + "epoch": 5.22, + "learning_rate": 5.680387103353298e-06, + "loss": 0.0072, + "step": 39875 + }, + { + "epoch": 5.22, + "learning_rate": 5.679908813416395e-06, + "loss": 0.007, + "step": 39876 + }, + { + "epoch": 5.22, + "learning_rate": 5.679430535629522e-06, + "loss": 0.0076, + "step": 39877 + }, + { + "epoch": 5.22, + "learning_rate": 5.67895226999402e-06, + "loss": 0.0065, + "step": 39878 + }, + { + "epoch": 5.22, + "learning_rate": 5.678474016511233e-06, + "loss": 0.0055, + "step": 39879 + }, + { + "epoch": 5.22, + "learning_rate": 5.6779957751825095e-06, + "loss": 0.0084, + "step": 39880 + }, + { + "epoch": 5.22, + "learning_rate": 5.677517546009194e-06, + "loss": 0.0067, + "step": 39881 + }, + { + "epoch": 5.22, + "learning_rate": 5.677039328992628e-06, + "loss": 0.0074, + "step": 39882 + }, + { + "epoch": 5.22, + "learning_rate": 5.6765611241341565e-06, + "loss": 0.0064, + "step": 39883 + }, + { + "epoch": 5.22, + "learning_rate": 5.676082931435132e-06, + "loss": 0.0051, + "step": 39884 + }, + { + "epoch": 5.22, + "learning_rate": 5.675604750896892e-06, + "loss": 0.0061, + "step": 39885 + }, + { + "epoch": 5.22, + "learning_rate": 5.675126582520779e-06, + "loss": 0.0066, + "step": 39886 + }, + { + "epoch": 5.22, + "learning_rate": 5.674648426308142e-06, + "loss": 0.0057, + "step": 39887 + }, + { + "epoch": 5.23, + "learning_rate": 5.6741702822603316e-06, + "loss": 0.0069, + "step": 39888 + }, + { + "epoch": 5.23, + "learning_rate": 5.6736921503786865e-06, + "loss": 0.0056, + "step": 39889 + }, + { + "epoch": 5.23, + "learning_rate": 5.673214030664551e-06, + "loss": 0.0075, + "step": 39890 + }, + { + "epoch": 5.23, + "learning_rate": 5.672735923119268e-06, + "loss": 0.0082, + "step": 39891 + }, + { + "epoch": 5.23, + "learning_rate": 5.672257827744187e-06, + "loss": 0.007, + "step": 39892 + }, + { + "epoch": 5.23, + "learning_rate": 5.671779744540652e-06, + "loss": 0.01, + "step": 39893 + }, + { + "epoch": 5.23, + "learning_rate": 5.671301673510001e-06, + "loss": 0.0073, + "step": 39894 + }, + { + "epoch": 5.23, + "learning_rate": 5.670823614653588e-06, + "loss": 0.012, + "step": 39895 + }, + { + "epoch": 5.23, + "learning_rate": 5.6703455679727525e-06, + "loss": 0.0079, + "step": 39896 + }, + { + "epoch": 5.23, + "learning_rate": 5.66986753346884e-06, + "loss": 0.0082, + "step": 39897 + }, + { + "epoch": 5.23, + "learning_rate": 5.669389511143191e-06, + "loss": 0.0093, + "step": 39898 + }, + { + "epoch": 5.23, + "learning_rate": 5.668911500997152e-06, + "loss": 0.0069, + "step": 39899 + }, + { + "epoch": 5.23, + "learning_rate": 5.668433503032074e-06, + "loss": 0.0067, + "step": 39900 + }, + { + "epoch": 5.23, + "learning_rate": 5.6679555172492954e-06, + "loss": 0.0071, + "step": 39901 + }, + { + "epoch": 5.23, + "learning_rate": 5.667477543650157e-06, + "loss": 0.0073, + "step": 39902 + }, + { + "epoch": 5.23, + "learning_rate": 5.666999582236013e-06, + "loss": 0.0089, + "step": 39903 + }, + { + "epoch": 5.23, + "learning_rate": 5.666521633008199e-06, + "loss": 0.0072, + "step": 39904 + }, + { + "epoch": 5.23, + "learning_rate": 5.666043695968064e-06, + "loss": 0.007, + "step": 39905 + }, + { + "epoch": 5.23, + "learning_rate": 5.665565771116946e-06, + "loss": 0.0058, + "step": 39906 + }, + { + "epoch": 5.23, + "learning_rate": 5.6650878584561975e-06, + "loss": 0.0086, + "step": 39907 + }, + { + "epoch": 5.23, + "learning_rate": 5.664609957987158e-06, + "loss": 0.0048, + "step": 39908 + }, + { + "epoch": 5.23, + "learning_rate": 5.664132069711169e-06, + "loss": 0.0085, + "step": 39909 + }, + { + "epoch": 5.23, + "learning_rate": 5.663654193629577e-06, + "loss": 0.0053, + "step": 39910 + }, + { + "epoch": 5.23, + "learning_rate": 5.663176329743732e-06, + "loss": 0.0099, + "step": 39911 + }, + { + "epoch": 5.23, + "learning_rate": 5.662698478054971e-06, + "loss": 0.0081, + "step": 39912 + }, + { + "epoch": 5.23, + "learning_rate": 5.662220638564639e-06, + "loss": 0.0079, + "step": 39913 + }, + { + "epoch": 5.23, + "learning_rate": 5.661742811274078e-06, + "loss": 0.0057, + "step": 39914 + }, + { + "epoch": 5.23, + "learning_rate": 5.661264996184639e-06, + "loss": 0.0068, + "step": 39915 + }, + { + "epoch": 5.23, + "learning_rate": 5.66078719329766e-06, + "loss": 0.0073, + "step": 39916 + }, + { + "epoch": 5.23, + "learning_rate": 5.660309402614481e-06, + "loss": 0.0059, + "step": 39917 + }, + { + "epoch": 5.23, + "learning_rate": 5.659831624136456e-06, + "loss": 0.0087, + "step": 39918 + }, + { + "epoch": 5.23, + "learning_rate": 5.659353857864924e-06, + "loss": 0.0073, + "step": 39919 + }, + { + "epoch": 5.23, + "learning_rate": 5.658876103801223e-06, + "loss": 0.008, + "step": 39920 + }, + { + "epoch": 5.23, + "learning_rate": 5.658398361946702e-06, + "loss": 0.0069, + "step": 39921 + }, + { + "epoch": 5.23, + "learning_rate": 5.657920632302708e-06, + "loss": 0.0083, + "step": 39922 + }, + { + "epoch": 5.23, + "learning_rate": 5.657442914870581e-06, + "loss": 0.01, + "step": 39923 + }, + { + "epoch": 5.23, + "learning_rate": 5.656965209651666e-06, + "loss": 0.0092, + "step": 39924 + }, + { + "epoch": 5.23, + "learning_rate": 5.6564875166473e-06, + "loss": 0.0106, + "step": 39925 + }, + { + "epoch": 5.23, + "learning_rate": 5.6560098358588354e-06, + "loss": 0.0063, + "step": 39926 + }, + { + "epoch": 5.23, + "learning_rate": 5.6555321672876115e-06, + "loss": 0.0065, + "step": 39927 + }, + { + "epoch": 5.23, + "learning_rate": 5.655054510934973e-06, + "loss": 0.0081, + "step": 39928 + }, + { + "epoch": 5.23, + "learning_rate": 5.654576866802257e-06, + "loss": 0.0089, + "step": 39929 + }, + { + "epoch": 5.23, + "learning_rate": 5.654099234890817e-06, + "loss": 0.0072, + "step": 39930 + }, + { + "epoch": 5.23, + "learning_rate": 5.653621615201988e-06, + "loss": 0.0075, + "step": 39931 + }, + { + "epoch": 5.23, + "learning_rate": 5.653144007737121e-06, + "loss": 0.0067, + "step": 39932 + }, + { + "epoch": 5.23, + "learning_rate": 5.6526664124975495e-06, + "loss": 0.0096, + "step": 39933 + }, + { + "epoch": 5.23, + "learning_rate": 5.652188829484627e-06, + "loss": 0.0055, + "step": 39934 + }, + { + "epoch": 5.23, + "learning_rate": 5.651711258699693e-06, + "loss": 0.0067, + "step": 39935 + }, + { + "epoch": 5.23, + "learning_rate": 5.651233700144083e-06, + "loss": 0.0077, + "step": 39936 + }, + { + "epoch": 5.23, + "learning_rate": 5.650756153819153e-06, + "loss": 0.0078, + "step": 39937 + }, + { + "epoch": 5.23, + "learning_rate": 5.650278619726237e-06, + "loss": 0.0082, + "step": 39938 + }, + { + "epoch": 5.23, + "learning_rate": 5.649801097866683e-06, + "loss": 0.0082, + "step": 39939 + }, + { + "epoch": 5.23, + "learning_rate": 5.649323588241828e-06, + "loss": 0.0081, + "step": 39940 + }, + { + "epoch": 5.23, + "learning_rate": 5.648846090853017e-06, + "loss": 0.0094, + "step": 39941 + }, + { + "epoch": 5.23, + "learning_rate": 5.6483686057015994e-06, + "loss": 0.0095, + "step": 39942 + }, + { + "epoch": 5.23, + "learning_rate": 5.647891132788914e-06, + "loss": 0.0068, + "step": 39943 + }, + { + "epoch": 5.23, + "learning_rate": 5.6474136721162984e-06, + "loss": 0.0066, + "step": 39944 + }, + { + "epoch": 5.23, + "learning_rate": 5.646936223685105e-06, + "loss": 0.0094, + "step": 39945 + }, + { + "epoch": 5.23, + "learning_rate": 5.64645878749667e-06, + "loss": 0.0057, + "step": 39946 + }, + { + "epoch": 5.23, + "learning_rate": 5.645981363552338e-06, + "loss": 0.0079, + "step": 39947 + }, + { + "epoch": 5.23, + "learning_rate": 5.645503951853448e-06, + "loss": 0.0081, + "step": 39948 + }, + { + "epoch": 5.23, + "learning_rate": 5.645026552401351e-06, + "loss": 0.0072, + "step": 39949 + }, + { + "epoch": 5.23, + "learning_rate": 5.644549165197385e-06, + "loss": 0.009, + "step": 39950 + }, + { + "epoch": 5.23, + "learning_rate": 5.644071790242888e-06, + "loss": 0.0069, + "step": 39951 + }, + { + "epoch": 5.23, + "learning_rate": 5.6435944275392075e-06, + "loss": 0.0065, + "step": 39952 + }, + { + "epoch": 5.23, + "learning_rate": 5.643117077087688e-06, + "loss": 0.0067, + "step": 39953 + }, + { + "epoch": 5.23, + "learning_rate": 5.64263973888967e-06, + "loss": 0.0091, + "step": 39954 + }, + { + "epoch": 5.23, + "learning_rate": 5.642162412946497e-06, + "loss": 0.0075, + "step": 39955 + }, + { + "epoch": 5.23, + "learning_rate": 5.641685099259506e-06, + "loss": 0.0105, + "step": 39956 + }, + { + "epoch": 5.23, + "learning_rate": 5.641207797830049e-06, + "loss": 0.009, + "step": 39957 + }, + { + "epoch": 5.23, + "learning_rate": 5.64073050865946e-06, + "loss": 0.0079, + "step": 39958 + }, + { + "epoch": 5.23, + "learning_rate": 5.640253231749082e-06, + "loss": 0.013, + "step": 39959 + }, + { + "epoch": 5.23, + "learning_rate": 5.639775967100263e-06, + "loss": 0.0081, + "step": 39960 + }, + { + "epoch": 5.23, + "learning_rate": 5.6392987147143416e-06, + "loss": 0.0118, + "step": 39961 + }, + { + "epoch": 5.23, + "learning_rate": 5.638821474592662e-06, + "loss": 0.0084, + "step": 39962 + }, + { + "epoch": 5.23, + "learning_rate": 5.6383442467365575e-06, + "loss": 0.0089, + "step": 39963 + }, + { + "epoch": 5.24, + "learning_rate": 5.6378670311473806e-06, + "loss": 0.0066, + "step": 39964 + }, + { + "epoch": 5.24, + "learning_rate": 5.637389827826474e-06, + "loss": 0.0048, + "step": 39965 + }, + { + "epoch": 5.24, + "learning_rate": 5.636912636775175e-06, + "loss": 0.0071, + "step": 39966 + }, + { + "epoch": 5.24, + "learning_rate": 5.636435457994823e-06, + "loss": 0.0077, + "step": 39967 + }, + { + "epoch": 5.24, + "learning_rate": 5.635958291486768e-06, + "loss": 0.0098, + "step": 39968 + }, + { + "epoch": 5.24, + "learning_rate": 5.635481137252349e-06, + "loss": 0.0063, + "step": 39969 + }, + { + "epoch": 5.24, + "learning_rate": 5.635003995292906e-06, + "loss": 0.0083, + "step": 39970 + }, + { + "epoch": 5.24, + "learning_rate": 5.634526865609778e-06, + "loss": 0.0092, + "step": 39971 + }, + { + "epoch": 5.24, + "learning_rate": 5.634049748204315e-06, + "loss": 0.0066, + "step": 39972 + }, + { + "epoch": 5.24, + "learning_rate": 5.633572643077854e-06, + "loss": 0.0087, + "step": 39973 + }, + { + "epoch": 5.24, + "learning_rate": 5.6330955502317335e-06, + "loss": 0.0075, + "step": 39974 + }, + { + "epoch": 5.24, + "learning_rate": 5.632618469667299e-06, + "loss": 0.0077, + "step": 39975 + }, + { + "epoch": 5.24, + "learning_rate": 5.632141401385898e-06, + "loss": 0.0098, + "step": 39976 + }, + { + "epoch": 5.24, + "learning_rate": 5.631664345388864e-06, + "loss": 0.0048, + "step": 39977 + }, + { + "epoch": 5.24, + "learning_rate": 5.631187301677543e-06, + "loss": 0.0063, + "step": 39978 + }, + { + "epoch": 5.24, + "learning_rate": 5.630710270253271e-06, + "loss": 0.0085, + "step": 39979 + }, + { + "epoch": 5.24, + "learning_rate": 5.630233251117398e-06, + "loss": 0.0075, + "step": 39980 + }, + { + "epoch": 5.24, + "learning_rate": 5.6297562442712606e-06, + "loss": 0.0092, + "step": 39981 + }, + { + "epoch": 5.24, + "learning_rate": 5.629279249716198e-06, + "loss": 0.0077, + "step": 39982 + }, + { + "epoch": 5.24, + "learning_rate": 5.628802267453558e-06, + "loss": 0.0061, + "step": 39983 + }, + { + "epoch": 5.24, + "learning_rate": 5.628325297484678e-06, + "loss": 0.0067, + "step": 39984 + }, + { + "epoch": 5.24, + "learning_rate": 5.6278483398108975e-06, + "loss": 0.006, + "step": 39985 + }, + { + "epoch": 5.24, + "learning_rate": 5.6273713944335605e-06, + "loss": 0.008, + "step": 39986 + }, + { + "epoch": 5.24, + "learning_rate": 5.626894461354014e-06, + "loss": 0.0101, + "step": 39987 + }, + { + "epoch": 5.24, + "learning_rate": 5.626417540573592e-06, + "loss": 0.0074, + "step": 39988 + }, + { + "epoch": 5.24, + "learning_rate": 5.6259406320936384e-06, + "loss": 0.006, + "step": 39989 + }, + { + "epoch": 5.24, + "learning_rate": 5.625463735915489e-06, + "loss": 0.0099, + "step": 39990 + }, + { + "epoch": 5.24, + "learning_rate": 5.624986852040495e-06, + "loss": 0.0062, + "step": 39991 + }, + { + "epoch": 5.24, + "learning_rate": 5.624509980469992e-06, + "loss": 0.0062, + "step": 39992 + }, + { + "epoch": 5.24, + "learning_rate": 5.624033121205321e-06, + "loss": 0.0053, + "step": 39993 + }, + { + "epoch": 5.24, + "learning_rate": 5.623556274247821e-06, + "loss": 0.005, + "step": 39994 + }, + { + "epoch": 5.24, + "learning_rate": 5.62307943959884e-06, + "loss": 0.0055, + "step": 39995 + }, + { + "epoch": 5.24, + "learning_rate": 5.62260261725971e-06, + "loss": 0.0078, + "step": 39996 + }, + { + "epoch": 5.24, + "learning_rate": 5.622125807231782e-06, + "loss": 0.0077, + "step": 39997 + }, + { + "epoch": 5.24, + "learning_rate": 5.621649009516388e-06, + "loss": 0.0074, + "step": 39998 + }, + { + "epoch": 5.24, + "learning_rate": 5.621172224114876e-06, + "loss": 0.0063, + "step": 39999 + }, + { + "epoch": 5.24, + "learning_rate": 5.620695451028584e-06, + "loss": 0.0065, + "step": 40000 + }, + { + "epoch": 5.24, + "learning_rate": 5.62021869025885e-06, + "loss": 0.0081, + "step": 40001 + }, + { + "epoch": 5.24, + "learning_rate": 5.619741941807021e-06, + "loss": 0.0069, + "step": 40002 + }, + { + "epoch": 5.24, + "learning_rate": 5.619265205674434e-06, + "loss": 0.0075, + "step": 40003 + }, + { + "epoch": 5.24, + "learning_rate": 5.6187884818624315e-06, + "loss": 0.006, + "step": 40004 + }, + { + "epoch": 5.24, + "learning_rate": 5.618311770372349e-06, + "loss": 0.0071, + "step": 40005 + }, + { + "epoch": 5.24, + "learning_rate": 5.617835071205534e-06, + "loss": 0.0053, + "step": 40006 + }, + { + "epoch": 5.24, + "learning_rate": 5.617358384363321e-06, + "loss": 0.0066, + "step": 40007 + }, + { + "epoch": 5.24, + "learning_rate": 5.616881709847059e-06, + "loss": 0.0072, + "step": 40008 + }, + { + "epoch": 5.24, + "learning_rate": 5.616405047658079e-06, + "loss": 0.0065, + "step": 40009 + }, + { + "epoch": 5.24, + "learning_rate": 5.615928397797731e-06, + "loss": 0.0073, + "step": 40010 + }, + { + "epoch": 5.24, + "learning_rate": 5.61545176026735e-06, + "loss": 0.0075, + "step": 40011 + }, + { + "epoch": 5.24, + "learning_rate": 5.614975135068278e-06, + "loss": 0.0084, + "step": 40012 + }, + { + "epoch": 5.24, + "learning_rate": 5.61449852220185e-06, + "loss": 0.0076, + "step": 40013 + }, + { + "epoch": 5.24, + "learning_rate": 5.614021921669417e-06, + "loss": 0.0057, + "step": 40014 + }, + { + "epoch": 5.24, + "learning_rate": 5.613545333472312e-06, + "loss": 0.0074, + "step": 40015 + }, + { + "epoch": 5.24, + "learning_rate": 5.613068757611875e-06, + "loss": 0.0089, + "step": 40016 + }, + { + "epoch": 5.24, + "learning_rate": 5.612592194089452e-06, + "loss": 0.0106, + "step": 40017 + }, + { + "epoch": 5.24, + "learning_rate": 5.612115642906375e-06, + "loss": 0.0078, + "step": 40018 + }, + { + "epoch": 5.24, + "learning_rate": 5.611639104063993e-06, + "loss": 0.0116, + "step": 40019 + }, + { + "epoch": 5.24, + "learning_rate": 5.6111625775636435e-06, + "loss": 0.0072, + "step": 40020 + }, + { + "epoch": 5.24, + "learning_rate": 5.610686063406661e-06, + "loss": 0.0053, + "step": 40021 + }, + { + "epoch": 5.24, + "learning_rate": 5.610209561594394e-06, + "loss": 0.0086, + "step": 40022 + }, + { + "epoch": 5.24, + "learning_rate": 5.6097330721281785e-06, + "loss": 0.0073, + "step": 40023 + }, + { + "epoch": 5.24, + "learning_rate": 5.609256595009351e-06, + "loss": 0.0065, + "step": 40024 + }, + { + "epoch": 5.24, + "learning_rate": 5.60878013023926e-06, + "loss": 0.0088, + "step": 40025 + }, + { + "epoch": 5.24, + "learning_rate": 5.608303677819242e-06, + "loss": 0.0062, + "step": 40026 + }, + { + "epoch": 5.24, + "learning_rate": 5.607827237750634e-06, + "loss": 0.008, + "step": 40027 + }, + { + "epoch": 5.24, + "learning_rate": 5.607350810034775e-06, + "loss": 0.0054, + "step": 40028 + }, + { + "epoch": 5.24, + "learning_rate": 5.606874394673007e-06, + "loss": 0.0127, + "step": 40029 + }, + { + "epoch": 5.24, + "learning_rate": 5.606397991666677e-06, + "loss": 0.0073, + "step": 40030 + }, + { + "epoch": 5.24, + "learning_rate": 5.605921601017117e-06, + "loss": 0.0066, + "step": 40031 + }, + { + "epoch": 5.24, + "learning_rate": 5.605445222725665e-06, + "loss": 0.0071, + "step": 40032 + }, + { + "epoch": 5.24, + "learning_rate": 5.604968856793669e-06, + "loss": 0.0083, + "step": 40033 + }, + { + "epoch": 5.24, + "learning_rate": 5.604492503222463e-06, + "loss": 0.0078, + "step": 40034 + }, + { + "epoch": 5.24, + "learning_rate": 5.604016162013387e-06, + "loss": 0.0066, + "step": 40035 + }, + { + "epoch": 5.24, + "learning_rate": 5.603539833167779e-06, + "loss": 0.0089, + "step": 40036 + }, + { + "epoch": 5.24, + "learning_rate": 5.603063516686984e-06, + "loss": 0.0064, + "step": 40037 + }, + { + "epoch": 5.24, + "learning_rate": 5.6025872125723395e-06, + "loss": 0.0095, + "step": 40038 + }, + { + "epoch": 5.24, + "learning_rate": 5.6021109208251786e-06, + "loss": 0.0058, + "step": 40039 + }, + { + "epoch": 5.24, + "learning_rate": 5.601634641446849e-06, + "loss": 0.0086, + "step": 40040 + }, + { + "epoch": 5.25, + "learning_rate": 5.601158374438691e-06, + "loss": 0.0066, + "step": 40041 + }, + { + "epoch": 5.25, + "learning_rate": 5.60068211980204e-06, + "loss": 0.0074, + "step": 40042 + }, + { + "epoch": 5.25, + "learning_rate": 5.600205877538236e-06, + "loss": 0.0088, + "step": 40043 + }, + { + "epoch": 5.25, + "learning_rate": 5.599729647648616e-06, + "loss": 0.0072, + "step": 40044 + }, + { + "epoch": 5.25, + "learning_rate": 5.5992534301345255e-06, + "loss": 0.0065, + "step": 40045 + }, + { + "epoch": 5.25, + "learning_rate": 5.598777224997301e-06, + "loss": 0.0069, + "step": 40046 + }, + { + "epoch": 5.25, + "learning_rate": 5.598301032238276e-06, + "loss": 0.0058, + "step": 40047 + }, + { + "epoch": 5.25, + "learning_rate": 5.5978248518587996e-06, + "loss": 0.0092, + "step": 40048 + }, + { + "epoch": 5.25, + "learning_rate": 5.597348683860206e-06, + "loss": 0.0084, + "step": 40049 + }, + { + "epoch": 5.25, + "learning_rate": 5.596872528243835e-06, + "loss": 0.0053, + "step": 40050 + }, + { + "epoch": 5.25, + "learning_rate": 5.5963963850110184e-06, + "loss": 0.0059, + "step": 40051 + }, + { + "epoch": 5.25, + "learning_rate": 5.595920254163109e-06, + "loss": 0.0054, + "step": 40052 + }, + { + "epoch": 5.25, + "learning_rate": 5.59544413570144e-06, + "loss": 0.0079, + "step": 40053 + }, + { + "epoch": 5.25, + "learning_rate": 5.5949680296273505e-06, + "loss": 0.0068, + "step": 40054 + }, + { + "epoch": 5.25, + "learning_rate": 5.594491935942172e-06, + "loss": 0.0064, + "step": 40055 + }, + { + "epoch": 5.25, + "learning_rate": 5.594015854647257e-06, + "loss": 0.0074, + "step": 40056 + }, + { + "epoch": 5.25, + "learning_rate": 5.593539785743938e-06, + "loss": 0.0084, + "step": 40057 + }, + { + "epoch": 5.25, + "learning_rate": 5.593063729233548e-06, + "loss": 0.0041, + "step": 40058 + }, + { + "epoch": 5.25, + "learning_rate": 5.592587685117437e-06, + "loss": 0.0056, + "step": 40059 + }, + { + "epoch": 5.25, + "learning_rate": 5.5921116533969375e-06, + "loss": 0.0065, + "step": 40060 + }, + { + "epoch": 5.25, + "learning_rate": 5.591635634073389e-06, + "loss": 0.0067, + "step": 40061 + }, + { + "epoch": 5.25, + "learning_rate": 5.591159627148128e-06, + "loss": 0.0075, + "step": 40062 + }, + { + "epoch": 5.25, + "learning_rate": 5.590683632622496e-06, + "loss": 0.0086, + "step": 40063 + }, + { + "epoch": 5.25, + "learning_rate": 5.590207650497833e-06, + "loss": 0.0078, + "step": 40064 + }, + { + "epoch": 5.25, + "learning_rate": 5.589731680775478e-06, + "loss": 0.0096, + "step": 40065 + }, + { + "epoch": 5.25, + "learning_rate": 5.589255723456764e-06, + "loss": 0.0068, + "step": 40066 + }, + { + "epoch": 5.25, + "learning_rate": 5.588779778543036e-06, + "loss": 0.0064, + "step": 40067 + }, + { + "epoch": 5.25, + "learning_rate": 5.58830384603563e-06, + "loss": 0.0099, + "step": 40068 + }, + { + "epoch": 5.25, + "learning_rate": 5.587827925935885e-06, + "loss": 0.0111, + "step": 40069 + }, + { + "epoch": 5.25, + "learning_rate": 5.587352018245136e-06, + "loss": 0.0075, + "step": 40070 + }, + { + "epoch": 5.25, + "learning_rate": 5.5868761229647276e-06, + "loss": 0.0089, + "step": 40071 + }, + { + "epoch": 5.25, + "learning_rate": 5.5864002400959905e-06, + "loss": 0.0094, + "step": 40072 + }, + { + "epoch": 5.25, + "learning_rate": 5.585924369640272e-06, + "loss": 0.0057, + "step": 40073 + }, + { + "epoch": 5.25, + "learning_rate": 5.585448511598903e-06, + "loss": 0.0059, + "step": 40074 + }, + { + "epoch": 5.25, + "learning_rate": 5.584972665973228e-06, + "loss": 0.0087, + "step": 40075 + }, + { + "epoch": 5.25, + "learning_rate": 5.584496832764584e-06, + "loss": 0.0084, + "step": 40076 + }, + { + "epoch": 5.25, + "learning_rate": 5.584021011974305e-06, + "loss": 0.0074, + "step": 40077 + }, + { + "epoch": 5.25, + "learning_rate": 5.583545203603728e-06, + "loss": 0.0093, + "step": 40078 + }, + { + "epoch": 5.25, + "learning_rate": 5.5830694076542004e-06, + "loss": 0.0077, + "step": 40079 + }, + { + "epoch": 5.25, + "learning_rate": 5.582593624127055e-06, + "loss": 0.0092, + "step": 40080 + }, + { + "epoch": 5.25, + "learning_rate": 5.582117853023625e-06, + "loss": 0.0092, + "step": 40081 + }, + { + "epoch": 5.25, + "learning_rate": 5.581642094345258e-06, + "loss": 0.0088, + "step": 40082 + }, + { + "epoch": 5.25, + "learning_rate": 5.581166348093282e-06, + "loss": 0.0103, + "step": 40083 + }, + { + "epoch": 5.25, + "learning_rate": 5.5806906142690454e-06, + "loss": 0.0059, + "step": 40084 + }, + { + "epoch": 5.25, + "learning_rate": 5.580214892873881e-06, + "loss": 0.0081, + "step": 40085 + }, + { + "epoch": 5.25, + "learning_rate": 5.579739183909122e-06, + "loss": 0.0093, + "step": 40086 + }, + { + "epoch": 5.25, + "learning_rate": 5.579263487376117e-06, + "loss": 0.0082, + "step": 40087 + }, + { + "epoch": 5.25, + "learning_rate": 5.578787803276195e-06, + "loss": 0.006, + "step": 40088 + }, + { + "epoch": 5.25, + "learning_rate": 5.578312131610694e-06, + "loss": 0.0083, + "step": 40089 + }, + { + "epoch": 5.25, + "learning_rate": 5.577836472380959e-06, + "loss": 0.0082, + "step": 40090 + }, + { + "epoch": 5.25, + "learning_rate": 5.577360825588325e-06, + "loss": 0.007, + "step": 40091 + }, + { + "epoch": 5.25, + "learning_rate": 5.576885191234127e-06, + "loss": 0.0067, + "step": 40092 + }, + { + "epoch": 5.25, + "learning_rate": 5.5764095693197e-06, + "loss": 0.0058, + "step": 40093 + }, + { + "epoch": 5.25, + "learning_rate": 5.5759339598463844e-06, + "loss": 0.0095, + "step": 40094 + }, + { + "epoch": 5.25, + "learning_rate": 5.575458362815525e-06, + "loss": 0.007, + "step": 40095 + }, + { + "epoch": 5.25, + "learning_rate": 5.5749827782284515e-06, + "loss": 0.0065, + "step": 40096 + }, + { + "epoch": 5.25, + "learning_rate": 5.5745072060865005e-06, + "loss": 0.0056, + "step": 40097 + }, + { + "epoch": 5.25, + "learning_rate": 5.574031646391016e-06, + "loss": 0.0076, + "step": 40098 + }, + { + "epoch": 5.25, + "learning_rate": 5.573556099143331e-06, + "loss": 0.007, + "step": 40099 + }, + { + "epoch": 5.25, + "learning_rate": 5.573080564344784e-06, + "loss": 0.0082, + "step": 40100 + }, + { + "epoch": 5.25, + "learning_rate": 5.5726050419967085e-06, + "loss": 0.0052, + "step": 40101 + }, + { + "epoch": 5.25, + "learning_rate": 5.57212953210045e-06, + "loss": 0.0075, + "step": 40102 + }, + { + "epoch": 5.25, + "learning_rate": 5.571654034657341e-06, + "loss": 0.0092, + "step": 40103 + }, + { + "epoch": 5.25, + "learning_rate": 5.571178549668714e-06, + "loss": 0.0094, + "step": 40104 + }, + { + "epoch": 5.25, + "learning_rate": 5.570703077135913e-06, + "loss": 0.006, + "step": 40105 + }, + { + "epoch": 5.25, + "learning_rate": 5.570227617060278e-06, + "loss": 0.0059, + "step": 40106 + }, + { + "epoch": 5.25, + "learning_rate": 5.5697521694431414e-06, + "loss": 0.0076, + "step": 40107 + }, + { + "epoch": 5.25, + "learning_rate": 5.569276734285835e-06, + "loss": 0.0071, + "step": 40108 + }, + { + "epoch": 5.25, + "learning_rate": 5.568801311589709e-06, + "loss": 0.0057, + "step": 40109 + }, + { + "epoch": 5.25, + "learning_rate": 5.5683259013560905e-06, + "loss": 0.0065, + "step": 40110 + }, + { + "epoch": 5.25, + "learning_rate": 5.567850503586321e-06, + "loss": 0.0072, + "step": 40111 + }, + { + "epoch": 5.25, + "learning_rate": 5.5673751182817325e-06, + "loss": 0.007, + "step": 40112 + }, + { + "epoch": 5.25, + "learning_rate": 5.566899745443668e-06, + "loss": 0.0074, + "step": 40113 + }, + { + "epoch": 5.25, + "learning_rate": 5.566424385073464e-06, + "loss": 0.0055, + "step": 40114 + }, + { + "epoch": 5.25, + "learning_rate": 5.565949037172454e-06, + "loss": 0.0066, + "step": 40115 + }, + { + "epoch": 5.25, + "learning_rate": 5.565473701741967e-06, + "loss": 0.0082, + "step": 40116 + }, + { + "epoch": 5.26, + "learning_rate": 5.564998378783357e-06, + "loss": 0.0097, + "step": 40117 + }, + { + "epoch": 5.26, + "learning_rate": 5.564523068297956e-06, + "loss": 0.0086, + "step": 40118 + }, + { + "epoch": 5.26, + "learning_rate": 5.564047770287094e-06, + "loss": 0.0066, + "step": 40119 + }, + { + "epoch": 5.26, + "learning_rate": 5.563572484752109e-06, + "loss": 0.0092, + "step": 40120 + }, + { + "epoch": 5.26, + "learning_rate": 5.5630972116943446e-06, + "loss": 0.0078, + "step": 40121 + }, + { + "epoch": 5.26, + "learning_rate": 5.5626219511151325e-06, + "loss": 0.0057, + "step": 40122 + }, + { + "epoch": 5.26, + "learning_rate": 5.562146703015806e-06, + "loss": 0.0077, + "step": 40123 + }, + { + "epoch": 5.26, + "learning_rate": 5.56167146739771e-06, + "loss": 0.0079, + "step": 40124 + }, + { + "epoch": 5.26, + "learning_rate": 5.561196244262176e-06, + "loss": 0.0066, + "step": 40125 + }, + { + "epoch": 5.26, + "learning_rate": 5.560721033610541e-06, + "loss": 0.0085, + "step": 40126 + }, + { + "epoch": 5.26, + "learning_rate": 5.5602458354441376e-06, + "loss": 0.0069, + "step": 40127 + }, + { + "epoch": 5.26, + "learning_rate": 5.559770649764308e-06, + "loss": 0.0066, + "step": 40128 + }, + { + "epoch": 5.26, + "learning_rate": 5.55929547657239e-06, + "loss": 0.0073, + "step": 40129 + }, + { + "epoch": 5.26, + "learning_rate": 5.558820315869717e-06, + "loss": 0.0081, + "step": 40130 + }, + { + "epoch": 5.26, + "learning_rate": 5.558345167657621e-06, + "loss": 0.0079, + "step": 40131 + }, + { + "epoch": 5.26, + "learning_rate": 5.557870031937447e-06, + "loss": 0.0065, + "step": 40132 + }, + { + "epoch": 5.26, + "learning_rate": 5.557394908710526e-06, + "loss": 0.0069, + "step": 40133 + }, + { + "epoch": 5.26, + "learning_rate": 5.556919797978197e-06, + "loss": 0.0084, + "step": 40134 + }, + { + "epoch": 5.26, + "learning_rate": 5.556444699741789e-06, + "loss": 0.01, + "step": 40135 + }, + { + "epoch": 5.26, + "learning_rate": 5.555969614002648e-06, + "loss": 0.0078, + "step": 40136 + }, + { + "epoch": 5.26, + "learning_rate": 5.555494540762105e-06, + "loss": 0.0102, + "step": 40137 + }, + { + "epoch": 5.26, + "learning_rate": 5.555019480021495e-06, + "loss": 0.0075, + "step": 40138 + }, + { + "epoch": 5.26, + "learning_rate": 5.554544431782154e-06, + "loss": 0.0102, + "step": 40139 + }, + { + "epoch": 5.26, + "learning_rate": 5.554069396045425e-06, + "loss": 0.0103, + "step": 40140 + }, + { + "epoch": 5.26, + "learning_rate": 5.553594372812639e-06, + "loss": 0.0073, + "step": 40141 + }, + { + "epoch": 5.26, + "learning_rate": 5.5531193620851336e-06, + "loss": 0.0069, + "step": 40142 + }, + { + "epoch": 5.26, + "learning_rate": 5.552644363864237e-06, + "loss": 0.0084, + "step": 40143 + }, + { + "epoch": 5.26, + "learning_rate": 5.552169378151295e-06, + "loss": 0.0097, + "step": 40144 + }, + { + "epoch": 5.26, + "learning_rate": 5.55169440494764e-06, + "loss": 0.0095, + "step": 40145 + }, + { + "epoch": 5.26, + "learning_rate": 5.5512194442546055e-06, + "loss": 0.0088, + "step": 40146 + }, + { + "epoch": 5.26, + "learning_rate": 5.550744496073532e-06, + "loss": 0.0068, + "step": 40147 + }, + { + "epoch": 5.26, + "learning_rate": 5.550269560405753e-06, + "loss": 0.0066, + "step": 40148 + }, + { + "epoch": 5.26, + "learning_rate": 5.5497946372526e-06, + "loss": 0.0069, + "step": 40149 + }, + { + "epoch": 5.26, + "learning_rate": 5.549319726615416e-06, + "loss": 0.0086, + "step": 40150 + }, + { + "epoch": 5.26, + "learning_rate": 5.54884482849553e-06, + "loss": 0.0074, + "step": 40151 + }, + { + "epoch": 5.26, + "learning_rate": 5.548369942894285e-06, + "loss": 0.0064, + "step": 40152 + }, + { + "epoch": 5.26, + "learning_rate": 5.5478950698130115e-06, + "loss": 0.0085, + "step": 40153 + }, + { + "epoch": 5.26, + "learning_rate": 5.547420209253043e-06, + "loss": 0.0088, + "step": 40154 + }, + { + "epoch": 5.26, + "learning_rate": 5.5469453612157206e-06, + "loss": 0.0061, + "step": 40155 + }, + { + "epoch": 5.26, + "learning_rate": 5.546470525702379e-06, + "loss": 0.0093, + "step": 40156 + }, + { + "epoch": 5.26, + "learning_rate": 5.545995702714351e-06, + "loss": 0.0096, + "step": 40157 + }, + { + "epoch": 5.26, + "learning_rate": 5.5455208922529694e-06, + "loss": 0.0094, + "step": 40158 + }, + { + "epoch": 5.26, + "learning_rate": 5.545046094319575e-06, + "loss": 0.0078, + "step": 40159 + }, + { + "epoch": 5.26, + "learning_rate": 5.544571308915499e-06, + "loss": 0.0065, + "step": 40160 + }, + { + "epoch": 5.26, + "learning_rate": 5.544096536042083e-06, + "loss": 0.01, + "step": 40161 + }, + { + "epoch": 5.26, + "learning_rate": 5.5436217757006534e-06, + "loss": 0.0072, + "step": 40162 + }, + { + "epoch": 5.26, + "learning_rate": 5.543147027892556e-06, + "loss": 0.0048, + "step": 40163 + }, + { + "epoch": 5.26, + "learning_rate": 5.542672292619119e-06, + "loss": 0.0043, + "step": 40164 + }, + { + "epoch": 5.26, + "learning_rate": 5.542197569881679e-06, + "loss": 0.0065, + "step": 40165 + }, + { + "epoch": 5.26, + "learning_rate": 5.541722859681567e-06, + "loss": 0.0067, + "step": 40166 + }, + { + "epoch": 5.26, + "learning_rate": 5.541248162020125e-06, + "loss": 0.006, + "step": 40167 + }, + { + "epoch": 5.26, + "learning_rate": 5.540773476898686e-06, + "loss": 0.0065, + "step": 40168 + }, + { + "epoch": 5.26, + "learning_rate": 5.5402988043185806e-06, + "loss": 0.0091, + "step": 40169 + }, + { + "epoch": 5.26, + "learning_rate": 5.539824144281151e-06, + "loss": 0.0081, + "step": 40170 + }, + { + "epoch": 5.26, + "learning_rate": 5.539349496787723e-06, + "loss": 0.0093, + "step": 40171 + }, + { + "epoch": 5.26, + "learning_rate": 5.538874861839642e-06, + "loss": 0.0083, + "step": 40172 + }, + { + "epoch": 5.26, + "learning_rate": 5.538400239438235e-06, + "loss": 0.0081, + "step": 40173 + }, + { + "epoch": 5.26, + "learning_rate": 5.537925629584844e-06, + "loss": 0.0059, + "step": 40174 + }, + { + "epoch": 5.26, + "learning_rate": 5.537451032280798e-06, + "loss": 0.0092, + "step": 40175 + }, + { + "epoch": 5.26, + "learning_rate": 5.536976447527433e-06, + "loss": 0.0078, + "step": 40176 + }, + { + "epoch": 5.26, + "learning_rate": 5.536501875326081e-06, + "loss": 0.0094, + "step": 40177 + }, + { + "epoch": 5.26, + "learning_rate": 5.536027315678083e-06, + "loss": 0.0071, + "step": 40178 + }, + { + "epoch": 5.26, + "learning_rate": 5.535552768584772e-06, + "loss": 0.005, + "step": 40179 + }, + { + "epoch": 5.26, + "learning_rate": 5.53507823404748e-06, + "loss": 0.0065, + "step": 40180 + }, + { + "epoch": 5.26, + "learning_rate": 5.534603712067539e-06, + "loss": 0.0059, + "step": 40181 + }, + { + "epoch": 5.26, + "learning_rate": 5.534129202646287e-06, + "loss": 0.0076, + "step": 40182 + }, + { + "epoch": 5.26, + "learning_rate": 5.5336547057850636e-06, + "loss": 0.008, + "step": 40183 + }, + { + "epoch": 5.26, + "learning_rate": 5.533180221485197e-06, + "loss": 0.0069, + "step": 40184 + }, + { + "epoch": 5.26, + "learning_rate": 5.53270574974802e-06, + "loss": 0.0085, + "step": 40185 + }, + { + "epoch": 5.26, + "learning_rate": 5.5322312905748744e-06, + "loss": 0.0108, + "step": 40186 + }, + { + "epoch": 5.26, + "learning_rate": 5.53175684396709e-06, + "loss": 0.0077, + "step": 40187 + }, + { + "epoch": 5.26, + "learning_rate": 5.5312824099259975e-06, + "loss": 0.0081, + "step": 40188 + }, + { + "epoch": 5.26, + "learning_rate": 5.53080798845294e-06, + "loss": 0.0072, + "step": 40189 + }, + { + "epoch": 5.26, + "learning_rate": 5.530333579549245e-06, + "loss": 0.0091, + "step": 40190 + }, + { + "epoch": 5.26, + "learning_rate": 5.529859183216251e-06, + "loss": 0.0072, + "step": 40191 + }, + { + "epoch": 5.26, + "learning_rate": 5.529384799455285e-06, + "loss": 0.0078, + "step": 40192 + }, + { + "epoch": 5.26, + "learning_rate": 5.528910428267687e-06, + "loss": 0.0083, + "step": 40193 + }, + { + "epoch": 5.27, + "learning_rate": 5.5284360696547935e-06, + "loss": 0.009, + "step": 40194 + }, + { + "epoch": 5.27, + "learning_rate": 5.527961723617936e-06, + "loss": 0.007, + "step": 40195 + }, + { + "epoch": 5.27, + "learning_rate": 5.527487390158444e-06, + "loss": 0.0093, + "step": 40196 + }, + { + "epoch": 5.27, + "learning_rate": 5.527013069277658e-06, + "loss": 0.0074, + "step": 40197 + }, + { + "epoch": 5.27, + "learning_rate": 5.526538760976911e-06, + "loss": 0.0099, + "step": 40198 + }, + { + "epoch": 5.27, + "learning_rate": 5.526064465257535e-06, + "loss": 0.008, + "step": 40199 + }, + { + "epoch": 5.27, + "learning_rate": 5.525590182120862e-06, + "loss": 0.0096, + "step": 40200 + }, + { + "epoch": 5.27, + "learning_rate": 5.525115911568231e-06, + "loss": 0.0084, + "step": 40201 + }, + { + "epoch": 5.27, + "learning_rate": 5.524641653600973e-06, + "loss": 0.0047, + "step": 40202 + }, + { + "epoch": 5.27, + "learning_rate": 5.524167408220417e-06, + "loss": 0.0085, + "step": 40203 + }, + { + "epoch": 5.27, + "learning_rate": 5.523693175427903e-06, + "loss": 0.0071, + "step": 40204 + }, + { + "epoch": 5.27, + "learning_rate": 5.523218955224768e-06, + "loss": 0.0074, + "step": 40205 + }, + { + "epoch": 5.27, + "learning_rate": 5.522744747612341e-06, + "loss": 0.0091, + "step": 40206 + }, + { + "epoch": 5.27, + "learning_rate": 5.522270552591955e-06, + "loss": 0.0085, + "step": 40207 + }, + { + "epoch": 5.27, + "learning_rate": 5.521796370164942e-06, + "loss": 0.0109, + "step": 40208 + }, + { + "epoch": 5.27, + "learning_rate": 5.521322200332642e-06, + "loss": 0.0093, + "step": 40209 + }, + { + "epoch": 5.27, + "learning_rate": 5.520848043096385e-06, + "loss": 0.0075, + "step": 40210 + }, + { + "epoch": 5.27, + "learning_rate": 5.520373898457499e-06, + "loss": 0.0061, + "step": 40211 + }, + { + "epoch": 5.27, + "learning_rate": 5.5198997664173286e-06, + "loss": 0.0134, + "step": 40212 + }, + { + "epoch": 5.27, + "learning_rate": 5.5194256469772e-06, + "loss": 0.0084, + "step": 40213 + }, + { + "epoch": 5.27, + "learning_rate": 5.518951540138445e-06, + "loss": 0.0087, + "step": 40214 + }, + { + "epoch": 5.27, + "learning_rate": 5.5184774459024035e-06, + "loss": 0.0066, + "step": 40215 + }, + { + "epoch": 5.27, + "learning_rate": 5.518003364270402e-06, + "loss": 0.0076, + "step": 40216 + }, + { + "epoch": 5.27, + "learning_rate": 5.517529295243782e-06, + "loss": 0.0096, + "step": 40217 + }, + { + "epoch": 5.27, + "learning_rate": 5.517055238823872e-06, + "loss": 0.0065, + "step": 40218 + }, + { + "epoch": 5.27, + "learning_rate": 5.516581195012001e-06, + "loss": 0.0086, + "step": 40219 + }, + { + "epoch": 5.27, + "learning_rate": 5.51610716380951e-06, + "loss": 0.005, + "step": 40220 + }, + { + "epoch": 5.27, + "learning_rate": 5.5156331452177294e-06, + "loss": 0.0074, + "step": 40221 + }, + { + "epoch": 5.27, + "learning_rate": 5.515159139237991e-06, + "loss": 0.0091, + "step": 40222 + }, + { + "epoch": 5.27, + "learning_rate": 5.514685145871625e-06, + "loss": 0.0071, + "step": 40223 + }, + { + "epoch": 5.27, + "learning_rate": 5.514211165119974e-06, + "loss": 0.0068, + "step": 40224 + }, + { + "epoch": 5.27, + "learning_rate": 5.513737196984359e-06, + "loss": 0.0085, + "step": 40225 + }, + { + "epoch": 5.27, + "learning_rate": 5.513263241466125e-06, + "loss": 0.0071, + "step": 40226 + }, + { + "epoch": 5.27, + "learning_rate": 5.512789298566595e-06, + "loss": 0.0068, + "step": 40227 + }, + { + "epoch": 5.27, + "learning_rate": 5.512315368287109e-06, + "loss": 0.0059, + "step": 40228 + }, + { + "epoch": 5.27, + "learning_rate": 5.511841450628996e-06, + "loss": 0.0073, + "step": 40229 + }, + { + "epoch": 5.27, + "learning_rate": 5.511367545593591e-06, + "loss": 0.0107, + "step": 40230 + }, + { + "epoch": 5.27, + "learning_rate": 5.510893653182223e-06, + "loss": 0.0092, + "step": 40231 + }, + { + "epoch": 5.27, + "learning_rate": 5.51041977339623e-06, + "loss": 0.0082, + "step": 40232 + }, + { + "epoch": 5.27, + "learning_rate": 5.509945906236945e-06, + "loss": 0.0073, + "step": 40233 + }, + { + "epoch": 5.27, + "learning_rate": 5.509472051705691e-06, + "loss": 0.007, + "step": 40234 + }, + { + "epoch": 5.27, + "learning_rate": 5.508998209803813e-06, + "loss": 0.0103, + "step": 40235 + }, + { + "epoch": 5.27, + "learning_rate": 5.508524380532634e-06, + "loss": 0.0069, + "step": 40236 + }, + { + "epoch": 5.27, + "learning_rate": 5.5080505638934946e-06, + "loss": 0.0054, + "step": 40237 + }, + { + "epoch": 5.27, + "learning_rate": 5.507576759887721e-06, + "loss": 0.006, + "step": 40238 + }, + { + "epoch": 5.27, + "learning_rate": 5.507102968516651e-06, + "loss": 0.0064, + "step": 40239 + }, + { + "epoch": 5.27, + "learning_rate": 5.5066291897816156e-06, + "loss": 0.0084, + "step": 40240 + }, + { + "epoch": 5.27, + "learning_rate": 5.506155423683946e-06, + "loss": 0.0043, + "step": 40241 + }, + { + "epoch": 5.27, + "learning_rate": 5.505681670224971e-06, + "loss": 0.006, + "step": 40242 + }, + { + "epoch": 5.27, + "learning_rate": 5.505207929406031e-06, + "loss": 0.006, + "step": 40243 + }, + { + "epoch": 5.27, + "learning_rate": 5.504734201228454e-06, + "loss": 0.0064, + "step": 40244 + }, + { + "epoch": 5.27, + "learning_rate": 5.504260485693573e-06, + "loss": 0.0101, + "step": 40245 + }, + { + "epoch": 5.27, + "learning_rate": 5.503786782802716e-06, + "loss": 0.0063, + "step": 40246 + }, + { + "epoch": 5.27, + "learning_rate": 5.50331309255722e-06, + "loss": 0.0099, + "step": 40247 + }, + { + "epoch": 5.27, + "learning_rate": 5.502839414958419e-06, + "loss": 0.0073, + "step": 40248 + }, + { + "epoch": 5.27, + "learning_rate": 5.502365750007643e-06, + "loss": 0.0095, + "step": 40249 + }, + { + "epoch": 5.27, + "learning_rate": 5.501892097706221e-06, + "loss": 0.0071, + "step": 40250 + }, + { + "epoch": 5.27, + "learning_rate": 5.501418458055491e-06, + "loss": 0.0069, + "step": 40251 + }, + { + "epoch": 5.27, + "learning_rate": 5.500944831056782e-06, + "loss": 0.008, + "step": 40252 + }, + { + "epoch": 5.27, + "learning_rate": 5.500471216711423e-06, + "loss": 0.0083, + "step": 40253 + }, + { + "epoch": 5.27, + "learning_rate": 5.499997615020752e-06, + "loss": 0.0094, + "step": 40254 + }, + { + "epoch": 5.27, + "learning_rate": 5.499524025986099e-06, + "loss": 0.0084, + "step": 40255 + }, + { + "epoch": 5.27, + "learning_rate": 5.499050449608794e-06, + "loss": 0.0065, + "step": 40256 + }, + { + "epoch": 5.27, + "learning_rate": 5.498576885890166e-06, + "loss": 0.0106, + "step": 40257 + }, + { + "epoch": 5.27, + "learning_rate": 5.49810333483155e-06, + "loss": 0.008, + "step": 40258 + }, + { + "epoch": 5.27, + "learning_rate": 5.497629796434285e-06, + "loss": 0.0064, + "step": 40259 + }, + { + "epoch": 5.27, + "learning_rate": 5.4971562706996954e-06, + "loss": 0.007, + "step": 40260 + }, + { + "epoch": 5.27, + "learning_rate": 5.4966827576291094e-06, + "loss": 0.0072, + "step": 40261 + }, + { + "epoch": 5.27, + "learning_rate": 5.496209257223868e-06, + "loss": 0.0103, + "step": 40262 + }, + { + "epoch": 5.27, + "learning_rate": 5.495735769485298e-06, + "loss": 0.0075, + "step": 40263 + }, + { + "epoch": 5.27, + "learning_rate": 5.4952622944147306e-06, + "loss": 0.0083, + "step": 40264 + }, + { + "epoch": 5.27, + "learning_rate": 5.494788832013495e-06, + "loss": 0.0073, + "step": 40265 + }, + { + "epoch": 5.27, + "learning_rate": 5.494315382282929e-06, + "loss": 0.0083, + "step": 40266 + }, + { + "epoch": 5.27, + "learning_rate": 5.493841945224362e-06, + "loss": 0.0058, + "step": 40267 + }, + { + "epoch": 5.27, + "learning_rate": 5.49336852083912e-06, + "loss": 0.0094, + "step": 40268 + }, + { + "epoch": 5.27, + "learning_rate": 5.49289510912854e-06, + "loss": 0.0072, + "step": 40269 + }, + { + "epoch": 5.28, + "learning_rate": 5.492421710093956e-06, + "loss": 0.0059, + "step": 40270 + }, + { + "epoch": 5.28, + "learning_rate": 5.491948323736696e-06, + "loss": 0.0084, + "step": 40271 + }, + { + "epoch": 5.28, + "learning_rate": 5.49147495005809e-06, + "loss": 0.0085, + "step": 40272 + }, + { + "epoch": 5.28, + "learning_rate": 5.491001589059468e-06, + "loss": 0.007, + "step": 40273 + }, + { + "epoch": 5.28, + "learning_rate": 5.490528240742167e-06, + "loss": 0.0048, + "step": 40274 + }, + { + "epoch": 5.28, + "learning_rate": 5.490054905107516e-06, + "loss": 0.0065, + "step": 40275 + }, + { + "epoch": 5.28, + "learning_rate": 5.48958158215684e-06, + "loss": 0.0071, + "step": 40276 + }, + { + "epoch": 5.28, + "learning_rate": 5.4891082718914816e-06, + "loss": 0.0054, + "step": 40277 + }, + { + "epoch": 5.28, + "learning_rate": 5.4886349743127655e-06, + "loss": 0.0113, + "step": 40278 + }, + { + "epoch": 5.28, + "learning_rate": 5.488161689422023e-06, + "loss": 0.0063, + "step": 40279 + }, + { + "epoch": 5.28, + "learning_rate": 5.48768841722058e-06, + "loss": 0.0074, + "step": 40280 + }, + { + "epoch": 5.28, + "learning_rate": 5.487215157709774e-06, + "loss": 0.0081, + "step": 40281 + }, + { + "epoch": 5.28, + "learning_rate": 5.486741910890941e-06, + "loss": 0.0069, + "step": 40282 + }, + { + "epoch": 5.28, + "learning_rate": 5.4862686767654024e-06, + "loss": 0.008, + "step": 40283 + }, + { + "epoch": 5.28, + "learning_rate": 5.485795455334492e-06, + "loss": 0.0089, + "step": 40284 + }, + { + "epoch": 5.28, + "learning_rate": 5.485322246599544e-06, + "loss": 0.0056, + "step": 40285 + }, + { + "epoch": 5.28, + "learning_rate": 5.484849050561887e-06, + "loss": 0.009, + "step": 40286 + }, + { + "epoch": 5.28, + "learning_rate": 5.484375867222851e-06, + "loss": 0.0076, + "step": 40287 + }, + { + "epoch": 5.28, + "learning_rate": 5.483902696583765e-06, + "loss": 0.0072, + "step": 40288 + }, + { + "epoch": 5.28, + "learning_rate": 5.483429538645964e-06, + "loss": 0.0066, + "step": 40289 + }, + { + "epoch": 5.28, + "learning_rate": 5.482956393410779e-06, + "loss": 0.0087, + "step": 40290 + }, + { + "epoch": 5.28, + "learning_rate": 5.482483260879533e-06, + "loss": 0.0047, + "step": 40291 + }, + { + "epoch": 5.28, + "learning_rate": 5.4820101410535634e-06, + "loss": 0.0071, + "step": 40292 + }, + { + "epoch": 5.28, + "learning_rate": 5.4815370339342036e-06, + "loss": 0.009, + "step": 40293 + }, + { + "epoch": 5.28, + "learning_rate": 5.481063939522779e-06, + "loss": 0.0056, + "step": 40294 + }, + { + "epoch": 5.28, + "learning_rate": 5.480590857820622e-06, + "loss": 0.0085, + "step": 40295 + }, + { + "epoch": 5.28, + "learning_rate": 5.480117788829061e-06, + "loss": 0.0076, + "step": 40296 + }, + { + "epoch": 5.28, + "learning_rate": 5.479644732549428e-06, + "loss": 0.0084, + "step": 40297 + }, + { + "epoch": 5.28, + "learning_rate": 5.479171688983056e-06, + "loss": 0.0058, + "step": 40298 + }, + { + "epoch": 5.28, + "learning_rate": 5.478698658131267e-06, + "loss": 0.0102, + "step": 40299 + }, + { + "epoch": 5.28, + "learning_rate": 5.4782256399954025e-06, + "loss": 0.0063, + "step": 40300 + }, + { + "epoch": 5.28, + "learning_rate": 5.477752634576788e-06, + "loss": 0.0076, + "step": 40301 + }, + { + "epoch": 5.28, + "learning_rate": 5.47727964187675e-06, + "loss": 0.009, + "step": 40302 + }, + { + "epoch": 5.28, + "learning_rate": 5.476806661896621e-06, + "loss": 0.0041, + "step": 40303 + }, + { + "epoch": 5.28, + "learning_rate": 5.476333694637736e-06, + "loss": 0.0072, + "step": 40304 + }, + { + "epoch": 5.28, + "learning_rate": 5.475860740101422e-06, + "loss": 0.0054, + "step": 40305 + }, + { + "epoch": 5.28, + "learning_rate": 5.4753877982890095e-06, + "loss": 0.0099, + "step": 40306 + }, + { + "epoch": 5.28, + "learning_rate": 5.474914869201823e-06, + "loss": 0.0096, + "step": 40307 + }, + { + "epoch": 5.28, + "learning_rate": 5.4744419528412005e-06, + "loss": 0.0096, + "step": 40308 + }, + { + "epoch": 5.28, + "learning_rate": 5.47396904920847e-06, + "loss": 0.0075, + "step": 40309 + }, + { + "epoch": 5.28, + "learning_rate": 5.4734961583049615e-06, + "loss": 0.009, + "step": 40310 + }, + { + "epoch": 5.28, + "learning_rate": 5.473023280132e-06, + "loss": 0.009, + "step": 40311 + }, + { + "epoch": 5.28, + "learning_rate": 5.472550414690923e-06, + "loss": 0.0067, + "step": 40312 + }, + { + "epoch": 5.28, + "learning_rate": 5.472077561983051e-06, + "loss": 0.0063, + "step": 40313 + }, + { + "epoch": 5.28, + "learning_rate": 5.471604722009726e-06, + "loss": 0.0057, + "step": 40314 + }, + { + "epoch": 5.28, + "learning_rate": 5.471131894772266e-06, + "loss": 0.0091, + "step": 40315 + }, + { + "epoch": 5.28, + "learning_rate": 5.470659080272012e-06, + "loss": 0.0065, + "step": 40316 + }, + { + "epoch": 5.28, + "learning_rate": 5.470186278510287e-06, + "loss": 0.0118, + "step": 40317 + }, + { + "epoch": 5.28, + "learning_rate": 5.4697134894884174e-06, + "loss": 0.0093, + "step": 40318 + }, + { + "epoch": 5.28, + "learning_rate": 5.4692407132077415e-06, + "loss": 0.007, + "step": 40319 + }, + { + "epoch": 5.28, + "learning_rate": 5.468767949669587e-06, + "loss": 0.0081, + "step": 40320 + }, + { + "epoch": 5.28, + "learning_rate": 5.468295198875279e-06, + "loss": 0.0072, + "step": 40321 + }, + { + "epoch": 5.28, + "learning_rate": 5.467822460826147e-06, + "loss": 0.0088, + "step": 40322 + }, + { + "epoch": 5.28, + "learning_rate": 5.467349735523526e-06, + "loss": 0.0066, + "step": 40323 + }, + { + "epoch": 5.28, + "learning_rate": 5.466877022968739e-06, + "loss": 0.0057, + "step": 40324 + }, + { + "epoch": 5.28, + "learning_rate": 5.466404323163121e-06, + "loss": 0.0114, + "step": 40325 + }, + { + "epoch": 5.28, + "learning_rate": 5.465931636107997e-06, + "loss": 0.0075, + "step": 40326 + }, + { + "epoch": 5.28, + "learning_rate": 5.4654589618047034e-06, + "loss": 0.0089, + "step": 40327 + }, + { + "epoch": 5.28, + "learning_rate": 5.464986300254565e-06, + "loss": 0.0059, + "step": 40328 + }, + { + "epoch": 5.28, + "learning_rate": 5.46451365145891e-06, + "loss": 0.0063, + "step": 40329 + }, + { + "epoch": 5.28, + "learning_rate": 5.464041015419066e-06, + "loss": 0.0067, + "step": 40330 + }, + { + "epoch": 5.28, + "learning_rate": 5.463568392136369e-06, + "loss": 0.0072, + "step": 40331 + }, + { + "epoch": 5.28, + "learning_rate": 5.463095781612143e-06, + "loss": 0.0069, + "step": 40332 + }, + { + "epoch": 5.28, + "learning_rate": 5.462623183847715e-06, + "loss": 0.0075, + "step": 40333 + }, + { + "epoch": 5.28, + "learning_rate": 5.462150598844418e-06, + "loss": 0.0068, + "step": 40334 + }, + { + "epoch": 5.28, + "learning_rate": 5.461678026603585e-06, + "loss": 0.0077, + "step": 40335 + }, + { + "epoch": 5.28, + "learning_rate": 5.46120546712654e-06, + "loss": 0.0113, + "step": 40336 + }, + { + "epoch": 5.28, + "learning_rate": 5.460732920414615e-06, + "loss": 0.0092, + "step": 40337 + }, + { + "epoch": 5.28, + "learning_rate": 5.460260386469132e-06, + "loss": 0.0084, + "step": 40338 + }, + { + "epoch": 5.28, + "learning_rate": 5.459787865291429e-06, + "loss": 0.0087, + "step": 40339 + }, + { + "epoch": 5.28, + "learning_rate": 5.459315356882828e-06, + "loss": 0.008, + "step": 40340 + }, + { + "epoch": 5.28, + "learning_rate": 5.45884286124466e-06, + "loss": 0.0062, + "step": 40341 + }, + { + "epoch": 5.28, + "learning_rate": 5.458370378378257e-06, + "loss": 0.0078, + "step": 40342 + }, + { + "epoch": 5.28, + "learning_rate": 5.457897908284946e-06, + "loss": 0.0075, + "step": 40343 + }, + { + "epoch": 5.28, + "learning_rate": 5.4574254509660566e-06, + "loss": 0.0063, + "step": 40344 + }, + { + "epoch": 5.28, + "learning_rate": 5.45695300642291e-06, + "loss": 0.0079, + "step": 40345 + }, + { + "epoch": 5.29, + "learning_rate": 5.456480574656841e-06, + "loss": 0.0051, + "step": 40346 + }, + { + "epoch": 5.29, + "learning_rate": 5.456008155669185e-06, + "loss": 0.0075, + "step": 40347 + }, + { + "epoch": 5.29, + "learning_rate": 5.4555357494612625e-06, + "loss": 0.0079, + "step": 40348 + }, + { + "epoch": 5.29, + "learning_rate": 5.4550633560344e-06, + "loss": 0.0088, + "step": 40349 + }, + { + "epoch": 5.29, + "learning_rate": 5.454590975389933e-06, + "loss": 0.007, + "step": 40350 + }, + { + "epoch": 5.29, + "learning_rate": 5.4541186075291875e-06, + "loss": 0.0066, + "step": 40351 + }, + { + "epoch": 5.29, + "learning_rate": 5.453646252453491e-06, + "loss": 0.0069, + "step": 40352 + }, + { + "epoch": 5.29, + "learning_rate": 5.453173910164169e-06, + "loss": 0.0071, + "step": 40353 + }, + { + "epoch": 5.29, + "learning_rate": 5.452701580662556e-06, + "loss": 0.0104, + "step": 40354 + }, + { + "epoch": 5.29, + "learning_rate": 5.45222926394998e-06, + "loss": 0.0077, + "step": 40355 + }, + { + "epoch": 5.29, + "learning_rate": 5.451756960027761e-06, + "loss": 0.0074, + "step": 40356 + }, + { + "epoch": 5.29, + "learning_rate": 5.451284668897233e-06, + "loss": 0.0065, + "step": 40357 + }, + { + "epoch": 5.29, + "learning_rate": 5.45081239055973e-06, + "loss": 0.0077, + "step": 40358 + }, + { + "epoch": 5.29, + "learning_rate": 5.450340125016573e-06, + "loss": 0.0072, + "step": 40359 + }, + { + "epoch": 5.29, + "learning_rate": 5.449867872269094e-06, + "loss": 0.0068, + "step": 40360 + }, + { + "epoch": 5.29, + "learning_rate": 5.449395632318615e-06, + "loss": 0.0093, + "step": 40361 + }, + { + "epoch": 5.29, + "learning_rate": 5.448923405166473e-06, + "loss": 0.0059, + "step": 40362 + }, + { + "epoch": 5.29, + "learning_rate": 5.44845119081399e-06, + "loss": 0.0088, + "step": 40363 + }, + { + "epoch": 5.29, + "learning_rate": 5.447978989262492e-06, + "loss": 0.0059, + "step": 40364 + }, + { + "epoch": 5.29, + "learning_rate": 5.4475068005133156e-06, + "loss": 0.0119, + "step": 40365 + }, + { + "epoch": 5.29, + "learning_rate": 5.447034624567783e-06, + "loss": 0.0061, + "step": 40366 + }, + { + "epoch": 5.29, + "learning_rate": 5.4465624614272195e-06, + "loss": 0.0061, + "step": 40367 + }, + { + "epoch": 5.29, + "learning_rate": 5.446090311092957e-06, + "loss": 0.0081, + "step": 40368 + }, + { + "epoch": 5.29, + "learning_rate": 5.445618173566326e-06, + "loss": 0.0064, + "step": 40369 + }, + { + "epoch": 5.29, + "learning_rate": 5.4451460488486516e-06, + "loss": 0.0062, + "step": 40370 + }, + { + "epoch": 5.29, + "learning_rate": 5.444673936941263e-06, + "loss": 0.0058, + "step": 40371 + }, + { + "epoch": 5.29, + "learning_rate": 5.44420183784548e-06, + "loss": 0.0062, + "step": 40372 + }, + { + "epoch": 5.29, + "learning_rate": 5.443729751562642e-06, + "loss": 0.0082, + "step": 40373 + }, + { + "epoch": 5.29, + "learning_rate": 5.4432576780940715e-06, + "loss": 0.0077, + "step": 40374 + }, + { + "epoch": 5.29, + "learning_rate": 5.442785617441091e-06, + "loss": 0.0052, + "step": 40375 + }, + { + "epoch": 5.29, + "learning_rate": 5.442313569605038e-06, + "loss": 0.0072, + "step": 40376 + }, + { + "epoch": 5.29, + "learning_rate": 5.4418415345872355e-06, + "loss": 0.0077, + "step": 40377 + }, + { + "epoch": 5.29, + "learning_rate": 5.441369512389008e-06, + "loss": 0.0061, + "step": 40378 + }, + { + "epoch": 5.29, + "learning_rate": 5.440897503011688e-06, + "loss": 0.0108, + "step": 40379 + }, + { + "epoch": 5.29, + "learning_rate": 5.440425506456597e-06, + "loss": 0.0063, + "step": 40380 + }, + { + "epoch": 5.29, + "learning_rate": 5.439953522725072e-06, + "loss": 0.0084, + "step": 40381 + }, + { + "epoch": 5.29, + "learning_rate": 5.4394815518184344e-06, + "loss": 0.0071, + "step": 40382 + }, + { + "epoch": 5.29, + "learning_rate": 5.439009593738007e-06, + "loss": 0.0043, + "step": 40383 + }, + { + "epoch": 5.29, + "learning_rate": 5.438537648485127e-06, + "loss": 0.0071, + "step": 40384 + }, + { + "epoch": 5.29, + "learning_rate": 5.438065716061119e-06, + "loss": 0.007, + "step": 40385 + }, + { + "epoch": 5.29, + "learning_rate": 5.437593796467304e-06, + "loss": 0.0083, + "step": 40386 + }, + { + "epoch": 5.29, + "learning_rate": 5.437121889705013e-06, + "loss": 0.0061, + "step": 40387 + }, + { + "epoch": 5.29, + "learning_rate": 5.4366499957755755e-06, + "loss": 0.0084, + "step": 40388 + }, + { + "epoch": 5.29, + "learning_rate": 5.436178114680315e-06, + "loss": 0.0075, + "step": 40389 + }, + { + "epoch": 5.29, + "learning_rate": 5.435706246420562e-06, + "loss": 0.0107, + "step": 40390 + }, + { + "epoch": 5.29, + "learning_rate": 5.43523439099764e-06, + "loss": 0.0077, + "step": 40391 + }, + { + "epoch": 5.29, + "learning_rate": 5.434762548412882e-06, + "loss": 0.0065, + "step": 40392 + }, + { + "epoch": 5.29, + "learning_rate": 5.4342907186676096e-06, + "loss": 0.0093, + "step": 40393 + }, + { + "epoch": 5.29, + "learning_rate": 5.433818901763154e-06, + "loss": 0.0082, + "step": 40394 + }, + { + "epoch": 5.29, + "learning_rate": 5.433347097700834e-06, + "loss": 0.009, + "step": 40395 + }, + { + "epoch": 5.29, + "learning_rate": 5.432875306481986e-06, + "loss": 0.0075, + "step": 40396 + }, + { + "epoch": 5.29, + "learning_rate": 5.432403528107933e-06, + "loss": 0.0084, + "step": 40397 + }, + { + "epoch": 5.29, + "learning_rate": 5.4319317625799985e-06, + "loss": 0.0092, + "step": 40398 + }, + { + "epoch": 5.29, + "learning_rate": 5.4314600098995165e-06, + "loss": 0.0073, + "step": 40399 + }, + { + "epoch": 5.29, + "learning_rate": 5.430988270067805e-06, + "loss": 0.0056, + "step": 40400 + }, + { + "epoch": 5.29, + "learning_rate": 5.430516543086202e-06, + "loss": 0.0066, + "step": 40401 + }, + { + "epoch": 5.29, + "learning_rate": 5.430044828956025e-06, + "loss": 0.0069, + "step": 40402 + }, + { + "epoch": 5.29, + "learning_rate": 5.429573127678601e-06, + "loss": 0.0083, + "step": 40403 + }, + { + "epoch": 5.29, + "learning_rate": 5.429101439255264e-06, + "loss": 0.007, + "step": 40404 + }, + { + "epoch": 5.29, + "learning_rate": 5.428629763687334e-06, + "loss": 0.0075, + "step": 40405 + }, + { + "epoch": 5.29, + "learning_rate": 5.428158100976136e-06, + "loss": 0.0077, + "step": 40406 + }, + { + "epoch": 5.29, + "learning_rate": 5.427686451123004e-06, + "loss": 0.01, + "step": 40407 + }, + { + "epoch": 5.29, + "learning_rate": 5.42721481412926e-06, + "loss": 0.0061, + "step": 40408 + }, + { + "epoch": 5.29, + "learning_rate": 5.426743189996232e-06, + "loss": 0.0108, + "step": 40409 + }, + { + "epoch": 5.29, + "learning_rate": 5.426271578725241e-06, + "loss": 0.0119, + "step": 40410 + }, + { + "epoch": 5.29, + "learning_rate": 5.425799980317616e-06, + "loss": 0.0063, + "step": 40411 + }, + { + "epoch": 5.29, + "learning_rate": 5.425328394774692e-06, + "loss": 0.0088, + "step": 40412 + }, + { + "epoch": 5.29, + "learning_rate": 5.424856822097786e-06, + "loss": 0.0068, + "step": 40413 + }, + { + "epoch": 5.29, + "learning_rate": 5.424385262288222e-06, + "loss": 0.0081, + "step": 40414 + }, + { + "epoch": 5.29, + "learning_rate": 5.423913715347336e-06, + "loss": 0.0069, + "step": 40415 + }, + { + "epoch": 5.29, + "learning_rate": 5.423442181276448e-06, + "loss": 0.0073, + "step": 40416 + }, + { + "epoch": 5.29, + "learning_rate": 5.422970660076886e-06, + "loss": 0.0074, + "step": 40417 + }, + { + "epoch": 5.29, + "learning_rate": 5.422499151749972e-06, + "loss": 0.0084, + "step": 40418 + }, + { + "epoch": 5.29, + "learning_rate": 5.422027656297038e-06, + "loss": 0.0075, + "step": 40419 + }, + { + "epoch": 5.29, + "learning_rate": 5.421556173719408e-06, + "loss": 0.0088, + "step": 40420 + }, + { + "epoch": 5.29, + "learning_rate": 5.421084704018402e-06, + "loss": 0.009, + "step": 40421 + }, + { + "epoch": 5.29, + "learning_rate": 5.420613247195354e-06, + "loss": 0.0098, + "step": 40422 + }, + { + "epoch": 5.3, + "learning_rate": 5.420141803251591e-06, + "loss": 0.0064, + "step": 40423 + }, + { + "epoch": 5.3, + "learning_rate": 5.419670372188432e-06, + "loss": 0.0072, + "step": 40424 + }, + { + "epoch": 5.3, + "learning_rate": 5.419198954007206e-06, + "loss": 0.008, + "step": 40425 + }, + { + "epoch": 5.3, + "learning_rate": 5.418727548709242e-06, + "loss": 0.0052, + "step": 40426 + }, + { + "epoch": 5.3, + "learning_rate": 5.418256156295863e-06, + "loss": 0.0059, + "step": 40427 + }, + { + "epoch": 5.3, + "learning_rate": 5.4177847767683934e-06, + "loss": 0.0076, + "step": 40428 + }, + { + "epoch": 5.3, + "learning_rate": 5.417313410128155e-06, + "loss": 0.0134, + "step": 40429 + }, + { + "epoch": 5.3, + "learning_rate": 5.4168420563764854e-06, + "loss": 0.0083, + "step": 40430 + }, + { + "epoch": 5.3, + "learning_rate": 5.416370715514703e-06, + "loss": 0.0113, + "step": 40431 + }, + { + "epoch": 5.3, + "learning_rate": 5.415899387544132e-06, + "loss": 0.0085, + "step": 40432 + }, + { + "epoch": 5.3, + "learning_rate": 5.4154280724660935e-06, + "loss": 0.0089, + "step": 40433 + }, + { + "epoch": 5.3, + "learning_rate": 5.414956770281928e-06, + "loss": 0.0067, + "step": 40434 + }, + { + "epoch": 5.3, + "learning_rate": 5.414485480992951e-06, + "loss": 0.008, + "step": 40435 + }, + { + "epoch": 5.3, + "learning_rate": 5.414014204600488e-06, + "loss": 0.0082, + "step": 40436 + }, + { + "epoch": 5.3, + "learning_rate": 5.4135429411058644e-06, + "loss": 0.0088, + "step": 40437 + }, + { + "epoch": 5.3, + "learning_rate": 5.41307169051041e-06, + "loss": 0.0058, + "step": 40438 + }, + { + "epoch": 5.3, + "learning_rate": 5.412600452815447e-06, + "loss": 0.0074, + "step": 40439 + }, + { + "epoch": 5.3, + "learning_rate": 5.412129228022298e-06, + "loss": 0.008, + "step": 40440 + }, + { + "epoch": 5.3, + "learning_rate": 5.411658016132295e-06, + "loss": 0.0067, + "step": 40441 + }, + { + "epoch": 5.3, + "learning_rate": 5.4111868171467585e-06, + "loss": 0.0073, + "step": 40442 + }, + { + "epoch": 5.3, + "learning_rate": 5.410715631067015e-06, + "loss": 0.0082, + "step": 40443 + }, + { + "epoch": 5.3, + "learning_rate": 5.410244457894386e-06, + "loss": 0.0072, + "step": 40444 + }, + { + "epoch": 5.3, + "learning_rate": 5.4097732976302e-06, + "loss": 0.0066, + "step": 40445 + }, + { + "epoch": 5.3, + "learning_rate": 5.409302150275786e-06, + "loss": 0.0071, + "step": 40446 + }, + { + "epoch": 5.3, + "learning_rate": 5.408831015832465e-06, + "loss": 0.006, + "step": 40447 + }, + { + "epoch": 5.3, + "learning_rate": 5.408359894301558e-06, + "loss": 0.0081, + "step": 40448 + }, + { + "epoch": 5.3, + "learning_rate": 5.407888785684399e-06, + "loss": 0.0112, + "step": 40449 + }, + { + "epoch": 5.3, + "learning_rate": 5.407417689982307e-06, + "loss": 0.0083, + "step": 40450 + }, + { + "epoch": 5.3, + "learning_rate": 5.406946607196609e-06, + "loss": 0.0083, + "step": 40451 + }, + { + "epoch": 5.3, + "learning_rate": 5.406475537328626e-06, + "loss": 0.0092, + "step": 40452 + }, + { + "epoch": 5.3, + "learning_rate": 5.406004480379688e-06, + "loss": 0.0079, + "step": 40453 + }, + { + "epoch": 5.3, + "learning_rate": 5.4055334363511205e-06, + "loss": 0.0109, + "step": 40454 + }, + { + "epoch": 5.3, + "learning_rate": 5.4050624052442394e-06, + "loss": 0.006, + "step": 40455 + }, + { + "epoch": 5.3, + "learning_rate": 5.404591387060376e-06, + "loss": 0.008, + "step": 40456 + }, + { + "epoch": 5.3, + "learning_rate": 5.404120381800858e-06, + "loss": 0.0082, + "step": 40457 + }, + { + "epoch": 5.3, + "learning_rate": 5.403649389467007e-06, + "loss": 0.0074, + "step": 40458 + }, + { + "epoch": 5.3, + "learning_rate": 5.403178410060148e-06, + "loss": 0.0076, + "step": 40459 + }, + { + "epoch": 5.3, + "learning_rate": 5.402707443581601e-06, + "loss": 0.009, + "step": 40460 + }, + { + "epoch": 5.3, + "learning_rate": 5.402236490032698e-06, + "loss": 0.0072, + "step": 40461 + }, + { + "epoch": 5.3, + "learning_rate": 5.401765549414759e-06, + "loss": 0.0076, + "step": 40462 + }, + { + "epoch": 5.3, + "learning_rate": 5.4012946217291075e-06, + "loss": 0.0076, + "step": 40463 + }, + { + "epoch": 5.3, + "learning_rate": 5.400823706977073e-06, + "loss": 0.0082, + "step": 40464 + }, + { + "epoch": 5.3, + "learning_rate": 5.400352805159973e-06, + "loss": 0.0063, + "step": 40465 + }, + { + "epoch": 5.3, + "learning_rate": 5.3998819162791395e-06, + "loss": 0.0091, + "step": 40466 + }, + { + "epoch": 5.3, + "learning_rate": 5.399411040335893e-06, + "loss": 0.0079, + "step": 40467 + }, + { + "epoch": 5.3, + "learning_rate": 5.398940177331555e-06, + "loss": 0.0059, + "step": 40468 + }, + { + "epoch": 5.3, + "learning_rate": 5.398469327267457e-06, + "loss": 0.0088, + "step": 40469 + }, + { + "epoch": 5.3, + "learning_rate": 5.397998490144918e-06, + "loss": 0.0081, + "step": 40470 + }, + { + "epoch": 5.3, + "learning_rate": 5.397527665965259e-06, + "loss": 0.0064, + "step": 40471 + }, + { + "epoch": 5.3, + "learning_rate": 5.397056854729812e-06, + "loss": 0.0075, + "step": 40472 + }, + { + "epoch": 5.3, + "learning_rate": 5.3965860564398985e-06, + "loss": 0.0076, + "step": 40473 + }, + { + "epoch": 5.3, + "learning_rate": 5.396115271096841e-06, + "loss": 0.0085, + "step": 40474 + }, + { + "epoch": 5.3, + "learning_rate": 5.39564449870196e-06, + "loss": 0.0081, + "step": 40475 + }, + { + "epoch": 5.3, + "learning_rate": 5.395173739256585e-06, + "loss": 0.0083, + "step": 40476 + }, + { + "epoch": 5.3, + "learning_rate": 5.394702992762041e-06, + "loss": 0.0077, + "step": 40477 + }, + { + "epoch": 5.3, + "learning_rate": 5.394232259219649e-06, + "loss": 0.0082, + "step": 40478 + }, + { + "epoch": 5.3, + "learning_rate": 5.393761538630732e-06, + "loss": 0.0084, + "step": 40479 + }, + { + "epoch": 5.3, + "learning_rate": 5.393290830996618e-06, + "loss": 0.007, + "step": 40480 + }, + { + "epoch": 5.3, + "learning_rate": 5.392820136318628e-06, + "loss": 0.0075, + "step": 40481 + }, + { + "epoch": 5.3, + "learning_rate": 5.392349454598086e-06, + "loss": 0.0066, + "step": 40482 + }, + { + "epoch": 5.3, + "learning_rate": 5.391878785836312e-06, + "loss": 0.0068, + "step": 40483 + }, + { + "epoch": 5.3, + "learning_rate": 5.391408130034638e-06, + "loss": 0.008, + "step": 40484 + }, + { + "epoch": 5.3, + "learning_rate": 5.3909374871943835e-06, + "loss": 0.0075, + "step": 40485 + }, + { + "epoch": 5.3, + "learning_rate": 5.390466857316866e-06, + "loss": 0.0063, + "step": 40486 + }, + { + "epoch": 5.3, + "learning_rate": 5.389996240403416e-06, + "loss": 0.0071, + "step": 40487 + }, + { + "epoch": 5.3, + "learning_rate": 5.389525636455361e-06, + "loss": 0.0089, + "step": 40488 + }, + { + "epoch": 5.3, + "learning_rate": 5.389055045474018e-06, + "loss": 0.0081, + "step": 40489 + }, + { + "epoch": 5.3, + "learning_rate": 5.388584467460709e-06, + "loss": 0.0068, + "step": 40490 + }, + { + "epoch": 5.3, + "learning_rate": 5.388113902416765e-06, + "loss": 0.0102, + "step": 40491 + }, + { + "epoch": 5.3, + "learning_rate": 5.387643350343505e-06, + "loss": 0.0062, + "step": 40492 + }, + { + "epoch": 5.3, + "learning_rate": 5.387172811242252e-06, + "loss": 0.0073, + "step": 40493 + }, + { + "epoch": 5.3, + "learning_rate": 5.386702285114326e-06, + "loss": 0.009, + "step": 40494 + }, + { + "epoch": 5.3, + "learning_rate": 5.386231771961058e-06, + "loss": 0.0061, + "step": 40495 + }, + { + "epoch": 5.3, + "learning_rate": 5.385761271783767e-06, + "loss": 0.0069, + "step": 40496 + }, + { + "epoch": 5.3, + "learning_rate": 5.385290784583778e-06, + "loss": 0.0075, + "step": 40497 + }, + { + "epoch": 5.3, + "learning_rate": 5.384820310362404e-06, + "loss": 0.0079, + "step": 40498 + }, + { + "epoch": 5.31, + "learning_rate": 5.384349849120984e-06, + "loss": 0.007, + "step": 40499 + }, + { + "epoch": 5.31, + "learning_rate": 5.383879400860836e-06, + "loss": 0.007, + "step": 40500 + }, + { + "epoch": 5.31, + "learning_rate": 5.383408965583281e-06, + "loss": 0.0079, + "step": 40501 + }, + { + "epoch": 5.31, + "learning_rate": 5.382938543289639e-06, + "loss": 0.0094, + "step": 40502 + }, + { + "epoch": 5.31, + "learning_rate": 5.38246813398124e-06, + "loss": 0.0066, + "step": 40503 + }, + { + "epoch": 5.31, + "learning_rate": 5.3819977376594025e-06, + "loss": 0.0096, + "step": 40504 + }, + { + "epoch": 5.31, + "learning_rate": 5.381527354325446e-06, + "loss": 0.0075, + "step": 40505 + }, + { + "epoch": 5.31, + "learning_rate": 5.381056983980704e-06, + "loss": 0.0094, + "step": 40506 + }, + { + "epoch": 5.31, + "learning_rate": 5.380586626626493e-06, + "loss": 0.0074, + "step": 40507 + }, + { + "epoch": 5.31, + "learning_rate": 5.380116282264135e-06, + "loss": 0.007, + "step": 40508 + }, + { + "epoch": 5.31, + "learning_rate": 5.379645950894951e-06, + "loss": 0.0063, + "step": 40509 + }, + { + "epoch": 5.31, + "learning_rate": 5.379175632520266e-06, + "loss": 0.007, + "step": 40510 + }, + { + "epoch": 5.31, + "learning_rate": 5.378705327141407e-06, + "loss": 0.0057, + "step": 40511 + }, + { + "epoch": 5.31, + "learning_rate": 5.378235034759693e-06, + "loss": 0.009, + "step": 40512 + }, + { + "epoch": 5.31, + "learning_rate": 5.377764755376445e-06, + "loss": 0.0072, + "step": 40513 + }, + { + "epoch": 5.31, + "learning_rate": 5.37729448899299e-06, + "loss": 0.0082, + "step": 40514 + }, + { + "epoch": 5.31, + "learning_rate": 5.376824235610647e-06, + "loss": 0.0077, + "step": 40515 + }, + { + "epoch": 5.31, + "learning_rate": 5.376353995230741e-06, + "loss": 0.01, + "step": 40516 + }, + { + "epoch": 5.31, + "learning_rate": 5.375883767854588e-06, + "loss": 0.0073, + "step": 40517 + }, + { + "epoch": 5.31, + "learning_rate": 5.375413553483522e-06, + "loss": 0.0057, + "step": 40518 + }, + { + "epoch": 5.31, + "learning_rate": 5.374943352118858e-06, + "loss": 0.0095, + "step": 40519 + }, + { + "epoch": 5.31, + "learning_rate": 5.3744731637619156e-06, + "loss": 0.009, + "step": 40520 + }, + { + "epoch": 5.31, + "learning_rate": 5.3740029884140195e-06, + "loss": 0.0088, + "step": 40521 + }, + { + "epoch": 5.31, + "learning_rate": 5.3735328260764995e-06, + "loss": 0.0079, + "step": 40522 + }, + { + "epoch": 5.31, + "learning_rate": 5.373062676750671e-06, + "loss": 0.0072, + "step": 40523 + }, + { + "epoch": 5.31, + "learning_rate": 5.372592540437857e-06, + "loss": 0.0074, + "step": 40524 + }, + { + "epoch": 5.31, + "learning_rate": 5.372122417139377e-06, + "loss": 0.008, + "step": 40525 + }, + { + "epoch": 5.31, + "learning_rate": 5.3716523068565605e-06, + "loss": 0.0049, + "step": 40526 + }, + { + "epoch": 5.31, + "learning_rate": 5.371182209590724e-06, + "loss": 0.0111, + "step": 40527 + }, + { + "epoch": 5.31, + "learning_rate": 5.37071212534319e-06, + "loss": 0.0113, + "step": 40528 + }, + { + "epoch": 5.31, + "learning_rate": 5.370242054115282e-06, + "loss": 0.0074, + "step": 40529 + }, + { + "epoch": 5.31, + "learning_rate": 5.369771995908324e-06, + "loss": 0.0074, + "step": 40530 + }, + { + "epoch": 5.31, + "learning_rate": 5.369301950723631e-06, + "loss": 0.0071, + "step": 40531 + }, + { + "epoch": 5.31, + "learning_rate": 5.368831918562533e-06, + "loss": 0.007, + "step": 40532 + }, + { + "epoch": 5.31, + "learning_rate": 5.368361899426346e-06, + "loss": 0.0071, + "step": 40533 + }, + { + "epoch": 5.31, + "learning_rate": 5.367891893316397e-06, + "loss": 0.0059, + "step": 40534 + }, + { + "epoch": 5.31, + "learning_rate": 5.367421900234006e-06, + "loss": 0.0059, + "step": 40535 + }, + { + "epoch": 5.31, + "learning_rate": 5.36695192018049e-06, + "loss": 0.0076, + "step": 40536 + }, + { + "epoch": 5.31, + "learning_rate": 5.36648195315718e-06, + "loss": 0.0094, + "step": 40537 + }, + { + "epoch": 5.31, + "learning_rate": 5.3660119991653925e-06, + "loss": 0.0085, + "step": 40538 + }, + { + "epoch": 5.31, + "learning_rate": 5.365542058206448e-06, + "loss": 0.0068, + "step": 40539 + }, + { + "epoch": 5.31, + "learning_rate": 5.365072130281668e-06, + "loss": 0.0071, + "step": 40540 + }, + { + "epoch": 5.31, + "learning_rate": 5.364602215392378e-06, + "loss": 0.0078, + "step": 40541 + }, + { + "epoch": 5.31, + "learning_rate": 5.3641323135398935e-06, + "loss": 0.0059, + "step": 40542 + }, + { + "epoch": 5.31, + "learning_rate": 5.363662424725544e-06, + "loss": 0.0072, + "step": 40543 + }, + { + "epoch": 5.31, + "learning_rate": 5.363192548950644e-06, + "loss": 0.0096, + "step": 40544 + }, + { + "epoch": 5.31, + "learning_rate": 5.362722686216522e-06, + "loss": 0.0064, + "step": 40545 + }, + { + "epoch": 5.31, + "learning_rate": 5.362252836524495e-06, + "loss": 0.0089, + "step": 40546 + }, + { + "epoch": 5.31, + "learning_rate": 5.3617829998758845e-06, + "loss": 0.0097, + "step": 40547 + }, + { + "epoch": 5.31, + "learning_rate": 5.3613131762720096e-06, + "loss": 0.0086, + "step": 40548 + }, + { + "epoch": 5.31, + "learning_rate": 5.360843365714197e-06, + "loss": 0.0069, + "step": 40549 + }, + { + "epoch": 5.31, + "learning_rate": 5.360373568203765e-06, + "loss": 0.008, + "step": 40550 + }, + { + "epoch": 5.31, + "learning_rate": 5.359903783742033e-06, + "loss": 0.0086, + "step": 40551 + }, + { + "epoch": 5.31, + "learning_rate": 5.359434012330327e-06, + "loss": 0.0072, + "step": 40552 + }, + { + "epoch": 5.31, + "learning_rate": 5.3589642539699625e-06, + "loss": 0.0062, + "step": 40553 + }, + { + "epoch": 5.31, + "learning_rate": 5.358494508662268e-06, + "loss": 0.0108, + "step": 40554 + }, + { + "epoch": 5.31, + "learning_rate": 5.358024776408557e-06, + "loss": 0.011, + "step": 40555 + }, + { + "epoch": 5.31, + "learning_rate": 5.357555057210157e-06, + "loss": 0.0089, + "step": 40556 + }, + { + "epoch": 5.31, + "learning_rate": 5.357085351068386e-06, + "loss": 0.0081, + "step": 40557 + }, + { + "epoch": 5.31, + "learning_rate": 5.356615657984565e-06, + "loss": 0.0073, + "step": 40558 + }, + { + "epoch": 5.31, + "learning_rate": 5.356145977960012e-06, + "loss": 0.0085, + "step": 40559 + }, + { + "epoch": 5.31, + "learning_rate": 5.355676310996057e-06, + "loss": 0.0076, + "step": 40560 + }, + { + "epoch": 5.31, + "learning_rate": 5.355206657094012e-06, + "loss": 0.0088, + "step": 40561 + }, + { + "epoch": 5.31, + "learning_rate": 5.354737016255203e-06, + "loss": 0.0056, + "step": 40562 + }, + { + "epoch": 5.31, + "learning_rate": 5.354267388480942e-06, + "loss": 0.0067, + "step": 40563 + }, + { + "epoch": 5.31, + "learning_rate": 5.35379777377256e-06, + "loss": 0.0067, + "step": 40564 + }, + { + "epoch": 5.31, + "learning_rate": 5.353328172131376e-06, + "loss": 0.0056, + "step": 40565 + }, + { + "epoch": 5.31, + "learning_rate": 5.35285858355871e-06, + "loss": 0.0102, + "step": 40566 + }, + { + "epoch": 5.31, + "learning_rate": 5.3523890080558785e-06, + "loss": 0.006, + "step": 40567 + }, + { + "epoch": 5.31, + "learning_rate": 5.351919445624207e-06, + "loss": 0.0076, + "step": 40568 + }, + { + "epoch": 5.31, + "learning_rate": 5.351449896265016e-06, + "loss": 0.0078, + "step": 40569 + }, + { + "epoch": 5.31, + "learning_rate": 5.35098035997962e-06, + "loss": 0.0078, + "step": 40570 + }, + { + "epoch": 5.31, + "learning_rate": 5.35051083676935e-06, + "loss": 0.0086, + "step": 40571 + }, + { + "epoch": 5.31, + "learning_rate": 5.35004132663552e-06, + "loss": 0.0072, + "step": 40572 + }, + { + "epoch": 5.31, + "learning_rate": 5.34957182957945e-06, + "loss": 0.0071, + "step": 40573 + }, + { + "epoch": 5.31, + "learning_rate": 5.349102345602459e-06, + "loss": 0.0075, + "step": 40574 + }, + { + "epoch": 5.32, + "learning_rate": 5.348632874705868e-06, + "loss": 0.0083, + "step": 40575 + }, + { + "epoch": 5.32, + "learning_rate": 5.348163416891004e-06, + "loss": 0.0082, + "step": 40576 + }, + { + "epoch": 5.32, + "learning_rate": 5.347693972159185e-06, + "loss": 0.0079, + "step": 40577 + }, + { + "epoch": 5.32, + "learning_rate": 5.347224540511722e-06, + "loss": 0.0098, + "step": 40578 + }, + { + "epoch": 5.32, + "learning_rate": 5.346755121949947e-06, + "loss": 0.0095, + "step": 40579 + }, + { + "epoch": 5.32, + "learning_rate": 5.346285716475177e-06, + "loss": 0.0079, + "step": 40580 + }, + { + "epoch": 5.32, + "learning_rate": 5.3458163240887284e-06, + "loss": 0.0062, + "step": 40581 + }, + { + "epoch": 5.32, + "learning_rate": 5.34534694479192e-06, + "loss": 0.0085, + "step": 40582 + }, + { + "epoch": 5.32, + "learning_rate": 5.344877578586081e-06, + "loss": 0.0069, + "step": 40583 + }, + { + "epoch": 5.32, + "learning_rate": 5.344408225472524e-06, + "loss": 0.0078, + "step": 40584 + }, + { + "epoch": 5.32, + "learning_rate": 5.343938885452569e-06, + "loss": 0.0073, + "step": 40585 + }, + { + "epoch": 5.32, + "learning_rate": 5.343469558527537e-06, + "loss": 0.0096, + "step": 40586 + }, + { + "epoch": 5.32, + "learning_rate": 5.343000244698751e-06, + "loss": 0.0074, + "step": 40587 + }, + { + "epoch": 5.32, + "learning_rate": 5.342530943967531e-06, + "loss": 0.0054, + "step": 40588 + }, + { + "epoch": 5.32, + "learning_rate": 5.342061656335194e-06, + "loss": 0.0074, + "step": 40589 + }, + { + "epoch": 5.32, + "learning_rate": 5.341592381803057e-06, + "loss": 0.0079, + "step": 40590 + }, + { + "epoch": 5.32, + "learning_rate": 5.341123120372447e-06, + "loss": 0.01, + "step": 40591 + }, + { + "epoch": 5.32, + "learning_rate": 5.34065387204468e-06, + "loss": 0.0074, + "step": 40592 + }, + { + "epoch": 5.32, + "learning_rate": 5.340184636821072e-06, + "loss": 0.0101, + "step": 40593 + }, + { + "epoch": 5.32, + "learning_rate": 5.33971541470295e-06, + "loss": 0.0068, + "step": 40594 + }, + { + "epoch": 5.32, + "learning_rate": 5.339246205691631e-06, + "loss": 0.0084, + "step": 40595 + }, + { + "epoch": 5.32, + "learning_rate": 5.338777009788434e-06, + "loss": 0.011, + "step": 40596 + }, + { + "epoch": 5.32, + "learning_rate": 5.338307826994673e-06, + "loss": 0.0068, + "step": 40597 + }, + { + "epoch": 5.32, + "learning_rate": 5.337838657311676e-06, + "loss": 0.0082, + "step": 40598 + }, + { + "epoch": 5.32, + "learning_rate": 5.3373695007407615e-06, + "loss": 0.008, + "step": 40599 + }, + { + "epoch": 5.32, + "learning_rate": 5.336900357283245e-06, + "loss": 0.007, + "step": 40600 + }, + { + "epoch": 5.32, + "learning_rate": 5.336431226940447e-06, + "loss": 0.0073, + "step": 40601 + }, + { + "epoch": 5.32, + "learning_rate": 5.33596210971369e-06, + "loss": 0.0064, + "step": 40602 + }, + { + "epoch": 5.32, + "learning_rate": 5.335493005604293e-06, + "loss": 0.0079, + "step": 40603 + }, + { + "epoch": 5.32, + "learning_rate": 5.335023914613573e-06, + "loss": 0.0094, + "step": 40604 + }, + { + "epoch": 5.32, + "learning_rate": 5.334554836742844e-06, + "loss": 0.0064, + "step": 40605 + }, + { + "epoch": 5.32, + "learning_rate": 5.334085771993437e-06, + "loss": 0.0092, + "step": 40606 + }, + { + "epoch": 5.32, + "learning_rate": 5.333616720366661e-06, + "loss": 0.0066, + "step": 40607 + }, + { + "epoch": 5.32, + "learning_rate": 5.333147681863843e-06, + "loss": 0.0073, + "step": 40608 + }, + { + "epoch": 5.32, + "learning_rate": 5.332678656486295e-06, + "loss": 0.0074, + "step": 40609 + }, + { + "epoch": 5.32, + "learning_rate": 5.332209644235342e-06, + "loss": 0.0114, + "step": 40610 + }, + { + "epoch": 5.32, + "learning_rate": 5.331740645112303e-06, + "loss": 0.011, + "step": 40611 + }, + { + "epoch": 5.32, + "learning_rate": 5.331271659118495e-06, + "loss": 0.0091, + "step": 40612 + }, + { + "epoch": 5.32, + "learning_rate": 5.330802686255232e-06, + "loss": 0.0082, + "step": 40613 + }, + { + "epoch": 5.32, + "learning_rate": 5.330333726523841e-06, + "loss": 0.0106, + "step": 40614 + }, + { + "epoch": 5.32, + "learning_rate": 5.329864779925638e-06, + "loss": 0.0095, + "step": 40615 + }, + { + "epoch": 5.32, + "learning_rate": 5.3293958464619384e-06, + "loss": 0.0094, + "step": 40616 + }, + { + "epoch": 5.32, + "learning_rate": 5.328926926134068e-06, + "loss": 0.0079, + "step": 40617 + }, + { + "epoch": 5.32, + "learning_rate": 5.328458018943339e-06, + "loss": 0.0084, + "step": 40618 + }, + { + "epoch": 5.32, + "learning_rate": 5.327989124891076e-06, + "loss": 0.0083, + "step": 40619 + }, + { + "epoch": 5.32, + "learning_rate": 5.32752024397859e-06, + "loss": 0.0077, + "step": 40620 + }, + { + "epoch": 5.32, + "learning_rate": 5.327051376207209e-06, + "loss": 0.0111, + "step": 40621 + }, + { + "epoch": 5.32, + "learning_rate": 5.326582521578248e-06, + "loss": 0.0078, + "step": 40622 + }, + { + "epoch": 5.32, + "learning_rate": 5.326113680093024e-06, + "loss": 0.0074, + "step": 40623 + }, + { + "epoch": 5.32, + "learning_rate": 5.325644851752851e-06, + "loss": 0.0067, + "step": 40624 + }, + { + "epoch": 5.32, + "learning_rate": 5.325176036559059e-06, + "loss": 0.0082, + "step": 40625 + }, + { + "epoch": 5.32, + "learning_rate": 5.32470723451296e-06, + "loss": 0.0068, + "step": 40626 + }, + { + "epoch": 5.32, + "learning_rate": 5.324238445615873e-06, + "loss": 0.0104, + "step": 40627 + }, + { + "epoch": 5.32, + "learning_rate": 5.323769669869112e-06, + "loss": 0.0092, + "step": 40628 + }, + { + "epoch": 5.32, + "learning_rate": 5.323300907273999e-06, + "loss": 0.0086, + "step": 40629 + }, + { + "epoch": 5.32, + "learning_rate": 5.322832157831859e-06, + "loss": 0.0087, + "step": 40630 + }, + { + "epoch": 5.32, + "learning_rate": 5.322363421544004e-06, + "loss": 0.0091, + "step": 40631 + }, + { + "epoch": 5.32, + "learning_rate": 5.321894698411747e-06, + "loss": 0.006, + "step": 40632 + }, + { + "epoch": 5.32, + "learning_rate": 5.321425988436417e-06, + "loss": 0.0068, + "step": 40633 + }, + { + "epoch": 5.32, + "learning_rate": 5.320957291619328e-06, + "loss": 0.0071, + "step": 40634 + }, + { + "epoch": 5.32, + "learning_rate": 5.320488607961793e-06, + "loss": 0.0088, + "step": 40635 + }, + { + "epoch": 5.32, + "learning_rate": 5.320019937465139e-06, + "loss": 0.01, + "step": 40636 + }, + { + "epoch": 5.32, + "learning_rate": 5.3195512801306796e-06, + "loss": 0.0088, + "step": 40637 + }, + { + "epoch": 5.32, + "learning_rate": 5.319082635959733e-06, + "loss": 0.0052, + "step": 40638 + }, + { + "epoch": 5.32, + "learning_rate": 5.318614004953613e-06, + "loss": 0.0068, + "step": 40639 + }, + { + "epoch": 5.32, + "learning_rate": 5.318145387113641e-06, + "loss": 0.0053, + "step": 40640 + }, + { + "epoch": 5.32, + "learning_rate": 5.317676782441141e-06, + "loss": 0.0088, + "step": 40641 + }, + { + "epoch": 5.32, + "learning_rate": 5.317208190937425e-06, + "loss": 0.0063, + "step": 40642 + }, + { + "epoch": 5.32, + "learning_rate": 5.316739612603807e-06, + "loss": 0.0076, + "step": 40643 + }, + { + "epoch": 5.32, + "learning_rate": 5.316271047441615e-06, + "loss": 0.0076, + "step": 40644 + }, + { + "epoch": 5.32, + "learning_rate": 5.31580249545216e-06, + "loss": 0.0069, + "step": 40645 + }, + { + "epoch": 5.32, + "learning_rate": 5.315333956636763e-06, + "loss": 0.0069, + "step": 40646 + }, + { + "epoch": 5.32, + "learning_rate": 5.314865430996734e-06, + "loss": 0.0086, + "step": 40647 + }, + { + "epoch": 5.32, + "learning_rate": 5.3143969185334e-06, + "loss": 0.0061, + "step": 40648 + }, + { + "epoch": 5.32, + "learning_rate": 5.3139284192480755e-06, + "loss": 0.0087, + "step": 40649 + }, + { + "epoch": 5.32, + "learning_rate": 5.313459933142076e-06, + "loss": 0.0095, + "step": 40650 + }, + { + "epoch": 5.32, + "learning_rate": 5.312991460216719e-06, + "loss": 0.0066, + "step": 40651 + }, + { + "epoch": 5.33, + "learning_rate": 5.312523000473326e-06, + "loss": 0.0061, + "step": 40652 + }, + { + "epoch": 5.33, + "learning_rate": 5.312054553913215e-06, + "loss": 0.0085, + "step": 40653 + }, + { + "epoch": 5.33, + "learning_rate": 5.3115861205377005e-06, + "loss": 0.0052, + "step": 40654 + }, + { + "epoch": 5.33, + "learning_rate": 5.311117700348096e-06, + "loss": 0.0066, + "step": 40655 + }, + { + "epoch": 5.33, + "learning_rate": 5.310649293345729e-06, + "loss": 0.0105, + "step": 40656 + }, + { + "epoch": 5.33, + "learning_rate": 5.3101808995319094e-06, + "loss": 0.0066, + "step": 40657 + }, + { + "epoch": 5.33, + "learning_rate": 5.3097125189079525e-06, + "loss": 0.0079, + "step": 40658 + }, + { + "epoch": 5.33, + "learning_rate": 5.309244151475183e-06, + "loss": 0.0082, + "step": 40659 + }, + { + "epoch": 5.33, + "learning_rate": 5.308775797234915e-06, + "loss": 0.0074, + "step": 40660 + }, + { + "epoch": 5.33, + "learning_rate": 5.308307456188464e-06, + "loss": 0.0075, + "step": 40661 + }, + { + "epoch": 5.33, + "learning_rate": 5.307839128337147e-06, + "loss": 0.0054, + "step": 40662 + }, + { + "epoch": 5.33, + "learning_rate": 5.30737081368228e-06, + "loss": 0.0063, + "step": 40663 + }, + { + "epoch": 5.33, + "learning_rate": 5.3069025122251885e-06, + "loss": 0.0077, + "step": 40664 + }, + { + "epoch": 5.33, + "learning_rate": 5.306434223967183e-06, + "loss": 0.0056, + "step": 40665 + }, + { + "epoch": 5.33, + "learning_rate": 5.305965948909577e-06, + "loss": 0.0071, + "step": 40666 + }, + { + "epoch": 5.33, + "learning_rate": 5.305497687053697e-06, + "loss": 0.0097, + "step": 40667 + }, + { + "epoch": 5.33, + "learning_rate": 5.305029438400853e-06, + "loss": 0.0075, + "step": 40668 + }, + { + "epoch": 5.33, + "learning_rate": 5.304561202952365e-06, + "loss": 0.0055, + "step": 40669 + }, + { + "epoch": 5.33, + "learning_rate": 5.304092980709544e-06, + "loss": 0.0066, + "step": 40670 + }, + { + "epoch": 5.33, + "learning_rate": 5.303624771673717e-06, + "loss": 0.0064, + "step": 40671 + }, + { + "epoch": 5.33, + "learning_rate": 5.3031565758461936e-06, + "loss": 0.0087, + "step": 40672 + }, + { + "epoch": 5.33, + "learning_rate": 5.302688393228288e-06, + "loss": 0.0078, + "step": 40673 + }, + { + "epoch": 5.33, + "learning_rate": 5.302220223821321e-06, + "loss": 0.0061, + "step": 40674 + }, + { + "epoch": 5.33, + "learning_rate": 5.301752067626614e-06, + "loss": 0.0089, + "step": 40675 + }, + { + "epoch": 5.33, + "learning_rate": 5.30128392464548e-06, + "loss": 0.0074, + "step": 40676 + }, + { + "epoch": 5.33, + "learning_rate": 5.300815794879234e-06, + "loss": 0.0083, + "step": 40677 + }, + { + "epoch": 5.33, + "learning_rate": 5.3003476783291894e-06, + "loss": 0.0073, + "step": 40678 + }, + { + "epoch": 5.33, + "learning_rate": 5.29987957499667e-06, + "loss": 0.008, + "step": 40679 + }, + { + "epoch": 5.33, + "learning_rate": 5.2994114848829905e-06, + "loss": 0.0086, + "step": 40680 + }, + { + "epoch": 5.33, + "learning_rate": 5.2989434079894605e-06, + "loss": 0.0064, + "step": 40681 + }, + { + "epoch": 5.33, + "learning_rate": 5.298475344317407e-06, + "loss": 0.0103, + "step": 40682 + }, + { + "epoch": 5.33, + "learning_rate": 5.298007293868139e-06, + "loss": 0.0084, + "step": 40683 + }, + { + "epoch": 5.33, + "learning_rate": 5.297539256642973e-06, + "loss": 0.0065, + "step": 40684 + }, + { + "epoch": 5.33, + "learning_rate": 5.297071232643227e-06, + "loss": 0.0061, + "step": 40685 + }, + { + "epoch": 5.33, + "learning_rate": 5.296603221870221e-06, + "loss": 0.0086, + "step": 40686 + }, + { + "epoch": 5.33, + "learning_rate": 5.29613522432527e-06, + "loss": 0.0082, + "step": 40687 + }, + { + "epoch": 5.33, + "learning_rate": 5.2956672400096855e-06, + "loss": 0.0091, + "step": 40688 + }, + { + "epoch": 5.33, + "learning_rate": 5.295199268924785e-06, + "loss": 0.0071, + "step": 40689 + }, + { + "epoch": 5.33, + "learning_rate": 5.294731311071888e-06, + "loss": 0.0064, + "step": 40690 + }, + { + "epoch": 5.33, + "learning_rate": 5.294263366452309e-06, + "loss": 0.0064, + "step": 40691 + }, + { + "epoch": 5.33, + "learning_rate": 5.293795435067363e-06, + "loss": 0.0077, + "step": 40692 + }, + { + "epoch": 5.33, + "learning_rate": 5.293327516918365e-06, + "loss": 0.008, + "step": 40693 + }, + { + "epoch": 5.33, + "learning_rate": 5.292859612006636e-06, + "loss": 0.007, + "step": 40694 + }, + { + "epoch": 5.33, + "learning_rate": 5.292391720333483e-06, + "loss": 0.0077, + "step": 40695 + }, + { + "epoch": 5.33, + "learning_rate": 5.291923841900233e-06, + "loss": 0.0105, + "step": 40696 + }, + { + "epoch": 5.33, + "learning_rate": 5.291455976708193e-06, + "loss": 0.0075, + "step": 40697 + }, + { + "epoch": 5.33, + "learning_rate": 5.290988124758684e-06, + "loss": 0.0083, + "step": 40698 + }, + { + "epoch": 5.33, + "learning_rate": 5.290520286053022e-06, + "loss": 0.0066, + "step": 40699 + }, + { + "epoch": 5.33, + "learning_rate": 5.290052460592518e-06, + "loss": 0.0081, + "step": 40700 + }, + { + "epoch": 5.33, + "learning_rate": 5.289584648378492e-06, + "loss": 0.0074, + "step": 40701 + }, + { + "epoch": 5.33, + "learning_rate": 5.289116849412259e-06, + "loss": 0.0103, + "step": 40702 + }, + { + "epoch": 5.33, + "learning_rate": 5.288649063695135e-06, + "loss": 0.0083, + "step": 40703 + }, + { + "epoch": 5.33, + "learning_rate": 5.28818129122843e-06, + "loss": 0.0096, + "step": 40704 + }, + { + "epoch": 5.33, + "learning_rate": 5.287713532013468e-06, + "loss": 0.0074, + "step": 40705 + }, + { + "epoch": 5.33, + "learning_rate": 5.287245786051558e-06, + "loss": 0.0064, + "step": 40706 + }, + { + "epoch": 5.33, + "learning_rate": 5.286778053344021e-06, + "loss": 0.0064, + "step": 40707 + }, + { + "epoch": 5.33, + "learning_rate": 5.286310333892166e-06, + "loss": 0.0091, + "step": 40708 + }, + { + "epoch": 5.33, + "learning_rate": 5.2858426276973165e-06, + "loss": 0.0081, + "step": 40709 + }, + { + "epoch": 5.33, + "learning_rate": 5.285374934760783e-06, + "loss": 0.0063, + "step": 40710 + }, + { + "epoch": 5.33, + "learning_rate": 5.284907255083882e-06, + "loss": 0.0083, + "step": 40711 + }, + { + "epoch": 5.33, + "learning_rate": 5.284439588667924e-06, + "loss": 0.0062, + "step": 40712 + }, + { + "epoch": 5.33, + "learning_rate": 5.283971935514231e-06, + "loss": 0.0119, + "step": 40713 + }, + { + "epoch": 5.33, + "learning_rate": 5.283504295624118e-06, + "loss": 0.0071, + "step": 40714 + }, + { + "epoch": 5.33, + "learning_rate": 5.283036668998893e-06, + "loss": 0.008, + "step": 40715 + }, + { + "epoch": 5.33, + "learning_rate": 5.282569055639881e-06, + "loss": 0.0067, + "step": 40716 + }, + { + "epoch": 5.33, + "learning_rate": 5.282101455548388e-06, + "loss": 0.0103, + "step": 40717 + }, + { + "epoch": 5.33, + "learning_rate": 5.281633868725737e-06, + "loss": 0.0071, + "step": 40718 + }, + { + "epoch": 5.33, + "learning_rate": 5.28116629517324e-06, + "loss": 0.0057, + "step": 40719 + }, + { + "epoch": 5.33, + "learning_rate": 5.280698734892207e-06, + "loss": 0.0085, + "step": 40720 + }, + { + "epoch": 5.33, + "learning_rate": 5.2802311878839615e-06, + "loss": 0.0091, + "step": 40721 + }, + { + "epoch": 5.33, + "learning_rate": 5.279763654149814e-06, + "loss": 0.0066, + "step": 40722 + }, + { + "epoch": 5.33, + "learning_rate": 5.279296133691078e-06, + "loss": 0.0065, + "step": 40723 + }, + { + "epoch": 5.33, + "learning_rate": 5.2788286265090715e-06, + "loss": 0.0082, + "step": 40724 + }, + { + "epoch": 5.33, + "learning_rate": 5.2783611326051095e-06, + "loss": 0.007, + "step": 40725 + }, + { + "epoch": 5.33, + "learning_rate": 5.277893651980506e-06, + "loss": 0.0072, + "step": 40726 + }, + { + "epoch": 5.33, + "learning_rate": 5.27742618463657e-06, + "loss": 0.0075, + "step": 40727 + }, + { + "epoch": 5.34, + "learning_rate": 5.276958730574622e-06, + "loss": 0.0074, + "step": 40728 + }, + { + "epoch": 5.34, + "learning_rate": 5.27649128979598e-06, + "loss": 0.0086, + "step": 40729 + }, + { + "epoch": 5.34, + "learning_rate": 5.276023862301953e-06, + "loss": 0.0077, + "step": 40730 + }, + { + "epoch": 5.34, + "learning_rate": 5.275556448093855e-06, + "loss": 0.009, + "step": 40731 + }, + { + "epoch": 5.34, + "learning_rate": 5.275089047173008e-06, + "loss": 0.0076, + "step": 40732 + }, + { + "epoch": 5.34, + "learning_rate": 5.27462165954072e-06, + "loss": 0.0079, + "step": 40733 + }, + { + "epoch": 5.34, + "learning_rate": 5.274154285198305e-06, + "loss": 0.0061, + "step": 40734 + }, + { + "epoch": 5.34, + "learning_rate": 5.273686924147079e-06, + "loss": 0.0071, + "step": 40735 + }, + { + "epoch": 5.34, + "learning_rate": 5.273219576388358e-06, + "loss": 0.0077, + "step": 40736 + }, + { + "epoch": 5.34, + "learning_rate": 5.272752241923458e-06, + "loss": 0.0088, + "step": 40737 + }, + { + "epoch": 5.34, + "learning_rate": 5.272284920753684e-06, + "loss": 0.007, + "step": 40738 + }, + { + "epoch": 5.34, + "learning_rate": 5.271817612880359e-06, + "loss": 0.0093, + "step": 40739 + }, + { + "epoch": 5.34, + "learning_rate": 5.271350318304798e-06, + "loss": 0.0064, + "step": 40740 + }, + { + "epoch": 5.34, + "learning_rate": 5.270883037028311e-06, + "loss": 0.0103, + "step": 40741 + }, + { + "epoch": 5.34, + "learning_rate": 5.270415769052215e-06, + "loss": 0.0057, + "step": 40742 + }, + { + "epoch": 5.34, + "learning_rate": 5.26994851437782e-06, + "loss": 0.0091, + "step": 40743 + }, + { + "epoch": 5.34, + "learning_rate": 5.2694812730064445e-06, + "loss": 0.0103, + "step": 40744 + }, + { + "epoch": 5.34, + "learning_rate": 5.269014044939403e-06, + "loss": 0.0059, + "step": 40745 + }, + { + "epoch": 5.34, + "learning_rate": 5.268546830178003e-06, + "loss": 0.0068, + "step": 40746 + }, + { + "epoch": 5.34, + "learning_rate": 5.268079628723568e-06, + "loss": 0.006, + "step": 40747 + }, + { + "epoch": 5.34, + "learning_rate": 5.267612440577405e-06, + "loss": 0.006, + "step": 40748 + }, + { + "epoch": 5.34, + "learning_rate": 5.267145265740827e-06, + "loss": 0.0053, + "step": 40749 + }, + { + "epoch": 5.34, + "learning_rate": 5.2666781042151515e-06, + "loss": 0.0066, + "step": 40750 + }, + { + "epoch": 5.34, + "learning_rate": 5.266210956001695e-06, + "loss": 0.0074, + "step": 40751 + }, + { + "epoch": 5.34, + "learning_rate": 5.2657438211017675e-06, + "loss": 0.0068, + "step": 40752 + }, + { + "epoch": 5.34, + "learning_rate": 5.265276699516685e-06, + "loss": 0.009, + "step": 40753 + }, + { + "epoch": 5.34, + "learning_rate": 5.264809591247755e-06, + "loss": 0.0087, + "step": 40754 + }, + { + "epoch": 5.34, + "learning_rate": 5.2643424962962995e-06, + "loss": 0.0074, + "step": 40755 + }, + { + "epoch": 5.34, + "learning_rate": 5.263875414663629e-06, + "loss": 0.0085, + "step": 40756 + }, + { + "epoch": 5.34, + "learning_rate": 5.263408346351052e-06, + "loss": 0.0093, + "step": 40757 + }, + { + "epoch": 5.34, + "learning_rate": 5.262941291359891e-06, + "loss": 0.0091, + "step": 40758 + }, + { + "epoch": 5.34, + "learning_rate": 5.262474249691456e-06, + "loss": 0.0076, + "step": 40759 + }, + { + "epoch": 5.34, + "learning_rate": 5.262007221347055e-06, + "loss": 0.0067, + "step": 40760 + }, + { + "epoch": 5.34, + "learning_rate": 5.261540206328011e-06, + "loss": 0.0076, + "step": 40761 + }, + { + "epoch": 5.34, + "learning_rate": 5.26107320463563e-06, + "loss": 0.0043, + "step": 40762 + }, + { + "epoch": 5.34, + "learning_rate": 5.260606216271229e-06, + "loss": 0.006, + "step": 40763 + }, + { + "epoch": 5.34, + "learning_rate": 5.260139241236123e-06, + "loss": 0.0123, + "step": 40764 + }, + { + "epoch": 5.34, + "learning_rate": 5.259672279531618e-06, + "loss": 0.008, + "step": 40765 + }, + { + "epoch": 5.34, + "learning_rate": 5.2592053311590376e-06, + "loss": 0.0072, + "step": 40766 + }, + { + "epoch": 5.34, + "learning_rate": 5.258738396119689e-06, + "loss": 0.0076, + "step": 40767 + }, + { + "epoch": 5.34, + "learning_rate": 5.258271474414887e-06, + "loss": 0.0097, + "step": 40768 + }, + { + "epoch": 5.34, + "learning_rate": 5.2578045660459385e-06, + "loss": 0.0062, + "step": 40769 + }, + { + "epoch": 5.34, + "learning_rate": 5.257337671014166e-06, + "loss": 0.0072, + "step": 40770 + }, + { + "epoch": 5.34, + "learning_rate": 5.256870789320876e-06, + "loss": 0.0068, + "step": 40771 + }, + { + "epoch": 5.34, + "learning_rate": 5.256403920967388e-06, + "loss": 0.0107, + "step": 40772 + }, + { + "epoch": 5.34, + "learning_rate": 5.255937065955006e-06, + "loss": 0.0061, + "step": 40773 + }, + { + "epoch": 5.34, + "learning_rate": 5.255470224285053e-06, + "loss": 0.0063, + "step": 40774 + }, + { + "epoch": 5.34, + "learning_rate": 5.255003395958839e-06, + "loss": 0.0079, + "step": 40775 + }, + { + "epoch": 5.34, + "learning_rate": 5.254536580977674e-06, + "loss": 0.0073, + "step": 40776 + }, + { + "epoch": 5.34, + "learning_rate": 5.254069779342867e-06, + "loss": 0.0075, + "step": 40777 + }, + { + "epoch": 5.34, + "learning_rate": 5.253602991055742e-06, + "loss": 0.0095, + "step": 40778 + }, + { + "epoch": 5.34, + "learning_rate": 5.253136216117604e-06, + "loss": 0.0097, + "step": 40779 + }, + { + "epoch": 5.34, + "learning_rate": 5.252669454529767e-06, + "loss": 0.0064, + "step": 40780 + }, + { + "epoch": 5.34, + "learning_rate": 5.252202706293546e-06, + "loss": 0.0071, + "step": 40781 + }, + { + "epoch": 5.34, + "learning_rate": 5.251735971410247e-06, + "loss": 0.0059, + "step": 40782 + }, + { + "epoch": 5.34, + "learning_rate": 5.251269249881193e-06, + "loss": 0.0079, + "step": 40783 + }, + { + "epoch": 5.34, + "learning_rate": 5.250802541707691e-06, + "loss": 0.0069, + "step": 40784 + }, + { + "epoch": 5.34, + "learning_rate": 5.25033584689105e-06, + "loss": 0.0082, + "step": 40785 + }, + { + "epoch": 5.34, + "learning_rate": 5.249869165432591e-06, + "loss": 0.0084, + "step": 40786 + }, + { + "epoch": 5.34, + "learning_rate": 5.249402497333621e-06, + "loss": 0.0092, + "step": 40787 + }, + { + "epoch": 5.34, + "learning_rate": 5.248935842595451e-06, + "loss": 0.0093, + "step": 40788 + }, + { + "epoch": 5.34, + "learning_rate": 5.248469201219399e-06, + "loss": 0.0083, + "step": 40789 + }, + { + "epoch": 5.34, + "learning_rate": 5.248002573206774e-06, + "loss": 0.0102, + "step": 40790 + }, + { + "epoch": 5.34, + "learning_rate": 5.247535958558889e-06, + "loss": 0.007, + "step": 40791 + }, + { + "epoch": 5.34, + "learning_rate": 5.247069357277052e-06, + "loss": 0.0078, + "step": 40792 + }, + { + "epoch": 5.34, + "learning_rate": 5.2466027693625785e-06, + "loss": 0.0085, + "step": 40793 + }, + { + "epoch": 5.34, + "learning_rate": 5.2461361948167865e-06, + "loss": 0.0058, + "step": 40794 + }, + { + "epoch": 5.34, + "learning_rate": 5.2456696336409844e-06, + "loss": 0.0083, + "step": 40795 + }, + { + "epoch": 5.34, + "learning_rate": 5.2452030858364775e-06, + "loss": 0.007, + "step": 40796 + }, + { + "epoch": 5.34, + "learning_rate": 5.244736551404588e-06, + "loss": 0.0071, + "step": 40797 + }, + { + "epoch": 5.34, + "learning_rate": 5.244270030346624e-06, + "loss": 0.0079, + "step": 40798 + }, + { + "epoch": 5.34, + "learning_rate": 5.243803522663897e-06, + "loss": 0.0058, + "step": 40799 + }, + { + "epoch": 5.34, + "learning_rate": 5.243337028357717e-06, + "loss": 0.0077, + "step": 40800 + }, + { + "epoch": 5.34, + "learning_rate": 5.2428705474294e-06, + "loss": 0.0084, + "step": 40801 + }, + { + "epoch": 5.34, + "learning_rate": 5.242404079880258e-06, + "loss": 0.0053, + "step": 40802 + }, + { + "epoch": 5.34, + "learning_rate": 5.241937625711596e-06, + "loss": 0.0075, + "step": 40803 + }, + { + "epoch": 5.35, + "learning_rate": 5.2414711849247315e-06, + "loss": 0.0089, + "step": 40804 + }, + { + "epoch": 5.35, + "learning_rate": 5.24100475752098e-06, + "loss": 0.0064, + "step": 40805 + }, + { + "epoch": 5.35, + "learning_rate": 5.240538343501648e-06, + "loss": 0.007, + "step": 40806 + }, + { + "epoch": 5.35, + "learning_rate": 5.240071942868045e-06, + "loss": 0.0066, + "step": 40807 + }, + { + "epoch": 5.35, + "learning_rate": 5.239605555621492e-06, + "loss": 0.0067, + "step": 40808 + }, + { + "epoch": 5.35, + "learning_rate": 5.239139181763293e-06, + "loss": 0.0062, + "step": 40809 + }, + { + "epoch": 5.35, + "learning_rate": 5.238672821294762e-06, + "loss": 0.0061, + "step": 40810 + }, + { + "epoch": 5.35, + "learning_rate": 5.238206474217207e-06, + "loss": 0.0055, + "step": 40811 + }, + { + "epoch": 5.35, + "learning_rate": 5.237740140531946e-06, + "loss": 0.0089, + "step": 40812 + }, + { + "epoch": 5.35, + "learning_rate": 5.237273820240287e-06, + "loss": 0.0054, + "step": 40813 + }, + { + "epoch": 5.35, + "learning_rate": 5.236807513343541e-06, + "loss": 0.0057, + "step": 40814 + }, + { + "epoch": 5.35, + "learning_rate": 5.236341219843013e-06, + "loss": 0.0094, + "step": 40815 + }, + { + "epoch": 5.35, + "learning_rate": 5.235874939740031e-06, + "loss": 0.0078, + "step": 40816 + }, + { + "epoch": 5.35, + "learning_rate": 5.235408673035896e-06, + "loss": 0.009, + "step": 40817 + }, + { + "epoch": 5.35, + "learning_rate": 5.234942419731919e-06, + "loss": 0.0056, + "step": 40818 + }, + { + "epoch": 5.35, + "learning_rate": 5.234476179829411e-06, + "loss": 0.0061, + "step": 40819 + }, + { + "epoch": 5.35, + "learning_rate": 5.234009953329687e-06, + "loss": 0.0066, + "step": 40820 + }, + { + "epoch": 5.35, + "learning_rate": 5.233543740234057e-06, + "loss": 0.008, + "step": 40821 + }, + { + "epoch": 5.35, + "learning_rate": 5.2330775405438275e-06, + "loss": 0.0085, + "step": 40822 + }, + { + "epoch": 5.35, + "learning_rate": 5.232611354260318e-06, + "loss": 0.0071, + "step": 40823 + }, + { + "epoch": 5.35, + "learning_rate": 5.232145181384833e-06, + "loss": 0.0075, + "step": 40824 + }, + { + "epoch": 5.35, + "learning_rate": 5.231679021918687e-06, + "loss": 0.007, + "step": 40825 + }, + { + "epoch": 5.35, + "learning_rate": 5.231212875863188e-06, + "loss": 0.008, + "step": 40826 + }, + { + "epoch": 5.35, + "learning_rate": 5.230746743219647e-06, + "loss": 0.01, + "step": 40827 + }, + { + "epoch": 5.35, + "learning_rate": 5.230280623989381e-06, + "loss": 0.0096, + "step": 40828 + }, + { + "epoch": 5.35, + "learning_rate": 5.229814518173697e-06, + "loss": 0.0073, + "step": 40829 + }, + { + "epoch": 5.35, + "learning_rate": 5.229348425773902e-06, + "loss": 0.0085, + "step": 40830 + }, + { + "epoch": 5.35, + "learning_rate": 5.228882346791313e-06, + "loss": 0.006, + "step": 40831 + }, + { + "epoch": 5.35, + "learning_rate": 5.228416281227238e-06, + "loss": 0.0064, + "step": 40832 + }, + { + "epoch": 5.35, + "learning_rate": 5.22795022908299e-06, + "loss": 0.0082, + "step": 40833 + }, + { + "epoch": 5.35, + "learning_rate": 5.227484190359873e-06, + "loss": 0.0058, + "step": 40834 + }, + { + "epoch": 5.35, + "learning_rate": 5.227018165059206e-06, + "loss": 0.012, + "step": 40835 + }, + { + "epoch": 5.35, + "learning_rate": 5.226552153182297e-06, + "loss": 0.0057, + "step": 40836 + }, + { + "epoch": 5.35, + "learning_rate": 5.226086154730452e-06, + "loss": 0.0068, + "step": 40837 + }, + { + "epoch": 5.35, + "learning_rate": 5.225620169704986e-06, + "loss": 0.0073, + "step": 40838 + }, + { + "epoch": 5.35, + "learning_rate": 5.225154198107212e-06, + "loss": 0.007, + "step": 40839 + }, + { + "epoch": 5.35, + "learning_rate": 5.224688239938437e-06, + "loss": 0.0077, + "step": 40840 + }, + { + "epoch": 5.35, + "learning_rate": 5.224222295199972e-06, + "loss": 0.0057, + "step": 40841 + }, + { + "epoch": 5.35, + "learning_rate": 5.223756363893124e-06, + "loss": 0.01, + "step": 40842 + }, + { + "epoch": 5.35, + "learning_rate": 5.2232904460192114e-06, + "loss": 0.0078, + "step": 40843 + }, + { + "epoch": 5.35, + "learning_rate": 5.222824541579541e-06, + "loss": 0.0062, + "step": 40844 + }, + { + "epoch": 5.35, + "learning_rate": 5.2223586505754165e-06, + "loss": 0.0076, + "step": 40845 + }, + { + "epoch": 5.35, + "learning_rate": 5.2218927730081585e-06, + "loss": 0.0066, + "step": 40846 + }, + { + "epoch": 5.35, + "learning_rate": 5.221426908879072e-06, + "loss": 0.0108, + "step": 40847 + }, + { + "epoch": 5.35, + "learning_rate": 5.220961058189464e-06, + "loss": 0.0059, + "step": 40848 + }, + { + "epoch": 5.35, + "learning_rate": 5.220495220940653e-06, + "loss": 0.0089, + "step": 40849 + }, + { + "epoch": 5.35, + "learning_rate": 5.220029397133941e-06, + "loss": 0.0097, + "step": 40850 + }, + { + "epoch": 5.35, + "learning_rate": 5.219563586770645e-06, + "loss": 0.0068, + "step": 40851 + }, + { + "epoch": 5.35, + "learning_rate": 5.219097789852072e-06, + "loss": 0.007, + "step": 40852 + }, + { + "epoch": 5.35, + "learning_rate": 5.218632006379528e-06, + "loss": 0.0063, + "step": 40853 + }, + { + "epoch": 5.35, + "learning_rate": 5.218166236354331e-06, + "loss": 0.0075, + "step": 40854 + }, + { + "epoch": 5.35, + "learning_rate": 5.217700479777786e-06, + "loss": 0.0082, + "step": 40855 + }, + { + "epoch": 5.35, + "learning_rate": 5.2172347366512044e-06, + "loss": 0.0078, + "step": 40856 + }, + { + "epoch": 5.35, + "learning_rate": 5.216769006975892e-06, + "loss": 0.0066, + "step": 40857 + }, + { + "epoch": 5.35, + "learning_rate": 5.216303290753165e-06, + "loss": 0.004, + "step": 40858 + }, + { + "epoch": 5.35, + "learning_rate": 5.215837587984327e-06, + "loss": 0.0082, + "step": 40859 + }, + { + "epoch": 5.35, + "learning_rate": 5.215371898670695e-06, + "loss": 0.0078, + "step": 40860 + }, + { + "epoch": 5.35, + "learning_rate": 5.21490622281357e-06, + "loss": 0.009, + "step": 40861 + }, + { + "epoch": 5.35, + "learning_rate": 5.214440560414271e-06, + "loss": 0.0096, + "step": 40862 + }, + { + "epoch": 5.35, + "learning_rate": 5.213974911474103e-06, + "loss": 0.0085, + "step": 40863 + }, + { + "epoch": 5.35, + "learning_rate": 5.213509275994375e-06, + "loss": 0.0072, + "step": 40864 + }, + { + "epoch": 5.35, + "learning_rate": 5.213043653976394e-06, + "loss": 0.0061, + "step": 40865 + }, + { + "epoch": 5.35, + "learning_rate": 5.2125780454214766e-06, + "loss": 0.0056, + "step": 40866 + }, + { + "epoch": 5.35, + "learning_rate": 5.212112450330929e-06, + "loss": 0.0094, + "step": 40867 + }, + { + "epoch": 5.35, + "learning_rate": 5.211646868706055e-06, + "loss": 0.0115, + "step": 40868 + }, + { + "epoch": 5.35, + "learning_rate": 5.211181300548171e-06, + "loss": 0.0128, + "step": 40869 + }, + { + "epoch": 5.35, + "learning_rate": 5.210715745858588e-06, + "loss": 0.0072, + "step": 40870 + }, + { + "epoch": 5.35, + "learning_rate": 5.2102502046386104e-06, + "loss": 0.0097, + "step": 40871 + }, + { + "epoch": 5.35, + "learning_rate": 5.209784676889547e-06, + "loss": 0.0086, + "step": 40872 + }, + { + "epoch": 5.35, + "learning_rate": 5.2093191626127115e-06, + "loss": 0.0067, + "step": 40873 + }, + { + "epoch": 5.35, + "learning_rate": 5.208853661809411e-06, + "loss": 0.0069, + "step": 40874 + }, + { + "epoch": 5.35, + "learning_rate": 5.208388174480955e-06, + "loss": 0.006, + "step": 40875 + }, + { + "epoch": 5.35, + "learning_rate": 5.207922700628648e-06, + "loss": 0.009, + "step": 40876 + }, + { + "epoch": 5.35, + "learning_rate": 5.207457240253807e-06, + "loss": 0.0097, + "step": 40877 + }, + { + "epoch": 5.35, + "learning_rate": 5.206991793357737e-06, + "loss": 0.0069, + "step": 40878 + }, + { + "epoch": 5.35, + "learning_rate": 5.206526359941748e-06, + "loss": 0.0096, + "step": 40879 + }, + { + "epoch": 5.35, + "learning_rate": 5.206060940007141e-06, + "loss": 0.0077, + "step": 40880 + }, + { + "epoch": 5.36, + "learning_rate": 5.205595533555239e-06, + "loss": 0.0064, + "step": 40881 + }, + { + "epoch": 5.36, + "learning_rate": 5.2051301405873445e-06, + "loss": 0.0099, + "step": 40882 + }, + { + "epoch": 5.36, + "learning_rate": 5.204664761104766e-06, + "loss": 0.0071, + "step": 40883 + }, + { + "epoch": 5.36, + "learning_rate": 5.204199395108809e-06, + "loss": 0.0068, + "step": 40884 + }, + { + "epoch": 5.36, + "learning_rate": 5.20373404260079e-06, + "loss": 0.0067, + "step": 40885 + }, + { + "epoch": 5.36, + "learning_rate": 5.203268703582013e-06, + "loss": 0.0063, + "step": 40886 + }, + { + "epoch": 5.36, + "learning_rate": 5.202803378053784e-06, + "loss": 0.0049, + "step": 40887 + }, + { + "epoch": 5.36, + "learning_rate": 5.202338066017418e-06, + "loss": 0.0097, + "step": 40888 + }, + { + "epoch": 5.36, + "learning_rate": 5.201872767474221e-06, + "loss": 0.0084, + "step": 40889 + }, + { + "epoch": 5.36, + "learning_rate": 5.2014074824255025e-06, + "loss": 0.0082, + "step": 40890 + }, + { + "epoch": 5.36, + "learning_rate": 5.2009422108725634e-06, + "loss": 0.0082, + "step": 40891 + }, + { + "epoch": 5.36, + "learning_rate": 5.200476952816721e-06, + "loss": 0.0073, + "step": 40892 + }, + { + "epoch": 5.36, + "learning_rate": 5.200011708259285e-06, + "loss": 0.0073, + "step": 40893 + }, + { + "epoch": 5.36, + "learning_rate": 5.19954647720156e-06, + "loss": 0.008, + "step": 40894 + }, + { + "epoch": 5.36, + "learning_rate": 5.199081259644851e-06, + "loss": 0.0066, + "step": 40895 + }, + { + "epoch": 5.36, + "learning_rate": 5.198616055590475e-06, + "loss": 0.0099, + "step": 40896 + }, + { + "epoch": 5.36, + "learning_rate": 5.198150865039735e-06, + "loss": 0.0085, + "step": 40897 + }, + { + "epoch": 5.36, + "learning_rate": 5.197685687993941e-06, + "loss": 0.0059, + "step": 40898 + }, + { + "epoch": 5.36, + "learning_rate": 5.197220524454395e-06, + "loss": 0.0084, + "step": 40899 + }, + { + "epoch": 5.36, + "learning_rate": 5.196755374422415e-06, + "loss": 0.0084, + "step": 40900 + }, + { + "epoch": 5.36, + "learning_rate": 5.196290237899304e-06, + "loss": 0.0058, + "step": 40901 + }, + { + "epoch": 5.36, + "learning_rate": 5.195825114886368e-06, + "loss": 0.0063, + "step": 40902 + }, + { + "epoch": 5.36, + "learning_rate": 5.195360005384919e-06, + "loss": 0.0076, + "step": 40903 + }, + { + "epoch": 5.36, + "learning_rate": 5.194894909396267e-06, + "loss": 0.0069, + "step": 40904 + }, + { + "epoch": 5.36, + "learning_rate": 5.194429826921716e-06, + "loss": 0.0083, + "step": 40905 + }, + { + "epoch": 5.36, + "learning_rate": 5.1939647579625756e-06, + "loss": 0.007, + "step": 40906 + }, + { + "epoch": 5.36, + "learning_rate": 5.1934997025201505e-06, + "loss": 0.0062, + "step": 40907 + }, + { + "epoch": 5.36, + "learning_rate": 5.193034660595755e-06, + "loss": 0.0075, + "step": 40908 + }, + { + "epoch": 5.36, + "learning_rate": 5.192569632190693e-06, + "loss": 0.0088, + "step": 40909 + }, + { + "epoch": 5.36, + "learning_rate": 5.192104617306269e-06, + "loss": 0.0079, + "step": 40910 + }, + { + "epoch": 5.36, + "learning_rate": 5.1916396159437995e-06, + "loss": 0.0075, + "step": 40911 + }, + { + "epoch": 5.36, + "learning_rate": 5.191174628104587e-06, + "loss": 0.0101, + "step": 40912 + }, + { + "epoch": 5.36, + "learning_rate": 5.190709653789936e-06, + "loss": 0.0075, + "step": 40913 + }, + { + "epoch": 5.36, + "learning_rate": 5.190244693001162e-06, + "loss": 0.0064, + "step": 40914 + }, + { + "epoch": 5.36, + "learning_rate": 5.189779745739565e-06, + "loss": 0.0046, + "step": 40915 + }, + { + "epoch": 5.36, + "learning_rate": 5.18931481200646e-06, + "loss": 0.0085, + "step": 40916 + }, + { + "epoch": 5.36, + "learning_rate": 5.188849891803151e-06, + "loss": 0.0052, + "step": 40917 + }, + { + "epoch": 5.36, + "learning_rate": 5.1883849851309405e-06, + "loss": 0.0072, + "step": 40918 + }, + { + "epoch": 5.36, + "learning_rate": 5.1879200919911455e-06, + "loss": 0.0071, + "step": 40919 + }, + { + "epoch": 5.36, + "learning_rate": 5.187455212385069e-06, + "loss": 0.0064, + "step": 40920 + }, + { + "epoch": 5.36, + "learning_rate": 5.186990346314019e-06, + "loss": 0.0072, + "step": 40921 + }, + { + "epoch": 5.36, + "learning_rate": 5.186525493779298e-06, + "loss": 0.0096, + "step": 40922 + }, + { + "epoch": 5.36, + "learning_rate": 5.186060654782221e-06, + "loss": 0.0055, + "step": 40923 + }, + { + "epoch": 5.36, + "learning_rate": 5.185595829324089e-06, + "loss": 0.0069, + "step": 40924 + }, + { + "epoch": 5.36, + "learning_rate": 5.185131017406217e-06, + "loss": 0.0069, + "step": 40925 + }, + { + "epoch": 5.36, + "learning_rate": 5.184666219029902e-06, + "loss": 0.0089, + "step": 40926 + }, + { + "epoch": 5.36, + "learning_rate": 5.184201434196462e-06, + "loss": 0.0072, + "step": 40927 + }, + { + "epoch": 5.36, + "learning_rate": 5.183736662907198e-06, + "loss": 0.0082, + "step": 40928 + }, + { + "epoch": 5.36, + "learning_rate": 5.183271905163418e-06, + "loss": 0.0072, + "step": 40929 + }, + { + "epoch": 5.36, + "learning_rate": 5.182807160966428e-06, + "loss": 0.0066, + "step": 40930 + }, + { + "epoch": 5.36, + "learning_rate": 5.182342430317538e-06, + "loss": 0.0073, + "step": 40931 + }, + { + "epoch": 5.36, + "learning_rate": 5.181877713218053e-06, + "loss": 0.0073, + "step": 40932 + }, + { + "epoch": 5.36, + "learning_rate": 5.181413009669277e-06, + "loss": 0.0066, + "step": 40933 + }, + { + "epoch": 5.36, + "learning_rate": 5.180948319672525e-06, + "loss": 0.0077, + "step": 40934 + }, + { + "epoch": 5.36, + "learning_rate": 5.180483643229094e-06, + "loss": 0.0066, + "step": 40935 + }, + { + "epoch": 5.36, + "learning_rate": 5.1800189803403e-06, + "loss": 0.007, + "step": 40936 + }, + { + "epoch": 5.36, + "learning_rate": 5.179554331007444e-06, + "loss": 0.0074, + "step": 40937 + }, + { + "epoch": 5.36, + "learning_rate": 5.179089695231836e-06, + "loss": 0.0061, + "step": 40938 + }, + { + "epoch": 5.36, + "learning_rate": 5.178625073014782e-06, + "loss": 0.0063, + "step": 40939 + }, + { + "epoch": 5.36, + "learning_rate": 5.17816046435759e-06, + "loss": 0.0084, + "step": 40940 + }, + { + "epoch": 5.36, + "learning_rate": 5.177695869261561e-06, + "loss": 0.0068, + "step": 40941 + }, + { + "epoch": 5.36, + "learning_rate": 5.1772312877280086e-06, + "loss": 0.0073, + "step": 40942 + }, + { + "epoch": 5.36, + "learning_rate": 5.176766719758235e-06, + "loss": 0.007, + "step": 40943 + }, + { + "epoch": 5.36, + "learning_rate": 5.176302165353551e-06, + "loss": 0.0079, + "step": 40944 + }, + { + "epoch": 5.36, + "learning_rate": 5.175837624515254e-06, + "loss": 0.0071, + "step": 40945 + }, + { + "epoch": 5.36, + "learning_rate": 5.175373097244657e-06, + "loss": 0.0083, + "step": 40946 + }, + { + "epoch": 5.36, + "learning_rate": 5.17490858354307e-06, + "loss": 0.0045, + "step": 40947 + }, + { + "epoch": 5.36, + "learning_rate": 5.174444083411797e-06, + "loss": 0.0072, + "step": 40948 + }, + { + "epoch": 5.36, + "learning_rate": 5.173979596852139e-06, + "loss": 0.0077, + "step": 40949 + }, + { + "epoch": 5.36, + "learning_rate": 5.173515123865409e-06, + "loss": 0.007, + "step": 40950 + }, + { + "epoch": 5.36, + "learning_rate": 5.173050664452912e-06, + "loss": 0.0066, + "step": 40951 + }, + { + "epoch": 5.36, + "learning_rate": 5.172586218615949e-06, + "loss": 0.0066, + "step": 40952 + }, + { + "epoch": 5.36, + "learning_rate": 5.172121786355833e-06, + "loss": 0.0087, + "step": 40953 + }, + { + "epoch": 5.36, + "learning_rate": 5.171657367673868e-06, + "loss": 0.0049, + "step": 40954 + }, + { + "epoch": 5.36, + "learning_rate": 5.1711929625713595e-06, + "loss": 0.007, + "step": 40955 + }, + { + "epoch": 5.36, + "learning_rate": 5.170728571049611e-06, + "loss": 0.0075, + "step": 40956 + }, + { + "epoch": 5.37, + "learning_rate": 5.170264193109928e-06, + "loss": 0.0113, + "step": 40957 + }, + { + "epoch": 5.37, + "learning_rate": 5.169799828753626e-06, + "loss": 0.0068, + "step": 40958 + }, + { + "epoch": 5.37, + "learning_rate": 5.169335477982005e-06, + "loss": 0.0061, + "step": 40959 + }, + { + "epoch": 5.37, + "learning_rate": 5.1688711407963665e-06, + "loss": 0.0109, + "step": 40960 + }, + { + "epoch": 5.37, + "learning_rate": 5.168406817198024e-06, + "loss": 0.0067, + "step": 40961 + }, + { + "epoch": 5.37, + "learning_rate": 5.16794250718828e-06, + "loss": 0.0082, + "step": 40962 + }, + { + "epoch": 5.37, + "learning_rate": 5.1674782107684414e-06, + "loss": 0.0064, + "step": 40963 + }, + { + "epoch": 5.37, + "learning_rate": 5.167013927939808e-06, + "loss": 0.0065, + "step": 40964 + }, + { + "epoch": 5.37, + "learning_rate": 5.1665496587036946e-06, + "loss": 0.0064, + "step": 40965 + }, + { + "epoch": 5.37, + "learning_rate": 5.166085403061402e-06, + "loss": 0.0071, + "step": 40966 + }, + { + "epoch": 5.37, + "learning_rate": 5.165621161014236e-06, + "loss": 0.0058, + "step": 40967 + }, + { + "epoch": 5.37, + "learning_rate": 5.1651569325635e-06, + "loss": 0.0038, + "step": 40968 + }, + { + "epoch": 5.37, + "learning_rate": 5.164692717710508e-06, + "loss": 0.0066, + "step": 40969 + }, + { + "epoch": 5.37, + "learning_rate": 5.164228516456559e-06, + "loss": 0.0099, + "step": 40970 + }, + { + "epoch": 5.37, + "learning_rate": 5.163764328802962e-06, + "loss": 0.0093, + "step": 40971 + }, + { + "epoch": 5.37, + "learning_rate": 5.163300154751014e-06, + "loss": 0.0061, + "step": 40972 + }, + { + "epoch": 5.37, + "learning_rate": 5.1628359943020315e-06, + "loss": 0.007, + "step": 40973 + }, + { + "epoch": 5.37, + "learning_rate": 5.162371847457316e-06, + "loss": 0.0064, + "step": 40974 + }, + { + "epoch": 5.37, + "learning_rate": 5.161907714218167e-06, + "loss": 0.0085, + "step": 40975 + }, + { + "epoch": 5.37, + "learning_rate": 5.1614435945859e-06, + "loss": 0.0065, + "step": 40976 + }, + { + "epoch": 5.37, + "learning_rate": 5.1609794885618144e-06, + "loss": 0.006, + "step": 40977 + }, + { + "epoch": 5.37, + "learning_rate": 5.160515396147216e-06, + "loss": 0.0073, + "step": 40978 + }, + { + "epoch": 5.37, + "learning_rate": 5.160051317343406e-06, + "loss": 0.0099, + "step": 40979 + }, + { + "epoch": 5.37, + "learning_rate": 5.159587252151696e-06, + "loss": 0.0095, + "step": 40980 + }, + { + "epoch": 5.37, + "learning_rate": 5.15912320057339e-06, + "loss": 0.0058, + "step": 40981 + }, + { + "epoch": 5.37, + "learning_rate": 5.158659162609796e-06, + "loss": 0.0067, + "step": 40982 + }, + { + "epoch": 5.37, + "learning_rate": 5.158195138262208e-06, + "loss": 0.0078, + "step": 40983 + }, + { + "epoch": 5.37, + "learning_rate": 5.157731127531944e-06, + "loss": 0.0071, + "step": 40984 + }, + { + "epoch": 5.37, + "learning_rate": 5.1572671304203025e-06, + "loss": 0.0075, + "step": 40985 + }, + { + "epoch": 5.37, + "learning_rate": 5.156803146928591e-06, + "loss": 0.0091, + "step": 40986 + }, + { + "epoch": 5.37, + "learning_rate": 5.1563391770581075e-06, + "loss": 0.0098, + "step": 40987 + }, + { + "epoch": 5.37, + "learning_rate": 5.1558752208101645e-06, + "loss": 0.0062, + "step": 40988 + }, + { + "epoch": 5.37, + "learning_rate": 5.155411278186066e-06, + "loss": 0.0073, + "step": 40989 + }, + { + "epoch": 5.37, + "learning_rate": 5.154947349187111e-06, + "loss": 0.0062, + "step": 40990 + }, + { + "epoch": 5.37, + "learning_rate": 5.15448343381461e-06, + "loss": 0.0076, + "step": 40991 + }, + { + "epoch": 5.37, + "learning_rate": 5.1540195320698695e-06, + "loss": 0.0085, + "step": 40992 + }, + { + "epoch": 5.37, + "learning_rate": 5.153555643954191e-06, + "loss": 0.0088, + "step": 40993 + }, + { + "epoch": 5.37, + "learning_rate": 5.153091769468878e-06, + "loss": 0.0065, + "step": 40994 + }, + { + "epoch": 5.37, + "learning_rate": 5.1526279086152335e-06, + "loss": 0.0071, + "step": 40995 + }, + { + "epoch": 5.37, + "learning_rate": 5.152164061394569e-06, + "loss": 0.0067, + "step": 40996 + }, + { + "epoch": 5.37, + "learning_rate": 5.151700227808184e-06, + "loss": 0.0069, + "step": 40997 + }, + { + "epoch": 5.37, + "learning_rate": 5.151236407857381e-06, + "loss": 0.0094, + "step": 40998 + }, + { + "epoch": 5.37, + "learning_rate": 5.15077260154347e-06, + "loss": 0.0062, + "step": 40999 + }, + { + "epoch": 5.37, + "learning_rate": 5.15030880886775e-06, + "loss": 0.0063, + "step": 41000 + }, + { + "epoch": 5.37, + "learning_rate": 5.149845029831531e-06, + "loss": 0.0065, + "step": 41001 + }, + { + "epoch": 5.37, + "learning_rate": 5.149381264436113e-06, + "loss": 0.0075, + "step": 41002 + }, + { + "epoch": 5.37, + "learning_rate": 5.1489175126828025e-06, + "loss": 0.0044, + "step": 41003 + }, + { + "epoch": 5.37, + "learning_rate": 5.148453774572906e-06, + "loss": 0.0054, + "step": 41004 + }, + { + "epoch": 5.37, + "learning_rate": 5.147990050107723e-06, + "loss": 0.0073, + "step": 41005 + }, + { + "epoch": 5.37, + "learning_rate": 5.147526339288558e-06, + "loss": 0.0066, + "step": 41006 + }, + { + "epoch": 5.37, + "learning_rate": 5.147062642116718e-06, + "loss": 0.0103, + "step": 41007 + }, + { + "epoch": 5.37, + "learning_rate": 5.146598958593507e-06, + "loss": 0.0097, + "step": 41008 + }, + { + "epoch": 5.37, + "learning_rate": 5.1461352887202286e-06, + "loss": 0.0059, + "step": 41009 + }, + { + "epoch": 5.37, + "learning_rate": 5.145671632498182e-06, + "loss": 0.007, + "step": 41010 + }, + { + "epoch": 5.37, + "learning_rate": 5.145207989928675e-06, + "loss": 0.0064, + "step": 41011 + }, + { + "epoch": 5.37, + "learning_rate": 5.144744361013016e-06, + "loss": 0.0077, + "step": 41012 + }, + { + "epoch": 5.37, + "learning_rate": 5.144280745752505e-06, + "loss": 0.0071, + "step": 41013 + }, + { + "epoch": 5.37, + "learning_rate": 5.143817144148442e-06, + "loss": 0.0053, + "step": 41014 + }, + { + "epoch": 5.37, + "learning_rate": 5.143353556202138e-06, + "loss": 0.0098, + "step": 41015 + }, + { + "epoch": 5.37, + "learning_rate": 5.142889981914893e-06, + "loss": 0.0071, + "step": 41016 + }, + { + "epoch": 5.37, + "learning_rate": 5.142426421288009e-06, + "loss": 0.0079, + "step": 41017 + }, + { + "epoch": 5.37, + "learning_rate": 5.141962874322796e-06, + "loss": 0.0044, + "step": 41018 + }, + { + "epoch": 5.37, + "learning_rate": 5.141499341020553e-06, + "loss": 0.0062, + "step": 41019 + }, + { + "epoch": 5.37, + "learning_rate": 5.141035821382584e-06, + "loss": 0.0084, + "step": 41020 + }, + { + "epoch": 5.37, + "learning_rate": 5.140572315410188e-06, + "loss": 0.0089, + "step": 41021 + }, + { + "epoch": 5.37, + "learning_rate": 5.140108823104676e-06, + "loss": 0.0099, + "step": 41022 + }, + { + "epoch": 5.37, + "learning_rate": 5.139645344467352e-06, + "loss": 0.0071, + "step": 41023 + }, + { + "epoch": 5.37, + "learning_rate": 5.139181879499517e-06, + "loss": 0.0079, + "step": 41024 + }, + { + "epoch": 5.37, + "learning_rate": 5.13871842820247e-06, + "loss": 0.0087, + "step": 41025 + }, + { + "epoch": 5.37, + "learning_rate": 5.138254990577522e-06, + "loss": 0.0071, + "step": 41026 + }, + { + "epoch": 5.37, + "learning_rate": 5.137791566625974e-06, + "loss": 0.0083, + "step": 41027 + }, + { + "epoch": 5.37, + "learning_rate": 5.137328156349126e-06, + "loss": 0.0062, + "step": 41028 + }, + { + "epoch": 5.37, + "learning_rate": 5.136864759748281e-06, + "loss": 0.0055, + "step": 41029 + }, + { + "epoch": 5.37, + "learning_rate": 5.1364013768247505e-06, + "loss": 0.0065, + "step": 41030 + }, + { + "epoch": 5.37, + "learning_rate": 5.13593800757983e-06, + "loss": 0.0058, + "step": 41031 + }, + { + "epoch": 5.37, + "learning_rate": 5.135474652014822e-06, + "loss": 0.0052, + "step": 41032 + }, + { + "epoch": 5.38, + "learning_rate": 5.1350113101310315e-06, + "loss": 0.0079, + "step": 41033 + }, + { + "epoch": 5.38, + "learning_rate": 5.134547981929767e-06, + "loss": 0.008, + "step": 41034 + }, + { + "epoch": 5.38, + "learning_rate": 5.134084667412328e-06, + "loss": 0.0056, + "step": 41035 + }, + { + "epoch": 5.38, + "learning_rate": 5.133621366580016e-06, + "loss": 0.0081, + "step": 41036 + }, + { + "epoch": 5.38, + "learning_rate": 5.133158079434131e-06, + "loss": 0.0096, + "step": 41037 + }, + { + "epoch": 5.38, + "learning_rate": 5.132694805975984e-06, + "loss": 0.0115, + "step": 41038 + }, + { + "epoch": 5.38, + "learning_rate": 5.132231546206871e-06, + "loss": 0.0072, + "step": 41039 + }, + { + "epoch": 5.38, + "learning_rate": 5.131768300128097e-06, + "loss": 0.0077, + "step": 41040 + }, + { + "epoch": 5.38, + "learning_rate": 5.1313050677409685e-06, + "loss": 0.0076, + "step": 41041 + }, + { + "epoch": 5.38, + "learning_rate": 5.130841849046785e-06, + "loss": 0.008, + "step": 41042 + }, + { + "epoch": 5.38, + "learning_rate": 5.1303786440468495e-06, + "loss": 0.0073, + "step": 41043 + }, + { + "epoch": 5.38, + "learning_rate": 5.12991545274246e-06, + "loss": 0.0086, + "step": 41044 + }, + { + "epoch": 5.38, + "learning_rate": 5.129452275134925e-06, + "loss": 0.009, + "step": 41045 + }, + { + "epoch": 5.38, + "learning_rate": 5.12898911122555e-06, + "loss": 0.0081, + "step": 41046 + }, + { + "epoch": 5.38, + "learning_rate": 5.128525961015633e-06, + "loss": 0.0103, + "step": 41047 + }, + { + "epoch": 5.38, + "learning_rate": 5.128062824506475e-06, + "loss": 0.0097, + "step": 41048 + }, + { + "epoch": 5.38, + "learning_rate": 5.127599701699383e-06, + "loss": 0.0049, + "step": 41049 + }, + { + "epoch": 5.38, + "learning_rate": 5.1271365925956575e-06, + "loss": 0.0088, + "step": 41050 + }, + { + "epoch": 5.38, + "learning_rate": 5.126673497196602e-06, + "loss": 0.0072, + "step": 41051 + }, + { + "epoch": 5.38, + "learning_rate": 5.126210415503512e-06, + "loss": 0.0074, + "step": 41052 + }, + { + "epoch": 5.38, + "learning_rate": 5.125747347517701e-06, + "loss": 0.0094, + "step": 41053 + }, + { + "epoch": 5.38, + "learning_rate": 5.125284293240465e-06, + "loss": 0.0085, + "step": 41054 + }, + { + "epoch": 5.38, + "learning_rate": 5.124821252673104e-06, + "loss": 0.0099, + "step": 41055 + }, + { + "epoch": 5.38, + "learning_rate": 5.124358225816924e-06, + "loss": 0.0073, + "step": 41056 + }, + { + "epoch": 5.38, + "learning_rate": 5.123895212673231e-06, + "loss": 0.0064, + "step": 41057 + }, + { + "epoch": 5.38, + "learning_rate": 5.123432213243323e-06, + "loss": 0.007, + "step": 41058 + }, + { + "epoch": 5.38, + "learning_rate": 5.122969227528501e-06, + "loss": 0.0057, + "step": 41059 + }, + { + "epoch": 5.38, + "learning_rate": 5.1225062555300665e-06, + "loss": 0.0101, + "step": 41060 + }, + { + "epoch": 5.38, + "learning_rate": 5.122043297249325e-06, + "loss": 0.009, + "step": 41061 + }, + { + "epoch": 5.38, + "learning_rate": 5.121580352687578e-06, + "loss": 0.0069, + "step": 41062 + }, + { + "epoch": 5.38, + "learning_rate": 5.121117421846123e-06, + "loss": 0.0077, + "step": 41063 + }, + { + "epoch": 5.38, + "learning_rate": 5.120654504726268e-06, + "loss": 0.0065, + "step": 41064 + }, + { + "epoch": 5.38, + "learning_rate": 5.120191601329314e-06, + "loss": 0.0071, + "step": 41065 + }, + { + "epoch": 5.38, + "learning_rate": 5.119728711656556e-06, + "loss": 0.0076, + "step": 41066 + }, + { + "epoch": 5.38, + "learning_rate": 5.119265835709302e-06, + "loss": 0.0065, + "step": 41067 + }, + { + "epoch": 5.38, + "learning_rate": 5.118802973488856e-06, + "loss": 0.0067, + "step": 41068 + }, + { + "epoch": 5.38, + "learning_rate": 5.118340124996517e-06, + "loss": 0.0066, + "step": 41069 + }, + { + "epoch": 5.38, + "learning_rate": 5.117877290233588e-06, + "loss": 0.0084, + "step": 41070 + }, + { + "epoch": 5.38, + "learning_rate": 5.117414469201363e-06, + "loss": 0.0091, + "step": 41071 + }, + { + "epoch": 5.38, + "learning_rate": 5.116951661901155e-06, + "loss": 0.0069, + "step": 41072 + }, + { + "epoch": 5.38, + "learning_rate": 5.1164888683342594e-06, + "loss": 0.0065, + "step": 41073 + }, + { + "epoch": 5.38, + "learning_rate": 5.116026088501975e-06, + "loss": 0.0074, + "step": 41074 + }, + { + "epoch": 5.38, + "learning_rate": 5.115563322405612e-06, + "loss": 0.0065, + "step": 41075 + }, + { + "epoch": 5.38, + "learning_rate": 5.115100570046467e-06, + "loss": 0.0101, + "step": 41076 + }, + { + "epoch": 5.38, + "learning_rate": 5.114637831425837e-06, + "loss": 0.0075, + "step": 41077 + }, + { + "epoch": 5.38, + "learning_rate": 5.114175106545032e-06, + "loss": 0.0081, + "step": 41078 + }, + { + "epoch": 5.38, + "learning_rate": 5.113712395405346e-06, + "loss": 0.006, + "step": 41079 + }, + { + "epoch": 5.38, + "learning_rate": 5.113249698008087e-06, + "loss": 0.0072, + "step": 41080 + }, + { + "epoch": 5.38, + "learning_rate": 5.112787014354552e-06, + "loss": 0.0078, + "step": 41081 + }, + { + "epoch": 5.38, + "learning_rate": 5.112324344446041e-06, + "loss": 0.0061, + "step": 41082 + }, + { + "epoch": 5.38, + "learning_rate": 5.111861688283861e-06, + "loss": 0.0083, + "step": 41083 + }, + { + "epoch": 5.38, + "learning_rate": 5.11139904586931e-06, + "loss": 0.0083, + "step": 41084 + }, + { + "epoch": 5.38, + "learning_rate": 5.110936417203687e-06, + "loss": 0.006, + "step": 41085 + }, + { + "epoch": 5.38, + "learning_rate": 5.110473802288294e-06, + "loss": 0.0073, + "step": 41086 + }, + { + "epoch": 5.38, + "learning_rate": 5.110011201124434e-06, + "loss": 0.0063, + "step": 41087 + }, + { + "epoch": 5.38, + "learning_rate": 5.109548613713404e-06, + "loss": 0.0079, + "step": 41088 + }, + { + "epoch": 5.38, + "learning_rate": 5.109086040056513e-06, + "loss": 0.0064, + "step": 41089 + }, + { + "epoch": 5.38, + "learning_rate": 5.108623480155053e-06, + "loss": 0.0084, + "step": 41090 + }, + { + "epoch": 5.38, + "learning_rate": 5.108160934010333e-06, + "loss": 0.0073, + "step": 41091 + }, + { + "epoch": 5.38, + "learning_rate": 5.107698401623649e-06, + "loss": 0.0071, + "step": 41092 + }, + { + "epoch": 5.38, + "learning_rate": 5.107235882996303e-06, + "loss": 0.0095, + "step": 41093 + }, + { + "epoch": 5.38, + "learning_rate": 5.106773378129592e-06, + "loss": 0.0096, + "step": 41094 + }, + { + "epoch": 5.38, + "learning_rate": 5.106310887024823e-06, + "loss": 0.0082, + "step": 41095 + }, + { + "epoch": 5.38, + "learning_rate": 5.1058484096832964e-06, + "loss": 0.0077, + "step": 41096 + }, + { + "epoch": 5.38, + "learning_rate": 5.105385946106305e-06, + "loss": 0.0105, + "step": 41097 + }, + { + "epoch": 5.38, + "learning_rate": 5.1049234962951585e-06, + "loss": 0.0063, + "step": 41098 + }, + { + "epoch": 5.38, + "learning_rate": 5.104461060251151e-06, + "loss": 0.0059, + "step": 41099 + }, + { + "epoch": 5.38, + "learning_rate": 5.103998637975589e-06, + "loss": 0.0064, + "step": 41100 + }, + { + "epoch": 5.38, + "learning_rate": 5.103536229469771e-06, + "loss": 0.0082, + "step": 41101 + }, + { + "epoch": 5.38, + "learning_rate": 5.1030738347349926e-06, + "loss": 0.008, + "step": 41102 + }, + { + "epoch": 5.38, + "learning_rate": 5.102611453772561e-06, + "loss": 0.0064, + "step": 41103 + }, + { + "epoch": 5.38, + "learning_rate": 5.102149086583776e-06, + "loss": 0.0065, + "step": 41104 + }, + { + "epoch": 5.38, + "learning_rate": 5.101686733169931e-06, + "loss": 0.0063, + "step": 41105 + }, + { + "epoch": 5.38, + "learning_rate": 5.101224393532335e-06, + "loss": 0.0069, + "step": 41106 + }, + { + "epoch": 5.38, + "learning_rate": 5.100762067672285e-06, + "loss": 0.0069, + "step": 41107 + }, + { + "epoch": 5.38, + "learning_rate": 5.100299755591079e-06, + "loss": 0.0103, + "step": 41108 + }, + { + "epoch": 5.38, + "learning_rate": 5.099837457290017e-06, + "loss": 0.0065, + "step": 41109 + }, + { + "epoch": 5.39, + "learning_rate": 5.0993751727704e-06, + "loss": 0.0068, + "step": 41110 + }, + { + "epoch": 5.39, + "learning_rate": 5.098912902033535e-06, + "loss": 0.0085, + "step": 41111 + }, + { + "epoch": 5.39, + "learning_rate": 5.098450645080715e-06, + "loss": 0.0064, + "step": 41112 + }, + { + "epoch": 5.39, + "learning_rate": 5.097988401913238e-06, + "loss": 0.0102, + "step": 41113 + }, + { + "epoch": 5.39, + "learning_rate": 5.0975261725324106e-06, + "loss": 0.0077, + "step": 41114 + }, + { + "epoch": 5.39, + "learning_rate": 5.097063956939531e-06, + "loss": 0.0082, + "step": 41115 + }, + { + "epoch": 5.39, + "learning_rate": 5.096601755135897e-06, + "loss": 0.0078, + "step": 41116 + }, + { + "epoch": 5.39, + "learning_rate": 5.0961395671228065e-06, + "loss": 0.0074, + "step": 41117 + }, + { + "epoch": 5.39, + "learning_rate": 5.095677392901566e-06, + "loss": 0.0062, + "step": 41118 + }, + { + "epoch": 5.39, + "learning_rate": 5.095215232473471e-06, + "loss": 0.008, + "step": 41119 + }, + { + "epoch": 5.39, + "learning_rate": 5.094753085839819e-06, + "loss": 0.0061, + "step": 41120 + }, + { + "epoch": 5.39, + "learning_rate": 5.094290953001913e-06, + "loss": 0.0101, + "step": 41121 + }, + { + "epoch": 5.39, + "learning_rate": 5.093828833961054e-06, + "loss": 0.0047, + "step": 41122 + }, + { + "epoch": 5.39, + "learning_rate": 5.093366728718542e-06, + "loss": 0.0097, + "step": 41123 + }, + { + "epoch": 5.39, + "learning_rate": 5.092904637275671e-06, + "loss": 0.0093, + "step": 41124 + }, + { + "epoch": 5.39, + "learning_rate": 5.092442559633748e-06, + "loss": 0.0082, + "step": 41125 + }, + { + "epoch": 5.39, + "learning_rate": 5.091980495794069e-06, + "loss": 0.0087, + "step": 41126 + }, + { + "epoch": 5.39, + "learning_rate": 5.091518445757933e-06, + "loss": 0.0059, + "step": 41127 + }, + { + "epoch": 5.39, + "learning_rate": 5.091056409526636e-06, + "loss": 0.0079, + "step": 41128 + }, + { + "epoch": 5.39, + "learning_rate": 5.0905943871014865e-06, + "loss": 0.0087, + "step": 41129 + }, + { + "epoch": 5.39, + "learning_rate": 5.090132378483778e-06, + "loss": 0.0083, + "step": 41130 + }, + { + "epoch": 5.39, + "learning_rate": 5.089670383674806e-06, + "loss": 0.0068, + "step": 41131 + }, + { + "epoch": 5.39, + "learning_rate": 5.089208402675876e-06, + "loss": 0.0091, + "step": 41132 + }, + { + "epoch": 5.39, + "learning_rate": 5.088746435488288e-06, + "loss": 0.0086, + "step": 41133 + }, + { + "epoch": 5.39, + "learning_rate": 5.088284482113342e-06, + "loss": 0.0086, + "step": 41134 + }, + { + "epoch": 5.39, + "learning_rate": 5.087822542552332e-06, + "loss": 0.0088, + "step": 41135 + }, + { + "epoch": 5.39, + "learning_rate": 5.087360616806556e-06, + "loss": 0.0081, + "step": 41136 + }, + { + "epoch": 5.39, + "learning_rate": 5.086898704877321e-06, + "loss": 0.0092, + "step": 41137 + }, + { + "epoch": 5.39, + "learning_rate": 5.086436806765922e-06, + "loss": 0.0108, + "step": 41138 + }, + { + "epoch": 5.39, + "learning_rate": 5.085974922473654e-06, + "loss": 0.0076, + "step": 41139 + }, + { + "epoch": 5.39, + "learning_rate": 5.085513052001822e-06, + "loss": 0.0074, + "step": 41140 + }, + { + "epoch": 5.39, + "learning_rate": 5.085051195351724e-06, + "loss": 0.009, + "step": 41141 + }, + { + "epoch": 5.39, + "learning_rate": 5.084589352524654e-06, + "loss": 0.0082, + "step": 41142 + }, + { + "epoch": 5.39, + "learning_rate": 5.084127523521918e-06, + "loss": 0.0074, + "step": 41143 + }, + { + "epoch": 5.39, + "learning_rate": 5.083665708344809e-06, + "loss": 0.0084, + "step": 41144 + }, + { + "epoch": 5.39, + "learning_rate": 5.083203906994631e-06, + "loss": 0.0073, + "step": 41145 + }, + { + "epoch": 5.39, + "learning_rate": 5.082742119472681e-06, + "loss": 0.0066, + "step": 41146 + }, + { + "epoch": 5.39, + "learning_rate": 5.082280345780253e-06, + "loss": 0.0086, + "step": 41147 + }, + { + "epoch": 5.39, + "learning_rate": 5.081818585918652e-06, + "loss": 0.006, + "step": 41148 + }, + { + "epoch": 5.39, + "learning_rate": 5.081356839889175e-06, + "loss": 0.0094, + "step": 41149 + }, + { + "epoch": 5.39, + "learning_rate": 5.080895107693122e-06, + "loss": 0.0059, + "step": 41150 + }, + { + "epoch": 5.39, + "learning_rate": 5.080433389331784e-06, + "loss": 0.007, + "step": 41151 + }, + { + "epoch": 5.39, + "learning_rate": 5.079971684806468e-06, + "loss": 0.007, + "step": 41152 + }, + { + "epoch": 5.39, + "learning_rate": 5.079509994118468e-06, + "loss": 0.0067, + "step": 41153 + }, + { + "epoch": 5.39, + "learning_rate": 5.079048317269086e-06, + "loss": 0.0086, + "step": 41154 + }, + { + "epoch": 5.39, + "learning_rate": 5.0785866542596165e-06, + "loss": 0.0091, + "step": 41155 + }, + { + "epoch": 5.39, + "learning_rate": 5.078125005091362e-06, + "loss": 0.0079, + "step": 41156 + }, + { + "epoch": 5.39, + "learning_rate": 5.07766336976562e-06, + "loss": 0.0072, + "step": 41157 + }, + { + "epoch": 5.39, + "learning_rate": 5.077201748283688e-06, + "loss": 0.0071, + "step": 41158 + }, + { + "epoch": 5.39, + "learning_rate": 5.076740140646858e-06, + "loss": 0.0062, + "step": 41159 + }, + { + "epoch": 5.39, + "learning_rate": 5.07627854685644e-06, + "loss": 0.0062, + "step": 41160 + }, + { + "epoch": 5.39, + "learning_rate": 5.075816966913727e-06, + "loss": 0.0066, + "step": 41161 + }, + { + "epoch": 5.39, + "learning_rate": 5.07535540082001e-06, + "loss": 0.009, + "step": 41162 + }, + { + "epoch": 5.39, + "learning_rate": 5.0748938485766e-06, + "loss": 0.0087, + "step": 41163 + }, + { + "epoch": 5.39, + "learning_rate": 5.074432310184783e-06, + "loss": 0.0057, + "step": 41164 + }, + { + "epoch": 5.39, + "learning_rate": 5.073970785645867e-06, + "loss": 0.0066, + "step": 41165 + }, + { + "epoch": 5.39, + "learning_rate": 5.073509274961148e-06, + "loss": 0.0081, + "step": 41166 + }, + { + "epoch": 5.39, + "learning_rate": 5.073047778131917e-06, + "loss": 0.0066, + "step": 41167 + }, + { + "epoch": 5.39, + "learning_rate": 5.072586295159479e-06, + "loss": 0.009, + "step": 41168 + }, + { + "epoch": 5.39, + "learning_rate": 5.072124826045131e-06, + "loss": 0.0086, + "step": 41169 + }, + { + "epoch": 5.39, + "learning_rate": 5.071663370790165e-06, + "loss": 0.008, + "step": 41170 + }, + { + "epoch": 5.39, + "learning_rate": 5.071201929395888e-06, + "loss": 0.0067, + "step": 41171 + }, + { + "epoch": 5.39, + "learning_rate": 5.070740501863592e-06, + "loss": 0.0065, + "step": 41172 + }, + { + "epoch": 5.39, + "learning_rate": 5.070279088194576e-06, + "loss": 0.0088, + "step": 41173 + }, + { + "epoch": 5.39, + "learning_rate": 5.069817688390135e-06, + "loss": 0.006, + "step": 41174 + }, + { + "epoch": 5.39, + "learning_rate": 5.069356302451569e-06, + "loss": 0.0053, + "step": 41175 + }, + { + "epoch": 5.39, + "learning_rate": 5.068894930380178e-06, + "loss": 0.0094, + "step": 41176 + }, + { + "epoch": 5.39, + "learning_rate": 5.0684335721772584e-06, + "loss": 0.0078, + "step": 41177 + }, + { + "epoch": 5.39, + "learning_rate": 5.067972227844104e-06, + "loss": 0.0077, + "step": 41178 + }, + { + "epoch": 5.39, + "learning_rate": 5.067510897382017e-06, + "loss": 0.0064, + "step": 41179 + }, + { + "epoch": 5.39, + "learning_rate": 5.067049580792294e-06, + "loss": 0.008, + "step": 41180 + }, + { + "epoch": 5.39, + "learning_rate": 5.066588278076231e-06, + "loss": 0.0077, + "step": 41181 + }, + { + "epoch": 5.39, + "learning_rate": 5.066126989235123e-06, + "loss": 0.0059, + "step": 41182 + }, + { + "epoch": 5.39, + "learning_rate": 5.065665714270272e-06, + "loss": 0.0072, + "step": 41183 + }, + { + "epoch": 5.39, + "learning_rate": 5.0652044531829745e-06, + "loss": 0.0079, + "step": 41184 + }, + { + "epoch": 5.39, + "learning_rate": 5.064743205974523e-06, + "loss": 0.0079, + "step": 41185 + }, + { + "epoch": 5.4, + "learning_rate": 5.064281972646218e-06, + "loss": 0.0074, + "step": 41186 + }, + { + "epoch": 5.4, + "learning_rate": 5.063820753199361e-06, + "loss": 0.0086, + "step": 41187 + }, + { + "epoch": 5.4, + "learning_rate": 5.063359547635245e-06, + "loss": 0.0078, + "step": 41188 + }, + { + "epoch": 5.4, + "learning_rate": 5.062898355955164e-06, + "loss": 0.008, + "step": 41189 + }, + { + "epoch": 5.4, + "learning_rate": 5.062437178160422e-06, + "loss": 0.006, + "step": 41190 + }, + { + "epoch": 5.4, + "learning_rate": 5.061976014252313e-06, + "loss": 0.009, + "step": 41191 + }, + { + "epoch": 5.4, + "learning_rate": 5.061514864232133e-06, + "loss": 0.0055, + "step": 41192 + }, + { + "epoch": 5.4, + "learning_rate": 5.061053728101176e-06, + "loss": 0.0087, + "step": 41193 + }, + { + "epoch": 5.4, + "learning_rate": 5.060592605860746e-06, + "loss": 0.0099, + "step": 41194 + }, + { + "epoch": 5.4, + "learning_rate": 5.060131497512136e-06, + "loss": 0.0076, + "step": 41195 + }, + { + "epoch": 5.4, + "learning_rate": 5.059670403056645e-06, + "loss": 0.0066, + "step": 41196 + }, + { + "epoch": 5.4, + "learning_rate": 5.059209322495558e-06, + "loss": 0.007, + "step": 41197 + }, + { + "epoch": 5.4, + "learning_rate": 5.0587482558301895e-06, + "loss": 0.0075, + "step": 41198 + }, + { + "epoch": 5.4, + "learning_rate": 5.058287203061829e-06, + "loss": 0.0051, + "step": 41199 + }, + { + "epoch": 5.4, + "learning_rate": 5.057826164191772e-06, + "loss": 0.0063, + "step": 41200 + }, + { + "epoch": 5.4, + "learning_rate": 5.057365139221313e-06, + "loss": 0.0068, + "step": 41201 + }, + { + "epoch": 5.4, + "learning_rate": 5.056904128151754e-06, + "loss": 0.0065, + "step": 41202 + }, + { + "epoch": 5.4, + "learning_rate": 5.0564431309843895e-06, + "loss": 0.0073, + "step": 41203 + }, + { + "epoch": 5.4, + "learning_rate": 5.055982147720511e-06, + "loss": 0.0082, + "step": 41204 + }, + { + "epoch": 5.4, + "learning_rate": 5.055521178361424e-06, + "loss": 0.0064, + "step": 41205 + }, + { + "epoch": 5.4, + "learning_rate": 5.05506022290842e-06, + "loss": 0.0083, + "step": 41206 + }, + { + "epoch": 5.4, + "learning_rate": 5.054599281362796e-06, + "loss": 0.0099, + "step": 41207 + }, + { + "epoch": 5.4, + "learning_rate": 5.054138353725844e-06, + "loss": 0.0058, + "step": 41208 + }, + { + "epoch": 5.4, + "learning_rate": 5.053677439998865e-06, + "loss": 0.0071, + "step": 41209 + }, + { + "epoch": 5.4, + "learning_rate": 5.053216540183159e-06, + "loss": 0.0076, + "step": 41210 + }, + { + "epoch": 5.4, + "learning_rate": 5.052755654280017e-06, + "loss": 0.0076, + "step": 41211 + }, + { + "epoch": 5.4, + "learning_rate": 5.052294782290732e-06, + "loss": 0.0074, + "step": 41212 + }, + { + "epoch": 5.4, + "learning_rate": 5.0518339242166095e-06, + "loss": 0.0058, + "step": 41213 + }, + { + "epoch": 5.4, + "learning_rate": 5.051373080058941e-06, + "loss": 0.0066, + "step": 41214 + }, + { + "epoch": 5.4, + "learning_rate": 5.050912249819021e-06, + "loss": 0.0069, + "step": 41215 + }, + { + "epoch": 5.4, + "learning_rate": 5.050451433498144e-06, + "loss": 0.0062, + "step": 41216 + }, + { + "epoch": 5.4, + "learning_rate": 5.049990631097611e-06, + "loss": 0.0053, + "step": 41217 + }, + { + "epoch": 5.4, + "learning_rate": 5.049529842618717e-06, + "loss": 0.0086, + "step": 41218 + }, + { + "epoch": 5.4, + "learning_rate": 5.049069068062753e-06, + "loss": 0.0084, + "step": 41219 + }, + { + "epoch": 5.4, + "learning_rate": 5.048608307431018e-06, + "loss": 0.0076, + "step": 41220 + }, + { + "epoch": 5.4, + "learning_rate": 5.048147560724812e-06, + "loss": 0.0072, + "step": 41221 + }, + { + "epoch": 5.4, + "learning_rate": 5.047686827945428e-06, + "loss": 0.0045, + "step": 41222 + }, + { + "epoch": 5.4, + "learning_rate": 5.047226109094161e-06, + "loss": 0.0089, + "step": 41223 + }, + { + "epoch": 5.4, + "learning_rate": 5.0467654041723026e-06, + "loss": 0.007, + "step": 41224 + }, + { + "epoch": 5.4, + "learning_rate": 5.046304713181156e-06, + "loss": 0.0077, + "step": 41225 + }, + { + "epoch": 5.4, + "learning_rate": 5.045844036122014e-06, + "loss": 0.006, + "step": 41226 + }, + { + "epoch": 5.4, + "learning_rate": 5.045383372996169e-06, + "loss": 0.0062, + "step": 41227 + }, + { + "epoch": 5.4, + "learning_rate": 5.044922723804921e-06, + "loss": 0.0053, + "step": 41228 + }, + { + "epoch": 5.4, + "learning_rate": 5.044462088549566e-06, + "loss": 0.0081, + "step": 41229 + }, + { + "epoch": 5.4, + "learning_rate": 5.044001467231394e-06, + "loss": 0.0056, + "step": 41230 + }, + { + "epoch": 5.4, + "learning_rate": 5.043540859851706e-06, + "loss": 0.0081, + "step": 41231 + }, + { + "epoch": 5.4, + "learning_rate": 5.043080266411793e-06, + "loss": 0.0068, + "step": 41232 + }, + { + "epoch": 5.4, + "learning_rate": 5.042619686912956e-06, + "loss": 0.006, + "step": 41233 + }, + { + "epoch": 5.4, + "learning_rate": 5.0421591213564865e-06, + "loss": 0.0077, + "step": 41234 + }, + { + "epoch": 5.4, + "learning_rate": 5.0416985697436786e-06, + "loss": 0.0073, + "step": 41235 + }, + { + "epoch": 5.4, + "learning_rate": 5.041238032075832e-06, + "loss": 0.0065, + "step": 41236 + }, + { + "epoch": 5.4, + "learning_rate": 5.04077750835424e-06, + "loss": 0.0087, + "step": 41237 + }, + { + "epoch": 5.4, + "learning_rate": 5.040316998580196e-06, + "loss": 0.0076, + "step": 41238 + }, + { + "epoch": 5.4, + "learning_rate": 5.0398565027549924e-06, + "loss": 0.0089, + "step": 41239 + }, + { + "epoch": 5.4, + "learning_rate": 5.039396020879933e-06, + "loss": 0.0075, + "step": 41240 + }, + { + "epoch": 5.4, + "learning_rate": 5.038935552956303e-06, + "loss": 0.0081, + "step": 41241 + }, + { + "epoch": 5.4, + "learning_rate": 5.038475098985408e-06, + "loss": 0.008, + "step": 41242 + }, + { + "epoch": 5.4, + "learning_rate": 5.038014658968533e-06, + "loss": 0.0084, + "step": 41243 + }, + { + "epoch": 5.4, + "learning_rate": 5.037554232906982e-06, + "loss": 0.0075, + "step": 41244 + }, + { + "epoch": 5.4, + "learning_rate": 5.037093820802045e-06, + "loss": 0.0055, + "step": 41245 + }, + { + "epoch": 5.4, + "learning_rate": 5.036633422655017e-06, + "loss": 0.0067, + "step": 41246 + }, + { + "epoch": 5.4, + "learning_rate": 5.0361730384671895e-06, + "loss": 0.0088, + "step": 41247 + }, + { + "epoch": 5.4, + "learning_rate": 5.035712668239865e-06, + "loss": 0.0078, + "step": 41248 + }, + { + "epoch": 5.4, + "learning_rate": 5.035252311974335e-06, + "loss": 0.0055, + "step": 41249 + }, + { + "epoch": 5.4, + "learning_rate": 5.034791969671888e-06, + "loss": 0.0061, + "step": 41250 + }, + { + "epoch": 5.4, + "learning_rate": 5.034331641333829e-06, + "loss": 0.0082, + "step": 41251 + }, + { + "epoch": 5.4, + "learning_rate": 5.033871326961444e-06, + "loss": 0.0061, + "step": 41252 + }, + { + "epoch": 5.4, + "learning_rate": 5.033411026556035e-06, + "loss": 0.0072, + "step": 41253 + }, + { + "epoch": 5.4, + "learning_rate": 5.032950740118889e-06, + "loss": 0.008, + "step": 41254 + }, + { + "epoch": 5.4, + "learning_rate": 5.032490467651309e-06, + "loss": 0.0112, + "step": 41255 + }, + { + "epoch": 5.4, + "learning_rate": 5.032030209154584e-06, + "loss": 0.0061, + "step": 41256 + }, + { + "epoch": 5.4, + "learning_rate": 5.0315699646300105e-06, + "loss": 0.0085, + "step": 41257 + }, + { + "epoch": 5.4, + "learning_rate": 5.031109734078878e-06, + "loss": 0.0103, + "step": 41258 + }, + { + "epoch": 5.4, + "learning_rate": 5.0306495175024884e-06, + "loss": 0.0048, + "step": 41259 + }, + { + "epoch": 5.4, + "learning_rate": 5.030189314902132e-06, + "loss": 0.0077, + "step": 41260 + }, + { + "epoch": 5.4, + "learning_rate": 5.029729126279104e-06, + "loss": 0.0083, + "step": 41261 + }, + { + "epoch": 5.41, + "learning_rate": 5.029268951634689e-06, + "loss": 0.0068, + "step": 41262 + }, + { + "epoch": 5.41, + "learning_rate": 5.028808790970201e-06, + "loss": 0.0068, + "step": 41263 + }, + { + "epoch": 5.41, + "learning_rate": 5.028348644286921e-06, + "loss": 0.0098, + "step": 41264 + }, + { + "epoch": 5.41, + "learning_rate": 5.027888511586148e-06, + "loss": 0.006, + "step": 41265 + }, + { + "epoch": 5.41, + "learning_rate": 5.027428392869167e-06, + "loss": 0.0048, + "step": 41266 + }, + { + "epoch": 5.41, + "learning_rate": 5.026968288137284e-06, + "loss": 0.0064, + "step": 41267 + }, + { + "epoch": 5.41, + "learning_rate": 5.026508197391788e-06, + "loss": 0.0064, + "step": 41268 + }, + { + "epoch": 5.41, + "learning_rate": 5.026048120633969e-06, + "loss": 0.0065, + "step": 41269 + }, + { + "epoch": 5.41, + "learning_rate": 5.02558805786513e-06, + "loss": 0.0052, + "step": 41270 + }, + { + "epoch": 5.41, + "learning_rate": 5.025128009086557e-06, + "loss": 0.0063, + "step": 41271 + }, + { + "epoch": 5.41, + "learning_rate": 5.0246679742995486e-06, + "loss": 0.0081, + "step": 41272 + }, + { + "epoch": 5.41, + "learning_rate": 5.024207953505391e-06, + "loss": 0.0057, + "step": 41273 + }, + { + "epoch": 5.41, + "learning_rate": 5.023747946705385e-06, + "loss": 0.0076, + "step": 41274 + }, + { + "epoch": 5.41, + "learning_rate": 5.023287953900826e-06, + "loss": 0.0079, + "step": 41275 + }, + { + "epoch": 5.41, + "learning_rate": 5.022827975093005e-06, + "loss": 0.0091, + "step": 41276 + }, + { + "epoch": 5.41, + "learning_rate": 5.022368010283212e-06, + "loss": 0.0068, + "step": 41277 + }, + { + "epoch": 5.41, + "learning_rate": 5.021908059472747e-06, + "loss": 0.0062, + "step": 41278 + }, + { + "epoch": 5.41, + "learning_rate": 5.0214481226629e-06, + "loss": 0.0079, + "step": 41279 + }, + { + "epoch": 5.41, + "learning_rate": 5.020988199854966e-06, + "loss": 0.0059, + "step": 41280 + }, + { + "epoch": 5.41, + "learning_rate": 5.020528291050233e-06, + "loss": 0.0091, + "step": 41281 + }, + { + "epoch": 5.41, + "learning_rate": 5.020068396250004e-06, + "loss": 0.0065, + "step": 41282 + }, + { + "epoch": 5.41, + "learning_rate": 5.0196085154555655e-06, + "loss": 0.0077, + "step": 41283 + }, + { + "epoch": 5.41, + "learning_rate": 5.019148648668211e-06, + "loss": 0.0093, + "step": 41284 + }, + { + "epoch": 5.41, + "learning_rate": 5.018688795889234e-06, + "loss": 0.0051, + "step": 41285 + }, + { + "epoch": 5.41, + "learning_rate": 5.018228957119935e-06, + "loss": 0.0055, + "step": 41286 + }, + { + "epoch": 5.41, + "learning_rate": 5.017769132361601e-06, + "loss": 0.0063, + "step": 41287 + }, + { + "epoch": 5.41, + "learning_rate": 5.017309321615524e-06, + "loss": 0.0071, + "step": 41288 + }, + { + "epoch": 5.41, + "learning_rate": 5.0168495248829984e-06, + "loss": 0.0083, + "step": 41289 + }, + { + "epoch": 5.41, + "learning_rate": 5.01638974216532e-06, + "loss": 0.0044, + "step": 41290 + }, + { + "epoch": 5.41, + "learning_rate": 5.015929973463781e-06, + "loss": 0.0088, + "step": 41291 + }, + { + "epoch": 5.41, + "learning_rate": 5.01547021877967e-06, + "loss": 0.0083, + "step": 41292 + }, + { + "epoch": 5.41, + "learning_rate": 5.015010478114286e-06, + "loss": 0.0073, + "step": 41293 + }, + { + "epoch": 5.41, + "learning_rate": 5.014550751468921e-06, + "loss": 0.0075, + "step": 41294 + }, + { + "epoch": 5.41, + "learning_rate": 5.014091038844862e-06, + "loss": 0.0047, + "step": 41295 + }, + { + "epoch": 5.41, + "learning_rate": 5.01363134024341e-06, + "loss": 0.0055, + "step": 41296 + }, + { + "epoch": 5.41, + "learning_rate": 5.013171655665852e-06, + "loss": 0.0085, + "step": 41297 + }, + { + "epoch": 5.41, + "learning_rate": 5.0127119851134855e-06, + "loss": 0.008, + "step": 41298 + }, + { + "epoch": 5.41, + "learning_rate": 5.012252328587601e-06, + "loss": 0.0071, + "step": 41299 + }, + { + "epoch": 5.41, + "learning_rate": 5.011792686089487e-06, + "loss": 0.01, + "step": 41300 + }, + { + "epoch": 5.41, + "learning_rate": 5.011333057620446e-06, + "loss": 0.0077, + "step": 41301 + }, + { + "epoch": 5.41, + "learning_rate": 5.010873443181764e-06, + "loss": 0.0086, + "step": 41302 + }, + { + "epoch": 5.41, + "learning_rate": 5.010413842774735e-06, + "loss": 0.0051, + "step": 41303 + }, + { + "epoch": 5.41, + "learning_rate": 5.009954256400648e-06, + "loss": 0.0066, + "step": 41304 + }, + { + "epoch": 5.41, + "learning_rate": 5.0094946840608016e-06, + "loss": 0.0077, + "step": 41305 + }, + { + "epoch": 5.41, + "learning_rate": 5.009035125756483e-06, + "loss": 0.0077, + "step": 41306 + }, + { + "epoch": 5.41, + "learning_rate": 5.0085755814889905e-06, + "loss": 0.0091, + "step": 41307 + }, + { + "epoch": 5.41, + "learning_rate": 5.008116051259608e-06, + "loss": 0.0058, + "step": 41308 + }, + { + "epoch": 5.41, + "learning_rate": 5.0076565350696396e-06, + "loss": 0.0068, + "step": 41309 + }, + { + "epoch": 5.41, + "learning_rate": 5.00719703292037e-06, + "loss": 0.0069, + "step": 41310 + }, + { + "epoch": 5.41, + "learning_rate": 5.006737544813094e-06, + "loss": 0.0053, + "step": 41311 + }, + { + "epoch": 5.41, + "learning_rate": 5.006278070749097e-06, + "loss": 0.0055, + "step": 41312 + }, + { + "epoch": 5.41, + "learning_rate": 5.005818610729682e-06, + "loss": 0.0079, + "step": 41313 + }, + { + "epoch": 5.41, + "learning_rate": 5.005359164756136e-06, + "loss": 0.0063, + "step": 41314 + }, + { + "epoch": 5.41, + "learning_rate": 5.004899732829749e-06, + "loss": 0.0068, + "step": 41315 + }, + { + "epoch": 5.41, + "learning_rate": 5.004440314951817e-06, + "loss": 0.0062, + "step": 41316 + }, + { + "epoch": 5.41, + "learning_rate": 5.003980911123629e-06, + "loss": 0.0063, + "step": 41317 + }, + { + "epoch": 5.41, + "learning_rate": 5.003521521346481e-06, + "loss": 0.0077, + "step": 41318 + }, + { + "epoch": 5.41, + "learning_rate": 5.0030621456216585e-06, + "loss": 0.0075, + "step": 41319 + }, + { + "epoch": 5.41, + "learning_rate": 5.002602783950463e-06, + "loss": 0.009, + "step": 41320 + }, + { + "epoch": 5.41, + "learning_rate": 5.0021434363341794e-06, + "loss": 0.0079, + "step": 41321 + }, + { + "epoch": 5.41, + "learning_rate": 5.001684102774103e-06, + "loss": 0.0086, + "step": 41322 + }, + { + "epoch": 5.41, + "learning_rate": 5.0012247832715176e-06, + "loss": 0.006, + "step": 41323 + }, + { + "epoch": 5.41, + "learning_rate": 5.000765477827727e-06, + "loss": 0.0105, + "step": 41324 + }, + { + "epoch": 5.41, + "learning_rate": 5.000306186444018e-06, + "loss": 0.0081, + "step": 41325 + }, + { + "epoch": 5.41, + "learning_rate": 4.9998469091216794e-06, + "loss": 0.0071, + "step": 41326 + }, + { + "epoch": 5.41, + "learning_rate": 4.999387645862002e-06, + "loss": 0.0071, + "step": 41327 + }, + { + "epoch": 5.41, + "learning_rate": 4.99892839666628e-06, + "loss": 0.0071, + "step": 41328 + }, + { + "epoch": 5.41, + "learning_rate": 4.998469161535811e-06, + "loss": 0.0067, + "step": 41329 + }, + { + "epoch": 5.41, + "learning_rate": 4.99800994047188e-06, + "loss": 0.0067, + "step": 41330 + }, + { + "epoch": 5.41, + "learning_rate": 4.997550733475776e-06, + "loss": 0.0069, + "step": 41331 + }, + { + "epoch": 5.41, + "learning_rate": 4.997091540548799e-06, + "loss": 0.0078, + "step": 41332 + }, + { + "epoch": 5.41, + "learning_rate": 4.996632361692234e-06, + "loss": 0.0085, + "step": 41333 + }, + { + "epoch": 5.41, + "learning_rate": 4.996173196907372e-06, + "loss": 0.0063, + "step": 41334 + }, + { + "epoch": 5.41, + "learning_rate": 4.995714046195508e-06, + "loss": 0.008, + "step": 41335 + }, + { + "epoch": 5.41, + "learning_rate": 4.995254909557933e-06, + "loss": 0.0078, + "step": 41336 + }, + { + "epoch": 5.41, + "learning_rate": 4.994795786995938e-06, + "loss": 0.0076, + "step": 41337 + }, + { + "epoch": 5.41, + "learning_rate": 4.9943366785108075e-06, + "loss": 0.0083, + "step": 41338 + }, + { + "epoch": 5.42, + "learning_rate": 4.99387758410384e-06, + "loss": 0.0079, + "step": 41339 + }, + { + "epoch": 5.42, + "learning_rate": 4.993418503776328e-06, + "loss": 0.0063, + "step": 41340 + }, + { + "epoch": 5.42, + "learning_rate": 4.992959437529561e-06, + "loss": 0.007, + "step": 41341 + }, + { + "epoch": 5.42, + "learning_rate": 4.992500385364825e-06, + "loss": 0.0061, + "step": 41342 + }, + { + "epoch": 5.42, + "learning_rate": 4.992041347283417e-06, + "loss": 0.0071, + "step": 41343 + }, + { + "epoch": 5.42, + "learning_rate": 4.991582323286628e-06, + "loss": 0.0068, + "step": 41344 + }, + { + "epoch": 5.42, + "learning_rate": 4.991123313375746e-06, + "loss": 0.0093, + "step": 41345 + }, + { + "epoch": 5.42, + "learning_rate": 4.99066431755206e-06, + "loss": 0.0061, + "step": 41346 + }, + { + "epoch": 5.42, + "learning_rate": 4.990205335816866e-06, + "loss": 0.008, + "step": 41347 + }, + { + "epoch": 5.42, + "learning_rate": 4.989746368171454e-06, + "loss": 0.0086, + "step": 41348 + }, + { + "epoch": 5.42, + "learning_rate": 4.98928741461711e-06, + "loss": 0.006, + "step": 41349 + }, + { + "epoch": 5.42, + "learning_rate": 4.988828475155128e-06, + "loss": 0.0069, + "step": 41350 + }, + { + "epoch": 5.42, + "learning_rate": 4.988369549786802e-06, + "loss": 0.0067, + "step": 41351 + }, + { + "epoch": 5.42, + "learning_rate": 4.987910638513421e-06, + "loss": 0.0065, + "step": 41352 + }, + { + "epoch": 5.42, + "learning_rate": 4.987451741336274e-06, + "loss": 0.0097, + "step": 41353 + }, + { + "epoch": 5.42, + "learning_rate": 4.9869928582566484e-06, + "loss": 0.0063, + "step": 41354 + }, + { + "epoch": 5.42, + "learning_rate": 4.986533989275841e-06, + "loss": 0.0079, + "step": 41355 + }, + { + "epoch": 5.42, + "learning_rate": 4.986075134395142e-06, + "loss": 0.007, + "step": 41356 + }, + { + "epoch": 5.42, + "learning_rate": 4.9856162936158335e-06, + "loss": 0.0096, + "step": 41357 + }, + { + "epoch": 5.42, + "learning_rate": 4.985157466939218e-06, + "loss": 0.0088, + "step": 41358 + }, + { + "epoch": 5.42, + "learning_rate": 4.984698654366578e-06, + "loss": 0.0073, + "step": 41359 + }, + { + "epoch": 5.42, + "learning_rate": 4.9842398558992075e-06, + "loss": 0.0048, + "step": 41360 + }, + { + "epoch": 5.42, + "learning_rate": 4.983781071538392e-06, + "loss": 0.0066, + "step": 41361 + }, + { + "epoch": 5.42, + "learning_rate": 4.9833223012854235e-06, + "loss": 0.0079, + "step": 41362 + }, + { + "epoch": 5.42, + "learning_rate": 4.982863545141599e-06, + "loss": 0.0069, + "step": 41363 + }, + { + "epoch": 5.42, + "learning_rate": 4.982404803108203e-06, + "loss": 0.0068, + "step": 41364 + }, + { + "epoch": 5.42, + "learning_rate": 4.981946075186523e-06, + "loss": 0.0066, + "step": 41365 + }, + { + "epoch": 5.42, + "learning_rate": 4.9814873613778545e-06, + "loss": 0.0072, + "step": 41366 + }, + { + "epoch": 5.42, + "learning_rate": 4.981028661683487e-06, + "loss": 0.0069, + "step": 41367 + }, + { + "epoch": 5.42, + "learning_rate": 4.980569976104709e-06, + "loss": 0.0135, + "step": 41368 + }, + { + "epoch": 5.42, + "learning_rate": 4.980111304642806e-06, + "loss": 0.0094, + "step": 41369 + }, + { + "epoch": 5.42, + "learning_rate": 4.979652647299078e-06, + "loss": 0.0056, + "step": 41370 + }, + { + "epoch": 5.42, + "learning_rate": 4.979194004074809e-06, + "loss": 0.0081, + "step": 41371 + }, + { + "epoch": 5.42, + "learning_rate": 4.978735374971285e-06, + "loss": 0.0072, + "step": 41372 + }, + { + "epoch": 5.42, + "learning_rate": 4.978276759989801e-06, + "loss": 0.008, + "step": 41373 + }, + { + "epoch": 5.42, + "learning_rate": 4.977818159131649e-06, + "loss": 0.0079, + "step": 41374 + }, + { + "epoch": 5.42, + "learning_rate": 4.977359572398117e-06, + "loss": 0.0097, + "step": 41375 + }, + { + "epoch": 5.42, + "learning_rate": 4.976900999790494e-06, + "loss": 0.0074, + "step": 41376 + }, + { + "epoch": 5.42, + "learning_rate": 4.9764424413100645e-06, + "loss": 0.0064, + "step": 41377 + }, + { + "epoch": 5.42, + "learning_rate": 4.9759838969581276e-06, + "loss": 0.0084, + "step": 41378 + }, + { + "epoch": 5.42, + "learning_rate": 4.975525366735968e-06, + "loss": 0.0071, + "step": 41379 + }, + { + "epoch": 5.42, + "learning_rate": 4.975066850644873e-06, + "loss": 0.0061, + "step": 41380 + }, + { + "epoch": 5.42, + "learning_rate": 4.974608348686137e-06, + "loss": 0.0067, + "step": 41381 + }, + { + "epoch": 5.42, + "learning_rate": 4.974149860861049e-06, + "loss": 0.007, + "step": 41382 + }, + { + "epoch": 5.42, + "learning_rate": 4.973691387170893e-06, + "loss": 0.0084, + "step": 41383 + }, + { + "epoch": 5.42, + "learning_rate": 4.973232927616961e-06, + "loss": 0.0067, + "step": 41384 + }, + { + "epoch": 5.42, + "learning_rate": 4.972774482200548e-06, + "loss": 0.0076, + "step": 41385 + }, + { + "epoch": 5.42, + "learning_rate": 4.972316050922939e-06, + "loss": 0.0096, + "step": 41386 + }, + { + "epoch": 5.42, + "learning_rate": 4.971857633785424e-06, + "loss": 0.0085, + "step": 41387 + }, + { + "epoch": 5.42, + "learning_rate": 4.971399230789288e-06, + "loss": 0.0087, + "step": 41388 + }, + { + "epoch": 5.42, + "learning_rate": 4.970940841935828e-06, + "loss": 0.0063, + "step": 41389 + }, + { + "epoch": 5.42, + "learning_rate": 4.970482467226327e-06, + "loss": 0.0055, + "step": 41390 + }, + { + "epoch": 5.42, + "learning_rate": 4.970024106662074e-06, + "loss": 0.0076, + "step": 41391 + }, + { + "epoch": 5.42, + "learning_rate": 4.969565760244364e-06, + "loss": 0.006, + "step": 41392 + }, + { + "epoch": 5.42, + "learning_rate": 4.969107427974479e-06, + "loss": 0.0083, + "step": 41393 + }, + { + "epoch": 5.42, + "learning_rate": 4.968649109853714e-06, + "loss": 0.0097, + "step": 41394 + }, + { + "epoch": 5.42, + "learning_rate": 4.968190805883356e-06, + "loss": 0.0065, + "step": 41395 + }, + { + "epoch": 5.42, + "learning_rate": 4.967732516064689e-06, + "loss": 0.0063, + "step": 41396 + }, + { + "epoch": 5.42, + "learning_rate": 4.967274240399011e-06, + "loss": 0.0081, + "step": 41397 + }, + { + "epoch": 5.42, + "learning_rate": 4.966815978887607e-06, + "loss": 0.0051, + "step": 41398 + }, + { + "epoch": 5.42, + "learning_rate": 4.96635773153176e-06, + "loss": 0.0067, + "step": 41399 + }, + { + "epoch": 5.42, + "learning_rate": 4.965899498332768e-06, + "loss": 0.007, + "step": 41400 + }, + { + "epoch": 5.42, + "learning_rate": 4.965441279291916e-06, + "loss": 0.0084, + "step": 41401 + }, + { + "epoch": 5.42, + "learning_rate": 4.964983074410491e-06, + "loss": 0.0079, + "step": 41402 + }, + { + "epoch": 5.42, + "learning_rate": 4.96452488368978e-06, + "loss": 0.0078, + "step": 41403 + }, + { + "epoch": 5.42, + "learning_rate": 4.964066707131074e-06, + "loss": 0.0078, + "step": 41404 + }, + { + "epoch": 5.42, + "learning_rate": 4.9636085447356675e-06, + "loss": 0.0068, + "step": 41405 + }, + { + "epoch": 5.42, + "learning_rate": 4.963150396504842e-06, + "loss": 0.0107, + "step": 41406 + }, + { + "epoch": 5.42, + "learning_rate": 4.962692262439885e-06, + "loss": 0.0068, + "step": 41407 + }, + { + "epoch": 5.42, + "learning_rate": 4.962234142542092e-06, + "loss": 0.0055, + "step": 41408 + }, + { + "epoch": 5.42, + "learning_rate": 4.961776036812746e-06, + "loss": 0.0066, + "step": 41409 + }, + { + "epoch": 5.42, + "learning_rate": 4.961317945253138e-06, + "loss": 0.0072, + "step": 41410 + }, + { + "epoch": 5.42, + "learning_rate": 4.960859867864549e-06, + "loss": 0.0091, + "step": 41411 + }, + { + "epoch": 5.42, + "learning_rate": 4.960401804648279e-06, + "loss": 0.0075, + "step": 41412 + }, + { + "epoch": 5.42, + "learning_rate": 4.959943755605609e-06, + "loss": 0.0095, + "step": 41413 + }, + { + "epoch": 5.42, + "learning_rate": 4.959485720737826e-06, + "loss": 0.0078, + "step": 41414 + }, + { + "epoch": 5.43, + "learning_rate": 4.959027700046221e-06, + "loss": 0.008, + "step": 41415 + }, + { + "epoch": 5.43, + "learning_rate": 4.958569693532087e-06, + "loss": 0.0085, + "step": 41416 + }, + { + "epoch": 5.43, + "learning_rate": 4.9581117011967065e-06, + "loss": 0.006, + "step": 41417 + }, + { + "epoch": 5.43, + "learning_rate": 4.9576537230413665e-06, + "loss": 0.0087, + "step": 41418 + }, + { + "epoch": 5.43, + "learning_rate": 4.957195759067356e-06, + "loss": 0.0097, + "step": 41419 + }, + { + "epoch": 5.43, + "learning_rate": 4.956737809275966e-06, + "loss": 0.0041, + "step": 41420 + }, + { + "epoch": 5.43, + "learning_rate": 4.956279873668482e-06, + "loss": 0.0076, + "step": 41421 + }, + { + "epoch": 5.43, + "learning_rate": 4.955821952246189e-06, + "loss": 0.0086, + "step": 41422 + }, + { + "epoch": 5.43, + "learning_rate": 4.955364045010382e-06, + "loss": 0.0063, + "step": 41423 + }, + { + "epoch": 5.43, + "learning_rate": 4.954906151962345e-06, + "loss": 0.0082, + "step": 41424 + }, + { + "epoch": 5.43, + "learning_rate": 4.9544482731033664e-06, + "loss": 0.0075, + "step": 41425 + }, + { + "epoch": 5.43, + "learning_rate": 4.953990408434728e-06, + "loss": 0.0075, + "step": 41426 + }, + { + "epoch": 5.43, + "learning_rate": 4.953532557957724e-06, + "loss": 0.0058, + "step": 41427 + }, + { + "epoch": 5.43, + "learning_rate": 4.953074721673644e-06, + "loss": 0.0067, + "step": 41428 + }, + { + "epoch": 5.43, + "learning_rate": 4.952616899583773e-06, + "loss": 0.0118, + "step": 41429 + }, + { + "epoch": 5.43, + "learning_rate": 4.952159091689395e-06, + "loss": 0.0079, + "step": 41430 + }, + { + "epoch": 5.43, + "learning_rate": 4.951701297991805e-06, + "loss": 0.0097, + "step": 41431 + }, + { + "epoch": 5.43, + "learning_rate": 4.951243518492284e-06, + "loss": 0.0096, + "step": 41432 + }, + { + "epoch": 5.43, + "learning_rate": 4.9507857531921236e-06, + "loss": 0.0045, + "step": 41433 + }, + { + "epoch": 5.43, + "learning_rate": 4.950328002092605e-06, + "loss": 0.0075, + "step": 41434 + }, + { + "epoch": 5.43, + "learning_rate": 4.949870265195023e-06, + "loss": 0.0044, + "step": 41435 + }, + { + "epoch": 5.43, + "learning_rate": 4.949412542500664e-06, + "loss": 0.008, + "step": 41436 + }, + { + "epoch": 5.43, + "learning_rate": 4.948954834010807e-06, + "loss": 0.0054, + "step": 41437 + }, + { + "epoch": 5.43, + "learning_rate": 4.948497139726747e-06, + "loss": 0.0078, + "step": 41438 + }, + { + "epoch": 5.43, + "learning_rate": 4.9480394596497725e-06, + "loss": 0.0075, + "step": 41439 + }, + { + "epoch": 5.43, + "learning_rate": 4.9475817937811696e-06, + "loss": 0.0063, + "step": 41440 + }, + { + "epoch": 5.43, + "learning_rate": 4.947124142122219e-06, + "loss": 0.009, + "step": 41441 + }, + { + "epoch": 5.43, + "learning_rate": 4.946666504674217e-06, + "loss": 0.0094, + "step": 41442 + }, + { + "epoch": 5.43, + "learning_rate": 4.946208881438446e-06, + "loss": 0.0076, + "step": 41443 + }, + { + "epoch": 5.43, + "learning_rate": 4.945751272416193e-06, + "loss": 0.007, + "step": 41444 + }, + { + "epoch": 5.43, + "learning_rate": 4.945293677608742e-06, + "loss": 0.007, + "step": 41445 + }, + { + "epoch": 5.43, + "learning_rate": 4.944836097017387e-06, + "loss": 0.0069, + "step": 41446 + }, + { + "epoch": 5.43, + "learning_rate": 4.9443785306434125e-06, + "loss": 0.0046, + "step": 41447 + }, + { + "epoch": 5.43, + "learning_rate": 4.943920978488098e-06, + "loss": 0.0069, + "step": 41448 + }, + { + "epoch": 5.43, + "learning_rate": 4.943463440552738e-06, + "loss": 0.0074, + "step": 41449 + }, + { + "epoch": 5.43, + "learning_rate": 4.943005916838622e-06, + "loss": 0.0066, + "step": 41450 + }, + { + "epoch": 5.43, + "learning_rate": 4.942548407347032e-06, + "loss": 0.0088, + "step": 41451 + }, + { + "epoch": 5.43, + "learning_rate": 4.942090912079255e-06, + "loss": 0.0066, + "step": 41452 + }, + { + "epoch": 5.43, + "learning_rate": 4.941633431036574e-06, + "loss": 0.0059, + "step": 41453 + }, + { + "epoch": 5.43, + "learning_rate": 4.9411759642202835e-06, + "loss": 0.0058, + "step": 41454 + }, + { + "epoch": 5.43, + "learning_rate": 4.940718511631667e-06, + "loss": 0.0087, + "step": 41455 + }, + { + "epoch": 5.43, + "learning_rate": 4.940261073272006e-06, + "loss": 0.0061, + "step": 41456 + }, + { + "epoch": 5.43, + "learning_rate": 4.939803649142595e-06, + "loss": 0.009, + "step": 41457 + }, + { + "epoch": 5.43, + "learning_rate": 4.939346239244716e-06, + "loss": 0.0062, + "step": 41458 + }, + { + "epoch": 5.43, + "learning_rate": 4.938888843579655e-06, + "loss": 0.0053, + "step": 41459 + }, + { + "epoch": 5.43, + "learning_rate": 4.938431462148701e-06, + "loss": 0.0072, + "step": 41460 + }, + { + "epoch": 5.43, + "learning_rate": 4.937974094953135e-06, + "loss": 0.0084, + "step": 41461 + }, + { + "epoch": 5.43, + "learning_rate": 4.9375167419942526e-06, + "loss": 0.0061, + "step": 41462 + }, + { + "epoch": 5.43, + "learning_rate": 4.937059403273334e-06, + "loss": 0.0068, + "step": 41463 + }, + { + "epoch": 5.43, + "learning_rate": 4.936602078791663e-06, + "loss": 0.009, + "step": 41464 + }, + { + "epoch": 5.43, + "learning_rate": 4.9361447685505314e-06, + "loss": 0.0074, + "step": 41465 + }, + { + "epoch": 5.43, + "learning_rate": 4.935687472551224e-06, + "loss": 0.0064, + "step": 41466 + }, + { + "epoch": 5.43, + "learning_rate": 4.935230190795026e-06, + "loss": 0.0078, + "step": 41467 + }, + { + "epoch": 5.43, + "learning_rate": 4.934772923283219e-06, + "loss": 0.0077, + "step": 41468 + }, + { + "epoch": 5.43, + "learning_rate": 4.934315670017098e-06, + "loss": 0.0085, + "step": 41469 + }, + { + "epoch": 5.43, + "learning_rate": 4.93385843099794e-06, + "loss": 0.008, + "step": 41470 + }, + { + "epoch": 5.43, + "learning_rate": 4.933401206227038e-06, + "loss": 0.0081, + "step": 41471 + }, + { + "epoch": 5.43, + "learning_rate": 4.932943995705674e-06, + "loss": 0.004, + "step": 41472 + }, + { + "epoch": 5.43, + "learning_rate": 4.932486799435138e-06, + "loss": 0.0063, + "step": 41473 + }, + { + "epoch": 5.43, + "learning_rate": 4.932029617416713e-06, + "loss": 0.0074, + "step": 41474 + }, + { + "epoch": 5.43, + "learning_rate": 4.931572449651684e-06, + "loss": 0.0072, + "step": 41475 + }, + { + "epoch": 5.43, + "learning_rate": 4.931115296141335e-06, + "loss": 0.0068, + "step": 41476 + }, + { + "epoch": 5.43, + "learning_rate": 4.930658156886957e-06, + "loss": 0.0076, + "step": 41477 + }, + { + "epoch": 5.43, + "learning_rate": 4.930201031889833e-06, + "loss": 0.0074, + "step": 41478 + }, + { + "epoch": 5.43, + "learning_rate": 4.929743921151245e-06, + "loss": 0.0056, + "step": 41479 + }, + { + "epoch": 5.43, + "learning_rate": 4.929286824672487e-06, + "loss": 0.0088, + "step": 41480 + }, + { + "epoch": 5.43, + "learning_rate": 4.9288297424548345e-06, + "loss": 0.0119, + "step": 41481 + }, + { + "epoch": 5.43, + "learning_rate": 4.928372674499584e-06, + "loss": 0.0069, + "step": 41482 + }, + { + "epoch": 5.43, + "learning_rate": 4.927915620808014e-06, + "loss": 0.0061, + "step": 41483 + }, + { + "epoch": 5.43, + "learning_rate": 4.9274585813814065e-06, + "loss": 0.0086, + "step": 41484 + }, + { + "epoch": 5.43, + "learning_rate": 4.9270015562210575e-06, + "loss": 0.009, + "step": 41485 + }, + { + "epoch": 5.43, + "learning_rate": 4.926544545328246e-06, + "loss": 0.0055, + "step": 41486 + }, + { + "epoch": 5.43, + "learning_rate": 4.926087548704252e-06, + "loss": 0.0054, + "step": 41487 + }, + { + "epoch": 5.43, + "learning_rate": 4.925630566350373e-06, + "loss": 0.0064, + "step": 41488 + }, + { + "epoch": 5.43, + "learning_rate": 4.925173598267886e-06, + "loss": 0.0069, + "step": 41489 + }, + { + "epoch": 5.43, + "learning_rate": 4.924716644458078e-06, + "loss": 0.0075, + "step": 41490 + }, + { + "epoch": 5.44, + "learning_rate": 4.924259704922232e-06, + "loss": 0.0074, + "step": 41491 + }, + { + "epoch": 5.44, + "learning_rate": 4.923802779661634e-06, + "loss": 0.0068, + "step": 41492 + }, + { + "epoch": 5.44, + "learning_rate": 4.923345868677574e-06, + "loss": 0.007, + "step": 41493 + }, + { + "epoch": 5.44, + "learning_rate": 4.922888971971333e-06, + "loss": 0.0058, + "step": 41494 + }, + { + "epoch": 5.44, + "learning_rate": 4.9224320895441944e-06, + "loss": 0.0092, + "step": 41495 + }, + { + "epoch": 5.44, + "learning_rate": 4.9219752213974485e-06, + "loss": 0.0087, + "step": 41496 + }, + { + "epoch": 5.44, + "learning_rate": 4.921518367532375e-06, + "loss": 0.0079, + "step": 41497 + }, + { + "epoch": 5.44, + "learning_rate": 4.921061527950263e-06, + "loss": 0.0065, + "step": 41498 + }, + { + "epoch": 5.44, + "learning_rate": 4.920604702652389e-06, + "loss": 0.0066, + "step": 41499 + }, + { + "epoch": 5.44, + "learning_rate": 4.920147891640049e-06, + "loss": 0.0064, + "step": 41500 + }, + { + "epoch": 5.44, + "learning_rate": 4.919691094914523e-06, + "loss": 0.0083, + "step": 41501 + }, + { + "epoch": 5.44, + "learning_rate": 4.919234312477091e-06, + "loss": 0.0057, + "step": 41502 + }, + { + "epoch": 5.44, + "learning_rate": 4.918777544329041e-06, + "loss": 0.009, + "step": 41503 + }, + { + "epoch": 5.44, + "learning_rate": 4.918320790471664e-06, + "loss": 0.0075, + "step": 41504 + }, + { + "epoch": 5.44, + "learning_rate": 4.917864050906237e-06, + "loss": 0.0087, + "step": 41505 + }, + { + "epoch": 5.44, + "learning_rate": 4.917407325634045e-06, + "loss": 0.0058, + "step": 41506 + }, + { + "epoch": 5.44, + "learning_rate": 4.916950614656376e-06, + "loss": 0.0072, + "step": 41507 + }, + { + "epoch": 5.44, + "learning_rate": 4.916493917974514e-06, + "loss": 0.0072, + "step": 41508 + }, + { + "epoch": 5.44, + "learning_rate": 4.916037235589742e-06, + "loss": 0.0083, + "step": 41509 + }, + { + "epoch": 5.44, + "learning_rate": 4.915580567503341e-06, + "loss": 0.0079, + "step": 41510 + }, + { + "epoch": 5.44, + "learning_rate": 4.915123913716603e-06, + "loss": 0.0074, + "step": 41511 + }, + { + "epoch": 5.44, + "learning_rate": 4.9146672742308065e-06, + "loss": 0.0073, + "step": 41512 + }, + { + "epoch": 5.44, + "learning_rate": 4.914210649047239e-06, + "loss": 0.0072, + "step": 41513 + }, + { + "epoch": 5.44, + "learning_rate": 4.913754038167175e-06, + "loss": 0.0063, + "step": 41514 + }, + { + "epoch": 5.44, + "learning_rate": 4.9132974415919155e-06, + "loss": 0.0067, + "step": 41515 + }, + { + "epoch": 5.44, + "learning_rate": 4.912840859322735e-06, + "loss": 0.0081, + "step": 41516 + }, + { + "epoch": 5.44, + "learning_rate": 4.912384291360919e-06, + "loss": 0.0074, + "step": 41517 + }, + { + "epoch": 5.44, + "learning_rate": 4.9119277377077465e-06, + "loss": 0.007, + "step": 41518 + }, + { + "epoch": 5.44, + "learning_rate": 4.9114711983645116e-06, + "loss": 0.0061, + "step": 41519 + }, + { + "epoch": 5.44, + "learning_rate": 4.911014673332493e-06, + "loss": 0.0072, + "step": 41520 + }, + { + "epoch": 5.44, + "learning_rate": 4.9105581626129695e-06, + "loss": 0.0055, + "step": 41521 + }, + { + "epoch": 5.44, + "learning_rate": 4.910101666207234e-06, + "loss": 0.0052, + "step": 41522 + }, + { + "epoch": 5.44, + "learning_rate": 4.909645184116568e-06, + "loss": 0.0081, + "step": 41523 + }, + { + "epoch": 5.44, + "learning_rate": 4.909188716342253e-06, + "loss": 0.0092, + "step": 41524 + }, + { + "epoch": 5.44, + "learning_rate": 4.908732262885569e-06, + "loss": 0.007, + "step": 41525 + }, + { + "epoch": 5.44, + "learning_rate": 4.908275823747806e-06, + "loss": 0.0073, + "step": 41526 + }, + { + "epoch": 5.44, + "learning_rate": 4.907819398930249e-06, + "loss": 0.0102, + "step": 41527 + }, + { + "epoch": 5.44, + "learning_rate": 4.907362988434179e-06, + "loss": 0.0087, + "step": 41528 + }, + { + "epoch": 5.44, + "learning_rate": 4.906906592260875e-06, + "loss": 0.0087, + "step": 41529 + }, + { + "epoch": 5.44, + "learning_rate": 4.90645021041163e-06, + "loss": 0.006, + "step": 41530 + }, + { + "epoch": 5.44, + "learning_rate": 4.905993842887721e-06, + "loss": 0.0071, + "step": 41531 + }, + { + "epoch": 5.44, + "learning_rate": 4.905537489690434e-06, + "loss": 0.0063, + "step": 41532 + }, + { + "epoch": 5.44, + "learning_rate": 4.905081150821047e-06, + "loss": 0.0062, + "step": 41533 + }, + { + "epoch": 5.44, + "learning_rate": 4.904624826280853e-06, + "loss": 0.0072, + "step": 41534 + }, + { + "epoch": 5.44, + "learning_rate": 4.904168516071126e-06, + "loss": 0.0069, + "step": 41535 + }, + { + "epoch": 5.44, + "learning_rate": 4.9037122201931566e-06, + "loss": 0.0072, + "step": 41536 + }, + { + "epoch": 5.44, + "learning_rate": 4.903255938648223e-06, + "loss": 0.0089, + "step": 41537 + }, + { + "epoch": 5.44, + "learning_rate": 4.902799671437613e-06, + "loss": 0.0075, + "step": 41538 + }, + { + "epoch": 5.44, + "learning_rate": 4.902343418562609e-06, + "loss": 0.0072, + "step": 41539 + }, + { + "epoch": 5.44, + "learning_rate": 4.90188718002449e-06, + "loss": 0.0066, + "step": 41540 + }, + { + "epoch": 5.44, + "learning_rate": 4.90143095582454e-06, + "loss": 0.0081, + "step": 41541 + }, + { + "epoch": 5.44, + "learning_rate": 4.9009747459640475e-06, + "loss": 0.0083, + "step": 41542 + }, + { + "epoch": 5.44, + "learning_rate": 4.900518550444292e-06, + "loss": 0.0077, + "step": 41543 + }, + { + "epoch": 5.44, + "learning_rate": 4.900062369266553e-06, + "loss": 0.0078, + "step": 41544 + }, + { + "epoch": 5.44, + "learning_rate": 4.89960620243212e-06, + "loss": 0.009, + "step": 41545 + }, + { + "epoch": 5.44, + "learning_rate": 4.899150049942269e-06, + "loss": 0.0073, + "step": 41546 + }, + { + "epoch": 5.44, + "learning_rate": 4.898693911798292e-06, + "loss": 0.0076, + "step": 41547 + }, + { + "epoch": 5.44, + "learning_rate": 4.898237788001466e-06, + "loss": 0.0079, + "step": 41548 + }, + { + "epoch": 5.44, + "learning_rate": 4.897781678553071e-06, + "loss": 0.0064, + "step": 41549 + }, + { + "epoch": 5.44, + "learning_rate": 4.897325583454397e-06, + "loss": 0.0072, + "step": 41550 + }, + { + "epoch": 5.44, + "learning_rate": 4.896869502706723e-06, + "loss": 0.0059, + "step": 41551 + }, + { + "epoch": 5.44, + "learning_rate": 4.896413436311327e-06, + "loss": 0.0076, + "step": 41552 + }, + { + "epoch": 5.44, + "learning_rate": 4.895957384269501e-06, + "loss": 0.0079, + "step": 41553 + }, + { + "epoch": 5.44, + "learning_rate": 4.895501346582523e-06, + "loss": 0.005, + "step": 41554 + }, + { + "epoch": 5.44, + "learning_rate": 4.895045323251676e-06, + "loss": 0.0099, + "step": 41555 + }, + { + "epoch": 5.44, + "learning_rate": 4.894589314278237e-06, + "loss": 0.0075, + "step": 41556 + }, + { + "epoch": 5.44, + "learning_rate": 4.894133319663496e-06, + "loss": 0.005, + "step": 41557 + }, + { + "epoch": 5.44, + "learning_rate": 4.893677339408735e-06, + "loss": 0.0099, + "step": 41558 + }, + { + "epoch": 5.44, + "learning_rate": 4.893221373515235e-06, + "loss": 0.0078, + "step": 41559 + }, + { + "epoch": 5.44, + "learning_rate": 4.892765421984275e-06, + "loss": 0.0054, + "step": 41560 + }, + { + "epoch": 5.44, + "learning_rate": 4.8923094848171435e-06, + "loss": 0.0061, + "step": 41561 + }, + { + "epoch": 5.44, + "learning_rate": 4.89185356201512e-06, + "loss": 0.0077, + "step": 41562 + }, + { + "epoch": 5.44, + "learning_rate": 4.891397653579485e-06, + "loss": 0.0078, + "step": 41563 + }, + { + "epoch": 5.44, + "learning_rate": 4.890941759511519e-06, + "loss": 0.0077, + "step": 41564 + }, + { + "epoch": 5.44, + "learning_rate": 4.890485879812512e-06, + "loss": 0.005, + "step": 41565 + }, + { + "epoch": 5.44, + "learning_rate": 4.89003001448374e-06, + "loss": 0.0081, + "step": 41566 + }, + { + "epoch": 5.44, + "learning_rate": 4.889574163526484e-06, + "loss": 0.0063, + "step": 41567 + }, + { + "epoch": 5.45, + "learning_rate": 4.889118326942027e-06, + "loss": 0.0086, + "step": 41568 + }, + { + "epoch": 5.45, + "learning_rate": 4.8886625047316576e-06, + "loss": 0.0072, + "step": 41569 + }, + { + "epoch": 5.45, + "learning_rate": 4.888206696896651e-06, + "loss": 0.0098, + "step": 41570 + }, + { + "epoch": 5.45, + "learning_rate": 4.887750903438288e-06, + "loss": 0.0103, + "step": 41571 + }, + { + "epoch": 5.45, + "learning_rate": 4.887295124357858e-06, + "loss": 0.0065, + "step": 41572 + }, + { + "epoch": 5.45, + "learning_rate": 4.886839359656637e-06, + "loss": 0.006, + "step": 41573 + }, + { + "epoch": 5.45, + "learning_rate": 4.886383609335909e-06, + "loss": 0.0094, + "step": 41574 + }, + { + "epoch": 5.45, + "learning_rate": 4.885927873396951e-06, + "loss": 0.0075, + "step": 41575 + }, + { + "epoch": 5.45, + "learning_rate": 4.8854721518410516e-06, + "loss": 0.0065, + "step": 41576 + }, + { + "epoch": 5.45, + "learning_rate": 4.8850164446694895e-06, + "loss": 0.0081, + "step": 41577 + }, + { + "epoch": 5.45, + "learning_rate": 4.884560751883546e-06, + "loss": 0.0086, + "step": 41578 + }, + { + "epoch": 5.45, + "learning_rate": 4.8841050734844965e-06, + "loss": 0.0065, + "step": 41579 + }, + { + "epoch": 5.45, + "learning_rate": 4.8836494094736355e-06, + "loss": 0.0059, + "step": 41580 + }, + { + "epoch": 5.45, + "learning_rate": 4.883193759852238e-06, + "loss": 0.0084, + "step": 41581 + }, + { + "epoch": 5.45, + "learning_rate": 4.882738124621586e-06, + "loss": 0.0066, + "step": 41582 + }, + { + "epoch": 5.45, + "learning_rate": 4.882282503782957e-06, + "loss": 0.0101, + "step": 41583 + }, + { + "epoch": 5.45, + "learning_rate": 4.88182689733764e-06, + "loss": 0.0088, + "step": 41584 + }, + { + "epoch": 5.45, + "learning_rate": 4.881371305286911e-06, + "loss": 0.007, + "step": 41585 + }, + { + "epoch": 5.45, + "learning_rate": 4.880915727632051e-06, + "loss": 0.0059, + "step": 41586 + }, + { + "epoch": 5.45, + "learning_rate": 4.8804601643743445e-06, + "loss": 0.0058, + "step": 41587 + }, + { + "epoch": 5.45, + "learning_rate": 4.880004615515072e-06, + "loss": 0.0058, + "step": 41588 + }, + { + "epoch": 5.45, + "learning_rate": 4.879549081055514e-06, + "loss": 0.0065, + "step": 41589 + }, + { + "epoch": 5.45, + "learning_rate": 4.879093560996947e-06, + "loss": 0.006, + "step": 41590 + }, + { + "epoch": 5.45, + "learning_rate": 4.878638055340658e-06, + "loss": 0.0087, + "step": 41591 + }, + { + "epoch": 5.45, + "learning_rate": 4.87818256408793e-06, + "loss": 0.0067, + "step": 41592 + }, + { + "epoch": 5.45, + "learning_rate": 4.877727087240039e-06, + "loss": 0.0062, + "step": 41593 + }, + { + "epoch": 5.45, + "learning_rate": 4.877271624798266e-06, + "loss": 0.0072, + "step": 41594 + }, + { + "epoch": 5.45, + "learning_rate": 4.876816176763897e-06, + "loss": 0.0065, + "step": 41595 + }, + { + "epoch": 5.45, + "learning_rate": 4.8763607431382095e-06, + "loss": 0.007, + "step": 41596 + }, + { + "epoch": 5.45, + "learning_rate": 4.875905323922485e-06, + "loss": 0.0074, + "step": 41597 + }, + { + "epoch": 5.45, + "learning_rate": 4.8754499191180004e-06, + "loss": 0.0058, + "step": 41598 + }, + { + "epoch": 5.45, + "learning_rate": 4.874994528726044e-06, + "loss": 0.0063, + "step": 41599 + }, + { + "epoch": 5.45, + "learning_rate": 4.874539152747891e-06, + "loss": 0.0052, + "step": 41600 + }, + { + "epoch": 5.45, + "learning_rate": 4.8740837911848216e-06, + "loss": 0.009, + "step": 41601 + }, + { + "epoch": 5.45, + "learning_rate": 4.873628444038118e-06, + "loss": 0.0069, + "step": 41602 + }, + { + "epoch": 5.45, + "learning_rate": 4.873173111309064e-06, + "loss": 0.0076, + "step": 41603 + }, + { + "epoch": 5.45, + "learning_rate": 4.872717792998939e-06, + "loss": 0.0077, + "step": 41604 + }, + { + "epoch": 5.45, + "learning_rate": 4.872262489109022e-06, + "loss": 0.0084, + "step": 41605 + }, + { + "epoch": 5.45, + "learning_rate": 4.871807199640589e-06, + "loss": 0.0102, + "step": 41606 + }, + { + "epoch": 5.45, + "learning_rate": 4.871351924594929e-06, + "loss": 0.0054, + "step": 41607 + }, + { + "epoch": 5.45, + "learning_rate": 4.87089666397332e-06, + "loss": 0.0058, + "step": 41608 + }, + { + "epoch": 5.45, + "learning_rate": 4.8704414177770365e-06, + "loss": 0.0071, + "step": 41609 + }, + { + "epoch": 5.45, + "learning_rate": 4.869986186007367e-06, + "loss": 0.0066, + "step": 41610 + }, + { + "epoch": 5.45, + "learning_rate": 4.869530968665588e-06, + "loss": 0.005, + "step": 41611 + }, + { + "epoch": 5.45, + "learning_rate": 4.8690757657529765e-06, + "loss": 0.0072, + "step": 41612 + }, + { + "epoch": 5.45, + "learning_rate": 4.868620577270819e-06, + "loss": 0.0055, + "step": 41613 + }, + { + "epoch": 5.45, + "learning_rate": 4.8681654032203906e-06, + "loss": 0.0071, + "step": 41614 + }, + { + "epoch": 5.45, + "learning_rate": 4.8677102436029765e-06, + "loss": 0.0061, + "step": 41615 + }, + { + "epoch": 5.45, + "learning_rate": 4.867255098419854e-06, + "loss": 0.0106, + "step": 41616 + }, + { + "epoch": 5.45, + "learning_rate": 4.8667999676722995e-06, + "loss": 0.0056, + "step": 41617 + }, + { + "epoch": 5.45, + "learning_rate": 4.8663448513616005e-06, + "loss": 0.0052, + "step": 41618 + }, + { + "epoch": 5.45, + "learning_rate": 4.865889749489035e-06, + "loss": 0.0048, + "step": 41619 + }, + { + "epoch": 5.45, + "learning_rate": 4.865434662055879e-06, + "loss": 0.0086, + "step": 41620 + }, + { + "epoch": 5.45, + "learning_rate": 4.864979589063411e-06, + "loss": 0.0044, + "step": 41621 + }, + { + "epoch": 5.45, + "learning_rate": 4.864524530512919e-06, + "loss": 0.0063, + "step": 41622 + }, + { + "epoch": 5.45, + "learning_rate": 4.864069486405674e-06, + "loss": 0.0095, + "step": 41623 + }, + { + "epoch": 5.45, + "learning_rate": 4.863614456742963e-06, + "loss": 0.0057, + "step": 41624 + }, + { + "epoch": 5.45, + "learning_rate": 4.863159441526061e-06, + "loss": 0.0076, + "step": 41625 + }, + { + "epoch": 5.45, + "learning_rate": 4.862704440756251e-06, + "loss": 0.0091, + "step": 41626 + }, + { + "epoch": 5.45, + "learning_rate": 4.862249454434813e-06, + "loss": 0.0095, + "step": 41627 + }, + { + "epoch": 5.45, + "learning_rate": 4.861794482563024e-06, + "loss": 0.005, + "step": 41628 + }, + { + "epoch": 5.45, + "learning_rate": 4.861339525142161e-06, + "loss": 0.0071, + "step": 41629 + }, + { + "epoch": 5.45, + "learning_rate": 4.860884582173512e-06, + "loss": 0.005, + "step": 41630 + }, + { + "epoch": 5.45, + "learning_rate": 4.8604296536583496e-06, + "loss": 0.0056, + "step": 41631 + }, + { + "epoch": 5.45, + "learning_rate": 4.859974739597951e-06, + "loss": 0.0069, + "step": 41632 + }, + { + "epoch": 5.45, + "learning_rate": 4.859519839993605e-06, + "loss": 0.007, + "step": 41633 + }, + { + "epoch": 5.45, + "learning_rate": 4.859064954846581e-06, + "loss": 0.0088, + "step": 41634 + }, + { + "epoch": 5.45, + "learning_rate": 4.858610084158167e-06, + "loss": 0.0086, + "step": 41635 + }, + { + "epoch": 5.45, + "learning_rate": 4.858155227929635e-06, + "loss": 0.0068, + "step": 41636 + }, + { + "epoch": 5.45, + "learning_rate": 4.857700386162271e-06, + "loss": 0.0059, + "step": 41637 + }, + { + "epoch": 5.45, + "learning_rate": 4.857245558857351e-06, + "loss": 0.007, + "step": 41638 + }, + { + "epoch": 5.45, + "learning_rate": 4.856790746016154e-06, + "loss": 0.0073, + "step": 41639 + }, + { + "epoch": 5.45, + "learning_rate": 4.856335947639954e-06, + "loss": 0.0067, + "step": 41640 + }, + { + "epoch": 5.45, + "learning_rate": 4.855881163730041e-06, + "loss": 0.0079, + "step": 41641 + }, + { + "epoch": 5.45, + "learning_rate": 4.855426394287687e-06, + "loss": 0.0091, + "step": 41642 + }, + { + "epoch": 5.45, + "learning_rate": 4.854971639314173e-06, + "loss": 0.0064, + "step": 41643 + }, + { + "epoch": 5.46, + "learning_rate": 4.854516898810773e-06, + "loss": 0.0078, + "step": 41644 + }, + { + "epoch": 5.46, + "learning_rate": 4.854062172778771e-06, + "loss": 0.0075, + "step": 41645 + }, + { + "epoch": 5.46, + "learning_rate": 4.853607461219449e-06, + "loss": 0.0067, + "step": 41646 + }, + { + "epoch": 5.46, + "learning_rate": 4.85315276413408e-06, + "loss": 0.008, + "step": 41647 + }, + { + "epoch": 5.46, + "learning_rate": 4.852698081523942e-06, + "loss": 0.0073, + "step": 41648 + }, + { + "epoch": 5.46, + "learning_rate": 4.852243413390321e-06, + "loss": 0.0066, + "step": 41649 + }, + { + "epoch": 5.46, + "learning_rate": 4.8517887597344905e-06, + "loss": 0.0087, + "step": 41650 + }, + { + "epoch": 5.46, + "learning_rate": 4.851334120557727e-06, + "loss": 0.0097, + "step": 41651 + }, + { + "epoch": 5.46, + "learning_rate": 4.850879495861316e-06, + "loss": 0.006, + "step": 41652 + }, + { + "epoch": 5.46, + "learning_rate": 4.850424885646531e-06, + "loss": 0.0076, + "step": 41653 + }, + { + "epoch": 5.46, + "learning_rate": 4.849970289914653e-06, + "loss": 0.0083, + "step": 41654 + }, + { + "epoch": 5.46, + "learning_rate": 4.849515708666954e-06, + "loss": 0.0059, + "step": 41655 + }, + { + "epoch": 5.46, + "learning_rate": 4.8490611419047195e-06, + "loss": 0.0064, + "step": 41656 + }, + { + "epoch": 5.46, + "learning_rate": 4.848606589629229e-06, + "loss": 0.0074, + "step": 41657 + }, + { + "epoch": 5.46, + "learning_rate": 4.848152051841758e-06, + "loss": 0.0068, + "step": 41658 + }, + { + "epoch": 5.46, + "learning_rate": 4.847697528543581e-06, + "loss": 0.0056, + "step": 41659 + }, + { + "epoch": 5.46, + "learning_rate": 4.8472430197359855e-06, + "loss": 0.0058, + "step": 41660 + }, + { + "epoch": 5.46, + "learning_rate": 4.846788525420244e-06, + "loss": 0.0068, + "step": 41661 + }, + { + "epoch": 5.46, + "learning_rate": 4.8463340455976345e-06, + "loss": 0.0051, + "step": 41662 + }, + { + "epoch": 5.46, + "learning_rate": 4.845879580269433e-06, + "loss": 0.0089, + "step": 41663 + }, + { + "epoch": 5.46, + "learning_rate": 4.845425129436924e-06, + "loss": 0.0045, + "step": 41664 + }, + { + "epoch": 5.46, + "learning_rate": 4.844970693101384e-06, + "loss": 0.0055, + "step": 41665 + }, + { + "epoch": 5.46, + "learning_rate": 4.844516271264084e-06, + "loss": 0.0065, + "step": 41666 + }, + { + "epoch": 5.46, + "learning_rate": 4.844061863926307e-06, + "loss": 0.0114, + "step": 41667 + }, + { + "epoch": 5.46, + "learning_rate": 4.843607471089337e-06, + "loss": 0.0099, + "step": 41668 + }, + { + "epoch": 5.46, + "learning_rate": 4.843153092754445e-06, + "loss": 0.0091, + "step": 41669 + }, + { + "epoch": 5.46, + "learning_rate": 4.842698728922911e-06, + "loss": 0.0068, + "step": 41670 + }, + { + "epoch": 5.46, + "learning_rate": 4.842244379596008e-06, + "loss": 0.0073, + "step": 41671 + }, + { + "epoch": 5.46, + "learning_rate": 4.841790044775022e-06, + "loss": 0.0072, + "step": 41672 + }, + { + "epoch": 5.46, + "learning_rate": 4.8413357244612265e-06, + "loss": 0.0058, + "step": 41673 + }, + { + "epoch": 5.46, + "learning_rate": 4.840881418655896e-06, + "loss": 0.0078, + "step": 41674 + }, + { + "epoch": 5.46, + "learning_rate": 4.840427127360315e-06, + "loss": 0.0075, + "step": 41675 + }, + { + "epoch": 5.46, + "learning_rate": 4.8399728505757594e-06, + "loss": 0.0067, + "step": 41676 + }, + { + "epoch": 5.46, + "learning_rate": 4.8395185883035005e-06, + "loss": 0.0089, + "step": 41677 + }, + { + "epoch": 5.46, + "learning_rate": 4.839064340544825e-06, + "loss": 0.0063, + "step": 41678 + }, + { + "epoch": 5.46, + "learning_rate": 4.838610107301002e-06, + "loss": 0.0071, + "step": 41679 + }, + { + "epoch": 5.46, + "learning_rate": 4.838155888573317e-06, + "loss": 0.0075, + "step": 41680 + }, + { + "epoch": 5.46, + "learning_rate": 4.837701684363044e-06, + "loss": 0.0059, + "step": 41681 + }, + { + "epoch": 5.46, + "learning_rate": 4.8372474946714555e-06, + "loss": 0.0093, + "step": 41682 + }, + { + "epoch": 5.46, + "learning_rate": 4.836793319499838e-06, + "loss": 0.0062, + "step": 41683 + }, + { + "epoch": 5.46, + "learning_rate": 4.8363391588494645e-06, + "loss": 0.0078, + "step": 41684 + }, + { + "epoch": 5.46, + "learning_rate": 4.835885012721612e-06, + "loss": 0.0059, + "step": 41685 + }, + { + "epoch": 5.46, + "learning_rate": 4.835430881117555e-06, + "loss": 0.006, + "step": 41686 + }, + { + "epoch": 5.46, + "learning_rate": 4.834976764038576e-06, + "loss": 0.0055, + "step": 41687 + }, + { + "epoch": 5.46, + "learning_rate": 4.834522661485945e-06, + "loss": 0.0069, + "step": 41688 + }, + { + "epoch": 5.46, + "learning_rate": 4.83406857346095e-06, + "loss": 0.0051, + "step": 41689 + }, + { + "epoch": 5.46, + "learning_rate": 4.833614499964857e-06, + "loss": 0.0086, + "step": 41690 + }, + { + "epoch": 5.46, + "learning_rate": 4.833160440998953e-06, + "loss": 0.0068, + "step": 41691 + }, + { + "epoch": 5.46, + "learning_rate": 4.83270639656451e-06, + "loss": 0.0082, + "step": 41692 + }, + { + "epoch": 5.46, + "learning_rate": 4.832252366662803e-06, + "loss": 0.0075, + "step": 41693 + }, + { + "epoch": 5.46, + "learning_rate": 4.831798351295109e-06, + "loss": 0.0076, + "step": 41694 + }, + { + "epoch": 5.46, + "learning_rate": 4.83134435046271e-06, + "loss": 0.0075, + "step": 41695 + }, + { + "epoch": 5.46, + "learning_rate": 4.830890364166881e-06, + "loss": 0.0087, + "step": 41696 + }, + { + "epoch": 5.46, + "learning_rate": 4.830436392408893e-06, + "loss": 0.0075, + "step": 41697 + }, + { + "epoch": 5.46, + "learning_rate": 4.82998243519003e-06, + "loss": 0.009, + "step": 41698 + }, + { + "epoch": 5.46, + "learning_rate": 4.8295284925115635e-06, + "loss": 0.0066, + "step": 41699 + }, + { + "epoch": 5.46, + "learning_rate": 4.829074564374777e-06, + "loss": 0.0082, + "step": 41700 + }, + { + "epoch": 5.46, + "learning_rate": 4.828620650780939e-06, + "loss": 0.0059, + "step": 41701 + }, + { + "epoch": 5.46, + "learning_rate": 4.8281667517313325e-06, + "loss": 0.0069, + "step": 41702 + }, + { + "epoch": 5.46, + "learning_rate": 4.827712867227233e-06, + "loss": 0.0084, + "step": 41703 + }, + { + "epoch": 5.46, + "learning_rate": 4.827258997269914e-06, + "loss": 0.0072, + "step": 41704 + }, + { + "epoch": 5.46, + "learning_rate": 4.826805141860652e-06, + "loss": 0.0074, + "step": 41705 + }, + { + "epoch": 5.46, + "learning_rate": 4.826351301000728e-06, + "loss": 0.007, + "step": 41706 + }, + { + "epoch": 5.46, + "learning_rate": 4.825897474691414e-06, + "loss": 0.0054, + "step": 41707 + }, + { + "epoch": 5.46, + "learning_rate": 4.825443662933987e-06, + "loss": 0.008, + "step": 41708 + }, + { + "epoch": 5.46, + "learning_rate": 4.824989865729726e-06, + "loss": 0.0071, + "step": 41709 + }, + { + "epoch": 5.46, + "learning_rate": 4.824536083079901e-06, + "loss": 0.0069, + "step": 41710 + }, + { + "epoch": 5.46, + "learning_rate": 4.824082314985797e-06, + "loss": 0.0073, + "step": 41711 + }, + { + "epoch": 5.46, + "learning_rate": 4.8236285614486865e-06, + "loss": 0.0064, + "step": 41712 + }, + { + "epoch": 5.46, + "learning_rate": 4.823174822469841e-06, + "loss": 0.0092, + "step": 41713 + }, + { + "epoch": 5.46, + "learning_rate": 4.822721098050544e-06, + "loss": 0.0068, + "step": 41714 + }, + { + "epoch": 5.46, + "learning_rate": 4.8222673881920675e-06, + "loss": 0.0044, + "step": 41715 + }, + { + "epoch": 5.46, + "learning_rate": 4.821813692895686e-06, + "loss": 0.009, + "step": 41716 + }, + { + "epoch": 5.46, + "learning_rate": 4.8213600121626815e-06, + "loss": 0.0071, + "step": 41717 + }, + { + "epoch": 5.46, + "learning_rate": 4.820906345994325e-06, + "loss": 0.0073, + "step": 41718 + }, + { + "epoch": 5.46, + "learning_rate": 4.820452694391895e-06, + "loss": 0.0057, + "step": 41719 + }, + { + "epoch": 5.47, + "learning_rate": 4.819999057356661e-06, + "loss": 0.0074, + "step": 41720 + }, + { + "epoch": 5.47, + "learning_rate": 4.819545434889904e-06, + "loss": 0.0085, + "step": 41721 + }, + { + "epoch": 5.47, + "learning_rate": 4.819091826992903e-06, + "loss": 0.0053, + "step": 41722 + }, + { + "epoch": 5.47, + "learning_rate": 4.818638233666931e-06, + "loss": 0.0068, + "step": 41723 + }, + { + "epoch": 5.47, + "learning_rate": 4.818184654913259e-06, + "loss": 0.0106, + "step": 41724 + }, + { + "epoch": 5.47, + "learning_rate": 4.817731090733171e-06, + "loss": 0.0087, + "step": 41725 + }, + { + "epoch": 5.47, + "learning_rate": 4.817277541127939e-06, + "loss": 0.0096, + "step": 41726 + }, + { + "epoch": 5.47, + "learning_rate": 4.8168240060988354e-06, + "loss": 0.0066, + "step": 41727 + }, + { + "epoch": 5.47, + "learning_rate": 4.816370485647137e-06, + "loss": 0.0079, + "step": 41728 + }, + { + "epoch": 5.47, + "learning_rate": 4.8159169797741235e-06, + "loss": 0.0056, + "step": 41729 + }, + { + "epoch": 5.47, + "learning_rate": 4.815463488481067e-06, + "loss": 0.0061, + "step": 41730 + }, + { + "epoch": 5.47, + "learning_rate": 4.81501001176924e-06, + "loss": 0.0068, + "step": 41731 + }, + { + "epoch": 5.47, + "learning_rate": 4.81455654963992e-06, + "loss": 0.0059, + "step": 41732 + }, + { + "epoch": 5.47, + "learning_rate": 4.81410310209439e-06, + "loss": 0.0063, + "step": 41733 + }, + { + "epoch": 5.47, + "learning_rate": 4.813649669133916e-06, + "loss": 0.0067, + "step": 41734 + }, + { + "epoch": 5.47, + "learning_rate": 4.813196250759777e-06, + "loss": 0.0082, + "step": 41735 + }, + { + "epoch": 5.47, + "learning_rate": 4.812742846973245e-06, + "loss": 0.0074, + "step": 41736 + }, + { + "epoch": 5.47, + "learning_rate": 4.8122894577756e-06, + "loss": 0.0071, + "step": 41737 + }, + { + "epoch": 5.47, + "learning_rate": 4.811836083168114e-06, + "loss": 0.0088, + "step": 41738 + }, + { + "epoch": 5.47, + "learning_rate": 4.8113827231520605e-06, + "loss": 0.0057, + "step": 41739 + }, + { + "epoch": 5.47, + "learning_rate": 4.81092937772872e-06, + "loss": 0.0074, + "step": 41740 + }, + { + "epoch": 5.47, + "learning_rate": 4.810476046899364e-06, + "loss": 0.0093, + "step": 41741 + }, + { + "epoch": 5.47, + "learning_rate": 4.810022730665268e-06, + "loss": 0.0076, + "step": 41742 + }, + { + "epoch": 5.47, + "learning_rate": 4.809569429027703e-06, + "loss": 0.0073, + "step": 41743 + }, + { + "epoch": 5.47, + "learning_rate": 4.809116141987947e-06, + "loss": 0.0096, + "step": 41744 + }, + { + "epoch": 5.47, + "learning_rate": 4.808662869547279e-06, + "loss": 0.0075, + "step": 41745 + }, + { + "epoch": 5.47, + "learning_rate": 4.808209611706971e-06, + "loss": 0.0065, + "step": 41746 + }, + { + "epoch": 5.47, + "learning_rate": 4.807756368468294e-06, + "loss": 0.0073, + "step": 41747 + }, + { + "epoch": 5.47, + "learning_rate": 4.807303139832528e-06, + "loss": 0.0064, + "step": 41748 + }, + { + "epoch": 5.47, + "learning_rate": 4.806849925800946e-06, + "loss": 0.009, + "step": 41749 + }, + { + "epoch": 5.47, + "learning_rate": 4.806396726374823e-06, + "loss": 0.0084, + "step": 41750 + }, + { + "epoch": 5.47, + "learning_rate": 4.805943541555427e-06, + "loss": 0.0089, + "step": 41751 + }, + { + "epoch": 5.47, + "learning_rate": 4.805490371344043e-06, + "loss": 0.0063, + "step": 41752 + }, + { + "epoch": 5.47, + "learning_rate": 4.8050372157419405e-06, + "loss": 0.0055, + "step": 41753 + }, + { + "epoch": 5.47, + "learning_rate": 4.80458407475039e-06, + "loss": 0.0058, + "step": 41754 + }, + { + "epoch": 5.47, + "learning_rate": 4.8041309483706696e-06, + "loss": 0.0101, + "step": 41755 + }, + { + "epoch": 5.47, + "learning_rate": 4.803677836604059e-06, + "loss": 0.006, + "step": 41756 + }, + { + "epoch": 5.47, + "learning_rate": 4.803224739451829e-06, + "loss": 0.0059, + "step": 41757 + }, + { + "epoch": 5.47, + "learning_rate": 4.802771656915246e-06, + "loss": 0.0053, + "step": 41758 + }, + { + "epoch": 5.47, + "learning_rate": 4.802318588995597e-06, + "loss": 0.0071, + "step": 41759 + }, + { + "epoch": 5.47, + "learning_rate": 4.801865535694149e-06, + "loss": 0.0053, + "step": 41760 + }, + { + "epoch": 5.47, + "learning_rate": 4.801412497012178e-06, + "loss": 0.007, + "step": 41761 + }, + { + "epoch": 5.47, + "learning_rate": 4.8009594729509535e-06, + "loss": 0.0065, + "step": 41762 + }, + { + "epoch": 5.47, + "learning_rate": 4.800506463511757e-06, + "loss": 0.0069, + "step": 41763 + }, + { + "epoch": 5.47, + "learning_rate": 4.80005346869586e-06, + "loss": 0.0052, + "step": 41764 + }, + { + "epoch": 5.47, + "learning_rate": 4.7996004885045325e-06, + "loss": 0.0066, + "step": 41765 + }, + { + "epoch": 5.47, + "learning_rate": 4.799147522939049e-06, + "loss": 0.0073, + "step": 41766 + }, + { + "epoch": 5.47, + "learning_rate": 4.7986945720006926e-06, + "loss": 0.0078, + "step": 41767 + }, + { + "epoch": 5.47, + "learning_rate": 4.79824163569073e-06, + "loss": 0.0073, + "step": 41768 + }, + { + "epoch": 5.47, + "learning_rate": 4.797788714010435e-06, + "loss": 0.0066, + "step": 41769 + }, + { + "epoch": 5.47, + "learning_rate": 4.797335806961079e-06, + "loss": 0.0073, + "step": 41770 + }, + { + "epoch": 5.47, + "learning_rate": 4.796882914543944e-06, + "loss": 0.0069, + "step": 41771 + }, + { + "epoch": 5.47, + "learning_rate": 4.796430036760296e-06, + "loss": 0.0066, + "step": 41772 + }, + { + "epoch": 5.47, + "learning_rate": 4.795977173611409e-06, + "loss": 0.0042, + "step": 41773 + }, + { + "epoch": 5.47, + "learning_rate": 4.795524325098564e-06, + "loss": 0.0076, + "step": 41774 + }, + { + "epoch": 5.47, + "learning_rate": 4.795071491223028e-06, + "loss": 0.0114, + "step": 41775 + }, + { + "epoch": 5.47, + "learning_rate": 4.794618671986073e-06, + "loss": 0.0077, + "step": 41776 + }, + { + "epoch": 5.47, + "learning_rate": 4.79416586738898e-06, + "loss": 0.0071, + "step": 41777 + }, + { + "epoch": 5.47, + "learning_rate": 4.7937130774330145e-06, + "loss": 0.0071, + "step": 41778 + }, + { + "epoch": 5.47, + "learning_rate": 4.793260302119457e-06, + "loss": 0.0069, + "step": 41779 + }, + { + "epoch": 5.47, + "learning_rate": 4.792807541449578e-06, + "loss": 0.0084, + "step": 41780 + }, + { + "epoch": 5.47, + "learning_rate": 4.792354795424646e-06, + "loss": 0.0076, + "step": 41781 + }, + { + "epoch": 5.47, + "learning_rate": 4.791902064045944e-06, + "loss": 0.007, + "step": 41782 + }, + { + "epoch": 5.47, + "learning_rate": 4.791449347314738e-06, + "loss": 0.0088, + "step": 41783 + }, + { + "epoch": 5.47, + "learning_rate": 4.790996645232304e-06, + "loss": 0.0055, + "step": 41784 + }, + { + "epoch": 5.47, + "learning_rate": 4.790543957799913e-06, + "loss": 0.0056, + "step": 41785 + }, + { + "epoch": 5.47, + "learning_rate": 4.790091285018841e-06, + "loss": 0.0063, + "step": 41786 + }, + { + "epoch": 5.47, + "learning_rate": 4.789638626890356e-06, + "loss": 0.0088, + "step": 41787 + }, + { + "epoch": 5.47, + "learning_rate": 4.789185983415739e-06, + "loss": 0.0102, + "step": 41788 + }, + { + "epoch": 5.47, + "learning_rate": 4.788733354596257e-06, + "loss": 0.0096, + "step": 41789 + }, + { + "epoch": 5.47, + "learning_rate": 4.788280740433187e-06, + "loss": 0.0073, + "step": 41790 + }, + { + "epoch": 5.47, + "learning_rate": 4.7878281409278e-06, + "loss": 0.0065, + "step": 41791 + }, + { + "epoch": 5.47, + "learning_rate": 4.787375556081369e-06, + "loss": 0.0082, + "step": 41792 + }, + { + "epoch": 5.47, + "learning_rate": 4.786922985895163e-06, + "loss": 0.0098, + "step": 41793 + }, + { + "epoch": 5.47, + "learning_rate": 4.786470430370462e-06, + "loss": 0.0067, + "step": 41794 + }, + { + "epoch": 5.47, + "learning_rate": 4.786017889508535e-06, + "loss": 0.009, + "step": 41795 + }, + { + "epoch": 5.47, + "learning_rate": 4.785565363310652e-06, + "loss": 0.008, + "step": 41796 + }, + { + "epoch": 5.48, + "learning_rate": 4.785112851778089e-06, + "loss": 0.0047, + "step": 41797 + }, + { + "epoch": 5.48, + "learning_rate": 4.784660354912121e-06, + "loss": 0.0059, + "step": 41798 + }, + { + "epoch": 5.48, + "learning_rate": 4.784207872714019e-06, + "loss": 0.0095, + "step": 41799 + }, + { + "epoch": 5.48, + "learning_rate": 4.783755405185054e-06, + "loss": 0.0113, + "step": 41800 + }, + { + "epoch": 5.48, + "learning_rate": 4.783302952326495e-06, + "loss": 0.009, + "step": 41801 + }, + { + "epoch": 5.48, + "learning_rate": 4.7828505141396235e-06, + "loss": 0.0069, + "step": 41802 + }, + { + "epoch": 5.48, + "learning_rate": 4.782398090625706e-06, + "loss": 0.0066, + "step": 41803 + }, + { + "epoch": 5.48, + "learning_rate": 4.7819456817860135e-06, + "loss": 0.0067, + "step": 41804 + }, + { + "epoch": 5.48, + "learning_rate": 4.781493287621824e-06, + "loss": 0.0095, + "step": 41805 + }, + { + "epoch": 5.48, + "learning_rate": 4.7810409081344066e-06, + "loss": 0.0071, + "step": 41806 + }, + { + "epoch": 5.48, + "learning_rate": 4.780588543325033e-06, + "loss": 0.0088, + "step": 41807 + }, + { + "epoch": 5.48, + "learning_rate": 4.78013619319497e-06, + "loss": 0.0071, + "step": 41808 + }, + { + "epoch": 5.48, + "learning_rate": 4.779683857745503e-06, + "loss": 0.0068, + "step": 41809 + }, + { + "epoch": 5.48, + "learning_rate": 4.779231536977898e-06, + "loss": 0.0064, + "step": 41810 + }, + { + "epoch": 5.48, + "learning_rate": 4.778779230893424e-06, + "loss": 0.0061, + "step": 41811 + }, + { + "epoch": 5.48, + "learning_rate": 4.778326939493354e-06, + "loss": 0.0064, + "step": 41812 + }, + { + "epoch": 5.48, + "learning_rate": 4.777874662778964e-06, + "loss": 0.009, + "step": 41813 + }, + { + "epoch": 5.48, + "learning_rate": 4.777422400751523e-06, + "loss": 0.0076, + "step": 41814 + }, + { + "epoch": 5.48, + "learning_rate": 4.776970153412305e-06, + "loss": 0.0079, + "step": 41815 + }, + { + "epoch": 5.48, + "learning_rate": 4.776517920762574e-06, + "loss": 0.0066, + "step": 41816 + }, + { + "epoch": 5.48, + "learning_rate": 4.776065702803614e-06, + "loss": 0.0056, + "step": 41817 + }, + { + "epoch": 5.48, + "learning_rate": 4.775613499536691e-06, + "loss": 0.0071, + "step": 41818 + }, + { + "epoch": 5.48, + "learning_rate": 4.775161310963072e-06, + "loss": 0.0089, + "step": 41819 + }, + { + "epoch": 5.48, + "learning_rate": 4.774709137084035e-06, + "loss": 0.0077, + "step": 41820 + }, + { + "epoch": 5.48, + "learning_rate": 4.774256977900853e-06, + "loss": 0.0086, + "step": 41821 + }, + { + "epoch": 5.48, + "learning_rate": 4.7738048334147946e-06, + "loss": 0.0074, + "step": 41822 + }, + { + "epoch": 5.48, + "learning_rate": 4.773352703627129e-06, + "loss": 0.0087, + "step": 41823 + }, + { + "epoch": 5.48, + "learning_rate": 4.772900588539134e-06, + "loss": 0.0099, + "step": 41824 + }, + { + "epoch": 5.48, + "learning_rate": 4.772448488152077e-06, + "loss": 0.0091, + "step": 41825 + }, + { + "epoch": 5.48, + "learning_rate": 4.7719964024672315e-06, + "loss": 0.0078, + "step": 41826 + }, + { + "epoch": 5.48, + "learning_rate": 4.771544331485863e-06, + "loss": 0.0075, + "step": 41827 + }, + { + "epoch": 5.48, + "learning_rate": 4.771092275209252e-06, + "loss": 0.0076, + "step": 41828 + }, + { + "epoch": 5.48, + "learning_rate": 4.770640233638665e-06, + "loss": 0.0055, + "step": 41829 + }, + { + "epoch": 5.48, + "learning_rate": 4.77018820677537e-06, + "loss": 0.0086, + "step": 41830 + }, + { + "epoch": 5.48, + "learning_rate": 4.769736194620643e-06, + "loss": 0.0063, + "step": 41831 + }, + { + "epoch": 5.48, + "learning_rate": 4.769284197175757e-06, + "loss": 0.0071, + "step": 41832 + }, + { + "epoch": 5.48, + "learning_rate": 4.768832214441982e-06, + "loss": 0.0081, + "step": 41833 + }, + { + "epoch": 5.48, + "learning_rate": 4.768380246420587e-06, + "loss": 0.0081, + "step": 41834 + }, + { + "epoch": 5.48, + "learning_rate": 4.76792829311284e-06, + "loss": 0.007, + "step": 41835 + }, + { + "epoch": 5.48, + "learning_rate": 4.767476354520021e-06, + "loss": 0.0072, + "step": 41836 + }, + { + "epoch": 5.48, + "learning_rate": 4.7670244306433945e-06, + "loss": 0.0094, + "step": 41837 + }, + { + "epoch": 5.48, + "learning_rate": 4.766572521484231e-06, + "loss": 0.0095, + "step": 41838 + }, + { + "epoch": 5.48, + "learning_rate": 4.766120627043806e-06, + "loss": 0.0089, + "step": 41839 + }, + { + "epoch": 5.48, + "learning_rate": 4.765668747323389e-06, + "loss": 0.0061, + "step": 41840 + }, + { + "epoch": 5.48, + "learning_rate": 4.765216882324245e-06, + "loss": 0.0084, + "step": 41841 + }, + { + "epoch": 5.48, + "learning_rate": 4.764765032047655e-06, + "loss": 0.0048, + "step": 41842 + }, + { + "epoch": 5.48, + "learning_rate": 4.764313196494879e-06, + "loss": 0.0082, + "step": 41843 + }, + { + "epoch": 5.48, + "learning_rate": 4.763861375667198e-06, + "loss": 0.0082, + "step": 41844 + }, + { + "epoch": 5.48, + "learning_rate": 4.7634095695658784e-06, + "loss": 0.0064, + "step": 41845 + }, + { + "epoch": 5.48, + "learning_rate": 4.762957778192187e-06, + "loss": 0.0083, + "step": 41846 + }, + { + "epoch": 5.48, + "learning_rate": 4.7625060015474e-06, + "loss": 0.0082, + "step": 41847 + }, + { + "epoch": 5.48, + "learning_rate": 4.762054239632787e-06, + "loss": 0.0087, + "step": 41848 + }, + { + "epoch": 5.48, + "learning_rate": 4.761602492449618e-06, + "loss": 0.0074, + "step": 41849 + }, + { + "epoch": 5.48, + "learning_rate": 4.761150759999159e-06, + "loss": 0.0105, + "step": 41850 + }, + { + "epoch": 5.48, + "learning_rate": 4.7606990422826885e-06, + "loss": 0.0087, + "step": 41851 + }, + { + "epoch": 5.48, + "learning_rate": 4.760247339301468e-06, + "loss": 0.0093, + "step": 41852 + }, + { + "epoch": 5.48, + "learning_rate": 4.759795651056778e-06, + "loss": 0.0058, + "step": 41853 + }, + { + "epoch": 5.48, + "learning_rate": 4.7593439775498795e-06, + "loss": 0.0047, + "step": 41854 + }, + { + "epoch": 5.48, + "learning_rate": 4.75889231878205e-06, + "loss": 0.0077, + "step": 41855 + }, + { + "epoch": 5.48, + "learning_rate": 4.758440674754557e-06, + "loss": 0.0046, + "step": 41856 + }, + { + "epoch": 5.48, + "learning_rate": 4.757989045468672e-06, + "loss": 0.0082, + "step": 41857 + }, + { + "epoch": 5.48, + "learning_rate": 4.757537430925658e-06, + "loss": 0.0084, + "step": 41858 + }, + { + "epoch": 5.48, + "learning_rate": 4.7570858311267954e-06, + "loss": 0.0057, + "step": 41859 + }, + { + "epoch": 5.48, + "learning_rate": 4.756634246073349e-06, + "loss": 0.0081, + "step": 41860 + }, + { + "epoch": 5.48, + "learning_rate": 4.756182675766586e-06, + "loss": 0.0078, + "step": 41861 + }, + { + "epoch": 5.48, + "learning_rate": 4.755731120207785e-06, + "loss": 0.0053, + "step": 41862 + }, + { + "epoch": 5.48, + "learning_rate": 4.755279579398207e-06, + "loss": 0.008, + "step": 41863 + }, + { + "epoch": 5.48, + "learning_rate": 4.754828053339128e-06, + "loss": 0.0096, + "step": 41864 + }, + { + "epoch": 5.48, + "learning_rate": 4.754376542031817e-06, + "loss": 0.0078, + "step": 41865 + }, + { + "epoch": 5.48, + "learning_rate": 4.753925045477539e-06, + "loss": 0.0049, + "step": 41866 + }, + { + "epoch": 5.48, + "learning_rate": 4.75347356367757e-06, + "loss": 0.0067, + "step": 41867 + }, + { + "epoch": 5.48, + "learning_rate": 4.753022096633177e-06, + "loss": 0.0087, + "step": 41868 + }, + { + "epoch": 5.48, + "learning_rate": 4.752570644345627e-06, + "loss": 0.0052, + "step": 41869 + }, + { + "epoch": 5.48, + "learning_rate": 4.752119206816198e-06, + "loss": 0.006, + "step": 41870 + }, + { + "epoch": 5.48, + "learning_rate": 4.751667784046152e-06, + "loss": 0.0078, + "step": 41871 + }, + { + "epoch": 5.48, + "learning_rate": 4.7512163760367625e-06, + "loss": 0.0096, + "step": 41872 + }, + { + "epoch": 5.49, + "learning_rate": 4.750764982789292e-06, + "loss": 0.0079, + "step": 41873 + }, + { + "epoch": 5.49, + "learning_rate": 4.750313604305017e-06, + "loss": 0.0043, + "step": 41874 + }, + { + "epoch": 5.49, + "learning_rate": 4.749862240585208e-06, + "loss": 0.0117, + "step": 41875 + }, + { + "epoch": 5.49, + "learning_rate": 4.749410891631132e-06, + "loss": 0.0079, + "step": 41876 + }, + { + "epoch": 5.49, + "learning_rate": 4.748959557444055e-06, + "loss": 0.0065, + "step": 41877 + }, + { + "epoch": 5.49, + "learning_rate": 4.748508238025253e-06, + "loss": 0.0059, + "step": 41878 + }, + { + "epoch": 5.49, + "learning_rate": 4.748056933375993e-06, + "loss": 0.011, + "step": 41879 + }, + { + "epoch": 5.49, + "learning_rate": 4.7476056434975416e-06, + "loss": 0.0078, + "step": 41880 + }, + { + "epoch": 5.49, + "learning_rate": 4.747154368391167e-06, + "loss": 0.0062, + "step": 41881 + }, + { + "epoch": 5.49, + "learning_rate": 4.746703108058144e-06, + "loss": 0.0071, + "step": 41882 + }, + { + "epoch": 5.49, + "learning_rate": 4.746251862499739e-06, + "loss": 0.0092, + "step": 41883 + }, + { + "epoch": 5.49, + "learning_rate": 4.745800631717218e-06, + "loss": 0.0041, + "step": 41884 + }, + { + "epoch": 5.49, + "learning_rate": 4.745349415711852e-06, + "loss": 0.0075, + "step": 41885 + }, + { + "epoch": 5.49, + "learning_rate": 4.744898214484914e-06, + "loss": 0.0053, + "step": 41886 + }, + { + "epoch": 5.49, + "learning_rate": 4.744447028037671e-06, + "loss": 0.0051, + "step": 41887 + }, + { + "epoch": 5.49, + "learning_rate": 4.7439958563713875e-06, + "loss": 0.0082, + "step": 41888 + }, + { + "epoch": 5.49, + "learning_rate": 4.743544699487338e-06, + "loss": 0.0061, + "step": 41889 + }, + { + "epoch": 5.49, + "learning_rate": 4.74309355738679e-06, + "loss": 0.0081, + "step": 41890 + }, + { + "epoch": 5.49, + "learning_rate": 4.742642430071012e-06, + "loss": 0.0072, + "step": 41891 + }, + { + "epoch": 5.49, + "learning_rate": 4.7421913175412674e-06, + "loss": 0.0073, + "step": 41892 + }, + { + "epoch": 5.49, + "learning_rate": 4.7417402197988335e-06, + "loss": 0.0087, + "step": 41893 + }, + { + "epoch": 5.49, + "learning_rate": 4.741289136844976e-06, + "loss": 0.0078, + "step": 41894 + }, + { + "epoch": 5.49, + "learning_rate": 4.740838068680962e-06, + "loss": 0.0057, + "step": 41895 + }, + { + "epoch": 5.49, + "learning_rate": 4.740387015308054e-06, + "loss": 0.0092, + "step": 41896 + }, + { + "epoch": 5.49, + "learning_rate": 4.739935976727535e-06, + "loss": 0.0056, + "step": 41897 + }, + { + "epoch": 5.49, + "learning_rate": 4.739484952940665e-06, + "loss": 0.0062, + "step": 41898 + }, + { + "epoch": 5.49, + "learning_rate": 4.739033943948714e-06, + "loss": 0.0062, + "step": 41899 + }, + { + "epoch": 5.49, + "learning_rate": 4.738582949752947e-06, + "loss": 0.0074, + "step": 41900 + }, + { + "epoch": 5.49, + "learning_rate": 4.738131970354637e-06, + "loss": 0.0086, + "step": 41901 + }, + { + "epoch": 5.49, + "learning_rate": 4.737681005755053e-06, + "loss": 0.0073, + "step": 41902 + }, + { + "epoch": 5.49, + "learning_rate": 4.737230055955455e-06, + "loss": 0.0078, + "step": 41903 + }, + { + "epoch": 5.49, + "learning_rate": 4.7367791209571224e-06, + "loss": 0.0087, + "step": 41904 + }, + { + "epoch": 5.49, + "learning_rate": 4.736328200761318e-06, + "loss": 0.0064, + "step": 41905 + }, + { + "epoch": 5.49, + "learning_rate": 4.73587729536931e-06, + "loss": 0.0059, + "step": 41906 + }, + { + "epoch": 5.49, + "learning_rate": 4.735426404782363e-06, + "loss": 0.0073, + "step": 41907 + }, + { + "epoch": 5.49, + "learning_rate": 4.734975529001748e-06, + "loss": 0.0083, + "step": 41908 + }, + { + "epoch": 5.49, + "learning_rate": 4.734524668028739e-06, + "loss": 0.0076, + "step": 41909 + }, + { + "epoch": 5.49, + "learning_rate": 4.734073821864599e-06, + "loss": 0.0089, + "step": 41910 + }, + { + "epoch": 5.49, + "learning_rate": 4.733622990510591e-06, + "loss": 0.0064, + "step": 41911 + }, + { + "epoch": 5.49, + "learning_rate": 4.733172173967992e-06, + "loss": 0.0071, + "step": 41912 + }, + { + "epoch": 5.49, + "learning_rate": 4.7327213722380675e-06, + "loss": 0.0078, + "step": 41913 + }, + { + "epoch": 5.49, + "learning_rate": 4.732270585322082e-06, + "loss": 0.0078, + "step": 41914 + }, + { + "epoch": 5.49, + "learning_rate": 4.731819813221301e-06, + "loss": 0.0064, + "step": 41915 + }, + { + "epoch": 5.49, + "learning_rate": 4.731369055937001e-06, + "loss": 0.0066, + "step": 41916 + }, + { + "epoch": 5.49, + "learning_rate": 4.730918313470444e-06, + "loss": 0.0068, + "step": 41917 + }, + { + "epoch": 5.49, + "learning_rate": 4.730467585822895e-06, + "loss": 0.0057, + "step": 41918 + }, + { + "epoch": 5.49, + "learning_rate": 4.730016872995626e-06, + "loss": 0.0062, + "step": 41919 + }, + { + "epoch": 5.49, + "learning_rate": 4.729566174989908e-06, + "loss": 0.0077, + "step": 41920 + }, + { + "epoch": 5.49, + "learning_rate": 4.729115491807004e-06, + "loss": 0.0077, + "step": 41921 + }, + { + "epoch": 5.49, + "learning_rate": 4.728664823448181e-06, + "loss": 0.0045, + "step": 41922 + }, + { + "epoch": 5.49, + "learning_rate": 4.728214169914704e-06, + "loss": 0.0083, + "step": 41923 + }, + { + "epoch": 5.49, + "learning_rate": 4.727763531207848e-06, + "loss": 0.0091, + "step": 41924 + }, + { + "epoch": 5.49, + "learning_rate": 4.7273129073288774e-06, + "loss": 0.0077, + "step": 41925 + }, + { + "epoch": 5.49, + "learning_rate": 4.726862298279053e-06, + "loss": 0.006, + "step": 41926 + }, + { + "epoch": 5.49, + "learning_rate": 4.726411704059653e-06, + "loss": 0.006, + "step": 41927 + }, + { + "epoch": 5.49, + "learning_rate": 4.725961124671934e-06, + "loss": 0.0062, + "step": 41928 + }, + { + "epoch": 5.49, + "learning_rate": 4.7255105601171735e-06, + "loss": 0.0111, + "step": 41929 + }, + { + "epoch": 5.49, + "learning_rate": 4.725060010396633e-06, + "loss": 0.0057, + "step": 41930 + }, + { + "epoch": 5.49, + "learning_rate": 4.724609475511577e-06, + "loss": 0.0076, + "step": 41931 + }, + { + "epoch": 5.49, + "learning_rate": 4.724158955463279e-06, + "loss": 0.0076, + "step": 41932 + }, + { + "epoch": 5.49, + "learning_rate": 4.723708450253003e-06, + "loss": 0.0072, + "step": 41933 + }, + { + "epoch": 5.49, + "learning_rate": 4.723257959882013e-06, + "loss": 0.006, + "step": 41934 + }, + { + "epoch": 5.49, + "learning_rate": 4.722807484351582e-06, + "loss": 0.0057, + "step": 41935 + }, + { + "epoch": 5.49, + "learning_rate": 4.722357023662973e-06, + "loss": 0.0045, + "step": 41936 + }, + { + "epoch": 5.49, + "learning_rate": 4.7219065778174555e-06, + "loss": 0.0056, + "step": 41937 + }, + { + "epoch": 5.49, + "learning_rate": 4.72145614681629e-06, + "loss": 0.0088, + "step": 41938 + }, + { + "epoch": 5.49, + "learning_rate": 4.721005730660748e-06, + "loss": 0.0052, + "step": 41939 + }, + { + "epoch": 5.49, + "learning_rate": 4.720555329352101e-06, + "loss": 0.0063, + "step": 41940 + }, + { + "epoch": 5.49, + "learning_rate": 4.72010494289161e-06, + "loss": 0.0066, + "step": 41941 + }, + { + "epoch": 5.49, + "learning_rate": 4.719654571280539e-06, + "loss": 0.0073, + "step": 41942 + }, + { + "epoch": 5.49, + "learning_rate": 4.719204214520164e-06, + "loss": 0.0048, + "step": 41943 + }, + { + "epoch": 5.49, + "learning_rate": 4.718753872611743e-06, + "loss": 0.0062, + "step": 41944 + }, + { + "epoch": 5.49, + "learning_rate": 4.718303545556547e-06, + "loss": 0.0056, + "step": 41945 + }, + { + "epoch": 5.49, + "learning_rate": 4.717853233355837e-06, + "loss": 0.0089, + "step": 41946 + }, + { + "epoch": 5.49, + "learning_rate": 4.717402936010887e-06, + "loss": 0.0072, + "step": 41947 + }, + { + "epoch": 5.49, + "learning_rate": 4.716952653522961e-06, + "loss": 0.0059, + "step": 41948 + }, + { + "epoch": 5.5, + "learning_rate": 4.716502385893319e-06, + "loss": 0.0084, + "step": 41949 + }, + { + "epoch": 5.5, + "learning_rate": 4.7160521331232335e-06, + "loss": 0.0063, + "step": 41950 + }, + { + "epoch": 5.5, + "learning_rate": 4.715601895213974e-06, + "loss": 0.0081, + "step": 41951 + }, + { + "epoch": 5.5, + "learning_rate": 4.715151672166802e-06, + "loss": 0.0054, + "step": 41952 + }, + { + "epoch": 5.5, + "learning_rate": 4.714701463982982e-06, + "loss": 0.0079, + "step": 41953 + }, + { + "epoch": 5.5, + "learning_rate": 4.714251270663785e-06, + "loss": 0.0085, + "step": 41954 + }, + { + "epoch": 5.5, + "learning_rate": 4.713801092210476e-06, + "loss": 0.0061, + "step": 41955 + }, + { + "epoch": 5.5, + "learning_rate": 4.713350928624319e-06, + "loss": 0.0039, + "step": 41956 + }, + { + "epoch": 5.5, + "learning_rate": 4.7129007799065774e-06, + "loss": 0.0089, + "step": 41957 + }, + { + "epoch": 5.5, + "learning_rate": 4.712450646058525e-06, + "loss": 0.0111, + "step": 41958 + }, + { + "epoch": 5.5, + "learning_rate": 4.712000527081423e-06, + "loss": 0.009, + "step": 41959 + }, + { + "epoch": 5.5, + "learning_rate": 4.711550422976538e-06, + "loss": 0.0085, + "step": 41960 + }, + { + "epoch": 5.5, + "learning_rate": 4.711100333745129e-06, + "loss": 0.0076, + "step": 41961 + }, + { + "epoch": 5.5, + "learning_rate": 4.710650259388475e-06, + "loss": 0.0093, + "step": 41962 + }, + { + "epoch": 5.5, + "learning_rate": 4.710200199907837e-06, + "loss": 0.0074, + "step": 41963 + }, + { + "epoch": 5.5, + "learning_rate": 4.709750155304478e-06, + "loss": 0.0071, + "step": 41964 + }, + { + "epoch": 5.5, + "learning_rate": 4.709300125579662e-06, + "loss": 0.0079, + "step": 41965 + }, + { + "epoch": 5.5, + "learning_rate": 4.708850110734661e-06, + "loss": 0.0087, + "step": 41966 + }, + { + "epoch": 5.5, + "learning_rate": 4.708400110770737e-06, + "loss": 0.0064, + "step": 41967 + }, + { + "epoch": 5.5, + "learning_rate": 4.707950125689153e-06, + "loss": 0.0065, + "step": 41968 + }, + { + "epoch": 5.5, + "learning_rate": 4.70750015549118e-06, + "loss": 0.0073, + "step": 41969 + }, + { + "epoch": 5.5, + "learning_rate": 4.707050200178082e-06, + "loss": 0.0079, + "step": 41970 + }, + { + "epoch": 5.5, + "learning_rate": 4.706600259751123e-06, + "loss": 0.0094, + "step": 41971 + }, + { + "epoch": 5.5, + "learning_rate": 4.7061503342115655e-06, + "loss": 0.0108, + "step": 41972 + }, + { + "epoch": 5.5, + "learning_rate": 4.705700423560677e-06, + "loss": 0.0056, + "step": 41973 + }, + { + "epoch": 5.5, + "learning_rate": 4.70525052779973e-06, + "loss": 0.0066, + "step": 41974 + }, + { + "epoch": 5.5, + "learning_rate": 4.704800646929982e-06, + "loss": 0.0089, + "step": 41975 + }, + { + "epoch": 5.5, + "learning_rate": 4.704350780952698e-06, + "loss": 0.0057, + "step": 41976 + }, + { + "epoch": 5.5, + "learning_rate": 4.7039009298691476e-06, + "loss": 0.0063, + "step": 41977 + }, + { + "epoch": 5.5, + "learning_rate": 4.703451093680595e-06, + "loss": 0.0047, + "step": 41978 + }, + { + "epoch": 5.5, + "learning_rate": 4.703001272388304e-06, + "loss": 0.0068, + "step": 41979 + }, + { + "epoch": 5.5, + "learning_rate": 4.702551465993536e-06, + "loss": 0.0071, + "step": 41980 + }, + { + "epoch": 5.5, + "learning_rate": 4.7021016744975634e-06, + "loss": 0.0075, + "step": 41981 + }, + { + "epoch": 5.5, + "learning_rate": 4.701651897901647e-06, + "loss": 0.0059, + "step": 41982 + }, + { + "epoch": 5.5, + "learning_rate": 4.701202136207049e-06, + "loss": 0.0058, + "step": 41983 + }, + { + "epoch": 5.5, + "learning_rate": 4.7007523894150385e-06, + "loss": 0.0054, + "step": 41984 + }, + { + "epoch": 5.5, + "learning_rate": 4.700302657526884e-06, + "loss": 0.0073, + "step": 41985 + }, + { + "epoch": 5.5, + "learning_rate": 4.699852940543846e-06, + "loss": 0.0066, + "step": 41986 + }, + { + "epoch": 5.5, + "learning_rate": 4.699403238467188e-06, + "loss": 0.0073, + "step": 41987 + }, + { + "epoch": 5.5, + "learning_rate": 4.698953551298173e-06, + "loss": 0.0054, + "step": 41988 + }, + { + "epoch": 5.5, + "learning_rate": 4.698503879038072e-06, + "loss": 0.0088, + "step": 41989 + }, + { + "epoch": 5.5, + "learning_rate": 4.698054221688147e-06, + "loss": 0.0069, + "step": 41990 + }, + { + "epoch": 5.5, + "learning_rate": 4.697604579249659e-06, + "loss": 0.0077, + "step": 41991 + }, + { + "epoch": 5.5, + "learning_rate": 4.697154951723879e-06, + "loss": 0.0073, + "step": 41992 + }, + { + "epoch": 5.5, + "learning_rate": 4.696705339112069e-06, + "loss": 0.0065, + "step": 41993 + }, + { + "epoch": 5.5, + "learning_rate": 4.696255741415489e-06, + "loss": 0.0064, + "step": 41994 + }, + { + "epoch": 5.5, + "learning_rate": 4.69580615863541e-06, + "loss": 0.0089, + "step": 41995 + }, + { + "epoch": 5.5, + "learning_rate": 4.695356590773091e-06, + "loss": 0.007, + "step": 41996 + }, + { + "epoch": 5.5, + "learning_rate": 4.6949070378298015e-06, + "loss": 0.0062, + "step": 41997 + }, + { + "epoch": 5.5, + "learning_rate": 4.694457499806804e-06, + "loss": 0.0063, + "step": 41998 + }, + { + "epoch": 5.5, + "learning_rate": 4.6940079767053595e-06, + "loss": 0.0066, + "step": 41999 + }, + { + "epoch": 5.5, + "learning_rate": 4.6935584685267376e-06, + "loss": 0.0083, + "step": 42000 + }, + { + "epoch": 5.5, + "learning_rate": 4.6931089752722e-06, + "loss": 0.0062, + "step": 42001 + }, + { + "epoch": 5.5, + "learning_rate": 4.6926594969430115e-06, + "loss": 0.0079, + "step": 42002 + }, + { + "epoch": 5.5, + "learning_rate": 4.692210033540431e-06, + "loss": 0.0094, + "step": 42003 + }, + { + "epoch": 5.5, + "learning_rate": 4.691760585065731e-06, + "loss": 0.0081, + "step": 42004 + }, + { + "epoch": 5.5, + "learning_rate": 4.691311151520168e-06, + "loss": 0.0069, + "step": 42005 + }, + { + "epoch": 5.5, + "learning_rate": 4.690861732905014e-06, + "loss": 0.0064, + "step": 42006 + }, + { + "epoch": 5.5, + "learning_rate": 4.690412329221524e-06, + "loss": 0.0088, + "step": 42007 + }, + { + "epoch": 5.5, + "learning_rate": 4.689962940470971e-06, + "loss": 0.008, + "step": 42008 + }, + { + "epoch": 5.5, + "learning_rate": 4.689513566654614e-06, + "loss": 0.0088, + "step": 42009 + }, + { + "epoch": 5.5, + "learning_rate": 4.689064207773718e-06, + "loss": 0.0052, + "step": 42010 + }, + { + "epoch": 5.5, + "learning_rate": 4.688614863829542e-06, + "loss": 0.0072, + "step": 42011 + }, + { + "epoch": 5.5, + "learning_rate": 4.6881655348233575e-06, + "loss": 0.006, + "step": 42012 + }, + { + "epoch": 5.5, + "learning_rate": 4.687716220756425e-06, + "loss": 0.0067, + "step": 42013 + }, + { + "epoch": 5.5, + "learning_rate": 4.687266921630003e-06, + "loss": 0.0081, + "step": 42014 + }, + { + "epoch": 5.5, + "learning_rate": 4.686817637445364e-06, + "loss": 0.0066, + "step": 42015 + }, + { + "epoch": 5.5, + "learning_rate": 4.686368368203764e-06, + "loss": 0.0069, + "step": 42016 + }, + { + "epoch": 5.5, + "learning_rate": 4.685919113906474e-06, + "loss": 0.0064, + "step": 42017 + }, + { + "epoch": 5.5, + "learning_rate": 4.68546987455475e-06, + "loss": 0.0064, + "step": 42018 + }, + { + "epoch": 5.5, + "learning_rate": 4.6850206501498615e-06, + "loss": 0.0054, + "step": 42019 + }, + { + "epoch": 5.5, + "learning_rate": 4.68457144069307e-06, + "loss": 0.0086, + "step": 42020 + }, + { + "epoch": 5.5, + "learning_rate": 4.6841222461856385e-06, + "loss": 0.0058, + "step": 42021 + }, + { + "epoch": 5.5, + "learning_rate": 4.6836730666288255e-06, + "loss": 0.0075, + "step": 42022 + }, + { + "epoch": 5.5, + "learning_rate": 4.683223902023903e-06, + "loss": 0.0063, + "step": 42023 + }, + { + "epoch": 5.5, + "learning_rate": 4.682774752372131e-06, + "loss": 0.0084, + "step": 42024 + }, + { + "epoch": 5.5, + "learning_rate": 4.682325617674768e-06, + "loss": 0.0074, + "step": 42025 + }, + { + "epoch": 5.51, + "learning_rate": 4.681876497933084e-06, + "loss": 0.0071, + "step": 42026 + }, + { + "epoch": 5.51, + "learning_rate": 4.681427393148335e-06, + "loss": 0.0065, + "step": 42027 + }, + { + "epoch": 5.51, + "learning_rate": 4.680978303321792e-06, + "loss": 0.0068, + "step": 42028 + }, + { + "epoch": 5.51, + "learning_rate": 4.680529228454715e-06, + "loss": 0.007, + "step": 42029 + }, + { + "epoch": 5.51, + "learning_rate": 4.6800801685483636e-06, + "loss": 0.0077, + "step": 42030 + }, + { + "epoch": 5.51, + "learning_rate": 4.679631123604006e-06, + "loss": 0.0075, + "step": 42031 + }, + { + "epoch": 5.51, + "learning_rate": 4.6791820936229014e-06, + "loss": 0.0058, + "step": 42032 + }, + { + "epoch": 5.51, + "learning_rate": 4.678733078606313e-06, + "loss": 0.008, + "step": 42033 + }, + { + "epoch": 5.51, + "learning_rate": 4.678284078555505e-06, + "loss": 0.0091, + "step": 42034 + }, + { + "epoch": 5.51, + "learning_rate": 4.677835093471741e-06, + "loss": 0.0066, + "step": 42035 + }, + { + "epoch": 5.51, + "learning_rate": 4.6773861233562824e-06, + "loss": 0.0068, + "step": 42036 + }, + { + "epoch": 5.51, + "learning_rate": 4.676937168210388e-06, + "loss": 0.0082, + "step": 42037 + }, + { + "epoch": 5.51, + "learning_rate": 4.676488228035326e-06, + "loss": 0.0101, + "step": 42038 + }, + { + "epoch": 5.51, + "learning_rate": 4.676039302832358e-06, + "loss": 0.0071, + "step": 42039 + }, + { + "epoch": 5.51, + "learning_rate": 4.675590392602748e-06, + "loss": 0.0089, + "step": 42040 + }, + { + "epoch": 5.51, + "learning_rate": 4.6751414973477515e-06, + "loss": 0.0065, + "step": 42041 + }, + { + "epoch": 5.51, + "learning_rate": 4.674692617068641e-06, + "loss": 0.0079, + "step": 42042 + }, + { + "epoch": 5.51, + "learning_rate": 4.6742437517666725e-06, + "loss": 0.008, + "step": 42043 + }, + { + "epoch": 5.51, + "learning_rate": 4.673794901443112e-06, + "loss": 0.007, + "step": 42044 + }, + { + "epoch": 5.51, + "learning_rate": 4.673346066099214e-06, + "loss": 0.0069, + "step": 42045 + }, + { + "epoch": 5.51, + "learning_rate": 4.672897245736251e-06, + "loss": 0.007, + "step": 42046 + }, + { + "epoch": 5.51, + "learning_rate": 4.67244844035548e-06, + "loss": 0.0073, + "step": 42047 + }, + { + "epoch": 5.51, + "learning_rate": 4.671999649958162e-06, + "loss": 0.0043, + "step": 42048 + }, + { + "epoch": 5.51, + "learning_rate": 4.671550874545561e-06, + "loss": 0.0086, + "step": 42049 + }, + { + "epoch": 5.51, + "learning_rate": 4.671102114118942e-06, + "loss": 0.0084, + "step": 42050 + }, + { + "epoch": 5.51, + "learning_rate": 4.670653368679564e-06, + "loss": 0.0103, + "step": 42051 + }, + { + "epoch": 5.51, + "learning_rate": 4.670204638228691e-06, + "loss": 0.0074, + "step": 42052 + }, + { + "epoch": 5.51, + "learning_rate": 4.669755922767579e-06, + "loss": 0.0068, + "step": 42053 + }, + { + "epoch": 5.51, + "learning_rate": 4.6693072222975e-06, + "loss": 0.0077, + "step": 42054 + }, + { + "epoch": 5.51, + "learning_rate": 4.668858536819709e-06, + "loss": 0.0069, + "step": 42055 + }, + { + "epoch": 5.51, + "learning_rate": 4.668409866335466e-06, + "loss": 0.008, + "step": 42056 + }, + { + "epoch": 5.51, + "learning_rate": 4.667961210846041e-06, + "loss": 0.0066, + "step": 42057 + }, + { + "epoch": 5.51, + "learning_rate": 4.667512570352689e-06, + "loss": 0.0088, + "step": 42058 + }, + { + "epoch": 5.51, + "learning_rate": 4.6670639448566724e-06, + "loss": 0.0082, + "step": 42059 + }, + { + "epoch": 5.51, + "learning_rate": 4.666615334359257e-06, + "loss": 0.0082, + "step": 42060 + }, + { + "epoch": 5.51, + "learning_rate": 4.666166738861698e-06, + "loss": 0.0068, + "step": 42061 + }, + { + "epoch": 5.51, + "learning_rate": 4.665718158365265e-06, + "loss": 0.0064, + "step": 42062 + }, + { + "epoch": 5.51, + "learning_rate": 4.6652695928712164e-06, + "loss": 0.0098, + "step": 42063 + }, + { + "epoch": 5.51, + "learning_rate": 4.664821042380808e-06, + "loss": 0.0078, + "step": 42064 + }, + { + "epoch": 5.51, + "learning_rate": 4.664372506895312e-06, + "loss": 0.007, + "step": 42065 + }, + { + "epoch": 5.51, + "learning_rate": 4.663923986415982e-06, + "loss": 0.0069, + "step": 42066 + }, + { + "epoch": 5.51, + "learning_rate": 4.663475480944083e-06, + "loss": 0.0083, + "step": 42067 + }, + { + "epoch": 5.51, + "learning_rate": 4.6630269904808716e-06, + "loss": 0.0067, + "step": 42068 + }, + { + "epoch": 5.51, + "learning_rate": 4.662578515027615e-06, + "loss": 0.0084, + "step": 42069 + }, + { + "epoch": 5.51, + "learning_rate": 4.662130054585569e-06, + "loss": 0.0056, + "step": 42070 + }, + { + "epoch": 5.51, + "learning_rate": 4.661681609156002e-06, + "loss": 0.0097, + "step": 42071 + }, + { + "epoch": 5.51, + "learning_rate": 4.661233178740166e-06, + "loss": 0.0057, + "step": 42072 + }, + { + "epoch": 5.51, + "learning_rate": 4.660784763339332e-06, + "loss": 0.0094, + "step": 42073 + }, + { + "epoch": 5.51, + "learning_rate": 4.660336362954757e-06, + "loss": 0.0072, + "step": 42074 + }, + { + "epoch": 5.51, + "learning_rate": 4.6598879775876985e-06, + "loss": 0.0074, + "step": 42075 + }, + { + "epoch": 5.51, + "learning_rate": 4.659439607239423e-06, + "loss": 0.005, + "step": 42076 + }, + { + "epoch": 5.51, + "learning_rate": 4.65899125191119e-06, + "loss": 0.006, + "step": 42077 + }, + { + "epoch": 5.51, + "learning_rate": 4.658542911604259e-06, + "loss": 0.0086, + "step": 42078 + }, + { + "epoch": 5.51, + "learning_rate": 4.658094586319888e-06, + "loss": 0.0087, + "step": 42079 + }, + { + "epoch": 5.51, + "learning_rate": 4.6576462760593455e-06, + "loss": 0.0058, + "step": 42080 + }, + { + "epoch": 5.51, + "learning_rate": 4.657197980823885e-06, + "loss": 0.0067, + "step": 42081 + }, + { + "epoch": 5.51, + "learning_rate": 4.656749700614774e-06, + "loss": 0.0067, + "step": 42082 + }, + { + "epoch": 5.51, + "learning_rate": 4.656301435433266e-06, + "loss": 0.007, + "step": 42083 + }, + { + "epoch": 5.51, + "learning_rate": 4.65585318528063e-06, + "loss": 0.0076, + "step": 42084 + }, + { + "epoch": 5.51, + "learning_rate": 4.655404950158122e-06, + "loss": 0.0075, + "step": 42085 + }, + { + "epoch": 5.51, + "learning_rate": 4.654956730067003e-06, + "loss": 0.0086, + "step": 42086 + }, + { + "epoch": 5.51, + "learning_rate": 4.65450852500853e-06, + "loss": 0.0055, + "step": 42087 + }, + { + "epoch": 5.51, + "learning_rate": 4.65406033498397e-06, + "loss": 0.0076, + "step": 42088 + }, + { + "epoch": 5.51, + "learning_rate": 4.653612159994581e-06, + "loss": 0.0075, + "step": 42089 + }, + { + "epoch": 5.51, + "learning_rate": 4.65316400004162e-06, + "loss": 0.0055, + "step": 42090 + }, + { + "epoch": 5.51, + "learning_rate": 4.652715855126353e-06, + "loss": 0.0079, + "step": 42091 + }, + { + "epoch": 5.51, + "learning_rate": 4.652267725250036e-06, + "loss": 0.0061, + "step": 42092 + }, + { + "epoch": 5.51, + "learning_rate": 4.651819610413934e-06, + "loss": 0.0077, + "step": 42093 + }, + { + "epoch": 5.51, + "learning_rate": 4.651371510619303e-06, + "loss": 0.0063, + "step": 42094 + }, + { + "epoch": 5.51, + "learning_rate": 4.650923425867403e-06, + "loss": 0.0072, + "step": 42095 + }, + { + "epoch": 5.51, + "learning_rate": 4.6504753561594994e-06, + "loss": 0.0082, + "step": 42096 + }, + { + "epoch": 5.51, + "learning_rate": 4.6500273014968485e-06, + "loss": 0.006, + "step": 42097 + }, + { + "epoch": 5.51, + "learning_rate": 4.649579261880708e-06, + "loss": 0.0058, + "step": 42098 + }, + { + "epoch": 5.51, + "learning_rate": 4.649131237312344e-06, + "loss": 0.0069, + "step": 42099 + }, + { + "epoch": 5.51, + "learning_rate": 4.648683227793014e-06, + "loss": 0.0066, + "step": 42100 + }, + { + "epoch": 5.51, + "learning_rate": 4.648235233323977e-06, + "loss": 0.0055, + "step": 42101 + }, + { + "epoch": 5.52, + "learning_rate": 4.64778725390649e-06, + "loss": 0.0073, + "step": 42102 + }, + { + "epoch": 5.52, + "learning_rate": 4.647339289541816e-06, + "loss": 0.0047, + "step": 42103 + }, + { + "epoch": 5.52, + "learning_rate": 4.646891340231219e-06, + "loss": 0.0074, + "step": 42104 + }, + { + "epoch": 5.52, + "learning_rate": 4.646443405975955e-06, + "loss": 0.0088, + "step": 42105 + }, + { + "epoch": 5.52, + "learning_rate": 4.64599548677728e-06, + "loss": 0.0076, + "step": 42106 + }, + { + "epoch": 5.52, + "learning_rate": 4.64554758263646e-06, + "loss": 0.0065, + "step": 42107 + }, + { + "epoch": 5.52, + "learning_rate": 4.645099693554754e-06, + "loss": 0.0069, + "step": 42108 + }, + { + "epoch": 5.52, + "learning_rate": 4.64465181953342e-06, + "loss": 0.0083, + "step": 42109 + }, + { + "epoch": 5.52, + "learning_rate": 4.6442039605737135e-06, + "loss": 0.0093, + "step": 42110 + }, + { + "epoch": 5.52, + "learning_rate": 4.643756116676902e-06, + "loss": 0.0056, + "step": 42111 + }, + { + "epoch": 5.52, + "learning_rate": 4.643308287844241e-06, + "loss": 0.0066, + "step": 42112 + }, + { + "epoch": 5.52, + "learning_rate": 4.6428604740769864e-06, + "loss": 0.0099, + "step": 42113 + }, + { + "epoch": 5.52, + "learning_rate": 4.642412675376402e-06, + "loss": 0.0089, + "step": 42114 + }, + { + "epoch": 5.52, + "learning_rate": 4.6419648917437496e-06, + "loss": 0.0081, + "step": 42115 + }, + { + "epoch": 5.52, + "learning_rate": 4.641517123180286e-06, + "loss": 0.0054, + "step": 42116 + }, + { + "epoch": 5.52, + "learning_rate": 4.64106936968727e-06, + "loss": 0.0082, + "step": 42117 + }, + { + "epoch": 5.52, + "learning_rate": 4.640621631265957e-06, + "loss": 0.0052, + "step": 42118 + }, + { + "epoch": 5.52, + "learning_rate": 4.640173907917614e-06, + "loss": 0.0055, + "step": 42119 + }, + { + "epoch": 5.52, + "learning_rate": 4.639726199643497e-06, + "loss": 0.0091, + "step": 42120 + }, + { + "epoch": 5.52, + "learning_rate": 4.639278506444861e-06, + "loss": 0.0099, + "step": 42121 + }, + { + "epoch": 5.52, + "learning_rate": 4.638830828322972e-06, + "loss": 0.0075, + "step": 42122 + }, + { + "epoch": 5.52, + "learning_rate": 4.638383165279086e-06, + "loss": 0.0065, + "step": 42123 + }, + { + "epoch": 5.52, + "learning_rate": 4.637935517314461e-06, + "loss": 0.0063, + "step": 42124 + }, + { + "epoch": 5.52, + "learning_rate": 4.637487884430351e-06, + "loss": 0.0049, + "step": 42125 + }, + { + "epoch": 5.52, + "learning_rate": 4.637040266628027e-06, + "loss": 0.0079, + "step": 42126 + }, + { + "epoch": 5.52, + "learning_rate": 4.636592663908742e-06, + "loss": 0.0063, + "step": 42127 + }, + { + "epoch": 5.52, + "learning_rate": 4.636145076273755e-06, + "loss": 0.0063, + "step": 42128 + }, + { + "epoch": 5.52, + "learning_rate": 4.63569750372432e-06, + "loss": 0.0063, + "step": 42129 + }, + { + "epoch": 5.52, + "learning_rate": 4.635249946261704e-06, + "loss": 0.0061, + "step": 42130 + }, + { + "epoch": 5.52, + "learning_rate": 4.634802403887161e-06, + "loss": 0.0109, + "step": 42131 + }, + { + "epoch": 5.52, + "learning_rate": 4.634354876601952e-06, + "loss": 0.0084, + "step": 42132 + }, + { + "epoch": 5.52, + "learning_rate": 4.6339073644073295e-06, + "loss": 0.0087, + "step": 42133 + }, + { + "epoch": 5.52, + "learning_rate": 4.633459867304561e-06, + "loss": 0.0068, + "step": 42134 + }, + { + "epoch": 5.52, + "learning_rate": 4.6330123852949e-06, + "loss": 0.0073, + "step": 42135 + }, + { + "epoch": 5.52, + "learning_rate": 4.632564918379603e-06, + "loss": 0.0084, + "step": 42136 + }, + { + "epoch": 5.52, + "learning_rate": 4.632117466559931e-06, + "loss": 0.0043, + "step": 42137 + }, + { + "epoch": 5.52, + "learning_rate": 4.631670029837146e-06, + "loss": 0.0064, + "step": 42138 + }, + { + "epoch": 5.52, + "learning_rate": 4.631222608212503e-06, + "loss": 0.005, + "step": 42139 + }, + { + "epoch": 5.52, + "learning_rate": 4.630775201687259e-06, + "loss": 0.0079, + "step": 42140 + }, + { + "epoch": 5.52, + "learning_rate": 4.630327810262676e-06, + "loss": 0.0082, + "step": 42141 + }, + { + "epoch": 5.52, + "learning_rate": 4.629880433940009e-06, + "loss": 0.0099, + "step": 42142 + }, + { + "epoch": 5.52, + "learning_rate": 4.629433072720518e-06, + "loss": 0.009, + "step": 42143 + }, + { + "epoch": 5.52, + "learning_rate": 4.6289857266054585e-06, + "loss": 0.0092, + "step": 42144 + }, + { + "epoch": 5.52, + "learning_rate": 4.628538395596093e-06, + "loss": 0.007, + "step": 42145 + }, + { + "epoch": 5.52, + "learning_rate": 4.628091079693676e-06, + "loss": 0.0057, + "step": 42146 + }, + { + "epoch": 5.52, + "learning_rate": 4.627643778899464e-06, + "loss": 0.0063, + "step": 42147 + }, + { + "epoch": 5.52, + "learning_rate": 4.627196493214718e-06, + "loss": 0.0057, + "step": 42148 + }, + { + "epoch": 5.52, + "learning_rate": 4.626749222640701e-06, + "loss": 0.0083, + "step": 42149 + }, + { + "epoch": 5.52, + "learning_rate": 4.626301967178664e-06, + "loss": 0.0058, + "step": 42150 + }, + { + "epoch": 5.52, + "learning_rate": 4.625854726829866e-06, + "loss": 0.0071, + "step": 42151 + }, + { + "epoch": 5.52, + "learning_rate": 4.625407501595562e-06, + "loss": 0.006, + "step": 42152 + }, + { + "epoch": 5.52, + "learning_rate": 4.6249602914770174e-06, + "loss": 0.0066, + "step": 42153 + }, + { + "epoch": 5.52, + "learning_rate": 4.624513096475486e-06, + "loss": 0.007, + "step": 42154 + }, + { + "epoch": 5.52, + "learning_rate": 4.624065916592221e-06, + "loss": 0.0065, + "step": 42155 + }, + { + "epoch": 5.52, + "learning_rate": 4.623618751828488e-06, + "loss": 0.0064, + "step": 42156 + }, + { + "epoch": 5.52, + "learning_rate": 4.623171602185542e-06, + "loss": 0.0058, + "step": 42157 + }, + { + "epoch": 5.52, + "learning_rate": 4.622724467664634e-06, + "loss": 0.009, + "step": 42158 + }, + { + "epoch": 5.52, + "learning_rate": 4.622277348267032e-06, + "loss": 0.0065, + "step": 42159 + }, + { + "epoch": 5.52, + "learning_rate": 4.621830243993986e-06, + "loss": 0.0089, + "step": 42160 + }, + { + "epoch": 5.52, + "learning_rate": 4.621383154846757e-06, + "loss": 0.0043, + "step": 42161 + }, + { + "epoch": 5.52, + "learning_rate": 4.620936080826604e-06, + "loss": 0.0092, + "step": 42162 + }, + { + "epoch": 5.52, + "learning_rate": 4.620489021934776e-06, + "loss": 0.0077, + "step": 42163 + }, + { + "epoch": 5.52, + "learning_rate": 4.620041978172542e-06, + "loss": 0.0077, + "step": 42164 + }, + { + "epoch": 5.52, + "learning_rate": 4.619594949541151e-06, + "loss": 0.0048, + "step": 42165 + }, + { + "epoch": 5.52, + "learning_rate": 4.619147936041864e-06, + "loss": 0.0064, + "step": 42166 + }, + { + "epoch": 5.52, + "learning_rate": 4.618700937675933e-06, + "loss": 0.0078, + "step": 42167 + }, + { + "epoch": 5.52, + "learning_rate": 4.618253954444623e-06, + "loss": 0.0063, + "step": 42168 + }, + { + "epoch": 5.52, + "learning_rate": 4.617806986349184e-06, + "loss": 0.0055, + "step": 42169 + }, + { + "epoch": 5.52, + "learning_rate": 4.617360033390879e-06, + "loss": 0.0068, + "step": 42170 + }, + { + "epoch": 5.52, + "learning_rate": 4.616913095570958e-06, + "loss": 0.006, + "step": 42171 + }, + { + "epoch": 5.52, + "learning_rate": 4.616466172890687e-06, + "loss": 0.0053, + "step": 42172 + }, + { + "epoch": 5.52, + "learning_rate": 4.616019265351318e-06, + "loss": 0.0076, + "step": 42173 + }, + { + "epoch": 5.52, + "learning_rate": 4.615572372954108e-06, + "loss": 0.0058, + "step": 42174 + }, + { + "epoch": 5.52, + "learning_rate": 4.61512549570031e-06, + "loss": 0.0073, + "step": 42175 + }, + { + "epoch": 5.52, + "learning_rate": 4.61467863359119e-06, + "loss": 0.006, + "step": 42176 + }, + { + "epoch": 5.52, + "learning_rate": 4.6142317866279975e-06, + "loss": 0.0124, + "step": 42177 + }, + { + "epoch": 5.53, + "learning_rate": 4.613784954811989e-06, + "loss": 0.0061, + "step": 42178 + }, + { + "epoch": 5.53, + "learning_rate": 4.613338138144426e-06, + "loss": 0.0064, + "step": 42179 + }, + { + "epoch": 5.53, + "learning_rate": 4.612891336626559e-06, + "loss": 0.008, + "step": 42180 + }, + { + "epoch": 5.53, + "learning_rate": 4.612444550259654e-06, + "loss": 0.003, + "step": 42181 + }, + { + "epoch": 5.53, + "learning_rate": 4.61199777904496e-06, + "loss": 0.0068, + "step": 42182 + }, + { + "epoch": 5.53, + "learning_rate": 4.611551022983731e-06, + "loss": 0.0066, + "step": 42183 + }, + { + "epoch": 5.53, + "learning_rate": 4.611104282077232e-06, + "loss": 0.0082, + "step": 42184 + }, + { + "epoch": 5.53, + "learning_rate": 4.610657556326716e-06, + "loss": 0.0062, + "step": 42185 + }, + { + "epoch": 5.53, + "learning_rate": 4.610210845733435e-06, + "loss": 0.0053, + "step": 42186 + }, + { + "epoch": 5.53, + "learning_rate": 4.609764150298652e-06, + "loss": 0.007, + "step": 42187 + }, + { + "epoch": 5.53, + "learning_rate": 4.609317470023621e-06, + "loss": 0.0071, + "step": 42188 + }, + { + "epoch": 5.53, + "learning_rate": 4.608870804909596e-06, + "loss": 0.0071, + "step": 42189 + }, + { + "epoch": 5.53, + "learning_rate": 4.608424154957829e-06, + "loss": 0.0095, + "step": 42190 + }, + { + "epoch": 5.53, + "learning_rate": 4.607977520169589e-06, + "loss": 0.0073, + "step": 42191 + }, + { + "epoch": 5.53, + "learning_rate": 4.607530900546125e-06, + "loss": 0.0067, + "step": 42192 + }, + { + "epoch": 5.53, + "learning_rate": 4.607084296088693e-06, + "loss": 0.0091, + "step": 42193 + }, + { + "epoch": 5.53, + "learning_rate": 4.606637706798546e-06, + "loss": 0.0046, + "step": 42194 + }, + { + "epoch": 5.53, + "learning_rate": 4.606191132676946e-06, + "loss": 0.0074, + "step": 42195 + }, + { + "epoch": 5.53, + "learning_rate": 4.605744573725146e-06, + "loss": 0.0073, + "step": 42196 + }, + { + "epoch": 5.53, + "learning_rate": 4.605298029944404e-06, + "loss": 0.0076, + "step": 42197 + }, + { + "epoch": 5.53, + "learning_rate": 4.604851501335969e-06, + "loss": 0.0059, + "step": 42198 + }, + { + "epoch": 5.53, + "learning_rate": 4.604404987901106e-06, + "loss": 0.0069, + "step": 42199 + }, + { + "epoch": 5.53, + "learning_rate": 4.603958489641066e-06, + "loss": 0.0072, + "step": 42200 + }, + { + "epoch": 5.53, + "learning_rate": 4.603512006557102e-06, + "loss": 0.0069, + "step": 42201 + }, + { + "epoch": 5.53, + "learning_rate": 4.603065538650473e-06, + "loss": 0.0079, + "step": 42202 + }, + { + "epoch": 5.53, + "learning_rate": 4.602619085922439e-06, + "loss": 0.0084, + "step": 42203 + }, + { + "epoch": 5.53, + "learning_rate": 4.60217264837425e-06, + "loss": 0.0067, + "step": 42204 + }, + { + "epoch": 5.53, + "learning_rate": 4.60172622600716e-06, + "loss": 0.0052, + "step": 42205 + }, + { + "epoch": 5.53, + "learning_rate": 4.6012798188224315e-06, + "loss": 0.0078, + "step": 42206 + }, + { + "epoch": 5.53, + "learning_rate": 4.600833426821316e-06, + "loss": 0.0071, + "step": 42207 + }, + { + "epoch": 5.53, + "learning_rate": 4.60038705000507e-06, + "loss": 0.0076, + "step": 42208 + }, + { + "epoch": 5.53, + "learning_rate": 4.599940688374942e-06, + "loss": 0.0109, + "step": 42209 + }, + { + "epoch": 5.53, + "learning_rate": 4.599494341932198e-06, + "loss": 0.0054, + "step": 42210 + }, + { + "epoch": 5.53, + "learning_rate": 4.599048010678089e-06, + "loss": 0.0068, + "step": 42211 + }, + { + "epoch": 5.53, + "learning_rate": 4.598601694613866e-06, + "loss": 0.0053, + "step": 42212 + }, + { + "epoch": 5.53, + "learning_rate": 4.598155393740788e-06, + "loss": 0.0053, + "step": 42213 + }, + { + "epoch": 5.53, + "learning_rate": 4.5977091080601136e-06, + "loss": 0.0091, + "step": 42214 + }, + { + "epoch": 5.53, + "learning_rate": 4.597262837573095e-06, + "loss": 0.009, + "step": 42215 + }, + { + "epoch": 5.53, + "learning_rate": 4.596816582280986e-06, + "loss": 0.0058, + "step": 42216 + }, + { + "epoch": 5.53, + "learning_rate": 4.59637034218504e-06, + "loss": 0.0062, + "step": 42217 + }, + { + "epoch": 5.53, + "learning_rate": 4.595924117286519e-06, + "loss": 0.0083, + "step": 42218 + }, + { + "epoch": 5.53, + "learning_rate": 4.595477907586672e-06, + "loss": 0.008, + "step": 42219 + }, + { + "epoch": 5.53, + "learning_rate": 4.5950317130867525e-06, + "loss": 0.0053, + "step": 42220 + }, + { + "epoch": 5.53, + "learning_rate": 4.594585533788023e-06, + "loss": 0.0054, + "step": 42221 + }, + { + "epoch": 5.53, + "learning_rate": 4.594139369691733e-06, + "loss": 0.008, + "step": 42222 + }, + { + "epoch": 5.53, + "learning_rate": 4.593693220799135e-06, + "loss": 0.0069, + "step": 42223 + }, + { + "epoch": 5.53, + "learning_rate": 4.59324708711149e-06, + "loss": 0.0066, + "step": 42224 + }, + { + "epoch": 5.53, + "learning_rate": 4.5928009686300456e-06, + "loss": 0.0064, + "step": 42225 + }, + { + "epoch": 5.53, + "learning_rate": 4.592354865356064e-06, + "loss": 0.0058, + "step": 42226 + }, + { + "epoch": 5.53, + "learning_rate": 4.591908777290797e-06, + "loss": 0.0068, + "step": 42227 + }, + { + "epoch": 5.53, + "learning_rate": 4.591462704435496e-06, + "loss": 0.007, + "step": 42228 + }, + { + "epoch": 5.53, + "learning_rate": 4.591016646791421e-06, + "loss": 0.0099, + "step": 42229 + }, + { + "epoch": 5.53, + "learning_rate": 4.590570604359823e-06, + "loss": 0.0053, + "step": 42230 + }, + { + "epoch": 5.53, + "learning_rate": 4.590124577141958e-06, + "loss": 0.0054, + "step": 42231 + }, + { + "epoch": 5.53, + "learning_rate": 4.5896785651390755e-06, + "loss": 0.008, + "step": 42232 + }, + { + "epoch": 5.53, + "learning_rate": 4.589232568352437e-06, + "loss": 0.0082, + "step": 42233 + }, + { + "epoch": 5.53, + "learning_rate": 4.5887865867832914e-06, + "loss": 0.0068, + "step": 42234 + }, + { + "epoch": 5.53, + "learning_rate": 4.588340620432898e-06, + "loss": 0.0072, + "step": 42235 + }, + { + "epoch": 5.53, + "learning_rate": 4.587894669302505e-06, + "loss": 0.0074, + "step": 42236 + }, + { + "epoch": 5.53, + "learning_rate": 4.587448733393374e-06, + "loss": 0.0061, + "step": 42237 + }, + { + "epoch": 5.53, + "learning_rate": 4.587002812706755e-06, + "loss": 0.0091, + "step": 42238 + }, + { + "epoch": 5.53, + "learning_rate": 4.586556907243902e-06, + "loss": 0.0079, + "step": 42239 + }, + { + "epoch": 5.53, + "learning_rate": 4.586111017006065e-06, + "loss": 0.0072, + "step": 42240 + }, + { + "epoch": 5.53, + "learning_rate": 4.585665141994508e-06, + "loss": 0.0065, + "step": 42241 + }, + { + "epoch": 5.53, + "learning_rate": 4.585219282210478e-06, + "loss": 0.0095, + "step": 42242 + }, + { + "epoch": 5.53, + "learning_rate": 4.584773437655228e-06, + "loss": 0.0058, + "step": 42243 + }, + { + "epoch": 5.53, + "learning_rate": 4.584327608330017e-06, + "loss": 0.0063, + "step": 42244 + }, + { + "epoch": 5.53, + "learning_rate": 4.583881794236093e-06, + "loss": 0.0055, + "step": 42245 + }, + { + "epoch": 5.53, + "learning_rate": 4.583435995374717e-06, + "loss": 0.0051, + "step": 42246 + }, + { + "epoch": 5.53, + "learning_rate": 4.582990211747138e-06, + "loss": 0.0067, + "step": 42247 + }, + { + "epoch": 5.53, + "learning_rate": 4.582544443354606e-06, + "loss": 0.0069, + "step": 42248 + }, + { + "epoch": 5.53, + "learning_rate": 4.582098690198384e-06, + "loss": 0.0054, + "step": 42249 + }, + { + "epoch": 5.53, + "learning_rate": 4.581652952279722e-06, + "loss": 0.0099, + "step": 42250 + }, + { + "epoch": 5.53, + "learning_rate": 4.581207229599868e-06, + "loss": 0.0061, + "step": 42251 + }, + { + "epoch": 5.53, + "learning_rate": 4.5807615221600835e-06, + "loss": 0.0079, + "step": 42252 + }, + { + "epoch": 5.53, + "learning_rate": 4.580315829961619e-06, + "loss": 0.0062, + "step": 42253 + }, + { + "epoch": 5.53, + "learning_rate": 4.579870153005728e-06, + "loss": 0.0069, + "step": 42254 + }, + { + "epoch": 5.54, + "learning_rate": 4.579424491293659e-06, + "loss": 0.0067, + "step": 42255 + }, + { + "epoch": 5.54, + "learning_rate": 4.578978844826671e-06, + "loss": 0.0075, + "step": 42256 + }, + { + "epoch": 5.54, + "learning_rate": 4.578533213606018e-06, + "loss": 0.0075, + "step": 42257 + }, + { + "epoch": 5.54, + "learning_rate": 4.578087597632953e-06, + "loss": 0.0066, + "step": 42258 + }, + { + "epoch": 5.54, + "learning_rate": 4.5776419969087235e-06, + "loss": 0.0074, + "step": 42259 + }, + { + "epoch": 5.54, + "learning_rate": 4.577196411434592e-06, + "loss": 0.006, + "step": 42260 + }, + { + "epoch": 5.54, + "learning_rate": 4.576750841211806e-06, + "loss": 0.0094, + "step": 42261 + }, + { + "epoch": 5.54, + "learning_rate": 4.57630528624162e-06, + "loss": 0.0061, + "step": 42262 + }, + { + "epoch": 5.54, + "learning_rate": 4.575859746525283e-06, + "loss": 0.0079, + "step": 42263 + }, + { + "epoch": 5.54, + "learning_rate": 4.575414222064054e-06, + "loss": 0.0051, + "step": 42264 + }, + { + "epoch": 5.54, + "learning_rate": 4.574968712859185e-06, + "loss": 0.0082, + "step": 42265 + }, + { + "epoch": 5.54, + "learning_rate": 4.574523218911923e-06, + "loss": 0.0067, + "step": 42266 + }, + { + "epoch": 5.54, + "learning_rate": 4.574077740223526e-06, + "loss": 0.0073, + "step": 42267 + }, + { + "epoch": 5.54, + "learning_rate": 4.57363227679525e-06, + "loss": 0.0072, + "step": 42268 + }, + { + "epoch": 5.54, + "learning_rate": 4.573186828628344e-06, + "loss": 0.0068, + "step": 42269 + }, + { + "epoch": 5.54, + "learning_rate": 4.572741395724057e-06, + "loss": 0.0062, + "step": 42270 + }, + { + "epoch": 5.54, + "learning_rate": 4.57229597808365e-06, + "loss": 0.0067, + "step": 42271 + }, + { + "epoch": 5.54, + "learning_rate": 4.571850575708373e-06, + "loss": 0.008, + "step": 42272 + }, + { + "epoch": 5.54, + "learning_rate": 4.571405188599476e-06, + "loss": 0.0062, + "step": 42273 + }, + { + "epoch": 5.54, + "learning_rate": 4.570959816758208e-06, + "loss": 0.0079, + "step": 42274 + }, + { + "epoch": 5.54, + "learning_rate": 4.570514460185831e-06, + "loss": 0.0056, + "step": 42275 + }, + { + "epoch": 5.54, + "learning_rate": 4.570069118883593e-06, + "loss": 0.009, + "step": 42276 + }, + { + "epoch": 5.54, + "learning_rate": 4.569623792852746e-06, + "loss": 0.0079, + "step": 42277 + }, + { + "epoch": 5.54, + "learning_rate": 4.569178482094536e-06, + "loss": 0.0069, + "step": 42278 + }, + { + "epoch": 5.54, + "learning_rate": 4.56873318661023e-06, + "loss": 0.0081, + "step": 42279 + }, + { + "epoch": 5.54, + "learning_rate": 4.568287906401072e-06, + "loss": 0.0045, + "step": 42280 + }, + { + "epoch": 5.54, + "learning_rate": 4.567842641468317e-06, + "loss": 0.0059, + "step": 42281 + }, + { + "epoch": 5.54, + "learning_rate": 4.5673973918132084e-06, + "loss": 0.0056, + "step": 42282 + }, + { + "epoch": 5.54, + "learning_rate": 4.566952157437011e-06, + "loss": 0.0087, + "step": 42283 + }, + { + "epoch": 5.54, + "learning_rate": 4.566506938340971e-06, + "loss": 0.0086, + "step": 42284 + }, + { + "epoch": 5.54, + "learning_rate": 4.566061734526337e-06, + "loss": 0.007, + "step": 42285 + }, + { + "epoch": 5.54, + "learning_rate": 4.565616545994369e-06, + "loss": 0.005, + "step": 42286 + }, + { + "epoch": 5.54, + "learning_rate": 4.565171372746315e-06, + "loss": 0.0089, + "step": 42287 + }, + { + "epoch": 5.54, + "learning_rate": 4.564726214783426e-06, + "loss": 0.006, + "step": 42288 + }, + { + "epoch": 5.54, + "learning_rate": 4.564281072106953e-06, + "loss": 0.0081, + "step": 42289 + }, + { + "epoch": 5.54, + "learning_rate": 4.5638359447181495e-06, + "loss": 0.0086, + "step": 42290 + }, + { + "epoch": 5.54, + "learning_rate": 4.5633908326182705e-06, + "loss": 0.0069, + "step": 42291 + }, + { + "epoch": 5.54, + "learning_rate": 4.562945735808566e-06, + "loss": 0.0048, + "step": 42292 + }, + { + "epoch": 5.54, + "learning_rate": 4.562500654290284e-06, + "loss": 0.0056, + "step": 42293 + }, + { + "epoch": 5.54, + "learning_rate": 4.562055588064682e-06, + "loss": 0.0068, + "step": 42294 + }, + { + "epoch": 5.54, + "learning_rate": 4.56161053713301e-06, + "loss": 0.0047, + "step": 42295 + }, + { + "epoch": 5.54, + "learning_rate": 4.561165501496519e-06, + "loss": 0.0074, + "step": 42296 + }, + { + "epoch": 5.54, + "learning_rate": 4.560720481156456e-06, + "loss": 0.0074, + "step": 42297 + }, + { + "epoch": 5.54, + "learning_rate": 4.560275476114081e-06, + "loss": 0.0078, + "step": 42298 + }, + { + "epoch": 5.54, + "learning_rate": 4.559830486370641e-06, + "loss": 0.0051, + "step": 42299 + }, + { + "epoch": 5.54, + "learning_rate": 4.559385511927385e-06, + "loss": 0.008, + "step": 42300 + }, + { + "epoch": 5.54, + "learning_rate": 4.5589405527855665e-06, + "loss": 0.0077, + "step": 42301 + }, + { + "epoch": 5.54, + "learning_rate": 4.558495608946443e-06, + "loss": 0.0062, + "step": 42302 + }, + { + "epoch": 5.54, + "learning_rate": 4.55805068041126e-06, + "loss": 0.0051, + "step": 42303 + }, + { + "epoch": 5.54, + "learning_rate": 4.55760576718127e-06, + "loss": 0.0067, + "step": 42304 + }, + { + "epoch": 5.54, + "learning_rate": 4.5571608692577195e-06, + "loss": 0.0061, + "step": 42305 + }, + { + "epoch": 5.54, + "learning_rate": 4.556715986641869e-06, + "loss": 0.0043, + "step": 42306 + }, + { + "epoch": 5.54, + "learning_rate": 4.556271119334964e-06, + "loss": 0.0089, + "step": 42307 + }, + { + "epoch": 5.54, + "learning_rate": 4.5558262673382535e-06, + "loss": 0.0075, + "step": 42308 + }, + { + "epoch": 5.54, + "learning_rate": 4.555381430652996e-06, + "loss": 0.0058, + "step": 42309 + }, + { + "epoch": 5.54, + "learning_rate": 4.554936609280437e-06, + "loss": 0.0085, + "step": 42310 + }, + { + "epoch": 5.54, + "learning_rate": 4.5544918032218254e-06, + "loss": 0.0098, + "step": 42311 + }, + { + "epoch": 5.54, + "learning_rate": 4.55404701247842e-06, + "loss": 0.0066, + "step": 42312 + }, + { + "epoch": 5.54, + "learning_rate": 4.5536022370514635e-06, + "loss": 0.0086, + "step": 42313 + }, + { + "epoch": 5.54, + "learning_rate": 4.553157476942213e-06, + "loss": 0.0079, + "step": 42314 + }, + { + "epoch": 5.54, + "learning_rate": 4.552712732151918e-06, + "loss": 0.007, + "step": 42315 + }, + { + "epoch": 5.54, + "learning_rate": 4.552268002681825e-06, + "loss": 0.0055, + "step": 42316 + }, + { + "epoch": 5.54, + "learning_rate": 4.551823288533191e-06, + "loss": 0.0089, + "step": 42317 + }, + { + "epoch": 5.54, + "learning_rate": 4.551378589707265e-06, + "loss": 0.0078, + "step": 42318 + }, + { + "epoch": 5.54, + "learning_rate": 4.550933906205296e-06, + "loss": 0.0068, + "step": 42319 + }, + { + "epoch": 5.54, + "learning_rate": 4.5504892380285315e-06, + "loss": 0.0053, + "step": 42320 + }, + { + "epoch": 5.54, + "learning_rate": 4.550044585178226e-06, + "loss": 0.0061, + "step": 42321 + }, + { + "epoch": 5.54, + "learning_rate": 4.549599947655633e-06, + "loss": 0.0093, + "step": 42322 + }, + { + "epoch": 5.54, + "learning_rate": 4.549155325462001e-06, + "loss": 0.0077, + "step": 42323 + }, + { + "epoch": 5.54, + "learning_rate": 4.548710718598575e-06, + "loss": 0.0069, + "step": 42324 + }, + { + "epoch": 5.54, + "learning_rate": 4.548266127066613e-06, + "loss": 0.0072, + "step": 42325 + }, + { + "epoch": 5.54, + "learning_rate": 4.547821550867363e-06, + "loss": 0.0068, + "step": 42326 + }, + { + "epoch": 5.54, + "learning_rate": 4.547376990002074e-06, + "loss": 0.007, + "step": 42327 + }, + { + "epoch": 5.54, + "learning_rate": 4.546932444471993e-06, + "loss": 0.0052, + "step": 42328 + }, + { + "epoch": 5.54, + "learning_rate": 4.5464879142783795e-06, + "loss": 0.006, + "step": 42329 + }, + { + "epoch": 5.54, + "learning_rate": 4.546043399422476e-06, + "loss": 0.006, + "step": 42330 + }, + { + "epoch": 5.55, + "learning_rate": 4.545598899905532e-06, + "loss": 0.0115, + "step": 42331 + }, + { + "epoch": 5.55, + "learning_rate": 4.545154415728802e-06, + "loss": 0.0065, + "step": 42332 + }, + { + "epoch": 5.55, + "learning_rate": 4.544709946893536e-06, + "loss": 0.0063, + "step": 42333 + }, + { + "epoch": 5.55, + "learning_rate": 4.5442654934009845e-06, + "loss": 0.007, + "step": 42334 + }, + { + "epoch": 5.55, + "learning_rate": 4.543821055252392e-06, + "loss": 0.0057, + "step": 42335 + }, + { + "epoch": 5.55, + "learning_rate": 4.543376632449016e-06, + "loss": 0.0088, + "step": 42336 + }, + { + "epoch": 5.55, + "learning_rate": 4.542932224992102e-06, + "loss": 0.0051, + "step": 42337 + }, + { + "epoch": 5.55, + "learning_rate": 4.542487832882901e-06, + "loss": 0.0047, + "step": 42338 + }, + { + "epoch": 5.55, + "learning_rate": 4.542043456122658e-06, + "loss": 0.0103, + "step": 42339 + }, + { + "epoch": 5.55, + "learning_rate": 4.541599094712631e-06, + "loss": 0.0074, + "step": 42340 + }, + { + "epoch": 5.55, + "learning_rate": 4.5411547486540666e-06, + "loss": 0.0083, + "step": 42341 + }, + { + "epoch": 5.55, + "learning_rate": 4.54071041794821e-06, + "loss": 0.0106, + "step": 42342 + }, + { + "epoch": 5.55, + "learning_rate": 4.540266102596313e-06, + "loss": 0.0062, + "step": 42343 + }, + { + "epoch": 5.55, + "learning_rate": 4.539821802599632e-06, + "loss": 0.0065, + "step": 42344 + }, + { + "epoch": 5.55, + "learning_rate": 4.539377517959411e-06, + "loss": 0.008, + "step": 42345 + }, + { + "epoch": 5.55, + "learning_rate": 4.5389332486769e-06, + "loss": 0.0061, + "step": 42346 + }, + { + "epoch": 5.55, + "learning_rate": 4.538488994753345e-06, + "loss": 0.0061, + "step": 42347 + }, + { + "epoch": 5.55, + "learning_rate": 4.538044756190002e-06, + "loss": 0.0072, + "step": 42348 + }, + { + "epoch": 5.55, + "learning_rate": 4.5376005329881175e-06, + "loss": 0.0047, + "step": 42349 + }, + { + "epoch": 5.55, + "learning_rate": 4.537156325148937e-06, + "loss": 0.0066, + "step": 42350 + }, + { + "epoch": 5.55, + "learning_rate": 4.536712132673717e-06, + "loss": 0.0084, + "step": 42351 + }, + { + "epoch": 5.55, + "learning_rate": 4.536267955563704e-06, + "loss": 0.0064, + "step": 42352 + }, + { + "epoch": 5.55, + "learning_rate": 4.535823793820146e-06, + "loss": 0.0082, + "step": 42353 + }, + { + "epoch": 5.55, + "learning_rate": 4.535379647444288e-06, + "loss": 0.007, + "step": 42354 + }, + { + "epoch": 5.55, + "learning_rate": 4.534935516437384e-06, + "loss": 0.0092, + "step": 42355 + }, + { + "epoch": 5.55, + "learning_rate": 4.534491400800688e-06, + "loss": 0.007, + "step": 42356 + }, + { + "epoch": 5.55, + "learning_rate": 4.534047300535443e-06, + "loss": 0.0063, + "step": 42357 + }, + { + "epoch": 5.55, + "learning_rate": 4.533603215642894e-06, + "loss": 0.0065, + "step": 42358 + }, + { + "epoch": 5.55, + "learning_rate": 4.5331591461242994e-06, + "loss": 0.0064, + "step": 42359 + }, + { + "epoch": 5.55, + "learning_rate": 4.5327150919809036e-06, + "loss": 0.0073, + "step": 42360 + }, + { + "epoch": 5.55, + "learning_rate": 4.532271053213956e-06, + "loss": 0.0075, + "step": 42361 + }, + { + "epoch": 5.55, + "learning_rate": 4.5318270298247e-06, + "loss": 0.0055, + "step": 42362 + }, + { + "epoch": 5.55, + "learning_rate": 4.531383021814395e-06, + "loss": 0.0063, + "step": 42363 + }, + { + "epoch": 5.55, + "learning_rate": 4.530939029184282e-06, + "loss": 0.0085, + "step": 42364 + }, + { + "epoch": 5.55, + "learning_rate": 4.53049505193561e-06, + "loss": 0.0073, + "step": 42365 + }, + { + "epoch": 5.55, + "learning_rate": 4.530051090069627e-06, + "loss": 0.0078, + "step": 42366 + }, + { + "epoch": 5.55, + "learning_rate": 4.5296071435875884e-06, + "loss": 0.0061, + "step": 42367 + }, + { + "epoch": 5.55, + "learning_rate": 4.529163212490739e-06, + "loss": 0.007, + "step": 42368 + }, + { + "epoch": 5.55, + "learning_rate": 4.5287192967803275e-06, + "loss": 0.0075, + "step": 42369 + }, + { + "epoch": 5.55, + "learning_rate": 4.5282753964575964e-06, + "loss": 0.0066, + "step": 42370 + }, + { + "epoch": 5.55, + "learning_rate": 4.527831511523803e-06, + "loss": 0.0072, + "step": 42371 + }, + { + "epoch": 5.55, + "learning_rate": 4.527387641980195e-06, + "loss": 0.0072, + "step": 42372 + }, + { + "epoch": 5.55, + "learning_rate": 4.526943787828011e-06, + "loss": 0.007, + "step": 42373 + }, + { + "epoch": 5.55, + "learning_rate": 4.52649994906851e-06, + "loss": 0.0059, + "step": 42374 + }, + { + "epoch": 5.55, + "learning_rate": 4.526056125702938e-06, + "loss": 0.0063, + "step": 42375 + }, + { + "epoch": 5.55, + "learning_rate": 4.525612317732536e-06, + "loss": 0.0086, + "step": 42376 + }, + { + "epoch": 5.55, + "learning_rate": 4.525168525158563e-06, + "loss": 0.0054, + "step": 42377 + }, + { + "epoch": 5.55, + "learning_rate": 4.524724747982258e-06, + "loss": 0.0077, + "step": 42378 + }, + { + "epoch": 5.55, + "learning_rate": 4.524280986204876e-06, + "loss": 0.0073, + "step": 42379 + }, + { + "epoch": 5.55, + "learning_rate": 4.523837239827663e-06, + "loss": 0.0038, + "step": 42380 + }, + { + "epoch": 5.55, + "learning_rate": 4.523393508851862e-06, + "loss": 0.0067, + "step": 42381 + }, + { + "epoch": 5.55, + "learning_rate": 4.522949793278729e-06, + "loss": 0.0068, + "step": 42382 + }, + { + "epoch": 5.55, + "learning_rate": 4.522506093109509e-06, + "loss": 0.0071, + "step": 42383 + }, + { + "epoch": 5.55, + "learning_rate": 4.5220624083454474e-06, + "loss": 0.0067, + "step": 42384 + }, + { + "epoch": 5.55, + "learning_rate": 4.52161873898779e-06, + "loss": 0.0086, + "step": 42385 + }, + { + "epoch": 5.55, + "learning_rate": 4.521175085037793e-06, + "loss": 0.0063, + "step": 42386 + }, + { + "epoch": 5.55, + "learning_rate": 4.520731446496695e-06, + "loss": 0.0076, + "step": 42387 + }, + { + "epoch": 5.55, + "learning_rate": 4.520287823365751e-06, + "loss": 0.0064, + "step": 42388 + }, + { + "epoch": 5.55, + "learning_rate": 4.519844215646203e-06, + "loss": 0.009, + "step": 42389 + }, + { + "epoch": 5.55, + "learning_rate": 4.519400623339305e-06, + "loss": 0.007, + "step": 42390 + }, + { + "epoch": 5.55, + "learning_rate": 4.5189570464463005e-06, + "loss": 0.0069, + "step": 42391 + }, + { + "epoch": 5.55, + "learning_rate": 4.518513484968433e-06, + "loss": 0.0075, + "step": 42392 + }, + { + "epoch": 5.55, + "learning_rate": 4.51806993890696e-06, + "loss": 0.0061, + "step": 42393 + }, + { + "epoch": 5.55, + "learning_rate": 4.517626408263121e-06, + "loss": 0.0061, + "step": 42394 + }, + { + "epoch": 5.55, + "learning_rate": 4.5171828930381675e-06, + "loss": 0.0061, + "step": 42395 + }, + { + "epoch": 5.55, + "learning_rate": 4.51673939323334e-06, + "loss": 0.0071, + "step": 42396 + }, + { + "epoch": 5.55, + "learning_rate": 4.516295908849896e-06, + "loss": 0.0091, + "step": 42397 + }, + { + "epoch": 5.55, + "learning_rate": 4.515852439889074e-06, + "loss": 0.0078, + "step": 42398 + }, + { + "epoch": 5.55, + "learning_rate": 4.515408986352128e-06, + "loss": 0.0063, + "step": 42399 + }, + { + "epoch": 5.55, + "learning_rate": 4.514965548240299e-06, + "loss": 0.0062, + "step": 42400 + }, + { + "epoch": 5.55, + "learning_rate": 4.514522125554841e-06, + "loss": 0.0067, + "step": 42401 + }, + { + "epoch": 5.55, + "learning_rate": 4.514078718296997e-06, + "loss": 0.007, + "step": 42402 + }, + { + "epoch": 5.55, + "learning_rate": 4.513635326468014e-06, + "loss": 0.0044, + "step": 42403 + }, + { + "epoch": 5.55, + "learning_rate": 4.513191950069137e-06, + "loss": 0.0061, + "step": 42404 + }, + { + "epoch": 5.55, + "learning_rate": 4.5127485891016185e-06, + "loss": 0.0066, + "step": 42405 + }, + { + "epoch": 5.55, + "learning_rate": 4.512305243566703e-06, + "loss": 0.0068, + "step": 42406 + }, + { + "epoch": 5.56, + "learning_rate": 4.511861913465632e-06, + "loss": 0.0083, + "step": 42407 + }, + { + "epoch": 5.56, + "learning_rate": 4.5114185987996605e-06, + "loss": 0.0073, + "step": 42408 + }, + { + "epoch": 5.56, + "learning_rate": 4.510975299570029e-06, + "loss": 0.0069, + "step": 42409 + }, + { + "epoch": 5.56, + "learning_rate": 4.5105320157779905e-06, + "loss": 0.0081, + "step": 42410 + }, + { + "epoch": 5.56, + "learning_rate": 4.510088747424789e-06, + "loss": 0.006, + "step": 42411 + }, + { + "epoch": 5.56, + "learning_rate": 4.509645494511667e-06, + "loss": 0.0062, + "step": 42412 + }, + { + "epoch": 5.56, + "learning_rate": 4.5092022570398775e-06, + "loss": 0.0079, + "step": 42413 + }, + { + "epoch": 5.56, + "learning_rate": 4.508759035010665e-06, + "loss": 0.0067, + "step": 42414 + }, + { + "epoch": 5.56, + "learning_rate": 4.5083158284252714e-06, + "loss": 0.0054, + "step": 42415 + }, + { + "epoch": 5.56, + "learning_rate": 4.507872637284951e-06, + "loss": 0.0072, + "step": 42416 + }, + { + "epoch": 5.56, + "learning_rate": 4.5074294615909455e-06, + "loss": 0.0072, + "step": 42417 + }, + { + "epoch": 5.56, + "learning_rate": 4.506986301344502e-06, + "loss": 0.0075, + "step": 42418 + }, + { + "epoch": 5.56, + "learning_rate": 4.506543156546864e-06, + "loss": 0.0048, + "step": 42419 + }, + { + "epoch": 5.56, + "learning_rate": 4.506100027199282e-06, + "loss": 0.0091, + "step": 42420 + }, + { + "epoch": 5.56, + "learning_rate": 4.505656913303003e-06, + "loss": 0.0055, + "step": 42421 + }, + { + "epoch": 5.56, + "learning_rate": 4.505213814859272e-06, + "loss": 0.0055, + "step": 42422 + }, + { + "epoch": 5.56, + "learning_rate": 4.5047707318693305e-06, + "loss": 0.0076, + "step": 42423 + }, + { + "epoch": 5.56, + "learning_rate": 4.504327664334432e-06, + "loss": 0.0075, + "step": 42424 + }, + { + "epoch": 5.56, + "learning_rate": 4.50388461225582e-06, + "loss": 0.0065, + "step": 42425 + }, + { + "epoch": 5.56, + "learning_rate": 4.503441575634741e-06, + "loss": 0.0067, + "step": 42426 + }, + { + "epoch": 5.56, + "learning_rate": 4.502998554472434e-06, + "loss": 0.0071, + "step": 42427 + }, + { + "epoch": 5.56, + "learning_rate": 4.502555548770155e-06, + "loss": 0.006, + "step": 42428 + }, + { + "epoch": 5.56, + "learning_rate": 4.502112558529147e-06, + "loss": 0.0062, + "step": 42429 + }, + { + "epoch": 5.56, + "learning_rate": 4.50166958375065e-06, + "loss": 0.0089, + "step": 42430 + }, + { + "epoch": 5.56, + "learning_rate": 4.501226624435915e-06, + "loss": 0.0058, + "step": 42431 + }, + { + "epoch": 5.56, + "learning_rate": 4.50078368058619e-06, + "loss": 0.0063, + "step": 42432 + }, + { + "epoch": 5.56, + "learning_rate": 4.500340752202719e-06, + "loss": 0.0062, + "step": 42433 + }, + { + "epoch": 5.56, + "learning_rate": 4.499897839286746e-06, + "loss": 0.0074, + "step": 42434 + }, + { + "epoch": 5.56, + "learning_rate": 4.499454941839514e-06, + "loss": 0.0068, + "step": 42435 + }, + { + "epoch": 5.56, + "learning_rate": 4.499012059862276e-06, + "loss": 0.0077, + "step": 42436 + }, + { + "epoch": 5.56, + "learning_rate": 4.498569193356274e-06, + "loss": 0.0067, + "step": 42437 + }, + { + "epoch": 5.56, + "learning_rate": 4.498126342322749e-06, + "loss": 0.0062, + "step": 42438 + }, + { + "epoch": 5.56, + "learning_rate": 4.497683506762954e-06, + "loss": 0.007, + "step": 42439 + }, + { + "epoch": 5.56, + "learning_rate": 4.4972406866781315e-06, + "loss": 0.0069, + "step": 42440 + }, + { + "epoch": 5.56, + "learning_rate": 4.496797882069527e-06, + "loss": 0.0072, + "step": 42441 + }, + { + "epoch": 5.56, + "learning_rate": 4.496355092938377e-06, + "loss": 0.0074, + "step": 42442 + }, + { + "epoch": 5.56, + "learning_rate": 4.495912319285943e-06, + "loss": 0.0075, + "step": 42443 + }, + { + "epoch": 5.56, + "learning_rate": 4.495469561113464e-06, + "loss": 0.0072, + "step": 42444 + }, + { + "epoch": 5.56, + "learning_rate": 4.495026818422181e-06, + "loss": 0.0071, + "step": 42445 + }, + { + "epoch": 5.56, + "learning_rate": 4.49458409121334e-06, + "loss": 0.0079, + "step": 42446 + }, + { + "epoch": 5.56, + "learning_rate": 4.494141379488192e-06, + "loss": 0.0066, + "step": 42447 + }, + { + "epoch": 5.56, + "learning_rate": 4.493698683247977e-06, + "loss": 0.0076, + "step": 42448 + }, + { + "epoch": 5.56, + "learning_rate": 4.493256002493942e-06, + "loss": 0.0074, + "step": 42449 + }, + { + "epoch": 5.56, + "learning_rate": 4.4928133372273275e-06, + "loss": 0.0063, + "step": 42450 + }, + { + "epoch": 5.56, + "learning_rate": 4.492370687449385e-06, + "loss": 0.0073, + "step": 42451 + }, + { + "epoch": 5.56, + "learning_rate": 4.491928053161357e-06, + "loss": 0.006, + "step": 42452 + }, + { + "epoch": 5.56, + "learning_rate": 4.4914854343644855e-06, + "loss": 0.0091, + "step": 42453 + }, + { + "epoch": 5.56, + "learning_rate": 4.491042831060017e-06, + "loss": 0.0082, + "step": 42454 + }, + { + "epoch": 5.56, + "learning_rate": 4.490600243249201e-06, + "loss": 0.0057, + "step": 42455 + }, + { + "epoch": 5.56, + "learning_rate": 4.4901576709332785e-06, + "loss": 0.0066, + "step": 42456 + }, + { + "epoch": 5.56, + "learning_rate": 4.48971511411349e-06, + "loss": 0.0076, + "step": 42457 + }, + { + "epoch": 5.56, + "learning_rate": 4.489272572791088e-06, + "loss": 0.0065, + "step": 42458 + }, + { + "epoch": 5.56, + "learning_rate": 4.488830046967314e-06, + "loss": 0.0058, + "step": 42459 + }, + { + "epoch": 5.56, + "learning_rate": 4.488387536643412e-06, + "loss": 0.0067, + "step": 42460 + }, + { + "epoch": 5.56, + "learning_rate": 4.487945041820622e-06, + "loss": 0.0072, + "step": 42461 + }, + { + "epoch": 5.56, + "learning_rate": 4.487502562500198e-06, + "loss": 0.0072, + "step": 42462 + }, + { + "epoch": 5.56, + "learning_rate": 4.487060098683375e-06, + "loss": 0.0074, + "step": 42463 + }, + { + "epoch": 5.56, + "learning_rate": 4.486617650371405e-06, + "loss": 0.0069, + "step": 42464 + }, + { + "epoch": 5.56, + "learning_rate": 4.486175217565526e-06, + "loss": 0.0064, + "step": 42465 + }, + { + "epoch": 5.56, + "learning_rate": 4.4857328002669895e-06, + "loss": 0.0074, + "step": 42466 + }, + { + "epoch": 5.56, + "learning_rate": 4.485290398477036e-06, + "loss": 0.0099, + "step": 42467 + }, + { + "epoch": 5.56, + "learning_rate": 4.48484801219691e-06, + "loss": 0.0064, + "step": 42468 + }, + { + "epoch": 5.56, + "learning_rate": 4.48440564142785e-06, + "loss": 0.0058, + "step": 42469 + }, + { + "epoch": 5.56, + "learning_rate": 4.483963286171111e-06, + "loss": 0.0058, + "step": 42470 + }, + { + "epoch": 5.56, + "learning_rate": 4.48352094642793e-06, + "loss": 0.0097, + "step": 42471 + }, + { + "epoch": 5.56, + "learning_rate": 4.483078622199549e-06, + "loss": 0.0071, + "step": 42472 + }, + { + "epoch": 5.56, + "learning_rate": 4.482636313487221e-06, + "loss": 0.0072, + "step": 42473 + }, + { + "epoch": 5.56, + "learning_rate": 4.482194020292179e-06, + "loss": 0.0081, + "step": 42474 + }, + { + "epoch": 5.56, + "learning_rate": 4.4817517426156766e-06, + "loss": 0.0056, + "step": 42475 + }, + { + "epoch": 5.56, + "learning_rate": 4.481309480458954e-06, + "loss": 0.0098, + "step": 42476 + }, + { + "epoch": 5.56, + "learning_rate": 4.480867233823251e-06, + "loss": 0.0085, + "step": 42477 + }, + { + "epoch": 5.56, + "learning_rate": 4.480425002709818e-06, + "loss": 0.0054, + "step": 42478 + }, + { + "epoch": 5.56, + "learning_rate": 4.4799827871198965e-06, + "loss": 0.0063, + "step": 42479 + }, + { + "epoch": 5.56, + "learning_rate": 4.479540587054725e-06, + "loss": 0.0091, + "step": 42480 + }, + { + "epoch": 5.56, + "learning_rate": 4.479098402515556e-06, + "loss": 0.0073, + "step": 42481 + }, + { + "epoch": 5.56, + "learning_rate": 4.478656233503628e-06, + "loss": 0.0061, + "step": 42482 + }, + { + "epoch": 5.56, + "learning_rate": 4.478214080020186e-06, + "loss": 0.0056, + "step": 42483 + }, + { + "epoch": 5.57, + "learning_rate": 4.477771942066469e-06, + "loss": 0.0093, + "step": 42484 + }, + { + "epoch": 5.57, + "learning_rate": 4.477329819643723e-06, + "loss": 0.008, + "step": 42485 + }, + { + "epoch": 5.57, + "learning_rate": 4.476887712753198e-06, + "loss": 0.007, + "step": 42486 + }, + { + "epoch": 5.57, + "learning_rate": 4.476445621396131e-06, + "loss": 0.0068, + "step": 42487 + }, + { + "epoch": 5.57, + "learning_rate": 4.476003545573765e-06, + "loss": 0.0065, + "step": 42488 + }, + { + "epoch": 5.57, + "learning_rate": 4.475561485287347e-06, + "loss": 0.0075, + "step": 42489 + }, + { + "epoch": 5.57, + "learning_rate": 4.475119440538118e-06, + "loss": 0.0086, + "step": 42490 + }, + { + "epoch": 5.57, + "learning_rate": 4.4746774113273215e-06, + "loss": 0.0062, + "step": 42491 + }, + { + "epoch": 5.57, + "learning_rate": 4.474235397656198e-06, + "loss": 0.0077, + "step": 42492 + }, + { + "epoch": 5.57, + "learning_rate": 4.473793399525995e-06, + "loss": 0.0083, + "step": 42493 + }, + { + "epoch": 5.57, + "learning_rate": 4.473351416937955e-06, + "loss": 0.0081, + "step": 42494 + }, + { + "epoch": 5.57, + "learning_rate": 4.472909449893316e-06, + "loss": 0.0052, + "step": 42495 + }, + { + "epoch": 5.57, + "learning_rate": 4.472467498393324e-06, + "loss": 0.0072, + "step": 42496 + }, + { + "epoch": 5.57, + "learning_rate": 4.472025562439228e-06, + "loss": 0.0044, + "step": 42497 + }, + { + "epoch": 5.57, + "learning_rate": 4.4715836420322654e-06, + "loss": 0.0091, + "step": 42498 + }, + { + "epoch": 5.57, + "learning_rate": 4.4711417371736785e-06, + "loss": 0.0077, + "step": 42499 + }, + { + "epoch": 5.57, + "learning_rate": 4.470699847864708e-06, + "loss": 0.0071, + "step": 42500 + }, + { + "epoch": 5.57, + "learning_rate": 4.470257974106603e-06, + "loss": 0.0058, + "step": 42501 + }, + { + "epoch": 5.57, + "learning_rate": 4.469816115900604e-06, + "loss": 0.0079, + "step": 42502 + }, + { + "epoch": 5.57, + "learning_rate": 4.469374273247947e-06, + "loss": 0.0092, + "step": 42503 + }, + { + "epoch": 5.57, + "learning_rate": 4.468932446149885e-06, + "loss": 0.0101, + "step": 42504 + }, + { + "epoch": 5.57, + "learning_rate": 4.4684906346076565e-06, + "loss": 0.0094, + "step": 42505 + }, + { + "epoch": 5.57, + "learning_rate": 4.468048838622503e-06, + "loss": 0.008, + "step": 42506 + }, + { + "epoch": 5.57, + "learning_rate": 4.467607058195661e-06, + "loss": 0.0076, + "step": 42507 + }, + { + "epoch": 5.57, + "learning_rate": 4.467165293328386e-06, + "loss": 0.0084, + "step": 42508 + }, + { + "epoch": 5.57, + "learning_rate": 4.466723544021913e-06, + "loss": 0.0057, + "step": 42509 + }, + { + "epoch": 5.57, + "learning_rate": 4.466281810277488e-06, + "loss": 0.0071, + "step": 42510 + }, + { + "epoch": 5.57, + "learning_rate": 4.465840092096346e-06, + "loss": 0.0064, + "step": 42511 + }, + { + "epoch": 5.57, + "learning_rate": 4.465398389479736e-06, + "loss": 0.0062, + "step": 42512 + }, + { + "epoch": 5.57, + "learning_rate": 4.4649567024289e-06, + "loss": 0.0071, + "step": 42513 + }, + { + "epoch": 5.57, + "learning_rate": 4.464515030945078e-06, + "loss": 0.0068, + "step": 42514 + }, + { + "epoch": 5.57, + "learning_rate": 4.46407337502951e-06, + "loss": 0.0064, + "step": 42515 + }, + { + "epoch": 5.57, + "learning_rate": 4.463631734683442e-06, + "loss": 0.0048, + "step": 42516 + }, + { + "epoch": 5.57, + "learning_rate": 4.463190109908118e-06, + "loss": 0.0061, + "step": 42517 + }, + { + "epoch": 5.57, + "learning_rate": 4.462748500704771e-06, + "loss": 0.0058, + "step": 42518 + }, + { + "epoch": 5.57, + "learning_rate": 4.462306907074649e-06, + "loss": 0.008, + "step": 42519 + }, + { + "epoch": 5.57, + "learning_rate": 4.461865329018999e-06, + "loss": 0.0088, + "step": 42520 + }, + { + "epoch": 5.57, + "learning_rate": 4.461423766539058e-06, + "loss": 0.0063, + "step": 42521 + }, + { + "epoch": 5.57, + "learning_rate": 4.4609822196360634e-06, + "loss": 0.0047, + "step": 42522 + }, + { + "epoch": 5.57, + "learning_rate": 4.460540688311264e-06, + "loss": 0.01, + "step": 42523 + }, + { + "epoch": 5.57, + "learning_rate": 4.4600991725659e-06, + "loss": 0.0045, + "step": 42524 + }, + { + "epoch": 5.57, + "learning_rate": 4.459657672401212e-06, + "loss": 0.0082, + "step": 42525 + }, + { + "epoch": 5.57, + "learning_rate": 4.459216187818438e-06, + "loss": 0.0073, + "step": 42526 + }, + { + "epoch": 5.57, + "learning_rate": 4.458774718818827e-06, + "loss": 0.0081, + "step": 42527 + }, + { + "epoch": 5.57, + "learning_rate": 4.458333265403617e-06, + "loss": 0.0067, + "step": 42528 + }, + { + "epoch": 5.57, + "learning_rate": 4.4578918275740455e-06, + "loss": 0.0084, + "step": 42529 + }, + { + "epoch": 5.57, + "learning_rate": 4.45745040533136e-06, + "loss": 0.0052, + "step": 42530 + }, + { + "epoch": 5.57, + "learning_rate": 4.457008998676802e-06, + "loss": 0.0078, + "step": 42531 + }, + { + "epoch": 5.57, + "learning_rate": 4.45656760761161e-06, + "loss": 0.0094, + "step": 42532 + }, + { + "epoch": 5.57, + "learning_rate": 4.456126232137028e-06, + "loss": 0.0071, + "step": 42533 + }, + { + "epoch": 5.57, + "learning_rate": 4.4556848722542926e-06, + "loss": 0.0076, + "step": 42534 + }, + { + "epoch": 5.57, + "learning_rate": 4.455243527964651e-06, + "loss": 0.0053, + "step": 42535 + }, + { + "epoch": 5.57, + "learning_rate": 4.454802199269341e-06, + "loss": 0.0069, + "step": 42536 + }, + { + "epoch": 5.57, + "learning_rate": 4.454360886169602e-06, + "loss": 0.0069, + "step": 42537 + }, + { + "epoch": 5.57, + "learning_rate": 4.453919588666681e-06, + "loss": 0.0059, + "step": 42538 + }, + { + "epoch": 5.57, + "learning_rate": 4.453478306761816e-06, + "loss": 0.0061, + "step": 42539 + }, + { + "epoch": 5.57, + "learning_rate": 4.453037040456244e-06, + "loss": 0.0083, + "step": 42540 + }, + { + "epoch": 5.57, + "learning_rate": 4.452595789751214e-06, + "loss": 0.0067, + "step": 42541 + }, + { + "epoch": 5.57, + "learning_rate": 4.452154554647959e-06, + "loss": 0.0064, + "step": 42542 + }, + { + "epoch": 5.57, + "learning_rate": 4.4517133351477275e-06, + "loss": 0.0076, + "step": 42543 + }, + { + "epoch": 5.57, + "learning_rate": 4.451272131251756e-06, + "loss": 0.0061, + "step": 42544 + }, + { + "epoch": 5.57, + "learning_rate": 4.450830942961283e-06, + "loss": 0.0056, + "step": 42545 + }, + { + "epoch": 5.57, + "learning_rate": 4.450389770277556e-06, + "loss": 0.0071, + "step": 42546 + }, + { + "epoch": 5.57, + "learning_rate": 4.449948613201814e-06, + "loss": 0.0084, + "step": 42547 + }, + { + "epoch": 5.57, + "learning_rate": 4.449507471735294e-06, + "loss": 0.0061, + "step": 42548 + }, + { + "epoch": 5.57, + "learning_rate": 4.449066345879235e-06, + "loss": 0.0075, + "step": 42549 + }, + { + "epoch": 5.57, + "learning_rate": 4.448625235634886e-06, + "loss": 0.0075, + "step": 42550 + }, + { + "epoch": 5.57, + "learning_rate": 4.448184141003479e-06, + "loss": 0.0057, + "step": 42551 + }, + { + "epoch": 5.57, + "learning_rate": 4.4477430619862615e-06, + "loss": 0.0085, + "step": 42552 + }, + { + "epoch": 5.57, + "learning_rate": 4.447301998584467e-06, + "loss": 0.0088, + "step": 42553 + }, + { + "epoch": 5.57, + "learning_rate": 4.446860950799345e-06, + "loss": 0.0073, + "step": 42554 + }, + { + "epoch": 5.57, + "learning_rate": 4.44641991863213e-06, + "loss": 0.0071, + "step": 42555 + }, + { + "epoch": 5.57, + "learning_rate": 4.445978902084062e-06, + "loss": 0.0062, + "step": 42556 + }, + { + "epoch": 5.57, + "learning_rate": 4.445537901156381e-06, + "loss": 0.0088, + "step": 42557 + }, + { + "epoch": 5.57, + "learning_rate": 4.445096915850332e-06, + "loss": 0.0051, + "step": 42558 + }, + { + "epoch": 5.57, + "learning_rate": 4.444655946167151e-06, + "loss": 0.0051, + "step": 42559 + }, + { + "epoch": 5.58, + "learning_rate": 4.444214992108077e-06, + "loss": 0.006, + "step": 42560 + }, + { + "epoch": 5.58, + "learning_rate": 4.4437740536743555e-06, + "loss": 0.007, + "step": 42561 + }, + { + "epoch": 5.58, + "learning_rate": 4.44333313086722e-06, + "loss": 0.0095, + "step": 42562 + }, + { + "epoch": 5.58, + "learning_rate": 4.442892223687918e-06, + "loss": 0.0086, + "step": 42563 + }, + { + "epoch": 5.58, + "learning_rate": 4.442451332137686e-06, + "loss": 0.0073, + "step": 42564 + }, + { + "epoch": 5.58, + "learning_rate": 4.442010456217759e-06, + "loss": 0.007, + "step": 42565 + }, + { + "epoch": 5.58, + "learning_rate": 4.441569595929386e-06, + "loss": 0.0054, + "step": 42566 + }, + { + "epoch": 5.58, + "learning_rate": 4.441128751273802e-06, + "loss": 0.0078, + "step": 42567 + }, + { + "epoch": 5.58, + "learning_rate": 4.440687922252244e-06, + "loss": 0.0059, + "step": 42568 + }, + { + "epoch": 5.58, + "learning_rate": 4.4402471088659595e-06, + "loss": 0.005, + "step": 42569 + }, + { + "epoch": 5.58, + "learning_rate": 4.439806311116184e-06, + "loss": 0.0078, + "step": 42570 + }, + { + "epoch": 5.58, + "learning_rate": 4.4393655290041556e-06, + "loss": 0.0051, + "step": 42571 + }, + { + "epoch": 5.58, + "learning_rate": 4.438924762531112e-06, + "loss": 0.007, + "step": 42572 + }, + { + "epoch": 5.58, + "learning_rate": 4.438484011698298e-06, + "loss": 0.0107, + "step": 42573 + }, + { + "epoch": 5.58, + "learning_rate": 4.438043276506954e-06, + "loss": 0.0042, + "step": 42574 + }, + { + "epoch": 5.58, + "learning_rate": 4.437602556958318e-06, + "loss": 0.0093, + "step": 42575 + }, + { + "epoch": 5.58, + "learning_rate": 4.437161853053624e-06, + "loss": 0.0066, + "step": 42576 + }, + { + "epoch": 5.58, + "learning_rate": 4.43672116479412e-06, + "loss": 0.0076, + "step": 42577 + }, + { + "epoch": 5.58, + "learning_rate": 4.436280492181041e-06, + "loss": 0.0062, + "step": 42578 + }, + { + "epoch": 5.58, + "learning_rate": 4.435839835215627e-06, + "loss": 0.0067, + "step": 42579 + }, + { + "epoch": 5.58, + "learning_rate": 4.435399193899115e-06, + "loss": 0.0054, + "step": 42580 + }, + { + "epoch": 5.58, + "learning_rate": 4.434958568232749e-06, + "loss": 0.0084, + "step": 42581 + }, + { + "epoch": 5.58, + "learning_rate": 4.434517958217765e-06, + "loss": 0.0094, + "step": 42582 + }, + { + "epoch": 5.58, + "learning_rate": 4.4340773638554005e-06, + "loss": 0.0063, + "step": 42583 + }, + { + "epoch": 5.58, + "learning_rate": 4.433636785146895e-06, + "loss": 0.0068, + "step": 42584 + }, + { + "epoch": 5.58, + "learning_rate": 4.433196222093496e-06, + "loss": 0.007, + "step": 42585 + }, + { + "epoch": 5.58, + "learning_rate": 4.432755674696434e-06, + "loss": 0.0072, + "step": 42586 + }, + { + "epoch": 5.58, + "learning_rate": 4.432315142956948e-06, + "loss": 0.0088, + "step": 42587 + }, + { + "epoch": 5.58, + "learning_rate": 4.431874626876281e-06, + "loss": 0.0056, + "step": 42588 + }, + { + "epoch": 5.58, + "learning_rate": 4.4314341264556715e-06, + "loss": 0.006, + "step": 42589 + }, + { + "epoch": 5.58, + "learning_rate": 4.430993641696357e-06, + "loss": 0.0065, + "step": 42590 + }, + { + "epoch": 5.58, + "learning_rate": 4.430553172599572e-06, + "loss": 0.0059, + "step": 42591 + }, + { + "epoch": 5.58, + "learning_rate": 4.430112719166564e-06, + "loss": 0.0061, + "step": 42592 + }, + { + "epoch": 5.58, + "learning_rate": 4.429672281398567e-06, + "loss": 0.0072, + "step": 42593 + }, + { + "epoch": 5.58, + "learning_rate": 4.429231859296817e-06, + "loss": 0.0071, + "step": 42594 + }, + { + "epoch": 5.58, + "learning_rate": 4.428791452862555e-06, + "loss": 0.0049, + "step": 42595 + }, + { + "epoch": 5.58, + "learning_rate": 4.428351062097023e-06, + "loss": 0.0095, + "step": 42596 + }, + { + "epoch": 5.58, + "learning_rate": 4.4279106870014585e-06, + "loss": 0.0058, + "step": 42597 + }, + { + "epoch": 5.58, + "learning_rate": 4.427470327577098e-06, + "loss": 0.0076, + "step": 42598 + }, + { + "epoch": 5.58, + "learning_rate": 4.427029983825176e-06, + "loss": 0.008, + "step": 42599 + }, + { + "epoch": 5.58, + "learning_rate": 4.42658965574694e-06, + "loss": 0.0054, + "step": 42600 + }, + { + "epoch": 5.58, + "learning_rate": 4.426149343343624e-06, + "loss": 0.0071, + "step": 42601 + }, + { + "epoch": 5.58, + "learning_rate": 4.425709046616462e-06, + "loss": 0.0066, + "step": 42602 + }, + { + "epoch": 5.58, + "learning_rate": 4.4252687655667005e-06, + "loss": 0.0073, + "step": 42603 + }, + { + "epoch": 5.58, + "learning_rate": 4.424828500195574e-06, + "loss": 0.0081, + "step": 42604 + }, + { + "epoch": 5.58, + "learning_rate": 4.424388250504316e-06, + "loss": 0.0061, + "step": 42605 + }, + { + "epoch": 5.58, + "learning_rate": 4.423948016494173e-06, + "loss": 0.007, + "step": 42606 + }, + { + "epoch": 5.58, + "learning_rate": 4.423507798166376e-06, + "loss": 0.0083, + "step": 42607 + }, + { + "epoch": 5.58, + "learning_rate": 4.423067595522169e-06, + "loss": 0.0092, + "step": 42608 + }, + { + "epoch": 5.58, + "learning_rate": 4.422627408562788e-06, + "loss": 0.0056, + "step": 42609 + }, + { + "epoch": 5.58, + "learning_rate": 4.422187237289467e-06, + "loss": 0.005, + "step": 42610 + }, + { + "epoch": 5.58, + "learning_rate": 4.421747081703451e-06, + "loss": 0.0071, + "step": 42611 + }, + { + "epoch": 5.58, + "learning_rate": 4.421306941805974e-06, + "loss": 0.0073, + "step": 42612 + }, + { + "epoch": 5.58, + "learning_rate": 4.4208668175982735e-06, + "loss": 0.0058, + "step": 42613 + }, + { + "epoch": 5.58, + "learning_rate": 4.420426709081586e-06, + "loss": 0.0095, + "step": 42614 + }, + { + "epoch": 5.58, + "learning_rate": 4.419986616257154e-06, + "loss": 0.0077, + "step": 42615 + }, + { + "epoch": 5.58, + "learning_rate": 4.419546539126208e-06, + "loss": 0.0068, + "step": 42616 + }, + { + "epoch": 5.58, + "learning_rate": 4.419106477689995e-06, + "loss": 0.006, + "step": 42617 + }, + { + "epoch": 5.58, + "learning_rate": 4.418666431949744e-06, + "loss": 0.0086, + "step": 42618 + }, + { + "epoch": 5.58, + "learning_rate": 4.4182264019067e-06, + "loss": 0.0071, + "step": 42619 + }, + { + "epoch": 5.58, + "learning_rate": 4.417786387562096e-06, + "loss": 0.0093, + "step": 42620 + }, + { + "epoch": 5.58, + "learning_rate": 4.417346388917172e-06, + "loss": 0.0087, + "step": 42621 + }, + { + "epoch": 5.58, + "learning_rate": 4.416906405973161e-06, + "loss": 0.0054, + "step": 42622 + }, + { + "epoch": 5.58, + "learning_rate": 4.416466438731305e-06, + "loss": 0.0055, + "step": 42623 + }, + { + "epoch": 5.58, + "learning_rate": 4.41602648719284e-06, + "loss": 0.0061, + "step": 42624 + }, + { + "epoch": 5.58, + "learning_rate": 4.415586551359e-06, + "loss": 0.0076, + "step": 42625 + }, + { + "epoch": 5.58, + "learning_rate": 4.41514663123103e-06, + "loss": 0.0088, + "step": 42626 + }, + { + "epoch": 5.58, + "learning_rate": 4.414706726810158e-06, + "loss": 0.0072, + "step": 42627 + }, + { + "epoch": 5.58, + "learning_rate": 4.4142668380976295e-06, + "loss": 0.008, + "step": 42628 + }, + { + "epoch": 5.58, + "learning_rate": 4.413826965094678e-06, + "loss": 0.006, + "step": 42629 + }, + { + "epoch": 5.58, + "learning_rate": 4.413387107802538e-06, + "loss": 0.0071, + "step": 42630 + }, + { + "epoch": 5.58, + "learning_rate": 4.412947266222453e-06, + "loss": 0.0071, + "step": 42631 + }, + { + "epoch": 5.58, + "learning_rate": 4.412507440355654e-06, + "loss": 0.0076, + "step": 42632 + }, + { + "epoch": 5.58, + "learning_rate": 4.4120676302033796e-06, + "loss": 0.0077, + "step": 42633 + }, + { + "epoch": 5.58, + "learning_rate": 4.411627835766869e-06, + "loss": 0.0078, + "step": 42634 + }, + { + "epoch": 5.58, + "learning_rate": 4.411188057047357e-06, + "loss": 0.0077, + "step": 42635 + }, + { + "epoch": 5.59, + "learning_rate": 4.410748294046084e-06, + "loss": 0.006, + "step": 42636 + }, + { + "epoch": 5.59, + "learning_rate": 4.410308546764277e-06, + "loss": 0.0057, + "step": 42637 + }, + { + "epoch": 5.59, + "learning_rate": 4.409868815203181e-06, + "loss": 0.009, + "step": 42638 + }, + { + "epoch": 5.59, + "learning_rate": 4.4094290993640356e-06, + "loss": 0.0061, + "step": 42639 + }, + { + "epoch": 5.59, + "learning_rate": 4.408989399248071e-06, + "loss": 0.0058, + "step": 42640 + }, + { + "epoch": 5.59, + "learning_rate": 4.408549714856524e-06, + "loss": 0.0061, + "step": 42641 + }, + { + "epoch": 5.59, + "learning_rate": 4.408110046190637e-06, + "loss": 0.0085, + "step": 42642 + }, + { + "epoch": 5.59, + "learning_rate": 4.4076703932516414e-06, + "loss": 0.0063, + "step": 42643 + }, + { + "epoch": 5.59, + "learning_rate": 4.407230756040775e-06, + "loss": 0.0047, + "step": 42644 + }, + { + "epoch": 5.59, + "learning_rate": 4.406791134559273e-06, + "loss": 0.0083, + "step": 42645 + }, + { + "epoch": 5.59, + "learning_rate": 4.4063515288083746e-06, + "loss": 0.0066, + "step": 42646 + }, + { + "epoch": 5.59, + "learning_rate": 4.405911938789315e-06, + "loss": 0.0057, + "step": 42647 + }, + { + "epoch": 5.59, + "learning_rate": 4.405472364503327e-06, + "loss": 0.006, + "step": 42648 + }, + { + "epoch": 5.59, + "learning_rate": 4.405032805951649e-06, + "loss": 0.0074, + "step": 42649 + }, + { + "epoch": 5.59, + "learning_rate": 4.404593263135523e-06, + "loss": 0.004, + "step": 42650 + }, + { + "epoch": 5.59, + "learning_rate": 4.40415373605618e-06, + "loss": 0.0062, + "step": 42651 + }, + { + "epoch": 5.59, + "learning_rate": 4.403714224714853e-06, + "loss": 0.0062, + "step": 42652 + }, + { + "epoch": 5.59, + "learning_rate": 4.403274729112785e-06, + "loss": 0.0065, + "step": 42653 + }, + { + "epoch": 5.59, + "learning_rate": 4.4028352492512095e-06, + "loss": 0.0095, + "step": 42654 + }, + { + "epoch": 5.59, + "learning_rate": 4.402395785131361e-06, + "loss": 0.0063, + "step": 42655 + }, + { + "epoch": 5.59, + "learning_rate": 4.401956336754474e-06, + "loss": 0.0059, + "step": 42656 + }, + { + "epoch": 5.59, + "learning_rate": 4.4015169041217885e-06, + "loss": 0.0052, + "step": 42657 + }, + { + "epoch": 5.59, + "learning_rate": 4.401077487234541e-06, + "loss": 0.0059, + "step": 42658 + }, + { + "epoch": 5.59, + "learning_rate": 4.400638086093959e-06, + "loss": 0.0058, + "step": 42659 + }, + { + "epoch": 5.59, + "learning_rate": 4.400198700701285e-06, + "loss": 0.0076, + "step": 42660 + }, + { + "epoch": 5.59, + "learning_rate": 4.399759331057758e-06, + "loss": 0.0058, + "step": 42661 + }, + { + "epoch": 5.59, + "learning_rate": 4.399319977164609e-06, + "loss": 0.0089, + "step": 42662 + }, + { + "epoch": 5.59, + "learning_rate": 4.398880639023074e-06, + "loss": 0.0071, + "step": 42663 + }, + { + "epoch": 5.59, + "learning_rate": 4.398441316634387e-06, + "loss": 0.0079, + "step": 42664 + }, + { + "epoch": 5.59, + "learning_rate": 4.398002009999787e-06, + "loss": 0.0042, + "step": 42665 + }, + { + "epoch": 5.59, + "learning_rate": 4.397562719120509e-06, + "loss": 0.0079, + "step": 42666 + }, + { + "epoch": 5.59, + "learning_rate": 4.397123443997784e-06, + "loss": 0.0064, + "step": 42667 + }, + { + "epoch": 5.59, + "learning_rate": 4.396684184632855e-06, + "loss": 0.0065, + "step": 42668 + }, + { + "epoch": 5.59, + "learning_rate": 4.396244941026952e-06, + "loss": 0.0089, + "step": 42669 + }, + { + "epoch": 5.59, + "learning_rate": 4.395805713181313e-06, + "loss": 0.0074, + "step": 42670 + }, + { + "epoch": 5.59, + "learning_rate": 4.395366501097167e-06, + "loss": 0.0104, + "step": 42671 + }, + { + "epoch": 5.59, + "learning_rate": 4.394927304775756e-06, + "loss": 0.0059, + "step": 42672 + }, + { + "epoch": 5.59, + "learning_rate": 4.394488124218316e-06, + "loss": 0.0082, + "step": 42673 + }, + { + "epoch": 5.59, + "learning_rate": 4.39404895942608e-06, + "loss": 0.0058, + "step": 42674 + }, + { + "epoch": 5.59, + "learning_rate": 4.393609810400278e-06, + "loss": 0.0042, + "step": 42675 + }, + { + "epoch": 5.59, + "learning_rate": 4.393170677142154e-06, + "loss": 0.0088, + "step": 42676 + }, + { + "epoch": 5.59, + "learning_rate": 4.39273155965294e-06, + "loss": 0.0064, + "step": 42677 + }, + { + "epoch": 5.59, + "learning_rate": 4.3922924579338686e-06, + "loss": 0.0072, + "step": 42678 + }, + { + "epoch": 5.59, + "learning_rate": 4.391853371986173e-06, + "loss": 0.0066, + "step": 42679 + }, + { + "epoch": 5.59, + "learning_rate": 4.391414301811095e-06, + "loss": 0.0096, + "step": 42680 + }, + { + "epoch": 5.59, + "learning_rate": 4.390975247409865e-06, + "loss": 0.0049, + "step": 42681 + }, + { + "epoch": 5.59, + "learning_rate": 4.390536208783715e-06, + "loss": 0.0064, + "step": 42682 + }, + { + "epoch": 5.59, + "learning_rate": 4.390097185933885e-06, + "loss": 0.0078, + "step": 42683 + }, + { + "epoch": 5.59, + "learning_rate": 4.389658178861609e-06, + "loss": 0.0068, + "step": 42684 + }, + { + "epoch": 5.59, + "learning_rate": 4.389219187568121e-06, + "loss": 0.0079, + "step": 42685 + }, + { + "epoch": 5.59, + "learning_rate": 4.3887802120546565e-06, + "loss": 0.005, + "step": 42686 + }, + { + "epoch": 5.59, + "learning_rate": 4.388341252322444e-06, + "loss": 0.0069, + "step": 42687 + }, + { + "epoch": 5.59, + "learning_rate": 4.387902308372728e-06, + "loss": 0.0084, + "step": 42688 + }, + { + "epoch": 5.59, + "learning_rate": 4.387463380206737e-06, + "loss": 0.0068, + "step": 42689 + }, + { + "epoch": 5.59, + "learning_rate": 4.3870244678257026e-06, + "loss": 0.007, + "step": 42690 + }, + { + "epoch": 5.59, + "learning_rate": 4.3865855712308655e-06, + "loss": 0.0066, + "step": 42691 + }, + { + "epoch": 5.59, + "learning_rate": 4.38614669042346e-06, + "loss": 0.0046, + "step": 42692 + }, + { + "epoch": 5.59, + "learning_rate": 4.3857078254047125e-06, + "loss": 0.0052, + "step": 42693 + }, + { + "epoch": 5.59, + "learning_rate": 4.385268976175866e-06, + "loss": 0.0055, + "step": 42694 + }, + { + "epoch": 5.59, + "learning_rate": 4.384830142738149e-06, + "loss": 0.0058, + "step": 42695 + }, + { + "epoch": 5.59, + "learning_rate": 4.3843913250928014e-06, + "loss": 0.0072, + "step": 42696 + }, + { + "epoch": 5.59, + "learning_rate": 4.383952523241055e-06, + "loss": 0.0055, + "step": 42697 + }, + { + "epoch": 5.59, + "learning_rate": 4.383513737184139e-06, + "loss": 0.0053, + "step": 42698 + }, + { + "epoch": 5.59, + "learning_rate": 4.383074966923294e-06, + "loss": 0.0063, + "step": 42699 + }, + { + "epoch": 5.59, + "learning_rate": 4.382636212459753e-06, + "loss": 0.0083, + "step": 42700 + }, + { + "epoch": 5.59, + "learning_rate": 4.382197473794748e-06, + "loss": 0.0054, + "step": 42701 + }, + { + "epoch": 5.59, + "learning_rate": 4.381758750929509e-06, + "loss": 0.0062, + "step": 42702 + }, + { + "epoch": 5.59, + "learning_rate": 4.3813200438652796e-06, + "loss": 0.0042, + "step": 42703 + }, + { + "epoch": 5.59, + "learning_rate": 4.380881352603283e-06, + "loss": 0.0059, + "step": 42704 + }, + { + "epoch": 5.59, + "learning_rate": 4.380442677144764e-06, + "loss": 0.0065, + "step": 42705 + }, + { + "epoch": 5.59, + "learning_rate": 4.3800040174909454e-06, + "loss": 0.0094, + "step": 42706 + }, + { + "epoch": 5.59, + "learning_rate": 4.37956537364307e-06, + "loss": 0.0064, + "step": 42707 + }, + { + "epoch": 5.59, + "learning_rate": 4.3791267456023675e-06, + "loss": 0.0064, + "step": 42708 + }, + { + "epoch": 5.59, + "learning_rate": 4.378688133370069e-06, + "loss": 0.0079, + "step": 42709 + }, + { + "epoch": 5.59, + "learning_rate": 4.378249536947414e-06, + "loss": 0.0067, + "step": 42710 + }, + { + "epoch": 5.59, + "learning_rate": 4.377810956335633e-06, + "loss": 0.0064, + "step": 42711 + }, + { + "epoch": 5.59, + "learning_rate": 4.3773723915359585e-06, + "loss": 0.0052, + "step": 42712 + }, + { + "epoch": 5.6, + "learning_rate": 4.376933842549622e-06, + "loss": 0.0048, + "step": 42713 + }, + { + "epoch": 5.6, + "learning_rate": 4.376495309377864e-06, + "loss": 0.0066, + "step": 42714 + }, + { + "epoch": 5.6, + "learning_rate": 4.376056792021909e-06, + "loss": 0.0056, + "step": 42715 + }, + { + "epoch": 5.6, + "learning_rate": 4.375618290482998e-06, + "loss": 0.0081, + "step": 42716 + }, + { + "epoch": 5.6, + "learning_rate": 4.375179804762358e-06, + "loss": 0.0091, + "step": 42717 + }, + { + "epoch": 5.6, + "learning_rate": 4.3747413348612276e-06, + "loss": 0.0052, + "step": 42718 + }, + { + "epoch": 5.6, + "learning_rate": 4.374302880780838e-06, + "loss": 0.0059, + "step": 42719 + }, + { + "epoch": 5.6, + "learning_rate": 4.373864442522423e-06, + "loss": 0.0054, + "step": 42720 + }, + { + "epoch": 5.6, + "learning_rate": 4.37342602008721e-06, + "loss": 0.0058, + "step": 42721 + }, + { + "epoch": 5.6, + "learning_rate": 4.372987613476442e-06, + "loss": 0.0066, + "step": 42722 + }, + { + "epoch": 5.6, + "learning_rate": 4.372549222691346e-06, + "loss": 0.0069, + "step": 42723 + }, + { + "epoch": 5.6, + "learning_rate": 4.372110847733151e-06, + "loss": 0.0059, + "step": 42724 + }, + { + "epoch": 5.6, + "learning_rate": 4.371672488603094e-06, + "loss": 0.0058, + "step": 42725 + }, + { + "epoch": 5.6, + "learning_rate": 4.371234145302414e-06, + "loss": 0.0073, + "step": 42726 + }, + { + "epoch": 5.6, + "learning_rate": 4.370795817832337e-06, + "loss": 0.0076, + "step": 42727 + }, + { + "epoch": 5.6, + "learning_rate": 4.370357506194097e-06, + "loss": 0.0094, + "step": 42728 + }, + { + "epoch": 5.6, + "learning_rate": 4.369919210388924e-06, + "loss": 0.0063, + "step": 42729 + }, + { + "epoch": 5.6, + "learning_rate": 4.369480930418056e-06, + "loss": 0.0055, + "step": 42730 + }, + { + "epoch": 5.6, + "learning_rate": 4.369042666282723e-06, + "loss": 0.0068, + "step": 42731 + }, + { + "epoch": 5.6, + "learning_rate": 4.368604417984156e-06, + "loss": 0.0056, + "step": 42732 + }, + { + "epoch": 5.6, + "learning_rate": 4.3681661855235905e-06, + "loss": 0.0084, + "step": 42733 + }, + { + "epoch": 5.6, + "learning_rate": 4.367727968902259e-06, + "loss": 0.01, + "step": 42734 + }, + { + "epoch": 5.6, + "learning_rate": 4.367289768121392e-06, + "loss": 0.0085, + "step": 42735 + }, + { + "epoch": 5.6, + "learning_rate": 4.366851583182219e-06, + "loss": 0.0053, + "step": 42736 + }, + { + "epoch": 5.6, + "learning_rate": 4.366413414085976e-06, + "loss": 0.0072, + "step": 42737 + }, + { + "epoch": 5.6, + "learning_rate": 4.365975260833899e-06, + "loss": 0.0065, + "step": 42738 + }, + { + "epoch": 5.6, + "learning_rate": 4.3655371234272155e-06, + "loss": 0.0077, + "step": 42739 + }, + { + "epoch": 5.6, + "learning_rate": 4.365099001867157e-06, + "loss": 0.008, + "step": 42740 + }, + { + "epoch": 5.6, + "learning_rate": 4.36466089615496e-06, + "loss": 0.0082, + "step": 42741 + }, + { + "epoch": 5.6, + "learning_rate": 4.364222806291855e-06, + "loss": 0.0084, + "step": 42742 + }, + { + "epoch": 5.6, + "learning_rate": 4.363784732279072e-06, + "loss": 0.0094, + "step": 42743 + }, + { + "epoch": 5.6, + "learning_rate": 4.363346674117841e-06, + "loss": 0.0058, + "step": 42744 + }, + { + "epoch": 5.6, + "learning_rate": 4.362908631809402e-06, + "loss": 0.008, + "step": 42745 + }, + { + "epoch": 5.6, + "learning_rate": 4.362470605354981e-06, + "loss": 0.0097, + "step": 42746 + }, + { + "epoch": 5.6, + "learning_rate": 4.362032594755808e-06, + "loss": 0.0062, + "step": 42747 + }, + { + "epoch": 5.6, + "learning_rate": 4.361594600013118e-06, + "loss": 0.0053, + "step": 42748 + }, + { + "epoch": 5.6, + "learning_rate": 4.361156621128147e-06, + "loss": 0.0072, + "step": 42749 + }, + { + "epoch": 5.6, + "learning_rate": 4.360718658102122e-06, + "loss": 0.009, + "step": 42750 + }, + { + "epoch": 5.6, + "learning_rate": 4.360280710936276e-06, + "loss": 0.0056, + "step": 42751 + }, + { + "epoch": 5.6, + "learning_rate": 4.359842779631836e-06, + "loss": 0.0109, + "step": 42752 + }, + { + "epoch": 5.6, + "learning_rate": 4.359404864190042e-06, + "loss": 0.0056, + "step": 42753 + }, + { + "epoch": 5.6, + "learning_rate": 4.358966964612121e-06, + "loss": 0.0058, + "step": 42754 + }, + { + "epoch": 5.6, + "learning_rate": 4.358529080899302e-06, + "loss": 0.0049, + "step": 42755 + }, + { + "epoch": 5.6, + "learning_rate": 4.358091213052823e-06, + "loss": 0.0076, + "step": 42756 + }, + { + "epoch": 5.6, + "learning_rate": 4.357653361073912e-06, + "loss": 0.0065, + "step": 42757 + }, + { + "epoch": 5.6, + "learning_rate": 4.357215524963797e-06, + "loss": 0.0085, + "step": 42758 + }, + { + "epoch": 5.6, + "learning_rate": 4.3567777047237125e-06, + "loss": 0.0064, + "step": 42759 + }, + { + "epoch": 5.6, + "learning_rate": 4.356339900354894e-06, + "loss": 0.0102, + "step": 42760 + }, + { + "epoch": 5.6, + "learning_rate": 4.3559021118585685e-06, + "loss": 0.0076, + "step": 42761 + }, + { + "epoch": 5.6, + "learning_rate": 4.355464339235968e-06, + "loss": 0.0057, + "step": 42762 + }, + { + "epoch": 5.6, + "learning_rate": 4.3550265824883195e-06, + "loss": 0.0051, + "step": 42763 + }, + { + "epoch": 5.6, + "learning_rate": 4.354588841616862e-06, + "loss": 0.0064, + "step": 42764 + }, + { + "epoch": 5.6, + "learning_rate": 4.354151116622823e-06, + "loss": 0.0097, + "step": 42765 + }, + { + "epoch": 5.6, + "learning_rate": 4.353713407507434e-06, + "loss": 0.0074, + "step": 42766 + }, + { + "epoch": 5.6, + "learning_rate": 4.3532757142719205e-06, + "loss": 0.0065, + "step": 42767 + }, + { + "epoch": 5.6, + "learning_rate": 4.352838036917523e-06, + "loss": 0.0064, + "step": 42768 + }, + { + "epoch": 5.6, + "learning_rate": 4.3524003754454625e-06, + "loss": 0.008, + "step": 42769 + }, + { + "epoch": 5.6, + "learning_rate": 4.35196272985698e-06, + "loss": 0.0092, + "step": 42770 + }, + { + "epoch": 5.6, + "learning_rate": 4.3515251001532975e-06, + "loss": 0.0065, + "step": 42771 + }, + { + "epoch": 5.6, + "learning_rate": 4.351087486335655e-06, + "loss": 0.0051, + "step": 42772 + }, + { + "epoch": 5.6, + "learning_rate": 4.3506498884052765e-06, + "loss": 0.0095, + "step": 42773 + }, + { + "epoch": 5.6, + "learning_rate": 4.3502123063633905e-06, + "loss": 0.0078, + "step": 42774 + }, + { + "epoch": 5.6, + "learning_rate": 4.349774740211237e-06, + "loss": 0.0053, + "step": 42775 + }, + { + "epoch": 5.6, + "learning_rate": 4.349337189950039e-06, + "loss": 0.0044, + "step": 42776 + }, + { + "epoch": 5.6, + "learning_rate": 4.34889965558103e-06, + "loss": 0.0103, + "step": 42777 + }, + { + "epoch": 5.6, + "learning_rate": 4.348462137105437e-06, + "loss": 0.0083, + "step": 42778 + }, + { + "epoch": 5.6, + "learning_rate": 4.348024634524496e-06, + "loss": 0.0086, + "step": 42779 + }, + { + "epoch": 5.6, + "learning_rate": 4.347587147839431e-06, + "loss": 0.0078, + "step": 42780 + }, + { + "epoch": 5.6, + "learning_rate": 4.347149677051481e-06, + "loss": 0.0071, + "step": 42781 + }, + { + "epoch": 5.6, + "learning_rate": 4.346712222161866e-06, + "loss": 0.0099, + "step": 42782 + }, + { + "epoch": 5.6, + "learning_rate": 4.3462747831718265e-06, + "loss": 0.0091, + "step": 42783 + }, + { + "epoch": 5.6, + "learning_rate": 4.345837360082589e-06, + "loss": 0.008, + "step": 42784 + }, + { + "epoch": 5.6, + "learning_rate": 4.345399952895381e-06, + "loss": 0.004, + "step": 42785 + }, + { + "epoch": 5.6, + "learning_rate": 4.344962561611432e-06, + "loss": 0.0062, + "step": 42786 + }, + { + "epoch": 5.6, + "learning_rate": 4.344525186231978e-06, + "loss": 0.0065, + "step": 42787 + }, + { + "epoch": 5.6, + "learning_rate": 4.344087826758245e-06, + "loss": 0.0061, + "step": 42788 + }, + { + "epoch": 5.61, + "learning_rate": 4.343650483191461e-06, + "loss": 0.0066, + "step": 42789 + }, + { + "epoch": 5.61, + "learning_rate": 4.343213155532863e-06, + "loss": 0.0058, + "step": 42790 + }, + { + "epoch": 5.61, + "learning_rate": 4.342775843783673e-06, + "loss": 0.0087, + "step": 42791 + }, + { + "epoch": 5.61, + "learning_rate": 4.342338547945127e-06, + "loss": 0.0075, + "step": 42792 + }, + { + "epoch": 5.61, + "learning_rate": 4.341901268018453e-06, + "loss": 0.0077, + "step": 42793 + }, + { + "epoch": 5.61, + "learning_rate": 4.341464004004877e-06, + "loss": 0.0079, + "step": 42794 + }, + { + "epoch": 5.61, + "learning_rate": 4.341026755905636e-06, + "loss": 0.0083, + "step": 42795 + }, + { + "epoch": 5.61, + "learning_rate": 4.3405895237219565e-06, + "loss": 0.0064, + "step": 42796 + }, + { + "epoch": 5.61, + "learning_rate": 4.340152307455064e-06, + "loss": 0.0085, + "step": 42797 + }, + { + "epoch": 5.61, + "learning_rate": 4.339715107106194e-06, + "loss": 0.0057, + "step": 42798 + }, + { + "epoch": 5.61, + "learning_rate": 4.339277922676574e-06, + "loss": 0.0083, + "step": 42799 + }, + { + "epoch": 5.61, + "learning_rate": 4.338840754167434e-06, + "loss": 0.0078, + "step": 42800 + }, + { + "epoch": 5.61, + "learning_rate": 4.338403601579999e-06, + "loss": 0.0047, + "step": 42801 + }, + { + "epoch": 5.61, + "learning_rate": 4.337966464915502e-06, + "loss": 0.0088, + "step": 42802 + }, + { + "epoch": 5.61, + "learning_rate": 4.3375293441751775e-06, + "loss": 0.0055, + "step": 42803 + }, + { + "epoch": 5.61, + "learning_rate": 4.337092239360249e-06, + "loss": 0.0074, + "step": 42804 + }, + { + "epoch": 5.61, + "learning_rate": 4.3366551504719435e-06, + "loss": 0.0073, + "step": 42805 + }, + { + "epoch": 5.61, + "learning_rate": 4.336218077511498e-06, + "loss": 0.0046, + "step": 42806 + }, + { + "epoch": 5.61, + "learning_rate": 4.3357810204801375e-06, + "loss": 0.0071, + "step": 42807 + }, + { + "epoch": 5.61, + "learning_rate": 4.335343979379091e-06, + "loss": 0.0062, + "step": 42808 + }, + { + "epoch": 5.61, + "learning_rate": 4.334906954209584e-06, + "loss": 0.0059, + "step": 42809 + }, + { + "epoch": 5.61, + "learning_rate": 4.334469944972855e-06, + "loss": 0.0067, + "step": 42810 + }, + { + "epoch": 5.61, + "learning_rate": 4.334032951670125e-06, + "loss": 0.0055, + "step": 42811 + }, + { + "epoch": 5.61, + "learning_rate": 4.333595974302623e-06, + "loss": 0.0057, + "step": 42812 + }, + { + "epoch": 5.61, + "learning_rate": 4.33315901287158e-06, + "loss": 0.0079, + "step": 42813 + }, + { + "epoch": 5.61, + "learning_rate": 4.332722067378231e-06, + "loss": 0.0059, + "step": 42814 + }, + { + "epoch": 5.61, + "learning_rate": 4.332285137823798e-06, + "loss": 0.0058, + "step": 42815 + }, + { + "epoch": 5.61, + "learning_rate": 4.331848224209511e-06, + "loss": 0.0059, + "step": 42816 + }, + { + "epoch": 5.61, + "learning_rate": 4.331411326536597e-06, + "loss": 0.0063, + "step": 42817 + }, + { + "epoch": 5.61, + "learning_rate": 4.330974444806289e-06, + "loss": 0.0079, + "step": 42818 + }, + { + "epoch": 5.61, + "learning_rate": 4.3305375790198124e-06, + "loss": 0.0061, + "step": 42819 + }, + { + "epoch": 5.61, + "learning_rate": 4.330100729178395e-06, + "loss": 0.0058, + "step": 42820 + }, + { + "epoch": 5.61, + "learning_rate": 4.32966389528327e-06, + "loss": 0.0064, + "step": 42821 + }, + { + "epoch": 5.61, + "learning_rate": 4.329227077335665e-06, + "loss": 0.0085, + "step": 42822 + }, + { + "epoch": 5.61, + "learning_rate": 4.328790275336805e-06, + "loss": 0.0066, + "step": 42823 + }, + { + "epoch": 5.61, + "learning_rate": 4.328353489287914e-06, + "loss": 0.0086, + "step": 42824 + }, + { + "epoch": 5.61, + "learning_rate": 4.327916719190234e-06, + "loss": 0.0071, + "step": 42825 + }, + { + "epoch": 5.61, + "learning_rate": 4.3274799650449865e-06, + "loss": 0.005, + "step": 42826 + }, + { + "epoch": 5.61, + "learning_rate": 4.327043226853398e-06, + "loss": 0.008, + "step": 42827 + }, + { + "epoch": 5.61, + "learning_rate": 4.326606504616696e-06, + "loss": 0.0085, + "step": 42828 + }, + { + "epoch": 5.61, + "learning_rate": 4.326169798336115e-06, + "loss": 0.0057, + "step": 42829 + }, + { + "epoch": 5.61, + "learning_rate": 4.3257331080128785e-06, + "loss": 0.0074, + "step": 42830 + }, + { + "epoch": 5.61, + "learning_rate": 4.325296433648215e-06, + "loss": 0.0072, + "step": 42831 + }, + { + "epoch": 5.61, + "learning_rate": 4.324859775243351e-06, + "loss": 0.0053, + "step": 42832 + }, + { + "epoch": 5.61, + "learning_rate": 4.324423132799519e-06, + "loss": 0.0062, + "step": 42833 + }, + { + "epoch": 5.61, + "learning_rate": 4.323986506317945e-06, + "loss": 0.0048, + "step": 42834 + }, + { + "epoch": 5.61, + "learning_rate": 4.323549895799854e-06, + "loss": 0.0059, + "step": 42835 + }, + { + "epoch": 5.61, + "learning_rate": 4.323113301246476e-06, + "loss": 0.0085, + "step": 42836 + }, + { + "epoch": 5.61, + "learning_rate": 4.322676722659043e-06, + "loss": 0.0075, + "step": 42837 + }, + { + "epoch": 5.61, + "learning_rate": 4.322240160038779e-06, + "loss": 0.0052, + "step": 42838 + }, + { + "epoch": 5.61, + "learning_rate": 4.321803613386909e-06, + "loss": 0.0058, + "step": 42839 + }, + { + "epoch": 5.61, + "learning_rate": 4.3213670827046675e-06, + "loss": 0.0083, + "step": 42840 + }, + { + "epoch": 5.61, + "learning_rate": 4.3209305679932774e-06, + "loss": 0.0085, + "step": 42841 + }, + { + "epoch": 5.61, + "learning_rate": 4.32049406925397e-06, + "loss": 0.0091, + "step": 42842 + }, + { + "epoch": 5.61, + "learning_rate": 4.320057586487966e-06, + "loss": 0.0088, + "step": 42843 + }, + { + "epoch": 5.61, + "learning_rate": 4.319621119696501e-06, + "loss": 0.0058, + "step": 42844 + }, + { + "epoch": 5.61, + "learning_rate": 4.319184668880799e-06, + "loss": 0.0091, + "step": 42845 + }, + { + "epoch": 5.61, + "learning_rate": 4.318748234042084e-06, + "loss": 0.0055, + "step": 42846 + }, + { + "epoch": 5.61, + "learning_rate": 4.318311815181587e-06, + "loss": 0.0054, + "step": 42847 + }, + { + "epoch": 5.61, + "learning_rate": 4.31787541230054e-06, + "loss": 0.0057, + "step": 42848 + }, + { + "epoch": 5.61, + "learning_rate": 4.317439025400164e-06, + "loss": 0.0067, + "step": 42849 + }, + { + "epoch": 5.61, + "learning_rate": 4.31700265448169e-06, + "loss": 0.007, + "step": 42850 + }, + { + "epoch": 5.61, + "learning_rate": 4.316566299546339e-06, + "loss": 0.0053, + "step": 42851 + }, + { + "epoch": 5.61, + "learning_rate": 4.316129960595347e-06, + "loss": 0.0067, + "step": 42852 + }, + { + "epoch": 5.61, + "learning_rate": 4.3156936376299356e-06, + "loss": 0.006, + "step": 42853 + }, + { + "epoch": 5.61, + "learning_rate": 4.315257330651329e-06, + "loss": 0.0062, + "step": 42854 + }, + { + "epoch": 5.61, + "learning_rate": 4.314821039660764e-06, + "loss": 0.0057, + "step": 42855 + }, + { + "epoch": 5.61, + "learning_rate": 4.3143847646594574e-06, + "loss": 0.0058, + "step": 42856 + }, + { + "epoch": 5.61, + "learning_rate": 4.313948505648644e-06, + "loss": 0.0051, + "step": 42857 + }, + { + "epoch": 5.61, + "learning_rate": 4.313512262629548e-06, + "loss": 0.0086, + "step": 42858 + }, + { + "epoch": 5.61, + "learning_rate": 4.313076035603393e-06, + "loss": 0.0066, + "step": 42859 + }, + { + "epoch": 5.61, + "learning_rate": 4.3126398245714114e-06, + "loss": 0.0059, + "step": 42860 + }, + { + "epoch": 5.61, + "learning_rate": 4.312203629534828e-06, + "loss": 0.0048, + "step": 42861 + }, + { + "epoch": 5.61, + "learning_rate": 4.311767450494866e-06, + "loss": 0.0071, + "step": 42862 + }, + { + "epoch": 5.61, + "learning_rate": 4.311331287452758e-06, + "loss": 0.0056, + "step": 42863 + }, + { + "epoch": 5.61, + "learning_rate": 4.310895140409728e-06, + "loss": 0.0071, + "step": 42864 + }, + { + "epoch": 5.62, + "learning_rate": 4.310459009367002e-06, + "loss": 0.0051, + "step": 42865 + }, + { + "epoch": 5.62, + "learning_rate": 4.310022894325804e-06, + "loss": 0.0054, + "step": 42866 + }, + { + "epoch": 5.62, + "learning_rate": 4.3095867952873626e-06, + "loss": 0.0061, + "step": 42867 + }, + { + "epoch": 5.62, + "learning_rate": 4.30915071225291e-06, + "loss": 0.0074, + "step": 42868 + }, + { + "epoch": 5.62, + "learning_rate": 4.308714645223667e-06, + "loss": 0.0071, + "step": 42869 + }, + { + "epoch": 5.62, + "learning_rate": 4.308278594200859e-06, + "loss": 0.0053, + "step": 42870 + }, + { + "epoch": 5.62, + "learning_rate": 4.307842559185717e-06, + "loss": 0.0066, + "step": 42871 + }, + { + "epoch": 5.62, + "learning_rate": 4.307406540179464e-06, + "loss": 0.0096, + "step": 42872 + }, + { + "epoch": 5.62, + "learning_rate": 4.306970537183328e-06, + "loss": 0.0066, + "step": 42873 + }, + { + "epoch": 5.62, + "learning_rate": 4.30653455019853e-06, + "loss": 0.0053, + "step": 42874 + }, + { + "epoch": 5.62, + "learning_rate": 4.3060985792263045e-06, + "loss": 0.0102, + "step": 42875 + }, + { + "epoch": 5.62, + "learning_rate": 4.305662624267873e-06, + "loss": 0.005, + "step": 42876 + }, + { + "epoch": 5.62, + "learning_rate": 4.3052266853244585e-06, + "loss": 0.0049, + "step": 42877 + }, + { + "epoch": 5.62, + "learning_rate": 4.3047907623972915e-06, + "loss": 0.0077, + "step": 42878 + }, + { + "epoch": 5.62, + "learning_rate": 4.3043548554876005e-06, + "loss": 0.0076, + "step": 42879 + }, + { + "epoch": 5.62, + "learning_rate": 4.303918964596607e-06, + "loss": 0.0066, + "step": 42880 + }, + { + "epoch": 5.62, + "learning_rate": 4.30348308972554e-06, + "loss": 0.0081, + "step": 42881 + }, + { + "epoch": 5.62, + "learning_rate": 4.3030472308756175e-06, + "loss": 0.0075, + "step": 42882 + }, + { + "epoch": 5.62, + "learning_rate": 4.302611388048077e-06, + "loss": 0.0061, + "step": 42883 + }, + { + "epoch": 5.62, + "learning_rate": 4.302175561244138e-06, + "loss": 0.0063, + "step": 42884 + }, + { + "epoch": 5.62, + "learning_rate": 4.301739750465023e-06, + "loss": 0.0062, + "step": 42885 + }, + { + "epoch": 5.62, + "learning_rate": 4.301303955711965e-06, + "loss": 0.0058, + "step": 42886 + }, + { + "epoch": 5.62, + "learning_rate": 4.300868176986186e-06, + "loss": 0.0069, + "step": 42887 + }, + { + "epoch": 5.62, + "learning_rate": 4.300432414288912e-06, + "loss": 0.0076, + "step": 42888 + }, + { + "epoch": 5.62, + "learning_rate": 4.2999966676213616e-06, + "loss": 0.0063, + "step": 42889 + }, + { + "epoch": 5.62, + "learning_rate": 4.299560936984773e-06, + "loss": 0.0076, + "step": 42890 + }, + { + "epoch": 5.62, + "learning_rate": 4.299125222380367e-06, + "loss": 0.0059, + "step": 42891 + }, + { + "epoch": 5.62, + "learning_rate": 4.298689523809367e-06, + "loss": 0.0065, + "step": 42892 + }, + { + "epoch": 5.62, + "learning_rate": 4.298253841272996e-06, + "loss": 0.0072, + "step": 42893 + }, + { + "epoch": 5.62, + "learning_rate": 4.297818174772485e-06, + "loss": 0.0053, + "step": 42894 + }, + { + "epoch": 5.62, + "learning_rate": 4.297382524309057e-06, + "loss": 0.0069, + "step": 42895 + }, + { + "epoch": 5.62, + "learning_rate": 4.296946889883939e-06, + "loss": 0.0074, + "step": 42896 + }, + { + "epoch": 5.62, + "learning_rate": 4.296511271498348e-06, + "loss": 0.008, + "step": 42897 + }, + { + "epoch": 5.62, + "learning_rate": 4.29607566915352e-06, + "loss": 0.0063, + "step": 42898 + }, + { + "epoch": 5.62, + "learning_rate": 4.295640082850675e-06, + "loss": 0.0057, + "step": 42899 + }, + { + "epoch": 5.62, + "learning_rate": 4.2952045125910344e-06, + "loss": 0.0068, + "step": 42900 + }, + { + "epoch": 5.62, + "learning_rate": 4.294768958375829e-06, + "loss": 0.0081, + "step": 42901 + }, + { + "epoch": 5.62, + "learning_rate": 4.294333420206285e-06, + "loss": 0.0074, + "step": 42902 + }, + { + "epoch": 5.62, + "learning_rate": 4.2938978980836245e-06, + "loss": 0.0084, + "step": 42903 + }, + { + "epoch": 5.62, + "learning_rate": 4.293462392009068e-06, + "loss": 0.0047, + "step": 42904 + }, + { + "epoch": 5.62, + "learning_rate": 4.2930269019838486e-06, + "loss": 0.0083, + "step": 42905 + }, + { + "epoch": 5.62, + "learning_rate": 4.292591428009188e-06, + "loss": 0.0069, + "step": 42906 + }, + { + "epoch": 5.62, + "learning_rate": 4.292155970086308e-06, + "loss": 0.0094, + "step": 42907 + }, + { + "epoch": 5.62, + "learning_rate": 4.291720528216433e-06, + "loss": 0.0054, + "step": 42908 + }, + { + "epoch": 5.62, + "learning_rate": 4.291285102400794e-06, + "loss": 0.006, + "step": 42909 + }, + { + "epoch": 5.62, + "learning_rate": 4.290849692640611e-06, + "loss": 0.0079, + "step": 42910 + }, + { + "epoch": 5.62, + "learning_rate": 4.290414298937105e-06, + "loss": 0.0063, + "step": 42911 + }, + { + "epoch": 5.62, + "learning_rate": 4.289978921291507e-06, + "loss": 0.0064, + "step": 42912 + }, + { + "epoch": 5.62, + "learning_rate": 4.289543559705041e-06, + "loss": 0.0072, + "step": 42913 + }, + { + "epoch": 5.62, + "learning_rate": 4.2891082141789294e-06, + "loss": 0.0089, + "step": 42914 + }, + { + "epoch": 5.62, + "learning_rate": 4.288672884714397e-06, + "loss": 0.0052, + "step": 42915 + }, + { + "epoch": 5.62, + "learning_rate": 4.288237571312666e-06, + "loss": 0.0073, + "step": 42916 + }, + { + "epoch": 5.62, + "learning_rate": 4.2878022739749645e-06, + "loss": 0.0049, + "step": 42917 + }, + { + "epoch": 5.62, + "learning_rate": 4.287366992702516e-06, + "loss": 0.0078, + "step": 42918 + }, + { + "epoch": 5.62, + "learning_rate": 4.286931727496539e-06, + "loss": 0.0071, + "step": 42919 + }, + { + "epoch": 5.62, + "learning_rate": 4.286496478358267e-06, + "loss": 0.0053, + "step": 42920 + }, + { + "epoch": 5.62, + "learning_rate": 4.286061245288918e-06, + "loss": 0.0064, + "step": 42921 + }, + { + "epoch": 5.62, + "learning_rate": 4.285626028289716e-06, + "loss": 0.0068, + "step": 42922 + }, + { + "epoch": 5.62, + "learning_rate": 4.285190827361888e-06, + "loss": 0.0053, + "step": 42923 + }, + { + "epoch": 5.62, + "learning_rate": 4.2847556425066535e-06, + "loss": 0.0087, + "step": 42924 + }, + { + "epoch": 5.62, + "learning_rate": 4.284320473725243e-06, + "loss": 0.0052, + "step": 42925 + }, + { + "epoch": 5.62, + "learning_rate": 4.283885321018877e-06, + "loss": 0.0065, + "step": 42926 + }, + { + "epoch": 5.62, + "learning_rate": 4.283450184388776e-06, + "loss": 0.008, + "step": 42927 + }, + { + "epoch": 5.62, + "learning_rate": 4.28301506383617e-06, + "loss": 0.0072, + "step": 42928 + }, + { + "epoch": 5.62, + "learning_rate": 4.2825799593622795e-06, + "loss": 0.0094, + "step": 42929 + }, + { + "epoch": 5.62, + "learning_rate": 4.282144870968328e-06, + "loss": 0.0055, + "step": 42930 + }, + { + "epoch": 5.62, + "learning_rate": 4.281709798655537e-06, + "loss": 0.0072, + "step": 42931 + }, + { + "epoch": 5.62, + "learning_rate": 4.281274742425135e-06, + "loss": 0.0074, + "step": 42932 + }, + { + "epoch": 5.62, + "learning_rate": 4.280839702278341e-06, + "loss": 0.0066, + "step": 42933 + }, + { + "epoch": 5.62, + "learning_rate": 4.280404678216383e-06, + "loss": 0.0072, + "step": 42934 + }, + { + "epoch": 5.62, + "learning_rate": 4.27996967024048e-06, + "loss": 0.0072, + "step": 42935 + }, + { + "epoch": 5.62, + "learning_rate": 4.279534678351861e-06, + "loss": 0.0046, + "step": 42936 + }, + { + "epoch": 5.62, + "learning_rate": 4.279099702551746e-06, + "loss": 0.0058, + "step": 42937 + }, + { + "epoch": 5.62, + "learning_rate": 4.278664742841358e-06, + "loss": 0.0054, + "step": 42938 + }, + { + "epoch": 5.62, + "learning_rate": 4.278229799221918e-06, + "loss": 0.0078, + "step": 42939 + }, + { + "epoch": 5.62, + "learning_rate": 4.277794871694655e-06, + "loss": 0.0056, + "step": 42940 + }, + { + "epoch": 5.62, + "learning_rate": 4.277359960260788e-06, + "loss": 0.0068, + "step": 42941 + }, + { + "epoch": 5.63, + "learning_rate": 4.27692506492154e-06, + "loss": 0.0064, + "step": 42942 + }, + { + "epoch": 5.63, + "learning_rate": 4.276490185678137e-06, + "loss": 0.0081, + "step": 42943 + }, + { + "epoch": 5.63, + "learning_rate": 4.276055322531799e-06, + "loss": 0.0053, + "step": 42944 + }, + { + "epoch": 5.63, + "learning_rate": 4.275620475483752e-06, + "loss": 0.0055, + "step": 42945 + }, + { + "epoch": 5.63, + "learning_rate": 4.275185644535219e-06, + "loss": 0.0085, + "step": 42946 + }, + { + "epoch": 5.63, + "learning_rate": 4.2747508296874184e-06, + "loss": 0.0077, + "step": 42947 + }, + { + "epoch": 5.63, + "learning_rate": 4.2743160309415786e-06, + "loss": 0.0075, + "step": 42948 + }, + { + "epoch": 5.63, + "learning_rate": 4.273881248298921e-06, + "loss": 0.0057, + "step": 42949 + }, + { + "epoch": 5.63, + "learning_rate": 4.2734464817606635e-06, + "loss": 0.0081, + "step": 42950 + }, + { + "epoch": 5.63, + "learning_rate": 4.273011731328036e-06, + "loss": 0.0086, + "step": 42951 + }, + { + "epoch": 5.63, + "learning_rate": 4.272576997002258e-06, + "loss": 0.0055, + "step": 42952 + }, + { + "epoch": 5.63, + "learning_rate": 4.272142278784552e-06, + "loss": 0.0079, + "step": 42953 + }, + { + "epoch": 5.63, + "learning_rate": 4.271707576676139e-06, + "loss": 0.0083, + "step": 42954 + }, + { + "epoch": 5.63, + "learning_rate": 4.271272890678242e-06, + "loss": 0.0106, + "step": 42955 + }, + { + "epoch": 5.63, + "learning_rate": 4.2708382207920884e-06, + "loss": 0.0077, + "step": 42956 + }, + { + "epoch": 5.63, + "learning_rate": 4.270403567018897e-06, + "loss": 0.009, + "step": 42957 + }, + { + "epoch": 5.63, + "learning_rate": 4.269968929359888e-06, + "loss": 0.0084, + "step": 42958 + }, + { + "epoch": 5.63, + "learning_rate": 4.269534307816289e-06, + "loss": 0.0061, + "step": 42959 + }, + { + "epoch": 5.63, + "learning_rate": 4.269099702389319e-06, + "loss": 0.0078, + "step": 42960 + }, + { + "epoch": 5.63, + "learning_rate": 4.268665113080202e-06, + "loss": 0.0068, + "step": 42961 + }, + { + "epoch": 5.63, + "learning_rate": 4.2682305398901545e-06, + "loss": 0.0061, + "step": 42962 + }, + { + "epoch": 5.63, + "learning_rate": 4.267795982820408e-06, + "loss": 0.0094, + "step": 42963 + }, + { + "epoch": 5.63, + "learning_rate": 4.26736144187218e-06, + "loss": 0.0065, + "step": 42964 + }, + { + "epoch": 5.63, + "learning_rate": 4.266926917046688e-06, + "loss": 0.0081, + "step": 42965 + }, + { + "epoch": 5.63, + "learning_rate": 4.26649240834516e-06, + "loss": 0.0081, + "step": 42966 + }, + { + "epoch": 5.63, + "learning_rate": 4.26605791576882e-06, + "loss": 0.0092, + "step": 42967 + }, + { + "epoch": 5.63, + "learning_rate": 4.265623439318887e-06, + "loss": 0.0061, + "step": 42968 + }, + { + "epoch": 5.63, + "learning_rate": 4.265188978996578e-06, + "loss": 0.0053, + "step": 42969 + }, + { + "epoch": 5.63, + "learning_rate": 4.2647545348031234e-06, + "loss": 0.0036, + "step": 42970 + }, + { + "epoch": 5.63, + "learning_rate": 4.264320106739743e-06, + "loss": 0.006, + "step": 42971 + }, + { + "epoch": 5.63, + "learning_rate": 4.263885694807656e-06, + "loss": 0.0076, + "step": 42972 + }, + { + "epoch": 5.63, + "learning_rate": 4.26345129900808e-06, + "loss": 0.0095, + "step": 42973 + }, + { + "epoch": 5.63, + "learning_rate": 4.263016919342247e-06, + "loss": 0.0076, + "step": 42974 + }, + { + "epoch": 5.63, + "learning_rate": 4.262582555811373e-06, + "loss": 0.006, + "step": 42975 + }, + { + "epoch": 5.63, + "learning_rate": 4.262148208416676e-06, + "loss": 0.0073, + "step": 42976 + }, + { + "epoch": 5.63, + "learning_rate": 4.2617138771593816e-06, + "loss": 0.0062, + "step": 42977 + }, + { + "epoch": 5.63, + "learning_rate": 4.2612795620407155e-06, + "loss": 0.0058, + "step": 42978 + }, + { + "epoch": 5.63, + "learning_rate": 4.260845263061895e-06, + "loss": 0.0076, + "step": 42979 + }, + { + "epoch": 5.63, + "learning_rate": 4.260410980224141e-06, + "loss": 0.0062, + "step": 42980 + }, + { + "epoch": 5.63, + "learning_rate": 4.259976713528672e-06, + "loss": 0.0051, + "step": 42981 + }, + { + "epoch": 5.63, + "learning_rate": 4.259542462976717e-06, + "loss": 0.0069, + "step": 42982 + }, + { + "epoch": 5.63, + "learning_rate": 4.259108228569492e-06, + "loss": 0.0055, + "step": 42983 + }, + { + "epoch": 5.63, + "learning_rate": 4.258674010308217e-06, + "loss": 0.0067, + "step": 42984 + }, + { + "epoch": 5.63, + "learning_rate": 4.258239808194118e-06, + "loss": 0.008, + "step": 42985 + }, + { + "epoch": 5.63, + "learning_rate": 4.257805622228415e-06, + "loss": 0.0065, + "step": 42986 + }, + { + "epoch": 5.63, + "learning_rate": 4.257371452412323e-06, + "loss": 0.0059, + "step": 42987 + }, + { + "epoch": 5.63, + "learning_rate": 4.256937298747072e-06, + "loss": 0.0076, + "step": 42988 + }, + { + "epoch": 5.63, + "learning_rate": 4.256503161233876e-06, + "loss": 0.0046, + "step": 42989 + }, + { + "epoch": 5.63, + "learning_rate": 4.2560690398739625e-06, + "loss": 0.0096, + "step": 42990 + }, + { + "epoch": 5.63, + "learning_rate": 4.255634934668549e-06, + "loss": 0.0046, + "step": 42991 + }, + { + "epoch": 5.63, + "learning_rate": 4.255200845618852e-06, + "loss": 0.0073, + "step": 42992 + }, + { + "epoch": 5.63, + "learning_rate": 4.254766772726101e-06, + "loss": 0.0075, + "step": 42993 + }, + { + "epoch": 5.63, + "learning_rate": 4.2543327159915125e-06, + "loss": 0.0065, + "step": 42994 + }, + { + "epoch": 5.63, + "learning_rate": 4.253898675416307e-06, + "loss": 0.0058, + "step": 42995 + }, + { + "epoch": 5.63, + "learning_rate": 4.253464651001702e-06, + "loss": 0.0068, + "step": 42996 + }, + { + "epoch": 5.63, + "learning_rate": 4.253030642748925e-06, + "loss": 0.0047, + "step": 42997 + }, + { + "epoch": 5.63, + "learning_rate": 4.25259665065919e-06, + "loss": 0.006, + "step": 42998 + }, + { + "epoch": 5.63, + "learning_rate": 4.252162674733724e-06, + "loss": 0.0061, + "step": 42999 + }, + { + "epoch": 5.63, + "learning_rate": 4.2517287149737415e-06, + "loss": 0.0084, + "step": 43000 + }, + { + "epoch": 5.63, + "learning_rate": 4.251294771380469e-06, + "loss": 0.008, + "step": 43001 + }, + { + "epoch": 5.63, + "learning_rate": 4.2508608439551234e-06, + "loss": 0.0048, + "step": 43002 + }, + { + "epoch": 5.63, + "learning_rate": 4.2504269326989266e-06, + "loss": 0.0059, + "step": 43003 + }, + { + "epoch": 5.63, + "learning_rate": 4.2499930376130935e-06, + "loss": 0.0054, + "step": 43004 + }, + { + "epoch": 5.63, + "learning_rate": 4.249559158698852e-06, + "loss": 0.0093, + "step": 43005 + }, + { + "epoch": 5.63, + "learning_rate": 4.2491252959574205e-06, + "loss": 0.0098, + "step": 43006 + }, + { + "epoch": 5.63, + "learning_rate": 4.248691449390013e-06, + "loss": 0.0068, + "step": 43007 + }, + { + "epoch": 5.63, + "learning_rate": 4.248257618997859e-06, + "loss": 0.0069, + "step": 43008 + }, + { + "epoch": 5.63, + "learning_rate": 4.2478238047821695e-06, + "loss": 0.0058, + "step": 43009 + }, + { + "epoch": 5.63, + "learning_rate": 4.247390006744174e-06, + "loss": 0.0053, + "step": 43010 + }, + { + "epoch": 5.63, + "learning_rate": 4.246956224885087e-06, + "loss": 0.0074, + "step": 43011 + }, + { + "epoch": 5.63, + "learning_rate": 4.246522459206126e-06, + "loss": 0.008, + "step": 43012 + }, + { + "epoch": 5.63, + "learning_rate": 4.246088709708518e-06, + "loss": 0.0099, + "step": 43013 + }, + { + "epoch": 5.63, + "learning_rate": 4.245654976393478e-06, + "loss": 0.0057, + "step": 43014 + }, + { + "epoch": 5.63, + "learning_rate": 4.245221259262224e-06, + "loss": 0.0081, + "step": 43015 + }, + { + "epoch": 5.63, + "learning_rate": 4.244787558315983e-06, + "loss": 0.0067, + "step": 43016 + }, + { + "epoch": 5.63, + "learning_rate": 4.24435387355597e-06, + "loss": 0.0053, + "step": 43017 + }, + { + "epoch": 5.64, + "learning_rate": 4.2439202049834046e-06, + "loss": 0.0098, + "step": 43018 + }, + { + "epoch": 5.64, + "learning_rate": 4.243486552599505e-06, + "loss": 0.0069, + "step": 43019 + }, + { + "epoch": 5.64, + "learning_rate": 4.2430529164054916e-06, + "loss": 0.0082, + "step": 43020 + }, + { + "epoch": 5.64, + "learning_rate": 4.242619296402589e-06, + "loss": 0.0068, + "step": 43021 + }, + { + "epoch": 5.64, + "learning_rate": 4.242185692592013e-06, + "loss": 0.0076, + "step": 43022 + }, + { + "epoch": 5.64, + "learning_rate": 4.2417521049749795e-06, + "loss": 0.0074, + "step": 43023 + }, + { + "epoch": 5.64, + "learning_rate": 4.241318533552717e-06, + "loss": 0.0066, + "step": 43024 + }, + { + "epoch": 5.64, + "learning_rate": 4.240884978326437e-06, + "loss": 0.0052, + "step": 43025 + }, + { + "epoch": 5.64, + "learning_rate": 4.2404514392973586e-06, + "loss": 0.0088, + "step": 43026 + }, + { + "epoch": 5.64, + "learning_rate": 4.240017916466708e-06, + "loss": 0.0081, + "step": 43027 + }, + { + "epoch": 5.64, + "learning_rate": 4.239584409835699e-06, + "loss": 0.0048, + "step": 43028 + }, + { + "epoch": 5.64, + "learning_rate": 4.239150919405554e-06, + "loss": 0.0088, + "step": 43029 + }, + { + "epoch": 5.64, + "learning_rate": 4.238717445177485e-06, + "loss": 0.0081, + "step": 43030 + }, + { + "epoch": 5.64, + "learning_rate": 4.238283987152717e-06, + "loss": 0.0066, + "step": 43031 + }, + { + "epoch": 5.64, + "learning_rate": 4.237850545332472e-06, + "loss": 0.0075, + "step": 43032 + }, + { + "epoch": 5.64, + "learning_rate": 4.237417119717965e-06, + "loss": 0.0076, + "step": 43033 + }, + { + "epoch": 5.64, + "learning_rate": 4.236983710310412e-06, + "loss": 0.0054, + "step": 43034 + }, + { + "epoch": 5.64, + "learning_rate": 4.236550317111039e-06, + "loss": 0.0071, + "step": 43035 + }, + { + "epoch": 5.64, + "learning_rate": 4.236116940121062e-06, + "loss": 0.0055, + "step": 43036 + }, + { + "epoch": 5.64, + "learning_rate": 4.235683579341699e-06, + "loss": 0.0054, + "step": 43037 + }, + { + "epoch": 5.64, + "learning_rate": 4.235250234774166e-06, + "loss": 0.0086, + "step": 43038 + }, + { + "epoch": 5.64, + "learning_rate": 4.234816906419687e-06, + "loss": 0.007, + "step": 43039 + }, + { + "epoch": 5.64, + "learning_rate": 4.234383594279479e-06, + "loss": 0.0091, + "step": 43040 + }, + { + "epoch": 5.64, + "learning_rate": 4.2339502983547556e-06, + "loss": 0.0095, + "step": 43041 + }, + { + "epoch": 5.64, + "learning_rate": 4.233517018646741e-06, + "loss": 0.008, + "step": 43042 + }, + { + "epoch": 5.64, + "learning_rate": 4.233083755156656e-06, + "loss": 0.0061, + "step": 43043 + }, + { + "epoch": 5.64, + "learning_rate": 4.232650507885716e-06, + "loss": 0.0074, + "step": 43044 + }, + { + "epoch": 5.64, + "learning_rate": 4.232217276835139e-06, + "loss": 0.0058, + "step": 43045 + }, + { + "epoch": 5.64, + "learning_rate": 4.23178406200614e-06, + "loss": 0.0058, + "step": 43046 + }, + { + "epoch": 5.64, + "learning_rate": 4.231350863399945e-06, + "loss": 0.0074, + "step": 43047 + }, + { + "epoch": 5.64, + "learning_rate": 4.230917681017769e-06, + "loss": 0.0086, + "step": 43048 + }, + { + "epoch": 5.64, + "learning_rate": 4.230484514860825e-06, + "loss": 0.0065, + "step": 43049 + }, + { + "epoch": 5.64, + "learning_rate": 4.230051364930341e-06, + "loss": 0.0056, + "step": 43050 + }, + { + "epoch": 5.64, + "learning_rate": 4.22961823122753e-06, + "loss": 0.0086, + "step": 43051 + }, + { + "epoch": 5.64, + "learning_rate": 4.22918511375361e-06, + "loss": 0.0078, + "step": 43052 + }, + { + "epoch": 5.64, + "learning_rate": 4.228752012509796e-06, + "loss": 0.0077, + "step": 43053 + }, + { + "epoch": 5.64, + "learning_rate": 4.228318927497309e-06, + "loss": 0.0069, + "step": 43054 + }, + { + "epoch": 5.64, + "learning_rate": 4.2278858587173725e-06, + "loss": 0.0102, + "step": 43055 + }, + { + "epoch": 5.64, + "learning_rate": 4.2274528061712e-06, + "loss": 0.0082, + "step": 43056 + }, + { + "epoch": 5.64, + "learning_rate": 4.227019769860003e-06, + "loss": 0.0047, + "step": 43057 + }, + { + "epoch": 5.64, + "learning_rate": 4.226586749785011e-06, + "loss": 0.0041, + "step": 43058 + }, + { + "epoch": 5.64, + "learning_rate": 4.2261537459474365e-06, + "loss": 0.0059, + "step": 43059 + }, + { + "epoch": 5.64, + "learning_rate": 4.225720758348496e-06, + "loss": 0.0057, + "step": 43060 + }, + { + "epoch": 5.64, + "learning_rate": 4.225287786989406e-06, + "loss": 0.007, + "step": 43061 + }, + { + "epoch": 5.64, + "learning_rate": 4.224854831871389e-06, + "loss": 0.0071, + "step": 43062 + }, + { + "epoch": 5.64, + "learning_rate": 4.224421892995662e-06, + "loss": 0.0062, + "step": 43063 + }, + { + "epoch": 5.64, + "learning_rate": 4.223988970363435e-06, + "loss": 0.0056, + "step": 43064 + }, + { + "epoch": 5.64, + "learning_rate": 4.223556063975932e-06, + "loss": 0.0089, + "step": 43065 + }, + { + "epoch": 5.64, + "learning_rate": 4.223123173834374e-06, + "loss": 0.0083, + "step": 43066 + }, + { + "epoch": 5.64, + "learning_rate": 4.222690299939975e-06, + "loss": 0.0082, + "step": 43067 + }, + { + "epoch": 5.64, + "learning_rate": 4.222257442293951e-06, + "loss": 0.0067, + "step": 43068 + }, + { + "epoch": 5.64, + "learning_rate": 4.221824600897517e-06, + "loss": 0.0087, + "step": 43069 + }, + { + "epoch": 5.64, + "learning_rate": 4.221391775751897e-06, + "loss": 0.0054, + "step": 43070 + }, + { + "epoch": 5.64, + "learning_rate": 4.2209589668583065e-06, + "loss": 0.0059, + "step": 43071 + }, + { + "epoch": 5.64, + "learning_rate": 4.220526174217955e-06, + "loss": 0.0059, + "step": 43072 + }, + { + "epoch": 5.64, + "learning_rate": 4.220093397832071e-06, + "loss": 0.005, + "step": 43073 + }, + { + "epoch": 5.64, + "learning_rate": 4.219660637701866e-06, + "loss": 0.0038, + "step": 43074 + }, + { + "epoch": 5.64, + "learning_rate": 4.2192278938285545e-06, + "loss": 0.0068, + "step": 43075 + }, + { + "epoch": 5.64, + "learning_rate": 4.218795166213356e-06, + "loss": 0.0072, + "step": 43076 + }, + { + "epoch": 5.64, + "learning_rate": 4.218362454857492e-06, + "loss": 0.0081, + "step": 43077 + }, + { + "epoch": 5.64, + "learning_rate": 4.217929759762176e-06, + "loss": 0.0074, + "step": 43078 + }, + { + "epoch": 5.64, + "learning_rate": 4.2174970809286245e-06, + "loss": 0.0065, + "step": 43079 + }, + { + "epoch": 5.64, + "learning_rate": 4.217064418358051e-06, + "loss": 0.007, + "step": 43080 + }, + { + "epoch": 5.64, + "learning_rate": 4.216631772051679e-06, + "loss": 0.0069, + "step": 43081 + }, + { + "epoch": 5.64, + "learning_rate": 4.216199142010723e-06, + "loss": 0.0055, + "step": 43082 + }, + { + "epoch": 5.64, + "learning_rate": 4.215766528236399e-06, + "loss": 0.0076, + "step": 43083 + }, + { + "epoch": 5.64, + "learning_rate": 4.215333930729919e-06, + "loss": 0.0068, + "step": 43084 + }, + { + "epoch": 5.64, + "learning_rate": 4.214901349492509e-06, + "loss": 0.0093, + "step": 43085 + }, + { + "epoch": 5.64, + "learning_rate": 4.214468784525376e-06, + "loss": 0.009, + "step": 43086 + }, + { + "epoch": 5.64, + "learning_rate": 4.214036235829746e-06, + "loss": 0.0049, + "step": 43087 + }, + { + "epoch": 5.64, + "learning_rate": 4.213603703406827e-06, + "loss": 0.0078, + "step": 43088 + }, + { + "epoch": 5.64, + "learning_rate": 4.213171187257843e-06, + "loss": 0.0051, + "step": 43089 + }, + { + "epoch": 5.64, + "learning_rate": 4.212738687384007e-06, + "loss": 0.0057, + "step": 43090 + }, + { + "epoch": 5.64, + "learning_rate": 4.212306203786531e-06, + "loss": 0.0073, + "step": 43091 + }, + { + "epoch": 5.64, + "learning_rate": 4.2118737364666405e-06, + "loss": 0.0051, + "step": 43092 + }, + { + "epoch": 5.64, + "learning_rate": 4.211441285425546e-06, + "loss": 0.007, + "step": 43093 + }, + { + "epoch": 5.65, + "learning_rate": 4.211008850664466e-06, + "loss": 0.0044, + "step": 43094 + }, + { + "epoch": 5.65, + "learning_rate": 4.21057643218461e-06, + "loss": 0.0064, + "step": 43095 + }, + { + "epoch": 5.65, + "learning_rate": 4.2101440299872034e-06, + "loss": 0.0058, + "step": 43096 + }, + { + "epoch": 5.65, + "learning_rate": 4.209711644073454e-06, + "loss": 0.0068, + "step": 43097 + }, + { + "epoch": 5.65, + "learning_rate": 4.209279274444588e-06, + "loss": 0.006, + "step": 43098 + }, + { + "epoch": 5.65, + "learning_rate": 4.2088469211018105e-06, + "loss": 0.0065, + "step": 43099 + }, + { + "epoch": 5.65, + "learning_rate": 4.208414584046346e-06, + "loss": 0.009, + "step": 43100 + }, + { + "epoch": 5.65, + "learning_rate": 4.207982263279409e-06, + "loss": 0.0054, + "step": 43101 + }, + { + "epoch": 5.65, + "learning_rate": 4.2075499588022115e-06, + "loss": 0.0067, + "step": 43102 + }, + { + "epoch": 5.65, + "learning_rate": 4.207117670615968e-06, + "loss": 0.0064, + "step": 43103 + }, + { + "epoch": 5.65, + "learning_rate": 4.206685398721901e-06, + "loss": 0.0076, + "step": 43104 + }, + { + "epoch": 5.65, + "learning_rate": 4.206253143121224e-06, + "loss": 0.0057, + "step": 43105 + }, + { + "epoch": 5.65, + "learning_rate": 4.205820903815148e-06, + "loss": 0.0059, + "step": 43106 + }, + { + "epoch": 5.65, + "learning_rate": 4.205388680804894e-06, + "loss": 0.0085, + "step": 43107 + }, + { + "epoch": 5.65, + "learning_rate": 4.204956474091673e-06, + "loss": 0.0085, + "step": 43108 + }, + { + "epoch": 5.65, + "learning_rate": 4.2045242836767076e-06, + "loss": 0.0074, + "step": 43109 + }, + { + "epoch": 5.65, + "learning_rate": 4.204092109561209e-06, + "loss": 0.0061, + "step": 43110 + }, + { + "epoch": 5.65, + "learning_rate": 4.203659951746389e-06, + "loss": 0.0047, + "step": 43111 + }, + { + "epoch": 5.65, + "learning_rate": 4.203227810233469e-06, + "loss": 0.0063, + "step": 43112 + }, + { + "epoch": 5.65, + "learning_rate": 4.202795685023663e-06, + "loss": 0.0065, + "step": 43113 + }, + { + "epoch": 5.65, + "learning_rate": 4.202363576118181e-06, + "loss": 0.0072, + "step": 43114 + }, + { + "epoch": 5.65, + "learning_rate": 4.201931483518248e-06, + "loss": 0.0087, + "step": 43115 + }, + { + "epoch": 5.65, + "learning_rate": 4.201499407225072e-06, + "loss": 0.0058, + "step": 43116 + }, + { + "epoch": 5.65, + "learning_rate": 4.201067347239871e-06, + "loss": 0.0063, + "step": 43117 + }, + { + "epoch": 5.65, + "learning_rate": 4.200635303563856e-06, + "loss": 0.0049, + "step": 43118 + }, + { + "epoch": 5.65, + "learning_rate": 4.200203276198246e-06, + "loss": 0.0068, + "step": 43119 + }, + { + "epoch": 5.65, + "learning_rate": 4.199771265144258e-06, + "loss": 0.007, + "step": 43120 + }, + { + "epoch": 5.65, + "learning_rate": 4.199339270403105e-06, + "loss": 0.0077, + "step": 43121 + }, + { + "epoch": 5.65, + "learning_rate": 4.198907291975996e-06, + "loss": 0.006, + "step": 43122 + }, + { + "epoch": 5.65, + "learning_rate": 4.198475329864157e-06, + "loss": 0.007, + "step": 43123 + }, + { + "epoch": 5.65, + "learning_rate": 4.198043384068796e-06, + "loss": 0.0085, + "step": 43124 + }, + { + "epoch": 5.65, + "learning_rate": 4.19761145459113e-06, + "loss": 0.0048, + "step": 43125 + }, + { + "epoch": 5.65, + "learning_rate": 4.197179541432369e-06, + "loss": 0.007, + "step": 43126 + }, + { + "epoch": 5.65, + "learning_rate": 4.196747644593735e-06, + "loss": 0.0093, + "step": 43127 + }, + { + "epoch": 5.65, + "learning_rate": 4.196315764076439e-06, + "loss": 0.0092, + "step": 43128 + }, + { + "epoch": 5.65, + "learning_rate": 4.195883899881692e-06, + "loss": 0.0055, + "step": 43129 + }, + { + "epoch": 5.65, + "learning_rate": 4.195452052010713e-06, + "loss": 0.0058, + "step": 43130 + }, + { + "epoch": 5.65, + "learning_rate": 4.195020220464719e-06, + "loss": 0.0065, + "step": 43131 + }, + { + "epoch": 5.65, + "learning_rate": 4.194588405244923e-06, + "loss": 0.005, + "step": 43132 + }, + { + "epoch": 5.65, + "learning_rate": 4.194156606352536e-06, + "loss": 0.0086, + "step": 43133 + }, + { + "epoch": 5.65, + "learning_rate": 4.193724823788772e-06, + "loss": 0.0072, + "step": 43134 + }, + { + "epoch": 5.65, + "learning_rate": 4.1932930575548505e-06, + "loss": 0.0054, + "step": 43135 + }, + { + "epoch": 5.65, + "learning_rate": 4.192861307651983e-06, + "loss": 0.0089, + "step": 43136 + }, + { + "epoch": 5.65, + "learning_rate": 4.192429574081381e-06, + "loss": 0.0067, + "step": 43137 + }, + { + "epoch": 5.65, + "learning_rate": 4.1919978568442645e-06, + "loss": 0.0069, + "step": 43138 + }, + { + "epoch": 5.65, + "learning_rate": 4.1915661559418455e-06, + "loss": 0.0061, + "step": 43139 + }, + { + "epoch": 5.65, + "learning_rate": 4.191134471375334e-06, + "loss": 0.0076, + "step": 43140 + }, + { + "epoch": 5.65, + "learning_rate": 4.190702803145946e-06, + "loss": 0.0084, + "step": 43141 + }, + { + "epoch": 5.65, + "learning_rate": 4.1902711512549014e-06, + "loss": 0.0088, + "step": 43142 + }, + { + "epoch": 5.65, + "learning_rate": 4.18983951570341e-06, + "loss": 0.0039, + "step": 43143 + }, + { + "epoch": 5.65, + "learning_rate": 4.189407896492685e-06, + "loss": 0.0085, + "step": 43144 + }, + { + "epoch": 5.65, + "learning_rate": 4.1889762936239375e-06, + "loss": 0.0041, + "step": 43145 + }, + { + "epoch": 5.65, + "learning_rate": 4.188544707098389e-06, + "loss": 0.0079, + "step": 43146 + }, + { + "epoch": 5.65, + "learning_rate": 4.188113136917248e-06, + "loss": 0.0066, + "step": 43147 + }, + { + "epoch": 5.65, + "learning_rate": 4.18768158308173e-06, + "loss": 0.0078, + "step": 43148 + }, + { + "epoch": 5.65, + "learning_rate": 4.187250045593045e-06, + "loss": 0.0052, + "step": 43149 + }, + { + "epoch": 5.65, + "learning_rate": 4.1868185244524115e-06, + "loss": 0.0091, + "step": 43150 + }, + { + "epoch": 5.65, + "learning_rate": 4.186387019661039e-06, + "loss": 0.005, + "step": 43151 + }, + { + "epoch": 5.65, + "learning_rate": 4.185955531220146e-06, + "loss": 0.0082, + "step": 43152 + }, + { + "epoch": 5.65, + "learning_rate": 4.185524059130941e-06, + "loss": 0.0067, + "step": 43153 + }, + { + "epoch": 5.65, + "learning_rate": 4.185092603394642e-06, + "loss": 0.0076, + "step": 43154 + }, + { + "epoch": 5.65, + "learning_rate": 4.1846611640124625e-06, + "loss": 0.0069, + "step": 43155 + }, + { + "epoch": 5.65, + "learning_rate": 4.184229740985609e-06, + "loss": 0.0052, + "step": 43156 + }, + { + "epoch": 5.65, + "learning_rate": 4.1837983343153035e-06, + "loss": 0.0072, + "step": 43157 + }, + { + "epoch": 5.65, + "learning_rate": 4.183366944002755e-06, + "loss": 0.008, + "step": 43158 + }, + { + "epoch": 5.65, + "learning_rate": 4.182935570049178e-06, + "loss": 0.0058, + "step": 43159 + }, + { + "epoch": 5.65, + "learning_rate": 4.1825042124557815e-06, + "loss": 0.0065, + "step": 43160 + }, + { + "epoch": 5.65, + "learning_rate": 4.182072871223785e-06, + "loss": 0.0063, + "step": 43161 + }, + { + "epoch": 5.65, + "learning_rate": 4.181641546354395e-06, + "loss": 0.0056, + "step": 43162 + }, + { + "epoch": 5.65, + "learning_rate": 4.181210237848832e-06, + "loss": 0.007, + "step": 43163 + }, + { + "epoch": 5.65, + "learning_rate": 4.1807789457083025e-06, + "loss": 0.0061, + "step": 43164 + }, + { + "epoch": 5.65, + "learning_rate": 4.180347669934025e-06, + "loss": 0.0076, + "step": 43165 + }, + { + "epoch": 5.65, + "learning_rate": 4.179916410527211e-06, + "loss": 0.0049, + "step": 43166 + }, + { + "epoch": 5.65, + "learning_rate": 4.1794851674890705e-06, + "loss": 0.0061, + "step": 43167 + }, + { + "epoch": 5.65, + "learning_rate": 4.179053940820815e-06, + "loss": 0.0056, + "step": 43168 + }, + { + "epoch": 5.65, + "learning_rate": 4.178622730523664e-06, + "loss": 0.0102, + "step": 43169 + }, + { + "epoch": 5.65, + "learning_rate": 4.178191536598827e-06, + "loss": 0.0042, + "step": 43170 + }, + { + "epoch": 5.66, + "learning_rate": 4.1777603590475125e-06, + "loss": 0.0041, + "step": 43171 + }, + { + "epoch": 5.66, + "learning_rate": 4.177329197870939e-06, + "loss": 0.0083, + "step": 43172 + }, + { + "epoch": 5.66, + "learning_rate": 4.176898053070315e-06, + "loss": 0.0071, + "step": 43173 + }, + { + "epoch": 5.66, + "learning_rate": 4.176466924646859e-06, + "loss": 0.0055, + "step": 43174 + }, + { + "epoch": 5.66, + "learning_rate": 4.17603581260178e-06, + "loss": 0.0095, + "step": 43175 + }, + { + "epoch": 5.66, + "learning_rate": 4.175604716936286e-06, + "loss": 0.0049, + "step": 43176 + }, + { + "epoch": 5.66, + "learning_rate": 4.1751736376515964e-06, + "loss": 0.0056, + "step": 43177 + }, + { + "epoch": 5.66, + "learning_rate": 4.174742574748921e-06, + "loss": 0.0046, + "step": 43178 + }, + { + "epoch": 5.66, + "learning_rate": 4.174311528229469e-06, + "loss": 0.0072, + "step": 43179 + }, + { + "epoch": 5.66, + "learning_rate": 4.173880498094458e-06, + "loss": 0.0068, + "step": 43180 + }, + { + "epoch": 5.66, + "learning_rate": 4.173449484345099e-06, + "loss": 0.0062, + "step": 43181 + }, + { + "epoch": 5.66, + "learning_rate": 4.1730184869826015e-06, + "loss": 0.0075, + "step": 43182 + }, + { + "epoch": 5.66, + "learning_rate": 4.172587506008178e-06, + "loss": 0.0063, + "step": 43183 + }, + { + "epoch": 5.66, + "learning_rate": 4.17215654142304e-06, + "loss": 0.0051, + "step": 43184 + }, + { + "epoch": 5.66, + "learning_rate": 4.171725593228405e-06, + "loss": 0.0057, + "step": 43185 + }, + { + "epoch": 5.66, + "learning_rate": 4.171294661425483e-06, + "loss": 0.0055, + "step": 43186 + }, + { + "epoch": 5.66, + "learning_rate": 4.170863746015478e-06, + "loss": 0.0067, + "step": 43187 + }, + { + "epoch": 5.66, + "learning_rate": 4.170432846999615e-06, + "loss": 0.0056, + "step": 43188 + }, + { + "epoch": 5.66, + "learning_rate": 4.170001964379098e-06, + "loss": 0.0089, + "step": 43189 + }, + { + "epoch": 5.66, + "learning_rate": 4.16957109815514e-06, + "loss": 0.0059, + "step": 43190 + }, + { + "epoch": 5.66, + "learning_rate": 4.169140248328948e-06, + "loss": 0.0063, + "step": 43191 + }, + { + "epoch": 5.66, + "learning_rate": 4.168709414901743e-06, + "loss": 0.008, + "step": 43192 + }, + { + "epoch": 5.66, + "learning_rate": 4.168278597874732e-06, + "loss": 0.0081, + "step": 43193 + }, + { + "epoch": 5.66, + "learning_rate": 4.167847797249124e-06, + "loss": 0.006, + "step": 43194 + }, + { + "epoch": 5.66, + "learning_rate": 4.1674170130261325e-06, + "loss": 0.0086, + "step": 43195 + }, + { + "epoch": 5.66, + "learning_rate": 4.166986245206975e-06, + "loss": 0.0053, + "step": 43196 + }, + { + "epoch": 5.66, + "learning_rate": 4.166555493792858e-06, + "loss": 0.0066, + "step": 43197 + }, + { + "epoch": 5.66, + "learning_rate": 4.166124758784992e-06, + "loss": 0.006, + "step": 43198 + }, + { + "epoch": 5.66, + "learning_rate": 4.165694040184586e-06, + "loss": 0.0051, + "step": 43199 + }, + { + "epoch": 5.66, + "learning_rate": 4.165263337992859e-06, + "loss": 0.0067, + "step": 43200 + }, + { + "epoch": 5.66, + "learning_rate": 4.164832652211017e-06, + "loss": 0.0059, + "step": 43201 + }, + { + "epoch": 5.66, + "learning_rate": 4.1644019828402695e-06, + "loss": 0.0087, + "step": 43202 + }, + { + "epoch": 5.66, + "learning_rate": 4.163971329881834e-06, + "loss": 0.008, + "step": 43203 + }, + { + "epoch": 5.66, + "learning_rate": 4.163540693336919e-06, + "loss": 0.0056, + "step": 43204 + }, + { + "epoch": 5.66, + "learning_rate": 4.163110073206734e-06, + "loss": 0.0053, + "step": 43205 + }, + { + "epoch": 5.66, + "learning_rate": 4.162679469492485e-06, + "loss": 0.0069, + "step": 43206 + }, + { + "epoch": 5.66, + "learning_rate": 4.162248882195395e-06, + "loss": 0.006, + "step": 43207 + }, + { + "epoch": 5.66, + "learning_rate": 4.161818311316671e-06, + "loss": 0.0066, + "step": 43208 + }, + { + "epoch": 5.66, + "learning_rate": 4.1613877568575215e-06, + "loss": 0.0062, + "step": 43209 + }, + { + "epoch": 5.66, + "learning_rate": 4.160957218819154e-06, + "loss": 0.0061, + "step": 43210 + }, + { + "epoch": 5.66, + "learning_rate": 4.160526697202787e-06, + "loss": 0.0084, + "step": 43211 + }, + { + "epoch": 5.66, + "learning_rate": 4.160096192009627e-06, + "loss": 0.0055, + "step": 43212 + }, + { + "epoch": 5.66, + "learning_rate": 4.159665703240886e-06, + "loss": 0.0075, + "step": 43213 + }, + { + "epoch": 5.66, + "learning_rate": 4.159235230897771e-06, + "loss": 0.0059, + "step": 43214 + }, + { + "epoch": 5.66, + "learning_rate": 4.1588047749815e-06, + "loss": 0.0075, + "step": 43215 + }, + { + "epoch": 5.66, + "learning_rate": 4.1583743354932786e-06, + "loss": 0.008, + "step": 43216 + }, + { + "epoch": 5.66, + "learning_rate": 4.157943912434315e-06, + "loss": 0.0065, + "step": 43217 + }, + { + "epoch": 5.66, + "learning_rate": 4.157513505805823e-06, + "loss": 0.0079, + "step": 43218 + }, + { + "epoch": 5.66, + "learning_rate": 4.157083115609016e-06, + "loss": 0.0061, + "step": 43219 + }, + { + "epoch": 5.66, + "learning_rate": 4.156652741845102e-06, + "loss": 0.0056, + "step": 43220 + }, + { + "epoch": 5.66, + "learning_rate": 4.156222384515287e-06, + "loss": 0.0044, + "step": 43221 + }, + { + "epoch": 5.66, + "learning_rate": 4.15579204362079e-06, + "loss": 0.0079, + "step": 43222 + }, + { + "epoch": 5.66, + "learning_rate": 4.155361719162816e-06, + "loss": 0.0074, + "step": 43223 + }, + { + "epoch": 5.66, + "learning_rate": 4.154931411142576e-06, + "loss": 0.0067, + "step": 43224 + }, + { + "epoch": 5.66, + "learning_rate": 4.154501119561277e-06, + "loss": 0.0087, + "step": 43225 + }, + { + "epoch": 5.66, + "learning_rate": 4.154070844420136e-06, + "loss": 0.0066, + "step": 43226 + }, + { + "epoch": 5.66, + "learning_rate": 4.153640585720359e-06, + "loss": 0.0068, + "step": 43227 + }, + { + "epoch": 5.66, + "learning_rate": 4.153210343463152e-06, + "loss": 0.0092, + "step": 43228 + }, + { + "epoch": 5.66, + "learning_rate": 4.152780117649731e-06, + "loss": 0.0067, + "step": 43229 + }, + { + "epoch": 5.66, + "learning_rate": 4.1523499082813075e-06, + "loss": 0.004, + "step": 43230 + }, + { + "epoch": 5.66, + "learning_rate": 4.151919715359088e-06, + "loss": 0.006, + "step": 43231 + }, + { + "epoch": 5.66, + "learning_rate": 4.151489538884284e-06, + "loss": 0.0085, + "step": 43232 + }, + { + "epoch": 5.66, + "learning_rate": 4.1510593788581e-06, + "loss": 0.0081, + "step": 43233 + }, + { + "epoch": 5.66, + "learning_rate": 4.150629235281754e-06, + "loss": 0.0069, + "step": 43234 + }, + { + "epoch": 5.66, + "learning_rate": 4.150199108156451e-06, + "loss": 0.006, + "step": 43235 + }, + { + "epoch": 5.66, + "learning_rate": 4.149768997483399e-06, + "loss": 0.0087, + "step": 43236 + }, + { + "epoch": 5.66, + "learning_rate": 4.149338903263812e-06, + "loss": 0.0049, + "step": 43237 + }, + { + "epoch": 5.66, + "learning_rate": 4.148908825498899e-06, + "loss": 0.0085, + "step": 43238 + }, + { + "epoch": 5.66, + "learning_rate": 4.148478764189865e-06, + "loss": 0.0049, + "step": 43239 + }, + { + "epoch": 5.66, + "learning_rate": 4.148048719337926e-06, + "loss": 0.0055, + "step": 43240 + }, + { + "epoch": 5.66, + "learning_rate": 4.147618690944285e-06, + "loss": 0.0044, + "step": 43241 + }, + { + "epoch": 5.66, + "learning_rate": 4.147188679010158e-06, + "loss": 0.0071, + "step": 43242 + }, + { + "epoch": 5.66, + "learning_rate": 4.146758683536751e-06, + "loss": 0.0053, + "step": 43243 + }, + { + "epoch": 5.66, + "learning_rate": 4.146328704525271e-06, + "loss": 0.0038, + "step": 43244 + }, + { + "epoch": 5.66, + "learning_rate": 4.145898741976932e-06, + "loss": 0.0041, + "step": 43245 + }, + { + "epoch": 5.66, + "learning_rate": 4.145468795892942e-06, + "loss": 0.0058, + "step": 43246 + }, + { + "epoch": 5.67, + "learning_rate": 4.145038866274509e-06, + "loss": 0.0066, + "step": 43247 + }, + { + "epoch": 5.67, + "learning_rate": 4.144608953122838e-06, + "loss": 0.0081, + "step": 43248 + }, + { + "epoch": 5.67, + "learning_rate": 4.144179056439144e-06, + "loss": 0.0095, + "step": 43249 + }, + { + "epoch": 5.67, + "learning_rate": 4.143749176224638e-06, + "loss": 0.0104, + "step": 43250 + }, + { + "epoch": 5.67, + "learning_rate": 4.143319312480524e-06, + "loss": 0.007, + "step": 43251 + }, + { + "epoch": 5.67, + "learning_rate": 4.14288946520801e-06, + "loss": 0.0055, + "step": 43252 + }, + { + "epoch": 5.67, + "learning_rate": 4.142459634408311e-06, + "loss": 0.007, + "step": 43253 + }, + { + "epoch": 5.67, + "learning_rate": 4.142029820082631e-06, + "loss": 0.0061, + "step": 43254 + }, + { + "epoch": 5.67, + "learning_rate": 4.1416000222321825e-06, + "loss": 0.0054, + "step": 43255 + }, + { + "epoch": 5.67, + "learning_rate": 4.141170240858168e-06, + "loss": 0.0073, + "step": 43256 + }, + { + "epoch": 5.67, + "learning_rate": 4.140740475961802e-06, + "loss": 0.0077, + "step": 43257 + }, + { + "epoch": 5.67, + "learning_rate": 4.140310727544294e-06, + "loss": 0.0074, + "step": 43258 + }, + { + "epoch": 5.67, + "learning_rate": 4.139880995606844e-06, + "loss": 0.0088, + "step": 43259 + }, + { + "epoch": 5.67, + "learning_rate": 4.139451280150667e-06, + "loss": 0.0067, + "step": 43260 + }, + { + "epoch": 5.67, + "learning_rate": 4.1390215811769764e-06, + "loss": 0.0078, + "step": 43261 + }, + { + "epoch": 5.67, + "learning_rate": 4.138591898686975e-06, + "loss": 0.0051, + "step": 43262 + }, + { + "epoch": 5.67, + "learning_rate": 4.138162232681871e-06, + "loss": 0.0052, + "step": 43263 + }, + { + "epoch": 5.67, + "learning_rate": 4.137732583162869e-06, + "loss": 0.0068, + "step": 43264 + }, + { + "epoch": 5.67, + "learning_rate": 4.137302950131187e-06, + "loss": 0.0071, + "step": 43265 + }, + { + "epoch": 5.67, + "learning_rate": 4.136873333588027e-06, + "loss": 0.0087, + "step": 43266 + }, + { + "epoch": 5.67, + "learning_rate": 4.136443733534596e-06, + "loss": 0.0066, + "step": 43267 + }, + { + "epoch": 5.67, + "learning_rate": 4.136014149972109e-06, + "loss": 0.0074, + "step": 43268 + }, + { + "epoch": 5.67, + "learning_rate": 4.135584582901769e-06, + "loss": 0.0069, + "step": 43269 + }, + { + "epoch": 5.67, + "learning_rate": 4.135155032324785e-06, + "loss": 0.0062, + "step": 43270 + }, + { + "epoch": 5.67, + "learning_rate": 4.134725498242358e-06, + "loss": 0.0089, + "step": 43271 + }, + { + "epoch": 5.67, + "learning_rate": 4.13429598065571e-06, + "loss": 0.0078, + "step": 43272 + }, + { + "epoch": 5.67, + "learning_rate": 4.133866479566043e-06, + "loss": 0.0075, + "step": 43273 + }, + { + "epoch": 5.67, + "learning_rate": 4.133436994974564e-06, + "loss": 0.0083, + "step": 43274 + }, + { + "epoch": 5.67, + "learning_rate": 4.1330075268824775e-06, + "loss": 0.0085, + "step": 43275 + }, + { + "epoch": 5.67, + "learning_rate": 4.132578075290998e-06, + "loss": 0.0055, + "step": 43276 + }, + { + "epoch": 5.67, + "learning_rate": 4.132148640201331e-06, + "loss": 0.0073, + "step": 43277 + }, + { + "epoch": 5.67, + "learning_rate": 4.131719221614683e-06, + "loss": 0.0069, + "step": 43278 + }, + { + "epoch": 5.67, + "learning_rate": 4.131289819532258e-06, + "loss": 0.0065, + "step": 43279 + }, + { + "epoch": 5.67, + "learning_rate": 4.130860433955272e-06, + "loss": 0.0054, + "step": 43280 + }, + { + "epoch": 5.67, + "learning_rate": 4.130431064884929e-06, + "loss": 0.0084, + "step": 43281 + }, + { + "epoch": 5.67, + "learning_rate": 4.130001712322432e-06, + "loss": 0.007, + "step": 43282 + }, + { + "epoch": 5.67, + "learning_rate": 4.129572376268992e-06, + "loss": 0.0058, + "step": 43283 + }, + { + "epoch": 5.67, + "learning_rate": 4.129143056725822e-06, + "loss": 0.0066, + "step": 43284 + }, + { + "epoch": 5.67, + "learning_rate": 4.128713753694123e-06, + "loss": 0.0082, + "step": 43285 + }, + { + "epoch": 5.67, + "learning_rate": 4.128284467175101e-06, + "loss": 0.0063, + "step": 43286 + }, + { + "epoch": 5.67, + "learning_rate": 4.12785519716997e-06, + "loss": 0.0068, + "step": 43287 + }, + { + "epoch": 5.67, + "learning_rate": 4.127425943679934e-06, + "loss": 0.0064, + "step": 43288 + }, + { + "epoch": 5.67, + "learning_rate": 4.126996706706199e-06, + "loss": 0.0053, + "step": 43289 + }, + { + "epoch": 5.67, + "learning_rate": 4.126567486249969e-06, + "loss": 0.0079, + "step": 43290 + }, + { + "epoch": 5.67, + "learning_rate": 4.126138282312459e-06, + "loss": 0.0061, + "step": 43291 + }, + { + "epoch": 5.67, + "learning_rate": 4.125709094894873e-06, + "loss": 0.0076, + "step": 43292 + }, + { + "epoch": 5.67, + "learning_rate": 4.125279923998414e-06, + "loss": 0.0082, + "step": 43293 + }, + { + "epoch": 5.67, + "learning_rate": 4.124850769624291e-06, + "loss": 0.0086, + "step": 43294 + }, + { + "epoch": 5.67, + "learning_rate": 4.124421631773717e-06, + "loss": 0.0087, + "step": 43295 + }, + { + "epoch": 5.67, + "learning_rate": 4.123992510447894e-06, + "loss": 0.0069, + "step": 43296 + }, + { + "epoch": 5.67, + "learning_rate": 4.123563405648029e-06, + "loss": 0.0085, + "step": 43297 + }, + { + "epoch": 5.67, + "learning_rate": 4.123134317375326e-06, + "loss": 0.0072, + "step": 43298 + }, + { + "epoch": 5.67, + "learning_rate": 4.122705245630998e-06, + "loss": 0.0044, + "step": 43299 + }, + { + "epoch": 5.67, + "learning_rate": 4.122276190416248e-06, + "loss": 0.0049, + "step": 43300 + }, + { + "epoch": 5.67, + "learning_rate": 4.12184715173228e-06, + "loss": 0.0064, + "step": 43301 + }, + { + "epoch": 5.67, + "learning_rate": 4.121418129580308e-06, + "loss": 0.0072, + "step": 43302 + }, + { + "epoch": 5.67, + "learning_rate": 4.120989123961534e-06, + "loss": 0.0072, + "step": 43303 + }, + { + "epoch": 5.67, + "learning_rate": 4.120560134877161e-06, + "loss": 0.0084, + "step": 43304 + }, + { + "epoch": 5.67, + "learning_rate": 4.120131162328404e-06, + "loss": 0.008, + "step": 43305 + }, + { + "epoch": 5.67, + "learning_rate": 4.11970220631646e-06, + "loss": 0.0069, + "step": 43306 + }, + { + "epoch": 5.67, + "learning_rate": 4.119273266842545e-06, + "loss": 0.0063, + "step": 43307 + }, + { + "epoch": 5.67, + "learning_rate": 4.118844343907861e-06, + "loss": 0.0075, + "step": 43308 + }, + { + "epoch": 5.67, + "learning_rate": 4.11841543751361e-06, + "loss": 0.0062, + "step": 43309 + }, + { + "epoch": 5.67, + "learning_rate": 4.117986547661006e-06, + "loss": 0.0054, + "step": 43310 + }, + { + "epoch": 5.67, + "learning_rate": 4.117557674351252e-06, + "loss": 0.0054, + "step": 43311 + }, + { + "epoch": 5.67, + "learning_rate": 4.117128817585553e-06, + "loss": 0.0074, + "step": 43312 + }, + { + "epoch": 5.67, + "learning_rate": 4.116699977365113e-06, + "loss": 0.0068, + "step": 43313 + }, + { + "epoch": 5.67, + "learning_rate": 4.116271153691144e-06, + "loss": 0.0075, + "step": 43314 + }, + { + "epoch": 5.67, + "learning_rate": 4.115842346564845e-06, + "loss": 0.0065, + "step": 43315 + }, + { + "epoch": 5.67, + "learning_rate": 4.115413555987431e-06, + "loss": 0.007, + "step": 43316 + }, + { + "epoch": 5.67, + "learning_rate": 4.114984781960099e-06, + "loss": 0.006, + "step": 43317 + }, + { + "epoch": 5.67, + "learning_rate": 4.1145560244840625e-06, + "loss": 0.0074, + "step": 43318 + }, + { + "epoch": 5.67, + "learning_rate": 4.114127283560524e-06, + "loss": 0.0056, + "step": 43319 + }, + { + "epoch": 5.67, + "learning_rate": 4.113698559190689e-06, + "loss": 0.0058, + "step": 43320 + }, + { + "epoch": 5.67, + "learning_rate": 4.113269851375758e-06, + "loss": 0.0074, + "step": 43321 + }, + { + "epoch": 5.67, + "learning_rate": 4.112841160116948e-06, + "loss": 0.0058, + "step": 43322 + }, + { + "epoch": 5.68, + "learning_rate": 4.112412485415458e-06, + "loss": 0.0051, + "step": 43323 + }, + { + "epoch": 5.68, + "learning_rate": 4.111983827272491e-06, + "loss": 0.0063, + "step": 43324 + }, + { + "epoch": 5.68, + "learning_rate": 4.111555185689259e-06, + "loss": 0.006, + "step": 43325 + }, + { + "epoch": 5.68, + "learning_rate": 4.111126560666961e-06, + "loss": 0.0066, + "step": 43326 + }, + { + "epoch": 5.68, + "learning_rate": 4.1106979522068095e-06, + "loss": 0.005, + "step": 43327 + }, + { + "epoch": 5.68, + "learning_rate": 4.1102693603100065e-06, + "loss": 0.0048, + "step": 43328 + }, + { + "epoch": 5.68, + "learning_rate": 4.109840784977754e-06, + "loss": 0.0047, + "step": 43329 + }, + { + "epoch": 5.68, + "learning_rate": 4.109412226211264e-06, + "loss": 0.0087, + "step": 43330 + }, + { + "epoch": 5.68, + "learning_rate": 4.108983684011737e-06, + "loss": 0.0052, + "step": 43331 + }, + { + "epoch": 5.68, + "learning_rate": 4.108555158380377e-06, + "loss": 0.0063, + "step": 43332 + }, + { + "epoch": 5.68, + "learning_rate": 4.108126649318397e-06, + "loss": 0.0055, + "step": 43333 + }, + { + "epoch": 5.68, + "learning_rate": 4.107698156826995e-06, + "loss": 0.0078, + "step": 43334 + }, + { + "epoch": 5.68, + "learning_rate": 4.107269680907379e-06, + "loss": 0.0094, + "step": 43335 + }, + { + "epoch": 5.68, + "learning_rate": 4.10684122156075e-06, + "loss": 0.008, + "step": 43336 + }, + { + "epoch": 5.68, + "learning_rate": 4.1064127787883155e-06, + "loss": 0.007, + "step": 43337 + }, + { + "epoch": 5.68, + "learning_rate": 4.105984352591286e-06, + "loss": 0.0077, + "step": 43338 + }, + { + "epoch": 5.68, + "learning_rate": 4.1055559429708605e-06, + "loss": 0.0075, + "step": 43339 + }, + { + "epoch": 5.68, + "learning_rate": 4.105127549928242e-06, + "loss": 0.0063, + "step": 43340 + }, + { + "epoch": 5.68, + "learning_rate": 4.104699173464641e-06, + "loss": 0.0053, + "step": 43341 + }, + { + "epoch": 5.68, + "learning_rate": 4.10427081358126e-06, + "loss": 0.0073, + "step": 43342 + }, + { + "epoch": 5.68, + "learning_rate": 4.1038424702793e-06, + "loss": 0.008, + "step": 43343 + }, + { + "epoch": 5.68, + "learning_rate": 4.103414143559973e-06, + "loss": 0.0048, + "step": 43344 + }, + { + "epoch": 5.68, + "learning_rate": 4.10298583342448e-06, + "loss": 0.0074, + "step": 43345 + }, + { + "epoch": 5.68, + "learning_rate": 4.102557539874023e-06, + "loss": 0.0101, + "step": 43346 + }, + { + "epoch": 5.68, + "learning_rate": 4.102129262909807e-06, + "loss": 0.0063, + "step": 43347 + }, + { + "epoch": 5.68, + "learning_rate": 4.10170100253304e-06, + "loss": 0.0064, + "step": 43348 + }, + { + "epoch": 5.68, + "learning_rate": 4.101272758744926e-06, + "loss": 0.0056, + "step": 43349 + }, + { + "epoch": 5.68, + "learning_rate": 4.10084453154667e-06, + "loss": 0.0075, + "step": 43350 + }, + { + "epoch": 5.68, + "learning_rate": 4.1004163209394685e-06, + "loss": 0.0059, + "step": 43351 + }, + { + "epoch": 5.68, + "learning_rate": 4.099988126924538e-06, + "loss": 0.0069, + "step": 43352 + }, + { + "epoch": 5.68, + "learning_rate": 4.099559949503076e-06, + "loss": 0.009, + "step": 43353 + }, + { + "epoch": 5.68, + "learning_rate": 4.099131788676287e-06, + "loss": 0.0067, + "step": 43354 + }, + { + "epoch": 5.68, + "learning_rate": 4.098703644445372e-06, + "loss": 0.0085, + "step": 43355 + }, + { + "epoch": 5.68, + "learning_rate": 4.098275516811543e-06, + "loss": 0.0055, + "step": 43356 + }, + { + "epoch": 5.68, + "learning_rate": 4.097847405775999e-06, + "loss": 0.0069, + "step": 43357 + }, + { + "epoch": 5.68, + "learning_rate": 4.097419311339942e-06, + "loss": 0.0091, + "step": 43358 + }, + { + "epoch": 5.68, + "learning_rate": 4.096991233504578e-06, + "loss": 0.0041, + "step": 43359 + }, + { + "epoch": 5.68, + "learning_rate": 4.0965631722711155e-06, + "loss": 0.0064, + "step": 43360 + }, + { + "epoch": 5.68, + "learning_rate": 4.0961351276407555e-06, + "loss": 0.0064, + "step": 43361 + }, + { + "epoch": 5.68, + "learning_rate": 4.095707099614701e-06, + "loss": 0.0047, + "step": 43362 + }, + { + "epoch": 5.68, + "learning_rate": 4.095279088194152e-06, + "loss": 0.0063, + "step": 43363 + }, + { + "epoch": 5.68, + "learning_rate": 4.0948510933803185e-06, + "loss": 0.0068, + "step": 43364 + }, + { + "epoch": 5.68, + "learning_rate": 4.094423115174403e-06, + "loss": 0.0077, + "step": 43365 + }, + { + "epoch": 5.68, + "learning_rate": 4.093995153577604e-06, + "loss": 0.0094, + "step": 43366 + }, + { + "epoch": 5.68, + "learning_rate": 4.093567208591134e-06, + "loss": 0.0087, + "step": 43367 + }, + { + "epoch": 5.68, + "learning_rate": 4.093139280216191e-06, + "loss": 0.0077, + "step": 43368 + }, + { + "epoch": 5.68, + "learning_rate": 4.092711368453979e-06, + "loss": 0.0055, + "step": 43369 + }, + { + "epoch": 5.68, + "learning_rate": 4.092283473305697e-06, + "loss": 0.0061, + "step": 43370 + }, + { + "epoch": 5.68, + "learning_rate": 4.091855594772554e-06, + "loss": 0.007, + "step": 43371 + }, + { + "epoch": 5.68, + "learning_rate": 4.091427732855757e-06, + "loss": 0.007, + "step": 43372 + }, + { + "epoch": 5.68, + "learning_rate": 4.090999887556504e-06, + "loss": 0.0042, + "step": 43373 + }, + { + "epoch": 5.68, + "learning_rate": 4.090572058875996e-06, + "loss": 0.0046, + "step": 43374 + }, + { + "epoch": 5.68, + "learning_rate": 4.090144246815442e-06, + "loss": 0.0062, + "step": 43375 + }, + { + "epoch": 5.68, + "learning_rate": 4.089716451376043e-06, + "loss": 0.0075, + "step": 43376 + }, + { + "epoch": 5.68, + "learning_rate": 4.089288672559003e-06, + "loss": 0.0057, + "step": 43377 + }, + { + "epoch": 5.68, + "learning_rate": 4.088860910365518e-06, + "loss": 0.0077, + "step": 43378 + }, + { + "epoch": 5.68, + "learning_rate": 4.088433164796802e-06, + "loss": 0.0063, + "step": 43379 + }, + { + "epoch": 5.68, + "learning_rate": 4.08800543585405e-06, + "loss": 0.0054, + "step": 43380 + }, + { + "epoch": 5.68, + "learning_rate": 4.087577723538471e-06, + "loss": 0.0095, + "step": 43381 + }, + { + "epoch": 5.68, + "learning_rate": 4.08715002785126e-06, + "loss": 0.0047, + "step": 43382 + }, + { + "epoch": 5.68, + "learning_rate": 4.08672234879363e-06, + "loss": 0.0076, + "step": 43383 + }, + { + "epoch": 5.68, + "learning_rate": 4.086294686366776e-06, + "loss": 0.0069, + "step": 43384 + }, + { + "epoch": 5.68, + "learning_rate": 4.085867040571905e-06, + "loss": 0.0089, + "step": 43385 + }, + { + "epoch": 5.68, + "learning_rate": 4.0854394114102145e-06, + "loss": 0.0077, + "step": 43386 + }, + { + "epoch": 5.68, + "learning_rate": 4.085011798882914e-06, + "loss": 0.004, + "step": 43387 + }, + { + "epoch": 5.68, + "learning_rate": 4.084584202991203e-06, + "loss": 0.0067, + "step": 43388 + }, + { + "epoch": 5.68, + "learning_rate": 4.08415662373628e-06, + "loss": 0.006, + "step": 43389 + }, + { + "epoch": 5.68, + "learning_rate": 4.083729061119354e-06, + "loss": 0.0058, + "step": 43390 + }, + { + "epoch": 5.68, + "learning_rate": 4.083301515141623e-06, + "loss": 0.01, + "step": 43391 + }, + { + "epoch": 5.68, + "learning_rate": 4.0828739858042945e-06, + "loss": 0.0067, + "step": 43392 + }, + { + "epoch": 5.68, + "learning_rate": 4.082446473108565e-06, + "loss": 0.0069, + "step": 43393 + }, + { + "epoch": 5.68, + "learning_rate": 4.082018977055641e-06, + "loss": 0.0063, + "step": 43394 + }, + { + "epoch": 5.68, + "learning_rate": 4.081591497646727e-06, + "loss": 0.0067, + "step": 43395 + }, + { + "epoch": 5.68, + "learning_rate": 4.081164034883018e-06, + "loss": 0.0102, + "step": 43396 + }, + { + "epoch": 5.68, + "learning_rate": 4.080736588765719e-06, + "loss": 0.0048, + "step": 43397 + }, + { + "epoch": 5.68, + "learning_rate": 4.080309159296035e-06, + "loss": 0.0049, + "step": 43398 + }, + { + "epoch": 5.68, + "learning_rate": 4.079881746475167e-06, + "loss": 0.0078, + "step": 43399 + }, + { + "epoch": 5.69, + "learning_rate": 4.079454350304317e-06, + "loss": 0.0082, + "step": 43400 + }, + { + "epoch": 5.69, + "learning_rate": 4.079026970784682e-06, + "loss": 0.0074, + "step": 43401 + }, + { + "epoch": 5.69, + "learning_rate": 4.078599607917468e-06, + "loss": 0.007, + "step": 43402 + }, + { + "epoch": 5.69, + "learning_rate": 4.078172261703882e-06, + "loss": 0.005, + "step": 43403 + }, + { + "epoch": 5.69, + "learning_rate": 4.07774493214512e-06, + "loss": 0.0073, + "step": 43404 + }, + { + "epoch": 5.69, + "learning_rate": 4.077317619242382e-06, + "loss": 0.0058, + "step": 43405 + }, + { + "epoch": 5.69, + "learning_rate": 4.076890322996876e-06, + "loss": 0.0076, + "step": 43406 + }, + { + "epoch": 5.69, + "learning_rate": 4.076463043409801e-06, + "loss": 0.0063, + "step": 43407 + }, + { + "epoch": 5.69, + "learning_rate": 4.076035780482355e-06, + "loss": 0.0071, + "step": 43408 + }, + { + "epoch": 5.69, + "learning_rate": 4.0756085342157455e-06, + "loss": 0.0079, + "step": 43409 + }, + { + "epoch": 5.69, + "learning_rate": 4.075181304611172e-06, + "loss": 0.0054, + "step": 43410 + }, + { + "epoch": 5.69, + "learning_rate": 4.074754091669836e-06, + "loss": 0.0074, + "step": 43411 + }, + { + "epoch": 5.69, + "learning_rate": 4.0743268953929334e-06, + "loss": 0.0078, + "step": 43412 + }, + { + "epoch": 5.69, + "learning_rate": 4.0738997157816725e-06, + "loss": 0.0052, + "step": 43413 + }, + { + "epoch": 5.69, + "learning_rate": 4.073472552837257e-06, + "loss": 0.009, + "step": 43414 + }, + { + "epoch": 5.69, + "learning_rate": 4.073045406560883e-06, + "loss": 0.007, + "step": 43415 + }, + { + "epoch": 5.69, + "learning_rate": 4.072618276953749e-06, + "loss": 0.0068, + "step": 43416 + }, + { + "epoch": 5.69, + "learning_rate": 4.0721911640170655e-06, + "loss": 0.0059, + "step": 43417 + }, + { + "epoch": 5.69, + "learning_rate": 4.071764067752029e-06, + "loss": 0.0045, + "step": 43418 + }, + { + "epoch": 5.69, + "learning_rate": 4.071336988159839e-06, + "loss": 0.0048, + "step": 43419 + }, + { + "epoch": 5.69, + "learning_rate": 4.070909925241695e-06, + "loss": 0.0039, + "step": 43420 + }, + { + "epoch": 5.69, + "learning_rate": 4.0704828789988045e-06, + "loss": 0.0074, + "step": 43421 + }, + { + "epoch": 5.69, + "learning_rate": 4.070055849432366e-06, + "loss": 0.0059, + "step": 43422 + }, + { + "epoch": 5.69, + "learning_rate": 4.069628836543576e-06, + "loss": 0.0058, + "step": 43423 + }, + { + "epoch": 5.69, + "learning_rate": 4.069201840333638e-06, + "loss": 0.0067, + "step": 43424 + }, + { + "epoch": 5.69, + "learning_rate": 4.068774860803758e-06, + "loss": 0.0054, + "step": 43425 + }, + { + "epoch": 5.69, + "learning_rate": 4.068347897955133e-06, + "loss": 0.0078, + "step": 43426 + }, + { + "epoch": 5.69, + "learning_rate": 4.067920951788963e-06, + "loss": 0.0075, + "step": 43427 + }, + { + "epoch": 5.69, + "learning_rate": 4.0674940223064465e-06, + "loss": 0.009, + "step": 43428 + }, + { + "epoch": 5.69, + "learning_rate": 4.06706710950879e-06, + "loss": 0.0071, + "step": 43429 + }, + { + "epoch": 5.69, + "learning_rate": 4.066640213397193e-06, + "loss": 0.0067, + "step": 43430 + }, + { + "epoch": 5.69, + "learning_rate": 4.06621333397285e-06, + "loss": 0.0045, + "step": 43431 + }, + { + "epoch": 5.69, + "learning_rate": 4.065786471236969e-06, + "loss": 0.0045, + "step": 43432 + }, + { + "epoch": 5.69, + "learning_rate": 4.065359625190747e-06, + "loss": 0.0058, + "step": 43433 + }, + { + "epoch": 5.69, + "learning_rate": 4.064932795835386e-06, + "loss": 0.0038, + "step": 43434 + }, + { + "epoch": 5.69, + "learning_rate": 4.064505983172082e-06, + "loss": 0.0055, + "step": 43435 + }, + { + "epoch": 5.69, + "learning_rate": 4.064079187202039e-06, + "loss": 0.0059, + "step": 43436 + }, + { + "epoch": 5.69, + "learning_rate": 4.06365240792646e-06, + "loss": 0.0054, + "step": 43437 + }, + { + "epoch": 5.69, + "learning_rate": 4.0632256453465424e-06, + "loss": 0.0067, + "step": 43438 + }, + { + "epoch": 5.69, + "learning_rate": 4.0627988994634835e-06, + "loss": 0.0083, + "step": 43439 + }, + { + "epoch": 5.69, + "learning_rate": 4.06237217027849e-06, + "loss": 0.0049, + "step": 43440 + }, + { + "epoch": 5.69, + "learning_rate": 4.0619454577927585e-06, + "loss": 0.0064, + "step": 43441 + }, + { + "epoch": 5.69, + "learning_rate": 4.0615187620074895e-06, + "loss": 0.0075, + "step": 43442 + }, + { + "epoch": 5.69, + "learning_rate": 4.061092082923879e-06, + "loss": 0.0058, + "step": 43443 + }, + { + "epoch": 5.69, + "learning_rate": 4.060665420543134e-06, + "loss": 0.0053, + "step": 43444 + }, + { + "epoch": 5.69, + "learning_rate": 4.060238774866452e-06, + "loss": 0.0092, + "step": 43445 + }, + { + "epoch": 5.69, + "learning_rate": 4.0598121458950275e-06, + "loss": 0.0074, + "step": 43446 + }, + { + "epoch": 5.69, + "learning_rate": 4.059385533630066e-06, + "loss": 0.0061, + "step": 43447 + }, + { + "epoch": 5.69, + "learning_rate": 4.058958938072769e-06, + "loss": 0.0074, + "step": 43448 + }, + { + "epoch": 5.69, + "learning_rate": 4.0585323592243345e-06, + "loss": 0.0099, + "step": 43449 + }, + { + "epoch": 5.69, + "learning_rate": 4.05810579708596e-06, + "loss": 0.0084, + "step": 43450 + }, + { + "epoch": 5.69, + "learning_rate": 4.057679251658844e-06, + "loss": 0.0063, + "step": 43451 + }, + { + "epoch": 5.69, + "learning_rate": 4.057252722944193e-06, + "loss": 0.0061, + "step": 43452 + }, + { + "epoch": 5.69, + "learning_rate": 4.0568262109432e-06, + "loss": 0.0061, + "step": 43453 + }, + { + "epoch": 5.69, + "learning_rate": 4.056399715657064e-06, + "loss": 0.0077, + "step": 43454 + }, + { + "epoch": 5.69, + "learning_rate": 4.055973237086992e-06, + "loss": 0.0054, + "step": 43455 + }, + { + "epoch": 5.69, + "learning_rate": 4.055546775234178e-06, + "loss": 0.0067, + "step": 43456 + }, + { + "epoch": 5.69, + "learning_rate": 4.055120330099819e-06, + "loss": 0.0054, + "step": 43457 + }, + { + "epoch": 5.69, + "learning_rate": 4.054693901685118e-06, + "loss": 0.0049, + "step": 43458 + }, + { + "epoch": 5.69, + "learning_rate": 4.054267489991276e-06, + "loss": 0.0082, + "step": 43459 + }, + { + "epoch": 5.69, + "learning_rate": 4.053841095019491e-06, + "loss": 0.0057, + "step": 43460 + }, + { + "epoch": 5.69, + "learning_rate": 4.0534147167709605e-06, + "loss": 0.0089, + "step": 43461 + }, + { + "epoch": 5.69, + "learning_rate": 4.052988355246881e-06, + "loss": 0.0056, + "step": 43462 + }, + { + "epoch": 5.69, + "learning_rate": 4.052562010448459e-06, + "loss": 0.0067, + "step": 43463 + }, + { + "epoch": 5.69, + "learning_rate": 4.052135682376888e-06, + "loss": 0.0066, + "step": 43464 + }, + { + "epoch": 5.69, + "learning_rate": 4.051709371033371e-06, + "loss": 0.007, + "step": 43465 + }, + { + "epoch": 5.69, + "learning_rate": 4.0512830764191006e-06, + "loss": 0.0072, + "step": 43466 + }, + { + "epoch": 5.69, + "learning_rate": 4.050856798535282e-06, + "loss": 0.0085, + "step": 43467 + }, + { + "epoch": 5.69, + "learning_rate": 4.05043053738311e-06, + "loss": 0.0074, + "step": 43468 + }, + { + "epoch": 5.69, + "learning_rate": 4.050004292963787e-06, + "loss": 0.0046, + "step": 43469 + }, + { + "epoch": 5.69, + "learning_rate": 4.0495780652785075e-06, + "loss": 0.0077, + "step": 43470 + }, + { + "epoch": 5.69, + "learning_rate": 4.049151854328477e-06, + "loss": 0.0072, + "step": 43471 + }, + { + "epoch": 5.69, + "learning_rate": 4.048725660114889e-06, + "loss": 0.0072, + "step": 43472 + }, + { + "epoch": 5.69, + "learning_rate": 4.04829948263894e-06, + "loss": 0.008, + "step": 43473 + }, + { + "epoch": 5.69, + "learning_rate": 4.047873321901836e-06, + "loss": 0.0057, + "step": 43474 + }, + { + "epoch": 5.69, + "learning_rate": 4.047447177904771e-06, + "loss": 0.0053, + "step": 43475 + }, + { + "epoch": 5.7, + "learning_rate": 4.047021050648943e-06, + "loss": 0.0052, + "step": 43476 + }, + { + "epoch": 5.7, + "learning_rate": 4.0465949401355475e-06, + "loss": 0.0061, + "step": 43477 + }, + { + "epoch": 5.7, + "learning_rate": 4.046168846365791e-06, + "loss": 0.0093, + "step": 43478 + }, + { + "epoch": 5.7, + "learning_rate": 4.045742769340864e-06, + "loss": 0.0058, + "step": 43479 + }, + { + "epoch": 5.7, + "learning_rate": 4.045316709061972e-06, + "loss": 0.0058, + "step": 43480 + }, + { + "epoch": 5.7, + "learning_rate": 4.044890665530306e-06, + "loss": 0.0071, + "step": 43481 + }, + { + "epoch": 5.7, + "learning_rate": 4.044464638747071e-06, + "loss": 0.0049, + "step": 43482 + }, + { + "epoch": 5.7, + "learning_rate": 4.044038628713462e-06, + "loss": 0.0067, + "step": 43483 + }, + { + "epoch": 5.7, + "learning_rate": 4.043612635430678e-06, + "loss": 0.0062, + "step": 43484 + }, + { + "epoch": 5.7, + "learning_rate": 4.0431866588999126e-06, + "loss": 0.0074, + "step": 43485 + }, + { + "epoch": 5.7, + "learning_rate": 4.042760699122371e-06, + "loss": 0.0046, + "step": 43486 + }, + { + "epoch": 5.7, + "learning_rate": 4.042334756099247e-06, + "loss": 0.0059, + "step": 43487 + }, + { + "epoch": 5.7, + "learning_rate": 4.041908829831736e-06, + "loss": 0.006, + "step": 43488 + }, + { + "epoch": 5.7, + "learning_rate": 4.041482920321043e-06, + "loss": 0.005, + "step": 43489 + }, + { + "epoch": 5.7, + "learning_rate": 4.041057027568359e-06, + "loss": 0.006, + "step": 43490 + }, + { + "epoch": 5.7, + "learning_rate": 4.040631151574887e-06, + "loss": 0.0048, + "step": 43491 + }, + { + "epoch": 5.7, + "learning_rate": 4.0402052923418235e-06, + "loss": 0.0055, + "step": 43492 + }, + { + "epoch": 5.7, + "learning_rate": 4.039779449870361e-06, + "loss": 0.009, + "step": 43493 + }, + { + "epoch": 5.7, + "learning_rate": 4.039353624161706e-06, + "loss": 0.006, + "step": 43494 + }, + { + "epoch": 5.7, + "learning_rate": 4.03892781521705e-06, + "loss": 0.0078, + "step": 43495 + }, + { + "epoch": 5.7, + "learning_rate": 4.038502023037589e-06, + "loss": 0.0065, + "step": 43496 + }, + { + "epoch": 5.7, + "learning_rate": 4.038076247624529e-06, + "loss": 0.0083, + "step": 43497 + }, + { + "epoch": 5.7, + "learning_rate": 4.037650488979059e-06, + "loss": 0.0069, + "step": 43498 + }, + { + "epoch": 5.7, + "learning_rate": 4.037224747102382e-06, + "loss": 0.007, + "step": 43499 + }, + { + "epoch": 5.7, + "learning_rate": 4.036799021995688e-06, + "loss": 0.0081, + "step": 43500 + }, + { + "epoch": 5.7, + "learning_rate": 4.036373313660179e-06, + "loss": 0.0049, + "step": 43501 + }, + { + "epoch": 5.7, + "learning_rate": 4.035947622097056e-06, + "loss": 0.009, + "step": 43502 + }, + { + "epoch": 5.7, + "learning_rate": 4.0355219473075134e-06, + "loss": 0.0074, + "step": 43503 + }, + { + "epoch": 5.7, + "learning_rate": 4.035096289292745e-06, + "loss": 0.0043, + "step": 43504 + }, + { + "epoch": 5.7, + "learning_rate": 4.0346706480539525e-06, + "loss": 0.0047, + "step": 43505 + }, + { + "epoch": 5.7, + "learning_rate": 4.034245023592332e-06, + "loss": 0.0099, + "step": 43506 + }, + { + "epoch": 5.7, + "learning_rate": 4.033819415909079e-06, + "loss": 0.0051, + "step": 43507 + }, + { + "epoch": 5.7, + "learning_rate": 4.033393825005388e-06, + "loss": 0.0052, + "step": 43508 + }, + { + "epoch": 5.7, + "learning_rate": 4.032968250882463e-06, + "loss": 0.0075, + "step": 43509 + }, + { + "epoch": 5.7, + "learning_rate": 4.032542693541497e-06, + "loss": 0.0055, + "step": 43510 + }, + { + "epoch": 5.7, + "learning_rate": 4.032117152983683e-06, + "loss": 0.0081, + "step": 43511 + }, + { + "epoch": 5.7, + "learning_rate": 4.031691629210222e-06, + "loss": 0.0079, + "step": 43512 + }, + { + "epoch": 5.7, + "learning_rate": 4.0312661222223135e-06, + "loss": 0.0073, + "step": 43513 + }, + { + "epoch": 5.7, + "learning_rate": 4.030840632021151e-06, + "loss": 0.0049, + "step": 43514 + }, + { + "epoch": 5.7, + "learning_rate": 4.030415158607932e-06, + "loss": 0.0063, + "step": 43515 + }, + { + "epoch": 5.7, + "learning_rate": 4.029989701983849e-06, + "loss": 0.0084, + "step": 43516 + }, + { + "epoch": 5.7, + "learning_rate": 4.029564262150105e-06, + "loss": 0.0054, + "step": 43517 + }, + { + "epoch": 5.7, + "learning_rate": 4.029138839107894e-06, + "loss": 0.007, + "step": 43518 + }, + { + "epoch": 5.7, + "learning_rate": 4.028713432858409e-06, + "loss": 0.0057, + "step": 43519 + }, + { + "epoch": 5.7, + "learning_rate": 4.028288043402852e-06, + "loss": 0.0082, + "step": 43520 + }, + { + "epoch": 5.7, + "learning_rate": 4.027862670742417e-06, + "loss": 0.008, + "step": 43521 + }, + { + "epoch": 5.7, + "learning_rate": 4.027437314878297e-06, + "loss": 0.0045, + "step": 43522 + }, + { + "epoch": 5.7, + "learning_rate": 4.027011975811692e-06, + "loss": 0.0066, + "step": 43523 + }, + { + "epoch": 5.7, + "learning_rate": 4.026586653543801e-06, + "loss": 0.0045, + "step": 43524 + }, + { + "epoch": 5.7, + "learning_rate": 4.026161348075817e-06, + "loss": 0.0042, + "step": 43525 + }, + { + "epoch": 5.7, + "learning_rate": 4.0257360594089355e-06, + "loss": 0.0078, + "step": 43526 + }, + { + "epoch": 5.7, + "learning_rate": 4.02531078754435e-06, + "loss": 0.0049, + "step": 43527 + }, + { + "epoch": 5.7, + "learning_rate": 4.0248855324832635e-06, + "loss": 0.008, + "step": 43528 + }, + { + "epoch": 5.7, + "learning_rate": 4.024460294226867e-06, + "loss": 0.0064, + "step": 43529 + }, + { + "epoch": 5.7, + "learning_rate": 4.024035072776359e-06, + "loss": 0.0082, + "step": 43530 + }, + { + "epoch": 5.7, + "learning_rate": 4.0236098681329305e-06, + "loss": 0.0074, + "step": 43531 + }, + { + "epoch": 5.7, + "learning_rate": 4.023184680297785e-06, + "loss": 0.006, + "step": 43532 + }, + { + "epoch": 5.7, + "learning_rate": 4.0227595092721096e-06, + "loss": 0.0052, + "step": 43533 + }, + { + "epoch": 5.7, + "learning_rate": 4.0223343550571085e-06, + "loss": 0.0097, + "step": 43534 + }, + { + "epoch": 5.7, + "learning_rate": 4.02190921765397e-06, + "loss": 0.0086, + "step": 43535 + }, + { + "epoch": 5.7, + "learning_rate": 4.021484097063898e-06, + "loss": 0.0063, + "step": 43536 + }, + { + "epoch": 5.7, + "learning_rate": 4.021058993288083e-06, + "loss": 0.0046, + "step": 43537 + }, + { + "epoch": 5.7, + "learning_rate": 4.020633906327718e-06, + "loss": 0.0076, + "step": 43538 + }, + { + "epoch": 5.7, + "learning_rate": 4.020208836184004e-06, + "loss": 0.0043, + "step": 43539 + }, + { + "epoch": 5.7, + "learning_rate": 4.019783782858136e-06, + "loss": 0.0071, + "step": 43540 + }, + { + "epoch": 5.7, + "learning_rate": 4.019358746351306e-06, + "loss": 0.0063, + "step": 43541 + }, + { + "epoch": 5.7, + "learning_rate": 4.018933726664709e-06, + "loss": 0.0103, + "step": 43542 + }, + { + "epoch": 5.7, + "learning_rate": 4.018508723799544e-06, + "loss": 0.0056, + "step": 43543 + }, + { + "epoch": 5.7, + "learning_rate": 4.018083737757003e-06, + "loss": 0.0064, + "step": 43544 + }, + { + "epoch": 5.7, + "learning_rate": 4.017658768538285e-06, + "loss": 0.0068, + "step": 43545 + }, + { + "epoch": 5.7, + "learning_rate": 4.017233816144581e-06, + "loss": 0.0061, + "step": 43546 + }, + { + "epoch": 5.7, + "learning_rate": 4.0168088805770914e-06, + "loss": 0.0066, + "step": 43547 + }, + { + "epoch": 5.7, + "learning_rate": 4.016383961837008e-06, + "loss": 0.0052, + "step": 43548 + }, + { + "epoch": 5.7, + "learning_rate": 4.015959059925526e-06, + "loss": 0.0037, + "step": 43549 + }, + { + "epoch": 5.7, + "learning_rate": 4.015534174843837e-06, + "loss": 0.007, + "step": 43550 + }, + { + "epoch": 5.7, + "learning_rate": 4.015109306593144e-06, + "loss": 0.0047, + "step": 43551 + }, + { + "epoch": 5.71, + "learning_rate": 4.014684455174635e-06, + "loss": 0.0066, + "step": 43552 + }, + { + "epoch": 5.71, + "learning_rate": 4.014259620589506e-06, + "loss": 0.0065, + "step": 43553 + }, + { + "epoch": 5.71, + "learning_rate": 4.013834802838955e-06, + "loss": 0.0049, + "step": 43554 + }, + { + "epoch": 5.71, + "learning_rate": 4.013410001924172e-06, + "loss": 0.0076, + "step": 43555 + }, + { + "epoch": 5.71, + "learning_rate": 4.012985217846358e-06, + "loss": 0.0065, + "step": 43556 + }, + { + "epoch": 5.71, + "learning_rate": 4.012560450606704e-06, + "loss": 0.0066, + "step": 43557 + }, + { + "epoch": 5.71, + "learning_rate": 4.012135700206402e-06, + "loss": 0.0049, + "step": 43558 + }, + { + "epoch": 5.71, + "learning_rate": 4.011710966646652e-06, + "loss": 0.0068, + "step": 43559 + }, + { + "epoch": 5.71, + "learning_rate": 4.0112862499286476e-06, + "loss": 0.007, + "step": 43560 + }, + { + "epoch": 5.71, + "learning_rate": 4.010861550053576e-06, + "loss": 0.0064, + "step": 43561 + }, + { + "epoch": 5.71, + "learning_rate": 4.0104368670226435e-06, + "loss": 0.005, + "step": 43562 + }, + { + "epoch": 5.71, + "learning_rate": 4.010012200837037e-06, + "loss": 0.0068, + "step": 43563 + }, + { + "epoch": 5.71, + "learning_rate": 4.009587551497952e-06, + "loss": 0.0051, + "step": 43564 + }, + { + "epoch": 5.71, + "learning_rate": 4.009162919006579e-06, + "loss": 0.0047, + "step": 43565 + }, + { + "epoch": 5.71, + "learning_rate": 4.0087383033641166e-06, + "loss": 0.0075, + "step": 43566 + }, + { + "epoch": 5.71, + "learning_rate": 4.008313704571763e-06, + "loss": 0.0084, + "step": 43567 + }, + { + "epoch": 5.71, + "learning_rate": 4.007889122630707e-06, + "loss": 0.0072, + "step": 43568 + }, + { + "epoch": 5.71, + "learning_rate": 4.007464557542141e-06, + "loss": 0.0076, + "step": 43569 + }, + { + "epoch": 5.71, + "learning_rate": 4.007040009307265e-06, + "loss": 0.0078, + "step": 43570 + }, + { + "epoch": 5.71, + "learning_rate": 4.006615477927271e-06, + "loss": 0.0063, + "step": 43571 + }, + { + "epoch": 5.71, + "learning_rate": 4.0061909634033505e-06, + "loss": 0.0062, + "step": 43572 + }, + { + "epoch": 5.71, + "learning_rate": 4.005766465736695e-06, + "loss": 0.0081, + "step": 43573 + }, + { + "epoch": 5.71, + "learning_rate": 4.005341984928506e-06, + "loss": 0.0095, + "step": 43574 + }, + { + "epoch": 5.71, + "learning_rate": 4.004917520979973e-06, + "loss": 0.0062, + "step": 43575 + }, + { + "epoch": 5.71, + "learning_rate": 4.0044930738922875e-06, + "loss": 0.0077, + "step": 43576 + }, + { + "epoch": 5.71, + "learning_rate": 4.004068643666646e-06, + "loss": 0.0055, + "step": 43577 + }, + { + "epoch": 5.71, + "learning_rate": 4.003644230304245e-06, + "loss": 0.0059, + "step": 43578 + }, + { + "epoch": 5.71, + "learning_rate": 4.003219833806276e-06, + "loss": 0.0058, + "step": 43579 + }, + { + "epoch": 5.71, + "learning_rate": 4.002795454173931e-06, + "loss": 0.0052, + "step": 43580 + }, + { + "epoch": 5.71, + "learning_rate": 4.0023710914084014e-06, + "loss": 0.0051, + "step": 43581 + }, + { + "epoch": 5.71, + "learning_rate": 4.001946745510887e-06, + "loss": 0.0076, + "step": 43582 + }, + { + "epoch": 5.71, + "learning_rate": 4.001522416482578e-06, + "loss": 0.0078, + "step": 43583 + }, + { + "epoch": 5.71, + "learning_rate": 4.001098104324665e-06, + "loss": 0.0055, + "step": 43584 + }, + { + "epoch": 5.71, + "learning_rate": 4.000673809038347e-06, + "loss": 0.0048, + "step": 43585 + }, + { + "epoch": 5.71, + "learning_rate": 4.000249530624813e-06, + "loss": 0.0045, + "step": 43586 + }, + { + "epoch": 5.71, + "learning_rate": 3.99982526908526e-06, + "loss": 0.0075, + "step": 43587 + }, + { + "epoch": 5.71, + "learning_rate": 3.99940102442087e-06, + "loss": 0.0076, + "step": 43588 + }, + { + "epoch": 5.71, + "learning_rate": 3.998976796632854e-06, + "loss": 0.0067, + "step": 43589 + }, + { + "epoch": 5.71, + "learning_rate": 3.998552585722395e-06, + "loss": 0.0076, + "step": 43590 + }, + { + "epoch": 5.71, + "learning_rate": 3.998128391690688e-06, + "loss": 0.0095, + "step": 43591 + }, + { + "epoch": 5.71, + "learning_rate": 3.9977042145389214e-06, + "loss": 0.0048, + "step": 43592 + }, + { + "epoch": 5.71, + "learning_rate": 3.997280054268296e-06, + "loss": 0.0062, + "step": 43593 + }, + { + "epoch": 5.71, + "learning_rate": 3.99685591088e-06, + "loss": 0.007, + "step": 43594 + }, + { + "epoch": 5.71, + "learning_rate": 3.996431784375227e-06, + "loss": 0.0079, + "step": 43595 + }, + { + "epoch": 5.71, + "learning_rate": 3.9960076747551655e-06, + "loss": 0.0075, + "step": 43596 + }, + { + "epoch": 5.71, + "learning_rate": 3.9955835820210186e-06, + "loss": 0.006, + "step": 43597 + }, + { + "epoch": 5.71, + "learning_rate": 3.995159506173971e-06, + "loss": 0.0079, + "step": 43598 + }, + { + "epoch": 5.71, + "learning_rate": 3.994735447215215e-06, + "loss": 0.0052, + "step": 43599 + }, + { + "epoch": 5.71, + "learning_rate": 3.994311405145946e-06, + "loss": 0.0058, + "step": 43600 + }, + { + "epoch": 5.71, + "learning_rate": 3.993887379967359e-06, + "loss": 0.007, + "step": 43601 + }, + { + "epoch": 5.71, + "learning_rate": 3.9934633716806445e-06, + "loss": 0.0051, + "step": 43602 + }, + { + "epoch": 5.71, + "learning_rate": 3.9930393802869915e-06, + "loss": 0.0055, + "step": 43603 + }, + { + "epoch": 5.71, + "learning_rate": 3.992615405787598e-06, + "loss": 0.006, + "step": 43604 + }, + { + "epoch": 5.71, + "learning_rate": 3.9921914481836545e-06, + "loss": 0.0075, + "step": 43605 + }, + { + "epoch": 5.71, + "learning_rate": 3.9917675074763514e-06, + "loss": 0.0063, + "step": 43606 + }, + { + "epoch": 5.71, + "learning_rate": 3.991343583666879e-06, + "loss": 0.0052, + "step": 43607 + }, + { + "epoch": 5.71, + "learning_rate": 3.990919676756438e-06, + "loss": 0.0074, + "step": 43608 + }, + { + "epoch": 5.71, + "learning_rate": 3.990495786746214e-06, + "loss": 0.0058, + "step": 43609 + }, + { + "epoch": 5.71, + "learning_rate": 3.990071913637397e-06, + "loss": 0.0057, + "step": 43610 + }, + { + "epoch": 5.71, + "learning_rate": 3.989648057431182e-06, + "loss": 0.0048, + "step": 43611 + }, + { + "epoch": 5.71, + "learning_rate": 3.989224218128766e-06, + "loss": 0.0062, + "step": 43612 + }, + { + "epoch": 5.71, + "learning_rate": 3.988800395731337e-06, + "loss": 0.0075, + "step": 43613 + }, + { + "epoch": 5.71, + "learning_rate": 3.988376590240087e-06, + "loss": 0.0036, + "step": 43614 + }, + { + "epoch": 5.71, + "learning_rate": 3.987952801656204e-06, + "loss": 0.0072, + "step": 43615 + }, + { + "epoch": 5.71, + "learning_rate": 3.987529029980887e-06, + "loss": 0.0057, + "step": 43616 + }, + { + "epoch": 5.71, + "learning_rate": 3.9871052752153235e-06, + "loss": 0.0057, + "step": 43617 + }, + { + "epoch": 5.71, + "learning_rate": 3.986681537360704e-06, + "loss": 0.0101, + "step": 43618 + }, + { + "epoch": 5.71, + "learning_rate": 3.986257816418226e-06, + "loss": 0.007, + "step": 43619 + }, + { + "epoch": 5.71, + "learning_rate": 3.985834112389078e-06, + "loss": 0.0057, + "step": 43620 + }, + { + "epoch": 5.71, + "learning_rate": 3.985410425274446e-06, + "loss": 0.0065, + "step": 43621 + }, + { + "epoch": 5.71, + "learning_rate": 3.984986755075532e-06, + "loss": 0.0057, + "step": 43622 + }, + { + "epoch": 5.71, + "learning_rate": 3.984563101793517e-06, + "loss": 0.0073, + "step": 43623 + }, + { + "epoch": 5.71, + "learning_rate": 3.984139465429602e-06, + "loss": 0.0073, + "step": 43624 + }, + { + "epoch": 5.71, + "learning_rate": 3.983715845984977e-06, + "loss": 0.0055, + "step": 43625 + }, + { + "epoch": 5.71, + "learning_rate": 3.9832922434608244e-06, + "loss": 0.0065, + "step": 43626 + }, + { + "epoch": 5.71, + "learning_rate": 3.9828686578583465e-06, + "loss": 0.0061, + "step": 43627 + }, + { + "epoch": 5.71, + "learning_rate": 3.9824450891787305e-06, + "loss": 0.006, + "step": 43628 + }, + { + "epoch": 5.72, + "learning_rate": 3.9820215374231665e-06, + "loss": 0.0096, + "step": 43629 + }, + { + "epoch": 5.72, + "learning_rate": 3.981598002592844e-06, + "loss": 0.0035, + "step": 43630 + }, + { + "epoch": 5.72, + "learning_rate": 3.981174484688959e-06, + "loss": 0.0051, + "step": 43631 + }, + { + "epoch": 5.72, + "learning_rate": 3.980750983712697e-06, + "loss": 0.0059, + "step": 43632 + }, + { + "epoch": 5.72, + "learning_rate": 3.980327499665256e-06, + "loss": 0.0062, + "step": 43633 + }, + { + "epoch": 5.72, + "learning_rate": 3.97990403254782e-06, + "loss": 0.0066, + "step": 43634 + }, + { + "epoch": 5.72, + "learning_rate": 3.9794805823615855e-06, + "loss": 0.0057, + "step": 43635 + }, + { + "epoch": 5.72, + "learning_rate": 3.9790571491077435e-06, + "loss": 0.0059, + "step": 43636 + }, + { + "epoch": 5.72, + "learning_rate": 3.978633732787482e-06, + "loss": 0.0058, + "step": 43637 + }, + { + "epoch": 5.72, + "learning_rate": 3.978210333401987e-06, + "loss": 0.0058, + "step": 43638 + }, + { + "epoch": 5.72, + "learning_rate": 3.97778695095246e-06, + "loss": 0.0082, + "step": 43639 + }, + { + "epoch": 5.72, + "learning_rate": 3.977363585440086e-06, + "loss": 0.0066, + "step": 43640 + }, + { + "epoch": 5.72, + "learning_rate": 3.976940236866053e-06, + "loss": 0.0055, + "step": 43641 + }, + { + "epoch": 5.72, + "learning_rate": 3.976516905231557e-06, + "loss": 0.0115, + "step": 43642 + }, + { + "epoch": 5.72, + "learning_rate": 3.976093590537785e-06, + "loss": 0.0056, + "step": 43643 + }, + { + "epoch": 5.72, + "learning_rate": 3.9756702927859315e-06, + "loss": 0.0073, + "step": 43644 + }, + { + "epoch": 5.72, + "learning_rate": 3.975247011977185e-06, + "loss": 0.0079, + "step": 43645 + }, + { + "epoch": 5.72, + "learning_rate": 3.974823748112731e-06, + "loss": 0.006, + "step": 43646 + }, + { + "epoch": 5.72, + "learning_rate": 3.974400501193769e-06, + "loss": 0.0081, + "step": 43647 + }, + { + "epoch": 5.72, + "learning_rate": 3.9739772712214845e-06, + "loss": 0.0082, + "step": 43648 + }, + { + "epoch": 5.72, + "learning_rate": 3.9735540581970634e-06, + "loss": 0.0048, + "step": 43649 + }, + { + "epoch": 5.72, + "learning_rate": 3.973130862121705e-06, + "loss": 0.0058, + "step": 43650 + }, + { + "epoch": 5.72, + "learning_rate": 3.972707682996595e-06, + "loss": 0.0073, + "step": 43651 + }, + { + "epoch": 5.72, + "learning_rate": 3.972284520822924e-06, + "loss": 0.0066, + "step": 43652 + }, + { + "epoch": 5.72, + "learning_rate": 3.971861375601874e-06, + "loss": 0.0091, + "step": 43653 + }, + { + "epoch": 5.72, + "learning_rate": 3.97143824733465e-06, + "loss": 0.0092, + "step": 43654 + }, + { + "epoch": 5.72, + "learning_rate": 3.971015136022435e-06, + "loss": 0.0058, + "step": 43655 + }, + { + "epoch": 5.72, + "learning_rate": 3.970592041666419e-06, + "loss": 0.0066, + "step": 43656 + }, + { + "epoch": 5.72, + "learning_rate": 3.970168964267789e-06, + "loss": 0.0052, + "step": 43657 + }, + { + "epoch": 5.72, + "learning_rate": 3.96974590382774e-06, + "loss": 0.006, + "step": 43658 + }, + { + "epoch": 5.72, + "learning_rate": 3.9693228603474594e-06, + "loss": 0.0077, + "step": 43659 + }, + { + "epoch": 5.72, + "learning_rate": 3.968899833828135e-06, + "loss": 0.0045, + "step": 43660 + }, + { + "epoch": 5.72, + "learning_rate": 3.968476824270962e-06, + "loss": 0.0051, + "step": 43661 + }, + { + "epoch": 5.72, + "learning_rate": 3.968053831677126e-06, + "loss": 0.0041, + "step": 43662 + }, + { + "epoch": 5.72, + "learning_rate": 3.967630856047817e-06, + "loss": 0.0073, + "step": 43663 + }, + { + "epoch": 5.72, + "learning_rate": 3.967207897384221e-06, + "loss": 0.0074, + "step": 43664 + }, + { + "epoch": 5.72, + "learning_rate": 3.966784955687533e-06, + "loss": 0.0051, + "step": 43665 + }, + { + "epoch": 5.72, + "learning_rate": 3.966362030958943e-06, + "loss": 0.0076, + "step": 43666 + }, + { + "epoch": 5.72, + "learning_rate": 3.96593912319964e-06, + "loss": 0.0072, + "step": 43667 + }, + { + "epoch": 5.72, + "learning_rate": 3.965516232410808e-06, + "loss": 0.0088, + "step": 43668 + }, + { + "epoch": 5.72, + "learning_rate": 3.965093358593643e-06, + "loss": 0.0074, + "step": 43669 + }, + { + "epoch": 5.72, + "learning_rate": 3.964670501749333e-06, + "loss": 0.0055, + "step": 43670 + }, + { + "epoch": 5.72, + "learning_rate": 3.9642476618790644e-06, + "loss": 0.0081, + "step": 43671 + }, + { + "epoch": 5.72, + "learning_rate": 3.963824838984025e-06, + "loss": 0.0064, + "step": 43672 + }, + { + "epoch": 5.72, + "learning_rate": 3.96340203306541e-06, + "loss": 0.0054, + "step": 43673 + }, + { + "epoch": 5.72, + "learning_rate": 3.962979244124407e-06, + "loss": 0.0037, + "step": 43674 + }, + { + "epoch": 5.72, + "learning_rate": 3.962556472162197e-06, + "loss": 0.0048, + "step": 43675 + }, + { + "epoch": 5.72, + "learning_rate": 3.962133717179977e-06, + "loss": 0.0086, + "step": 43676 + }, + { + "epoch": 5.72, + "learning_rate": 3.961710979178938e-06, + "loss": 0.006, + "step": 43677 + }, + { + "epoch": 5.72, + "learning_rate": 3.961288258160264e-06, + "loss": 0.0065, + "step": 43678 + }, + { + "epoch": 5.72, + "learning_rate": 3.960865554125147e-06, + "loss": 0.0061, + "step": 43679 + }, + { + "epoch": 5.72, + "learning_rate": 3.960442867074769e-06, + "loss": 0.0043, + "step": 43680 + }, + { + "epoch": 5.72, + "learning_rate": 3.960020197010328e-06, + "loss": 0.0058, + "step": 43681 + }, + { + "epoch": 5.72, + "learning_rate": 3.95959754393301e-06, + "loss": 0.0049, + "step": 43682 + }, + { + "epoch": 5.72, + "learning_rate": 3.959174907843997e-06, + "loss": 0.0053, + "step": 43683 + }, + { + "epoch": 5.72, + "learning_rate": 3.958752288744486e-06, + "loss": 0.005, + "step": 43684 + }, + { + "epoch": 5.72, + "learning_rate": 3.958329686635664e-06, + "loss": 0.008, + "step": 43685 + }, + { + "epoch": 5.72, + "learning_rate": 3.957907101518713e-06, + "loss": 0.0064, + "step": 43686 + }, + { + "epoch": 5.72, + "learning_rate": 3.957484533394831e-06, + "loss": 0.007, + "step": 43687 + }, + { + "epoch": 5.72, + "learning_rate": 3.957061982265198e-06, + "loss": 0.0065, + "step": 43688 + }, + { + "epoch": 5.72, + "learning_rate": 3.95663944813101e-06, + "loss": 0.0084, + "step": 43689 + }, + { + "epoch": 5.72, + "learning_rate": 3.956216930993452e-06, + "loss": 0.0072, + "step": 43690 + }, + { + "epoch": 5.72, + "learning_rate": 3.955794430853707e-06, + "loss": 0.0068, + "step": 43691 + }, + { + "epoch": 5.72, + "learning_rate": 3.955371947712973e-06, + "loss": 0.0091, + "step": 43692 + }, + { + "epoch": 5.72, + "learning_rate": 3.9549494815724345e-06, + "loss": 0.0074, + "step": 43693 + }, + { + "epoch": 5.72, + "learning_rate": 3.9545270324332775e-06, + "loss": 0.0044, + "step": 43694 + }, + { + "epoch": 5.72, + "learning_rate": 3.954104600296687e-06, + "loss": 0.005, + "step": 43695 + }, + { + "epoch": 5.72, + "learning_rate": 3.953682185163861e-06, + "loss": 0.004, + "step": 43696 + }, + { + "epoch": 5.72, + "learning_rate": 3.953259787035977e-06, + "loss": 0.0049, + "step": 43697 + }, + { + "epoch": 5.72, + "learning_rate": 3.952837405914231e-06, + "loss": 0.008, + "step": 43698 + }, + { + "epoch": 5.72, + "learning_rate": 3.952415041799805e-06, + "loss": 0.0079, + "step": 43699 + }, + { + "epoch": 5.72, + "learning_rate": 3.951992694693893e-06, + "loss": 0.0065, + "step": 43700 + }, + { + "epoch": 5.72, + "learning_rate": 3.95157036459768e-06, + "loss": 0.0068, + "step": 43701 + }, + { + "epoch": 5.72, + "learning_rate": 3.9511480515123525e-06, + "loss": 0.0067, + "step": 43702 + }, + { + "epoch": 5.72, + "learning_rate": 3.950725755439096e-06, + "loss": 0.0056, + "step": 43703 + }, + { + "epoch": 5.72, + "learning_rate": 3.950303476379104e-06, + "loss": 0.0051, + "step": 43704 + }, + { + "epoch": 5.73, + "learning_rate": 3.9498812143335615e-06, + "loss": 0.0046, + "step": 43705 + }, + { + "epoch": 5.73, + "learning_rate": 3.949458969303653e-06, + "loss": 0.0085, + "step": 43706 + }, + { + "epoch": 5.73, + "learning_rate": 3.949036741290572e-06, + "loss": 0.0049, + "step": 43707 + }, + { + "epoch": 5.73, + "learning_rate": 3.948614530295499e-06, + "loss": 0.0074, + "step": 43708 + }, + { + "epoch": 5.73, + "learning_rate": 3.9481923363196295e-06, + "loss": 0.0063, + "step": 43709 + }, + { + "epoch": 5.73, + "learning_rate": 3.947770159364142e-06, + "loss": 0.0055, + "step": 43710 + }, + { + "epoch": 5.73, + "learning_rate": 3.947347999430235e-06, + "loss": 0.0069, + "step": 43711 + }, + { + "epoch": 5.73, + "learning_rate": 3.946925856519086e-06, + "loss": 0.0079, + "step": 43712 + }, + { + "epoch": 5.73, + "learning_rate": 3.9465037306318874e-06, + "loss": 0.006, + "step": 43713 + }, + { + "epoch": 5.73, + "learning_rate": 3.946081621769822e-06, + "loss": 0.0074, + "step": 43714 + }, + { + "epoch": 5.73, + "learning_rate": 3.945659529934082e-06, + "loss": 0.009, + "step": 43715 + }, + { + "epoch": 5.73, + "learning_rate": 3.945237455125852e-06, + "loss": 0.0077, + "step": 43716 + }, + { + "epoch": 5.73, + "learning_rate": 3.94481539734632e-06, + "loss": 0.0054, + "step": 43717 + }, + { + "epoch": 5.73, + "learning_rate": 3.9443933565966686e-06, + "loss": 0.0057, + "step": 43718 + }, + { + "epoch": 5.73, + "learning_rate": 3.943971332878088e-06, + "loss": 0.0071, + "step": 43719 + }, + { + "epoch": 5.73, + "learning_rate": 3.943549326191769e-06, + "loss": 0.0044, + "step": 43720 + }, + { + "epoch": 5.73, + "learning_rate": 3.9431273365388955e-06, + "loss": 0.0065, + "step": 43721 + }, + { + "epoch": 5.73, + "learning_rate": 3.942705363920649e-06, + "loss": 0.005, + "step": 43722 + }, + { + "epoch": 5.73, + "learning_rate": 3.9422834083382265e-06, + "loss": 0.0063, + "step": 43723 + }, + { + "epoch": 5.73, + "learning_rate": 3.941861469792808e-06, + "loss": 0.0073, + "step": 43724 + }, + { + "epoch": 5.73, + "learning_rate": 3.941439548285578e-06, + "loss": 0.0055, + "step": 43725 + }, + { + "epoch": 5.73, + "learning_rate": 3.9410176438177305e-06, + "loss": 0.0066, + "step": 43726 + }, + { + "epoch": 5.73, + "learning_rate": 3.940595756390449e-06, + "loss": 0.0074, + "step": 43727 + }, + { + "epoch": 5.73, + "learning_rate": 3.940173886004919e-06, + "loss": 0.0059, + "step": 43728 + }, + { + "epoch": 5.73, + "learning_rate": 3.939752032662323e-06, + "loss": 0.0063, + "step": 43729 + }, + { + "epoch": 5.73, + "learning_rate": 3.939330196363852e-06, + "loss": 0.007, + "step": 43730 + }, + { + "epoch": 5.73, + "learning_rate": 3.938908377110695e-06, + "loss": 0.0069, + "step": 43731 + }, + { + "epoch": 5.73, + "learning_rate": 3.938486574904036e-06, + "loss": 0.0062, + "step": 43732 + }, + { + "epoch": 5.73, + "learning_rate": 3.9380647897450575e-06, + "loss": 0.0073, + "step": 43733 + }, + { + "epoch": 5.73, + "learning_rate": 3.9376430216349515e-06, + "loss": 0.0065, + "step": 43734 + }, + { + "epoch": 5.73, + "learning_rate": 3.937221270574902e-06, + "loss": 0.0065, + "step": 43735 + }, + { + "epoch": 5.73, + "learning_rate": 3.936799536566096e-06, + "loss": 0.0059, + "step": 43736 + }, + { + "epoch": 5.73, + "learning_rate": 3.936377819609713e-06, + "loss": 0.0108, + "step": 43737 + }, + { + "epoch": 5.73, + "learning_rate": 3.93595611970695e-06, + "loss": 0.0078, + "step": 43738 + }, + { + "epoch": 5.73, + "learning_rate": 3.935534436858986e-06, + "loss": 0.0045, + "step": 43739 + }, + { + "epoch": 5.73, + "learning_rate": 3.935112771067006e-06, + "loss": 0.0055, + "step": 43740 + }, + { + "epoch": 5.73, + "learning_rate": 3.934691122332197e-06, + "loss": 0.005, + "step": 43741 + }, + { + "epoch": 5.73, + "learning_rate": 3.934269490655751e-06, + "loss": 0.0063, + "step": 43742 + }, + { + "epoch": 5.73, + "learning_rate": 3.933847876038849e-06, + "loss": 0.0078, + "step": 43743 + }, + { + "epoch": 5.73, + "learning_rate": 3.933426278482676e-06, + "loss": 0.0071, + "step": 43744 + }, + { + "epoch": 5.73, + "learning_rate": 3.933004697988415e-06, + "loss": 0.0051, + "step": 43745 + }, + { + "epoch": 5.73, + "learning_rate": 3.93258313455726e-06, + "loss": 0.007, + "step": 43746 + }, + { + "epoch": 5.73, + "learning_rate": 3.9321615881903895e-06, + "loss": 0.008, + "step": 43747 + }, + { + "epoch": 5.73, + "learning_rate": 3.93174005888899e-06, + "loss": 0.0054, + "step": 43748 + }, + { + "epoch": 5.73, + "learning_rate": 3.931318546654252e-06, + "loss": 0.008, + "step": 43749 + }, + { + "epoch": 5.73, + "learning_rate": 3.930897051487356e-06, + "loss": 0.0058, + "step": 43750 + }, + { + "epoch": 5.73, + "learning_rate": 3.93047557338949e-06, + "loss": 0.0063, + "step": 43751 + }, + { + "epoch": 5.73, + "learning_rate": 3.930054112361833e-06, + "loss": 0.0038, + "step": 43752 + }, + { + "epoch": 5.73, + "learning_rate": 3.929632668405577e-06, + "loss": 0.0062, + "step": 43753 + }, + { + "epoch": 5.73, + "learning_rate": 3.9292112415219095e-06, + "loss": 0.0072, + "step": 43754 + }, + { + "epoch": 5.73, + "learning_rate": 3.928789831712011e-06, + "loss": 0.007, + "step": 43755 + }, + { + "epoch": 5.73, + "learning_rate": 3.928368438977066e-06, + "loss": 0.0073, + "step": 43756 + }, + { + "epoch": 5.73, + "learning_rate": 3.927947063318264e-06, + "loss": 0.0056, + "step": 43757 + }, + { + "epoch": 5.73, + "learning_rate": 3.927525704736787e-06, + "loss": 0.0058, + "step": 43758 + }, + { + "epoch": 5.73, + "learning_rate": 3.927104363233822e-06, + "loss": 0.0056, + "step": 43759 + }, + { + "epoch": 5.73, + "learning_rate": 3.926683038810547e-06, + "loss": 0.0052, + "step": 43760 + }, + { + "epoch": 5.73, + "learning_rate": 3.926261731468157e-06, + "loss": 0.0054, + "step": 43761 + }, + { + "epoch": 5.73, + "learning_rate": 3.925840441207833e-06, + "loss": 0.0078, + "step": 43762 + }, + { + "epoch": 5.73, + "learning_rate": 3.9254191680307556e-06, + "loss": 0.006, + "step": 43763 + }, + { + "epoch": 5.73, + "learning_rate": 3.924997911938113e-06, + "loss": 0.0074, + "step": 43764 + }, + { + "epoch": 5.73, + "learning_rate": 3.924576672931093e-06, + "loss": 0.007, + "step": 43765 + }, + { + "epoch": 5.73, + "learning_rate": 3.924155451010879e-06, + "loss": 0.0051, + "step": 43766 + }, + { + "epoch": 5.73, + "learning_rate": 3.923734246178653e-06, + "loss": 0.0068, + "step": 43767 + }, + { + "epoch": 5.73, + "learning_rate": 3.9233130584355985e-06, + "loss": 0.0059, + "step": 43768 + }, + { + "epoch": 5.73, + "learning_rate": 3.922891887782906e-06, + "loss": 0.0067, + "step": 43769 + }, + { + "epoch": 5.73, + "learning_rate": 3.922470734221755e-06, + "loss": 0.0053, + "step": 43770 + }, + { + "epoch": 5.73, + "learning_rate": 3.922049597753329e-06, + "loss": 0.0051, + "step": 43771 + }, + { + "epoch": 5.73, + "learning_rate": 3.921628478378819e-06, + "loss": 0.0053, + "step": 43772 + }, + { + "epoch": 5.73, + "learning_rate": 3.921207376099404e-06, + "loss": 0.0068, + "step": 43773 + }, + { + "epoch": 5.73, + "learning_rate": 3.920786290916267e-06, + "loss": 0.0079, + "step": 43774 + }, + { + "epoch": 5.73, + "learning_rate": 3.920365222830593e-06, + "loss": 0.0052, + "step": 43775 + }, + { + "epoch": 5.73, + "learning_rate": 3.919944171843573e-06, + "loss": 0.0076, + "step": 43776 + }, + { + "epoch": 5.73, + "learning_rate": 3.919523137956386e-06, + "loss": 0.0071, + "step": 43777 + }, + { + "epoch": 5.73, + "learning_rate": 3.919102121170216e-06, + "loss": 0.0055, + "step": 43778 + }, + { + "epoch": 5.73, + "learning_rate": 3.918681121486244e-06, + "loss": 0.0053, + "step": 43779 + }, + { + "epoch": 5.73, + "learning_rate": 3.9182601389056615e-06, + "loss": 0.0077, + "step": 43780 + }, + { + "epoch": 5.74, + "learning_rate": 3.917839173429649e-06, + "loss": 0.0069, + "step": 43781 + }, + { + "epoch": 5.74, + "learning_rate": 3.917418225059388e-06, + "loss": 0.005, + "step": 43782 + }, + { + "epoch": 5.74, + "learning_rate": 3.916997293796063e-06, + "loss": 0.0048, + "step": 43783 + }, + { + "epoch": 5.74, + "learning_rate": 3.916576379640858e-06, + "loss": 0.0047, + "step": 43784 + }, + { + "epoch": 5.74, + "learning_rate": 3.916155482594962e-06, + "loss": 0.0098, + "step": 43785 + }, + { + "epoch": 5.74, + "learning_rate": 3.9157346026595545e-06, + "loss": 0.006, + "step": 43786 + }, + { + "epoch": 5.74, + "learning_rate": 3.915313739835815e-06, + "loss": 0.0058, + "step": 43787 + }, + { + "epoch": 5.74, + "learning_rate": 3.914892894124937e-06, + "loss": 0.0071, + "step": 43788 + }, + { + "epoch": 5.74, + "learning_rate": 3.914472065528096e-06, + "loss": 0.0063, + "step": 43789 + }, + { + "epoch": 5.74, + "learning_rate": 3.914051254046477e-06, + "loss": 0.0063, + "step": 43790 + }, + { + "epoch": 5.74, + "learning_rate": 3.913630459681268e-06, + "loss": 0.0092, + "step": 43791 + }, + { + "epoch": 5.74, + "learning_rate": 3.913209682433649e-06, + "loss": 0.0097, + "step": 43792 + }, + { + "epoch": 5.74, + "learning_rate": 3.912788922304803e-06, + "loss": 0.0058, + "step": 43793 + }, + { + "epoch": 5.74, + "learning_rate": 3.912368179295911e-06, + "loss": 0.0065, + "step": 43794 + }, + { + "epoch": 5.74, + "learning_rate": 3.911947453408159e-06, + "loss": 0.005, + "step": 43795 + }, + { + "epoch": 5.74, + "learning_rate": 3.911526744642734e-06, + "loss": 0.0068, + "step": 43796 + }, + { + "epoch": 5.74, + "learning_rate": 3.911106053000816e-06, + "loss": 0.0076, + "step": 43797 + }, + { + "epoch": 5.74, + "learning_rate": 3.910685378483585e-06, + "loss": 0.0056, + "step": 43798 + }, + { + "epoch": 5.74, + "learning_rate": 3.910264721092231e-06, + "loss": 0.0059, + "step": 43799 + }, + { + "epoch": 5.74, + "learning_rate": 3.9098440808279316e-06, + "loss": 0.0057, + "step": 43800 + }, + { + "epoch": 5.74, + "learning_rate": 3.909423457691872e-06, + "loss": 0.0113, + "step": 43801 + }, + { + "epoch": 5.74, + "learning_rate": 3.9090028516852315e-06, + "loss": 0.0099, + "step": 43802 + }, + { + "epoch": 5.74, + "learning_rate": 3.9085822628092e-06, + "loss": 0.0059, + "step": 43803 + }, + { + "epoch": 5.74, + "learning_rate": 3.908161691064956e-06, + "loss": 0.0091, + "step": 43804 + }, + { + "epoch": 5.74, + "learning_rate": 3.9077411364536785e-06, + "loss": 0.0076, + "step": 43805 + }, + { + "epoch": 5.74, + "learning_rate": 3.907320598976555e-06, + "loss": 0.005, + "step": 43806 + }, + { + "epoch": 5.74, + "learning_rate": 3.906900078634772e-06, + "loss": 0.0079, + "step": 43807 + }, + { + "epoch": 5.74, + "learning_rate": 3.9064795754295074e-06, + "loss": 0.0076, + "step": 43808 + }, + { + "epoch": 5.74, + "learning_rate": 3.9060590893619455e-06, + "loss": 0.004, + "step": 43809 + }, + { + "epoch": 5.74, + "learning_rate": 3.905638620433263e-06, + "loss": 0.0056, + "step": 43810 + }, + { + "epoch": 5.74, + "learning_rate": 3.905218168644652e-06, + "loss": 0.0047, + "step": 43811 + }, + { + "epoch": 5.74, + "learning_rate": 3.904797733997289e-06, + "loss": 0.0064, + "step": 43812 + }, + { + "epoch": 5.74, + "learning_rate": 3.904377316492356e-06, + "loss": 0.0052, + "step": 43813 + }, + { + "epoch": 5.74, + "learning_rate": 3.90395691613104e-06, + "loss": 0.0071, + "step": 43814 + }, + { + "epoch": 5.74, + "learning_rate": 3.90353653291452e-06, + "loss": 0.0056, + "step": 43815 + }, + { + "epoch": 5.74, + "learning_rate": 3.903116166843979e-06, + "loss": 0.0054, + "step": 43816 + }, + { + "epoch": 5.74, + "learning_rate": 3.902695817920596e-06, + "loss": 0.0064, + "step": 43817 + }, + { + "epoch": 5.74, + "learning_rate": 3.902275486145556e-06, + "loss": 0.0059, + "step": 43818 + }, + { + "epoch": 5.74, + "learning_rate": 3.901855171520045e-06, + "loss": 0.0062, + "step": 43819 + }, + { + "epoch": 5.74, + "learning_rate": 3.901434874045242e-06, + "loss": 0.0077, + "step": 43820 + }, + { + "epoch": 5.74, + "learning_rate": 3.901014593722325e-06, + "loss": 0.0079, + "step": 43821 + }, + { + "epoch": 5.74, + "learning_rate": 3.900594330552483e-06, + "loss": 0.004, + "step": 43822 + }, + { + "epoch": 5.74, + "learning_rate": 3.900174084536894e-06, + "loss": 0.0064, + "step": 43823 + }, + { + "epoch": 5.74, + "learning_rate": 3.899753855676742e-06, + "loss": 0.0059, + "step": 43824 + }, + { + "epoch": 5.74, + "learning_rate": 3.899333643973203e-06, + "loss": 0.0079, + "step": 43825 + }, + { + "epoch": 5.74, + "learning_rate": 3.898913449427467e-06, + "loss": 0.0057, + "step": 43826 + }, + { + "epoch": 5.74, + "learning_rate": 3.898493272040712e-06, + "loss": 0.0078, + "step": 43827 + }, + { + "epoch": 5.74, + "learning_rate": 3.898073111814116e-06, + "loss": 0.006, + "step": 43828 + }, + { + "epoch": 5.74, + "learning_rate": 3.897652968748864e-06, + "loss": 0.0077, + "step": 43829 + }, + { + "epoch": 5.74, + "learning_rate": 3.897232842846143e-06, + "loss": 0.0062, + "step": 43830 + }, + { + "epoch": 5.74, + "learning_rate": 3.896812734107128e-06, + "loss": 0.0068, + "step": 43831 + }, + { + "epoch": 5.74, + "learning_rate": 3.896392642533004e-06, + "loss": 0.0066, + "step": 43832 + }, + { + "epoch": 5.74, + "learning_rate": 3.895972568124945e-06, + "loss": 0.0065, + "step": 43833 + }, + { + "epoch": 5.74, + "learning_rate": 3.895552510884143e-06, + "loss": 0.006, + "step": 43834 + }, + { + "epoch": 5.74, + "learning_rate": 3.895132470811774e-06, + "loss": 0.0065, + "step": 43835 + }, + { + "epoch": 5.74, + "learning_rate": 3.894712447909017e-06, + "loss": 0.0058, + "step": 43836 + }, + { + "epoch": 5.74, + "learning_rate": 3.894292442177059e-06, + "loss": 0.0078, + "step": 43837 + }, + { + "epoch": 5.74, + "learning_rate": 3.893872453617078e-06, + "loss": 0.009, + "step": 43838 + }, + { + "epoch": 5.74, + "learning_rate": 3.893452482230252e-06, + "loss": 0.0086, + "step": 43839 + }, + { + "epoch": 5.74, + "learning_rate": 3.8930325280177674e-06, + "loss": 0.0048, + "step": 43840 + }, + { + "epoch": 5.74, + "learning_rate": 3.892612590980806e-06, + "loss": 0.0088, + "step": 43841 + }, + { + "epoch": 5.74, + "learning_rate": 3.892192671120546e-06, + "loss": 0.005, + "step": 43842 + }, + { + "epoch": 5.74, + "learning_rate": 3.891772768438169e-06, + "loss": 0.0044, + "step": 43843 + }, + { + "epoch": 5.74, + "learning_rate": 3.891352882934852e-06, + "loss": 0.0056, + "step": 43844 + }, + { + "epoch": 5.74, + "learning_rate": 3.890933014611784e-06, + "loss": 0.0072, + "step": 43845 + }, + { + "epoch": 5.74, + "learning_rate": 3.890513163470142e-06, + "loss": 0.0059, + "step": 43846 + }, + { + "epoch": 5.74, + "learning_rate": 3.890093329511106e-06, + "loss": 0.0059, + "step": 43847 + }, + { + "epoch": 5.74, + "learning_rate": 3.889673512735852e-06, + "loss": 0.0057, + "step": 43848 + }, + { + "epoch": 5.74, + "learning_rate": 3.889253713145571e-06, + "loss": 0.005, + "step": 43849 + }, + { + "epoch": 5.74, + "learning_rate": 3.888833930741435e-06, + "loss": 0.0048, + "step": 43850 + }, + { + "epoch": 5.74, + "learning_rate": 3.888414165524631e-06, + "loss": 0.0058, + "step": 43851 + }, + { + "epoch": 5.74, + "learning_rate": 3.887994417496334e-06, + "loss": 0.0068, + "step": 43852 + }, + { + "epoch": 5.74, + "learning_rate": 3.88757468665773e-06, + "loss": 0.0056, + "step": 43853 + }, + { + "epoch": 5.74, + "learning_rate": 3.8871549730099965e-06, + "loss": 0.0067, + "step": 43854 + }, + { + "epoch": 5.74, + "learning_rate": 3.886735276554311e-06, + "loss": 0.0065, + "step": 43855 + }, + { + "epoch": 5.74, + "learning_rate": 3.8863155972918606e-06, + "loss": 0.0058, + "step": 43856 + }, + { + "epoch": 5.74, + "learning_rate": 3.885895935223822e-06, + "loss": 0.0064, + "step": 43857 + }, + { + "epoch": 5.75, + "learning_rate": 3.885476290351375e-06, + "loss": 0.0064, + "step": 43858 + }, + { + "epoch": 5.75, + "learning_rate": 3.885056662675698e-06, + "loss": 0.0055, + "step": 43859 + }, + { + "epoch": 5.75, + "learning_rate": 3.8846370521979755e-06, + "loss": 0.0062, + "step": 43860 + }, + { + "epoch": 5.75, + "learning_rate": 3.8842174589193836e-06, + "loss": 0.0076, + "step": 43861 + }, + { + "epoch": 5.75, + "learning_rate": 3.883797882841107e-06, + "loss": 0.0085, + "step": 43862 + }, + { + "epoch": 5.75, + "learning_rate": 3.88337832396432e-06, + "loss": 0.0059, + "step": 43863 + }, + { + "epoch": 5.75, + "learning_rate": 3.882958782290209e-06, + "loss": 0.0089, + "step": 43864 + }, + { + "epoch": 5.75, + "learning_rate": 3.882539257819951e-06, + "loss": 0.0062, + "step": 43865 + }, + { + "epoch": 5.75, + "learning_rate": 3.882119750554725e-06, + "loss": 0.0061, + "step": 43866 + }, + { + "epoch": 5.75, + "learning_rate": 3.8817002604957075e-06, + "loss": 0.006, + "step": 43867 + }, + { + "epoch": 5.75, + "learning_rate": 3.8812807876440875e-06, + "loss": 0.005, + "step": 43868 + }, + { + "epoch": 5.75, + "learning_rate": 3.880861332001038e-06, + "loss": 0.0094, + "step": 43869 + }, + { + "epoch": 5.75, + "learning_rate": 3.880441893567737e-06, + "loss": 0.0068, + "step": 43870 + }, + { + "epoch": 5.75, + "learning_rate": 3.88002247234537e-06, + "loss": 0.0106, + "step": 43871 + }, + { + "epoch": 5.75, + "learning_rate": 3.8796030683351114e-06, + "loss": 0.004, + "step": 43872 + }, + { + "epoch": 5.75, + "learning_rate": 3.879183681538147e-06, + "loss": 0.0058, + "step": 43873 + }, + { + "epoch": 5.75, + "learning_rate": 3.878764311955652e-06, + "loss": 0.0062, + "step": 43874 + }, + { + "epoch": 5.75, + "learning_rate": 3.878344959588803e-06, + "loss": 0.0068, + "step": 43875 + }, + { + "epoch": 5.75, + "learning_rate": 3.877925624438786e-06, + "loss": 0.006, + "step": 43876 + }, + { + "epoch": 5.75, + "learning_rate": 3.877506306506778e-06, + "loss": 0.006, + "step": 43877 + }, + { + "epoch": 5.75, + "learning_rate": 3.877087005793953e-06, + "loss": 0.005, + "step": 43878 + }, + { + "epoch": 5.75, + "learning_rate": 3.876667722301498e-06, + "loss": 0.0063, + "step": 43879 + }, + { + "epoch": 5.75, + "learning_rate": 3.8762484560305904e-06, + "loss": 0.0054, + "step": 43880 + }, + { + "epoch": 5.75, + "learning_rate": 3.875829206982406e-06, + "loss": 0.0038, + "step": 43881 + }, + { + "epoch": 5.75, + "learning_rate": 3.875409975158125e-06, + "loss": 0.0056, + "step": 43882 + }, + { + "epoch": 5.75, + "learning_rate": 3.8749907605589245e-06, + "loss": 0.0061, + "step": 43883 + }, + { + "epoch": 5.75, + "learning_rate": 3.874571563185991e-06, + "loss": 0.0043, + "step": 43884 + }, + { + "epoch": 5.75, + "learning_rate": 3.874152383040499e-06, + "loss": 0.011, + "step": 43885 + }, + { + "epoch": 5.75, + "learning_rate": 3.873733220123623e-06, + "loss": 0.0094, + "step": 43886 + }, + { + "epoch": 5.75, + "learning_rate": 3.8733140744365505e-06, + "loss": 0.0049, + "step": 43887 + }, + { + "epoch": 5.75, + "learning_rate": 3.8728949459804555e-06, + "loss": 0.0041, + "step": 43888 + }, + { + "epoch": 5.75, + "learning_rate": 3.872475834756517e-06, + "loss": 0.0064, + "step": 43889 + }, + { + "epoch": 5.75, + "learning_rate": 3.872056740765911e-06, + "loss": 0.0047, + "step": 43890 + }, + { + "epoch": 5.75, + "learning_rate": 3.871637664009822e-06, + "loss": 0.0054, + "step": 43891 + }, + { + "epoch": 5.75, + "learning_rate": 3.871218604489425e-06, + "loss": 0.0042, + "step": 43892 + }, + { + "epoch": 5.75, + "learning_rate": 3.870799562205897e-06, + "loss": 0.0057, + "step": 43893 + }, + { + "epoch": 5.75, + "learning_rate": 3.870380537160419e-06, + "loss": 0.0051, + "step": 43894 + }, + { + "epoch": 5.75, + "learning_rate": 3.8699615293541715e-06, + "loss": 0.0083, + "step": 43895 + }, + { + "epoch": 5.75, + "learning_rate": 3.869542538788333e-06, + "loss": 0.0054, + "step": 43896 + }, + { + "epoch": 5.75, + "learning_rate": 3.869123565464077e-06, + "loss": 0.0058, + "step": 43897 + }, + { + "epoch": 5.75, + "learning_rate": 3.868704609382582e-06, + "loss": 0.005, + "step": 43898 + }, + { + "epoch": 5.75, + "learning_rate": 3.8682856705450324e-06, + "loss": 0.0064, + "step": 43899 + }, + { + "epoch": 5.75, + "learning_rate": 3.867866748952604e-06, + "loss": 0.0072, + "step": 43900 + }, + { + "epoch": 5.75, + "learning_rate": 3.867447844606468e-06, + "loss": 0.0056, + "step": 43901 + }, + { + "epoch": 5.75, + "learning_rate": 3.8670289575078135e-06, + "loss": 0.005, + "step": 43902 + }, + { + "epoch": 5.75, + "learning_rate": 3.866610087657813e-06, + "loss": 0.0085, + "step": 43903 + }, + { + "epoch": 5.75, + "learning_rate": 3.866191235057645e-06, + "loss": 0.0083, + "step": 43904 + }, + { + "epoch": 5.75, + "learning_rate": 3.86577239970848e-06, + "loss": 0.0062, + "step": 43905 + }, + { + "epoch": 5.75, + "learning_rate": 3.8653535816115116e-06, + "loss": 0.0071, + "step": 43906 + }, + { + "epoch": 5.75, + "learning_rate": 3.864934780767909e-06, + "loss": 0.0072, + "step": 43907 + }, + { + "epoch": 5.75, + "learning_rate": 3.8645159971788505e-06, + "loss": 0.0081, + "step": 43908 + }, + { + "epoch": 5.75, + "learning_rate": 3.86409723084551e-06, + "loss": 0.0084, + "step": 43909 + }, + { + "epoch": 5.75, + "learning_rate": 3.863678481769073e-06, + "loss": 0.0071, + "step": 43910 + }, + { + "epoch": 5.75, + "learning_rate": 3.863259749950714e-06, + "loss": 0.0073, + "step": 43911 + }, + { + "epoch": 5.75, + "learning_rate": 3.86284103539161e-06, + "loss": 0.0061, + "step": 43912 + }, + { + "epoch": 5.75, + "learning_rate": 3.862422338092936e-06, + "loss": 0.0053, + "step": 43913 + }, + { + "epoch": 5.75, + "learning_rate": 3.862003658055875e-06, + "loss": 0.0092, + "step": 43914 + }, + { + "epoch": 5.75, + "learning_rate": 3.861584995281597e-06, + "loss": 0.0055, + "step": 43915 + }, + { + "epoch": 5.75, + "learning_rate": 3.861166349771291e-06, + "loss": 0.0061, + "step": 43916 + }, + { + "epoch": 5.75, + "learning_rate": 3.860747721526121e-06, + "loss": 0.0088, + "step": 43917 + }, + { + "epoch": 5.75, + "learning_rate": 3.860329110547277e-06, + "loss": 0.0064, + "step": 43918 + }, + { + "epoch": 5.75, + "learning_rate": 3.85991051683593e-06, + "loss": 0.006, + "step": 43919 + }, + { + "epoch": 5.75, + "learning_rate": 3.859491940393253e-06, + "loss": 0.0063, + "step": 43920 + }, + { + "epoch": 5.75, + "learning_rate": 3.859073381220432e-06, + "loss": 0.0049, + "step": 43921 + }, + { + "epoch": 5.75, + "learning_rate": 3.8586548393186404e-06, + "loss": 0.0064, + "step": 43922 + }, + { + "epoch": 5.75, + "learning_rate": 3.8582363146890545e-06, + "loss": 0.0076, + "step": 43923 + }, + { + "epoch": 5.75, + "learning_rate": 3.857817807332849e-06, + "loss": 0.0059, + "step": 43924 + }, + { + "epoch": 5.75, + "learning_rate": 3.857399317251207e-06, + "loss": 0.0037, + "step": 43925 + }, + { + "epoch": 5.75, + "learning_rate": 3.856980844445298e-06, + "loss": 0.0042, + "step": 43926 + }, + { + "epoch": 5.75, + "learning_rate": 3.856562388916308e-06, + "loss": 0.0067, + "step": 43927 + }, + { + "epoch": 5.75, + "learning_rate": 3.856143950665405e-06, + "loss": 0.0065, + "step": 43928 + }, + { + "epoch": 5.75, + "learning_rate": 3.855725529693774e-06, + "loss": 0.0081, + "step": 43929 + }, + { + "epoch": 5.75, + "learning_rate": 3.855307126002588e-06, + "loss": 0.0052, + "step": 43930 + }, + { + "epoch": 5.75, + "learning_rate": 3.854888739593022e-06, + "loss": 0.007, + "step": 43931 + }, + { + "epoch": 5.75, + "learning_rate": 3.854470370466253e-06, + "loss": 0.0062, + "step": 43932 + }, + { + "epoch": 5.75, + "learning_rate": 3.85405201862346e-06, + "loss": 0.0079, + "step": 43933 + }, + { + "epoch": 5.76, + "learning_rate": 3.853633684065819e-06, + "loss": 0.0075, + "step": 43934 + }, + { + "epoch": 5.76, + "learning_rate": 3.853215366794502e-06, + "loss": 0.0073, + "step": 43935 + }, + { + "epoch": 5.76, + "learning_rate": 3.852797066810694e-06, + "loss": 0.0057, + "step": 43936 + }, + { + "epoch": 5.76, + "learning_rate": 3.852378784115564e-06, + "loss": 0.0102, + "step": 43937 + }, + { + "epoch": 5.76, + "learning_rate": 3.851960518710294e-06, + "loss": 0.006, + "step": 43938 + }, + { + "epoch": 5.76, + "learning_rate": 3.851542270596057e-06, + "loss": 0.0051, + "step": 43939 + }, + { + "epoch": 5.76, + "learning_rate": 3.851124039774027e-06, + "loss": 0.0055, + "step": 43940 + }, + { + "epoch": 5.76, + "learning_rate": 3.850705826245386e-06, + "loss": 0.0072, + "step": 43941 + }, + { + "epoch": 5.76, + "learning_rate": 3.8502876300113075e-06, + "loss": 0.0062, + "step": 43942 + }, + { + "epoch": 5.76, + "learning_rate": 3.849869451072965e-06, + "loss": 0.0032, + "step": 43943 + }, + { + "epoch": 5.76, + "learning_rate": 3.84945128943154e-06, + "loss": 0.0056, + "step": 43944 + }, + { + "epoch": 5.76, + "learning_rate": 3.849033145088207e-06, + "loss": 0.0065, + "step": 43945 + }, + { + "epoch": 5.76, + "learning_rate": 3.84861501804414e-06, + "loss": 0.0059, + "step": 43946 + }, + { + "epoch": 5.76, + "learning_rate": 3.848196908300511e-06, + "loss": 0.0065, + "step": 43947 + }, + { + "epoch": 5.76, + "learning_rate": 3.8477788158585015e-06, + "loss": 0.0048, + "step": 43948 + }, + { + "epoch": 5.76, + "learning_rate": 3.84736074071929e-06, + "loss": 0.0052, + "step": 43949 + }, + { + "epoch": 5.76, + "learning_rate": 3.846942682884048e-06, + "loss": 0.0068, + "step": 43950 + }, + { + "epoch": 5.76, + "learning_rate": 3.846524642353949e-06, + "loss": 0.0071, + "step": 43951 + }, + { + "epoch": 5.76, + "learning_rate": 3.846106619130177e-06, + "loss": 0.0048, + "step": 43952 + }, + { + "epoch": 5.76, + "learning_rate": 3.845688613213901e-06, + "loss": 0.0056, + "step": 43953 + }, + { + "epoch": 5.76, + "learning_rate": 3.8452706246062965e-06, + "loss": 0.0091, + "step": 43954 + }, + { + "epoch": 5.76, + "learning_rate": 3.844852653308538e-06, + "loss": 0.0084, + "step": 43955 + }, + { + "epoch": 5.76, + "learning_rate": 3.844434699321809e-06, + "loss": 0.0046, + "step": 43956 + }, + { + "epoch": 5.76, + "learning_rate": 3.844016762647278e-06, + "loss": 0.0049, + "step": 43957 + }, + { + "epoch": 5.76, + "learning_rate": 3.8435988432861185e-06, + "loss": 0.0053, + "step": 43958 + }, + { + "epoch": 5.76, + "learning_rate": 3.84318094123951e-06, + "loss": 0.0083, + "step": 43959 + }, + { + "epoch": 5.76, + "learning_rate": 3.84276305650863e-06, + "loss": 0.0047, + "step": 43960 + }, + { + "epoch": 5.76, + "learning_rate": 3.842345189094652e-06, + "loss": 0.0057, + "step": 43961 + }, + { + "epoch": 5.76, + "learning_rate": 3.841927338998749e-06, + "loss": 0.005, + "step": 43962 + }, + { + "epoch": 5.76, + "learning_rate": 3.841509506222095e-06, + "loss": 0.0048, + "step": 43963 + }, + { + "epoch": 5.76, + "learning_rate": 3.84109169076587e-06, + "loss": 0.0079, + "step": 43964 + }, + { + "epoch": 5.76, + "learning_rate": 3.840673892631248e-06, + "loss": 0.0079, + "step": 43965 + }, + { + "epoch": 5.76, + "learning_rate": 3.8402561118193985e-06, + "loss": 0.0062, + "step": 43966 + }, + { + "epoch": 5.76, + "learning_rate": 3.839838348331504e-06, + "loss": 0.0064, + "step": 43967 + }, + { + "epoch": 5.76, + "learning_rate": 3.839420602168737e-06, + "loss": 0.0049, + "step": 43968 + }, + { + "epoch": 5.76, + "learning_rate": 3.83900287333227e-06, + "loss": 0.007, + "step": 43969 + }, + { + "epoch": 5.76, + "learning_rate": 3.838585161823273e-06, + "loss": 0.0083, + "step": 43970 + }, + { + "epoch": 5.76, + "learning_rate": 3.838167467642933e-06, + "loss": 0.0052, + "step": 43971 + }, + { + "epoch": 5.76, + "learning_rate": 3.8377497907924214e-06, + "loss": 0.0073, + "step": 43972 + }, + { + "epoch": 5.76, + "learning_rate": 3.837332131272908e-06, + "loss": 0.0066, + "step": 43973 + }, + { + "epoch": 5.76, + "learning_rate": 3.836914489085568e-06, + "loss": 0.006, + "step": 43974 + }, + { + "epoch": 5.76, + "learning_rate": 3.836496864231581e-06, + "loss": 0.0058, + "step": 43975 + }, + { + "epoch": 5.76, + "learning_rate": 3.836079256712117e-06, + "loss": 0.0071, + "step": 43976 + }, + { + "epoch": 5.76, + "learning_rate": 3.8356616665283495e-06, + "loss": 0.0065, + "step": 43977 + }, + { + "epoch": 5.76, + "learning_rate": 3.835244093681458e-06, + "loss": 0.0066, + "step": 43978 + }, + { + "epoch": 5.76, + "learning_rate": 3.834826538172616e-06, + "loss": 0.0071, + "step": 43979 + }, + { + "epoch": 5.76, + "learning_rate": 3.834409000002994e-06, + "loss": 0.009, + "step": 43980 + }, + { + "epoch": 5.76, + "learning_rate": 3.833991479173766e-06, + "loss": 0.0081, + "step": 43981 + }, + { + "epoch": 5.76, + "learning_rate": 3.833573975686108e-06, + "loss": 0.0064, + "step": 43982 + }, + { + "epoch": 5.76, + "learning_rate": 3.833156489541199e-06, + "loss": 0.0085, + "step": 43983 + }, + { + "epoch": 5.76, + "learning_rate": 3.832739020740208e-06, + "loss": 0.0057, + "step": 43984 + }, + { + "epoch": 5.76, + "learning_rate": 3.832321569284308e-06, + "loss": 0.0031, + "step": 43985 + }, + { + "epoch": 5.76, + "learning_rate": 3.831904135174677e-06, + "loss": 0.0042, + "step": 43986 + }, + { + "epoch": 5.76, + "learning_rate": 3.831486718412487e-06, + "loss": 0.0046, + "step": 43987 + }, + { + "epoch": 5.76, + "learning_rate": 3.831069318998913e-06, + "loss": 0.0066, + "step": 43988 + }, + { + "epoch": 5.76, + "learning_rate": 3.830651936935125e-06, + "loss": 0.0047, + "step": 43989 + }, + { + "epoch": 5.76, + "learning_rate": 3.830234572222304e-06, + "loss": 0.0077, + "step": 43990 + }, + { + "epoch": 5.76, + "learning_rate": 3.829817224861617e-06, + "loss": 0.0077, + "step": 43991 + }, + { + "epoch": 5.76, + "learning_rate": 3.829399894854239e-06, + "loss": 0.0054, + "step": 43992 + }, + { + "epoch": 5.76, + "learning_rate": 3.828982582201345e-06, + "loss": 0.0072, + "step": 43993 + }, + { + "epoch": 5.76, + "learning_rate": 3.8285652869041125e-06, + "loss": 0.0072, + "step": 43994 + }, + { + "epoch": 5.76, + "learning_rate": 3.82814800896371e-06, + "loss": 0.0054, + "step": 43995 + }, + { + "epoch": 5.76, + "learning_rate": 3.827730748381314e-06, + "loss": 0.0064, + "step": 43996 + }, + { + "epoch": 5.76, + "learning_rate": 3.8273135051580936e-06, + "loss": 0.0076, + "step": 43997 + }, + { + "epoch": 5.76, + "learning_rate": 3.826896279295227e-06, + "loss": 0.0054, + "step": 43998 + }, + { + "epoch": 5.76, + "learning_rate": 3.826479070793887e-06, + "loss": 0.0058, + "step": 43999 + }, + { + "epoch": 5.76, + "learning_rate": 3.826061879655242e-06, + "loss": 0.0054, + "step": 44000 + }, + { + "epoch": 5.76, + "learning_rate": 3.825644705880474e-06, + "loss": 0.008, + "step": 44001 + }, + { + "epoch": 5.76, + "learning_rate": 3.825227549470749e-06, + "loss": 0.0069, + "step": 44002 + }, + { + "epoch": 5.76, + "learning_rate": 3.824810410427241e-06, + "loss": 0.0068, + "step": 44003 + }, + { + "epoch": 5.76, + "learning_rate": 3.824393288751128e-06, + "loss": 0.0055, + "step": 44004 + }, + { + "epoch": 5.76, + "learning_rate": 3.823976184443576e-06, + "loss": 0.0063, + "step": 44005 + }, + { + "epoch": 5.76, + "learning_rate": 3.823559097505765e-06, + "loss": 0.0051, + "step": 44006 + }, + { + "epoch": 5.76, + "learning_rate": 3.823142027938867e-06, + "loss": 0.0086, + "step": 44007 + }, + { + "epoch": 5.76, + "learning_rate": 3.8227249757440485e-06, + "loss": 0.0069, + "step": 44008 + }, + { + "epoch": 5.76, + "learning_rate": 3.82230794092249e-06, + "loss": 0.0063, + "step": 44009 + }, + { + "epoch": 5.76, + "learning_rate": 3.821890923475363e-06, + "loss": 0.0064, + "step": 44010 + }, + { + "epoch": 5.77, + "learning_rate": 3.8214739234038375e-06, + "loss": 0.0057, + "step": 44011 + }, + { + "epoch": 5.77, + "learning_rate": 3.821056940709085e-06, + "loss": 0.0066, + "step": 44012 + }, + { + "epoch": 5.77, + "learning_rate": 3.820639975392284e-06, + "loss": 0.0056, + "step": 44013 + }, + { + "epoch": 5.77, + "learning_rate": 3.8202230274546e-06, + "loss": 0.0057, + "step": 44014 + }, + { + "epoch": 5.77, + "learning_rate": 3.819806096897214e-06, + "loss": 0.0059, + "step": 44015 + }, + { + "epoch": 5.77, + "learning_rate": 3.81938918372129e-06, + "loss": 0.006, + "step": 44016 + }, + { + "epoch": 5.77, + "learning_rate": 3.818972287928008e-06, + "loss": 0.0041, + "step": 44017 + }, + { + "epoch": 5.77, + "learning_rate": 3.818555409518537e-06, + "loss": 0.0052, + "step": 44018 + }, + { + "epoch": 5.77, + "learning_rate": 3.818138548494051e-06, + "loss": 0.0058, + "step": 44019 + }, + { + "epoch": 5.77, + "learning_rate": 3.8177217048557166e-06, + "loss": 0.0097, + "step": 44020 + }, + { + "epoch": 5.77, + "learning_rate": 3.8173048786047154e-06, + "loss": 0.0063, + "step": 44021 + }, + { + "epoch": 5.77, + "learning_rate": 3.816888069742214e-06, + "loss": 0.0051, + "step": 44022 + }, + { + "epoch": 5.77, + "learning_rate": 3.816471278269383e-06, + "loss": 0.0077, + "step": 44023 + }, + { + "epoch": 5.77, + "learning_rate": 3.8160545041874e-06, + "loss": 0.0071, + "step": 44024 + }, + { + "epoch": 5.77, + "learning_rate": 3.8156377474974304e-06, + "loss": 0.0057, + "step": 44025 + }, + { + "epoch": 5.77, + "learning_rate": 3.815221008200655e-06, + "loss": 0.0077, + "step": 44026 + }, + { + "epoch": 5.77, + "learning_rate": 3.814804286298237e-06, + "loss": 0.0066, + "step": 44027 + }, + { + "epoch": 5.77, + "learning_rate": 3.814387581791356e-06, + "loss": 0.0078, + "step": 44028 + }, + { + "epoch": 5.77, + "learning_rate": 3.8139708946811804e-06, + "loss": 0.0051, + "step": 44029 + }, + { + "epoch": 5.77, + "learning_rate": 3.8135542249688826e-06, + "loss": 0.0082, + "step": 44030 + }, + { + "epoch": 5.77, + "learning_rate": 3.813137572655631e-06, + "loss": 0.0053, + "step": 44031 + }, + { + "epoch": 5.77, + "learning_rate": 3.8127209377426035e-06, + "loss": 0.0071, + "step": 44032 + }, + { + "epoch": 5.77, + "learning_rate": 3.8123043202309684e-06, + "loss": 0.0045, + "step": 44033 + }, + { + "epoch": 5.77, + "learning_rate": 3.8118877201218984e-06, + "loss": 0.0072, + "step": 44034 + }, + { + "epoch": 5.77, + "learning_rate": 3.8114711374165616e-06, + "loss": 0.0064, + "step": 44035 + }, + { + "epoch": 5.77, + "learning_rate": 3.811054572116132e-06, + "loss": 0.0077, + "step": 44036 + }, + { + "epoch": 5.77, + "learning_rate": 3.8106380242217857e-06, + "loss": 0.0082, + "step": 44037 + }, + { + "epoch": 5.77, + "learning_rate": 3.810221493734689e-06, + "loss": 0.0052, + "step": 44038 + }, + { + "epoch": 5.77, + "learning_rate": 3.8098049806560132e-06, + "loss": 0.0058, + "step": 44039 + }, + { + "epoch": 5.77, + "learning_rate": 3.809388484986933e-06, + "loss": 0.0053, + "step": 44040 + }, + { + "epoch": 5.77, + "learning_rate": 3.8089720067286196e-06, + "loss": 0.0062, + "step": 44041 + }, + { + "epoch": 5.77, + "learning_rate": 3.808555545882239e-06, + "loss": 0.0049, + "step": 44042 + }, + { + "epoch": 5.77, + "learning_rate": 3.8081391024489687e-06, + "loss": 0.0049, + "step": 44043 + }, + { + "epoch": 5.77, + "learning_rate": 3.807722676429978e-06, + "loss": 0.0068, + "step": 44044 + }, + { + "epoch": 5.77, + "learning_rate": 3.807306267826438e-06, + "loss": 0.0061, + "step": 44045 + }, + { + "epoch": 5.77, + "learning_rate": 3.806889876639516e-06, + "loss": 0.0071, + "step": 44046 + }, + { + "epoch": 5.77, + "learning_rate": 3.806473502870385e-06, + "loss": 0.0045, + "step": 44047 + }, + { + "epoch": 5.77, + "learning_rate": 3.8060571465202223e-06, + "loss": 0.0089, + "step": 44048 + }, + { + "epoch": 5.77, + "learning_rate": 3.805640807590193e-06, + "loss": 0.0043, + "step": 44049 + }, + { + "epoch": 5.77, + "learning_rate": 3.8052244860814656e-06, + "loss": 0.0078, + "step": 44050 + }, + { + "epoch": 5.77, + "learning_rate": 3.8048081819952187e-06, + "loss": 0.0068, + "step": 44051 + }, + { + "epoch": 5.77, + "learning_rate": 3.8043918953326187e-06, + "loss": 0.0061, + "step": 44052 + }, + { + "epoch": 5.77, + "learning_rate": 3.803975626094837e-06, + "loss": 0.0072, + "step": 44053 + }, + { + "epoch": 5.77, + "learning_rate": 3.803559374283039e-06, + "loss": 0.0057, + "step": 44054 + }, + { + "epoch": 5.77, + "learning_rate": 3.8031431398984054e-06, + "loss": 0.0041, + "step": 44055 + }, + { + "epoch": 5.77, + "learning_rate": 3.8027269229421006e-06, + "loss": 0.0071, + "step": 44056 + }, + { + "epoch": 5.77, + "learning_rate": 3.8023107234152935e-06, + "loss": 0.0061, + "step": 44057 + }, + { + "epoch": 5.77, + "learning_rate": 3.8018945413191565e-06, + "loss": 0.0081, + "step": 44058 + }, + { + "epoch": 5.77, + "learning_rate": 3.8014783766548647e-06, + "loss": 0.0073, + "step": 44059 + }, + { + "epoch": 5.77, + "learning_rate": 3.8010622294235855e-06, + "loss": 0.007, + "step": 44060 + }, + { + "epoch": 5.77, + "learning_rate": 3.800646099626488e-06, + "loss": 0.0078, + "step": 44061 + }, + { + "epoch": 5.77, + "learning_rate": 3.8002299872647407e-06, + "loss": 0.0056, + "step": 44062 + }, + { + "epoch": 5.77, + "learning_rate": 3.7998138923395188e-06, + "loss": 0.0051, + "step": 44063 + }, + { + "epoch": 5.77, + "learning_rate": 3.799397814851992e-06, + "loss": 0.0051, + "step": 44064 + }, + { + "epoch": 5.77, + "learning_rate": 3.7989817548033236e-06, + "loss": 0.0054, + "step": 44065 + }, + { + "epoch": 5.77, + "learning_rate": 3.798565712194692e-06, + "loss": 0.007, + "step": 44066 + }, + { + "epoch": 5.77, + "learning_rate": 3.7981496870272648e-06, + "loss": 0.0068, + "step": 44067 + }, + { + "epoch": 5.77, + "learning_rate": 3.7977336793022078e-06, + "loss": 0.0068, + "step": 44068 + }, + { + "epoch": 5.77, + "learning_rate": 3.7973176890206974e-06, + "loss": 0.0053, + "step": 44069 + }, + { + "epoch": 5.77, + "learning_rate": 3.7969017161838984e-06, + "loss": 0.006, + "step": 44070 + }, + { + "epoch": 5.77, + "learning_rate": 3.796485760792985e-06, + "loss": 0.0043, + "step": 44071 + }, + { + "epoch": 5.77, + "learning_rate": 3.7960698228491268e-06, + "loss": 0.0075, + "step": 44072 + }, + { + "epoch": 5.77, + "learning_rate": 3.7956539023534865e-06, + "loss": 0.0044, + "step": 44073 + }, + { + "epoch": 5.77, + "learning_rate": 3.7952379993072442e-06, + "loss": 0.0035, + "step": 44074 + }, + { + "epoch": 5.77, + "learning_rate": 3.7948221137115637e-06, + "loss": 0.007, + "step": 44075 + }, + { + "epoch": 5.77, + "learning_rate": 3.7944062455676156e-06, + "loss": 0.0089, + "step": 44076 + }, + { + "epoch": 5.77, + "learning_rate": 3.7939903948765666e-06, + "loss": 0.0051, + "step": 44077 + }, + { + "epoch": 5.77, + "learning_rate": 3.7935745616395915e-06, + "loss": 0.0052, + "step": 44078 + }, + { + "epoch": 5.77, + "learning_rate": 3.7931587458578543e-06, + "loss": 0.0067, + "step": 44079 + }, + { + "epoch": 5.77, + "learning_rate": 3.792742947532532e-06, + "loss": 0.0096, + "step": 44080 + }, + { + "epoch": 5.77, + "learning_rate": 3.7923271666647855e-06, + "loss": 0.0064, + "step": 44081 + }, + { + "epoch": 5.77, + "learning_rate": 3.7919114032557925e-06, + "loss": 0.005, + "step": 44082 + }, + { + "epoch": 5.77, + "learning_rate": 3.7914956573067173e-06, + "loss": 0.0065, + "step": 44083 + }, + { + "epoch": 5.77, + "learning_rate": 3.7910799288187294e-06, + "loss": 0.0052, + "step": 44084 + }, + { + "epoch": 5.77, + "learning_rate": 3.7906642177929966e-06, + "loss": 0.007, + "step": 44085 + }, + { + "epoch": 5.77, + "learning_rate": 3.790248524230692e-06, + "loss": 0.005, + "step": 44086 + }, + { + "epoch": 5.78, + "learning_rate": 3.7898328481329836e-06, + "loss": 0.0083, + "step": 44087 + }, + { + "epoch": 5.78, + "learning_rate": 3.7894171895010357e-06, + "loss": 0.0064, + "step": 44088 + }, + { + "epoch": 5.78, + "learning_rate": 3.7890015483360245e-06, + "loss": 0.0055, + "step": 44089 + }, + { + "epoch": 5.78, + "learning_rate": 3.7885859246391133e-06, + "loss": 0.0058, + "step": 44090 + }, + { + "epoch": 5.78, + "learning_rate": 3.788170318411476e-06, + "loss": 0.0074, + "step": 44091 + }, + { + "epoch": 5.78, + "learning_rate": 3.787754729654275e-06, + "loss": 0.0066, + "step": 44092 + }, + { + "epoch": 5.78, + "learning_rate": 3.787339158368687e-06, + "loss": 0.0047, + "step": 44093 + }, + { + "epoch": 5.78, + "learning_rate": 3.7869236045558766e-06, + "loss": 0.0052, + "step": 44094 + }, + { + "epoch": 5.78, + "learning_rate": 3.786508068217013e-06, + "loss": 0.0051, + "step": 44095 + }, + { + "epoch": 5.78, + "learning_rate": 3.786092549353262e-06, + "loss": 0.0066, + "step": 44096 + }, + { + "epoch": 5.78, + "learning_rate": 3.7856770479657966e-06, + "loss": 0.0059, + "step": 44097 + }, + { + "epoch": 5.78, + "learning_rate": 3.7852615640557842e-06, + "loss": 0.0084, + "step": 44098 + }, + { + "epoch": 5.78, + "learning_rate": 3.784846097624393e-06, + "loss": 0.0088, + "step": 44099 + }, + { + "epoch": 5.78, + "learning_rate": 3.7844306486727876e-06, + "loss": 0.007, + "step": 44100 + }, + { + "epoch": 5.78, + "learning_rate": 3.7840152172021395e-06, + "loss": 0.0056, + "step": 44101 + }, + { + "epoch": 5.78, + "learning_rate": 3.783599803213621e-06, + "loss": 0.0095, + "step": 44102 + }, + { + "epoch": 5.78, + "learning_rate": 3.783184406708398e-06, + "loss": 0.008, + "step": 44103 + }, + { + "epoch": 5.78, + "learning_rate": 3.782769027687634e-06, + "loss": 0.0061, + "step": 44104 + }, + { + "epoch": 5.78, + "learning_rate": 3.782353666152504e-06, + "loss": 0.0051, + "step": 44105 + }, + { + "epoch": 5.78, + "learning_rate": 3.781938322104173e-06, + "loss": 0.0049, + "step": 44106 + }, + { + "epoch": 5.78, + "learning_rate": 3.781522995543806e-06, + "loss": 0.0055, + "step": 44107 + }, + { + "epoch": 5.78, + "learning_rate": 3.781107686472578e-06, + "loss": 0.0051, + "step": 44108 + }, + { + "epoch": 5.78, + "learning_rate": 3.780692394891654e-06, + "loss": 0.0053, + "step": 44109 + }, + { + "epoch": 5.78, + "learning_rate": 3.7802771208022017e-06, + "loss": 0.006, + "step": 44110 + }, + { + "epoch": 5.78, + "learning_rate": 3.779861864205384e-06, + "loss": 0.0093, + "step": 44111 + }, + { + "epoch": 5.78, + "learning_rate": 3.7794466251023743e-06, + "loss": 0.0066, + "step": 44112 + }, + { + "epoch": 5.78, + "learning_rate": 3.779031403494343e-06, + "loss": 0.0052, + "step": 44113 + }, + { + "epoch": 5.78, + "learning_rate": 3.7786161993824554e-06, + "loss": 0.0063, + "step": 44114 + }, + { + "epoch": 5.78, + "learning_rate": 3.778201012767874e-06, + "loss": 0.0066, + "step": 44115 + }, + { + "epoch": 5.78, + "learning_rate": 3.7777858436517746e-06, + "loss": 0.0053, + "step": 44116 + }, + { + "epoch": 5.78, + "learning_rate": 3.7773706920353213e-06, + "loss": 0.0047, + "step": 44117 + }, + { + "epoch": 5.78, + "learning_rate": 3.7769555579196813e-06, + "loss": 0.007, + "step": 44118 + }, + { + "epoch": 5.78, + "learning_rate": 3.7765404413060193e-06, + "loss": 0.0058, + "step": 44119 + }, + { + "epoch": 5.78, + "learning_rate": 3.776125342195509e-06, + "loss": 0.0113, + "step": 44120 + }, + { + "epoch": 5.78, + "learning_rate": 3.7757102605893148e-06, + "loss": 0.0063, + "step": 44121 + }, + { + "epoch": 5.78, + "learning_rate": 3.775295196488601e-06, + "loss": 0.0075, + "step": 44122 + }, + { + "epoch": 5.78, + "learning_rate": 3.7748801498945375e-06, + "loss": 0.0049, + "step": 44123 + }, + { + "epoch": 5.78, + "learning_rate": 3.774465120808296e-06, + "loss": 0.0103, + "step": 44124 + }, + { + "epoch": 5.78, + "learning_rate": 3.77405010923104e-06, + "loss": 0.0059, + "step": 44125 + }, + { + "epoch": 5.78, + "learning_rate": 3.773635115163935e-06, + "loss": 0.0058, + "step": 44126 + }, + { + "epoch": 5.78, + "learning_rate": 3.7732201386081478e-06, + "loss": 0.0082, + "step": 44127 + }, + { + "epoch": 5.78, + "learning_rate": 3.7728051795648502e-06, + "loss": 0.0045, + "step": 44128 + }, + { + "epoch": 5.78, + "learning_rate": 3.772390238035206e-06, + "loss": 0.0043, + "step": 44129 + }, + { + "epoch": 5.78, + "learning_rate": 3.7719753140203797e-06, + "loss": 0.0067, + "step": 44130 + }, + { + "epoch": 5.78, + "learning_rate": 3.771560407521545e-06, + "loss": 0.0047, + "step": 44131 + }, + { + "epoch": 5.78, + "learning_rate": 3.771145518539865e-06, + "loss": 0.0054, + "step": 44132 + }, + { + "epoch": 5.78, + "learning_rate": 3.7707306470765058e-06, + "loss": 0.0061, + "step": 44133 + }, + { + "epoch": 5.78, + "learning_rate": 3.7703157931326315e-06, + "loss": 0.0051, + "step": 44134 + }, + { + "epoch": 5.78, + "learning_rate": 3.7699009567094123e-06, + "loss": 0.0056, + "step": 44135 + }, + { + "epoch": 5.78, + "learning_rate": 3.7694861378080183e-06, + "loss": 0.0076, + "step": 44136 + }, + { + "epoch": 5.78, + "learning_rate": 3.769071336429613e-06, + "loss": 0.008, + "step": 44137 + }, + { + "epoch": 5.78, + "learning_rate": 3.7686565525753594e-06, + "loss": 0.0069, + "step": 44138 + }, + { + "epoch": 5.78, + "learning_rate": 3.7682417862464306e-06, + "loss": 0.0048, + "step": 44139 + }, + { + "epoch": 5.78, + "learning_rate": 3.767827037443991e-06, + "loss": 0.0057, + "step": 44140 + }, + { + "epoch": 5.78, + "learning_rate": 3.7674123061692047e-06, + "loss": 0.0079, + "step": 44141 + }, + { + "epoch": 5.78, + "learning_rate": 3.766997592423236e-06, + "loss": 0.0067, + "step": 44142 + }, + { + "epoch": 5.78, + "learning_rate": 3.7665828962072582e-06, + "loss": 0.0059, + "step": 44143 + }, + { + "epoch": 5.78, + "learning_rate": 3.7661682175224345e-06, + "loss": 0.0068, + "step": 44144 + }, + { + "epoch": 5.78, + "learning_rate": 3.7657535563699265e-06, + "loss": 0.0069, + "step": 44145 + }, + { + "epoch": 5.78, + "learning_rate": 3.7653389127509056e-06, + "loss": 0.004, + "step": 44146 + }, + { + "epoch": 5.78, + "learning_rate": 3.764924286666539e-06, + "loss": 0.0059, + "step": 44147 + }, + { + "epoch": 5.78, + "learning_rate": 3.7645096781179924e-06, + "loss": 0.0042, + "step": 44148 + }, + { + "epoch": 5.78, + "learning_rate": 3.76409508710643e-06, + "loss": 0.0061, + "step": 44149 + }, + { + "epoch": 5.78, + "learning_rate": 3.7636805136330136e-06, + "loss": 0.0066, + "step": 44150 + }, + { + "epoch": 5.78, + "learning_rate": 3.7632659576989183e-06, + "loss": 0.0081, + "step": 44151 + }, + { + "epoch": 5.78, + "learning_rate": 3.7628514193053046e-06, + "loss": 0.0076, + "step": 44152 + }, + { + "epoch": 5.78, + "learning_rate": 3.7624368984533354e-06, + "loss": 0.0069, + "step": 44153 + }, + { + "epoch": 5.78, + "learning_rate": 3.7620223951441847e-06, + "loss": 0.0058, + "step": 44154 + }, + { + "epoch": 5.78, + "learning_rate": 3.7616079093790137e-06, + "loss": 0.0061, + "step": 44155 + }, + { + "epoch": 5.78, + "learning_rate": 3.7611934411589846e-06, + "loss": 0.0072, + "step": 44156 + }, + { + "epoch": 5.78, + "learning_rate": 3.7607789904852663e-06, + "loss": 0.0067, + "step": 44157 + }, + { + "epoch": 5.78, + "learning_rate": 3.7603645573590286e-06, + "loss": 0.0074, + "step": 44158 + }, + { + "epoch": 5.78, + "learning_rate": 3.759950141781433e-06, + "loss": 0.0091, + "step": 44159 + }, + { + "epoch": 5.78, + "learning_rate": 3.759535743753645e-06, + "loss": 0.006, + "step": 44160 + }, + { + "epoch": 5.78, + "learning_rate": 3.7591213632768275e-06, + "loss": 0.0069, + "step": 44161 + }, + { + "epoch": 5.78, + "learning_rate": 3.758707000352153e-06, + "loss": 0.0059, + "step": 44162 + }, + { + "epoch": 5.79, + "learning_rate": 3.758292654980782e-06, + "loss": 0.0067, + "step": 44163 + }, + { + "epoch": 5.79, + "learning_rate": 3.7578783271638796e-06, + "loss": 0.0056, + "step": 44164 + }, + { + "epoch": 5.79, + "learning_rate": 3.75746401690261e-06, + "loss": 0.0061, + "step": 44165 + }, + { + "epoch": 5.79, + "learning_rate": 3.7570497241981428e-06, + "loss": 0.0074, + "step": 44166 + }, + { + "epoch": 5.79, + "learning_rate": 3.756635449051639e-06, + "loss": 0.0072, + "step": 44167 + }, + { + "epoch": 5.79, + "learning_rate": 3.7562211914642677e-06, + "loss": 0.0045, + "step": 44168 + }, + { + "epoch": 5.79, + "learning_rate": 3.7558069514371876e-06, + "loss": 0.0085, + "step": 44169 + }, + { + "epoch": 5.79, + "learning_rate": 3.755392728971573e-06, + "loss": 0.0052, + "step": 44170 + }, + { + "epoch": 5.79, + "learning_rate": 3.754978524068582e-06, + "loss": 0.0056, + "step": 44171 + }, + { + "epoch": 5.79, + "learning_rate": 3.7545643367293795e-06, + "loss": 0.008, + "step": 44172 + }, + { + "epoch": 5.79, + "learning_rate": 3.754150166955135e-06, + "loss": 0.0052, + "step": 44173 + }, + { + "epoch": 5.79, + "learning_rate": 3.7537360147470104e-06, + "loss": 0.0058, + "step": 44174 + }, + { + "epoch": 5.79, + "learning_rate": 3.7533218801061712e-06, + "loss": 0.0096, + "step": 44175 + }, + { + "epoch": 5.79, + "learning_rate": 3.7529077630337775e-06, + "loss": 0.0058, + "step": 44176 + }, + { + "epoch": 5.79, + "learning_rate": 3.7524936635309982e-06, + "loss": 0.0047, + "step": 44177 + }, + { + "epoch": 5.79, + "learning_rate": 3.7520795815990018e-06, + "loss": 0.0052, + "step": 44178 + }, + { + "epoch": 5.79, + "learning_rate": 3.751665517238948e-06, + "loss": 0.0045, + "step": 44179 + }, + { + "epoch": 5.79, + "learning_rate": 3.751251470452e-06, + "loss": 0.0074, + "step": 44180 + }, + { + "epoch": 5.79, + "learning_rate": 3.7508374412393265e-06, + "loss": 0.0056, + "step": 44181 + }, + { + "epoch": 5.79, + "learning_rate": 3.7504234296020904e-06, + "loss": 0.0082, + "step": 44182 + }, + { + "epoch": 5.79, + "learning_rate": 3.7500094355414553e-06, + "loss": 0.0068, + "step": 44183 + }, + { + "epoch": 5.79, + "learning_rate": 3.7495954590585827e-06, + "loss": 0.0074, + "step": 44184 + }, + { + "epoch": 5.79, + "learning_rate": 3.749181500154644e-06, + "loss": 0.0048, + "step": 44185 + }, + { + "epoch": 5.79, + "learning_rate": 3.7487675588307994e-06, + "loss": 0.0063, + "step": 44186 + }, + { + "epoch": 5.79, + "learning_rate": 3.7483536350882088e-06, + "loss": 0.0068, + "step": 44187 + }, + { + "epoch": 5.79, + "learning_rate": 3.7479397289280405e-06, + "loss": 0.0065, + "step": 44188 + }, + { + "epoch": 5.79, + "learning_rate": 3.7475258403514634e-06, + "loss": 0.0062, + "step": 44189 + }, + { + "epoch": 5.79, + "learning_rate": 3.747111969359636e-06, + "loss": 0.0057, + "step": 44190 + }, + { + "epoch": 5.79, + "learning_rate": 3.746698115953724e-06, + "loss": 0.0069, + "step": 44191 + }, + { + "epoch": 5.79, + "learning_rate": 3.7462842801348864e-06, + "loss": 0.0045, + "step": 44192 + }, + { + "epoch": 5.79, + "learning_rate": 3.7458704619042952e-06, + "loss": 0.0068, + "step": 44193 + }, + { + "epoch": 5.79, + "learning_rate": 3.7454566612631094e-06, + "loss": 0.0064, + "step": 44194 + }, + { + "epoch": 5.79, + "learning_rate": 3.7450428782124914e-06, + "loss": 0.0037, + "step": 44195 + }, + { + "epoch": 5.79, + "learning_rate": 3.7446291127536093e-06, + "loss": 0.0062, + "step": 44196 + }, + { + "epoch": 5.79, + "learning_rate": 3.7442153648876256e-06, + "loss": 0.0043, + "step": 44197 + }, + { + "epoch": 5.79, + "learning_rate": 3.7438016346157034e-06, + "loss": 0.0046, + "step": 44198 + }, + { + "epoch": 5.79, + "learning_rate": 3.7433879219390022e-06, + "loss": 0.0074, + "step": 44199 + }, + { + "epoch": 5.79, + "learning_rate": 3.742974226858689e-06, + "loss": 0.0069, + "step": 44200 + }, + { + "epoch": 5.79, + "learning_rate": 3.742560549375931e-06, + "loss": 0.0046, + "step": 44201 + }, + { + "epoch": 5.79, + "learning_rate": 3.7421468894918876e-06, + "loss": 0.0049, + "step": 44202 + }, + { + "epoch": 5.79, + "learning_rate": 3.7417332472077205e-06, + "loss": 0.0071, + "step": 44203 + }, + { + "epoch": 5.79, + "learning_rate": 3.7413196225245983e-06, + "loss": 0.0076, + "step": 44204 + }, + { + "epoch": 5.79, + "learning_rate": 3.7409060154436817e-06, + "loss": 0.0062, + "step": 44205 + }, + { + "epoch": 5.79, + "learning_rate": 3.7404924259661334e-06, + "loss": 0.0042, + "step": 44206 + }, + { + "epoch": 5.79, + "learning_rate": 3.740078854093113e-06, + "loss": 0.0089, + "step": 44207 + }, + { + "epoch": 5.79, + "learning_rate": 3.739665299825792e-06, + "loss": 0.0058, + "step": 44208 + }, + { + "epoch": 5.79, + "learning_rate": 3.739251763165328e-06, + "loss": 0.0078, + "step": 44209 + }, + { + "epoch": 5.79, + "learning_rate": 3.7388382441128823e-06, + "loss": 0.0058, + "step": 44210 + }, + { + "epoch": 5.79, + "learning_rate": 3.7384247426696196e-06, + "loss": 0.0072, + "step": 44211 + }, + { + "epoch": 5.79, + "learning_rate": 3.738011258836708e-06, + "loss": 0.0066, + "step": 44212 + }, + { + "epoch": 5.79, + "learning_rate": 3.7375977926153064e-06, + "loss": 0.0065, + "step": 44213 + }, + { + "epoch": 5.79, + "learning_rate": 3.7371843440065767e-06, + "loss": 0.0073, + "step": 44214 + }, + { + "epoch": 5.79, + "learning_rate": 3.7367709130116804e-06, + "loss": 0.0057, + "step": 44215 + }, + { + "epoch": 5.79, + "learning_rate": 3.736357499631784e-06, + "loss": 0.0059, + "step": 44216 + }, + { + "epoch": 5.79, + "learning_rate": 3.73594410386805e-06, + "loss": 0.0078, + "step": 44217 + }, + { + "epoch": 5.79, + "learning_rate": 3.7355307257216357e-06, + "loss": 0.0064, + "step": 44218 + }, + { + "epoch": 5.79, + "learning_rate": 3.735117365193711e-06, + "loss": 0.0082, + "step": 44219 + }, + { + "epoch": 5.79, + "learning_rate": 3.7347040222854357e-06, + "loss": 0.0055, + "step": 44220 + }, + { + "epoch": 5.79, + "learning_rate": 3.7342906969979675e-06, + "loss": 0.0059, + "step": 44221 + }, + { + "epoch": 5.79, + "learning_rate": 3.733877389332473e-06, + "loss": 0.0065, + "step": 44222 + }, + { + "epoch": 5.79, + "learning_rate": 3.7334640992901196e-06, + "loss": 0.0057, + "step": 44223 + }, + { + "epoch": 5.79, + "learning_rate": 3.7330508268720632e-06, + "loss": 0.0055, + "step": 44224 + }, + { + "epoch": 5.79, + "learning_rate": 3.7326375720794673e-06, + "loss": 0.0034, + "step": 44225 + }, + { + "epoch": 5.79, + "learning_rate": 3.7322243349134924e-06, + "loss": 0.0063, + "step": 44226 + }, + { + "epoch": 5.79, + "learning_rate": 3.7318111153753057e-06, + "loss": 0.0069, + "step": 44227 + }, + { + "epoch": 5.79, + "learning_rate": 3.7313979134660662e-06, + "loss": 0.0063, + "step": 44228 + }, + { + "epoch": 5.79, + "learning_rate": 3.7309847291869363e-06, + "loss": 0.0066, + "step": 44229 + }, + { + "epoch": 5.79, + "learning_rate": 3.7305715625390737e-06, + "loss": 0.0056, + "step": 44230 + }, + { + "epoch": 5.79, + "learning_rate": 3.730158413523649e-06, + "loss": 0.0073, + "step": 44231 + }, + { + "epoch": 5.79, + "learning_rate": 3.729745282141817e-06, + "loss": 0.0045, + "step": 44232 + }, + { + "epoch": 5.79, + "learning_rate": 3.7293321683947447e-06, + "loss": 0.007, + "step": 44233 + }, + { + "epoch": 5.79, + "learning_rate": 3.728919072283589e-06, + "loss": 0.007, + "step": 44234 + }, + { + "epoch": 5.79, + "learning_rate": 3.728505993809518e-06, + "loss": 0.0053, + "step": 44235 + }, + { + "epoch": 5.79, + "learning_rate": 3.7280929329736903e-06, + "loss": 0.0063, + "step": 44236 + }, + { + "epoch": 5.79, + "learning_rate": 3.7276798897772625e-06, + "loss": 0.0062, + "step": 44237 + }, + { + "epoch": 5.79, + "learning_rate": 3.727266864221406e-06, + "loss": 0.005, + "step": 44238 + }, + { + "epoch": 5.79, + "learning_rate": 3.7268538563072766e-06, + "loss": 0.0051, + "step": 44239 + }, + { + "epoch": 5.8, + "learning_rate": 3.7264408660360364e-06, + "loss": 0.0049, + "step": 44240 + }, + { + "epoch": 5.8, + "learning_rate": 3.7260278934088436e-06, + "loss": 0.0074, + "step": 44241 + }, + { + "epoch": 5.8, + "learning_rate": 3.725614938426867e-06, + "loss": 0.0052, + "step": 44242 + }, + { + "epoch": 5.8, + "learning_rate": 3.7252020010912613e-06, + "loss": 0.0047, + "step": 44243 + }, + { + "epoch": 5.8, + "learning_rate": 3.724789081403194e-06, + "loss": 0.0091, + "step": 44244 + }, + { + "epoch": 5.8, + "learning_rate": 3.724376179363819e-06, + "loss": 0.006, + "step": 44245 + }, + { + "epoch": 5.8, + "learning_rate": 3.7239632949743067e-06, + "loss": 0.0069, + "step": 44246 + }, + { + "epoch": 5.8, + "learning_rate": 3.7235504282358127e-06, + "loss": 0.0046, + "step": 44247 + }, + { + "epoch": 5.8, + "learning_rate": 3.7231375791494984e-06, + "loss": 0.0072, + "step": 44248 + }, + { + "epoch": 5.8, + "learning_rate": 3.7227247477165217e-06, + "loss": 0.0081, + "step": 44249 + }, + { + "epoch": 5.8, + "learning_rate": 3.722311933938052e-06, + "loss": 0.0061, + "step": 44250 + }, + { + "epoch": 5.8, + "learning_rate": 3.7218991378152447e-06, + "loss": 0.0053, + "step": 44251 + }, + { + "epoch": 5.8, + "learning_rate": 3.721486359349259e-06, + "loss": 0.0062, + "step": 44252 + }, + { + "epoch": 5.8, + "learning_rate": 3.7210735985412617e-06, + "loss": 0.0061, + "step": 44253 + }, + { + "epoch": 5.8, + "learning_rate": 3.7206608553924073e-06, + "loss": 0.0048, + "step": 44254 + }, + { + "epoch": 5.8, + "learning_rate": 3.720248129903863e-06, + "loss": 0.0053, + "step": 44255 + }, + { + "epoch": 5.8, + "learning_rate": 3.7198354220767864e-06, + "loss": 0.0066, + "step": 44256 + }, + { + "epoch": 5.8, + "learning_rate": 3.7194227319123357e-06, + "loss": 0.0073, + "step": 44257 + }, + { + "epoch": 5.8, + "learning_rate": 3.7190100594116763e-06, + "loss": 0.0047, + "step": 44258 + }, + { + "epoch": 5.8, + "learning_rate": 3.7185974045759665e-06, + "loss": 0.0062, + "step": 44259 + }, + { + "epoch": 5.8, + "learning_rate": 3.718184767406364e-06, + "loss": 0.0056, + "step": 44260 + }, + { + "epoch": 5.8, + "learning_rate": 3.7177721479040362e-06, + "loss": 0.0049, + "step": 44261 + }, + { + "epoch": 5.8, + "learning_rate": 3.7173595460701396e-06, + "loss": 0.0055, + "step": 44262 + }, + { + "epoch": 5.8, + "learning_rate": 3.716946961905834e-06, + "loss": 0.0059, + "step": 44263 + }, + { + "epoch": 5.8, + "learning_rate": 3.716534395412278e-06, + "loss": 0.0072, + "step": 44264 + }, + { + "epoch": 5.8, + "learning_rate": 3.7161218465906337e-06, + "loss": 0.0061, + "step": 44265 + }, + { + "epoch": 5.8, + "learning_rate": 3.7157093154420664e-06, + "loss": 0.0065, + "step": 44266 + }, + { + "epoch": 5.8, + "learning_rate": 3.7152968019677304e-06, + "loss": 0.0057, + "step": 44267 + }, + { + "epoch": 5.8, + "learning_rate": 3.7148843061687855e-06, + "loss": 0.0052, + "step": 44268 + }, + { + "epoch": 5.8, + "learning_rate": 3.7144718280463967e-06, + "loss": 0.0066, + "step": 44269 + }, + { + "epoch": 5.8, + "learning_rate": 3.714059367601721e-06, + "loss": 0.0053, + "step": 44270 + }, + { + "epoch": 5.8, + "learning_rate": 3.7136469248359185e-06, + "loss": 0.0059, + "step": 44271 + }, + { + "epoch": 5.8, + "learning_rate": 3.713234499750147e-06, + "loss": 0.0069, + "step": 44272 + }, + { + "epoch": 5.8, + "learning_rate": 3.7128220923455703e-06, + "loss": 0.0076, + "step": 44273 + }, + { + "epoch": 5.8, + "learning_rate": 3.7124097026233473e-06, + "loss": 0.0078, + "step": 44274 + }, + { + "epoch": 5.8, + "learning_rate": 3.7119973305846345e-06, + "loss": 0.0057, + "step": 44275 + }, + { + "epoch": 5.8, + "learning_rate": 3.711584976230593e-06, + "loss": 0.0093, + "step": 44276 + }, + { + "epoch": 5.8, + "learning_rate": 3.7111726395623893e-06, + "loss": 0.0097, + "step": 44277 + }, + { + "epoch": 5.8, + "learning_rate": 3.710760320581176e-06, + "loss": 0.006, + "step": 44278 + }, + { + "epoch": 5.8, + "learning_rate": 3.7103480192881145e-06, + "loss": 0.0086, + "step": 44279 + }, + { + "epoch": 5.8, + "learning_rate": 3.709935735684361e-06, + "loss": 0.0067, + "step": 44280 + }, + { + "epoch": 5.8, + "learning_rate": 3.7095234697710826e-06, + "loss": 0.0049, + "step": 44281 + }, + { + "epoch": 5.8, + "learning_rate": 3.709111221549434e-06, + "loss": 0.0047, + "step": 44282 + }, + { + "epoch": 5.8, + "learning_rate": 3.7086989910205706e-06, + "loss": 0.0055, + "step": 44283 + }, + { + "epoch": 5.8, + "learning_rate": 3.7082867781856614e-06, + "loss": 0.0063, + "step": 44284 + }, + { + "epoch": 5.8, + "learning_rate": 3.7078745830458596e-06, + "loss": 0.0074, + "step": 44285 + }, + { + "epoch": 5.8, + "learning_rate": 3.7074624056023247e-06, + "loss": 0.0059, + "step": 44286 + }, + { + "epoch": 5.8, + "learning_rate": 3.707050245856211e-06, + "loss": 0.0082, + "step": 44287 + }, + { + "epoch": 5.8, + "learning_rate": 3.7066381038086895e-06, + "loss": 0.0047, + "step": 44288 + }, + { + "epoch": 5.8, + "learning_rate": 3.7062259794609135e-06, + "loss": 0.0062, + "step": 44289 + }, + { + "epoch": 5.8, + "learning_rate": 3.7058138728140424e-06, + "loss": 0.0078, + "step": 44290 + }, + { + "epoch": 5.8, + "learning_rate": 3.7054017838692302e-06, + "loss": 0.0063, + "step": 44291 + }, + { + "epoch": 5.8, + "learning_rate": 3.7049897126276447e-06, + "loss": 0.0056, + "step": 44292 + }, + { + "epoch": 5.8, + "learning_rate": 3.704577659090439e-06, + "loss": 0.0079, + "step": 44293 + }, + { + "epoch": 5.8, + "learning_rate": 3.7041656232587706e-06, + "loss": 0.0062, + "step": 44294 + }, + { + "epoch": 5.8, + "learning_rate": 3.703753605133805e-06, + "loss": 0.0072, + "step": 44295 + }, + { + "epoch": 5.8, + "learning_rate": 3.703341604716697e-06, + "loss": 0.0065, + "step": 44296 + }, + { + "epoch": 5.8, + "learning_rate": 3.7029296220086053e-06, + "loss": 0.0061, + "step": 44297 + }, + { + "epoch": 5.8, + "learning_rate": 3.7025176570106846e-06, + "loss": 0.0054, + "step": 44298 + }, + { + "epoch": 5.8, + "learning_rate": 3.7021057097240988e-06, + "loss": 0.0061, + "step": 44299 + }, + { + "epoch": 5.8, + "learning_rate": 3.7016937801500074e-06, + "loss": 0.0043, + "step": 44300 + }, + { + "epoch": 5.8, + "learning_rate": 3.701281868289568e-06, + "loss": 0.0046, + "step": 44301 + }, + { + "epoch": 5.8, + "learning_rate": 3.7008699741439348e-06, + "loss": 0.0059, + "step": 44302 + }, + { + "epoch": 5.8, + "learning_rate": 3.700458097714271e-06, + "loss": 0.0076, + "step": 44303 + }, + { + "epoch": 5.8, + "learning_rate": 3.7000462390017346e-06, + "loss": 0.005, + "step": 44304 + }, + { + "epoch": 5.8, + "learning_rate": 3.6996343980074824e-06, + "loss": 0.0063, + "step": 44305 + }, + { + "epoch": 5.8, + "learning_rate": 3.699222574732669e-06, + "loss": 0.0077, + "step": 44306 + }, + { + "epoch": 5.8, + "learning_rate": 3.6988107691784614e-06, + "loss": 0.0074, + "step": 44307 + }, + { + "epoch": 5.8, + "learning_rate": 3.6983989813460088e-06, + "loss": 0.0071, + "step": 44308 + }, + { + "epoch": 5.8, + "learning_rate": 3.6979872112364767e-06, + "loss": 0.0054, + "step": 44309 + }, + { + "epoch": 5.8, + "learning_rate": 3.697575458851016e-06, + "loss": 0.0069, + "step": 44310 + }, + { + "epoch": 5.8, + "learning_rate": 3.697163724190793e-06, + "loss": 0.0079, + "step": 44311 + }, + { + "epoch": 5.8, + "learning_rate": 3.696752007256962e-06, + "loss": 0.0065, + "step": 44312 + }, + { + "epoch": 5.8, + "learning_rate": 3.6963403080506786e-06, + "loss": 0.005, + "step": 44313 + }, + { + "epoch": 5.8, + "learning_rate": 3.6959286265731e-06, + "loss": 0.0053, + "step": 44314 + }, + { + "epoch": 5.8, + "learning_rate": 3.6955169628253895e-06, + "loss": 0.0084, + "step": 44315 + }, + { + "epoch": 5.81, + "learning_rate": 3.6951053168087024e-06, + "loss": 0.0066, + "step": 44316 + }, + { + "epoch": 5.81, + "learning_rate": 3.6946936885241914e-06, + "loss": 0.0078, + "step": 44317 + }, + { + "epoch": 5.81, + "learning_rate": 3.6942820779730228e-06, + "loss": 0.0065, + "step": 44318 + }, + { + "epoch": 5.81, + "learning_rate": 3.693870485156347e-06, + "loss": 0.0056, + "step": 44319 + }, + { + "epoch": 5.81, + "learning_rate": 3.693458910075327e-06, + "loss": 0.0052, + "step": 44320 + }, + { + "epoch": 5.81, + "learning_rate": 3.6930473527311185e-06, + "loss": 0.0041, + "step": 44321 + }, + { + "epoch": 5.81, + "learning_rate": 3.6926358131248753e-06, + "loss": 0.0063, + "step": 44322 + }, + { + "epoch": 5.81, + "learning_rate": 3.692224291257761e-06, + "loss": 0.0064, + "step": 44323 + }, + { + "epoch": 5.81, + "learning_rate": 3.6918127871309304e-06, + "loss": 0.0037, + "step": 44324 + }, + { + "epoch": 5.81, + "learning_rate": 3.6914013007455364e-06, + "loss": 0.006, + "step": 44325 + }, + { + "epoch": 5.81, + "learning_rate": 3.6909898321027437e-06, + "loss": 0.0049, + "step": 44326 + }, + { + "epoch": 5.81, + "learning_rate": 3.6905783812037065e-06, + "loss": 0.0061, + "step": 44327 + }, + { + "epoch": 5.81, + "learning_rate": 3.6901669480495817e-06, + "loss": 0.0057, + "step": 44328 + }, + { + "epoch": 5.81, + "learning_rate": 3.6897555326415236e-06, + "loss": 0.0074, + "step": 44329 + }, + { + "epoch": 5.81, + "learning_rate": 3.6893441349806903e-06, + "loss": 0.0044, + "step": 44330 + }, + { + "epoch": 5.81, + "learning_rate": 3.6889327550682453e-06, + "loss": 0.0045, + "step": 44331 + }, + { + "epoch": 5.81, + "learning_rate": 3.6885213929053397e-06, + "loss": 0.0079, + "step": 44332 + }, + { + "epoch": 5.81, + "learning_rate": 3.68811004849313e-06, + "loss": 0.0055, + "step": 44333 + }, + { + "epoch": 5.81, + "learning_rate": 3.6876987218327763e-06, + "loss": 0.0076, + "step": 44334 + }, + { + "epoch": 5.81, + "learning_rate": 3.6872874129254342e-06, + "loss": 0.005, + "step": 44335 + }, + { + "epoch": 5.81, + "learning_rate": 3.6868761217722606e-06, + "loss": 0.004, + "step": 44336 + }, + { + "epoch": 5.81, + "learning_rate": 3.686464848374408e-06, + "loss": 0.0059, + "step": 44337 + }, + { + "epoch": 5.81, + "learning_rate": 3.68605359273304e-06, + "loss": 0.005, + "step": 44338 + }, + { + "epoch": 5.81, + "learning_rate": 3.6856423548493102e-06, + "loss": 0.0052, + "step": 44339 + }, + { + "epoch": 5.81, + "learning_rate": 3.685231134724372e-06, + "loss": 0.0043, + "step": 44340 + }, + { + "epoch": 5.81, + "learning_rate": 3.684819932359385e-06, + "loss": 0.0051, + "step": 44341 + }, + { + "epoch": 5.81, + "learning_rate": 3.684408747755508e-06, + "loss": 0.0061, + "step": 44342 + }, + { + "epoch": 5.81, + "learning_rate": 3.683997580913896e-06, + "loss": 0.005, + "step": 44343 + }, + { + "epoch": 5.81, + "learning_rate": 3.6835864318357007e-06, + "loss": 0.0055, + "step": 44344 + }, + { + "epoch": 5.81, + "learning_rate": 3.683175300522086e-06, + "loss": 0.0069, + "step": 44345 + }, + { + "epoch": 5.81, + "learning_rate": 3.6827641869742035e-06, + "loss": 0.0049, + "step": 44346 + }, + { + "epoch": 5.81, + "learning_rate": 3.6823530911932115e-06, + "loss": 0.0041, + "step": 44347 + }, + { + "epoch": 5.81, + "learning_rate": 3.681942013180261e-06, + "loss": 0.0058, + "step": 44348 + }, + { + "epoch": 5.81, + "learning_rate": 3.681530952936515e-06, + "loss": 0.0058, + "step": 44349 + }, + { + "epoch": 5.81, + "learning_rate": 3.6811199104631278e-06, + "loss": 0.0058, + "step": 44350 + }, + { + "epoch": 5.81, + "learning_rate": 3.680708885761254e-06, + "loss": 0.0041, + "step": 44351 + }, + { + "epoch": 5.81, + "learning_rate": 3.680297878832042e-06, + "loss": 0.0043, + "step": 44352 + }, + { + "epoch": 5.81, + "learning_rate": 3.6798868896766636e-06, + "loss": 0.0077, + "step": 44353 + }, + { + "epoch": 5.81, + "learning_rate": 3.679475918296267e-06, + "loss": 0.0061, + "step": 44354 + }, + { + "epoch": 5.81, + "learning_rate": 3.6790649646920073e-06, + "loss": 0.0073, + "step": 44355 + }, + { + "epoch": 5.81, + "learning_rate": 3.6786540288650376e-06, + "loss": 0.0066, + "step": 44356 + }, + { + "epoch": 5.81, + "learning_rate": 3.6782431108165195e-06, + "loss": 0.0061, + "step": 44357 + }, + { + "epoch": 5.81, + "learning_rate": 3.6778322105476063e-06, + "loss": 0.0068, + "step": 44358 + }, + { + "epoch": 5.81, + "learning_rate": 3.6774213280594507e-06, + "loss": 0.005, + "step": 44359 + }, + { + "epoch": 5.81, + "learning_rate": 3.6770104633532132e-06, + "loss": 0.0058, + "step": 44360 + }, + { + "epoch": 5.81, + "learning_rate": 3.676599616430048e-06, + "loss": 0.0052, + "step": 44361 + }, + { + "epoch": 5.81, + "learning_rate": 3.6761887872911097e-06, + "loss": 0.0058, + "step": 44362 + }, + { + "epoch": 5.81, + "learning_rate": 3.675777975937549e-06, + "loss": 0.0058, + "step": 44363 + }, + { + "epoch": 5.81, + "learning_rate": 3.675367182370527e-06, + "loss": 0.0051, + "step": 44364 + }, + { + "epoch": 5.81, + "learning_rate": 3.6749564065912015e-06, + "loss": 0.0041, + "step": 44365 + }, + { + "epoch": 5.81, + "learning_rate": 3.6745456486007235e-06, + "loss": 0.013, + "step": 44366 + }, + { + "epoch": 5.81, + "learning_rate": 3.6741349084002465e-06, + "loss": 0.0052, + "step": 44367 + }, + { + "epoch": 5.81, + "learning_rate": 3.673724185990931e-06, + "loss": 0.0043, + "step": 44368 + }, + { + "epoch": 5.81, + "learning_rate": 3.673313481373929e-06, + "loss": 0.009, + "step": 44369 + }, + { + "epoch": 5.81, + "learning_rate": 3.672902794550397e-06, + "loss": 0.0058, + "step": 44370 + }, + { + "epoch": 5.81, + "learning_rate": 3.6724921255214853e-06, + "loss": 0.0063, + "step": 44371 + }, + { + "epoch": 5.81, + "learning_rate": 3.672081474288355e-06, + "loss": 0.0055, + "step": 44372 + }, + { + "epoch": 5.81, + "learning_rate": 3.6716708408521596e-06, + "loss": 0.0058, + "step": 44373 + }, + { + "epoch": 5.81, + "learning_rate": 3.6712602252140485e-06, + "loss": 0.0042, + "step": 44374 + }, + { + "epoch": 5.81, + "learning_rate": 3.6708496273751816e-06, + "loss": 0.0046, + "step": 44375 + }, + { + "epoch": 5.81, + "learning_rate": 3.6704390473367167e-06, + "loss": 0.0053, + "step": 44376 + }, + { + "epoch": 5.81, + "learning_rate": 3.6700284850998045e-06, + "loss": 0.0049, + "step": 44377 + }, + { + "epoch": 5.81, + "learning_rate": 3.669617940665601e-06, + "loss": 0.006, + "step": 44378 + }, + { + "epoch": 5.81, + "learning_rate": 3.669207414035255e-06, + "loss": 0.005, + "step": 44379 + }, + { + "epoch": 5.81, + "learning_rate": 3.6687969052099304e-06, + "loss": 0.004, + "step": 44380 + }, + { + "epoch": 5.81, + "learning_rate": 3.6683864141907776e-06, + "loss": 0.0076, + "step": 44381 + }, + { + "epoch": 5.81, + "learning_rate": 3.667975940978947e-06, + "loss": 0.005, + "step": 44382 + }, + { + "epoch": 5.81, + "learning_rate": 3.6675654855756003e-06, + "loss": 0.0049, + "step": 44383 + }, + { + "epoch": 5.81, + "learning_rate": 3.6671550479818885e-06, + "loss": 0.0052, + "step": 44384 + }, + { + "epoch": 5.81, + "learning_rate": 3.6667446281989626e-06, + "loss": 0.0049, + "step": 44385 + }, + { + "epoch": 5.81, + "learning_rate": 3.6663342262279834e-06, + "loss": 0.0075, + "step": 44386 + }, + { + "epoch": 5.81, + "learning_rate": 3.665923842070098e-06, + "loss": 0.0048, + "step": 44387 + }, + { + "epoch": 5.81, + "learning_rate": 3.6655134757264686e-06, + "loss": 0.007, + "step": 44388 + }, + { + "epoch": 5.81, + "learning_rate": 3.665103127198245e-06, + "loss": 0.0059, + "step": 44389 + }, + { + "epoch": 5.81, + "learning_rate": 3.6646927964865785e-06, + "loss": 0.0061, + "step": 44390 + }, + { + "epoch": 5.81, + "learning_rate": 3.664282483592628e-06, + "loss": 0.0051, + "step": 44391 + }, + { + "epoch": 5.82, + "learning_rate": 3.663872188517548e-06, + "loss": 0.0054, + "step": 44392 + }, + { + "epoch": 5.82, + "learning_rate": 3.663461911262488e-06, + "loss": 0.005, + "step": 44393 + }, + { + "epoch": 5.82, + "learning_rate": 3.6630516518286017e-06, + "loss": 0.0064, + "step": 44394 + }, + { + "epoch": 5.82, + "learning_rate": 3.6626414102170483e-06, + "loss": 0.0056, + "step": 44395 + }, + { + "epoch": 5.82, + "learning_rate": 3.662231186428975e-06, + "loss": 0.005, + "step": 44396 + }, + { + "epoch": 5.82, + "learning_rate": 3.6618209804655426e-06, + "loss": 0.004, + "step": 44397 + }, + { + "epoch": 5.82, + "learning_rate": 3.6614107923278974e-06, + "loss": 0.0047, + "step": 44398 + }, + { + "epoch": 5.82, + "learning_rate": 3.6610006220172e-06, + "loss": 0.0042, + "step": 44399 + }, + { + "epoch": 5.82, + "learning_rate": 3.660590469534601e-06, + "loss": 0.0048, + "step": 44400 + }, + { + "epoch": 5.82, + "learning_rate": 3.6601803348812537e-06, + "loss": 0.0043, + "step": 44401 + }, + { + "epoch": 5.82, + "learning_rate": 3.6597702180583083e-06, + "loss": 0.0051, + "step": 44402 + }, + { + "epoch": 5.82, + "learning_rate": 3.659360119066925e-06, + "loss": 0.0063, + "step": 44403 + }, + { + "epoch": 5.82, + "learning_rate": 3.6589500379082543e-06, + "loss": 0.0059, + "step": 44404 + }, + { + "epoch": 5.82, + "learning_rate": 3.6585399745834437e-06, + "loss": 0.0043, + "step": 44405 + }, + { + "epoch": 5.82, + "learning_rate": 3.6581299290936567e-06, + "loss": 0.0064, + "step": 44406 + }, + { + "epoch": 5.82, + "learning_rate": 3.6577199014400377e-06, + "loss": 0.0056, + "step": 44407 + }, + { + "epoch": 5.82, + "learning_rate": 3.657309891623747e-06, + "loss": 0.0062, + "step": 44408 + }, + { + "epoch": 5.82, + "learning_rate": 3.6568998996459315e-06, + "loss": 0.0074, + "step": 44409 + }, + { + "epoch": 5.82, + "learning_rate": 3.6564899255077503e-06, + "loss": 0.0063, + "step": 44410 + }, + { + "epoch": 5.82, + "learning_rate": 3.656079969210353e-06, + "loss": 0.0082, + "step": 44411 + }, + { + "epoch": 5.82, + "learning_rate": 3.6556700307548943e-06, + "loss": 0.0055, + "step": 44412 + }, + { + "epoch": 5.82, + "learning_rate": 3.6552601101425212e-06, + "loss": 0.0068, + "step": 44413 + }, + { + "epoch": 5.82, + "learning_rate": 3.654850207374395e-06, + "loss": 0.0068, + "step": 44414 + }, + { + "epoch": 5.82, + "learning_rate": 3.6544403224516643e-06, + "loss": 0.0067, + "step": 44415 + }, + { + "epoch": 5.82, + "learning_rate": 3.654030455375482e-06, + "loss": 0.0068, + "step": 44416 + }, + { + "epoch": 5.82, + "learning_rate": 3.6536206061469992e-06, + "loss": 0.006, + "step": 44417 + }, + { + "epoch": 5.82, + "learning_rate": 3.6532107747673694e-06, + "loss": 0.0057, + "step": 44418 + }, + { + "epoch": 5.82, + "learning_rate": 3.652800961237749e-06, + "loss": 0.0067, + "step": 44419 + }, + { + "epoch": 5.82, + "learning_rate": 3.652391165559289e-06, + "loss": 0.0049, + "step": 44420 + }, + { + "epoch": 5.82, + "learning_rate": 3.6519813877331365e-06, + "loss": 0.0059, + "step": 44421 + }, + { + "epoch": 5.82, + "learning_rate": 3.651571627760453e-06, + "loss": 0.004, + "step": 44422 + }, + { + "epoch": 5.82, + "learning_rate": 3.651161885642386e-06, + "loss": 0.0051, + "step": 44423 + }, + { + "epoch": 5.82, + "learning_rate": 3.6507521613800834e-06, + "loss": 0.0065, + "step": 44424 + }, + { + "epoch": 5.82, + "learning_rate": 3.6503424549747067e-06, + "loss": 0.0048, + "step": 44425 + }, + { + "epoch": 5.82, + "learning_rate": 3.649932766427403e-06, + "loss": 0.0066, + "step": 44426 + }, + { + "epoch": 5.82, + "learning_rate": 3.649523095739326e-06, + "loss": 0.0045, + "step": 44427 + }, + { + "epoch": 5.82, + "learning_rate": 3.6491134429116236e-06, + "loss": 0.0079, + "step": 44428 + }, + { + "epoch": 5.82, + "learning_rate": 3.648703807945451e-06, + "loss": 0.006, + "step": 44429 + }, + { + "epoch": 5.82, + "learning_rate": 3.6482941908419645e-06, + "loss": 0.0073, + "step": 44430 + }, + { + "epoch": 5.82, + "learning_rate": 3.647884591602312e-06, + "loss": 0.0078, + "step": 44431 + }, + { + "epoch": 5.82, + "learning_rate": 3.6474750102276422e-06, + "loss": 0.007, + "step": 44432 + }, + { + "epoch": 5.82, + "learning_rate": 3.6470654467191145e-06, + "loss": 0.0047, + "step": 44433 + }, + { + "epoch": 5.82, + "learning_rate": 3.646655901077878e-06, + "loss": 0.0058, + "step": 44434 + }, + { + "epoch": 5.82, + "learning_rate": 3.6462463733050813e-06, + "loss": 0.0068, + "step": 44435 + }, + { + "epoch": 5.82, + "learning_rate": 3.645836863401876e-06, + "loss": 0.0052, + "step": 44436 + }, + { + "epoch": 5.82, + "learning_rate": 3.64542737136942e-06, + "loss": 0.0031, + "step": 44437 + }, + { + "epoch": 5.82, + "learning_rate": 3.64501789720886e-06, + "loss": 0.0047, + "step": 44438 + }, + { + "epoch": 5.82, + "learning_rate": 3.644608440921347e-06, + "loss": 0.0052, + "step": 44439 + }, + { + "epoch": 5.82, + "learning_rate": 3.644199002508032e-06, + "loss": 0.0053, + "step": 44440 + }, + { + "epoch": 5.82, + "learning_rate": 3.643789581970074e-06, + "loss": 0.0047, + "step": 44441 + }, + { + "epoch": 5.82, + "learning_rate": 3.643380179308619e-06, + "loss": 0.006, + "step": 44442 + }, + { + "epoch": 5.82, + "learning_rate": 3.6429707945248185e-06, + "loss": 0.0055, + "step": 44443 + }, + { + "epoch": 5.82, + "learning_rate": 3.6425614276198206e-06, + "loss": 0.0043, + "step": 44444 + }, + { + "epoch": 5.82, + "learning_rate": 3.642152078594784e-06, + "loss": 0.0062, + "step": 44445 + }, + { + "epoch": 5.82, + "learning_rate": 3.641742747450855e-06, + "loss": 0.0065, + "step": 44446 + }, + { + "epoch": 5.82, + "learning_rate": 3.6413334341891837e-06, + "loss": 0.0061, + "step": 44447 + }, + { + "epoch": 5.82, + "learning_rate": 3.6409241388109263e-06, + "loss": 0.0064, + "step": 44448 + }, + { + "epoch": 5.82, + "learning_rate": 3.6405148613172313e-06, + "loss": 0.0091, + "step": 44449 + }, + { + "epoch": 5.82, + "learning_rate": 3.640105601709246e-06, + "loss": 0.0044, + "step": 44450 + }, + { + "epoch": 5.82, + "learning_rate": 3.6396963599881284e-06, + "loss": 0.0062, + "step": 44451 + }, + { + "epoch": 5.82, + "learning_rate": 3.639287136155023e-06, + "loss": 0.0067, + "step": 44452 + }, + { + "epoch": 5.82, + "learning_rate": 3.6388779302110876e-06, + "loss": 0.0047, + "step": 44453 + }, + { + "epoch": 5.82, + "learning_rate": 3.6384687421574684e-06, + "loss": 0.0074, + "step": 44454 + }, + { + "epoch": 5.82, + "learning_rate": 3.6380595719953136e-06, + "loss": 0.0068, + "step": 44455 + }, + { + "epoch": 5.82, + "learning_rate": 3.6376504197257825e-06, + "loss": 0.0055, + "step": 44456 + }, + { + "epoch": 5.82, + "learning_rate": 3.6372412853500192e-06, + "loss": 0.0064, + "step": 44457 + }, + { + "epoch": 5.82, + "learning_rate": 3.6368321688691764e-06, + "loss": 0.0047, + "step": 44458 + }, + { + "epoch": 5.82, + "learning_rate": 3.6364230702844007e-06, + "loss": 0.0061, + "step": 44459 + }, + { + "epoch": 5.82, + "learning_rate": 3.6360139895968493e-06, + "loss": 0.0048, + "step": 44460 + }, + { + "epoch": 5.82, + "learning_rate": 3.6356049268076678e-06, + "loss": 0.0071, + "step": 44461 + }, + { + "epoch": 5.82, + "learning_rate": 3.63519588191801e-06, + "loss": 0.0067, + "step": 44462 + }, + { + "epoch": 5.82, + "learning_rate": 3.6347868549290222e-06, + "loss": 0.0044, + "step": 44463 + }, + { + "epoch": 5.82, + "learning_rate": 3.634377845841861e-06, + "loss": 0.007, + "step": 44464 + }, + { + "epoch": 5.82, + "learning_rate": 3.6339688546576734e-06, + "loss": 0.0068, + "step": 44465 + }, + { + "epoch": 5.82, + "learning_rate": 3.6335598813776084e-06, + "loss": 0.0066, + "step": 44466 + }, + { + "epoch": 5.82, + "learning_rate": 3.633150926002814e-06, + "loss": 0.0061, + "step": 44467 + }, + { + "epoch": 5.82, + "learning_rate": 3.6327419885344484e-06, + "loss": 0.0051, + "step": 44468 + }, + { + "epoch": 5.83, + "learning_rate": 3.632333068973656e-06, + "loss": 0.0048, + "step": 44469 + }, + { + "epoch": 5.83, + "learning_rate": 3.6319241673215833e-06, + "loss": 0.0052, + "step": 44470 + }, + { + "epoch": 5.83, + "learning_rate": 3.6315152835793887e-06, + "loss": 0.0055, + "step": 44471 + }, + { + "epoch": 5.83, + "learning_rate": 3.6311064177482157e-06, + "loss": 0.0053, + "step": 44472 + }, + { + "epoch": 5.83, + "learning_rate": 3.630697569829219e-06, + "loss": 0.0061, + "step": 44473 + }, + { + "epoch": 5.83, + "learning_rate": 3.6302887398235444e-06, + "loss": 0.0071, + "step": 44474 + }, + { + "epoch": 5.83, + "learning_rate": 3.629879927732345e-06, + "loss": 0.0064, + "step": 44475 + }, + { + "epoch": 5.83, + "learning_rate": 3.62947113355677e-06, + "loss": 0.005, + "step": 44476 + }, + { + "epoch": 5.83, + "learning_rate": 3.6290623572979687e-06, + "loss": 0.0066, + "step": 44477 + }, + { + "epoch": 5.83, + "learning_rate": 3.628653598957086e-06, + "loss": 0.0071, + "step": 44478 + }, + { + "epoch": 5.83, + "learning_rate": 3.6282448585352792e-06, + "loss": 0.0082, + "step": 44479 + }, + { + "epoch": 5.83, + "learning_rate": 3.6278361360336946e-06, + "loss": 0.0057, + "step": 44480 + }, + { + "epoch": 5.83, + "learning_rate": 3.627427431453482e-06, + "loss": 0.0069, + "step": 44481 + }, + { + "epoch": 5.83, + "learning_rate": 3.627018744795786e-06, + "loss": 0.0066, + "step": 44482 + }, + { + "epoch": 5.83, + "learning_rate": 3.6266100760617617e-06, + "loss": 0.0056, + "step": 44483 + }, + { + "epoch": 5.83, + "learning_rate": 3.6262014252525603e-06, + "loss": 0.0052, + "step": 44484 + }, + { + "epoch": 5.83, + "learning_rate": 3.625792792369327e-06, + "loss": 0.0043, + "step": 44485 + }, + { + "epoch": 5.83, + "learning_rate": 3.625384177413209e-06, + "loss": 0.0046, + "step": 44486 + }, + { + "epoch": 5.83, + "learning_rate": 3.6249755803853627e-06, + "loss": 0.0051, + "step": 44487 + }, + { + "epoch": 5.83, + "learning_rate": 3.6245670012869316e-06, + "loss": 0.0063, + "step": 44488 + }, + { + "epoch": 5.83, + "learning_rate": 3.6241584401190644e-06, + "loss": 0.0052, + "step": 44489 + }, + { + "epoch": 5.83, + "learning_rate": 3.6237498968829144e-06, + "loss": 0.0065, + "step": 44490 + }, + { + "epoch": 5.83, + "learning_rate": 3.6233413715796284e-06, + "loss": 0.0061, + "step": 44491 + }, + { + "epoch": 5.83, + "learning_rate": 3.6229328642103554e-06, + "loss": 0.0074, + "step": 44492 + }, + { + "epoch": 5.83, + "learning_rate": 3.622524374776242e-06, + "loss": 0.0059, + "step": 44493 + }, + { + "epoch": 5.83, + "learning_rate": 3.622115903278437e-06, + "loss": 0.0077, + "step": 44494 + }, + { + "epoch": 5.83, + "learning_rate": 3.621707449718096e-06, + "loss": 0.0076, + "step": 44495 + }, + { + "epoch": 5.83, + "learning_rate": 3.621299014096363e-06, + "loss": 0.0053, + "step": 44496 + }, + { + "epoch": 5.83, + "learning_rate": 3.620890596414384e-06, + "loss": 0.0039, + "step": 44497 + }, + { + "epoch": 5.83, + "learning_rate": 3.620482196673313e-06, + "loss": 0.0068, + "step": 44498 + }, + { + "epoch": 5.83, + "learning_rate": 3.6200738148742955e-06, + "loss": 0.0056, + "step": 44499 + }, + { + "epoch": 5.83, + "learning_rate": 3.6196654510184814e-06, + "loss": 0.007, + "step": 44500 + }, + { + "epoch": 5.83, + "learning_rate": 3.6192571051070158e-06, + "loss": 0.005, + "step": 44501 + }, + { + "epoch": 5.83, + "learning_rate": 3.618848777141052e-06, + "loss": 0.0069, + "step": 44502 + }, + { + "epoch": 5.83, + "learning_rate": 3.6184404671217367e-06, + "loss": 0.0044, + "step": 44503 + }, + { + "epoch": 5.83, + "learning_rate": 3.6180321750502145e-06, + "loss": 0.0053, + "step": 44504 + }, + { + "epoch": 5.83, + "learning_rate": 3.6176239009276372e-06, + "loss": 0.0063, + "step": 44505 + }, + { + "epoch": 5.83, + "learning_rate": 3.6172156447551556e-06, + "loss": 0.0061, + "step": 44506 + }, + { + "epoch": 5.83, + "learning_rate": 3.6168074065339156e-06, + "loss": 0.007, + "step": 44507 + }, + { + "epoch": 5.83, + "learning_rate": 3.6163991862650637e-06, + "loss": 0.0047, + "step": 44508 + }, + { + "epoch": 5.83, + "learning_rate": 3.6159909839497465e-06, + "loss": 0.0067, + "step": 44509 + }, + { + "epoch": 5.83, + "learning_rate": 3.615582799589119e-06, + "loss": 0.0054, + "step": 44510 + }, + { + "epoch": 5.83, + "learning_rate": 3.615174633184324e-06, + "loss": 0.0057, + "step": 44511 + }, + { + "epoch": 5.83, + "learning_rate": 3.6147664847365062e-06, + "loss": 0.0062, + "step": 44512 + }, + { + "epoch": 5.83, + "learning_rate": 3.614358354246822e-06, + "loss": 0.0083, + "step": 44513 + }, + { + "epoch": 5.83, + "learning_rate": 3.6139502417164152e-06, + "loss": 0.0068, + "step": 44514 + }, + { + "epoch": 5.83, + "learning_rate": 3.613542147146433e-06, + "loss": 0.0061, + "step": 44515 + }, + { + "epoch": 5.83, + "learning_rate": 3.613134070538019e-06, + "loss": 0.0054, + "step": 44516 + }, + { + "epoch": 5.83, + "learning_rate": 3.6127260118923257e-06, + "loss": 0.0083, + "step": 44517 + }, + { + "epoch": 5.83, + "learning_rate": 3.612317971210505e-06, + "loss": 0.0052, + "step": 44518 + }, + { + "epoch": 5.83, + "learning_rate": 3.611909948493698e-06, + "loss": 0.0053, + "step": 44519 + }, + { + "epoch": 5.83, + "learning_rate": 3.611501943743052e-06, + "loss": 0.0041, + "step": 44520 + }, + { + "epoch": 5.83, + "learning_rate": 3.6110939569597202e-06, + "loss": 0.0064, + "step": 44521 + }, + { + "epoch": 5.83, + "learning_rate": 3.610685988144845e-06, + "loss": 0.008, + "step": 44522 + }, + { + "epoch": 5.83, + "learning_rate": 3.610278037299576e-06, + "loss": 0.0069, + "step": 44523 + }, + { + "epoch": 5.83, + "learning_rate": 3.609870104425057e-06, + "loss": 0.0048, + "step": 44524 + }, + { + "epoch": 5.83, + "learning_rate": 3.6094621895224413e-06, + "loss": 0.0063, + "step": 44525 + }, + { + "epoch": 5.83, + "learning_rate": 3.609054292592872e-06, + "loss": 0.0049, + "step": 44526 + }, + { + "epoch": 5.83, + "learning_rate": 3.6086464136374945e-06, + "loss": 0.0068, + "step": 44527 + }, + { + "epoch": 5.83, + "learning_rate": 3.608238552657458e-06, + "loss": 0.0055, + "step": 44528 + }, + { + "epoch": 5.83, + "learning_rate": 3.607830709653913e-06, + "loss": 0.0044, + "step": 44529 + }, + { + "epoch": 5.83, + "learning_rate": 3.607422884628006e-06, + "loss": 0.0064, + "step": 44530 + }, + { + "epoch": 5.83, + "learning_rate": 3.60701507758088e-06, + "loss": 0.0068, + "step": 44531 + }, + { + "epoch": 5.83, + "learning_rate": 3.6066072885136807e-06, + "loss": 0.0041, + "step": 44532 + }, + { + "epoch": 5.83, + "learning_rate": 3.606199517427561e-06, + "loss": 0.0087, + "step": 44533 + }, + { + "epoch": 5.83, + "learning_rate": 3.605791764323664e-06, + "loss": 0.0061, + "step": 44534 + }, + { + "epoch": 5.83, + "learning_rate": 3.605384029203135e-06, + "loss": 0.0046, + "step": 44535 + }, + { + "epoch": 5.83, + "learning_rate": 3.6049763120671255e-06, + "loss": 0.0075, + "step": 44536 + }, + { + "epoch": 5.83, + "learning_rate": 3.60456861291678e-06, + "loss": 0.0067, + "step": 44537 + }, + { + "epoch": 5.83, + "learning_rate": 3.6041609317532423e-06, + "loss": 0.0076, + "step": 44538 + }, + { + "epoch": 5.83, + "learning_rate": 3.60375326857766e-06, + "loss": 0.0088, + "step": 44539 + }, + { + "epoch": 5.83, + "learning_rate": 3.6033456233911847e-06, + "loss": 0.0052, + "step": 44540 + }, + { + "epoch": 5.83, + "learning_rate": 3.6029379961949595e-06, + "loss": 0.0056, + "step": 44541 + }, + { + "epoch": 5.83, + "learning_rate": 3.6025303869901296e-06, + "loss": 0.0067, + "step": 44542 + }, + { + "epoch": 5.83, + "learning_rate": 3.6021227957778392e-06, + "loss": 0.0053, + "step": 44543 + }, + { + "epoch": 5.83, + "learning_rate": 3.601715222559242e-06, + "loss": 0.0054, + "step": 44544 + }, + { + "epoch": 5.84, + "learning_rate": 3.60130766733548e-06, + "loss": 0.0063, + "step": 44545 + }, + { + "epoch": 5.84, + "learning_rate": 3.6009001301076985e-06, + "loss": 0.0061, + "step": 44546 + }, + { + "epoch": 5.84, + "learning_rate": 3.600492610877041e-06, + "loss": 0.006, + "step": 44547 + }, + { + "epoch": 5.84, + "learning_rate": 3.600085109644662e-06, + "loss": 0.0063, + "step": 44548 + }, + { + "epoch": 5.84, + "learning_rate": 3.5996776264116982e-06, + "loss": 0.0085, + "step": 44549 + }, + { + "epoch": 5.84, + "learning_rate": 3.5992701611793047e-06, + "loss": 0.0066, + "step": 44550 + }, + { + "epoch": 5.84, + "learning_rate": 3.5988627139486186e-06, + "loss": 0.0035, + "step": 44551 + }, + { + "epoch": 5.84, + "learning_rate": 3.5984552847207945e-06, + "loss": 0.0061, + "step": 44552 + }, + { + "epoch": 5.84, + "learning_rate": 3.5980478734969737e-06, + "loss": 0.0068, + "step": 44553 + }, + { + "epoch": 5.84, + "learning_rate": 3.597640480278298e-06, + "loss": 0.0072, + "step": 44554 + }, + { + "epoch": 5.84, + "learning_rate": 3.5972331050659226e-06, + "loss": 0.006, + "step": 44555 + }, + { + "epoch": 5.84, + "learning_rate": 3.596825747860987e-06, + "loss": 0.0056, + "step": 44556 + }, + { + "epoch": 5.84, + "learning_rate": 3.5964184086646382e-06, + "loss": 0.0071, + "step": 44557 + }, + { + "epoch": 5.84, + "learning_rate": 3.5960110874780184e-06, + "loss": 0.0064, + "step": 44558 + }, + { + "epoch": 5.84, + "learning_rate": 3.5956037843022795e-06, + "loss": 0.0048, + "step": 44559 + }, + { + "epoch": 5.84, + "learning_rate": 3.59519649913856e-06, + "loss": 0.0066, + "step": 44560 + }, + { + "epoch": 5.84, + "learning_rate": 3.594789231988014e-06, + "loss": 0.0054, + "step": 44561 + }, + { + "epoch": 5.84, + "learning_rate": 3.594381982851778e-06, + "loss": 0.0048, + "step": 44562 + }, + { + "epoch": 5.84, + "learning_rate": 3.5939747517310053e-06, + "loss": 0.0075, + "step": 44563 + }, + { + "epoch": 5.84, + "learning_rate": 3.593567538626838e-06, + "loss": 0.0059, + "step": 44564 + }, + { + "epoch": 5.84, + "learning_rate": 3.5931603435404193e-06, + "loss": 0.0062, + "step": 44565 + }, + { + "epoch": 5.84, + "learning_rate": 3.5927531664728933e-06, + "loss": 0.0074, + "step": 44566 + }, + { + "epoch": 5.84, + "learning_rate": 3.5923460074254113e-06, + "loss": 0.0086, + "step": 44567 + }, + { + "epoch": 5.84, + "learning_rate": 3.591938866399115e-06, + "loss": 0.0069, + "step": 44568 + }, + { + "epoch": 5.84, + "learning_rate": 3.591531743395146e-06, + "loss": 0.0058, + "step": 44569 + }, + { + "epoch": 5.84, + "learning_rate": 3.5911246384146557e-06, + "loss": 0.0053, + "step": 44570 + }, + { + "epoch": 5.84, + "learning_rate": 3.5907175514587823e-06, + "loss": 0.0078, + "step": 44571 + }, + { + "epoch": 5.84, + "learning_rate": 3.5903104825286784e-06, + "loss": 0.0048, + "step": 44572 + }, + { + "epoch": 5.84, + "learning_rate": 3.5899034316254855e-06, + "loss": 0.0042, + "step": 44573 + }, + { + "epoch": 5.84, + "learning_rate": 3.589496398750344e-06, + "loss": 0.0077, + "step": 44574 + }, + { + "epoch": 5.84, + "learning_rate": 3.589089383904406e-06, + "loss": 0.0047, + "step": 44575 + }, + { + "epoch": 5.84, + "learning_rate": 3.588682387088812e-06, + "loss": 0.0069, + "step": 44576 + }, + { + "epoch": 5.84, + "learning_rate": 3.5882754083047046e-06, + "loss": 0.0044, + "step": 44577 + }, + { + "epoch": 5.84, + "learning_rate": 3.5878684475532345e-06, + "loss": 0.0045, + "step": 44578 + }, + { + "epoch": 5.84, + "learning_rate": 3.587461504835543e-06, + "loss": 0.0076, + "step": 44579 + }, + { + "epoch": 5.84, + "learning_rate": 3.5870545801527746e-06, + "loss": 0.0077, + "step": 44580 + }, + { + "epoch": 5.84, + "learning_rate": 3.5866476735060706e-06, + "loss": 0.008, + "step": 44581 + }, + { + "epoch": 5.84, + "learning_rate": 3.5862407848965774e-06, + "loss": 0.0061, + "step": 44582 + }, + { + "epoch": 5.84, + "learning_rate": 3.585833914325445e-06, + "loss": 0.0054, + "step": 44583 + }, + { + "epoch": 5.84, + "learning_rate": 3.5854270617938124e-06, + "loss": 0.0046, + "step": 44584 + }, + { + "epoch": 5.84, + "learning_rate": 3.5850202273028212e-06, + "loss": 0.0073, + "step": 44585 + }, + { + "epoch": 5.84, + "learning_rate": 3.5846134108536225e-06, + "loss": 0.0054, + "step": 44586 + }, + { + "epoch": 5.84, + "learning_rate": 3.5842066124473573e-06, + "loss": 0.0079, + "step": 44587 + }, + { + "epoch": 5.84, + "learning_rate": 3.5837998320851684e-06, + "loss": 0.0052, + "step": 44588 + }, + { + "epoch": 5.84, + "learning_rate": 3.5833930697681983e-06, + "loss": 0.006, + "step": 44589 + }, + { + "epoch": 5.84, + "learning_rate": 3.5829863254975972e-06, + "loss": 0.0055, + "step": 44590 + }, + { + "epoch": 5.84, + "learning_rate": 3.5825795992745048e-06, + "loss": 0.0043, + "step": 44591 + }, + { + "epoch": 5.84, + "learning_rate": 3.582172891100062e-06, + "loss": 0.0062, + "step": 44592 + }, + { + "epoch": 5.84, + "learning_rate": 3.581766200975416e-06, + "loss": 0.0054, + "step": 44593 + }, + { + "epoch": 5.84, + "learning_rate": 3.5813595289017143e-06, + "loss": 0.0064, + "step": 44594 + }, + { + "epoch": 5.84, + "learning_rate": 3.580952874880097e-06, + "loss": 0.0046, + "step": 44595 + }, + { + "epoch": 5.84, + "learning_rate": 3.5805462389117087e-06, + "loss": 0.0057, + "step": 44596 + }, + { + "epoch": 5.84, + "learning_rate": 3.580139620997688e-06, + "loss": 0.0068, + "step": 44597 + }, + { + "epoch": 5.84, + "learning_rate": 3.5797330211391866e-06, + "loss": 0.005, + "step": 44598 + }, + { + "epoch": 5.84, + "learning_rate": 3.579326439337344e-06, + "loss": 0.0059, + "step": 44599 + }, + { + "epoch": 5.84, + "learning_rate": 3.5789198755933008e-06, + "loss": 0.0063, + "step": 44600 + }, + { + "epoch": 5.84, + "learning_rate": 3.5785133299082066e-06, + "loss": 0.0061, + "step": 44601 + }, + { + "epoch": 5.84, + "learning_rate": 3.5781068022832012e-06, + "loss": 0.0077, + "step": 44602 + }, + { + "epoch": 5.84, + "learning_rate": 3.577700292719426e-06, + "loss": 0.0065, + "step": 44603 + }, + { + "epoch": 5.84, + "learning_rate": 3.5772938012180257e-06, + "loss": 0.0052, + "step": 44604 + }, + { + "epoch": 5.84, + "learning_rate": 3.5768873277801486e-06, + "loss": 0.0043, + "step": 44605 + }, + { + "epoch": 5.84, + "learning_rate": 3.5764808724069346e-06, + "loss": 0.0064, + "step": 44606 + }, + { + "epoch": 5.84, + "learning_rate": 3.5760744350995247e-06, + "loss": 0.0056, + "step": 44607 + }, + { + "epoch": 5.84, + "learning_rate": 3.575668015859061e-06, + "loss": 0.0045, + "step": 44608 + }, + { + "epoch": 5.84, + "learning_rate": 3.5752616146866915e-06, + "loss": 0.0061, + "step": 44609 + }, + { + "epoch": 5.84, + "learning_rate": 3.574855231583557e-06, + "loss": 0.0069, + "step": 44610 + }, + { + "epoch": 5.84, + "learning_rate": 3.574448866550796e-06, + "loss": 0.0076, + "step": 44611 + }, + { + "epoch": 5.84, + "learning_rate": 3.574042519589559e-06, + "loss": 0.0075, + "step": 44612 + }, + { + "epoch": 5.84, + "learning_rate": 3.573636190700985e-06, + "loss": 0.004, + "step": 44613 + }, + { + "epoch": 5.84, + "learning_rate": 3.5732298798862154e-06, + "loss": 0.005, + "step": 44614 + }, + { + "epoch": 5.84, + "learning_rate": 3.5728235871463957e-06, + "loss": 0.0065, + "step": 44615 + }, + { + "epoch": 5.84, + "learning_rate": 3.5724173124826656e-06, + "loss": 0.0061, + "step": 44616 + }, + { + "epoch": 5.84, + "learning_rate": 3.5720110558961717e-06, + "loss": 0.0058, + "step": 44617 + }, + { + "epoch": 5.84, + "learning_rate": 3.571604817388055e-06, + "loss": 0.0053, + "step": 44618 + }, + { + "epoch": 5.84, + "learning_rate": 3.571198596959453e-06, + "loss": 0.0064, + "step": 44619 + }, + { + "epoch": 5.84, + "learning_rate": 3.5707923946115175e-06, + "loss": 0.0056, + "step": 44620 + }, + { + "epoch": 5.85, + "learning_rate": 3.5703862103453856e-06, + "loss": 0.0063, + "step": 44621 + }, + { + "epoch": 5.85, + "learning_rate": 3.5699800441622e-06, + "loss": 0.0058, + "step": 44622 + }, + { + "epoch": 5.85, + "learning_rate": 3.5695738960630987e-06, + "loss": 0.0049, + "step": 44623 + }, + { + "epoch": 5.85, + "learning_rate": 3.569167766049233e-06, + "loss": 0.007, + "step": 44624 + }, + { + "epoch": 5.85, + "learning_rate": 3.5687616541217373e-06, + "loss": 0.0085, + "step": 44625 + }, + { + "epoch": 5.85, + "learning_rate": 3.5683555602817597e-06, + "loss": 0.0051, + "step": 44626 + }, + { + "epoch": 5.85, + "learning_rate": 3.5679494845304363e-06, + "loss": 0.0075, + "step": 44627 + }, + { + "epoch": 5.85, + "learning_rate": 3.567543426868916e-06, + "loss": 0.0057, + "step": 44628 + }, + { + "epoch": 5.85, + "learning_rate": 3.5671373872983363e-06, + "loss": 0.0067, + "step": 44629 + }, + { + "epoch": 5.85, + "learning_rate": 3.566731365819841e-06, + "loss": 0.0059, + "step": 44630 + }, + { + "epoch": 5.85, + "learning_rate": 3.5663253624345685e-06, + "loss": 0.0066, + "step": 44631 + }, + { + "epoch": 5.85, + "learning_rate": 3.565919377143665e-06, + "loss": 0.0047, + "step": 44632 + }, + { + "epoch": 5.85, + "learning_rate": 3.5655134099482714e-06, + "loss": 0.0062, + "step": 44633 + }, + { + "epoch": 5.85, + "learning_rate": 3.565107460849526e-06, + "loss": 0.0066, + "step": 44634 + }, + { + "epoch": 5.85, + "learning_rate": 3.5647015298485753e-06, + "loss": 0.0064, + "step": 44635 + }, + { + "epoch": 5.85, + "learning_rate": 3.564295616946557e-06, + "loss": 0.0054, + "step": 44636 + }, + { + "epoch": 5.85, + "learning_rate": 3.5638897221446166e-06, + "loss": 0.0067, + "step": 44637 + }, + { + "epoch": 5.85, + "learning_rate": 3.5634838454438936e-06, + "loss": 0.0073, + "step": 44638 + }, + { + "epoch": 5.85, + "learning_rate": 3.5630779868455267e-06, + "loss": 0.0038, + "step": 44639 + }, + { + "epoch": 5.85, + "learning_rate": 3.562672146350664e-06, + "loss": 0.006, + "step": 44640 + }, + { + "epoch": 5.85, + "learning_rate": 3.5622663239604427e-06, + "loss": 0.0056, + "step": 44641 + }, + { + "epoch": 5.85, + "learning_rate": 3.5618605196760002e-06, + "loss": 0.0041, + "step": 44642 + }, + { + "epoch": 5.85, + "learning_rate": 3.561454733498487e-06, + "loss": 0.0069, + "step": 44643 + }, + { + "epoch": 5.85, + "learning_rate": 3.561048965429039e-06, + "loss": 0.0065, + "step": 44644 + }, + { + "epoch": 5.85, + "learning_rate": 3.5606432154687976e-06, + "loss": 0.0053, + "step": 44645 + }, + { + "epoch": 5.85, + "learning_rate": 3.5602374836189002e-06, + "loss": 0.0051, + "step": 44646 + }, + { + "epoch": 5.85, + "learning_rate": 3.559831769880493e-06, + "loss": 0.0057, + "step": 44647 + }, + { + "epoch": 5.85, + "learning_rate": 3.5594260742547194e-06, + "loss": 0.0065, + "step": 44648 + }, + { + "epoch": 5.85, + "learning_rate": 3.559020396742716e-06, + "loss": 0.0059, + "step": 44649 + }, + { + "epoch": 5.85, + "learning_rate": 3.5586147373456226e-06, + "loss": 0.006, + "step": 44650 + }, + { + "epoch": 5.85, + "learning_rate": 3.5582090960645855e-06, + "loss": 0.0064, + "step": 44651 + }, + { + "epoch": 5.85, + "learning_rate": 3.5578034729007416e-06, + "loss": 0.0061, + "step": 44652 + }, + { + "epoch": 5.85, + "learning_rate": 3.5573978678552322e-06, + "loss": 0.0068, + "step": 44653 + }, + { + "epoch": 5.85, + "learning_rate": 3.5569922809291955e-06, + "loss": 0.0059, + "step": 44654 + }, + { + "epoch": 5.85, + "learning_rate": 3.5565867121237774e-06, + "loss": 0.0057, + "step": 44655 + }, + { + "epoch": 5.85, + "learning_rate": 3.5561811614401163e-06, + "loss": 0.0045, + "step": 44656 + }, + { + "epoch": 5.85, + "learning_rate": 3.55577562887935e-06, + "loss": 0.0046, + "step": 44657 + }, + { + "epoch": 5.85, + "learning_rate": 3.555370114442621e-06, + "loss": 0.0056, + "step": 44658 + }, + { + "epoch": 5.85, + "learning_rate": 3.5549646181310737e-06, + "loss": 0.0064, + "step": 44659 + }, + { + "epoch": 5.85, + "learning_rate": 3.5545591399458456e-06, + "loss": 0.0053, + "step": 44660 + }, + { + "epoch": 5.85, + "learning_rate": 3.5541536798880725e-06, + "loss": 0.0049, + "step": 44661 + }, + { + "epoch": 5.85, + "learning_rate": 3.5537482379589026e-06, + "loss": 0.0058, + "step": 44662 + }, + { + "epoch": 5.85, + "learning_rate": 3.5533428141594726e-06, + "loss": 0.0061, + "step": 44663 + }, + { + "epoch": 5.85, + "learning_rate": 3.552937408490923e-06, + "loss": 0.0075, + "step": 44664 + }, + { + "epoch": 5.85, + "learning_rate": 3.55253202095439e-06, + "loss": 0.007, + "step": 44665 + }, + { + "epoch": 5.85, + "learning_rate": 3.55212665155102e-06, + "loss": 0.0088, + "step": 44666 + }, + { + "epoch": 5.85, + "learning_rate": 3.551721300281952e-06, + "loss": 0.0048, + "step": 44667 + }, + { + "epoch": 5.85, + "learning_rate": 3.5513159671483233e-06, + "loss": 0.0056, + "step": 44668 + }, + { + "epoch": 5.85, + "learning_rate": 3.550910652151268e-06, + "loss": 0.0051, + "step": 44669 + }, + { + "epoch": 5.85, + "learning_rate": 3.55050535529194e-06, + "loss": 0.0058, + "step": 44670 + }, + { + "epoch": 5.85, + "learning_rate": 3.550100076571473e-06, + "loss": 0.0067, + "step": 44671 + }, + { + "epoch": 5.85, + "learning_rate": 3.549694815991006e-06, + "loss": 0.0061, + "step": 44672 + }, + { + "epoch": 5.85, + "learning_rate": 3.549289573551674e-06, + "loss": 0.0033, + "step": 44673 + }, + { + "epoch": 5.85, + "learning_rate": 3.5488843492546265e-06, + "loss": 0.0036, + "step": 44674 + }, + { + "epoch": 5.85, + "learning_rate": 3.548479143100998e-06, + "loss": 0.0044, + "step": 44675 + }, + { + "epoch": 5.85, + "learning_rate": 3.5480739550919243e-06, + "loss": 0.006, + "step": 44676 + }, + { + "epoch": 5.85, + "learning_rate": 3.547668785228553e-06, + "loss": 0.0057, + "step": 44677 + }, + { + "epoch": 5.85, + "learning_rate": 3.547263633512019e-06, + "loss": 0.0082, + "step": 44678 + }, + { + "epoch": 5.85, + "learning_rate": 3.546858499943463e-06, + "loss": 0.0068, + "step": 44679 + }, + { + "epoch": 5.85, + "learning_rate": 3.54645338452402e-06, + "loss": 0.0038, + "step": 44680 + }, + { + "epoch": 5.85, + "learning_rate": 3.546048287254834e-06, + "loss": 0.0063, + "step": 44681 + }, + { + "epoch": 5.85, + "learning_rate": 3.545643208137046e-06, + "loss": 0.0084, + "step": 44682 + }, + { + "epoch": 5.85, + "learning_rate": 3.5452381471717935e-06, + "loss": 0.0062, + "step": 44683 + }, + { + "epoch": 5.85, + "learning_rate": 3.5448331043602102e-06, + "loss": 0.0061, + "step": 44684 + }, + { + "epoch": 5.85, + "learning_rate": 3.5444280797034446e-06, + "loss": 0.0079, + "step": 44685 + }, + { + "epoch": 5.85, + "learning_rate": 3.5440230732026315e-06, + "loss": 0.0055, + "step": 44686 + }, + { + "epoch": 5.85, + "learning_rate": 3.5436180848589086e-06, + "loss": 0.0049, + "step": 44687 + }, + { + "epoch": 5.85, + "learning_rate": 3.543213114673414e-06, + "loss": 0.0049, + "step": 44688 + }, + { + "epoch": 5.85, + "learning_rate": 3.542808162647291e-06, + "loss": 0.0056, + "step": 44689 + }, + { + "epoch": 5.85, + "learning_rate": 3.5424032287816768e-06, + "loss": 0.0077, + "step": 44690 + }, + { + "epoch": 5.85, + "learning_rate": 3.541998313077706e-06, + "loss": 0.0042, + "step": 44691 + }, + { + "epoch": 5.85, + "learning_rate": 3.5415934155365205e-06, + "loss": 0.0055, + "step": 44692 + }, + { + "epoch": 5.85, + "learning_rate": 3.5411885361592634e-06, + "loss": 0.0064, + "step": 44693 + }, + { + "epoch": 5.85, + "learning_rate": 3.540783674947069e-06, + "loss": 0.0101, + "step": 44694 + }, + { + "epoch": 5.85, + "learning_rate": 3.540378831901078e-06, + "loss": 0.0057, + "step": 44695 + }, + { + "epoch": 5.85, + "learning_rate": 3.539974007022423e-06, + "loss": 0.0053, + "step": 44696 + }, + { + "epoch": 5.85, + "learning_rate": 3.5395692003122507e-06, + "loss": 0.0044, + "step": 44697 + }, + { + "epoch": 5.86, + "learning_rate": 3.5391644117716963e-06, + "loss": 0.0074, + "step": 44698 + }, + { + "epoch": 5.86, + "learning_rate": 3.538759641401894e-06, + "loss": 0.0064, + "step": 44699 + }, + { + "epoch": 5.86, + "learning_rate": 3.5383548892039897e-06, + "loss": 0.004, + "step": 44700 + }, + { + "epoch": 5.86, + "learning_rate": 3.5379501551791186e-06, + "loss": 0.0058, + "step": 44701 + }, + { + "epoch": 5.86, + "learning_rate": 3.5375454393284148e-06, + "loss": 0.0078, + "step": 44702 + }, + { + "epoch": 5.86, + "learning_rate": 3.5371407416530233e-06, + "loss": 0.0047, + "step": 44703 + }, + { + "epoch": 5.86, + "learning_rate": 3.536736062154077e-06, + "loss": 0.0084, + "step": 44704 + }, + { + "epoch": 5.86, + "learning_rate": 3.536331400832719e-06, + "loss": 0.0063, + "step": 44705 + }, + { + "epoch": 5.86, + "learning_rate": 3.535926757690085e-06, + "loss": 0.0071, + "step": 44706 + }, + { + "epoch": 5.86, + "learning_rate": 3.535522132727309e-06, + "loss": 0.0076, + "step": 44707 + }, + { + "epoch": 5.86, + "learning_rate": 3.5351175259455363e-06, + "loss": 0.0097, + "step": 44708 + }, + { + "epoch": 5.86, + "learning_rate": 3.534712937345901e-06, + "loss": 0.0063, + "step": 44709 + }, + { + "epoch": 5.86, + "learning_rate": 3.534308366929542e-06, + "loss": 0.006, + "step": 44710 + }, + { + "epoch": 5.86, + "learning_rate": 3.5339038146975924e-06, + "loss": 0.008, + "step": 44711 + }, + { + "epoch": 5.86, + "learning_rate": 3.5334992806511948e-06, + "loss": 0.004, + "step": 44712 + }, + { + "epoch": 5.86, + "learning_rate": 3.533094764791488e-06, + "loss": 0.0061, + "step": 44713 + }, + { + "epoch": 5.86, + "learning_rate": 3.5326902671196095e-06, + "loss": 0.0059, + "step": 44714 + }, + { + "epoch": 5.86, + "learning_rate": 3.532285787636691e-06, + "loss": 0.0042, + "step": 44715 + }, + { + "epoch": 5.86, + "learning_rate": 3.531881326343878e-06, + "loss": 0.0056, + "step": 44716 + }, + { + "epoch": 5.86, + "learning_rate": 3.531476883242303e-06, + "loss": 0.0057, + "step": 44717 + }, + { + "epoch": 5.86, + "learning_rate": 3.5310724583331057e-06, + "loss": 0.0074, + "step": 44718 + }, + { + "epoch": 5.86, + "learning_rate": 3.53066805161742e-06, + "loss": 0.0073, + "step": 44719 + }, + { + "epoch": 5.86, + "learning_rate": 3.5302636630963884e-06, + "loss": 0.0057, + "step": 44720 + }, + { + "epoch": 5.86, + "learning_rate": 3.529859292771146e-06, + "loss": 0.0066, + "step": 44721 + }, + { + "epoch": 5.86, + "learning_rate": 3.529454940642826e-06, + "loss": 0.0042, + "step": 44722 + }, + { + "epoch": 5.86, + "learning_rate": 3.5290506067125694e-06, + "loss": 0.0057, + "step": 44723 + }, + { + "epoch": 5.86, + "learning_rate": 3.5286462909815176e-06, + "loss": 0.0073, + "step": 44724 + }, + { + "epoch": 5.86, + "learning_rate": 3.528241993450803e-06, + "loss": 0.0058, + "step": 44725 + }, + { + "epoch": 5.86, + "learning_rate": 3.5278377141215604e-06, + "loss": 0.0058, + "step": 44726 + }, + { + "epoch": 5.86, + "learning_rate": 3.5274334529949318e-06, + "loss": 0.0049, + "step": 44727 + }, + { + "epoch": 5.86, + "learning_rate": 3.5270292100720526e-06, + "loss": 0.0068, + "step": 44728 + }, + { + "epoch": 5.86, + "learning_rate": 3.526624985354059e-06, + "loss": 0.0084, + "step": 44729 + }, + { + "epoch": 5.86, + "learning_rate": 3.526220778842084e-06, + "loss": 0.0073, + "step": 44730 + }, + { + "epoch": 5.86, + "learning_rate": 3.525816590537272e-06, + "loss": 0.0085, + "step": 44731 + }, + { + "epoch": 5.86, + "learning_rate": 3.525412420440756e-06, + "loss": 0.0052, + "step": 44732 + }, + { + "epoch": 5.86, + "learning_rate": 3.5250082685536733e-06, + "loss": 0.0067, + "step": 44733 + }, + { + "epoch": 5.86, + "learning_rate": 3.524604134877153e-06, + "loss": 0.0042, + "step": 44734 + }, + { + "epoch": 5.86, + "learning_rate": 3.5242000194123448e-06, + "loss": 0.0057, + "step": 44735 + }, + { + "epoch": 5.86, + "learning_rate": 3.52379592216038e-06, + "loss": 0.0046, + "step": 44736 + }, + { + "epoch": 5.86, + "learning_rate": 3.5233918431223947e-06, + "loss": 0.0068, + "step": 44737 + }, + { + "epoch": 5.86, + "learning_rate": 3.5229877822995205e-06, + "loss": 0.0051, + "step": 44738 + }, + { + "epoch": 5.86, + "learning_rate": 3.522583739692902e-06, + "loss": 0.006, + "step": 44739 + }, + { + "epoch": 5.86, + "learning_rate": 3.522179715303672e-06, + "loss": 0.0067, + "step": 44740 + }, + { + "epoch": 5.86, + "learning_rate": 3.5217757091329628e-06, + "loss": 0.0047, + "step": 44741 + }, + { + "epoch": 5.86, + "learning_rate": 3.5213717211819186e-06, + "loss": 0.0057, + "step": 44742 + }, + { + "epoch": 5.86, + "learning_rate": 3.5209677514516704e-06, + "loss": 0.0076, + "step": 44743 + }, + { + "epoch": 5.86, + "learning_rate": 3.520563799943356e-06, + "loss": 0.0079, + "step": 44744 + }, + { + "epoch": 5.86, + "learning_rate": 3.520159866658107e-06, + "loss": 0.0058, + "step": 44745 + }, + { + "epoch": 5.86, + "learning_rate": 3.5197559515970635e-06, + "loss": 0.0075, + "step": 44746 + }, + { + "epoch": 5.86, + "learning_rate": 3.5193520547613645e-06, + "loss": 0.0056, + "step": 44747 + }, + { + "epoch": 5.86, + "learning_rate": 3.518948176152144e-06, + "loss": 0.0065, + "step": 44748 + }, + { + "epoch": 5.86, + "learning_rate": 3.5185443157705325e-06, + "loss": 0.0041, + "step": 44749 + }, + { + "epoch": 5.86, + "learning_rate": 3.518140473617673e-06, + "loss": 0.0044, + "step": 44750 + }, + { + "epoch": 5.86, + "learning_rate": 3.5177366496946984e-06, + "loss": 0.0073, + "step": 44751 + }, + { + "epoch": 5.86, + "learning_rate": 3.5173328440027443e-06, + "loss": 0.0087, + "step": 44752 + }, + { + "epoch": 5.86, + "learning_rate": 3.5169290565429427e-06, + "loss": 0.007, + "step": 44753 + }, + { + "epoch": 5.86, + "learning_rate": 3.516525287316436e-06, + "loss": 0.0054, + "step": 44754 + }, + { + "epoch": 5.86, + "learning_rate": 3.516121536324357e-06, + "loss": 0.0058, + "step": 44755 + }, + { + "epoch": 5.86, + "learning_rate": 3.5157178035678386e-06, + "loss": 0.0059, + "step": 44756 + }, + { + "epoch": 5.86, + "learning_rate": 3.5153140890480177e-06, + "loss": 0.0063, + "step": 44757 + }, + { + "epoch": 5.86, + "learning_rate": 3.5149103927660333e-06, + "loss": 0.0051, + "step": 44758 + }, + { + "epoch": 5.86, + "learning_rate": 3.5145067147230184e-06, + "loss": 0.006, + "step": 44759 + }, + { + "epoch": 5.86, + "learning_rate": 3.5141030549201084e-06, + "loss": 0.0083, + "step": 44760 + }, + { + "epoch": 5.86, + "learning_rate": 3.5136994133584345e-06, + "loss": 0.0058, + "step": 44761 + }, + { + "epoch": 5.86, + "learning_rate": 3.5132957900391398e-06, + "loss": 0.0062, + "step": 44762 + }, + { + "epoch": 5.86, + "learning_rate": 3.5128921849633533e-06, + "loss": 0.008, + "step": 44763 + }, + { + "epoch": 5.86, + "learning_rate": 3.512488598132211e-06, + "loss": 0.0052, + "step": 44764 + }, + { + "epoch": 5.86, + "learning_rate": 3.5120850295468513e-06, + "loss": 0.0086, + "step": 44765 + }, + { + "epoch": 5.86, + "learning_rate": 3.5116814792084066e-06, + "loss": 0.0064, + "step": 44766 + }, + { + "epoch": 5.86, + "learning_rate": 3.5112779471180094e-06, + "loss": 0.0055, + "step": 44767 + }, + { + "epoch": 5.86, + "learning_rate": 3.5108744332768007e-06, + "loss": 0.0047, + "step": 44768 + }, + { + "epoch": 5.86, + "learning_rate": 3.5104709376859082e-06, + "loss": 0.0068, + "step": 44769 + }, + { + "epoch": 5.86, + "learning_rate": 3.5100674603464748e-06, + "loss": 0.005, + "step": 44770 + }, + { + "epoch": 5.86, + "learning_rate": 3.5096640012596306e-06, + "loss": 0.0062, + "step": 44771 + }, + { + "epoch": 5.86, + "learning_rate": 3.5092605604265074e-06, + "loss": 0.0084, + "step": 44772 + }, + { + "epoch": 5.86, + "learning_rate": 3.5088571378482474e-06, + "loss": 0.0116, + "step": 44773 + }, + { + "epoch": 5.87, + "learning_rate": 3.5084537335259792e-06, + "loss": 0.0069, + "step": 44774 + }, + { + "epoch": 5.87, + "learning_rate": 3.508050347460841e-06, + "loss": 0.0054, + "step": 44775 + }, + { + "epoch": 5.87, + "learning_rate": 3.5076469796539615e-06, + "loss": 0.0088, + "step": 44776 + }, + { + "epoch": 5.87, + "learning_rate": 3.5072436301064828e-06, + "loss": 0.0048, + "step": 44777 + }, + { + "epoch": 5.87, + "learning_rate": 3.506840298819532e-06, + "loss": 0.0067, + "step": 44778 + }, + { + "epoch": 5.87, + "learning_rate": 3.5064369857942495e-06, + "loss": 0.0067, + "step": 44779 + }, + { + "epoch": 5.87, + "learning_rate": 3.506033691031765e-06, + "loss": 0.0062, + "step": 44780 + }, + { + "epoch": 5.87, + "learning_rate": 3.5056304145332178e-06, + "loss": 0.0057, + "step": 44781 + }, + { + "epoch": 5.87, + "learning_rate": 3.5052271562997384e-06, + "loss": 0.0038, + "step": 44782 + }, + { + "epoch": 5.87, + "learning_rate": 3.5048239163324636e-06, + "loss": 0.0053, + "step": 44783 + }, + { + "epoch": 5.87, + "learning_rate": 3.504420694632521e-06, + "loss": 0.0061, + "step": 44784 + }, + { + "epoch": 5.87, + "learning_rate": 3.504017491201053e-06, + "loss": 0.0049, + "step": 44785 + }, + { + "epoch": 5.87, + "learning_rate": 3.5036143060391903e-06, + "loss": 0.0068, + "step": 44786 + }, + { + "epoch": 5.87, + "learning_rate": 3.5032111391480627e-06, + "loss": 0.0056, + "step": 44787 + }, + { + "epoch": 5.87, + "learning_rate": 3.5028079905288104e-06, + "loss": 0.005, + "step": 44788 + }, + { + "epoch": 5.87, + "learning_rate": 3.5024048601825623e-06, + "loss": 0.0035, + "step": 44789 + }, + { + "epoch": 5.87, + "learning_rate": 3.5020017481104574e-06, + "loss": 0.0047, + "step": 44790 + }, + { + "epoch": 5.87, + "learning_rate": 3.5015986543136237e-06, + "loss": 0.0078, + "step": 44791 + }, + { + "epoch": 5.87, + "learning_rate": 3.5011955787932005e-06, + "loss": 0.009, + "step": 44792 + }, + { + "epoch": 5.87, + "learning_rate": 3.5007925215503193e-06, + "loss": 0.0062, + "step": 44793 + }, + { + "epoch": 5.87, + "learning_rate": 3.500389482586112e-06, + "loss": 0.0083, + "step": 44794 + }, + { + "epoch": 5.87, + "learning_rate": 3.499986461901711e-06, + "loss": 0.0064, + "step": 44795 + }, + { + "epoch": 5.87, + "learning_rate": 3.4995834594982546e-06, + "loss": 0.0079, + "step": 44796 + }, + { + "epoch": 5.87, + "learning_rate": 3.4991804753768732e-06, + "loss": 0.005, + "step": 44797 + }, + { + "epoch": 5.87, + "learning_rate": 3.498777509538701e-06, + "loss": 0.0056, + "step": 44798 + }, + { + "epoch": 5.87, + "learning_rate": 3.4983745619848674e-06, + "loss": 0.0062, + "step": 44799 + }, + { + "epoch": 5.87, + "learning_rate": 3.4979716327165092e-06, + "loss": 0.008, + "step": 44800 + }, + { + "epoch": 5.87, + "learning_rate": 3.497568721734763e-06, + "loss": 0.0048, + "step": 44801 + }, + { + "epoch": 5.87, + "learning_rate": 3.4971658290407586e-06, + "loss": 0.0048, + "step": 44802 + }, + { + "epoch": 5.87, + "learning_rate": 3.496762954635625e-06, + "loss": 0.0064, + "step": 44803 + }, + { + "epoch": 5.87, + "learning_rate": 3.4963600985205026e-06, + "loss": 0.0049, + "step": 44804 + }, + { + "epoch": 5.87, + "learning_rate": 3.495957260696522e-06, + "loss": 0.0073, + "step": 44805 + }, + { + "epoch": 5.87, + "learning_rate": 3.4955544411648113e-06, + "loss": 0.0043, + "step": 44806 + }, + { + "epoch": 5.87, + "learning_rate": 3.4951516399265105e-06, + "loss": 0.0074, + "step": 44807 + }, + { + "epoch": 5.87, + "learning_rate": 3.4947488569827505e-06, + "loss": 0.0057, + "step": 44808 + }, + { + "epoch": 5.87, + "learning_rate": 3.4943460923346615e-06, + "loss": 0.0078, + "step": 44809 + }, + { + "epoch": 5.87, + "learning_rate": 3.493943345983375e-06, + "loss": 0.0069, + "step": 44810 + }, + { + "epoch": 5.87, + "learning_rate": 3.4935406179300256e-06, + "loss": 0.0053, + "step": 44811 + }, + { + "epoch": 5.87, + "learning_rate": 3.4931379081757512e-06, + "loss": 0.0083, + "step": 44812 + }, + { + "epoch": 5.87, + "learning_rate": 3.4927352167216787e-06, + "loss": 0.0053, + "step": 44813 + }, + { + "epoch": 5.87, + "learning_rate": 3.4923325435689393e-06, + "loss": 0.0081, + "step": 44814 + }, + { + "epoch": 5.87, + "learning_rate": 3.4919298887186714e-06, + "loss": 0.0073, + "step": 44815 + }, + { + "epoch": 5.87, + "learning_rate": 3.4915272521720046e-06, + "loss": 0.0061, + "step": 44816 + }, + { + "epoch": 5.87, + "learning_rate": 3.49112463393007e-06, + "loss": 0.0062, + "step": 44817 + }, + { + "epoch": 5.87, + "learning_rate": 3.4907220339939975e-06, + "loss": 0.0083, + "step": 44818 + }, + { + "epoch": 5.87, + "learning_rate": 3.4903194523649265e-06, + "loss": 0.0078, + "step": 44819 + }, + { + "epoch": 5.87, + "learning_rate": 3.489916889043985e-06, + "loss": 0.0064, + "step": 44820 + }, + { + "epoch": 5.87, + "learning_rate": 3.489514344032302e-06, + "loss": 0.0058, + "step": 44821 + }, + { + "epoch": 5.87, + "learning_rate": 3.4891118173310134e-06, + "loss": 0.0075, + "step": 44822 + }, + { + "epoch": 5.87, + "learning_rate": 3.488709308941255e-06, + "loss": 0.0048, + "step": 44823 + }, + { + "epoch": 5.87, + "learning_rate": 3.4883068188641545e-06, + "loss": 0.0041, + "step": 44824 + }, + { + "epoch": 5.87, + "learning_rate": 3.4879043471008434e-06, + "loss": 0.0081, + "step": 44825 + }, + { + "epoch": 5.87, + "learning_rate": 3.487501893652453e-06, + "loss": 0.0052, + "step": 44826 + }, + { + "epoch": 5.87, + "learning_rate": 3.487099458520119e-06, + "loss": 0.0089, + "step": 44827 + }, + { + "epoch": 5.87, + "learning_rate": 3.4866970417049705e-06, + "loss": 0.0055, + "step": 44828 + }, + { + "epoch": 5.87, + "learning_rate": 3.486294643208137e-06, + "loss": 0.0047, + "step": 44829 + }, + { + "epoch": 5.87, + "learning_rate": 3.4858922630307557e-06, + "loss": 0.007, + "step": 44830 + }, + { + "epoch": 5.87, + "learning_rate": 3.4854899011739564e-06, + "loss": 0.006, + "step": 44831 + }, + { + "epoch": 5.87, + "learning_rate": 3.4850875576388686e-06, + "loss": 0.0051, + "step": 44832 + }, + { + "epoch": 5.87, + "learning_rate": 3.4846852324266235e-06, + "loss": 0.0083, + "step": 44833 + }, + { + "epoch": 5.87, + "learning_rate": 3.4842829255383526e-06, + "loss": 0.0059, + "step": 44834 + }, + { + "epoch": 5.87, + "learning_rate": 3.483880636975193e-06, + "loss": 0.006, + "step": 44835 + }, + { + "epoch": 5.87, + "learning_rate": 3.483478366738272e-06, + "loss": 0.006, + "step": 44836 + }, + { + "epoch": 5.87, + "learning_rate": 3.4830761148287185e-06, + "loss": 0.0053, + "step": 44837 + }, + { + "epoch": 5.87, + "learning_rate": 3.482673881247669e-06, + "loss": 0.0084, + "step": 44838 + }, + { + "epoch": 5.87, + "learning_rate": 3.482271665996252e-06, + "loss": 0.0043, + "step": 44839 + }, + { + "epoch": 5.87, + "learning_rate": 3.4818694690755996e-06, + "loss": 0.0067, + "step": 44840 + }, + { + "epoch": 5.87, + "learning_rate": 3.4814672904868387e-06, + "loss": 0.0076, + "step": 44841 + }, + { + "epoch": 5.87, + "learning_rate": 3.481065130231106e-06, + "loss": 0.0044, + "step": 44842 + }, + { + "epoch": 5.87, + "learning_rate": 3.480662988309528e-06, + "loss": 0.0062, + "step": 44843 + }, + { + "epoch": 5.87, + "learning_rate": 3.4802608647232415e-06, + "loss": 0.0058, + "step": 44844 + }, + { + "epoch": 5.87, + "learning_rate": 3.479858759473371e-06, + "loss": 0.0073, + "step": 44845 + }, + { + "epoch": 5.87, + "learning_rate": 3.4794566725610533e-06, + "loss": 0.0067, + "step": 44846 + }, + { + "epoch": 5.87, + "learning_rate": 3.4790546039874173e-06, + "loss": 0.0041, + "step": 44847 + }, + { + "epoch": 5.87, + "learning_rate": 3.478652553753592e-06, + "loss": 0.0058, + "step": 44848 + }, + { + "epoch": 5.87, + "learning_rate": 3.4782505218607064e-06, + "loss": 0.0073, + "step": 44849 + }, + { + "epoch": 5.88, + "learning_rate": 3.4778485083098978e-06, + "loss": 0.0062, + "step": 44850 + }, + { + "epoch": 5.88, + "learning_rate": 3.4774465131022915e-06, + "loss": 0.0083, + "step": 44851 + }, + { + "epoch": 5.88, + "learning_rate": 3.477044536239017e-06, + "loss": 0.0039, + "step": 44852 + }, + { + "epoch": 5.88, + "learning_rate": 3.476642577721211e-06, + "loss": 0.005, + "step": 44853 + }, + { + "epoch": 5.88, + "learning_rate": 3.476240637549997e-06, + "loss": 0.0073, + "step": 44854 + }, + { + "epoch": 5.88, + "learning_rate": 3.4758387157265118e-06, + "loss": 0.0049, + "step": 44855 + }, + { + "epoch": 5.88, + "learning_rate": 3.4754368122518787e-06, + "loss": 0.0063, + "step": 44856 + }, + { + "epoch": 5.88, + "learning_rate": 3.4750349271272366e-06, + "loss": 0.0085, + "step": 44857 + }, + { + "epoch": 5.88, + "learning_rate": 3.4746330603537114e-06, + "loss": 0.0054, + "step": 44858 + }, + { + "epoch": 5.88, + "learning_rate": 3.474231211932433e-06, + "loss": 0.0054, + "step": 44859 + }, + { + "epoch": 5.88, + "learning_rate": 3.4738293818645274e-06, + "loss": 0.0037, + "step": 44860 + }, + { + "epoch": 5.88, + "learning_rate": 3.4734275701511335e-06, + "loss": 0.0048, + "step": 44861 + }, + { + "epoch": 5.88, + "learning_rate": 3.4730257767933763e-06, + "loss": 0.0076, + "step": 44862 + }, + { + "epoch": 5.88, + "learning_rate": 3.472624001792387e-06, + "loss": 0.0078, + "step": 44863 + }, + { + "epoch": 5.88, + "learning_rate": 3.4722222451492913e-06, + "loss": 0.0054, + "step": 44864 + }, + { + "epoch": 5.88, + "learning_rate": 3.4718205068652243e-06, + "loss": 0.0035, + "step": 44865 + }, + { + "epoch": 5.88, + "learning_rate": 3.4714187869413163e-06, + "loss": 0.0049, + "step": 44866 + }, + { + "epoch": 5.88, + "learning_rate": 3.471017085378696e-06, + "loss": 0.0045, + "step": 44867 + }, + { + "epoch": 5.88, + "learning_rate": 3.470615402178489e-06, + "loss": 0.0044, + "step": 44868 + }, + { + "epoch": 5.88, + "learning_rate": 3.470213737341832e-06, + "loss": 0.0058, + "step": 44869 + }, + { + "epoch": 5.88, + "learning_rate": 3.4698120908698506e-06, + "loss": 0.008, + "step": 44870 + }, + { + "epoch": 5.88, + "learning_rate": 3.469410462763673e-06, + "loss": 0.0072, + "step": 44871 + }, + { + "epoch": 5.88, + "learning_rate": 3.4690088530244326e-06, + "loss": 0.0064, + "step": 44872 + }, + { + "epoch": 5.88, + "learning_rate": 3.4686072616532573e-06, + "loss": 0.0055, + "step": 44873 + }, + { + "epoch": 5.88, + "learning_rate": 3.468205688651276e-06, + "loss": 0.0061, + "step": 44874 + }, + { + "epoch": 5.88, + "learning_rate": 3.4678041340196156e-06, + "loss": 0.0039, + "step": 44875 + }, + { + "epoch": 5.88, + "learning_rate": 3.4674025977594074e-06, + "loss": 0.0068, + "step": 44876 + }, + { + "epoch": 5.88, + "learning_rate": 3.4670010798717856e-06, + "loss": 0.0044, + "step": 44877 + }, + { + "epoch": 5.88, + "learning_rate": 3.4665995803578754e-06, + "loss": 0.0063, + "step": 44878 + }, + { + "epoch": 5.88, + "learning_rate": 3.4661980992188017e-06, + "loss": 0.0065, + "step": 44879 + }, + { + "epoch": 5.88, + "learning_rate": 3.465796636455703e-06, + "loss": 0.0082, + "step": 44880 + }, + { + "epoch": 5.88, + "learning_rate": 3.4653951920697017e-06, + "loss": 0.0061, + "step": 44881 + }, + { + "epoch": 5.88, + "learning_rate": 3.4649937660619294e-06, + "loss": 0.0046, + "step": 44882 + }, + { + "epoch": 5.88, + "learning_rate": 3.4645923584335096e-06, + "loss": 0.0073, + "step": 44883 + }, + { + "epoch": 5.88, + "learning_rate": 3.4641909691855802e-06, + "loss": 0.0093, + "step": 44884 + }, + { + "epoch": 5.88, + "learning_rate": 3.463789598319265e-06, + "loss": 0.0051, + "step": 44885 + }, + { + "epoch": 5.88, + "learning_rate": 3.4633882458356905e-06, + "loss": 0.0042, + "step": 44886 + }, + { + "epoch": 5.88, + "learning_rate": 3.4629869117359884e-06, + "loss": 0.0073, + "step": 44887 + }, + { + "epoch": 5.88, + "learning_rate": 3.4625855960212905e-06, + "loss": 0.0079, + "step": 44888 + }, + { + "epoch": 5.88, + "learning_rate": 3.4621842986927235e-06, + "loss": 0.0051, + "step": 44889 + }, + { + "epoch": 5.88, + "learning_rate": 3.4617830197514135e-06, + "loss": 0.0039, + "step": 44890 + }, + { + "epoch": 5.88, + "learning_rate": 3.4613817591984886e-06, + "loss": 0.0056, + "step": 44891 + }, + { + "epoch": 5.88, + "learning_rate": 3.460980517035082e-06, + "loss": 0.0073, + "step": 44892 + }, + { + "epoch": 5.88, + "learning_rate": 3.4605792932623194e-06, + "loss": 0.0052, + "step": 44893 + }, + { + "epoch": 5.88, + "learning_rate": 3.460178087881326e-06, + "loss": 0.0079, + "step": 44894 + }, + { + "epoch": 5.88, + "learning_rate": 3.459776900893237e-06, + "loss": 0.0068, + "step": 44895 + }, + { + "epoch": 5.88, + "learning_rate": 3.459375732299177e-06, + "loss": 0.0074, + "step": 44896 + }, + { + "epoch": 5.88, + "learning_rate": 3.4589745821002753e-06, + "loss": 0.0072, + "step": 44897 + }, + { + "epoch": 5.88, + "learning_rate": 3.4585734502976544e-06, + "loss": 0.0059, + "step": 44898 + }, + { + "epoch": 5.88, + "learning_rate": 3.458172336892448e-06, + "loss": 0.0038, + "step": 44899 + }, + { + "epoch": 5.88, + "learning_rate": 3.4577712418857867e-06, + "loss": 0.0065, + "step": 44900 + }, + { + "epoch": 5.88, + "learning_rate": 3.457370165278796e-06, + "loss": 0.0064, + "step": 44901 + }, + { + "epoch": 5.88, + "learning_rate": 3.4569691070725996e-06, + "loss": 0.0057, + "step": 44902 + }, + { + "epoch": 5.88, + "learning_rate": 3.456568067268332e-06, + "loss": 0.0065, + "step": 44903 + }, + { + "epoch": 5.88, + "learning_rate": 3.4561670458671193e-06, + "loss": 0.0064, + "step": 44904 + }, + { + "epoch": 5.88, + "learning_rate": 3.4557660428700878e-06, + "loss": 0.0035, + "step": 44905 + }, + { + "epoch": 5.88, + "learning_rate": 3.4553650582783614e-06, + "loss": 0.006, + "step": 44906 + }, + { + "epoch": 5.88, + "learning_rate": 3.454964092093077e-06, + "loss": 0.0058, + "step": 44907 + }, + { + "epoch": 5.88, + "learning_rate": 3.454563144315357e-06, + "loss": 0.0064, + "step": 44908 + }, + { + "epoch": 5.88, + "learning_rate": 3.454162214946327e-06, + "loss": 0.0076, + "step": 44909 + }, + { + "epoch": 5.88, + "learning_rate": 3.4537613039871163e-06, + "loss": 0.0064, + "step": 44910 + }, + { + "epoch": 5.88, + "learning_rate": 3.4533604114388574e-06, + "loss": 0.0065, + "step": 44911 + }, + { + "epoch": 5.88, + "learning_rate": 3.4529595373026727e-06, + "loss": 0.0061, + "step": 44912 + }, + { + "epoch": 5.88, + "learning_rate": 3.452558681579691e-06, + "loss": 0.0056, + "step": 44913 + }, + { + "epoch": 5.88, + "learning_rate": 3.4521578442710367e-06, + "loss": 0.0068, + "step": 44914 + }, + { + "epoch": 5.88, + "learning_rate": 3.451757025377842e-06, + "loss": 0.0043, + "step": 44915 + }, + { + "epoch": 5.88, + "learning_rate": 3.4513562249012334e-06, + "loss": 0.0045, + "step": 44916 + }, + { + "epoch": 5.88, + "learning_rate": 3.450955442842332e-06, + "loss": 0.0074, + "step": 44917 + }, + { + "epoch": 5.88, + "learning_rate": 3.4505546792022737e-06, + "loss": 0.005, + "step": 44918 + }, + { + "epoch": 5.88, + "learning_rate": 3.450153933982181e-06, + "loss": 0.0043, + "step": 44919 + }, + { + "epoch": 5.88, + "learning_rate": 3.4497532071831787e-06, + "loss": 0.0055, + "step": 44920 + }, + { + "epoch": 5.88, + "learning_rate": 3.4493524988063963e-06, + "loss": 0.0057, + "step": 44921 + }, + { + "epoch": 5.88, + "learning_rate": 3.448951808852965e-06, + "loss": 0.0049, + "step": 44922 + }, + { + "epoch": 5.88, + "learning_rate": 3.448551137324008e-06, + "loss": 0.0063, + "step": 44923 + }, + { + "epoch": 5.88, + "learning_rate": 3.4481504842206515e-06, + "loss": 0.0037, + "step": 44924 + }, + { + "epoch": 5.88, + "learning_rate": 3.44774984954402e-06, + "loss": 0.0062, + "step": 44925 + }, + { + "epoch": 5.88, + "learning_rate": 3.4473492332952454e-06, + "loss": 0.006, + "step": 44926 + }, + { + "epoch": 5.89, + "learning_rate": 3.4469486354754535e-06, + "loss": 0.0053, + "step": 44927 + }, + { + "epoch": 5.89, + "learning_rate": 3.4465480560857657e-06, + "loss": 0.0048, + "step": 44928 + }, + { + "epoch": 5.89, + "learning_rate": 3.4461474951273153e-06, + "loss": 0.005, + "step": 44929 + }, + { + "epoch": 5.89, + "learning_rate": 3.4457469526012265e-06, + "loss": 0.0071, + "step": 44930 + }, + { + "epoch": 5.89, + "learning_rate": 3.4453464285086223e-06, + "loss": 0.004, + "step": 44931 + }, + { + "epoch": 5.89, + "learning_rate": 3.4449459228506356e-06, + "loss": 0.0067, + "step": 44932 + }, + { + "epoch": 5.89, + "learning_rate": 3.4445454356283858e-06, + "loss": 0.0083, + "step": 44933 + }, + { + "epoch": 5.89, + "learning_rate": 3.444144966843006e-06, + "loss": 0.0074, + "step": 44934 + }, + { + "epoch": 5.89, + "learning_rate": 3.4437445164956196e-06, + "loss": 0.0084, + "step": 44935 + }, + { + "epoch": 5.89, + "learning_rate": 3.443344084587349e-06, + "loss": 0.0066, + "step": 44936 + }, + { + "epoch": 5.89, + "learning_rate": 3.4429436711193277e-06, + "loss": 0.0054, + "step": 44937 + }, + { + "epoch": 5.89, + "learning_rate": 3.442543276092678e-06, + "loss": 0.005, + "step": 44938 + }, + { + "epoch": 5.89, + "learning_rate": 3.442142899508526e-06, + "loss": 0.0061, + "step": 44939 + }, + { + "epoch": 5.89, + "learning_rate": 3.4417425413679952e-06, + "loss": 0.0035, + "step": 44940 + }, + { + "epoch": 5.89, + "learning_rate": 3.4413422016722164e-06, + "loss": 0.005, + "step": 44941 + }, + { + "epoch": 5.89, + "learning_rate": 3.4409418804223106e-06, + "loss": 0.0065, + "step": 44942 + }, + { + "epoch": 5.89, + "learning_rate": 3.4405415776194105e-06, + "loss": 0.005, + "step": 44943 + }, + { + "epoch": 5.89, + "learning_rate": 3.440141293264634e-06, + "loss": 0.0064, + "step": 44944 + }, + { + "epoch": 5.89, + "learning_rate": 3.4397410273591137e-06, + "loss": 0.0052, + "step": 44945 + }, + { + "epoch": 5.89, + "learning_rate": 3.4393407799039723e-06, + "loss": 0.0059, + "step": 44946 + }, + { + "epoch": 5.89, + "learning_rate": 3.438940550900337e-06, + "loss": 0.0063, + "step": 44947 + }, + { + "epoch": 5.89, + "learning_rate": 3.438540340349328e-06, + "loss": 0.0056, + "step": 44948 + }, + { + "epoch": 5.89, + "learning_rate": 3.438140148252077e-06, + "loss": 0.0071, + "step": 44949 + }, + { + "epoch": 5.89, + "learning_rate": 3.4377399746097084e-06, + "loss": 0.0054, + "step": 44950 + }, + { + "epoch": 5.89, + "learning_rate": 3.437339819423343e-06, + "loss": 0.0065, + "step": 44951 + }, + { + "epoch": 5.89, + "learning_rate": 3.4369396826941137e-06, + "loss": 0.0051, + "step": 44952 + }, + { + "epoch": 5.89, + "learning_rate": 3.436539564423139e-06, + "loss": 0.0063, + "step": 44953 + }, + { + "epoch": 5.89, + "learning_rate": 3.4361394646115498e-06, + "loss": 0.004, + "step": 44954 + }, + { + "epoch": 5.89, + "learning_rate": 3.4357393832604692e-06, + "loss": 0.0076, + "step": 44955 + }, + { + "epoch": 5.89, + "learning_rate": 3.4353393203710185e-06, + "loss": 0.005, + "step": 44956 + }, + { + "epoch": 5.89, + "learning_rate": 3.43493927594433e-06, + "loss": 0.0067, + "step": 44957 + }, + { + "epoch": 5.89, + "learning_rate": 3.4345392499815243e-06, + "loss": 0.0066, + "step": 44958 + }, + { + "epoch": 5.89, + "learning_rate": 3.434139242483725e-06, + "loss": 0.0084, + "step": 44959 + }, + { + "epoch": 5.89, + "learning_rate": 3.433739253452063e-06, + "loss": 0.0049, + "step": 44960 + }, + { + "epoch": 5.89, + "learning_rate": 3.433339282887659e-06, + "loss": 0.0068, + "step": 44961 + }, + { + "epoch": 5.89, + "learning_rate": 3.432939330791638e-06, + "loss": 0.0074, + "step": 44962 + }, + { + "epoch": 5.89, + "learning_rate": 3.4325393971651234e-06, + "loss": 0.0042, + "step": 44963 + }, + { + "epoch": 5.89, + "learning_rate": 3.432139482009241e-06, + "loss": 0.0049, + "step": 44964 + }, + { + "epoch": 5.89, + "learning_rate": 3.43173958532512e-06, + "loss": 0.0043, + "step": 44965 + }, + { + "epoch": 5.89, + "learning_rate": 3.431339707113882e-06, + "loss": 0.0058, + "step": 44966 + }, + { + "epoch": 5.89, + "learning_rate": 3.4309398473766475e-06, + "loss": 0.0057, + "step": 44967 + }, + { + "epoch": 5.89, + "learning_rate": 3.430540006114549e-06, + "loss": 0.0054, + "step": 44968 + }, + { + "epoch": 5.89, + "learning_rate": 3.430140183328706e-06, + "loss": 0.0039, + "step": 44969 + }, + { + "epoch": 5.89, + "learning_rate": 3.4297403790202445e-06, + "loss": 0.0065, + "step": 44970 + }, + { + "epoch": 5.89, + "learning_rate": 3.4293405931902844e-06, + "loss": 0.005, + "step": 44971 + }, + { + "epoch": 5.89, + "learning_rate": 3.428940825839958e-06, + "loss": 0.0066, + "step": 44972 + }, + { + "epoch": 5.89, + "learning_rate": 3.4285410769703853e-06, + "loss": 0.0054, + "step": 44973 + }, + { + "epoch": 5.89, + "learning_rate": 3.428141346582687e-06, + "loss": 0.0083, + "step": 44974 + }, + { + "epoch": 5.89, + "learning_rate": 3.4277416346779926e-06, + "loss": 0.0038, + "step": 44975 + }, + { + "epoch": 5.89, + "learning_rate": 3.427341941257427e-06, + "loss": 0.0096, + "step": 44976 + }, + { + "epoch": 5.89, + "learning_rate": 3.4269422663221118e-06, + "loss": 0.0052, + "step": 44977 + }, + { + "epoch": 5.89, + "learning_rate": 3.426542609873169e-06, + "loss": 0.0042, + "step": 44978 + }, + { + "epoch": 5.89, + "learning_rate": 3.426142971911728e-06, + "loss": 0.0066, + "step": 44979 + }, + { + "epoch": 5.89, + "learning_rate": 3.4257433524389107e-06, + "loss": 0.0046, + "step": 44980 + }, + { + "epoch": 5.89, + "learning_rate": 3.4253437514558385e-06, + "loss": 0.0062, + "step": 44981 + }, + { + "epoch": 5.89, + "learning_rate": 3.4249441689636344e-06, + "loss": 0.0065, + "step": 44982 + }, + { + "epoch": 5.89, + "learning_rate": 3.424544604963428e-06, + "loss": 0.0054, + "step": 44983 + }, + { + "epoch": 5.89, + "learning_rate": 3.42414505945634e-06, + "loss": 0.0039, + "step": 44984 + }, + { + "epoch": 5.89, + "learning_rate": 3.42374553244349e-06, + "loss": 0.0058, + "step": 44985 + }, + { + "epoch": 5.89, + "learning_rate": 3.423346023926006e-06, + "loss": 0.0067, + "step": 44986 + }, + { + "epoch": 5.89, + "learning_rate": 3.422946533905014e-06, + "loss": 0.0043, + "step": 44987 + }, + { + "epoch": 5.89, + "learning_rate": 3.4225470623816346e-06, + "loss": 0.0071, + "step": 44988 + }, + { + "epoch": 5.89, + "learning_rate": 3.422147609356992e-06, + "loss": 0.0057, + "step": 44989 + }, + { + "epoch": 5.89, + "learning_rate": 3.4217481748322046e-06, + "loss": 0.0037, + "step": 44990 + }, + { + "epoch": 5.89, + "learning_rate": 3.421348758808404e-06, + "loss": 0.0072, + "step": 44991 + }, + { + "epoch": 5.89, + "learning_rate": 3.4209493612867096e-06, + "loss": 0.0056, + "step": 44992 + }, + { + "epoch": 5.89, + "learning_rate": 3.420549982268242e-06, + "loss": 0.0061, + "step": 44993 + }, + { + "epoch": 5.89, + "learning_rate": 3.42015062175413e-06, + "loss": 0.0052, + "step": 44994 + }, + { + "epoch": 5.89, + "learning_rate": 3.4197512797454934e-06, + "loss": 0.0069, + "step": 44995 + }, + { + "epoch": 5.89, + "learning_rate": 3.4193519562434542e-06, + "loss": 0.0046, + "step": 44996 + }, + { + "epoch": 5.89, + "learning_rate": 3.4189526512491387e-06, + "loss": 0.0059, + "step": 44997 + }, + { + "epoch": 5.89, + "learning_rate": 3.418553364763666e-06, + "loss": 0.0056, + "step": 44998 + }, + { + "epoch": 5.89, + "learning_rate": 3.4181540967881642e-06, + "loss": 0.006, + "step": 44999 + }, + { + "epoch": 5.89, + "learning_rate": 3.417754847323754e-06, + "loss": 0.0044, + "step": 45000 + }, + { + "epoch": 5.89, + "learning_rate": 3.417355616371555e-06, + "loss": 0.004, + "step": 45001 + }, + { + "epoch": 5.89, + "learning_rate": 3.416956403932695e-06, + "loss": 0.0058, + "step": 45002 + }, + { + "epoch": 5.9, + "learning_rate": 3.4165572100082954e-06, + "loss": 0.0051, + "step": 45003 + }, + { + "epoch": 5.9, + "learning_rate": 3.4161580345994783e-06, + "loss": 0.0047, + "step": 45004 + }, + { + "epoch": 5.9, + "learning_rate": 3.4157588777073612e-06, + "loss": 0.0059, + "step": 45005 + }, + { + "epoch": 5.9, + "learning_rate": 3.4153597393330763e-06, + "loss": 0.0073, + "step": 45006 + }, + { + "epoch": 5.9, + "learning_rate": 3.4149606194777383e-06, + "loss": 0.0085, + "step": 45007 + }, + { + "epoch": 5.9, + "learning_rate": 3.414561518142475e-06, + "loss": 0.0055, + "step": 45008 + }, + { + "epoch": 5.9, + "learning_rate": 3.4141624353284053e-06, + "loss": 0.0054, + "step": 45009 + }, + { + "epoch": 5.9, + "learning_rate": 3.4137633710366545e-06, + "loss": 0.0055, + "step": 45010 + }, + { + "epoch": 5.9, + "learning_rate": 3.4133643252683446e-06, + "loss": 0.0063, + "step": 45011 + }, + { + "epoch": 5.9, + "learning_rate": 3.4129652980245964e-06, + "loss": 0.0058, + "step": 45012 + }, + { + "epoch": 5.9, + "learning_rate": 3.412566289306529e-06, + "loss": 0.0074, + "step": 45013 + }, + { + "epoch": 5.9, + "learning_rate": 3.4121672991152712e-06, + "loss": 0.0063, + "step": 45014 + }, + { + "epoch": 5.9, + "learning_rate": 3.411768327451943e-06, + "loss": 0.0067, + "step": 45015 + }, + { + "epoch": 5.9, + "learning_rate": 3.4113693743176612e-06, + "loss": 0.0084, + "step": 45016 + }, + { + "epoch": 5.9, + "learning_rate": 3.410970439713556e-06, + "loss": 0.0037, + "step": 45017 + }, + { + "epoch": 5.9, + "learning_rate": 3.410571523640742e-06, + "loss": 0.006, + "step": 45018 + }, + { + "epoch": 5.9, + "learning_rate": 3.410172626100349e-06, + "loss": 0.005, + "step": 45019 + }, + { + "epoch": 5.9, + "learning_rate": 3.4097737470934935e-06, + "loss": 0.0051, + "step": 45020 + }, + { + "epoch": 5.9, + "learning_rate": 3.409374886621295e-06, + "loss": 0.0055, + "step": 45021 + }, + { + "epoch": 5.9, + "learning_rate": 3.408976044684882e-06, + "loss": 0.008, + "step": 45022 + }, + { + "epoch": 5.9, + "learning_rate": 3.408577221285374e-06, + "loss": 0.0029, + "step": 45023 + }, + { + "epoch": 5.9, + "learning_rate": 3.4081784164238874e-06, + "loss": 0.0046, + "step": 45024 + }, + { + "epoch": 5.9, + "learning_rate": 3.407779630101551e-06, + "loss": 0.0047, + "step": 45025 + }, + { + "epoch": 5.9, + "learning_rate": 3.407380862319484e-06, + "loss": 0.0042, + "step": 45026 + }, + { + "epoch": 5.9, + "learning_rate": 3.4069821130788062e-06, + "loss": 0.0055, + "step": 45027 + }, + { + "epoch": 5.9, + "learning_rate": 3.4065833823806373e-06, + "loss": 0.005, + "step": 45028 + }, + { + "epoch": 5.9, + "learning_rate": 3.4061846702261017e-06, + "loss": 0.0081, + "step": 45029 + }, + { + "epoch": 5.9, + "learning_rate": 3.4057859766163236e-06, + "loss": 0.0068, + "step": 45030 + }, + { + "epoch": 5.9, + "learning_rate": 3.405387301552422e-06, + "loss": 0.0053, + "step": 45031 + }, + { + "epoch": 5.9, + "learning_rate": 3.404988645035514e-06, + "loss": 0.0052, + "step": 45032 + }, + { + "epoch": 5.9, + "learning_rate": 3.404590007066727e-06, + "loss": 0.0076, + "step": 45033 + }, + { + "epoch": 5.9, + "learning_rate": 3.40419138764718e-06, + "loss": 0.0059, + "step": 45034 + }, + { + "epoch": 5.9, + "learning_rate": 3.4037927867779928e-06, + "loss": 0.0062, + "step": 45035 + }, + { + "epoch": 5.9, + "learning_rate": 3.403394204460284e-06, + "loss": 0.0049, + "step": 45036 + }, + { + "epoch": 5.9, + "learning_rate": 3.4029956406951813e-06, + "loss": 0.0075, + "step": 45037 + }, + { + "epoch": 5.9, + "learning_rate": 3.4025970954838015e-06, + "loss": 0.0057, + "step": 45038 + }, + { + "epoch": 5.9, + "learning_rate": 3.402198568827263e-06, + "loss": 0.0041, + "step": 45039 + }, + { + "epoch": 5.9, + "learning_rate": 3.40180006072669e-06, + "loss": 0.0054, + "step": 45040 + }, + { + "epoch": 5.9, + "learning_rate": 3.401401571183206e-06, + "loss": 0.0071, + "step": 45041 + }, + { + "epoch": 5.9, + "learning_rate": 3.4010031001979282e-06, + "loss": 0.0058, + "step": 45042 + }, + { + "epoch": 5.9, + "learning_rate": 3.400604647771976e-06, + "loss": 0.0044, + "step": 45043 + }, + { + "epoch": 5.9, + "learning_rate": 3.4002062139064727e-06, + "loss": 0.0059, + "step": 45044 + }, + { + "epoch": 5.9, + "learning_rate": 3.39980779860254e-06, + "loss": 0.005, + "step": 45045 + }, + { + "epoch": 5.9, + "learning_rate": 3.3994094018612954e-06, + "loss": 0.0049, + "step": 45046 + }, + { + "epoch": 5.9, + "learning_rate": 3.399011023683857e-06, + "loss": 0.0049, + "step": 45047 + }, + { + "epoch": 5.9, + "learning_rate": 3.3986126640713525e-06, + "loss": 0.0055, + "step": 45048 + }, + { + "epoch": 5.9, + "learning_rate": 3.398214323024899e-06, + "loss": 0.0063, + "step": 45049 + }, + { + "epoch": 5.9, + "learning_rate": 3.397816000545614e-06, + "loss": 0.0071, + "step": 45050 + }, + { + "epoch": 5.9, + "learning_rate": 3.397417696634615e-06, + "loss": 0.006, + "step": 45051 + }, + { + "epoch": 5.9, + "learning_rate": 3.3970194112930332e-06, + "loss": 0.0058, + "step": 45052 + }, + { + "epoch": 5.9, + "learning_rate": 3.396621144521982e-06, + "loss": 0.0051, + "step": 45053 + }, + { + "epoch": 5.9, + "learning_rate": 3.3962228963225843e-06, + "loss": 0.0066, + "step": 45054 + }, + { + "epoch": 5.9, + "learning_rate": 3.395824666695953e-06, + "loss": 0.0065, + "step": 45055 + }, + { + "epoch": 5.9, + "learning_rate": 3.395426455643216e-06, + "loss": 0.0047, + "step": 45056 + }, + { + "epoch": 5.9, + "learning_rate": 3.395028263165492e-06, + "loss": 0.0046, + "step": 45057 + }, + { + "epoch": 5.9, + "learning_rate": 3.394630089263895e-06, + "loss": 0.0048, + "step": 45058 + }, + { + "epoch": 5.9, + "learning_rate": 3.3942319339395524e-06, + "loss": 0.0082, + "step": 45059 + }, + { + "epoch": 5.9, + "learning_rate": 3.393833797193581e-06, + "loss": 0.0045, + "step": 45060 + }, + { + "epoch": 5.9, + "learning_rate": 3.3934356790270995e-06, + "loss": 0.005, + "step": 45061 + }, + { + "epoch": 5.9, + "learning_rate": 3.3930375794412262e-06, + "loss": 0.0047, + "step": 45062 + }, + { + "epoch": 5.9, + "learning_rate": 3.3926394984370816e-06, + "loss": 0.0058, + "step": 45063 + }, + { + "epoch": 5.9, + "learning_rate": 3.3922414360157907e-06, + "loss": 0.0084, + "step": 45064 + }, + { + "epoch": 5.9, + "learning_rate": 3.3918433921784687e-06, + "loss": 0.0062, + "step": 45065 + }, + { + "epoch": 5.9, + "learning_rate": 3.3914453669262317e-06, + "loss": 0.0059, + "step": 45066 + }, + { + "epoch": 5.9, + "learning_rate": 3.391047360260207e-06, + "loss": 0.0068, + "step": 45067 + }, + { + "epoch": 5.9, + "learning_rate": 3.3906493721815083e-06, + "loss": 0.0056, + "step": 45068 + }, + { + "epoch": 5.9, + "learning_rate": 3.390251402691257e-06, + "loss": 0.006, + "step": 45069 + }, + { + "epoch": 5.9, + "learning_rate": 3.3898534517905678e-06, + "loss": 0.0065, + "step": 45070 + }, + { + "epoch": 5.9, + "learning_rate": 3.389455519480568e-06, + "loss": 0.0059, + "step": 45071 + }, + { + "epoch": 5.9, + "learning_rate": 3.3890576057623713e-06, + "loss": 0.006, + "step": 45072 + }, + { + "epoch": 5.9, + "learning_rate": 3.3886597106370954e-06, + "loss": 0.0064, + "step": 45073 + }, + { + "epoch": 5.9, + "learning_rate": 3.3882618341058627e-06, + "loss": 0.0087, + "step": 45074 + }, + { + "epoch": 5.9, + "learning_rate": 3.387863976169794e-06, + "loss": 0.0069, + "step": 45075 + }, + { + "epoch": 5.9, + "learning_rate": 3.3874661368300056e-06, + "loss": 0.0068, + "step": 45076 + }, + { + "epoch": 5.9, + "learning_rate": 3.387068316087617e-06, + "loss": 0.0075, + "step": 45077 + }, + { + "epoch": 5.9, + "learning_rate": 3.3866705139437427e-06, + "loss": 0.0067, + "step": 45078 + }, + { + "epoch": 5.91, + "learning_rate": 3.386272730399509e-06, + "loss": 0.007, + "step": 45079 + }, + { + "epoch": 5.91, + "learning_rate": 3.3858749654560317e-06, + "loss": 0.0072, + "step": 45080 + }, + { + "epoch": 5.91, + "learning_rate": 3.385477219114425e-06, + "loss": 0.006, + "step": 45081 + }, + { + "epoch": 5.91, + "learning_rate": 3.3850794913758144e-06, + "loss": 0.005, + "step": 45082 + }, + { + "epoch": 5.91, + "learning_rate": 3.384681782241316e-06, + "loss": 0.0061, + "step": 45083 + }, + { + "epoch": 5.91, + "learning_rate": 3.3842840917120446e-06, + "loss": 0.0077, + "step": 45084 + }, + { + "epoch": 5.91, + "learning_rate": 3.383886419789124e-06, + "loss": 0.0069, + "step": 45085 + }, + { + "epoch": 5.91, + "learning_rate": 3.383488766473668e-06, + "loss": 0.0048, + "step": 45086 + }, + { + "epoch": 5.91, + "learning_rate": 3.3830911317668015e-06, + "loss": 0.0066, + "step": 45087 + }, + { + "epoch": 5.91, + "learning_rate": 3.3826935156696373e-06, + "loss": 0.0057, + "step": 45088 + }, + { + "epoch": 5.91, + "learning_rate": 3.3822959181832924e-06, + "loss": 0.0061, + "step": 45089 + }, + { + "epoch": 5.91, + "learning_rate": 3.3818983393088924e-06, + "loss": 0.0048, + "step": 45090 + }, + { + "epoch": 5.91, + "learning_rate": 3.3815007790475496e-06, + "loss": 0.0076, + "step": 45091 + }, + { + "epoch": 5.91, + "learning_rate": 3.381103237400384e-06, + "loss": 0.0052, + "step": 45092 + }, + { + "epoch": 5.91, + "learning_rate": 3.3807057143685096e-06, + "loss": 0.0052, + "step": 45093 + }, + { + "epoch": 5.91, + "learning_rate": 3.3803082099530502e-06, + "loss": 0.0069, + "step": 45094 + }, + { + "epoch": 5.91, + "learning_rate": 3.37991072415512e-06, + "loss": 0.0036, + "step": 45095 + }, + { + "epoch": 5.91, + "learning_rate": 3.3795132569758406e-06, + "loss": 0.0067, + "step": 45096 + }, + { + "epoch": 5.91, + "learning_rate": 3.3791158084163245e-06, + "loss": 0.0042, + "step": 45097 + }, + { + "epoch": 5.91, + "learning_rate": 3.3787183784776956e-06, + "loss": 0.0057, + "step": 45098 + }, + { + "epoch": 5.91, + "learning_rate": 3.3783209671610682e-06, + "loss": 0.0058, + "step": 45099 + }, + { + "epoch": 5.91, + "learning_rate": 3.377923574467561e-06, + "loss": 0.0054, + "step": 45100 + }, + { + "epoch": 5.91, + "learning_rate": 3.3775262003982876e-06, + "loss": 0.0062, + "step": 45101 + }, + { + "epoch": 5.91, + "learning_rate": 3.377128844954374e-06, + "loss": 0.0045, + "step": 45102 + }, + { + "epoch": 5.91, + "learning_rate": 3.376731508136931e-06, + "loss": 0.0052, + "step": 45103 + }, + { + "epoch": 5.91, + "learning_rate": 3.3763341899470747e-06, + "loss": 0.0072, + "step": 45104 + }, + { + "epoch": 5.91, + "learning_rate": 3.3759368903859268e-06, + "loss": 0.0038, + "step": 45105 + }, + { + "epoch": 5.91, + "learning_rate": 3.3755396094546057e-06, + "loss": 0.0063, + "step": 45106 + }, + { + "epoch": 5.91, + "learning_rate": 3.375142347154228e-06, + "loss": 0.0093, + "step": 45107 + }, + { + "epoch": 5.91, + "learning_rate": 3.374745103485906e-06, + "loss": 0.0058, + "step": 45108 + }, + { + "epoch": 5.91, + "learning_rate": 3.3743478784507633e-06, + "loss": 0.0084, + "step": 45109 + }, + { + "epoch": 5.91, + "learning_rate": 3.373950672049914e-06, + "loss": 0.0053, + "step": 45110 + }, + { + "epoch": 5.91, + "learning_rate": 3.3735534842844765e-06, + "loss": 0.0054, + "step": 45111 + }, + { + "epoch": 5.91, + "learning_rate": 3.3731563151555637e-06, + "loss": 0.0057, + "step": 45112 + }, + { + "epoch": 5.91, + "learning_rate": 3.372759164664299e-06, + "loss": 0.0069, + "step": 45113 + }, + { + "epoch": 5.91, + "learning_rate": 3.3723620328117955e-06, + "loss": 0.0051, + "step": 45114 + }, + { + "epoch": 5.91, + "learning_rate": 3.371964919599171e-06, + "loss": 0.0046, + "step": 45115 + }, + { + "epoch": 5.91, + "learning_rate": 3.3715678250275363e-06, + "loss": 0.004, + "step": 45116 + }, + { + "epoch": 5.91, + "learning_rate": 3.3711707490980204e-06, + "loss": 0.0048, + "step": 45117 + }, + { + "epoch": 5.91, + "learning_rate": 3.370773691811734e-06, + "loss": 0.006, + "step": 45118 + }, + { + "epoch": 5.91, + "learning_rate": 3.3703766531697935e-06, + "loss": 0.0052, + "step": 45119 + }, + { + "epoch": 5.91, + "learning_rate": 3.3699796331733115e-06, + "loss": 0.0064, + "step": 45120 + }, + { + "epoch": 5.91, + "learning_rate": 3.3695826318234127e-06, + "loss": 0.0058, + "step": 45121 + }, + { + "epoch": 5.91, + "learning_rate": 3.3691856491212094e-06, + "loss": 0.0054, + "step": 45122 + }, + { + "epoch": 5.91, + "learning_rate": 3.3687886850678153e-06, + "loss": 0.0052, + "step": 45123 + }, + { + "epoch": 5.91, + "learning_rate": 3.368391739664353e-06, + "loss": 0.0051, + "step": 45124 + }, + { + "epoch": 5.91, + "learning_rate": 3.3679948129119364e-06, + "loss": 0.0054, + "step": 45125 + }, + { + "epoch": 5.91, + "learning_rate": 3.36759790481168e-06, + "loss": 0.005, + "step": 45126 + }, + { + "epoch": 5.91, + "learning_rate": 3.3672010153646995e-06, + "loss": 0.0075, + "step": 45127 + }, + { + "epoch": 5.91, + "learning_rate": 3.366804144572112e-06, + "loss": 0.0063, + "step": 45128 + }, + { + "epoch": 5.91, + "learning_rate": 3.3664072924350377e-06, + "loss": 0.0056, + "step": 45129 + }, + { + "epoch": 5.91, + "learning_rate": 3.3660104589545905e-06, + "loss": 0.004, + "step": 45130 + }, + { + "epoch": 5.91, + "learning_rate": 3.365613644131881e-06, + "loss": 0.0041, + "step": 45131 + }, + { + "epoch": 5.91, + "learning_rate": 3.3652168479680348e-06, + "loss": 0.0055, + "step": 45132 + }, + { + "epoch": 5.91, + "learning_rate": 3.364820070464162e-06, + "loss": 0.0049, + "step": 45133 + }, + { + "epoch": 5.91, + "learning_rate": 3.3644233116213786e-06, + "loss": 0.0055, + "step": 45134 + }, + { + "epoch": 5.91, + "learning_rate": 3.364026571440799e-06, + "loss": 0.0085, + "step": 45135 + }, + { + "epoch": 5.91, + "learning_rate": 3.363629849923544e-06, + "loss": 0.0065, + "step": 45136 + }, + { + "epoch": 5.91, + "learning_rate": 3.363233147070728e-06, + "loss": 0.005, + "step": 45137 + }, + { + "epoch": 5.91, + "learning_rate": 3.362836462883461e-06, + "loss": 0.0063, + "step": 45138 + }, + { + "epoch": 5.91, + "learning_rate": 3.362439797362863e-06, + "loss": 0.0063, + "step": 45139 + }, + { + "epoch": 5.91, + "learning_rate": 3.3620431505100525e-06, + "loss": 0.0069, + "step": 45140 + }, + { + "epoch": 5.91, + "learning_rate": 3.361646522326143e-06, + "loss": 0.0066, + "step": 45141 + }, + { + "epoch": 5.91, + "learning_rate": 3.361249912812249e-06, + "loss": 0.0068, + "step": 45142 + }, + { + "epoch": 5.91, + "learning_rate": 3.3608533219694816e-06, + "loss": 0.0067, + "step": 45143 + }, + { + "epoch": 5.91, + "learning_rate": 3.360456749798965e-06, + "loss": 0.0064, + "step": 45144 + }, + { + "epoch": 5.91, + "learning_rate": 3.36006019630181e-06, + "loss": 0.0037, + "step": 45145 + }, + { + "epoch": 5.91, + "learning_rate": 3.359663661479129e-06, + "loss": 0.0065, + "step": 45146 + }, + { + "epoch": 5.91, + "learning_rate": 3.3592671453320436e-06, + "loss": 0.0069, + "step": 45147 + }, + { + "epoch": 5.91, + "learning_rate": 3.3588706478616664e-06, + "loss": 0.0054, + "step": 45148 + }, + { + "epoch": 5.91, + "learning_rate": 3.3584741690691068e-06, + "loss": 0.0042, + "step": 45149 + }, + { + "epoch": 5.91, + "learning_rate": 3.3580777089554893e-06, + "loss": 0.0053, + "step": 45150 + }, + { + "epoch": 5.91, + "learning_rate": 3.3576812675219218e-06, + "loss": 0.0063, + "step": 45151 + }, + { + "epoch": 5.91, + "learning_rate": 3.357284844769524e-06, + "loss": 0.0047, + "step": 45152 + }, + { + "epoch": 5.91, + "learning_rate": 3.3568884406994095e-06, + "loss": 0.007, + "step": 45153 + }, + { + "epoch": 5.91, + "learning_rate": 3.356492055312689e-06, + "loss": 0.0036, + "step": 45154 + }, + { + "epoch": 5.91, + "learning_rate": 3.3560956886104846e-06, + "loss": 0.0067, + "step": 45155 + }, + { + "epoch": 5.92, + "learning_rate": 3.3556993405939074e-06, + "loss": 0.0049, + "step": 45156 + }, + { + "epoch": 5.92, + "learning_rate": 3.3553030112640717e-06, + "loss": 0.0067, + "step": 45157 + }, + { + "epoch": 5.92, + "learning_rate": 3.354906700622088e-06, + "loss": 0.0063, + "step": 45158 + }, + { + "epoch": 5.92, + "learning_rate": 3.3545104086690805e-06, + "loss": 0.006, + "step": 45159 + }, + { + "epoch": 5.92, + "learning_rate": 3.3541141354061535e-06, + "loss": 0.0045, + "step": 45160 + }, + { + "epoch": 5.92, + "learning_rate": 3.353717880834432e-06, + "loss": 0.005, + "step": 45161 + }, + { + "epoch": 5.92, + "learning_rate": 3.35332164495502e-06, + "loss": 0.0057, + "step": 45162 + }, + { + "epoch": 5.92, + "learning_rate": 3.3529254277690404e-06, + "loss": 0.0055, + "step": 45163 + }, + { + "epoch": 5.92, + "learning_rate": 3.3525292292776046e-06, + "loss": 0.0053, + "step": 45164 + }, + { + "epoch": 5.92, + "learning_rate": 3.352133049481826e-06, + "loss": 0.0063, + "step": 45165 + }, + { + "epoch": 5.92, + "learning_rate": 3.351736888382816e-06, + "loss": 0.0044, + "step": 45166 + }, + { + "epoch": 5.92, + "learning_rate": 3.351340745981695e-06, + "loss": 0.0069, + "step": 45167 + }, + { + "epoch": 5.92, + "learning_rate": 3.350944622279575e-06, + "loss": 0.0037, + "step": 45168 + }, + { + "epoch": 5.92, + "learning_rate": 3.3505485172775643e-06, + "loss": 0.0054, + "step": 45169 + }, + { + "epoch": 5.92, + "learning_rate": 3.3501524309767863e-06, + "loss": 0.0052, + "step": 45170 + }, + { + "epoch": 5.92, + "learning_rate": 3.349756363378347e-06, + "loss": 0.006, + "step": 45171 + }, + { + "epoch": 5.92, + "learning_rate": 3.3493603144833664e-06, + "loss": 0.0063, + "step": 45172 + }, + { + "epoch": 5.92, + "learning_rate": 3.3489642842929536e-06, + "loss": 0.0036, + "step": 45173 + }, + { + "epoch": 5.92, + "learning_rate": 3.348568272808227e-06, + "loss": 0.0042, + "step": 45174 + }, + { + "epoch": 5.92, + "learning_rate": 3.3481722800302986e-06, + "loss": 0.0061, + "step": 45175 + }, + { + "epoch": 5.92, + "learning_rate": 3.3477763059602807e-06, + "loss": 0.0059, + "step": 45176 + }, + { + "epoch": 5.92, + "learning_rate": 3.3473803505992843e-06, + "loss": 0.0061, + "step": 45177 + }, + { + "epoch": 5.92, + "learning_rate": 3.346984413948431e-06, + "loss": 0.005, + "step": 45178 + }, + { + "epoch": 5.92, + "learning_rate": 3.346588496008828e-06, + "loss": 0.0079, + "step": 45179 + }, + { + "epoch": 5.92, + "learning_rate": 3.3461925967815913e-06, + "loss": 0.0066, + "step": 45180 + }, + { + "epoch": 5.92, + "learning_rate": 3.3457967162678307e-06, + "loss": 0.0046, + "step": 45181 + }, + { + "epoch": 5.92, + "learning_rate": 3.345400854468662e-06, + "loss": 0.0047, + "step": 45182 + }, + { + "epoch": 5.92, + "learning_rate": 3.345005011385203e-06, + "loss": 0.0073, + "step": 45183 + }, + { + "epoch": 5.92, + "learning_rate": 3.3446091870185616e-06, + "loss": 0.0061, + "step": 45184 + }, + { + "epoch": 5.92, + "learning_rate": 3.344213381369851e-06, + "loss": 0.0063, + "step": 45185 + }, + { + "epoch": 5.92, + "learning_rate": 3.3438175944401875e-06, + "loss": 0.0079, + "step": 45186 + }, + { + "epoch": 5.92, + "learning_rate": 3.343421826230683e-06, + "loss": 0.0076, + "step": 45187 + }, + { + "epoch": 5.92, + "learning_rate": 3.343026076742447e-06, + "loss": 0.0061, + "step": 45188 + }, + { + "epoch": 5.92, + "learning_rate": 3.3426303459765996e-06, + "loss": 0.005, + "step": 45189 + }, + { + "epoch": 5.92, + "learning_rate": 3.3422346339342482e-06, + "loss": 0.0057, + "step": 45190 + }, + { + "epoch": 5.92, + "learning_rate": 3.3418389406165084e-06, + "loss": 0.0095, + "step": 45191 + }, + { + "epoch": 5.92, + "learning_rate": 3.341443266024489e-06, + "loss": 0.0057, + "step": 45192 + }, + { + "epoch": 5.92, + "learning_rate": 3.3410476101593057e-06, + "loss": 0.0056, + "step": 45193 + }, + { + "epoch": 5.92, + "learning_rate": 3.3406519730220743e-06, + "loss": 0.0055, + "step": 45194 + }, + { + "epoch": 5.92, + "learning_rate": 3.3402563546139044e-06, + "loss": 0.0058, + "step": 45195 + }, + { + "epoch": 5.92, + "learning_rate": 3.3398607549359063e-06, + "loss": 0.0051, + "step": 45196 + }, + { + "epoch": 5.92, + "learning_rate": 3.3394651739891983e-06, + "loss": 0.0062, + "step": 45197 + }, + { + "epoch": 5.92, + "learning_rate": 3.339069611774889e-06, + "loss": 0.006, + "step": 45198 + }, + { + "epoch": 5.92, + "learning_rate": 3.3386740682940922e-06, + "loss": 0.0038, + "step": 45199 + }, + { + "epoch": 5.92, + "learning_rate": 3.338278543547917e-06, + "loss": 0.0048, + "step": 45200 + }, + { + "epoch": 5.92, + "learning_rate": 3.3378830375374814e-06, + "loss": 0.0067, + "step": 45201 + }, + { + "epoch": 5.92, + "learning_rate": 3.337487550263895e-06, + "loss": 0.0071, + "step": 45202 + }, + { + "epoch": 5.92, + "learning_rate": 3.3370920817282672e-06, + "loss": 0.0043, + "step": 45203 + }, + { + "epoch": 5.92, + "learning_rate": 3.336696631931713e-06, + "loss": 0.0041, + "step": 45204 + }, + { + "epoch": 5.92, + "learning_rate": 3.336301200875347e-06, + "loss": 0.0069, + "step": 45205 + }, + { + "epoch": 5.92, + "learning_rate": 3.3359057885602796e-06, + "loss": 0.0056, + "step": 45206 + }, + { + "epoch": 5.92, + "learning_rate": 3.3355103949876234e-06, + "loss": 0.0047, + "step": 45207 + }, + { + "epoch": 5.92, + "learning_rate": 3.335115020158485e-06, + "loss": 0.0067, + "step": 45208 + }, + { + "epoch": 5.92, + "learning_rate": 3.334719664073983e-06, + "loss": 0.0049, + "step": 45209 + }, + { + "epoch": 5.92, + "learning_rate": 3.3343243267352288e-06, + "loss": 0.008, + "step": 45210 + }, + { + "epoch": 5.92, + "learning_rate": 3.3339290081433283e-06, + "loss": 0.0061, + "step": 45211 + }, + { + "epoch": 5.92, + "learning_rate": 3.3335337082994014e-06, + "loss": 0.0076, + "step": 45212 + }, + { + "epoch": 5.92, + "learning_rate": 3.333138427204555e-06, + "loss": 0.0042, + "step": 45213 + }, + { + "epoch": 5.92, + "learning_rate": 3.3327431648599017e-06, + "loss": 0.0049, + "step": 45214 + }, + { + "epoch": 5.92, + "learning_rate": 3.332347921266551e-06, + "loss": 0.0054, + "step": 45215 + }, + { + "epoch": 5.92, + "learning_rate": 3.3319526964256154e-06, + "loss": 0.0072, + "step": 45216 + }, + { + "epoch": 5.92, + "learning_rate": 3.331557490338212e-06, + "loss": 0.006, + "step": 45217 + }, + { + "epoch": 5.92, + "learning_rate": 3.3311623030054474e-06, + "loss": 0.008, + "step": 45218 + }, + { + "epoch": 5.92, + "learning_rate": 3.330767134428431e-06, + "loss": 0.0055, + "step": 45219 + }, + { + "epoch": 5.92, + "learning_rate": 3.3303719846082805e-06, + "loss": 0.0059, + "step": 45220 + }, + { + "epoch": 5.92, + "learning_rate": 3.329976853546103e-06, + "loss": 0.0054, + "step": 45221 + }, + { + "epoch": 5.92, + "learning_rate": 3.32958174124301e-06, + "loss": 0.0047, + "step": 45222 + }, + { + "epoch": 5.92, + "learning_rate": 3.32918664770011e-06, + "loss": 0.0074, + "step": 45223 + }, + { + "epoch": 5.92, + "learning_rate": 3.328791572918522e-06, + "loss": 0.0063, + "step": 45224 + }, + { + "epoch": 5.92, + "learning_rate": 3.32839651689935e-06, + "loss": 0.0067, + "step": 45225 + }, + { + "epoch": 5.92, + "learning_rate": 3.3280014796437053e-06, + "loss": 0.006, + "step": 45226 + }, + { + "epoch": 5.92, + "learning_rate": 3.3276064611527016e-06, + "loss": 0.0059, + "step": 45227 + }, + { + "epoch": 5.92, + "learning_rate": 3.327211461427453e-06, + "loss": 0.0067, + "step": 45228 + }, + { + "epoch": 5.92, + "learning_rate": 3.3268164804690662e-06, + "loss": 0.0047, + "step": 45229 + }, + { + "epoch": 5.92, + "learning_rate": 3.326421518278653e-06, + "loss": 0.005, + "step": 45230 + }, + { + "epoch": 5.92, + "learning_rate": 3.3260265748573194e-06, + "loss": 0.0065, + "step": 45231 + }, + { + "epoch": 5.93, + "learning_rate": 3.325631650206186e-06, + "loss": 0.0067, + "step": 45232 + }, + { + "epoch": 5.93, + "learning_rate": 3.3252367443263565e-06, + "loss": 0.0055, + "step": 45233 + }, + { + "epoch": 5.93, + "learning_rate": 3.32484185721894e-06, + "loss": 0.0043, + "step": 45234 + }, + { + "epoch": 5.93, + "learning_rate": 3.324446988885055e-06, + "loss": 0.0057, + "step": 45235 + }, + { + "epoch": 5.93, + "learning_rate": 3.324052139325803e-06, + "loss": 0.0064, + "step": 45236 + }, + { + "epoch": 5.93, + "learning_rate": 3.323657308542302e-06, + "loss": 0.0046, + "step": 45237 + }, + { + "epoch": 5.93, + "learning_rate": 3.3232624965356564e-06, + "loss": 0.0079, + "step": 45238 + }, + { + "epoch": 5.93, + "learning_rate": 3.3228677033069823e-06, + "loss": 0.0054, + "step": 45239 + }, + { + "epoch": 5.93, + "learning_rate": 3.3224729288573875e-06, + "loss": 0.0062, + "step": 45240 + }, + { + "epoch": 5.93, + "learning_rate": 3.3220781731879814e-06, + "loss": 0.0065, + "step": 45241 + }, + { + "epoch": 5.93, + "learning_rate": 3.321683436299873e-06, + "loss": 0.0055, + "step": 45242 + }, + { + "epoch": 5.93, + "learning_rate": 3.3212887181941766e-06, + "loss": 0.0065, + "step": 45243 + }, + { + "epoch": 5.93, + "learning_rate": 3.3208940188719996e-06, + "loss": 0.0054, + "step": 45244 + }, + { + "epoch": 5.93, + "learning_rate": 3.3204993383344532e-06, + "loss": 0.0049, + "step": 45245 + }, + { + "epoch": 5.93, + "learning_rate": 3.320104676582643e-06, + "loss": 0.0046, + "step": 45246 + }, + { + "epoch": 5.93, + "learning_rate": 3.3197100336176836e-06, + "loss": 0.0064, + "step": 45247 + }, + { + "epoch": 5.93, + "learning_rate": 3.3193154094406867e-06, + "loss": 0.0064, + "step": 45248 + }, + { + "epoch": 5.93, + "learning_rate": 3.318920804052759e-06, + "loss": 0.0073, + "step": 45249 + }, + { + "epoch": 5.93, + "learning_rate": 3.318526217455007e-06, + "loss": 0.0055, + "step": 45250 + }, + { + "epoch": 5.93, + "learning_rate": 3.318131649648548e-06, + "loss": 0.0047, + "step": 45251 + }, + { + "epoch": 5.93, + "learning_rate": 3.3177371006344894e-06, + "loss": 0.0059, + "step": 45252 + }, + { + "epoch": 5.93, + "learning_rate": 3.317342570413934e-06, + "loss": 0.007, + "step": 45253 + }, + { + "epoch": 5.93, + "learning_rate": 3.316948058988001e-06, + "loss": 0.0038, + "step": 45254 + }, + { + "epoch": 5.93, + "learning_rate": 3.3165535663577965e-06, + "loss": 0.0062, + "step": 45255 + }, + { + "epoch": 5.93, + "learning_rate": 3.316159092524428e-06, + "loss": 0.0067, + "step": 45256 + }, + { + "epoch": 5.93, + "learning_rate": 3.3157646374890027e-06, + "loss": 0.0044, + "step": 45257 + }, + { + "epoch": 5.93, + "learning_rate": 3.3153702012526334e-06, + "loss": 0.0057, + "step": 45258 + }, + { + "epoch": 5.93, + "learning_rate": 3.314975783816433e-06, + "loss": 0.0058, + "step": 45259 + }, + { + "epoch": 5.93, + "learning_rate": 3.314581385181508e-06, + "loss": 0.0046, + "step": 45260 + }, + { + "epoch": 5.93, + "learning_rate": 3.314187005348962e-06, + "loss": 0.0055, + "step": 45261 + }, + { + "epoch": 5.93, + "learning_rate": 3.313792644319914e-06, + "loss": 0.0052, + "step": 45262 + }, + { + "epoch": 5.93, + "learning_rate": 3.3133983020954673e-06, + "loss": 0.0069, + "step": 45263 + }, + { + "epoch": 5.93, + "learning_rate": 3.313003978676731e-06, + "loss": 0.0052, + "step": 45264 + }, + { + "epoch": 5.93, + "learning_rate": 3.312609674064813e-06, + "loss": 0.0077, + "step": 45265 + }, + { + "epoch": 5.93, + "learning_rate": 3.3122153882608266e-06, + "loss": 0.0061, + "step": 45266 + }, + { + "epoch": 5.93, + "learning_rate": 3.311821121265879e-06, + "loss": 0.0068, + "step": 45267 + }, + { + "epoch": 5.93, + "learning_rate": 3.311426873081074e-06, + "loss": 0.0045, + "step": 45268 + }, + { + "epoch": 5.93, + "learning_rate": 3.311032643707526e-06, + "loss": 0.0065, + "step": 45269 + }, + { + "epoch": 5.93, + "learning_rate": 3.310638433146345e-06, + "loss": 0.0084, + "step": 45270 + }, + { + "epoch": 5.93, + "learning_rate": 3.310244241398637e-06, + "loss": 0.0063, + "step": 45271 + }, + { + "epoch": 5.93, + "learning_rate": 3.309850068465511e-06, + "loss": 0.0036, + "step": 45272 + }, + { + "epoch": 5.93, + "learning_rate": 3.309455914348072e-06, + "loss": 0.0043, + "step": 45273 + }, + { + "epoch": 5.93, + "learning_rate": 3.3090617790474357e-06, + "loss": 0.007, + "step": 45274 + }, + { + "epoch": 5.93, + "learning_rate": 3.3086676625647063e-06, + "loss": 0.0046, + "step": 45275 + }, + { + "epoch": 5.93, + "learning_rate": 3.30827356490099e-06, + "loss": 0.0061, + "step": 45276 + }, + { + "epoch": 5.93, + "learning_rate": 3.3078794860574005e-06, + "loss": 0.0058, + "step": 45277 + }, + { + "epoch": 5.93, + "learning_rate": 3.3074854260350443e-06, + "loss": 0.0049, + "step": 45278 + }, + { + "epoch": 5.93, + "learning_rate": 3.3070913848350285e-06, + "loss": 0.0058, + "step": 45279 + }, + { + "epoch": 5.93, + "learning_rate": 3.3066973624584587e-06, + "loss": 0.0092, + "step": 45280 + }, + { + "epoch": 5.93, + "learning_rate": 3.3063033589064465e-06, + "loss": 0.0051, + "step": 45281 + }, + { + "epoch": 5.93, + "learning_rate": 3.305909374180102e-06, + "loss": 0.007, + "step": 45282 + }, + { + "epoch": 5.93, + "learning_rate": 3.3055154082805317e-06, + "loss": 0.0074, + "step": 45283 + }, + { + "epoch": 5.93, + "learning_rate": 3.305121461208839e-06, + "loss": 0.006, + "step": 45284 + }, + { + "epoch": 5.93, + "learning_rate": 3.3047275329661387e-06, + "loss": 0.0069, + "step": 45285 + }, + { + "epoch": 5.93, + "learning_rate": 3.304333623553536e-06, + "loss": 0.0047, + "step": 45286 + }, + { + "epoch": 5.93, + "learning_rate": 3.303939732972139e-06, + "loss": 0.0043, + "step": 45287 + }, + { + "epoch": 5.93, + "learning_rate": 3.303545861223051e-06, + "loss": 0.0056, + "step": 45288 + }, + { + "epoch": 5.93, + "learning_rate": 3.3031520083073874e-06, + "loss": 0.0056, + "step": 45289 + }, + { + "epoch": 5.93, + "learning_rate": 3.3027581742262515e-06, + "loss": 0.0057, + "step": 45290 + }, + { + "epoch": 5.93, + "learning_rate": 3.3023643589807486e-06, + "loss": 0.0074, + "step": 45291 + }, + { + "epoch": 5.93, + "learning_rate": 3.30197056257199e-06, + "loss": 0.004, + "step": 45292 + }, + { + "epoch": 5.93, + "learning_rate": 3.301576785001085e-06, + "loss": 0.0043, + "step": 45293 + }, + { + "epoch": 5.93, + "learning_rate": 3.301183026269139e-06, + "loss": 0.0057, + "step": 45294 + }, + { + "epoch": 5.93, + "learning_rate": 3.3007892863772584e-06, + "loss": 0.0042, + "step": 45295 + }, + { + "epoch": 5.93, + "learning_rate": 3.3003955653265487e-06, + "loss": 0.006, + "step": 45296 + }, + { + "epoch": 5.93, + "learning_rate": 3.300001863118122e-06, + "loss": 0.0051, + "step": 45297 + }, + { + "epoch": 5.93, + "learning_rate": 3.299608179753084e-06, + "loss": 0.0081, + "step": 45298 + }, + { + "epoch": 5.93, + "learning_rate": 3.299214515232537e-06, + "loss": 0.0048, + "step": 45299 + }, + { + "epoch": 5.93, + "learning_rate": 3.298820869557596e-06, + "loss": 0.0034, + "step": 45300 + }, + { + "epoch": 5.93, + "learning_rate": 3.2984272427293652e-06, + "loss": 0.0077, + "step": 45301 + }, + { + "epoch": 5.93, + "learning_rate": 3.2980336347489472e-06, + "loss": 0.0059, + "step": 45302 + }, + { + "epoch": 5.93, + "learning_rate": 3.297640045617453e-06, + "loss": 0.0072, + "step": 45303 + }, + { + "epoch": 5.93, + "learning_rate": 3.2972464753359922e-06, + "loss": 0.0054, + "step": 45304 + }, + { + "epoch": 5.93, + "learning_rate": 3.296852923905669e-06, + "loss": 0.0047, + "step": 45305 + }, + { + "epoch": 5.93, + "learning_rate": 3.296459391327589e-06, + "loss": 0.0045, + "step": 45306 + }, + { + "epoch": 5.93, + "learning_rate": 3.2960658776028574e-06, + "loss": 0.0046, + "step": 45307 + }, + { + "epoch": 5.94, + "learning_rate": 3.2956723827325874e-06, + "loss": 0.0043, + "step": 45308 + }, + { + "epoch": 5.94, + "learning_rate": 3.2952789067178816e-06, + "loss": 0.0066, + "step": 45309 + }, + { + "epoch": 5.94, + "learning_rate": 3.2948854495598427e-06, + "loss": 0.0058, + "step": 45310 + }, + { + "epoch": 5.94, + "learning_rate": 3.2944920112595846e-06, + "loss": 0.0079, + "step": 45311 + }, + { + "epoch": 5.94, + "learning_rate": 3.294098591818211e-06, + "loss": 0.0043, + "step": 45312 + }, + { + "epoch": 5.94, + "learning_rate": 3.2937051912368255e-06, + "loss": 0.0058, + "step": 45313 + }, + { + "epoch": 5.94, + "learning_rate": 3.2933118095165396e-06, + "loss": 0.0045, + "step": 45314 + }, + { + "epoch": 5.94, + "learning_rate": 3.2929184466584542e-06, + "loss": 0.0049, + "step": 45315 + }, + { + "epoch": 5.94, + "learning_rate": 3.292525102663681e-06, + "loss": 0.0065, + "step": 45316 + }, + { + "epoch": 5.94, + "learning_rate": 3.292131777533324e-06, + "loss": 0.0042, + "step": 45317 + }, + { + "epoch": 5.94, + "learning_rate": 3.2917384712684862e-06, + "loss": 0.0058, + "step": 45318 + }, + { + "epoch": 5.94, + "learning_rate": 3.2913451838702804e-06, + "loss": 0.0062, + "step": 45319 + }, + { + "epoch": 5.94, + "learning_rate": 3.2909519153398085e-06, + "loss": 0.0049, + "step": 45320 + }, + { + "epoch": 5.94, + "learning_rate": 3.2905586656781775e-06, + "loss": 0.0051, + "step": 45321 + }, + { + "epoch": 5.94, + "learning_rate": 3.2901654348864886e-06, + "loss": 0.0045, + "step": 45322 + }, + { + "epoch": 5.94, + "learning_rate": 3.2897722229658556e-06, + "loss": 0.0067, + "step": 45323 + }, + { + "epoch": 5.94, + "learning_rate": 3.289379029917379e-06, + "loss": 0.0048, + "step": 45324 + }, + { + "epoch": 5.94, + "learning_rate": 3.2889858557421684e-06, + "loss": 0.0074, + "step": 45325 + }, + { + "epoch": 5.94, + "learning_rate": 3.2885927004413244e-06, + "loss": 0.0058, + "step": 45326 + }, + { + "epoch": 5.94, + "learning_rate": 3.2881995640159605e-06, + "loss": 0.0047, + "step": 45327 + }, + { + "epoch": 5.94, + "learning_rate": 3.2878064464671768e-06, + "loss": 0.0061, + "step": 45328 + }, + { + "epoch": 5.94, + "learning_rate": 3.28741334779608e-06, + "loss": 0.006, + "step": 45329 + }, + { + "epoch": 5.94, + "learning_rate": 3.2870202680037734e-06, + "loss": 0.0073, + "step": 45330 + }, + { + "epoch": 5.94, + "learning_rate": 3.2866272070913675e-06, + "loss": 0.0038, + "step": 45331 + }, + { + "epoch": 5.94, + "learning_rate": 3.286234165059965e-06, + "loss": 0.0068, + "step": 45332 + }, + { + "epoch": 5.94, + "learning_rate": 3.285841141910667e-06, + "loss": 0.0062, + "step": 45333 + }, + { + "epoch": 5.94, + "learning_rate": 3.2854481376445878e-06, + "loss": 0.0048, + "step": 45334 + }, + { + "epoch": 5.94, + "learning_rate": 3.285055152262825e-06, + "loss": 0.0056, + "step": 45335 + }, + { + "epoch": 5.94, + "learning_rate": 3.2846621857664907e-06, + "loss": 0.0063, + "step": 45336 + }, + { + "epoch": 5.94, + "learning_rate": 3.2842692381566855e-06, + "loss": 0.008, + "step": 45337 + }, + { + "epoch": 5.94, + "learning_rate": 3.2838763094345118e-06, + "loss": 0.0056, + "step": 45338 + }, + { + "epoch": 5.94, + "learning_rate": 3.283483399601083e-06, + "loss": 0.0055, + "step": 45339 + }, + { + "epoch": 5.94, + "learning_rate": 3.283090508657498e-06, + "loss": 0.0041, + "step": 45340 + }, + { + "epoch": 5.94, + "learning_rate": 3.28269763660486e-06, + "loss": 0.0058, + "step": 45341 + }, + { + "epoch": 5.94, + "learning_rate": 3.282304783444281e-06, + "loss": 0.0045, + "step": 45342 + }, + { + "epoch": 5.94, + "learning_rate": 3.2819119491768625e-06, + "loss": 0.0064, + "step": 45343 + }, + { + "epoch": 5.94, + "learning_rate": 3.281519133803709e-06, + "loss": 0.0068, + "step": 45344 + }, + { + "epoch": 5.94, + "learning_rate": 3.2811263373259204e-06, + "loss": 0.0078, + "step": 45345 + }, + { + "epoch": 5.94, + "learning_rate": 3.2807335597446077e-06, + "loss": 0.0056, + "step": 45346 + }, + { + "epoch": 5.94, + "learning_rate": 3.2803408010608774e-06, + "loss": 0.0052, + "step": 45347 + }, + { + "epoch": 5.94, + "learning_rate": 3.2799480612758294e-06, + "loss": 0.007, + "step": 45348 + }, + { + "epoch": 5.94, + "learning_rate": 3.2795553403905666e-06, + "loss": 0.0038, + "step": 45349 + }, + { + "epoch": 5.94, + "learning_rate": 3.2791626384062002e-06, + "loss": 0.005, + "step": 45350 + }, + { + "epoch": 5.94, + "learning_rate": 3.2787699553238315e-06, + "loss": 0.0048, + "step": 45351 + }, + { + "epoch": 5.94, + "learning_rate": 3.2783772911445645e-06, + "loss": 0.0077, + "step": 45352 + }, + { + "epoch": 5.94, + "learning_rate": 3.2779846458694987e-06, + "loss": 0.0039, + "step": 45353 + }, + { + "epoch": 5.94, + "learning_rate": 3.277592019499748e-06, + "loss": 0.0033, + "step": 45354 + }, + { + "epoch": 5.94, + "learning_rate": 3.2771994120364103e-06, + "loss": 0.0045, + "step": 45355 + }, + { + "epoch": 5.94, + "learning_rate": 3.2768068234805895e-06, + "loss": 0.0066, + "step": 45356 + }, + { + "epoch": 5.94, + "learning_rate": 3.27641425383339e-06, + "loss": 0.0056, + "step": 45357 + }, + { + "epoch": 5.94, + "learning_rate": 3.2760217030959207e-06, + "loss": 0.005, + "step": 45358 + }, + { + "epoch": 5.94, + "learning_rate": 3.2756291712692833e-06, + "loss": 0.0048, + "step": 45359 + }, + { + "epoch": 5.94, + "learning_rate": 3.275236658354577e-06, + "loss": 0.0053, + "step": 45360 + }, + { + "epoch": 5.94, + "learning_rate": 3.2748441643529126e-06, + "loss": 0.006, + "step": 45361 + }, + { + "epoch": 5.94, + "learning_rate": 3.2744516892653912e-06, + "loss": 0.0038, + "step": 45362 + }, + { + "epoch": 5.94, + "learning_rate": 3.2740592330931164e-06, + "loss": 0.0047, + "step": 45363 + }, + { + "epoch": 5.94, + "learning_rate": 3.2736667958371883e-06, + "loss": 0.0077, + "step": 45364 + }, + { + "epoch": 5.94, + "learning_rate": 3.2732743774987175e-06, + "loss": 0.0062, + "step": 45365 + }, + { + "epoch": 5.94, + "learning_rate": 3.2728819780788034e-06, + "loss": 0.0074, + "step": 45366 + }, + { + "epoch": 5.94, + "learning_rate": 3.272489597578552e-06, + "loss": 0.009, + "step": 45367 + }, + { + "epoch": 5.94, + "learning_rate": 3.272097235999058e-06, + "loss": 0.0076, + "step": 45368 + }, + { + "epoch": 5.94, + "learning_rate": 3.2717048933414388e-06, + "loss": 0.0054, + "step": 45369 + }, + { + "epoch": 5.94, + "learning_rate": 3.2713125696067916e-06, + "loss": 0.0061, + "step": 45370 + }, + { + "epoch": 5.94, + "learning_rate": 3.270920264796219e-06, + "loss": 0.0052, + "step": 45371 + }, + { + "epoch": 5.94, + "learning_rate": 3.2705279789108212e-06, + "loss": 0.0038, + "step": 45372 + }, + { + "epoch": 5.94, + "learning_rate": 3.2701357119517096e-06, + "loss": 0.0054, + "step": 45373 + }, + { + "epoch": 5.94, + "learning_rate": 3.269743463919982e-06, + "loss": 0.007, + "step": 45374 + }, + { + "epoch": 5.94, + "learning_rate": 3.2693512348167387e-06, + "loss": 0.0062, + "step": 45375 + }, + { + "epoch": 5.94, + "learning_rate": 3.2689590246430915e-06, + "loss": 0.0061, + "step": 45376 + }, + { + "epoch": 5.94, + "learning_rate": 3.2685668334001373e-06, + "loss": 0.0059, + "step": 45377 + }, + { + "epoch": 5.94, + "learning_rate": 3.2681746610889776e-06, + "loss": 0.0048, + "step": 45378 + }, + { + "epoch": 5.94, + "learning_rate": 3.2677825077107217e-06, + "loss": 0.0068, + "step": 45379 + }, + { + "epoch": 5.94, + "learning_rate": 3.267390373266466e-06, + "loss": 0.0049, + "step": 45380 + }, + { + "epoch": 5.94, + "learning_rate": 3.2669982577573188e-06, + "loss": 0.0042, + "step": 45381 + }, + { + "epoch": 5.94, + "learning_rate": 3.2666061611843804e-06, + "loss": 0.0041, + "step": 45382 + }, + { + "epoch": 5.94, + "learning_rate": 3.266214083548751e-06, + "loss": 0.006, + "step": 45383 + }, + { + "epoch": 5.94, + "learning_rate": 3.2658220248515383e-06, + "loss": 0.0042, + "step": 45384 + }, + { + "epoch": 5.95, + "learning_rate": 3.265429985093843e-06, + "loss": 0.0054, + "step": 45385 + }, + { + "epoch": 5.95, + "learning_rate": 3.2650379642767673e-06, + "loss": 0.0069, + "step": 45386 + }, + { + "epoch": 5.95, + "learning_rate": 3.2646459624014103e-06, + "loss": 0.006, + "step": 45387 + }, + { + "epoch": 5.95, + "learning_rate": 3.264253979468881e-06, + "loss": 0.0051, + "step": 45388 + }, + { + "epoch": 5.95, + "learning_rate": 3.263862015480276e-06, + "loss": 0.0067, + "step": 45389 + }, + { + "epoch": 5.95, + "learning_rate": 3.2634700704367037e-06, + "loss": 0.0046, + "step": 45390 + }, + { + "epoch": 5.95, + "learning_rate": 3.2630781443392588e-06, + "loss": 0.0052, + "step": 45391 + }, + { + "epoch": 5.95, + "learning_rate": 3.2626862371890523e-06, + "loss": 0.0038, + "step": 45392 + }, + { + "epoch": 5.95, + "learning_rate": 3.2622943489871804e-06, + "loss": 0.0073, + "step": 45393 + }, + { + "epoch": 5.95, + "learning_rate": 3.261902479734749e-06, + "loss": 0.0039, + "step": 45394 + }, + { + "epoch": 5.95, + "learning_rate": 3.2615106294328524e-06, + "loss": 0.0058, + "step": 45395 + }, + { + "epoch": 5.95, + "learning_rate": 3.2611187980826033e-06, + "loss": 0.0048, + "step": 45396 + }, + { + "epoch": 5.95, + "learning_rate": 3.260726985685099e-06, + "loss": 0.0066, + "step": 45397 + }, + { + "epoch": 5.95, + "learning_rate": 3.2603351922414363e-06, + "loss": 0.0073, + "step": 45398 + }, + { + "epoch": 5.95, + "learning_rate": 3.2599434177527257e-06, + "loss": 0.0051, + "step": 45399 + }, + { + "epoch": 5.95, + "learning_rate": 3.259551662220063e-06, + "loss": 0.0051, + "step": 45400 + }, + { + "epoch": 5.95, + "learning_rate": 3.259159925644555e-06, + "loss": 0.0078, + "step": 45401 + }, + { + "epoch": 5.95, + "learning_rate": 3.2587682080272998e-06, + "loss": 0.0045, + "step": 45402 + }, + { + "epoch": 5.95, + "learning_rate": 3.2583765093693977e-06, + "loss": 0.0035, + "step": 45403 + }, + { + "epoch": 5.95, + "learning_rate": 3.2579848296719562e-06, + "loss": 0.0057, + "step": 45404 + }, + { + "epoch": 5.95, + "learning_rate": 3.2575931689360728e-06, + "loss": 0.0039, + "step": 45405 + }, + { + "epoch": 5.95, + "learning_rate": 3.2572015271628464e-06, + "loss": 0.0064, + "step": 45406 + }, + { + "epoch": 5.95, + "learning_rate": 3.2568099043533853e-06, + "loss": 0.0056, + "step": 45407 + }, + { + "epoch": 5.95, + "learning_rate": 3.256418300508787e-06, + "loss": 0.0059, + "step": 45408 + }, + { + "epoch": 5.95, + "learning_rate": 3.256026715630153e-06, + "loss": 0.0062, + "step": 45409 + }, + { + "epoch": 5.95, + "learning_rate": 3.2556351497185814e-06, + "loss": 0.0056, + "step": 45410 + }, + { + "epoch": 5.95, + "learning_rate": 3.2552436027751768e-06, + "loss": 0.0072, + "step": 45411 + }, + { + "epoch": 5.95, + "learning_rate": 3.2548520748010437e-06, + "loss": 0.0044, + "step": 45412 + }, + { + "epoch": 5.95, + "learning_rate": 3.2544605657972794e-06, + "loss": 0.0067, + "step": 45413 + }, + { + "epoch": 5.95, + "learning_rate": 3.254069075764983e-06, + "loss": 0.0071, + "step": 45414 + }, + { + "epoch": 5.95, + "learning_rate": 3.2536776047052607e-06, + "loss": 0.0048, + "step": 45415 + }, + { + "epoch": 5.95, + "learning_rate": 3.2532861526192116e-06, + "loss": 0.0057, + "step": 45416 + }, + { + "epoch": 5.95, + "learning_rate": 3.2528947195079362e-06, + "loss": 0.0059, + "step": 45417 + }, + { + "epoch": 5.95, + "learning_rate": 3.252503305372531e-06, + "loss": 0.0056, + "step": 45418 + }, + { + "epoch": 5.95, + "learning_rate": 3.252111910214104e-06, + "loss": 0.005, + "step": 45419 + }, + { + "epoch": 5.95, + "learning_rate": 3.251720534033754e-06, + "loss": 0.0076, + "step": 45420 + }, + { + "epoch": 5.95, + "learning_rate": 3.251329176832577e-06, + "loss": 0.0063, + "step": 45421 + }, + { + "epoch": 5.95, + "learning_rate": 3.2509378386116763e-06, + "loss": 0.0048, + "step": 45422 + }, + { + "epoch": 5.95, + "learning_rate": 3.2505465193721574e-06, + "loss": 0.0037, + "step": 45423 + }, + { + "epoch": 5.95, + "learning_rate": 3.2501552191151174e-06, + "loss": 0.0061, + "step": 45424 + }, + { + "epoch": 5.95, + "learning_rate": 3.2497639378416536e-06, + "loss": 0.0066, + "step": 45425 + }, + { + "epoch": 5.95, + "learning_rate": 3.249372675552871e-06, + "loss": 0.0056, + "step": 45426 + }, + { + "epoch": 5.95, + "learning_rate": 3.2489814322498693e-06, + "loss": 0.0055, + "step": 45427 + }, + { + "epoch": 5.95, + "learning_rate": 3.2485902079337474e-06, + "loss": 0.0046, + "step": 45428 + }, + { + "epoch": 5.95, + "learning_rate": 3.2481990026056032e-06, + "loss": 0.0055, + "step": 45429 + }, + { + "epoch": 5.95, + "learning_rate": 3.2478078162665427e-06, + "loss": 0.0095, + "step": 45430 + }, + { + "epoch": 5.95, + "learning_rate": 3.2474166489176626e-06, + "loss": 0.0049, + "step": 45431 + }, + { + "epoch": 5.95, + "learning_rate": 3.2470255005600647e-06, + "loss": 0.0051, + "step": 45432 + }, + { + "epoch": 5.95, + "learning_rate": 3.246634371194841e-06, + "loss": 0.0069, + "step": 45433 + }, + { + "epoch": 5.95, + "learning_rate": 3.2462432608231044e-06, + "loss": 0.0071, + "step": 45434 + }, + { + "epoch": 5.95, + "learning_rate": 3.24585216944595e-06, + "loss": 0.0082, + "step": 45435 + }, + { + "epoch": 5.95, + "learning_rate": 3.245461097064476e-06, + "loss": 0.0063, + "step": 45436 + }, + { + "epoch": 5.95, + "learning_rate": 3.24507004367978e-06, + "loss": 0.0048, + "step": 45437 + }, + { + "epoch": 5.95, + "learning_rate": 3.244679009292967e-06, + "loss": 0.0081, + "step": 45438 + }, + { + "epoch": 5.95, + "learning_rate": 3.244287993905135e-06, + "loss": 0.0073, + "step": 45439 + }, + { + "epoch": 5.95, + "learning_rate": 3.2438969975173817e-06, + "loss": 0.0057, + "step": 45440 + }, + { + "epoch": 5.95, + "learning_rate": 3.24350602013081e-06, + "loss": 0.005, + "step": 45441 + }, + { + "epoch": 5.95, + "learning_rate": 3.243115061746518e-06, + "loss": 0.0068, + "step": 45442 + }, + { + "epoch": 5.95, + "learning_rate": 3.2427241223656048e-06, + "loss": 0.0072, + "step": 45443 + }, + { + "epoch": 5.95, + "learning_rate": 3.2423332019891675e-06, + "loss": 0.0057, + "step": 45444 + }, + { + "epoch": 5.95, + "learning_rate": 3.241942300618308e-06, + "loss": 0.005, + "step": 45445 + }, + { + "epoch": 5.95, + "learning_rate": 3.2415514182541287e-06, + "loss": 0.0054, + "step": 45446 + }, + { + "epoch": 5.95, + "learning_rate": 3.2411605548977264e-06, + "loss": 0.0104, + "step": 45447 + }, + { + "epoch": 5.95, + "learning_rate": 3.240769710550198e-06, + "loss": 0.0064, + "step": 45448 + }, + { + "epoch": 5.95, + "learning_rate": 3.2403788852126462e-06, + "loss": 0.0052, + "step": 45449 + }, + { + "epoch": 5.95, + "learning_rate": 3.2399880788861706e-06, + "loss": 0.0075, + "step": 45450 + }, + { + "epoch": 5.95, + "learning_rate": 3.239597291571868e-06, + "loss": 0.0089, + "step": 45451 + }, + { + "epoch": 5.95, + "learning_rate": 3.2392065232708336e-06, + "loss": 0.0059, + "step": 45452 + }, + { + "epoch": 5.95, + "learning_rate": 3.2388157739841754e-06, + "loss": 0.0081, + "step": 45453 + }, + { + "epoch": 5.95, + "learning_rate": 3.238425043712987e-06, + "loss": 0.009, + "step": 45454 + }, + { + "epoch": 5.95, + "learning_rate": 3.2380343324583653e-06, + "loss": 0.0061, + "step": 45455 + }, + { + "epoch": 5.95, + "learning_rate": 3.2376436402214106e-06, + "loss": 0.0052, + "step": 45456 + }, + { + "epoch": 5.95, + "learning_rate": 3.237252967003228e-06, + "loss": 0.0077, + "step": 45457 + }, + { + "epoch": 5.95, + "learning_rate": 3.2368623128049104e-06, + "loss": 0.0065, + "step": 45458 + }, + { + "epoch": 5.95, + "learning_rate": 3.2364716776275586e-06, + "loss": 0.0061, + "step": 45459 + }, + { + "epoch": 5.95, + "learning_rate": 3.236081061472265e-06, + "loss": 0.0062, + "step": 45460 + }, + { + "epoch": 5.96, + "learning_rate": 3.235690464340138e-06, + "loss": 0.0056, + "step": 45461 + }, + { + "epoch": 5.96, + "learning_rate": 3.2352998862322704e-06, + "loss": 0.0072, + "step": 45462 + }, + { + "epoch": 5.96, + "learning_rate": 3.234909327149759e-06, + "loss": 0.0064, + "step": 45463 + }, + { + "epoch": 5.96, + "learning_rate": 3.234518787093708e-06, + "loss": 0.0057, + "step": 45464 + }, + { + "epoch": 5.96, + "learning_rate": 3.2341282660652117e-06, + "loss": 0.0047, + "step": 45465 + }, + { + "epoch": 5.96, + "learning_rate": 3.2337377640653666e-06, + "loss": 0.0063, + "step": 45466 + }, + { + "epoch": 5.96, + "learning_rate": 3.233347281095277e-06, + "loss": 0.0071, + "step": 45467 + }, + { + "epoch": 5.96, + "learning_rate": 3.2329568171560355e-06, + "loss": 0.0043, + "step": 45468 + }, + { + "epoch": 5.96, + "learning_rate": 3.2325663722487445e-06, + "loss": 0.0068, + "step": 45469 + }, + { + "epoch": 5.96, + "learning_rate": 3.2321759463745007e-06, + "loss": 0.0068, + "step": 45470 + }, + { + "epoch": 5.96, + "learning_rate": 3.231785539534399e-06, + "loss": 0.0045, + "step": 45471 + }, + { + "epoch": 5.96, + "learning_rate": 3.2313951517295427e-06, + "loss": 0.0042, + "step": 45472 + }, + { + "epoch": 5.96, + "learning_rate": 3.231004782961027e-06, + "loss": 0.0053, + "step": 45473 + }, + { + "epoch": 5.96, + "learning_rate": 3.2306144332299506e-06, + "loss": 0.007, + "step": 45474 + }, + { + "epoch": 5.96, + "learning_rate": 3.230224102537406e-06, + "loss": 0.0057, + "step": 45475 + }, + { + "epoch": 5.96, + "learning_rate": 3.2298337908845e-06, + "loss": 0.0065, + "step": 45476 + }, + { + "epoch": 5.96, + "learning_rate": 3.2294434982723223e-06, + "loss": 0.0055, + "step": 45477 + }, + { + "epoch": 5.96, + "learning_rate": 3.229053224701978e-06, + "loss": 0.0041, + "step": 45478 + }, + { + "epoch": 5.96, + "learning_rate": 3.2286629701745564e-06, + "loss": 0.0068, + "step": 45479 + }, + { + "epoch": 5.96, + "learning_rate": 3.2282727346911637e-06, + "loss": 0.0072, + "step": 45480 + }, + { + "epoch": 5.96, + "learning_rate": 3.2278825182528926e-06, + "loss": 0.0042, + "step": 45481 + }, + { + "epoch": 5.96, + "learning_rate": 3.2274923208608422e-06, + "loss": 0.0036, + "step": 45482 + }, + { + "epoch": 5.96, + "learning_rate": 3.227102142516105e-06, + "loss": 0.0051, + "step": 45483 + }, + { + "epoch": 5.96, + "learning_rate": 3.2267119832197867e-06, + "loss": 0.0049, + "step": 45484 + }, + { + "epoch": 5.96, + "learning_rate": 3.22632184297298e-06, + "loss": 0.0057, + "step": 45485 + }, + { + "epoch": 5.96, + "learning_rate": 3.2259317217767784e-06, + "loss": 0.0057, + "step": 45486 + }, + { + "epoch": 5.96, + "learning_rate": 3.2255416196322866e-06, + "loss": 0.0053, + "step": 45487 + }, + { + "epoch": 5.96, + "learning_rate": 3.225151536540595e-06, + "loss": 0.0053, + "step": 45488 + }, + { + "epoch": 5.96, + "learning_rate": 3.2247614725028077e-06, + "loss": 0.004, + "step": 45489 + }, + { + "epoch": 5.96, + "learning_rate": 3.2243714275200145e-06, + "loss": 0.0046, + "step": 45490 + }, + { + "epoch": 5.96, + "learning_rate": 3.223981401593318e-06, + "loss": 0.0046, + "step": 45491 + }, + { + "epoch": 5.96, + "learning_rate": 3.223591394723814e-06, + "loss": 0.0061, + "step": 45492 + }, + { + "epoch": 5.96, + "learning_rate": 3.2232014069125985e-06, + "loss": 0.0049, + "step": 45493 + }, + { + "epoch": 5.96, + "learning_rate": 3.2228114381607644e-06, + "loss": 0.0057, + "step": 45494 + }, + { + "epoch": 5.96, + "learning_rate": 3.222421488469416e-06, + "loss": 0.0077, + "step": 45495 + }, + { + "epoch": 5.96, + "learning_rate": 3.2220315578396454e-06, + "loss": 0.0056, + "step": 45496 + }, + { + "epoch": 5.96, + "learning_rate": 3.221641646272551e-06, + "loss": 0.009, + "step": 45497 + }, + { + "epoch": 5.96, + "learning_rate": 3.2212517537692247e-06, + "loss": 0.0029, + "step": 45498 + }, + { + "epoch": 5.96, + "learning_rate": 3.220861880330767e-06, + "loss": 0.0054, + "step": 45499 + }, + { + "epoch": 5.96, + "learning_rate": 3.2204720259582767e-06, + "loss": 0.006, + "step": 45500 + }, + { + "epoch": 5.96, + "learning_rate": 3.220082190652849e-06, + "loss": 0.0046, + "step": 45501 + }, + { + "epoch": 5.96, + "learning_rate": 3.219692374415575e-06, + "loss": 0.0078, + "step": 45502 + }, + { + "epoch": 5.96, + "learning_rate": 3.2193025772475595e-06, + "loss": 0.008, + "step": 45503 + }, + { + "epoch": 5.96, + "learning_rate": 3.2189127991498937e-06, + "loss": 0.0052, + "step": 45504 + }, + { + "epoch": 5.96, + "learning_rate": 3.218523040123671e-06, + "loss": 0.0054, + "step": 45505 + }, + { + "epoch": 5.96, + "learning_rate": 3.2181333001699943e-06, + "loss": 0.0048, + "step": 45506 + }, + { + "epoch": 5.96, + "learning_rate": 3.217743579289957e-06, + "loss": 0.0036, + "step": 45507 + }, + { + "epoch": 5.96, + "learning_rate": 3.2173538774846547e-06, + "loss": 0.005, + "step": 45508 + }, + { + "epoch": 5.96, + "learning_rate": 3.21696419475518e-06, + "loss": 0.0062, + "step": 45509 + }, + { + "epoch": 5.96, + "learning_rate": 3.216574531102632e-06, + "loss": 0.0055, + "step": 45510 + }, + { + "epoch": 5.96, + "learning_rate": 3.216184886528111e-06, + "loss": 0.0074, + "step": 45511 + }, + { + "epoch": 5.96, + "learning_rate": 3.215795261032708e-06, + "loss": 0.0056, + "step": 45512 + }, + { + "epoch": 5.96, + "learning_rate": 3.215405654617516e-06, + "loss": 0.0053, + "step": 45513 + }, + { + "epoch": 5.96, + "learning_rate": 3.2150160672836384e-06, + "loss": 0.0065, + "step": 45514 + }, + { + "epoch": 5.96, + "learning_rate": 3.2146264990321663e-06, + "loss": 0.0066, + "step": 45515 + }, + { + "epoch": 5.96, + "learning_rate": 3.2142369498641957e-06, + "loss": 0.0077, + "step": 45516 + }, + { + "epoch": 5.96, + "learning_rate": 3.21384741978082e-06, + "loss": 0.0049, + "step": 45517 + }, + { + "epoch": 5.96, + "learning_rate": 3.2134579087831396e-06, + "loss": 0.0046, + "step": 45518 + }, + { + "epoch": 5.96, + "learning_rate": 3.213068416872247e-06, + "loss": 0.0054, + "step": 45519 + }, + { + "epoch": 5.96, + "learning_rate": 3.2126789440492355e-06, + "loss": 0.005, + "step": 45520 + }, + { + "epoch": 5.96, + "learning_rate": 3.2122894903152025e-06, + "loss": 0.007, + "step": 45521 + }, + { + "epoch": 5.96, + "learning_rate": 3.2119000556712478e-06, + "loss": 0.005, + "step": 45522 + }, + { + "epoch": 5.96, + "learning_rate": 3.2115106401184628e-06, + "loss": 0.0063, + "step": 45523 + }, + { + "epoch": 5.96, + "learning_rate": 3.211121243657941e-06, + "loss": 0.0073, + "step": 45524 + }, + { + "epoch": 5.96, + "learning_rate": 3.2107318662907773e-06, + "loss": 0.0038, + "step": 45525 + }, + { + "epoch": 5.96, + "learning_rate": 3.2103425080180715e-06, + "loss": 0.0064, + "step": 45526 + }, + { + "epoch": 5.96, + "learning_rate": 3.209953168840916e-06, + "loss": 0.006, + "step": 45527 + }, + { + "epoch": 5.96, + "learning_rate": 3.209563848760403e-06, + "loss": 0.003, + "step": 45528 + }, + { + "epoch": 5.96, + "learning_rate": 3.209174547777633e-06, + "loss": 0.0042, + "step": 45529 + }, + { + "epoch": 5.96, + "learning_rate": 3.2087852658936967e-06, + "loss": 0.0051, + "step": 45530 + }, + { + "epoch": 5.96, + "learning_rate": 3.2083960031096883e-06, + "loss": 0.004, + "step": 45531 + }, + { + "epoch": 5.96, + "learning_rate": 3.2080067594267073e-06, + "loss": 0.0043, + "step": 45532 + }, + { + "epoch": 5.96, + "learning_rate": 3.2076175348458415e-06, + "loss": 0.0042, + "step": 45533 + }, + { + "epoch": 5.96, + "learning_rate": 3.207228329368194e-06, + "loss": 0.0055, + "step": 45534 + }, + { + "epoch": 5.96, + "learning_rate": 3.2068391429948553e-06, + "loss": 0.0054, + "step": 45535 + }, + { + "epoch": 5.96, + "learning_rate": 3.2064499757269152e-06, + "loss": 0.0032, + "step": 45536 + }, + { + "epoch": 5.97, + "learning_rate": 3.206060827565477e-06, + "loss": 0.0072, + "step": 45537 + }, + { + "epoch": 5.97, + "learning_rate": 3.205671698511631e-06, + "loss": 0.0071, + "step": 45538 + }, + { + "epoch": 5.97, + "learning_rate": 3.205282588566472e-06, + "loss": 0.0071, + "step": 45539 + }, + { + "epoch": 5.97, + "learning_rate": 3.20489349773109e-06, + "loss": 0.0059, + "step": 45540 + }, + { + "epoch": 5.97, + "learning_rate": 3.2045044260065863e-06, + "loss": 0.0047, + "step": 45541 + }, + { + "epoch": 5.97, + "learning_rate": 3.204115373394048e-06, + "loss": 0.0061, + "step": 45542 + }, + { + "epoch": 5.97, + "learning_rate": 3.203726339894577e-06, + "loss": 0.0043, + "step": 45543 + }, + { + "epoch": 5.97, + "learning_rate": 3.2033373255092613e-06, + "loss": 0.0059, + "step": 45544 + }, + { + "epoch": 5.97, + "learning_rate": 3.2029483302392005e-06, + "loss": 0.0059, + "step": 45545 + }, + { + "epoch": 5.97, + "learning_rate": 3.202559354085485e-06, + "loss": 0.0065, + "step": 45546 + }, + { + "epoch": 5.97, + "learning_rate": 3.20217039704921e-06, + "loss": 0.0054, + "step": 45547 + }, + { + "epoch": 5.97, + "learning_rate": 3.2017814591314644e-06, + "loss": 0.0045, + "step": 45548 + }, + { + "epoch": 5.97, + "learning_rate": 3.201392540333351e-06, + "loss": 0.0065, + "step": 45549 + }, + { + "epoch": 5.97, + "learning_rate": 3.2010036406559584e-06, + "loss": 0.0042, + "step": 45550 + }, + { + "epoch": 5.97, + "learning_rate": 3.2006147601003777e-06, + "loss": 0.0048, + "step": 45551 + }, + { + "epoch": 5.97, + "learning_rate": 3.2002258986677093e-06, + "loss": 0.006, + "step": 45552 + }, + { + "epoch": 5.97, + "learning_rate": 3.19983705635904e-06, + "loss": 0.0078, + "step": 45553 + }, + { + "epoch": 5.97, + "learning_rate": 3.1994482331754707e-06, + "loss": 0.0052, + "step": 45554 + }, + { + "epoch": 5.97, + "learning_rate": 3.1990594291180867e-06, + "loss": 0.0097, + "step": 45555 + }, + { + "epoch": 5.97, + "learning_rate": 3.19867064418799e-06, + "loss": 0.0065, + "step": 45556 + }, + { + "epoch": 5.97, + "learning_rate": 3.19828187838627e-06, + "loss": 0.005, + "step": 45557 + }, + { + "epoch": 5.97, + "learning_rate": 3.1978931317140192e-06, + "loss": 0.0035, + "step": 45558 + }, + { + "epoch": 5.97, + "learning_rate": 3.197504404172329e-06, + "loss": 0.0091, + "step": 45559 + }, + { + "epoch": 5.97, + "learning_rate": 3.1971156957622997e-06, + "loss": 0.0059, + "step": 45560 + }, + { + "epoch": 5.97, + "learning_rate": 3.196727006485019e-06, + "loss": 0.0053, + "step": 45561 + }, + { + "epoch": 5.97, + "learning_rate": 3.1963383363415812e-06, + "loss": 0.0083, + "step": 45562 + }, + { + "epoch": 5.97, + "learning_rate": 3.195949685333076e-06, + "loss": 0.0068, + "step": 45563 + }, + { + "epoch": 5.97, + "learning_rate": 3.195561053460601e-06, + "loss": 0.0054, + "step": 45564 + }, + { + "epoch": 5.97, + "learning_rate": 3.1951724407252503e-06, + "loss": 0.0061, + "step": 45565 + }, + { + "epoch": 5.97, + "learning_rate": 3.1947838471281157e-06, + "loss": 0.0057, + "step": 45566 + }, + { + "epoch": 5.97, + "learning_rate": 3.1943952726702865e-06, + "loss": 0.0073, + "step": 45567 + }, + { + "epoch": 5.97, + "learning_rate": 3.19400671735286e-06, + "loss": 0.0057, + "step": 45568 + }, + { + "epoch": 5.97, + "learning_rate": 3.1936181811769274e-06, + "loss": 0.0058, + "step": 45569 + }, + { + "epoch": 5.97, + "learning_rate": 3.193229664143579e-06, + "loss": 0.0062, + "step": 45570 + }, + { + "epoch": 5.97, + "learning_rate": 3.1928411662539115e-06, + "loss": 0.0052, + "step": 45571 + }, + { + "epoch": 5.97, + "learning_rate": 3.1924526875090168e-06, + "loss": 0.0046, + "step": 45572 + }, + { + "epoch": 5.97, + "learning_rate": 3.1920642279099855e-06, + "loss": 0.0055, + "step": 45573 + }, + { + "epoch": 5.97, + "learning_rate": 3.1916757874579084e-06, + "loss": 0.0052, + "step": 45574 + }, + { + "epoch": 5.97, + "learning_rate": 3.19128736615388e-06, + "loss": 0.007, + "step": 45575 + }, + { + "epoch": 5.97, + "learning_rate": 3.1908989639989974e-06, + "loss": 0.0073, + "step": 45576 + }, + { + "epoch": 5.97, + "learning_rate": 3.190510580994348e-06, + "loss": 0.0048, + "step": 45577 + }, + { + "epoch": 5.97, + "learning_rate": 3.1901222171410227e-06, + "loss": 0.0067, + "step": 45578 + }, + { + "epoch": 5.97, + "learning_rate": 3.1897338724401185e-06, + "loss": 0.0085, + "step": 45579 + }, + { + "epoch": 5.97, + "learning_rate": 3.189345546892726e-06, + "loss": 0.005, + "step": 45580 + }, + { + "epoch": 5.97, + "learning_rate": 3.188957240499936e-06, + "loss": 0.0048, + "step": 45581 + }, + { + "epoch": 5.97, + "learning_rate": 3.188568953262837e-06, + "loss": 0.0061, + "step": 45582 + }, + { + "epoch": 5.97, + "learning_rate": 3.188180685182529e-06, + "loss": 0.0055, + "step": 45583 + }, + { + "epoch": 5.97, + "learning_rate": 3.1877924362601e-06, + "loss": 0.0069, + "step": 45584 + }, + { + "epoch": 5.97, + "learning_rate": 3.1874042064966383e-06, + "loss": 0.0058, + "step": 45585 + }, + { + "epoch": 5.97, + "learning_rate": 3.18701599589324e-06, + "loss": 0.0052, + "step": 45586 + }, + { + "epoch": 5.97, + "learning_rate": 3.1866278044510003e-06, + "loss": 0.0057, + "step": 45587 + }, + { + "epoch": 5.97, + "learning_rate": 3.1862396321710052e-06, + "loss": 0.0051, + "step": 45588 + }, + { + "epoch": 5.97, + "learning_rate": 3.1858514790543504e-06, + "loss": 0.0051, + "step": 45589 + }, + { + "epoch": 5.97, + "learning_rate": 3.185463345102121e-06, + "loss": 0.0037, + "step": 45590 + }, + { + "epoch": 5.97, + "learning_rate": 3.1850752303154164e-06, + "loss": 0.0061, + "step": 45591 + }, + { + "epoch": 5.97, + "learning_rate": 3.1846871346953246e-06, + "loss": 0.0078, + "step": 45592 + }, + { + "epoch": 5.97, + "learning_rate": 3.1842990582429345e-06, + "loss": 0.0041, + "step": 45593 + }, + { + "epoch": 5.97, + "learning_rate": 3.1839110009593434e-06, + "loss": 0.005, + "step": 45594 + }, + { + "epoch": 5.97, + "learning_rate": 3.1835229628456397e-06, + "loss": 0.0052, + "step": 45595 + }, + { + "epoch": 5.97, + "learning_rate": 3.1831349439029145e-06, + "loss": 0.0074, + "step": 45596 + }, + { + "epoch": 5.97, + "learning_rate": 3.1827469441322566e-06, + "loss": 0.0071, + "step": 45597 + }, + { + "epoch": 5.97, + "learning_rate": 3.1823589635347596e-06, + "loss": 0.006, + "step": 45598 + }, + { + "epoch": 5.97, + "learning_rate": 3.181971002111518e-06, + "loss": 0.0071, + "step": 45599 + }, + { + "epoch": 5.97, + "learning_rate": 3.1815830598636198e-06, + "loss": 0.0062, + "step": 45600 + }, + { + "epoch": 5.97, + "learning_rate": 3.1811951367921535e-06, + "loss": 0.0041, + "step": 45601 + }, + { + "epoch": 5.97, + "learning_rate": 3.1808072328982154e-06, + "loss": 0.0042, + "step": 45602 + }, + { + "epoch": 5.97, + "learning_rate": 3.180419348182894e-06, + "loss": 0.0066, + "step": 45603 + }, + { + "epoch": 5.97, + "learning_rate": 3.180031482647279e-06, + "loss": 0.0058, + "step": 45604 + }, + { + "epoch": 5.97, + "learning_rate": 3.17964363629246e-06, + "loss": 0.0043, + "step": 45605 + }, + { + "epoch": 5.97, + "learning_rate": 3.1792558091195336e-06, + "loss": 0.0031, + "step": 45606 + }, + { + "epoch": 5.97, + "learning_rate": 3.1788680011295868e-06, + "loss": 0.0044, + "step": 45607 + }, + { + "epoch": 5.97, + "learning_rate": 3.1784802123237067e-06, + "loss": 0.0086, + "step": 45608 + }, + { + "epoch": 5.97, + "learning_rate": 3.178092442702989e-06, + "loss": 0.0053, + "step": 45609 + }, + { + "epoch": 5.97, + "learning_rate": 3.1777046922685253e-06, + "loss": 0.0054, + "step": 45610 + }, + { + "epoch": 5.97, + "learning_rate": 3.1773169610214037e-06, + "loss": 0.0089, + "step": 45611 + }, + { + "epoch": 5.97, + "learning_rate": 3.1769292489627146e-06, + "loss": 0.0046, + "step": 45612 + }, + { + "epoch": 5.97, + "learning_rate": 3.1765415560935463e-06, + "loss": 0.0037, + "step": 45613 + }, + { + "epoch": 5.98, + "learning_rate": 3.176153882414994e-06, + "loss": 0.0076, + "step": 45614 + }, + { + "epoch": 5.98, + "learning_rate": 3.1757662279281467e-06, + "loss": 0.0063, + "step": 45615 + }, + { + "epoch": 5.98, + "learning_rate": 3.1753785926340884e-06, + "loss": 0.0059, + "step": 45616 + }, + { + "epoch": 5.98, + "learning_rate": 3.174990976533919e-06, + "loss": 0.0045, + "step": 45617 + }, + { + "epoch": 5.98, + "learning_rate": 3.1746033796287235e-06, + "loss": 0.0058, + "step": 45618 + }, + { + "epoch": 5.98, + "learning_rate": 3.1742158019195903e-06, + "loss": 0.0046, + "step": 45619 + }, + { + "epoch": 5.98, + "learning_rate": 3.1738282434076105e-06, + "loss": 0.0053, + "step": 45620 + }, + { + "epoch": 5.98, + "learning_rate": 3.1734407040938786e-06, + "loss": 0.007, + "step": 45621 + }, + { + "epoch": 5.98, + "learning_rate": 3.173053183979481e-06, + "loss": 0.0051, + "step": 45622 + }, + { + "epoch": 5.98, + "learning_rate": 3.1726656830655077e-06, + "loss": 0.0085, + "step": 45623 + }, + { + "epoch": 5.98, + "learning_rate": 3.1722782013530464e-06, + "loss": 0.0061, + "step": 45624 + }, + { + "epoch": 5.98, + "learning_rate": 3.171890738843192e-06, + "loss": 0.0063, + "step": 45625 + }, + { + "epoch": 5.98, + "learning_rate": 3.1715032955370306e-06, + "loss": 0.0052, + "step": 45626 + }, + { + "epoch": 5.98, + "learning_rate": 3.1711158714356495e-06, + "loss": 0.0033, + "step": 45627 + }, + { + "epoch": 5.98, + "learning_rate": 3.170728466540145e-06, + "loss": 0.0077, + "step": 45628 + }, + { + "epoch": 5.98, + "learning_rate": 3.170341080851602e-06, + "loss": 0.005, + "step": 45629 + }, + { + "epoch": 5.98, + "learning_rate": 3.169953714371109e-06, + "loss": 0.0052, + "step": 45630 + }, + { + "epoch": 5.98, + "learning_rate": 3.16956636709976e-06, + "loss": 0.0063, + "step": 45631 + }, + { + "epoch": 5.98, + "learning_rate": 3.169179039038639e-06, + "loss": 0.0056, + "step": 45632 + }, + { + "epoch": 5.98, + "learning_rate": 3.1687917301888417e-06, + "loss": 0.005, + "step": 45633 + }, + { + "epoch": 5.98, + "learning_rate": 3.1684044405514536e-06, + "loss": 0.0047, + "step": 45634 + }, + { + "epoch": 5.98, + "learning_rate": 3.168017170127561e-06, + "loss": 0.0057, + "step": 45635 + }, + { + "epoch": 5.98, + "learning_rate": 3.167629918918259e-06, + "loss": 0.0041, + "step": 45636 + }, + { + "epoch": 5.98, + "learning_rate": 3.1672426869246343e-06, + "loss": 0.0056, + "step": 45637 + }, + { + "epoch": 5.98, + "learning_rate": 3.1668554741477752e-06, + "loss": 0.0057, + "step": 45638 + }, + { + "epoch": 5.98, + "learning_rate": 3.166468280588769e-06, + "loss": 0.0042, + "step": 45639 + }, + { + "epoch": 5.98, + "learning_rate": 3.166081106248706e-06, + "loss": 0.0045, + "step": 45640 + }, + { + "epoch": 5.98, + "learning_rate": 3.1656939511286798e-06, + "loss": 0.0051, + "step": 45641 + }, + { + "epoch": 5.98, + "learning_rate": 3.165306815229775e-06, + "loss": 0.0048, + "step": 45642 + }, + { + "epoch": 5.98, + "learning_rate": 3.1649196985530774e-06, + "loss": 0.0053, + "step": 45643 + }, + { + "epoch": 5.98, + "learning_rate": 3.1645326010996826e-06, + "loss": 0.0064, + "step": 45644 + }, + { + "epoch": 5.98, + "learning_rate": 3.1641455228706763e-06, + "loss": 0.0047, + "step": 45645 + }, + { + "epoch": 5.98, + "learning_rate": 3.163758463867146e-06, + "loss": 0.004, + "step": 45646 + }, + { + "epoch": 5.98, + "learning_rate": 3.1633714240901783e-06, + "loss": 0.0033, + "step": 45647 + }, + { + "epoch": 5.98, + "learning_rate": 3.1629844035408673e-06, + "loss": 0.0047, + "step": 45648 + }, + { + "epoch": 5.98, + "learning_rate": 3.1625974022202986e-06, + "loss": 0.008, + "step": 45649 + }, + { + "epoch": 5.98, + "learning_rate": 3.162210420129558e-06, + "loss": 0.0068, + "step": 45650 + }, + { + "epoch": 5.98, + "learning_rate": 3.1618234572697358e-06, + "loss": 0.0054, + "step": 45651 + }, + { + "epoch": 5.98, + "learning_rate": 3.1614365136419245e-06, + "loss": 0.0043, + "step": 45652 + }, + { + "epoch": 5.98, + "learning_rate": 3.1610495892472084e-06, + "loss": 0.0039, + "step": 45653 + }, + { + "epoch": 5.98, + "learning_rate": 3.1606626840866763e-06, + "loss": 0.0075, + "step": 45654 + }, + { + "epoch": 5.98, + "learning_rate": 3.160275798161413e-06, + "loss": 0.0074, + "step": 45655 + }, + { + "epoch": 5.98, + "learning_rate": 3.1598889314725125e-06, + "loss": 0.0055, + "step": 45656 + }, + { + "epoch": 5.98, + "learning_rate": 3.159502084021061e-06, + "loss": 0.0056, + "step": 45657 + }, + { + "epoch": 5.98, + "learning_rate": 3.1591152558081418e-06, + "loss": 0.0036, + "step": 45658 + }, + { + "epoch": 5.98, + "learning_rate": 3.15872844683485e-06, + "loss": 0.0064, + "step": 45659 + }, + { + "epoch": 5.98, + "learning_rate": 3.158341657102271e-06, + "loss": 0.0054, + "step": 45660 + }, + { + "epoch": 5.98, + "learning_rate": 3.1579548866114906e-06, + "loss": 0.0049, + "step": 45661 + }, + { + "epoch": 5.98, + "learning_rate": 3.1575681353635946e-06, + "loss": 0.005, + "step": 45662 + }, + { + "epoch": 5.98, + "learning_rate": 3.1571814033596748e-06, + "loss": 0.004, + "step": 45663 + }, + { + "epoch": 5.98, + "learning_rate": 3.1567946906008195e-06, + "loss": 0.0033, + "step": 45664 + }, + { + "epoch": 5.98, + "learning_rate": 3.1564079970881156e-06, + "loss": 0.0068, + "step": 45665 + }, + { + "epoch": 5.98, + "learning_rate": 3.1560213228226464e-06, + "loss": 0.0057, + "step": 45666 + }, + { + "epoch": 5.98, + "learning_rate": 3.155634667805507e-06, + "loss": 0.0057, + "step": 45667 + }, + { + "epoch": 5.98, + "learning_rate": 3.15524803203778e-06, + "loss": 0.0056, + "step": 45668 + }, + { + "epoch": 5.98, + "learning_rate": 3.1548614155205525e-06, + "loss": 0.0058, + "step": 45669 + }, + { + "epoch": 5.98, + "learning_rate": 3.1544748182549103e-06, + "loss": 0.0075, + "step": 45670 + }, + { + "epoch": 5.98, + "learning_rate": 3.154088240241947e-06, + "loss": 0.0052, + "step": 45671 + }, + { + "epoch": 5.98, + "learning_rate": 3.153701681482745e-06, + "loss": 0.0062, + "step": 45672 + }, + { + "epoch": 5.98, + "learning_rate": 3.1533151419783893e-06, + "loss": 0.0063, + "step": 45673 + }, + { + "epoch": 5.98, + "learning_rate": 3.1529286217299703e-06, + "loss": 0.006, + "step": 45674 + }, + { + "epoch": 5.98, + "learning_rate": 3.1525421207385774e-06, + "loss": 0.0083, + "step": 45675 + }, + { + "epoch": 5.98, + "learning_rate": 3.1521556390052967e-06, + "loss": 0.0066, + "step": 45676 + }, + { + "epoch": 5.98, + "learning_rate": 3.1517691765312096e-06, + "loss": 0.0058, + "step": 45677 + }, + { + "epoch": 5.98, + "learning_rate": 3.1513827333174095e-06, + "loss": 0.0042, + "step": 45678 + }, + { + "epoch": 5.98, + "learning_rate": 3.150996309364982e-06, + "loss": 0.0059, + "step": 45679 + }, + { + "epoch": 5.98, + "learning_rate": 3.1506099046750115e-06, + "loss": 0.0061, + "step": 45680 + }, + { + "epoch": 5.98, + "learning_rate": 3.150223519248584e-06, + "loss": 0.0066, + "step": 45681 + }, + { + "epoch": 5.98, + "learning_rate": 3.14983715308679e-06, + "loss": 0.0038, + "step": 45682 + }, + { + "epoch": 5.98, + "learning_rate": 3.1494508061907147e-06, + "loss": 0.0049, + "step": 45683 + }, + { + "epoch": 5.98, + "learning_rate": 3.149064478561441e-06, + "loss": 0.0055, + "step": 45684 + }, + { + "epoch": 5.98, + "learning_rate": 3.1486781702000593e-06, + "loss": 0.0069, + "step": 45685 + }, + { + "epoch": 5.98, + "learning_rate": 3.148291881107658e-06, + "loss": 0.0052, + "step": 45686 + }, + { + "epoch": 5.98, + "learning_rate": 3.14790561128532e-06, + "loss": 0.0044, + "step": 45687 + }, + { + "epoch": 5.98, + "learning_rate": 3.1475193607341335e-06, + "loss": 0.0057, + "step": 45688 + }, + { + "epoch": 5.98, + "learning_rate": 3.1471331294551812e-06, + "loss": 0.0058, + "step": 45689 + }, + { + "epoch": 5.99, + "learning_rate": 3.146746917449555e-06, + "loss": 0.0054, + "step": 45690 + }, + { + "epoch": 5.99, + "learning_rate": 3.1463607247183368e-06, + "loss": 0.0054, + "step": 45691 + }, + { + "epoch": 5.99, + "learning_rate": 3.1459745512626126e-06, + "loss": 0.0079, + "step": 45692 + }, + { + "epoch": 5.99, + "learning_rate": 3.1455883970834723e-06, + "loss": 0.0084, + "step": 45693 + }, + { + "epoch": 5.99, + "learning_rate": 3.1452022621819998e-06, + "loss": 0.0083, + "step": 45694 + }, + { + "epoch": 5.99, + "learning_rate": 3.1448161465592775e-06, + "loss": 0.0037, + "step": 45695 + }, + { + "epoch": 5.99, + "learning_rate": 3.144430050216398e-06, + "loss": 0.0057, + "step": 45696 + }, + { + "epoch": 5.99, + "learning_rate": 3.1440439731544413e-06, + "loss": 0.0091, + "step": 45697 + }, + { + "epoch": 5.99, + "learning_rate": 3.1436579153744982e-06, + "loss": 0.0044, + "step": 45698 + }, + { + "epoch": 5.99, + "learning_rate": 3.143271876877653e-06, + "loss": 0.0055, + "step": 45699 + }, + { + "epoch": 5.99, + "learning_rate": 3.1428858576649855e-06, + "loss": 0.0061, + "step": 45700 + }, + { + "epoch": 5.99, + "learning_rate": 3.142499857737592e-06, + "loss": 0.0029, + "step": 45701 + }, + { + "epoch": 5.99, + "learning_rate": 3.142113877096551e-06, + "loss": 0.0056, + "step": 45702 + }, + { + "epoch": 5.99, + "learning_rate": 3.141727915742949e-06, + "loss": 0.0035, + "step": 45703 + }, + { + "epoch": 5.99, + "learning_rate": 3.1413419736778707e-06, + "loss": 0.0048, + "step": 45704 + }, + { + "epoch": 5.99, + "learning_rate": 3.140956050902404e-06, + "loss": 0.0059, + "step": 45705 + }, + { + "epoch": 5.99, + "learning_rate": 3.1405701474176307e-06, + "loss": 0.0059, + "step": 45706 + }, + { + "epoch": 5.99, + "learning_rate": 3.1401842632246427e-06, + "loss": 0.006, + "step": 45707 + }, + { + "epoch": 5.99, + "learning_rate": 3.1397983983245173e-06, + "loss": 0.0056, + "step": 45708 + }, + { + "epoch": 5.99, + "learning_rate": 3.139412552718346e-06, + "loss": 0.004, + "step": 45709 + }, + { + "epoch": 5.99, + "learning_rate": 3.139026726407213e-06, + "loss": 0.0075, + "step": 45710 + }, + { + "epoch": 5.99, + "learning_rate": 3.1386409193922006e-06, + "loss": 0.0066, + "step": 45711 + }, + { + "epoch": 5.99, + "learning_rate": 3.1382551316743925e-06, + "loss": 0.0065, + "step": 45712 + }, + { + "epoch": 5.99, + "learning_rate": 3.1378693632548794e-06, + "loss": 0.0061, + "step": 45713 + }, + { + "epoch": 5.99, + "learning_rate": 3.1374836141347444e-06, + "loss": 0.0056, + "step": 45714 + }, + { + "epoch": 5.99, + "learning_rate": 3.137097884315067e-06, + "loss": 0.0072, + "step": 45715 + }, + { + "epoch": 5.99, + "learning_rate": 3.13671217379694e-06, + "loss": 0.0071, + "step": 45716 + }, + { + "epoch": 5.99, + "learning_rate": 3.136326482581441e-06, + "loss": 0.0079, + "step": 45717 + }, + { + "epoch": 5.99, + "learning_rate": 3.135940810669662e-06, + "loss": 0.0054, + "step": 45718 + }, + { + "epoch": 5.99, + "learning_rate": 3.1355551580626833e-06, + "loss": 0.0044, + "step": 45719 + }, + { + "epoch": 5.99, + "learning_rate": 3.1351695247615877e-06, + "loss": 0.0043, + "step": 45720 + }, + { + "epoch": 5.99, + "learning_rate": 3.134783910767465e-06, + "loss": 0.0061, + "step": 45721 + }, + { + "epoch": 5.99, + "learning_rate": 3.1343983160813963e-06, + "loss": 0.0044, + "step": 45722 + }, + { + "epoch": 5.99, + "learning_rate": 3.134012740704464e-06, + "loss": 0.0042, + "step": 45723 + }, + { + "epoch": 5.99, + "learning_rate": 3.1336271846377585e-06, + "loss": 0.0045, + "step": 45724 + }, + { + "epoch": 5.99, + "learning_rate": 3.1332416478823613e-06, + "loss": 0.0051, + "step": 45725 + }, + { + "epoch": 5.99, + "learning_rate": 3.1328561304393556e-06, + "loss": 0.0066, + "step": 45726 + }, + { + "epoch": 5.99, + "learning_rate": 3.132470632309824e-06, + "loss": 0.0056, + "step": 45727 + }, + { + "epoch": 5.99, + "learning_rate": 3.1320851534948526e-06, + "loss": 0.0037, + "step": 45728 + }, + { + "epoch": 5.99, + "learning_rate": 3.131699693995529e-06, + "loss": 0.0055, + "step": 45729 + }, + { + "epoch": 5.99, + "learning_rate": 3.131314253812934e-06, + "loss": 0.0096, + "step": 45730 + }, + { + "epoch": 5.99, + "learning_rate": 3.130928832948149e-06, + "loss": 0.0062, + "step": 45731 + }, + { + "epoch": 5.99, + "learning_rate": 3.1305434314022644e-06, + "loss": 0.0043, + "step": 45732 + }, + { + "epoch": 5.99, + "learning_rate": 3.1301580491763606e-06, + "loss": 0.0064, + "step": 45733 + }, + { + "epoch": 5.99, + "learning_rate": 3.1297726862715205e-06, + "loss": 0.0064, + "step": 45734 + }, + { + "epoch": 5.99, + "learning_rate": 3.1293873426888268e-06, + "loss": 0.004, + "step": 45735 + }, + { + "epoch": 5.99, + "learning_rate": 3.1290020184293677e-06, + "loss": 0.0034, + "step": 45736 + }, + { + "epoch": 5.99, + "learning_rate": 3.1286167134942246e-06, + "loss": 0.0038, + "step": 45737 + }, + { + "epoch": 5.99, + "learning_rate": 3.1282314278844783e-06, + "loss": 0.0057, + "step": 45738 + }, + { + "epoch": 5.99, + "learning_rate": 3.1278461616012147e-06, + "loss": 0.0067, + "step": 45739 + }, + { + "epoch": 5.99, + "learning_rate": 3.1274609146455213e-06, + "loss": 0.0062, + "step": 45740 + }, + { + "epoch": 5.99, + "learning_rate": 3.127075687018478e-06, + "loss": 0.0051, + "step": 45741 + }, + { + "epoch": 5.99, + "learning_rate": 3.1266904787211647e-06, + "loss": 0.0051, + "step": 45742 + }, + { + "epoch": 5.99, + "learning_rate": 3.126305289754673e-06, + "loss": 0.0039, + "step": 45743 + }, + { + "epoch": 5.99, + "learning_rate": 3.1259201201200806e-06, + "loss": 0.006, + "step": 45744 + }, + { + "epoch": 5.99, + "learning_rate": 3.125534969818471e-06, + "loss": 0.0039, + "step": 45745 + }, + { + "epoch": 5.99, + "learning_rate": 3.125149838850926e-06, + "loss": 0.0047, + "step": 45746 + }, + { + "epoch": 5.99, + "learning_rate": 3.1247647272185343e-06, + "loss": 0.0055, + "step": 45747 + }, + { + "epoch": 5.99, + "learning_rate": 3.1243796349223754e-06, + "loss": 0.0064, + "step": 45748 + }, + { + "epoch": 5.99, + "learning_rate": 3.1239945619635327e-06, + "loss": 0.005, + "step": 45749 + }, + { + "epoch": 5.99, + "learning_rate": 3.123609508343083e-06, + "loss": 0.0051, + "step": 45750 + }, + { + "epoch": 5.99, + "learning_rate": 3.123224474062122e-06, + "loss": 0.005, + "step": 45751 + }, + { + "epoch": 5.99, + "learning_rate": 3.1228394591217246e-06, + "loss": 0.0055, + "step": 45752 + }, + { + "epoch": 5.99, + "learning_rate": 3.1224544635229758e-06, + "loss": 0.0041, + "step": 45753 + }, + { + "epoch": 5.99, + "learning_rate": 3.1220694872669544e-06, + "loss": 0.0063, + "step": 45754 + }, + { + "epoch": 5.99, + "learning_rate": 3.1216845303547494e-06, + "loss": 0.0066, + "step": 45755 + }, + { + "epoch": 5.99, + "learning_rate": 3.121299592787441e-06, + "loss": 0.0054, + "step": 45756 + }, + { + "epoch": 5.99, + "learning_rate": 3.1209146745661066e-06, + "loss": 0.0061, + "step": 45757 + }, + { + "epoch": 5.99, + "learning_rate": 3.1205297756918384e-06, + "loss": 0.0062, + "step": 45758 + }, + { + "epoch": 5.99, + "learning_rate": 3.120144896165712e-06, + "loss": 0.0049, + "step": 45759 + }, + { + "epoch": 5.99, + "learning_rate": 3.1197600359888137e-06, + "loss": 0.005, + "step": 45760 + }, + { + "epoch": 5.99, + "learning_rate": 3.1193751951622187e-06, + "loss": 0.005, + "step": 45761 + }, + { + "epoch": 5.99, + "learning_rate": 3.118990373687015e-06, + "loss": 0.008, + "step": 45762 + }, + { + "epoch": 5.99, + "learning_rate": 3.1186055715642883e-06, + "loss": 0.0049, + "step": 45763 + }, + { + "epoch": 5.99, + "learning_rate": 3.1182207887951167e-06, + "loss": 0.0054, + "step": 45764 + }, + { + "epoch": 5.99, + "learning_rate": 3.1178360253805797e-06, + "loss": 0.0088, + "step": 45765 + }, + { + "epoch": 6.0, + "learning_rate": 3.117451281321765e-06, + "loss": 0.0062, + "step": 45766 + }, + { + "epoch": 6.0, + "learning_rate": 3.1170665566197533e-06, + "loss": 0.0057, + "step": 45767 + }, + { + "epoch": 6.0, + "learning_rate": 3.1166818512756238e-06, + "loss": 0.0045, + "step": 45768 + }, + { + "epoch": 6.0, + "learning_rate": 3.1162971652904572e-06, + "loss": 0.0045, + "step": 45769 + }, + { + "epoch": 6.0, + "learning_rate": 3.115912498665342e-06, + "loss": 0.0051, + "step": 45770 + }, + { + "epoch": 6.0, + "learning_rate": 3.115527851401353e-06, + "loss": 0.0065, + "step": 45771 + }, + { + "epoch": 6.0, + "learning_rate": 3.115143223499578e-06, + "loss": 0.0075, + "step": 45772 + }, + { + "epoch": 6.0, + "learning_rate": 3.114758614961093e-06, + "loss": 0.0055, + "step": 45773 + }, + { + "epoch": 6.0, + "learning_rate": 3.114374025786987e-06, + "loss": 0.0087, + "step": 45774 + }, + { + "epoch": 6.0, + "learning_rate": 3.1139894559783355e-06, + "loss": 0.0067, + "step": 45775 + }, + { + "epoch": 6.0, + "learning_rate": 3.113604905536224e-06, + "loss": 0.0053, + "step": 45776 + }, + { + "epoch": 6.0, + "learning_rate": 3.1132203744617275e-06, + "loss": 0.0038, + "step": 45777 + }, + { + "epoch": 6.0, + "learning_rate": 3.1128358627559353e-06, + "loss": 0.0046, + "step": 45778 + }, + { + "epoch": 6.0, + "learning_rate": 3.1124513704199253e-06, + "loss": 0.005, + "step": 45779 + }, + { + "epoch": 6.0, + "learning_rate": 3.112066897454776e-06, + "loss": 0.0064, + "step": 45780 + }, + { + "epoch": 6.0, + "learning_rate": 3.1116824438615754e-06, + "loss": 0.0045, + "step": 45781 + }, + { + "epoch": 6.0, + "learning_rate": 3.111298009641397e-06, + "loss": 0.004, + "step": 45782 + }, + { + "epoch": 6.0, + "learning_rate": 3.1109135947953306e-06, + "loss": 0.0065, + "step": 45783 + }, + { + "epoch": 6.0, + "learning_rate": 3.110529199324451e-06, + "loss": 0.0064, + "step": 45784 + }, + { + "epoch": 6.0, + "learning_rate": 3.110144823229839e-06, + "loss": 0.0072, + "step": 45785 + }, + { + "epoch": 6.0, + "learning_rate": 3.1097604665125814e-06, + "loss": 0.0042, + "step": 45786 + }, + { + "epoch": 6.0, + "learning_rate": 3.1093761291737556e-06, + "loss": 0.006, + "step": 45787 + }, + { + "epoch": 6.0, + "learning_rate": 3.108991811214439e-06, + "loss": 0.0044, + "step": 45788 + }, + { + "epoch": 6.0, + "learning_rate": 3.1086075126357185e-06, + "loss": 0.0041, + "step": 45789 + }, + { + "epoch": 6.0, + "learning_rate": 3.108223233438673e-06, + "loss": 0.0056, + "step": 45790 + }, + { + "epoch": 6.0, + "learning_rate": 3.107838973624383e-06, + "loss": 0.0055, + "step": 45791 + }, + { + "epoch": 6.0, + "learning_rate": 3.107454733193925e-06, + "loss": 0.004, + "step": 45792 + }, + { + "epoch": 6.0, + "learning_rate": 3.1070705121483837e-06, + "loss": 0.0064, + "step": 45793 + }, + { + "epoch": 6.0, + "learning_rate": 3.106686310488842e-06, + "loss": 0.0058, + "step": 45794 + }, + { + "epoch": 6.0, + "learning_rate": 3.1063021282163795e-06, + "loss": 0.006, + "step": 45795 + }, + { + "epoch": 6.0, + "learning_rate": 3.1059179653320713e-06, + "loss": 0.0058, + "step": 45796 + }, + { + "epoch": 6.0, + "learning_rate": 3.1055338218370044e-06, + "loss": 0.0079, + "step": 45797 + }, + { + "epoch": 6.0, + "learning_rate": 3.105149697732257e-06, + "loss": 0.0045, + "step": 45798 + }, + { + "epoch": 6.0, + "learning_rate": 3.104765593018909e-06, + "loss": 0.0062, + "step": 45799 + }, + { + "epoch": 6.0, + "learning_rate": 3.1043815076980367e-06, + "loss": 0.004, + "step": 45800 + }, + { + "epoch": 6.0, + "learning_rate": 3.1039974417707273e-06, + "loss": 0.0072, + "step": 45801 + }, + { + "epoch": 6.0, + "learning_rate": 3.103613395238059e-06, + "loss": 0.0059, + "step": 45802 + }, + { + "epoch": 6.0, + "learning_rate": 3.1032293681011074e-06, + "loss": 0.0045, + "step": 45803 + }, + { + "epoch": 6.0, + "learning_rate": 3.1028453603609554e-06, + "loss": 0.0042, + "step": 45804 + }, + { + "epoch": 6.0, + "learning_rate": 3.102461372018686e-06, + "loss": 0.0022, + "step": 45805 + }, + { + "epoch": 6.0, + "learning_rate": 3.1020774030753777e-06, + "loss": 0.0035, + "step": 45806 + }, + { + "epoch": 6.0, + "learning_rate": 3.1016934535321063e-06, + "loss": 0.0031, + "step": 45807 + }, + { + "epoch": 6.0, + "learning_rate": 3.1013095233899572e-06, + "loss": 0.002, + "step": 45808 + }, + { + "epoch": 6.0, + "learning_rate": 3.1009256126500076e-06, + "loss": 0.0046, + "step": 45809 + }, + { + "epoch": 6.0, + "learning_rate": 3.100541721313338e-06, + "loss": 0.002, + "step": 45810 + }, + { + "epoch": 6.0, + "learning_rate": 3.1001578493810237e-06, + "loss": 0.0018, + "step": 45811 + }, + { + "epoch": 6.0, + "learning_rate": 3.099773996854152e-06, + "loss": 0.0023, + "step": 45812 + }, + { + "epoch": 6.0, + "learning_rate": 3.0993901637337977e-06, + "loss": 0.0034, + "step": 45813 + }, + { + "epoch": 6.0, + "learning_rate": 3.099006350021041e-06, + "loss": 0.0022, + "step": 45814 + }, + { + "epoch": 6.0, + "learning_rate": 3.098622555716956e-06, + "loss": 0.0015, + "step": 45815 + }, + { + "epoch": 6.0, + "learning_rate": 3.0982387808226332e-06, + "loss": 0.0023, + "step": 45816 + }, + { + "epoch": 6.0, + "learning_rate": 3.097855025339146e-06, + "loss": 0.0029, + "step": 45817 + }, + { + "epoch": 6.0, + "learning_rate": 3.097471289267574e-06, + "loss": 0.0028, + "step": 45818 + }, + { + "epoch": 6.0, + "learning_rate": 3.0970875726089944e-06, + "loss": 0.0018, + "step": 45819 + }, + { + "epoch": 6.0, + "learning_rate": 3.0967038753644907e-06, + "loss": 0.0019, + "step": 45820 + }, + { + "epoch": 6.0, + "learning_rate": 3.09632019753514e-06, + "loss": 0.0021, + "step": 45821 + }, + { + "epoch": 6.0, + "learning_rate": 3.0959365391220176e-06, + "loss": 0.0027, + "step": 45822 + }, + { + "epoch": 6.0, + "learning_rate": 3.0955529001262086e-06, + "loss": 0.0027, + "step": 45823 + }, + { + "epoch": 6.0, + "learning_rate": 3.095169280548791e-06, + "loss": 0.0028, + "step": 45824 + }, + { + "epoch": 6.0, + "learning_rate": 3.0947856803908406e-06, + "loss": 0.0024, + "step": 45825 + }, + { + "epoch": 6.0, + "learning_rate": 3.094402099653435e-06, + "loss": 0.0028, + "step": 45826 + }, + { + "epoch": 6.0, + "learning_rate": 3.0940185383376566e-06, + "loss": 0.0026, + "step": 45827 + }, + { + "epoch": 6.0, + "learning_rate": 3.0936349964445854e-06, + "loss": 0.0027, + "step": 45828 + }, + { + "epoch": 6.0, + "learning_rate": 3.0932514739752984e-06, + "loss": 0.0022, + "step": 45829 + }, + { + "epoch": 6.0, + "learning_rate": 3.0928679709308707e-06, + "loss": 0.0032, + "step": 45830 + }, + { + "epoch": 6.0, + "learning_rate": 3.0924844873123872e-06, + "loss": 0.0017, + "step": 45831 + }, + { + "epoch": 6.0, + "learning_rate": 3.0921010231209224e-06, + "loss": 0.0021, + "step": 45832 + }, + { + "epoch": 6.0, + "learning_rate": 3.091717578357557e-06, + "loss": 0.0027, + "step": 45833 + }, + { + "epoch": 6.0, + "learning_rate": 3.091334153023364e-06, + "loss": 0.002, + "step": 45834 + }, + { + "epoch": 6.0, + "learning_rate": 3.090950747119429e-06, + "loss": 0.0027, + "step": 45835 + }, + { + "epoch": 6.0, + "learning_rate": 3.0905673606468277e-06, + "loss": 0.0013, + "step": 45836 + }, + { + "epoch": 6.0, + "learning_rate": 3.0901839936066335e-06, + "loss": 0.0039, + "step": 45837 + }, + { + "epoch": 6.0, + "learning_rate": 3.0898006459999298e-06, + "loss": 0.0017, + "step": 45838 + }, + { + "epoch": 6.0, + "learning_rate": 3.0894173178277965e-06, + "loss": 0.0017, + "step": 45839 + }, + { + "epoch": 6.0, + "learning_rate": 3.0890340090913097e-06, + "loss": 0.0013, + "step": 45840 + }, + { + "epoch": 6.0, + "learning_rate": 3.088650719791546e-06, + "loss": 0.0026, + "step": 45841 + }, + { + "epoch": 6.0, + "learning_rate": 3.0882674499295807e-06, + "loss": 0.0016, + "step": 45842 + }, + { + "epoch": 6.01, + "learning_rate": 3.087884199506499e-06, + "loss": 0.0019, + "step": 45843 + }, + { + "epoch": 6.01, + "learning_rate": 3.0875009685233747e-06, + "loss": 0.0035, + "step": 45844 + }, + { + "epoch": 6.01, + "learning_rate": 3.0871177569812814e-06, + "loss": 0.0017, + "step": 45845 + }, + { + "epoch": 6.01, + "learning_rate": 3.0867345648813065e-06, + "loss": 0.002, + "step": 45846 + }, + { + "epoch": 6.01, + "learning_rate": 3.0863513922245214e-06, + "loss": 0.0017, + "step": 45847 + }, + { + "epoch": 6.01, + "learning_rate": 3.0859682390120015e-06, + "loss": 0.0019, + "step": 45848 + }, + { + "epoch": 6.01, + "learning_rate": 3.0855851052448317e-06, + "loss": 0.0025, + "step": 45849 + }, + { + "epoch": 6.01, + "learning_rate": 3.0852019909240815e-06, + "loss": 0.0032, + "step": 45850 + }, + { + "epoch": 6.01, + "learning_rate": 3.0848188960508364e-06, + "loss": 0.0037, + "step": 45851 + }, + { + "epoch": 6.01, + "learning_rate": 3.0844358206261695e-06, + "loss": 0.0019, + "step": 45852 + }, + { + "epoch": 6.01, + "learning_rate": 3.0840527646511553e-06, + "loss": 0.0021, + "step": 45853 + }, + { + "epoch": 6.01, + "learning_rate": 3.083669728126878e-06, + "loss": 0.0037, + "step": 45854 + }, + { + "epoch": 6.01, + "learning_rate": 3.083286711054411e-06, + "loss": 0.0013, + "step": 45855 + }, + { + "epoch": 6.01, + "learning_rate": 3.0829037134348315e-06, + "loss": 0.003, + "step": 45856 + }, + { + "epoch": 6.01, + "learning_rate": 3.0825207352692142e-06, + "loss": 0.0039, + "step": 45857 + }, + { + "epoch": 6.01, + "learning_rate": 3.0821377765586414e-06, + "loss": 0.0021, + "step": 45858 + }, + { + "epoch": 6.01, + "learning_rate": 3.081754837304185e-06, + "loss": 0.0023, + "step": 45859 + }, + { + "epoch": 6.01, + "learning_rate": 3.081371917506929e-06, + "loss": 0.0013, + "step": 45860 + }, + { + "epoch": 6.01, + "learning_rate": 3.0809890171679413e-06, + "loss": 0.0027, + "step": 45861 + }, + { + "epoch": 6.01, + "learning_rate": 3.0806061362883076e-06, + "loss": 0.003, + "step": 45862 + }, + { + "epoch": 6.01, + "learning_rate": 3.0802232748690995e-06, + "loss": 0.0015, + "step": 45863 + }, + { + "epoch": 6.01, + "learning_rate": 3.0798404329113963e-06, + "loss": 0.0017, + "step": 45864 + }, + { + "epoch": 6.01, + "learning_rate": 3.0794576104162698e-06, + "loss": 0.0022, + "step": 45865 + }, + { + "epoch": 6.01, + "learning_rate": 3.0790748073848033e-06, + "loss": 0.0021, + "step": 45866 + }, + { + "epoch": 6.01, + "learning_rate": 3.0786920238180706e-06, + "loss": 0.0025, + "step": 45867 + }, + { + "epoch": 6.01, + "learning_rate": 3.078309259717144e-06, + "loss": 0.0017, + "step": 45868 + }, + { + "epoch": 6.01, + "learning_rate": 3.0779265150831083e-06, + "loss": 0.0018, + "step": 45869 + }, + { + "epoch": 6.01, + "learning_rate": 3.0775437899170314e-06, + "loss": 0.0028, + "step": 45870 + }, + { + "epoch": 6.01, + "learning_rate": 3.077161084219997e-06, + "loss": 0.0012, + "step": 45871 + }, + { + "epoch": 6.01, + "learning_rate": 3.0767783979930765e-06, + "loss": 0.002, + "step": 45872 + }, + { + "epoch": 6.01, + "learning_rate": 3.0763957312373507e-06, + "loss": 0.0022, + "step": 45873 + }, + { + "epoch": 6.01, + "learning_rate": 3.0760130839538925e-06, + "loss": 0.003, + "step": 45874 + }, + { + "epoch": 6.01, + "learning_rate": 3.07563045614378e-06, + "loss": 0.002, + "step": 45875 + }, + { + "epoch": 6.01, + "learning_rate": 3.0752478478080837e-06, + "loss": 0.002, + "step": 45876 + }, + { + "epoch": 6.01, + "learning_rate": 3.074865258947888e-06, + "loss": 0.002, + "step": 45877 + }, + { + "epoch": 6.01, + "learning_rate": 3.0744826895642656e-06, + "loss": 0.0013, + "step": 45878 + }, + { + "epoch": 6.01, + "learning_rate": 3.0741001396582904e-06, + "loss": 0.002, + "step": 45879 + }, + { + "epoch": 6.01, + "learning_rate": 3.0737176092310373e-06, + "loss": 0.0015, + "step": 45880 + }, + { + "epoch": 6.01, + "learning_rate": 3.073335098283584e-06, + "loss": 0.0022, + "step": 45881 + }, + { + "epoch": 6.01, + "learning_rate": 3.0729526068170114e-06, + "loss": 0.0021, + "step": 45882 + }, + { + "epoch": 6.01, + "learning_rate": 3.072570134832389e-06, + "loss": 0.0022, + "step": 45883 + }, + { + "epoch": 6.01, + "learning_rate": 3.0721876823307918e-06, + "loss": 0.0016, + "step": 45884 + }, + { + "epoch": 6.01, + "learning_rate": 3.0718052493133012e-06, + "loss": 0.002, + "step": 45885 + }, + { + "epoch": 6.01, + "learning_rate": 3.0714228357809892e-06, + "loss": 0.0015, + "step": 45886 + }, + { + "epoch": 6.01, + "learning_rate": 3.071040441734928e-06, + "loss": 0.0023, + "step": 45887 + }, + { + "epoch": 6.01, + "learning_rate": 3.0706580671762008e-06, + "loss": 0.0033, + "step": 45888 + }, + { + "epoch": 6.01, + "learning_rate": 3.070275712105878e-06, + "loss": 0.0018, + "step": 45889 + }, + { + "epoch": 6.01, + "learning_rate": 3.069893376525036e-06, + "loss": 0.0017, + "step": 45890 + }, + { + "epoch": 6.01, + "learning_rate": 3.0695110604347455e-06, + "loss": 0.0021, + "step": 45891 + }, + { + "epoch": 6.01, + "learning_rate": 3.069128763836088e-06, + "loss": 0.0027, + "step": 45892 + }, + { + "epoch": 6.01, + "learning_rate": 3.068746486730139e-06, + "loss": 0.0018, + "step": 45893 + }, + { + "epoch": 6.01, + "learning_rate": 3.0683642291179704e-06, + "loss": 0.0028, + "step": 45894 + }, + { + "epoch": 6.01, + "learning_rate": 3.067981991000657e-06, + "loss": 0.0018, + "step": 45895 + }, + { + "epoch": 6.01, + "learning_rate": 3.0675997723792773e-06, + "loss": 0.0015, + "step": 45896 + }, + { + "epoch": 6.01, + "learning_rate": 3.0672175732549037e-06, + "loss": 0.002, + "step": 45897 + }, + { + "epoch": 6.01, + "learning_rate": 3.0668353936286123e-06, + "loss": 0.003, + "step": 45898 + }, + { + "epoch": 6.01, + "learning_rate": 3.066453233501473e-06, + "loss": 0.0019, + "step": 45899 + }, + { + "epoch": 6.01, + "learning_rate": 3.0660710928745686e-06, + "loss": 0.0025, + "step": 45900 + }, + { + "epoch": 6.01, + "learning_rate": 3.0656889717489703e-06, + "loss": 0.0022, + "step": 45901 + }, + { + "epoch": 6.01, + "learning_rate": 3.0653068701257482e-06, + "loss": 0.0028, + "step": 45902 + }, + { + "epoch": 6.01, + "learning_rate": 3.064924788005982e-06, + "loss": 0.0028, + "step": 45903 + }, + { + "epoch": 6.01, + "learning_rate": 3.064542725390749e-06, + "loss": 0.0025, + "step": 45904 + }, + { + "epoch": 6.01, + "learning_rate": 3.0641606822811198e-06, + "loss": 0.0012, + "step": 45905 + }, + { + "epoch": 6.01, + "learning_rate": 3.0637786586781694e-06, + "loss": 0.0032, + "step": 45906 + }, + { + "epoch": 6.01, + "learning_rate": 3.0633966545829684e-06, + "loss": 0.0034, + "step": 45907 + }, + { + "epoch": 6.01, + "learning_rate": 3.063014669996599e-06, + "loss": 0.0016, + "step": 45908 + }, + { + "epoch": 6.01, + "learning_rate": 3.062632704920131e-06, + "loss": 0.0019, + "step": 45909 + }, + { + "epoch": 6.01, + "learning_rate": 3.062250759354636e-06, + "loss": 0.0019, + "step": 45910 + }, + { + "epoch": 6.01, + "learning_rate": 3.061868833301195e-06, + "loss": 0.0018, + "step": 45911 + }, + { + "epoch": 6.01, + "learning_rate": 3.061486926760877e-06, + "loss": 0.0027, + "step": 45912 + }, + { + "epoch": 6.01, + "learning_rate": 3.0611050397347553e-06, + "loss": 0.0017, + "step": 45913 + }, + { + "epoch": 6.01, + "learning_rate": 3.0607231722239093e-06, + "loss": 0.0016, + "step": 45914 + }, + { + "epoch": 6.01, + "learning_rate": 3.0603413242294067e-06, + "loss": 0.002, + "step": 45915 + }, + { + "epoch": 6.01, + "learning_rate": 3.059959495752327e-06, + "loss": 0.0013, + "step": 45916 + }, + { + "epoch": 6.01, + "learning_rate": 3.059577686793743e-06, + "loss": 0.0017, + "step": 45917 + }, + { + "epoch": 6.01, + "learning_rate": 3.059195897354723e-06, + "loss": 0.0013, + "step": 45918 + }, + { + "epoch": 6.02, + "learning_rate": 3.058814127436348e-06, + "loss": 0.0021, + "step": 45919 + }, + { + "epoch": 6.02, + "learning_rate": 3.0584323770396897e-06, + "loss": 0.0021, + "step": 45920 + }, + { + "epoch": 6.02, + "learning_rate": 3.0580506461658197e-06, + "loss": 0.0012, + "step": 45921 + }, + { + "epoch": 6.02, + "learning_rate": 3.0576689348158093e-06, + "loss": 0.0032, + "step": 45922 + }, + { + "epoch": 6.02, + "learning_rate": 3.0572872429907397e-06, + "loss": 0.002, + "step": 45923 + }, + { + "epoch": 6.02, + "learning_rate": 3.0569055706916752e-06, + "loss": 0.002, + "step": 45924 + }, + { + "epoch": 6.02, + "learning_rate": 3.0565239179196983e-06, + "loss": 0.0022, + "step": 45925 + }, + { + "epoch": 6.02, + "learning_rate": 3.0561422846758746e-06, + "loss": 0.0024, + "step": 45926 + }, + { + "epoch": 6.02, + "learning_rate": 3.055760670961284e-06, + "loss": 0.0009, + "step": 45927 + }, + { + "epoch": 6.02, + "learning_rate": 3.0553790767769976e-06, + "loss": 0.0025, + "step": 45928 + }, + { + "epoch": 6.02, + "learning_rate": 3.054997502124086e-06, + "loss": 0.0036, + "step": 45929 + }, + { + "epoch": 6.02, + "learning_rate": 3.054615947003623e-06, + "loss": 0.0016, + "step": 45930 + }, + { + "epoch": 6.02, + "learning_rate": 3.0542344114166842e-06, + "loss": 0.0019, + "step": 45931 + }, + { + "epoch": 6.02, + "learning_rate": 3.053852895364342e-06, + "loss": 0.002, + "step": 45932 + }, + { + "epoch": 6.02, + "learning_rate": 3.053471398847665e-06, + "loss": 0.0018, + "step": 45933 + }, + { + "epoch": 6.02, + "learning_rate": 3.0530899218677334e-06, + "loss": 0.002, + "step": 45934 + }, + { + "epoch": 6.02, + "learning_rate": 3.052708464425613e-06, + "loss": 0.003, + "step": 45935 + }, + { + "epoch": 6.02, + "learning_rate": 3.052327026522384e-06, + "loss": 0.0017, + "step": 45936 + }, + { + "epoch": 6.02, + "learning_rate": 3.0519456081591116e-06, + "loss": 0.0018, + "step": 45937 + }, + { + "epoch": 6.02, + "learning_rate": 3.0515642093368745e-06, + "loss": 0.0016, + "step": 45938 + }, + { + "epoch": 6.02, + "learning_rate": 3.0511828300567447e-06, + "loss": 0.0025, + "step": 45939 + }, + { + "epoch": 6.02, + "learning_rate": 3.050801470319792e-06, + "loss": 0.0025, + "step": 45940 + }, + { + "epoch": 6.02, + "learning_rate": 3.0504201301270876e-06, + "loss": 0.0031, + "step": 45941 + }, + { + "epoch": 6.02, + "learning_rate": 3.0500388094797083e-06, + "loss": 0.0021, + "step": 45942 + }, + { + "epoch": 6.02, + "learning_rate": 3.049657508378726e-06, + "loss": 0.0025, + "step": 45943 + }, + { + "epoch": 6.02, + "learning_rate": 3.0492762268252084e-06, + "loss": 0.0034, + "step": 45944 + }, + { + "epoch": 6.02, + "learning_rate": 3.048894964820235e-06, + "loss": 0.0023, + "step": 45945 + }, + { + "epoch": 6.02, + "learning_rate": 3.0485137223648697e-06, + "loss": 0.0014, + "step": 45946 + }, + { + "epoch": 6.02, + "learning_rate": 3.0481324994601933e-06, + "loss": 0.0018, + "step": 45947 + }, + { + "epoch": 6.02, + "learning_rate": 3.047751296107274e-06, + "loss": 0.0033, + "step": 45948 + }, + { + "epoch": 6.02, + "learning_rate": 3.047370112307181e-06, + "loss": 0.0026, + "step": 45949 + }, + { + "epoch": 6.02, + "learning_rate": 3.0469889480609927e-06, + "loss": 0.0018, + "step": 45950 + }, + { + "epoch": 6.02, + "learning_rate": 3.0466078033697775e-06, + "loss": 0.002, + "step": 45951 + }, + { + "epoch": 6.02, + "learning_rate": 3.046226678234604e-06, + "loss": 0.0023, + "step": 45952 + }, + { + "epoch": 6.02, + "learning_rate": 3.0458455726565516e-06, + "loss": 0.002, + "step": 45953 + }, + { + "epoch": 6.02, + "learning_rate": 3.045464486636688e-06, + "loss": 0.0025, + "step": 45954 + }, + { + "epoch": 6.02, + "learning_rate": 3.0450834201760836e-06, + "loss": 0.0022, + "step": 45955 + }, + { + "epoch": 6.02, + "learning_rate": 3.0447023732758107e-06, + "loss": 0.0024, + "step": 45956 + }, + { + "epoch": 6.02, + "learning_rate": 3.0443213459369405e-06, + "loss": 0.0017, + "step": 45957 + }, + { + "epoch": 6.02, + "learning_rate": 3.04394033816055e-06, + "loss": 0.0019, + "step": 45958 + }, + { + "epoch": 6.02, + "learning_rate": 3.043559349947707e-06, + "loss": 0.0014, + "step": 45959 + }, + { + "epoch": 6.02, + "learning_rate": 3.0431783812994797e-06, + "loss": 0.0033, + "step": 45960 + }, + { + "epoch": 6.02, + "learning_rate": 3.042797432216945e-06, + "loss": 0.0019, + "step": 45961 + }, + { + "epoch": 6.02, + "learning_rate": 3.0424165027011733e-06, + "loss": 0.003, + "step": 45962 + }, + { + "epoch": 6.02, + "learning_rate": 3.042035592753234e-06, + "loss": 0.0019, + "step": 45963 + }, + { + "epoch": 6.02, + "learning_rate": 3.041654702374196e-06, + "loss": 0.0024, + "step": 45964 + }, + { + "epoch": 6.02, + "learning_rate": 3.041273831565137e-06, + "loss": 0.0026, + "step": 45965 + }, + { + "epoch": 6.02, + "learning_rate": 3.0408929803271236e-06, + "loss": 0.0016, + "step": 45966 + }, + { + "epoch": 6.02, + "learning_rate": 3.0405121486612265e-06, + "loss": 0.0037, + "step": 45967 + }, + { + "epoch": 6.02, + "learning_rate": 3.040131336568517e-06, + "loss": 0.0025, + "step": 45968 + }, + { + "epoch": 6.02, + "learning_rate": 3.0397505440500718e-06, + "loss": 0.002, + "step": 45969 + }, + { + "epoch": 6.02, + "learning_rate": 3.0393697711069568e-06, + "loss": 0.0018, + "step": 45970 + }, + { + "epoch": 6.02, + "learning_rate": 3.0389890177402436e-06, + "loss": 0.0024, + "step": 45971 + }, + { + "epoch": 6.02, + "learning_rate": 3.0386082839510013e-06, + "loss": 0.0025, + "step": 45972 + }, + { + "epoch": 6.02, + "learning_rate": 3.038227569740304e-06, + "loss": 0.0025, + "step": 45973 + }, + { + "epoch": 6.02, + "learning_rate": 3.0378468751092218e-06, + "loss": 0.001, + "step": 45974 + }, + { + "epoch": 6.02, + "learning_rate": 3.0374662000588204e-06, + "loss": 0.0028, + "step": 45975 + }, + { + "epoch": 6.02, + "learning_rate": 3.037085544590178e-06, + "loss": 0.0023, + "step": 45976 + }, + { + "epoch": 6.02, + "learning_rate": 3.036704908704363e-06, + "loss": 0.0035, + "step": 45977 + }, + { + "epoch": 6.02, + "learning_rate": 3.036324292402444e-06, + "loss": 0.0025, + "step": 45978 + }, + { + "epoch": 6.02, + "learning_rate": 3.0359436956854883e-06, + "loss": 0.0024, + "step": 45979 + }, + { + "epoch": 6.02, + "learning_rate": 3.03556311855457e-06, + "loss": 0.0029, + "step": 45980 + }, + { + "epoch": 6.02, + "learning_rate": 3.035182561010763e-06, + "loss": 0.0021, + "step": 45981 + }, + { + "epoch": 6.02, + "learning_rate": 3.0348020230551346e-06, + "loss": 0.0025, + "step": 45982 + }, + { + "epoch": 6.02, + "learning_rate": 3.0344215046887515e-06, + "loss": 0.0024, + "step": 45983 + }, + { + "epoch": 6.02, + "learning_rate": 3.034041005912689e-06, + "loss": 0.0018, + "step": 45984 + }, + { + "epoch": 6.02, + "learning_rate": 3.033660526728016e-06, + "loss": 0.0025, + "step": 45985 + }, + { + "epoch": 6.02, + "learning_rate": 3.033280067135802e-06, + "loss": 0.0017, + "step": 45986 + }, + { + "epoch": 6.02, + "learning_rate": 3.0328996271371137e-06, + "loss": 0.0031, + "step": 45987 + }, + { + "epoch": 6.02, + "learning_rate": 3.0325192067330267e-06, + "loss": 0.0018, + "step": 45988 + }, + { + "epoch": 6.02, + "learning_rate": 3.032138805924609e-06, + "loss": 0.0023, + "step": 45989 + }, + { + "epoch": 6.02, + "learning_rate": 3.0317584247129263e-06, + "loss": 0.0034, + "step": 45990 + }, + { + "epoch": 6.02, + "learning_rate": 3.031378063099052e-06, + "loss": 0.0027, + "step": 45991 + }, + { + "epoch": 6.02, + "learning_rate": 3.030997721084059e-06, + "loss": 0.0021, + "step": 45992 + }, + { + "epoch": 6.02, + "learning_rate": 3.0306173986690134e-06, + "loss": 0.0018, + "step": 45993 + }, + { + "epoch": 6.02, + "learning_rate": 3.030237095854983e-06, + "loss": 0.0028, + "step": 45994 + }, + { + "epoch": 6.03, + "learning_rate": 3.0298568126430416e-06, + "loss": 0.0028, + "step": 45995 + }, + { + "epoch": 6.03, + "learning_rate": 3.029476549034258e-06, + "loss": 0.0037, + "step": 45996 + }, + { + "epoch": 6.03, + "learning_rate": 3.0290963050296995e-06, + "loss": 0.0016, + "step": 45997 + }, + { + "epoch": 6.03, + "learning_rate": 3.0287160806304337e-06, + "loss": 0.0017, + "step": 45998 + }, + { + "epoch": 6.03, + "learning_rate": 3.0283358758375357e-06, + "loss": 0.0023, + "step": 45999 + }, + { + "epoch": 6.03, + "learning_rate": 3.0279556906520714e-06, + "loss": 0.0019, + "step": 46000 + }, + { + "epoch": 6.03, + "learning_rate": 3.027575525075108e-06, + "loss": 0.0024, + "step": 46001 + }, + { + "epoch": 6.03, + "learning_rate": 3.027195379107717e-06, + "loss": 0.0016, + "step": 46002 + }, + { + "epoch": 6.03, + "learning_rate": 3.0268152527509708e-06, + "loss": 0.0022, + "step": 46003 + }, + { + "epoch": 6.03, + "learning_rate": 3.026435146005935e-06, + "loss": 0.0018, + "step": 46004 + }, + { + "epoch": 6.03, + "learning_rate": 3.0260550588736793e-06, + "loss": 0.0034, + "step": 46005 + }, + { + "epoch": 6.03, + "learning_rate": 3.0256749913552686e-06, + "loss": 0.0017, + "step": 46006 + }, + { + "epoch": 6.03, + "learning_rate": 3.02529494345178e-06, + "loss": 0.0021, + "step": 46007 + }, + { + "epoch": 6.03, + "learning_rate": 3.0249149151642766e-06, + "loss": 0.0019, + "step": 46008 + }, + { + "epoch": 6.03, + "learning_rate": 3.0245349064938255e-06, + "loss": 0.0024, + "step": 46009 + }, + { + "epoch": 6.03, + "learning_rate": 3.024154917441502e-06, + "loss": 0.0029, + "step": 46010 + }, + { + "epoch": 6.03, + "learning_rate": 3.0237749480083724e-06, + "loss": 0.0021, + "step": 46011 + }, + { + "epoch": 6.03, + "learning_rate": 3.023394998195499e-06, + "loss": 0.0023, + "step": 46012 + }, + { + "epoch": 6.03, + "learning_rate": 3.0230150680039604e-06, + "loss": 0.0019, + "step": 46013 + }, + { + "epoch": 6.03, + "learning_rate": 3.0226351574348168e-06, + "loss": 0.0022, + "step": 46014 + }, + { + "epoch": 6.03, + "learning_rate": 3.0222552664891437e-06, + "loss": 0.0024, + "step": 46015 + }, + { + "epoch": 6.03, + "learning_rate": 3.0218753951680056e-06, + "loss": 0.0019, + "step": 46016 + }, + { + "epoch": 6.03, + "learning_rate": 3.021495543472468e-06, + "loss": 0.0019, + "step": 46017 + }, + { + "epoch": 6.03, + "learning_rate": 3.0211157114036062e-06, + "loss": 0.0023, + "step": 46018 + }, + { + "epoch": 6.03, + "learning_rate": 3.0207358989624845e-06, + "loss": 0.0023, + "step": 46019 + }, + { + "epoch": 6.03, + "learning_rate": 3.020356106150172e-06, + "loss": 0.0016, + "step": 46020 + }, + { + "epoch": 6.03, + "learning_rate": 3.019976332967732e-06, + "loss": 0.0023, + "step": 46021 + }, + { + "epoch": 6.03, + "learning_rate": 3.0195965794162407e-06, + "loss": 0.0017, + "step": 46022 + }, + { + "epoch": 6.03, + "learning_rate": 3.0192168454967597e-06, + "loss": 0.0019, + "step": 46023 + }, + { + "epoch": 6.03, + "learning_rate": 3.0188371312103614e-06, + "loss": 0.0011, + "step": 46024 + }, + { + "epoch": 6.03, + "learning_rate": 3.0184574365581098e-06, + "loss": 0.0022, + "step": 46025 + }, + { + "epoch": 6.03, + "learning_rate": 3.018077761541077e-06, + "loss": 0.0036, + "step": 46026 + }, + { + "epoch": 6.03, + "learning_rate": 3.0176981061603307e-06, + "loss": 0.0013, + "step": 46027 + }, + { + "epoch": 6.03, + "learning_rate": 3.0173184704169344e-06, + "loss": 0.0022, + "step": 46028 + }, + { + "epoch": 6.03, + "learning_rate": 3.016938854311956e-06, + "loss": 0.0019, + "step": 46029 + }, + { + "epoch": 6.03, + "learning_rate": 3.0165592578464688e-06, + "loss": 0.0022, + "step": 46030 + }, + { + "epoch": 6.03, + "learning_rate": 3.016179681021536e-06, + "loss": 0.0015, + "step": 46031 + }, + { + "epoch": 6.03, + "learning_rate": 3.0158001238382227e-06, + "loss": 0.0034, + "step": 46032 + }, + { + "epoch": 6.03, + "learning_rate": 3.0154205862976007e-06, + "loss": 0.0033, + "step": 46033 + }, + { + "epoch": 6.03, + "learning_rate": 3.015041068400738e-06, + "loss": 0.0021, + "step": 46034 + }, + { + "epoch": 6.03, + "learning_rate": 3.0146615701487016e-06, + "loss": 0.003, + "step": 46035 + }, + { + "epoch": 6.03, + "learning_rate": 3.0142820915425564e-06, + "loss": 0.0026, + "step": 46036 + }, + { + "epoch": 6.03, + "learning_rate": 3.013902632583369e-06, + "loss": 0.0013, + "step": 46037 + }, + { + "epoch": 6.03, + "learning_rate": 3.01352319327221e-06, + "loss": 0.0027, + "step": 46038 + }, + { + "epoch": 6.03, + "learning_rate": 3.0131437736101452e-06, + "loss": 0.0014, + "step": 46039 + }, + { + "epoch": 6.03, + "learning_rate": 3.012764373598239e-06, + "loss": 0.0025, + "step": 46040 + }, + { + "epoch": 6.03, + "learning_rate": 3.012384993237565e-06, + "loss": 0.0027, + "step": 46041 + }, + { + "epoch": 6.03, + "learning_rate": 3.0120056325291845e-06, + "loss": 0.0023, + "step": 46042 + }, + { + "epoch": 6.03, + "learning_rate": 3.011626291474167e-06, + "loss": 0.0018, + "step": 46043 + }, + { + "epoch": 6.03, + "learning_rate": 3.011246970073571e-06, + "loss": 0.002, + "step": 46044 + }, + { + "epoch": 6.03, + "learning_rate": 3.0108676683284775e-06, + "loss": 0.0015, + "step": 46045 + }, + { + "epoch": 6.03, + "learning_rate": 3.010488386239947e-06, + "loss": 0.0013, + "step": 46046 + }, + { + "epoch": 6.03, + "learning_rate": 3.010109123809045e-06, + "loss": 0.0029, + "step": 46047 + }, + { + "epoch": 6.03, + "learning_rate": 3.009729881036836e-06, + "loss": 0.0016, + "step": 46048 + }, + { + "epoch": 6.03, + "learning_rate": 3.0093506579243925e-06, + "loss": 0.003, + "step": 46049 + }, + { + "epoch": 6.03, + "learning_rate": 3.008971454472779e-06, + "loss": 0.0037, + "step": 46050 + }, + { + "epoch": 6.03, + "learning_rate": 3.0085922706830596e-06, + "loss": 0.0027, + "step": 46051 + }, + { + "epoch": 6.03, + "learning_rate": 3.0082131065563005e-06, + "loss": 0.002, + "step": 46052 + }, + { + "epoch": 6.03, + "learning_rate": 3.0078339620935713e-06, + "loss": 0.0017, + "step": 46053 + }, + { + "epoch": 6.03, + "learning_rate": 3.0074548372959368e-06, + "loss": 0.0027, + "step": 46054 + }, + { + "epoch": 6.03, + "learning_rate": 3.0070757321644604e-06, + "loss": 0.0045, + "step": 46055 + }, + { + "epoch": 6.03, + "learning_rate": 3.006696646700211e-06, + "loss": 0.0017, + "step": 46056 + }, + { + "epoch": 6.03, + "learning_rate": 3.0063175809042587e-06, + "loss": 0.002, + "step": 46057 + }, + { + "epoch": 6.03, + "learning_rate": 3.005938534777666e-06, + "loss": 0.0014, + "step": 46058 + }, + { + "epoch": 6.03, + "learning_rate": 3.005559508321495e-06, + "loss": 0.002, + "step": 46059 + }, + { + "epoch": 6.03, + "learning_rate": 3.0051805015368185e-06, + "loss": 0.0024, + "step": 46060 + }, + { + "epoch": 6.03, + "learning_rate": 3.0048015144246987e-06, + "loss": 0.0022, + "step": 46061 + }, + { + "epoch": 6.03, + "learning_rate": 3.0044225469862032e-06, + "loss": 0.002, + "step": 46062 + }, + { + "epoch": 6.03, + "learning_rate": 3.0040435992223926e-06, + "loss": 0.0023, + "step": 46063 + }, + { + "epoch": 6.03, + "learning_rate": 3.003664671134341e-06, + "loss": 0.0024, + "step": 46064 + }, + { + "epoch": 6.03, + "learning_rate": 3.0032857627231094e-06, + "loss": 0.0019, + "step": 46065 + }, + { + "epoch": 6.03, + "learning_rate": 3.0029068739897605e-06, + "loss": 0.0016, + "step": 46066 + }, + { + "epoch": 6.03, + "learning_rate": 3.0025280049353645e-06, + "loss": 0.0023, + "step": 46067 + }, + { + "epoch": 6.03, + "learning_rate": 3.0021491555609884e-06, + "loss": 0.0027, + "step": 46068 + }, + { + "epoch": 6.03, + "learning_rate": 3.0017703258676947e-06, + "loss": 0.0015, + "step": 46069 + }, + { + "epoch": 6.03, + "learning_rate": 3.0013915158565498e-06, + "loss": 0.0011, + "step": 46070 + }, + { + "epoch": 6.03, + "learning_rate": 3.0010127255286147e-06, + "loss": 0.0024, + "step": 46071 + }, + { + "epoch": 6.04, + "learning_rate": 3.000633954884963e-06, + "loss": 0.0015, + "step": 46072 + }, + { + "epoch": 6.04, + "learning_rate": 3.000255203926654e-06, + "loss": 0.0029, + "step": 46073 + }, + { + "epoch": 6.04, + "learning_rate": 2.9998764726547523e-06, + "loss": 0.0017, + "step": 46074 + }, + { + "epoch": 6.04, + "learning_rate": 2.999497761070328e-06, + "loss": 0.0026, + "step": 46075 + }, + { + "epoch": 6.04, + "learning_rate": 2.999119069174444e-06, + "loss": 0.0018, + "step": 46076 + }, + { + "epoch": 6.04, + "learning_rate": 2.998740396968162e-06, + "loss": 0.0024, + "step": 46077 + }, + { + "epoch": 6.04, + "learning_rate": 2.9983617444525516e-06, + "loss": 0.0029, + "step": 46078 + }, + { + "epoch": 6.04, + "learning_rate": 2.997983111628674e-06, + "loss": 0.0039, + "step": 46079 + }, + { + "epoch": 6.04, + "learning_rate": 2.9976044984975983e-06, + "loss": 0.0019, + "step": 46080 + }, + { + "epoch": 6.04, + "learning_rate": 2.997225905060387e-06, + "loss": 0.0022, + "step": 46081 + }, + { + "epoch": 6.04, + "learning_rate": 2.9968473313181025e-06, + "loss": 0.0016, + "step": 46082 + }, + { + "epoch": 6.04, + "learning_rate": 2.9964687772718147e-06, + "loss": 0.0022, + "step": 46083 + }, + { + "epoch": 6.04, + "learning_rate": 2.9960902429225857e-06, + "loss": 0.001, + "step": 46084 + }, + { + "epoch": 6.04, + "learning_rate": 2.9957117282714797e-06, + "loss": 0.0025, + "step": 46085 + }, + { + "epoch": 6.04, + "learning_rate": 2.9953332333195585e-06, + "loss": 0.0009, + "step": 46086 + }, + { + "epoch": 6.04, + "learning_rate": 2.9949547580678928e-06, + "loss": 0.0015, + "step": 46087 + }, + { + "epoch": 6.04, + "learning_rate": 2.99457630251754e-06, + "loss": 0.0024, + "step": 46088 + }, + { + "epoch": 6.04, + "learning_rate": 2.9941978666695725e-06, + "loss": 0.0016, + "step": 46089 + }, + { + "epoch": 6.04, + "learning_rate": 2.993819450525046e-06, + "loss": 0.0023, + "step": 46090 + }, + { + "epoch": 6.04, + "learning_rate": 2.9934410540850333e-06, + "loss": 0.0025, + "step": 46091 + }, + { + "epoch": 6.04, + "learning_rate": 2.9930626773505944e-06, + "loss": 0.0015, + "step": 46092 + }, + { + "epoch": 6.04, + "learning_rate": 2.992684320322793e-06, + "loss": 0.0031, + "step": 46093 + }, + { + "epoch": 6.04, + "learning_rate": 2.9923059830026914e-06, + "loss": 0.0017, + "step": 46094 + }, + { + "epoch": 6.04, + "learning_rate": 2.991927665391359e-06, + "loss": 0.0022, + "step": 46095 + }, + { + "epoch": 6.04, + "learning_rate": 2.9915493674898566e-06, + "loss": 0.0024, + "step": 46096 + }, + { + "epoch": 6.04, + "learning_rate": 2.991171089299245e-06, + "loss": 0.0023, + "step": 46097 + }, + { + "epoch": 6.04, + "learning_rate": 2.9907928308205948e-06, + "loss": 0.0033, + "step": 46098 + }, + { + "epoch": 6.04, + "learning_rate": 2.990414592054964e-06, + "loss": 0.0019, + "step": 46099 + }, + { + "epoch": 6.04, + "learning_rate": 2.990036373003421e-06, + "loss": 0.0016, + "step": 46100 + }, + { + "epoch": 6.04, + "learning_rate": 2.9896581736670284e-06, + "loss": 0.0025, + "step": 46101 + }, + { + "epoch": 6.04, + "learning_rate": 2.989279994046845e-06, + "loss": 0.0027, + "step": 46102 + }, + { + "epoch": 6.04, + "learning_rate": 2.9889018341439414e-06, + "loss": 0.0027, + "step": 46103 + }, + { + "epoch": 6.04, + "learning_rate": 2.988523693959379e-06, + "loss": 0.0018, + "step": 46104 + }, + { + "epoch": 6.04, + "learning_rate": 2.9881455734942156e-06, + "loss": 0.0028, + "step": 46105 + }, + { + "epoch": 6.04, + "learning_rate": 2.987767472749523e-06, + "loss": 0.0027, + "step": 46106 + }, + { + "epoch": 6.04, + "learning_rate": 2.987389391726362e-06, + "loss": 0.0016, + "step": 46107 + }, + { + "epoch": 6.04, + "learning_rate": 2.9870113304257943e-06, + "loss": 0.0028, + "step": 46108 + }, + { + "epoch": 6.04, + "learning_rate": 2.9866332888488803e-06, + "loss": 0.0016, + "step": 46109 + }, + { + "epoch": 6.04, + "learning_rate": 2.986255266996687e-06, + "loss": 0.002, + "step": 46110 + }, + { + "epoch": 6.04, + "learning_rate": 2.9858772648702804e-06, + "loss": 0.0029, + "step": 46111 + }, + { + "epoch": 6.04, + "learning_rate": 2.9854992824707205e-06, + "loss": 0.0027, + "step": 46112 + }, + { + "epoch": 6.04, + "learning_rate": 2.985121319799067e-06, + "loss": 0.0027, + "step": 46113 + }, + { + "epoch": 6.04, + "learning_rate": 2.9847433768563895e-06, + "loss": 0.0013, + "step": 46114 + }, + { + "epoch": 6.04, + "learning_rate": 2.9843654536437473e-06, + "loss": 0.0028, + "step": 46115 + }, + { + "epoch": 6.04, + "learning_rate": 2.983987550162204e-06, + "loss": 0.002, + "step": 46116 + }, + { + "epoch": 6.04, + "learning_rate": 2.983609666412819e-06, + "loss": 0.0017, + "step": 46117 + }, + { + "epoch": 6.04, + "learning_rate": 2.983231802396661e-06, + "loss": 0.0018, + "step": 46118 + }, + { + "epoch": 6.04, + "learning_rate": 2.9828539581147907e-06, + "loss": 0.0023, + "step": 46119 + }, + { + "epoch": 6.04, + "learning_rate": 2.9824761335682663e-06, + "loss": 0.0012, + "step": 46120 + }, + { + "epoch": 6.04, + "learning_rate": 2.982098328758154e-06, + "loss": 0.0023, + "step": 46121 + }, + { + "epoch": 6.04, + "learning_rate": 2.981720543685519e-06, + "loss": 0.0023, + "step": 46122 + }, + { + "epoch": 6.04, + "learning_rate": 2.9813427783514226e-06, + "loss": 0.0032, + "step": 46123 + }, + { + "epoch": 6.04, + "learning_rate": 2.980965032756922e-06, + "loss": 0.0021, + "step": 46124 + }, + { + "epoch": 6.04, + "learning_rate": 2.9805873069030864e-06, + "loss": 0.0023, + "step": 46125 + }, + { + "epoch": 6.04, + "learning_rate": 2.9802096007909754e-06, + "loss": 0.0011, + "step": 46126 + }, + { + "epoch": 6.04, + "learning_rate": 2.9798319144216502e-06, + "loss": 0.0021, + "step": 46127 + }, + { + "epoch": 6.04, + "learning_rate": 2.9794542477961718e-06, + "loss": 0.0025, + "step": 46128 + }, + { + "epoch": 6.04, + "learning_rate": 2.979076600915608e-06, + "loss": 0.0025, + "step": 46129 + }, + { + "epoch": 6.04, + "learning_rate": 2.978698973781017e-06, + "loss": 0.0018, + "step": 46130 + }, + { + "epoch": 6.04, + "learning_rate": 2.9783213663934605e-06, + "loss": 0.0023, + "step": 46131 + }, + { + "epoch": 6.04, + "learning_rate": 2.9779437787539954e-06, + "loss": 0.0016, + "step": 46132 + }, + { + "epoch": 6.04, + "learning_rate": 2.977566210863696e-06, + "loss": 0.0026, + "step": 46133 + }, + { + "epoch": 6.04, + "learning_rate": 2.9771886627236167e-06, + "loss": 0.0026, + "step": 46134 + }, + { + "epoch": 6.04, + "learning_rate": 2.9768111343348215e-06, + "loss": 0.0023, + "step": 46135 + }, + { + "epoch": 6.04, + "learning_rate": 2.976433625698367e-06, + "loss": 0.0018, + "step": 46136 + }, + { + "epoch": 6.04, + "learning_rate": 2.976056136815323e-06, + "loss": 0.0023, + "step": 46137 + }, + { + "epoch": 6.04, + "learning_rate": 2.9756786676867455e-06, + "loss": 0.002, + "step": 46138 + }, + { + "epoch": 6.04, + "learning_rate": 2.9753012183136966e-06, + "loss": 0.0018, + "step": 46139 + }, + { + "epoch": 6.04, + "learning_rate": 2.9749237886972402e-06, + "loss": 0.0029, + "step": 46140 + }, + { + "epoch": 6.04, + "learning_rate": 2.9745463788384376e-06, + "loss": 0.0029, + "step": 46141 + }, + { + "epoch": 6.04, + "learning_rate": 2.9741689887383484e-06, + "loss": 0.0011, + "step": 46142 + }, + { + "epoch": 6.04, + "learning_rate": 2.973791618398033e-06, + "loss": 0.0024, + "step": 46143 + }, + { + "epoch": 6.04, + "learning_rate": 2.973414267818553e-06, + "loss": 0.001, + "step": 46144 + }, + { + "epoch": 6.04, + "learning_rate": 2.9730369370009747e-06, + "loss": 0.0016, + "step": 46145 + }, + { + "epoch": 6.04, + "learning_rate": 2.9726596259463568e-06, + "loss": 0.0025, + "step": 46146 + }, + { + "epoch": 6.04, + "learning_rate": 2.972282334655755e-06, + "loss": 0.001, + "step": 46147 + }, + { + "epoch": 6.05, + "learning_rate": 2.9719050631302395e-06, + "loss": 0.0024, + "step": 46148 + }, + { + "epoch": 6.05, + "learning_rate": 2.9715278113708658e-06, + "loss": 0.0028, + "step": 46149 + }, + { + "epoch": 6.05, + "learning_rate": 2.9711505793786953e-06, + "loss": 0.0023, + "step": 46150 + }, + { + "epoch": 6.05, + "learning_rate": 2.970773367154787e-06, + "loss": 0.0025, + "step": 46151 + }, + { + "epoch": 6.05, + "learning_rate": 2.9703961747002073e-06, + "loss": 0.0038, + "step": 46152 + }, + { + "epoch": 6.05, + "learning_rate": 2.9700190020160145e-06, + "loss": 0.0024, + "step": 46153 + }, + { + "epoch": 6.05, + "learning_rate": 2.969641849103265e-06, + "loss": 0.0019, + "step": 46154 + }, + { + "epoch": 6.05, + "learning_rate": 2.969264715963024e-06, + "loss": 0.0021, + "step": 46155 + }, + { + "epoch": 6.05, + "learning_rate": 2.968887602596354e-06, + "loss": 0.0026, + "step": 46156 + }, + { + "epoch": 6.05, + "learning_rate": 2.9685105090043143e-06, + "loss": 0.004, + "step": 46157 + }, + { + "epoch": 6.05, + "learning_rate": 2.9681334351879632e-06, + "loss": 0.0023, + "step": 46158 + }, + { + "epoch": 6.05, + "learning_rate": 2.96775638114836e-06, + "loss": 0.0018, + "step": 46159 + }, + { + "epoch": 6.05, + "learning_rate": 2.9673793468865707e-06, + "loss": 0.0022, + "step": 46160 + }, + { + "epoch": 6.05, + "learning_rate": 2.9670023324036524e-06, + "loss": 0.0016, + "step": 46161 + }, + { + "epoch": 6.05, + "learning_rate": 2.9666253377006614e-06, + "loss": 0.0017, + "step": 46162 + }, + { + "epoch": 6.05, + "learning_rate": 2.9662483627786663e-06, + "loss": 0.002, + "step": 46163 + }, + { + "epoch": 6.05, + "learning_rate": 2.9658714076387196e-06, + "loss": 0.0012, + "step": 46164 + }, + { + "epoch": 6.05, + "learning_rate": 2.965494472281888e-06, + "loss": 0.0021, + "step": 46165 + }, + { + "epoch": 6.05, + "learning_rate": 2.9651175567092283e-06, + "loss": 0.0025, + "step": 46166 + }, + { + "epoch": 6.05, + "learning_rate": 2.9647406609217987e-06, + "loss": 0.0019, + "step": 46167 + }, + { + "epoch": 6.05, + "learning_rate": 2.964363784920663e-06, + "loss": 0.0028, + "step": 46168 + }, + { + "epoch": 6.05, + "learning_rate": 2.963986928706881e-06, + "loss": 0.0016, + "step": 46169 + }, + { + "epoch": 6.05, + "learning_rate": 2.963610092281507e-06, + "loss": 0.0026, + "step": 46170 + }, + { + "epoch": 6.05, + "learning_rate": 2.9632332756456085e-06, + "loss": 0.0033, + "step": 46171 + }, + { + "epoch": 6.05, + "learning_rate": 2.962856478800241e-06, + "loss": 0.0015, + "step": 46172 + }, + { + "epoch": 6.05, + "learning_rate": 2.9624797017464655e-06, + "loss": 0.0017, + "step": 46173 + }, + { + "epoch": 6.05, + "learning_rate": 2.962102944485339e-06, + "loss": 0.0028, + "step": 46174 + }, + { + "epoch": 6.05, + "learning_rate": 2.961726207017921e-06, + "loss": 0.0016, + "step": 46175 + }, + { + "epoch": 6.05, + "learning_rate": 2.9613494893452775e-06, + "loss": 0.0027, + "step": 46176 + }, + { + "epoch": 6.05, + "learning_rate": 2.9609727914684636e-06, + "loss": 0.0019, + "step": 46177 + }, + { + "epoch": 6.05, + "learning_rate": 2.9605961133885364e-06, + "loss": 0.0021, + "step": 46178 + }, + { + "epoch": 6.05, + "learning_rate": 2.96021945510656e-06, + "loss": 0.0015, + "step": 46179 + }, + { + "epoch": 6.05, + "learning_rate": 2.9598428166235915e-06, + "loss": 0.0024, + "step": 46180 + }, + { + "epoch": 6.05, + "learning_rate": 2.95946619794069e-06, + "loss": 0.0024, + "step": 46181 + }, + { + "epoch": 6.05, + "learning_rate": 2.959089599058913e-06, + "loss": 0.0009, + "step": 46182 + }, + { + "epoch": 6.05, + "learning_rate": 2.9587130199793234e-06, + "loss": 0.0037, + "step": 46183 + }, + { + "epoch": 6.05, + "learning_rate": 2.9583364607029796e-06, + "loss": 0.0017, + "step": 46184 + }, + { + "epoch": 6.05, + "learning_rate": 2.957959921230935e-06, + "loss": 0.0009, + "step": 46185 + }, + { + "epoch": 6.05, + "learning_rate": 2.9575834015642545e-06, + "loss": 0.0013, + "step": 46186 + }, + { + "epoch": 6.05, + "learning_rate": 2.9572069017039985e-06, + "loss": 0.0028, + "step": 46187 + }, + { + "epoch": 6.05, + "learning_rate": 2.956830421651222e-06, + "loss": 0.0022, + "step": 46188 + }, + { + "epoch": 6.05, + "learning_rate": 2.956453961406982e-06, + "loss": 0.0036, + "step": 46189 + }, + { + "epoch": 6.05, + "learning_rate": 2.9560775209723436e-06, + "loss": 0.0029, + "step": 46190 + }, + { + "epoch": 6.05, + "learning_rate": 2.9557011003483627e-06, + "loss": 0.0017, + "step": 46191 + }, + { + "epoch": 6.05, + "learning_rate": 2.9553246995360963e-06, + "loss": 0.0033, + "step": 46192 + }, + { + "epoch": 6.05, + "learning_rate": 2.9549483185366013e-06, + "loss": 0.0016, + "step": 46193 + }, + { + "epoch": 6.05, + "learning_rate": 2.9545719573509426e-06, + "loss": 0.0012, + "step": 46194 + }, + { + "epoch": 6.05, + "learning_rate": 2.954195615980174e-06, + "loss": 0.001, + "step": 46195 + }, + { + "epoch": 6.05, + "learning_rate": 2.953819294425356e-06, + "loss": 0.0054, + "step": 46196 + }, + { + "epoch": 6.05, + "learning_rate": 2.9534429926875397e-06, + "loss": 0.0019, + "step": 46197 + }, + { + "epoch": 6.05, + "learning_rate": 2.953066710767796e-06, + "loss": 0.0017, + "step": 46198 + }, + { + "epoch": 6.05, + "learning_rate": 2.952690448667176e-06, + "loss": 0.0018, + "step": 46199 + }, + { + "epoch": 6.05, + "learning_rate": 2.9523142063867393e-06, + "loss": 0.0014, + "step": 46200 + }, + { + "epoch": 6.05, + "learning_rate": 2.9519379839275408e-06, + "loss": 0.0025, + "step": 46201 + }, + { + "epoch": 6.05, + "learning_rate": 2.9515617812906437e-06, + "loss": 0.0014, + "step": 46202 + }, + { + "epoch": 6.05, + "learning_rate": 2.9511855984771056e-06, + "loss": 0.0021, + "step": 46203 + }, + { + "epoch": 6.05, + "learning_rate": 2.950809435487978e-06, + "loss": 0.0028, + "step": 46204 + }, + { + "epoch": 6.05, + "learning_rate": 2.950433292324326e-06, + "loss": 0.003, + "step": 46205 + }, + { + "epoch": 6.05, + "learning_rate": 2.950057168987206e-06, + "loss": 0.0026, + "step": 46206 + }, + { + "epoch": 6.05, + "learning_rate": 2.9496810654776755e-06, + "loss": 0.0018, + "step": 46207 + }, + { + "epoch": 6.05, + "learning_rate": 2.9493049817967866e-06, + "loss": 0.0025, + "step": 46208 + }, + { + "epoch": 6.05, + "learning_rate": 2.9489289179456026e-06, + "loss": 0.0025, + "step": 46209 + }, + { + "epoch": 6.05, + "learning_rate": 2.9485528739251845e-06, + "loss": 0.0024, + "step": 46210 + }, + { + "epoch": 6.05, + "learning_rate": 2.9481768497365847e-06, + "loss": 0.0028, + "step": 46211 + }, + { + "epoch": 6.05, + "learning_rate": 2.9478008453808603e-06, + "loss": 0.0021, + "step": 46212 + }, + { + "epoch": 6.05, + "learning_rate": 2.9474248608590726e-06, + "loss": 0.0019, + "step": 46213 + }, + { + "epoch": 6.05, + "learning_rate": 2.9470488961722766e-06, + "loss": 0.0018, + "step": 46214 + }, + { + "epoch": 6.05, + "learning_rate": 2.9466729513215307e-06, + "loss": 0.0032, + "step": 46215 + }, + { + "epoch": 6.05, + "learning_rate": 2.946297026307888e-06, + "loss": 0.0026, + "step": 46216 + }, + { + "epoch": 6.05, + "learning_rate": 2.9459211211324136e-06, + "loss": 0.0022, + "step": 46217 + }, + { + "epoch": 6.05, + "learning_rate": 2.9455452357961588e-06, + "loss": 0.0018, + "step": 46218 + }, + { + "epoch": 6.05, + "learning_rate": 2.9451693703001794e-06, + "loss": 0.0015, + "step": 46219 + }, + { + "epoch": 6.05, + "learning_rate": 2.944793524645536e-06, + "loss": 0.0025, + "step": 46220 + }, + { + "epoch": 6.05, + "learning_rate": 2.9444176988332883e-06, + "loss": 0.0035, + "step": 46221 + }, + { + "epoch": 6.05, + "learning_rate": 2.94404189286449e-06, + "loss": 0.0032, + "step": 46222 + }, + { + "epoch": 6.05, + "learning_rate": 2.943666106740197e-06, + "loss": 0.0022, + "step": 46223 + }, + { + "epoch": 6.06, + "learning_rate": 2.9432903404614656e-06, + "loss": 0.0029, + "step": 46224 + }, + { + "epoch": 6.06, + "learning_rate": 2.942914594029357e-06, + "loss": 0.0013, + "step": 46225 + }, + { + "epoch": 6.06, + "learning_rate": 2.9425388674449263e-06, + "loss": 0.0015, + "step": 46226 + }, + { + "epoch": 6.06, + "learning_rate": 2.942163160709225e-06, + "loss": 0.0015, + "step": 46227 + }, + { + "epoch": 6.06, + "learning_rate": 2.9417874738233166e-06, + "loss": 0.0018, + "step": 46228 + }, + { + "epoch": 6.06, + "learning_rate": 2.941411806788256e-06, + "loss": 0.0013, + "step": 46229 + }, + { + "epoch": 6.06, + "learning_rate": 2.941036159605095e-06, + "loss": 0.0039, + "step": 46230 + }, + { + "epoch": 6.06, + "learning_rate": 2.940660532274897e-06, + "loss": 0.0021, + "step": 46231 + }, + { + "epoch": 6.06, + "learning_rate": 2.940284924798713e-06, + "loss": 0.0019, + "step": 46232 + }, + { + "epoch": 6.06, + "learning_rate": 2.9399093371776045e-06, + "loss": 0.0025, + "step": 46233 + }, + { + "epoch": 6.06, + "learning_rate": 2.939533769412625e-06, + "loss": 0.0026, + "step": 46234 + }, + { + "epoch": 6.06, + "learning_rate": 2.9391582215048275e-06, + "loss": 0.0013, + "step": 46235 + }, + { + "epoch": 6.06, + "learning_rate": 2.938782693455274e-06, + "loss": 0.0017, + "step": 46236 + }, + { + "epoch": 6.06, + "learning_rate": 2.9384071852650186e-06, + "loss": 0.0029, + "step": 46237 + }, + { + "epoch": 6.06, + "learning_rate": 2.9380316969351175e-06, + "loss": 0.0016, + "step": 46238 + }, + { + "epoch": 6.06, + "learning_rate": 2.937656228466623e-06, + "loss": 0.0027, + "step": 46239 + }, + { + "epoch": 6.06, + "learning_rate": 2.937280779860596e-06, + "loss": 0.0036, + "step": 46240 + }, + { + "epoch": 6.06, + "learning_rate": 2.936905351118089e-06, + "loss": 0.0015, + "step": 46241 + }, + { + "epoch": 6.06, + "learning_rate": 2.936529942240163e-06, + "loss": 0.0021, + "step": 46242 + }, + { + "epoch": 6.06, + "learning_rate": 2.936154553227867e-06, + "loss": 0.0013, + "step": 46243 + }, + { + "epoch": 6.06, + "learning_rate": 2.9357791840822624e-06, + "loss": 0.0029, + "step": 46244 + }, + { + "epoch": 6.06, + "learning_rate": 2.9354038348044035e-06, + "loss": 0.0023, + "step": 46245 + }, + { + "epoch": 6.06, + "learning_rate": 2.9350285053953444e-06, + "loss": 0.0026, + "step": 46246 + }, + { + "epoch": 6.06, + "learning_rate": 2.9346531958561387e-06, + "loss": 0.0016, + "step": 46247 + }, + { + "epoch": 6.06, + "learning_rate": 2.9342779061878466e-06, + "loss": 0.0021, + "step": 46248 + }, + { + "epoch": 6.06, + "learning_rate": 2.9339026363915234e-06, + "loss": 0.0019, + "step": 46249 + }, + { + "epoch": 6.06, + "learning_rate": 2.933527386468219e-06, + "loss": 0.0022, + "step": 46250 + }, + { + "epoch": 6.06, + "learning_rate": 2.9331521564189957e-06, + "loss": 0.0013, + "step": 46251 + }, + { + "epoch": 6.06, + "learning_rate": 2.932776946244902e-06, + "loss": 0.0023, + "step": 46252 + }, + { + "epoch": 6.06, + "learning_rate": 2.9324017559469996e-06, + "loss": 0.0019, + "step": 46253 + }, + { + "epoch": 6.06, + "learning_rate": 2.9320265855263385e-06, + "loss": 0.0024, + "step": 46254 + }, + { + "epoch": 6.06, + "learning_rate": 2.9316514349839796e-06, + "loss": 0.0026, + "step": 46255 + }, + { + "epoch": 6.06, + "learning_rate": 2.9312763043209745e-06, + "loss": 0.0014, + "step": 46256 + }, + { + "epoch": 6.06, + "learning_rate": 2.930901193538378e-06, + "loss": 0.002, + "step": 46257 + }, + { + "epoch": 6.06, + "learning_rate": 2.930526102637242e-06, + "loss": 0.0011, + "step": 46258 + }, + { + "epoch": 6.06, + "learning_rate": 2.9301510316186287e-06, + "loss": 0.0019, + "step": 46259 + }, + { + "epoch": 6.06, + "learning_rate": 2.929775980483589e-06, + "loss": 0.0016, + "step": 46260 + }, + { + "epoch": 6.06, + "learning_rate": 2.9294009492331744e-06, + "loss": 0.0019, + "step": 46261 + }, + { + "epoch": 6.06, + "learning_rate": 2.929025937868447e-06, + "loss": 0.0016, + "step": 46262 + }, + { + "epoch": 6.06, + "learning_rate": 2.928650946390453e-06, + "loss": 0.003, + "step": 46263 + }, + { + "epoch": 6.06, + "learning_rate": 2.9282759748002567e-06, + "loss": 0.0013, + "step": 46264 + }, + { + "epoch": 6.06, + "learning_rate": 2.927901023098906e-06, + "loss": 0.0021, + "step": 46265 + }, + { + "epoch": 6.06, + "learning_rate": 2.927526091287455e-06, + "loss": 0.0023, + "step": 46266 + }, + { + "epoch": 6.06, + "learning_rate": 2.9271511793669628e-06, + "loss": 0.0024, + "step": 46267 + }, + { + "epoch": 6.06, + "learning_rate": 2.926776287338481e-06, + "loss": 0.0016, + "step": 46268 + }, + { + "epoch": 6.06, + "learning_rate": 2.926401415203062e-06, + "loss": 0.002, + "step": 46269 + }, + { + "epoch": 6.06, + "learning_rate": 2.9260265629617644e-06, + "loss": 0.0016, + "step": 46270 + }, + { + "epoch": 6.06, + "learning_rate": 2.9256517306156407e-06, + "loss": 0.0025, + "step": 46271 + }, + { + "epoch": 6.06, + "learning_rate": 2.925276918165746e-06, + "loss": 0.0025, + "step": 46272 + }, + { + "epoch": 6.06, + "learning_rate": 2.9249021256131283e-06, + "loss": 0.0017, + "step": 46273 + }, + { + "epoch": 6.06, + "learning_rate": 2.924527352958847e-06, + "loss": 0.0027, + "step": 46274 + }, + { + "epoch": 6.06, + "learning_rate": 2.92415260020396e-06, + "loss": 0.0013, + "step": 46275 + }, + { + "epoch": 6.06, + "learning_rate": 2.9237778673495156e-06, + "loss": 0.0024, + "step": 46276 + }, + { + "epoch": 6.06, + "learning_rate": 2.923403154396567e-06, + "loss": 0.0031, + "step": 46277 + }, + { + "epoch": 6.06, + "learning_rate": 2.9230284613461722e-06, + "loss": 0.0015, + "step": 46278 + }, + { + "epoch": 6.06, + "learning_rate": 2.9226537881993844e-06, + "loss": 0.0016, + "step": 46279 + }, + { + "epoch": 6.06, + "learning_rate": 2.9222791349572555e-06, + "loss": 0.0023, + "step": 46280 + }, + { + "epoch": 6.06, + "learning_rate": 2.921904501620836e-06, + "loss": 0.0021, + "step": 46281 + }, + { + "epoch": 6.06, + "learning_rate": 2.921529888191186e-06, + "loss": 0.0015, + "step": 46282 + }, + { + "epoch": 6.06, + "learning_rate": 2.9211552946693567e-06, + "loss": 0.0023, + "step": 46283 + }, + { + "epoch": 6.06, + "learning_rate": 2.9207807210563987e-06, + "loss": 0.0026, + "step": 46284 + }, + { + "epoch": 6.06, + "learning_rate": 2.920406167353367e-06, + "loss": 0.0017, + "step": 46285 + }, + { + "epoch": 6.06, + "learning_rate": 2.9200316335613197e-06, + "loss": 0.0024, + "step": 46286 + }, + { + "epoch": 6.06, + "learning_rate": 2.919657119681306e-06, + "loss": 0.0036, + "step": 46287 + }, + { + "epoch": 6.06, + "learning_rate": 2.9192826257143792e-06, + "loss": 0.002, + "step": 46288 + }, + { + "epoch": 6.06, + "learning_rate": 2.91890815166159e-06, + "loss": 0.0019, + "step": 46289 + }, + { + "epoch": 6.06, + "learning_rate": 2.9185336975239985e-06, + "loss": 0.0023, + "step": 46290 + }, + { + "epoch": 6.06, + "learning_rate": 2.9181592633026533e-06, + "loss": 0.0015, + "step": 46291 + }, + { + "epoch": 6.06, + "learning_rate": 2.9177848489986048e-06, + "loss": 0.0016, + "step": 46292 + }, + { + "epoch": 6.06, + "learning_rate": 2.9174104546129124e-06, + "loss": 0.0024, + "step": 46293 + }, + { + "epoch": 6.06, + "learning_rate": 2.917036080146626e-06, + "loss": 0.0017, + "step": 46294 + }, + { + "epoch": 6.06, + "learning_rate": 2.916661725600799e-06, + "loss": 0.003, + "step": 46295 + }, + { + "epoch": 6.06, + "learning_rate": 2.91628739097648e-06, + "loss": 0.0025, + "step": 46296 + }, + { + "epoch": 6.06, + "learning_rate": 2.9159130762747257e-06, + "loss": 0.0014, + "step": 46297 + }, + { + "epoch": 6.06, + "learning_rate": 2.9155387814965907e-06, + "loss": 0.0016, + "step": 46298 + }, + { + "epoch": 6.06, + "learning_rate": 2.9151645066431265e-06, + "loss": 0.0036, + "step": 46299 + }, + { + "epoch": 6.06, + "learning_rate": 2.9147902517153823e-06, + "loss": 0.002, + "step": 46300 + }, + { + "epoch": 6.07, + "learning_rate": 2.914416016714415e-06, + "loss": 0.002, + "step": 46301 + }, + { + "epoch": 6.07, + "learning_rate": 2.9140418016412762e-06, + "loss": 0.0019, + "step": 46302 + }, + { + "epoch": 6.07, + "learning_rate": 2.9136676064970184e-06, + "loss": 0.0012, + "step": 46303 + }, + { + "epoch": 6.07, + "learning_rate": 2.9132934312826876e-06, + "loss": 0.0033, + "step": 46304 + }, + { + "epoch": 6.07, + "learning_rate": 2.9129192759993462e-06, + "loss": 0.0021, + "step": 46305 + }, + { + "epoch": 6.07, + "learning_rate": 2.9125451406480386e-06, + "loss": 0.0016, + "step": 46306 + }, + { + "epoch": 6.07, + "learning_rate": 2.9121710252298243e-06, + "loss": 0.0023, + "step": 46307 + }, + { + "epoch": 6.07, + "learning_rate": 2.9117969297457473e-06, + "loss": 0.0026, + "step": 46308 + }, + { + "epoch": 6.07, + "learning_rate": 2.911422854196868e-06, + "loss": 0.0016, + "step": 46309 + }, + { + "epoch": 6.07, + "learning_rate": 2.911048798584234e-06, + "loss": 0.0013, + "step": 46310 + }, + { + "epoch": 6.07, + "learning_rate": 2.910674762908894e-06, + "loss": 0.0018, + "step": 46311 + }, + { + "epoch": 6.07, + "learning_rate": 2.9103007471719068e-06, + "loss": 0.0022, + "step": 46312 + }, + { + "epoch": 6.07, + "learning_rate": 2.9099267513743225e-06, + "loss": 0.004, + "step": 46313 + }, + { + "epoch": 6.07, + "learning_rate": 2.9095527755171902e-06, + "loss": 0.0021, + "step": 46314 + }, + { + "epoch": 6.07, + "learning_rate": 2.9091788196015615e-06, + "loss": 0.002, + "step": 46315 + }, + { + "epoch": 6.07, + "learning_rate": 2.908804883628492e-06, + "loss": 0.0019, + "step": 46316 + }, + { + "epoch": 6.07, + "learning_rate": 2.9084309675990274e-06, + "loss": 0.0023, + "step": 46317 + }, + { + "epoch": 6.07, + "learning_rate": 2.9080570715142277e-06, + "loss": 0.0021, + "step": 46318 + }, + { + "epoch": 6.07, + "learning_rate": 2.907683195375135e-06, + "loss": 0.0026, + "step": 46319 + }, + { + "epoch": 6.07, + "learning_rate": 2.90730933918281e-06, + "loss": 0.0026, + "step": 46320 + }, + { + "epoch": 6.07, + "learning_rate": 2.9069355029382994e-06, + "loss": 0.0025, + "step": 46321 + }, + { + "epoch": 6.07, + "learning_rate": 2.906561686642655e-06, + "loss": 0.0012, + "step": 46322 + }, + { + "epoch": 6.07, + "learning_rate": 2.906187890296924e-06, + "loss": 0.0014, + "step": 46323 + }, + { + "epoch": 6.07, + "learning_rate": 2.9058141139021668e-06, + "loss": 0.0018, + "step": 46324 + }, + { + "epoch": 6.07, + "learning_rate": 2.905440357459428e-06, + "loss": 0.0033, + "step": 46325 + }, + { + "epoch": 6.07, + "learning_rate": 2.9050666209697587e-06, + "loss": 0.0026, + "step": 46326 + }, + { + "epoch": 6.07, + "learning_rate": 2.904692904434214e-06, + "loss": 0.0027, + "step": 46327 + }, + { + "epoch": 6.07, + "learning_rate": 2.9043192078538395e-06, + "loss": 0.0028, + "step": 46328 + }, + { + "epoch": 6.07, + "learning_rate": 2.903945531229693e-06, + "loss": 0.0017, + "step": 46329 + }, + { + "epoch": 6.07, + "learning_rate": 2.9035718745628206e-06, + "loss": 0.0021, + "step": 46330 + }, + { + "epoch": 6.07, + "learning_rate": 2.9031982378542724e-06, + "loss": 0.002, + "step": 46331 + }, + { + "epoch": 6.07, + "learning_rate": 2.9028246211051035e-06, + "loss": 0.002, + "step": 46332 + }, + { + "epoch": 6.07, + "learning_rate": 2.902451024316363e-06, + "loss": 0.0012, + "step": 46333 + }, + { + "epoch": 6.07, + "learning_rate": 2.9020774474890977e-06, + "loss": 0.0019, + "step": 46334 + }, + { + "epoch": 6.07, + "learning_rate": 2.9017038906243654e-06, + "loss": 0.0028, + "step": 46335 + }, + { + "epoch": 6.07, + "learning_rate": 2.901330353723212e-06, + "loss": 0.004, + "step": 46336 + }, + { + "epoch": 6.07, + "learning_rate": 2.900956836786689e-06, + "loss": 0.0022, + "step": 46337 + }, + { + "epoch": 6.07, + "learning_rate": 2.9005833398158444e-06, + "loss": 0.0046, + "step": 46338 + }, + { + "epoch": 6.07, + "learning_rate": 2.9002098628117305e-06, + "loss": 0.0019, + "step": 46339 + }, + { + "epoch": 6.07, + "learning_rate": 2.8998364057754015e-06, + "loss": 0.0026, + "step": 46340 + }, + { + "epoch": 6.07, + "learning_rate": 2.8994629687079047e-06, + "loss": 0.0019, + "step": 46341 + }, + { + "epoch": 6.07, + "learning_rate": 2.8990895516102866e-06, + "loss": 0.0026, + "step": 46342 + }, + { + "epoch": 6.07, + "learning_rate": 2.8987161544836052e-06, + "loss": 0.0032, + "step": 46343 + }, + { + "epoch": 6.07, + "learning_rate": 2.8983427773289053e-06, + "loss": 0.0027, + "step": 46344 + }, + { + "epoch": 6.07, + "learning_rate": 2.8979694201472385e-06, + "loss": 0.003, + "step": 46345 + }, + { + "epoch": 6.07, + "learning_rate": 2.897596082939652e-06, + "loss": 0.0022, + "step": 46346 + }, + { + "epoch": 6.07, + "learning_rate": 2.8972227657072004e-06, + "loss": 0.0015, + "step": 46347 + }, + { + "epoch": 6.07, + "learning_rate": 2.8968494684509318e-06, + "loss": 0.0016, + "step": 46348 + }, + { + "epoch": 6.07, + "learning_rate": 2.896476191171893e-06, + "loss": 0.0023, + "step": 46349 + }, + { + "epoch": 6.07, + "learning_rate": 2.8961029338711366e-06, + "loss": 0.0023, + "step": 46350 + }, + { + "epoch": 6.07, + "learning_rate": 2.895729696549716e-06, + "loss": 0.0022, + "step": 46351 + }, + { + "epoch": 6.07, + "learning_rate": 2.8953564792086764e-06, + "loss": 0.0022, + "step": 46352 + }, + { + "epoch": 6.07, + "learning_rate": 2.894983281849069e-06, + "loss": 0.0019, + "step": 46353 + }, + { + "epoch": 6.07, + "learning_rate": 2.894610104471939e-06, + "loss": 0.0029, + "step": 46354 + }, + { + "epoch": 6.07, + "learning_rate": 2.8942369470783426e-06, + "loss": 0.0017, + "step": 46355 + }, + { + "epoch": 6.07, + "learning_rate": 2.8938638096693274e-06, + "loss": 0.0011, + "step": 46356 + }, + { + "epoch": 6.07, + "learning_rate": 2.893490692245938e-06, + "loss": 0.0041, + "step": 46357 + }, + { + "epoch": 6.07, + "learning_rate": 2.8931175948092307e-06, + "loss": 0.0022, + "step": 46358 + }, + { + "epoch": 6.07, + "learning_rate": 2.892744517360252e-06, + "loss": 0.0022, + "step": 46359 + }, + { + "epoch": 6.07, + "learning_rate": 2.892371459900051e-06, + "loss": 0.0024, + "step": 46360 + }, + { + "epoch": 6.07, + "learning_rate": 2.8919984224296706e-06, + "loss": 0.0024, + "step": 46361 + }, + { + "epoch": 6.07, + "learning_rate": 2.8916254049501713e-06, + "loss": 0.0022, + "step": 46362 + }, + { + "epoch": 6.07, + "learning_rate": 2.891252407462598e-06, + "loss": 0.0014, + "step": 46363 + }, + { + "epoch": 6.07, + "learning_rate": 2.890879429967999e-06, + "loss": 0.0038, + "step": 46364 + }, + { + "epoch": 6.07, + "learning_rate": 2.8905064724674193e-06, + "loss": 0.0019, + "step": 46365 + }, + { + "epoch": 6.07, + "learning_rate": 2.8901335349619143e-06, + "loss": 0.0016, + "step": 46366 + }, + { + "epoch": 6.07, + "learning_rate": 2.8897606174525308e-06, + "loss": 0.0018, + "step": 46367 + }, + { + "epoch": 6.07, + "learning_rate": 2.8893877199403165e-06, + "loss": 0.0024, + "step": 46368 + }, + { + "epoch": 6.07, + "learning_rate": 2.8890148424263177e-06, + "loss": 0.0017, + "step": 46369 + }, + { + "epoch": 6.07, + "learning_rate": 2.8886419849115887e-06, + "loss": 0.0023, + "step": 46370 + }, + { + "epoch": 6.07, + "learning_rate": 2.888269147397176e-06, + "loss": 0.002, + "step": 46371 + }, + { + "epoch": 6.07, + "learning_rate": 2.8878963298841246e-06, + "loss": 0.0024, + "step": 46372 + }, + { + "epoch": 6.07, + "learning_rate": 2.8875235323734853e-06, + "loss": 0.0015, + "step": 46373 + }, + { + "epoch": 6.07, + "learning_rate": 2.8871507548663114e-06, + "loss": 0.0017, + "step": 46374 + }, + { + "epoch": 6.07, + "learning_rate": 2.886777997363646e-06, + "loss": 0.0015, + "step": 46375 + }, + { + "epoch": 6.07, + "learning_rate": 2.886405259866536e-06, + "loss": 0.0026, + "step": 46376 + }, + { + "epoch": 6.08, + "learning_rate": 2.886032542376036e-06, + "loss": 0.0021, + "step": 46377 + }, + { + "epoch": 6.08, + "learning_rate": 2.8856598448931905e-06, + "loss": 0.0015, + "step": 46378 + }, + { + "epoch": 6.08, + "learning_rate": 2.8852871674190474e-06, + "loss": 0.0018, + "step": 46379 + }, + { + "epoch": 6.08, + "learning_rate": 2.884914509954653e-06, + "loss": 0.0021, + "step": 46380 + }, + { + "epoch": 6.08, + "learning_rate": 2.8845418725010597e-06, + "loss": 0.0016, + "step": 46381 + }, + { + "epoch": 6.08, + "learning_rate": 2.8841692550593137e-06, + "loss": 0.0021, + "step": 46382 + }, + { + "epoch": 6.08, + "learning_rate": 2.88379665763046e-06, + "loss": 0.0027, + "step": 46383 + }, + { + "epoch": 6.08, + "learning_rate": 2.883424080215549e-06, + "loss": 0.0021, + "step": 46384 + }, + { + "epoch": 6.08, + "learning_rate": 2.883051522815632e-06, + "loss": 0.0015, + "step": 46385 + }, + { + "epoch": 6.08, + "learning_rate": 2.8826789854317537e-06, + "loss": 0.0039, + "step": 46386 + }, + { + "epoch": 6.08, + "learning_rate": 2.8823064680649616e-06, + "loss": 0.0027, + "step": 46387 + }, + { + "epoch": 6.08, + "learning_rate": 2.8819339707163e-06, + "loss": 0.0023, + "step": 46388 + }, + { + "epoch": 6.08, + "learning_rate": 2.8815614933868243e-06, + "loss": 0.0027, + "step": 46389 + }, + { + "epoch": 6.08, + "learning_rate": 2.8811890360775764e-06, + "loss": 0.0018, + "step": 46390 + }, + { + "epoch": 6.08, + "learning_rate": 2.880816598789603e-06, + "loss": 0.0022, + "step": 46391 + }, + { + "epoch": 6.08, + "learning_rate": 2.880444181523957e-06, + "loss": 0.0032, + "step": 46392 + }, + { + "epoch": 6.08, + "learning_rate": 2.8800717842816815e-06, + "loss": 0.0039, + "step": 46393 + }, + { + "epoch": 6.08, + "learning_rate": 2.879699407063823e-06, + "loss": 0.003, + "step": 46394 + }, + { + "epoch": 6.08, + "learning_rate": 2.8793270498714333e-06, + "loss": 0.0015, + "step": 46395 + }, + { + "epoch": 6.08, + "learning_rate": 2.878954712705554e-06, + "loss": 0.001, + "step": 46396 + }, + { + "epoch": 6.08, + "learning_rate": 2.8785823955672385e-06, + "loss": 0.0025, + "step": 46397 + }, + { + "epoch": 6.08, + "learning_rate": 2.878210098457531e-06, + "loss": 0.001, + "step": 46398 + }, + { + "epoch": 6.08, + "learning_rate": 2.877837821377475e-06, + "loss": 0.0023, + "step": 46399 + }, + { + "epoch": 6.08, + "learning_rate": 2.8774655643281245e-06, + "loss": 0.0015, + "step": 46400 + }, + { + "epoch": 6.08, + "learning_rate": 2.8770933273105227e-06, + "loss": 0.0023, + "step": 46401 + }, + { + "epoch": 6.08, + "learning_rate": 2.8767211103257164e-06, + "loss": 0.0018, + "step": 46402 + }, + { + "epoch": 6.08, + "learning_rate": 2.876348913374749e-06, + "loss": 0.0017, + "step": 46403 + }, + { + "epoch": 6.08, + "learning_rate": 2.8759767364586742e-06, + "loss": 0.0012, + "step": 46404 + }, + { + "epoch": 6.08, + "learning_rate": 2.8756045795785326e-06, + "loss": 0.0019, + "step": 46405 + }, + { + "epoch": 6.08, + "learning_rate": 2.8752324427353772e-06, + "loss": 0.0023, + "step": 46406 + }, + { + "epoch": 6.08, + "learning_rate": 2.8748603259302478e-06, + "loss": 0.0025, + "step": 46407 + }, + { + "epoch": 6.08, + "learning_rate": 2.8744882291641975e-06, + "loss": 0.0013, + "step": 46408 + }, + { + "epoch": 6.08, + "learning_rate": 2.8741161524382687e-06, + "loss": 0.0033, + "step": 46409 + }, + { + "epoch": 6.08, + "learning_rate": 2.87374409575351e-06, + "loss": 0.0016, + "step": 46410 + }, + { + "epoch": 6.08, + "learning_rate": 2.8733720591109624e-06, + "loss": 0.0024, + "step": 46411 + }, + { + "epoch": 6.08, + "learning_rate": 2.87300004251168e-06, + "loss": 0.0016, + "step": 46412 + }, + { + "epoch": 6.08, + "learning_rate": 2.8726280459567057e-06, + "loss": 0.0015, + "step": 46413 + }, + { + "epoch": 6.08, + "learning_rate": 2.872256069447081e-06, + "loss": 0.0016, + "step": 46414 + }, + { + "epoch": 6.08, + "learning_rate": 2.8718841129838613e-06, + "loss": 0.0014, + "step": 46415 + }, + { + "epoch": 6.08, + "learning_rate": 2.8715121765680843e-06, + "loss": 0.0017, + "step": 46416 + }, + { + "epoch": 6.08, + "learning_rate": 2.871140260200802e-06, + "loss": 0.0019, + "step": 46417 + }, + { + "epoch": 6.08, + "learning_rate": 2.8707683638830596e-06, + "loss": 0.0025, + "step": 46418 + }, + { + "epoch": 6.08, + "learning_rate": 2.8703964876158973e-06, + "loss": 0.0027, + "step": 46419 + }, + { + "epoch": 6.08, + "learning_rate": 2.870024631400369e-06, + "loss": 0.0024, + "step": 46420 + }, + { + "epoch": 6.08, + "learning_rate": 2.8696527952375163e-06, + "loss": 0.0015, + "step": 46421 + }, + { + "epoch": 6.08, + "learning_rate": 2.8692809791283817e-06, + "loss": 0.0029, + "step": 46422 + }, + { + "epoch": 6.08, + "learning_rate": 2.868909183074019e-06, + "loss": 0.0019, + "step": 46423 + }, + { + "epoch": 6.08, + "learning_rate": 2.868537407075468e-06, + "loss": 0.002, + "step": 46424 + }, + { + "epoch": 6.08, + "learning_rate": 2.8681656511337765e-06, + "loss": 0.0021, + "step": 46425 + }, + { + "epoch": 6.08, + "learning_rate": 2.867793915249987e-06, + "loss": 0.0024, + "step": 46426 + }, + { + "epoch": 6.08, + "learning_rate": 2.867422199425146e-06, + "loss": 0.0036, + "step": 46427 + }, + { + "epoch": 6.08, + "learning_rate": 2.8670505036603047e-06, + "loss": 0.002, + "step": 46428 + }, + { + "epoch": 6.08, + "learning_rate": 2.8666788279565026e-06, + "loss": 0.0016, + "step": 46429 + }, + { + "epoch": 6.08, + "learning_rate": 2.866307172314784e-06, + "loss": 0.002, + "step": 46430 + }, + { + "epoch": 6.08, + "learning_rate": 2.8659355367361996e-06, + "loss": 0.0024, + "step": 46431 + }, + { + "epoch": 6.08, + "learning_rate": 2.8655639212217914e-06, + "loss": 0.0015, + "step": 46432 + }, + { + "epoch": 6.08, + "learning_rate": 2.8651923257726056e-06, + "loss": 0.0025, + "step": 46433 + }, + { + "epoch": 6.08, + "learning_rate": 2.864820750389682e-06, + "loss": 0.0014, + "step": 46434 + }, + { + "epoch": 6.08, + "learning_rate": 2.8644491950740738e-06, + "loss": 0.002, + "step": 46435 + }, + { + "epoch": 6.08, + "learning_rate": 2.864077659826823e-06, + "loss": 0.0014, + "step": 46436 + }, + { + "epoch": 6.08, + "learning_rate": 2.86370614464897e-06, + "loss": 0.0026, + "step": 46437 + }, + { + "epoch": 6.08, + "learning_rate": 2.8633346495415636e-06, + "loss": 0.0021, + "step": 46438 + }, + { + "epoch": 6.08, + "learning_rate": 2.8629631745056507e-06, + "loss": 0.0022, + "step": 46439 + }, + { + "epoch": 6.08, + "learning_rate": 2.8625917195422748e-06, + "loss": 0.002, + "step": 46440 + }, + { + "epoch": 6.08, + "learning_rate": 2.862220284652476e-06, + "loss": 0.0024, + "step": 46441 + }, + { + "epoch": 6.08, + "learning_rate": 2.861848869837306e-06, + "loss": 0.0016, + "step": 46442 + }, + { + "epoch": 6.08, + "learning_rate": 2.861477475097806e-06, + "loss": 0.002, + "step": 46443 + }, + { + "epoch": 6.08, + "learning_rate": 2.8611061004350193e-06, + "loss": 0.0034, + "step": 46444 + }, + { + "epoch": 6.08, + "learning_rate": 2.86073474584999e-06, + "loss": 0.0021, + "step": 46445 + }, + { + "epoch": 6.08, + "learning_rate": 2.8603634113437663e-06, + "loss": 0.0019, + "step": 46446 + }, + { + "epoch": 6.08, + "learning_rate": 2.85999209691739e-06, + "loss": 0.0027, + "step": 46447 + }, + { + "epoch": 6.08, + "learning_rate": 2.859620802571903e-06, + "loss": 0.0032, + "step": 46448 + }, + { + "epoch": 6.08, + "learning_rate": 2.859249528308353e-06, + "loss": 0.0025, + "step": 46449 + }, + { + "epoch": 6.08, + "learning_rate": 2.8588782741277855e-06, + "loss": 0.0042, + "step": 46450 + }, + { + "epoch": 6.08, + "learning_rate": 2.858507040031243e-06, + "loss": 0.0025, + "step": 46451 + }, + { + "epoch": 6.08, + "learning_rate": 2.858135826019769e-06, + "loss": 0.0019, + "step": 46452 + }, + { + "epoch": 6.09, + "learning_rate": 2.8577646320944052e-06, + "loss": 0.0019, + "step": 46453 + }, + { + "epoch": 6.09, + "learning_rate": 2.8573934582562e-06, + "loss": 0.002, + "step": 46454 + }, + { + "epoch": 6.09, + "learning_rate": 2.857022304506196e-06, + "loss": 0.0018, + "step": 46455 + }, + { + "epoch": 6.09, + "learning_rate": 2.8566511708454335e-06, + "loss": 0.0017, + "step": 46456 + }, + { + "epoch": 6.09, + "learning_rate": 2.8562800572749618e-06, + "loss": 0.0027, + "step": 46457 + }, + { + "epoch": 6.09, + "learning_rate": 2.8559089637958214e-06, + "loss": 0.0019, + "step": 46458 + }, + { + "epoch": 6.09, + "learning_rate": 2.8555378904090547e-06, + "loss": 0.0017, + "step": 46459 + }, + { + "epoch": 6.09, + "learning_rate": 2.8551668371157094e-06, + "loss": 0.0019, + "step": 46460 + }, + { + "epoch": 6.09, + "learning_rate": 2.8547958039168245e-06, + "loss": 0.0018, + "step": 46461 + }, + { + "epoch": 6.09, + "learning_rate": 2.8544247908134494e-06, + "loss": 0.0018, + "step": 46462 + }, + { + "epoch": 6.09, + "learning_rate": 2.854053797806623e-06, + "loss": 0.0038, + "step": 46463 + }, + { + "epoch": 6.09, + "learning_rate": 2.8536828248973868e-06, + "loss": 0.0016, + "step": 46464 + }, + { + "epoch": 6.09, + "learning_rate": 2.85331187208679e-06, + "loss": 0.0028, + "step": 46465 + }, + { + "epoch": 6.09, + "learning_rate": 2.8529409393758745e-06, + "loss": 0.0014, + "step": 46466 + }, + { + "epoch": 6.09, + "learning_rate": 2.852570026765681e-06, + "loss": 0.0025, + "step": 46467 + }, + { + "epoch": 6.09, + "learning_rate": 2.85219913425725e-06, + "loss": 0.0027, + "step": 46468 + }, + { + "epoch": 6.09, + "learning_rate": 2.8518282618516314e-06, + "loss": 0.0019, + "step": 46469 + }, + { + "epoch": 6.09, + "learning_rate": 2.8514574095498626e-06, + "loss": 0.0033, + "step": 46470 + }, + { + "epoch": 6.09, + "learning_rate": 2.851086577352993e-06, + "loss": 0.002, + "step": 46471 + }, + { + "epoch": 6.09, + "learning_rate": 2.8507157652620576e-06, + "loss": 0.0031, + "step": 46472 + }, + { + "epoch": 6.09, + "learning_rate": 2.8503449732781064e-06, + "loss": 0.0033, + "step": 46473 + }, + { + "epoch": 6.09, + "learning_rate": 2.849974201402179e-06, + "loss": 0.0022, + "step": 46474 + }, + { + "epoch": 6.09, + "learning_rate": 2.84960344963532e-06, + "loss": 0.0018, + "step": 46475 + }, + { + "epoch": 6.09, + "learning_rate": 2.849232717978565e-06, + "loss": 0.0015, + "step": 46476 + }, + { + "epoch": 6.09, + "learning_rate": 2.8488620064329676e-06, + "loss": 0.0013, + "step": 46477 + }, + { + "epoch": 6.09, + "learning_rate": 2.848491314999563e-06, + "loss": 0.0026, + "step": 46478 + }, + { + "epoch": 6.09, + "learning_rate": 2.848120643679394e-06, + "loss": 0.0027, + "step": 46479 + }, + { + "epoch": 6.09, + "learning_rate": 2.8477499924735074e-06, + "loss": 0.0019, + "step": 46480 + }, + { + "epoch": 6.09, + "learning_rate": 2.8473793613829402e-06, + "loss": 0.0028, + "step": 46481 + }, + { + "epoch": 6.09, + "learning_rate": 2.8470087504087406e-06, + "loss": 0.0014, + "step": 46482 + }, + { + "epoch": 6.09, + "learning_rate": 2.8466381595519487e-06, + "loss": 0.0015, + "step": 46483 + }, + { + "epoch": 6.09, + "learning_rate": 2.8462675888136017e-06, + "loss": 0.0023, + "step": 46484 + }, + { + "epoch": 6.09, + "learning_rate": 2.8458970381947504e-06, + "loss": 0.0017, + "step": 46485 + }, + { + "epoch": 6.09, + "learning_rate": 2.845526507696431e-06, + "loss": 0.0012, + "step": 46486 + }, + { + "epoch": 6.09, + "learning_rate": 2.845155997319686e-06, + "loss": 0.003, + "step": 46487 + }, + { + "epoch": 6.09, + "learning_rate": 2.8447855070655607e-06, + "loss": 0.0015, + "step": 46488 + }, + { + "epoch": 6.09, + "learning_rate": 2.8444150369350966e-06, + "loss": 0.0029, + "step": 46489 + }, + { + "epoch": 6.09, + "learning_rate": 2.844044586929332e-06, + "loss": 0.0022, + "step": 46490 + }, + { + "epoch": 6.09, + "learning_rate": 2.8436741570493097e-06, + "loss": 0.0028, + "step": 46491 + }, + { + "epoch": 6.09, + "learning_rate": 2.843303747296071e-06, + "loss": 0.0016, + "step": 46492 + }, + { + "epoch": 6.09, + "learning_rate": 2.8429333576706643e-06, + "loss": 0.0025, + "step": 46493 + }, + { + "epoch": 6.09, + "learning_rate": 2.8425629881741255e-06, + "loss": 0.0032, + "step": 46494 + }, + { + "epoch": 6.09, + "learning_rate": 2.8421926388074937e-06, + "loss": 0.0023, + "step": 46495 + }, + { + "epoch": 6.09, + "learning_rate": 2.841822309571818e-06, + "loss": 0.0025, + "step": 46496 + }, + { + "epoch": 6.09, + "learning_rate": 2.8414520004681345e-06, + "loss": 0.0034, + "step": 46497 + }, + { + "epoch": 6.09, + "learning_rate": 2.8410817114974875e-06, + "loss": 0.0025, + "step": 46498 + }, + { + "epoch": 6.09, + "learning_rate": 2.8407114426609127e-06, + "loss": 0.0027, + "step": 46499 + }, + { + "epoch": 6.09, + "learning_rate": 2.8403411939594596e-06, + "loss": 0.0033, + "step": 46500 + }, + { + "epoch": 6.09, + "learning_rate": 2.8399709653941655e-06, + "loss": 0.0012, + "step": 46501 + }, + { + "epoch": 6.09, + "learning_rate": 2.8396007569660687e-06, + "loss": 0.0035, + "step": 46502 + }, + { + "epoch": 6.09, + "learning_rate": 2.8392305686762133e-06, + "loss": 0.0018, + "step": 46503 + }, + { + "epoch": 6.09, + "learning_rate": 2.8388604005256436e-06, + "loss": 0.0024, + "step": 46504 + }, + { + "epoch": 6.09, + "learning_rate": 2.838490252515397e-06, + "loss": 0.0021, + "step": 46505 + }, + { + "epoch": 6.09, + "learning_rate": 2.838120124646513e-06, + "loss": 0.0029, + "step": 46506 + }, + { + "epoch": 6.09, + "learning_rate": 2.837750016920038e-06, + "loss": 0.0029, + "step": 46507 + }, + { + "epoch": 6.09, + "learning_rate": 2.8373799293370086e-06, + "loss": 0.004, + "step": 46508 + }, + { + "epoch": 6.09, + "learning_rate": 2.8370098618984675e-06, + "loss": 0.0024, + "step": 46509 + }, + { + "epoch": 6.09, + "learning_rate": 2.836639814605452e-06, + "loss": 0.002, + "step": 46510 + }, + { + "epoch": 6.09, + "learning_rate": 2.8362697874590085e-06, + "loss": 0.0021, + "step": 46511 + }, + { + "epoch": 6.09, + "learning_rate": 2.835899780460174e-06, + "loss": 0.0013, + "step": 46512 + }, + { + "epoch": 6.09, + "learning_rate": 2.83552979360999e-06, + "loss": 0.002, + "step": 46513 + }, + { + "epoch": 6.09, + "learning_rate": 2.8351598269094916e-06, + "loss": 0.0025, + "step": 46514 + }, + { + "epoch": 6.09, + "learning_rate": 2.8347898803597295e-06, + "loss": 0.0016, + "step": 46515 + }, + { + "epoch": 6.09, + "learning_rate": 2.83441995396174e-06, + "loss": 0.0019, + "step": 46516 + }, + { + "epoch": 6.09, + "learning_rate": 2.8340500477165635e-06, + "loss": 0.0017, + "step": 46517 + }, + { + "epoch": 6.09, + "learning_rate": 2.8336801616252353e-06, + "loss": 0.0014, + "step": 46518 + }, + { + "epoch": 6.09, + "learning_rate": 2.8333102956888036e-06, + "loss": 0.0024, + "step": 46519 + }, + { + "epoch": 6.09, + "learning_rate": 2.8329404499083057e-06, + "loss": 0.0029, + "step": 46520 + }, + { + "epoch": 6.09, + "learning_rate": 2.832570624284777e-06, + "loss": 0.0021, + "step": 46521 + }, + { + "epoch": 6.09, + "learning_rate": 2.832200818819265e-06, + "loss": 0.0025, + "step": 46522 + }, + { + "epoch": 6.09, + "learning_rate": 2.8318310335128073e-06, + "loss": 0.0019, + "step": 46523 + }, + { + "epoch": 6.09, + "learning_rate": 2.831461268366442e-06, + "loss": 0.0018, + "step": 46524 + }, + { + "epoch": 6.09, + "learning_rate": 2.8310915233812086e-06, + "loss": 0.0016, + "step": 46525 + }, + { + "epoch": 6.09, + "learning_rate": 2.830721798558147e-06, + "loss": 0.0028, + "step": 46526 + }, + { + "epoch": 6.09, + "learning_rate": 2.830352093898302e-06, + "loss": 0.0026, + "step": 46527 + }, + { + "epoch": 6.09, + "learning_rate": 2.8299824094027104e-06, + "loss": 0.0027, + "step": 46528 + }, + { + "epoch": 6.09, + "learning_rate": 2.8296127450724077e-06, + "loss": 0.0028, + "step": 46529 + }, + { + "epoch": 6.1, + "learning_rate": 2.8292431009084407e-06, + "loss": 0.001, + "step": 46530 + }, + { + "epoch": 6.1, + "learning_rate": 2.8288734769118454e-06, + "loss": 0.002, + "step": 46531 + }, + { + "epoch": 6.1, + "learning_rate": 2.8285038730836624e-06, + "loss": 0.0018, + "step": 46532 + }, + { + "epoch": 6.1, + "learning_rate": 2.828134289424926e-06, + "loss": 0.0023, + "step": 46533 + }, + { + "epoch": 6.1, + "learning_rate": 2.8277647259366847e-06, + "loss": 0.0018, + "step": 46534 + }, + { + "epoch": 6.1, + "learning_rate": 2.827395182619972e-06, + "loss": 0.0018, + "step": 46535 + }, + { + "epoch": 6.1, + "learning_rate": 2.8270256594758248e-06, + "loss": 0.0015, + "step": 46536 + }, + { + "epoch": 6.1, + "learning_rate": 2.8266561565052875e-06, + "loss": 0.0015, + "step": 46537 + }, + { + "epoch": 6.1, + "learning_rate": 2.8262866737093996e-06, + "loss": 0.0026, + "step": 46538 + }, + { + "epoch": 6.1, + "learning_rate": 2.8259172110891987e-06, + "loss": 0.0015, + "step": 46539 + }, + { + "epoch": 6.1, + "learning_rate": 2.8255477686457245e-06, + "loss": 0.0015, + "step": 46540 + }, + { + "epoch": 6.1, + "learning_rate": 2.825178346380011e-06, + "loss": 0.0018, + "step": 46541 + }, + { + "epoch": 6.1, + "learning_rate": 2.8248089442931046e-06, + "loss": 0.0026, + "step": 46542 + }, + { + "epoch": 6.1, + "learning_rate": 2.824439562386041e-06, + "loss": 0.0015, + "step": 46543 + }, + { + "epoch": 6.1, + "learning_rate": 2.8240702006598553e-06, + "loss": 0.0044, + "step": 46544 + }, + { + "epoch": 6.1, + "learning_rate": 2.8237008591155946e-06, + "loss": 0.0018, + "step": 46545 + }, + { + "epoch": 6.1, + "learning_rate": 2.8233315377542914e-06, + "loss": 0.0036, + "step": 46546 + }, + { + "epoch": 6.1, + "learning_rate": 2.8229622365769837e-06, + "loss": 0.0026, + "step": 46547 + }, + { + "epoch": 6.1, + "learning_rate": 2.8225929555847164e-06, + "loss": 0.0026, + "step": 46548 + }, + { + "epoch": 6.1, + "learning_rate": 2.82222369477852e-06, + "loss": 0.0017, + "step": 46549 + }, + { + "epoch": 6.1, + "learning_rate": 2.8218544541594406e-06, + "loss": 0.0013, + "step": 46550 + }, + { + "epoch": 6.1, + "learning_rate": 2.8214852337285137e-06, + "loss": 0.0016, + "step": 46551 + }, + { + "epoch": 6.1, + "learning_rate": 2.8211160334867737e-06, + "loss": 0.0029, + "step": 46552 + }, + { + "epoch": 6.1, + "learning_rate": 2.8207468534352656e-06, + "loss": 0.0023, + "step": 46553 + }, + { + "epoch": 6.1, + "learning_rate": 2.820377693575025e-06, + "loss": 0.0019, + "step": 46554 + }, + { + "epoch": 6.1, + "learning_rate": 2.82000855390709e-06, + "loss": 0.002, + "step": 46555 + }, + { + "epoch": 6.1, + "learning_rate": 2.819639434432495e-06, + "loss": 0.0016, + "step": 46556 + }, + { + "epoch": 6.1, + "learning_rate": 2.8192703351522854e-06, + "loss": 0.002, + "step": 46557 + }, + { + "epoch": 6.1, + "learning_rate": 2.818901256067492e-06, + "loss": 0.0021, + "step": 46558 + }, + { + "epoch": 6.1, + "learning_rate": 2.81853219717916e-06, + "loss": 0.0032, + "step": 46559 + }, + { + "epoch": 6.1, + "learning_rate": 2.8181631584883195e-06, + "loss": 0.0024, + "step": 46560 + }, + { + "epoch": 6.1, + "learning_rate": 2.8177941399960175e-06, + "loss": 0.003, + "step": 46561 + }, + { + "epoch": 6.1, + "learning_rate": 2.817425141703285e-06, + "loss": 0.0021, + "step": 46562 + }, + { + "epoch": 6.1, + "learning_rate": 2.817056163611164e-06, + "loss": 0.0029, + "step": 46563 + }, + { + "epoch": 6.1, + "learning_rate": 2.816687205720685e-06, + "loss": 0.0024, + "step": 46564 + }, + { + "epoch": 6.1, + "learning_rate": 2.8163182680328948e-06, + "loss": 0.0026, + "step": 46565 + }, + { + "epoch": 6.1, + "learning_rate": 2.8159493505488266e-06, + "loss": 0.0026, + "step": 46566 + }, + { + "epoch": 6.1, + "learning_rate": 2.815580453269515e-06, + "loss": 0.0022, + "step": 46567 + }, + { + "epoch": 6.1, + "learning_rate": 2.8152115761960007e-06, + "loss": 0.0022, + "step": 46568 + }, + { + "epoch": 6.1, + "learning_rate": 2.814842719329325e-06, + "loss": 0.004, + "step": 46569 + }, + { + "epoch": 6.1, + "learning_rate": 2.814473882670521e-06, + "loss": 0.0025, + "step": 46570 + }, + { + "epoch": 6.1, + "learning_rate": 2.8141050662206235e-06, + "loss": 0.003, + "step": 46571 + }, + { + "epoch": 6.1, + "learning_rate": 2.813736269980677e-06, + "loss": 0.0017, + "step": 46572 + }, + { + "epoch": 6.1, + "learning_rate": 2.813367493951713e-06, + "loss": 0.0038, + "step": 46573 + }, + { + "epoch": 6.1, + "learning_rate": 2.8129987381347713e-06, + "loss": 0.0012, + "step": 46574 + }, + { + "epoch": 6.1, + "learning_rate": 2.812630002530884e-06, + "loss": 0.0018, + "step": 46575 + }, + { + "epoch": 6.1, + "learning_rate": 2.812261287141096e-06, + "loss": 0.0019, + "step": 46576 + }, + { + "epoch": 6.1, + "learning_rate": 2.8118925919664396e-06, + "loss": 0.0024, + "step": 46577 + }, + { + "epoch": 6.1, + "learning_rate": 2.81152391700795e-06, + "loss": 0.0028, + "step": 46578 + }, + { + "epoch": 6.1, + "learning_rate": 2.811155262266666e-06, + "loss": 0.0023, + "step": 46579 + }, + { + "epoch": 6.1, + "learning_rate": 2.810786627743628e-06, + "loss": 0.0025, + "step": 46580 + }, + { + "epoch": 6.1, + "learning_rate": 2.810418013439871e-06, + "loss": 0.0023, + "step": 46581 + }, + { + "epoch": 6.1, + "learning_rate": 2.810049419356429e-06, + "loss": 0.0016, + "step": 46582 + }, + { + "epoch": 6.1, + "learning_rate": 2.809680845494337e-06, + "loss": 0.001, + "step": 46583 + }, + { + "epoch": 6.1, + "learning_rate": 2.809312291854639e-06, + "loss": 0.0027, + "step": 46584 + }, + { + "epoch": 6.1, + "learning_rate": 2.8089437584383663e-06, + "loss": 0.0023, + "step": 46585 + }, + { + "epoch": 6.1, + "learning_rate": 2.8085752452465543e-06, + "loss": 0.0029, + "step": 46586 + }, + { + "epoch": 6.1, + "learning_rate": 2.808206752280244e-06, + "loss": 0.003, + "step": 46587 + }, + { + "epoch": 6.1, + "learning_rate": 2.807838279540468e-06, + "loss": 0.0015, + "step": 46588 + }, + { + "epoch": 6.1, + "learning_rate": 2.807469827028265e-06, + "loss": 0.0018, + "step": 46589 + }, + { + "epoch": 6.1, + "learning_rate": 2.807101394744667e-06, + "loss": 0.0035, + "step": 46590 + }, + { + "epoch": 6.1, + "learning_rate": 2.806732982690713e-06, + "loss": 0.0021, + "step": 46591 + }, + { + "epoch": 6.1, + "learning_rate": 2.806364590867443e-06, + "loss": 0.0017, + "step": 46592 + }, + { + "epoch": 6.1, + "learning_rate": 2.8059962192758895e-06, + "loss": 0.0017, + "step": 46593 + }, + { + "epoch": 6.1, + "learning_rate": 2.8056278679170845e-06, + "loss": 0.0008, + "step": 46594 + }, + { + "epoch": 6.1, + "learning_rate": 2.805259536792072e-06, + "loss": 0.0025, + "step": 46595 + }, + { + "epoch": 6.1, + "learning_rate": 2.8048912259018833e-06, + "loss": 0.0014, + "step": 46596 + }, + { + "epoch": 6.1, + "learning_rate": 2.8045229352475554e-06, + "loss": 0.0022, + "step": 46597 + }, + { + "epoch": 6.1, + "learning_rate": 2.804154664830121e-06, + "loss": 0.0012, + "step": 46598 + }, + { + "epoch": 6.1, + "learning_rate": 2.8037864146506198e-06, + "loss": 0.0014, + "step": 46599 + }, + { + "epoch": 6.1, + "learning_rate": 2.8034181847100872e-06, + "loss": 0.0016, + "step": 46600 + }, + { + "epoch": 6.1, + "learning_rate": 2.803049975009555e-06, + "loss": 0.0038, + "step": 46601 + }, + { + "epoch": 6.1, + "learning_rate": 2.802681785550061e-06, + "loss": 0.0031, + "step": 46602 + }, + { + "epoch": 6.1, + "learning_rate": 2.8023136163326435e-06, + "loss": 0.0018, + "step": 46603 + }, + { + "epoch": 6.1, + "learning_rate": 2.801945467358337e-06, + "loss": 0.0018, + "step": 46604 + }, + { + "epoch": 6.1, + "learning_rate": 2.8015773386281755e-06, + "loss": 0.0027, + "step": 46605 + }, + { + "epoch": 6.11, + "learning_rate": 2.8012092301431903e-06, + "loss": 0.0018, + "step": 46606 + }, + { + "epoch": 6.11, + "learning_rate": 2.8008411419044245e-06, + "loss": 0.0029, + "step": 46607 + }, + { + "epoch": 6.11, + "learning_rate": 2.8004730739129094e-06, + "loss": 0.0024, + "step": 46608 + }, + { + "epoch": 6.11, + "learning_rate": 2.800105026169677e-06, + "loss": 0.0029, + "step": 46609 + }, + { + "epoch": 6.11, + "learning_rate": 2.7997369986757695e-06, + "loss": 0.0012, + "step": 46610 + }, + { + "epoch": 6.11, + "learning_rate": 2.7993689914322177e-06, + "loss": 0.0026, + "step": 46611 + }, + { + "epoch": 6.11, + "learning_rate": 2.7990010044400553e-06, + "loss": 0.0013, + "step": 46612 + }, + { + "epoch": 6.11, + "learning_rate": 2.79863303770032e-06, + "loss": 0.0024, + "step": 46613 + }, + { + "epoch": 6.11, + "learning_rate": 2.798265091214044e-06, + "loss": 0.0031, + "step": 46614 + }, + { + "epoch": 6.11, + "learning_rate": 2.797897164982267e-06, + "loss": 0.0016, + "step": 46615 + }, + { + "epoch": 6.11, + "learning_rate": 2.797529259006021e-06, + "loss": 0.0013, + "step": 46616 + }, + { + "epoch": 6.11, + "learning_rate": 2.7971613732863366e-06, + "loss": 0.0012, + "step": 46617 + }, + { + "epoch": 6.11, + "learning_rate": 2.7967935078242557e-06, + "loss": 0.0025, + "step": 46618 + }, + { + "epoch": 6.11, + "learning_rate": 2.7964256626208097e-06, + "loss": 0.002, + "step": 46619 + }, + { + "epoch": 6.11, + "learning_rate": 2.796057837677033e-06, + "loss": 0.0019, + "step": 46620 + }, + { + "epoch": 6.11, + "learning_rate": 2.7956900329939563e-06, + "loss": 0.0019, + "step": 46621 + }, + { + "epoch": 6.11, + "learning_rate": 2.7953222485726218e-06, + "loss": 0.0023, + "step": 46622 + }, + { + "epoch": 6.11, + "learning_rate": 2.794954484414055e-06, + "loss": 0.0025, + "step": 46623 + }, + { + "epoch": 6.11, + "learning_rate": 2.7945867405192995e-06, + "loss": 0.0018, + "step": 46624 + }, + { + "epoch": 6.11, + "learning_rate": 2.7942190168893814e-06, + "loss": 0.0022, + "step": 46625 + }, + { + "epoch": 6.11, + "learning_rate": 2.793851313525342e-06, + "loss": 0.0011, + "step": 46626 + }, + { + "epoch": 6.11, + "learning_rate": 2.793483630428212e-06, + "loss": 0.0021, + "step": 46627 + }, + { + "epoch": 6.11, + "learning_rate": 2.7931159675990225e-06, + "loss": 0.0056, + "step": 46628 + }, + { + "epoch": 6.11, + "learning_rate": 2.7927483250388134e-06, + "loss": 0.0014, + "step": 46629 + }, + { + "epoch": 6.11, + "learning_rate": 2.792380702748616e-06, + "loss": 0.0036, + "step": 46630 + }, + { + "epoch": 6.11, + "learning_rate": 2.7920131007294637e-06, + "loss": 0.0027, + "step": 46631 + }, + { + "epoch": 6.11, + "learning_rate": 2.7916455189823886e-06, + "loss": 0.0018, + "step": 46632 + }, + { + "epoch": 6.11, + "learning_rate": 2.791277957508428e-06, + "loss": 0.0031, + "step": 46633 + }, + { + "epoch": 6.11, + "learning_rate": 2.790910416308613e-06, + "loss": 0.0022, + "step": 46634 + }, + { + "epoch": 6.11, + "learning_rate": 2.790542895383981e-06, + "loss": 0.0023, + "step": 46635 + }, + { + "epoch": 6.11, + "learning_rate": 2.790175394735559e-06, + "loss": 0.0029, + "step": 46636 + }, + { + "epoch": 6.11, + "learning_rate": 2.78980791436439e-06, + "loss": 0.0018, + "step": 46637 + }, + { + "epoch": 6.11, + "learning_rate": 2.7894404542715002e-06, + "loss": 0.0023, + "step": 46638 + }, + { + "epoch": 6.11, + "learning_rate": 2.7890730144579258e-06, + "loss": 0.0018, + "step": 46639 + }, + { + "epoch": 6.11, + "learning_rate": 2.788705594924697e-06, + "loss": 0.0012, + "step": 46640 + }, + { + "epoch": 6.11, + "learning_rate": 2.788338195672852e-06, + "loss": 0.0035, + "step": 46641 + }, + { + "epoch": 6.11, + "learning_rate": 2.7879708167034226e-06, + "loss": 0.0021, + "step": 46642 + }, + { + "epoch": 6.11, + "learning_rate": 2.7876034580174373e-06, + "loss": 0.0017, + "step": 46643 + }, + { + "epoch": 6.11, + "learning_rate": 2.787236119615937e-06, + "loss": 0.0022, + "step": 46644 + }, + { + "epoch": 6.11, + "learning_rate": 2.786868801499948e-06, + "loss": 0.0022, + "step": 46645 + }, + { + "epoch": 6.11, + "learning_rate": 2.786501503670509e-06, + "loss": 0.0031, + "step": 46646 + }, + { + "epoch": 6.11, + "learning_rate": 2.7861342261286496e-06, + "loss": 0.0015, + "step": 46647 + }, + { + "epoch": 6.11, + "learning_rate": 2.785766968875402e-06, + "loss": 0.0024, + "step": 46648 + }, + { + "epoch": 6.11, + "learning_rate": 2.785399731911802e-06, + "loss": 0.001, + "step": 46649 + }, + { + "epoch": 6.11, + "learning_rate": 2.7850325152388823e-06, + "loss": 0.0014, + "step": 46650 + }, + { + "epoch": 6.11, + "learning_rate": 2.7846653188576712e-06, + "loss": 0.0021, + "step": 46651 + }, + { + "epoch": 6.11, + "learning_rate": 2.784298142769207e-06, + "loss": 0.0015, + "step": 46652 + }, + { + "epoch": 6.11, + "learning_rate": 2.7839309869745213e-06, + "loss": 0.0016, + "step": 46653 + }, + { + "epoch": 6.11, + "learning_rate": 2.7835638514746444e-06, + "loss": 0.0016, + "step": 46654 + }, + { + "epoch": 6.11, + "learning_rate": 2.783196736270608e-06, + "loss": 0.0028, + "step": 46655 + }, + { + "epoch": 6.11, + "learning_rate": 2.7828296413634446e-06, + "loss": 0.0047, + "step": 46656 + }, + { + "epoch": 6.11, + "learning_rate": 2.7824625667541926e-06, + "loss": 0.0019, + "step": 46657 + }, + { + "epoch": 6.11, + "learning_rate": 2.7820955124438807e-06, + "loss": 0.0026, + "step": 46658 + }, + { + "epoch": 6.11, + "learning_rate": 2.7817284784335375e-06, + "loss": 0.0046, + "step": 46659 + }, + { + "epoch": 6.11, + "learning_rate": 2.7813614647242015e-06, + "loss": 0.002, + "step": 46660 + }, + { + "epoch": 6.11, + "learning_rate": 2.7809944713169014e-06, + "loss": 0.0013, + "step": 46661 + }, + { + "epoch": 6.11, + "learning_rate": 2.78062749821267e-06, + "loss": 0.0018, + "step": 46662 + }, + { + "epoch": 6.11, + "learning_rate": 2.7802605454125375e-06, + "loss": 0.0017, + "step": 46663 + }, + { + "epoch": 6.11, + "learning_rate": 2.7798936129175404e-06, + "loss": 0.002, + "step": 46664 + }, + { + "epoch": 6.11, + "learning_rate": 2.7795267007287064e-06, + "loss": 0.0013, + "step": 46665 + }, + { + "epoch": 6.11, + "learning_rate": 2.779159808847067e-06, + "loss": 0.0015, + "step": 46666 + }, + { + "epoch": 6.11, + "learning_rate": 2.7787929372736565e-06, + "loss": 0.0019, + "step": 46667 + }, + { + "epoch": 6.11, + "learning_rate": 2.778426086009509e-06, + "loss": 0.0012, + "step": 46668 + }, + { + "epoch": 6.11, + "learning_rate": 2.778059255055654e-06, + "loss": 0.0018, + "step": 46669 + }, + { + "epoch": 6.11, + "learning_rate": 2.7776924444131228e-06, + "loss": 0.003, + "step": 46670 + }, + { + "epoch": 6.11, + "learning_rate": 2.777325654082943e-06, + "loss": 0.0026, + "step": 46671 + }, + { + "epoch": 6.11, + "learning_rate": 2.7769588840661543e-06, + "loss": 0.0032, + "step": 46672 + }, + { + "epoch": 6.11, + "learning_rate": 2.7765921343637836e-06, + "loss": 0.0017, + "step": 46673 + }, + { + "epoch": 6.11, + "learning_rate": 2.77622540497686e-06, + "loss": 0.0017, + "step": 46674 + }, + { + "epoch": 6.11, + "learning_rate": 2.7758586959064203e-06, + "loss": 0.0038, + "step": 46675 + }, + { + "epoch": 6.11, + "learning_rate": 2.775492007153493e-06, + "loss": 0.0018, + "step": 46676 + }, + { + "epoch": 6.11, + "learning_rate": 2.7751253387191104e-06, + "loss": 0.0018, + "step": 46677 + }, + { + "epoch": 6.11, + "learning_rate": 2.7747586906042966e-06, + "loss": 0.0023, + "step": 46678 + }, + { + "epoch": 6.11, + "learning_rate": 2.7743920628100962e-06, + "loss": 0.0018, + "step": 46679 + }, + { + "epoch": 6.11, + "learning_rate": 2.774025455337532e-06, + "loss": 0.0024, + "step": 46680 + }, + { + "epoch": 6.11, + "learning_rate": 2.7736588681876374e-06, + "loss": 0.0021, + "step": 46681 + }, + { + "epoch": 6.12, + "learning_rate": 2.773292301361439e-06, + "loss": 0.0021, + "step": 46682 + }, + { + "epoch": 6.12, + "learning_rate": 2.772925754859974e-06, + "loss": 0.0029, + "step": 46683 + }, + { + "epoch": 6.12, + "learning_rate": 2.7725592286842716e-06, + "loss": 0.0016, + "step": 46684 + }, + { + "epoch": 6.12, + "learning_rate": 2.772192722835361e-06, + "loss": 0.0022, + "step": 46685 + }, + { + "epoch": 6.12, + "learning_rate": 2.771826237314269e-06, + "loss": 0.0022, + "step": 46686 + }, + { + "epoch": 6.12, + "learning_rate": 2.771459772122035e-06, + "loss": 0.0011, + "step": 46687 + }, + { + "epoch": 6.12, + "learning_rate": 2.771093327259685e-06, + "loss": 0.0035, + "step": 46688 + }, + { + "epoch": 6.12, + "learning_rate": 2.770726902728248e-06, + "loss": 0.0016, + "step": 46689 + }, + { + "epoch": 6.12, + "learning_rate": 2.770360498528756e-06, + "loss": 0.0013, + "step": 46690 + }, + { + "epoch": 6.12, + "learning_rate": 2.7699941146622423e-06, + "loss": 0.0014, + "step": 46691 + }, + { + "epoch": 6.12, + "learning_rate": 2.7696277511297363e-06, + "loss": 0.0026, + "step": 46692 + }, + { + "epoch": 6.12, + "learning_rate": 2.7692614079322633e-06, + "loss": 0.0011, + "step": 46693 + }, + { + "epoch": 6.12, + "learning_rate": 2.7688950850708616e-06, + "loss": 0.0015, + "step": 46694 + }, + { + "epoch": 6.12, + "learning_rate": 2.7685287825465567e-06, + "loss": 0.0031, + "step": 46695 + }, + { + "epoch": 6.12, + "learning_rate": 2.7681625003603794e-06, + "loss": 0.0021, + "step": 46696 + }, + { + "epoch": 6.12, + "learning_rate": 2.767796238513357e-06, + "loss": 0.002, + "step": 46697 + }, + { + "epoch": 6.12, + "learning_rate": 2.7674299970065254e-06, + "loss": 0.002, + "step": 46698 + }, + { + "epoch": 6.12, + "learning_rate": 2.767063775840909e-06, + "loss": 0.0012, + "step": 46699 + }, + { + "epoch": 6.12, + "learning_rate": 2.766697575017544e-06, + "loss": 0.0019, + "step": 46700 + }, + { + "epoch": 6.12, + "learning_rate": 2.7663313945374537e-06, + "loss": 0.0018, + "step": 46701 + }, + { + "epoch": 6.12, + "learning_rate": 2.765965234401674e-06, + "loss": 0.0012, + "step": 46702 + }, + { + "epoch": 6.12, + "learning_rate": 2.765599094611231e-06, + "loss": 0.0018, + "step": 46703 + }, + { + "epoch": 6.12, + "learning_rate": 2.765232975167157e-06, + "loss": 0.0011, + "step": 46704 + }, + { + "epoch": 6.12, + "learning_rate": 2.7648668760704754e-06, + "loss": 0.0029, + "step": 46705 + }, + { + "epoch": 6.12, + "learning_rate": 2.764500797322224e-06, + "loss": 0.0017, + "step": 46706 + }, + { + "epoch": 6.12, + "learning_rate": 2.7641347389234298e-06, + "loss": 0.0024, + "step": 46707 + }, + { + "epoch": 6.12, + "learning_rate": 2.763768700875117e-06, + "loss": 0.0017, + "step": 46708 + }, + { + "epoch": 6.12, + "learning_rate": 2.7634026831783222e-06, + "loss": 0.0018, + "step": 46709 + }, + { + "epoch": 6.12, + "learning_rate": 2.7630366858340696e-06, + "loss": 0.0032, + "step": 46710 + }, + { + "epoch": 6.12, + "learning_rate": 2.7626707088433935e-06, + "loss": 0.0021, + "step": 46711 + }, + { + "epoch": 6.12, + "learning_rate": 2.762304752207321e-06, + "loss": 0.0024, + "step": 46712 + }, + { + "epoch": 6.12, + "learning_rate": 2.761938815926879e-06, + "loss": 0.0022, + "step": 46713 + }, + { + "epoch": 6.12, + "learning_rate": 2.7615729000031e-06, + "loss": 0.0026, + "step": 46714 + }, + { + "epoch": 6.12, + "learning_rate": 2.761207004437012e-06, + "loss": 0.0017, + "step": 46715 + }, + { + "epoch": 6.12, + "learning_rate": 2.760841129229641e-06, + "loss": 0.0032, + "step": 46716 + }, + { + "epoch": 6.12, + "learning_rate": 2.760475274382023e-06, + "loss": 0.0015, + "step": 46717 + }, + { + "epoch": 6.12, + "learning_rate": 2.760109439895181e-06, + "loss": 0.0033, + "step": 46718 + }, + { + "epoch": 6.12, + "learning_rate": 2.759743625770146e-06, + "loss": 0.0038, + "step": 46719 + }, + { + "epoch": 6.12, + "learning_rate": 2.7593778320079446e-06, + "loss": 0.0029, + "step": 46720 + }, + { + "epoch": 6.12, + "learning_rate": 2.759012058609606e-06, + "loss": 0.0023, + "step": 46721 + }, + { + "epoch": 6.12, + "learning_rate": 2.758646305576165e-06, + "loss": 0.002, + "step": 46722 + }, + { + "epoch": 6.12, + "learning_rate": 2.7582805729086438e-06, + "loss": 0.0017, + "step": 46723 + }, + { + "epoch": 6.12, + "learning_rate": 2.757914860608071e-06, + "loss": 0.0021, + "step": 46724 + }, + { + "epoch": 6.12, + "learning_rate": 2.7575491686754795e-06, + "loss": 0.0016, + "step": 46725 + }, + { + "epoch": 6.12, + "learning_rate": 2.7571834971118962e-06, + "loss": 0.0019, + "step": 46726 + }, + { + "epoch": 6.12, + "learning_rate": 2.7568178459183483e-06, + "loss": 0.0014, + "step": 46727 + }, + { + "epoch": 6.12, + "learning_rate": 2.7564522150958606e-06, + "loss": 0.0026, + "step": 46728 + }, + { + "epoch": 6.12, + "learning_rate": 2.7560866046454695e-06, + "loss": 0.0028, + "step": 46729 + }, + { + "epoch": 6.12, + "learning_rate": 2.7557210145681977e-06, + "loss": 0.0027, + "step": 46730 + }, + { + "epoch": 6.12, + "learning_rate": 2.7553554448650723e-06, + "loss": 0.0012, + "step": 46731 + }, + { + "epoch": 6.12, + "learning_rate": 2.7549898955371245e-06, + "loss": 0.0018, + "step": 46732 + }, + { + "epoch": 6.12, + "learning_rate": 2.7546243665853844e-06, + "loss": 0.0021, + "step": 46733 + }, + { + "epoch": 6.12, + "learning_rate": 2.7542588580108777e-06, + "loss": 0.0017, + "step": 46734 + }, + { + "epoch": 6.12, + "learning_rate": 2.753893369814631e-06, + "loss": 0.0015, + "step": 46735 + }, + { + "epoch": 6.12, + "learning_rate": 2.7535279019976713e-06, + "loss": 0.0036, + "step": 46736 + }, + { + "epoch": 6.12, + "learning_rate": 2.7531624545610316e-06, + "loss": 0.0031, + "step": 46737 + }, + { + "epoch": 6.12, + "learning_rate": 2.752797027505736e-06, + "loss": 0.0026, + "step": 46738 + }, + { + "epoch": 6.12, + "learning_rate": 2.7524316208328105e-06, + "loss": 0.0015, + "step": 46739 + }, + { + "epoch": 6.12, + "learning_rate": 2.752066234543288e-06, + "loss": 0.0019, + "step": 46740 + }, + { + "epoch": 6.12, + "learning_rate": 2.7517008686381928e-06, + "loss": 0.0025, + "step": 46741 + }, + { + "epoch": 6.12, + "learning_rate": 2.7513355231185525e-06, + "loss": 0.0018, + "step": 46742 + }, + { + "epoch": 6.12, + "learning_rate": 2.7509701979853896e-06, + "loss": 0.0031, + "step": 46743 + }, + { + "epoch": 6.12, + "learning_rate": 2.750604893239743e-06, + "loss": 0.0027, + "step": 46744 + }, + { + "epoch": 6.12, + "learning_rate": 2.7502396088826342e-06, + "loss": 0.0019, + "step": 46745 + }, + { + "epoch": 6.12, + "learning_rate": 2.74987434491509e-06, + "loss": 0.0016, + "step": 46746 + }, + { + "epoch": 6.12, + "learning_rate": 2.749509101338136e-06, + "loss": 0.002, + "step": 46747 + }, + { + "epoch": 6.12, + "learning_rate": 2.7491438781528037e-06, + "loss": 0.0024, + "step": 46748 + }, + { + "epoch": 6.12, + "learning_rate": 2.7487786753601187e-06, + "loss": 0.0017, + "step": 46749 + }, + { + "epoch": 6.12, + "learning_rate": 2.748413492961107e-06, + "loss": 0.0026, + "step": 46750 + }, + { + "epoch": 6.12, + "learning_rate": 2.7480483309567927e-06, + "loss": 0.0021, + "step": 46751 + }, + { + "epoch": 6.12, + "learning_rate": 2.7476831893482105e-06, + "loss": 0.0018, + "step": 46752 + }, + { + "epoch": 6.12, + "learning_rate": 2.747318068136383e-06, + "loss": 0.0044, + "step": 46753 + }, + { + "epoch": 6.12, + "learning_rate": 2.746952967322334e-06, + "loss": 0.0015, + "step": 46754 + }, + { + "epoch": 6.12, + "learning_rate": 2.7465878869070927e-06, + "loss": 0.0012, + "step": 46755 + }, + { + "epoch": 6.12, + "learning_rate": 2.7462228268916903e-06, + "loss": 0.0016, + "step": 46756 + }, + { + "epoch": 6.12, + "learning_rate": 2.7458577872771497e-06, + "loss": 0.0017, + "step": 46757 + }, + { + "epoch": 6.12, + "learning_rate": 2.745492768064495e-06, + "loss": 0.0011, + "step": 46758 + }, + { + "epoch": 6.13, + "learning_rate": 2.7451277692547583e-06, + "loss": 0.0013, + "step": 46759 + }, + { + "epoch": 6.13, + "learning_rate": 2.7447627908489636e-06, + "loss": 0.0038, + "step": 46760 + }, + { + "epoch": 6.13, + "learning_rate": 2.744397832848138e-06, + "loss": 0.001, + "step": 46761 + }, + { + "epoch": 6.13, + "learning_rate": 2.7440328952533034e-06, + "loss": 0.0015, + "step": 46762 + }, + { + "epoch": 6.13, + "learning_rate": 2.7436679780654928e-06, + "loss": 0.0033, + "step": 46763 + }, + { + "epoch": 6.13, + "learning_rate": 2.7433030812857297e-06, + "loss": 0.0028, + "step": 46764 + }, + { + "epoch": 6.13, + "learning_rate": 2.7429382049150367e-06, + "loss": 0.0026, + "step": 46765 + }, + { + "epoch": 6.13, + "learning_rate": 2.742573348954445e-06, + "loss": 0.003, + "step": 46766 + }, + { + "epoch": 6.13, + "learning_rate": 2.7422085134049805e-06, + "loss": 0.0016, + "step": 46767 + }, + { + "epoch": 6.13, + "learning_rate": 2.741843698267669e-06, + "loss": 0.0017, + "step": 46768 + }, + { + "epoch": 6.13, + "learning_rate": 2.7414789035435353e-06, + "loss": 0.0019, + "step": 46769 + }, + { + "epoch": 6.13, + "learning_rate": 2.741114129233603e-06, + "loss": 0.0019, + "step": 46770 + }, + { + "epoch": 6.13, + "learning_rate": 2.7407493753389037e-06, + "loss": 0.0021, + "step": 46771 + }, + { + "epoch": 6.13, + "learning_rate": 2.7403846418604594e-06, + "loss": 0.0014, + "step": 46772 + }, + { + "epoch": 6.13, + "learning_rate": 2.7400199287992945e-06, + "loss": 0.0023, + "step": 46773 + }, + { + "epoch": 6.13, + "learning_rate": 2.73965523615644e-06, + "loss": 0.0021, + "step": 46774 + }, + { + "epoch": 6.13, + "learning_rate": 2.7392905639329182e-06, + "loss": 0.0026, + "step": 46775 + }, + { + "epoch": 6.13, + "learning_rate": 2.7389259121297516e-06, + "loss": 0.0017, + "step": 46776 + }, + { + "epoch": 6.13, + "learning_rate": 2.738561280747972e-06, + "loss": 0.0014, + "step": 46777 + }, + { + "epoch": 6.13, + "learning_rate": 2.7381966697886e-06, + "loss": 0.0021, + "step": 46778 + }, + { + "epoch": 6.13, + "learning_rate": 2.737832079252666e-06, + "loss": 0.0029, + "step": 46779 + }, + { + "epoch": 6.13, + "learning_rate": 2.737467509141193e-06, + "loss": 0.0023, + "step": 46780 + }, + { + "epoch": 6.13, + "learning_rate": 2.7371029594552013e-06, + "loss": 0.0013, + "step": 46781 + }, + { + "epoch": 6.13, + "learning_rate": 2.7367384301957254e-06, + "loss": 0.0011, + "step": 46782 + }, + { + "epoch": 6.13, + "learning_rate": 2.7363739213637853e-06, + "loss": 0.0018, + "step": 46783 + }, + { + "epoch": 6.13, + "learning_rate": 2.736009432960406e-06, + "loss": 0.0024, + "step": 46784 + }, + { + "epoch": 6.13, + "learning_rate": 2.7356449649866114e-06, + "loss": 0.0021, + "step": 46785 + }, + { + "epoch": 6.13, + "learning_rate": 2.7352805174434305e-06, + "loss": 0.0034, + "step": 46786 + }, + { + "epoch": 6.13, + "learning_rate": 2.734916090331884e-06, + "loss": 0.004, + "step": 46787 + }, + { + "epoch": 6.13, + "learning_rate": 2.7345516836530016e-06, + "loss": 0.0018, + "step": 46788 + }, + { + "epoch": 6.13, + "learning_rate": 2.734187297407803e-06, + "loss": 0.0021, + "step": 46789 + }, + { + "epoch": 6.13, + "learning_rate": 2.73382293159732e-06, + "loss": 0.0019, + "step": 46790 + }, + { + "epoch": 6.13, + "learning_rate": 2.7334585862225715e-06, + "loss": 0.0011, + "step": 46791 + }, + { + "epoch": 6.13, + "learning_rate": 2.7330942612845836e-06, + "loss": 0.0022, + "step": 46792 + }, + { + "epoch": 6.13, + "learning_rate": 2.732729956784379e-06, + "loss": 0.0017, + "step": 46793 + }, + { + "epoch": 6.13, + "learning_rate": 2.7323656727229865e-06, + "loss": 0.0019, + "step": 46794 + }, + { + "epoch": 6.13, + "learning_rate": 2.732001409101429e-06, + "loss": 0.0018, + "step": 46795 + }, + { + "epoch": 6.13, + "learning_rate": 2.7316371659207276e-06, + "loss": 0.0014, + "step": 46796 + }, + { + "epoch": 6.13, + "learning_rate": 2.731272943181913e-06, + "loss": 0.0021, + "step": 46797 + }, + { + "epoch": 6.13, + "learning_rate": 2.7309087408860023e-06, + "loss": 0.0037, + "step": 46798 + }, + { + "epoch": 6.13, + "learning_rate": 2.730544559034026e-06, + "loss": 0.0029, + "step": 46799 + }, + { + "epoch": 6.13, + "learning_rate": 2.730180397627008e-06, + "loss": 0.0012, + "step": 46800 + }, + { + "epoch": 6.13, + "learning_rate": 2.729816256665966e-06, + "loss": 0.0016, + "step": 46801 + }, + { + "epoch": 6.13, + "learning_rate": 2.729452136151932e-06, + "loss": 0.002, + "step": 46802 + }, + { + "epoch": 6.13, + "learning_rate": 2.7290880360859273e-06, + "loss": 0.0036, + "step": 46803 + }, + { + "epoch": 6.13, + "learning_rate": 2.7287239564689715e-06, + "loss": 0.0016, + "step": 46804 + }, + { + "epoch": 6.13, + "learning_rate": 2.7283598973020953e-06, + "loss": 0.0021, + "step": 46805 + }, + { + "epoch": 6.13, + "learning_rate": 2.727995858586321e-06, + "loss": 0.0016, + "step": 46806 + }, + { + "epoch": 6.13, + "learning_rate": 2.72763184032267e-06, + "loss": 0.0029, + "step": 46807 + }, + { + "epoch": 6.13, + "learning_rate": 2.7272678425121633e-06, + "loss": 0.0019, + "step": 46808 + }, + { + "epoch": 6.13, + "learning_rate": 2.72690386515583e-06, + "loss": 0.0022, + "step": 46809 + }, + { + "epoch": 6.13, + "learning_rate": 2.726539908254695e-06, + "loss": 0.0019, + "step": 46810 + }, + { + "epoch": 6.13, + "learning_rate": 2.726175971809779e-06, + "loss": 0.0016, + "step": 46811 + }, + { + "epoch": 6.13, + "learning_rate": 2.7258120558221022e-06, + "loss": 0.0015, + "step": 46812 + }, + { + "epoch": 6.13, + "learning_rate": 2.725448160292695e-06, + "loss": 0.002, + "step": 46813 + }, + { + "epoch": 6.13, + "learning_rate": 2.725084285222578e-06, + "loss": 0.0031, + "step": 46814 + }, + { + "epoch": 6.13, + "learning_rate": 2.724720430612773e-06, + "loss": 0.0016, + "step": 46815 + }, + { + "epoch": 6.13, + "learning_rate": 2.7243565964643026e-06, + "loss": 0.0017, + "step": 46816 + }, + { + "epoch": 6.13, + "learning_rate": 2.7239927827781933e-06, + "loss": 0.0017, + "step": 46817 + }, + { + "epoch": 6.13, + "learning_rate": 2.7236289895554678e-06, + "loss": 0.0026, + "step": 46818 + }, + { + "epoch": 6.13, + "learning_rate": 2.723265216797145e-06, + "loss": 0.0021, + "step": 46819 + }, + { + "epoch": 6.13, + "learning_rate": 2.722901464504252e-06, + "loss": 0.0024, + "step": 46820 + }, + { + "epoch": 6.13, + "learning_rate": 2.7225377326778134e-06, + "loss": 0.0017, + "step": 46821 + }, + { + "epoch": 6.13, + "learning_rate": 2.7221740213188495e-06, + "loss": 0.0024, + "step": 46822 + }, + { + "epoch": 6.13, + "learning_rate": 2.7218103304283817e-06, + "loss": 0.0018, + "step": 46823 + }, + { + "epoch": 6.13, + "learning_rate": 2.7214466600074363e-06, + "loss": 0.0016, + "step": 46824 + }, + { + "epoch": 6.13, + "learning_rate": 2.7210830100570363e-06, + "loss": 0.0035, + "step": 46825 + }, + { + "epoch": 6.13, + "learning_rate": 2.7207193805782017e-06, + "loss": 0.0015, + "step": 46826 + }, + { + "epoch": 6.13, + "learning_rate": 2.720355771571953e-06, + "loss": 0.0021, + "step": 46827 + }, + { + "epoch": 6.13, + "learning_rate": 2.7199921830393196e-06, + "loss": 0.0015, + "step": 46828 + }, + { + "epoch": 6.13, + "learning_rate": 2.71962861498132e-06, + "loss": 0.002, + "step": 46829 + }, + { + "epoch": 6.13, + "learning_rate": 2.719265067398974e-06, + "loss": 0.0015, + "step": 46830 + }, + { + "epoch": 6.13, + "learning_rate": 2.7189015402933085e-06, + "loss": 0.0022, + "step": 46831 + }, + { + "epoch": 6.13, + "learning_rate": 2.7185380336653466e-06, + "loss": 0.002, + "step": 46832 + }, + { + "epoch": 6.13, + "learning_rate": 2.7181745475161092e-06, + "loss": 0.0016, + "step": 46833 + }, + { + "epoch": 6.13, + "learning_rate": 2.7178110818466174e-06, + "loss": 0.001, + "step": 46834 + }, + { + "epoch": 6.14, + "learning_rate": 2.7174476366578918e-06, + "loss": 0.0021, + "step": 46835 + }, + { + "epoch": 6.14, + "learning_rate": 2.71708421195096e-06, + "loss": 0.0027, + "step": 46836 + }, + { + "epoch": 6.14, + "learning_rate": 2.71672080772684e-06, + "loss": 0.0036, + "step": 46837 + }, + { + "epoch": 6.14, + "learning_rate": 2.716357423986553e-06, + "loss": 0.0034, + "step": 46838 + }, + { + "epoch": 6.14, + "learning_rate": 2.715994060731126e-06, + "loss": 0.0013, + "step": 46839 + }, + { + "epoch": 6.14, + "learning_rate": 2.715630717961577e-06, + "loss": 0.0011, + "step": 46840 + }, + { + "epoch": 6.14, + "learning_rate": 2.7152673956789266e-06, + "loss": 0.0017, + "step": 46841 + }, + { + "epoch": 6.14, + "learning_rate": 2.7149040938842007e-06, + "loss": 0.0022, + "step": 46842 + }, + { + "epoch": 6.14, + "learning_rate": 2.7145408125784166e-06, + "loss": 0.0017, + "step": 46843 + }, + { + "epoch": 6.14, + "learning_rate": 2.7141775517626013e-06, + "loss": 0.0014, + "step": 46844 + }, + { + "epoch": 6.14, + "learning_rate": 2.7138143114377734e-06, + "loss": 0.0014, + "step": 46845 + }, + { + "epoch": 6.14, + "learning_rate": 2.7134510916049527e-06, + "loss": 0.0018, + "step": 46846 + }, + { + "epoch": 6.14, + "learning_rate": 2.713087892265165e-06, + "loss": 0.002, + "step": 46847 + }, + { + "epoch": 6.14, + "learning_rate": 2.71272471341943e-06, + "loss": 0.0013, + "step": 46848 + }, + { + "epoch": 6.14, + "learning_rate": 2.7123615550687677e-06, + "loss": 0.0021, + "step": 46849 + }, + { + "epoch": 6.14, + "learning_rate": 2.711998417214199e-06, + "loss": 0.0019, + "step": 46850 + }, + { + "epoch": 6.14, + "learning_rate": 2.7116352998567487e-06, + "loss": 0.0013, + "step": 46851 + }, + { + "epoch": 6.14, + "learning_rate": 2.7112722029974325e-06, + "loss": 0.0025, + "step": 46852 + }, + { + "epoch": 6.14, + "learning_rate": 2.710909126637279e-06, + "loss": 0.0025, + "step": 46853 + }, + { + "epoch": 6.14, + "learning_rate": 2.710546070777301e-06, + "loss": 0.0024, + "step": 46854 + }, + { + "epoch": 6.14, + "learning_rate": 2.7101830354185286e-06, + "loss": 0.0015, + "step": 46855 + }, + { + "epoch": 6.14, + "learning_rate": 2.7098200205619772e-06, + "loss": 0.0023, + "step": 46856 + }, + { + "epoch": 6.14, + "learning_rate": 2.7094570262086693e-06, + "loss": 0.0026, + "step": 46857 + }, + { + "epoch": 6.14, + "learning_rate": 2.709094052359623e-06, + "loss": 0.0021, + "step": 46858 + }, + { + "epoch": 6.14, + "learning_rate": 2.7087310990158635e-06, + "loss": 0.0019, + "step": 46859 + }, + { + "epoch": 6.14, + "learning_rate": 2.70836816617841e-06, + "loss": 0.0018, + "step": 46860 + }, + { + "epoch": 6.14, + "learning_rate": 2.7080052538482794e-06, + "loss": 0.0021, + "step": 46861 + }, + { + "epoch": 6.14, + "learning_rate": 2.707642362026498e-06, + "loss": 0.0023, + "step": 46862 + }, + { + "epoch": 6.14, + "learning_rate": 2.707279490714082e-06, + "loss": 0.0018, + "step": 46863 + }, + { + "epoch": 6.14, + "learning_rate": 2.706916639912056e-06, + "loss": 0.0029, + "step": 46864 + }, + { + "epoch": 6.14, + "learning_rate": 2.7065538096214406e-06, + "loss": 0.0019, + "step": 46865 + }, + { + "epoch": 6.14, + "learning_rate": 2.706190999843249e-06, + "loss": 0.002, + "step": 46866 + }, + { + "epoch": 6.14, + "learning_rate": 2.7058282105785117e-06, + "loss": 0.0028, + "step": 46867 + }, + { + "epoch": 6.14, + "learning_rate": 2.705465441828242e-06, + "loss": 0.0049, + "step": 46868 + }, + { + "epoch": 6.14, + "learning_rate": 2.705102693593461e-06, + "loss": 0.0016, + "step": 46869 + }, + { + "epoch": 6.14, + "learning_rate": 2.7047399658751927e-06, + "loss": 0.0021, + "step": 46870 + }, + { + "epoch": 6.14, + "learning_rate": 2.7043772586744553e-06, + "loss": 0.0021, + "step": 46871 + }, + { + "epoch": 6.14, + "learning_rate": 2.7040145719922672e-06, + "loss": 0.0024, + "step": 46872 + }, + { + "epoch": 6.14, + "learning_rate": 2.703651905829647e-06, + "loss": 0.0052, + "step": 46873 + }, + { + "epoch": 6.14, + "learning_rate": 2.7032892601876182e-06, + "loss": 0.001, + "step": 46874 + }, + { + "epoch": 6.14, + "learning_rate": 2.702926635067201e-06, + "loss": 0.0019, + "step": 46875 + }, + { + "epoch": 6.14, + "learning_rate": 2.7025640304694154e-06, + "loss": 0.0024, + "step": 46876 + }, + { + "epoch": 6.14, + "learning_rate": 2.7022014463952773e-06, + "loss": 0.0027, + "step": 46877 + }, + { + "epoch": 6.14, + "learning_rate": 2.701838882845811e-06, + "loss": 0.0023, + "step": 46878 + }, + { + "epoch": 6.14, + "learning_rate": 2.7014763398220346e-06, + "loss": 0.0025, + "step": 46879 + }, + { + "epoch": 6.14, + "learning_rate": 2.7011138173249675e-06, + "loss": 0.0023, + "step": 46880 + }, + { + "epoch": 6.14, + "learning_rate": 2.7007513153556252e-06, + "loss": 0.0017, + "step": 46881 + }, + { + "epoch": 6.14, + "learning_rate": 2.700388833915034e-06, + "loss": 0.0021, + "step": 46882 + }, + { + "epoch": 6.14, + "learning_rate": 2.700026373004212e-06, + "loss": 0.0019, + "step": 46883 + }, + { + "epoch": 6.14, + "learning_rate": 2.6996639326241724e-06, + "loss": 0.0029, + "step": 46884 + }, + { + "epoch": 6.14, + "learning_rate": 2.6993015127759403e-06, + "loss": 0.0015, + "step": 46885 + }, + { + "epoch": 6.14, + "learning_rate": 2.698939113460537e-06, + "loss": 0.002, + "step": 46886 + }, + { + "epoch": 6.14, + "learning_rate": 2.698576734678978e-06, + "loss": 0.0017, + "step": 46887 + }, + { + "epoch": 6.14, + "learning_rate": 2.698214376432281e-06, + "loss": 0.0029, + "step": 46888 + }, + { + "epoch": 6.14, + "learning_rate": 2.697852038721469e-06, + "loss": 0.0016, + "step": 46889 + }, + { + "epoch": 6.14, + "learning_rate": 2.6974897215475605e-06, + "loss": 0.0018, + "step": 46890 + }, + { + "epoch": 6.14, + "learning_rate": 2.697127424911572e-06, + "loss": 0.0015, + "step": 46891 + }, + { + "epoch": 6.14, + "learning_rate": 2.696765148814522e-06, + "loss": 0.0032, + "step": 46892 + }, + { + "epoch": 6.14, + "learning_rate": 2.6964028932574327e-06, + "loss": 0.0036, + "step": 46893 + }, + { + "epoch": 6.14, + "learning_rate": 2.6960406582413222e-06, + "loss": 0.0019, + "step": 46894 + }, + { + "epoch": 6.14, + "learning_rate": 2.6956784437672057e-06, + "loss": 0.0015, + "step": 46895 + }, + { + "epoch": 6.14, + "learning_rate": 2.695316249836104e-06, + "loss": 0.0023, + "step": 46896 + }, + { + "epoch": 6.14, + "learning_rate": 2.69495407644904e-06, + "loss": 0.0014, + "step": 46897 + }, + { + "epoch": 6.14, + "learning_rate": 2.6945919236070283e-06, + "loss": 0.0023, + "step": 46898 + }, + { + "epoch": 6.14, + "learning_rate": 2.694229791311087e-06, + "loss": 0.0023, + "step": 46899 + }, + { + "epoch": 6.14, + "learning_rate": 2.6938676795622333e-06, + "loss": 0.0026, + "step": 46900 + }, + { + "epoch": 6.14, + "learning_rate": 2.6935055883614904e-06, + "loss": 0.0019, + "step": 46901 + }, + { + "epoch": 6.14, + "learning_rate": 2.693143517709874e-06, + "loss": 0.0018, + "step": 46902 + }, + { + "epoch": 6.14, + "learning_rate": 2.6927814676083997e-06, + "loss": 0.0015, + "step": 46903 + }, + { + "epoch": 6.14, + "learning_rate": 2.69241943805809e-06, + "loss": 0.0025, + "step": 46904 + }, + { + "epoch": 6.14, + "learning_rate": 2.6920574290599632e-06, + "loss": 0.002, + "step": 46905 + }, + { + "epoch": 6.14, + "learning_rate": 2.6916954406150353e-06, + "loss": 0.0019, + "step": 46906 + }, + { + "epoch": 6.14, + "learning_rate": 2.6913334727243214e-06, + "loss": 0.0018, + "step": 46907 + }, + { + "epoch": 6.14, + "learning_rate": 2.6909715253888424e-06, + "loss": 0.0019, + "step": 46908 + }, + { + "epoch": 6.14, + "learning_rate": 2.69060959860962e-06, + "loss": 0.0025, + "step": 46909 + }, + { + "epoch": 6.14, + "learning_rate": 2.6902476923876695e-06, + "loss": 0.0028, + "step": 46910 + }, + { + "epoch": 6.15, + "learning_rate": 2.689885806724005e-06, + "loss": 0.0024, + "step": 46911 + }, + { + "epoch": 6.15, + "learning_rate": 2.6895239416196505e-06, + "loss": 0.0035, + "step": 46912 + }, + { + "epoch": 6.15, + "learning_rate": 2.6891620970756194e-06, + "loss": 0.002, + "step": 46913 + }, + { + "epoch": 6.15, + "learning_rate": 2.688800273092932e-06, + "loss": 0.0021, + "step": 46914 + }, + { + "epoch": 6.15, + "learning_rate": 2.6884384696726006e-06, + "loss": 0.0012, + "step": 46915 + }, + { + "epoch": 6.15, + "learning_rate": 2.6880766868156504e-06, + "loss": 0.0014, + "step": 46916 + }, + { + "epoch": 6.15, + "learning_rate": 2.687714924523095e-06, + "loss": 0.0023, + "step": 46917 + }, + { + "epoch": 6.15, + "learning_rate": 2.687353182795949e-06, + "loss": 0.0034, + "step": 46918 + }, + { + "epoch": 6.15, + "learning_rate": 2.686991461635231e-06, + "loss": 0.0024, + "step": 46919 + }, + { + "epoch": 6.15, + "learning_rate": 2.6866297610419655e-06, + "loss": 0.0043, + "step": 46920 + }, + { + "epoch": 6.15, + "learning_rate": 2.6862680810171627e-06, + "loss": 0.0022, + "step": 46921 + }, + { + "epoch": 6.15, + "learning_rate": 2.6859064215618423e-06, + "loss": 0.0029, + "step": 46922 + }, + { + "epoch": 6.15, + "learning_rate": 2.6855447826770177e-06, + "loss": 0.0021, + "step": 46923 + }, + { + "epoch": 6.15, + "learning_rate": 2.6851831643637118e-06, + "loss": 0.0024, + "step": 46924 + }, + { + "epoch": 6.15, + "learning_rate": 2.6848215666229383e-06, + "loss": 0.0014, + "step": 46925 + }, + { + "epoch": 6.15, + "learning_rate": 2.6844599894557112e-06, + "loss": 0.0024, + "step": 46926 + }, + { + "epoch": 6.15, + "learning_rate": 2.6840984328630547e-06, + "loss": 0.0018, + "step": 46927 + }, + { + "epoch": 6.15, + "learning_rate": 2.6837368968459808e-06, + "loss": 0.0016, + "step": 46928 + }, + { + "epoch": 6.15, + "learning_rate": 2.683375381405505e-06, + "loss": 0.0018, + "step": 46929 + }, + { + "epoch": 6.15, + "learning_rate": 2.6830138865426493e-06, + "loss": 0.0027, + "step": 46930 + }, + { + "epoch": 6.15, + "learning_rate": 2.682652412258424e-06, + "loss": 0.0021, + "step": 46931 + }, + { + "epoch": 6.15, + "learning_rate": 2.6822909585538526e-06, + "loss": 0.0016, + "step": 46932 + }, + { + "epoch": 6.15, + "learning_rate": 2.6819295254299484e-06, + "loss": 0.003, + "step": 46933 + }, + { + "epoch": 6.15, + "learning_rate": 2.6815681128877247e-06, + "loss": 0.0017, + "step": 46934 + }, + { + "epoch": 6.15, + "learning_rate": 2.6812067209282033e-06, + "loss": 0.0017, + "step": 46935 + }, + { + "epoch": 6.15, + "learning_rate": 2.680845349552398e-06, + "loss": 0.0041, + "step": 46936 + }, + { + "epoch": 6.15, + "learning_rate": 2.680483998761326e-06, + "loss": 0.0017, + "step": 46937 + }, + { + "epoch": 6.15, + "learning_rate": 2.680122668556e-06, + "loss": 0.0034, + "step": 46938 + }, + { + "epoch": 6.15, + "learning_rate": 2.679761358937442e-06, + "loss": 0.0028, + "step": 46939 + }, + { + "epoch": 6.15, + "learning_rate": 2.6794000699066626e-06, + "loss": 0.0016, + "step": 46940 + }, + { + "epoch": 6.15, + "learning_rate": 2.679038801464683e-06, + "loss": 0.0018, + "step": 46941 + }, + { + "epoch": 6.15, + "learning_rate": 2.678677553612514e-06, + "loss": 0.0021, + "step": 46942 + }, + { + "epoch": 6.15, + "learning_rate": 2.678316326351178e-06, + "loss": 0.0019, + "step": 46943 + }, + { + "epoch": 6.15, + "learning_rate": 2.677955119681688e-06, + "loss": 0.0031, + "step": 46944 + }, + { + "epoch": 6.15, + "learning_rate": 2.677593933605055e-06, + "loss": 0.0015, + "step": 46945 + }, + { + "epoch": 6.15, + "learning_rate": 2.6772327681223033e-06, + "loss": 0.0016, + "step": 46946 + }, + { + "epoch": 6.15, + "learning_rate": 2.6768716232344437e-06, + "loss": 0.0013, + "step": 46947 + }, + { + "epoch": 6.15, + "learning_rate": 2.676510498942493e-06, + "loss": 0.0015, + "step": 46948 + }, + { + "epoch": 6.15, + "learning_rate": 2.6761493952474626e-06, + "loss": 0.0028, + "step": 46949 + }, + { + "epoch": 6.15, + "learning_rate": 2.6757883121503757e-06, + "loss": 0.0022, + "step": 46950 + }, + { + "epoch": 6.15, + "learning_rate": 2.6754272496522415e-06, + "loss": 0.0032, + "step": 46951 + }, + { + "epoch": 6.15, + "learning_rate": 2.675066207754081e-06, + "loss": 0.003, + "step": 46952 + }, + { + "epoch": 6.15, + "learning_rate": 2.674705186456904e-06, + "loss": 0.0021, + "step": 46953 + }, + { + "epoch": 6.15, + "learning_rate": 2.6743441857617304e-06, + "loss": 0.0058, + "step": 46954 + }, + { + "epoch": 6.15, + "learning_rate": 2.6739832056695746e-06, + "loss": 0.001, + "step": 46955 + }, + { + "epoch": 6.15, + "learning_rate": 2.673622246181451e-06, + "loss": 0.0023, + "step": 46956 + }, + { + "epoch": 6.15, + "learning_rate": 2.6732613072983714e-06, + "loss": 0.0017, + "step": 46957 + }, + { + "epoch": 6.15, + "learning_rate": 2.6729003890213566e-06, + "loss": 0.0019, + "step": 46958 + }, + { + "epoch": 6.15, + "learning_rate": 2.67253949135142e-06, + "loss": 0.0021, + "step": 46959 + }, + { + "epoch": 6.15, + "learning_rate": 2.6721786142895723e-06, + "loss": 0.0015, + "step": 46960 + }, + { + "epoch": 6.15, + "learning_rate": 2.6718177578368333e-06, + "loss": 0.002, + "step": 46961 + }, + { + "epoch": 6.15, + "learning_rate": 2.671456921994219e-06, + "loss": 0.0021, + "step": 46962 + }, + { + "epoch": 6.15, + "learning_rate": 2.6710961067627407e-06, + "loss": 0.003, + "step": 46963 + }, + { + "epoch": 6.15, + "learning_rate": 2.6707353121434153e-06, + "loss": 0.0031, + "step": 46964 + }, + { + "epoch": 6.15, + "learning_rate": 2.6703745381372537e-06, + "loss": 0.0022, + "step": 46965 + }, + { + "epoch": 6.15, + "learning_rate": 2.670013784745276e-06, + "loss": 0.0037, + "step": 46966 + }, + { + "epoch": 6.15, + "learning_rate": 2.6696530519684937e-06, + "loss": 0.0025, + "step": 46967 + }, + { + "epoch": 6.15, + "learning_rate": 2.6692923398079198e-06, + "loss": 0.0021, + "step": 46968 + }, + { + "epoch": 6.15, + "learning_rate": 2.6689316482645733e-06, + "loss": 0.0022, + "step": 46969 + }, + { + "epoch": 6.15, + "learning_rate": 2.6685709773394665e-06, + "loss": 0.0011, + "step": 46970 + }, + { + "epoch": 6.15, + "learning_rate": 2.6682103270336137e-06, + "loss": 0.0015, + "step": 46971 + }, + { + "epoch": 6.15, + "learning_rate": 2.667849697348025e-06, + "loss": 0.0019, + "step": 46972 + }, + { + "epoch": 6.15, + "learning_rate": 2.667489088283719e-06, + "loss": 0.0025, + "step": 46973 + }, + { + "epoch": 6.15, + "learning_rate": 2.6671284998417123e-06, + "loss": 0.0019, + "step": 46974 + }, + { + "epoch": 6.15, + "learning_rate": 2.6667679320230153e-06, + "loss": 0.0027, + "step": 46975 + }, + { + "epoch": 6.15, + "learning_rate": 2.666407384828642e-06, + "loss": 0.0014, + "step": 46976 + }, + { + "epoch": 6.15, + "learning_rate": 2.6660468582596087e-06, + "loss": 0.0024, + "step": 46977 + }, + { + "epoch": 6.15, + "learning_rate": 2.6656863523169286e-06, + "loss": 0.0021, + "step": 46978 + }, + { + "epoch": 6.15, + "learning_rate": 2.6653258670016158e-06, + "loss": 0.0012, + "step": 46979 + }, + { + "epoch": 6.15, + "learning_rate": 2.66496540231468e-06, + "loss": 0.0014, + "step": 46980 + }, + { + "epoch": 6.15, + "learning_rate": 2.664604958257141e-06, + "loss": 0.0016, + "step": 46981 + }, + { + "epoch": 6.15, + "learning_rate": 2.6642445348300095e-06, + "loss": 0.0018, + "step": 46982 + }, + { + "epoch": 6.15, + "learning_rate": 2.6638841320342977e-06, + "loss": 0.0016, + "step": 46983 + }, + { + "epoch": 6.15, + "learning_rate": 2.6635237498710197e-06, + "loss": 0.0013, + "step": 46984 + }, + { + "epoch": 6.15, + "learning_rate": 2.6631633883411945e-06, + "loss": 0.003, + "step": 46985 + }, + { + "epoch": 6.15, + "learning_rate": 2.6628030474458323e-06, + "loss": 0.0018, + "step": 46986 + }, + { + "epoch": 6.15, + "learning_rate": 2.6624427271859443e-06, + "loss": 0.0015, + "step": 46987 + }, + { + "epoch": 6.16, + "learning_rate": 2.6620824275625436e-06, + "loss": 0.0014, + "step": 46988 + }, + { + "epoch": 6.16, + "learning_rate": 2.6617221485766475e-06, + "loss": 0.0023, + "step": 46989 + }, + { + "epoch": 6.16, + "learning_rate": 2.661361890229267e-06, + "loss": 0.0021, + "step": 46990 + }, + { + "epoch": 6.16, + "learning_rate": 2.6610016525214132e-06, + "loss": 0.002, + "step": 46991 + }, + { + "epoch": 6.16, + "learning_rate": 2.6606414354541033e-06, + "loss": 0.002, + "step": 46992 + }, + { + "epoch": 6.16, + "learning_rate": 2.6602812390283493e-06, + "loss": 0.004, + "step": 46993 + }, + { + "epoch": 6.16, + "learning_rate": 2.65992106324516e-06, + "loss": 0.0018, + "step": 46994 + }, + { + "epoch": 6.16, + "learning_rate": 2.6595609081055517e-06, + "loss": 0.0019, + "step": 46995 + }, + { + "epoch": 6.16, + "learning_rate": 2.659200773610542e-06, + "loss": 0.0022, + "step": 46996 + }, + { + "epoch": 6.16, + "learning_rate": 2.658840659761137e-06, + "loss": 0.0021, + "step": 46997 + }, + { + "epoch": 6.16, + "learning_rate": 2.6584805665583536e-06, + "loss": 0.0014, + "step": 46998 + }, + { + "epoch": 6.16, + "learning_rate": 2.658120494003198e-06, + "loss": 0.0018, + "step": 46999 + }, + { + "epoch": 6.16, + "learning_rate": 2.657760442096692e-06, + "loss": 0.0009, + "step": 47000 + }, + { + "epoch": 6.16, + "learning_rate": 2.6574004108398422e-06, + "loss": 0.0019, + "step": 47001 + }, + { + "epoch": 6.16, + "learning_rate": 2.6570404002336635e-06, + "loss": 0.0026, + "step": 47002 + }, + { + "epoch": 6.16, + "learning_rate": 2.6566804102791653e-06, + "loss": 0.0026, + "step": 47003 + }, + { + "epoch": 6.16, + "learning_rate": 2.6563204409773645e-06, + "loss": 0.0038, + "step": 47004 + }, + { + "epoch": 6.16, + "learning_rate": 2.655960492329269e-06, + "loss": 0.0016, + "step": 47005 + }, + { + "epoch": 6.16, + "learning_rate": 2.6556005643358964e-06, + "loss": 0.0021, + "step": 47006 + }, + { + "epoch": 6.16, + "learning_rate": 2.6552406569982526e-06, + "loss": 0.0014, + "step": 47007 + }, + { + "epoch": 6.16, + "learning_rate": 2.6548807703173575e-06, + "loss": 0.0019, + "step": 47008 + }, + { + "epoch": 6.16, + "learning_rate": 2.654520904294218e-06, + "loss": 0.0022, + "step": 47009 + }, + { + "epoch": 6.16, + "learning_rate": 2.654161058929844e-06, + "loss": 0.0018, + "step": 47010 + }, + { + "epoch": 6.16, + "learning_rate": 2.653801234225254e-06, + "loss": 0.0017, + "step": 47011 + }, + { + "epoch": 6.16, + "learning_rate": 2.653441430181458e-06, + "loss": 0.0021, + "step": 47012 + }, + { + "epoch": 6.16, + "learning_rate": 2.653081646799466e-06, + "loss": 0.0029, + "step": 47013 + }, + { + "epoch": 6.16, + "learning_rate": 2.652721884080287e-06, + "loss": 0.0017, + "step": 47014 + }, + { + "epoch": 6.16, + "learning_rate": 2.65236214202494e-06, + "loss": 0.001, + "step": 47015 + }, + { + "epoch": 6.16, + "learning_rate": 2.6520024206344307e-06, + "loss": 0.0024, + "step": 47016 + }, + { + "epoch": 6.16, + "learning_rate": 2.651642719909775e-06, + "loss": 0.004, + "step": 47017 + }, + { + "epoch": 6.16, + "learning_rate": 2.651283039851981e-06, + "loss": 0.002, + "step": 47018 + }, + { + "epoch": 6.16, + "learning_rate": 2.650923380462065e-06, + "loss": 0.0028, + "step": 47019 + }, + { + "epoch": 6.16, + "learning_rate": 2.6505637417410346e-06, + "loss": 0.002, + "step": 47020 + }, + { + "epoch": 6.16, + "learning_rate": 2.6502041236899035e-06, + "loss": 0.0027, + "step": 47021 + }, + { + "epoch": 6.16, + "learning_rate": 2.649844526309678e-06, + "loss": 0.0035, + "step": 47022 + }, + { + "epoch": 6.16, + "learning_rate": 2.6494849496013774e-06, + "loss": 0.0019, + "step": 47023 + }, + { + "epoch": 6.16, + "learning_rate": 2.6491253935660076e-06, + "loss": 0.0022, + "step": 47024 + }, + { + "epoch": 6.16, + "learning_rate": 2.6487658582045796e-06, + "loss": 0.0017, + "step": 47025 + }, + { + "epoch": 6.16, + "learning_rate": 2.6484063435181086e-06, + "loss": 0.0025, + "step": 47026 + }, + { + "epoch": 6.16, + "learning_rate": 2.6480468495075996e-06, + "loss": 0.0024, + "step": 47027 + }, + { + "epoch": 6.16, + "learning_rate": 2.647687376174071e-06, + "loss": 0.0019, + "step": 47028 + }, + { + "epoch": 6.16, + "learning_rate": 2.64732792351853e-06, + "loss": 0.0024, + "step": 47029 + }, + { + "epoch": 6.16, + "learning_rate": 2.6469684915419846e-06, + "loss": 0.0017, + "step": 47030 + }, + { + "epoch": 6.16, + "learning_rate": 2.646609080245451e-06, + "loss": 0.0023, + "step": 47031 + }, + { + "epoch": 6.16, + "learning_rate": 2.6462496896299396e-06, + "loss": 0.0022, + "step": 47032 + }, + { + "epoch": 6.16, + "learning_rate": 2.6458903196964547e-06, + "loss": 0.0029, + "step": 47033 + }, + { + "epoch": 6.16, + "learning_rate": 2.645530970446015e-06, + "loss": 0.0041, + "step": 47034 + }, + { + "epoch": 6.16, + "learning_rate": 2.645171641879628e-06, + "loss": 0.0014, + "step": 47035 + }, + { + "epoch": 6.16, + "learning_rate": 2.6448123339983044e-06, + "loss": 0.0022, + "step": 47036 + }, + { + "epoch": 6.16, + "learning_rate": 2.6444530468030516e-06, + "loss": 0.0018, + "step": 47037 + }, + { + "epoch": 6.16, + "learning_rate": 2.644093780294882e-06, + "loss": 0.0022, + "step": 47038 + }, + { + "epoch": 6.16, + "learning_rate": 2.64373453447481e-06, + "loss": 0.0012, + "step": 47039 + }, + { + "epoch": 6.16, + "learning_rate": 2.643375309343843e-06, + "loss": 0.0022, + "step": 47040 + }, + { + "epoch": 6.16, + "learning_rate": 2.643016104902989e-06, + "loss": 0.0017, + "step": 47041 + }, + { + "epoch": 6.16, + "learning_rate": 2.6426569211532626e-06, + "loss": 0.0017, + "step": 47042 + }, + { + "epoch": 6.16, + "learning_rate": 2.642297758095672e-06, + "loss": 0.0028, + "step": 47043 + }, + { + "epoch": 6.16, + "learning_rate": 2.6419386157312275e-06, + "loss": 0.0038, + "step": 47044 + }, + { + "epoch": 6.16, + "learning_rate": 2.6415794940609353e-06, + "loss": 0.0034, + "step": 47045 + }, + { + "epoch": 6.16, + "learning_rate": 2.6412203930858116e-06, + "loss": 0.0024, + "step": 47046 + }, + { + "epoch": 6.16, + "learning_rate": 2.6408613128068648e-06, + "loss": 0.0023, + "step": 47047 + }, + { + "epoch": 6.16, + "learning_rate": 2.6405022532251003e-06, + "loss": 0.0023, + "step": 47048 + }, + { + "epoch": 6.16, + "learning_rate": 2.6401432143415307e-06, + "loss": 0.0021, + "step": 47049 + }, + { + "epoch": 6.16, + "learning_rate": 2.639784196157169e-06, + "loss": 0.0027, + "step": 47050 + }, + { + "epoch": 6.16, + "learning_rate": 2.6394251986730234e-06, + "loss": 0.0019, + "step": 47051 + }, + { + "epoch": 6.16, + "learning_rate": 2.639066221890102e-06, + "loss": 0.0027, + "step": 47052 + }, + { + "epoch": 6.16, + "learning_rate": 2.6387072658094114e-06, + "loss": 0.0021, + "step": 47053 + }, + { + "epoch": 6.16, + "learning_rate": 2.638348330431969e-06, + "loss": 0.0023, + "step": 47054 + }, + { + "epoch": 6.16, + "learning_rate": 2.6379894157587783e-06, + "loss": 0.0021, + "step": 47055 + }, + { + "epoch": 6.16, + "learning_rate": 2.6376305217908484e-06, + "loss": 0.0033, + "step": 47056 + }, + { + "epoch": 6.16, + "learning_rate": 2.6372716485291925e-06, + "loss": 0.0024, + "step": 47057 + }, + { + "epoch": 6.16, + "learning_rate": 2.6369127959748198e-06, + "loss": 0.0021, + "step": 47058 + }, + { + "epoch": 6.16, + "learning_rate": 2.6365539641287365e-06, + "loss": 0.0026, + "step": 47059 + }, + { + "epoch": 6.16, + "learning_rate": 2.636195152991947e-06, + "loss": 0.0019, + "step": 47060 + }, + { + "epoch": 6.16, + "learning_rate": 2.6358363625654726e-06, + "loss": 0.0016, + "step": 47061 + }, + { + "epoch": 6.16, + "learning_rate": 2.635477592850316e-06, + "loss": 0.0015, + "step": 47062 + }, + { + "epoch": 6.16, + "learning_rate": 2.635118843847487e-06, + "loss": 0.003, + "step": 47063 + }, + { + "epoch": 6.17, + "learning_rate": 2.6347601155579906e-06, + "loss": 0.0019, + "step": 47064 + }, + { + "epoch": 6.17, + "learning_rate": 2.6344014079828427e-06, + "loss": 0.0028, + "step": 47065 + }, + { + "epoch": 6.17, + "learning_rate": 2.634042721123048e-06, + "loss": 0.0023, + "step": 47066 + }, + { + "epoch": 6.17, + "learning_rate": 2.633684054979616e-06, + "loss": 0.0027, + "step": 47067 + }, + { + "epoch": 6.17, + "learning_rate": 2.6333254095535534e-06, + "loss": 0.0028, + "step": 47068 + }, + { + "epoch": 6.17, + "learning_rate": 2.6329667848458728e-06, + "loss": 0.0015, + "step": 47069 + }, + { + "epoch": 6.17, + "learning_rate": 2.632608180857581e-06, + "loss": 0.002, + "step": 47070 + }, + { + "epoch": 6.17, + "learning_rate": 2.6322495975896834e-06, + "loss": 0.002, + "step": 47071 + }, + { + "epoch": 6.17, + "learning_rate": 2.6318910350431914e-06, + "loss": 0.0034, + "step": 47072 + }, + { + "epoch": 6.17, + "learning_rate": 2.6315324932191157e-06, + "loss": 0.0012, + "step": 47073 + }, + { + "epoch": 6.17, + "learning_rate": 2.631173972118464e-06, + "loss": 0.0021, + "step": 47074 + }, + { + "epoch": 6.17, + "learning_rate": 2.6308154717422386e-06, + "loss": 0.0029, + "step": 47075 + }, + { + "epoch": 6.17, + "learning_rate": 2.6304569920914557e-06, + "loss": 0.0022, + "step": 47076 + }, + { + "epoch": 6.17, + "learning_rate": 2.63009853316712e-06, + "loss": 0.0028, + "step": 47077 + }, + { + "epoch": 6.17, + "learning_rate": 2.62974009497024e-06, + "loss": 0.0018, + "step": 47078 + }, + { + "epoch": 6.17, + "learning_rate": 2.629381677501821e-06, + "loss": 0.0018, + "step": 47079 + }, + { + "epoch": 6.17, + "learning_rate": 2.629023280762876e-06, + "loss": 0.0022, + "step": 47080 + }, + { + "epoch": 6.17, + "learning_rate": 2.6286649047544113e-06, + "loss": 0.0019, + "step": 47081 + }, + { + "epoch": 6.17, + "learning_rate": 2.6283065494774295e-06, + "loss": 0.0016, + "step": 47082 + }, + { + "epoch": 6.17, + "learning_rate": 2.6279482149329437e-06, + "loss": 0.0011, + "step": 47083 + }, + { + "epoch": 6.17, + "learning_rate": 2.627589901121964e-06, + "loss": 0.0034, + "step": 47084 + }, + { + "epoch": 6.17, + "learning_rate": 2.6272316080454953e-06, + "loss": 0.0019, + "step": 47085 + }, + { + "epoch": 6.17, + "learning_rate": 2.626873335704544e-06, + "loss": 0.0018, + "step": 47086 + }, + { + "epoch": 6.17, + "learning_rate": 2.6265150841001174e-06, + "loss": 0.0018, + "step": 47087 + }, + { + "epoch": 6.17, + "learning_rate": 2.626156853233226e-06, + "loss": 0.0016, + "step": 47088 + }, + { + "epoch": 6.17, + "learning_rate": 2.6257986431048764e-06, + "loss": 0.0033, + "step": 47089 + }, + { + "epoch": 6.17, + "learning_rate": 2.6254404537160727e-06, + "loss": 0.0019, + "step": 47090 + }, + { + "epoch": 6.17, + "learning_rate": 2.625082285067828e-06, + "loss": 0.0019, + "step": 47091 + }, + { + "epoch": 6.17, + "learning_rate": 2.624724137161143e-06, + "loss": 0.0012, + "step": 47092 + }, + { + "epoch": 6.17, + "learning_rate": 2.6243660099970324e-06, + "loss": 0.0024, + "step": 47093 + }, + { + "epoch": 6.17, + "learning_rate": 2.624007903576499e-06, + "loss": 0.0031, + "step": 47094 + }, + { + "epoch": 6.17, + "learning_rate": 2.6236498179005476e-06, + "loss": 0.0032, + "step": 47095 + }, + { + "epoch": 6.17, + "learning_rate": 2.6232917529701916e-06, + "loss": 0.0022, + "step": 47096 + }, + { + "epoch": 6.17, + "learning_rate": 2.6229337087864347e-06, + "loss": 0.0039, + "step": 47097 + }, + { + "epoch": 6.17, + "learning_rate": 2.62257568535028e-06, + "loss": 0.0015, + "step": 47098 + }, + { + "epoch": 6.17, + "learning_rate": 2.622217682662743e-06, + "loss": 0.0019, + "step": 47099 + }, + { + "epoch": 6.17, + "learning_rate": 2.621859700724825e-06, + "loss": 0.0016, + "step": 47100 + }, + { + "epoch": 6.17, + "learning_rate": 2.621501739537534e-06, + "loss": 0.0032, + "step": 47101 + }, + { + "epoch": 6.17, + "learning_rate": 2.621143799101873e-06, + "loss": 0.002, + "step": 47102 + }, + { + "epoch": 6.17, + "learning_rate": 2.6207858794188524e-06, + "loss": 0.0029, + "step": 47103 + }, + { + "epoch": 6.17, + "learning_rate": 2.6204279804894806e-06, + "loss": 0.0022, + "step": 47104 + }, + { + "epoch": 6.17, + "learning_rate": 2.6200701023147632e-06, + "loss": 0.0015, + "step": 47105 + }, + { + "epoch": 6.17, + "learning_rate": 2.6197122448957022e-06, + "loss": 0.0018, + "step": 47106 + }, + { + "epoch": 6.17, + "learning_rate": 2.619354408233311e-06, + "loss": 0.0018, + "step": 47107 + }, + { + "epoch": 6.17, + "learning_rate": 2.618996592328592e-06, + "loss": 0.0014, + "step": 47108 + }, + { + "epoch": 6.17, + "learning_rate": 2.6186387971825524e-06, + "loss": 0.002, + "step": 47109 + }, + { + "epoch": 6.17, + "learning_rate": 2.618281022796194e-06, + "loss": 0.0024, + "step": 47110 + }, + { + "epoch": 6.17, + "learning_rate": 2.6179232691705314e-06, + "loss": 0.0031, + "step": 47111 + }, + { + "epoch": 6.17, + "learning_rate": 2.6175655363065654e-06, + "loss": 0.0029, + "step": 47112 + }, + { + "epoch": 6.17, + "learning_rate": 2.617207824205301e-06, + "loss": 0.0028, + "step": 47113 + }, + { + "epoch": 6.17, + "learning_rate": 2.6168501328677465e-06, + "loss": 0.0031, + "step": 47114 + }, + { + "epoch": 6.17, + "learning_rate": 2.6164924622949104e-06, + "loss": 0.0027, + "step": 47115 + }, + { + "epoch": 6.17, + "learning_rate": 2.616134812487796e-06, + "loss": 0.0022, + "step": 47116 + }, + { + "epoch": 6.17, + "learning_rate": 2.61577718344741e-06, + "loss": 0.0019, + "step": 47117 + }, + { + "epoch": 6.17, + "learning_rate": 2.6154195751747534e-06, + "loss": 0.0023, + "step": 47118 + }, + { + "epoch": 6.17, + "learning_rate": 2.6150619876708395e-06, + "loss": 0.0029, + "step": 47119 + }, + { + "epoch": 6.17, + "learning_rate": 2.614704420936671e-06, + "loss": 0.0018, + "step": 47120 + }, + { + "epoch": 6.17, + "learning_rate": 2.6143468749732494e-06, + "loss": 0.0031, + "step": 47121 + }, + { + "epoch": 6.17, + "learning_rate": 2.613989349781587e-06, + "loss": 0.0016, + "step": 47122 + }, + { + "epoch": 6.17, + "learning_rate": 2.6136318453626864e-06, + "loss": 0.0023, + "step": 47123 + }, + { + "epoch": 6.17, + "learning_rate": 2.6132743617175526e-06, + "loss": 0.0024, + "step": 47124 + }, + { + "epoch": 6.17, + "learning_rate": 2.6129168988471855e-06, + "loss": 0.002, + "step": 47125 + }, + { + "epoch": 6.17, + "learning_rate": 2.6125594567526016e-06, + "loss": 0.0034, + "step": 47126 + }, + { + "epoch": 6.17, + "learning_rate": 2.612202035434802e-06, + "loss": 0.0029, + "step": 47127 + }, + { + "epoch": 6.17, + "learning_rate": 2.61184463489479e-06, + "loss": 0.0017, + "step": 47128 + }, + { + "epoch": 6.17, + "learning_rate": 2.611487255133568e-06, + "loss": 0.0025, + "step": 47129 + }, + { + "epoch": 6.17, + "learning_rate": 2.611129896152149e-06, + "loss": 0.001, + "step": 47130 + }, + { + "epoch": 6.17, + "learning_rate": 2.610772557951533e-06, + "loss": 0.002, + "step": 47131 + }, + { + "epoch": 6.17, + "learning_rate": 2.6104152405327255e-06, + "loss": 0.0036, + "step": 47132 + }, + { + "epoch": 6.17, + "learning_rate": 2.610057943896729e-06, + "loss": 0.0023, + "step": 47133 + }, + { + "epoch": 6.17, + "learning_rate": 2.6097006680445536e-06, + "loss": 0.0016, + "step": 47134 + }, + { + "epoch": 6.17, + "learning_rate": 2.6093434129772012e-06, + "loss": 0.0016, + "step": 47135 + }, + { + "epoch": 6.17, + "learning_rate": 2.608986178695675e-06, + "loss": 0.0016, + "step": 47136 + }, + { + "epoch": 6.17, + "learning_rate": 2.60862896520098e-06, + "loss": 0.0024, + "step": 47137 + }, + { + "epoch": 6.17, + "learning_rate": 2.6082717724941266e-06, + "loss": 0.0013, + "step": 47138 + }, + { + "epoch": 6.17, + "learning_rate": 2.6079146005761146e-06, + "loss": 0.002, + "step": 47139 + }, + { + "epoch": 6.18, + "learning_rate": 2.6075574494479474e-06, + "loss": 0.002, + "step": 47140 + }, + { + "epoch": 6.18, + "learning_rate": 2.6072003191106333e-06, + "loss": 0.0016, + "step": 47141 + }, + { + "epoch": 6.18, + "learning_rate": 2.606843209565174e-06, + "loss": 0.0015, + "step": 47142 + }, + { + "epoch": 6.18, + "learning_rate": 2.6064861208125756e-06, + "loss": 0.0017, + "step": 47143 + }, + { + "epoch": 6.18, + "learning_rate": 2.6061290528538374e-06, + "loss": 0.0019, + "step": 47144 + }, + { + "epoch": 6.18, + "learning_rate": 2.6057720056899703e-06, + "loss": 0.0019, + "step": 47145 + }, + { + "epoch": 6.18, + "learning_rate": 2.605414979321976e-06, + "loss": 0.0025, + "step": 47146 + }, + { + "epoch": 6.18, + "learning_rate": 2.6050579737508563e-06, + "loss": 0.0017, + "step": 47147 + }, + { + "epoch": 6.18, + "learning_rate": 2.6047009889776166e-06, + "loss": 0.0016, + "step": 47148 + }, + { + "epoch": 6.18, + "learning_rate": 2.6043440250032638e-06, + "loss": 0.0023, + "step": 47149 + }, + { + "epoch": 6.18, + "learning_rate": 2.6039870818288006e-06, + "loss": 0.002, + "step": 47150 + }, + { + "epoch": 6.18, + "learning_rate": 2.603630159455229e-06, + "loss": 0.0026, + "step": 47151 + }, + { + "epoch": 6.18, + "learning_rate": 2.603273257883552e-06, + "loss": 0.0023, + "step": 47152 + }, + { + "epoch": 6.18, + "learning_rate": 2.602916377114777e-06, + "loss": 0.0018, + "step": 47153 + }, + { + "epoch": 6.18, + "learning_rate": 2.6025595171499052e-06, + "loss": 0.0029, + "step": 47154 + }, + { + "epoch": 6.18, + "learning_rate": 2.602202677989939e-06, + "loss": 0.002, + "step": 47155 + }, + { + "epoch": 6.18, + "learning_rate": 2.6018458596358865e-06, + "loss": 0.0015, + "step": 47156 + }, + { + "epoch": 6.18, + "learning_rate": 2.601489062088748e-06, + "loss": 0.003, + "step": 47157 + }, + { + "epoch": 6.18, + "learning_rate": 2.601132285349525e-06, + "loss": 0.0031, + "step": 47158 + }, + { + "epoch": 6.18, + "learning_rate": 2.600775529419226e-06, + "loss": 0.0026, + "step": 47159 + }, + { + "epoch": 6.18, + "learning_rate": 2.60041879429885e-06, + "loss": 0.0024, + "step": 47160 + }, + { + "epoch": 6.18, + "learning_rate": 2.600062079989404e-06, + "loss": 0.0021, + "step": 47161 + }, + { + "epoch": 6.18, + "learning_rate": 2.59970538649189e-06, + "loss": 0.0021, + "step": 47162 + }, + { + "epoch": 6.18, + "learning_rate": 2.599348713807307e-06, + "loss": 0.0031, + "step": 47163 + }, + { + "epoch": 6.18, + "learning_rate": 2.598992061936665e-06, + "loss": 0.0018, + "step": 47164 + }, + { + "epoch": 6.18, + "learning_rate": 2.5986354308809625e-06, + "loss": 0.0022, + "step": 47165 + }, + { + "epoch": 6.18, + "learning_rate": 2.598278820641205e-06, + "loss": 0.001, + "step": 47166 + }, + { + "epoch": 6.18, + "learning_rate": 2.5979222312183905e-06, + "loss": 0.0024, + "step": 47167 + }, + { + "epoch": 6.18, + "learning_rate": 2.5975656626135293e-06, + "loss": 0.0024, + "step": 47168 + }, + { + "epoch": 6.18, + "learning_rate": 2.597209114827617e-06, + "loss": 0.0019, + "step": 47169 + }, + { + "epoch": 6.18, + "learning_rate": 2.596852587861662e-06, + "loss": 0.0021, + "step": 47170 + }, + { + "epoch": 6.18, + "learning_rate": 2.596496081716663e-06, + "loss": 0.0023, + "step": 47171 + }, + { + "epoch": 6.18, + "learning_rate": 2.5961395963936277e-06, + "loss": 0.0017, + "step": 47172 + }, + { + "epoch": 6.18, + "learning_rate": 2.5957831318935546e-06, + "loss": 0.0033, + "step": 47173 + }, + { + "epoch": 6.18, + "learning_rate": 2.595426688217448e-06, + "loss": 0.0011, + "step": 47174 + }, + { + "epoch": 6.18, + "learning_rate": 2.595070265366306e-06, + "loss": 0.0026, + "step": 47175 + }, + { + "epoch": 6.18, + "learning_rate": 2.5947138633411373e-06, + "loss": 0.0025, + "step": 47176 + }, + { + "epoch": 6.18, + "learning_rate": 2.5943574821429417e-06, + "loss": 0.0039, + "step": 47177 + }, + { + "epoch": 6.18, + "learning_rate": 2.594001121772719e-06, + "loss": 0.0016, + "step": 47178 + }, + { + "epoch": 6.18, + "learning_rate": 2.593644782231476e-06, + "loss": 0.0034, + "step": 47179 + }, + { + "epoch": 6.18, + "learning_rate": 2.5932884635202104e-06, + "loss": 0.0021, + "step": 47180 + }, + { + "epoch": 6.18, + "learning_rate": 2.5929321656399286e-06, + "loss": 0.0021, + "step": 47181 + }, + { + "epoch": 6.18, + "learning_rate": 2.5925758885916307e-06, + "loss": 0.0019, + "step": 47182 + }, + { + "epoch": 6.18, + "learning_rate": 2.592219632376316e-06, + "loss": 0.003, + "step": 47183 + }, + { + "epoch": 6.18, + "learning_rate": 2.5918633969949926e-06, + "loss": 0.0015, + "step": 47184 + }, + { + "epoch": 6.18, + "learning_rate": 2.591507182448658e-06, + "loss": 0.0028, + "step": 47185 + }, + { + "epoch": 6.18, + "learning_rate": 2.5911509887383124e-06, + "loss": 0.0018, + "step": 47186 + }, + { + "epoch": 6.18, + "learning_rate": 2.590794815864963e-06, + "loss": 0.0014, + "step": 47187 + }, + { + "epoch": 6.18, + "learning_rate": 2.5904386638296086e-06, + "loss": 0.0011, + "step": 47188 + }, + { + "epoch": 6.18, + "learning_rate": 2.5900825326332513e-06, + "loss": 0.0027, + "step": 47189 + }, + { + "epoch": 6.18, + "learning_rate": 2.589726422276888e-06, + "loss": 0.0022, + "step": 47190 + }, + { + "epoch": 6.18, + "learning_rate": 2.589370332761526e-06, + "loss": 0.0037, + "step": 47191 + }, + { + "epoch": 6.18, + "learning_rate": 2.589014264088168e-06, + "loss": 0.001, + "step": 47192 + }, + { + "epoch": 6.18, + "learning_rate": 2.588658216257812e-06, + "loss": 0.0027, + "step": 47193 + }, + { + "epoch": 6.18, + "learning_rate": 2.5883021892714576e-06, + "loss": 0.0017, + "step": 47194 + }, + { + "epoch": 6.18, + "learning_rate": 2.587946183130111e-06, + "loss": 0.0031, + "step": 47195 + }, + { + "epoch": 6.18, + "learning_rate": 2.5875901978347716e-06, + "loss": 0.0016, + "step": 47196 + }, + { + "epoch": 6.18, + "learning_rate": 2.5872342333864397e-06, + "loss": 0.0031, + "step": 47197 + }, + { + "epoch": 6.18, + "learning_rate": 2.5868782897861144e-06, + "loss": 0.0015, + "step": 47198 + }, + { + "epoch": 6.18, + "learning_rate": 2.5865223670348017e-06, + "loss": 0.002, + "step": 47199 + }, + { + "epoch": 6.18, + "learning_rate": 2.5861664651335006e-06, + "loss": 0.0021, + "step": 47200 + }, + { + "epoch": 6.18, + "learning_rate": 2.585810584083208e-06, + "loss": 0.0028, + "step": 47201 + }, + { + "epoch": 6.18, + "learning_rate": 2.585454723884928e-06, + "loss": 0.0021, + "step": 47202 + }, + { + "epoch": 6.18, + "learning_rate": 2.5850988845396652e-06, + "loss": 0.0015, + "step": 47203 + }, + { + "epoch": 6.18, + "learning_rate": 2.584743066048416e-06, + "loss": 0.0026, + "step": 47204 + }, + { + "epoch": 6.18, + "learning_rate": 2.584387268412181e-06, + "loss": 0.0025, + "step": 47205 + }, + { + "epoch": 6.18, + "learning_rate": 2.584031491631963e-06, + "loss": 0.0019, + "step": 47206 + }, + { + "epoch": 6.18, + "learning_rate": 2.5836757357087625e-06, + "loss": 0.0019, + "step": 47207 + }, + { + "epoch": 6.18, + "learning_rate": 2.5833200006435786e-06, + "loss": 0.0023, + "step": 47208 + }, + { + "epoch": 6.18, + "learning_rate": 2.5829642864374094e-06, + "loss": 0.0024, + "step": 47209 + }, + { + "epoch": 6.18, + "learning_rate": 2.5826085930912612e-06, + "loss": 0.0029, + "step": 47210 + }, + { + "epoch": 6.18, + "learning_rate": 2.582252920606132e-06, + "loss": 0.0011, + "step": 47211 + }, + { + "epoch": 6.18, + "learning_rate": 2.581897268983017e-06, + "loss": 0.0032, + "step": 47212 + }, + { + "epoch": 6.18, + "learning_rate": 2.581541638222922e-06, + "loss": 0.0033, + "step": 47213 + }, + { + "epoch": 6.18, + "learning_rate": 2.5811860283268485e-06, + "loss": 0.0021, + "step": 47214 + }, + { + "epoch": 6.18, + "learning_rate": 2.580830439295795e-06, + "loss": 0.002, + "step": 47215 + }, + { + "epoch": 6.18, + "learning_rate": 2.5804748711307604e-06, + "loss": 0.0018, + "step": 47216 + }, + { + "epoch": 6.19, + "learning_rate": 2.5801193238327414e-06, + "loss": 0.002, + "step": 47217 + }, + { + "epoch": 6.19, + "learning_rate": 2.5797637974027457e-06, + "loss": 0.0016, + "step": 47218 + }, + { + "epoch": 6.19, + "learning_rate": 2.5794082918417686e-06, + "loss": 0.0023, + "step": 47219 + }, + { + "epoch": 6.19, + "learning_rate": 2.579052807150807e-06, + "loss": 0.0028, + "step": 47220 + }, + { + "epoch": 6.19, + "learning_rate": 2.578697343330868e-06, + "loss": 0.0024, + "step": 47221 + }, + { + "epoch": 6.19, + "learning_rate": 2.578341900382948e-06, + "loss": 0.0014, + "step": 47222 + }, + { + "epoch": 6.19, + "learning_rate": 2.5779864783080456e-06, + "loss": 0.0022, + "step": 47223 + }, + { + "epoch": 6.19, + "learning_rate": 2.577631077107158e-06, + "loss": 0.0016, + "step": 47224 + }, + { + "epoch": 6.19, + "learning_rate": 2.5772756967812873e-06, + "loss": 0.0028, + "step": 47225 + }, + { + "epoch": 6.19, + "learning_rate": 2.5769203373314366e-06, + "loss": 0.0029, + "step": 47226 + }, + { + "epoch": 6.19, + "learning_rate": 2.5765649987586026e-06, + "loss": 0.0018, + "step": 47227 + }, + { + "epoch": 6.19, + "learning_rate": 2.5762096810637806e-06, + "loss": 0.0016, + "step": 47228 + }, + { + "epoch": 6.19, + "learning_rate": 2.575854384247977e-06, + "loss": 0.0021, + "step": 47229 + }, + { + "epoch": 6.19, + "learning_rate": 2.575499108312187e-06, + "loss": 0.0028, + "step": 47230 + }, + { + "epoch": 6.19, + "learning_rate": 2.5751438532574115e-06, + "loss": 0.001, + "step": 47231 + }, + { + "epoch": 6.19, + "learning_rate": 2.574788619084644e-06, + "loss": 0.002, + "step": 47232 + }, + { + "epoch": 6.19, + "learning_rate": 2.574433405794893e-06, + "loss": 0.0027, + "step": 47233 + }, + { + "epoch": 6.19, + "learning_rate": 2.5740782133891474e-06, + "loss": 0.0019, + "step": 47234 + }, + { + "epoch": 6.19, + "learning_rate": 2.5737230418684155e-06, + "loss": 0.003, + "step": 47235 + }, + { + "epoch": 6.19, + "learning_rate": 2.573367891233689e-06, + "loss": 0.0014, + "step": 47236 + }, + { + "epoch": 6.19, + "learning_rate": 2.5730127614859735e-06, + "loss": 0.0022, + "step": 47237 + }, + { + "epoch": 6.19, + "learning_rate": 2.5726576526262636e-06, + "loss": 0.0021, + "step": 47238 + }, + { + "epoch": 6.19, + "learning_rate": 2.5723025646555577e-06, + "loss": 0.0022, + "step": 47239 + }, + { + "epoch": 6.19, + "learning_rate": 2.571947497574854e-06, + "loss": 0.0021, + "step": 47240 + }, + { + "epoch": 6.19, + "learning_rate": 2.5715924513851543e-06, + "loss": 0.0023, + "step": 47241 + }, + { + "epoch": 6.19, + "learning_rate": 2.571237426087455e-06, + "loss": 0.0033, + "step": 47242 + }, + { + "epoch": 6.19, + "learning_rate": 2.570882421682752e-06, + "loss": 0.0036, + "step": 47243 + }, + { + "epoch": 6.19, + "learning_rate": 2.57052743817205e-06, + "loss": 0.0014, + "step": 47244 + }, + { + "epoch": 6.19, + "learning_rate": 2.5701724755563407e-06, + "loss": 0.0024, + "step": 47245 + }, + { + "epoch": 6.19, + "learning_rate": 2.5698175338366283e-06, + "loss": 0.0022, + "step": 47246 + }, + { + "epoch": 6.19, + "learning_rate": 2.569462613013909e-06, + "loss": 0.0036, + "step": 47247 + }, + { + "epoch": 6.19, + "learning_rate": 2.5691077130891773e-06, + "loss": 0.0014, + "step": 47248 + }, + { + "epoch": 6.19, + "learning_rate": 2.568752834063437e-06, + "loss": 0.0019, + "step": 47249 + }, + { + "epoch": 6.19, + "learning_rate": 2.5683979759376843e-06, + "loss": 0.0023, + "step": 47250 + }, + { + "epoch": 6.19, + "learning_rate": 2.5680431387129123e-06, + "loss": 0.0032, + "step": 47251 + }, + { + "epoch": 6.19, + "learning_rate": 2.567688322390127e-06, + "loss": 0.0026, + "step": 47252 + }, + { + "epoch": 6.19, + "learning_rate": 2.5673335269703226e-06, + "loss": 0.0017, + "step": 47253 + }, + { + "epoch": 6.19, + "learning_rate": 2.5669787524544977e-06, + "loss": 0.0021, + "step": 47254 + }, + { + "epoch": 6.19, + "learning_rate": 2.566623998843645e-06, + "loss": 0.0031, + "step": 47255 + }, + { + "epoch": 6.19, + "learning_rate": 2.566269266138767e-06, + "loss": 0.0022, + "step": 47256 + }, + { + "epoch": 6.19, + "learning_rate": 2.5659145543408625e-06, + "loss": 0.002, + "step": 47257 + }, + { + "epoch": 6.19, + "learning_rate": 2.5655598634509293e-06, + "loss": 0.0013, + "step": 47258 + }, + { + "epoch": 6.19, + "learning_rate": 2.5652051934699594e-06, + "loss": 0.0021, + "step": 47259 + }, + { + "epoch": 6.19, + "learning_rate": 2.5648505443989567e-06, + "loss": 0.002, + "step": 47260 + }, + { + "epoch": 6.19, + "learning_rate": 2.564495916238916e-06, + "loss": 0.0015, + "step": 47261 + }, + { + "epoch": 6.19, + "learning_rate": 2.5641413089908317e-06, + "loss": 0.0017, + "step": 47262 + }, + { + "epoch": 6.19, + "learning_rate": 2.5637867226557067e-06, + "loss": 0.0015, + "step": 47263 + }, + { + "epoch": 6.19, + "learning_rate": 2.5634321572345353e-06, + "loss": 0.0018, + "step": 47264 + }, + { + "epoch": 6.19, + "learning_rate": 2.563077612728315e-06, + "loss": 0.0016, + "step": 47265 + }, + { + "epoch": 6.19, + "learning_rate": 2.5627230891380404e-06, + "loss": 0.003, + "step": 47266 + }, + { + "epoch": 6.19, + "learning_rate": 2.562368586464712e-06, + "loss": 0.0037, + "step": 47267 + }, + { + "epoch": 6.19, + "learning_rate": 2.562014104709327e-06, + "loss": 0.0029, + "step": 47268 + }, + { + "epoch": 6.19, + "learning_rate": 2.5616596438728825e-06, + "loss": 0.0026, + "step": 47269 + }, + { + "epoch": 6.19, + "learning_rate": 2.5613052039563704e-06, + "loss": 0.0028, + "step": 47270 + }, + { + "epoch": 6.19, + "learning_rate": 2.560950784960795e-06, + "loss": 0.0028, + "step": 47271 + }, + { + "epoch": 6.19, + "learning_rate": 2.5605963868871488e-06, + "loss": 0.003, + "step": 47272 + }, + { + "epoch": 6.19, + "learning_rate": 2.560242009736429e-06, + "loss": 0.0018, + "step": 47273 + }, + { + "epoch": 6.19, + "learning_rate": 2.5598876535096295e-06, + "loss": 0.0034, + "step": 47274 + }, + { + "epoch": 6.19, + "learning_rate": 2.5595333182077532e-06, + "loss": 0.0032, + "step": 47275 + }, + { + "epoch": 6.19, + "learning_rate": 2.559179003831792e-06, + "loss": 0.0011, + "step": 47276 + }, + { + "epoch": 6.19, + "learning_rate": 2.558824710382741e-06, + "loss": 0.0013, + "step": 47277 + }, + { + "epoch": 6.19, + "learning_rate": 2.5584704378615998e-06, + "loss": 0.0028, + "step": 47278 + }, + { + "epoch": 6.19, + "learning_rate": 2.558116186269368e-06, + "loss": 0.0041, + "step": 47279 + }, + { + "epoch": 6.19, + "learning_rate": 2.5577619556070364e-06, + "loss": 0.0015, + "step": 47280 + }, + { + "epoch": 6.19, + "learning_rate": 2.557407745875604e-06, + "loss": 0.0013, + "step": 47281 + }, + { + "epoch": 6.19, + "learning_rate": 2.5570535570760623e-06, + "loss": 0.0019, + "step": 47282 + }, + { + "epoch": 6.19, + "learning_rate": 2.556699389209415e-06, + "loss": 0.0026, + "step": 47283 + }, + { + "epoch": 6.19, + "learning_rate": 2.5563452422766545e-06, + "loss": 0.0028, + "step": 47284 + }, + { + "epoch": 6.19, + "learning_rate": 2.5559911162787722e-06, + "loss": 0.0016, + "step": 47285 + }, + { + "epoch": 6.19, + "learning_rate": 2.555637011216773e-06, + "loss": 0.0021, + "step": 47286 + }, + { + "epoch": 6.19, + "learning_rate": 2.5552829270916468e-06, + "loss": 0.0019, + "step": 47287 + }, + { + "epoch": 6.19, + "learning_rate": 2.554928863904391e-06, + "loss": 0.0018, + "step": 47288 + }, + { + "epoch": 6.19, + "learning_rate": 2.5545748216559996e-06, + "loss": 0.0014, + "step": 47289 + }, + { + "epoch": 6.19, + "learning_rate": 2.5542208003474687e-06, + "loss": 0.0016, + "step": 47290 + }, + { + "epoch": 6.19, + "learning_rate": 2.5538667999797996e-06, + "loss": 0.002, + "step": 47291 + }, + { + "epoch": 6.19, + "learning_rate": 2.553512820553983e-06, + "loss": 0.0013, + "step": 47292 + }, + { + "epoch": 6.2, + "learning_rate": 2.5531588620710123e-06, + "loss": 0.0027, + "step": 47293 + }, + { + "epoch": 6.2, + "learning_rate": 2.5528049245318876e-06, + "loss": 0.0014, + "step": 47294 + }, + { + "epoch": 6.2, + "learning_rate": 2.552451007937603e-06, + "loss": 0.0028, + "step": 47295 + }, + { + "epoch": 6.2, + "learning_rate": 2.552097112289155e-06, + "loss": 0.0028, + "step": 47296 + }, + { + "epoch": 6.2, + "learning_rate": 2.551743237587532e-06, + "loss": 0.002, + "step": 47297 + }, + { + "epoch": 6.2, + "learning_rate": 2.551389383833738e-06, + "loss": 0.0023, + "step": 47298 + }, + { + "epoch": 6.2, + "learning_rate": 2.5510355510287654e-06, + "loss": 0.0022, + "step": 47299 + }, + { + "epoch": 6.2, + "learning_rate": 2.550681739173605e-06, + "loss": 0.0023, + "step": 47300 + }, + { + "epoch": 6.2, + "learning_rate": 2.5503279482692554e-06, + "loss": 0.0019, + "step": 47301 + }, + { + "epoch": 6.2, + "learning_rate": 2.5499741783167152e-06, + "loss": 0.0033, + "step": 47302 + }, + { + "epoch": 6.2, + "learning_rate": 2.549620429316976e-06, + "loss": 0.0027, + "step": 47303 + }, + { + "epoch": 6.2, + "learning_rate": 2.5492667012710324e-06, + "loss": 0.0021, + "step": 47304 + }, + { + "epoch": 6.2, + "learning_rate": 2.548912994179876e-06, + "loss": 0.0021, + "step": 47305 + }, + { + "epoch": 6.2, + "learning_rate": 2.5485593080445093e-06, + "loss": 0.0012, + "step": 47306 + }, + { + "epoch": 6.2, + "learning_rate": 2.5482056428659217e-06, + "loss": 0.0018, + "step": 47307 + }, + { + "epoch": 6.2, + "learning_rate": 2.5478519986451057e-06, + "loss": 0.0032, + "step": 47308 + }, + { + "epoch": 6.2, + "learning_rate": 2.547498375383063e-06, + "loss": 0.003, + "step": 47309 + }, + { + "epoch": 6.2, + "learning_rate": 2.547144773080783e-06, + "loss": 0.0026, + "step": 47310 + }, + { + "epoch": 6.2, + "learning_rate": 2.5467911917392597e-06, + "loss": 0.0026, + "step": 47311 + }, + { + "epoch": 6.2, + "learning_rate": 2.5464376313594885e-06, + "loss": 0.0021, + "step": 47312 + }, + { + "epoch": 6.2, + "learning_rate": 2.546084091942468e-06, + "loss": 0.0022, + "step": 47313 + }, + { + "epoch": 6.2, + "learning_rate": 2.545730573489188e-06, + "loss": 0.0015, + "step": 47314 + }, + { + "epoch": 6.2, + "learning_rate": 2.5453770760006447e-06, + "loss": 0.0018, + "step": 47315 + }, + { + "epoch": 6.2, + "learning_rate": 2.545023599477828e-06, + "loss": 0.0037, + "step": 47316 + }, + { + "epoch": 6.2, + "learning_rate": 2.5446701439217393e-06, + "loss": 0.0027, + "step": 47317 + }, + { + "epoch": 6.2, + "learning_rate": 2.5443167093333675e-06, + "loss": 0.0015, + "step": 47318 + }, + { + "epoch": 6.2, + "learning_rate": 2.5439632957137096e-06, + "loss": 0.0027, + "step": 47319 + }, + { + "epoch": 6.2, + "learning_rate": 2.543609903063754e-06, + "loss": 0.0021, + "step": 47320 + }, + { + "epoch": 6.2, + "learning_rate": 2.5432565313845003e-06, + "loss": 0.0019, + "step": 47321 + }, + { + "epoch": 6.2, + "learning_rate": 2.542903180676939e-06, + "loss": 0.0022, + "step": 47322 + }, + { + "epoch": 6.2, + "learning_rate": 2.5425498509420675e-06, + "loss": 0.0028, + "step": 47323 + }, + { + "epoch": 6.2, + "learning_rate": 2.5421965421808758e-06, + "loss": 0.0025, + "step": 47324 + }, + { + "epoch": 6.2, + "learning_rate": 2.5418432543943607e-06, + "loss": 0.0016, + "step": 47325 + }, + { + "epoch": 6.2, + "learning_rate": 2.5414899875835154e-06, + "loss": 0.0017, + "step": 47326 + }, + { + "epoch": 6.2, + "learning_rate": 2.5411367417493284e-06, + "loss": 0.0029, + "step": 47327 + }, + { + "epoch": 6.2, + "learning_rate": 2.540783516892801e-06, + "loss": 0.0011, + "step": 47328 + }, + { + "epoch": 6.2, + "learning_rate": 2.540430313014921e-06, + "loss": 0.0024, + "step": 47329 + }, + { + "epoch": 6.2, + "learning_rate": 2.540077130116685e-06, + "loss": 0.0018, + "step": 47330 + }, + { + "epoch": 6.2, + "learning_rate": 2.5397239681990817e-06, + "loss": 0.0023, + "step": 47331 + }, + { + "epoch": 6.2, + "learning_rate": 2.53937082726311e-06, + "loss": 0.0023, + "step": 47332 + }, + { + "epoch": 6.2, + "learning_rate": 2.5390177073097575e-06, + "loss": 0.0017, + "step": 47333 + }, + { + "epoch": 6.2, + "learning_rate": 2.5386646083400247e-06, + "loss": 0.0019, + "step": 47334 + }, + { + "epoch": 6.2, + "learning_rate": 2.5383115303548954e-06, + "loss": 0.0068, + "step": 47335 + }, + { + "epoch": 6.2, + "learning_rate": 2.537958473355372e-06, + "loss": 0.0011, + "step": 47336 + }, + { + "epoch": 6.2, + "learning_rate": 2.537605437342442e-06, + "loss": 0.0026, + "step": 47337 + }, + { + "epoch": 6.2, + "learning_rate": 2.5372524223171e-06, + "loss": 0.0025, + "step": 47338 + }, + { + "epoch": 6.2, + "learning_rate": 2.536899428280335e-06, + "loss": 0.0022, + "step": 47339 + }, + { + "epoch": 6.2, + "learning_rate": 2.5365464552331465e-06, + "loss": 0.0022, + "step": 47340 + }, + { + "epoch": 6.2, + "learning_rate": 2.5361935031765227e-06, + "loss": 0.0026, + "step": 47341 + }, + { + "epoch": 6.2, + "learning_rate": 2.5358405721114544e-06, + "loss": 0.0011, + "step": 47342 + }, + { + "epoch": 6.2, + "learning_rate": 2.5354876620389403e-06, + "loss": 0.0027, + "step": 47343 + }, + { + "epoch": 6.2, + "learning_rate": 2.535134772959966e-06, + "loss": 0.0016, + "step": 47344 + }, + { + "epoch": 6.2, + "learning_rate": 2.5347819048755317e-06, + "loss": 0.003, + "step": 47345 + }, + { + "epoch": 6.2, + "learning_rate": 2.5344290577866248e-06, + "loss": 0.0016, + "step": 47346 + }, + { + "epoch": 6.2, + "learning_rate": 2.5340762316942357e-06, + "loss": 0.0016, + "step": 47347 + }, + { + "epoch": 6.2, + "learning_rate": 2.5337234265993627e-06, + "loss": 0.003, + "step": 47348 + }, + { + "epoch": 6.2, + "learning_rate": 2.533370642502996e-06, + "loss": 0.0013, + "step": 47349 + }, + { + "epoch": 6.2, + "learning_rate": 2.533017879406122e-06, + "loss": 0.0009, + "step": 47350 + }, + { + "epoch": 6.2, + "learning_rate": 2.532665137309742e-06, + "loss": 0.0031, + "step": 47351 + }, + { + "epoch": 6.2, + "learning_rate": 2.5323124162148427e-06, + "loss": 0.003, + "step": 47352 + }, + { + "epoch": 6.2, + "learning_rate": 2.5319597161224175e-06, + "loss": 0.002, + "step": 47353 + }, + { + "epoch": 6.2, + "learning_rate": 2.5316070370334545e-06, + "loss": 0.0022, + "step": 47354 + }, + { + "epoch": 6.2, + "learning_rate": 2.5312543789489498e-06, + "loss": 0.0036, + "step": 47355 + }, + { + "epoch": 6.2, + "learning_rate": 2.5309017418698965e-06, + "loss": 0.0014, + "step": 47356 + }, + { + "epoch": 6.2, + "learning_rate": 2.5305491257972845e-06, + "loss": 0.0024, + "step": 47357 + }, + { + "epoch": 6.2, + "learning_rate": 2.530196530732103e-06, + "loss": 0.0017, + "step": 47358 + }, + { + "epoch": 6.2, + "learning_rate": 2.5298439566753484e-06, + "loss": 0.0033, + "step": 47359 + }, + { + "epoch": 6.2, + "learning_rate": 2.5294914036280105e-06, + "loss": 0.0012, + "step": 47360 + }, + { + "epoch": 6.2, + "learning_rate": 2.5291388715910792e-06, + "loss": 0.002, + "step": 47361 + }, + { + "epoch": 6.2, + "learning_rate": 2.5287863605655447e-06, + "loss": 0.0015, + "step": 47362 + }, + { + "epoch": 6.2, + "learning_rate": 2.5284338705524036e-06, + "loss": 0.001, + "step": 47363 + }, + { + "epoch": 6.2, + "learning_rate": 2.5280814015526443e-06, + "loss": 0.0023, + "step": 47364 + }, + { + "epoch": 6.2, + "learning_rate": 2.5277289535672544e-06, + "loss": 0.0023, + "step": 47365 + }, + { + "epoch": 6.2, + "learning_rate": 2.5273765265972306e-06, + "loss": 0.0021, + "step": 47366 + }, + { + "epoch": 6.2, + "learning_rate": 2.527024120643564e-06, + "loss": 0.0019, + "step": 47367 + }, + { + "epoch": 6.2, + "learning_rate": 2.5266717357072445e-06, + "loss": 0.0018, + "step": 47368 + }, + { + "epoch": 6.21, + "learning_rate": 2.526319371789262e-06, + "loss": 0.0016, + "step": 47369 + }, + { + "epoch": 6.21, + "learning_rate": 2.5259670288906067e-06, + "loss": 0.0021, + "step": 47370 + }, + { + "epoch": 6.21, + "learning_rate": 2.5256147070122728e-06, + "loss": 0.0024, + "step": 47371 + }, + { + "epoch": 6.21, + "learning_rate": 2.52526240615525e-06, + "loss": 0.003, + "step": 47372 + }, + { + "epoch": 6.21, + "learning_rate": 2.524910126320527e-06, + "loss": 0.0024, + "step": 47373 + }, + { + "epoch": 6.21, + "learning_rate": 2.5245578675090974e-06, + "loss": 0.002, + "step": 47374 + }, + { + "epoch": 6.21, + "learning_rate": 2.5242056297219506e-06, + "loss": 0.0018, + "step": 47375 + }, + { + "epoch": 6.21, + "learning_rate": 2.5238534129600755e-06, + "loss": 0.002, + "step": 47376 + }, + { + "epoch": 6.21, + "learning_rate": 2.5235012172244643e-06, + "loss": 0.002, + "step": 47377 + }, + { + "epoch": 6.21, + "learning_rate": 2.5231490425161097e-06, + "loss": 0.0031, + "step": 47378 + }, + { + "epoch": 6.21, + "learning_rate": 2.522796888836001e-06, + "loss": 0.0022, + "step": 47379 + }, + { + "epoch": 6.21, + "learning_rate": 2.5224447561851283e-06, + "loss": 0.0028, + "step": 47380 + }, + { + "epoch": 6.21, + "learning_rate": 2.522092644564478e-06, + "loss": 0.0025, + "step": 47381 + }, + { + "epoch": 6.21, + "learning_rate": 2.521740553975047e-06, + "loss": 0.0017, + "step": 47382 + }, + { + "epoch": 6.21, + "learning_rate": 2.5213884844178226e-06, + "loss": 0.0023, + "step": 47383 + }, + { + "epoch": 6.21, + "learning_rate": 2.5210364358937945e-06, + "loss": 0.0018, + "step": 47384 + }, + { + "epoch": 6.21, + "learning_rate": 2.5206844084039505e-06, + "loss": 0.0035, + "step": 47385 + }, + { + "epoch": 6.21, + "learning_rate": 2.520332401949287e-06, + "loss": 0.002, + "step": 47386 + }, + { + "epoch": 6.21, + "learning_rate": 2.519980416530786e-06, + "loss": 0.0026, + "step": 47387 + }, + { + "epoch": 6.21, + "learning_rate": 2.5196284521494453e-06, + "loss": 0.0014, + "step": 47388 + }, + { + "epoch": 6.21, + "learning_rate": 2.519276508806249e-06, + "loss": 0.0023, + "step": 47389 + }, + { + "epoch": 6.21, + "learning_rate": 2.5189245865021903e-06, + "loss": 0.0019, + "step": 47390 + }, + { + "epoch": 6.21, + "learning_rate": 2.5185726852382588e-06, + "loss": 0.0035, + "step": 47391 + }, + { + "epoch": 6.21, + "learning_rate": 2.51822080501544e-06, + "loss": 0.0015, + "step": 47392 + }, + { + "epoch": 6.21, + "learning_rate": 2.5178689458347306e-06, + "loss": 0.0024, + "step": 47393 + }, + { + "epoch": 6.21, + "learning_rate": 2.517517107697116e-06, + "loss": 0.0019, + "step": 47394 + }, + { + "epoch": 6.21, + "learning_rate": 2.517165290603586e-06, + "loss": 0.0022, + "step": 47395 + }, + { + "epoch": 6.21, + "learning_rate": 2.5168134945551257e-06, + "loss": 0.0024, + "step": 47396 + }, + { + "epoch": 6.21, + "learning_rate": 2.516461719552734e-06, + "loss": 0.0024, + "step": 47397 + }, + { + "epoch": 6.21, + "learning_rate": 2.5161099655973907e-06, + "loss": 0.0026, + "step": 47398 + }, + { + "epoch": 6.21, + "learning_rate": 2.515758232690094e-06, + "loss": 0.0035, + "step": 47399 + }, + { + "epoch": 6.21, + "learning_rate": 2.5154065208318245e-06, + "loss": 0.0027, + "step": 47400 + }, + { + "epoch": 6.21, + "learning_rate": 2.515054830023579e-06, + "loss": 0.0022, + "step": 47401 + }, + { + "epoch": 6.21, + "learning_rate": 2.5147031602663443e-06, + "loss": 0.0018, + "step": 47402 + }, + { + "epoch": 6.21, + "learning_rate": 2.514351511561106e-06, + "loss": 0.0033, + "step": 47403 + }, + { + "epoch": 6.21, + "learning_rate": 2.513999883908854e-06, + "loss": 0.001, + "step": 47404 + }, + { + "epoch": 6.21, + "learning_rate": 2.513648277310582e-06, + "loss": 0.0022, + "step": 47405 + }, + { + "epoch": 6.21, + "learning_rate": 2.513296691767275e-06, + "loss": 0.0022, + "step": 47406 + }, + { + "epoch": 6.21, + "learning_rate": 2.512945127279919e-06, + "loss": 0.002, + "step": 47407 + }, + { + "epoch": 6.21, + "learning_rate": 2.5125935838495097e-06, + "loss": 0.0025, + "step": 47408 + }, + { + "epoch": 6.21, + "learning_rate": 2.5122420614770293e-06, + "loss": 0.0032, + "step": 47409 + }, + { + "epoch": 6.21, + "learning_rate": 2.5118905601634715e-06, + "loss": 0.004, + "step": 47410 + }, + { + "epoch": 6.21, + "learning_rate": 2.5115390799098226e-06, + "loss": 0.0018, + "step": 47411 + }, + { + "epoch": 6.21, + "learning_rate": 2.5111876207170695e-06, + "loss": 0.002, + "step": 47412 + }, + { + "epoch": 6.21, + "learning_rate": 2.5108361825862047e-06, + "loss": 0.0015, + "step": 47413 + }, + { + "epoch": 6.21, + "learning_rate": 2.510484765518214e-06, + "loss": 0.0022, + "step": 47414 + }, + { + "epoch": 6.21, + "learning_rate": 2.510133369514084e-06, + "loss": 0.0025, + "step": 47415 + }, + { + "epoch": 6.21, + "learning_rate": 2.509781994574808e-06, + "loss": 0.0019, + "step": 47416 + }, + { + "epoch": 6.21, + "learning_rate": 2.50943064070137e-06, + "loss": 0.0024, + "step": 47417 + }, + { + "epoch": 6.21, + "learning_rate": 2.5090793078947605e-06, + "loss": 0.0034, + "step": 47418 + }, + { + "epoch": 6.21, + "learning_rate": 2.508727996155963e-06, + "loss": 0.0015, + "step": 47419 + }, + { + "epoch": 6.21, + "learning_rate": 2.50837670548597e-06, + "loss": 0.0017, + "step": 47420 + }, + { + "epoch": 6.21, + "learning_rate": 2.5080254358857703e-06, + "loss": 0.0011, + "step": 47421 + }, + { + "epoch": 6.21, + "learning_rate": 2.5076741873563516e-06, + "loss": 0.0019, + "step": 47422 + }, + { + "epoch": 6.21, + "learning_rate": 2.507322959898696e-06, + "loss": 0.0032, + "step": 47423 + }, + { + "epoch": 6.21, + "learning_rate": 2.506971753513798e-06, + "loss": 0.0017, + "step": 47424 + }, + { + "epoch": 6.21, + "learning_rate": 2.5066205682026445e-06, + "loss": 0.0028, + "step": 47425 + }, + { + "epoch": 6.21, + "learning_rate": 2.5062694039662203e-06, + "loss": 0.0034, + "step": 47426 + }, + { + "epoch": 6.21, + "learning_rate": 2.505918260805512e-06, + "loss": 0.0019, + "step": 47427 + }, + { + "epoch": 6.21, + "learning_rate": 2.505567138721513e-06, + "loss": 0.0018, + "step": 47428 + }, + { + "epoch": 6.21, + "learning_rate": 2.505216037715207e-06, + "loss": 0.0017, + "step": 47429 + }, + { + "epoch": 6.21, + "learning_rate": 2.504864957787578e-06, + "loss": 0.0041, + "step": 47430 + }, + { + "epoch": 6.21, + "learning_rate": 2.504513898939618e-06, + "loss": 0.0035, + "step": 47431 + }, + { + "epoch": 6.21, + "learning_rate": 2.504162861172317e-06, + "loss": 0.003, + "step": 47432 + }, + { + "epoch": 6.21, + "learning_rate": 2.5038118444866575e-06, + "loss": 0.0021, + "step": 47433 + }, + { + "epoch": 6.21, + "learning_rate": 2.5034608488836287e-06, + "loss": 0.0038, + "step": 47434 + }, + { + "epoch": 6.21, + "learning_rate": 2.5031098743642134e-06, + "loss": 0.0022, + "step": 47435 + }, + { + "epoch": 6.21, + "learning_rate": 2.5027589209294057e-06, + "loss": 0.0016, + "step": 47436 + }, + { + "epoch": 6.21, + "learning_rate": 2.5024079885801898e-06, + "loss": 0.0018, + "step": 47437 + }, + { + "epoch": 6.21, + "learning_rate": 2.502057077317549e-06, + "loss": 0.0015, + "step": 47438 + }, + { + "epoch": 6.21, + "learning_rate": 2.5017061871424754e-06, + "loss": 0.0019, + "step": 47439 + }, + { + "epoch": 6.21, + "learning_rate": 2.501355318055955e-06, + "loss": 0.002, + "step": 47440 + }, + { + "epoch": 6.21, + "learning_rate": 2.501004470058973e-06, + "loss": 0.0022, + "step": 47441 + }, + { + "epoch": 6.21, + "learning_rate": 2.500653643152511e-06, + "loss": 0.0019, + "step": 47442 + }, + { + "epoch": 6.21, + "learning_rate": 2.500302837337566e-06, + "loss": 0.0014, + "step": 47443 + }, + { + "epoch": 6.21, + "learning_rate": 2.499952052615121e-06, + "loss": 0.0014, + "step": 47444 + }, + { + "epoch": 6.21, + "learning_rate": 2.499601288986161e-06, + "loss": 0.0051, + "step": 47445 + }, + { + "epoch": 6.22, + "learning_rate": 2.4992505464516705e-06, + "loss": 0.0033, + "step": 47446 + }, + { + "epoch": 6.22, + "learning_rate": 2.4988998250126407e-06, + "loss": 0.0022, + "step": 47447 + }, + { + "epoch": 6.22, + "learning_rate": 2.4985491246700566e-06, + "loss": 0.0024, + "step": 47448 + }, + { + "epoch": 6.22, + "learning_rate": 2.498198445424903e-06, + "loss": 0.0012, + "step": 47449 + }, + { + "epoch": 6.22, + "learning_rate": 2.4978477872781638e-06, + "loss": 0.0013, + "step": 47450 + }, + { + "epoch": 6.22, + "learning_rate": 2.497497150230831e-06, + "loss": 0.0023, + "step": 47451 + }, + { + "epoch": 6.22, + "learning_rate": 2.4971465342838884e-06, + "loss": 0.0012, + "step": 47452 + }, + { + "epoch": 6.22, + "learning_rate": 2.4967959394383177e-06, + "loss": 0.0017, + "step": 47453 + }, + { + "epoch": 6.22, + "learning_rate": 2.4964453656951105e-06, + "loss": 0.0013, + "step": 47454 + }, + { + "epoch": 6.22, + "learning_rate": 2.4960948130552522e-06, + "loss": 0.002, + "step": 47455 + }, + { + "epoch": 6.22, + "learning_rate": 2.4957442815197285e-06, + "loss": 0.0009, + "step": 47456 + }, + { + "epoch": 6.22, + "learning_rate": 2.4953937710895215e-06, + "loss": 0.0037, + "step": 47457 + }, + { + "epoch": 6.22, + "learning_rate": 2.4950432817656234e-06, + "loss": 0.0019, + "step": 47458 + }, + { + "epoch": 6.22, + "learning_rate": 2.4946928135490155e-06, + "loss": 0.0015, + "step": 47459 + }, + { + "epoch": 6.22, + "learning_rate": 2.494342366440685e-06, + "loss": 0.0024, + "step": 47460 + }, + { + "epoch": 6.22, + "learning_rate": 2.4939919404416135e-06, + "loss": 0.0031, + "step": 47461 + }, + { + "epoch": 6.22, + "learning_rate": 2.4936415355527933e-06, + "loss": 0.0029, + "step": 47462 + }, + { + "epoch": 6.22, + "learning_rate": 2.4932911517752055e-06, + "loss": 0.0025, + "step": 47463 + }, + { + "epoch": 6.22, + "learning_rate": 2.4929407891098347e-06, + "loss": 0.0023, + "step": 47464 + }, + { + "epoch": 6.22, + "learning_rate": 2.4925904475576677e-06, + "loss": 0.002, + "step": 47465 + }, + { + "epoch": 6.22, + "learning_rate": 2.492240127119694e-06, + "loss": 0.0028, + "step": 47466 + }, + { + "epoch": 6.22, + "learning_rate": 2.491889827796894e-06, + "loss": 0.002, + "step": 47467 + }, + { + "epoch": 6.22, + "learning_rate": 2.491539549590255e-06, + "loss": 0.0021, + "step": 47468 + }, + { + "epoch": 6.22, + "learning_rate": 2.4911892925007575e-06, + "loss": 0.0022, + "step": 47469 + }, + { + "epoch": 6.22, + "learning_rate": 2.4908390565293936e-06, + "loss": 0.0015, + "step": 47470 + }, + { + "epoch": 6.22, + "learning_rate": 2.4904888416771446e-06, + "loss": 0.0015, + "step": 47471 + }, + { + "epoch": 6.22, + "learning_rate": 2.490138647944993e-06, + "loss": 0.0017, + "step": 47472 + }, + { + "epoch": 6.22, + "learning_rate": 2.4897884753339297e-06, + "loss": 0.0022, + "step": 47473 + }, + { + "epoch": 6.22, + "learning_rate": 2.489438323844936e-06, + "loss": 0.0011, + "step": 47474 + }, + { + "epoch": 6.22, + "learning_rate": 2.4890881934789945e-06, + "loss": 0.0029, + "step": 47475 + }, + { + "epoch": 6.22, + "learning_rate": 2.4887380842370956e-06, + "loss": 0.0022, + "step": 47476 + }, + { + "epoch": 6.22, + "learning_rate": 2.488387996120217e-06, + "loss": 0.0029, + "step": 47477 + }, + { + "epoch": 6.22, + "learning_rate": 2.4880379291293512e-06, + "loss": 0.0023, + "step": 47478 + }, + { + "epoch": 6.22, + "learning_rate": 2.4876878832654784e-06, + "loss": 0.0013, + "step": 47479 + }, + { + "epoch": 6.22, + "learning_rate": 2.4873378585295805e-06, + "loss": 0.0031, + "step": 47480 + }, + { + "epoch": 6.22, + "learning_rate": 2.4869878549226466e-06, + "loss": 0.002, + "step": 47481 + }, + { + "epoch": 6.22, + "learning_rate": 2.4866378724456607e-06, + "loss": 0.0028, + "step": 47482 + }, + { + "epoch": 6.22, + "learning_rate": 2.4862879110996053e-06, + "loss": 0.0013, + "step": 47483 + }, + { + "epoch": 6.22, + "learning_rate": 2.4859379708854612e-06, + "loss": 0.0014, + "step": 47484 + }, + { + "epoch": 6.22, + "learning_rate": 2.485588051804221e-06, + "loss": 0.0024, + "step": 47485 + }, + { + "epoch": 6.22, + "learning_rate": 2.485238153856859e-06, + "loss": 0.0032, + "step": 47486 + }, + { + "epoch": 6.22, + "learning_rate": 2.4848882770443696e-06, + "loss": 0.0021, + "step": 47487 + }, + { + "epoch": 6.22, + "learning_rate": 2.4845384213677284e-06, + "loss": 0.0018, + "step": 47488 + }, + { + "epoch": 6.22, + "learning_rate": 2.484188586827925e-06, + "loss": 0.0022, + "step": 47489 + }, + { + "epoch": 6.22, + "learning_rate": 2.4838387734259405e-06, + "loss": 0.0012, + "step": 47490 + }, + { + "epoch": 6.22, + "learning_rate": 2.4834889811627593e-06, + "loss": 0.001, + "step": 47491 + }, + { + "epoch": 6.22, + "learning_rate": 2.4831392100393635e-06, + "loss": 0.0014, + "step": 47492 + }, + { + "epoch": 6.22, + "learning_rate": 2.4827894600567403e-06, + "loss": 0.0017, + "step": 47493 + }, + { + "epoch": 6.22, + "learning_rate": 2.482439731215871e-06, + "loss": 0.0012, + "step": 47494 + }, + { + "epoch": 6.22, + "learning_rate": 2.482090023517737e-06, + "loss": 0.0027, + "step": 47495 + }, + { + "epoch": 6.22, + "learning_rate": 2.481740336963325e-06, + "loss": 0.0028, + "step": 47496 + }, + { + "epoch": 6.22, + "learning_rate": 2.4813906715536205e-06, + "loss": 0.0022, + "step": 47497 + }, + { + "epoch": 6.22, + "learning_rate": 2.4810410272896047e-06, + "loss": 0.002, + "step": 47498 + }, + { + "epoch": 6.22, + "learning_rate": 2.4806914041722606e-06, + "loss": 0.0016, + "step": 47499 + }, + { + "epoch": 6.22, + "learning_rate": 2.4803418022025684e-06, + "loss": 0.0009, + "step": 47500 + }, + { + "epoch": 6.22, + "learning_rate": 2.4799922213815164e-06, + "loss": 0.0012, + "step": 47501 + }, + { + "epoch": 6.22, + "learning_rate": 2.4796426617100868e-06, + "loss": 0.0018, + "step": 47502 + }, + { + "epoch": 6.22, + "learning_rate": 2.479293123189259e-06, + "loss": 0.0017, + "step": 47503 + }, + { + "epoch": 6.22, + "learning_rate": 2.4789436058200213e-06, + "loss": 0.0018, + "step": 47504 + }, + { + "epoch": 6.22, + "learning_rate": 2.478594109603354e-06, + "loss": 0.0016, + "step": 47505 + }, + { + "epoch": 6.22, + "learning_rate": 2.478244634540241e-06, + "loss": 0.0017, + "step": 47506 + }, + { + "epoch": 6.22, + "learning_rate": 2.4778951806316574e-06, + "loss": 0.0021, + "step": 47507 + }, + { + "epoch": 6.22, + "learning_rate": 2.4775457478786e-06, + "loss": 0.0032, + "step": 47508 + }, + { + "epoch": 6.22, + "learning_rate": 2.4771963362820438e-06, + "loss": 0.0023, + "step": 47509 + }, + { + "epoch": 6.22, + "learning_rate": 2.476846945842973e-06, + "loss": 0.0028, + "step": 47510 + }, + { + "epoch": 6.22, + "learning_rate": 2.476497576562367e-06, + "loss": 0.0015, + "step": 47511 + }, + { + "epoch": 6.22, + "learning_rate": 2.4761482284412118e-06, + "loss": 0.0011, + "step": 47512 + }, + { + "epoch": 6.22, + "learning_rate": 2.4757989014804907e-06, + "loss": 0.0029, + "step": 47513 + }, + { + "epoch": 6.22, + "learning_rate": 2.475449595681184e-06, + "loss": 0.0017, + "step": 47514 + }, + { + "epoch": 6.22, + "learning_rate": 2.4751003110442717e-06, + "loss": 0.0026, + "step": 47515 + }, + { + "epoch": 6.22, + "learning_rate": 2.4747510475707426e-06, + "loss": 0.0012, + "step": 47516 + }, + { + "epoch": 6.22, + "learning_rate": 2.474401805261575e-06, + "loss": 0.0019, + "step": 47517 + }, + { + "epoch": 6.22, + "learning_rate": 2.4740525841177474e-06, + "loss": 0.0026, + "step": 47518 + }, + { + "epoch": 6.22, + "learning_rate": 2.473703384140248e-06, + "loss": 0.0014, + "step": 47519 + }, + { + "epoch": 6.22, + "learning_rate": 2.4733542053300596e-06, + "loss": 0.0015, + "step": 47520 + }, + { + "epoch": 6.22, + "learning_rate": 2.4730050476881606e-06, + "loss": 0.0018, + "step": 47521 + }, + { + "epoch": 6.23, + "learning_rate": 2.4726559112155325e-06, + "loss": 0.0031, + "step": 47522 + }, + { + "epoch": 6.23, + "learning_rate": 2.4723067959131606e-06, + "loss": 0.0031, + "step": 47523 + }, + { + "epoch": 6.23, + "learning_rate": 2.4719577017820253e-06, + "loss": 0.0021, + "step": 47524 + }, + { + "epoch": 6.23, + "learning_rate": 2.471608628823109e-06, + "loss": 0.0026, + "step": 47525 + }, + { + "epoch": 6.23, + "learning_rate": 2.471259577037388e-06, + "loss": 0.0026, + "step": 47526 + }, + { + "epoch": 6.23, + "learning_rate": 2.470910546425852e-06, + "loss": 0.0014, + "step": 47527 + }, + { + "epoch": 6.23, + "learning_rate": 2.4705615369894797e-06, + "loss": 0.0023, + "step": 47528 + }, + { + "epoch": 6.23, + "learning_rate": 2.4702125487292495e-06, + "loss": 0.0031, + "step": 47529 + }, + { + "epoch": 6.23, + "learning_rate": 2.469863581646145e-06, + "loss": 0.0024, + "step": 47530 + }, + { + "epoch": 6.23, + "learning_rate": 2.4695146357411505e-06, + "loss": 0.0039, + "step": 47531 + }, + { + "epoch": 6.23, + "learning_rate": 2.4691657110152456e-06, + "loss": 0.0019, + "step": 47532 + }, + { + "epoch": 6.23, + "learning_rate": 2.468816807469411e-06, + "loss": 0.0025, + "step": 47533 + }, + { + "epoch": 6.23, + "learning_rate": 2.4684679251046253e-06, + "loss": 0.0025, + "step": 47534 + }, + { + "epoch": 6.23, + "learning_rate": 2.4681190639218754e-06, + "loss": 0.0018, + "step": 47535 + }, + { + "epoch": 6.23, + "learning_rate": 2.467770223922139e-06, + "loss": 0.0016, + "step": 47536 + }, + { + "epoch": 6.23, + "learning_rate": 2.467421405106395e-06, + "loss": 0.0023, + "step": 47537 + }, + { + "epoch": 6.23, + "learning_rate": 2.46707260747563e-06, + "loss": 0.0014, + "step": 47538 + }, + { + "epoch": 6.23, + "learning_rate": 2.466723831030823e-06, + "loss": 0.0023, + "step": 47539 + }, + { + "epoch": 6.23, + "learning_rate": 2.466375075772951e-06, + "loss": 0.0028, + "step": 47540 + }, + { + "epoch": 6.23, + "learning_rate": 2.4660263417030004e-06, + "loss": 0.0018, + "step": 47541 + }, + { + "epoch": 6.23, + "learning_rate": 2.4656776288219465e-06, + "loss": 0.0015, + "step": 47542 + }, + { + "epoch": 6.23, + "learning_rate": 2.4653289371307753e-06, + "loss": 0.002, + "step": 47543 + }, + { + "epoch": 6.23, + "learning_rate": 2.464980266630467e-06, + "loss": 0.0015, + "step": 47544 + }, + { + "epoch": 6.23, + "learning_rate": 2.4646316173219965e-06, + "loss": 0.0023, + "step": 47545 + }, + { + "epoch": 6.23, + "learning_rate": 2.4642829892063502e-06, + "loss": 0.0033, + "step": 47546 + }, + { + "epoch": 6.23, + "learning_rate": 2.4639343822845084e-06, + "loss": 0.002, + "step": 47547 + }, + { + "epoch": 6.23, + "learning_rate": 2.4635857965574495e-06, + "loss": 0.0031, + "step": 47548 + }, + { + "epoch": 6.23, + "learning_rate": 2.46323723202615e-06, + "loss": 0.0019, + "step": 47549 + }, + { + "epoch": 6.23, + "learning_rate": 2.4628886886915992e-06, + "loss": 0.0017, + "step": 47550 + }, + { + "epoch": 6.23, + "learning_rate": 2.462540166554769e-06, + "loss": 0.0019, + "step": 47551 + }, + { + "epoch": 6.23, + "learning_rate": 2.4621916656166455e-06, + "loss": 0.0017, + "step": 47552 + }, + { + "epoch": 6.23, + "learning_rate": 2.461843185878203e-06, + "loss": 0.0012, + "step": 47553 + }, + { + "epoch": 6.23, + "learning_rate": 2.46149472734043e-06, + "loss": 0.0023, + "step": 47554 + }, + { + "epoch": 6.23, + "learning_rate": 2.4611462900043004e-06, + "loss": 0.0023, + "step": 47555 + }, + { + "epoch": 6.23, + "learning_rate": 2.460797873870795e-06, + "loss": 0.0021, + "step": 47556 + }, + { + "epoch": 6.23, + "learning_rate": 2.460449478940893e-06, + "loss": 0.004, + "step": 47557 + }, + { + "epoch": 6.23, + "learning_rate": 2.4601011052155767e-06, + "loss": 0.0022, + "step": 47558 + }, + { + "epoch": 6.23, + "learning_rate": 2.4597527526958243e-06, + "loss": 0.0012, + "step": 47559 + }, + { + "epoch": 6.23, + "learning_rate": 2.459404421382614e-06, + "loss": 0.0028, + "step": 47560 + }, + { + "epoch": 6.23, + "learning_rate": 2.4590561112769294e-06, + "loss": 0.0013, + "step": 47561 + }, + { + "epoch": 6.23, + "learning_rate": 2.4587078223797444e-06, + "loss": 0.0028, + "step": 47562 + }, + { + "epoch": 6.23, + "learning_rate": 2.458359554692046e-06, + "loss": 0.0027, + "step": 47563 + }, + { + "epoch": 6.23, + "learning_rate": 2.458011308214809e-06, + "loss": 0.0017, + "step": 47564 + }, + { + "epoch": 6.23, + "learning_rate": 2.457663082949011e-06, + "loss": 0.0015, + "step": 47565 + }, + { + "epoch": 6.23, + "learning_rate": 2.457314878895637e-06, + "loss": 0.0021, + "step": 47566 + }, + { + "epoch": 6.23, + "learning_rate": 2.456966696055664e-06, + "loss": 0.0013, + "step": 47567 + }, + { + "epoch": 6.23, + "learning_rate": 2.456618534430066e-06, + "loss": 0.0037, + "step": 47568 + }, + { + "epoch": 6.23, + "learning_rate": 2.456270394019831e-06, + "loss": 0.0011, + "step": 47569 + }, + { + "epoch": 6.23, + "learning_rate": 2.4559222748259327e-06, + "loss": 0.0026, + "step": 47570 + }, + { + "epoch": 6.23, + "learning_rate": 2.4555741768493515e-06, + "loss": 0.0029, + "step": 47571 + }, + { + "epoch": 6.23, + "learning_rate": 2.4552261000910637e-06, + "loss": 0.0013, + "step": 47572 + }, + { + "epoch": 6.23, + "learning_rate": 2.4548780445520503e-06, + "loss": 0.0014, + "step": 47573 + }, + { + "epoch": 6.23, + "learning_rate": 2.454530010233294e-06, + "loss": 0.0015, + "step": 47574 + }, + { + "epoch": 6.23, + "learning_rate": 2.4541819971357696e-06, + "loss": 0.0015, + "step": 47575 + }, + { + "epoch": 6.23, + "learning_rate": 2.453834005260454e-06, + "loss": 0.0026, + "step": 47576 + }, + { + "epoch": 6.23, + "learning_rate": 2.45348603460833e-06, + "loss": 0.0017, + "step": 47577 + }, + { + "epoch": 6.23, + "learning_rate": 2.4531380851803764e-06, + "loss": 0.0024, + "step": 47578 + }, + { + "epoch": 6.23, + "learning_rate": 2.4527901569775657e-06, + "loss": 0.0026, + "step": 47579 + }, + { + "epoch": 6.23, + "learning_rate": 2.452442250000885e-06, + "loss": 0.0014, + "step": 47580 + }, + { + "epoch": 6.23, + "learning_rate": 2.4520943642513084e-06, + "loss": 0.0027, + "step": 47581 + }, + { + "epoch": 6.23, + "learning_rate": 2.4517464997298133e-06, + "loss": 0.0023, + "step": 47582 + }, + { + "epoch": 6.23, + "learning_rate": 2.451398656437377e-06, + "loss": 0.0029, + "step": 47583 + }, + { + "epoch": 6.23, + "learning_rate": 2.4510508343749805e-06, + "loss": 0.0018, + "step": 47584 + }, + { + "epoch": 6.23, + "learning_rate": 2.4507030335436046e-06, + "loss": 0.0027, + "step": 47585 + }, + { + "epoch": 6.23, + "learning_rate": 2.4503552539442233e-06, + "loss": 0.0018, + "step": 47586 + }, + { + "epoch": 6.23, + "learning_rate": 2.450007495577813e-06, + "loss": 0.0013, + "step": 47587 + }, + { + "epoch": 6.23, + "learning_rate": 2.4496597584453584e-06, + "loss": 0.0018, + "step": 47588 + }, + { + "epoch": 6.23, + "learning_rate": 2.4493120425478332e-06, + "loss": 0.0014, + "step": 47589 + }, + { + "epoch": 6.23, + "learning_rate": 2.4489643478862156e-06, + "loss": 0.0017, + "step": 47590 + }, + { + "epoch": 6.23, + "learning_rate": 2.4486166744614804e-06, + "loss": 0.0021, + "step": 47591 + }, + { + "epoch": 6.23, + "learning_rate": 2.448269022274612e-06, + "loss": 0.0013, + "step": 47592 + }, + { + "epoch": 6.23, + "learning_rate": 2.4479213913265855e-06, + "loss": 0.0037, + "step": 47593 + }, + { + "epoch": 6.23, + "learning_rate": 2.447573781618374e-06, + "loss": 0.0019, + "step": 47594 + }, + { + "epoch": 6.23, + "learning_rate": 2.4472261931509587e-06, + "loss": 0.0018, + "step": 47595 + }, + { + "epoch": 6.23, + "learning_rate": 2.4468786259253218e-06, + "loss": 0.0019, + "step": 47596 + }, + { + "epoch": 6.23, + "learning_rate": 2.446531079942436e-06, + "loss": 0.0035, + "step": 47597 + }, + { + "epoch": 6.24, + "learning_rate": 2.4461835552032787e-06, + "loss": 0.002, + "step": 47598 + }, + { + "epoch": 6.24, + "learning_rate": 2.4458360517088255e-06, + "loss": 0.0029, + "step": 47599 + }, + { + "epoch": 6.24, + "learning_rate": 2.4454885694600592e-06, + "loss": 0.0029, + "step": 47600 + }, + { + "epoch": 6.24, + "learning_rate": 2.4451411084579536e-06, + "loss": 0.002, + "step": 47601 + }, + { + "epoch": 6.24, + "learning_rate": 2.4447936687034835e-06, + "loss": 0.0015, + "step": 47602 + }, + { + "epoch": 6.24, + "learning_rate": 2.444446250197632e-06, + "loss": 0.0017, + "step": 47603 + }, + { + "epoch": 6.24, + "learning_rate": 2.444098852941372e-06, + "loss": 0.0029, + "step": 47604 + }, + { + "epoch": 6.24, + "learning_rate": 2.443751476935683e-06, + "loss": 0.0022, + "step": 47605 + }, + { + "epoch": 6.24, + "learning_rate": 2.4434041221815375e-06, + "loss": 0.0021, + "step": 47606 + }, + { + "epoch": 6.24, + "learning_rate": 2.443056788679915e-06, + "loss": 0.0016, + "step": 47607 + }, + { + "epoch": 6.24, + "learning_rate": 2.4427094764317962e-06, + "loss": 0.0027, + "step": 47608 + }, + { + "epoch": 6.24, + "learning_rate": 2.442362185438154e-06, + "loss": 0.0023, + "step": 47609 + }, + { + "epoch": 6.24, + "learning_rate": 2.4420149156999627e-06, + "loss": 0.0024, + "step": 47610 + }, + { + "epoch": 6.24, + "learning_rate": 2.441667667218205e-06, + "loss": 0.0025, + "step": 47611 + }, + { + "epoch": 6.24, + "learning_rate": 2.441320439993855e-06, + "loss": 0.0035, + "step": 47612 + }, + { + "epoch": 6.24, + "learning_rate": 2.440973234027888e-06, + "loss": 0.0046, + "step": 47613 + }, + { + "epoch": 6.24, + "learning_rate": 2.4406260493212795e-06, + "loss": 0.0017, + "step": 47614 + }, + { + "epoch": 6.24, + "learning_rate": 2.4402788858750093e-06, + "loss": 0.002, + "step": 47615 + }, + { + "epoch": 6.24, + "learning_rate": 2.4399317436900525e-06, + "loss": 0.0017, + "step": 47616 + }, + { + "epoch": 6.24, + "learning_rate": 2.4395846227673824e-06, + "loss": 0.0014, + "step": 47617 + }, + { + "epoch": 6.24, + "learning_rate": 2.4392375231079778e-06, + "loss": 0.0027, + "step": 47618 + }, + { + "epoch": 6.24, + "learning_rate": 2.438890444712818e-06, + "loss": 0.0026, + "step": 47619 + }, + { + "epoch": 6.24, + "learning_rate": 2.438543387582877e-06, + "loss": 0.0013, + "step": 47620 + }, + { + "epoch": 6.24, + "learning_rate": 2.4381963517191286e-06, + "loss": 0.0031, + "step": 47621 + }, + { + "epoch": 6.24, + "learning_rate": 2.4378493371225486e-06, + "loss": 0.0027, + "step": 47622 + }, + { + "epoch": 6.24, + "learning_rate": 2.437502343794117e-06, + "loss": 0.0035, + "step": 47623 + }, + { + "epoch": 6.24, + "learning_rate": 2.437155371734807e-06, + "loss": 0.0016, + "step": 47624 + }, + { + "epoch": 6.24, + "learning_rate": 2.436808420945592e-06, + "loss": 0.0018, + "step": 47625 + }, + { + "epoch": 6.24, + "learning_rate": 2.4364614914274544e-06, + "loss": 0.0024, + "step": 47626 + }, + { + "epoch": 6.24, + "learning_rate": 2.4361145831813615e-06, + "loss": 0.0025, + "step": 47627 + }, + { + "epoch": 6.24, + "learning_rate": 2.435767696208298e-06, + "loss": 0.0023, + "step": 47628 + }, + { + "epoch": 6.24, + "learning_rate": 2.4354208305092307e-06, + "loss": 0.0011, + "step": 47629 + }, + { + "epoch": 6.24, + "learning_rate": 2.4350739860851434e-06, + "loss": 0.0017, + "step": 47630 + }, + { + "epoch": 6.24, + "learning_rate": 2.434727162937007e-06, + "loss": 0.0024, + "step": 47631 + }, + { + "epoch": 6.24, + "learning_rate": 2.434380361065799e-06, + "loss": 0.0021, + "step": 47632 + }, + { + "epoch": 6.24, + "learning_rate": 2.434033580472489e-06, + "loss": 0.0014, + "step": 47633 + }, + { + "epoch": 6.24, + "learning_rate": 2.43368682115806e-06, + "loss": 0.0029, + "step": 47634 + }, + { + "epoch": 6.24, + "learning_rate": 2.4333400831234833e-06, + "loss": 0.0021, + "step": 47635 + }, + { + "epoch": 6.24, + "learning_rate": 2.432993366369736e-06, + "loss": 0.002, + "step": 47636 + }, + { + "epoch": 6.24, + "learning_rate": 2.432646670897788e-06, + "loss": 0.0052, + "step": 47637 + }, + { + "epoch": 6.24, + "learning_rate": 2.432299996708618e-06, + "loss": 0.0018, + "step": 47638 + }, + { + "epoch": 6.24, + "learning_rate": 2.431953343803205e-06, + "loss": 0.0021, + "step": 47639 + }, + { + "epoch": 6.24, + "learning_rate": 2.4316067121825194e-06, + "loss": 0.0019, + "step": 47640 + }, + { + "epoch": 6.24, + "learning_rate": 2.4312601018475347e-06, + "loss": 0.0017, + "step": 47641 + }, + { + "epoch": 6.24, + "learning_rate": 2.43091351279923e-06, + "loss": 0.0025, + "step": 47642 + }, + { + "epoch": 6.24, + "learning_rate": 2.4305669450385793e-06, + "loss": 0.0028, + "step": 47643 + }, + { + "epoch": 6.24, + "learning_rate": 2.4302203985665517e-06, + "loss": 0.0027, + "step": 47644 + }, + { + "epoch": 6.24, + "learning_rate": 2.4298738733841297e-06, + "loss": 0.0028, + "step": 47645 + }, + { + "epoch": 6.24, + "learning_rate": 2.429527369492284e-06, + "loss": 0.0024, + "step": 47646 + }, + { + "epoch": 6.24, + "learning_rate": 2.4291808868919895e-06, + "loss": 0.0012, + "step": 47647 + }, + { + "epoch": 6.24, + "learning_rate": 2.428834425584218e-06, + "loss": 0.0014, + "step": 47648 + }, + { + "epoch": 6.24, + "learning_rate": 2.428487985569946e-06, + "loss": 0.0021, + "step": 47649 + }, + { + "epoch": 6.24, + "learning_rate": 2.4281415668501516e-06, + "loss": 0.0022, + "step": 47650 + }, + { + "epoch": 6.24, + "learning_rate": 2.427795169425806e-06, + "loss": 0.0025, + "step": 47651 + }, + { + "epoch": 6.24, + "learning_rate": 2.4274487932978796e-06, + "loss": 0.0014, + "step": 47652 + }, + { + "epoch": 6.24, + "learning_rate": 2.4271024384673536e-06, + "loss": 0.0032, + "step": 47653 + }, + { + "epoch": 6.24, + "learning_rate": 2.4267561049351983e-06, + "loss": 0.0024, + "step": 47654 + }, + { + "epoch": 6.24, + "learning_rate": 2.4264097927023886e-06, + "loss": 0.0021, + "step": 47655 + }, + { + "epoch": 6.24, + "learning_rate": 2.4260635017698954e-06, + "loss": 0.002, + "step": 47656 + }, + { + "epoch": 6.24, + "learning_rate": 2.425717232138698e-06, + "loss": 0.0016, + "step": 47657 + }, + { + "epoch": 6.24, + "learning_rate": 2.4253709838097673e-06, + "loss": 0.0033, + "step": 47658 + }, + { + "epoch": 6.24, + "learning_rate": 2.4250247567840748e-06, + "loss": 0.0025, + "step": 47659 + }, + { + "epoch": 6.24, + "learning_rate": 2.424678551062596e-06, + "loss": 0.0014, + "step": 47660 + }, + { + "epoch": 6.24, + "learning_rate": 2.424332366646309e-06, + "loss": 0.0021, + "step": 47661 + }, + { + "epoch": 6.24, + "learning_rate": 2.423986203536183e-06, + "loss": 0.0022, + "step": 47662 + }, + { + "epoch": 6.24, + "learning_rate": 2.4236400617331933e-06, + "loss": 0.0028, + "step": 47663 + }, + { + "epoch": 6.24, + "learning_rate": 2.4232939412383094e-06, + "loss": 0.0022, + "step": 47664 + }, + { + "epoch": 6.24, + "learning_rate": 2.42294784205251e-06, + "loss": 0.0011, + "step": 47665 + }, + { + "epoch": 6.24, + "learning_rate": 2.4226017641767664e-06, + "loss": 0.0014, + "step": 47666 + }, + { + "epoch": 6.24, + "learning_rate": 2.42225570761205e-06, + "loss": 0.0031, + "step": 47667 + }, + { + "epoch": 6.24, + "learning_rate": 2.421909672359337e-06, + "loss": 0.0018, + "step": 47668 + }, + { + "epoch": 6.24, + "learning_rate": 2.421563658419601e-06, + "loss": 0.0018, + "step": 47669 + }, + { + "epoch": 6.24, + "learning_rate": 2.4212176657938137e-06, + "loss": 0.0015, + "step": 47670 + }, + { + "epoch": 6.24, + "learning_rate": 2.4208716944829447e-06, + "loss": 0.0015, + "step": 47671 + }, + { + "epoch": 6.24, + "learning_rate": 2.4205257444879703e-06, + "loss": 0.0014, + "step": 47672 + }, + { + "epoch": 6.24, + "learning_rate": 2.4201798158098667e-06, + "loss": 0.0015, + "step": 47673 + }, + { + "epoch": 6.24, + "learning_rate": 2.4198339084496037e-06, + "loss": 0.0021, + "step": 47674 + }, + { + "epoch": 6.25, + "learning_rate": 2.4194880224081507e-06, + "loss": 0.0018, + "step": 47675 + }, + { + "epoch": 6.25, + "learning_rate": 2.419142157686487e-06, + "loss": 0.002, + "step": 47676 + }, + { + "epoch": 6.25, + "learning_rate": 2.418796314285583e-06, + "loss": 0.0013, + "step": 47677 + }, + { + "epoch": 6.25, + "learning_rate": 2.41845049220641e-06, + "loss": 0.0033, + "step": 47678 + }, + { + "epoch": 6.25, + "learning_rate": 2.4181046914499396e-06, + "loss": 0.0014, + "step": 47679 + }, + { + "epoch": 6.25, + "learning_rate": 2.4177589120171473e-06, + "loss": 0.0014, + "step": 47680 + }, + { + "epoch": 6.25, + "learning_rate": 2.417413153909005e-06, + "loss": 0.0018, + "step": 47681 + }, + { + "epoch": 6.25, + "learning_rate": 2.417067417126482e-06, + "loss": 0.0015, + "step": 47682 + }, + { + "epoch": 6.25, + "learning_rate": 2.4167217016705523e-06, + "loss": 0.0023, + "step": 47683 + }, + { + "epoch": 6.25, + "learning_rate": 2.4163760075421915e-06, + "loss": 0.0067, + "step": 47684 + }, + { + "epoch": 6.25, + "learning_rate": 2.4160303347423697e-06, + "loss": 0.0023, + "step": 47685 + }, + { + "epoch": 6.25, + "learning_rate": 2.4156846832720594e-06, + "loss": 0.0023, + "step": 47686 + }, + { + "epoch": 6.25, + "learning_rate": 2.4153390531322287e-06, + "loss": 0.0021, + "step": 47687 + }, + { + "epoch": 6.25, + "learning_rate": 2.4149934443238555e-06, + "loss": 0.0011, + "step": 47688 + }, + { + "epoch": 6.25, + "learning_rate": 2.4146478568479106e-06, + "loss": 0.0024, + "step": 47689 + }, + { + "epoch": 6.25, + "learning_rate": 2.4143022907053603e-06, + "loss": 0.0012, + "step": 47690 + }, + { + "epoch": 6.25, + "learning_rate": 2.4139567458971847e-06, + "loss": 0.0022, + "step": 47691 + }, + { + "epoch": 6.25, + "learning_rate": 2.413611222424351e-06, + "loss": 0.0051, + "step": 47692 + }, + { + "epoch": 6.25, + "learning_rate": 2.4132657202878294e-06, + "loss": 0.0016, + "step": 47693 + }, + { + "epoch": 6.25, + "learning_rate": 2.4129202394885932e-06, + "loss": 0.0033, + "step": 47694 + }, + { + "epoch": 6.25, + "learning_rate": 2.4125747800276188e-06, + "loss": 0.0013, + "step": 47695 + }, + { + "epoch": 6.25, + "learning_rate": 2.412229341905873e-06, + "loss": 0.004, + "step": 47696 + }, + { + "epoch": 6.25, + "learning_rate": 2.4118839251243285e-06, + "loss": 0.0063, + "step": 47697 + }, + { + "epoch": 6.25, + "learning_rate": 2.4115385296839533e-06, + "loss": 0.0042, + "step": 47698 + }, + { + "epoch": 6.25, + "learning_rate": 2.4111931555857247e-06, + "loss": 0.0032, + "step": 47699 + }, + { + "epoch": 6.25, + "learning_rate": 2.4108478028306113e-06, + "loss": 0.0027, + "step": 47700 + }, + { + "epoch": 6.25, + "learning_rate": 2.4105024714195835e-06, + "loss": 0.0035, + "step": 47701 + }, + { + "epoch": 6.25, + "learning_rate": 2.4101571613536114e-06, + "loss": 0.0038, + "step": 47702 + }, + { + "epoch": 6.25, + "learning_rate": 2.4098118726336718e-06, + "loss": 0.0031, + "step": 47703 + }, + { + "epoch": 6.25, + "learning_rate": 2.4094666052607276e-06, + "loss": 0.003, + "step": 47704 + }, + { + "epoch": 6.25, + "learning_rate": 2.4091213592357577e-06, + "loss": 0.0018, + "step": 47705 + }, + { + "epoch": 6.25, + "learning_rate": 2.408776134559727e-06, + "loss": 0.0017, + "step": 47706 + }, + { + "epoch": 6.25, + "learning_rate": 2.408430931233612e-06, + "loss": 0.0017, + "step": 47707 + }, + { + "epoch": 6.25, + "learning_rate": 2.4080857492583808e-06, + "loss": 0.0016, + "step": 47708 + }, + { + "epoch": 6.25, + "learning_rate": 2.4077405886350012e-06, + "loss": 0.0031, + "step": 47709 + }, + { + "epoch": 6.25, + "learning_rate": 2.4073954493644503e-06, + "loss": 0.0016, + "step": 47710 + }, + { + "epoch": 6.25, + "learning_rate": 2.4070503314476943e-06, + "loss": 0.0019, + "step": 47711 + }, + { + "epoch": 6.25, + "learning_rate": 2.406705234885706e-06, + "loss": 0.0044, + "step": 47712 + }, + { + "epoch": 6.25, + "learning_rate": 2.4063601596794517e-06, + "loss": 0.002, + "step": 47713 + }, + { + "epoch": 6.25, + "learning_rate": 2.4060151058299076e-06, + "loss": 0.002, + "step": 47714 + }, + { + "epoch": 6.25, + "learning_rate": 2.405670073338039e-06, + "loss": 0.0016, + "step": 47715 + }, + { + "epoch": 6.25, + "learning_rate": 2.405325062204822e-06, + "loss": 0.0028, + "step": 47716 + }, + { + "epoch": 6.25, + "learning_rate": 2.4049800724312202e-06, + "loss": 0.0025, + "step": 47717 + }, + { + "epoch": 6.25, + "learning_rate": 2.4046351040182105e-06, + "loss": 0.0018, + "step": 47718 + }, + { + "epoch": 6.25, + "learning_rate": 2.4042901569667597e-06, + "loss": 0.0013, + "step": 47719 + }, + { + "epoch": 6.25, + "learning_rate": 2.4039452312778398e-06, + "loss": 0.0031, + "step": 47720 + }, + { + "epoch": 6.25, + "learning_rate": 2.4036003269524156e-06, + "loss": 0.0014, + "step": 47721 + }, + { + "epoch": 6.25, + "learning_rate": 2.4032554439914637e-06, + "loss": 0.0027, + "step": 47722 + }, + { + "epoch": 6.25, + "learning_rate": 2.402910582395952e-06, + "loss": 0.0013, + "step": 47723 + }, + { + "epoch": 6.25, + "learning_rate": 2.4025657421668457e-06, + "loss": 0.0022, + "step": 47724 + }, + { + "epoch": 6.25, + "learning_rate": 2.402220923305123e-06, + "loss": 0.0012, + "step": 47725 + }, + { + "epoch": 6.25, + "learning_rate": 2.401876125811745e-06, + "loss": 0.0023, + "step": 47726 + }, + { + "epoch": 6.25, + "learning_rate": 2.401531349687689e-06, + "loss": 0.0016, + "step": 47727 + }, + { + "epoch": 6.25, + "learning_rate": 2.401186594933922e-06, + "loss": 0.0017, + "step": 47728 + }, + { + "epoch": 6.25, + "learning_rate": 2.4008418615514105e-06, + "loss": 0.0021, + "step": 47729 + }, + { + "epoch": 6.25, + "learning_rate": 2.400497149541129e-06, + "loss": 0.0025, + "step": 47730 + }, + { + "epoch": 6.25, + "learning_rate": 2.4001524589040446e-06, + "loss": 0.0034, + "step": 47731 + }, + { + "epoch": 6.25, + "learning_rate": 2.399807789641123e-06, + "loss": 0.0019, + "step": 47732 + }, + { + "epoch": 6.25, + "learning_rate": 2.399463141753341e-06, + "loss": 0.0015, + "step": 47733 + }, + { + "epoch": 6.25, + "learning_rate": 2.399118515241664e-06, + "loss": 0.0027, + "step": 47734 + }, + { + "epoch": 6.25, + "learning_rate": 2.398773910107062e-06, + "loss": 0.0018, + "step": 47735 + }, + { + "epoch": 6.25, + "learning_rate": 2.3984293263504998e-06, + "loss": 0.0013, + "step": 47736 + }, + { + "epoch": 6.25, + "learning_rate": 2.398084763972951e-06, + "loss": 0.0024, + "step": 47737 + }, + { + "epoch": 6.25, + "learning_rate": 2.3977402229753867e-06, + "loss": 0.002, + "step": 47738 + }, + { + "epoch": 6.25, + "learning_rate": 2.3973957033587725e-06, + "loss": 0.0024, + "step": 47739 + }, + { + "epoch": 6.25, + "learning_rate": 2.3970512051240757e-06, + "loss": 0.0033, + "step": 47740 + }, + { + "epoch": 6.25, + "learning_rate": 2.39670672827227e-06, + "loss": 0.0033, + "step": 47741 + }, + { + "epoch": 6.25, + "learning_rate": 2.396362272804322e-06, + "loss": 0.0019, + "step": 47742 + }, + { + "epoch": 6.25, + "learning_rate": 2.396017838721201e-06, + "loss": 0.0027, + "step": 47743 + }, + { + "epoch": 6.25, + "learning_rate": 2.39567342602387e-06, + "loss": 0.002, + "step": 47744 + }, + { + "epoch": 6.25, + "learning_rate": 2.3953290347133074e-06, + "loss": 0.0015, + "step": 47745 + }, + { + "epoch": 6.25, + "learning_rate": 2.3949846647904763e-06, + "loss": 0.0016, + "step": 47746 + }, + { + "epoch": 6.25, + "learning_rate": 2.3946403162563427e-06, + "loss": 0.0029, + "step": 47747 + }, + { + "epoch": 6.25, + "learning_rate": 2.3942959891118776e-06, + "loss": 0.001, + "step": 47748 + }, + { + "epoch": 6.25, + "learning_rate": 2.3939516833580544e-06, + "loss": 0.0031, + "step": 47749 + }, + { + "epoch": 6.25, + "learning_rate": 2.3936073989958352e-06, + "loss": 0.003, + "step": 47750 + }, + { + "epoch": 6.26, + "learning_rate": 2.3932631360261917e-06, + "loss": 0.0027, + "step": 47751 + }, + { + "epoch": 6.26, + "learning_rate": 2.3929188944500868e-06, + "loss": 0.002, + "step": 47752 + }, + { + "epoch": 6.26, + "learning_rate": 2.3925746742684963e-06, + "loss": 0.0017, + "step": 47753 + }, + { + "epoch": 6.26, + "learning_rate": 2.3922304754823834e-06, + "loss": 0.0028, + "step": 47754 + }, + { + "epoch": 6.26, + "learning_rate": 2.391886298092715e-06, + "loss": 0.0017, + "step": 47755 + }, + { + "epoch": 6.26, + "learning_rate": 2.3915421421004637e-06, + "loss": 0.0019, + "step": 47756 + }, + { + "epoch": 6.26, + "learning_rate": 2.3911980075065945e-06, + "loss": 0.0014, + "step": 47757 + }, + { + "epoch": 6.26, + "learning_rate": 2.390853894312073e-06, + "loss": 0.0017, + "step": 47758 + }, + { + "epoch": 6.26, + "learning_rate": 2.390509802517871e-06, + "loss": 0.0028, + "step": 47759 + }, + { + "epoch": 6.26, + "learning_rate": 2.390165732124956e-06, + "loss": 0.0044, + "step": 47760 + }, + { + "epoch": 6.26, + "learning_rate": 2.3898216831342958e-06, + "loss": 0.0025, + "step": 47761 + }, + { + "epoch": 6.26, + "learning_rate": 2.389477655546857e-06, + "loss": 0.0021, + "step": 47762 + }, + { + "epoch": 6.26, + "learning_rate": 2.389133649363603e-06, + "loss": 0.0025, + "step": 47763 + }, + { + "epoch": 6.26, + "learning_rate": 2.388789664585509e-06, + "loss": 0.0019, + "step": 47764 + }, + { + "epoch": 6.26, + "learning_rate": 2.3884457012135386e-06, + "loss": 0.002, + "step": 47765 + }, + { + "epoch": 6.26, + "learning_rate": 2.388101759248659e-06, + "loss": 0.0026, + "step": 47766 + }, + { + "epoch": 6.26, + "learning_rate": 2.3877578386918353e-06, + "loss": 0.0019, + "step": 47767 + }, + { + "epoch": 6.26, + "learning_rate": 2.38741393954404e-06, + "loss": 0.0016, + "step": 47768 + }, + { + "epoch": 6.26, + "learning_rate": 2.387070061806235e-06, + "loss": 0.0027, + "step": 47769 + }, + { + "epoch": 6.26, + "learning_rate": 2.3867262054793927e-06, + "loss": 0.003, + "step": 47770 + }, + { + "epoch": 6.26, + "learning_rate": 2.386382370564475e-06, + "loss": 0.0029, + "step": 47771 + }, + { + "epoch": 6.26, + "learning_rate": 2.3860385570624544e-06, + "loss": 0.0028, + "step": 47772 + }, + { + "epoch": 6.26, + "learning_rate": 2.3856947649742945e-06, + "loss": 0.0023, + "step": 47773 + }, + { + "epoch": 6.26, + "learning_rate": 2.38535099430096e-06, + "loss": 0.0023, + "step": 47774 + }, + { + "epoch": 6.26, + "learning_rate": 2.3850072450434227e-06, + "loss": 0.0018, + "step": 47775 + }, + { + "epoch": 6.26, + "learning_rate": 2.3846635172026466e-06, + "loss": 0.0017, + "step": 47776 + }, + { + "epoch": 6.26, + "learning_rate": 2.3843198107795997e-06, + "loss": 0.0016, + "step": 47777 + }, + { + "epoch": 6.26, + "learning_rate": 2.383976125775245e-06, + "loss": 0.0028, + "step": 47778 + }, + { + "epoch": 6.26, + "learning_rate": 2.3836324621905538e-06, + "loss": 0.0023, + "step": 47779 + }, + { + "epoch": 6.26, + "learning_rate": 2.383288820026489e-06, + "loss": 0.0019, + "step": 47780 + }, + { + "epoch": 6.26, + "learning_rate": 2.3829451992840214e-06, + "loss": 0.0025, + "step": 47781 + }, + { + "epoch": 6.26, + "learning_rate": 2.382601599964112e-06, + "loss": 0.0027, + "step": 47782 + }, + { + "epoch": 6.26, + "learning_rate": 2.3822580220677317e-06, + "loss": 0.0016, + "step": 47783 + }, + { + "epoch": 6.26, + "learning_rate": 2.381914465595846e-06, + "loss": 0.0018, + "step": 47784 + }, + { + "epoch": 6.26, + "learning_rate": 2.3815709305494194e-06, + "loss": 0.0049, + "step": 47785 + }, + { + "epoch": 6.26, + "learning_rate": 2.3812274169294167e-06, + "loss": 0.0028, + "step": 47786 + }, + { + "epoch": 6.26, + "learning_rate": 2.3808839247368086e-06, + "loss": 0.0012, + "step": 47787 + }, + { + "epoch": 6.26, + "learning_rate": 2.3805404539725594e-06, + "loss": 0.0018, + "step": 47788 + }, + { + "epoch": 6.26, + "learning_rate": 2.380197004637631e-06, + "loss": 0.0034, + "step": 47789 + }, + { + "epoch": 6.26, + "learning_rate": 2.3798535767329957e-06, + "loss": 0.0022, + "step": 47790 + }, + { + "epoch": 6.26, + "learning_rate": 2.3795101702596136e-06, + "loss": 0.0012, + "step": 47791 + }, + { + "epoch": 6.26, + "learning_rate": 2.3791667852184563e-06, + "loss": 0.0025, + "step": 47792 + }, + { + "epoch": 6.26, + "learning_rate": 2.3788234216104864e-06, + "loss": 0.0024, + "step": 47793 + }, + { + "epoch": 6.26, + "learning_rate": 2.378480079436667e-06, + "loss": 0.0024, + "step": 47794 + }, + { + "epoch": 6.26, + "learning_rate": 2.37813675869797e-06, + "loss": 0.0024, + "step": 47795 + }, + { + "epoch": 6.26, + "learning_rate": 2.3777934593953565e-06, + "loss": 0.0031, + "step": 47796 + }, + { + "epoch": 6.26, + "learning_rate": 2.37745018152979e-06, + "loss": 0.0024, + "step": 47797 + }, + { + "epoch": 6.26, + "learning_rate": 2.377106925102243e-06, + "loss": 0.0024, + "step": 47798 + }, + { + "epoch": 6.26, + "learning_rate": 2.376763690113676e-06, + "loss": 0.0046, + "step": 47799 + }, + { + "epoch": 6.26, + "learning_rate": 2.3764204765650557e-06, + "loss": 0.0035, + "step": 47800 + }, + { + "epoch": 6.26, + "learning_rate": 2.3760772844573433e-06, + "loss": 0.0013, + "step": 47801 + }, + { + "epoch": 6.26, + "learning_rate": 2.375734113791508e-06, + "loss": 0.002, + "step": 47802 + }, + { + "epoch": 6.26, + "learning_rate": 2.375390964568517e-06, + "loss": 0.0014, + "step": 47803 + }, + { + "epoch": 6.26, + "learning_rate": 2.3750478367893327e-06, + "loss": 0.0017, + "step": 47804 + }, + { + "epoch": 6.26, + "learning_rate": 2.3747047304549176e-06, + "loss": 0.0029, + "step": 47805 + }, + { + "epoch": 6.26, + "learning_rate": 2.374361645566242e-06, + "loss": 0.0019, + "step": 47806 + }, + { + "epoch": 6.26, + "learning_rate": 2.374018582124269e-06, + "loss": 0.0023, + "step": 47807 + }, + { + "epoch": 6.26, + "learning_rate": 2.373675540129963e-06, + "loss": 0.002, + "step": 47808 + }, + { + "epoch": 6.26, + "learning_rate": 2.373332519584284e-06, + "loss": 0.002, + "step": 47809 + }, + { + "epoch": 6.26, + "learning_rate": 2.372989520488205e-06, + "loss": 0.0019, + "step": 47810 + }, + { + "epoch": 6.26, + "learning_rate": 2.372646542842686e-06, + "loss": 0.002, + "step": 47811 + }, + { + "epoch": 6.26, + "learning_rate": 2.372303586648691e-06, + "loss": 0.0023, + "step": 47812 + }, + { + "epoch": 6.26, + "learning_rate": 2.3719606519071846e-06, + "loss": 0.0028, + "step": 47813 + }, + { + "epoch": 6.26, + "learning_rate": 2.3716177386191362e-06, + "loss": 0.0024, + "step": 47814 + }, + { + "epoch": 6.26, + "learning_rate": 2.3712748467855063e-06, + "loss": 0.0023, + "step": 47815 + }, + { + "epoch": 6.26, + "learning_rate": 2.37093197640726e-06, + "loss": 0.0023, + "step": 47816 + }, + { + "epoch": 6.26, + "learning_rate": 2.3705891274853578e-06, + "loss": 0.0018, + "step": 47817 + }, + { + "epoch": 6.26, + "learning_rate": 2.370246300020771e-06, + "loss": 0.0037, + "step": 47818 + }, + { + "epoch": 6.26, + "learning_rate": 2.3699034940144594e-06, + "loss": 0.0019, + "step": 47819 + }, + { + "epoch": 6.26, + "learning_rate": 2.3695607094673854e-06, + "loss": 0.0038, + "step": 47820 + }, + { + "epoch": 6.26, + "learning_rate": 2.3692179463805187e-06, + "loss": 0.0052, + "step": 47821 + }, + { + "epoch": 6.26, + "learning_rate": 2.3688752047548204e-06, + "loss": 0.0012, + "step": 47822 + }, + { + "epoch": 6.26, + "learning_rate": 2.368532484591254e-06, + "loss": 0.0016, + "step": 47823 + }, + { + "epoch": 6.26, + "learning_rate": 2.3681897858907764e-06, + "loss": 0.0022, + "step": 47824 + }, + { + "epoch": 6.26, + "learning_rate": 2.367847108654366e-06, + "loss": 0.0042, + "step": 47825 + }, + { + "epoch": 6.26, + "learning_rate": 2.367504452882979e-06, + "loss": 0.0025, + "step": 47826 + }, + { + "epoch": 6.26, + "learning_rate": 2.3671618185775792e-06, + "loss": 0.0018, + "step": 47827 + }, + { + "epoch": 6.27, + "learning_rate": 2.366819205739127e-06, + "loss": 0.0025, + "step": 47828 + }, + { + "epoch": 6.27, + "learning_rate": 2.3664766143685925e-06, + "loss": 0.0015, + "step": 47829 + }, + { + "epoch": 6.27, + "learning_rate": 2.3661340444669357e-06, + "loss": 0.0025, + "step": 47830 + }, + { + "epoch": 6.27, + "learning_rate": 2.3657914960351205e-06, + "loss": 0.0019, + "step": 47831 + }, + { + "epoch": 6.27, + "learning_rate": 2.3654489690741077e-06, + "loss": 0.0017, + "step": 47832 + }, + { + "epoch": 6.27, + "learning_rate": 2.365106463584865e-06, + "loss": 0.0014, + "step": 47833 + }, + { + "epoch": 6.27, + "learning_rate": 2.3647639795683554e-06, + "loss": 0.0016, + "step": 47834 + }, + { + "epoch": 6.27, + "learning_rate": 2.3644215170255357e-06, + "loss": 0.0022, + "step": 47835 + }, + { + "epoch": 6.27, + "learning_rate": 2.364079075957375e-06, + "loss": 0.0015, + "step": 47836 + }, + { + "epoch": 6.27, + "learning_rate": 2.3637366563648377e-06, + "loss": 0.0026, + "step": 47837 + }, + { + "epoch": 6.27, + "learning_rate": 2.3633942582488856e-06, + "loss": 0.0027, + "step": 47838 + }, + { + "epoch": 6.27, + "learning_rate": 2.3630518816104763e-06, + "loss": 0.0025, + "step": 47839 + }, + { + "epoch": 6.27, + "learning_rate": 2.3627095264505808e-06, + "loss": 0.0019, + "step": 47840 + }, + { + "epoch": 6.27, + "learning_rate": 2.3623671927701575e-06, + "loss": 0.0024, + "step": 47841 + }, + { + "epoch": 6.27, + "learning_rate": 2.3620248805701697e-06, + "loss": 0.0018, + "step": 47842 + }, + { + "epoch": 6.27, + "learning_rate": 2.3616825898515773e-06, + "loss": 0.0012, + "step": 47843 + }, + { + "epoch": 6.27, + "learning_rate": 2.3613403206153486e-06, + "loss": 0.0022, + "step": 47844 + }, + { + "epoch": 6.27, + "learning_rate": 2.3609980728624427e-06, + "loss": 0.0036, + "step": 47845 + }, + { + "epoch": 6.27, + "learning_rate": 2.360655846593821e-06, + "loss": 0.0018, + "step": 47846 + }, + { + "epoch": 6.27, + "learning_rate": 2.360313641810448e-06, + "loss": 0.0023, + "step": 47847 + }, + { + "epoch": 6.27, + "learning_rate": 2.3599714585132883e-06, + "loss": 0.0023, + "step": 47848 + }, + { + "epoch": 6.27, + "learning_rate": 2.359629296703303e-06, + "loss": 0.0023, + "step": 47849 + }, + { + "epoch": 6.27, + "learning_rate": 2.3592871563814524e-06, + "loss": 0.0015, + "step": 47850 + }, + { + "epoch": 6.27, + "learning_rate": 2.3589450375486975e-06, + "loss": 0.002, + "step": 47851 + }, + { + "epoch": 6.27, + "learning_rate": 2.358602940206005e-06, + "loss": 0.0023, + "step": 47852 + }, + { + "epoch": 6.27, + "learning_rate": 2.3582608643543358e-06, + "loss": 0.0023, + "step": 47853 + }, + { + "epoch": 6.27, + "learning_rate": 2.357918809994648e-06, + "loss": 0.0019, + "step": 47854 + }, + { + "epoch": 6.27, + "learning_rate": 2.357576777127909e-06, + "loss": 0.0038, + "step": 47855 + }, + { + "epoch": 6.27, + "learning_rate": 2.357234765755079e-06, + "loss": 0.0028, + "step": 47856 + }, + { + "epoch": 6.27, + "learning_rate": 2.3568927758771153e-06, + "loss": 0.0016, + "step": 47857 + }, + { + "epoch": 6.27, + "learning_rate": 2.3565508074949874e-06, + "loss": 0.0021, + "step": 47858 + }, + { + "epoch": 6.27, + "learning_rate": 2.356208860609649e-06, + "loss": 0.0014, + "step": 47859 + }, + { + "epoch": 6.27, + "learning_rate": 2.355866935222071e-06, + "loss": 0.002, + "step": 47860 + }, + { + "epoch": 6.27, + "learning_rate": 2.3555250313332088e-06, + "loss": 0.0021, + "step": 47861 + }, + { + "epoch": 6.27, + "learning_rate": 2.3551831489440226e-06, + "loss": 0.0033, + "step": 47862 + }, + { + "epoch": 6.27, + "learning_rate": 2.3548412880554795e-06, + "loss": 0.0035, + "step": 47863 + }, + { + "epoch": 6.27, + "learning_rate": 2.354499448668539e-06, + "loss": 0.0029, + "step": 47864 + }, + { + "epoch": 6.27, + "learning_rate": 2.3541576307841607e-06, + "loss": 0.0009, + "step": 47865 + }, + { + "epoch": 6.27, + "learning_rate": 2.3538158344033044e-06, + "loss": 0.0031, + "step": 47866 + }, + { + "epoch": 6.27, + "learning_rate": 2.3534740595269367e-06, + "loss": 0.0012, + "step": 47867 + }, + { + "epoch": 6.27, + "learning_rate": 2.353132306156014e-06, + "loss": 0.0018, + "step": 47868 + }, + { + "epoch": 6.27, + "learning_rate": 2.3527905742915014e-06, + "loss": 0.0042, + "step": 47869 + }, + { + "epoch": 6.27, + "learning_rate": 2.3524488639343557e-06, + "loss": 0.002, + "step": 47870 + }, + { + "epoch": 6.27, + "learning_rate": 2.352107175085543e-06, + "loss": 0.0036, + "step": 47871 + }, + { + "epoch": 6.27, + "learning_rate": 2.351765507746021e-06, + "loss": 0.0029, + "step": 47872 + }, + { + "epoch": 6.27, + "learning_rate": 2.351423861916752e-06, + "loss": 0.0038, + "step": 47873 + }, + { + "epoch": 6.27, + "learning_rate": 2.351082237598692e-06, + "loss": 0.0016, + "step": 47874 + }, + { + "epoch": 6.27, + "learning_rate": 2.3507406347928097e-06, + "loss": 0.0029, + "step": 47875 + }, + { + "epoch": 6.27, + "learning_rate": 2.3503990535000622e-06, + "loss": 0.0019, + "step": 47876 + }, + { + "epoch": 6.27, + "learning_rate": 2.350057493721407e-06, + "loss": 0.0034, + "step": 47877 + }, + { + "epoch": 6.27, + "learning_rate": 2.349715955457811e-06, + "loss": 0.0027, + "step": 47878 + }, + { + "epoch": 6.27, + "learning_rate": 2.349374438710228e-06, + "loss": 0.0021, + "step": 47879 + }, + { + "epoch": 6.27, + "learning_rate": 2.3490329434796244e-06, + "loss": 0.0014, + "step": 47880 + }, + { + "epoch": 6.27, + "learning_rate": 2.3486914697669595e-06, + "loss": 0.0022, + "step": 47881 + }, + { + "epoch": 6.27, + "learning_rate": 2.3483500175731877e-06, + "loss": 0.0024, + "step": 47882 + }, + { + "epoch": 6.27, + "learning_rate": 2.3480085868992786e-06, + "loss": 0.002, + "step": 47883 + }, + { + "epoch": 6.27, + "learning_rate": 2.3476671777461867e-06, + "loss": 0.0016, + "step": 47884 + }, + { + "epoch": 6.27, + "learning_rate": 2.3473257901148718e-06, + "loss": 0.0012, + "step": 47885 + }, + { + "epoch": 6.27, + "learning_rate": 2.346984424006297e-06, + "loss": 0.0015, + "step": 47886 + }, + { + "epoch": 6.27, + "learning_rate": 2.346643079421421e-06, + "loss": 0.0017, + "step": 47887 + }, + { + "epoch": 6.27, + "learning_rate": 2.3463017563612044e-06, + "loss": 0.0015, + "step": 47888 + }, + { + "epoch": 6.27, + "learning_rate": 2.3459604548266012e-06, + "loss": 0.0016, + "step": 47889 + }, + { + "epoch": 6.27, + "learning_rate": 2.345619174818582e-06, + "loss": 0.0018, + "step": 47890 + }, + { + "epoch": 6.27, + "learning_rate": 2.3452779163381014e-06, + "loss": 0.0032, + "step": 47891 + }, + { + "epoch": 6.27, + "learning_rate": 2.3449366793861182e-06, + "loss": 0.0023, + "step": 47892 + }, + { + "epoch": 6.27, + "learning_rate": 2.3445954639635903e-06, + "loss": 0.0017, + "step": 47893 + }, + { + "epoch": 6.27, + "learning_rate": 2.3442542700714833e-06, + "loss": 0.0027, + "step": 47894 + }, + { + "epoch": 6.27, + "learning_rate": 2.343913097710754e-06, + "loss": 0.0022, + "step": 47895 + }, + { + "epoch": 6.27, + "learning_rate": 2.343571946882357e-06, + "loss": 0.0024, + "step": 47896 + }, + { + "epoch": 6.27, + "learning_rate": 2.343230817587261e-06, + "loss": 0.0025, + "step": 47897 + }, + { + "epoch": 6.27, + "learning_rate": 2.342889709826419e-06, + "loss": 0.0032, + "step": 47898 + }, + { + "epoch": 6.27, + "learning_rate": 2.3425486236007923e-06, + "loss": 0.0022, + "step": 47899 + }, + { + "epoch": 6.27, + "learning_rate": 2.342207558911337e-06, + "loss": 0.0014, + "step": 47900 + }, + { + "epoch": 6.27, + "learning_rate": 2.341866515759016e-06, + "loss": 0.0017, + "step": 47901 + }, + { + "epoch": 6.27, + "learning_rate": 2.3415254941447896e-06, + "loss": 0.0022, + "step": 47902 + }, + { + "epoch": 6.27, + "learning_rate": 2.341184494069615e-06, + "loss": 0.0023, + "step": 47903 + }, + { + "epoch": 6.28, + "learning_rate": 2.340843515534448e-06, + "loss": 0.0021, + "step": 47904 + }, + { + "epoch": 6.28, + "learning_rate": 2.340502558540254e-06, + "loss": 0.001, + "step": 47905 + }, + { + "epoch": 6.28, + "learning_rate": 2.340161623087989e-06, + "loss": 0.0021, + "step": 47906 + }, + { + "epoch": 6.28, + "learning_rate": 2.3398207091786105e-06, + "loss": 0.0021, + "step": 47907 + }, + { + "epoch": 6.28, + "learning_rate": 2.339479816813076e-06, + "loss": 0.0015, + "step": 47908 + }, + { + "epoch": 6.28, + "learning_rate": 2.33913894599235e-06, + "loss": 0.0013, + "step": 47909 + }, + { + "epoch": 6.28, + "learning_rate": 2.3387980967173863e-06, + "loss": 0.0025, + "step": 47910 + }, + { + "epoch": 6.28, + "learning_rate": 2.3384572689891428e-06, + "loss": 0.0024, + "step": 47911 + }, + { + "epoch": 6.28, + "learning_rate": 2.3381164628085805e-06, + "loss": 0.002, + "step": 47912 + }, + { + "epoch": 6.28, + "learning_rate": 2.33777567817666e-06, + "loss": 0.0019, + "step": 47913 + }, + { + "epoch": 6.28, + "learning_rate": 2.3374349150943387e-06, + "loss": 0.0017, + "step": 47914 + }, + { + "epoch": 6.28, + "learning_rate": 2.337094173562573e-06, + "loss": 0.0017, + "step": 47915 + }, + { + "epoch": 6.28, + "learning_rate": 2.336753453582319e-06, + "loss": 0.0023, + "step": 47916 + }, + { + "epoch": 6.28, + "learning_rate": 2.3364127551545403e-06, + "loss": 0.0014, + "step": 47917 + }, + { + "epoch": 6.28, + "learning_rate": 2.3360720782801927e-06, + "loss": 0.0022, + "step": 47918 + }, + { + "epoch": 6.28, + "learning_rate": 2.335731422960231e-06, + "loss": 0.0016, + "step": 47919 + }, + { + "epoch": 6.28, + "learning_rate": 2.3353907891956198e-06, + "loss": 0.0033, + "step": 47920 + }, + { + "epoch": 6.28, + "learning_rate": 2.3350501769873135e-06, + "loss": 0.0021, + "step": 47921 + }, + { + "epoch": 6.28, + "learning_rate": 2.334709586336268e-06, + "loss": 0.0024, + "step": 47922 + }, + { + "epoch": 6.28, + "learning_rate": 2.3343690172434464e-06, + "loss": 0.0022, + "step": 47923 + }, + { + "epoch": 6.28, + "learning_rate": 2.3340284697098016e-06, + "loss": 0.0014, + "step": 47924 + }, + { + "epoch": 6.28, + "learning_rate": 2.3336879437362946e-06, + "loss": 0.002, + "step": 47925 + }, + { + "epoch": 6.28, + "learning_rate": 2.333347439323883e-06, + "loss": 0.0025, + "step": 47926 + }, + { + "epoch": 6.28, + "learning_rate": 2.3330069564735212e-06, + "loss": 0.0027, + "step": 47927 + }, + { + "epoch": 6.28, + "learning_rate": 2.332666495186171e-06, + "loss": 0.0017, + "step": 47928 + }, + { + "epoch": 6.28, + "learning_rate": 2.332326055462788e-06, + "loss": 0.0019, + "step": 47929 + }, + { + "epoch": 6.28, + "learning_rate": 2.3319856373043302e-06, + "loss": 0.002, + "step": 47930 + }, + { + "epoch": 6.28, + "learning_rate": 2.3316452407117517e-06, + "loss": 0.0016, + "step": 47931 + }, + { + "epoch": 6.28, + "learning_rate": 2.3313048656860147e-06, + "loss": 0.0019, + "step": 47932 + }, + { + "epoch": 6.28, + "learning_rate": 2.3309645122280723e-06, + "loss": 0.003, + "step": 47933 + }, + { + "epoch": 6.28, + "learning_rate": 2.3306241803388863e-06, + "loss": 0.0016, + "step": 47934 + }, + { + "epoch": 6.28, + "learning_rate": 2.3302838700194073e-06, + "loss": 0.0016, + "step": 47935 + }, + { + "epoch": 6.28, + "learning_rate": 2.3299435812706007e-06, + "loss": 0.0014, + "step": 47936 + }, + { + "epoch": 6.28, + "learning_rate": 2.3296033140934193e-06, + "loss": 0.0021, + "step": 47937 + }, + { + "epoch": 6.28, + "learning_rate": 2.3292630684888195e-06, + "loss": 0.002, + "step": 47938 + }, + { + "epoch": 6.28, + "learning_rate": 2.328922844457757e-06, + "loss": 0.0027, + "step": 47939 + }, + { + "epoch": 6.28, + "learning_rate": 2.3285826420011926e-06, + "loss": 0.0016, + "step": 47940 + }, + { + "epoch": 6.28, + "learning_rate": 2.3282424611200815e-06, + "loss": 0.0029, + "step": 47941 + }, + { + "epoch": 6.28, + "learning_rate": 2.3279023018153767e-06, + "loss": 0.0017, + "step": 47942 + }, + { + "epoch": 6.28, + "learning_rate": 2.327562164088041e-06, + "loss": 0.0034, + "step": 47943 + }, + { + "epoch": 6.28, + "learning_rate": 2.327222047939025e-06, + "loss": 0.0017, + "step": 47944 + }, + { + "epoch": 6.28, + "learning_rate": 2.326881953369291e-06, + "loss": 0.0012, + "step": 47945 + }, + { + "epoch": 6.28, + "learning_rate": 2.3265418803797913e-06, + "loss": 0.0015, + "step": 47946 + }, + { + "epoch": 6.28, + "learning_rate": 2.326201828971486e-06, + "loss": 0.0018, + "step": 47947 + }, + { + "epoch": 6.28, + "learning_rate": 2.3258617991453293e-06, + "loss": 0.0017, + "step": 47948 + }, + { + "epoch": 6.28, + "learning_rate": 2.3255217909022778e-06, + "loss": 0.003, + "step": 47949 + }, + { + "epoch": 6.28, + "learning_rate": 2.325181804243284e-06, + "loss": 0.0013, + "step": 47950 + }, + { + "epoch": 6.28, + "learning_rate": 2.3248418391693117e-06, + "loss": 0.0018, + "step": 47951 + }, + { + "epoch": 6.28, + "learning_rate": 2.3245018956813124e-06, + "loss": 0.0023, + "step": 47952 + }, + { + "epoch": 6.28, + "learning_rate": 2.3241619737802424e-06, + "loss": 0.0018, + "step": 47953 + }, + { + "epoch": 6.28, + "learning_rate": 2.3238220734670557e-06, + "loss": 0.0013, + "step": 47954 + }, + { + "epoch": 6.28, + "learning_rate": 2.323482194742711e-06, + "loss": 0.0029, + "step": 47955 + }, + { + "epoch": 6.28, + "learning_rate": 2.3231423376081663e-06, + "loss": 0.0027, + "step": 47956 + }, + { + "epoch": 6.28, + "learning_rate": 2.3228025020643753e-06, + "loss": 0.0022, + "step": 47957 + }, + { + "epoch": 6.28, + "learning_rate": 2.322462688112289e-06, + "loss": 0.002, + "step": 47958 + }, + { + "epoch": 6.28, + "learning_rate": 2.322122895752872e-06, + "loss": 0.0009, + "step": 47959 + }, + { + "epoch": 6.28, + "learning_rate": 2.3217831249870747e-06, + "loss": 0.002, + "step": 47960 + }, + { + "epoch": 6.28, + "learning_rate": 2.3214433758158518e-06, + "loss": 0.0009, + "step": 47961 + }, + { + "epoch": 6.28, + "learning_rate": 2.321103648240163e-06, + "loss": 0.0012, + "step": 47962 + }, + { + "epoch": 6.28, + "learning_rate": 2.320763942260961e-06, + "loss": 0.0025, + "step": 47963 + }, + { + "epoch": 6.28, + "learning_rate": 2.3204242578792004e-06, + "loss": 0.003, + "step": 47964 + }, + { + "epoch": 6.28, + "learning_rate": 2.320084595095836e-06, + "loss": 0.002, + "step": 47965 + }, + { + "epoch": 6.28, + "learning_rate": 2.3197449539118245e-06, + "loss": 0.0031, + "step": 47966 + }, + { + "epoch": 6.28, + "learning_rate": 2.3194053343281243e-06, + "loss": 0.0022, + "step": 47967 + }, + { + "epoch": 6.28, + "learning_rate": 2.319065736345687e-06, + "loss": 0.0015, + "step": 47968 + }, + { + "epoch": 6.28, + "learning_rate": 2.318726159965465e-06, + "loss": 0.0027, + "step": 47969 + }, + { + "epoch": 6.28, + "learning_rate": 2.318386605188421e-06, + "loss": 0.001, + "step": 47970 + }, + { + "epoch": 6.28, + "learning_rate": 2.3180470720155047e-06, + "loss": 0.0023, + "step": 47971 + }, + { + "epoch": 6.28, + "learning_rate": 2.317707560447672e-06, + "loss": 0.0024, + "step": 47972 + }, + { + "epoch": 6.28, + "learning_rate": 2.3173680704858747e-06, + "loss": 0.003, + "step": 47973 + }, + { + "epoch": 6.28, + "learning_rate": 2.317028602131074e-06, + "loss": 0.0019, + "step": 47974 + }, + { + "epoch": 6.28, + "learning_rate": 2.316689155384221e-06, + "loss": 0.0023, + "step": 47975 + }, + { + "epoch": 6.28, + "learning_rate": 2.3163497302462667e-06, + "loss": 0.0027, + "step": 47976 + }, + { + "epoch": 6.28, + "learning_rate": 2.316010326718171e-06, + "loss": 0.0033, + "step": 47977 + }, + { + "epoch": 6.28, + "learning_rate": 2.3156709448008885e-06, + "loss": 0.0016, + "step": 47978 + }, + { + "epoch": 6.28, + "learning_rate": 2.315331584495374e-06, + "loss": 0.0024, + "step": 47979 + }, + { + "epoch": 6.29, + "learning_rate": 2.3149922458025785e-06, + "loss": 0.0011, + "step": 47980 + }, + { + "epoch": 6.29, + "learning_rate": 2.3146529287234563e-06, + "loss": 0.0017, + "step": 47981 + }, + { + "epoch": 6.29, + "learning_rate": 2.3143136332589666e-06, + "loss": 0.0018, + "step": 47982 + }, + { + "epoch": 6.29, + "learning_rate": 2.3139743594100593e-06, + "loss": 0.0017, + "step": 47983 + }, + { + "epoch": 6.29, + "learning_rate": 2.3136351071776873e-06, + "loss": 0.0017, + "step": 47984 + }, + { + "epoch": 6.29, + "learning_rate": 2.3132958765628113e-06, + "loss": 0.0021, + "step": 47985 + }, + { + "epoch": 6.29, + "learning_rate": 2.3129566675663806e-06, + "loss": 0.0015, + "step": 47986 + }, + { + "epoch": 6.29, + "learning_rate": 2.3126174801893495e-06, + "loss": 0.0022, + "step": 47987 + }, + { + "epoch": 6.29, + "learning_rate": 2.31227831443267e-06, + "loss": 0.0017, + "step": 47988 + }, + { + "epoch": 6.29, + "learning_rate": 2.311939170297297e-06, + "loss": 0.0018, + "step": 47989 + }, + { + "epoch": 6.29, + "learning_rate": 2.3116000477841906e-06, + "loss": 0.0014, + "step": 47990 + }, + { + "epoch": 6.29, + "learning_rate": 2.311260946894298e-06, + "loss": 0.0025, + "step": 47991 + }, + { + "epoch": 6.29, + "learning_rate": 2.3109218676285726e-06, + "loss": 0.0032, + "step": 47992 + }, + { + "epoch": 6.29, + "learning_rate": 2.310582809987972e-06, + "loss": 0.0037, + "step": 47993 + }, + { + "epoch": 6.29, + "learning_rate": 2.3102437739734486e-06, + "loss": 0.003, + "step": 47994 + }, + { + "epoch": 6.29, + "learning_rate": 2.3099047595859546e-06, + "loss": 0.0018, + "step": 47995 + }, + { + "epoch": 6.29, + "learning_rate": 2.3095657668264416e-06, + "loss": 0.0014, + "step": 47996 + }, + { + "epoch": 6.29, + "learning_rate": 2.3092267956958682e-06, + "loss": 0.0018, + "step": 47997 + }, + { + "epoch": 6.29, + "learning_rate": 2.3088878461951845e-06, + "loss": 0.0027, + "step": 47998 + }, + { + "epoch": 6.29, + "learning_rate": 2.3085489183253405e-06, + "loss": 0.0019, + "step": 47999 + }, + { + "epoch": 6.29, + "learning_rate": 2.308210012087294e-06, + "loss": 0.0016, + "step": 48000 + }, + { + "epoch": 6.29, + "learning_rate": 2.307871127482001e-06, + "loss": 0.0028, + "step": 48001 + }, + { + "epoch": 6.29, + "learning_rate": 2.3075322645104094e-06, + "loss": 0.0019, + "step": 48002 + }, + { + "epoch": 6.29, + "learning_rate": 2.3071934231734737e-06, + "loss": 0.0027, + "step": 48003 + }, + { + "epoch": 6.29, + "learning_rate": 2.306854603472145e-06, + "loss": 0.002, + "step": 48004 + }, + { + "epoch": 6.29, + "learning_rate": 2.30651580540738e-06, + "loss": 0.0019, + "step": 48005 + }, + { + "epoch": 6.29, + "learning_rate": 2.30617702898013e-06, + "loss": 0.0028, + "step": 48006 + }, + { + "epoch": 6.29, + "learning_rate": 2.3058382741913445e-06, + "loss": 0.0025, + "step": 48007 + }, + { + "epoch": 6.29, + "learning_rate": 2.305499541041981e-06, + "loss": 0.0029, + "step": 48008 + }, + { + "epoch": 6.29, + "learning_rate": 2.3051608295329917e-06, + "loss": 0.0021, + "step": 48009 + }, + { + "epoch": 6.29, + "learning_rate": 2.3048221396653248e-06, + "loss": 0.0019, + "step": 48010 + }, + { + "epoch": 6.29, + "learning_rate": 2.3044834714399354e-06, + "loss": 0.0013, + "step": 48011 + }, + { + "epoch": 6.29, + "learning_rate": 2.3041448248577793e-06, + "loss": 0.0018, + "step": 48012 + }, + { + "epoch": 6.29, + "learning_rate": 2.303806199919806e-06, + "loss": 0.0046, + "step": 48013 + }, + { + "epoch": 6.29, + "learning_rate": 2.303467596626968e-06, + "loss": 0.0019, + "step": 48014 + }, + { + "epoch": 6.29, + "learning_rate": 2.3031290149802145e-06, + "loss": 0.0034, + "step": 48015 + }, + { + "epoch": 6.29, + "learning_rate": 2.3027904549805034e-06, + "loss": 0.0012, + "step": 48016 + }, + { + "epoch": 6.29, + "learning_rate": 2.302451916628785e-06, + "loss": 0.002, + "step": 48017 + }, + { + "epoch": 6.29, + "learning_rate": 2.3021133999260103e-06, + "loss": 0.0018, + "step": 48018 + }, + { + "epoch": 6.29, + "learning_rate": 2.3017749048731285e-06, + "loss": 0.0021, + "step": 48019 + }, + { + "epoch": 6.29, + "learning_rate": 2.301436431471096e-06, + "loss": 0.002, + "step": 48020 + }, + { + "epoch": 6.29, + "learning_rate": 2.3010979797208655e-06, + "loss": 0.0014, + "step": 48021 + }, + { + "epoch": 6.29, + "learning_rate": 2.3007595496233868e-06, + "loss": 0.003, + "step": 48022 + }, + { + "epoch": 6.29, + "learning_rate": 2.3004211411796086e-06, + "loss": 0.0019, + "step": 48023 + }, + { + "epoch": 6.29, + "learning_rate": 2.3000827543904904e-06, + "loss": 0.0023, + "step": 48024 + }, + { + "epoch": 6.29, + "learning_rate": 2.299744389256978e-06, + "loss": 0.002, + "step": 48025 + }, + { + "epoch": 6.29, + "learning_rate": 2.2994060457800214e-06, + "loss": 0.0018, + "step": 48026 + }, + { + "epoch": 6.29, + "learning_rate": 2.2990677239605786e-06, + "loss": 0.0024, + "step": 48027 + }, + { + "epoch": 6.29, + "learning_rate": 2.2987294237995985e-06, + "loss": 0.0012, + "step": 48028 + }, + { + "epoch": 6.29, + "learning_rate": 2.2983911452980313e-06, + "loss": 0.0023, + "step": 48029 + }, + { + "epoch": 6.29, + "learning_rate": 2.2980528884568255e-06, + "loss": 0.0025, + "step": 48030 + }, + { + "epoch": 6.29, + "learning_rate": 2.2977146532769355e-06, + "loss": 0.0022, + "step": 48031 + }, + { + "epoch": 6.29, + "learning_rate": 2.2973764397593168e-06, + "loss": 0.003, + "step": 48032 + }, + { + "epoch": 6.29, + "learning_rate": 2.297038247904916e-06, + "loss": 0.0024, + "step": 48033 + }, + { + "epoch": 6.29, + "learning_rate": 2.296700077714682e-06, + "loss": 0.0014, + "step": 48034 + }, + { + "epoch": 6.29, + "learning_rate": 2.2963619291895723e-06, + "loss": 0.0031, + "step": 48035 + }, + { + "epoch": 6.29, + "learning_rate": 2.2960238023305337e-06, + "loss": 0.0035, + "step": 48036 + }, + { + "epoch": 6.29, + "learning_rate": 2.2956856971385187e-06, + "loss": 0.0016, + "step": 48037 + }, + { + "epoch": 6.29, + "learning_rate": 2.295347613614474e-06, + "loss": 0.0025, + "step": 48038 + }, + { + "epoch": 6.29, + "learning_rate": 2.2950095517593562e-06, + "loss": 0.0021, + "step": 48039 + }, + { + "epoch": 6.29, + "learning_rate": 2.2946715115741146e-06, + "loss": 0.0025, + "step": 48040 + }, + { + "epoch": 6.29, + "learning_rate": 2.294333493059696e-06, + "loss": 0.0036, + "step": 48041 + }, + { + "epoch": 6.29, + "learning_rate": 2.293995496217054e-06, + "loss": 0.0031, + "step": 48042 + }, + { + "epoch": 6.29, + "learning_rate": 2.293657521047142e-06, + "loss": 0.0016, + "step": 48043 + }, + { + "epoch": 6.29, + "learning_rate": 2.2933195675509078e-06, + "loss": 0.0026, + "step": 48044 + }, + { + "epoch": 6.29, + "learning_rate": 2.292981635729302e-06, + "loss": 0.0029, + "step": 48045 + }, + { + "epoch": 6.29, + "learning_rate": 2.2926437255832724e-06, + "loss": 0.0024, + "step": 48046 + }, + { + "epoch": 6.29, + "learning_rate": 2.2923058371137743e-06, + "loss": 0.0013, + "step": 48047 + }, + { + "epoch": 6.29, + "learning_rate": 2.2919679703217555e-06, + "loss": 0.0045, + "step": 48048 + }, + { + "epoch": 6.29, + "learning_rate": 2.291630125208164e-06, + "loss": 0.0022, + "step": 48049 + }, + { + "epoch": 6.29, + "learning_rate": 2.291292301773955e-06, + "loss": 0.0029, + "step": 48050 + }, + { + "epoch": 6.29, + "learning_rate": 2.2909545000200762e-06, + "loss": 0.0018, + "step": 48051 + }, + { + "epoch": 6.29, + "learning_rate": 2.290616719947476e-06, + "loss": 0.0034, + "step": 48052 + }, + { + "epoch": 6.29, + "learning_rate": 2.2902789615571043e-06, + "loss": 0.0009, + "step": 48053 + }, + { + "epoch": 6.29, + "learning_rate": 2.289941224849912e-06, + "loss": 0.0029, + "step": 48054 + }, + { + "epoch": 6.29, + "learning_rate": 2.289603509826852e-06, + "loss": 0.0024, + "step": 48055 + }, + { + "epoch": 6.29, + "learning_rate": 2.2892658164888713e-06, + "loss": 0.0024, + "step": 48056 + }, + { + "epoch": 6.3, + "learning_rate": 2.2889281448369184e-06, + "loss": 0.0013, + "step": 48057 + }, + { + "epoch": 6.3, + "learning_rate": 2.2885904948719463e-06, + "loss": 0.0018, + "step": 48058 + }, + { + "epoch": 6.3, + "learning_rate": 2.2882528665949022e-06, + "loss": 0.0026, + "step": 48059 + }, + { + "epoch": 6.3, + "learning_rate": 2.287915260006737e-06, + "loss": 0.0013, + "step": 48060 + }, + { + "epoch": 6.3, + "learning_rate": 2.287577675108397e-06, + "loss": 0.0019, + "step": 48061 + }, + { + "epoch": 6.3, + "learning_rate": 2.287240111900836e-06, + "loss": 0.0025, + "step": 48062 + }, + { + "epoch": 6.3, + "learning_rate": 2.2869025703850013e-06, + "loss": 0.001, + "step": 48063 + }, + { + "epoch": 6.3, + "learning_rate": 2.28656505056184e-06, + "loss": 0.0017, + "step": 48064 + }, + { + "epoch": 6.3, + "learning_rate": 2.2862275524323042e-06, + "loss": 0.002, + "step": 48065 + }, + { + "epoch": 6.3, + "learning_rate": 2.2858900759973446e-06, + "loss": 0.0036, + "step": 48066 + }, + { + "epoch": 6.3, + "learning_rate": 2.2855526212579094e-06, + "loss": 0.0015, + "step": 48067 + }, + { + "epoch": 6.3, + "learning_rate": 2.2852151882149454e-06, + "loss": 0.0024, + "step": 48068 + }, + { + "epoch": 6.3, + "learning_rate": 2.2848777768694007e-06, + "loss": 0.0017, + "step": 48069 + }, + { + "epoch": 6.3, + "learning_rate": 2.2845403872222292e-06, + "loss": 0.003, + "step": 48070 + }, + { + "epoch": 6.3, + "learning_rate": 2.284203019274377e-06, + "loss": 0.0017, + "step": 48071 + }, + { + "epoch": 6.3, + "learning_rate": 2.2838656730267905e-06, + "loss": 0.0006, + "step": 48072 + }, + { + "epoch": 6.3, + "learning_rate": 2.283528348480423e-06, + "loss": 0.0022, + "step": 48073 + }, + { + "epoch": 6.3, + "learning_rate": 2.2831910456362217e-06, + "loss": 0.0018, + "step": 48074 + }, + { + "epoch": 6.3, + "learning_rate": 2.2828537644951308e-06, + "loss": 0.0025, + "step": 48075 + }, + { + "epoch": 6.3, + "learning_rate": 2.2825165050581033e-06, + "loss": 0.0022, + "step": 48076 + }, + { + "epoch": 6.3, + "learning_rate": 2.28217926732609e-06, + "loss": 0.0021, + "step": 48077 + }, + { + "epoch": 6.3, + "learning_rate": 2.2818420513000374e-06, + "loss": 0.001, + "step": 48078 + }, + { + "epoch": 6.3, + "learning_rate": 2.2815048569808916e-06, + "loss": 0.0008, + "step": 48079 + }, + { + "epoch": 6.3, + "learning_rate": 2.2811676843696007e-06, + "loss": 0.0022, + "step": 48080 + }, + { + "epoch": 6.3, + "learning_rate": 2.2808305334671166e-06, + "loss": 0.0015, + "step": 48081 + }, + { + "epoch": 6.3, + "learning_rate": 2.2804934042743855e-06, + "loss": 0.0032, + "step": 48082 + }, + { + "epoch": 6.3, + "learning_rate": 2.280156296792356e-06, + "loss": 0.0013, + "step": 48083 + }, + { + "epoch": 6.3, + "learning_rate": 2.279819211021973e-06, + "loss": 0.0021, + "step": 48084 + }, + { + "epoch": 6.3, + "learning_rate": 2.2794821469641905e-06, + "loss": 0.0026, + "step": 48085 + }, + { + "epoch": 6.3, + "learning_rate": 2.27914510461995e-06, + "loss": 0.0024, + "step": 48086 + }, + { + "epoch": 6.3, + "learning_rate": 2.278808083990206e-06, + "loss": 0.0019, + "step": 48087 + }, + { + "epoch": 6.3, + "learning_rate": 2.2784710850758996e-06, + "loss": 0.0018, + "step": 48088 + }, + { + "epoch": 6.3, + "learning_rate": 2.2781341078779852e-06, + "loss": 0.0014, + "step": 48089 + }, + { + "epoch": 6.3, + "learning_rate": 2.2777971523974074e-06, + "loss": 0.0024, + "step": 48090 + }, + { + "epoch": 6.3, + "learning_rate": 2.2774602186351113e-06, + "loss": 0.0026, + "step": 48091 + }, + { + "epoch": 6.3, + "learning_rate": 2.2771233065920496e-06, + "loss": 0.0029, + "step": 48092 + }, + { + "epoch": 6.3, + "learning_rate": 2.276786416269168e-06, + "loss": 0.0022, + "step": 48093 + }, + { + "epoch": 6.3, + "learning_rate": 2.276449547667412e-06, + "loss": 0.0028, + "step": 48094 + }, + { + "epoch": 6.3, + "learning_rate": 2.2761127007877284e-06, + "loss": 0.0024, + "step": 48095 + }, + { + "epoch": 6.3, + "learning_rate": 2.275775875631069e-06, + "loss": 0.0016, + "step": 48096 + }, + { + "epoch": 6.3, + "learning_rate": 2.2754390721983765e-06, + "loss": 0.0022, + "step": 48097 + }, + { + "epoch": 6.3, + "learning_rate": 2.2751022904906016e-06, + "loss": 0.0023, + "step": 48098 + }, + { + "epoch": 6.3, + "learning_rate": 2.274765530508689e-06, + "loss": 0.0029, + "step": 48099 + }, + { + "epoch": 6.3, + "learning_rate": 2.274428792253589e-06, + "loss": 0.0021, + "step": 48100 + }, + { + "epoch": 6.3, + "learning_rate": 2.274092075726246e-06, + "loss": 0.0019, + "step": 48101 + }, + { + "epoch": 6.3, + "learning_rate": 2.273755380927609e-06, + "loss": 0.0015, + "step": 48102 + }, + { + "epoch": 6.3, + "learning_rate": 2.2734187078586203e-06, + "loss": 0.0032, + "step": 48103 + }, + { + "epoch": 6.3, + "learning_rate": 2.2730820565202316e-06, + "loss": 0.0015, + "step": 48104 + }, + { + "epoch": 6.3, + "learning_rate": 2.2727454269133896e-06, + "loss": 0.002, + "step": 48105 + }, + { + "epoch": 6.3, + "learning_rate": 2.272408819039036e-06, + "loss": 0.0017, + "step": 48106 + }, + { + "epoch": 6.3, + "learning_rate": 2.272072232898125e-06, + "loss": 0.0016, + "step": 48107 + }, + { + "epoch": 6.3, + "learning_rate": 2.2717356684915946e-06, + "loss": 0.0022, + "step": 48108 + }, + { + "epoch": 6.3, + "learning_rate": 2.2713991258204006e-06, + "loss": 0.0021, + "step": 48109 + }, + { + "epoch": 6.3, + "learning_rate": 2.2710626048854845e-06, + "loss": 0.0014, + "step": 48110 + }, + { + "epoch": 6.3, + "learning_rate": 2.2707261056877893e-06, + "loss": 0.0018, + "step": 48111 + }, + { + "epoch": 6.3, + "learning_rate": 2.270389628228269e-06, + "loss": 0.0018, + "step": 48112 + }, + { + "epoch": 6.3, + "learning_rate": 2.2700531725078666e-06, + "loss": 0.0015, + "step": 48113 + }, + { + "epoch": 6.3, + "learning_rate": 2.2697167385275253e-06, + "loss": 0.0022, + "step": 48114 + }, + { + "epoch": 6.3, + "learning_rate": 2.2693803262881964e-06, + "loss": 0.0044, + "step": 48115 + }, + { + "epoch": 6.3, + "learning_rate": 2.2690439357908234e-06, + "loss": 0.0013, + "step": 48116 + }, + { + "epoch": 6.3, + "learning_rate": 2.2687075670363534e-06, + "loss": 0.002, + "step": 48117 + }, + { + "epoch": 6.3, + "learning_rate": 2.268371220025728e-06, + "loss": 0.0018, + "step": 48118 + }, + { + "epoch": 6.3, + "learning_rate": 2.2680348947598973e-06, + "loss": 0.0017, + "step": 48119 + }, + { + "epoch": 6.3, + "learning_rate": 2.2676985912398086e-06, + "loss": 0.0023, + "step": 48120 + }, + { + "epoch": 6.3, + "learning_rate": 2.2673623094664077e-06, + "loss": 0.004, + "step": 48121 + }, + { + "epoch": 6.3, + "learning_rate": 2.2670260494406337e-06, + "loss": 0.0018, + "step": 48122 + }, + { + "epoch": 6.3, + "learning_rate": 2.2666898111634416e-06, + "loss": 0.0019, + "step": 48123 + }, + { + "epoch": 6.3, + "learning_rate": 2.266353594635772e-06, + "loss": 0.0018, + "step": 48124 + }, + { + "epoch": 6.3, + "learning_rate": 2.26601739985857e-06, + "loss": 0.002, + "step": 48125 + }, + { + "epoch": 6.3, + "learning_rate": 2.265681226832781e-06, + "loss": 0.0028, + "step": 48126 + }, + { + "epoch": 6.3, + "learning_rate": 2.2653450755593533e-06, + "loss": 0.0028, + "step": 48127 + }, + { + "epoch": 6.3, + "learning_rate": 2.265008946039231e-06, + "loss": 0.0017, + "step": 48128 + }, + { + "epoch": 6.3, + "learning_rate": 2.2646728382733575e-06, + "loss": 0.0025, + "step": 48129 + }, + { + "epoch": 6.3, + "learning_rate": 2.264336752262678e-06, + "loss": 0.0017, + "step": 48130 + }, + { + "epoch": 6.3, + "learning_rate": 2.264000688008143e-06, + "loss": 0.0015, + "step": 48131 + }, + { + "epoch": 6.3, + "learning_rate": 2.2636646455106935e-06, + "loss": 0.0024, + "step": 48132 + }, + { + "epoch": 6.31, + "learning_rate": 2.263328624771276e-06, + "loss": 0.0014, + "step": 48133 + }, + { + "epoch": 6.31, + "learning_rate": 2.262992625790832e-06, + "loss": 0.0027, + "step": 48134 + }, + { + "epoch": 6.31, + "learning_rate": 2.2626566485703118e-06, + "loss": 0.0021, + "step": 48135 + }, + { + "epoch": 6.31, + "learning_rate": 2.262320693110658e-06, + "loss": 0.0022, + "step": 48136 + }, + { + "epoch": 6.31, + "learning_rate": 2.261984759412812e-06, + "loss": 0.0052, + "step": 48137 + }, + { + "epoch": 6.31, + "learning_rate": 2.2616488474777244e-06, + "loss": 0.0012, + "step": 48138 + }, + { + "epoch": 6.31, + "learning_rate": 2.2613129573063377e-06, + "loss": 0.0022, + "step": 48139 + }, + { + "epoch": 6.31, + "learning_rate": 2.2609770888995964e-06, + "loss": 0.0023, + "step": 48140 + }, + { + "epoch": 6.31, + "learning_rate": 2.260641242258439e-06, + "loss": 0.0023, + "step": 48141 + }, + { + "epoch": 6.31, + "learning_rate": 2.2603054173838214e-06, + "loss": 0.0011, + "step": 48142 + }, + { + "epoch": 6.31, + "learning_rate": 2.2599696142766825e-06, + "loss": 0.002, + "step": 48143 + }, + { + "epoch": 6.31, + "learning_rate": 2.2596338329379673e-06, + "loss": 0.0031, + "step": 48144 + }, + { + "epoch": 6.31, + "learning_rate": 2.2592980733686167e-06, + "loss": 0.0018, + "step": 48145 + }, + { + "epoch": 6.31, + "learning_rate": 2.258962335569581e-06, + "loss": 0.0033, + "step": 48146 + }, + { + "epoch": 6.31, + "learning_rate": 2.2586266195418016e-06, + "loss": 0.0017, + "step": 48147 + }, + { + "epoch": 6.31, + "learning_rate": 2.258290925286222e-06, + "loss": 0.0018, + "step": 48148 + }, + { + "epoch": 6.31, + "learning_rate": 2.257955252803784e-06, + "loss": 0.0019, + "step": 48149 + }, + { + "epoch": 6.31, + "learning_rate": 2.257619602095439e-06, + "loss": 0.0025, + "step": 48150 + }, + { + "epoch": 6.31, + "learning_rate": 2.2572839731621223e-06, + "loss": 0.0016, + "step": 48151 + }, + { + "epoch": 6.31, + "learning_rate": 2.256948366004785e-06, + "loss": 0.0019, + "step": 48152 + }, + { + "epoch": 6.31, + "learning_rate": 2.2566127806243653e-06, + "loss": 0.0014, + "step": 48153 + }, + { + "epoch": 6.31, + "learning_rate": 2.256277217021813e-06, + "loss": 0.0015, + "step": 48154 + }, + { + "epoch": 6.31, + "learning_rate": 2.255941675198069e-06, + "loss": 0.0025, + "step": 48155 + }, + { + "epoch": 6.31, + "learning_rate": 2.2556061551540732e-06, + "loss": 0.0017, + "step": 48156 + }, + { + "epoch": 6.31, + "learning_rate": 2.2552706568907755e-06, + "loss": 0.0019, + "step": 48157 + }, + { + "epoch": 6.31, + "learning_rate": 2.254935180409117e-06, + "loss": 0.0022, + "step": 48158 + }, + { + "epoch": 6.31, + "learning_rate": 2.2545997257100403e-06, + "loss": 0.002, + "step": 48159 + }, + { + "epoch": 6.31, + "learning_rate": 2.254264292794487e-06, + "loss": 0.0013, + "step": 48160 + }, + { + "epoch": 6.31, + "learning_rate": 2.2539288816634053e-06, + "loss": 0.0024, + "step": 48161 + }, + { + "epoch": 6.31, + "learning_rate": 2.253593492317734e-06, + "loss": 0.0023, + "step": 48162 + }, + { + "epoch": 6.31, + "learning_rate": 2.253258124758421e-06, + "loss": 0.002, + "step": 48163 + }, + { + "epoch": 6.31, + "learning_rate": 2.252922778986404e-06, + "loss": 0.0016, + "step": 48164 + }, + { + "epoch": 6.31, + "learning_rate": 2.2525874550026315e-06, + "loss": 0.0014, + "step": 48165 + }, + { + "epoch": 6.31, + "learning_rate": 2.2522521528080454e-06, + "loss": 0.0022, + "step": 48166 + }, + { + "epoch": 6.31, + "learning_rate": 2.251916872403587e-06, + "loss": 0.0017, + "step": 48167 + }, + { + "epoch": 6.31, + "learning_rate": 2.251581613790197e-06, + "loss": 0.0019, + "step": 48168 + }, + { + "epoch": 6.31, + "learning_rate": 2.251246376968823e-06, + "loss": 0.0017, + "step": 48169 + }, + { + "epoch": 6.31, + "learning_rate": 2.2509111619404065e-06, + "loss": 0.0018, + "step": 48170 + }, + { + "epoch": 6.31, + "learning_rate": 2.250575968705887e-06, + "loss": 0.0022, + "step": 48171 + }, + { + "epoch": 6.31, + "learning_rate": 2.2502407972662133e-06, + "loss": 0.0043, + "step": 48172 + }, + { + "epoch": 6.31, + "learning_rate": 2.249905647622321e-06, + "loss": 0.0015, + "step": 48173 + }, + { + "epoch": 6.31, + "learning_rate": 2.2495705197751596e-06, + "loss": 0.0012, + "step": 48174 + }, + { + "epoch": 6.31, + "learning_rate": 2.249235413725668e-06, + "loss": 0.0019, + "step": 48175 + }, + { + "epoch": 6.31, + "learning_rate": 2.248900329474787e-06, + "loss": 0.0014, + "step": 48176 + }, + { + "epoch": 6.31, + "learning_rate": 2.2485652670234627e-06, + "loss": 0.0018, + "step": 48177 + }, + { + "epoch": 6.31, + "learning_rate": 2.248230226372636e-06, + "loss": 0.0021, + "step": 48178 + }, + { + "epoch": 6.31, + "learning_rate": 2.2478952075232476e-06, + "loss": 0.0025, + "step": 48179 + }, + { + "epoch": 6.31, + "learning_rate": 2.2475602104762428e-06, + "loss": 0.0016, + "step": 48180 + }, + { + "epoch": 6.31, + "learning_rate": 2.247225235232562e-06, + "loss": 0.0029, + "step": 48181 + }, + { + "epoch": 6.31, + "learning_rate": 2.246890281793147e-06, + "loss": 0.0028, + "step": 48182 + }, + { + "epoch": 6.31, + "learning_rate": 2.2465553501589388e-06, + "loss": 0.0012, + "step": 48183 + }, + { + "epoch": 6.31, + "learning_rate": 2.2462204403308797e-06, + "loss": 0.0021, + "step": 48184 + }, + { + "epoch": 6.31, + "learning_rate": 2.2458855523099154e-06, + "loss": 0.0028, + "step": 48185 + }, + { + "epoch": 6.31, + "learning_rate": 2.2455506860969855e-06, + "loss": 0.0028, + "step": 48186 + }, + { + "epoch": 6.31, + "learning_rate": 2.245215841693029e-06, + "loss": 0.0015, + "step": 48187 + }, + { + "epoch": 6.31, + "learning_rate": 2.244881019098992e-06, + "loss": 0.0018, + "step": 48188 + }, + { + "epoch": 6.31, + "learning_rate": 2.244546218315814e-06, + "loss": 0.003, + "step": 48189 + }, + { + "epoch": 6.31, + "learning_rate": 2.2442114393444368e-06, + "loss": 0.0017, + "step": 48190 + }, + { + "epoch": 6.31, + "learning_rate": 2.2438766821858006e-06, + "loss": 0.0023, + "step": 48191 + }, + { + "epoch": 6.31, + "learning_rate": 2.2435419468408493e-06, + "loss": 0.0011, + "step": 48192 + }, + { + "epoch": 6.31, + "learning_rate": 2.243207233310524e-06, + "loss": 0.0012, + "step": 48193 + }, + { + "epoch": 6.31, + "learning_rate": 2.242872541595762e-06, + "loss": 0.0021, + "step": 48194 + }, + { + "epoch": 6.31, + "learning_rate": 2.242537871697508e-06, + "loss": 0.0015, + "step": 48195 + }, + { + "epoch": 6.31, + "learning_rate": 2.242203223616706e-06, + "loss": 0.0022, + "step": 48196 + }, + { + "epoch": 6.31, + "learning_rate": 2.2418685973542944e-06, + "loss": 0.0027, + "step": 48197 + }, + { + "epoch": 6.31, + "learning_rate": 2.2415339929112144e-06, + "loss": 0.0019, + "step": 48198 + }, + { + "epoch": 6.31, + "learning_rate": 2.2411994102884037e-06, + "loss": 0.0021, + "step": 48199 + }, + { + "epoch": 6.31, + "learning_rate": 2.240864849486809e-06, + "loss": 0.0026, + "step": 48200 + }, + { + "epoch": 6.31, + "learning_rate": 2.2405303105073694e-06, + "loss": 0.0048, + "step": 48201 + }, + { + "epoch": 6.31, + "learning_rate": 2.240195793351022e-06, + "loss": 0.0016, + "step": 48202 + }, + { + "epoch": 6.31, + "learning_rate": 2.239861298018714e-06, + "loss": 0.0014, + "step": 48203 + }, + { + "epoch": 6.31, + "learning_rate": 2.239526824511382e-06, + "loss": 0.0015, + "step": 48204 + }, + { + "epoch": 6.31, + "learning_rate": 2.239192372829968e-06, + "loss": 0.0028, + "step": 48205 + }, + { + "epoch": 6.31, + "learning_rate": 2.2388579429754067e-06, + "loss": 0.0019, + "step": 48206 + }, + { + "epoch": 6.31, + "learning_rate": 2.238523534948649e-06, + "loss": 0.0025, + "step": 48207 + }, + { + "epoch": 6.31, + "learning_rate": 2.2381891487506313e-06, + "loss": 0.002, + "step": 48208 + }, + { + "epoch": 6.32, + "learning_rate": 2.2378547843822927e-06, + "loss": 0.0012, + "step": 48209 + }, + { + "epoch": 6.32, + "learning_rate": 2.2375204418445716e-06, + "loss": 0.0032, + "step": 48210 + }, + { + "epoch": 6.32, + "learning_rate": 2.237186121138414e-06, + "loss": 0.0025, + "step": 48211 + }, + { + "epoch": 6.32, + "learning_rate": 2.236851822264757e-06, + "loss": 0.0018, + "step": 48212 + }, + { + "epoch": 6.32, + "learning_rate": 2.236517545224538e-06, + "loss": 0.0034, + "step": 48213 + }, + { + "epoch": 6.32, + "learning_rate": 2.236183290018702e-06, + "loss": 0.003, + "step": 48214 + }, + { + "epoch": 6.32, + "learning_rate": 2.2358490566481885e-06, + "loss": 0.0009, + "step": 48215 + }, + { + "epoch": 6.32, + "learning_rate": 2.2355148451139352e-06, + "loss": 0.0028, + "step": 48216 + }, + { + "epoch": 6.32, + "learning_rate": 2.2351806554168796e-06, + "loss": 0.0011, + "step": 48217 + }, + { + "epoch": 6.32, + "learning_rate": 2.2348464875579655e-06, + "loss": 0.0035, + "step": 48218 + }, + { + "epoch": 6.32, + "learning_rate": 2.2345123415381354e-06, + "loss": 0.002, + "step": 48219 + }, + { + "epoch": 6.32, + "learning_rate": 2.234178217358325e-06, + "loss": 0.0023, + "step": 48220 + }, + { + "epoch": 6.32, + "learning_rate": 2.233844115019472e-06, + "loss": 0.0027, + "step": 48221 + }, + { + "epoch": 6.32, + "learning_rate": 2.2335100345225224e-06, + "loss": 0.0014, + "step": 48222 + }, + { + "epoch": 6.32, + "learning_rate": 2.2331759758684115e-06, + "loss": 0.0027, + "step": 48223 + }, + { + "epoch": 6.32, + "learning_rate": 2.23284193905808e-06, + "loss": 0.0021, + "step": 48224 + }, + { + "epoch": 6.32, + "learning_rate": 2.232507924092464e-06, + "loss": 0.0018, + "step": 48225 + }, + { + "epoch": 6.32, + "learning_rate": 2.2321739309725087e-06, + "loss": 0.0011, + "step": 48226 + }, + { + "epoch": 6.32, + "learning_rate": 2.23183995969915e-06, + "loss": 0.0013, + "step": 48227 + }, + { + "epoch": 6.32, + "learning_rate": 2.2315060102733254e-06, + "loss": 0.0015, + "step": 48228 + }, + { + "epoch": 6.32, + "learning_rate": 2.231172082695976e-06, + "loss": 0.0029, + "step": 48229 + }, + { + "epoch": 6.32, + "learning_rate": 2.2308381769680455e-06, + "loss": 0.0017, + "step": 48230 + }, + { + "epoch": 6.32, + "learning_rate": 2.2305042930904676e-06, + "loss": 0.0019, + "step": 48231 + }, + { + "epoch": 6.32, + "learning_rate": 2.2301704310641838e-06, + "loss": 0.0046, + "step": 48232 + }, + { + "epoch": 6.32, + "learning_rate": 2.2298365908901276e-06, + "loss": 0.0023, + "step": 48233 + }, + { + "epoch": 6.32, + "learning_rate": 2.2295027725692454e-06, + "loss": 0.0022, + "step": 48234 + }, + { + "epoch": 6.32, + "learning_rate": 2.229168976102474e-06, + "loss": 0.0026, + "step": 48235 + }, + { + "epoch": 6.32, + "learning_rate": 2.2288352014907466e-06, + "loss": 0.0031, + "step": 48236 + }, + { + "epoch": 6.32, + "learning_rate": 2.228501448735011e-06, + "loss": 0.0027, + "step": 48237 + }, + { + "epoch": 6.32, + "learning_rate": 2.228167717836199e-06, + "loss": 0.0019, + "step": 48238 + }, + { + "epoch": 6.32, + "learning_rate": 2.22783400879525e-06, + "loss": 0.0025, + "step": 48239 + }, + { + "epoch": 6.32, + "learning_rate": 2.2275003216131075e-06, + "loss": 0.0028, + "step": 48240 + }, + { + "epoch": 6.32, + "learning_rate": 2.2271666562907023e-06, + "loss": 0.0028, + "step": 48241 + }, + { + "epoch": 6.32, + "learning_rate": 2.2268330128289805e-06, + "loss": 0.0017, + "step": 48242 + }, + { + "epoch": 6.32, + "learning_rate": 2.2264993912288757e-06, + "loss": 0.0017, + "step": 48243 + }, + { + "epoch": 6.32, + "learning_rate": 2.2261657914913258e-06, + "loss": 0.0028, + "step": 48244 + }, + { + "epoch": 6.32, + "learning_rate": 2.2258322136172738e-06, + "loss": 0.0025, + "step": 48245 + }, + { + "epoch": 6.32, + "learning_rate": 2.225498657607653e-06, + "loss": 0.0013, + "step": 48246 + }, + { + "epoch": 6.32, + "learning_rate": 2.225165123463404e-06, + "loss": 0.0017, + "step": 48247 + }, + { + "epoch": 6.32, + "learning_rate": 2.224831611185462e-06, + "loss": 0.0019, + "step": 48248 + }, + { + "epoch": 6.32, + "learning_rate": 2.2244981207747696e-06, + "loss": 0.0019, + "step": 48249 + }, + { + "epoch": 6.32, + "learning_rate": 2.224164652232259e-06, + "loss": 0.0036, + "step": 48250 + }, + { + "epoch": 6.32, + "learning_rate": 2.223831205558874e-06, + "loss": 0.0029, + "step": 48251 + }, + { + "epoch": 6.32, + "learning_rate": 2.223497780755547e-06, + "loss": 0.0029, + "step": 48252 + }, + { + "epoch": 6.32, + "learning_rate": 2.2231643778232216e-06, + "loss": 0.0019, + "step": 48253 + }, + { + "epoch": 6.32, + "learning_rate": 2.2228309967628314e-06, + "loss": 0.0027, + "step": 48254 + }, + { + "epoch": 6.32, + "learning_rate": 2.222497637575315e-06, + "loss": 0.0028, + "step": 48255 + }, + { + "epoch": 6.32, + "learning_rate": 2.2221643002616076e-06, + "loss": 0.0016, + "step": 48256 + }, + { + "epoch": 6.32, + "learning_rate": 2.2218309848226517e-06, + "loss": 0.0017, + "step": 48257 + }, + { + "epoch": 6.32, + "learning_rate": 2.221497691259381e-06, + "loss": 0.0019, + "step": 48258 + }, + { + "epoch": 6.32, + "learning_rate": 2.2211644195727313e-06, + "loss": 0.0026, + "step": 48259 + }, + { + "epoch": 6.32, + "learning_rate": 2.2208311697636454e-06, + "loss": 0.0023, + "step": 48260 + }, + { + "epoch": 6.32, + "learning_rate": 2.2204979418330554e-06, + "loss": 0.0024, + "step": 48261 + }, + { + "epoch": 6.32, + "learning_rate": 2.2201647357819033e-06, + "loss": 0.0031, + "step": 48262 + }, + { + "epoch": 6.32, + "learning_rate": 2.21983155161112e-06, + "loss": 0.0012, + "step": 48263 + }, + { + "epoch": 6.32, + "learning_rate": 2.2194983893216494e-06, + "loss": 0.0029, + "step": 48264 + }, + { + "epoch": 6.32, + "learning_rate": 2.2191652489144254e-06, + "loss": 0.0012, + "step": 48265 + }, + { + "epoch": 6.32, + "learning_rate": 2.218832130390385e-06, + "loss": 0.0024, + "step": 48266 + }, + { + "epoch": 6.32, + "learning_rate": 2.218499033750461e-06, + "loss": 0.0019, + "step": 48267 + }, + { + "epoch": 6.32, + "learning_rate": 2.218165958995597e-06, + "loss": 0.0023, + "step": 48268 + }, + { + "epoch": 6.32, + "learning_rate": 2.217832906126728e-06, + "loss": 0.0017, + "step": 48269 + }, + { + "epoch": 6.32, + "learning_rate": 2.2174998751447884e-06, + "loss": 0.0015, + "step": 48270 + }, + { + "epoch": 6.32, + "learning_rate": 2.217166866050714e-06, + "loss": 0.0021, + "step": 48271 + }, + { + "epoch": 6.32, + "learning_rate": 2.2168338788454435e-06, + "loss": 0.0021, + "step": 48272 + }, + { + "epoch": 6.32, + "learning_rate": 2.2165009135299154e-06, + "loss": 0.0029, + "step": 48273 + }, + { + "epoch": 6.32, + "learning_rate": 2.2161679701050644e-06, + "loss": 0.0025, + "step": 48274 + }, + { + "epoch": 6.32, + "learning_rate": 2.2158350485718227e-06, + "loss": 0.0017, + "step": 48275 + }, + { + "epoch": 6.32, + "learning_rate": 2.2155021489311335e-06, + "loss": 0.0032, + "step": 48276 + }, + { + "epoch": 6.32, + "learning_rate": 2.215169271183931e-06, + "loss": 0.001, + "step": 48277 + }, + { + "epoch": 6.32, + "learning_rate": 2.2148364153311463e-06, + "loss": 0.0023, + "step": 48278 + }, + { + "epoch": 6.32, + "learning_rate": 2.2145035813737235e-06, + "loss": 0.0019, + "step": 48279 + }, + { + "epoch": 6.32, + "learning_rate": 2.2141707693125946e-06, + "loss": 0.0025, + "step": 48280 + }, + { + "epoch": 6.32, + "learning_rate": 2.213837979148695e-06, + "loss": 0.0017, + "step": 48281 + }, + { + "epoch": 6.32, + "learning_rate": 2.213505210882959e-06, + "loss": 0.0028, + "step": 48282 + }, + { + "epoch": 6.32, + "learning_rate": 2.2131724645163253e-06, + "loss": 0.0017, + "step": 48283 + }, + { + "epoch": 6.32, + "learning_rate": 2.2128397400497314e-06, + "loss": 0.0022, + "step": 48284 + }, + { + "epoch": 6.32, + "learning_rate": 2.212507037484112e-06, + "loss": 0.0034, + "step": 48285 + }, + { + "epoch": 6.33, + "learning_rate": 2.212174356820399e-06, + "loss": 0.0031, + "step": 48286 + }, + { + "epoch": 6.33, + "learning_rate": 2.2118416980595324e-06, + "loss": 0.0023, + "step": 48287 + }, + { + "epoch": 6.33, + "learning_rate": 2.211509061202448e-06, + "loss": 0.0015, + "step": 48288 + }, + { + "epoch": 6.33, + "learning_rate": 2.2111764462500784e-06, + "loss": 0.0021, + "step": 48289 + }, + { + "epoch": 6.33, + "learning_rate": 2.2108438532033584e-06, + "loss": 0.0011, + "step": 48290 + }, + { + "epoch": 6.33, + "learning_rate": 2.2105112820632267e-06, + "loss": 0.0032, + "step": 48291 + }, + { + "epoch": 6.33, + "learning_rate": 2.210178732830618e-06, + "loss": 0.0017, + "step": 48292 + }, + { + "epoch": 6.33, + "learning_rate": 2.209846205506464e-06, + "loss": 0.002, + "step": 48293 + }, + { + "epoch": 6.33, + "learning_rate": 2.209513700091703e-06, + "loss": 0.002, + "step": 48294 + }, + { + "epoch": 6.33, + "learning_rate": 2.209181216587273e-06, + "loss": 0.0013, + "step": 48295 + }, + { + "epoch": 6.33, + "learning_rate": 2.208848754994105e-06, + "loss": 0.0023, + "step": 48296 + }, + { + "epoch": 6.33, + "learning_rate": 2.208516315313135e-06, + "loss": 0.0027, + "step": 48297 + }, + { + "epoch": 6.33, + "learning_rate": 2.208183897545295e-06, + "loss": 0.0019, + "step": 48298 + }, + { + "epoch": 6.33, + "learning_rate": 2.207851501691526e-06, + "loss": 0.0019, + "step": 48299 + }, + { + "epoch": 6.33, + "learning_rate": 2.2075191277527595e-06, + "loss": 0.0033, + "step": 48300 + }, + { + "epoch": 6.33, + "learning_rate": 2.2071867757299283e-06, + "loss": 0.004, + "step": 48301 + }, + { + "epoch": 6.33, + "learning_rate": 2.2068544456239717e-06, + "loss": 0.0018, + "step": 48302 + }, + { + "epoch": 6.33, + "learning_rate": 2.2065221374358227e-06, + "loss": 0.0027, + "step": 48303 + }, + { + "epoch": 6.33, + "learning_rate": 2.2061898511664113e-06, + "loss": 0.0034, + "step": 48304 + }, + { + "epoch": 6.33, + "learning_rate": 2.2058575868166787e-06, + "loss": 0.002, + "step": 48305 + }, + { + "epoch": 6.33, + "learning_rate": 2.2055253443875548e-06, + "loss": 0.0011, + "step": 48306 + }, + { + "epoch": 6.33, + "learning_rate": 2.2051931238799772e-06, + "loss": 0.0014, + "step": 48307 + }, + { + "epoch": 6.33, + "learning_rate": 2.2048609252948796e-06, + "loss": 0.0019, + "step": 48308 + }, + { + "epoch": 6.33, + "learning_rate": 2.2045287486331935e-06, + "loss": 0.0016, + "step": 48309 + }, + { + "epoch": 6.33, + "learning_rate": 2.2041965938958566e-06, + "loss": 0.002, + "step": 48310 + }, + { + "epoch": 6.33, + "learning_rate": 2.2038644610838024e-06, + "loss": 0.0034, + "step": 48311 + }, + { + "epoch": 6.33, + "learning_rate": 2.2035323501979634e-06, + "loss": 0.0026, + "step": 48312 + }, + { + "epoch": 6.33, + "learning_rate": 2.2032002612392723e-06, + "loss": 0.0015, + "step": 48313 + }, + { + "epoch": 6.33, + "learning_rate": 2.2028681942086684e-06, + "loss": 0.0018, + "step": 48314 + }, + { + "epoch": 6.33, + "learning_rate": 2.2025361491070796e-06, + "loss": 0.0024, + "step": 48315 + }, + { + "epoch": 6.33, + "learning_rate": 2.2022041259354453e-06, + "loss": 0.0016, + "step": 48316 + }, + { + "epoch": 6.33, + "learning_rate": 2.201872124694694e-06, + "loss": 0.0022, + "step": 48317 + }, + { + "epoch": 6.33, + "learning_rate": 2.2015401453857645e-06, + "loss": 0.002, + "step": 48318 + }, + { + "epoch": 6.33, + "learning_rate": 2.2012081880095882e-06, + "loss": 0.0025, + "step": 48319 + }, + { + "epoch": 6.33, + "learning_rate": 2.200876252567098e-06, + "loss": 0.0017, + "step": 48320 + }, + { + "epoch": 6.33, + "learning_rate": 2.2005443390592253e-06, + "loss": 0.0018, + "step": 48321 + }, + { + "epoch": 6.33, + "learning_rate": 2.20021244748691e-06, + "loss": 0.0035, + "step": 48322 + }, + { + "epoch": 6.33, + "learning_rate": 2.1998805778510814e-06, + "loss": 0.0014, + "step": 48323 + }, + { + "epoch": 6.33, + "learning_rate": 2.19954873015267e-06, + "loss": 0.003, + "step": 48324 + }, + { + "epoch": 6.33, + "learning_rate": 2.199216904392616e-06, + "loss": 0.002, + "step": 48325 + }, + { + "epoch": 6.33, + "learning_rate": 2.198885100571846e-06, + "loss": 0.0017, + "step": 48326 + }, + { + "epoch": 6.33, + "learning_rate": 2.198553318691299e-06, + "loss": 0.0028, + "step": 48327 + }, + { + "epoch": 6.33, + "learning_rate": 2.198221558751902e-06, + "loss": 0.0025, + "step": 48328 + }, + { + "epoch": 6.33, + "learning_rate": 2.197889820754595e-06, + "loss": 0.0031, + "step": 48329 + }, + { + "epoch": 6.33, + "learning_rate": 2.197558104700307e-06, + "loss": 0.0013, + "step": 48330 + }, + { + "epoch": 6.33, + "learning_rate": 2.197226410589971e-06, + "loss": 0.0022, + "step": 48331 + }, + { + "epoch": 6.33, + "learning_rate": 2.1968947384245175e-06, + "loss": 0.0032, + "step": 48332 + }, + { + "epoch": 6.33, + "learning_rate": 2.1965630882048848e-06, + "loss": 0.0033, + "step": 48333 + }, + { + "epoch": 6.33, + "learning_rate": 2.196231459932003e-06, + "loss": 0.0023, + "step": 48334 + }, + { + "epoch": 6.33, + "learning_rate": 2.1958998536068053e-06, + "loss": 0.0042, + "step": 48335 + }, + { + "epoch": 6.33, + "learning_rate": 2.1955682692302204e-06, + "loss": 0.003, + "step": 48336 + }, + { + "epoch": 6.33, + "learning_rate": 2.1952367068031843e-06, + "loss": 0.0017, + "step": 48337 + }, + { + "epoch": 6.33, + "learning_rate": 2.194905166326632e-06, + "loss": 0.0018, + "step": 48338 + }, + { + "epoch": 6.33, + "learning_rate": 2.194573647801492e-06, + "loss": 0.0023, + "step": 48339 + }, + { + "epoch": 6.33, + "learning_rate": 2.194242151228697e-06, + "loss": 0.0016, + "step": 48340 + }, + { + "epoch": 6.33, + "learning_rate": 2.1939106766091823e-06, + "loss": 0.0022, + "step": 48341 + }, + { + "epoch": 6.33, + "learning_rate": 2.1935792239438782e-06, + "loss": 0.0014, + "step": 48342 + }, + { + "epoch": 6.33, + "learning_rate": 2.1932477932337147e-06, + "loss": 0.0018, + "step": 48343 + }, + { + "epoch": 6.33, + "learning_rate": 2.1929163844796277e-06, + "loss": 0.0015, + "step": 48344 + }, + { + "epoch": 6.33, + "learning_rate": 2.1925849976825486e-06, + "loss": 0.0013, + "step": 48345 + }, + { + "epoch": 6.33, + "learning_rate": 2.1922536328434073e-06, + "loss": 0.0015, + "step": 48346 + }, + { + "epoch": 6.33, + "learning_rate": 2.1919222899631353e-06, + "loss": 0.0034, + "step": 48347 + }, + { + "epoch": 6.33, + "learning_rate": 2.191590969042665e-06, + "loss": 0.0013, + "step": 48348 + }, + { + "epoch": 6.33, + "learning_rate": 2.1912596700829327e-06, + "loss": 0.0018, + "step": 48349 + }, + { + "epoch": 6.33, + "learning_rate": 2.190928393084867e-06, + "loss": 0.0025, + "step": 48350 + }, + { + "epoch": 6.33, + "learning_rate": 2.190597138049395e-06, + "loss": 0.002, + "step": 48351 + }, + { + "epoch": 6.33, + "learning_rate": 2.1902659049774565e-06, + "loss": 0.0014, + "step": 48352 + }, + { + "epoch": 6.33, + "learning_rate": 2.1899346938699794e-06, + "loss": 0.003, + "step": 48353 + }, + { + "epoch": 6.33, + "learning_rate": 2.189603504727895e-06, + "loss": 0.0022, + "step": 48354 + }, + { + "epoch": 6.33, + "learning_rate": 2.1892723375521328e-06, + "loss": 0.0013, + "step": 48355 + }, + { + "epoch": 6.33, + "learning_rate": 2.1889411923436276e-06, + "loss": 0.0018, + "step": 48356 + }, + { + "epoch": 6.33, + "learning_rate": 2.1886100691033097e-06, + "loss": 0.0026, + "step": 48357 + }, + { + "epoch": 6.33, + "learning_rate": 2.188278967832108e-06, + "loss": 0.0018, + "step": 48358 + }, + { + "epoch": 6.33, + "learning_rate": 2.1879478885309545e-06, + "loss": 0.0012, + "step": 48359 + }, + { + "epoch": 6.33, + "learning_rate": 2.187616831200785e-06, + "loss": 0.003, + "step": 48360 + }, + { + "epoch": 6.33, + "learning_rate": 2.1872857958425276e-06, + "loss": 0.0021, + "step": 48361 + }, + { + "epoch": 6.34, + "learning_rate": 2.1869547824571124e-06, + "loss": 0.0024, + "step": 48362 + }, + { + "epoch": 6.34, + "learning_rate": 2.186623791045468e-06, + "loss": 0.0013, + "step": 48363 + }, + { + "epoch": 6.34, + "learning_rate": 2.186292821608531e-06, + "loss": 0.0031, + "step": 48364 + }, + { + "epoch": 6.34, + "learning_rate": 2.185961874147229e-06, + "loss": 0.0014, + "step": 48365 + }, + { + "epoch": 6.34, + "learning_rate": 2.1856309486624906e-06, + "loss": 0.0028, + "step": 48366 + }, + { + "epoch": 6.34, + "learning_rate": 2.185300045155252e-06, + "loss": 0.0019, + "step": 48367 + }, + { + "epoch": 6.34, + "learning_rate": 2.1849691636264405e-06, + "loss": 0.002, + "step": 48368 + }, + { + "epoch": 6.34, + "learning_rate": 2.184638304076987e-06, + "loss": 0.0016, + "step": 48369 + }, + { + "epoch": 6.34, + "learning_rate": 2.184307466507819e-06, + "loss": 0.0018, + "step": 48370 + }, + { + "epoch": 6.34, + "learning_rate": 2.1839766509198714e-06, + "loss": 0.0025, + "step": 48371 + }, + { + "epoch": 6.34, + "learning_rate": 2.183645857314074e-06, + "loss": 0.0032, + "step": 48372 + }, + { + "epoch": 6.34, + "learning_rate": 2.1833150856913575e-06, + "loss": 0.0029, + "step": 48373 + }, + { + "epoch": 6.34, + "learning_rate": 2.182984336052649e-06, + "loss": 0.0024, + "step": 48374 + }, + { + "epoch": 6.34, + "learning_rate": 2.1826536083988824e-06, + "loss": 0.0018, + "step": 48375 + }, + { + "epoch": 6.34, + "learning_rate": 2.1823229027309866e-06, + "loss": 0.002, + "step": 48376 + }, + { + "epoch": 6.34, + "learning_rate": 2.181992219049891e-06, + "loss": 0.0016, + "step": 48377 + }, + { + "epoch": 6.34, + "learning_rate": 2.181661557356524e-06, + "loss": 0.0011, + "step": 48378 + }, + { + "epoch": 6.34, + "learning_rate": 2.1813309176518194e-06, + "loss": 0.0018, + "step": 48379 + }, + { + "epoch": 6.34, + "learning_rate": 2.1810002999367065e-06, + "loss": 0.0028, + "step": 48380 + }, + { + "epoch": 6.34, + "learning_rate": 2.1806697042121093e-06, + "loss": 0.0018, + "step": 48381 + }, + { + "epoch": 6.34, + "learning_rate": 2.1803391304789632e-06, + "loss": 0.0035, + "step": 48382 + }, + { + "epoch": 6.34, + "learning_rate": 2.1800085787381997e-06, + "loss": 0.0013, + "step": 48383 + }, + { + "epoch": 6.34, + "learning_rate": 2.1796780489907453e-06, + "loss": 0.0017, + "step": 48384 + }, + { + "epoch": 6.34, + "learning_rate": 2.179347541237531e-06, + "loss": 0.0025, + "step": 48385 + }, + { + "epoch": 6.34, + "learning_rate": 2.179017055479481e-06, + "loss": 0.0021, + "step": 48386 + }, + { + "epoch": 6.34, + "learning_rate": 2.1786865917175328e-06, + "loss": 0.0017, + "step": 48387 + }, + { + "epoch": 6.34, + "learning_rate": 2.1783561499526116e-06, + "loss": 0.0018, + "step": 48388 + }, + { + "epoch": 6.34, + "learning_rate": 2.1780257301856443e-06, + "loss": 0.0021, + "step": 48389 + }, + { + "epoch": 6.34, + "learning_rate": 2.1776953324175664e-06, + "loss": 0.002, + "step": 48390 + }, + { + "epoch": 6.34, + "learning_rate": 2.1773649566493037e-06, + "loss": 0.0016, + "step": 48391 + }, + { + "epoch": 6.34, + "learning_rate": 2.1770346028817825e-06, + "loss": 0.0012, + "step": 48392 + }, + { + "epoch": 6.34, + "learning_rate": 2.176704271115935e-06, + "loss": 0.0018, + "step": 48393 + }, + { + "epoch": 6.34, + "learning_rate": 2.1763739613526926e-06, + "loss": 0.0022, + "step": 48394 + }, + { + "epoch": 6.34, + "learning_rate": 2.1760436735929814e-06, + "loss": 0.0021, + "step": 48395 + }, + { + "epoch": 6.34, + "learning_rate": 2.175713407837732e-06, + "loss": 0.0021, + "step": 48396 + }, + { + "epoch": 6.34, + "learning_rate": 2.1753831640878677e-06, + "loss": 0.0016, + "step": 48397 + }, + { + "epoch": 6.34, + "learning_rate": 2.1750529423443256e-06, + "loss": 0.0022, + "step": 48398 + }, + { + "epoch": 6.34, + "learning_rate": 2.1747227426080297e-06, + "loss": 0.0024, + "step": 48399 + }, + { + "epoch": 6.34, + "learning_rate": 2.1743925648799093e-06, + "loss": 0.0016, + "step": 48400 + }, + { + "epoch": 6.34, + "learning_rate": 2.174062409160891e-06, + "loss": 0.0025, + "step": 48401 + }, + { + "epoch": 6.34, + "learning_rate": 2.1737322754519077e-06, + "loss": 0.0017, + "step": 48402 + }, + { + "epoch": 6.34, + "learning_rate": 2.1734021637538828e-06, + "loss": 0.0032, + "step": 48403 + }, + { + "epoch": 6.34, + "learning_rate": 2.17307207406775e-06, + "loss": 0.0013, + "step": 48404 + }, + { + "epoch": 6.34, + "learning_rate": 2.172742006394433e-06, + "loss": 0.0014, + "step": 48405 + }, + { + "epoch": 6.34, + "learning_rate": 2.172411960734865e-06, + "loss": 0.0014, + "step": 48406 + }, + { + "epoch": 6.34, + "learning_rate": 2.1720819370899714e-06, + "loss": 0.0035, + "step": 48407 + }, + { + "epoch": 6.34, + "learning_rate": 2.1717519354606774e-06, + "loss": 0.002, + "step": 48408 + }, + { + "epoch": 6.34, + "learning_rate": 2.171421955847918e-06, + "loss": 0.0023, + "step": 48409 + }, + { + "epoch": 6.34, + "learning_rate": 2.1710919982526155e-06, + "loss": 0.0018, + "step": 48410 + }, + { + "epoch": 6.34, + "learning_rate": 2.170762062675702e-06, + "loss": 0.0016, + "step": 48411 + }, + { + "epoch": 6.34, + "learning_rate": 2.170432149118099e-06, + "loss": 0.0025, + "step": 48412 + }, + { + "epoch": 6.34, + "learning_rate": 2.170102257580743e-06, + "loss": 0.0016, + "step": 48413 + }, + { + "epoch": 6.34, + "learning_rate": 2.1697723880645526e-06, + "loss": 0.0014, + "step": 48414 + }, + { + "epoch": 6.34, + "learning_rate": 2.1694425405704645e-06, + "loss": 0.0013, + "step": 48415 + }, + { + "epoch": 6.34, + "learning_rate": 2.1691127150993995e-06, + "loss": 0.0019, + "step": 48416 + }, + { + "epoch": 6.34, + "learning_rate": 2.1687829116522897e-06, + "loss": 0.0018, + "step": 48417 + }, + { + "epoch": 6.34, + "learning_rate": 2.1684531302300614e-06, + "loss": 0.0015, + "step": 48418 + }, + { + "epoch": 6.34, + "learning_rate": 2.1681233708336425e-06, + "loss": 0.0015, + "step": 48419 + }, + { + "epoch": 6.34, + "learning_rate": 2.167793633463956e-06, + "loss": 0.0021, + "step": 48420 + }, + { + "epoch": 6.34, + "learning_rate": 2.167463918121936e-06, + "loss": 0.0017, + "step": 48421 + }, + { + "epoch": 6.34, + "learning_rate": 2.1671342248085057e-06, + "loss": 0.0012, + "step": 48422 + }, + { + "epoch": 6.34, + "learning_rate": 2.1668045535245917e-06, + "loss": 0.0023, + "step": 48423 + }, + { + "epoch": 6.34, + "learning_rate": 2.166474904271122e-06, + "loss": 0.0018, + "step": 48424 + }, + { + "epoch": 6.34, + "learning_rate": 2.1661452770490277e-06, + "loss": 0.0023, + "step": 48425 + }, + { + "epoch": 6.34, + "learning_rate": 2.1658156718592318e-06, + "loss": 0.0021, + "step": 48426 + }, + { + "epoch": 6.34, + "learning_rate": 2.165486088702663e-06, + "loss": 0.0036, + "step": 48427 + }, + { + "epoch": 6.34, + "learning_rate": 2.1651565275802442e-06, + "loss": 0.0012, + "step": 48428 + }, + { + "epoch": 6.34, + "learning_rate": 2.1648269884929085e-06, + "loss": 0.0018, + "step": 48429 + }, + { + "epoch": 6.34, + "learning_rate": 2.164497471441579e-06, + "loss": 0.0016, + "step": 48430 + }, + { + "epoch": 6.34, + "learning_rate": 2.1641679764271806e-06, + "loss": 0.002, + "step": 48431 + }, + { + "epoch": 6.34, + "learning_rate": 2.163838503450646e-06, + "loss": 0.0008, + "step": 48432 + }, + { + "epoch": 6.34, + "learning_rate": 2.163509052512898e-06, + "loss": 0.0018, + "step": 48433 + }, + { + "epoch": 6.34, + "learning_rate": 2.1631796236148627e-06, + "loss": 0.0027, + "step": 48434 + }, + { + "epoch": 6.34, + "learning_rate": 2.162850216757465e-06, + "loss": 0.0019, + "step": 48435 + }, + { + "epoch": 6.34, + "learning_rate": 2.1625208319416337e-06, + "loss": 0.0042, + "step": 48436 + }, + { + "epoch": 6.34, + "learning_rate": 2.1621914691682975e-06, + "loss": 0.0009, + "step": 48437 + }, + { + "epoch": 6.35, + "learning_rate": 2.1618621284383813e-06, + "loss": 0.0027, + "step": 48438 + }, + { + "epoch": 6.35, + "learning_rate": 2.1615328097528076e-06, + "loss": 0.0033, + "step": 48439 + }, + { + "epoch": 6.35, + "learning_rate": 2.1612035131125077e-06, + "loss": 0.0018, + "step": 48440 + }, + { + "epoch": 6.35, + "learning_rate": 2.1608742385184056e-06, + "loss": 0.0017, + "step": 48441 + }, + { + "epoch": 6.35, + "learning_rate": 2.1605449859714277e-06, + "loss": 0.0021, + "step": 48442 + }, + { + "epoch": 6.35, + "learning_rate": 2.1602157554724967e-06, + "loss": 0.0049, + "step": 48443 + }, + { + "epoch": 6.35, + "learning_rate": 2.159886547022544e-06, + "loss": 0.001, + "step": 48444 + }, + { + "epoch": 6.35, + "learning_rate": 2.159557360622493e-06, + "loss": 0.0029, + "step": 48445 + }, + { + "epoch": 6.35, + "learning_rate": 2.1592281962732666e-06, + "loss": 0.0011, + "step": 48446 + }, + { + "epoch": 6.35, + "learning_rate": 2.1588990539757926e-06, + "loss": 0.0015, + "step": 48447 + }, + { + "epoch": 6.35, + "learning_rate": 2.1585699337310005e-06, + "loss": 0.0021, + "step": 48448 + }, + { + "epoch": 6.35, + "learning_rate": 2.158240835539813e-06, + "loss": 0.0021, + "step": 48449 + }, + { + "epoch": 6.35, + "learning_rate": 2.157911759403156e-06, + "loss": 0.002, + "step": 48450 + }, + { + "epoch": 6.35, + "learning_rate": 2.1575827053219523e-06, + "loss": 0.0016, + "step": 48451 + }, + { + "epoch": 6.35, + "learning_rate": 2.1572536732971316e-06, + "loss": 0.0018, + "step": 48452 + }, + { + "epoch": 6.35, + "learning_rate": 2.156924663329617e-06, + "loss": 0.0012, + "step": 48453 + }, + { + "epoch": 6.35, + "learning_rate": 2.156595675420332e-06, + "loss": 0.0015, + "step": 48454 + }, + { + "epoch": 6.35, + "learning_rate": 2.1562667095702063e-06, + "loss": 0.003, + "step": 48455 + }, + { + "epoch": 6.35, + "learning_rate": 2.1559377657801627e-06, + "loss": 0.0017, + "step": 48456 + }, + { + "epoch": 6.35, + "learning_rate": 2.1556088440511236e-06, + "loss": 0.0017, + "step": 48457 + }, + { + "epoch": 6.35, + "learning_rate": 2.155279944384017e-06, + "loss": 0.0017, + "step": 48458 + }, + { + "epoch": 6.35, + "learning_rate": 2.154951066779771e-06, + "loss": 0.0036, + "step": 48459 + }, + { + "epoch": 6.35, + "learning_rate": 2.1546222112393066e-06, + "loss": 0.0016, + "step": 48460 + }, + { + "epoch": 6.35, + "learning_rate": 2.15429337776355e-06, + "loss": 0.0019, + "step": 48461 + }, + { + "epoch": 6.35, + "learning_rate": 2.1539645663534224e-06, + "loss": 0.005, + "step": 48462 + }, + { + "epoch": 6.35, + "learning_rate": 2.153635777009855e-06, + "loss": 0.002, + "step": 48463 + }, + { + "epoch": 6.35, + "learning_rate": 2.1533070097337682e-06, + "loss": 0.0021, + "step": 48464 + }, + { + "epoch": 6.35, + "learning_rate": 2.1529782645260887e-06, + "loss": 0.0016, + "step": 48465 + }, + { + "epoch": 6.35, + "learning_rate": 2.152649541387737e-06, + "loss": 0.0023, + "step": 48466 + }, + { + "epoch": 6.35, + "learning_rate": 2.1523208403196426e-06, + "loss": 0.0022, + "step": 48467 + }, + { + "epoch": 6.35, + "learning_rate": 2.1519921613227248e-06, + "loss": 0.0032, + "step": 48468 + }, + { + "epoch": 6.35, + "learning_rate": 2.151663504397915e-06, + "loss": 0.001, + "step": 48469 + }, + { + "epoch": 6.35, + "learning_rate": 2.1513348695461302e-06, + "loss": 0.0025, + "step": 48470 + }, + { + "epoch": 6.35, + "learning_rate": 2.151006256768301e-06, + "loss": 0.0026, + "step": 48471 + }, + { + "epoch": 6.35, + "learning_rate": 2.1506776660653495e-06, + "loss": 0.0026, + "step": 48472 + }, + { + "epoch": 6.35, + "learning_rate": 2.150349097438195e-06, + "loss": 0.0023, + "step": 48473 + }, + { + "epoch": 6.35, + "learning_rate": 2.1500205508877693e-06, + "loss": 0.0027, + "step": 48474 + }, + { + "epoch": 6.35, + "learning_rate": 2.1496920264149924e-06, + "loss": 0.0021, + "step": 48475 + }, + { + "epoch": 6.35, + "learning_rate": 2.1493635240207887e-06, + "loss": 0.0022, + "step": 48476 + }, + { + "epoch": 6.35, + "learning_rate": 2.1490350437060793e-06, + "loss": 0.0019, + "step": 48477 + }, + { + "epoch": 6.35, + "learning_rate": 2.1487065854717935e-06, + "loss": 0.0025, + "step": 48478 + }, + { + "epoch": 6.35, + "learning_rate": 2.1483781493188493e-06, + "loss": 0.0022, + "step": 48479 + }, + { + "epoch": 6.35, + "learning_rate": 2.1480497352481765e-06, + "loss": 0.0008, + "step": 48480 + }, + { + "epoch": 6.35, + "learning_rate": 2.1477213432606924e-06, + "loss": 0.0016, + "step": 48481 + }, + { + "epoch": 6.35, + "learning_rate": 2.1473929733573274e-06, + "loss": 0.0016, + "step": 48482 + }, + { + "epoch": 6.35, + "learning_rate": 2.1470646255390014e-06, + "loss": 0.0018, + "step": 48483 + }, + { + "epoch": 6.35, + "learning_rate": 2.1467362998066377e-06, + "loss": 0.0025, + "step": 48484 + }, + { + "epoch": 6.35, + "learning_rate": 2.1464079961611574e-06, + "loss": 0.0021, + "step": 48485 + }, + { + "epoch": 6.35, + "learning_rate": 2.1460797146034885e-06, + "loss": 0.001, + "step": 48486 + }, + { + "epoch": 6.35, + "learning_rate": 2.1457514551345525e-06, + "loss": 0.0023, + "step": 48487 + }, + { + "epoch": 6.35, + "learning_rate": 2.1454232177552693e-06, + "loss": 0.0015, + "step": 48488 + }, + { + "epoch": 6.35, + "learning_rate": 2.1450950024665685e-06, + "loss": 0.0019, + "step": 48489 + }, + { + "epoch": 6.35, + "learning_rate": 2.144766809269366e-06, + "loss": 0.0022, + "step": 48490 + }, + { + "epoch": 6.35, + "learning_rate": 2.144438638164592e-06, + "loss": 0.0024, + "step": 48491 + }, + { + "epoch": 6.35, + "learning_rate": 2.144110489153166e-06, + "loss": 0.0021, + "step": 48492 + }, + { + "epoch": 6.35, + "learning_rate": 2.143782362236008e-06, + "loss": 0.0015, + "step": 48493 + }, + { + "epoch": 6.35, + "learning_rate": 2.143454257414046e-06, + "loss": 0.0017, + "step": 48494 + }, + { + "epoch": 6.35, + "learning_rate": 2.1431261746882013e-06, + "loss": 0.0015, + "step": 48495 + }, + { + "epoch": 6.35, + "learning_rate": 2.1427981140593933e-06, + "loss": 0.0021, + "step": 48496 + }, + { + "epoch": 6.35, + "learning_rate": 2.1424700755285488e-06, + "loss": 0.004, + "step": 48497 + }, + { + "epoch": 6.35, + "learning_rate": 2.1421420590965893e-06, + "loss": 0.0024, + "step": 48498 + }, + { + "epoch": 6.35, + "learning_rate": 2.141814064764437e-06, + "loss": 0.0018, + "step": 48499 + }, + { + "epoch": 6.35, + "learning_rate": 2.1414860925330117e-06, + "loss": 0.002, + "step": 48500 + }, + { + "epoch": 6.35, + "learning_rate": 2.141158142403238e-06, + "loss": 0.0021, + "step": 48501 + }, + { + "epoch": 6.35, + "learning_rate": 2.1408302143760407e-06, + "loss": 0.0032, + "step": 48502 + }, + { + "epoch": 6.35, + "learning_rate": 2.14050230845234e-06, + "loss": 0.002, + "step": 48503 + }, + { + "epoch": 6.35, + "learning_rate": 2.1401744246330557e-06, + "loss": 0.0019, + "step": 48504 + }, + { + "epoch": 6.35, + "learning_rate": 2.1398465629191146e-06, + "loss": 0.0021, + "step": 48505 + }, + { + "epoch": 6.35, + "learning_rate": 2.139518723311437e-06, + "loss": 0.0026, + "step": 48506 + }, + { + "epoch": 6.35, + "learning_rate": 2.139190905810944e-06, + "loss": 0.0029, + "step": 48507 + }, + { + "epoch": 6.35, + "learning_rate": 2.138863110418555e-06, + "loss": 0.0012, + "step": 48508 + }, + { + "epoch": 6.35, + "learning_rate": 2.1385353371351993e-06, + "loss": 0.0027, + "step": 48509 + }, + { + "epoch": 6.35, + "learning_rate": 2.1382075859617924e-06, + "loss": 0.0022, + "step": 48510 + }, + { + "epoch": 6.35, + "learning_rate": 2.137879856899255e-06, + "loss": 0.002, + "step": 48511 + }, + { + "epoch": 6.35, + "learning_rate": 2.137552149948513e-06, + "loss": 0.0028, + "step": 48512 + }, + { + "epoch": 6.35, + "learning_rate": 2.13722446511049e-06, + "loss": 0.003, + "step": 48513 + }, + { + "epoch": 6.35, + "learning_rate": 2.1368968023861037e-06, + "loss": 0.0035, + "step": 48514 + }, + { + "epoch": 6.36, + "learning_rate": 2.1365691617762763e-06, + "loss": 0.0025, + "step": 48515 + }, + { + "epoch": 6.36, + "learning_rate": 2.1362415432819273e-06, + "loss": 0.0029, + "step": 48516 + }, + { + "epoch": 6.36, + "learning_rate": 2.1359139469039825e-06, + "loss": 0.0024, + "step": 48517 + }, + { + "epoch": 6.36, + "learning_rate": 2.13558637264336e-06, + "loss": 0.002, + "step": 48518 + }, + { + "epoch": 6.36, + "learning_rate": 2.1352588205009804e-06, + "loss": 0.0031, + "step": 48519 + }, + { + "epoch": 6.36, + "learning_rate": 2.1349312904777697e-06, + "loss": 0.0018, + "step": 48520 + }, + { + "epoch": 6.36, + "learning_rate": 2.1346037825746448e-06, + "loss": 0.0028, + "step": 48521 + }, + { + "epoch": 6.36, + "learning_rate": 2.134276296792528e-06, + "loss": 0.002, + "step": 48522 + }, + { + "epoch": 6.36, + "learning_rate": 2.1339488331323354e-06, + "loss": 0.0021, + "step": 48523 + }, + { + "epoch": 6.36, + "learning_rate": 2.133621391594998e-06, + "loss": 0.0018, + "step": 48524 + }, + { + "epoch": 6.36, + "learning_rate": 2.1332939721814316e-06, + "loss": 0.0025, + "step": 48525 + }, + { + "epoch": 6.36, + "learning_rate": 2.1329665748925566e-06, + "loss": 0.002, + "step": 48526 + }, + { + "epoch": 6.36, + "learning_rate": 2.1326391997292917e-06, + "loss": 0.0023, + "step": 48527 + }, + { + "epoch": 6.36, + "learning_rate": 2.132311846692563e-06, + "loss": 0.0026, + "step": 48528 + }, + { + "epoch": 6.36, + "learning_rate": 2.1319845157832886e-06, + "loss": 0.0024, + "step": 48529 + }, + { + "epoch": 6.36, + "learning_rate": 2.1316572070023854e-06, + "loss": 0.0022, + "step": 48530 + }, + { + "epoch": 6.36, + "learning_rate": 2.13132992035078e-06, + "loss": 0.0016, + "step": 48531 + }, + { + "epoch": 6.36, + "learning_rate": 2.1310026558293905e-06, + "loss": 0.0018, + "step": 48532 + }, + { + "epoch": 6.36, + "learning_rate": 2.1306754134391373e-06, + "loss": 0.0017, + "step": 48533 + }, + { + "epoch": 6.36, + "learning_rate": 2.130348193180937e-06, + "loss": 0.002, + "step": 48534 + }, + { + "epoch": 6.36, + "learning_rate": 2.130020995055714e-06, + "loss": 0.0011, + "step": 48535 + }, + { + "epoch": 6.36, + "learning_rate": 2.129693819064389e-06, + "loss": 0.002, + "step": 48536 + }, + { + "epoch": 6.36, + "learning_rate": 2.1293666652078826e-06, + "loss": 0.0026, + "step": 48537 + }, + { + "epoch": 6.36, + "learning_rate": 2.1290395334871104e-06, + "loss": 0.0012, + "step": 48538 + }, + { + "epoch": 6.36, + "learning_rate": 2.1287124239029976e-06, + "loss": 0.0028, + "step": 48539 + }, + { + "epoch": 6.36, + "learning_rate": 2.1283853364564622e-06, + "loss": 0.0012, + "step": 48540 + }, + { + "epoch": 6.36, + "learning_rate": 2.1280582711484242e-06, + "loss": 0.0021, + "step": 48541 + }, + { + "epoch": 6.36, + "learning_rate": 2.1277312279798e-06, + "loss": 0.0025, + "step": 48542 + }, + { + "epoch": 6.36, + "learning_rate": 2.1274042069515144e-06, + "loss": 0.002, + "step": 48543 + }, + { + "epoch": 6.36, + "learning_rate": 2.1270772080644866e-06, + "loss": 0.0018, + "step": 48544 + }, + { + "epoch": 6.36, + "learning_rate": 2.126750231319632e-06, + "loss": 0.0029, + "step": 48545 + }, + { + "epoch": 6.36, + "learning_rate": 2.1264232767178717e-06, + "loss": 0.0014, + "step": 48546 + }, + { + "epoch": 6.36, + "learning_rate": 2.126096344260131e-06, + "loss": 0.0026, + "step": 48547 + }, + { + "epoch": 6.36, + "learning_rate": 2.1257694339473246e-06, + "loss": 0.0033, + "step": 48548 + }, + { + "epoch": 6.36, + "learning_rate": 2.125442545780372e-06, + "loss": 0.0019, + "step": 48549 + }, + { + "epoch": 6.36, + "learning_rate": 2.12511567976019e-06, + "loss": 0.0018, + "step": 48550 + }, + { + "epoch": 6.36, + "learning_rate": 2.1247888358877043e-06, + "loss": 0.0018, + "step": 48551 + }, + { + "epoch": 6.36, + "learning_rate": 2.1244620141638295e-06, + "loss": 0.0028, + "step": 48552 + }, + { + "epoch": 6.36, + "learning_rate": 2.1241352145894843e-06, + "loss": 0.0013, + "step": 48553 + }, + { + "epoch": 6.36, + "learning_rate": 2.123808437165591e-06, + "loss": 0.0016, + "step": 48554 + }, + { + "epoch": 6.36, + "learning_rate": 2.123481681893066e-06, + "loss": 0.0016, + "step": 48555 + }, + { + "epoch": 6.36, + "learning_rate": 2.123154948772831e-06, + "loss": 0.0015, + "step": 48556 + }, + { + "epoch": 6.36, + "learning_rate": 2.122828237805803e-06, + "loss": 0.0016, + "step": 48557 + }, + { + "epoch": 6.36, + "learning_rate": 2.122501548992898e-06, + "loss": 0.0046, + "step": 48558 + }, + { + "epoch": 6.36, + "learning_rate": 2.1221748823350417e-06, + "loss": 0.0017, + "step": 48559 + }, + { + "epoch": 6.36, + "learning_rate": 2.1218482378331485e-06, + "loss": 0.0018, + "step": 48560 + }, + { + "epoch": 6.36, + "learning_rate": 2.1215216154881346e-06, + "loss": 0.0021, + "step": 48561 + }, + { + "epoch": 6.36, + "learning_rate": 2.121195015300924e-06, + "loss": 0.0024, + "step": 48562 + }, + { + "epoch": 6.36, + "learning_rate": 2.1208684372724342e-06, + "loss": 0.0021, + "step": 48563 + }, + { + "epoch": 6.36, + "learning_rate": 2.1205418814035816e-06, + "loss": 0.0027, + "step": 48564 + }, + { + "epoch": 6.36, + "learning_rate": 2.1202153476952827e-06, + "loss": 0.0012, + "step": 48565 + }, + { + "epoch": 6.36, + "learning_rate": 2.119888836148458e-06, + "loss": 0.0025, + "step": 48566 + }, + { + "epoch": 6.36, + "learning_rate": 2.11956234676403e-06, + "loss": 0.0011, + "step": 48567 + }, + { + "epoch": 6.36, + "learning_rate": 2.1192358795429123e-06, + "loss": 0.0021, + "step": 48568 + }, + { + "epoch": 6.36, + "learning_rate": 2.118909434486022e-06, + "loss": 0.0023, + "step": 48569 + }, + { + "epoch": 6.36, + "learning_rate": 2.118583011594282e-06, + "loss": 0.0037, + "step": 48570 + }, + { + "epoch": 6.36, + "learning_rate": 2.1182566108686065e-06, + "loss": 0.0027, + "step": 48571 + }, + { + "epoch": 6.36, + "learning_rate": 2.117930232309916e-06, + "loss": 0.0028, + "step": 48572 + }, + { + "epoch": 6.36, + "learning_rate": 2.117603875919123e-06, + "loss": 0.0021, + "step": 48573 + }, + { + "epoch": 6.36, + "learning_rate": 2.1172775416971524e-06, + "loss": 0.0019, + "step": 48574 + }, + { + "epoch": 6.36, + "learning_rate": 2.11695122964492e-06, + "loss": 0.0035, + "step": 48575 + }, + { + "epoch": 6.36, + "learning_rate": 2.1166249397633386e-06, + "loss": 0.0019, + "step": 48576 + }, + { + "epoch": 6.36, + "learning_rate": 2.1162986720533294e-06, + "loss": 0.0021, + "step": 48577 + }, + { + "epoch": 6.36, + "learning_rate": 2.1159724265158146e-06, + "loss": 0.0008, + "step": 48578 + }, + { + "epoch": 6.36, + "learning_rate": 2.1156462031517067e-06, + "loss": 0.0022, + "step": 48579 + }, + { + "epoch": 6.36, + "learning_rate": 2.11532000196192e-06, + "loss": 0.0013, + "step": 48580 + }, + { + "epoch": 6.36, + "learning_rate": 2.114993822947381e-06, + "loss": 0.002, + "step": 48581 + }, + { + "epoch": 6.36, + "learning_rate": 2.1146676661090005e-06, + "loss": 0.0024, + "step": 48582 + }, + { + "epoch": 6.36, + "learning_rate": 2.114341531447698e-06, + "loss": 0.0013, + "step": 48583 + }, + { + "epoch": 6.36, + "learning_rate": 2.114015418964386e-06, + "loss": 0.0022, + "step": 48584 + }, + { + "epoch": 6.36, + "learning_rate": 2.113689328659989e-06, + "loss": 0.0019, + "step": 48585 + }, + { + "epoch": 6.36, + "learning_rate": 2.1133632605354215e-06, + "loss": 0.0015, + "step": 48586 + }, + { + "epoch": 6.36, + "learning_rate": 2.1130372145915988e-06, + "loss": 0.0023, + "step": 48587 + }, + { + "epoch": 6.36, + "learning_rate": 2.112711190829434e-06, + "loss": 0.0017, + "step": 48588 + }, + { + "epoch": 6.36, + "learning_rate": 2.1123851892498547e-06, + "loss": 0.0014, + "step": 48589 + }, + { + "epoch": 6.36, + "learning_rate": 2.1120592098537717e-06, + "loss": 0.0021, + "step": 48590 + }, + { + "epoch": 6.37, + "learning_rate": 2.111733252642103e-06, + "loss": 0.0021, + "step": 48591 + }, + { + "epoch": 6.37, + "learning_rate": 2.1114073176157603e-06, + "loss": 0.0023, + "step": 48592 + }, + { + "epoch": 6.37, + "learning_rate": 2.1110814047756678e-06, + "loss": 0.002, + "step": 48593 + }, + { + "epoch": 6.37, + "learning_rate": 2.110755514122739e-06, + "loss": 0.0022, + "step": 48594 + }, + { + "epoch": 6.37, + "learning_rate": 2.1104296456578867e-06, + "loss": 0.0018, + "step": 48595 + }, + { + "epoch": 6.37, + "learning_rate": 2.1101037993820337e-06, + "loss": 0.0024, + "step": 48596 + }, + { + "epoch": 6.37, + "learning_rate": 2.1097779752960947e-06, + "loss": 0.0018, + "step": 48597 + }, + { + "epoch": 6.37, + "learning_rate": 2.1094521734009845e-06, + "loss": 0.003, + "step": 48598 + }, + { + "epoch": 6.37, + "learning_rate": 2.109126393697617e-06, + "loss": 0.0017, + "step": 48599 + }, + { + "epoch": 6.37, + "learning_rate": 2.1088006361869108e-06, + "loss": 0.0018, + "step": 48600 + }, + { + "epoch": 6.37, + "learning_rate": 2.1084749008697858e-06, + "loss": 0.0018, + "step": 48601 + }, + { + "epoch": 6.37, + "learning_rate": 2.108149187747155e-06, + "loss": 0.0024, + "step": 48602 + }, + { + "epoch": 6.37, + "learning_rate": 2.1078234968199305e-06, + "loss": 0.0021, + "step": 48603 + }, + { + "epoch": 6.37, + "learning_rate": 2.1074978280890358e-06, + "loss": 0.0022, + "step": 48604 + }, + { + "epoch": 6.37, + "learning_rate": 2.1071721815553835e-06, + "loss": 0.001, + "step": 48605 + }, + { + "epoch": 6.37, + "learning_rate": 2.106846557219888e-06, + "loss": 0.0033, + "step": 48606 + }, + { + "epoch": 6.37, + "learning_rate": 2.1065209550834653e-06, + "loss": 0.0014, + "step": 48607 + }, + { + "epoch": 6.37, + "learning_rate": 2.106195375147033e-06, + "loss": 0.002, + "step": 48608 + }, + { + "epoch": 6.37, + "learning_rate": 2.1058698174115065e-06, + "loss": 0.0019, + "step": 48609 + }, + { + "epoch": 6.37, + "learning_rate": 2.105544281877798e-06, + "loss": 0.0012, + "step": 48610 + }, + { + "epoch": 6.37, + "learning_rate": 2.1052187685468263e-06, + "loss": 0.0024, + "step": 48611 + }, + { + "epoch": 6.37, + "learning_rate": 2.1048932774195088e-06, + "loss": 0.0032, + "step": 48612 + }, + { + "epoch": 6.37, + "learning_rate": 2.104567808496758e-06, + "loss": 0.0017, + "step": 48613 + }, + { + "epoch": 6.37, + "learning_rate": 2.1042423617794905e-06, + "loss": 0.0019, + "step": 48614 + }, + { + "epoch": 6.37, + "learning_rate": 2.1039169372686172e-06, + "loss": 0.0013, + "step": 48615 + }, + { + "epoch": 6.37, + "learning_rate": 2.1035915349650615e-06, + "loss": 0.002, + "step": 48616 + }, + { + "epoch": 6.37, + "learning_rate": 2.103266154869733e-06, + "loss": 0.0022, + "step": 48617 + }, + { + "epoch": 6.37, + "learning_rate": 2.1029407969835446e-06, + "loss": 0.0015, + "step": 48618 + }, + { + "epoch": 6.37, + "learning_rate": 2.102615461307418e-06, + "loss": 0.0016, + "step": 48619 + }, + { + "epoch": 6.37, + "learning_rate": 2.102290147842265e-06, + "loss": 0.0017, + "step": 48620 + }, + { + "epoch": 6.37, + "learning_rate": 2.101964856588997e-06, + "loss": 0.0027, + "step": 48621 + }, + { + "epoch": 6.37, + "learning_rate": 2.101639587548536e-06, + "loss": 0.0013, + "step": 48622 + }, + { + "epoch": 6.37, + "learning_rate": 2.1013143407217894e-06, + "loss": 0.0027, + "step": 48623 + }, + { + "epoch": 6.37, + "learning_rate": 2.1009891161096795e-06, + "loss": 0.0017, + "step": 48624 + }, + { + "epoch": 6.37, + "learning_rate": 2.1006639137131157e-06, + "loss": 0.0016, + "step": 48625 + }, + { + "epoch": 6.37, + "learning_rate": 2.100338733533012e-06, + "loss": 0.0026, + "step": 48626 + }, + { + "epoch": 6.37, + "learning_rate": 2.100013575570288e-06, + "loss": 0.0014, + "step": 48627 + }, + { + "epoch": 6.37, + "learning_rate": 2.0996884398258542e-06, + "loss": 0.0009, + "step": 48628 + }, + { + "epoch": 6.37, + "learning_rate": 2.0993633263006263e-06, + "loss": 0.0029, + "step": 48629 + }, + { + "epoch": 6.37, + "learning_rate": 2.099038234995515e-06, + "loss": 0.0013, + "step": 48630 + }, + { + "epoch": 6.37, + "learning_rate": 2.0987131659114415e-06, + "loss": 0.0024, + "step": 48631 + }, + { + "epoch": 6.37, + "learning_rate": 2.0983881190493137e-06, + "loss": 0.0011, + "step": 48632 + }, + { + "epoch": 6.37, + "learning_rate": 2.0980630944100512e-06, + "loss": 0.0015, + "step": 48633 + }, + { + "epoch": 6.37, + "learning_rate": 2.0977380919945624e-06, + "loss": 0.0016, + "step": 48634 + }, + { + "epoch": 6.37, + "learning_rate": 2.0974131118037676e-06, + "loss": 0.0013, + "step": 48635 + }, + { + "epoch": 6.37, + "learning_rate": 2.097088153838577e-06, + "loss": 0.0014, + "step": 48636 + }, + { + "epoch": 6.37, + "learning_rate": 2.096763218099905e-06, + "loss": 0.002, + "step": 48637 + }, + { + "epoch": 6.37, + "learning_rate": 2.0964383045886637e-06, + "loss": 0.0015, + "step": 48638 + }, + { + "epoch": 6.37, + "learning_rate": 2.0961134133057713e-06, + "loss": 0.0013, + "step": 48639 + }, + { + "epoch": 6.37, + "learning_rate": 2.0957885442521387e-06, + "loss": 0.0011, + "step": 48640 + }, + { + "epoch": 6.37, + "learning_rate": 2.095463697428677e-06, + "loss": 0.001, + "step": 48641 + }, + { + "epoch": 6.37, + "learning_rate": 2.0951388728363053e-06, + "loss": 0.001, + "step": 48642 + }, + { + "epoch": 6.37, + "learning_rate": 2.0948140704759324e-06, + "loss": 0.0014, + "step": 48643 + }, + { + "epoch": 6.37, + "learning_rate": 2.094489290348477e-06, + "loss": 0.0016, + "step": 48644 + }, + { + "epoch": 6.37, + "learning_rate": 2.0941645324548456e-06, + "loss": 0.0016, + "step": 48645 + }, + { + "epoch": 6.37, + "learning_rate": 2.093839796795959e-06, + "loss": 0.0042, + "step": 48646 + }, + { + "epoch": 6.37, + "learning_rate": 2.093515083372727e-06, + "loss": 0.002, + "step": 48647 + }, + { + "epoch": 6.37, + "learning_rate": 2.093190392186062e-06, + "loss": 0.0035, + "step": 48648 + }, + { + "epoch": 6.37, + "learning_rate": 2.0928657232368763e-06, + "loss": 0.0025, + "step": 48649 + }, + { + "epoch": 6.37, + "learning_rate": 2.092541076526087e-06, + "loss": 0.0016, + "step": 48650 + }, + { + "epoch": 6.37, + "learning_rate": 2.092216452054604e-06, + "loss": 0.0018, + "step": 48651 + }, + { + "epoch": 6.37, + "learning_rate": 2.091891849823342e-06, + "loss": 0.0022, + "step": 48652 + }, + { + "epoch": 6.37, + "learning_rate": 2.09156726983321e-06, + "loss": 0.002, + "step": 48653 + }, + { + "epoch": 6.37, + "learning_rate": 2.091242712085124e-06, + "loss": 0.0029, + "step": 48654 + }, + { + "epoch": 6.37, + "learning_rate": 2.090918176579999e-06, + "loss": 0.0038, + "step": 48655 + }, + { + "epoch": 6.37, + "learning_rate": 2.090593663318746e-06, + "loss": 0.0028, + "step": 48656 + }, + { + "epoch": 6.37, + "learning_rate": 2.090269172302275e-06, + "loss": 0.0033, + "step": 48657 + }, + { + "epoch": 6.37, + "learning_rate": 2.0899447035315023e-06, + "loss": 0.0019, + "step": 48658 + }, + { + "epoch": 6.37, + "learning_rate": 2.08962025700734e-06, + "loss": 0.0014, + "step": 48659 + }, + { + "epoch": 6.37, + "learning_rate": 2.0892958327306957e-06, + "loss": 0.0019, + "step": 48660 + }, + { + "epoch": 6.37, + "learning_rate": 2.088971430702489e-06, + "loss": 0.0052, + "step": 48661 + }, + { + "epoch": 6.37, + "learning_rate": 2.088647050923628e-06, + "loss": 0.0024, + "step": 48662 + }, + { + "epoch": 6.37, + "learning_rate": 2.088322693395026e-06, + "loss": 0.0024, + "step": 48663 + }, + { + "epoch": 6.37, + "learning_rate": 2.0879983581175935e-06, + "loss": 0.0037, + "step": 48664 + }, + { + "epoch": 6.37, + "learning_rate": 2.087674045092244e-06, + "loss": 0.0015, + "step": 48665 + }, + { + "epoch": 6.37, + "learning_rate": 2.087349754319892e-06, + "loss": 0.0018, + "step": 48666 + }, + { + "epoch": 6.38, + "learning_rate": 2.087025485801447e-06, + "loss": 0.0018, + "step": 48667 + }, + { + "epoch": 6.38, + "learning_rate": 2.0867012395378196e-06, + "loss": 0.003, + "step": 48668 + }, + { + "epoch": 6.38, + "learning_rate": 2.0863770155299256e-06, + "loss": 0.0024, + "step": 48669 + }, + { + "epoch": 6.38, + "learning_rate": 2.086052813778676e-06, + "loss": 0.0023, + "step": 48670 + }, + { + "epoch": 6.38, + "learning_rate": 2.08572863428498e-06, + "loss": 0.0023, + "step": 48671 + }, + { + "epoch": 6.38, + "learning_rate": 2.0854044770497496e-06, + "loss": 0.0036, + "step": 48672 + }, + { + "epoch": 6.38, + "learning_rate": 2.0850803420738996e-06, + "loss": 0.002, + "step": 48673 + }, + { + "epoch": 6.38, + "learning_rate": 2.084756229358339e-06, + "loss": 0.0015, + "step": 48674 + }, + { + "epoch": 6.38, + "learning_rate": 2.084432138903978e-06, + "loss": 0.0028, + "step": 48675 + }, + { + "epoch": 6.38, + "learning_rate": 2.08410807071173e-06, + "loss": 0.0019, + "step": 48676 + }, + { + "epoch": 6.38, + "learning_rate": 2.0837840247825093e-06, + "loss": 0.0012, + "step": 48677 + }, + { + "epoch": 6.38, + "learning_rate": 2.0834600011172235e-06, + "loss": 0.0015, + "step": 48678 + }, + { + "epoch": 6.38, + "learning_rate": 2.0831359997167854e-06, + "loss": 0.0024, + "step": 48679 + }, + { + "epoch": 6.38, + "learning_rate": 2.0828120205821032e-06, + "loss": 0.0022, + "step": 48680 + }, + { + "epoch": 6.38, + "learning_rate": 2.0824880637140932e-06, + "loss": 0.0022, + "step": 48681 + }, + { + "epoch": 6.38, + "learning_rate": 2.082164129113664e-06, + "loss": 0.0024, + "step": 48682 + }, + { + "epoch": 6.38, + "learning_rate": 2.081840216781724e-06, + "loss": 0.0023, + "step": 48683 + }, + { + "epoch": 6.38, + "learning_rate": 2.0815163267191895e-06, + "loss": 0.0014, + "step": 48684 + }, + { + "epoch": 6.38, + "learning_rate": 2.0811924589269683e-06, + "loss": 0.0021, + "step": 48685 + }, + { + "epoch": 6.38, + "learning_rate": 2.0808686134059696e-06, + "loss": 0.0023, + "step": 48686 + }, + { + "epoch": 6.38, + "learning_rate": 2.0805447901571075e-06, + "loss": 0.0012, + "step": 48687 + }, + { + "epoch": 6.38, + "learning_rate": 2.08022098918129e-06, + "loss": 0.0017, + "step": 48688 + }, + { + "epoch": 6.38, + "learning_rate": 2.0798972104794313e-06, + "loss": 0.002, + "step": 48689 + }, + { + "epoch": 6.38, + "learning_rate": 2.0795734540524402e-06, + "loss": 0.0014, + "step": 48690 + }, + { + "epoch": 6.38, + "learning_rate": 2.079249719901224e-06, + "loss": 0.0017, + "step": 48691 + }, + { + "epoch": 6.38, + "learning_rate": 2.0789260080266993e-06, + "loss": 0.0015, + "step": 48692 + }, + { + "epoch": 6.38, + "learning_rate": 2.078602318429773e-06, + "loss": 0.002, + "step": 48693 + }, + { + "epoch": 6.38, + "learning_rate": 2.0782786511113563e-06, + "loss": 0.0019, + "step": 48694 + }, + { + "epoch": 6.38, + "learning_rate": 2.077955006072355e-06, + "loss": 0.002, + "step": 48695 + }, + { + "epoch": 6.38, + "learning_rate": 2.0776313833136875e-06, + "loss": 0.0025, + "step": 48696 + }, + { + "epoch": 6.38, + "learning_rate": 2.0773077828362566e-06, + "loss": 0.0019, + "step": 48697 + }, + { + "epoch": 6.38, + "learning_rate": 2.0769842046409783e-06, + "loss": 0.0017, + "step": 48698 + }, + { + "epoch": 6.38, + "learning_rate": 2.076660648728758e-06, + "loss": 0.003, + "step": 48699 + }, + { + "epoch": 6.38, + "learning_rate": 2.0763371151005098e-06, + "loss": 0.0029, + "step": 48700 + }, + { + "epoch": 6.38, + "learning_rate": 2.076013603757141e-06, + "loss": 0.0019, + "step": 48701 + }, + { + "epoch": 6.38, + "learning_rate": 2.075690114699562e-06, + "loss": 0.0024, + "step": 48702 + }, + { + "epoch": 6.38, + "learning_rate": 2.0753666479286817e-06, + "loss": 0.0025, + "step": 48703 + }, + { + "epoch": 6.38, + "learning_rate": 2.0750432034454116e-06, + "loss": 0.0013, + "step": 48704 + }, + { + "epoch": 6.38, + "learning_rate": 2.0747197812506602e-06, + "loss": 0.003, + "step": 48705 + }, + { + "epoch": 6.38, + "learning_rate": 2.074396381345336e-06, + "loss": 0.0021, + "step": 48706 + }, + { + "epoch": 6.38, + "learning_rate": 2.0740730037303527e-06, + "loss": 0.0032, + "step": 48707 + }, + { + "epoch": 6.38, + "learning_rate": 2.0737496484066133e-06, + "loss": 0.002, + "step": 48708 + }, + { + "epoch": 6.38, + "learning_rate": 2.0734263153750333e-06, + "loss": 0.0017, + "step": 48709 + }, + { + "epoch": 6.38, + "learning_rate": 2.0731030046365173e-06, + "loss": 0.0012, + "step": 48710 + }, + { + "epoch": 6.38, + "learning_rate": 2.0727797161919804e-06, + "loss": 0.0019, + "step": 48711 + }, + { + "epoch": 6.38, + "learning_rate": 2.072456450042328e-06, + "loss": 0.0019, + "step": 48712 + }, + { + "epoch": 6.38, + "learning_rate": 2.072133206188469e-06, + "loss": 0.0031, + "step": 48713 + }, + { + "epoch": 6.38, + "learning_rate": 2.071809984631311e-06, + "loss": 0.0023, + "step": 48714 + }, + { + "epoch": 6.38, + "learning_rate": 2.071486785371769e-06, + "loss": 0.002, + "step": 48715 + }, + { + "epoch": 6.38, + "learning_rate": 2.071163608410747e-06, + "loss": 0.0013, + "step": 48716 + }, + { + "epoch": 6.38, + "learning_rate": 2.0708404537491556e-06, + "loss": 0.0018, + "step": 48717 + }, + { + "epoch": 6.38, + "learning_rate": 2.070517321387899e-06, + "loss": 0.002, + "step": 48718 + }, + { + "epoch": 6.38, + "learning_rate": 2.0701942113278915e-06, + "loss": 0.0014, + "step": 48719 + }, + { + "epoch": 6.38, + "learning_rate": 2.069871123570043e-06, + "loss": 0.0023, + "step": 48720 + }, + { + "epoch": 6.38, + "learning_rate": 2.0695480581152603e-06, + "loss": 0.0028, + "step": 48721 + }, + { + "epoch": 6.38, + "learning_rate": 2.0692250149644477e-06, + "loss": 0.0023, + "step": 48722 + }, + { + "epoch": 6.38, + "learning_rate": 2.0689019941185207e-06, + "loss": 0.0026, + "step": 48723 + }, + { + "epoch": 6.38, + "learning_rate": 2.0685789955783843e-06, + "loss": 0.0026, + "step": 48724 + }, + { + "epoch": 6.38, + "learning_rate": 2.068256019344944e-06, + "loss": 0.0025, + "step": 48725 + }, + { + "epoch": 6.38, + "learning_rate": 2.067933065419114e-06, + "loss": 0.0024, + "step": 48726 + }, + { + "epoch": 6.38, + "learning_rate": 2.0676101338018007e-06, + "loss": 0.0028, + "step": 48727 + }, + { + "epoch": 6.38, + "learning_rate": 2.067287224493911e-06, + "loss": 0.0026, + "step": 48728 + }, + { + "epoch": 6.38, + "learning_rate": 2.066964337496351e-06, + "loss": 0.0018, + "step": 48729 + }, + { + "epoch": 6.38, + "learning_rate": 2.0666414728100315e-06, + "loss": 0.0019, + "step": 48730 + }, + { + "epoch": 6.38, + "learning_rate": 2.0663186304358616e-06, + "loss": 0.0031, + "step": 48731 + }, + { + "epoch": 6.38, + "learning_rate": 2.0659958103747503e-06, + "loss": 0.0015, + "step": 48732 + }, + { + "epoch": 6.38, + "learning_rate": 2.0656730126275992e-06, + "loss": 0.0047, + "step": 48733 + }, + { + "epoch": 6.38, + "learning_rate": 2.065350237195323e-06, + "loss": 0.0015, + "step": 48734 + }, + { + "epoch": 6.38, + "learning_rate": 2.0650274840788265e-06, + "loss": 0.0024, + "step": 48735 + }, + { + "epoch": 6.38, + "learning_rate": 2.064704753279018e-06, + "loss": 0.0032, + "step": 48736 + }, + { + "epoch": 6.38, + "learning_rate": 2.064382044796802e-06, + "loss": 0.0017, + "step": 48737 + }, + { + "epoch": 6.38, + "learning_rate": 2.0640593586330915e-06, + "loss": 0.0016, + "step": 48738 + }, + { + "epoch": 6.38, + "learning_rate": 2.0637366947887915e-06, + "loss": 0.0019, + "step": 48739 + }, + { + "epoch": 6.38, + "learning_rate": 2.063414053264807e-06, + "loss": 0.0015, + "step": 48740 + }, + { + "epoch": 6.38, + "learning_rate": 2.063091434062048e-06, + "loss": 0.002, + "step": 48741 + }, + { + "epoch": 6.38, + "learning_rate": 2.0627688371814246e-06, + "loss": 0.0024, + "step": 48742 + }, + { + "epoch": 6.38, + "learning_rate": 2.0624462626238404e-06, + "loss": 0.0023, + "step": 48743 + }, + { + "epoch": 6.39, + "learning_rate": 2.0621237103902035e-06, + "loss": 0.0022, + "step": 48744 + }, + { + "epoch": 6.39, + "learning_rate": 2.0618011804814187e-06, + "loss": 0.0024, + "step": 48745 + }, + { + "epoch": 6.39, + "learning_rate": 2.0614786728983972e-06, + "loss": 0.0026, + "step": 48746 + }, + { + "epoch": 6.39, + "learning_rate": 2.061156187642045e-06, + "loss": 0.0011, + "step": 48747 + }, + { + "epoch": 6.39, + "learning_rate": 2.060833724713265e-06, + "loss": 0.0027, + "step": 48748 + }, + { + "epoch": 6.39, + "learning_rate": 2.0605112841129703e-06, + "loss": 0.0028, + "step": 48749 + }, + { + "epoch": 6.39, + "learning_rate": 2.0601888658420655e-06, + "loss": 0.0016, + "step": 48750 + }, + { + "epoch": 6.39, + "learning_rate": 2.059866469901456e-06, + "loss": 0.0025, + "step": 48751 + }, + { + "epoch": 6.39, + "learning_rate": 2.0595440962920453e-06, + "loss": 0.0013, + "step": 48752 + }, + { + "epoch": 6.39, + "learning_rate": 2.0592217450147457e-06, + "loss": 0.0012, + "step": 48753 + }, + { + "epoch": 6.39, + "learning_rate": 2.0588994160704646e-06, + "loss": 0.0019, + "step": 48754 + }, + { + "epoch": 6.39, + "learning_rate": 2.058577109460106e-06, + "loss": 0.0012, + "step": 48755 + }, + { + "epoch": 6.39, + "learning_rate": 2.0582548251845725e-06, + "loss": 0.0018, + "step": 48756 + }, + { + "epoch": 6.39, + "learning_rate": 2.057932563244779e-06, + "loss": 0.0027, + "step": 48757 + }, + { + "epoch": 6.39, + "learning_rate": 2.0576103236416256e-06, + "loss": 0.003, + "step": 48758 + }, + { + "epoch": 6.39, + "learning_rate": 2.057288106376021e-06, + "loss": 0.0015, + "step": 48759 + }, + { + "epoch": 6.39, + "learning_rate": 2.056965911448867e-06, + "loss": 0.0035, + "step": 48760 + }, + { + "epoch": 6.39, + "learning_rate": 2.056643738861077e-06, + "loss": 0.0022, + "step": 48761 + }, + { + "epoch": 6.39, + "learning_rate": 2.0563215886135535e-06, + "loss": 0.0017, + "step": 48762 + }, + { + "epoch": 6.39, + "learning_rate": 2.0559994607071997e-06, + "loss": 0.0031, + "step": 48763 + }, + { + "epoch": 6.39, + "learning_rate": 2.0556773551429243e-06, + "loss": 0.0022, + "step": 48764 + }, + { + "epoch": 6.39, + "learning_rate": 2.0553552719216353e-06, + "loss": 0.0035, + "step": 48765 + }, + { + "epoch": 6.39, + "learning_rate": 2.0550332110442372e-06, + "loss": 0.003, + "step": 48766 + }, + { + "epoch": 6.39, + "learning_rate": 2.0547111725116353e-06, + "loss": 0.0019, + "step": 48767 + }, + { + "epoch": 6.39, + "learning_rate": 2.054389156324732e-06, + "loss": 0.0016, + "step": 48768 + }, + { + "epoch": 6.39, + "learning_rate": 2.054067162484439e-06, + "loss": 0.003, + "step": 48769 + }, + { + "epoch": 6.39, + "learning_rate": 2.0537451909916585e-06, + "loss": 0.0015, + "step": 48770 + }, + { + "epoch": 6.39, + "learning_rate": 2.0534232418472943e-06, + "loss": 0.002, + "step": 48771 + }, + { + "epoch": 6.39, + "learning_rate": 2.0531013150522574e-06, + "loss": 0.0024, + "step": 48772 + }, + { + "epoch": 6.39, + "learning_rate": 2.0527794106074485e-06, + "loss": 0.0022, + "step": 48773 + }, + { + "epoch": 6.39, + "learning_rate": 2.052457528513773e-06, + "loss": 0.0017, + "step": 48774 + }, + { + "epoch": 6.39, + "learning_rate": 2.052135668772136e-06, + "loss": 0.0014, + "step": 48775 + }, + { + "epoch": 6.39, + "learning_rate": 2.051813831383448e-06, + "loss": 0.0021, + "step": 48776 + }, + { + "epoch": 6.39, + "learning_rate": 2.0514920163486096e-06, + "loss": 0.0014, + "step": 48777 + }, + { + "epoch": 6.39, + "learning_rate": 2.051170223668526e-06, + "loss": 0.0041, + "step": 48778 + }, + { + "epoch": 6.39, + "learning_rate": 2.050848453344102e-06, + "loss": 0.0023, + "step": 48779 + }, + { + "epoch": 6.39, + "learning_rate": 2.0505267053762447e-06, + "loss": 0.0014, + "step": 48780 + }, + { + "epoch": 6.39, + "learning_rate": 2.0502049797658586e-06, + "loss": 0.0036, + "step": 48781 + }, + { + "epoch": 6.39, + "learning_rate": 2.0498832765138466e-06, + "loss": 0.002, + "step": 48782 + }, + { + "epoch": 6.39, + "learning_rate": 2.0495615956211123e-06, + "loss": 0.0013, + "step": 48783 + }, + { + "epoch": 6.39, + "learning_rate": 2.049239937088565e-06, + "loss": 0.0025, + "step": 48784 + }, + { + "epoch": 6.39, + "learning_rate": 2.0489183009171053e-06, + "loss": 0.0038, + "step": 48785 + }, + { + "epoch": 6.39, + "learning_rate": 2.048596687107641e-06, + "loss": 0.0017, + "step": 48786 + }, + { + "epoch": 6.39, + "learning_rate": 2.048275095661073e-06, + "loss": 0.0016, + "step": 48787 + }, + { + "epoch": 6.39, + "learning_rate": 2.0479535265783112e-06, + "loss": 0.0023, + "step": 48788 + }, + { + "epoch": 6.39, + "learning_rate": 2.047631979860256e-06, + "loss": 0.003, + "step": 48789 + }, + { + "epoch": 6.39, + "learning_rate": 2.0473104555078096e-06, + "loss": 0.0031, + "step": 48790 + }, + { + "epoch": 6.39, + "learning_rate": 2.0469889535218812e-06, + "loss": 0.0016, + "step": 48791 + }, + { + "epoch": 6.39, + "learning_rate": 2.0466674739033732e-06, + "loss": 0.0019, + "step": 48792 + }, + { + "epoch": 6.39, + "learning_rate": 2.04634601665319e-06, + "loss": 0.0023, + "step": 48793 + }, + { + "epoch": 6.39, + "learning_rate": 2.046024581772231e-06, + "loss": 0.0015, + "step": 48794 + }, + { + "epoch": 6.39, + "learning_rate": 2.045703169261408e-06, + "loss": 0.0016, + "step": 48795 + }, + { + "epoch": 6.39, + "learning_rate": 2.045381779121619e-06, + "loss": 0.0027, + "step": 48796 + }, + { + "epoch": 6.39, + "learning_rate": 2.045060411353772e-06, + "loss": 0.0023, + "step": 48797 + }, + { + "epoch": 6.39, + "learning_rate": 2.0447390659587674e-06, + "loss": 0.0026, + "step": 48798 + }, + { + "epoch": 6.39, + "learning_rate": 2.0444177429375122e-06, + "loss": 0.0016, + "step": 48799 + }, + { + "epoch": 6.39, + "learning_rate": 2.044096442290908e-06, + "loss": 0.0016, + "step": 48800 + }, + { + "epoch": 6.39, + "learning_rate": 2.04377516401986e-06, + "loss": 0.0031, + "step": 48801 + }, + { + "epoch": 6.39, + "learning_rate": 2.043453908125267e-06, + "loss": 0.0021, + "step": 48802 + }, + { + "epoch": 6.39, + "learning_rate": 2.0431326746080393e-06, + "loss": 0.0013, + "step": 48803 + }, + { + "epoch": 6.39, + "learning_rate": 2.0428114634690774e-06, + "loss": 0.003, + "step": 48804 + }, + { + "epoch": 6.39, + "learning_rate": 2.042490274709282e-06, + "loss": 0.0023, + "step": 48805 + }, + { + "epoch": 6.39, + "learning_rate": 2.0421691083295613e-06, + "loss": 0.0011, + "step": 48806 + }, + { + "epoch": 6.39, + "learning_rate": 2.041847964330813e-06, + "loss": 0.0026, + "step": 48807 + }, + { + "epoch": 6.39, + "learning_rate": 2.0415268427139467e-06, + "loss": 0.0025, + "step": 48808 + }, + { + "epoch": 6.39, + "learning_rate": 2.0412057434798626e-06, + "loss": 0.0023, + "step": 48809 + }, + { + "epoch": 6.39, + "learning_rate": 2.0408846666294614e-06, + "loss": 0.002, + "step": 48810 + }, + { + "epoch": 6.39, + "learning_rate": 2.0405636121636494e-06, + "loss": 0.0013, + "step": 48811 + }, + { + "epoch": 6.39, + "learning_rate": 2.0402425800833292e-06, + "loss": 0.0026, + "step": 48812 + }, + { + "epoch": 6.39, + "learning_rate": 2.0399215703894004e-06, + "loss": 0.0018, + "step": 48813 + }, + { + "epoch": 6.39, + "learning_rate": 2.0396005830827704e-06, + "loss": 0.0014, + "step": 48814 + }, + { + "epoch": 6.39, + "learning_rate": 2.0392796181643406e-06, + "loss": 0.0009, + "step": 48815 + }, + { + "epoch": 6.39, + "learning_rate": 2.038958675635012e-06, + "loss": 0.0022, + "step": 48816 + }, + { + "epoch": 6.39, + "learning_rate": 2.038637755495686e-06, + "loss": 0.0045, + "step": 48817 + }, + { + "epoch": 6.39, + "learning_rate": 2.0383168577472665e-06, + "loss": 0.0023, + "step": 48818 + }, + { + "epoch": 6.39, + "learning_rate": 2.0379959823906603e-06, + "loss": 0.0014, + "step": 48819 + }, + { + "epoch": 6.4, + "learning_rate": 2.0376751294267673e-06, + "loss": 0.0023, + "step": 48820 + }, + { + "epoch": 6.4, + "learning_rate": 2.037354298856485e-06, + "loss": 0.0022, + "step": 48821 + }, + { + "epoch": 6.4, + "learning_rate": 2.0370334906807234e-06, + "loss": 0.0017, + "step": 48822 + }, + { + "epoch": 6.4, + "learning_rate": 2.0367127049003798e-06, + "loss": 0.0021, + "step": 48823 + }, + { + "epoch": 6.4, + "learning_rate": 2.036391941516358e-06, + "loss": 0.0021, + "step": 48824 + }, + { + "epoch": 6.4, + "learning_rate": 2.036071200529558e-06, + "loss": 0.003, + "step": 48825 + }, + { + "epoch": 6.4, + "learning_rate": 2.0357504819408858e-06, + "loss": 0.0013, + "step": 48826 + }, + { + "epoch": 6.4, + "learning_rate": 2.0354297857512407e-06, + "loss": 0.0019, + "step": 48827 + }, + { + "epoch": 6.4, + "learning_rate": 2.035109111961523e-06, + "loss": 0.0009, + "step": 48828 + }, + { + "epoch": 6.4, + "learning_rate": 2.0347884605726363e-06, + "loss": 0.0027, + "step": 48829 + }, + { + "epoch": 6.4, + "learning_rate": 2.034467831585486e-06, + "loss": 0.0017, + "step": 48830 + }, + { + "epoch": 6.4, + "learning_rate": 2.034147225000971e-06, + "loss": 0.0018, + "step": 48831 + }, + { + "epoch": 6.4, + "learning_rate": 2.0338266408199927e-06, + "loss": 0.0022, + "step": 48832 + }, + { + "epoch": 6.4, + "learning_rate": 2.033506079043449e-06, + "loss": 0.0018, + "step": 48833 + }, + { + "epoch": 6.4, + "learning_rate": 2.033185539672249e-06, + "loss": 0.0021, + "step": 48834 + }, + { + "epoch": 6.4, + "learning_rate": 2.03286502270729e-06, + "loss": 0.0022, + "step": 48835 + }, + { + "epoch": 6.4, + "learning_rate": 2.0325445281494714e-06, + "loss": 0.0035, + "step": 48836 + }, + { + "epoch": 6.4, + "learning_rate": 2.0322240559996995e-06, + "loss": 0.0019, + "step": 48837 + }, + { + "epoch": 6.4, + "learning_rate": 2.0319036062588737e-06, + "loss": 0.002, + "step": 48838 + }, + { + "epoch": 6.4, + "learning_rate": 2.031583178927892e-06, + "loss": 0.0016, + "step": 48839 + }, + { + "epoch": 6.4, + "learning_rate": 2.0312627740076574e-06, + "loss": 0.0012, + "step": 48840 + }, + { + "epoch": 6.4, + "learning_rate": 2.0309423914990744e-06, + "loss": 0.0023, + "step": 48841 + }, + { + "epoch": 6.4, + "learning_rate": 2.030622031403042e-06, + "loss": 0.002, + "step": 48842 + }, + { + "epoch": 6.4, + "learning_rate": 2.030301693720461e-06, + "loss": 0.002, + "step": 48843 + }, + { + "epoch": 6.4, + "learning_rate": 2.0299813784522292e-06, + "loss": 0.0015, + "step": 48844 + }, + { + "epoch": 6.4, + "learning_rate": 2.0296610855992525e-06, + "loss": 0.0018, + "step": 48845 + }, + { + "epoch": 6.4, + "learning_rate": 2.0293408151624306e-06, + "loss": 0.0028, + "step": 48846 + }, + { + "epoch": 6.4, + "learning_rate": 2.02902056714266e-06, + "loss": 0.0025, + "step": 48847 + }, + { + "epoch": 6.4, + "learning_rate": 2.0287003415408467e-06, + "loss": 0.0014, + "step": 48848 + }, + { + "epoch": 6.4, + "learning_rate": 2.0283801383578906e-06, + "loss": 0.0022, + "step": 48849 + }, + { + "epoch": 6.4, + "learning_rate": 2.028059957594687e-06, + "loss": 0.0011, + "step": 48850 + }, + { + "epoch": 6.4, + "learning_rate": 2.0277397992521432e-06, + "loss": 0.0023, + "step": 48851 + }, + { + "epoch": 6.4, + "learning_rate": 2.027419663331154e-06, + "loss": 0.0015, + "step": 48852 + }, + { + "epoch": 6.4, + "learning_rate": 2.0270995498326253e-06, + "loss": 0.0038, + "step": 48853 + }, + { + "epoch": 6.4, + "learning_rate": 2.026779458757454e-06, + "loss": 0.0025, + "step": 48854 + }, + { + "epoch": 6.4, + "learning_rate": 2.026459390106539e-06, + "loss": 0.0016, + "step": 48855 + }, + { + "epoch": 6.4, + "learning_rate": 2.0261393438807843e-06, + "loss": 0.0021, + "step": 48856 + }, + { + "epoch": 6.4, + "learning_rate": 2.025819320081088e-06, + "loss": 0.0022, + "step": 48857 + }, + { + "epoch": 6.4, + "learning_rate": 2.0254993187083514e-06, + "loss": 0.0013, + "step": 48858 + }, + { + "epoch": 6.4, + "learning_rate": 2.02517933976347e-06, + "loss": 0.0024, + "step": 48859 + }, + { + "epoch": 6.4, + "learning_rate": 2.0248593832473497e-06, + "loss": 0.002, + "step": 48860 + }, + { + "epoch": 6.4, + "learning_rate": 2.024539449160884e-06, + "loss": 0.0018, + "step": 48861 + }, + { + "epoch": 6.4, + "learning_rate": 2.02421953750498e-06, + "loss": 0.0023, + "step": 48862 + }, + { + "epoch": 6.4, + "learning_rate": 2.0238996482805308e-06, + "loss": 0.0011, + "step": 48863 + }, + { + "epoch": 6.4, + "learning_rate": 2.0235797814884426e-06, + "loss": 0.0024, + "step": 48864 + }, + { + "epoch": 6.4, + "learning_rate": 2.02325993712961e-06, + "loss": 0.0023, + "step": 48865 + }, + { + "epoch": 6.4, + "learning_rate": 2.0229401152049355e-06, + "loss": 0.0023, + "step": 48866 + }, + { + "epoch": 6.4, + "learning_rate": 2.0226203157153134e-06, + "loss": 0.0035, + "step": 48867 + }, + { + "epoch": 6.4, + "learning_rate": 2.02230053866165e-06, + "loss": 0.0035, + "step": 48868 + }, + { + "epoch": 6.4, + "learning_rate": 2.021980784044841e-06, + "loss": 0.003, + "step": 48869 + }, + { + "epoch": 6.4, + "learning_rate": 2.0216610518657853e-06, + "loss": 0.0024, + "step": 48870 + }, + { + "epoch": 6.4, + "learning_rate": 2.0213413421253836e-06, + "loss": 0.0018, + "step": 48871 + }, + { + "epoch": 6.4, + "learning_rate": 2.021021654824533e-06, + "loss": 0.0034, + "step": 48872 + }, + { + "epoch": 6.4, + "learning_rate": 2.020701989964137e-06, + "loss": 0.0013, + "step": 48873 + }, + { + "epoch": 6.4, + "learning_rate": 2.02038234754509e-06, + "loss": 0.0014, + "step": 48874 + }, + { + "epoch": 6.4, + "learning_rate": 2.0200627275682917e-06, + "loss": 0.0019, + "step": 48875 + }, + { + "epoch": 6.4, + "learning_rate": 2.0197431300346435e-06, + "loss": 0.0012, + "step": 48876 + }, + { + "epoch": 6.4, + "learning_rate": 2.0194235549450438e-06, + "loss": 0.0008, + "step": 48877 + }, + { + "epoch": 6.4, + "learning_rate": 2.0191040023003873e-06, + "loss": 0.0016, + "step": 48878 + }, + { + "epoch": 6.4, + "learning_rate": 2.018784472101578e-06, + "loss": 0.0022, + "step": 48879 + }, + { + "epoch": 6.4, + "learning_rate": 2.0184649643495135e-06, + "loss": 0.0025, + "step": 48880 + }, + { + "epoch": 6.4, + "learning_rate": 2.0181454790450906e-06, + "loss": 0.001, + "step": 48881 + }, + { + "epoch": 6.4, + "learning_rate": 2.017826016189205e-06, + "loss": 0.0018, + "step": 48882 + }, + { + "epoch": 6.4, + "learning_rate": 2.0175065757827596e-06, + "loss": 0.0028, + "step": 48883 + }, + { + "epoch": 6.4, + "learning_rate": 2.0171871578266544e-06, + "loss": 0.0026, + "step": 48884 + }, + { + "epoch": 6.4, + "learning_rate": 2.0168677623217847e-06, + "loss": 0.004, + "step": 48885 + }, + { + "epoch": 6.4, + "learning_rate": 2.0165483892690475e-06, + "loss": 0.0016, + "step": 48886 + }, + { + "epoch": 6.4, + "learning_rate": 2.016229038669345e-06, + "loss": 0.0033, + "step": 48887 + }, + { + "epoch": 6.4, + "learning_rate": 2.015909710523574e-06, + "loss": 0.0013, + "step": 48888 + }, + { + "epoch": 6.4, + "learning_rate": 2.01559040483263e-06, + "loss": 0.0019, + "step": 48889 + }, + { + "epoch": 6.4, + "learning_rate": 2.015271121597411e-06, + "loss": 0.0018, + "step": 48890 + }, + { + "epoch": 6.4, + "learning_rate": 2.01495186081882e-06, + "loss": 0.0024, + "step": 48891 + }, + { + "epoch": 6.4, + "learning_rate": 2.014632622497751e-06, + "loss": 0.0028, + "step": 48892 + }, + { + "epoch": 6.4, + "learning_rate": 2.0143134066351e-06, + "loss": 0.0024, + "step": 48893 + }, + { + "epoch": 6.4, + "learning_rate": 2.013994213231767e-06, + "loss": 0.0015, + "step": 48894 + }, + { + "epoch": 6.4, + "learning_rate": 2.013675042288653e-06, + "loss": 0.0019, + "step": 48895 + }, + { + "epoch": 6.41, + "learning_rate": 2.0133558938066532e-06, + "loss": 0.0019, + "step": 48896 + }, + { + "epoch": 6.41, + "learning_rate": 2.013036767786661e-06, + "loss": 0.002, + "step": 48897 + }, + { + "epoch": 6.41, + "learning_rate": 2.012717664229581e-06, + "loss": 0.0014, + "step": 48898 + }, + { + "epoch": 6.41, + "learning_rate": 2.0123985831363067e-06, + "loss": 0.0019, + "step": 48899 + }, + { + "epoch": 6.41, + "learning_rate": 2.012079524507736e-06, + "loss": 0.0012, + "step": 48900 + }, + { + "epoch": 6.41, + "learning_rate": 2.0117604883447638e-06, + "loss": 0.0026, + "step": 48901 + }, + { + "epoch": 6.41, + "learning_rate": 2.0114414746482926e-06, + "loss": 0.0016, + "step": 48902 + }, + { + "epoch": 6.41, + "learning_rate": 2.0111224834192165e-06, + "loss": 0.0022, + "step": 48903 + }, + { + "epoch": 6.41, + "learning_rate": 2.0108035146584327e-06, + "loss": 0.0022, + "step": 48904 + }, + { + "epoch": 6.41, + "learning_rate": 2.010484568366834e-06, + "loss": 0.0022, + "step": 48905 + }, + { + "epoch": 6.41, + "learning_rate": 2.0101656445453267e-06, + "loss": 0.0007, + "step": 48906 + }, + { + "epoch": 6.41, + "learning_rate": 2.0098467431948032e-06, + "loss": 0.0032, + "step": 48907 + }, + { + "epoch": 6.41, + "learning_rate": 2.009527864316161e-06, + "loss": 0.0013, + "step": 48908 + }, + { + "epoch": 6.41, + "learning_rate": 2.0092090079102923e-06, + "loss": 0.0011, + "step": 48909 + }, + { + "epoch": 6.41, + "learning_rate": 2.008890173978101e-06, + "loss": 0.0023, + "step": 48910 + }, + { + "epoch": 6.41, + "learning_rate": 2.0085713625204804e-06, + "loss": 0.0014, + "step": 48911 + }, + { + "epoch": 6.41, + "learning_rate": 2.008252573538325e-06, + "loss": 0.0014, + "step": 48912 + }, + { + "epoch": 6.41, + "learning_rate": 2.007933807032536e-06, + "loss": 0.002, + "step": 48913 + }, + { + "epoch": 6.41, + "learning_rate": 2.007615063004007e-06, + "loss": 0.0013, + "step": 48914 + }, + { + "epoch": 6.41, + "learning_rate": 2.007296341453636e-06, + "loss": 0.0044, + "step": 48915 + }, + { + "epoch": 6.41, + "learning_rate": 2.006977642382315e-06, + "loss": 0.001, + "step": 48916 + }, + { + "epoch": 6.41, + "learning_rate": 2.006658965790944e-06, + "loss": 0.0012, + "step": 48917 + }, + { + "epoch": 6.41, + "learning_rate": 2.0063403116804224e-06, + "loss": 0.001, + "step": 48918 + }, + { + "epoch": 6.41, + "learning_rate": 2.0060216800516417e-06, + "loss": 0.0012, + "step": 48919 + }, + { + "epoch": 6.41, + "learning_rate": 2.005703070905498e-06, + "loss": 0.0011, + "step": 48920 + }, + { + "epoch": 6.41, + "learning_rate": 2.0053844842428904e-06, + "loss": 0.0016, + "step": 48921 + }, + { + "epoch": 6.41, + "learning_rate": 2.005065920064714e-06, + "loss": 0.0017, + "step": 48922 + }, + { + "epoch": 6.41, + "learning_rate": 2.004747378371863e-06, + "loss": 0.0019, + "step": 48923 + }, + { + "epoch": 6.41, + "learning_rate": 2.0044288591652316e-06, + "loss": 0.0017, + "step": 48924 + }, + { + "epoch": 6.41, + "learning_rate": 2.0041103624457216e-06, + "loss": 0.0021, + "step": 48925 + }, + { + "epoch": 6.41, + "learning_rate": 2.0037918882142248e-06, + "loss": 0.0018, + "step": 48926 + }, + { + "epoch": 6.41, + "learning_rate": 2.0034734364716345e-06, + "loss": 0.0022, + "step": 48927 + }, + { + "epoch": 6.41, + "learning_rate": 2.00315500721885e-06, + "loss": 0.001, + "step": 48928 + }, + { + "epoch": 6.41, + "learning_rate": 2.0028366004567688e-06, + "loss": 0.0018, + "step": 48929 + }, + { + "epoch": 6.41, + "learning_rate": 2.0025182161862845e-06, + "loss": 0.0017, + "step": 48930 + }, + { + "epoch": 6.41, + "learning_rate": 2.002199854408291e-06, + "loss": 0.0014, + "step": 48931 + }, + { + "epoch": 6.41, + "learning_rate": 2.0018815151236814e-06, + "loss": 0.0018, + "step": 48932 + }, + { + "epoch": 6.41, + "learning_rate": 2.001563198333357e-06, + "loss": 0.0025, + "step": 48933 + }, + { + "epoch": 6.41, + "learning_rate": 2.0012449040382108e-06, + "loss": 0.0024, + "step": 48934 + }, + { + "epoch": 6.41, + "learning_rate": 2.000926632239134e-06, + "loss": 0.0013, + "step": 48935 + }, + { + "epoch": 6.41, + "learning_rate": 2.000608382937028e-06, + "loss": 0.0015, + "step": 48936 + }, + { + "epoch": 6.41, + "learning_rate": 2.0002901561327847e-06, + "loss": 0.0014, + "step": 48937 + }, + { + "epoch": 6.41, + "learning_rate": 1.9999719518272974e-06, + "loss": 0.0032, + "step": 48938 + }, + { + "epoch": 6.41, + "learning_rate": 1.9996537700214646e-06, + "loss": 0.0015, + "step": 48939 + }, + { + "epoch": 6.41, + "learning_rate": 1.9993356107161777e-06, + "loss": 0.0011, + "step": 48940 + }, + { + "epoch": 6.41, + "learning_rate": 1.999017473912336e-06, + "loss": 0.0017, + "step": 48941 + }, + { + "epoch": 6.41, + "learning_rate": 1.998699359610832e-06, + "loss": 0.0009, + "step": 48942 + }, + { + "epoch": 6.41, + "learning_rate": 1.998381267812556e-06, + "loss": 0.0032, + "step": 48943 + }, + { + "epoch": 6.41, + "learning_rate": 1.9980631985184097e-06, + "loss": 0.0026, + "step": 48944 + }, + { + "epoch": 6.41, + "learning_rate": 1.9977451517292847e-06, + "loss": 0.0023, + "step": 48945 + }, + { + "epoch": 6.41, + "learning_rate": 1.9974271274460755e-06, + "loss": 0.0044, + "step": 48946 + }, + { + "epoch": 6.41, + "learning_rate": 1.997109125669673e-06, + "loss": 0.0027, + "step": 48947 + }, + { + "epoch": 6.41, + "learning_rate": 1.996791146400976e-06, + "loss": 0.0009, + "step": 48948 + }, + { + "epoch": 6.41, + "learning_rate": 1.9964731896408794e-06, + "loss": 0.0018, + "step": 48949 + }, + { + "epoch": 6.41, + "learning_rate": 1.9961552553902764e-06, + "loss": 0.001, + "step": 48950 + }, + { + "epoch": 6.41, + "learning_rate": 1.9958373436500568e-06, + "loss": 0.0019, + "step": 48951 + }, + { + "epoch": 6.41, + "learning_rate": 1.995519454421122e-06, + "loss": 0.0017, + "step": 48952 + }, + { + "epoch": 6.41, + "learning_rate": 1.9952015877043616e-06, + "loss": 0.0013, + "step": 48953 + }, + { + "epoch": 6.41, + "learning_rate": 1.994883743500672e-06, + "loss": 0.0017, + "step": 48954 + }, + { + "epoch": 6.41, + "learning_rate": 1.9945659218109416e-06, + "loss": 0.0032, + "step": 48955 + }, + { + "epoch": 6.41, + "learning_rate": 1.994248122636071e-06, + "loss": 0.0038, + "step": 48956 + }, + { + "epoch": 6.41, + "learning_rate": 1.993930345976952e-06, + "loss": 0.0021, + "step": 48957 + }, + { + "epoch": 6.41, + "learning_rate": 1.993612591834474e-06, + "loss": 0.0013, + "step": 48958 + }, + { + "epoch": 6.41, + "learning_rate": 1.993294860209535e-06, + "loss": 0.0012, + "step": 48959 + }, + { + "epoch": 6.41, + "learning_rate": 1.9929771511030306e-06, + "loss": 0.0011, + "step": 48960 + }, + { + "epoch": 6.41, + "learning_rate": 1.99265946451585e-06, + "loss": 0.0019, + "step": 48961 + }, + { + "epoch": 6.41, + "learning_rate": 1.992341800448887e-06, + "loss": 0.0024, + "step": 48962 + }, + { + "epoch": 6.41, + "learning_rate": 1.9920241589030387e-06, + "loss": 0.0014, + "step": 48963 + }, + { + "epoch": 6.41, + "learning_rate": 1.991706539879197e-06, + "loss": 0.0017, + "step": 48964 + }, + { + "epoch": 6.41, + "learning_rate": 1.991388943378253e-06, + "loss": 0.0025, + "step": 48965 + }, + { + "epoch": 6.41, + "learning_rate": 1.9910713694010973e-06, + "loss": 0.0025, + "step": 48966 + }, + { + "epoch": 6.41, + "learning_rate": 1.9907538179486317e-06, + "loss": 0.001, + "step": 48967 + }, + { + "epoch": 6.41, + "learning_rate": 1.990436289021743e-06, + "loss": 0.0012, + "step": 48968 + }, + { + "epoch": 6.41, + "learning_rate": 1.9901187826213266e-06, + "loss": 0.0021, + "step": 48969 + }, + { + "epoch": 6.41, + "learning_rate": 1.9898012987482683e-06, + "loss": 0.0022, + "step": 48970 + }, + { + "epoch": 6.41, + "learning_rate": 1.9894838374034743e-06, + "loss": 0.0023, + "step": 48971 + }, + { + "epoch": 6.41, + "learning_rate": 1.9891663985878284e-06, + "loss": 0.002, + "step": 48972 + }, + { + "epoch": 6.42, + "learning_rate": 1.9888489823022272e-06, + "loss": 0.0019, + "step": 48973 + }, + { + "epoch": 6.42, + "learning_rate": 1.9885315885475576e-06, + "loss": 0.0022, + "step": 48974 + }, + { + "epoch": 6.42, + "learning_rate": 1.98821421732472e-06, + "loss": 0.0026, + "step": 48975 + }, + { + "epoch": 6.42, + "learning_rate": 1.9878968686346023e-06, + "loss": 0.0016, + "step": 48976 + }, + { + "epoch": 6.42, + "learning_rate": 1.9875795424780954e-06, + "loss": 0.0021, + "step": 48977 + }, + { + "epoch": 6.42, + "learning_rate": 1.987262238856097e-06, + "loss": 0.0033, + "step": 48978 + }, + { + "epoch": 6.42, + "learning_rate": 1.9869449577694976e-06, + "loss": 0.0012, + "step": 48979 + }, + { + "epoch": 6.42, + "learning_rate": 1.986627699219188e-06, + "loss": 0.0013, + "step": 48980 + }, + { + "epoch": 6.42, + "learning_rate": 1.9863104632060593e-06, + "loss": 0.0021, + "step": 48981 + }, + { + "epoch": 6.42, + "learning_rate": 1.985993249731005e-06, + "loss": 0.0024, + "step": 48982 + }, + { + "epoch": 6.42, + "learning_rate": 1.985676058794921e-06, + "loss": 0.0021, + "step": 48983 + }, + { + "epoch": 6.42, + "learning_rate": 1.9853588903986965e-06, + "loss": 0.0011, + "step": 48984 + }, + { + "epoch": 6.42, + "learning_rate": 1.9850417445432203e-06, + "loss": 0.0019, + "step": 48985 + }, + { + "epoch": 6.42, + "learning_rate": 1.98472462122939e-06, + "loss": 0.0026, + "step": 48986 + }, + { + "epoch": 6.42, + "learning_rate": 1.9844075204580947e-06, + "loss": 0.0019, + "step": 48987 + }, + { + "epoch": 6.42, + "learning_rate": 1.9840904422302266e-06, + "loss": 0.0021, + "step": 48988 + }, + { + "epoch": 6.42, + "learning_rate": 1.9837733865466744e-06, + "loss": 0.0018, + "step": 48989 + }, + { + "epoch": 6.42, + "learning_rate": 1.983456353408336e-06, + "loss": 0.0019, + "step": 48990 + }, + { + "epoch": 6.42, + "learning_rate": 1.9831393428160996e-06, + "loss": 0.0021, + "step": 48991 + }, + { + "epoch": 6.42, + "learning_rate": 1.9828223547708536e-06, + "loss": 0.0019, + "step": 48992 + }, + { + "epoch": 6.42, + "learning_rate": 1.9825053892734934e-06, + "loss": 0.0017, + "step": 48993 + }, + { + "epoch": 6.42, + "learning_rate": 1.9821884463249117e-06, + "loss": 0.0018, + "step": 48994 + }, + { + "epoch": 6.42, + "learning_rate": 1.9818715259259993e-06, + "loss": 0.0038, + "step": 48995 + }, + { + "epoch": 6.42, + "learning_rate": 1.9815546280776445e-06, + "loss": 0.0014, + "step": 48996 + }, + { + "epoch": 6.42, + "learning_rate": 1.981237752780739e-06, + "loss": 0.0008, + "step": 48997 + }, + { + "epoch": 6.42, + "learning_rate": 1.980920900036177e-06, + "loss": 0.0018, + "step": 48998 + }, + { + "epoch": 6.42, + "learning_rate": 1.9806040698448493e-06, + "loss": 0.0024, + "step": 48999 + }, + { + "epoch": 6.42, + "learning_rate": 1.9802872622076417e-06, + "loss": 0.0018, + "step": 49000 + }, + { + "epoch": 6.42, + "learning_rate": 1.9799704771254524e-06, + "loss": 0.0005, + "step": 49001 + }, + { + "epoch": 6.42, + "learning_rate": 1.9796537145991677e-06, + "loss": 0.0016, + "step": 49002 + }, + { + "epoch": 6.42, + "learning_rate": 1.9793369746296787e-06, + "loss": 0.0024, + "step": 49003 + }, + { + "epoch": 6.42, + "learning_rate": 1.9790202572178795e-06, + "loss": 0.0022, + "step": 49004 + }, + { + "epoch": 6.42, + "learning_rate": 1.978703562364657e-06, + "loss": 0.002, + "step": 49005 + }, + { + "epoch": 6.42, + "learning_rate": 1.9783868900709048e-06, + "loss": 0.0028, + "step": 49006 + }, + { + "epoch": 6.42, + "learning_rate": 1.9780702403375128e-06, + "loss": 0.0017, + "step": 49007 + }, + { + "epoch": 6.42, + "learning_rate": 1.9777536131653685e-06, + "loss": 0.0011, + "step": 49008 + }, + { + "epoch": 6.42, + "learning_rate": 1.9774370085553672e-06, + "loss": 0.0019, + "step": 49009 + }, + { + "epoch": 6.42, + "learning_rate": 1.9771204265083975e-06, + "loss": 0.0026, + "step": 49010 + }, + { + "epoch": 6.42, + "learning_rate": 1.9768038670253487e-06, + "loss": 0.0013, + "step": 49011 + }, + { + "epoch": 6.42, + "learning_rate": 1.97648733010711e-06, + "loss": 0.0033, + "step": 49012 + }, + { + "epoch": 6.42, + "learning_rate": 1.9761708157545754e-06, + "loss": 0.0017, + "step": 49013 + }, + { + "epoch": 6.42, + "learning_rate": 1.9758543239686313e-06, + "loss": 0.0027, + "step": 49014 + }, + { + "epoch": 6.42, + "learning_rate": 1.975537854750171e-06, + "loss": 0.0015, + "step": 49015 + }, + { + "epoch": 6.42, + "learning_rate": 1.975221408100082e-06, + "loss": 0.0014, + "step": 49016 + }, + { + "epoch": 6.42, + "learning_rate": 1.974904984019257e-06, + "loss": 0.0025, + "step": 49017 + }, + { + "epoch": 6.42, + "learning_rate": 1.9745885825085844e-06, + "loss": 0.0014, + "step": 49018 + }, + { + "epoch": 6.42, + "learning_rate": 1.9742722035689545e-06, + "loss": 0.002, + "step": 49019 + }, + { + "epoch": 6.42, + "learning_rate": 1.9739558472012543e-06, + "loss": 0.0024, + "step": 49020 + }, + { + "epoch": 6.42, + "learning_rate": 1.9736395134063777e-06, + "loss": 0.001, + "step": 49021 + }, + { + "epoch": 6.42, + "learning_rate": 1.973323202185212e-06, + "loss": 0.0024, + "step": 49022 + }, + { + "epoch": 6.42, + "learning_rate": 1.9730069135386464e-06, + "loss": 0.0023, + "step": 49023 + }, + { + "epoch": 6.42, + "learning_rate": 1.9726906474675734e-06, + "loss": 0.002, + "step": 49024 + }, + { + "epoch": 6.42, + "learning_rate": 1.9723744039728774e-06, + "loss": 0.002, + "step": 49025 + }, + { + "epoch": 6.42, + "learning_rate": 1.972058183055454e-06, + "loss": 0.0037, + "step": 49026 + }, + { + "epoch": 6.42, + "learning_rate": 1.9717419847161866e-06, + "loss": 0.0025, + "step": 49027 + }, + { + "epoch": 6.42, + "learning_rate": 1.971425808955969e-06, + "loss": 0.0018, + "step": 49028 + }, + { + "epoch": 6.42, + "learning_rate": 1.9711096557756904e-06, + "loss": 0.002, + "step": 49029 + }, + { + "epoch": 6.42, + "learning_rate": 1.970793525176238e-06, + "loss": 0.0015, + "step": 49030 + }, + { + "epoch": 6.42, + "learning_rate": 1.9704774171584974e-06, + "loss": 0.002, + "step": 49031 + }, + { + "epoch": 6.42, + "learning_rate": 1.9701613317233647e-06, + "loss": 0.0019, + "step": 49032 + }, + { + "epoch": 6.42, + "learning_rate": 1.969845268871726e-06, + "loss": 0.0018, + "step": 49033 + }, + { + "epoch": 6.42, + "learning_rate": 1.96952922860447e-06, + "loss": 0.0015, + "step": 49034 + }, + { + "epoch": 6.42, + "learning_rate": 1.969213210922483e-06, + "loss": 0.0025, + "step": 49035 + }, + { + "epoch": 6.42, + "learning_rate": 1.968897215826655e-06, + "loss": 0.0012, + "step": 49036 + }, + { + "epoch": 6.42, + "learning_rate": 1.9685812433178795e-06, + "loss": 0.0029, + "step": 49037 + }, + { + "epoch": 6.42, + "learning_rate": 1.9682652933970412e-06, + "loss": 0.0014, + "step": 49038 + }, + { + "epoch": 6.42, + "learning_rate": 1.9679493660650263e-06, + "loss": 0.0022, + "step": 49039 + }, + { + "epoch": 6.42, + "learning_rate": 1.9676334613227286e-06, + "loss": 0.0013, + "step": 49040 + }, + { + "epoch": 6.42, + "learning_rate": 1.9673175791710343e-06, + "loss": 0.0027, + "step": 49041 + }, + { + "epoch": 6.42, + "learning_rate": 1.9670017196108293e-06, + "loss": 0.0023, + "step": 49042 + }, + { + "epoch": 6.42, + "learning_rate": 1.966685882643007e-06, + "loss": 0.0011, + "step": 49043 + }, + { + "epoch": 6.42, + "learning_rate": 1.9663700682684516e-06, + "loss": 0.002, + "step": 49044 + }, + { + "epoch": 6.42, + "learning_rate": 1.966054276488053e-06, + "loss": 0.0017, + "step": 49045 + }, + { + "epoch": 6.42, + "learning_rate": 1.965738507302697e-06, + "loss": 0.0027, + "step": 49046 + }, + { + "epoch": 6.42, + "learning_rate": 1.9654227607132727e-06, + "loss": 0.003, + "step": 49047 + }, + { + "epoch": 6.42, + "learning_rate": 1.9651070367206727e-06, + "loss": 0.0022, + "step": 49048 + }, + { + "epoch": 6.43, + "learning_rate": 1.964791335325781e-06, + "loss": 0.0015, + "step": 49049 + }, + { + "epoch": 6.43, + "learning_rate": 1.964475656529482e-06, + "loss": 0.0017, + "step": 49050 + }, + { + "epoch": 6.43, + "learning_rate": 1.9641600003326713e-06, + "loss": 0.0013, + "step": 49051 + }, + { + "epoch": 6.43, + "learning_rate": 1.9638443667362317e-06, + "loss": 0.0016, + "step": 49052 + }, + { + "epoch": 6.43, + "learning_rate": 1.963528755741052e-06, + "loss": 0.0014, + "step": 49053 + }, + { + "epoch": 6.43, + "learning_rate": 1.963213167348018e-06, + "loss": 0.0033, + "step": 49054 + }, + { + "epoch": 6.43, + "learning_rate": 1.96289760155802e-06, + "loss": 0.0011, + "step": 49055 + }, + { + "epoch": 6.43, + "learning_rate": 1.9625820583719457e-06, + "loss": 0.0032, + "step": 49056 + }, + { + "epoch": 6.43, + "learning_rate": 1.9622665377906793e-06, + "loss": 0.0012, + "step": 49057 + }, + { + "epoch": 6.43, + "learning_rate": 1.9619510398151087e-06, + "loss": 0.0009, + "step": 49058 + }, + { + "epoch": 6.43, + "learning_rate": 1.961635564446126e-06, + "loss": 0.0012, + "step": 49059 + }, + { + "epoch": 6.43, + "learning_rate": 1.961320111684616e-06, + "loss": 0.0014, + "step": 49060 + }, + { + "epoch": 6.43, + "learning_rate": 1.9610046815314643e-06, + "loss": 0.0016, + "step": 49061 + }, + { + "epoch": 6.43, + "learning_rate": 1.9606892739875562e-06, + "loss": 0.0011, + "step": 49062 + }, + { + "epoch": 6.43, + "learning_rate": 1.960373889053784e-06, + "loss": 0.0033, + "step": 49063 + }, + { + "epoch": 6.43, + "learning_rate": 1.9600585267310324e-06, + "loss": 0.003, + "step": 49064 + }, + { + "epoch": 6.43, + "learning_rate": 1.9597431870201845e-06, + "loss": 0.0021, + "step": 49065 + }, + { + "epoch": 6.43, + "learning_rate": 1.9594278699221347e-06, + "loss": 0.0025, + "step": 49066 + }, + { + "epoch": 6.43, + "learning_rate": 1.959112575437765e-06, + "loss": 0.0011, + "step": 49067 + }, + { + "epoch": 6.43, + "learning_rate": 1.958797303567963e-06, + "loss": 0.0013, + "step": 49068 + }, + { + "epoch": 6.43, + "learning_rate": 1.958482054313613e-06, + "loss": 0.0016, + "step": 49069 + }, + { + "epoch": 6.43, + "learning_rate": 1.958166827675604e-06, + "loss": 0.0021, + "step": 49070 + }, + { + "epoch": 6.43, + "learning_rate": 1.957851623654825e-06, + "loss": 0.0032, + "step": 49071 + }, + { + "epoch": 6.43, + "learning_rate": 1.95753644225216e-06, + "loss": 0.0039, + "step": 49072 + }, + { + "epoch": 6.43, + "learning_rate": 1.9572212834684934e-06, + "loss": 0.0021, + "step": 49073 + }, + { + "epoch": 6.43, + "learning_rate": 1.9569061473047157e-06, + "loss": 0.0037, + "step": 49074 + }, + { + "epoch": 6.43, + "learning_rate": 1.956591033761711e-06, + "loss": 0.0018, + "step": 49075 + }, + { + "epoch": 6.43, + "learning_rate": 1.9562759428403655e-06, + "loss": 0.0014, + "step": 49076 + }, + { + "epoch": 6.43, + "learning_rate": 1.955960874541564e-06, + "loss": 0.0016, + "step": 49077 + }, + { + "epoch": 6.43, + "learning_rate": 1.955645828866196e-06, + "loss": 0.0027, + "step": 49078 + }, + { + "epoch": 6.43, + "learning_rate": 1.9553308058151434e-06, + "loss": 0.0016, + "step": 49079 + }, + { + "epoch": 6.43, + "learning_rate": 1.9550158053892963e-06, + "loss": 0.0027, + "step": 49080 + }, + { + "epoch": 6.43, + "learning_rate": 1.9547008275895373e-06, + "loss": 0.002, + "step": 49081 + }, + { + "epoch": 6.43, + "learning_rate": 1.954385872416756e-06, + "loss": 0.0022, + "step": 49082 + }, + { + "epoch": 6.43, + "learning_rate": 1.954070939871836e-06, + "loss": 0.0018, + "step": 49083 + }, + { + "epoch": 6.43, + "learning_rate": 1.9537560299556635e-06, + "loss": 0.0026, + "step": 49084 + }, + { + "epoch": 6.43, + "learning_rate": 1.95344114266912e-06, + "loss": 0.0016, + "step": 49085 + }, + { + "epoch": 6.43, + "learning_rate": 1.953126278013098e-06, + "loss": 0.0023, + "step": 49086 + }, + { + "epoch": 6.43, + "learning_rate": 1.9528114359884797e-06, + "loss": 0.0018, + "step": 49087 + }, + { + "epoch": 6.43, + "learning_rate": 1.9524966165961477e-06, + "loss": 0.0012, + "step": 49088 + }, + { + "epoch": 6.43, + "learning_rate": 1.9521818198369947e-06, + "loss": 0.0012, + "step": 49089 + }, + { + "epoch": 6.43, + "learning_rate": 1.951867045711897e-06, + "loss": 0.0022, + "step": 49090 + }, + { + "epoch": 6.43, + "learning_rate": 1.951552294221749e-06, + "loss": 0.0023, + "step": 49091 + }, + { + "epoch": 6.43, + "learning_rate": 1.9512375653674275e-06, + "loss": 0.0026, + "step": 49092 + }, + { + "epoch": 6.43, + "learning_rate": 1.950922859149825e-06, + "loss": 0.0018, + "step": 49093 + }, + { + "epoch": 6.43, + "learning_rate": 1.9506081755698238e-06, + "loss": 0.002, + "step": 49094 + }, + { + "epoch": 6.43, + "learning_rate": 1.9502935146283076e-06, + "loss": 0.0021, + "step": 49095 + }, + { + "epoch": 6.43, + "learning_rate": 1.9499788763261598e-06, + "loss": 0.0016, + "step": 49096 + }, + { + "epoch": 6.43, + "learning_rate": 1.94966426066427e-06, + "loss": 0.0031, + "step": 49097 + }, + { + "epoch": 6.43, + "learning_rate": 1.949349667643522e-06, + "loss": 0.002, + "step": 49098 + }, + { + "epoch": 6.43, + "learning_rate": 1.9490350972647986e-06, + "loss": 0.0014, + "step": 49099 + }, + { + "epoch": 6.43, + "learning_rate": 1.948720549528983e-06, + "loss": 0.0025, + "step": 49100 + }, + { + "epoch": 6.43, + "learning_rate": 1.948406024436962e-06, + "loss": 0.0024, + "step": 49101 + }, + { + "epoch": 6.43, + "learning_rate": 1.9480915219896226e-06, + "loss": 0.0027, + "step": 49102 + }, + { + "epoch": 6.43, + "learning_rate": 1.9477770421878462e-06, + "loss": 0.0023, + "step": 49103 + }, + { + "epoch": 6.43, + "learning_rate": 1.947462585032517e-06, + "loss": 0.0012, + "step": 49104 + }, + { + "epoch": 6.43, + "learning_rate": 1.9471481505245217e-06, + "loss": 0.0011, + "step": 49105 + }, + { + "epoch": 6.43, + "learning_rate": 1.946833738664744e-06, + "loss": 0.0017, + "step": 49106 + }, + { + "epoch": 6.43, + "learning_rate": 1.946519349454065e-06, + "loss": 0.0011, + "step": 49107 + }, + { + "epoch": 6.43, + "learning_rate": 1.946204982893374e-06, + "loss": 0.0029, + "step": 49108 + }, + { + "epoch": 6.43, + "learning_rate": 1.945890638983553e-06, + "loss": 0.0019, + "step": 49109 + }, + { + "epoch": 6.43, + "learning_rate": 1.945576317725485e-06, + "loss": 0.0011, + "step": 49110 + }, + { + "epoch": 6.43, + "learning_rate": 1.9452620191200523e-06, + "loss": 0.0024, + "step": 49111 + }, + { + "epoch": 6.43, + "learning_rate": 1.944947743168142e-06, + "loss": 0.0016, + "step": 49112 + }, + { + "epoch": 6.43, + "learning_rate": 1.9446334898706387e-06, + "loss": 0.0024, + "step": 49113 + }, + { + "epoch": 6.43, + "learning_rate": 1.9443192592284256e-06, + "loss": 0.0018, + "step": 49114 + }, + { + "epoch": 6.43, + "learning_rate": 1.9440050512423824e-06, + "loss": 0.0018, + "step": 49115 + }, + { + "epoch": 6.43, + "learning_rate": 1.943690865913398e-06, + "loss": 0.0032, + "step": 49116 + }, + { + "epoch": 6.43, + "learning_rate": 1.943376703242356e-06, + "loss": 0.0022, + "step": 49117 + }, + { + "epoch": 6.43, + "learning_rate": 1.943062563230136e-06, + "loss": 0.0018, + "step": 49118 + }, + { + "epoch": 6.43, + "learning_rate": 1.9427484458776213e-06, + "loss": 0.003, + "step": 49119 + }, + { + "epoch": 6.43, + "learning_rate": 1.9424343511857e-06, + "loss": 0.0021, + "step": 49120 + }, + { + "epoch": 6.43, + "learning_rate": 1.942120279155254e-06, + "loss": 0.002, + "step": 49121 + }, + { + "epoch": 6.43, + "learning_rate": 1.941806229787162e-06, + "loss": 0.0023, + "step": 49122 + }, + { + "epoch": 6.43, + "learning_rate": 1.9414922030823114e-06, + "loss": 0.0023, + "step": 49123 + }, + { + "epoch": 6.43, + "learning_rate": 1.941178199041588e-06, + "loss": 0.0024, + "step": 49124 + }, + { + "epoch": 6.44, + "learning_rate": 1.940864217665871e-06, + "loss": 0.0017, + "step": 49125 + }, + { + "epoch": 6.44, + "learning_rate": 1.9405502589560443e-06, + "loss": 0.0016, + "step": 49126 + }, + { + "epoch": 6.44, + "learning_rate": 1.9402363229129883e-06, + "loss": 0.0018, + "step": 49127 + }, + { + "epoch": 6.44, + "learning_rate": 1.939922409537591e-06, + "loss": 0.0028, + "step": 49128 + }, + { + "epoch": 6.44, + "learning_rate": 1.9396085188307324e-06, + "loss": 0.0014, + "step": 49129 + }, + { + "epoch": 6.44, + "learning_rate": 1.939294650793293e-06, + "loss": 0.0012, + "step": 49130 + }, + { + "epoch": 6.44, + "learning_rate": 1.9389808054261615e-06, + "loss": 0.0018, + "step": 49131 + }, + { + "epoch": 6.44, + "learning_rate": 1.938666982730216e-06, + "loss": 0.0017, + "step": 49132 + }, + { + "epoch": 6.44, + "learning_rate": 1.9383531827063418e-06, + "loss": 0.0019, + "step": 49133 + }, + { + "epoch": 6.44, + "learning_rate": 1.9380394053554163e-06, + "loss": 0.0024, + "step": 49134 + }, + { + "epoch": 6.44, + "learning_rate": 1.9377256506783258e-06, + "loss": 0.0028, + "step": 49135 + }, + { + "epoch": 6.44, + "learning_rate": 1.9374119186759553e-06, + "loss": 0.0021, + "step": 49136 + }, + { + "epoch": 6.44, + "learning_rate": 1.937098209349184e-06, + "loss": 0.0025, + "step": 49137 + }, + { + "epoch": 6.44, + "learning_rate": 1.9367845226988935e-06, + "loss": 0.0023, + "step": 49138 + }, + { + "epoch": 6.44, + "learning_rate": 1.9364708587259686e-06, + "loss": 0.0026, + "step": 49139 + }, + { + "epoch": 6.44, + "learning_rate": 1.936157217431289e-06, + "loss": 0.0011, + "step": 49140 + }, + { + "epoch": 6.44, + "learning_rate": 1.93584359881574e-06, + "loss": 0.0027, + "step": 49141 + }, + { + "epoch": 6.44, + "learning_rate": 1.935530002880197e-06, + "loss": 0.0013, + "step": 49142 + }, + { + "epoch": 6.44, + "learning_rate": 1.9352164296255502e-06, + "loss": 0.0017, + "step": 49143 + }, + { + "epoch": 6.44, + "learning_rate": 1.934902879052677e-06, + "loss": 0.0019, + "step": 49144 + }, + { + "epoch": 6.44, + "learning_rate": 1.934589351162458e-06, + "loss": 0.0019, + "step": 49145 + }, + { + "epoch": 6.44, + "learning_rate": 1.9342758459557765e-06, + "loss": 0.0037, + "step": 49146 + }, + { + "epoch": 6.44, + "learning_rate": 1.933962363433517e-06, + "loss": 0.0019, + "step": 49147 + }, + { + "epoch": 6.44, + "learning_rate": 1.933648903596559e-06, + "loss": 0.0024, + "step": 49148 + }, + { + "epoch": 6.44, + "learning_rate": 1.9333354664457836e-06, + "loss": 0.0028, + "step": 49149 + }, + { + "epoch": 6.44, + "learning_rate": 1.9330220519820708e-06, + "loss": 0.0026, + "step": 49150 + }, + { + "epoch": 6.44, + "learning_rate": 1.932708660206306e-06, + "loss": 0.0017, + "step": 49151 + }, + { + "epoch": 6.44, + "learning_rate": 1.932395291119369e-06, + "loss": 0.0016, + "step": 49152 + }, + { + "epoch": 6.44, + "learning_rate": 1.9320819447221373e-06, + "loss": 0.0029, + "step": 49153 + }, + { + "epoch": 6.44, + "learning_rate": 1.9317686210154984e-06, + "loss": 0.0023, + "step": 49154 + }, + { + "epoch": 6.44, + "learning_rate": 1.9314553200003305e-06, + "loss": 0.0007, + "step": 49155 + }, + { + "epoch": 6.44, + "learning_rate": 1.931142041677513e-06, + "loss": 0.005, + "step": 49156 + }, + { + "epoch": 6.44, + "learning_rate": 1.930828786047928e-06, + "loss": 0.0025, + "step": 49157 + }, + { + "epoch": 6.44, + "learning_rate": 1.9305155531124597e-06, + "loss": 0.0017, + "step": 49158 + }, + { + "epoch": 6.44, + "learning_rate": 1.9302023428719874e-06, + "loss": 0.002, + "step": 49159 + }, + { + "epoch": 6.44, + "learning_rate": 1.9298891553273915e-06, + "loss": 0.0013, + "step": 49160 + }, + { + "epoch": 6.44, + "learning_rate": 1.9295759904795486e-06, + "loss": 0.0019, + "step": 49161 + }, + { + "epoch": 6.44, + "learning_rate": 1.9292628483293474e-06, + "loss": 0.0026, + "step": 49162 + }, + { + "epoch": 6.44, + "learning_rate": 1.9289497288776636e-06, + "loss": 0.0021, + "step": 49163 + }, + { + "epoch": 6.44, + "learning_rate": 1.928636632125377e-06, + "loss": 0.0021, + "step": 49164 + }, + { + "epoch": 6.44, + "learning_rate": 1.928323558073373e-06, + "loss": 0.0025, + "step": 49165 + }, + { + "epoch": 6.44, + "learning_rate": 1.9280105067225286e-06, + "loss": 0.0012, + "step": 49166 + }, + { + "epoch": 6.44, + "learning_rate": 1.9276974780737225e-06, + "loss": 0.0019, + "step": 49167 + }, + { + "epoch": 6.44, + "learning_rate": 1.927384472127839e-06, + "loss": 0.0023, + "step": 49168 + }, + { + "epoch": 6.44, + "learning_rate": 1.9270714888857543e-06, + "loss": 0.0018, + "step": 49169 + }, + { + "epoch": 6.44, + "learning_rate": 1.926758528348355e-06, + "loss": 0.0023, + "step": 49170 + }, + { + "epoch": 6.44, + "learning_rate": 1.9264455905165168e-06, + "loss": 0.0018, + "step": 49171 + }, + { + "epoch": 6.44, + "learning_rate": 1.926132675391118e-06, + "loss": 0.0017, + "step": 49172 + }, + { + "epoch": 6.44, + "learning_rate": 1.925819782973043e-06, + "loss": 0.0018, + "step": 49173 + }, + { + "epoch": 6.44, + "learning_rate": 1.9255069132631698e-06, + "loss": 0.0012, + "step": 49174 + }, + { + "epoch": 6.44, + "learning_rate": 1.925194066262379e-06, + "loss": 0.0015, + "step": 49175 + }, + { + "epoch": 6.44, + "learning_rate": 1.924881241971548e-06, + "loss": 0.0014, + "step": 49176 + }, + { + "epoch": 6.44, + "learning_rate": 1.92456844039156e-06, + "loss": 0.0028, + "step": 49177 + }, + { + "epoch": 6.44, + "learning_rate": 1.924255661523291e-06, + "loss": 0.0033, + "step": 49178 + }, + { + "epoch": 6.44, + "learning_rate": 1.923942905367625e-06, + "loss": 0.0019, + "step": 49179 + }, + { + "epoch": 6.44, + "learning_rate": 1.923630171925437e-06, + "loss": 0.0024, + "step": 49180 + }, + { + "epoch": 6.44, + "learning_rate": 1.9233174611976124e-06, + "loss": 0.0023, + "step": 49181 + }, + { + "epoch": 6.44, + "learning_rate": 1.9230047731850266e-06, + "loss": 0.0019, + "step": 49182 + }, + { + "epoch": 6.44, + "learning_rate": 1.922692107888561e-06, + "loss": 0.0018, + "step": 49183 + }, + { + "epoch": 6.44, + "learning_rate": 1.92237946530909e-06, + "loss": 0.0013, + "step": 49184 + }, + { + "epoch": 6.44, + "learning_rate": 1.922066845447499e-06, + "loss": 0.0024, + "step": 49185 + }, + { + "epoch": 6.44, + "learning_rate": 1.9217542483046646e-06, + "loss": 0.0019, + "step": 49186 + }, + { + "epoch": 6.44, + "learning_rate": 1.9214416738814646e-06, + "loss": 0.0019, + "step": 49187 + }, + { + "epoch": 6.44, + "learning_rate": 1.921129122178782e-06, + "loss": 0.003, + "step": 49188 + }, + { + "epoch": 6.44, + "learning_rate": 1.920816593197491e-06, + "loss": 0.0014, + "step": 49189 + }, + { + "epoch": 6.44, + "learning_rate": 1.920504086938475e-06, + "loss": 0.0046, + "step": 49190 + }, + { + "epoch": 6.44, + "learning_rate": 1.9201916034026113e-06, + "loss": 0.002, + "step": 49191 + }, + { + "epoch": 6.44, + "learning_rate": 1.919879142590776e-06, + "loss": 0.0013, + "step": 49192 + }, + { + "epoch": 6.44, + "learning_rate": 1.9195667045038526e-06, + "loss": 0.002, + "step": 49193 + }, + { + "epoch": 6.44, + "learning_rate": 1.919254289142718e-06, + "loss": 0.0015, + "step": 49194 + }, + { + "epoch": 6.44, + "learning_rate": 1.918941896508247e-06, + "loss": 0.0021, + "step": 49195 + }, + { + "epoch": 6.44, + "learning_rate": 1.9186295266013255e-06, + "loss": 0.0021, + "step": 49196 + }, + { + "epoch": 6.44, + "learning_rate": 1.918317179422827e-06, + "loss": 0.0031, + "step": 49197 + }, + { + "epoch": 6.44, + "learning_rate": 1.918004854973632e-06, + "loss": 0.0014, + "step": 49198 + }, + { + "epoch": 6.44, + "learning_rate": 1.917692553254614e-06, + "loss": 0.0017, + "step": 49199 + }, + { + "epoch": 6.44, + "learning_rate": 1.917380274266657e-06, + "loss": 0.0026, + "step": 49200 + }, + { + "epoch": 6.44, + "learning_rate": 1.9170680180106392e-06, + "loss": 0.0013, + "step": 49201 + }, + { + "epoch": 6.45, + "learning_rate": 1.9167557844874375e-06, + "loss": 0.0012, + "step": 49202 + }, + { + "epoch": 6.45, + "learning_rate": 1.916443573697926e-06, + "loss": 0.0021, + "step": 49203 + }, + { + "epoch": 6.45, + "learning_rate": 1.91613138564299e-06, + "loss": 0.0013, + "step": 49204 + }, + { + "epoch": 6.45, + "learning_rate": 1.9158192203235037e-06, + "loss": 0.0031, + "step": 49205 + }, + { + "epoch": 6.45, + "learning_rate": 1.915507077740346e-06, + "loss": 0.0019, + "step": 49206 + }, + { + "epoch": 6.45, + "learning_rate": 1.9151949578943907e-06, + "loss": 0.0021, + "step": 49207 + }, + { + "epoch": 6.45, + "learning_rate": 1.9148828607865223e-06, + "loss": 0.0021, + "step": 49208 + }, + { + "epoch": 6.45, + "learning_rate": 1.9145707864176143e-06, + "loss": 0.0019, + "step": 49209 + }, + { + "epoch": 6.45, + "learning_rate": 1.9142587347885433e-06, + "loss": 0.0032, + "step": 49210 + }, + { + "epoch": 6.45, + "learning_rate": 1.9139467059001893e-06, + "loss": 0.0014, + "step": 49211 + }, + { + "epoch": 6.45, + "learning_rate": 1.913634699753433e-06, + "loss": 0.003, + "step": 49212 + }, + { + "epoch": 6.45, + "learning_rate": 1.9133227163491463e-06, + "loss": 0.0025, + "step": 49213 + }, + { + "epoch": 6.45, + "learning_rate": 1.9130107556882082e-06, + "loss": 0.0012, + "step": 49214 + }, + { + "epoch": 6.45, + "learning_rate": 1.9126988177714977e-06, + "loss": 0.0017, + "step": 49215 + }, + { + "epoch": 6.45, + "learning_rate": 1.912386902599892e-06, + "loss": 0.0012, + "step": 49216 + }, + { + "epoch": 6.45, + "learning_rate": 1.912075010174268e-06, + "loss": 0.003, + "step": 49217 + }, + { + "epoch": 6.45, + "learning_rate": 1.9117631404954984e-06, + "loss": 0.002, + "step": 49218 + }, + { + "epoch": 6.45, + "learning_rate": 1.911451293564468e-06, + "loss": 0.0032, + "step": 49219 + }, + { + "epoch": 6.45, + "learning_rate": 1.9111394693820497e-06, + "loss": 0.0024, + "step": 49220 + }, + { + "epoch": 6.45, + "learning_rate": 1.910827667949118e-06, + "loss": 0.0021, + "step": 49221 + }, + { + "epoch": 6.45, + "learning_rate": 1.9105158892665533e-06, + "loss": 0.0035, + "step": 49222 + }, + { + "epoch": 6.45, + "learning_rate": 1.910204133335234e-06, + "loss": 0.0023, + "step": 49223 + }, + { + "epoch": 6.45, + "learning_rate": 1.909892400156035e-06, + "loss": 0.0013, + "step": 49224 + }, + { + "epoch": 6.45, + "learning_rate": 1.909580689729832e-06, + "loss": 0.0022, + "step": 49225 + }, + { + "epoch": 6.45, + "learning_rate": 1.909269002057501e-06, + "loss": 0.0015, + "step": 49226 + }, + { + "epoch": 6.45, + "learning_rate": 1.9089573371399218e-06, + "loss": 0.0023, + "step": 49227 + }, + { + "epoch": 6.45, + "learning_rate": 1.9086456949779696e-06, + "loss": 0.0014, + "step": 49228 + }, + { + "epoch": 6.45, + "learning_rate": 1.908334075572519e-06, + "loss": 0.0028, + "step": 49229 + }, + { + "epoch": 6.45, + "learning_rate": 1.908022478924448e-06, + "loss": 0.0018, + "step": 49230 + }, + { + "epoch": 6.45, + "learning_rate": 1.907710905034635e-06, + "loss": 0.0016, + "step": 49231 + }, + { + "epoch": 6.45, + "learning_rate": 1.9073993539039504e-06, + "loss": 0.001, + "step": 49232 + }, + { + "epoch": 6.45, + "learning_rate": 1.907087825533278e-06, + "loss": 0.0024, + "step": 49233 + }, + { + "epoch": 6.45, + "learning_rate": 1.9067763199234868e-06, + "loss": 0.0021, + "step": 49234 + }, + { + "epoch": 6.45, + "learning_rate": 1.9064648370754578e-06, + "loss": 0.0024, + "step": 49235 + }, + { + "epoch": 6.45, + "learning_rate": 1.9061533769900664e-06, + "loss": 0.002, + "step": 49236 + }, + { + "epoch": 6.45, + "learning_rate": 1.905841939668185e-06, + "loss": 0.0024, + "step": 49237 + }, + { + "epoch": 6.45, + "learning_rate": 1.9055305251106948e-06, + "loss": 0.0012, + "step": 49238 + }, + { + "epoch": 6.45, + "learning_rate": 1.9052191333184688e-06, + "loss": 0.0016, + "step": 49239 + }, + { + "epoch": 6.45, + "learning_rate": 1.904907764292383e-06, + "loss": 0.0018, + "step": 49240 + }, + { + "epoch": 6.45, + "learning_rate": 1.904596418033311e-06, + "loss": 0.0017, + "step": 49241 + }, + { + "epoch": 6.45, + "learning_rate": 1.904285094542132e-06, + "loss": 0.0012, + "step": 49242 + }, + { + "epoch": 6.45, + "learning_rate": 1.9039737938197178e-06, + "loss": 0.002, + "step": 49243 + }, + { + "epoch": 6.45, + "learning_rate": 1.9036625158669486e-06, + "loss": 0.003, + "step": 49244 + }, + { + "epoch": 6.45, + "learning_rate": 1.903351260684695e-06, + "loss": 0.0017, + "step": 49245 + }, + { + "epoch": 6.45, + "learning_rate": 1.903040028273838e-06, + "loss": 0.0006, + "step": 49246 + }, + { + "epoch": 6.45, + "learning_rate": 1.9027288186352488e-06, + "loss": 0.0023, + "step": 49247 + }, + { + "epoch": 6.45, + "learning_rate": 1.902417631769804e-06, + "loss": 0.0011, + "step": 49248 + }, + { + "epoch": 6.45, + "learning_rate": 1.9021064676783762e-06, + "loss": 0.0025, + "step": 49249 + }, + { + "epoch": 6.45, + "learning_rate": 1.901795326361845e-06, + "loss": 0.002, + "step": 49250 + }, + { + "epoch": 6.45, + "learning_rate": 1.9014842078210827e-06, + "loss": 0.0022, + "step": 49251 + }, + { + "epoch": 6.45, + "learning_rate": 1.901173112056962e-06, + "loss": 0.0015, + "step": 49252 + }, + { + "epoch": 6.45, + "learning_rate": 1.9008620390703646e-06, + "loss": 0.002, + "step": 49253 + }, + { + "epoch": 6.45, + "learning_rate": 1.9005509888621575e-06, + "loss": 0.0015, + "step": 49254 + }, + { + "epoch": 6.45, + "learning_rate": 1.9002399614332222e-06, + "loss": 0.0026, + "step": 49255 + }, + { + "epoch": 6.45, + "learning_rate": 1.8999289567844315e-06, + "loss": 0.0014, + "step": 49256 + }, + { + "epoch": 6.45, + "learning_rate": 1.8996179749166565e-06, + "loss": 0.0021, + "step": 49257 + }, + { + "epoch": 6.45, + "learning_rate": 1.899307015830777e-06, + "loss": 0.0018, + "step": 49258 + }, + { + "epoch": 6.45, + "learning_rate": 1.898996079527664e-06, + "loss": 0.0017, + "step": 49259 + }, + { + "epoch": 6.45, + "learning_rate": 1.8986851660081918e-06, + "loss": 0.0026, + "step": 49260 + }, + { + "epoch": 6.45, + "learning_rate": 1.8983742752732393e-06, + "loss": 0.0023, + "step": 49261 + }, + { + "epoch": 6.45, + "learning_rate": 1.8980634073236759e-06, + "loss": 0.0035, + "step": 49262 + }, + { + "epoch": 6.45, + "learning_rate": 1.8977525621603786e-06, + "loss": 0.0031, + "step": 49263 + }, + { + "epoch": 6.45, + "learning_rate": 1.8974417397842182e-06, + "loss": 0.0025, + "step": 49264 + }, + { + "epoch": 6.45, + "learning_rate": 1.8971309401960714e-06, + "loss": 0.0017, + "step": 49265 + }, + { + "epoch": 6.45, + "learning_rate": 1.8968201633968152e-06, + "loss": 0.0025, + "step": 49266 + }, + { + "epoch": 6.45, + "learning_rate": 1.8965094093873203e-06, + "loss": 0.0028, + "step": 49267 + }, + { + "epoch": 6.45, + "learning_rate": 1.8961986781684594e-06, + "loss": 0.0029, + "step": 49268 + }, + { + "epoch": 6.45, + "learning_rate": 1.8958879697411104e-06, + "loss": 0.0017, + "step": 49269 + }, + { + "epoch": 6.45, + "learning_rate": 1.8955772841061447e-06, + "loss": 0.0024, + "step": 49270 + }, + { + "epoch": 6.45, + "learning_rate": 1.8952666212644355e-06, + "loss": 0.0008, + "step": 49271 + }, + { + "epoch": 6.45, + "learning_rate": 1.894955981216856e-06, + "loss": 0.0028, + "step": 49272 + }, + { + "epoch": 6.45, + "learning_rate": 1.8946453639642836e-06, + "loss": 0.0025, + "step": 49273 + }, + { + "epoch": 6.45, + "learning_rate": 1.8943347695075886e-06, + "loss": 0.0012, + "step": 49274 + }, + { + "epoch": 6.45, + "learning_rate": 1.8940241978476437e-06, + "loss": 0.0031, + "step": 49275 + }, + { + "epoch": 6.45, + "learning_rate": 1.8937136489853237e-06, + "loss": 0.0019, + "step": 49276 + }, + { + "epoch": 6.45, + "learning_rate": 1.8934031229215043e-06, + "loss": 0.0015, + "step": 49277 + }, + { + "epoch": 6.46, + "learning_rate": 1.8930926196570586e-06, + "loss": 0.004, + "step": 49278 + }, + { + "epoch": 6.46, + "learning_rate": 1.8927821391928546e-06, + "loss": 0.0013, + "step": 49279 + }, + { + "epoch": 6.46, + "learning_rate": 1.8924716815297716e-06, + "loss": 0.0029, + "step": 49280 + }, + { + "epoch": 6.46, + "learning_rate": 1.89216124666868e-06, + "loss": 0.0018, + "step": 49281 + }, + { + "epoch": 6.46, + "learning_rate": 1.8918508346104537e-06, + "loss": 0.0019, + "step": 49282 + }, + { + "epoch": 6.46, + "learning_rate": 1.8915404453559627e-06, + "loss": 0.0022, + "step": 49283 + }, + { + "epoch": 6.46, + "learning_rate": 1.8912300789060845e-06, + "loss": 0.0031, + "step": 49284 + }, + { + "epoch": 6.46, + "learning_rate": 1.8909197352616892e-06, + "loss": 0.0016, + "step": 49285 + }, + { + "epoch": 6.46, + "learning_rate": 1.890609414423652e-06, + "loss": 0.0018, + "step": 49286 + }, + { + "epoch": 6.46, + "learning_rate": 1.8902991163928375e-06, + "loss": 0.0032, + "step": 49287 + }, + { + "epoch": 6.46, + "learning_rate": 1.8899888411701306e-06, + "loss": 0.002, + "step": 49288 + }, + { + "epoch": 6.46, + "learning_rate": 1.8896785887563984e-06, + "loss": 0.0016, + "step": 49289 + }, + { + "epoch": 6.46, + "learning_rate": 1.8893683591525125e-06, + "loss": 0.0024, + "step": 49290 + }, + { + "epoch": 6.46, + "learning_rate": 1.8890581523593443e-06, + "loss": 0.0015, + "step": 49291 + }, + { + "epoch": 6.46, + "learning_rate": 1.8887479683777699e-06, + "loss": 0.002, + "step": 49292 + }, + { + "epoch": 6.46, + "learning_rate": 1.8884378072086607e-06, + "loss": 0.0019, + "step": 49293 + }, + { + "epoch": 6.46, + "learning_rate": 1.8881276688528849e-06, + "loss": 0.0012, + "step": 49294 + }, + { + "epoch": 6.46, + "learning_rate": 1.8878175533113219e-06, + "loss": 0.0017, + "step": 49295 + }, + { + "epoch": 6.46, + "learning_rate": 1.8875074605848387e-06, + "loss": 0.0018, + "step": 49296 + }, + { + "epoch": 6.46, + "learning_rate": 1.8871973906743091e-06, + "loss": 0.0032, + "step": 49297 + }, + { + "epoch": 6.46, + "learning_rate": 1.8868873435806024e-06, + "loss": 0.0018, + "step": 49298 + }, + { + "epoch": 6.46, + "learning_rate": 1.8865773193045933e-06, + "loss": 0.0031, + "step": 49299 + }, + { + "epoch": 6.46, + "learning_rate": 1.8862673178471557e-06, + "loss": 0.0025, + "step": 49300 + }, + { + "epoch": 6.46, + "learning_rate": 1.885957339209159e-06, + "loss": 0.0038, + "step": 49301 + }, + { + "epoch": 6.46, + "learning_rate": 1.8856473833914724e-06, + "loss": 0.0034, + "step": 49302 + }, + { + "epoch": 6.46, + "learning_rate": 1.8853374503949739e-06, + "loss": 0.0011, + "step": 49303 + }, + { + "epoch": 6.46, + "learning_rate": 1.8850275402205309e-06, + "loss": 0.0022, + "step": 49304 + }, + { + "epoch": 6.46, + "learning_rate": 1.8847176528690158e-06, + "loss": 0.0023, + "step": 49305 + }, + { + "epoch": 6.46, + "learning_rate": 1.8844077883412982e-06, + "loss": 0.0018, + "step": 49306 + }, + { + "epoch": 6.46, + "learning_rate": 1.8840979466382536e-06, + "loss": 0.0016, + "step": 49307 + }, + { + "epoch": 6.46, + "learning_rate": 1.8837881277607517e-06, + "loss": 0.0009, + "step": 49308 + }, + { + "epoch": 6.46, + "learning_rate": 1.8834783317096606e-06, + "loss": 0.0016, + "step": 49309 + }, + { + "epoch": 6.46, + "learning_rate": 1.883168558485854e-06, + "loss": 0.002, + "step": 49310 + }, + { + "epoch": 6.46, + "learning_rate": 1.8828588080902066e-06, + "loss": 0.0015, + "step": 49311 + }, + { + "epoch": 6.46, + "learning_rate": 1.8825490805235868e-06, + "loss": 0.0024, + "step": 49312 + }, + { + "epoch": 6.46, + "learning_rate": 1.8822393757868652e-06, + "loss": 0.0027, + "step": 49313 + }, + { + "epoch": 6.46, + "learning_rate": 1.8819296938809106e-06, + "loss": 0.0019, + "step": 49314 + }, + { + "epoch": 6.46, + "learning_rate": 1.8816200348065994e-06, + "loss": 0.0026, + "step": 49315 + }, + { + "epoch": 6.46, + "learning_rate": 1.8813103985647985e-06, + "loss": 0.0017, + "step": 49316 + }, + { + "epoch": 6.46, + "learning_rate": 1.8810007851563782e-06, + "loss": 0.0012, + "step": 49317 + }, + { + "epoch": 6.46, + "learning_rate": 1.8806911945822126e-06, + "loss": 0.0027, + "step": 49318 + }, + { + "epoch": 6.46, + "learning_rate": 1.8803816268431708e-06, + "loss": 0.0015, + "step": 49319 + }, + { + "epoch": 6.46, + "learning_rate": 1.8800720819401209e-06, + "loss": 0.0022, + "step": 49320 + }, + { + "epoch": 6.46, + "learning_rate": 1.8797625598739387e-06, + "loss": 0.0017, + "step": 49321 + }, + { + "epoch": 6.46, + "learning_rate": 1.8794530606454886e-06, + "loss": 0.0021, + "step": 49322 + }, + { + "epoch": 6.46, + "learning_rate": 1.879143584255647e-06, + "loss": 0.0017, + "step": 49323 + }, + { + "epoch": 6.46, + "learning_rate": 1.8788341307052827e-06, + "loss": 0.0019, + "step": 49324 + }, + { + "epoch": 6.46, + "learning_rate": 1.8785246999952612e-06, + "loss": 0.0038, + "step": 49325 + }, + { + "epoch": 6.46, + "learning_rate": 1.8782152921264595e-06, + "loss": 0.001, + "step": 49326 + }, + { + "epoch": 6.46, + "learning_rate": 1.877905907099744e-06, + "loss": 0.002, + "step": 49327 + }, + { + "epoch": 6.46, + "learning_rate": 1.8775965449159871e-06, + "loss": 0.0015, + "step": 49328 + }, + { + "epoch": 6.46, + "learning_rate": 1.8772872055760538e-06, + "loss": 0.0022, + "step": 49329 + }, + { + "epoch": 6.46, + "learning_rate": 1.87697788908082e-06, + "loss": 0.0027, + "step": 49330 + }, + { + "epoch": 6.46, + "learning_rate": 1.8766685954311515e-06, + "loss": 0.0045, + "step": 49331 + }, + { + "epoch": 6.46, + "learning_rate": 1.8763593246279222e-06, + "loss": 0.0029, + "step": 49332 + }, + { + "epoch": 6.46, + "learning_rate": 1.8760500766719968e-06, + "loss": 0.0023, + "step": 49333 + }, + { + "epoch": 6.46, + "learning_rate": 1.8757408515642516e-06, + "loss": 0.0014, + "step": 49334 + }, + { + "epoch": 6.46, + "learning_rate": 1.8754316493055513e-06, + "loss": 0.0025, + "step": 49335 + }, + { + "epoch": 6.46, + "learning_rate": 1.8751224698967675e-06, + "loss": 0.0025, + "step": 49336 + }, + { + "epoch": 6.46, + "learning_rate": 1.874813313338767e-06, + "loss": 0.0012, + "step": 49337 + }, + { + "epoch": 6.46, + "learning_rate": 1.8745041796324226e-06, + "loss": 0.0013, + "step": 49338 + }, + { + "epoch": 6.46, + "learning_rate": 1.8741950687786038e-06, + "loss": 0.0017, + "step": 49339 + }, + { + "epoch": 6.46, + "learning_rate": 1.873885980778175e-06, + "loss": 0.0018, + "step": 49340 + }, + { + "epoch": 6.46, + "learning_rate": 1.8735769156320127e-06, + "loss": 0.0017, + "step": 49341 + }, + { + "epoch": 6.46, + "learning_rate": 1.8732678733409792e-06, + "loss": 0.0021, + "step": 49342 + }, + { + "epoch": 6.46, + "learning_rate": 1.8729588539059496e-06, + "loss": 0.0012, + "step": 49343 + }, + { + "epoch": 6.46, + "learning_rate": 1.8726498573277884e-06, + "loss": 0.0024, + "step": 49344 + }, + { + "epoch": 6.46, + "learning_rate": 1.8723408836073686e-06, + "loss": 0.003, + "step": 49345 + }, + { + "epoch": 6.46, + "learning_rate": 1.8720319327455582e-06, + "loss": 0.0016, + "step": 49346 + }, + { + "epoch": 6.46, + "learning_rate": 1.8717230047432244e-06, + "loss": 0.0024, + "step": 49347 + }, + { + "epoch": 6.46, + "learning_rate": 1.8714140996012341e-06, + "loss": 0.0028, + "step": 49348 + }, + { + "epoch": 6.46, + "learning_rate": 1.8711052173204624e-06, + "loss": 0.0016, + "step": 49349 + }, + { + "epoch": 6.46, + "learning_rate": 1.8707963579017728e-06, + "loss": 0.0015, + "step": 49350 + }, + { + "epoch": 6.46, + "learning_rate": 1.8704875213460372e-06, + "loss": 0.0025, + "step": 49351 + }, + { + "epoch": 6.46, + "learning_rate": 1.8701787076541166e-06, + "loss": 0.0016, + "step": 49352 + }, + { + "epoch": 6.46, + "learning_rate": 1.8698699168268907e-06, + "loss": 0.0018, + "step": 49353 + }, + { + "epoch": 6.47, + "learning_rate": 1.869561148865222e-06, + "loss": 0.0026, + "step": 49354 + }, + { + "epoch": 6.47, + "learning_rate": 1.86925240376998e-06, + "loss": 0.0015, + "step": 49355 + }, + { + "epoch": 6.47, + "learning_rate": 1.8689436815420303e-06, + "loss": 0.0041, + "step": 49356 + }, + { + "epoch": 6.47, + "learning_rate": 1.8686349821822457e-06, + "loss": 0.0021, + "step": 49357 + }, + { + "epoch": 6.47, + "learning_rate": 1.8683263056914935e-06, + "loss": 0.0023, + "step": 49358 + }, + { + "epoch": 6.47, + "learning_rate": 1.8680176520706362e-06, + "loss": 0.0018, + "step": 49359 + }, + { + "epoch": 6.47, + "learning_rate": 1.8677090213205496e-06, + "loss": 0.0025, + "step": 49360 + }, + { + "epoch": 6.47, + "learning_rate": 1.8674004134420987e-06, + "loss": 0.0008, + "step": 49361 + }, + { + "epoch": 6.47, + "learning_rate": 1.8670918284361505e-06, + "loss": 0.0021, + "step": 49362 + }, + { + "epoch": 6.47, + "learning_rate": 1.8667832663035712e-06, + "loss": 0.0022, + "step": 49363 + }, + { + "epoch": 6.47, + "learning_rate": 1.8664747270452298e-06, + "loss": 0.0023, + "step": 49364 + }, + { + "epoch": 6.47, + "learning_rate": 1.8661662106619982e-06, + "loss": 0.0014, + "step": 49365 + }, + { + "epoch": 6.47, + "learning_rate": 1.8658577171547422e-06, + "loss": 0.0017, + "step": 49366 + }, + { + "epoch": 6.47, + "learning_rate": 1.865549246524324e-06, + "loss": 0.0025, + "step": 49367 + }, + { + "epoch": 6.47, + "learning_rate": 1.8652407987716192e-06, + "loss": 0.0022, + "step": 49368 + }, + { + "epoch": 6.47, + "learning_rate": 1.86493237389749e-06, + "loss": 0.0028, + "step": 49369 + }, + { + "epoch": 6.47, + "learning_rate": 1.864623971902807e-06, + "loss": 0.0027, + "step": 49370 + }, + { + "epoch": 6.47, + "learning_rate": 1.8643155927884315e-06, + "loss": 0.0022, + "step": 49371 + }, + { + "epoch": 6.47, + "learning_rate": 1.8640072365552386e-06, + "loss": 0.0021, + "step": 49372 + }, + { + "epoch": 6.47, + "learning_rate": 1.863698903204092e-06, + "loss": 0.0021, + "step": 49373 + }, + { + "epoch": 6.47, + "learning_rate": 1.8633905927358565e-06, + "loss": 0.0011, + "step": 49374 + }, + { + "epoch": 6.47, + "learning_rate": 1.8630823051514023e-06, + "loss": 0.0021, + "step": 49375 + }, + { + "epoch": 6.47, + "learning_rate": 1.8627740404515971e-06, + "loss": 0.0023, + "step": 49376 + }, + { + "epoch": 6.47, + "learning_rate": 1.8624657986373074e-06, + "loss": 0.0012, + "step": 49377 + }, + { + "epoch": 6.47, + "learning_rate": 1.8621575797093993e-06, + "loss": 0.0037, + "step": 49378 + }, + { + "epoch": 6.47, + "learning_rate": 1.8618493836687367e-06, + "loss": 0.0018, + "step": 49379 + }, + { + "epoch": 6.47, + "learning_rate": 1.8615412105161923e-06, + "loss": 0.0021, + "step": 49380 + }, + { + "epoch": 6.47, + "learning_rate": 1.8612330602526296e-06, + "loss": 0.0015, + "step": 49381 + }, + { + "epoch": 6.47, + "learning_rate": 1.8609249328789126e-06, + "loss": 0.001, + "step": 49382 + }, + { + "epoch": 6.47, + "learning_rate": 1.8606168283959137e-06, + "loss": 0.0031, + "step": 49383 + }, + { + "epoch": 6.47, + "learning_rate": 1.8603087468044966e-06, + "loss": 0.0014, + "step": 49384 + }, + { + "epoch": 6.47, + "learning_rate": 1.8600006881055244e-06, + "loss": 0.002, + "step": 49385 + }, + { + "epoch": 6.47, + "learning_rate": 1.8596926522998704e-06, + "loss": 0.0022, + "step": 49386 + }, + { + "epoch": 6.47, + "learning_rate": 1.859384639388394e-06, + "loss": 0.0021, + "step": 49387 + }, + { + "epoch": 6.47, + "learning_rate": 1.8590766493719682e-06, + "loss": 0.0017, + "step": 49388 + }, + { + "epoch": 6.47, + "learning_rate": 1.858768682251455e-06, + "loss": 0.0011, + "step": 49389 + }, + { + "epoch": 6.47, + "learning_rate": 1.8584607380277187e-06, + "loss": 0.0019, + "step": 49390 + }, + { + "epoch": 6.47, + "learning_rate": 1.8581528167016317e-06, + "loss": 0.0019, + "step": 49391 + }, + { + "epoch": 6.47, + "learning_rate": 1.8578449182740555e-06, + "loss": 0.0015, + "step": 49392 + }, + { + "epoch": 6.47, + "learning_rate": 1.857537042745856e-06, + "loss": 0.0023, + "step": 49393 + }, + { + "epoch": 6.47, + "learning_rate": 1.8572291901178985e-06, + "loss": 0.0024, + "step": 49394 + }, + { + "epoch": 6.47, + "learning_rate": 1.856921360391053e-06, + "loss": 0.0016, + "step": 49395 + }, + { + "epoch": 6.47, + "learning_rate": 1.85661355356618e-06, + "loss": 0.0039, + "step": 49396 + }, + { + "epoch": 6.47, + "learning_rate": 1.8563057696441488e-06, + "loss": 0.0023, + "step": 49397 + }, + { + "epoch": 6.47, + "learning_rate": 1.8559980086258234e-06, + "loss": 0.0025, + "step": 49398 + }, + { + "epoch": 6.47, + "learning_rate": 1.8556902705120716e-06, + "loss": 0.0021, + "step": 49399 + }, + { + "epoch": 6.47, + "learning_rate": 1.8553825553037564e-06, + "loss": 0.0019, + "step": 49400 + }, + { + "epoch": 6.47, + "learning_rate": 1.8550748630017457e-06, + "loss": 0.0019, + "step": 49401 + }, + { + "epoch": 6.47, + "learning_rate": 1.854767193606899e-06, + "loss": 0.0027, + "step": 49402 + }, + { + "epoch": 6.47, + "learning_rate": 1.8544595471200888e-06, + "loss": 0.0026, + "step": 49403 + }, + { + "epoch": 6.47, + "learning_rate": 1.854151923542178e-06, + "loss": 0.0014, + "step": 49404 + }, + { + "epoch": 6.47, + "learning_rate": 1.8538443228740287e-06, + "loss": 0.0017, + "step": 49405 + }, + { + "epoch": 6.47, + "learning_rate": 1.85353674511651e-06, + "loss": 0.0016, + "step": 49406 + }, + { + "epoch": 6.47, + "learning_rate": 1.8532291902704836e-06, + "loss": 0.0028, + "step": 49407 + }, + { + "epoch": 6.47, + "learning_rate": 1.852921658336818e-06, + "loss": 0.0013, + "step": 49408 + }, + { + "epoch": 6.47, + "learning_rate": 1.852614149316375e-06, + "loss": 0.0036, + "step": 49409 + }, + { + "epoch": 6.47, + "learning_rate": 1.8523066632100228e-06, + "loss": 0.0029, + "step": 49410 + }, + { + "epoch": 6.47, + "learning_rate": 1.8519992000186238e-06, + "loss": 0.0028, + "step": 49411 + }, + { + "epoch": 6.47, + "learning_rate": 1.8516917597430438e-06, + "loss": 0.0019, + "step": 49412 + }, + { + "epoch": 6.47, + "learning_rate": 1.8513843423841438e-06, + "loss": 0.0016, + "step": 49413 + }, + { + "epoch": 6.47, + "learning_rate": 1.8510769479427937e-06, + "loss": 0.0023, + "step": 49414 + }, + { + "epoch": 6.47, + "learning_rate": 1.8507695764198563e-06, + "loss": 0.002, + "step": 49415 + }, + { + "epoch": 6.47, + "learning_rate": 1.8504622278161954e-06, + "loss": 0.0012, + "step": 49416 + }, + { + "epoch": 6.47, + "learning_rate": 1.8501549021326727e-06, + "loss": 0.0023, + "step": 49417 + }, + { + "epoch": 6.47, + "learning_rate": 1.849847599370156e-06, + "loss": 0.0015, + "step": 49418 + }, + { + "epoch": 6.47, + "learning_rate": 1.8495403195295102e-06, + "loss": 0.0032, + "step": 49419 + }, + { + "epoch": 6.47, + "learning_rate": 1.8492330626115996e-06, + "loss": 0.0017, + "step": 49420 + }, + { + "epoch": 6.47, + "learning_rate": 1.848925828617284e-06, + "loss": 0.002, + "step": 49421 + }, + { + "epoch": 6.47, + "learning_rate": 1.848618617547432e-06, + "loss": 0.0013, + "step": 49422 + }, + { + "epoch": 6.47, + "learning_rate": 1.848311429402907e-06, + "loss": 0.0026, + "step": 49423 + }, + { + "epoch": 6.47, + "learning_rate": 1.8480042641845697e-06, + "loss": 0.0018, + "step": 49424 + }, + { + "epoch": 6.47, + "learning_rate": 1.8476971218932883e-06, + "loss": 0.0021, + "step": 49425 + }, + { + "epoch": 6.47, + "learning_rate": 1.8473900025299252e-06, + "loss": 0.0017, + "step": 49426 + }, + { + "epoch": 6.47, + "learning_rate": 1.8470829060953421e-06, + "loss": 0.0018, + "step": 49427 + }, + { + "epoch": 6.47, + "learning_rate": 1.8467758325904028e-06, + "loss": 0.0016, + "step": 49428 + }, + { + "epoch": 6.47, + "learning_rate": 1.8464687820159721e-06, + "loss": 0.0014, + "step": 49429 + }, + { + "epoch": 6.47, + "learning_rate": 1.846161754372916e-06, + "loss": 0.002, + "step": 49430 + }, + { + "epoch": 6.48, + "learning_rate": 1.845854749662096e-06, + "loss": 0.0016, + "step": 49431 + }, + { + "epoch": 6.48, + "learning_rate": 1.8455477678843737e-06, + "loss": 0.0033, + "step": 49432 + }, + { + "epoch": 6.48, + "learning_rate": 1.8452408090406149e-06, + "loss": 0.0014, + "step": 49433 + }, + { + "epoch": 6.48, + "learning_rate": 1.8449338731316834e-06, + "loss": 0.0011, + "step": 49434 + }, + { + "epoch": 6.48, + "learning_rate": 1.84462696015844e-06, + "loss": 0.0014, + "step": 49435 + }, + { + "epoch": 6.48, + "learning_rate": 1.8443200701217468e-06, + "loss": 0.0019, + "step": 49436 + }, + { + "epoch": 6.48, + "learning_rate": 1.8440132030224722e-06, + "loss": 0.0016, + "step": 49437 + }, + { + "epoch": 6.48, + "learning_rate": 1.8437063588614757e-06, + "loss": 0.0017, + "step": 49438 + }, + { + "epoch": 6.48, + "learning_rate": 1.8433995376396186e-06, + "loss": 0.0038, + "step": 49439 + }, + { + "epoch": 6.48, + "learning_rate": 1.8430927393577647e-06, + "loss": 0.0024, + "step": 49440 + }, + { + "epoch": 6.48, + "learning_rate": 1.8427859640167821e-06, + "loss": 0.0024, + "step": 49441 + }, + { + "epoch": 6.48, + "learning_rate": 1.8424792116175283e-06, + "loss": 0.0019, + "step": 49442 + }, + { + "epoch": 6.48, + "learning_rate": 1.842172482160869e-06, + "loss": 0.0014, + "step": 49443 + }, + { + "epoch": 6.48, + "learning_rate": 1.841865775647661e-06, + "loss": 0.0009, + "step": 49444 + }, + { + "epoch": 6.48, + "learning_rate": 1.841559092078774e-06, + "loss": 0.0006, + "step": 49445 + }, + { + "epoch": 6.48, + "learning_rate": 1.8412524314550684e-06, + "loss": 0.0013, + "step": 49446 + }, + { + "epoch": 6.48, + "learning_rate": 1.8409457937774022e-06, + "loss": 0.0032, + "step": 49447 + }, + { + "epoch": 6.48, + "learning_rate": 1.840639179046645e-06, + "loss": 0.0016, + "step": 49448 + }, + { + "epoch": 6.48, + "learning_rate": 1.8403325872636546e-06, + "loss": 0.0035, + "step": 49449 + }, + { + "epoch": 6.48, + "learning_rate": 1.840026018429295e-06, + "loss": 0.0015, + "step": 49450 + }, + { + "epoch": 6.48, + "learning_rate": 1.8397194725444246e-06, + "loss": 0.0011, + "step": 49451 + }, + { + "epoch": 6.48, + "learning_rate": 1.8394129496099088e-06, + "loss": 0.0024, + "step": 49452 + }, + { + "epoch": 6.48, + "learning_rate": 1.839106449626612e-06, + "loss": 0.0023, + "step": 49453 + }, + { + "epoch": 6.48, + "learning_rate": 1.8387999725953943e-06, + "loss": 0.0039, + "step": 49454 + }, + { + "epoch": 6.48, + "learning_rate": 1.8384935185171139e-06, + "loss": 0.003, + "step": 49455 + }, + { + "epoch": 6.48, + "learning_rate": 1.8381870873926388e-06, + "loss": 0.0021, + "step": 49456 + }, + { + "epoch": 6.48, + "learning_rate": 1.8378806792228276e-06, + "loss": 0.0028, + "step": 49457 + }, + { + "epoch": 6.48, + "learning_rate": 1.8375742940085416e-06, + "loss": 0.0019, + "step": 49458 + }, + { + "epoch": 6.48, + "learning_rate": 1.8372679317506425e-06, + "loss": 0.0026, + "step": 49459 + }, + { + "epoch": 6.48, + "learning_rate": 1.836961592449994e-06, + "loss": 0.0013, + "step": 49460 + }, + { + "epoch": 6.48, + "learning_rate": 1.8366552761074564e-06, + "loss": 0.0018, + "step": 49461 + }, + { + "epoch": 6.48, + "learning_rate": 1.8363489827238889e-06, + "loss": 0.0017, + "step": 49462 + }, + { + "epoch": 6.48, + "learning_rate": 1.8360427123001546e-06, + "loss": 0.002, + "step": 49463 + }, + { + "epoch": 6.48, + "learning_rate": 1.8357364648371178e-06, + "loss": 0.002, + "step": 49464 + }, + { + "epoch": 6.48, + "learning_rate": 1.8354302403356383e-06, + "loss": 0.0024, + "step": 49465 + }, + { + "epoch": 6.48, + "learning_rate": 1.8351240387965762e-06, + "loss": 0.0017, + "step": 49466 + }, + { + "epoch": 6.48, + "learning_rate": 1.83481786022079e-06, + "loss": 0.0007, + "step": 49467 + }, + { + "epoch": 6.48, + "learning_rate": 1.8345117046091466e-06, + "loss": 0.0007, + "step": 49468 + }, + { + "epoch": 6.48, + "learning_rate": 1.834205571962504e-06, + "loss": 0.0017, + "step": 49469 + }, + { + "epoch": 6.48, + "learning_rate": 1.8338994622817207e-06, + "loss": 0.0024, + "step": 49470 + }, + { + "epoch": 6.48, + "learning_rate": 1.8335933755676628e-06, + "loss": 0.0024, + "step": 49471 + }, + { + "epoch": 6.48, + "learning_rate": 1.833287311821189e-06, + "loss": 0.0014, + "step": 49472 + }, + { + "epoch": 6.48, + "learning_rate": 1.8329812710431572e-06, + "loss": 0.0015, + "step": 49473 + }, + { + "epoch": 6.48, + "learning_rate": 1.8326752532344305e-06, + "loss": 0.0023, + "step": 49474 + }, + { + "epoch": 6.48, + "learning_rate": 1.832369258395873e-06, + "loss": 0.0017, + "step": 49475 + }, + { + "epoch": 6.48, + "learning_rate": 1.8320632865283406e-06, + "loss": 0.0024, + "step": 49476 + }, + { + "epoch": 6.48, + "learning_rate": 1.8317573376326958e-06, + "loss": 0.0024, + "step": 49477 + }, + { + "epoch": 6.48, + "learning_rate": 1.831451411709797e-06, + "loss": 0.0017, + "step": 49478 + }, + { + "epoch": 6.48, + "learning_rate": 1.831145508760508e-06, + "loss": 0.0012, + "step": 49479 + }, + { + "epoch": 6.48, + "learning_rate": 1.8308396287856867e-06, + "loss": 0.0024, + "step": 49480 + }, + { + "epoch": 6.48, + "learning_rate": 1.8305337717861915e-06, + "loss": 0.0017, + "step": 49481 + }, + { + "epoch": 6.48, + "learning_rate": 1.8302279377628873e-06, + "loss": 0.0018, + "step": 49482 + }, + { + "epoch": 6.48, + "learning_rate": 1.82992212671663e-06, + "loss": 0.0019, + "step": 49483 + }, + { + "epoch": 6.48, + "learning_rate": 1.8296163386482845e-06, + "loss": 0.0015, + "step": 49484 + }, + { + "epoch": 6.48, + "learning_rate": 1.8293105735587069e-06, + "loss": 0.0012, + "step": 49485 + }, + { + "epoch": 6.48, + "learning_rate": 1.8290048314487564e-06, + "loss": 0.0028, + "step": 49486 + }, + { + "epoch": 6.48, + "learning_rate": 1.8286991123192977e-06, + "loss": 0.0022, + "step": 49487 + }, + { + "epoch": 6.48, + "learning_rate": 1.828393416171187e-06, + "loss": 0.0017, + "step": 49488 + }, + { + "epoch": 6.48, + "learning_rate": 1.8280877430052813e-06, + "loss": 0.0021, + "step": 49489 + }, + { + "epoch": 6.48, + "learning_rate": 1.8277820928224478e-06, + "loss": 0.0008, + "step": 49490 + }, + { + "epoch": 6.48, + "learning_rate": 1.8274764656235411e-06, + "loss": 0.0023, + "step": 49491 + }, + { + "epoch": 6.48, + "learning_rate": 1.8271708614094207e-06, + "loss": 0.0028, + "step": 49492 + }, + { + "epoch": 6.48, + "learning_rate": 1.826865280180946e-06, + "loss": 0.0024, + "step": 49493 + }, + { + "epoch": 6.48, + "learning_rate": 1.8265597219389764e-06, + "loss": 0.0017, + "step": 49494 + }, + { + "epoch": 6.48, + "learning_rate": 1.8262541866843753e-06, + "loss": 0.0017, + "step": 49495 + }, + { + "epoch": 6.48, + "learning_rate": 1.8259486744179999e-06, + "loss": 0.0028, + "step": 49496 + }, + { + "epoch": 6.48, + "learning_rate": 1.8256431851407053e-06, + "loss": 0.0017, + "step": 49497 + }, + { + "epoch": 6.48, + "learning_rate": 1.825337718853356e-06, + "loss": 0.0016, + "step": 49498 + }, + { + "epoch": 6.48, + "learning_rate": 1.8250322755568096e-06, + "loss": 0.0021, + "step": 49499 + }, + { + "epoch": 6.48, + "learning_rate": 1.8247268552519237e-06, + "loss": 0.0023, + "step": 49500 + }, + { + "epoch": 6.48, + "learning_rate": 1.824421457939557e-06, + "loss": 0.0023, + "step": 49501 + }, + { + "epoch": 6.48, + "learning_rate": 1.8241160836205707e-06, + "loss": 0.0035, + "step": 49502 + }, + { + "epoch": 6.48, + "learning_rate": 1.8238107322958233e-06, + "loss": 0.0018, + "step": 49503 + }, + { + "epoch": 6.48, + "learning_rate": 1.8235054039661705e-06, + "loss": 0.0025, + "step": 49504 + }, + { + "epoch": 6.48, + "learning_rate": 1.8232000986324726e-06, + "loss": 0.0026, + "step": 49505 + }, + { + "epoch": 6.48, + "learning_rate": 1.8228948162955929e-06, + "loss": 0.0015, + "step": 49506 + }, + { + "epoch": 6.49, + "learning_rate": 1.8225895569563844e-06, + "loss": 0.0019, + "step": 49507 + }, + { + "epoch": 6.49, + "learning_rate": 1.8222843206157082e-06, + "loss": 0.0029, + "step": 49508 + }, + { + "epoch": 6.49, + "learning_rate": 1.8219791072744198e-06, + "loss": 0.0025, + "step": 49509 + }, + { + "epoch": 6.49, + "learning_rate": 1.8216739169333809e-06, + "loss": 0.004, + "step": 49510 + }, + { + "epoch": 6.49, + "learning_rate": 1.8213687495934496e-06, + "loss": 0.0033, + "step": 49511 + }, + { + "epoch": 6.49, + "learning_rate": 1.8210636052554798e-06, + "loss": 0.001, + "step": 49512 + }, + { + "epoch": 6.49, + "learning_rate": 1.8207584839203363e-06, + "loss": 0.0019, + "step": 49513 + }, + { + "epoch": 6.49, + "learning_rate": 1.820453385588874e-06, + "loss": 0.0011, + "step": 49514 + }, + { + "epoch": 6.49, + "learning_rate": 1.8201483102619511e-06, + "loss": 0.0018, + "step": 49515 + }, + { + "epoch": 6.49, + "learning_rate": 1.8198432579404224e-06, + "loss": 0.0012, + "step": 49516 + }, + { + "epoch": 6.49, + "learning_rate": 1.8195382286251483e-06, + "loss": 0.0023, + "step": 49517 + }, + { + "epoch": 6.49, + "learning_rate": 1.8192332223169905e-06, + "loss": 0.0015, + "step": 49518 + }, + { + "epoch": 6.49, + "learning_rate": 1.818928239016804e-06, + "loss": 0.001, + "step": 49519 + }, + { + "epoch": 6.49, + "learning_rate": 1.8186232787254433e-06, + "loss": 0.0019, + "step": 49520 + }, + { + "epoch": 6.49, + "learning_rate": 1.8183183414437722e-06, + "loss": 0.0025, + "step": 49521 + }, + { + "epoch": 6.49, + "learning_rate": 1.8180134271726434e-06, + "loss": 0.004, + "step": 49522 + }, + { + "epoch": 6.49, + "learning_rate": 1.8177085359129176e-06, + "loss": 0.0028, + "step": 49523 + }, + { + "epoch": 6.49, + "learning_rate": 1.817403667665447e-06, + "loss": 0.0013, + "step": 49524 + }, + { + "epoch": 6.49, + "learning_rate": 1.817098822431097e-06, + "loss": 0.0013, + "step": 49525 + }, + { + "epoch": 6.49, + "learning_rate": 1.8167940002107197e-06, + "loss": 0.0016, + "step": 49526 + }, + { + "epoch": 6.49, + "learning_rate": 1.8164892010051716e-06, + "loss": 0.0029, + "step": 49527 + }, + { + "epoch": 6.49, + "learning_rate": 1.8161844248153115e-06, + "loss": 0.0029, + "step": 49528 + }, + { + "epoch": 6.49, + "learning_rate": 1.815879671641999e-06, + "loss": 0.0026, + "step": 49529 + }, + { + "epoch": 6.49, + "learning_rate": 1.815574941486089e-06, + "loss": 0.0018, + "step": 49530 + }, + { + "epoch": 6.49, + "learning_rate": 1.8152702343484374e-06, + "loss": 0.0024, + "step": 49531 + }, + { + "epoch": 6.49, + "learning_rate": 1.8149655502299035e-06, + "loss": 0.0019, + "step": 49532 + }, + { + "epoch": 6.49, + "learning_rate": 1.8146608891313433e-06, + "loss": 0.0036, + "step": 49533 + }, + { + "epoch": 6.49, + "learning_rate": 1.814356251053614e-06, + "loss": 0.0023, + "step": 49534 + }, + { + "epoch": 6.49, + "learning_rate": 1.814051635997569e-06, + "loss": 0.0018, + "step": 49535 + }, + { + "epoch": 6.49, + "learning_rate": 1.8137470439640715e-06, + "loss": 0.0019, + "step": 49536 + }, + { + "epoch": 6.49, + "learning_rate": 1.8134424749539725e-06, + "loss": 0.0019, + "step": 49537 + }, + { + "epoch": 6.49, + "learning_rate": 1.8131379289681294e-06, + "loss": 0.0013, + "step": 49538 + }, + { + "epoch": 6.49, + "learning_rate": 1.8128334060073993e-06, + "loss": 0.0015, + "step": 49539 + }, + { + "epoch": 6.49, + "learning_rate": 1.8125289060726424e-06, + "loss": 0.001, + "step": 49540 + }, + { + "epoch": 6.49, + "learning_rate": 1.8122244291647117e-06, + "loss": 0.0019, + "step": 49541 + }, + { + "epoch": 6.49, + "learning_rate": 1.811919975284464e-06, + "loss": 0.0017, + "step": 49542 + }, + { + "epoch": 6.49, + "learning_rate": 1.811615544432752e-06, + "loss": 0.001, + "step": 49543 + }, + { + "epoch": 6.49, + "learning_rate": 1.8113111366104385e-06, + "loss": 0.0013, + "step": 49544 + }, + { + "epoch": 6.49, + "learning_rate": 1.811006751818376e-06, + "loss": 0.0023, + "step": 49545 + }, + { + "epoch": 6.49, + "learning_rate": 1.8107023900574183e-06, + "loss": 0.0036, + "step": 49546 + }, + { + "epoch": 6.49, + "learning_rate": 1.810398051328427e-06, + "loss": 0.0015, + "step": 49547 + }, + { + "epoch": 6.49, + "learning_rate": 1.8100937356322545e-06, + "loss": 0.0024, + "step": 49548 + }, + { + "epoch": 6.49, + "learning_rate": 1.809789442969755e-06, + "loss": 0.0017, + "step": 49549 + }, + { + "epoch": 6.49, + "learning_rate": 1.8094851733417895e-06, + "loss": 0.0017, + "step": 49550 + }, + { + "epoch": 6.49, + "learning_rate": 1.8091809267492077e-06, + "loss": 0.0019, + "step": 49551 + }, + { + "epoch": 6.49, + "learning_rate": 1.8088767031928711e-06, + "loss": 0.0018, + "step": 49552 + }, + { + "epoch": 6.49, + "learning_rate": 1.8085725026736323e-06, + "loss": 0.0026, + "step": 49553 + }, + { + "epoch": 6.49, + "learning_rate": 1.808268325192344e-06, + "loss": 0.0022, + "step": 49554 + }, + { + "epoch": 6.49, + "learning_rate": 1.8079641707498685e-06, + "loss": 0.0014, + "step": 49555 + }, + { + "epoch": 6.49, + "learning_rate": 1.8076600393470568e-06, + "loss": 0.0029, + "step": 49556 + }, + { + "epoch": 6.49, + "learning_rate": 1.8073559309847654e-06, + "loss": 0.0018, + "step": 49557 + }, + { + "epoch": 6.49, + "learning_rate": 1.8070518456638463e-06, + "loss": 0.0014, + "step": 49558 + }, + { + "epoch": 6.49, + "learning_rate": 1.806747783385161e-06, + "loss": 0.0024, + "step": 49559 + }, + { + "epoch": 6.49, + "learning_rate": 1.8064437441495575e-06, + "loss": 0.003, + "step": 49560 + }, + { + "epoch": 6.49, + "learning_rate": 1.8061397279578975e-06, + "loss": 0.0022, + "step": 49561 + }, + { + "epoch": 6.49, + "learning_rate": 1.805835734811031e-06, + "loss": 0.0016, + "step": 49562 + }, + { + "epoch": 6.49, + "learning_rate": 1.805531764709817e-06, + "loss": 0.0007, + "step": 49563 + }, + { + "epoch": 6.49, + "learning_rate": 1.8052278176551098e-06, + "loss": 0.0019, + "step": 49564 + }, + { + "epoch": 6.49, + "learning_rate": 1.804923893647762e-06, + "loss": 0.0015, + "step": 49565 + }, + { + "epoch": 6.49, + "learning_rate": 1.8046199926886265e-06, + "loss": 0.0013, + "step": 49566 + }, + { + "epoch": 6.49, + "learning_rate": 1.8043161147785637e-06, + "loss": 0.0027, + "step": 49567 + }, + { + "epoch": 6.49, + "learning_rate": 1.8040122599184263e-06, + "loss": 0.0013, + "step": 49568 + }, + { + "epoch": 6.49, + "learning_rate": 1.8037084281090644e-06, + "loss": 0.0024, + "step": 49569 + }, + { + "epoch": 6.49, + "learning_rate": 1.8034046193513388e-06, + "loss": 0.0013, + "step": 49570 + }, + { + "epoch": 6.49, + "learning_rate": 1.8031008336460986e-06, + "loss": 0.0021, + "step": 49571 + }, + { + "epoch": 6.49, + "learning_rate": 1.8027970709942033e-06, + "loss": 0.0014, + "step": 49572 + }, + { + "epoch": 6.49, + "learning_rate": 1.8024933313965054e-06, + "loss": 0.0009, + "step": 49573 + }, + { + "epoch": 6.49, + "learning_rate": 1.8021896148538553e-06, + "loss": 0.0025, + "step": 49574 + }, + { + "epoch": 6.49, + "learning_rate": 1.8018859213671125e-06, + "loss": 0.0014, + "step": 49575 + }, + { + "epoch": 6.49, + "learning_rate": 1.8015822509371295e-06, + "loss": 0.0023, + "step": 49576 + }, + { + "epoch": 6.49, + "learning_rate": 1.8012786035647577e-06, + "loss": 0.0018, + "step": 49577 + }, + { + "epoch": 6.49, + "learning_rate": 1.8009749792508547e-06, + "loss": 0.0015, + "step": 49578 + }, + { + "epoch": 6.49, + "learning_rate": 1.8006713779962736e-06, + "loss": 0.0018, + "step": 49579 + }, + { + "epoch": 6.49, + "learning_rate": 1.8003677998018676e-06, + "loss": 0.0021, + "step": 49580 + }, + { + "epoch": 6.49, + "learning_rate": 1.800064244668488e-06, + "loss": 0.0023, + "step": 49581 + }, + { + "epoch": 6.49, + "learning_rate": 1.7997607125969906e-06, + "loss": 0.0022, + "step": 49582 + }, + { + "epoch": 6.5, + "learning_rate": 1.799457203588233e-06, + "loss": 0.0014, + "step": 49583 + }, + { + "epoch": 6.5, + "learning_rate": 1.799153717643065e-06, + "loss": 0.0021, + "step": 49584 + }, + { + "epoch": 6.5, + "learning_rate": 1.7988502547623377e-06, + "loss": 0.0021, + "step": 49585 + }, + { + "epoch": 6.5, + "learning_rate": 1.7985468149469098e-06, + "loss": 0.0029, + "step": 49586 + }, + { + "epoch": 6.5, + "learning_rate": 1.7982433981976321e-06, + "loss": 0.0011, + "step": 49587 + }, + { + "epoch": 6.5, + "learning_rate": 1.7979400045153584e-06, + "loss": 0.0013, + "step": 49588 + }, + { + "epoch": 6.5, + "learning_rate": 1.7976366339009387e-06, + "loss": 0.0018, + "step": 49589 + }, + { + "epoch": 6.5, + "learning_rate": 1.7973332863552318e-06, + "loss": 0.0016, + "step": 49590 + }, + { + "epoch": 6.5, + "learning_rate": 1.7970299618790888e-06, + "loss": 0.0029, + "step": 49591 + }, + { + "epoch": 6.5, + "learning_rate": 1.796726660473359e-06, + "loss": 0.002, + "step": 49592 + }, + { + "epoch": 6.5, + "learning_rate": 1.7964233821388988e-06, + "loss": 0.0025, + "step": 49593 + }, + { + "epoch": 6.5, + "learning_rate": 1.7961201268765626e-06, + "loss": 0.0032, + "step": 49594 + }, + { + "epoch": 6.5, + "learning_rate": 1.7958168946872034e-06, + "loss": 0.002, + "step": 49595 + }, + { + "epoch": 6.5, + "learning_rate": 1.7955136855716682e-06, + "loss": 0.0024, + "step": 49596 + }, + { + "epoch": 6.5, + "learning_rate": 1.7952104995308172e-06, + "loss": 0.0017, + "step": 49597 + }, + { + "epoch": 6.5, + "learning_rate": 1.7949073365655e-06, + "loss": 0.0018, + "step": 49598 + }, + { + "epoch": 6.5, + "learning_rate": 1.7946041966765681e-06, + "loss": 0.0016, + "step": 49599 + }, + { + "epoch": 6.5, + "learning_rate": 1.794301079864873e-06, + "loss": 0.0012, + "step": 49600 + }, + { + "epoch": 6.5, + "learning_rate": 1.7939979861312717e-06, + "loss": 0.0021, + "step": 49601 + }, + { + "epoch": 6.5, + "learning_rate": 1.7936949154766148e-06, + "loss": 0.0016, + "step": 49602 + }, + { + "epoch": 6.5, + "learning_rate": 1.7933918679017526e-06, + "loss": 0.0028, + "step": 49603 + }, + { + "epoch": 6.5, + "learning_rate": 1.7930888434075343e-06, + "loss": 0.0016, + "step": 49604 + }, + { + "epoch": 6.5, + "learning_rate": 1.7927858419948218e-06, + "loss": 0.002, + "step": 49605 + }, + { + "epoch": 6.5, + "learning_rate": 1.792482863664461e-06, + "loss": 0.0017, + "step": 49606 + }, + { + "epoch": 6.5, + "learning_rate": 1.7921799084173064e-06, + "loss": 0.0018, + "step": 49607 + }, + { + "epoch": 6.5, + "learning_rate": 1.7918769762542055e-06, + "loss": 0.0014, + "step": 49608 + }, + { + "epoch": 6.5, + "learning_rate": 1.7915740671760162e-06, + "loss": 0.0018, + "step": 49609 + }, + { + "epoch": 6.5, + "learning_rate": 1.791271181183587e-06, + "loss": 0.0037, + "step": 49610 + }, + { + "epoch": 6.5, + "learning_rate": 1.7909683182777694e-06, + "loss": 0.0024, + "step": 49611 + }, + { + "epoch": 6.5, + "learning_rate": 1.790665478459418e-06, + "loss": 0.0022, + "step": 49612 + }, + { + "epoch": 6.5, + "learning_rate": 1.7903626617293824e-06, + "loss": 0.0021, + "step": 49613 + }, + { + "epoch": 6.5, + "learning_rate": 1.7900598680885127e-06, + "loss": 0.0014, + "step": 49614 + }, + { + "epoch": 6.5, + "learning_rate": 1.7897570975376644e-06, + "loss": 0.0032, + "step": 49615 + }, + { + "epoch": 6.5, + "learning_rate": 1.7894543500776852e-06, + "loss": 0.0031, + "step": 49616 + }, + { + "epoch": 6.5, + "learning_rate": 1.7891516257094298e-06, + "loss": 0.0025, + "step": 49617 + }, + { + "epoch": 6.5, + "learning_rate": 1.7888489244337492e-06, + "loss": 0.0024, + "step": 49618 + }, + { + "epoch": 6.5, + "learning_rate": 1.7885462462514912e-06, + "loss": 0.001, + "step": 49619 + }, + { + "epoch": 6.5, + "learning_rate": 1.788243591163512e-06, + "loss": 0.0011, + "step": 49620 + }, + { + "epoch": 6.5, + "learning_rate": 1.7879409591706599e-06, + "loss": 0.0022, + "step": 49621 + }, + { + "epoch": 6.5, + "learning_rate": 1.7876383502737871e-06, + "loss": 0.0019, + "step": 49622 + }, + { + "epoch": 6.5, + "learning_rate": 1.787335764473741e-06, + "loss": 0.0043, + "step": 49623 + }, + { + "epoch": 6.5, + "learning_rate": 1.7870332017713788e-06, + "loss": 0.0023, + "step": 49624 + }, + { + "epoch": 6.5, + "learning_rate": 1.7867306621675461e-06, + "loss": 0.0016, + "step": 49625 + }, + { + "epoch": 6.5, + "learning_rate": 1.7864281456630973e-06, + "loss": 0.002, + "step": 49626 + }, + { + "epoch": 6.5, + "learning_rate": 1.786125652258881e-06, + "loss": 0.0022, + "step": 49627 + }, + { + "epoch": 6.5, + "learning_rate": 1.7858231819557503e-06, + "loss": 0.0031, + "step": 49628 + }, + { + "epoch": 6.5, + "learning_rate": 1.7855207347545545e-06, + "loss": 0.0017, + "step": 49629 + }, + { + "epoch": 6.5, + "learning_rate": 1.785218310656145e-06, + "loss": 0.0017, + "step": 49630 + }, + { + "epoch": 6.5, + "learning_rate": 1.7849159096613678e-06, + "loss": 0.0023, + "step": 49631 + }, + { + "epoch": 6.5, + "learning_rate": 1.784613531771081e-06, + "loss": 0.0019, + "step": 49632 + }, + { + "epoch": 6.5, + "learning_rate": 1.7843111769861298e-06, + "loss": 0.0011, + "step": 49633 + }, + { + "epoch": 6.5, + "learning_rate": 1.7840088453073645e-06, + "loss": 0.0015, + "step": 49634 + }, + { + "epoch": 6.5, + "learning_rate": 1.7837065367356388e-06, + "loss": 0.0027, + "step": 49635 + }, + { + "epoch": 6.5, + "learning_rate": 1.7834042512717987e-06, + "loss": 0.0036, + "step": 49636 + }, + { + "epoch": 6.5, + "learning_rate": 1.783101988916699e-06, + "loss": 0.002, + "step": 49637 + }, + { + "epoch": 6.5, + "learning_rate": 1.782799749671187e-06, + "loss": 0.0025, + "step": 49638 + }, + { + "epoch": 6.5, + "learning_rate": 1.7824975335361106e-06, + "loss": 0.0018, + "step": 49639 + }, + { + "epoch": 6.5, + "learning_rate": 1.7821953405123248e-06, + "loss": 0.0014, + "step": 49640 + }, + { + "epoch": 6.5, + "learning_rate": 1.7818931706006781e-06, + "loss": 0.0034, + "step": 49641 + }, + { + "epoch": 6.5, + "learning_rate": 1.781591023802015e-06, + "loss": 0.0019, + "step": 49642 + }, + { + "epoch": 6.5, + "learning_rate": 1.781288900117193e-06, + "loss": 0.0022, + "step": 49643 + }, + { + "epoch": 6.5, + "learning_rate": 1.7809867995470586e-06, + "loss": 0.0021, + "step": 49644 + }, + { + "epoch": 6.5, + "learning_rate": 1.7806847220924606e-06, + "loss": 0.0022, + "step": 49645 + }, + { + "epoch": 6.5, + "learning_rate": 1.7803826677542469e-06, + "loss": 0.0029, + "step": 49646 + }, + { + "epoch": 6.5, + "learning_rate": 1.780080636533269e-06, + "loss": 0.0027, + "step": 49647 + }, + { + "epoch": 6.5, + "learning_rate": 1.7797786284303786e-06, + "loss": 0.0021, + "step": 49648 + }, + { + "epoch": 6.5, + "learning_rate": 1.779476643446424e-06, + "loss": 0.0018, + "step": 49649 + }, + { + "epoch": 6.5, + "learning_rate": 1.7791746815822509e-06, + "loss": 0.0034, + "step": 49650 + }, + { + "epoch": 6.5, + "learning_rate": 1.7788727428387132e-06, + "loss": 0.0009, + "step": 49651 + }, + { + "epoch": 6.5, + "learning_rate": 1.778570827216658e-06, + "loss": 0.0019, + "step": 49652 + }, + { + "epoch": 6.5, + "learning_rate": 1.7782689347169346e-06, + "loss": 0.0027, + "step": 49653 + }, + { + "epoch": 6.5, + "learning_rate": 1.7779670653403902e-06, + "loss": 0.0014, + "step": 49654 + }, + { + "epoch": 6.5, + "learning_rate": 1.7776652190878774e-06, + "loss": 0.0017, + "step": 49655 + }, + { + "epoch": 6.5, + "learning_rate": 1.7773633959602444e-06, + "loss": 0.0029, + "step": 49656 + }, + { + "epoch": 6.5, + "learning_rate": 1.7770615959583348e-06, + "loss": 0.0014, + "step": 49657 + }, + { + "epoch": 6.5, + "learning_rate": 1.7767598190830026e-06, + "loss": 0.002, + "step": 49658 + }, + { + "epoch": 6.5, + "learning_rate": 1.7764580653350982e-06, + "loss": 0.0014, + "step": 49659 + }, + { + "epoch": 6.51, + "learning_rate": 1.7761563347154665e-06, + "loss": 0.003, + "step": 49660 + }, + { + "epoch": 6.51, + "learning_rate": 1.7758546272249555e-06, + "loss": 0.0019, + "step": 49661 + }, + { + "epoch": 6.51, + "learning_rate": 1.775552942864418e-06, + "loss": 0.0017, + "step": 49662 + }, + { + "epoch": 6.51, + "learning_rate": 1.7752512816346989e-06, + "loss": 0.0017, + "step": 49663 + }, + { + "epoch": 6.51, + "learning_rate": 1.7749496435366487e-06, + "loss": 0.0016, + "step": 49664 + }, + { + "epoch": 6.51, + "learning_rate": 1.774648028571112e-06, + "loss": 0.0019, + "step": 49665 + }, + { + "epoch": 6.51, + "learning_rate": 1.7743464367389418e-06, + "loss": 0.0017, + "step": 49666 + }, + { + "epoch": 6.51, + "learning_rate": 1.774044868040984e-06, + "loss": 0.0018, + "step": 49667 + }, + { + "epoch": 6.51, + "learning_rate": 1.7737433224780875e-06, + "loss": 0.0016, + "step": 49668 + }, + { + "epoch": 6.51, + "learning_rate": 1.7734418000510946e-06, + "loss": 0.0009, + "step": 49669 + }, + { + "epoch": 6.51, + "learning_rate": 1.773140300760864e-06, + "loss": 0.0023, + "step": 49670 + }, + { + "epoch": 6.51, + "learning_rate": 1.7728388246082372e-06, + "loss": 0.0015, + "step": 49671 + }, + { + "epoch": 6.51, + "learning_rate": 1.772537371594063e-06, + "loss": 0.0008, + "step": 49672 + }, + { + "epoch": 6.51, + "learning_rate": 1.7722359417191881e-06, + "loss": 0.0018, + "step": 49673 + }, + { + "epoch": 6.51, + "learning_rate": 1.771934534984463e-06, + "loss": 0.0019, + "step": 49674 + }, + { + "epoch": 6.51, + "learning_rate": 1.7716331513907337e-06, + "loss": 0.0019, + "step": 49675 + }, + { + "epoch": 6.51, + "learning_rate": 1.771331790938846e-06, + "loss": 0.0022, + "step": 49676 + }, + { + "epoch": 6.51, + "learning_rate": 1.7710304536296518e-06, + "loss": 0.0015, + "step": 49677 + }, + { + "epoch": 6.51, + "learning_rate": 1.770729139463997e-06, + "loss": 0.0012, + "step": 49678 + }, + { + "epoch": 6.51, + "learning_rate": 1.7704278484427272e-06, + "loss": 0.0016, + "step": 49679 + }, + { + "epoch": 6.51, + "learning_rate": 1.770126580566689e-06, + "loss": 0.0018, + "step": 49680 + }, + { + "epoch": 6.51, + "learning_rate": 1.7698253358367301e-06, + "loss": 0.0012, + "step": 49681 + }, + { + "epoch": 6.51, + "learning_rate": 1.7695241142537035e-06, + "loss": 0.0019, + "step": 49682 + }, + { + "epoch": 6.51, + "learning_rate": 1.7692229158184515e-06, + "loss": 0.0032, + "step": 49683 + }, + { + "epoch": 6.51, + "learning_rate": 1.768921740531818e-06, + "loss": 0.0011, + "step": 49684 + }, + { + "epoch": 6.51, + "learning_rate": 1.7686205883946572e-06, + "loss": 0.0031, + "step": 49685 + }, + { + "epoch": 6.51, + "learning_rate": 1.7683194594078134e-06, + "loss": 0.0013, + "step": 49686 + }, + { + "epoch": 6.51, + "learning_rate": 1.7680183535721318e-06, + "loss": 0.0021, + "step": 49687 + }, + { + "epoch": 6.51, + "learning_rate": 1.7677172708884582e-06, + "loss": 0.0026, + "step": 49688 + }, + { + "epoch": 6.51, + "learning_rate": 1.7674162113576432e-06, + "loss": 0.0012, + "step": 49689 + }, + { + "epoch": 6.51, + "learning_rate": 1.7671151749805327e-06, + "loss": 0.0027, + "step": 49690 + }, + { + "epoch": 6.51, + "learning_rate": 1.7668141617579693e-06, + "loss": 0.0016, + "step": 49691 + }, + { + "epoch": 6.51, + "learning_rate": 1.7665131716908023e-06, + "loss": 0.0016, + "step": 49692 + }, + { + "epoch": 6.51, + "learning_rate": 1.7662122047798814e-06, + "loss": 0.0039, + "step": 49693 + }, + { + "epoch": 6.51, + "learning_rate": 1.7659112610260498e-06, + "loss": 0.0015, + "step": 49694 + }, + { + "epoch": 6.51, + "learning_rate": 1.765610340430155e-06, + "loss": 0.0016, + "step": 49695 + }, + { + "epoch": 6.51, + "learning_rate": 1.7653094429930396e-06, + "loss": 0.0015, + "step": 49696 + }, + { + "epoch": 6.51, + "learning_rate": 1.765008568715556e-06, + "loss": 0.0022, + "step": 49697 + }, + { + "epoch": 6.51, + "learning_rate": 1.764707717598546e-06, + "loss": 0.0025, + "step": 49698 + }, + { + "epoch": 6.51, + "learning_rate": 1.7644068896428558e-06, + "loss": 0.0021, + "step": 49699 + }, + { + "epoch": 6.51, + "learning_rate": 1.7641060848493341e-06, + "loss": 0.0015, + "step": 49700 + }, + { + "epoch": 6.51, + "learning_rate": 1.7638053032188263e-06, + "loss": 0.0029, + "step": 49701 + }, + { + "epoch": 6.51, + "learning_rate": 1.7635045447521749e-06, + "loss": 0.0017, + "step": 49702 + }, + { + "epoch": 6.51, + "learning_rate": 1.7632038094502303e-06, + "loss": 0.0015, + "step": 49703 + }, + { + "epoch": 6.51, + "learning_rate": 1.762903097313834e-06, + "loss": 0.0015, + "step": 49704 + }, + { + "epoch": 6.51, + "learning_rate": 1.7626024083438375e-06, + "loss": 0.0024, + "step": 49705 + }, + { + "epoch": 6.51, + "learning_rate": 1.7623017425410815e-06, + "loss": 0.0027, + "step": 49706 + }, + { + "epoch": 6.51, + "learning_rate": 1.762001099906412e-06, + "loss": 0.0011, + "step": 49707 + }, + { + "epoch": 6.51, + "learning_rate": 1.7617004804406778e-06, + "loss": 0.0031, + "step": 49708 + }, + { + "epoch": 6.51, + "learning_rate": 1.7613998841447221e-06, + "loss": 0.0019, + "step": 49709 + }, + { + "epoch": 6.51, + "learning_rate": 1.7610993110193908e-06, + "loss": 0.0018, + "step": 49710 + }, + { + "epoch": 6.51, + "learning_rate": 1.7607987610655252e-06, + "loss": 0.0023, + "step": 49711 + }, + { + "epoch": 6.51, + "learning_rate": 1.7604982342839782e-06, + "loss": 0.0026, + "step": 49712 + }, + { + "epoch": 6.51, + "learning_rate": 1.7601977306755879e-06, + "loss": 0.0016, + "step": 49713 + }, + { + "epoch": 6.51, + "learning_rate": 1.759897250241206e-06, + "loss": 0.0011, + "step": 49714 + }, + { + "epoch": 6.51, + "learning_rate": 1.7595967929816714e-06, + "loss": 0.0013, + "step": 49715 + }, + { + "epoch": 6.51, + "learning_rate": 1.759296358897834e-06, + "loss": 0.0019, + "step": 49716 + }, + { + "epoch": 6.51, + "learning_rate": 1.758995947990537e-06, + "loss": 0.0012, + "step": 49717 + }, + { + "epoch": 6.51, + "learning_rate": 1.7586955602606259e-06, + "loss": 0.0013, + "step": 49718 + }, + { + "epoch": 6.51, + "learning_rate": 1.7583951957089418e-06, + "loss": 0.0024, + "step": 49719 + }, + { + "epoch": 6.51, + "learning_rate": 1.758094854336334e-06, + "loss": 0.0012, + "step": 49720 + }, + { + "epoch": 6.51, + "learning_rate": 1.7577945361436456e-06, + "loss": 0.0015, + "step": 49721 + }, + { + "epoch": 6.51, + "learning_rate": 1.7574942411317196e-06, + "loss": 0.0019, + "step": 49722 + }, + { + "epoch": 6.51, + "learning_rate": 1.7571939693014039e-06, + "loss": 0.0024, + "step": 49723 + }, + { + "epoch": 6.51, + "learning_rate": 1.7568937206535386e-06, + "loss": 0.0008, + "step": 49724 + }, + { + "epoch": 6.51, + "learning_rate": 1.756593495188974e-06, + "loss": 0.0018, + "step": 49725 + }, + { + "epoch": 6.51, + "learning_rate": 1.7562932929085475e-06, + "loss": 0.0014, + "step": 49726 + }, + { + "epoch": 6.51, + "learning_rate": 1.7559931138131102e-06, + "loss": 0.0022, + "step": 49727 + }, + { + "epoch": 6.51, + "learning_rate": 1.7556929579035042e-06, + "loss": 0.0023, + "step": 49728 + }, + { + "epoch": 6.51, + "learning_rate": 1.7553928251805718e-06, + "loss": 0.0015, + "step": 49729 + }, + { + "epoch": 6.51, + "learning_rate": 1.7550927156451548e-06, + "loss": 0.002, + "step": 49730 + }, + { + "epoch": 6.51, + "learning_rate": 1.7547926292981043e-06, + "loss": 0.0019, + "step": 49731 + }, + { + "epoch": 6.51, + "learning_rate": 1.75449256614026e-06, + "loss": 0.002, + "step": 49732 + }, + { + "epoch": 6.51, + "learning_rate": 1.7541925261724657e-06, + "loss": 0.0019, + "step": 49733 + }, + { + "epoch": 6.51, + "learning_rate": 1.7538925093955638e-06, + "loss": 0.0013, + "step": 49734 + }, + { + "epoch": 6.51, + "learning_rate": 1.7535925158104006e-06, + "loss": 0.0018, + "step": 49735 + }, + { + "epoch": 6.52, + "learning_rate": 1.7532925454178208e-06, + "loss": 0.002, + "step": 49736 + }, + { + "epoch": 6.52, + "learning_rate": 1.7529925982186669e-06, + "loss": 0.0025, + "step": 49737 + }, + { + "epoch": 6.52, + "learning_rate": 1.7526926742137795e-06, + "loss": 0.0025, + "step": 49738 + }, + { + "epoch": 6.52, + "learning_rate": 1.7523927734040081e-06, + "loss": 0.0026, + "step": 49739 + }, + { + "epoch": 6.52, + "learning_rate": 1.7520928957901916e-06, + "loss": 0.0028, + "step": 49740 + }, + { + "epoch": 6.52, + "learning_rate": 1.7517930413731732e-06, + "loss": 0.0008, + "step": 49741 + }, + { + "epoch": 6.52, + "learning_rate": 1.7514932101537996e-06, + "loss": 0.0039, + "step": 49742 + }, + { + "epoch": 6.52, + "learning_rate": 1.7511934021329124e-06, + "loss": 0.0023, + "step": 49743 + }, + { + "epoch": 6.52, + "learning_rate": 1.7508936173113544e-06, + "loss": 0.0019, + "step": 49744 + }, + { + "epoch": 6.52, + "learning_rate": 1.750593855689966e-06, + "loss": 0.0013, + "step": 49745 + }, + { + "epoch": 6.52, + "learning_rate": 1.7502941172695932e-06, + "loss": 0.0016, + "step": 49746 + }, + { + "epoch": 6.52, + "learning_rate": 1.7499944020510818e-06, + "loss": 0.0018, + "step": 49747 + }, + { + "epoch": 6.52, + "learning_rate": 1.7496947100352713e-06, + "loss": 0.0022, + "step": 49748 + }, + { + "epoch": 6.52, + "learning_rate": 1.749395041223002e-06, + "loss": 0.0017, + "step": 49749 + }, + { + "epoch": 6.52, + "learning_rate": 1.7490953956151236e-06, + "loss": 0.0028, + "step": 49750 + }, + { + "epoch": 6.52, + "learning_rate": 1.7487957732124739e-06, + "loss": 0.0019, + "step": 49751 + }, + { + "epoch": 6.52, + "learning_rate": 1.7484961740158969e-06, + "loss": 0.0016, + "step": 49752 + }, + { + "epoch": 6.52, + "learning_rate": 1.7481965980262328e-06, + "loss": 0.0016, + "step": 49753 + }, + { + "epoch": 6.52, + "learning_rate": 1.747897045244329e-06, + "loss": 0.0017, + "step": 49754 + }, + { + "epoch": 6.52, + "learning_rate": 1.7475975156710246e-06, + "loss": 0.0009, + "step": 49755 + }, + { + "epoch": 6.52, + "learning_rate": 1.7472980093071611e-06, + "loss": 0.0011, + "step": 49756 + }, + { + "epoch": 6.52, + "learning_rate": 1.7469985261535827e-06, + "loss": 0.0025, + "step": 49757 + }, + { + "epoch": 6.52, + "learning_rate": 1.7466990662111326e-06, + "loss": 0.0012, + "step": 49758 + }, + { + "epoch": 6.52, + "learning_rate": 1.7463996294806528e-06, + "loss": 0.005, + "step": 49759 + }, + { + "epoch": 6.52, + "learning_rate": 1.7461002159629847e-06, + "loss": 0.0017, + "step": 49760 + }, + { + "epoch": 6.52, + "learning_rate": 1.7458008256589664e-06, + "loss": 0.0013, + "step": 49761 + }, + { + "epoch": 6.52, + "learning_rate": 1.7455014585694474e-06, + "loss": 0.0013, + "step": 49762 + }, + { + "epoch": 6.52, + "learning_rate": 1.7452021146952657e-06, + "loss": 0.0012, + "step": 49763 + }, + { + "epoch": 6.52, + "learning_rate": 1.744902794037261e-06, + "loss": 0.0015, + "step": 49764 + }, + { + "epoch": 6.52, + "learning_rate": 1.7446034965962798e-06, + "loss": 0.0013, + "step": 49765 + }, + { + "epoch": 6.52, + "learning_rate": 1.744304222373162e-06, + "loss": 0.0024, + "step": 49766 + }, + { + "epoch": 6.52, + "learning_rate": 1.7440049713687468e-06, + "loss": 0.0021, + "step": 49767 + }, + { + "epoch": 6.52, + "learning_rate": 1.74370574358388e-06, + "loss": 0.0022, + "step": 49768 + }, + { + "epoch": 6.52, + "learning_rate": 1.7434065390193977e-06, + "loss": 0.0024, + "step": 49769 + }, + { + "epoch": 6.52, + "learning_rate": 1.743107357676148e-06, + "loss": 0.0021, + "step": 49770 + }, + { + "epoch": 6.52, + "learning_rate": 1.7428081995549695e-06, + "loss": 0.0018, + "step": 49771 + }, + { + "epoch": 6.52, + "learning_rate": 1.7425090646567011e-06, + "loss": 0.0013, + "step": 49772 + }, + { + "epoch": 6.52, + "learning_rate": 1.7422099529821878e-06, + "loss": 0.0015, + "step": 49773 + }, + { + "epoch": 6.52, + "learning_rate": 1.7419108645322702e-06, + "loss": 0.001, + "step": 49774 + }, + { + "epoch": 6.52, + "learning_rate": 1.7416117993077875e-06, + "loss": 0.002, + "step": 49775 + }, + { + "epoch": 6.52, + "learning_rate": 1.7413127573095801e-06, + "loss": 0.0014, + "step": 49776 + }, + { + "epoch": 6.52, + "learning_rate": 1.741013738538493e-06, + "loss": 0.0015, + "step": 49777 + }, + { + "epoch": 6.52, + "learning_rate": 1.740714742995362e-06, + "loss": 0.0013, + "step": 49778 + }, + { + "epoch": 6.52, + "learning_rate": 1.7404157706810343e-06, + "loss": 0.0031, + "step": 49779 + }, + { + "epoch": 6.52, + "learning_rate": 1.7401168215963438e-06, + "loss": 0.0022, + "step": 49780 + }, + { + "epoch": 6.52, + "learning_rate": 1.7398178957421386e-06, + "loss": 0.0033, + "step": 49781 + }, + { + "epoch": 6.52, + "learning_rate": 1.7395189931192546e-06, + "loss": 0.0025, + "step": 49782 + }, + { + "epoch": 6.52, + "learning_rate": 1.7392201137285348e-06, + "loss": 0.0033, + "step": 49783 + }, + { + "epoch": 6.52, + "learning_rate": 1.7389212575708148e-06, + "loss": 0.0027, + "step": 49784 + }, + { + "epoch": 6.52, + "learning_rate": 1.738622424646942e-06, + "loss": 0.0018, + "step": 49785 + }, + { + "epoch": 6.52, + "learning_rate": 1.7383236149577531e-06, + "loss": 0.0025, + "step": 49786 + }, + { + "epoch": 6.52, + "learning_rate": 1.7380248285040869e-06, + "loss": 0.0027, + "step": 49787 + }, + { + "epoch": 6.52, + "learning_rate": 1.737726065286789e-06, + "loss": 0.0019, + "step": 49788 + }, + { + "epoch": 6.52, + "learning_rate": 1.737427325306693e-06, + "loss": 0.0018, + "step": 49789 + }, + { + "epoch": 6.52, + "learning_rate": 1.7371286085646456e-06, + "loss": 0.0052, + "step": 49790 + }, + { + "epoch": 6.52, + "learning_rate": 1.736829915061482e-06, + "loss": 0.0016, + "step": 49791 + }, + { + "epoch": 6.52, + "learning_rate": 1.7365312447980453e-06, + "loss": 0.0029, + "step": 49792 + }, + { + "epoch": 6.52, + "learning_rate": 1.736232597775176e-06, + "loss": 0.002, + "step": 49793 + }, + { + "epoch": 6.52, + "learning_rate": 1.735933973993712e-06, + "loss": 0.0014, + "step": 49794 + }, + { + "epoch": 6.52, + "learning_rate": 1.7356353734544906e-06, + "loss": 0.0015, + "step": 49795 + }, + { + "epoch": 6.52, + "learning_rate": 1.7353367961583579e-06, + "loss": 0.002, + "step": 49796 + }, + { + "epoch": 6.52, + "learning_rate": 1.7350382421061497e-06, + "loss": 0.0018, + "step": 49797 + }, + { + "epoch": 6.52, + "learning_rate": 1.7347397112987064e-06, + "loss": 0.0014, + "step": 49798 + }, + { + "epoch": 6.52, + "learning_rate": 1.7344412037368652e-06, + "loss": 0.0054, + "step": 49799 + }, + { + "epoch": 6.52, + "learning_rate": 1.7341427194214678e-06, + "loss": 0.0011, + "step": 49800 + }, + { + "epoch": 6.52, + "learning_rate": 1.7338442583533555e-06, + "loss": 0.0014, + "step": 49801 + }, + { + "epoch": 6.52, + "learning_rate": 1.7335458205333677e-06, + "loss": 0.0047, + "step": 49802 + }, + { + "epoch": 6.52, + "learning_rate": 1.7332474059623384e-06, + "loss": 0.002, + "step": 49803 + }, + { + "epoch": 6.52, + "learning_rate": 1.7329490146411132e-06, + "loss": 0.002, + "step": 49804 + }, + { + "epoch": 6.52, + "learning_rate": 1.7326506465705284e-06, + "loss": 0.002, + "step": 49805 + }, + { + "epoch": 6.52, + "learning_rate": 1.7323523017514222e-06, + "loss": 0.0022, + "step": 49806 + }, + { + "epoch": 6.52, + "learning_rate": 1.7320539801846359e-06, + "loss": 0.0014, + "step": 49807 + }, + { + "epoch": 6.52, + "learning_rate": 1.731755681871008e-06, + "loss": 0.0025, + "step": 49808 + }, + { + "epoch": 6.52, + "learning_rate": 1.7314574068113777e-06, + "loss": 0.0017, + "step": 49809 + }, + { + "epoch": 6.52, + "learning_rate": 1.7311591550065798e-06, + "loss": 0.0025, + "step": 49810 + }, + { + "epoch": 6.52, + "learning_rate": 1.730860926457456e-06, + "loss": 0.0026, + "step": 49811 + }, + { + "epoch": 6.53, + "learning_rate": 1.730562721164849e-06, + "loss": 0.0027, + "step": 49812 + }, + { + "epoch": 6.53, + "learning_rate": 1.7302645391295935e-06, + "loss": 0.0025, + "step": 49813 + }, + { + "epoch": 6.53, + "learning_rate": 1.7299663803525258e-06, + "loss": 0.0019, + "step": 49814 + }, + { + "epoch": 6.53, + "learning_rate": 1.7296682448344904e-06, + "loss": 0.0017, + "step": 49815 + }, + { + "epoch": 6.53, + "learning_rate": 1.7293701325763223e-06, + "loss": 0.0026, + "step": 49816 + }, + { + "epoch": 6.53, + "learning_rate": 1.72907204357886e-06, + "loss": 0.0042, + "step": 49817 + }, + { + "epoch": 6.53, + "learning_rate": 1.7287739778429403e-06, + "loss": 0.0022, + "step": 49818 + }, + { + "epoch": 6.53, + "learning_rate": 1.7284759353694047e-06, + "loss": 0.0023, + "step": 49819 + }, + { + "epoch": 6.53, + "learning_rate": 1.7281779161590905e-06, + "loss": 0.0022, + "step": 49820 + }, + { + "epoch": 6.53, + "learning_rate": 1.7278799202128338e-06, + "loss": 0.0014, + "step": 49821 + }, + { + "epoch": 6.53, + "learning_rate": 1.7275819475314736e-06, + "loss": 0.0018, + "step": 49822 + }, + { + "epoch": 6.53, + "learning_rate": 1.7272839981158518e-06, + "loss": 0.0019, + "step": 49823 + }, + { + "epoch": 6.53, + "learning_rate": 1.7269860719668018e-06, + "loss": 0.002, + "step": 49824 + }, + { + "epoch": 6.53, + "learning_rate": 1.7266881690851644e-06, + "loss": 0.0016, + "step": 49825 + }, + { + "epoch": 6.53, + "learning_rate": 1.726390289471772e-06, + "loss": 0.0019, + "step": 49826 + }, + { + "epoch": 6.53, + "learning_rate": 1.7260924331274698e-06, + "loss": 0.0032, + "step": 49827 + }, + { + "epoch": 6.53, + "learning_rate": 1.7257946000530922e-06, + "loss": 0.002, + "step": 49828 + }, + { + "epoch": 6.53, + "learning_rate": 1.7254967902494735e-06, + "loss": 0.0013, + "step": 49829 + }, + { + "epoch": 6.53, + "learning_rate": 1.7251990037174571e-06, + "loss": 0.0012, + "step": 49830 + }, + { + "epoch": 6.53, + "learning_rate": 1.7249012404578779e-06, + "loss": 0.0019, + "step": 49831 + }, + { + "epoch": 6.53, + "learning_rate": 1.7246035004715733e-06, + "loss": 0.0016, + "step": 49832 + }, + { + "epoch": 6.53, + "learning_rate": 1.724305783759378e-06, + "loss": 0.0019, + "step": 49833 + }, + { + "epoch": 6.53, + "learning_rate": 1.7240080903221324e-06, + "loss": 0.0047, + "step": 49834 + }, + { + "epoch": 6.53, + "learning_rate": 1.7237104201606758e-06, + "loss": 0.001, + "step": 49835 + }, + { + "epoch": 6.53, + "learning_rate": 1.7234127732758432e-06, + "loss": 0.0013, + "step": 49836 + }, + { + "epoch": 6.53, + "learning_rate": 1.7231151496684683e-06, + "loss": 0.0019, + "step": 49837 + }, + { + "epoch": 6.53, + "learning_rate": 1.7228175493393939e-06, + "loss": 0.0027, + "step": 49838 + }, + { + "epoch": 6.53, + "learning_rate": 1.7225199722894537e-06, + "loss": 0.0012, + "step": 49839 + }, + { + "epoch": 6.53, + "learning_rate": 1.7222224185194857e-06, + "loss": 0.0016, + "step": 49840 + }, + { + "epoch": 6.53, + "learning_rate": 1.721924888030324e-06, + "loss": 0.0019, + "step": 49841 + }, + { + "epoch": 6.53, + "learning_rate": 1.7216273808228112e-06, + "loss": 0.0028, + "step": 49842 + }, + { + "epoch": 6.53, + "learning_rate": 1.72132989689778e-06, + "loss": 0.0019, + "step": 49843 + }, + { + "epoch": 6.53, + "learning_rate": 1.7210324362560638e-06, + "loss": 0.0034, + "step": 49844 + }, + { + "epoch": 6.53, + "learning_rate": 1.7207349988985044e-06, + "loss": 0.0027, + "step": 49845 + }, + { + "epoch": 6.53, + "learning_rate": 1.7204375848259391e-06, + "loss": 0.0022, + "step": 49846 + }, + { + "epoch": 6.53, + "learning_rate": 1.7201401940392015e-06, + "loss": 0.002, + "step": 49847 + }, + { + "epoch": 6.53, + "learning_rate": 1.7198428265391299e-06, + "loss": 0.0034, + "step": 49848 + }, + { + "epoch": 6.53, + "learning_rate": 1.7195454823265556e-06, + "loss": 0.0026, + "step": 49849 + }, + { + "epoch": 6.53, + "learning_rate": 1.7192481614023215e-06, + "loss": 0.0026, + "step": 49850 + }, + { + "epoch": 6.53, + "learning_rate": 1.7189508637672614e-06, + "loss": 0.001, + "step": 49851 + }, + { + "epoch": 6.53, + "learning_rate": 1.718653589422209e-06, + "loss": 0.0018, + "step": 49852 + }, + { + "epoch": 6.53, + "learning_rate": 1.7183563383680036e-06, + "loss": 0.0015, + "step": 49853 + }, + { + "epoch": 6.53, + "learning_rate": 1.7180591106054812e-06, + "loss": 0.0024, + "step": 49854 + }, + { + "epoch": 6.53, + "learning_rate": 1.7177619061354734e-06, + "loss": 0.0023, + "step": 49855 + }, + { + "epoch": 6.53, + "learning_rate": 1.7174647249588182e-06, + "loss": 0.0026, + "step": 49856 + }, + { + "epoch": 6.53, + "learning_rate": 1.7171675670763567e-06, + "loss": 0.0034, + "step": 49857 + }, + { + "epoch": 6.53, + "learning_rate": 1.7168704324889196e-06, + "loss": 0.0031, + "step": 49858 + }, + { + "epoch": 6.53, + "learning_rate": 1.7165733211973423e-06, + "loss": 0.0012, + "step": 49859 + }, + { + "epoch": 6.53, + "learning_rate": 1.7162762332024606e-06, + "loss": 0.0017, + "step": 49860 + }, + { + "epoch": 6.53, + "learning_rate": 1.715979168505112e-06, + "loss": 0.0012, + "step": 49861 + }, + { + "epoch": 6.53, + "learning_rate": 1.715682127106132e-06, + "loss": 0.002, + "step": 49862 + }, + { + "epoch": 6.53, + "learning_rate": 1.715385109006351e-06, + "loss": 0.0024, + "step": 49863 + }, + { + "epoch": 6.53, + "learning_rate": 1.7150881142066123e-06, + "loss": 0.0015, + "step": 49864 + }, + { + "epoch": 6.53, + "learning_rate": 1.7147911427077457e-06, + "loss": 0.0026, + "step": 49865 + }, + { + "epoch": 6.53, + "learning_rate": 1.7144941945105863e-06, + "loss": 0.002, + "step": 49866 + }, + { + "epoch": 6.53, + "learning_rate": 1.7141972696159725e-06, + "loss": 0.0014, + "step": 49867 + }, + { + "epoch": 6.53, + "learning_rate": 1.7139003680247358e-06, + "loss": 0.0011, + "step": 49868 + }, + { + "epoch": 6.53, + "learning_rate": 1.7136034897377151e-06, + "loss": 0.0017, + "step": 49869 + }, + { + "epoch": 6.53, + "learning_rate": 1.7133066347557437e-06, + "loss": 0.0017, + "step": 49870 + }, + { + "epoch": 6.53, + "learning_rate": 1.7130098030796527e-06, + "loss": 0.0019, + "step": 49871 + }, + { + "epoch": 6.53, + "learning_rate": 1.7127129947102838e-06, + "loss": 0.0013, + "step": 49872 + }, + { + "epoch": 6.53, + "learning_rate": 1.7124162096484676e-06, + "loss": 0.0011, + "step": 49873 + }, + { + "epoch": 6.53, + "learning_rate": 1.7121194478950398e-06, + "loss": 0.0012, + "step": 49874 + }, + { + "epoch": 6.53, + "learning_rate": 1.7118227094508322e-06, + "loss": 0.0014, + "step": 49875 + }, + { + "epoch": 6.53, + "learning_rate": 1.7115259943166818e-06, + "loss": 0.002, + "step": 49876 + }, + { + "epoch": 6.53, + "learning_rate": 1.7112293024934257e-06, + "loss": 0.0014, + "step": 49877 + }, + { + "epoch": 6.53, + "learning_rate": 1.7109326339818955e-06, + "loss": 0.0023, + "step": 49878 + }, + { + "epoch": 6.53, + "learning_rate": 1.7106359887829239e-06, + "loss": 0.0027, + "step": 49879 + }, + { + "epoch": 6.53, + "learning_rate": 1.7103393668973489e-06, + "loss": 0.0031, + "step": 49880 + }, + { + "epoch": 6.53, + "learning_rate": 1.7100427683260036e-06, + "loss": 0.0019, + "step": 49881 + }, + { + "epoch": 6.53, + "learning_rate": 1.7097461930697213e-06, + "loss": 0.0018, + "step": 49882 + }, + { + "epoch": 6.53, + "learning_rate": 1.7094496411293337e-06, + "loss": 0.0041, + "step": 49883 + }, + { + "epoch": 6.53, + "learning_rate": 1.7091531125056804e-06, + "loss": 0.0019, + "step": 49884 + }, + { + "epoch": 6.53, + "learning_rate": 1.7088566071995926e-06, + "loss": 0.002, + "step": 49885 + }, + { + "epoch": 6.53, + "learning_rate": 1.708560125211901e-06, + "loss": 0.0013, + "step": 49886 + }, + { + "epoch": 6.53, + "learning_rate": 1.7082636665434427e-06, + "loss": 0.0018, + "step": 49887 + }, + { + "epoch": 6.53, + "learning_rate": 1.7079672311950546e-06, + "loss": 0.0035, + "step": 49888 + }, + { + "epoch": 6.54, + "learning_rate": 1.707670819167566e-06, + "loss": 0.0017, + "step": 49889 + }, + { + "epoch": 6.54, + "learning_rate": 1.7073744304618112e-06, + "loss": 0.002, + "step": 49890 + }, + { + "epoch": 6.54, + "learning_rate": 1.7070780650786234e-06, + "loss": 0.0015, + "step": 49891 + }, + { + "epoch": 6.54, + "learning_rate": 1.7067817230188378e-06, + "loss": 0.0023, + "step": 49892 + }, + { + "epoch": 6.54, + "learning_rate": 1.706485404283288e-06, + "loss": 0.0017, + "step": 49893 + }, + { + "epoch": 6.54, + "learning_rate": 1.7061891088728032e-06, + "loss": 0.0013, + "step": 49894 + }, + { + "epoch": 6.54, + "learning_rate": 1.7058928367882233e-06, + "loss": 0.0021, + "step": 49895 + }, + { + "epoch": 6.54, + "learning_rate": 1.705596588030377e-06, + "loss": 0.0017, + "step": 49896 + }, + { + "epoch": 6.54, + "learning_rate": 1.7053003626000985e-06, + "loss": 0.002, + "step": 49897 + }, + { + "epoch": 6.54, + "learning_rate": 1.7050041604982193e-06, + "loss": 0.0012, + "step": 49898 + }, + { + "epoch": 6.54, + "learning_rate": 1.7047079817255741e-06, + "loss": 0.002, + "step": 49899 + }, + { + "epoch": 6.54, + "learning_rate": 1.7044118262829978e-06, + "loss": 0.0027, + "step": 49900 + }, + { + "epoch": 6.54, + "learning_rate": 1.7041156941713222e-06, + "loss": 0.0027, + "step": 49901 + }, + { + "epoch": 6.54, + "learning_rate": 1.7038195853913763e-06, + "loss": 0.0021, + "step": 49902 + }, + { + "epoch": 6.54, + "learning_rate": 1.7035234999439999e-06, + "loss": 0.0023, + "step": 49903 + }, + { + "epoch": 6.54, + "learning_rate": 1.7032274378300207e-06, + "loss": 0.0017, + "step": 49904 + }, + { + "epoch": 6.54, + "learning_rate": 1.7029313990502728e-06, + "loss": 0.0025, + "step": 49905 + }, + { + "epoch": 6.54, + "learning_rate": 1.7026353836055853e-06, + "loss": 0.0008, + "step": 49906 + }, + { + "epoch": 6.54, + "learning_rate": 1.7023393914967967e-06, + "loss": 0.002, + "step": 49907 + }, + { + "epoch": 6.54, + "learning_rate": 1.7020434227247374e-06, + "loss": 0.0013, + "step": 49908 + }, + { + "epoch": 6.54, + "learning_rate": 1.7017474772902353e-06, + "loss": 0.0017, + "step": 49909 + }, + { + "epoch": 6.54, + "learning_rate": 1.701451555194128e-06, + "loss": 0.0016, + "step": 49910 + }, + { + "epoch": 6.54, + "learning_rate": 1.7011556564372477e-06, + "loss": 0.0011, + "step": 49911 + }, + { + "epoch": 6.54, + "learning_rate": 1.7008597810204254e-06, + "loss": 0.0021, + "step": 49912 + }, + { + "epoch": 6.54, + "learning_rate": 1.700563928944491e-06, + "loss": 0.0015, + "step": 49913 + }, + { + "epoch": 6.54, + "learning_rate": 1.7002681002102805e-06, + "loss": 0.0014, + "step": 49914 + }, + { + "epoch": 6.54, + "learning_rate": 1.6999722948186248e-06, + "loss": 0.0011, + "step": 49915 + }, + { + "epoch": 6.54, + "learning_rate": 1.6996765127703542e-06, + "loss": 0.0022, + "step": 49916 + }, + { + "epoch": 6.54, + "learning_rate": 1.6993807540662988e-06, + "loss": 0.0015, + "step": 49917 + }, + { + "epoch": 6.54, + "learning_rate": 1.699085018707296e-06, + "loss": 0.0015, + "step": 49918 + }, + { + "epoch": 6.54, + "learning_rate": 1.698789306694174e-06, + "loss": 0.0026, + "step": 49919 + }, + { + "epoch": 6.54, + "learning_rate": 1.6984936180277633e-06, + "loss": 0.0031, + "step": 49920 + }, + { + "epoch": 6.54, + "learning_rate": 1.6981979527088966e-06, + "loss": 0.0027, + "step": 49921 + }, + { + "epoch": 6.54, + "learning_rate": 1.6979023107384096e-06, + "loss": 0.0026, + "step": 49922 + }, + { + "epoch": 6.54, + "learning_rate": 1.6976066921171285e-06, + "loss": 0.0026, + "step": 49923 + }, + { + "epoch": 6.54, + "learning_rate": 1.6973110968458884e-06, + "loss": 0.003, + "step": 49924 + }, + { + "epoch": 6.54, + "learning_rate": 1.697015524925515e-06, + "loss": 0.0013, + "step": 49925 + }, + { + "epoch": 6.54, + "learning_rate": 1.6967199763568453e-06, + "loss": 0.0022, + "step": 49926 + }, + { + "epoch": 6.54, + "learning_rate": 1.6964244511407092e-06, + "loss": 0.0034, + "step": 49927 + }, + { + "epoch": 6.54, + "learning_rate": 1.6961289492779342e-06, + "loss": 0.0014, + "step": 49928 + }, + { + "epoch": 6.54, + "learning_rate": 1.695833470769358e-06, + "loss": 0.0023, + "step": 49929 + }, + { + "epoch": 6.54, + "learning_rate": 1.6955380156158063e-06, + "loss": 0.0027, + "step": 49930 + }, + { + "epoch": 6.54, + "learning_rate": 1.6952425838181109e-06, + "loss": 0.0027, + "step": 49931 + }, + { + "epoch": 6.54, + "learning_rate": 1.6949471753771052e-06, + "loss": 0.0015, + "step": 49932 + }, + { + "epoch": 6.54, + "learning_rate": 1.6946517902936156e-06, + "loss": 0.0026, + "step": 49933 + }, + { + "epoch": 6.54, + "learning_rate": 1.6943564285684776e-06, + "loss": 0.0016, + "step": 49934 + }, + { + "epoch": 6.54, + "learning_rate": 1.6940610902025202e-06, + "loss": 0.0016, + "step": 49935 + }, + { + "epoch": 6.54, + "learning_rate": 1.693765775196572e-06, + "loss": 0.0015, + "step": 49936 + }, + { + "epoch": 6.54, + "learning_rate": 1.693470483551467e-06, + "loss": 0.003, + "step": 49937 + }, + { + "epoch": 6.54, + "learning_rate": 1.6931752152680335e-06, + "loss": 0.002, + "step": 49938 + }, + { + "epoch": 6.54, + "learning_rate": 1.6928799703471032e-06, + "loss": 0.0019, + "step": 49939 + }, + { + "epoch": 6.54, + "learning_rate": 1.6925847487895032e-06, + "loss": 0.002, + "step": 49940 + }, + { + "epoch": 6.54, + "learning_rate": 1.692289550596068e-06, + "loss": 0.0026, + "step": 49941 + }, + { + "epoch": 6.54, + "learning_rate": 1.691994375767624e-06, + "loss": 0.0012, + "step": 49942 + }, + { + "epoch": 6.54, + "learning_rate": 1.6916992243050057e-06, + "loss": 0.0026, + "step": 49943 + }, + { + "epoch": 6.54, + "learning_rate": 1.6914040962090383e-06, + "loss": 0.0031, + "step": 49944 + }, + { + "epoch": 6.54, + "learning_rate": 1.6911089914805568e-06, + "loss": 0.0012, + "step": 49945 + }, + { + "epoch": 6.54, + "learning_rate": 1.690813910120389e-06, + "loss": 0.0016, + "step": 49946 + }, + { + "epoch": 6.54, + "learning_rate": 1.6905188521293647e-06, + "loss": 0.0034, + "step": 49947 + }, + { + "epoch": 6.54, + "learning_rate": 1.6902238175083109e-06, + "loss": 0.002, + "step": 49948 + }, + { + "epoch": 6.54, + "learning_rate": 1.6899288062580621e-06, + "loss": 0.0021, + "step": 49949 + }, + { + "epoch": 6.54, + "learning_rate": 1.689633818379447e-06, + "loss": 0.0017, + "step": 49950 + }, + { + "epoch": 6.54, + "learning_rate": 1.6893388538732913e-06, + "loss": 0.0021, + "step": 49951 + }, + { + "epoch": 6.54, + "learning_rate": 1.6890439127404301e-06, + "loss": 0.0023, + "step": 49952 + }, + { + "epoch": 6.54, + "learning_rate": 1.6887489949816871e-06, + "loss": 0.0018, + "step": 49953 + }, + { + "epoch": 6.54, + "learning_rate": 1.6884541005978983e-06, + "loss": 0.0018, + "step": 49954 + }, + { + "epoch": 6.54, + "learning_rate": 1.6881592295898896e-06, + "loss": 0.0033, + "step": 49955 + }, + { + "epoch": 6.54, + "learning_rate": 1.6878643819584873e-06, + "loss": 0.0026, + "step": 49956 + }, + { + "epoch": 6.54, + "learning_rate": 1.687569557704527e-06, + "loss": 0.0016, + "step": 49957 + }, + { + "epoch": 6.54, + "learning_rate": 1.687274756828835e-06, + "loss": 0.0013, + "step": 49958 + }, + { + "epoch": 6.54, + "learning_rate": 1.6869799793322362e-06, + "loss": 0.0016, + "step": 49959 + }, + { + "epoch": 6.54, + "learning_rate": 1.6866852252155675e-06, + "loss": 0.0024, + "step": 49960 + }, + { + "epoch": 6.54, + "learning_rate": 1.6863904944796528e-06, + "loss": 0.0013, + "step": 49961 + }, + { + "epoch": 6.54, + "learning_rate": 1.6860957871253224e-06, + "loss": 0.0014, + "step": 49962 + }, + { + "epoch": 6.54, + "learning_rate": 1.6858011031534026e-06, + "loss": 0.0014, + "step": 49963 + }, + { + "epoch": 6.54, + "learning_rate": 1.685506442564725e-06, + "loss": 0.0045, + "step": 49964 + }, + { + "epoch": 6.55, + "learning_rate": 1.6852118053601196e-06, + "loss": 0.0016, + "step": 49965 + }, + { + "epoch": 6.55, + "learning_rate": 1.6849171915404127e-06, + "loss": 0.0007, + "step": 49966 + }, + { + "epoch": 6.55, + "learning_rate": 1.6846226011064315e-06, + "loss": 0.0021, + "step": 49967 + }, + { + "epoch": 6.55, + "learning_rate": 1.6843280340590085e-06, + "loss": 0.0028, + "step": 49968 + }, + { + "epoch": 6.55, + "learning_rate": 1.6840334903989708e-06, + "loss": 0.0028, + "step": 49969 + }, + { + "epoch": 6.55, + "learning_rate": 1.6837389701271456e-06, + "loss": 0.0015, + "step": 49970 + }, + { + "epoch": 6.55, + "learning_rate": 1.683444473244359e-06, + "loss": 0.0017, + "step": 49971 + }, + { + "epoch": 6.55, + "learning_rate": 1.6831499997514445e-06, + "loss": 0.0018, + "step": 49972 + }, + { + "epoch": 6.55, + "learning_rate": 1.6828555496492283e-06, + "loss": 0.0016, + "step": 49973 + }, + { + "epoch": 6.55, + "learning_rate": 1.6825611229385352e-06, + "loss": 0.0031, + "step": 49974 + }, + { + "epoch": 6.55, + "learning_rate": 1.6822667196201959e-06, + "loss": 0.001, + "step": 49975 + }, + { + "epoch": 6.55, + "learning_rate": 1.6819723396950404e-06, + "loss": 0.0011, + "step": 49976 + }, + { + "epoch": 6.55, + "learning_rate": 1.6816779831638952e-06, + "loss": 0.0015, + "step": 49977 + }, + { + "epoch": 6.55, + "learning_rate": 1.681383650027586e-06, + "loss": 0.0028, + "step": 49978 + }, + { + "epoch": 6.55, + "learning_rate": 1.6810893402869432e-06, + "loss": 0.0037, + "step": 49979 + }, + { + "epoch": 6.55, + "learning_rate": 1.6807950539427953e-06, + "loss": 0.0019, + "step": 49980 + }, + { + "epoch": 6.55, + "learning_rate": 1.680500790995967e-06, + "loss": 0.002, + "step": 49981 + }, + { + "epoch": 6.55, + "learning_rate": 1.6802065514472855e-06, + "loss": 0.0021, + "step": 49982 + }, + { + "epoch": 6.55, + "learning_rate": 1.6799123352975821e-06, + "loss": 0.0021, + "step": 49983 + }, + { + "epoch": 6.55, + "learning_rate": 1.6796181425476821e-06, + "loss": 0.0022, + "step": 49984 + }, + { + "epoch": 6.55, + "learning_rate": 1.6793239731984134e-06, + "loss": 0.0027, + "step": 49985 + }, + { + "epoch": 6.55, + "learning_rate": 1.679029827250598e-06, + "loss": 0.0016, + "step": 49986 + }, + { + "epoch": 6.55, + "learning_rate": 1.6787357047050724e-06, + "loss": 0.0025, + "step": 49987 + }, + { + "epoch": 6.55, + "learning_rate": 1.6784416055626596e-06, + "loss": 0.0025, + "step": 49988 + }, + { + "epoch": 6.55, + "learning_rate": 1.6781475298241868e-06, + "loss": 0.0026, + "step": 49989 + }, + { + "epoch": 6.55, + "learning_rate": 1.6778534774904786e-06, + "loss": 0.0022, + "step": 49990 + }, + { + "epoch": 6.55, + "learning_rate": 1.6775594485623669e-06, + "loss": 0.0013, + "step": 49991 + }, + { + "epoch": 6.55, + "learning_rate": 1.6772654430406755e-06, + "loss": 0.002, + "step": 49992 + }, + { + "epoch": 6.55, + "learning_rate": 1.67697146092623e-06, + "loss": 0.0022, + "step": 49993 + }, + { + "epoch": 6.55, + "learning_rate": 1.676677502219861e-06, + "loss": 0.0013, + "step": 49994 + }, + { + "epoch": 6.55, + "learning_rate": 1.6763835669223937e-06, + "loss": 0.0021, + "step": 49995 + }, + { + "epoch": 6.55, + "learning_rate": 1.676089655034654e-06, + "loss": 0.0027, + "step": 49996 + }, + { + "epoch": 6.55, + "learning_rate": 1.6757957665574665e-06, + "loss": 0.0014, + "step": 49997 + }, + { + "epoch": 6.55, + "learning_rate": 1.6755019014916606e-06, + "loss": 0.0017, + "step": 49998 + }, + { + "epoch": 6.55, + "learning_rate": 1.6752080598380637e-06, + "loss": 0.0017, + "step": 49999 + }, + { + "epoch": 6.55, + "learning_rate": 1.6749142415975016e-06, + "loss": 0.0012, + "step": 50000 + }, + { + "epoch": 6.55, + "learning_rate": 1.674620446770797e-06, + "loss": 0.0011, + "step": 50001 + }, + { + "epoch": 6.55, + "learning_rate": 1.6743266753587827e-06, + "loss": 0.0024, + "step": 50002 + }, + { + "epoch": 6.55, + "learning_rate": 1.6740329273622801e-06, + "loss": 0.0015, + "step": 50003 + }, + { + "epoch": 6.55, + "learning_rate": 1.6737392027821164e-06, + "loss": 0.0015, + "step": 50004 + }, + { + "epoch": 6.55, + "learning_rate": 1.6734455016191164e-06, + "loss": 0.0014, + "step": 50005 + }, + { + "epoch": 6.55, + "learning_rate": 1.6731518238741085e-06, + "loss": 0.0016, + "step": 50006 + }, + { + "epoch": 6.55, + "learning_rate": 1.6728581695479163e-06, + "loss": 0.002, + "step": 50007 + }, + { + "epoch": 6.55, + "learning_rate": 1.672564538641369e-06, + "loss": 0.0014, + "step": 50008 + }, + { + "epoch": 6.55, + "learning_rate": 1.6722709311552886e-06, + "loss": 0.0015, + "step": 50009 + }, + { + "epoch": 6.55, + "learning_rate": 1.6719773470905043e-06, + "loss": 0.0014, + "step": 50010 + }, + { + "epoch": 6.55, + "learning_rate": 1.6716837864478408e-06, + "loss": 0.0017, + "step": 50011 + }, + { + "epoch": 6.55, + "learning_rate": 1.671390249228123e-06, + "loss": 0.0021, + "step": 50012 + }, + { + "epoch": 6.55, + "learning_rate": 1.6710967354321739e-06, + "loss": 0.003, + "step": 50013 + }, + { + "epoch": 6.55, + "learning_rate": 1.6708032450608248e-06, + "loss": 0.0014, + "step": 50014 + }, + { + "epoch": 6.55, + "learning_rate": 1.670509778114897e-06, + "loss": 0.0019, + "step": 50015 + }, + { + "epoch": 6.55, + "learning_rate": 1.670216334595216e-06, + "loss": 0.0031, + "step": 50016 + }, + { + "epoch": 6.55, + "learning_rate": 1.6699229145026086e-06, + "loss": 0.0007, + "step": 50017 + }, + { + "epoch": 6.55, + "learning_rate": 1.6696295178378984e-06, + "loss": 0.0017, + "step": 50018 + }, + { + "epoch": 6.55, + "learning_rate": 1.669336144601913e-06, + "loss": 0.0015, + "step": 50019 + }, + { + "epoch": 6.55, + "learning_rate": 1.6690427947954769e-06, + "loss": 0.0021, + "step": 50020 + }, + { + "epoch": 6.55, + "learning_rate": 1.6687494684194116e-06, + "loss": 0.0021, + "step": 50021 + }, + { + "epoch": 6.55, + "learning_rate": 1.6684561654745467e-06, + "loss": 0.0032, + "step": 50022 + }, + { + "epoch": 6.55, + "learning_rate": 1.668162885961706e-06, + "loss": 0.0021, + "step": 50023 + }, + { + "epoch": 6.55, + "learning_rate": 1.6678696298817098e-06, + "loss": 0.0031, + "step": 50024 + }, + { + "epoch": 6.55, + "learning_rate": 1.6675763972353898e-06, + "loss": 0.0014, + "step": 50025 + }, + { + "epoch": 6.55, + "learning_rate": 1.6672831880235673e-06, + "loss": 0.002, + "step": 50026 + }, + { + "epoch": 6.55, + "learning_rate": 1.6669900022470675e-06, + "loss": 0.0024, + "step": 50027 + }, + { + "epoch": 6.55, + "learning_rate": 1.6666968399067119e-06, + "loss": 0.002, + "step": 50028 + }, + { + "epoch": 6.55, + "learning_rate": 1.6664037010033274e-06, + "loss": 0.0017, + "step": 50029 + }, + { + "epoch": 6.55, + "learning_rate": 1.6661105855377414e-06, + "loss": 0.0034, + "step": 50030 + }, + { + "epoch": 6.55, + "learning_rate": 1.665817493510775e-06, + "loss": 0.003, + "step": 50031 + }, + { + "epoch": 6.55, + "learning_rate": 1.6655244249232516e-06, + "loss": 0.0024, + "step": 50032 + }, + { + "epoch": 6.55, + "learning_rate": 1.6652313797759988e-06, + "loss": 0.0023, + "step": 50033 + }, + { + "epoch": 6.55, + "learning_rate": 1.6649383580698387e-06, + "loss": 0.0016, + "step": 50034 + }, + { + "epoch": 6.55, + "learning_rate": 1.6646453598055968e-06, + "loss": 0.0014, + "step": 50035 + }, + { + "epoch": 6.55, + "learning_rate": 1.6643523849840915e-06, + "loss": 0.0025, + "step": 50036 + }, + { + "epoch": 6.55, + "learning_rate": 1.6640594336061556e-06, + "loss": 0.002, + "step": 50037 + }, + { + "epoch": 6.55, + "learning_rate": 1.663766505672607e-06, + "loss": 0.0032, + "step": 50038 + }, + { + "epoch": 6.55, + "learning_rate": 1.66347360118427e-06, + "loss": 0.001, + "step": 50039 + }, + { + "epoch": 6.55, + "learning_rate": 1.6631807201419692e-06, + "loss": 0.0024, + "step": 50040 + }, + { + "epoch": 6.56, + "learning_rate": 1.6628878625465307e-06, + "loss": 0.0017, + "step": 50041 + }, + { + "epoch": 6.56, + "learning_rate": 1.6625950283987756e-06, + "loss": 0.0018, + "step": 50042 + }, + { + "epoch": 6.56, + "learning_rate": 1.6623022176995273e-06, + "loss": 0.0018, + "step": 50043 + }, + { + "epoch": 6.56, + "learning_rate": 1.6620094304496115e-06, + "loss": 0.0018, + "step": 50044 + }, + { + "epoch": 6.56, + "learning_rate": 1.6617166666498496e-06, + "loss": 0.0018, + "step": 50045 + }, + { + "epoch": 6.56, + "learning_rate": 1.6614239263010656e-06, + "loss": 0.0015, + "step": 50046 + }, + { + "epoch": 6.56, + "learning_rate": 1.661131209404081e-06, + "loss": 0.0018, + "step": 50047 + }, + { + "epoch": 6.56, + "learning_rate": 1.660838515959723e-06, + "loss": 0.0028, + "step": 50048 + }, + { + "epoch": 6.56, + "learning_rate": 1.6605458459688128e-06, + "loss": 0.0015, + "step": 50049 + }, + { + "epoch": 6.56, + "learning_rate": 1.6602531994321735e-06, + "loss": 0.0021, + "step": 50050 + }, + { + "epoch": 6.56, + "learning_rate": 1.6599605763506222e-06, + "loss": 0.0015, + "step": 50051 + }, + { + "epoch": 6.56, + "learning_rate": 1.6596679767249936e-06, + "loss": 0.0016, + "step": 50052 + }, + { + "epoch": 6.56, + "learning_rate": 1.6593754005561037e-06, + "loss": 0.0026, + "step": 50053 + }, + { + "epoch": 6.56, + "learning_rate": 1.6590828478447763e-06, + "loss": 0.0011, + "step": 50054 + }, + { + "epoch": 6.56, + "learning_rate": 1.658790318591832e-06, + "loss": 0.0023, + "step": 50055 + }, + { + "epoch": 6.56, + "learning_rate": 1.6584978127980988e-06, + "loss": 0.0023, + "step": 50056 + }, + { + "epoch": 6.56, + "learning_rate": 1.6582053304643953e-06, + "loss": 0.0044, + "step": 50057 + }, + { + "epoch": 6.56, + "learning_rate": 1.6579128715915437e-06, + "loss": 0.0024, + "step": 50058 + }, + { + "epoch": 6.56, + "learning_rate": 1.6576204361803693e-06, + "loss": 0.0013, + "step": 50059 + }, + { + "epoch": 6.56, + "learning_rate": 1.6573280242316936e-06, + "loss": 0.0023, + "step": 50060 + }, + { + "epoch": 6.56, + "learning_rate": 1.6570356357463391e-06, + "loss": 0.0027, + "step": 50061 + }, + { + "epoch": 6.56, + "learning_rate": 1.6567432707251242e-06, + "loss": 0.0014, + "step": 50062 + }, + { + "epoch": 6.56, + "learning_rate": 1.6564509291688756e-06, + "loss": 0.0011, + "step": 50063 + }, + { + "epoch": 6.56, + "learning_rate": 1.6561586110784156e-06, + "loss": 0.002, + "step": 50064 + }, + { + "epoch": 6.56, + "learning_rate": 1.6558663164545663e-06, + "loss": 0.0021, + "step": 50065 + }, + { + "epoch": 6.56, + "learning_rate": 1.655574045298145e-06, + "loss": 0.0015, + "step": 50066 + }, + { + "epoch": 6.56, + "learning_rate": 1.655281797609981e-06, + "loss": 0.0011, + "step": 50067 + }, + { + "epoch": 6.56, + "learning_rate": 1.6549895733908928e-06, + "loss": 0.0013, + "step": 50068 + }, + { + "epoch": 6.56, + "learning_rate": 1.6546973726417003e-06, + "loss": 0.0018, + "step": 50069 + }, + { + "epoch": 6.56, + "learning_rate": 1.6544051953632256e-06, + "loss": 0.0021, + "step": 50070 + }, + { + "epoch": 6.56, + "learning_rate": 1.6541130415562946e-06, + "loss": 0.0011, + "step": 50071 + }, + { + "epoch": 6.56, + "learning_rate": 1.6538209112217262e-06, + "loss": 0.0021, + "step": 50072 + }, + { + "epoch": 6.56, + "learning_rate": 1.6535288043603393e-06, + "loss": 0.0019, + "step": 50073 + }, + { + "epoch": 6.56, + "learning_rate": 1.6532367209729583e-06, + "loss": 0.0023, + "step": 50074 + }, + { + "epoch": 6.56, + "learning_rate": 1.6529446610604062e-06, + "loss": 0.0011, + "step": 50075 + }, + { + "epoch": 6.56, + "learning_rate": 1.652652624623504e-06, + "loss": 0.0016, + "step": 50076 + }, + { + "epoch": 6.56, + "learning_rate": 1.6523606116630707e-06, + "loss": 0.0031, + "step": 50077 + }, + { + "epoch": 6.56, + "learning_rate": 1.6520686221799264e-06, + "loss": 0.0027, + "step": 50078 + }, + { + "epoch": 6.56, + "learning_rate": 1.651776656174897e-06, + "loss": 0.0023, + "step": 50079 + }, + { + "epoch": 6.56, + "learning_rate": 1.6514847136488022e-06, + "loss": 0.002, + "step": 50080 + }, + { + "epoch": 6.56, + "learning_rate": 1.6511927946024586e-06, + "loss": 0.0024, + "step": 50081 + }, + { + "epoch": 6.56, + "learning_rate": 1.6509008990366938e-06, + "loss": 0.0014, + "step": 50082 + }, + { + "epoch": 6.56, + "learning_rate": 1.6506090269523245e-06, + "loss": 0.0018, + "step": 50083 + }, + { + "epoch": 6.56, + "learning_rate": 1.6503171783501714e-06, + "loss": 0.0021, + "step": 50084 + }, + { + "epoch": 6.56, + "learning_rate": 1.6500253532310583e-06, + "loss": 0.0018, + "step": 50085 + }, + { + "epoch": 6.56, + "learning_rate": 1.6497335515958025e-06, + "loss": 0.0019, + "step": 50086 + }, + { + "epoch": 6.56, + "learning_rate": 1.6494417734452283e-06, + "loss": 0.0027, + "step": 50087 + }, + { + "epoch": 6.56, + "learning_rate": 1.6491500187801546e-06, + "loss": 0.0018, + "step": 50088 + }, + { + "epoch": 6.56, + "learning_rate": 1.6488582876013993e-06, + "loss": 0.0016, + "step": 50089 + }, + { + "epoch": 6.56, + "learning_rate": 1.6485665799097873e-06, + "loss": 0.0015, + "step": 50090 + }, + { + "epoch": 6.56, + "learning_rate": 1.6482748957061367e-06, + "loss": 0.0021, + "step": 50091 + }, + { + "epoch": 6.56, + "learning_rate": 1.6479832349912694e-06, + "loss": 0.0015, + "step": 50092 + }, + { + "epoch": 6.56, + "learning_rate": 1.6476915977660014e-06, + "loss": 0.0023, + "step": 50093 + }, + { + "epoch": 6.56, + "learning_rate": 1.6473999840311584e-06, + "loss": 0.0021, + "step": 50094 + }, + { + "epoch": 6.56, + "learning_rate": 1.6471083937875565e-06, + "loss": 0.002, + "step": 50095 + }, + { + "epoch": 6.56, + "learning_rate": 1.6468168270360197e-06, + "loss": 0.0017, + "step": 50096 + }, + { + "epoch": 6.56, + "learning_rate": 1.6465252837773627e-06, + "loss": 0.0037, + "step": 50097 + }, + { + "epoch": 6.56, + "learning_rate": 1.6462337640124104e-06, + "loss": 0.001, + "step": 50098 + }, + { + "epoch": 6.56, + "learning_rate": 1.6459422677419823e-06, + "loss": 0.0023, + "step": 50099 + }, + { + "epoch": 6.56, + "learning_rate": 1.6456507949668953e-06, + "loss": 0.0019, + "step": 50100 + }, + { + "epoch": 6.56, + "learning_rate": 1.645359345687969e-06, + "loss": 0.0016, + "step": 50101 + }, + { + "epoch": 6.56, + "learning_rate": 1.645067919906026e-06, + "loss": 0.0015, + "step": 50102 + }, + { + "epoch": 6.56, + "learning_rate": 1.6447765176218854e-06, + "loss": 0.003, + "step": 50103 + }, + { + "epoch": 6.56, + "learning_rate": 1.6444851388363637e-06, + "loss": 0.0022, + "step": 50104 + }, + { + "epoch": 6.56, + "learning_rate": 1.6441937835502852e-06, + "loss": 0.0038, + "step": 50105 + }, + { + "epoch": 6.56, + "learning_rate": 1.643902451764463e-06, + "loss": 0.0031, + "step": 50106 + }, + { + "epoch": 6.56, + "learning_rate": 1.6436111434797242e-06, + "loss": 0.0014, + "step": 50107 + }, + { + "epoch": 6.56, + "learning_rate": 1.64331985869688e-06, + "loss": 0.0009, + "step": 50108 + }, + { + "epoch": 6.56, + "learning_rate": 1.643028597416757e-06, + "loss": 0.0016, + "step": 50109 + }, + { + "epoch": 6.56, + "learning_rate": 1.6427373596401708e-06, + "loss": 0.0009, + "step": 50110 + }, + { + "epoch": 6.56, + "learning_rate": 1.6424461453679409e-06, + "loss": 0.002, + "step": 50111 + }, + { + "epoch": 6.56, + "learning_rate": 1.6421549546008841e-06, + "loss": 0.0013, + "step": 50112 + }, + { + "epoch": 6.56, + "learning_rate": 1.6418637873398236e-06, + "loss": 0.0021, + "step": 50113 + }, + { + "epoch": 6.56, + "learning_rate": 1.6415726435855762e-06, + "loss": 0.0019, + "step": 50114 + }, + { + "epoch": 6.56, + "learning_rate": 1.64128152333896e-06, + "loss": 0.0007, + "step": 50115 + }, + { + "epoch": 6.56, + "learning_rate": 1.6409904266007925e-06, + "loss": 0.0019, + "step": 50116 + }, + { + "epoch": 6.56, + "learning_rate": 1.640699353371894e-06, + "loss": 0.0012, + "step": 50117 + }, + { + "epoch": 6.57, + "learning_rate": 1.640408303653086e-06, + "loss": 0.0028, + "step": 50118 + }, + { + "epoch": 6.57, + "learning_rate": 1.6401172774451834e-06, + "loss": 0.0017, + "step": 50119 + }, + { + "epoch": 6.57, + "learning_rate": 1.6398262747490045e-06, + "loss": 0.0018, + "step": 50120 + }, + { + "epoch": 6.57, + "learning_rate": 1.639535295565372e-06, + "loss": 0.0026, + "step": 50121 + }, + { + "epoch": 6.57, + "learning_rate": 1.6392443398950997e-06, + "loss": 0.0017, + "step": 50122 + }, + { + "epoch": 6.57, + "learning_rate": 1.6389534077390068e-06, + "loss": 0.0012, + "step": 50123 + }, + { + "epoch": 6.57, + "learning_rate": 1.6386624990979127e-06, + "loss": 0.0009, + "step": 50124 + }, + { + "epoch": 6.57, + "learning_rate": 1.6383716139726368e-06, + "loss": 0.0016, + "step": 50125 + }, + { + "epoch": 6.57, + "learning_rate": 1.638080752363994e-06, + "loss": 0.0032, + "step": 50126 + }, + { + "epoch": 6.57, + "learning_rate": 1.6377899142728027e-06, + "loss": 0.002, + "step": 50127 + }, + { + "epoch": 6.57, + "learning_rate": 1.6374990996998807e-06, + "loss": 0.0029, + "step": 50128 + }, + { + "epoch": 6.57, + "learning_rate": 1.6372083086460511e-06, + "loss": 0.0013, + "step": 50129 + }, + { + "epoch": 6.57, + "learning_rate": 1.6369175411121264e-06, + "loss": 0.0008, + "step": 50130 + }, + { + "epoch": 6.57, + "learning_rate": 1.6366267970989235e-06, + "loss": 0.0016, + "step": 50131 + }, + { + "epoch": 6.57, + "learning_rate": 1.6363360766072656e-06, + "loss": 0.0017, + "step": 50132 + }, + { + "epoch": 6.57, + "learning_rate": 1.636045379637966e-06, + "loss": 0.0032, + "step": 50133 + }, + { + "epoch": 6.57, + "learning_rate": 1.6357547061918444e-06, + "loss": 0.003, + "step": 50134 + }, + { + "epoch": 6.57, + "learning_rate": 1.6354640562697145e-06, + "loss": 0.0027, + "step": 50135 + }, + { + "epoch": 6.57, + "learning_rate": 1.6351734298723987e-06, + "loss": 0.0024, + "step": 50136 + }, + { + "epoch": 6.57, + "learning_rate": 1.6348828270007122e-06, + "loss": 0.0019, + "step": 50137 + }, + { + "epoch": 6.57, + "learning_rate": 1.6345922476554698e-06, + "loss": 0.0016, + "step": 50138 + }, + { + "epoch": 6.57, + "learning_rate": 1.634301691837491e-06, + "loss": 0.003, + "step": 50139 + }, + { + "epoch": 6.57, + "learning_rate": 1.634011159547596e-06, + "loss": 0.0023, + "step": 50140 + }, + { + "epoch": 6.57, + "learning_rate": 1.6337206507865987e-06, + "loss": 0.0021, + "step": 50141 + }, + { + "epoch": 6.57, + "learning_rate": 1.6334301655553175e-06, + "loss": 0.0013, + "step": 50142 + }, + { + "epoch": 6.57, + "learning_rate": 1.633139703854565e-06, + "loss": 0.0019, + "step": 50143 + }, + { + "epoch": 6.57, + "learning_rate": 1.6328492656851647e-06, + "loss": 0.0016, + "step": 50144 + }, + { + "epoch": 6.57, + "learning_rate": 1.6325588510479295e-06, + "loss": 0.0018, + "step": 50145 + }, + { + "epoch": 6.57, + "learning_rate": 1.6322684599436754e-06, + "loss": 0.002, + "step": 50146 + }, + { + "epoch": 6.57, + "learning_rate": 1.631978092373222e-06, + "loss": 0.0015, + "step": 50147 + }, + { + "epoch": 6.57, + "learning_rate": 1.6316877483373861e-06, + "loss": 0.0022, + "step": 50148 + }, + { + "epoch": 6.57, + "learning_rate": 1.6313974278369793e-06, + "loss": 0.0016, + "step": 50149 + }, + { + "epoch": 6.57, + "learning_rate": 1.6311071308728232e-06, + "loss": 0.0024, + "step": 50150 + }, + { + "epoch": 6.57, + "learning_rate": 1.6308168574457318e-06, + "loss": 0.0018, + "step": 50151 + }, + { + "epoch": 6.57, + "learning_rate": 1.630526607556524e-06, + "loss": 0.0013, + "step": 50152 + }, + { + "epoch": 6.57, + "learning_rate": 1.630236381206014e-06, + "loss": 0.0016, + "step": 50153 + }, + { + "epoch": 6.57, + "learning_rate": 1.6299461783950166e-06, + "loss": 0.0016, + "step": 50154 + }, + { + "epoch": 6.57, + "learning_rate": 1.629655999124352e-06, + "loss": 0.0015, + "step": 50155 + }, + { + "epoch": 6.57, + "learning_rate": 1.6293658433948346e-06, + "loss": 0.0012, + "step": 50156 + }, + { + "epoch": 6.57, + "learning_rate": 1.6290757112072796e-06, + "loss": 0.0013, + "step": 50157 + }, + { + "epoch": 6.57, + "learning_rate": 1.6287856025625015e-06, + "loss": 0.0014, + "step": 50158 + }, + { + "epoch": 6.57, + "learning_rate": 1.6284955174613205e-06, + "loss": 0.0027, + "step": 50159 + }, + { + "epoch": 6.57, + "learning_rate": 1.6282054559045468e-06, + "loss": 0.0028, + "step": 50160 + }, + { + "epoch": 6.57, + "learning_rate": 1.6279154178930023e-06, + "loss": 0.002, + "step": 50161 + }, + { + "epoch": 6.57, + "learning_rate": 1.6276254034274973e-06, + "loss": 0.0027, + "step": 50162 + }, + { + "epoch": 6.57, + "learning_rate": 1.6273354125088525e-06, + "loss": 0.0012, + "step": 50163 + }, + { + "epoch": 6.57, + "learning_rate": 1.6270454451378813e-06, + "loss": 0.0025, + "step": 50164 + }, + { + "epoch": 6.57, + "learning_rate": 1.6267555013153979e-06, + "loss": 0.0012, + "step": 50165 + }, + { + "epoch": 6.57, + "learning_rate": 1.626465581042218e-06, + "loss": 0.0011, + "step": 50166 + }, + { + "epoch": 6.57, + "learning_rate": 1.626175684319159e-06, + "loss": 0.001, + "step": 50167 + }, + { + "epoch": 6.57, + "learning_rate": 1.6258858111470355e-06, + "loss": 0.0015, + "step": 50168 + }, + { + "epoch": 6.57, + "learning_rate": 1.6255959615266593e-06, + "loss": 0.0028, + "step": 50169 + }, + { + "epoch": 6.57, + "learning_rate": 1.6253061354588507e-06, + "loss": 0.0027, + "step": 50170 + }, + { + "epoch": 6.57, + "learning_rate": 1.6250163329444202e-06, + "loss": 0.0031, + "step": 50171 + }, + { + "epoch": 6.57, + "learning_rate": 1.6247265539841883e-06, + "loss": 0.0016, + "step": 50172 + }, + { + "epoch": 6.57, + "learning_rate": 1.6244367985789643e-06, + "loss": 0.0015, + "step": 50173 + }, + { + "epoch": 6.57, + "learning_rate": 1.6241470667295678e-06, + "loss": 0.002, + "step": 50174 + }, + { + "epoch": 6.57, + "learning_rate": 1.6238573584368123e-06, + "loss": 0.0026, + "step": 50175 + }, + { + "epoch": 6.57, + "learning_rate": 1.6235676737015116e-06, + "loss": 0.0017, + "step": 50176 + }, + { + "epoch": 6.57, + "learning_rate": 1.6232780125244784e-06, + "loss": 0.0012, + "step": 50177 + }, + { + "epoch": 6.57, + "learning_rate": 1.6229883749065323e-06, + "loss": 0.0013, + "step": 50178 + }, + { + "epoch": 6.57, + "learning_rate": 1.6226987608484846e-06, + "loss": 0.0021, + "step": 50179 + }, + { + "epoch": 6.57, + "learning_rate": 1.622409170351149e-06, + "loss": 0.0015, + "step": 50180 + }, + { + "epoch": 6.57, + "learning_rate": 1.622119603415343e-06, + "loss": 0.0016, + "step": 50181 + }, + { + "epoch": 6.57, + "learning_rate": 1.621830060041878e-06, + "loss": 0.0023, + "step": 50182 + }, + { + "epoch": 6.57, + "learning_rate": 1.621540540231572e-06, + "loss": 0.0022, + "step": 50183 + }, + { + "epoch": 6.57, + "learning_rate": 1.621251043985237e-06, + "loss": 0.0022, + "step": 50184 + }, + { + "epoch": 6.57, + "learning_rate": 1.6209615713036842e-06, + "loss": 0.0019, + "step": 50185 + }, + { + "epoch": 6.57, + "learning_rate": 1.620672122187733e-06, + "loss": 0.0014, + "step": 50186 + }, + { + "epoch": 6.57, + "learning_rate": 1.6203826966381964e-06, + "loss": 0.0021, + "step": 50187 + }, + { + "epoch": 6.57, + "learning_rate": 1.6200932946558845e-06, + "loss": 0.0033, + "step": 50188 + }, + { + "epoch": 6.57, + "learning_rate": 1.6198039162416157e-06, + "loss": 0.0011, + "step": 50189 + }, + { + "epoch": 6.57, + "learning_rate": 1.6195145613962027e-06, + "loss": 0.0015, + "step": 50190 + }, + { + "epoch": 6.57, + "learning_rate": 1.619225230120458e-06, + "loss": 0.0014, + "step": 50191 + }, + { + "epoch": 6.57, + "learning_rate": 1.6189359224151945e-06, + "loss": 0.0016, + "step": 50192 + }, + { + "epoch": 6.57, + "learning_rate": 1.618646638281227e-06, + "loss": 0.0012, + "step": 50193 + }, + { + "epoch": 6.58, + "learning_rate": 1.6183573777193716e-06, + "loss": 0.0014, + "step": 50194 + }, + { + "epoch": 6.58, + "learning_rate": 1.618068140730441e-06, + "loss": 0.0014, + "step": 50195 + }, + { + "epoch": 6.58, + "learning_rate": 1.6177789273152433e-06, + "loss": 0.002, + "step": 50196 + }, + { + "epoch": 6.58, + "learning_rate": 1.617489737474599e-06, + "loss": 0.0013, + "step": 50197 + }, + { + "epoch": 6.58, + "learning_rate": 1.6172005712093196e-06, + "loss": 0.0009, + "step": 50198 + }, + { + "epoch": 6.58, + "learning_rate": 1.616911428520216e-06, + "loss": 0.003, + "step": 50199 + }, + { + "epoch": 6.58, + "learning_rate": 1.6166223094081e-06, + "loss": 0.0023, + "step": 50200 + }, + { + "epoch": 6.58, + "learning_rate": 1.6163332138737896e-06, + "loss": 0.0015, + "step": 50201 + }, + { + "epoch": 6.58, + "learning_rate": 1.616044141918096e-06, + "loss": 0.0018, + "step": 50202 + }, + { + "epoch": 6.58, + "learning_rate": 1.6157550935418297e-06, + "loss": 0.0024, + "step": 50203 + }, + { + "epoch": 6.58, + "learning_rate": 1.6154660687458057e-06, + "loss": 0.0015, + "step": 50204 + }, + { + "epoch": 6.58, + "learning_rate": 1.6151770675308387e-06, + "loss": 0.0015, + "step": 50205 + }, + { + "epoch": 6.58, + "learning_rate": 1.6148880898977404e-06, + "loss": 0.0014, + "step": 50206 + }, + { + "epoch": 6.58, + "learning_rate": 1.6145991358473223e-06, + "loss": 0.002, + "step": 50207 + }, + { + "epoch": 6.58, + "learning_rate": 1.6143102053803949e-06, + "loss": 0.0019, + "step": 50208 + }, + { + "epoch": 6.58, + "learning_rate": 1.6140212984977766e-06, + "loss": 0.0018, + "step": 50209 + }, + { + "epoch": 6.58, + "learning_rate": 1.6137324152002766e-06, + "loss": 0.0025, + "step": 50210 + }, + { + "epoch": 6.58, + "learning_rate": 1.6134435554887052e-06, + "loss": 0.002, + "step": 50211 + }, + { + "epoch": 6.58, + "learning_rate": 1.6131547193638797e-06, + "loss": 0.0009, + "step": 50212 + }, + { + "epoch": 6.58, + "learning_rate": 1.6128659068266095e-06, + "loss": 0.0022, + "step": 50213 + }, + { + "epoch": 6.58, + "learning_rate": 1.6125771178777071e-06, + "loss": 0.0016, + "step": 50214 + }, + { + "epoch": 6.58, + "learning_rate": 1.6122883525179834e-06, + "loss": 0.0008, + "step": 50215 + }, + { + "epoch": 6.58, + "learning_rate": 1.6119996107482516e-06, + "loss": 0.0025, + "step": 50216 + }, + { + "epoch": 6.58, + "learning_rate": 1.6117108925693259e-06, + "loss": 0.0014, + "step": 50217 + }, + { + "epoch": 6.58, + "learning_rate": 1.6114221979820177e-06, + "loss": 0.0015, + "step": 50218 + }, + { + "epoch": 6.58, + "learning_rate": 1.6111335269871354e-06, + "loss": 0.0017, + "step": 50219 + }, + { + "epoch": 6.58, + "learning_rate": 1.6108448795854947e-06, + "loss": 0.0007, + "step": 50220 + }, + { + "epoch": 6.58, + "learning_rate": 1.6105562557779065e-06, + "loss": 0.0035, + "step": 50221 + }, + { + "epoch": 6.58, + "learning_rate": 1.610267655565182e-06, + "loss": 0.0018, + "step": 50222 + }, + { + "epoch": 6.58, + "learning_rate": 1.6099790789481307e-06, + "loss": 0.0012, + "step": 50223 + }, + { + "epoch": 6.58, + "learning_rate": 1.6096905259275697e-06, + "loss": 0.0009, + "step": 50224 + }, + { + "epoch": 6.58, + "learning_rate": 1.609401996504306e-06, + "loss": 0.0022, + "step": 50225 + }, + { + "epoch": 6.58, + "learning_rate": 1.6091134906791506e-06, + "loss": 0.0023, + "step": 50226 + }, + { + "epoch": 6.58, + "learning_rate": 1.6088250084529166e-06, + "loss": 0.0011, + "step": 50227 + }, + { + "epoch": 6.58, + "learning_rate": 1.608536549826417e-06, + "loss": 0.0019, + "step": 50228 + }, + { + "epoch": 6.58, + "learning_rate": 1.6082481148004626e-06, + "loss": 0.0018, + "step": 50229 + }, + { + "epoch": 6.58, + "learning_rate": 1.6079597033758609e-06, + "loss": 0.0023, + "step": 50230 + }, + { + "epoch": 6.58, + "learning_rate": 1.6076713155534273e-06, + "loss": 0.0018, + "step": 50231 + }, + { + "epoch": 6.58, + "learning_rate": 1.6073829513339723e-06, + "loss": 0.0016, + "step": 50232 + }, + { + "epoch": 6.58, + "learning_rate": 1.6070946107183049e-06, + "loss": 0.0014, + "step": 50233 + }, + { + "epoch": 6.58, + "learning_rate": 1.606806293707235e-06, + "loss": 0.0014, + "step": 50234 + }, + { + "epoch": 6.58, + "learning_rate": 1.6065180003015779e-06, + "loss": 0.0014, + "step": 50235 + }, + { + "epoch": 6.58, + "learning_rate": 1.6062297305021413e-06, + "loss": 0.0025, + "step": 50236 + }, + { + "epoch": 6.58, + "learning_rate": 1.6059414843097355e-06, + "loss": 0.0026, + "step": 50237 + }, + { + "epoch": 6.58, + "learning_rate": 1.6056532617251709e-06, + "loss": 0.0023, + "step": 50238 + }, + { + "epoch": 6.58, + "learning_rate": 1.6053650627492623e-06, + "loss": 0.0022, + "step": 50239 + }, + { + "epoch": 6.58, + "learning_rate": 1.6050768873828182e-06, + "loss": 0.0022, + "step": 50240 + }, + { + "epoch": 6.58, + "learning_rate": 1.6047887356266468e-06, + "loss": 0.002, + "step": 50241 + }, + { + "epoch": 6.58, + "learning_rate": 1.6045006074815584e-06, + "loss": 0.0024, + "step": 50242 + }, + { + "epoch": 6.58, + "learning_rate": 1.604212502948368e-06, + "loss": 0.0015, + "step": 50243 + }, + { + "epoch": 6.58, + "learning_rate": 1.6039244220278817e-06, + "loss": 0.0024, + "step": 50244 + }, + { + "epoch": 6.58, + "learning_rate": 1.6036363647209097e-06, + "loss": 0.0017, + "step": 50245 + }, + { + "epoch": 6.58, + "learning_rate": 1.603348331028265e-06, + "loss": 0.0019, + "step": 50246 + }, + { + "epoch": 6.58, + "learning_rate": 1.6030603209507555e-06, + "loss": 0.0019, + "step": 50247 + }, + { + "epoch": 6.58, + "learning_rate": 1.6027723344891898e-06, + "loss": 0.0011, + "step": 50248 + }, + { + "epoch": 6.58, + "learning_rate": 1.6024843716443817e-06, + "loss": 0.0022, + "step": 50249 + }, + { + "epoch": 6.58, + "learning_rate": 1.602196432417137e-06, + "loss": 0.0043, + "step": 50250 + }, + { + "epoch": 6.58, + "learning_rate": 1.6019085168082704e-06, + "loss": 0.0018, + "step": 50251 + }, + { + "epoch": 6.58, + "learning_rate": 1.6016206248185884e-06, + "loss": 0.0013, + "step": 50252 + }, + { + "epoch": 6.58, + "learning_rate": 1.601332756448899e-06, + "loss": 0.0021, + "step": 50253 + }, + { + "epoch": 6.58, + "learning_rate": 1.601044911700016e-06, + "loss": 0.0013, + "step": 50254 + }, + { + "epoch": 6.58, + "learning_rate": 1.6007570905727476e-06, + "loss": 0.0019, + "step": 50255 + }, + { + "epoch": 6.58, + "learning_rate": 1.6004692930679023e-06, + "loss": 0.0025, + "step": 50256 + }, + { + "epoch": 6.58, + "learning_rate": 1.6001815191862868e-06, + "loss": 0.0018, + "step": 50257 + }, + { + "epoch": 6.58, + "learning_rate": 1.5998937689287163e-06, + "loss": 0.0019, + "step": 50258 + }, + { + "epoch": 6.58, + "learning_rate": 1.5996060422959959e-06, + "loss": 0.002, + "step": 50259 + }, + { + "epoch": 6.58, + "learning_rate": 1.5993183392889367e-06, + "loss": 0.0017, + "step": 50260 + }, + { + "epoch": 6.58, + "learning_rate": 1.599030659908346e-06, + "loss": 0.0013, + "step": 50261 + }, + { + "epoch": 6.58, + "learning_rate": 1.5987430041550366e-06, + "loss": 0.0021, + "step": 50262 + }, + { + "epoch": 6.58, + "learning_rate": 1.5984553720298146e-06, + "loss": 0.0031, + "step": 50263 + }, + { + "epoch": 6.58, + "learning_rate": 1.5981677635334902e-06, + "loss": 0.0006, + "step": 50264 + }, + { + "epoch": 6.58, + "learning_rate": 1.5978801786668685e-06, + "loss": 0.0016, + "step": 50265 + }, + { + "epoch": 6.58, + "learning_rate": 1.5975926174307632e-06, + "loss": 0.003, + "step": 50266 + }, + { + "epoch": 6.58, + "learning_rate": 1.5973050798259825e-06, + "loss": 0.0026, + "step": 50267 + }, + { + "epoch": 6.58, + "learning_rate": 1.5970175658533304e-06, + "loss": 0.0019, + "step": 50268 + }, + { + "epoch": 6.58, + "learning_rate": 1.5967300755136218e-06, + "loss": 0.0016, + "step": 50269 + }, + { + "epoch": 6.59, + "learning_rate": 1.5964426088076602e-06, + "loss": 0.0029, + "step": 50270 + }, + { + "epoch": 6.59, + "learning_rate": 1.5961551657362584e-06, + "loss": 0.0015, + "step": 50271 + }, + { + "epoch": 6.59, + "learning_rate": 1.5958677463002225e-06, + "loss": 0.0045, + "step": 50272 + }, + { + "epoch": 6.59, + "learning_rate": 1.5955803505003586e-06, + "loss": 0.0011, + "step": 50273 + }, + { + "epoch": 6.59, + "learning_rate": 1.5952929783374805e-06, + "loss": 0.0018, + "step": 50274 + }, + { + "epoch": 6.59, + "learning_rate": 1.5950056298123929e-06, + "loss": 0.0029, + "step": 50275 + }, + { + "epoch": 6.59, + "learning_rate": 1.5947183049259007e-06, + "loss": 0.0033, + "step": 50276 + }, + { + "epoch": 6.59, + "learning_rate": 1.59443100367882e-06, + "loss": 0.0014, + "step": 50277 + }, + { + "epoch": 6.59, + "learning_rate": 1.5941437260719539e-06, + "loss": 0.0008, + "step": 50278 + }, + { + "epoch": 6.59, + "learning_rate": 1.5938564721061101e-06, + "loss": 0.0017, + "step": 50279 + }, + { + "epoch": 6.59, + "learning_rate": 1.5935692417820937e-06, + "loss": 0.0012, + "step": 50280 + }, + { + "epoch": 6.59, + "learning_rate": 1.5932820351007194e-06, + "loss": 0.0016, + "step": 50281 + }, + { + "epoch": 6.59, + "learning_rate": 1.5929948520627935e-06, + "loss": 0.0017, + "step": 50282 + }, + { + "epoch": 6.59, + "learning_rate": 1.592707692669121e-06, + "loss": 0.0024, + "step": 50283 + }, + { + "epoch": 6.59, + "learning_rate": 1.5924205569205076e-06, + "loss": 0.0019, + "step": 50284 + }, + { + "epoch": 6.59, + "learning_rate": 1.592133444817766e-06, + "loss": 0.0018, + "step": 50285 + }, + { + "epoch": 6.59, + "learning_rate": 1.5918463563617025e-06, + "loss": 0.0014, + "step": 50286 + }, + { + "epoch": 6.59, + "learning_rate": 1.591559291553123e-06, + "loss": 0.0016, + "step": 50287 + }, + { + "epoch": 6.59, + "learning_rate": 1.5912722503928323e-06, + "loss": 0.0017, + "step": 50288 + }, + { + "epoch": 6.59, + "learning_rate": 1.5909852328816432e-06, + "loss": 0.0017, + "step": 50289 + }, + { + "epoch": 6.59, + "learning_rate": 1.5906982390203607e-06, + "loss": 0.0028, + "step": 50290 + }, + { + "epoch": 6.59, + "learning_rate": 1.5904112688097883e-06, + "loss": 0.0016, + "step": 50291 + }, + { + "epoch": 6.59, + "learning_rate": 1.5901243222507367e-06, + "loss": 0.0016, + "step": 50292 + }, + { + "epoch": 6.59, + "learning_rate": 1.5898373993440153e-06, + "loss": 0.0022, + "step": 50293 + }, + { + "epoch": 6.59, + "learning_rate": 1.5895505000904288e-06, + "loss": 0.0014, + "step": 50294 + }, + { + "epoch": 6.59, + "learning_rate": 1.58926362449078e-06, + "loss": 0.0025, + "step": 50295 + }, + { + "epoch": 6.59, + "learning_rate": 1.5889767725458826e-06, + "loss": 0.0011, + "step": 50296 + }, + { + "epoch": 6.59, + "learning_rate": 1.5886899442565395e-06, + "loss": 0.0014, + "step": 50297 + }, + { + "epoch": 6.59, + "learning_rate": 1.588403139623559e-06, + "loss": 0.0011, + "step": 50298 + }, + { + "epoch": 6.59, + "learning_rate": 1.5881163586477432e-06, + "loss": 0.0025, + "step": 50299 + }, + { + "epoch": 6.59, + "learning_rate": 1.5878296013299055e-06, + "loss": 0.0016, + "step": 50300 + }, + { + "epoch": 6.59, + "learning_rate": 1.5875428676708482e-06, + "loss": 0.0025, + "step": 50301 + }, + { + "epoch": 6.59, + "learning_rate": 1.5872561576713775e-06, + "loss": 0.0013, + "step": 50302 + }, + { + "epoch": 6.59, + "learning_rate": 1.5869694713323003e-06, + "loss": 0.0023, + "step": 50303 + }, + { + "epoch": 6.59, + "learning_rate": 1.586682808654425e-06, + "loss": 0.0012, + "step": 50304 + }, + { + "epoch": 6.59, + "learning_rate": 1.5863961696385565e-06, + "loss": 0.0021, + "step": 50305 + }, + { + "epoch": 6.59, + "learning_rate": 1.586109554285501e-06, + "loss": 0.0022, + "step": 50306 + }, + { + "epoch": 6.59, + "learning_rate": 1.585822962596062e-06, + "loss": 0.0033, + "step": 50307 + }, + { + "epoch": 6.59, + "learning_rate": 1.5855363945710512e-06, + "loss": 0.0009, + "step": 50308 + }, + { + "epoch": 6.59, + "learning_rate": 1.58524985021127e-06, + "loss": 0.0025, + "step": 50309 + }, + { + "epoch": 6.59, + "learning_rate": 1.5849633295175227e-06, + "loss": 0.0017, + "step": 50310 + }, + { + "epoch": 6.59, + "learning_rate": 1.5846768324906215e-06, + "loss": 0.0033, + "step": 50311 + }, + { + "epoch": 6.59, + "learning_rate": 1.5843903591313681e-06, + "loss": 0.0026, + "step": 50312 + }, + { + "epoch": 6.59, + "learning_rate": 1.5841039094405663e-06, + "loss": 0.0015, + "step": 50313 + }, + { + "epoch": 6.59, + "learning_rate": 1.5838174834190257e-06, + "loss": 0.0011, + "step": 50314 + }, + { + "epoch": 6.59, + "learning_rate": 1.5835310810675487e-06, + "loss": 0.0022, + "step": 50315 + }, + { + "epoch": 6.59, + "learning_rate": 1.5832447023869447e-06, + "loss": 0.0027, + "step": 50316 + }, + { + "epoch": 6.59, + "learning_rate": 1.5829583473780163e-06, + "loss": 0.0014, + "step": 50317 + }, + { + "epoch": 6.59, + "learning_rate": 1.5826720160415675e-06, + "loss": 0.0023, + "step": 50318 + }, + { + "epoch": 6.59, + "learning_rate": 1.5823857083784066e-06, + "loss": 0.0016, + "step": 50319 + }, + { + "epoch": 6.59, + "learning_rate": 1.5820994243893394e-06, + "loss": 0.0013, + "step": 50320 + }, + { + "epoch": 6.59, + "learning_rate": 1.5818131640751678e-06, + "loss": 0.0023, + "step": 50321 + }, + { + "epoch": 6.59, + "learning_rate": 1.5815269274366973e-06, + "loss": 0.0016, + "step": 50322 + }, + { + "epoch": 6.59, + "learning_rate": 1.5812407144747356e-06, + "loss": 0.0019, + "step": 50323 + }, + { + "epoch": 6.59, + "learning_rate": 1.5809545251900838e-06, + "loss": 0.0026, + "step": 50324 + }, + { + "epoch": 6.59, + "learning_rate": 1.5806683595835516e-06, + "loss": 0.0024, + "step": 50325 + }, + { + "epoch": 6.59, + "learning_rate": 1.580382217655938e-06, + "loss": 0.0028, + "step": 50326 + }, + { + "epoch": 6.59, + "learning_rate": 1.580096099408054e-06, + "loss": 0.0035, + "step": 50327 + }, + { + "epoch": 6.59, + "learning_rate": 1.5798100048407018e-06, + "loss": 0.0018, + "step": 50328 + }, + { + "epoch": 6.59, + "learning_rate": 1.5795239339546853e-06, + "loss": 0.0017, + "step": 50329 + }, + { + "epoch": 6.59, + "learning_rate": 1.5792378867508063e-06, + "loss": 0.0011, + "step": 50330 + }, + { + "epoch": 6.59, + "learning_rate": 1.578951863229875e-06, + "loss": 0.0024, + "step": 50331 + }, + { + "epoch": 6.59, + "learning_rate": 1.5786658633926944e-06, + "loss": 0.0021, + "step": 50332 + }, + { + "epoch": 6.59, + "learning_rate": 1.5783798872400635e-06, + "loss": 0.0022, + "step": 50333 + }, + { + "epoch": 6.59, + "learning_rate": 1.578093934772794e-06, + "loss": 0.0013, + "step": 50334 + }, + { + "epoch": 6.59, + "learning_rate": 1.5778080059916845e-06, + "loss": 0.002, + "step": 50335 + }, + { + "epoch": 6.59, + "learning_rate": 1.5775221008975428e-06, + "loss": 0.0013, + "step": 50336 + }, + { + "epoch": 6.59, + "learning_rate": 1.5772362194911727e-06, + "loss": 0.0021, + "step": 50337 + }, + { + "epoch": 6.59, + "learning_rate": 1.576950361773375e-06, + "loss": 0.002, + "step": 50338 + }, + { + "epoch": 6.59, + "learning_rate": 1.5766645277449577e-06, + "loss": 0.0013, + "step": 50339 + }, + { + "epoch": 6.59, + "learning_rate": 1.5763787174067224e-06, + "loss": 0.0015, + "step": 50340 + }, + { + "epoch": 6.59, + "learning_rate": 1.576092930759472e-06, + "loss": 0.0014, + "step": 50341 + }, + { + "epoch": 6.59, + "learning_rate": 1.5758071678040133e-06, + "loss": 0.0014, + "step": 50342 + }, + { + "epoch": 6.59, + "learning_rate": 1.575521428541149e-06, + "loss": 0.0017, + "step": 50343 + }, + { + "epoch": 6.59, + "learning_rate": 1.5752357129716812e-06, + "loss": 0.0016, + "step": 50344 + }, + { + "epoch": 6.59, + "learning_rate": 1.5749500210964119e-06, + "loss": 0.0025, + "step": 50345 + }, + { + "epoch": 6.59, + "learning_rate": 1.5746643529161477e-06, + "loss": 0.0033, + "step": 50346 + }, + { + "epoch": 6.6, + "learning_rate": 1.5743787084316941e-06, + "loss": 0.0014, + "step": 50347 + }, + { + "epoch": 6.6, + "learning_rate": 1.574093087643851e-06, + "loss": 0.0019, + "step": 50348 + }, + { + "epoch": 6.6, + "learning_rate": 1.5738074905534195e-06, + "loss": 0.003, + "step": 50349 + }, + { + "epoch": 6.6, + "learning_rate": 1.5735219171612093e-06, + "loss": 0.0009, + "step": 50350 + }, + { + "epoch": 6.6, + "learning_rate": 1.5732363674680184e-06, + "loss": 0.0033, + "step": 50351 + }, + { + "epoch": 6.6, + "learning_rate": 1.572950841474653e-06, + "loss": 0.0006, + "step": 50352 + }, + { + "epoch": 6.6, + "learning_rate": 1.5726653391819113e-06, + "loss": 0.0012, + "step": 50353 + }, + { + "epoch": 6.6, + "learning_rate": 1.5723798605906026e-06, + "loss": 0.0026, + "step": 50354 + }, + { + "epoch": 6.6, + "learning_rate": 1.5720944057015263e-06, + "loss": 0.0021, + "step": 50355 + }, + { + "epoch": 6.6, + "learning_rate": 1.571808974515483e-06, + "loss": 0.0024, + "step": 50356 + }, + { + "epoch": 6.6, + "learning_rate": 1.5715235670332784e-06, + "loss": 0.0021, + "step": 50357 + }, + { + "epoch": 6.6, + "learning_rate": 1.5712381832557178e-06, + "loss": 0.0015, + "step": 50358 + }, + { + "epoch": 6.6, + "learning_rate": 1.5709528231836003e-06, + "loss": 0.0025, + "step": 50359 + }, + { + "epoch": 6.6, + "learning_rate": 1.5706674868177274e-06, + "loss": 0.0014, + "step": 50360 + }, + { + "epoch": 6.6, + "learning_rate": 1.5703821741589055e-06, + "loss": 0.0018, + "step": 50361 + }, + { + "epoch": 6.6, + "learning_rate": 1.5700968852079346e-06, + "loss": 0.0014, + "step": 50362 + }, + { + "epoch": 6.6, + "learning_rate": 1.5698116199656166e-06, + "loss": 0.0024, + "step": 50363 + }, + { + "epoch": 6.6, + "learning_rate": 1.569526378432753e-06, + "loss": 0.0022, + "step": 50364 + }, + { + "epoch": 6.6, + "learning_rate": 1.5692411606101498e-06, + "loss": 0.0031, + "step": 50365 + }, + { + "epoch": 6.6, + "learning_rate": 1.5689559664986064e-06, + "loss": 0.0017, + "step": 50366 + }, + { + "epoch": 6.6, + "learning_rate": 1.5686707960989257e-06, + "loss": 0.0024, + "step": 50367 + }, + { + "epoch": 6.6, + "learning_rate": 1.5683856494119054e-06, + "loss": 0.0013, + "step": 50368 + }, + { + "epoch": 6.6, + "learning_rate": 1.5681005264383553e-06, + "loss": 0.0021, + "step": 50369 + }, + { + "epoch": 6.6, + "learning_rate": 1.5678154271790736e-06, + "loss": 0.0017, + "step": 50370 + }, + { + "epoch": 6.6, + "learning_rate": 1.5675303516348617e-06, + "loss": 0.0015, + "step": 50371 + }, + { + "epoch": 6.6, + "learning_rate": 1.567245299806519e-06, + "loss": 0.0015, + "step": 50372 + }, + { + "epoch": 6.6, + "learning_rate": 1.5669602716948529e-06, + "loss": 0.0024, + "step": 50373 + }, + { + "epoch": 6.6, + "learning_rate": 1.5666752673006625e-06, + "loss": 0.0019, + "step": 50374 + }, + { + "epoch": 6.6, + "learning_rate": 1.566390286624746e-06, + "loss": 0.0021, + "step": 50375 + }, + { + "epoch": 6.6, + "learning_rate": 1.5661053296679097e-06, + "loss": 0.0018, + "step": 50376 + }, + { + "epoch": 6.6, + "learning_rate": 1.5658203964309526e-06, + "loss": 0.0013, + "step": 50377 + }, + { + "epoch": 6.6, + "learning_rate": 1.5655354869146766e-06, + "loss": 0.0009, + "step": 50378 + }, + { + "epoch": 6.6, + "learning_rate": 1.565250601119881e-06, + "loss": 0.0015, + "step": 50379 + }, + { + "epoch": 6.6, + "learning_rate": 1.5649657390473694e-06, + "loss": 0.0016, + "step": 50380 + }, + { + "epoch": 6.6, + "learning_rate": 1.5646809006979445e-06, + "loss": 0.0014, + "step": 50381 + }, + { + "epoch": 6.6, + "learning_rate": 1.564396086072405e-06, + "loss": 0.0016, + "step": 50382 + }, + { + "epoch": 6.6, + "learning_rate": 1.5641112951715498e-06, + "loss": 0.0015, + "step": 50383 + }, + { + "epoch": 6.6, + "learning_rate": 1.563826527996185e-06, + "loss": 0.0019, + "step": 50384 + }, + { + "epoch": 6.6, + "learning_rate": 1.563541784547109e-06, + "loss": 0.0014, + "step": 50385 + }, + { + "epoch": 6.6, + "learning_rate": 1.5632570648251221e-06, + "loss": 0.0009, + "step": 50386 + }, + { + "epoch": 6.6, + "learning_rate": 1.5629723688310228e-06, + "loss": 0.0021, + "step": 50387 + }, + { + "epoch": 6.6, + "learning_rate": 1.5626876965656178e-06, + "loss": 0.0014, + "step": 50388 + }, + { + "epoch": 6.6, + "learning_rate": 1.5624030480297037e-06, + "loss": 0.0021, + "step": 50389 + }, + { + "epoch": 6.6, + "learning_rate": 1.5621184232240795e-06, + "loss": 0.0017, + "step": 50390 + }, + { + "epoch": 6.6, + "learning_rate": 1.5618338221495478e-06, + "loss": 0.0015, + "step": 50391 + }, + { + "epoch": 6.6, + "learning_rate": 1.5615492448069114e-06, + "loss": 0.0026, + "step": 50392 + }, + { + "epoch": 6.6, + "learning_rate": 1.5612646911969675e-06, + "loss": 0.0012, + "step": 50393 + }, + { + "epoch": 6.6, + "learning_rate": 1.5609801613205188e-06, + "loss": 0.002, + "step": 50394 + }, + { + "epoch": 6.6, + "learning_rate": 1.5606956551783614e-06, + "loss": 0.0017, + "step": 50395 + }, + { + "epoch": 6.6, + "learning_rate": 1.5604111727713002e-06, + "loss": 0.0017, + "step": 50396 + }, + { + "epoch": 6.6, + "learning_rate": 1.560126714100132e-06, + "loss": 0.002, + "step": 50397 + }, + { + "epoch": 6.6, + "learning_rate": 1.5598422791656564e-06, + "loss": 0.0016, + "step": 50398 + }, + { + "epoch": 6.6, + "learning_rate": 1.5595578679686785e-06, + "loss": 0.002, + "step": 50399 + }, + { + "epoch": 6.6, + "learning_rate": 1.559273480509993e-06, + "loss": 0.0017, + "step": 50400 + }, + { + "epoch": 6.6, + "learning_rate": 1.5589891167903992e-06, + "loss": 0.0019, + "step": 50401 + }, + { + "epoch": 6.6, + "learning_rate": 1.5587047768107022e-06, + "loss": 0.0024, + "step": 50402 + }, + { + "epoch": 6.6, + "learning_rate": 1.5584204605716957e-06, + "loss": 0.0024, + "step": 50403 + }, + { + "epoch": 6.6, + "learning_rate": 1.5581361680741847e-06, + "loss": 0.0018, + "step": 50404 + }, + { + "epoch": 6.6, + "learning_rate": 1.5578518993189651e-06, + "loss": 0.0011, + "step": 50405 + }, + { + "epoch": 6.6, + "learning_rate": 1.5575676543068363e-06, + "loss": 0.0014, + "step": 50406 + }, + { + "epoch": 6.6, + "learning_rate": 1.557283433038601e-06, + "loss": 0.0011, + "step": 50407 + }, + { + "epoch": 6.6, + "learning_rate": 1.5569992355150566e-06, + "loss": 0.0014, + "step": 50408 + }, + { + "epoch": 6.6, + "learning_rate": 1.5567150617370018e-06, + "loss": 0.0018, + "step": 50409 + }, + { + "epoch": 6.6, + "learning_rate": 1.5564309117052333e-06, + "loss": 0.002, + "step": 50410 + }, + { + "epoch": 6.6, + "learning_rate": 1.5561467854205548e-06, + "loss": 0.001, + "step": 50411 + }, + { + "epoch": 6.6, + "learning_rate": 1.5558626828837643e-06, + "loss": 0.0025, + "step": 50412 + }, + { + "epoch": 6.6, + "learning_rate": 1.5555786040956623e-06, + "loss": 0.0019, + "step": 50413 + }, + { + "epoch": 6.6, + "learning_rate": 1.5552945490570415e-06, + "loss": 0.0025, + "step": 50414 + }, + { + "epoch": 6.6, + "learning_rate": 1.5550105177687092e-06, + "loss": 0.0016, + "step": 50415 + }, + { + "epoch": 6.6, + "learning_rate": 1.554726510231459e-06, + "loss": 0.0023, + "step": 50416 + }, + { + "epoch": 6.6, + "learning_rate": 1.5544425264460905e-06, + "loss": 0.0024, + "step": 50417 + }, + { + "epoch": 6.6, + "learning_rate": 1.5541585664134007e-06, + "loss": 0.0022, + "step": 50418 + }, + { + "epoch": 6.6, + "learning_rate": 1.5538746301341935e-06, + "loss": 0.0033, + "step": 50419 + }, + { + "epoch": 6.6, + "learning_rate": 1.5535907176092625e-06, + "loss": 0.0014, + "step": 50420 + }, + { + "epoch": 6.6, + "learning_rate": 1.5533068288394071e-06, + "loss": 0.0024, + "step": 50421 + }, + { + "epoch": 6.6, + "learning_rate": 1.5530229638254247e-06, + "loss": 0.0011, + "step": 50422 + }, + { + "epoch": 6.61, + "learning_rate": 1.5527391225681188e-06, + "loss": 0.0027, + "step": 50423 + }, + { + "epoch": 6.61, + "learning_rate": 1.5524553050682845e-06, + "loss": 0.0015, + "step": 50424 + }, + { + "epoch": 6.61, + "learning_rate": 1.5521715113267165e-06, + "loss": 0.0031, + "step": 50425 + }, + { + "epoch": 6.61, + "learning_rate": 1.551887741344219e-06, + "loss": 0.003, + "step": 50426 + }, + { + "epoch": 6.61, + "learning_rate": 1.5516039951215879e-06, + "loss": 0.0015, + "step": 50427 + }, + { + "epoch": 6.61, + "learning_rate": 1.5513202726596198e-06, + "loss": 0.0018, + "step": 50428 + }, + { + "epoch": 6.61, + "learning_rate": 1.5510365739591116e-06, + "loss": 0.0018, + "step": 50429 + }, + { + "epoch": 6.61, + "learning_rate": 1.5507528990208654e-06, + "loss": 0.0015, + "step": 50430 + }, + { + "epoch": 6.61, + "learning_rate": 1.5504692478456772e-06, + "loss": 0.0018, + "step": 50431 + }, + { + "epoch": 6.61, + "learning_rate": 1.5501856204343436e-06, + "loss": 0.002, + "step": 50432 + }, + { + "epoch": 6.61, + "learning_rate": 1.5499020167876588e-06, + "loss": 0.0012, + "step": 50433 + }, + { + "epoch": 6.61, + "learning_rate": 1.5496184369064283e-06, + "loss": 0.0016, + "step": 50434 + }, + { + "epoch": 6.61, + "learning_rate": 1.5493348807914467e-06, + "loss": 0.0023, + "step": 50435 + }, + { + "epoch": 6.61, + "learning_rate": 1.549051348443511e-06, + "loss": 0.0011, + "step": 50436 + }, + { + "epoch": 6.61, + "learning_rate": 1.5487678398634154e-06, + "loss": 0.0025, + "step": 50437 + }, + { + "epoch": 6.61, + "learning_rate": 1.5484843550519623e-06, + "loss": 0.0008, + "step": 50438 + }, + { + "epoch": 6.61, + "learning_rate": 1.5482008940099479e-06, + "loss": 0.0017, + "step": 50439 + }, + { + "epoch": 6.61, + "learning_rate": 1.547917456738165e-06, + "loss": 0.0012, + "step": 50440 + }, + { + "epoch": 6.61, + "learning_rate": 1.5476340432374171e-06, + "loss": 0.0019, + "step": 50441 + }, + { + "epoch": 6.61, + "learning_rate": 1.547350653508498e-06, + "loss": 0.0021, + "step": 50442 + }, + { + "epoch": 6.61, + "learning_rate": 1.5470672875522064e-06, + "loss": 0.0023, + "step": 50443 + }, + { + "epoch": 6.61, + "learning_rate": 1.5467839453693333e-06, + "loss": 0.0017, + "step": 50444 + }, + { + "epoch": 6.61, + "learning_rate": 1.546500626960682e-06, + "loss": 0.0012, + "step": 50445 + }, + { + "epoch": 6.61, + "learning_rate": 1.546217332327049e-06, + "loss": 0.0029, + "step": 50446 + }, + { + "epoch": 6.61, + "learning_rate": 1.5459340614692297e-06, + "loss": 0.0017, + "step": 50447 + }, + { + "epoch": 6.61, + "learning_rate": 1.545650814388019e-06, + "loss": 0.0025, + "step": 50448 + }, + { + "epoch": 6.61, + "learning_rate": 1.545367591084216e-06, + "loss": 0.0013, + "step": 50449 + }, + { + "epoch": 6.61, + "learning_rate": 1.5450843915586179e-06, + "loss": 0.0024, + "step": 50450 + }, + { + "epoch": 6.61, + "learning_rate": 1.5448012158120184e-06, + "loss": 0.0011, + "step": 50451 + }, + { + "epoch": 6.61, + "learning_rate": 1.544518063845214e-06, + "loss": 0.0014, + "step": 50452 + }, + { + "epoch": 6.61, + "learning_rate": 1.5442349356590047e-06, + "loss": 0.0014, + "step": 50453 + }, + { + "epoch": 6.61, + "learning_rate": 1.5439518312541834e-06, + "loss": 0.0029, + "step": 50454 + }, + { + "epoch": 6.61, + "learning_rate": 1.543668750631545e-06, + "loss": 0.002, + "step": 50455 + }, + { + "epoch": 6.61, + "learning_rate": 1.543385693791889e-06, + "loss": 0.0024, + "step": 50456 + }, + { + "epoch": 6.61, + "learning_rate": 1.543102660736011e-06, + "loss": 0.0011, + "step": 50457 + }, + { + "epoch": 6.61, + "learning_rate": 1.5428196514647076e-06, + "loss": 0.0019, + "step": 50458 + }, + { + "epoch": 6.61, + "learning_rate": 1.5425366659787732e-06, + "loss": 0.0016, + "step": 50459 + }, + { + "epoch": 6.61, + "learning_rate": 1.542253704279002e-06, + "loss": 0.0014, + "step": 50460 + }, + { + "epoch": 6.61, + "learning_rate": 1.5419707663661942e-06, + "loss": 0.0015, + "step": 50461 + }, + { + "epoch": 6.61, + "learning_rate": 1.5416878522411426e-06, + "loss": 0.0013, + "step": 50462 + }, + { + "epoch": 6.61, + "learning_rate": 1.5414049619046423e-06, + "loss": 0.0016, + "step": 50463 + }, + { + "epoch": 6.61, + "learning_rate": 1.5411220953574913e-06, + "loss": 0.001, + "step": 50464 + }, + { + "epoch": 6.61, + "learning_rate": 1.5408392526004845e-06, + "loss": 0.0029, + "step": 50465 + }, + { + "epoch": 6.61, + "learning_rate": 1.5405564336344147e-06, + "loss": 0.0018, + "step": 50466 + }, + { + "epoch": 6.61, + "learning_rate": 1.5402736384600814e-06, + "loss": 0.0021, + "step": 50467 + }, + { + "epoch": 6.61, + "learning_rate": 1.5399908670782759e-06, + "loss": 0.0009, + "step": 50468 + }, + { + "epoch": 6.61, + "learning_rate": 1.5397081194897978e-06, + "loss": 0.0018, + "step": 50469 + }, + { + "epoch": 6.61, + "learning_rate": 1.5394253956954396e-06, + "loss": 0.0024, + "step": 50470 + }, + { + "epoch": 6.61, + "learning_rate": 1.5391426956959953e-06, + "loss": 0.0022, + "step": 50471 + }, + { + "epoch": 6.61, + "learning_rate": 1.538860019492263e-06, + "loss": 0.0024, + "step": 50472 + }, + { + "epoch": 6.61, + "learning_rate": 1.5385773670850367e-06, + "loss": 0.0017, + "step": 50473 + }, + { + "epoch": 6.61, + "learning_rate": 1.5382947384751113e-06, + "loss": 0.0011, + "step": 50474 + }, + { + "epoch": 6.61, + "learning_rate": 1.538012133663279e-06, + "loss": 0.0021, + "step": 50475 + }, + { + "epoch": 6.61, + "learning_rate": 1.537729552650339e-06, + "loss": 0.0016, + "step": 50476 + }, + { + "epoch": 6.61, + "learning_rate": 1.537446995437082e-06, + "loss": 0.0024, + "step": 50477 + }, + { + "epoch": 6.61, + "learning_rate": 1.5371644620243064e-06, + "loss": 0.0023, + "step": 50478 + }, + { + "epoch": 6.61, + "learning_rate": 1.5368819524128032e-06, + "loss": 0.0024, + "step": 50479 + }, + { + "epoch": 6.61, + "learning_rate": 1.5365994666033712e-06, + "loss": 0.0016, + "step": 50480 + }, + { + "epoch": 6.61, + "learning_rate": 1.5363170045968023e-06, + "loss": 0.0016, + "step": 50481 + }, + { + "epoch": 6.61, + "learning_rate": 1.5360345663938913e-06, + "loss": 0.0017, + "step": 50482 + }, + { + "epoch": 6.61, + "learning_rate": 1.5357521519954299e-06, + "loss": 0.0013, + "step": 50483 + }, + { + "epoch": 6.61, + "learning_rate": 1.5354697614022175e-06, + "loss": 0.0021, + "step": 50484 + }, + { + "epoch": 6.61, + "learning_rate": 1.5351873946150454e-06, + "loss": 0.002, + "step": 50485 + }, + { + "epoch": 6.61, + "learning_rate": 1.5349050516347052e-06, + "loss": 0.0022, + "step": 50486 + }, + { + "epoch": 6.61, + "learning_rate": 1.5346227324619967e-06, + "loss": 0.0024, + "step": 50487 + }, + { + "epoch": 6.61, + "learning_rate": 1.5343404370977088e-06, + "loss": 0.0027, + "step": 50488 + }, + { + "epoch": 6.61, + "learning_rate": 1.534058165542639e-06, + "loss": 0.0015, + "step": 50489 + }, + { + "epoch": 6.61, + "learning_rate": 1.5337759177975775e-06, + "loss": 0.0024, + "step": 50490 + }, + { + "epoch": 6.61, + "learning_rate": 1.5334936938633238e-06, + "loss": 0.0009, + "step": 50491 + }, + { + "epoch": 6.61, + "learning_rate": 1.5332114937406673e-06, + "loss": 0.0021, + "step": 50492 + }, + { + "epoch": 6.61, + "learning_rate": 1.5329293174304028e-06, + "loss": 0.0025, + "step": 50493 + }, + { + "epoch": 6.61, + "learning_rate": 1.5326471649333218e-06, + "loss": 0.0029, + "step": 50494 + }, + { + "epoch": 6.61, + "learning_rate": 1.5323650362502217e-06, + "loss": 0.0023, + "step": 50495 + }, + { + "epoch": 6.61, + "learning_rate": 1.532082931381894e-06, + "loss": 0.0022, + "step": 50496 + }, + { + "epoch": 6.61, + "learning_rate": 1.5318008503291293e-06, + "loss": 0.0018, + "step": 50497 + }, + { + "epoch": 6.61, + "learning_rate": 1.5315187930927268e-06, + "loss": 0.0019, + "step": 50498 + }, + { + "epoch": 6.62, + "learning_rate": 1.5312367596734745e-06, + "loss": 0.0007, + "step": 50499 + }, + { + "epoch": 6.62, + "learning_rate": 1.53095475007217e-06, + "loss": 0.0016, + "step": 50500 + }, + { + "epoch": 6.62, + "learning_rate": 1.5306727642896036e-06, + "loss": 0.001, + "step": 50501 + }, + { + "epoch": 6.62, + "learning_rate": 1.5303908023265669e-06, + "loss": 0.0017, + "step": 50502 + }, + { + "epoch": 6.62, + "learning_rate": 1.530108864183858e-06, + "loss": 0.0011, + "step": 50503 + }, + { + "epoch": 6.62, + "learning_rate": 1.5298269498622664e-06, + "loss": 0.0011, + "step": 50504 + }, + { + "epoch": 6.62, + "learning_rate": 1.5295450593625826e-06, + "loss": 0.0018, + "step": 50505 + }, + { + "epoch": 6.62, + "learning_rate": 1.529263192685606e-06, + "loss": 0.0016, + "step": 50506 + }, + { + "epoch": 6.62, + "learning_rate": 1.5289813498321248e-06, + "loss": 0.001, + "step": 50507 + }, + { + "epoch": 6.62, + "learning_rate": 1.5286995308029318e-06, + "loss": 0.0013, + "step": 50508 + }, + { + "epoch": 6.62, + "learning_rate": 1.5284177355988183e-06, + "loss": 0.0018, + "step": 50509 + }, + { + "epoch": 6.62, + "learning_rate": 1.5281359642205795e-06, + "loss": 0.0013, + "step": 50510 + }, + { + "epoch": 6.62, + "learning_rate": 1.527854216669009e-06, + "loss": 0.0019, + "step": 50511 + }, + { + "epoch": 6.62, + "learning_rate": 1.5275724929448976e-06, + "loss": 0.0018, + "step": 50512 + }, + { + "epoch": 6.62, + "learning_rate": 1.5272907930490345e-06, + "loss": 0.0029, + "step": 50513 + }, + { + "epoch": 6.62, + "learning_rate": 1.5270091169822177e-06, + "loss": 0.0016, + "step": 50514 + }, + { + "epoch": 6.62, + "learning_rate": 1.5267274647452357e-06, + "loss": 0.0019, + "step": 50515 + }, + { + "epoch": 6.62, + "learning_rate": 1.5264458363388824e-06, + "loss": 0.0026, + "step": 50516 + }, + { + "epoch": 6.62, + "learning_rate": 1.5261642317639458e-06, + "loss": 0.0012, + "step": 50517 + }, + { + "epoch": 6.62, + "learning_rate": 1.5258826510212242e-06, + "loss": 0.0025, + "step": 50518 + }, + { + "epoch": 6.62, + "learning_rate": 1.5256010941115062e-06, + "loss": 0.0019, + "step": 50519 + }, + { + "epoch": 6.62, + "learning_rate": 1.5253195610355808e-06, + "loss": 0.0016, + "step": 50520 + }, + { + "epoch": 6.62, + "learning_rate": 1.5250380517942442e-06, + "loss": 0.0039, + "step": 50521 + }, + { + "epoch": 6.62, + "learning_rate": 1.524756566388288e-06, + "loss": 0.0011, + "step": 50522 + }, + { + "epoch": 6.62, + "learning_rate": 1.5244751048185024e-06, + "loss": 0.0019, + "step": 50523 + }, + { + "epoch": 6.62, + "learning_rate": 1.5241936670856805e-06, + "loss": 0.0026, + "step": 50524 + }, + { + "epoch": 6.62, + "learning_rate": 1.5239122531906091e-06, + "loss": 0.001, + "step": 50525 + }, + { + "epoch": 6.62, + "learning_rate": 1.5236308631340857e-06, + "loss": 0.0013, + "step": 50526 + }, + { + "epoch": 6.62, + "learning_rate": 1.5233494969168995e-06, + "loss": 0.0008, + "step": 50527 + }, + { + "epoch": 6.62, + "learning_rate": 1.5230681545398384e-06, + "loss": 0.0009, + "step": 50528 + }, + { + "epoch": 6.62, + "learning_rate": 1.522786836003699e-06, + "loss": 0.002, + "step": 50529 + }, + { + "epoch": 6.62, + "learning_rate": 1.5225055413092715e-06, + "loss": 0.0013, + "step": 50530 + }, + { + "epoch": 6.62, + "learning_rate": 1.5222242704573442e-06, + "loss": 0.0023, + "step": 50531 + }, + { + "epoch": 6.62, + "learning_rate": 1.5219430234487075e-06, + "loss": 0.0027, + "step": 50532 + }, + { + "epoch": 6.62, + "learning_rate": 1.5216618002841555e-06, + "loss": 0.0022, + "step": 50533 + }, + { + "epoch": 6.62, + "learning_rate": 1.5213806009644805e-06, + "loss": 0.0016, + "step": 50534 + }, + { + "epoch": 6.62, + "learning_rate": 1.5210994254904698e-06, + "loss": 0.0017, + "step": 50535 + }, + { + "epoch": 6.62, + "learning_rate": 1.520818273862914e-06, + "loss": 0.0016, + "step": 50536 + }, + { + "epoch": 6.62, + "learning_rate": 1.520537146082608e-06, + "loss": 0.0016, + "step": 50537 + }, + { + "epoch": 6.62, + "learning_rate": 1.5202560421503387e-06, + "loss": 0.0012, + "step": 50538 + }, + { + "epoch": 6.62, + "learning_rate": 1.5199749620668979e-06, + "loss": 0.0013, + "step": 50539 + }, + { + "epoch": 6.62, + "learning_rate": 1.519693905833074e-06, + "loss": 0.002, + "step": 50540 + }, + { + "epoch": 6.62, + "learning_rate": 1.5194128734496616e-06, + "loss": 0.0015, + "step": 50541 + }, + { + "epoch": 6.62, + "learning_rate": 1.519131864917447e-06, + "loss": 0.0015, + "step": 50542 + }, + { + "epoch": 6.62, + "learning_rate": 1.5188508802372237e-06, + "loss": 0.0024, + "step": 50543 + }, + { + "epoch": 6.62, + "learning_rate": 1.518569919409779e-06, + "loss": 0.0014, + "step": 50544 + }, + { + "epoch": 6.62, + "learning_rate": 1.518288982435907e-06, + "loss": 0.0015, + "step": 50545 + }, + { + "epoch": 6.62, + "learning_rate": 1.5180080693163957e-06, + "loss": 0.0017, + "step": 50546 + }, + { + "epoch": 6.62, + "learning_rate": 1.5177271800520322e-06, + "loss": 0.001, + "step": 50547 + }, + { + "epoch": 6.62, + "learning_rate": 1.5174463146436124e-06, + "loss": 0.0019, + "step": 50548 + }, + { + "epoch": 6.62, + "learning_rate": 1.5171654730919228e-06, + "loss": 0.0008, + "step": 50549 + }, + { + "epoch": 6.62, + "learning_rate": 1.5168846553977535e-06, + "loss": 0.0012, + "step": 50550 + }, + { + "epoch": 6.62, + "learning_rate": 1.5166038615618917e-06, + "loss": 0.002, + "step": 50551 + }, + { + "epoch": 6.62, + "learning_rate": 1.5163230915851323e-06, + "loss": 0.0019, + "step": 50552 + }, + { + "epoch": 6.62, + "learning_rate": 1.5160423454682605e-06, + "loss": 0.003, + "step": 50553 + }, + { + "epoch": 6.62, + "learning_rate": 1.5157616232120697e-06, + "loss": 0.0016, + "step": 50554 + }, + { + "epoch": 6.62, + "learning_rate": 1.5154809248173453e-06, + "loss": 0.0031, + "step": 50555 + }, + { + "epoch": 6.62, + "learning_rate": 1.5152002502848807e-06, + "loss": 0.0024, + "step": 50556 + }, + { + "epoch": 6.62, + "learning_rate": 1.5149195996154642e-06, + "loss": 0.0023, + "step": 50557 + }, + { + "epoch": 6.62, + "learning_rate": 1.5146389728098842e-06, + "loss": 0.0036, + "step": 50558 + }, + { + "epoch": 6.62, + "learning_rate": 1.5143583698689278e-06, + "loss": 0.0012, + "step": 50559 + }, + { + "epoch": 6.62, + "learning_rate": 1.51407779079339e-06, + "loss": 0.0032, + "step": 50560 + }, + { + "epoch": 6.62, + "learning_rate": 1.5137972355840557e-06, + "loss": 0.0019, + "step": 50561 + }, + { + "epoch": 6.62, + "learning_rate": 1.513516704241712e-06, + "loss": 0.0015, + "step": 50562 + }, + { + "epoch": 6.62, + "learning_rate": 1.513236196767154e-06, + "loss": 0.0011, + "step": 50563 + }, + { + "epoch": 6.62, + "learning_rate": 1.5129557131611638e-06, + "loss": 0.002, + "step": 50564 + }, + { + "epoch": 6.62, + "learning_rate": 1.512675253424537e-06, + "loss": 0.002, + "step": 50565 + }, + { + "epoch": 6.62, + "learning_rate": 1.5123948175580582e-06, + "loss": 0.0014, + "step": 50566 + }, + { + "epoch": 6.62, + "learning_rate": 1.5121144055625147e-06, + "loss": 0.0011, + "step": 50567 + }, + { + "epoch": 6.62, + "learning_rate": 1.5118340174387003e-06, + "loss": 0.0022, + "step": 50568 + }, + { + "epoch": 6.62, + "learning_rate": 1.511553653187401e-06, + "loss": 0.0017, + "step": 50569 + }, + { + "epoch": 6.62, + "learning_rate": 1.5112733128094027e-06, + "loss": 0.0021, + "step": 50570 + }, + { + "epoch": 6.62, + "learning_rate": 1.510992996305497e-06, + "loss": 0.0017, + "step": 50571 + }, + { + "epoch": 6.62, + "learning_rate": 1.5107127036764723e-06, + "loss": 0.0018, + "step": 50572 + }, + { + "epoch": 6.62, + "learning_rate": 1.510432434923117e-06, + "loss": 0.0015, + "step": 50573 + }, + { + "epoch": 6.62, + "learning_rate": 1.5101521900462147e-06, + "loss": 0.0019, + "step": 50574 + }, + { + "epoch": 6.62, + "learning_rate": 1.509871969046558e-06, + "loss": 0.0024, + "step": 50575 + }, + { + "epoch": 6.63, + "learning_rate": 1.5095917719249365e-06, + "loss": 0.0022, + "step": 50576 + }, + { + "epoch": 6.63, + "learning_rate": 1.509311598682136e-06, + "loss": 0.0014, + "step": 50577 + }, + { + "epoch": 6.63, + "learning_rate": 1.5090314493189417e-06, + "loss": 0.0015, + "step": 50578 + }, + { + "epoch": 6.63, + "learning_rate": 1.5087513238361473e-06, + "loss": 0.0016, + "step": 50579 + }, + { + "epoch": 6.63, + "learning_rate": 1.5084712222345365e-06, + "loss": 0.001, + "step": 50580 + }, + { + "epoch": 6.63, + "learning_rate": 1.5081911445149e-06, + "loss": 0.0021, + "step": 50581 + }, + { + "epoch": 6.63, + "learning_rate": 1.5079110906780204e-06, + "loss": 0.0013, + "step": 50582 + }, + { + "epoch": 6.63, + "learning_rate": 1.5076310607246902e-06, + "loss": 0.0028, + "step": 50583 + }, + { + "epoch": 6.63, + "learning_rate": 1.5073510546556957e-06, + "loss": 0.0017, + "step": 50584 + }, + { + "epoch": 6.63, + "learning_rate": 1.507071072471823e-06, + "loss": 0.0023, + "step": 50585 + }, + { + "epoch": 6.63, + "learning_rate": 1.5067911141738589e-06, + "loss": 0.0023, + "step": 50586 + }, + { + "epoch": 6.63, + "learning_rate": 1.5065111797625964e-06, + "loss": 0.001, + "step": 50587 + }, + { + "epoch": 6.63, + "learning_rate": 1.506231269238818e-06, + "loss": 0.0015, + "step": 50588 + }, + { + "epoch": 6.63, + "learning_rate": 1.5059513826033122e-06, + "loss": 0.001, + "step": 50589 + }, + { + "epoch": 6.63, + "learning_rate": 1.5056715198568627e-06, + "loss": 0.0025, + "step": 50590 + }, + { + "epoch": 6.63, + "learning_rate": 1.5053916810002632e-06, + "loss": 0.0028, + "step": 50591 + }, + { + "epoch": 6.63, + "learning_rate": 1.5051118660342966e-06, + "loss": 0.0024, + "step": 50592 + }, + { + "epoch": 6.63, + "learning_rate": 1.504832074959748e-06, + "loss": 0.0024, + "step": 50593 + }, + { + "epoch": 6.63, + "learning_rate": 1.5045523077774094e-06, + "loss": 0.0009, + "step": 50594 + }, + { + "epoch": 6.63, + "learning_rate": 1.5042725644880652e-06, + "loss": 0.0011, + "step": 50595 + }, + { + "epoch": 6.63, + "learning_rate": 1.5039928450925012e-06, + "loss": 0.0016, + "step": 50596 + }, + { + "epoch": 6.63, + "learning_rate": 1.5037131495915014e-06, + "loss": 0.0012, + "step": 50597 + }, + { + "epoch": 6.63, + "learning_rate": 1.5034334779858594e-06, + "loss": 0.0011, + "step": 50598 + }, + { + "epoch": 6.63, + "learning_rate": 1.5031538302763594e-06, + "loss": 0.0024, + "step": 50599 + }, + { + "epoch": 6.63, + "learning_rate": 1.5028742064637858e-06, + "loss": 0.0013, + "step": 50600 + }, + { + "epoch": 6.63, + "learning_rate": 1.5025946065489239e-06, + "loss": 0.0013, + "step": 50601 + }, + { + "epoch": 6.63, + "learning_rate": 1.502315030532565e-06, + "loss": 0.002, + "step": 50602 + }, + { + "epoch": 6.63, + "learning_rate": 1.5020354784154922e-06, + "loss": 0.0026, + "step": 50603 + }, + { + "epoch": 6.63, + "learning_rate": 1.5017559501984925e-06, + "loss": 0.0025, + "step": 50604 + }, + { + "epoch": 6.63, + "learning_rate": 1.5014764458823482e-06, + "loss": 0.0022, + "step": 50605 + }, + { + "epoch": 6.63, + "learning_rate": 1.5011969654678515e-06, + "loss": 0.002, + "step": 50606 + }, + { + "epoch": 6.63, + "learning_rate": 1.500917508955787e-06, + "loss": 0.0018, + "step": 50607 + }, + { + "epoch": 6.63, + "learning_rate": 1.5006380763469353e-06, + "loss": 0.001, + "step": 50608 + }, + { + "epoch": 6.63, + "learning_rate": 1.5003586676420878e-06, + "loss": 0.0011, + "step": 50609 + }, + { + "epoch": 6.63, + "learning_rate": 1.5000792828420306e-06, + "loss": 0.001, + "step": 50610 + }, + { + "epoch": 6.63, + "learning_rate": 1.4997999219475479e-06, + "loss": 0.002, + "step": 50611 + }, + { + "epoch": 6.63, + "learning_rate": 1.499520584959423e-06, + "loss": 0.0019, + "step": 50612 + }, + { + "epoch": 6.63, + "learning_rate": 1.4992412718784455e-06, + "loss": 0.0019, + "step": 50613 + }, + { + "epoch": 6.63, + "learning_rate": 1.4989619827054003e-06, + "loss": 0.0009, + "step": 50614 + }, + { + "epoch": 6.63, + "learning_rate": 1.4986827174410713e-06, + "loss": 0.0015, + "step": 50615 + }, + { + "epoch": 6.63, + "learning_rate": 1.498403476086242e-06, + "loss": 0.0012, + "step": 50616 + }, + { + "epoch": 6.63, + "learning_rate": 1.4981242586417022e-06, + "loss": 0.0037, + "step": 50617 + }, + { + "epoch": 6.63, + "learning_rate": 1.4978450651082355e-06, + "loss": 0.0014, + "step": 50618 + }, + { + "epoch": 6.63, + "learning_rate": 1.4975658954866246e-06, + "loss": 0.0019, + "step": 50619 + }, + { + "epoch": 6.63, + "learning_rate": 1.4972867497776567e-06, + "loss": 0.002, + "step": 50620 + }, + { + "epoch": 6.63, + "learning_rate": 1.49700762798212e-06, + "loss": 0.0018, + "step": 50621 + }, + { + "epoch": 6.63, + "learning_rate": 1.496728530100795e-06, + "loss": 0.0016, + "step": 50622 + }, + { + "epoch": 6.63, + "learning_rate": 1.49644945613447e-06, + "loss": 0.002, + "step": 50623 + }, + { + "epoch": 6.63, + "learning_rate": 1.4961704060839244e-06, + "loss": 0.0017, + "step": 50624 + }, + { + "epoch": 6.63, + "learning_rate": 1.4958913799499497e-06, + "loss": 0.0047, + "step": 50625 + }, + { + "epoch": 6.63, + "learning_rate": 1.4956123777333275e-06, + "loss": 0.001, + "step": 50626 + }, + { + "epoch": 6.63, + "learning_rate": 1.4953333994348396e-06, + "loss": 0.0018, + "step": 50627 + }, + { + "epoch": 6.63, + "learning_rate": 1.4950544450552762e-06, + "loss": 0.0012, + "step": 50628 + }, + { + "epoch": 6.63, + "learning_rate": 1.49477551459542e-06, + "loss": 0.0017, + "step": 50629 + }, + { + "epoch": 6.63, + "learning_rate": 1.4944966080560518e-06, + "loss": 0.0017, + "step": 50630 + }, + { + "epoch": 6.63, + "learning_rate": 1.494217725437962e-06, + "loss": 0.0022, + "step": 50631 + }, + { + "epoch": 6.63, + "learning_rate": 1.4939388667419286e-06, + "loss": 0.002, + "step": 50632 + }, + { + "epoch": 6.63, + "learning_rate": 1.4936600319687422e-06, + "loss": 0.0015, + "step": 50633 + }, + { + "epoch": 6.63, + "learning_rate": 1.4933812211191834e-06, + "loss": 0.0021, + "step": 50634 + }, + { + "epoch": 6.63, + "learning_rate": 1.4931024341940348e-06, + "loss": 0.0026, + "step": 50635 + }, + { + "epoch": 6.63, + "learning_rate": 1.4928236711940846e-06, + "loss": 0.0013, + "step": 50636 + }, + { + "epoch": 6.63, + "learning_rate": 1.4925449321201157e-06, + "loss": 0.0019, + "step": 50637 + }, + { + "epoch": 6.63, + "learning_rate": 1.4922662169729096e-06, + "loss": 0.0012, + "step": 50638 + }, + { + "epoch": 6.63, + "learning_rate": 1.4919875257532512e-06, + "loss": 0.0024, + "step": 50639 + }, + { + "epoch": 6.63, + "learning_rate": 1.4917088584619254e-06, + "loss": 0.0014, + "step": 50640 + }, + { + "epoch": 6.63, + "learning_rate": 1.491430215099714e-06, + "loss": 0.0021, + "step": 50641 + }, + { + "epoch": 6.63, + "learning_rate": 1.491151595667405e-06, + "loss": 0.0016, + "step": 50642 + }, + { + "epoch": 6.63, + "learning_rate": 1.4908730001657757e-06, + "loss": 0.0011, + "step": 50643 + }, + { + "epoch": 6.63, + "learning_rate": 1.4905944285956152e-06, + "loss": 0.0028, + "step": 50644 + }, + { + "epoch": 6.63, + "learning_rate": 1.4903158809577056e-06, + "loss": 0.0015, + "step": 50645 + }, + { + "epoch": 6.63, + "learning_rate": 1.4900373572528281e-06, + "loss": 0.0011, + "step": 50646 + }, + { + "epoch": 6.63, + "learning_rate": 1.4897588574817657e-06, + "loss": 0.0017, + "step": 50647 + }, + { + "epoch": 6.63, + "learning_rate": 1.4894803816453063e-06, + "loss": 0.0034, + "step": 50648 + }, + { + "epoch": 6.63, + "learning_rate": 1.4892019297442296e-06, + "loss": 0.0019, + "step": 50649 + }, + { + "epoch": 6.63, + "learning_rate": 1.488923501779317e-06, + "loss": 0.0009, + "step": 50650 + }, + { + "epoch": 6.63, + "learning_rate": 1.4886450977513555e-06, + "loss": 0.0022, + "step": 50651 + }, + { + "epoch": 6.64, + "learning_rate": 1.488366717661125e-06, + "loss": 0.0015, + "step": 50652 + }, + { + "epoch": 6.64, + "learning_rate": 1.4880883615094111e-06, + "loss": 0.0028, + "step": 50653 + }, + { + "epoch": 6.64, + "learning_rate": 1.4878100292969954e-06, + "loss": 0.003, + "step": 50654 + }, + { + "epoch": 6.64, + "learning_rate": 1.4875317210246598e-06, + "loss": 0.0031, + "step": 50655 + }, + { + "epoch": 6.64, + "learning_rate": 1.487253436693189e-06, + "loss": 0.0013, + "step": 50656 + }, + { + "epoch": 6.64, + "learning_rate": 1.4869751763033647e-06, + "loss": 0.0014, + "step": 50657 + }, + { + "epoch": 6.64, + "learning_rate": 1.4866969398559682e-06, + "loss": 0.0021, + "step": 50658 + }, + { + "epoch": 6.64, + "learning_rate": 1.486418727351784e-06, + "loss": 0.0026, + "step": 50659 + }, + { + "epoch": 6.64, + "learning_rate": 1.486140538791595e-06, + "loss": 0.0019, + "step": 50660 + }, + { + "epoch": 6.64, + "learning_rate": 1.4858623741761813e-06, + "loss": 0.0016, + "step": 50661 + }, + { + "epoch": 6.64, + "learning_rate": 1.4855842335063242e-06, + "loss": 0.0029, + "step": 50662 + }, + { + "epoch": 6.64, + "learning_rate": 1.4853061167828087e-06, + "loss": 0.0018, + "step": 50663 + }, + { + "epoch": 6.64, + "learning_rate": 1.4850280240064175e-06, + "loss": 0.0022, + "step": 50664 + }, + { + "epoch": 6.64, + "learning_rate": 1.4847499551779332e-06, + "loss": 0.0017, + "step": 50665 + }, + { + "epoch": 6.64, + "learning_rate": 1.484471910298132e-06, + "loss": 0.0017, + "step": 50666 + }, + { + "epoch": 6.64, + "learning_rate": 1.4841938893678032e-06, + "loss": 0.002, + "step": 50667 + }, + { + "epoch": 6.64, + "learning_rate": 1.483915892387726e-06, + "loss": 0.0014, + "step": 50668 + }, + { + "epoch": 6.64, + "learning_rate": 1.4836379193586815e-06, + "loss": 0.0017, + "step": 50669 + }, + { + "epoch": 6.64, + "learning_rate": 1.4833599702814495e-06, + "loss": 0.0027, + "step": 50670 + }, + { + "epoch": 6.64, + "learning_rate": 1.483082045156816e-06, + "loss": 0.0008, + "step": 50671 + }, + { + "epoch": 6.64, + "learning_rate": 1.482804143985561e-06, + "loss": 0.0031, + "step": 50672 + }, + { + "epoch": 6.64, + "learning_rate": 1.4825262667684636e-06, + "loss": 0.0017, + "step": 50673 + }, + { + "epoch": 6.64, + "learning_rate": 1.4822484135063075e-06, + "loss": 0.0025, + "step": 50674 + }, + { + "epoch": 6.64, + "learning_rate": 1.4819705841998767e-06, + "loss": 0.0033, + "step": 50675 + }, + { + "epoch": 6.64, + "learning_rate": 1.481692778849949e-06, + "loss": 0.0015, + "step": 50676 + }, + { + "epoch": 6.64, + "learning_rate": 1.4814149974573055e-06, + "loss": 0.0013, + "step": 50677 + }, + { + "epoch": 6.64, + "learning_rate": 1.4811372400227309e-06, + "loss": 0.0016, + "step": 50678 + }, + { + "epoch": 6.64, + "learning_rate": 1.4808595065470032e-06, + "loss": 0.0018, + "step": 50679 + }, + { + "epoch": 6.64, + "learning_rate": 1.4805817970309055e-06, + "loss": 0.0021, + "step": 50680 + }, + { + "epoch": 6.64, + "learning_rate": 1.4803041114752147e-06, + "loss": 0.0018, + "step": 50681 + }, + { + "epoch": 6.64, + "learning_rate": 1.4800264498807181e-06, + "loss": 0.0028, + "step": 50682 + }, + { + "epoch": 6.64, + "learning_rate": 1.479748812248193e-06, + "loss": 0.0025, + "step": 50683 + }, + { + "epoch": 6.64, + "learning_rate": 1.4794711985784182e-06, + "loss": 0.0016, + "step": 50684 + }, + { + "epoch": 6.64, + "learning_rate": 1.479193608872177e-06, + "loss": 0.0025, + "step": 50685 + }, + { + "epoch": 6.64, + "learning_rate": 1.4789160431302519e-06, + "loss": 0.0015, + "step": 50686 + }, + { + "epoch": 6.64, + "learning_rate": 1.478638501353422e-06, + "loss": 0.0013, + "step": 50687 + }, + { + "epoch": 6.64, + "learning_rate": 1.4783609835424684e-06, + "loss": 0.0018, + "step": 50688 + }, + { + "epoch": 6.64, + "learning_rate": 1.4780834896981667e-06, + "loss": 0.0018, + "step": 50689 + }, + { + "epoch": 6.64, + "learning_rate": 1.4778060198213051e-06, + "loss": 0.0015, + "step": 50690 + }, + { + "epoch": 6.64, + "learning_rate": 1.4775285739126598e-06, + "loss": 0.0022, + "step": 50691 + }, + { + "epoch": 6.64, + "learning_rate": 1.4772511519730094e-06, + "loss": 0.0016, + "step": 50692 + }, + { + "epoch": 6.64, + "learning_rate": 1.4769737540031381e-06, + "loss": 0.0025, + "step": 50693 + }, + { + "epoch": 6.64, + "learning_rate": 1.4766963800038237e-06, + "loss": 0.0016, + "step": 50694 + }, + { + "epoch": 6.64, + "learning_rate": 1.4764190299758451e-06, + "loss": 0.0023, + "step": 50695 + }, + { + "epoch": 6.64, + "learning_rate": 1.4761417039199865e-06, + "loss": 0.0026, + "step": 50696 + }, + { + "epoch": 6.64, + "learning_rate": 1.4758644018370238e-06, + "loss": 0.0027, + "step": 50697 + }, + { + "epoch": 6.64, + "learning_rate": 1.4755871237277397e-06, + "loss": 0.0016, + "step": 50698 + }, + { + "epoch": 6.64, + "learning_rate": 1.4753098695929135e-06, + "loss": 0.0031, + "step": 50699 + }, + { + "epoch": 6.64, + "learning_rate": 1.4750326394333226e-06, + "loss": 0.0027, + "step": 50700 + }, + { + "epoch": 6.64, + "learning_rate": 1.4747554332497494e-06, + "loss": 0.0018, + "step": 50701 + }, + { + "epoch": 6.64, + "learning_rate": 1.4744782510429734e-06, + "loss": 0.0015, + "step": 50702 + }, + { + "epoch": 6.64, + "learning_rate": 1.4742010928137729e-06, + "loss": 0.0015, + "step": 50703 + }, + { + "epoch": 6.64, + "learning_rate": 1.473923958562926e-06, + "loss": 0.0021, + "step": 50704 + }, + { + "epoch": 6.64, + "learning_rate": 1.473646848291216e-06, + "loss": 0.0026, + "step": 50705 + }, + { + "epoch": 6.64, + "learning_rate": 1.473369761999417e-06, + "loss": 0.0016, + "step": 50706 + }, + { + "epoch": 6.64, + "learning_rate": 1.4730926996883155e-06, + "loss": 0.0025, + "step": 50707 + }, + { + "epoch": 6.64, + "learning_rate": 1.472815661358683e-06, + "loss": 0.0011, + "step": 50708 + }, + { + "epoch": 6.64, + "learning_rate": 1.4725386470113045e-06, + "loss": 0.0029, + "step": 50709 + }, + { + "epoch": 6.64, + "learning_rate": 1.4722616566469584e-06, + "loss": 0.0018, + "step": 50710 + }, + { + "epoch": 6.64, + "learning_rate": 1.4719846902664215e-06, + "loss": 0.0018, + "step": 50711 + }, + { + "epoch": 6.64, + "learning_rate": 1.471707747870471e-06, + "loss": 0.0011, + "step": 50712 + }, + { + "epoch": 6.64, + "learning_rate": 1.4714308294598901e-06, + "loss": 0.0023, + "step": 50713 + }, + { + "epoch": 6.64, + "learning_rate": 1.4711539350354565e-06, + "loss": 0.0014, + "step": 50714 + }, + { + "epoch": 6.64, + "learning_rate": 1.4708770645979464e-06, + "loss": 0.0019, + "step": 50715 + }, + { + "epoch": 6.64, + "learning_rate": 1.4706002181481415e-06, + "loss": 0.0029, + "step": 50716 + }, + { + "epoch": 6.64, + "learning_rate": 1.470323395686818e-06, + "loss": 0.0013, + "step": 50717 + }, + { + "epoch": 6.64, + "learning_rate": 1.4700465972147583e-06, + "loss": 0.0023, + "step": 50718 + }, + { + "epoch": 6.64, + "learning_rate": 1.4697698227327372e-06, + "loss": 0.0026, + "step": 50719 + }, + { + "epoch": 6.64, + "learning_rate": 1.4694930722415324e-06, + "loss": 0.0013, + "step": 50720 + }, + { + "epoch": 6.64, + "learning_rate": 1.4692163457419263e-06, + "loss": 0.0015, + "step": 50721 + }, + { + "epoch": 6.64, + "learning_rate": 1.4689396432346948e-06, + "loss": 0.0008, + "step": 50722 + }, + { + "epoch": 6.64, + "learning_rate": 1.468662964720614e-06, + "loss": 0.0016, + "step": 50723 + }, + { + "epoch": 6.64, + "learning_rate": 1.468386310200467e-06, + "loss": 0.002, + "step": 50724 + }, + { + "epoch": 6.64, + "learning_rate": 1.4681096796750283e-06, + "loss": 0.0021, + "step": 50725 + }, + { + "epoch": 6.64, + "learning_rate": 1.4678330731450774e-06, + "loss": 0.0026, + "step": 50726 + }, + { + "epoch": 6.64, + "learning_rate": 1.4675564906113893e-06, + "loss": 0.0022, + "step": 50727 + }, + { + "epoch": 6.65, + "learning_rate": 1.4672799320747445e-06, + "loss": 0.0012, + "step": 50728 + }, + { + "epoch": 6.65, + "learning_rate": 1.4670033975359222e-06, + "loss": 0.002, + "step": 50729 + }, + { + "epoch": 6.65, + "learning_rate": 1.4667268869956984e-06, + "loss": 0.0019, + "step": 50730 + }, + { + "epoch": 6.65, + "learning_rate": 1.4664504004548496e-06, + "loss": 0.0022, + "step": 50731 + }, + { + "epoch": 6.65, + "learning_rate": 1.466173937914156e-06, + "loss": 0.0012, + "step": 50732 + }, + { + "epoch": 6.65, + "learning_rate": 1.4658974993743936e-06, + "loss": 0.0022, + "step": 50733 + }, + { + "epoch": 6.65, + "learning_rate": 1.4656210848363407e-06, + "loss": 0.0027, + "step": 50734 + }, + { + "epoch": 6.65, + "learning_rate": 1.465344694300771e-06, + "loss": 0.002, + "step": 50735 + }, + { + "epoch": 6.65, + "learning_rate": 1.4650683277684674e-06, + "loss": 0.0021, + "step": 50736 + }, + { + "epoch": 6.65, + "learning_rate": 1.4647919852402048e-06, + "loss": 0.0025, + "step": 50737 + }, + { + "epoch": 6.65, + "learning_rate": 1.464515666716757e-06, + "loss": 0.0028, + "step": 50738 + }, + { + "epoch": 6.65, + "learning_rate": 1.4642393721989056e-06, + "loss": 0.0023, + "step": 50739 + }, + { + "epoch": 6.65, + "learning_rate": 1.4639631016874278e-06, + "loss": 0.0016, + "step": 50740 + }, + { + "epoch": 6.65, + "learning_rate": 1.4636868551830996e-06, + "loss": 0.0012, + "step": 50741 + }, + { + "epoch": 6.65, + "learning_rate": 1.4634106326866958e-06, + "loss": 0.0019, + "step": 50742 + }, + { + "epoch": 6.65, + "learning_rate": 1.4631344341989962e-06, + "loss": 0.0018, + "step": 50743 + }, + { + "epoch": 6.65, + "learning_rate": 1.4628582597207775e-06, + "loss": 0.0011, + "step": 50744 + }, + { + "epoch": 6.65, + "learning_rate": 1.4625821092528148e-06, + "loss": 0.002, + "step": 50745 + }, + { + "epoch": 6.65, + "learning_rate": 1.462305982795883e-06, + "loss": 0.0011, + "step": 50746 + }, + { + "epoch": 6.65, + "learning_rate": 1.4620298803507626e-06, + "loss": 0.0014, + "step": 50747 + }, + { + "epoch": 6.65, + "learning_rate": 1.4617538019182298e-06, + "loss": 0.0019, + "step": 50748 + }, + { + "epoch": 6.65, + "learning_rate": 1.4614777474990582e-06, + "loss": 0.0012, + "step": 50749 + }, + { + "epoch": 6.65, + "learning_rate": 1.4612017170940228e-06, + "loss": 0.0023, + "step": 50750 + }, + { + "epoch": 6.65, + "learning_rate": 1.4609257107039066e-06, + "loss": 0.0023, + "step": 50751 + }, + { + "epoch": 6.65, + "learning_rate": 1.4606497283294819e-06, + "loss": 0.003, + "step": 50752 + }, + { + "epoch": 6.65, + "learning_rate": 1.4603737699715248e-06, + "loss": 0.0013, + "step": 50753 + }, + { + "epoch": 6.65, + "learning_rate": 1.4600978356308105e-06, + "loss": 0.0024, + "step": 50754 + }, + { + "epoch": 6.65, + "learning_rate": 1.459821925308118e-06, + "loss": 0.0018, + "step": 50755 + }, + { + "epoch": 6.65, + "learning_rate": 1.4595460390042216e-06, + "loss": 0.0017, + "step": 50756 + }, + { + "epoch": 6.65, + "learning_rate": 1.4592701767198948e-06, + "loss": 0.001, + "step": 50757 + }, + { + "epoch": 6.65, + "learning_rate": 1.458994338455918e-06, + "loss": 0.0016, + "step": 50758 + }, + { + "epoch": 6.65, + "learning_rate": 1.4587185242130641e-06, + "loss": 0.0017, + "step": 50759 + }, + { + "epoch": 6.65, + "learning_rate": 1.4584427339921104e-06, + "loss": 0.0012, + "step": 50760 + }, + { + "epoch": 6.65, + "learning_rate": 1.4581669677938293e-06, + "loss": 0.0021, + "step": 50761 + }, + { + "epoch": 6.65, + "learning_rate": 1.4578912256189993e-06, + "loss": 0.0022, + "step": 50762 + }, + { + "epoch": 6.65, + "learning_rate": 1.4576155074683961e-06, + "loss": 0.0018, + "step": 50763 + }, + { + "epoch": 6.65, + "learning_rate": 1.4573398133427963e-06, + "loss": 0.0013, + "step": 50764 + }, + { + "epoch": 6.65, + "learning_rate": 1.45706414324297e-06, + "loss": 0.0019, + "step": 50765 + }, + { + "epoch": 6.65, + "learning_rate": 1.4567884971696978e-06, + "loss": 0.0016, + "step": 50766 + }, + { + "epoch": 6.65, + "learning_rate": 1.4565128751237534e-06, + "loss": 0.0026, + "step": 50767 + }, + { + "epoch": 6.65, + "learning_rate": 1.4562372771059119e-06, + "loss": 0.0016, + "step": 50768 + }, + { + "epoch": 6.65, + "learning_rate": 1.4559617031169448e-06, + "loss": 0.0013, + "step": 50769 + }, + { + "epoch": 6.65, + "learning_rate": 1.4556861531576338e-06, + "loss": 0.0015, + "step": 50770 + }, + { + "epoch": 6.65, + "learning_rate": 1.4554106272287493e-06, + "loss": 0.0018, + "step": 50771 + }, + { + "epoch": 6.65, + "learning_rate": 1.4551351253310663e-06, + "loss": 0.0013, + "step": 50772 + }, + { + "epoch": 6.65, + "learning_rate": 1.4548596474653598e-06, + "loss": 0.0024, + "step": 50773 + }, + { + "epoch": 6.65, + "learning_rate": 1.454584193632409e-06, + "loss": 0.0029, + "step": 50774 + }, + { + "epoch": 6.65, + "learning_rate": 1.4543087638329833e-06, + "loss": 0.0017, + "step": 50775 + }, + { + "epoch": 6.65, + "learning_rate": 1.4540333580678612e-06, + "loss": 0.002, + "step": 50776 + }, + { + "epoch": 6.65, + "learning_rate": 1.4537579763378107e-06, + "loss": 0.0022, + "step": 50777 + }, + { + "epoch": 6.65, + "learning_rate": 1.4534826186436146e-06, + "loss": 0.0024, + "step": 50778 + }, + { + "epoch": 6.65, + "learning_rate": 1.4532072849860434e-06, + "loss": 0.0026, + "step": 50779 + }, + { + "epoch": 6.65, + "learning_rate": 1.4529319753658688e-06, + "loss": 0.0012, + "step": 50780 + }, + { + "epoch": 6.65, + "learning_rate": 1.4526566897838701e-06, + "loss": 0.0016, + "step": 50781 + }, + { + "epoch": 6.65, + "learning_rate": 1.45238142824082e-06, + "loss": 0.0026, + "step": 50782 + }, + { + "epoch": 6.65, + "learning_rate": 1.452106190737489e-06, + "loss": 0.003, + "step": 50783 + }, + { + "epoch": 6.65, + "learning_rate": 1.4518309772746564e-06, + "loss": 0.0033, + "step": 50784 + }, + { + "epoch": 6.65, + "learning_rate": 1.4515557878530918e-06, + "loss": 0.0028, + "step": 50785 + }, + { + "epoch": 6.65, + "learning_rate": 1.4512806224735731e-06, + "loss": 0.0022, + "step": 50786 + }, + { + "epoch": 6.65, + "learning_rate": 1.4510054811368734e-06, + "loss": 0.0018, + "step": 50787 + }, + { + "epoch": 6.65, + "learning_rate": 1.450730363843763e-06, + "loss": 0.0017, + "step": 50788 + }, + { + "epoch": 6.65, + "learning_rate": 1.4504552705950204e-06, + "loss": 0.0028, + "step": 50789 + }, + { + "epoch": 6.65, + "learning_rate": 1.4501802013914169e-06, + "loss": 0.0014, + "step": 50790 + }, + { + "epoch": 6.65, + "learning_rate": 1.4499051562337262e-06, + "loss": 0.002, + "step": 50791 + }, + { + "epoch": 6.65, + "learning_rate": 1.4496301351227194e-06, + "loss": 0.0012, + "step": 50792 + }, + { + "epoch": 6.65, + "learning_rate": 1.4493551380591754e-06, + "loss": 0.0015, + "step": 50793 + }, + { + "epoch": 6.65, + "learning_rate": 1.4490801650438623e-06, + "loss": 0.0019, + "step": 50794 + }, + { + "epoch": 6.65, + "learning_rate": 1.4488052160775578e-06, + "loss": 0.0028, + "step": 50795 + }, + { + "epoch": 6.65, + "learning_rate": 1.448530291161031e-06, + "loss": 0.0016, + "step": 50796 + }, + { + "epoch": 6.65, + "learning_rate": 1.4482553902950603e-06, + "loss": 0.0025, + "step": 50797 + }, + { + "epoch": 6.65, + "learning_rate": 1.4479805134804147e-06, + "loss": 0.0016, + "step": 50798 + }, + { + "epoch": 6.65, + "learning_rate": 1.4477056607178697e-06, + "loss": 0.0023, + "step": 50799 + }, + { + "epoch": 6.65, + "learning_rate": 1.4474308320081941e-06, + "loss": 0.0017, + "step": 50800 + }, + { + "epoch": 6.65, + "learning_rate": 1.4471560273521668e-06, + "loss": 0.0016, + "step": 50801 + }, + { + "epoch": 6.65, + "learning_rate": 1.4468812467505565e-06, + "loss": 0.0014, + "step": 50802 + }, + { + "epoch": 6.65, + "learning_rate": 1.4466064902041355e-06, + "loss": 0.0026, + "step": 50803 + }, + { + "epoch": 6.65, + "learning_rate": 1.4463317577136794e-06, + "loss": 0.001, + "step": 50804 + }, + { + "epoch": 6.66, + "learning_rate": 1.4460570492799609e-06, + "loss": 0.0025, + "step": 50805 + }, + { + "epoch": 6.66, + "learning_rate": 1.4457823649037517e-06, + "loss": 0.0018, + "step": 50806 + }, + { + "epoch": 6.66, + "learning_rate": 1.4455077045858213e-06, + "loss": 0.001, + "step": 50807 + }, + { + "epoch": 6.66, + "learning_rate": 1.4452330683269478e-06, + "loss": 0.001, + "step": 50808 + }, + { + "epoch": 6.66, + "learning_rate": 1.4449584561279007e-06, + "loss": 0.0021, + "step": 50809 + }, + { + "epoch": 6.66, + "learning_rate": 1.4446838679894515e-06, + "loss": 0.0017, + "step": 50810 + }, + { + "epoch": 6.66, + "learning_rate": 1.4444093039123719e-06, + "loss": 0.0023, + "step": 50811 + }, + { + "epoch": 6.66, + "learning_rate": 1.4441347638974378e-06, + "loss": 0.0014, + "step": 50812 + }, + { + "epoch": 6.66, + "learning_rate": 1.443860247945419e-06, + "loss": 0.0014, + "step": 50813 + }, + { + "epoch": 6.66, + "learning_rate": 1.4435857560570843e-06, + "loss": 0.0027, + "step": 50814 + }, + { + "epoch": 6.66, + "learning_rate": 1.44331128823321e-06, + "loss": 0.0021, + "step": 50815 + }, + { + "epoch": 6.66, + "learning_rate": 1.443036844474569e-06, + "loss": 0.0015, + "step": 50816 + }, + { + "epoch": 6.66, + "learning_rate": 1.4427624247819317e-06, + "loss": 0.0014, + "step": 50817 + }, + { + "epoch": 6.66, + "learning_rate": 1.4424880291560695e-06, + "loss": 0.0023, + "step": 50818 + }, + { + "epoch": 6.66, + "learning_rate": 1.4422136575977508e-06, + "loss": 0.0013, + "step": 50819 + }, + { + "epoch": 6.66, + "learning_rate": 1.4419393101077527e-06, + "loss": 0.0018, + "step": 50820 + }, + { + "epoch": 6.66, + "learning_rate": 1.441664986686846e-06, + "loss": 0.0021, + "step": 50821 + }, + { + "epoch": 6.66, + "learning_rate": 1.4413906873357974e-06, + "loss": 0.0025, + "step": 50822 + }, + { + "epoch": 6.66, + "learning_rate": 1.4411164120553844e-06, + "loss": 0.0028, + "step": 50823 + }, + { + "epoch": 6.66, + "learning_rate": 1.440842160846375e-06, + "loss": 0.0016, + "step": 50824 + }, + { + "epoch": 6.66, + "learning_rate": 1.4405679337095413e-06, + "loss": 0.0014, + "step": 50825 + }, + { + "epoch": 6.66, + "learning_rate": 1.4402937306456533e-06, + "loss": 0.0018, + "step": 50826 + }, + { + "epoch": 6.66, + "learning_rate": 1.4400195516554827e-06, + "loss": 0.0019, + "step": 50827 + }, + { + "epoch": 6.66, + "learning_rate": 1.4397453967398033e-06, + "loss": 0.0013, + "step": 50828 + }, + { + "epoch": 6.66, + "learning_rate": 1.4394712658993836e-06, + "loss": 0.0021, + "step": 50829 + }, + { + "epoch": 6.66, + "learning_rate": 1.439197159134994e-06, + "loss": 0.0014, + "step": 50830 + }, + { + "epoch": 6.66, + "learning_rate": 1.4389230764474083e-06, + "loss": 0.0016, + "step": 50831 + }, + { + "epoch": 6.66, + "learning_rate": 1.4386490178373957e-06, + "loss": 0.0014, + "step": 50832 + }, + { + "epoch": 6.66, + "learning_rate": 1.4383749833057258e-06, + "loss": 0.0027, + "step": 50833 + }, + { + "epoch": 6.66, + "learning_rate": 1.4381009728531692e-06, + "loss": 0.0019, + "step": 50834 + }, + { + "epoch": 6.66, + "learning_rate": 1.4378269864804994e-06, + "loss": 0.002, + "step": 50835 + }, + { + "epoch": 6.66, + "learning_rate": 1.437553024188486e-06, + "loss": 0.0029, + "step": 50836 + }, + { + "epoch": 6.66, + "learning_rate": 1.4372790859778962e-06, + "loss": 0.0013, + "step": 50837 + }, + { + "epoch": 6.66, + "learning_rate": 1.4370051718495027e-06, + "loss": 0.002, + "step": 50838 + }, + { + "epoch": 6.66, + "learning_rate": 1.436731281804078e-06, + "loss": 0.0025, + "step": 50839 + }, + { + "epoch": 6.66, + "learning_rate": 1.4364574158423905e-06, + "loss": 0.0022, + "step": 50840 + }, + { + "epoch": 6.66, + "learning_rate": 1.436183573965212e-06, + "loss": 0.0026, + "step": 50841 + }, + { + "epoch": 6.66, + "learning_rate": 1.435909756173307e-06, + "loss": 0.0018, + "step": 50842 + }, + { + "epoch": 6.66, + "learning_rate": 1.4356359624674532e-06, + "loss": 0.0018, + "step": 50843 + }, + { + "epoch": 6.66, + "learning_rate": 1.4353621928484163e-06, + "loss": 0.0019, + "step": 50844 + }, + { + "epoch": 6.66, + "learning_rate": 1.4350884473169658e-06, + "loss": 0.0017, + "step": 50845 + }, + { + "epoch": 6.66, + "learning_rate": 1.4348147258738743e-06, + "loss": 0.0011, + "step": 50846 + }, + { + "epoch": 6.66, + "learning_rate": 1.4345410285199113e-06, + "loss": 0.0025, + "step": 50847 + }, + { + "epoch": 6.66, + "learning_rate": 1.4342673552558427e-06, + "loss": 0.003, + "step": 50848 + }, + { + "epoch": 6.66, + "learning_rate": 1.4339937060824428e-06, + "loss": 0.0013, + "step": 50849 + }, + { + "epoch": 6.66, + "learning_rate": 1.4337200810004782e-06, + "loss": 0.002, + "step": 50850 + }, + { + "epoch": 6.66, + "learning_rate": 1.4334464800107217e-06, + "loss": 0.002, + "step": 50851 + }, + { + "epoch": 6.66, + "learning_rate": 1.4331729031139397e-06, + "loss": 0.0031, + "step": 50852 + }, + { + "epoch": 6.66, + "learning_rate": 1.4328993503109012e-06, + "loss": 0.0026, + "step": 50853 + }, + { + "epoch": 6.66, + "learning_rate": 1.4326258216023791e-06, + "loss": 0.0019, + "step": 50854 + }, + { + "epoch": 6.66, + "learning_rate": 1.4323523169891419e-06, + "loss": 0.0021, + "step": 50855 + }, + { + "epoch": 6.66, + "learning_rate": 1.4320788364719563e-06, + "loss": 0.0029, + "step": 50856 + }, + { + "epoch": 6.66, + "learning_rate": 1.4318053800515907e-06, + "loss": 0.0018, + "step": 50857 + }, + { + "epoch": 6.66, + "learning_rate": 1.4315319477288182e-06, + "loss": 0.0035, + "step": 50858 + }, + { + "epoch": 6.66, + "learning_rate": 1.4312585395044043e-06, + "loss": 0.0008, + "step": 50859 + }, + { + "epoch": 6.66, + "learning_rate": 1.430985155379121e-06, + "loss": 0.0033, + "step": 50860 + }, + { + "epoch": 6.66, + "learning_rate": 1.430711795353733e-06, + "loss": 0.0015, + "step": 50861 + }, + { + "epoch": 6.66, + "learning_rate": 1.4304384594290143e-06, + "loss": 0.0011, + "step": 50862 + }, + { + "epoch": 6.66, + "learning_rate": 1.430165147605732e-06, + "loss": 0.0021, + "step": 50863 + }, + { + "epoch": 6.66, + "learning_rate": 1.42989185988465e-06, + "loss": 0.0027, + "step": 50864 + }, + { + "epoch": 6.66, + "learning_rate": 1.4296185962665443e-06, + "loss": 0.0021, + "step": 50865 + }, + { + "epoch": 6.66, + "learning_rate": 1.4293453567521799e-06, + "loss": 0.0013, + "step": 50866 + }, + { + "epoch": 6.66, + "learning_rate": 1.4290721413423248e-06, + "loss": 0.0028, + "step": 50867 + }, + { + "epoch": 6.66, + "learning_rate": 1.4287989500377453e-06, + "loss": 0.002, + "step": 50868 + }, + { + "epoch": 6.66, + "learning_rate": 1.4285257828392151e-06, + "loss": 0.0015, + "step": 50869 + }, + { + "epoch": 6.66, + "learning_rate": 1.4282526397474972e-06, + "loss": 0.0023, + "step": 50870 + }, + { + "epoch": 6.66, + "learning_rate": 1.4279795207633651e-06, + "loss": 0.0016, + "step": 50871 + }, + { + "epoch": 6.66, + "learning_rate": 1.4277064258875806e-06, + "loss": 0.002, + "step": 50872 + }, + { + "epoch": 6.66, + "learning_rate": 1.4274333551209185e-06, + "loss": 0.0014, + "step": 50873 + }, + { + "epoch": 6.66, + "learning_rate": 1.4271603084641439e-06, + "loss": 0.0031, + "step": 50874 + }, + { + "epoch": 6.66, + "learning_rate": 1.4268872859180238e-06, + "loss": 0.0012, + "step": 50875 + }, + { + "epoch": 6.66, + "learning_rate": 1.4266142874833245e-06, + "loss": 0.002, + "step": 50876 + }, + { + "epoch": 6.66, + "learning_rate": 1.4263413131608184e-06, + "loss": 0.0024, + "step": 50877 + }, + { + "epoch": 6.66, + "learning_rate": 1.4260683629512705e-06, + "loss": 0.0011, + "step": 50878 + }, + { + "epoch": 6.66, + "learning_rate": 1.425795436855446e-06, + "loss": 0.0014, + "step": 50879 + }, + { + "epoch": 6.66, + "learning_rate": 1.4255225348741187e-06, + "loss": 0.002, + "step": 50880 + }, + { + "epoch": 6.67, + "learning_rate": 1.42524965700805e-06, + "loss": 0.0034, + "step": 50881 + }, + { + "epoch": 6.67, + "learning_rate": 1.4249768032580113e-06, + "loss": 0.0017, + "step": 50882 + }, + { + "epoch": 6.67, + "learning_rate": 1.4247039736247692e-06, + "loss": 0.0017, + "step": 50883 + }, + { + "epoch": 6.67, + "learning_rate": 1.4244311681090884e-06, + "loss": 0.0019, + "step": 50884 + }, + { + "epoch": 6.67, + "learning_rate": 1.4241583867117415e-06, + "loss": 0.0026, + "step": 50885 + }, + { + "epoch": 6.67, + "learning_rate": 1.4238856294334912e-06, + "loss": 0.0012, + "step": 50886 + }, + { + "epoch": 6.67, + "learning_rate": 1.4236128962751038e-06, + "loss": 0.0021, + "step": 50887 + }, + { + "epoch": 6.67, + "learning_rate": 1.4233401872373509e-06, + "loss": 0.0024, + "step": 50888 + }, + { + "epoch": 6.67, + "learning_rate": 1.4230675023209972e-06, + "loss": 0.0022, + "step": 50889 + }, + { + "epoch": 6.67, + "learning_rate": 1.422794841526809e-06, + "loss": 0.0016, + "step": 50890 + }, + { + "epoch": 6.67, + "learning_rate": 1.422522204855551e-06, + "loss": 0.0041, + "step": 50891 + }, + { + "epoch": 6.67, + "learning_rate": 1.4222495923079926e-06, + "loss": 0.0018, + "step": 50892 + }, + { + "epoch": 6.67, + "learning_rate": 1.4219770038849034e-06, + "loss": 0.0018, + "step": 50893 + }, + { + "epoch": 6.67, + "learning_rate": 1.4217044395870461e-06, + "loss": 0.0018, + "step": 50894 + }, + { + "epoch": 6.67, + "learning_rate": 1.4214318994151866e-06, + "loss": 0.0008, + "step": 50895 + }, + { + "epoch": 6.67, + "learning_rate": 1.4211593833700954e-06, + "loss": 0.0015, + "step": 50896 + }, + { + "epoch": 6.67, + "learning_rate": 1.4208868914525365e-06, + "loss": 0.0011, + "step": 50897 + }, + { + "epoch": 6.67, + "learning_rate": 1.420614423663277e-06, + "loss": 0.0024, + "step": 50898 + }, + { + "epoch": 6.67, + "learning_rate": 1.4203419800030794e-06, + "loss": 0.0014, + "step": 50899 + }, + { + "epoch": 6.67, + "learning_rate": 1.4200695604727154e-06, + "loss": 0.0013, + "step": 50900 + }, + { + "epoch": 6.67, + "learning_rate": 1.4197971650729492e-06, + "loss": 0.0022, + "step": 50901 + }, + { + "epoch": 6.67, + "learning_rate": 1.4195247938045432e-06, + "loss": 0.0025, + "step": 50902 + }, + { + "epoch": 6.67, + "learning_rate": 1.4192524466682677e-06, + "loss": 0.0012, + "step": 50903 + }, + { + "epoch": 6.67, + "learning_rate": 1.4189801236648903e-06, + "loss": 0.0023, + "step": 50904 + }, + { + "epoch": 6.67, + "learning_rate": 1.4187078247951736e-06, + "loss": 0.0022, + "step": 50905 + }, + { + "epoch": 6.67, + "learning_rate": 1.4184355500598846e-06, + "loss": 0.0022, + "step": 50906 + }, + { + "epoch": 6.67, + "learning_rate": 1.418163299459787e-06, + "loss": 0.0018, + "step": 50907 + }, + { + "epoch": 6.67, + "learning_rate": 1.4178910729956496e-06, + "loss": 0.0022, + "step": 50908 + }, + { + "epoch": 6.67, + "learning_rate": 1.417618870668236e-06, + "loss": 0.0021, + "step": 50909 + }, + { + "epoch": 6.67, + "learning_rate": 1.4173466924783108e-06, + "loss": 0.0014, + "step": 50910 + }, + { + "epoch": 6.67, + "learning_rate": 1.4170745384266438e-06, + "loss": 0.0018, + "step": 50911 + }, + { + "epoch": 6.67, + "learning_rate": 1.4168024085139965e-06, + "loss": 0.0014, + "step": 50912 + }, + { + "epoch": 6.67, + "learning_rate": 1.416530302741136e-06, + "loss": 0.0014, + "step": 50913 + }, + { + "epoch": 6.67, + "learning_rate": 1.4162582211088216e-06, + "loss": 0.0022, + "step": 50914 + }, + { + "epoch": 6.67, + "learning_rate": 1.4159861636178286e-06, + "loss": 0.0033, + "step": 50915 + }, + { + "epoch": 6.67, + "learning_rate": 1.4157141302689182e-06, + "loss": 0.0021, + "step": 50916 + }, + { + "epoch": 6.67, + "learning_rate": 1.4154421210628543e-06, + "loss": 0.003, + "step": 50917 + }, + { + "epoch": 6.67, + "learning_rate": 1.4151701360004e-06, + "loss": 0.0012, + "step": 50918 + }, + { + "epoch": 6.67, + "learning_rate": 1.4148981750823242e-06, + "loss": 0.0021, + "step": 50919 + }, + { + "epoch": 6.67, + "learning_rate": 1.41462623830939e-06, + "loss": 0.0021, + "step": 50920 + }, + { + "epoch": 6.67, + "learning_rate": 1.4143543256823634e-06, + "loss": 0.001, + "step": 50921 + }, + { + "epoch": 6.67, + "learning_rate": 1.4140824372020056e-06, + "loss": 0.003, + "step": 50922 + }, + { + "epoch": 6.67, + "learning_rate": 1.4138105728690854e-06, + "loss": 0.0021, + "step": 50923 + }, + { + "epoch": 6.67, + "learning_rate": 1.4135387326843653e-06, + "loss": 0.0013, + "step": 50924 + }, + { + "epoch": 6.67, + "learning_rate": 1.413266916648609e-06, + "loss": 0.0021, + "step": 50925 + }, + { + "epoch": 6.67, + "learning_rate": 1.4129951247625818e-06, + "loss": 0.0018, + "step": 50926 + }, + { + "epoch": 6.67, + "learning_rate": 1.4127233570270514e-06, + "loss": 0.0019, + "step": 50927 + }, + { + "epoch": 6.67, + "learning_rate": 1.412451613442778e-06, + "loss": 0.0015, + "step": 50928 + }, + { + "epoch": 6.67, + "learning_rate": 1.4121798940105258e-06, + "loss": 0.0014, + "step": 50929 + }, + { + "epoch": 6.67, + "learning_rate": 1.4119081987310624e-06, + "loss": 0.0013, + "step": 50930 + }, + { + "epoch": 6.67, + "learning_rate": 1.4116365276051502e-06, + "loss": 0.002, + "step": 50931 + }, + { + "epoch": 6.67, + "learning_rate": 1.4113648806335523e-06, + "loss": 0.0014, + "step": 50932 + }, + { + "epoch": 6.67, + "learning_rate": 1.4110932578170312e-06, + "loss": 0.0015, + "step": 50933 + }, + { + "epoch": 6.67, + "learning_rate": 1.410821659156355e-06, + "loss": 0.0012, + "step": 50934 + }, + { + "epoch": 6.67, + "learning_rate": 1.4105500846522834e-06, + "loss": 0.0014, + "step": 50935 + }, + { + "epoch": 6.67, + "learning_rate": 1.4102785343055846e-06, + "loss": 0.0032, + "step": 50936 + }, + { + "epoch": 6.67, + "learning_rate": 1.410007008117018e-06, + "loss": 0.0013, + "step": 50937 + }, + { + "epoch": 6.67, + "learning_rate": 1.4097355060873508e-06, + "loss": 0.0019, + "step": 50938 + }, + { + "epoch": 6.67, + "learning_rate": 1.4094640282173454e-06, + "loss": 0.0017, + "step": 50939 + }, + { + "epoch": 6.67, + "learning_rate": 1.409192574507765e-06, + "loss": 0.0015, + "step": 50940 + }, + { + "epoch": 6.67, + "learning_rate": 1.408921144959371e-06, + "loss": 0.0015, + "step": 50941 + }, + { + "epoch": 6.67, + "learning_rate": 1.4086497395729315e-06, + "loss": 0.0015, + "step": 50942 + }, + { + "epoch": 6.67, + "learning_rate": 1.408378358349206e-06, + "loss": 0.0009, + "step": 50943 + }, + { + "epoch": 6.67, + "learning_rate": 1.4081070012889564e-06, + "loss": 0.0017, + "step": 50944 + }, + { + "epoch": 6.67, + "learning_rate": 1.4078356683929518e-06, + "loss": 0.0017, + "step": 50945 + }, + { + "epoch": 6.67, + "learning_rate": 1.407564359661948e-06, + "loss": 0.0014, + "step": 50946 + }, + { + "epoch": 6.67, + "learning_rate": 1.407293075096715e-06, + "loss": 0.0015, + "step": 50947 + }, + { + "epoch": 6.67, + "learning_rate": 1.407021814698013e-06, + "loss": 0.0015, + "step": 50948 + }, + { + "epoch": 6.67, + "learning_rate": 1.4067505784666024e-06, + "loss": 0.002, + "step": 50949 + }, + { + "epoch": 6.67, + "learning_rate": 1.4064793664032494e-06, + "loss": 0.0013, + "step": 50950 + }, + { + "epoch": 6.67, + "learning_rate": 1.4062081785087168e-06, + "loss": 0.0018, + "step": 50951 + }, + { + "epoch": 6.67, + "learning_rate": 1.4059370147837625e-06, + "loss": 0.0016, + "step": 50952 + }, + { + "epoch": 6.67, + "learning_rate": 1.4056658752291562e-06, + "loss": 0.001, + "step": 50953 + }, + { + "epoch": 6.67, + "learning_rate": 1.4053947598456564e-06, + "loss": 0.0018, + "step": 50954 + }, + { + "epoch": 6.67, + "learning_rate": 1.4051236686340253e-06, + "loss": 0.0018, + "step": 50955 + }, + { + "epoch": 6.67, + "learning_rate": 1.4048526015950248e-06, + "loss": 0.0022, + "step": 50956 + }, + { + "epoch": 6.68, + "learning_rate": 1.4045815587294187e-06, + "loss": 0.0007, + "step": 50957 + }, + { + "epoch": 6.68, + "learning_rate": 1.4043105400379707e-06, + "loss": 0.0025, + "step": 50958 + }, + { + "epoch": 6.68, + "learning_rate": 1.4040395455214428e-06, + "loss": 0.0019, + "step": 50959 + }, + { + "epoch": 6.68, + "learning_rate": 1.403768575180593e-06, + "loss": 0.0013, + "step": 50960 + }, + { + "epoch": 6.68, + "learning_rate": 1.4034976290161885e-06, + "loss": 0.0019, + "step": 50961 + }, + { + "epoch": 6.68, + "learning_rate": 1.4032267070289895e-06, + "loss": 0.0017, + "step": 50962 + }, + { + "epoch": 6.68, + "learning_rate": 1.402955809219757e-06, + "loss": 0.002, + "step": 50963 + }, + { + "epoch": 6.68, + "learning_rate": 1.4026849355892514e-06, + "loss": 0.0017, + "step": 50964 + }, + { + "epoch": 6.68, + "learning_rate": 1.4024140861382396e-06, + "loss": 0.0016, + "step": 50965 + }, + { + "epoch": 6.68, + "learning_rate": 1.402143260867479e-06, + "loss": 0.0021, + "step": 50966 + }, + { + "epoch": 6.68, + "learning_rate": 1.401872459777731e-06, + "loss": 0.0016, + "step": 50967 + }, + { + "epoch": 6.68, + "learning_rate": 1.4016016828697586e-06, + "loss": 0.0016, + "step": 50968 + }, + { + "epoch": 6.68, + "learning_rate": 1.4013309301443268e-06, + "loss": 0.0017, + "step": 50969 + }, + { + "epoch": 6.68, + "learning_rate": 1.4010602016021934e-06, + "loss": 0.0022, + "step": 50970 + }, + { + "epoch": 6.68, + "learning_rate": 1.4007894972441195e-06, + "loss": 0.0029, + "step": 50971 + }, + { + "epoch": 6.68, + "learning_rate": 1.400518817070866e-06, + "loss": 0.002, + "step": 50972 + }, + { + "epoch": 6.68, + "learning_rate": 1.4002481610831965e-06, + "loss": 0.0009, + "step": 50973 + }, + { + "epoch": 6.68, + "learning_rate": 1.3999775292818719e-06, + "loss": 0.001, + "step": 50974 + }, + { + "epoch": 6.68, + "learning_rate": 1.3997069216676495e-06, + "loss": 0.0014, + "step": 50975 + }, + { + "epoch": 6.68, + "learning_rate": 1.3994363382412968e-06, + "loss": 0.0033, + "step": 50976 + }, + { + "epoch": 6.68, + "learning_rate": 1.3991657790035706e-06, + "loss": 0.0014, + "step": 50977 + }, + { + "epoch": 6.68, + "learning_rate": 1.3988952439552328e-06, + "loss": 0.0018, + "step": 50978 + }, + { + "epoch": 6.68, + "learning_rate": 1.3986247330970394e-06, + "loss": 0.0015, + "step": 50979 + }, + { + "epoch": 6.68, + "learning_rate": 1.3983542464297607e-06, + "loss": 0.0005, + "step": 50980 + }, + { + "epoch": 6.68, + "learning_rate": 1.3980837839541527e-06, + "loss": 0.0011, + "step": 50981 + }, + { + "epoch": 6.68, + "learning_rate": 1.3978133456709763e-06, + "loss": 0.0026, + "step": 50982 + }, + { + "epoch": 6.68, + "learning_rate": 1.3975429315809885e-06, + "loss": 0.0026, + "step": 50983 + }, + { + "epoch": 6.68, + "learning_rate": 1.3972725416849563e-06, + "loss": 0.0016, + "step": 50984 + }, + { + "epoch": 6.68, + "learning_rate": 1.3970021759836362e-06, + "loss": 0.002, + "step": 50985 + }, + { + "epoch": 6.68, + "learning_rate": 1.3967318344777892e-06, + "loss": 0.0014, + "step": 50986 + }, + { + "epoch": 6.68, + "learning_rate": 1.3964615171681739e-06, + "loss": 0.0013, + "step": 50987 + }, + { + "epoch": 6.68, + "learning_rate": 1.3961912240555552e-06, + "loss": 0.0015, + "step": 50988 + }, + { + "epoch": 6.68, + "learning_rate": 1.3959209551406904e-06, + "loss": 0.0017, + "step": 50989 + }, + { + "epoch": 6.68, + "learning_rate": 1.3956507104243366e-06, + "loss": 0.0016, + "step": 50990 + }, + { + "epoch": 6.68, + "learning_rate": 1.3953804899072577e-06, + "loss": 0.0016, + "step": 50991 + }, + { + "epoch": 6.68, + "learning_rate": 1.395110293590215e-06, + "loss": 0.0023, + "step": 50992 + }, + { + "epoch": 6.68, + "learning_rate": 1.3948401214739659e-06, + "loss": 0.0012, + "step": 50993 + }, + { + "epoch": 6.68, + "learning_rate": 1.3945699735592687e-06, + "loss": 0.0017, + "step": 50994 + }, + { + "epoch": 6.68, + "learning_rate": 1.3942998498468884e-06, + "loss": 0.0016, + "step": 50995 + }, + { + "epoch": 6.68, + "learning_rate": 1.39402975033758e-06, + "loss": 0.0017, + "step": 50996 + }, + { + "epoch": 6.68, + "learning_rate": 1.3937596750321058e-06, + "loss": 0.0035, + "step": 50997 + }, + { + "epoch": 6.68, + "learning_rate": 1.3934896239312213e-06, + "loss": 0.001, + "step": 50998 + }, + { + "epoch": 6.68, + "learning_rate": 1.3932195970356922e-06, + "loss": 0.0021, + "step": 50999 + }, + { + "epoch": 6.68, + "learning_rate": 1.3929495943462734e-06, + "loss": 0.0014, + "step": 51000 + }, + { + "epoch": 6.68, + "learning_rate": 1.3926796158637234e-06, + "loss": 0.0021, + "step": 51001 + }, + { + "epoch": 6.68, + "learning_rate": 1.392409661588804e-06, + "loss": 0.0026, + "step": 51002 + }, + { + "epoch": 6.68, + "learning_rate": 1.3921397315222763e-06, + "loss": 0.0011, + "step": 51003 + }, + { + "epoch": 6.68, + "learning_rate": 1.3918698256648966e-06, + "loss": 0.0027, + "step": 51004 + }, + { + "epoch": 6.68, + "learning_rate": 1.3915999440174255e-06, + "loss": 0.0015, + "step": 51005 + }, + { + "epoch": 6.68, + "learning_rate": 1.3913300865806178e-06, + "loss": 0.002, + "step": 51006 + }, + { + "epoch": 6.68, + "learning_rate": 1.3910602533552397e-06, + "loss": 0.0013, + "step": 51007 + }, + { + "epoch": 6.68, + "learning_rate": 1.390790444342045e-06, + "loss": 0.001, + "step": 51008 + }, + { + "epoch": 6.68, + "learning_rate": 1.3905206595417919e-06, + "loss": 0.0018, + "step": 51009 + }, + { + "epoch": 6.68, + "learning_rate": 1.390250898955242e-06, + "loss": 0.0015, + "step": 51010 + }, + { + "epoch": 6.68, + "learning_rate": 1.389981162583155e-06, + "loss": 0.002, + "step": 51011 + }, + { + "epoch": 6.68, + "learning_rate": 1.3897114504262831e-06, + "loss": 0.0025, + "step": 51012 + }, + { + "epoch": 6.68, + "learning_rate": 1.389441762485393e-06, + "loss": 0.0011, + "step": 51013 + }, + { + "epoch": 6.68, + "learning_rate": 1.3891720987612367e-06, + "loss": 0.0018, + "step": 51014 + }, + { + "epoch": 6.68, + "learning_rate": 1.3889024592545775e-06, + "loss": 0.0021, + "step": 51015 + }, + { + "epoch": 6.68, + "learning_rate": 1.3886328439661723e-06, + "loss": 0.0013, + "step": 51016 + }, + { + "epoch": 6.68, + "learning_rate": 1.388363252896775e-06, + "loss": 0.0023, + "step": 51017 + }, + { + "epoch": 6.68, + "learning_rate": 1.3880936860471506e-06, + "loss": 0.0017, + "step": 51018 + }, + { + "epoch": 6.68, + "learning_rate": 1.3878241434180538e-06, + "loss": 0.0006, + "step": 51019 + }, + { + "epoch": 6.68, + "learning_rate": 1.3875546250102435e-06, + "loss": 0.0031, + "step": 51020 + }, + { + "epoch": 6.68, + "learning_rate": 1.387285130824474e-06, + "loss": 0.0014, + "step": 51021 + }, + { + "epoch": 6.68, + "learning_rate": 1.3870156608615092e-06, + "loss": 0.0022, + "step": 51022 + }, + { + "epoch": 6.68, + "learning_rate": 1.3867462151221024e-06, + "loss": 0.0019, + "step": 51023 + }, + { + "epoch": 6.68, + "learning_rate": 1.3864767936070145e-06, + "loss": 0.002, + "step": 51024 + }, + { + "epoch": 6.68, + "learning_rate": 1.3862073963169996e-06, + "loss": 0.0021, + "step": 51025 + }, + { + "epoch": 6.68, + "learning_rate": 1.3859380232528208e-06, + "loss": 0.0024, + "step": 51026 + }, + { + "epoch": 6.68, + "learning_rate": 1.3856686744152325e-06, + "loss": 0.0017, + "step": 51027 + }, + { + "epoch": 6.68, + "learning_rate": 1.3853993498049923e-06, + "loss": 0.0027, + "step": 51028 + }, + { + "epoch": 6.68, + "learning_rate": 1.3851300494228549e-06, + "loss": 0.0008, + "step": 51029 + }, + { + "epoch": 6.68, + "learning_rate": 1.3848607732695828e-06, + "loss": 0.0014, + "step": 51030 + }, + { + "epoch": 6.68, + "learning_rate": 1.3845915213459315e-06, + "loss": 0.0012, + "step": 51031 + }, + { + "epoch": 6.68, + "learning_rate": 1.3843222936526547e-06, + "loss": 0.0025, + "step": 51032 + }, + { + "epoch": 6.68, + "learning_rate": 1.3840530901905158e-06, + "loss": 0.0014, + "step": 51033 + }, + { + "epoch": 6.69, + "learning_rate": 1.3837839109602659e-06, + "loss": 0.0016, + "step": 51034 + }, + { + "epoch": 6.69, + "learning_rate": 1.3835147559626672e-06, + "loss": 0.0023, + "step": 51035 + }, + { + "epoch": 6.69, + "learning_rate": 1.383245625198475e-06, + "loss": 0.0029, + "step": 51036 + }, + { + "epoch": 6.69, + "learning_rate": 1.382976518668443e-06, + "loss": 0.001, + "step": 51037 + }, + { + "epoch": 6.69, + "learning_rate": 1.3827074363733327e-06, + "loss": 0.0023, + "step": 51038 + }, + { + "epoch": 6.69, + "learning_rate": 1.3824383783138983e-06, + "loss": 0.0015, + "step": 51039 + }, + { + "epoch": 6.69, + "learning_rate": 1.3821693444908957e-06, + "loss": 0.0023, + "step": 51040 + }, + { + "epoch": 6.69, + "learning_rate": 1.3819003349050852e-06, + "loss": 0.0015, + "step": 51041 + }, + { + "epoch": 6.69, + "learning_rate": 1.381631349557221e-06, + "loss": 0.0032, + "step": 51042 + }, + { + "epoch": 6.69, + "learning_rate": 1.3813623884480588e-06, + "loss": 0.0011, + "step": 51043 + }, + { + "epoch": 6.69, + "learning_rate": 1.3810934515783548e-06, + "loss": 0.0019, + "step": 51044 + }, + { + "epoch": 6.69, + "learning_rate": 1.380824538948865e-06, + "loss": 0.002, + "step": 51045 + }, + { + "epoch": 6.69, + "learning_rate": 1.3805556505603513e-06, + "loss": 0.002, + "step": 51046 + }, + { + "epoch": 6.69, + "learning_rate": 1.380286786413565e-06, + "loss": 0.0023, + "step": 51047 + }, + { + "epoch": 6.69, + "learning_rate": 1.38001794650926e-06, + "loss": 0.0011, + "step": 51048 + }, + { + "epoch": 6.69, + "learning_rate": 1.3797491308481991e-06, + "loss": 0.0016, + "step": 51049 + }, + { + "epoch": 6.69, + "learning_rate": 1.379480339431134e-06, + "loss": 0.0019, + "step": 51050 + }, + { + "epoch": 6.69, + "learning_rate": 1.3792115722588218e-06, + "loss": 0.0016, + "step": 51051 + }, + { + "epoch": 6.69, + "learning_rate": 1.378942829332015e-06, + "loss": 0.0022, + "step": 51052 + }, + { + "epoch": 6.69, + "learning_rate": 1.3786741106514746e-06, + "loss": 0.0013, + "step": 51053 + }, + { + "epoch": 6.69, + "learning_rate": 1.3784054162179549e-06, + "loss": 0.0025, + "step": 51054 + }, + { + "epoch": 6.69, + "learning_rate": 1.378136746032208e-06, + "loss": 0.0015, + "step": 51055 + }, + { + "epoch": 6.69, + "learning_rate": 1.3778681000949933e-06, + "loss": 0.0017, + "step": 51056 + }, + { + "epoch": 6.69, + "learning_rate": 1.3775994784070668e-06, + "loss": 0.002, + "step": 51057 + }, + { + "epoch": 6.69, + "learning_rate": 1.377330880969182e-06, + "loss": 0.0019, + "step": 51058 + }, + { + "epoch": 6.69, + "learning_rate": 1.3770623077820944e-06, + "loss": 0.0013, + "step": 51059 + }, + { + "epoch": 6.69, + "learning_rate": 1.3767937588465608e-06, + "loss": 0.0029, + "step": 51060 + }, + { + "epoch": 6.69, + "learning_rate": 1.3765252341633362e-06, + "loss": 0.0009, + "step": 51061 + }, + { + "epoch": 6.69, + "learning_rate": 1.3762567337331745e-06, + "loss": 0.0015, + "step": 51062 + }, + { + "epoch": 6.69, + "learning_rate": 1.3759882575568296e-06, + "loss": 0.0013, + "step": 51063 + }, + { + "epoch": 6.69, + "learning_rate": 1.3757198056350606e-06, + "loss": 0.0016, + "step": 51064 + }, + { + "epoch": 6.69, + "learning_rate": 1.3754513779686206e-06, + "loss": 0.002, + "step": 51065 + }, + { + "epoch": 6.69, + "learning_rate": 1.3751829745582646e-06, + "loss": 0.0017, + "step": 51066 + }, + { + "epoch": 6.69, + "learning_rate": 1.3749145954047427e-06, + "loss": 0.0014, + "step": 51067 + }, + { + "epoch": 6.69, + "learning_rate": 1.374646240508818e-06, + "loss": 0.0021, + "step": 51068 + }, + { + "epoch": 6.69, + "learning_rate": 1.3743779098712417e-06, + "loss": 0.002, + "step": 51069 + }, + { + "epoch": 6.69, + "learning_rate": 1.374109603492768e-06, + "loss": 0.0014, + "step": 51070 + }, + { + "epoch": 6.69, + "learning_rate": 1.3738413213741496e-06, + "loss": 0.0018, + "step": 51071 + }, + { + "epoch": 6.69, + "learning_rate": 1.373573063516146e-06, + "loss": 0.0018, + "step": 51072 + }, + { + "epoch": 6.69, + "learning_rate": 1.3733048299195073e-06, + "loss": 0.001, + "step": 51073 + }, + { + "epoch": 6.69, + "learning_rate": 1.3730366205849888e-06, + "loss": 0.0023, + "step": 51074 + }, + { + "epoch": 6.69, + "learning_rate": 1.3727684355133464e-06, + "loss": 0.0031, + "step": 51075 + }, + { + "epoch": 6.69, + "learning_rate": 1.372500274705334e-06, + "loss": 0.0012, + "step": 51076 + }, + { + "epoch": 6.69, + "learning_rate": 1.372232138161702e-06, + "loss": 0.0011, + "step": 51077 + }, + { + "epoch": 6.69, + "learning_rate": 1.3719640258832112e-06, + "loss": 0.0012, + "step": 51078 + }, + { + "epoch": 6.69, + "learning_rate": 1.3716959378706085e-06, + "loss": 0.0015, + "step": 51079 + }, + { + "epoch": 6.69, + "learning_rate": 1.3714278741246545e-06, + "loss": 0.0014, + "step": 51080 + }, + { + "epoch": 6.69, + "learning_rate": 1.3711598346461007e-06, + "loss": 0.0016, + "step": 51081 + }, + { + "epoch": 6.69, + "learning_rate": 1.370891819435698e-06, + "loss": 0.0022, + "step": 51082 + }, + { + "epoch": 6.69, + "learning_rate": 1.3706238284942042e-06, + "loss": 0.0021, + "step": 51083 + }, + { + "epoch": 6.69, + "learning_rate": 1.370355861822371e-06, + "loss": 0.0012, + "step": 51084 + }, + { + "epoch": 6.69, + "learning_rate": 1.3700879194209526e-06, + "loss": 0.0018, + "step": 51085 + }, + { + "epoch": 6.69, + "learning_rate": 1.3698200012907005e-06, + "loss": 0.0022, + "step": 51086 + }, + { + "epoch": 6.69, + "learning_rate": 1.3695521074323714e-06, + "loss": 0.0017, + "step": 51087 + }, + { + "epoch": 6.69, + "learning_rate": 1.3692842378467164e-06, + "loss": 0.0025, + "step": 51088 + }, + { + "epoch": 6.69, + "learning_rate": 1.3690163925344912e-06, + "loss": 0.0014, + "step": 51089 + }, + { + "epoch": 6.69, + "learning_rate": 1.3687485714964467e-06, + "loss": 0.0022, + "step": 51090 + }, + { + "epoch": 6.69, + "learning_rate": 1.3684807747333385e-06, + "loss": 0.002, + "step": 51091 + }, + { + "epoch": 6.69, + "learning_rate": 1.3682130022459183e-06, + "loss": 0.0013, + "step": 51092 + }, + { + "epoch": 6.69, + "learning_rate": 1.3679452540349402e-06, + "loss": 0.0019, + "step": 51093 + }, + { + "epoch": 6.69, + "learning_rate": 1.3676775301011536e-06, + "loss": 0.0014, + "step": 51094 + }, + { + "epoch": 6.69, + "learning_rate": 1.3674098304453165e-06, + "loss": 0.0024, + "step": 51095 + }, + { + "epoch": 6.69, + "learning_rate": 1.3671421550681795e-06, + "loss": 0.0023, + "step": 51096 + }, + { + "epoch": 6.69, + "learning_rate": 1.3668745039704934e-06, + "loss": 0.003, + "step": 51097 + }, + { + "epoch": 6.69, + "learning_rate": 1.366606877153015e-06, + "loss": 0.0023, + "step": 51098 + }, + { + "epoch": 6.69, + "learning_rate": 1.3663392746164927e-06, + "loss": 0.0015, + "step": 51099 + }, + { + "epoch": 6.69, + "learning_rate": 1.3660716963616839e-06, + "loss": 0.0029, + "step": 51100 + }, + { + "epoch": 6.69, + "learning_rate": 1.3658041423893386e-06, + "loss": 0.0011, + "step": 51101 + }, + { + "epoch": 6.69, + "learning_rate": 1.3655366127002067e-06, + "loss": 0.0018, + "step": 51102 + }, + { + "epoch": 6.69, + "learning_rate": 1.3652691072950463e-06, + "loss": 0.0027, + "step": 51103 + }, + { + "epoch": 6.69, + "learning_rate": 1.3650016261746057e-06, + "loss": 0.0017, + "step": 51104 + }, + { + "epoch": 6.69, + "learning_rate": 1.3647341693396365e-06, + "loss": 0.0008, + "step": 51105 + }, + { + "epoch": 6.69, + "learning_rate": 1.364466736790896e-06, + "loss": 0.0009, + "step": 51106 + }, + { + "epoch": 6.69, + "learning_rate": 1.364199328529131e-06, + "loss": 0.0023, + "step": 51107 + }, + { + "epoch": 6.69, + "learning_rate": 1.363931944555097e-06, + "loss": 0.0012, + "step": 51108 + }, + { + "epoch": 6.69, + "learning_rate": 1.3636645848695418e-06, + "loss": 0.002, + "step": 51109 + }, + { + "epoch": 6.7, + "learning_rate": 1.3633972494732195e-06, + "loss": 0.0011, + "step": 51110 + }, + { + "epoch": 6.7, + "learning_rate": 1.363129938366885e-06, + "loss": 0.0016, + "step": 51111 + }, + { + "epoch": 6.7, + "learning_rate": 1.3628626515512876e-06, + "loss": 0.0015, + "step": 51112 + }, + { + "epoch": 6.7, + "learning_rate": 1.362595389027177e-06, + "loss": 0.0019, + "step": 51113 + }, + { + "epoch": 6.7, + "learning_rate": 1.362328150795309e-06, + "loss": 0.0022, + "step": 51114 + }, + { + "epoch": 6.7, + "learning_rate": 1.3620609368564331e-06, + "loss": 0.0009, + "step": 51115 + }, + { + "epoch": 6.7, + "learning_rate": 1.3617937472113007e-06, + "loss": 0.0025, + "step": 51116 + }, + { + "epoch": 6.7, + "learning_rate": 1.3615265818606615e-06, + "loss": 0.002, + "step": 51117 + }, + { + "epoch": 6.7, + "learning_rate": 1.3612594408052715e-06, + "loss": 0.0022, + "step": 51118 + }, + { + "epoch": 6.7, + "learning_rate": 1.3609923240458777e-06, + "loss": 0.0011, + "step": 51119 + }, + { + "epoch": 6.7, + "learning_rate": 1.360725231583232e-06, + "loss": 0.0017, + "step": 51120 + }, + { + "epoch": 6.7, + "learning_rate": 1.3604581634180858e-06, + "loss": 0.0021, + "step": 51121 + }, + { + "epoch": 6.7, + "learning_rate": 1.3601911195511942e-06, + "loss": 0.0021, + "step": 51122 + }, + { + "epoch": 6.7, + "learning_rate": 1.3599240999833042e-06, + "loss": 0.0018, + "step": 51123 + }, + { + "epoch": 6.7, + "learning_rate": 1.3596571047151653e-06, + "loss": 0.0014, + "step": 51124 + }, + { + "epoch": 6.7, + "learning_rate": 1.3593901337475324e-06, + "loss": 0.002, + "step": 51125 + }, + { + "epoch": 6.7, + "learning_rate": 1.3591231870811561e-06, + "loss": 0.0013, + "step": 51126 + }, + { + "epoch": 6.7, + "learning_rate": 1.3588562647167847e-06, + "loss": 0.0015, + "step": 51127 + }, + { + "epoch": 6.7, + "learning_rate": 1.3585893666551675e-06, + "loss": 0.0014, + "step": 51128 + }, + { + "epoch": 6.7, + "learning_rate": 1.3583224928970596e-06, + "loss": 0.0027, + "step": 51129 + }, + { + "epoch": 6.7, + "learning_rate": 1.35805564344321e-06, + "loss": 0.0024, + "step": 51130 + }, + { + "epoch": 6.7, + "learning_rate": 1.3577888182943666e-06, + "loss": 0.0018, + "step": 51131 + }, + { + "epoch": 6.7, + "learning_rate": 1.3575220174512816e-06, + "loss": 0.0008, + "step": 51132 + }, + { + "epoch": 6.7, + "learning_rate": 1.3572552409147077e-06, + "loss": 0.0011, + "step": 51133 + }, + { + "epoch": 6.7, + "learning_rate": 1.3569884886853935e-06, + "loss": 0.003, + "step": 51134 + }, + { + "epoch": 6.7, + "learning_rate": 1.3567217607640892e-06, + "loss": 0.0018, + "step": 51135 + }, + { + "epoch": 6.7, + "learning_rate": 1.3564550571515422e-06, + "loss": 0.0023, + "step": 51136 + }, + { + "epoch": 6.7, + "learning_rate": 1.3561883778485073e-06, + "loss": 0.0018, + "step": 51137 + }, + { + "epoch": 6.7, + "learning_rate": 1.355921722855732e-06, + "loss": 0.0012, + "step": 51138 + }, + { + "epoch": 6.7, + "learning_rate": 1.3556550921739642e-06, + "loss": 0.0021, + "step": 51139 + }, + { + "epoch": 6.7, + "learning_rate": 1.355388485803959e-06, + "loss": 0.0014, + "step": 51140 + }, + { + "epoch": 6.7, + "learning_rate": 1.3551219037464625e-06, + "loss": 0.002, + "step": 51141 + }, + { + "epoch": 6.7, + "learning_rate": 1.3548553460022262e-06, + "loss": 0.001, + "step": 51142 + }, + { + "epoch": 6.7, + "learning_rate": 1.3545888125719965e-06, + "loss": 0.0022, + "step": 51143 + }, + { + "epoch": 6.7, + "learning_rate": 1.3543223034565245e-06, + "loss": 0.0013, + "step": 51144 + }, + { + "epoch": 6.7, + "learning_rate": 1.3540558186565633e-06, + "loss": 0.0019, + "step": 51145 + }, + { + "epoch": 6.7, + "learning_rate": 1.3537893581728602e-06, + "loss": 0.0019, + "step": 51146 + }, + { + "epoch": 6.7, + "learning_rate": 1.3535229220061607e-06, + "loss": 0.0016, + "step": 51147 + }, + { + "epoch": 6.7, + "learning_rate": 1.3532565101572204e-06, + "loss": 0.0022, + "step": 51148 + }, + { + "epoch": 6.7, + "learning_rate": 1.3529901226267861e-06, + "loss": 0.0021, + "step": 51149 + }, + { + "epoch": 6.7, + "learning_rate": 1.3527237594156061e-06, + "loss": 0.0016, + "step": 51150 + }, + { + "epoch": 6.7, + "learning_rate": 1.3524574205244278e-06, + "loss": 0.0016, + "step": 51151 + }, + { + "epoch": 6.7, + "learning_rate": 1.352191105954005e-06, + "loss": 0.002, + "step": 51152 + }, + { + "epoch": 6.7, + "learning_rate": 1.3519248157050847e-06, + "loss": 0.0015, + "step": 51153 + }, + { + "epoch": 6.7, + "learning_rate": 1.3516585497784118e-06, + "loss": 0.0024, + "step": 51154 + }, + { + "epoch": 6.7, + "learning_rate": 1.3513923081747392e-06, + "loss": 0.0026, + "step": 51155 + }, + { + "epoch": 6.7, + "learning_rate": 1.3511260908948175e-06, + "loss": 0.0022, + "step": 51156 + }, + { + "epoch": 6.7, + "learning_rate": 1.3508598979393938e-06, + "loss": 0.0014, + "step": 51157 + }, + { + "epoch": 6.7, + "learning_rate": 1.3505937293092154e-06, + "loss": 0.0011, + "step": 51158 + }, + { + "epoch": 6.7, + "learning_rate": 1.350327585005029e-06, + "loss": 0.001, + "step": 51159 + }, + { + "epoch": 6.7, + "learning_rate": 1.350061465027589e-06, + "loss": 0.0019, + "step": 51160 + }, + { + "epoch": 6.7, + "learning_rate": 1.3497953693776389e-06, + "loss": 0.0014, + "step": 51161 + }, + { + "epoch": 6.7, + "learning_rate": 1.3495292980559272e-06, + "loss": 0.001, + "step": 51162 + }, + { + "epoch": 6.7, + "learning_rate": 1.3492632510632064e-06, + "loss": 0.002, + "step": 51163 + }, + { + "epoch": 6.7, + "learning_rate": 1.3489972284002218e-06, + "loss": 0.0017, + "step": 51164 + }, + { + "epoch": 6.7, + "learning_rate": 1.3487312300677191e-06, + "loss": 0.0023, + "step": 51165 + }, + { + "epoch": 6.7, + "learning_rate": 1.3484652560664513e-06, + "loss": 0.0014, + "step": 51166 + }, + { + "epoch": 6.7, + "learning_rate": 1.3481993063971622e-06, + "loss": 0.0016, + "step": 51167 + }, + { + "epoch": 6.7, + "learning_rate": 1.3479333810606043e-06, + "loss": 0.0019, + "step": 51168 + }, + { + "epoch": 6.7, + "learning_rate": 1.3476674800575229e-06, + "loss": 0.0019, + "step": 51169 + }, + { + "epoch": 6.7, + "learning_rate": 1.3474016033886638e-06, + "loss": 0.0016, + "step": 51170 + }, + { + "epoch": 6.7, + "learning_rate": 1.3471357510547789e-06, + "loss": 0.0009, + "step": 51171 + }, + { + "epoch": 6.7, + "learning_rate": 1.346869923056614e-06, + "loss": 0.0012, + "step": 51172 + }, + { + "epoch": 6.7, + "learning_rate": 1.3466041193949164e-06, + "loss": 0.0014, + "step": 51173 + }, + { + "epoch": 6.7, + "learning_rate": 1.3463383400704322e-06, + "loss": 0.001, + "step": 51174 + }, + { + "epoch": 6.7, + "learning_rate": 1.3460725850839118e-06, + "loss": 0.0007, + "step": 51175 + }, + { + "epoch": 6.7, + "learning_rate": 1.3458068544361002e-06, + "loss": 0.0008, + "step": 51176 + }, + { + "epoch": 6.7, + "learning_rate": 1.345541148127748e-06, + "loss": 0.0019, + "step": 51177 + }, + { + "epoch": 6.7, + "learning_rate": 1.345275466159598e-06, + "loss": 0.0019, + "step": 51178 + }, + { + "epoch": 6.7, + "learning_rate": 1.3450098085324026e-06, + "loss": 0.0021, + "step": 51179 + }, + { + "epoch": 6.7, + "learning_rate": 1.344744175246906e-06, + "loss": 0.0015, + "step": 51180 + }, + { + "epoch": 6.7, + "learning_rate": 1.344478566303853e-06, + "loss": 0.002, + "step": 51181 + }, + { + "epoch": 6.7, + "learning_rate": 1.344212981703995e-06, + "loss": 0.0012, + "step": 51182 + }, + { + "epoch": 6.7, + "learning_rate": 1.3439474214480774e-06, + "loss": 0.0014, + "step": 51183 + }, + { + "epoch": 6.7, + "learning_rate": 1.3436818855368471e-06, + "loss": 0.0028, + "step": 51184 + }, + { + "epoch": 6.7, + "learning_rate": 1.3434163739710472e-06, + "loss": 0.0027, + "step": 51185 + }, + { + "epoch": 6.71, + "learning_rate": 1.343150886751431e-06, + "loss": 0.0014, + "step": 51186 + }, + { + "epoch": 6.71, + "learning_rate": 1.3428854238787391e-06, + "loss": 0.0035, + "step": 51187 + }, + { + "epoch": 6.71, + "learning_rate": 1.3426199853537235e-06, + "loss": 0.0037, + "step": 51188 + }, + { + "epoch": 6.71, + "learning_rate": 1.3423545711771257e-06, + "loss": 0.0022, + "step": 51189 + }, + { + "epoch": 6.71, + "learning_rate": 1.3420891813496972e-06, + "loss": 0.0023, + "step": 51190 + }, + { + "epoch": 6.71, + "learning_rate": 1.3418238158721819e-06, + "loss": 0.0025, + "step": 51191 + }, + { + "epoch": 6.71, + "learning_rate": 1.3415584747453258e-06, + "loss": 0.0014, + "step": 51192 + }, + { + "epoch": 6.71, + "learning_rate": 1.3412931579698729e-06, + "loss": 0.0011, + "step": 51193 + }, + { + "epoch": 6.71, + "learning_rate": 1.3410278655465748e-06, + "loss": 0.0015, + "step": 51194 + }, + { + "epoch": 6.71, + "learning_rate": 1.340762597476174e-06, + "loss": 0.0017, + "step": 51195 + }, + { + "epoch": 6.71, + "learning_rate": 1.3404973537594156e-06, + "loss": 0.0009, + "step": 51196 + }, + { + "epoch": 6.71, + "learning_rate": 1.3402321343970492e-06, + "loss": 0.0016, + "step": 51197 + }, + { + "epoch": 6.71, + "learning_rate": 1.3399669393898163e-06, + "loss": 0.0022, + "step": 51198 + }, + { + "epoch": 6.71, + "learning_rate": 1.3397017687384684e-06, + "loss": 0.003, + "step": 51199 + }, + { + "epoch": 6.71, + "learning_rate": 1.3394366224437471e-06, + "loss": 0.0024, + "step": 51200 + }, + { + "epoch": 6.71, + "learning_rate": 1.3391715005063965e-06, + "loss": 0.0011, + "step": 51201 + }, + { + "epoch": 6.71, + "learning_rate": 1.3389064029271681e-06, + "loss": 0.0012, + "step": 51202 + }, + { + "epoch": 6.71, + "learning_rate": 1.3386413297068036e-06, + "loss": 0.0022, + "step": 51203 + }, + { + "epoch": 6.71, + "learning_rate": 1.3383762808460477e-06, + "loss": 0.002, + "step": 51204 + }, + { + "epoch": 6.71, + "learning_rate": 1.3381112563456489e-06, + "loss": 0.003, + "step": 51205 + }, + { + "epoch": 6.71, + "learning_rate": 1.337846256206351e-06, + "loss": 0.003, + "step": 51206 + }, + { + "epoch": 6.71, + "learning_rate": 1.337581280428899e-06, + "loss": 0.0023, + "step": 51207 + }, + { + "epoch": 6.71, + "learning_rate": 1.3373163290140357e-06, + "loss": 0.0016, + "step": 51208 + }, + { + "epoch": 6.71, + "learning_rate": 1.3370514019625103e-06, + "loss": 0.0034, + "step": 51209 + }, + { + "epoch": 6.71, + "learning_rate": 1.336786499275068e-06, + "loss": 0.0023, + "step": 51210 + }, + { + "epoch": 6.71, + "learning_rate": 1.3365216209524512e-06, + "loss": 0.0014, + "step": 51211 + }, + { + "epoch": 6.71, + "learning_rate": 1.3362567669954053e-06, + "loss": 0.0012, + "step": 51212 + }, + { + "epoch": 6.71, + "learning_rate": 1.335991937404677e-06, + "loss": 0.0011, + "step": 51213 + }, + { + "epoch": 6.71, + "learning_rate": 1.3357271321810107e-06, + "loss": 0.0028, + "step": 51214 + }, + { + "epoch": 6.71, + "learning_rate": 1.3354623513251508e-06, + "loss": 0.0013, + "step": 51215 + }, + { + "epoch": 6.71, + "learning_rate": 1.3351975948378394e-06, + "loss": 0.0024, + "step": 51216 + }, + { + "epoch": 6.71, + "learning_rate": 1.3349328627198244e-06, + "loss": 0.002, + "step": 51217 + }, + { + "epoch": 6.71, + "learning_rate": 1.3346681549718498e-06, + "loss": 0.002, + "step": 51218 + }, + { + "epoch": 6.71, + "learning_rate": 1.3344034715946574e-06, + "loss": 0.0013, + "step": 51219 + }, + { + "epoch": 6.71, + "learning_rate": 1.3341388125889944e-06, + "loss": 0.0017, + "step": 51220 + }, + { + "epoch": 6.71, + "learning_rate": 1.3338741779556063e-06, + "loss": 0.0014, + "step": 51221 + }, + { + "epoch": 6.71, + "learning_rate": 1.3336095676952354e-06, + "loss": 0.0016, + "step": 51222 + }, + { + "epoch": 6.71, + "learning_rate": 1.3333449818086253e-06, + "loss": 0.0017, + "step": 51223 + }, + { + "epoch": 6.71, + "learning_rate": 1.3330804202965197e-06, + "loss": 0.0017, + "step": 51224 + }, + { + "epoch": 6.71, + "learning_rate": 1.332815883159666e-06, + "loss": 0.0026, + "step": 51225 + }, + { + "epoch": 6.71, + "learning_rate": 1.3325513703988057e-06, + "loss": 0.0018, + "step": 51226 + }, + { + "epoch": 6.71, + "learning_rate": 1.3322868820146816e-06, + "loss": 0.002, + "step": 51227 + }, + { + "epoch": 6.71, + "learning_rate": 1.332022418008041e-06, + "loss": 0.0019, + "step": 51228 + }, + { + "epoch": 6.71, + "learning_rate": 1.3317579783796253e-06, + "loss": 0.0025, + "step": 51229 + }, + { + "epoch": 6.71, + "learning_rate": 1.3314935631301772e-06, + "loss": 0.0017, + "step": 51230 + }, + { + "epoch": 6.71, + "learning_rate": 1.3312291722604408e-06, + "loss": 0.0027, + "step": 51231 + }, + { + "epoch": 6.71, + "learning_rate": 1.3309648057711644e-06, + "loss": 0.0026, + "step": 51232 + }, + { + "epoch": 6.71, + "learning_rate": 1.330700463663086e-06, + "loss": 0.0022, + "step": 51233 + }, + { + "epoch": 6.71, + "learning_rate": 1.330436145936952e-06, + "loss": 0.0029, + "step": 51234 + }, + { + "epoch": 6.71, + "learning_rate": 1.3301718525935026e-06, + "loss": 0.0016, + "step": 51235 + }, + { + "epoch": 6.71, + "learning_rate": 1.3299075836334852e-06, + "loss": 0.0025, + "step": 51236 + }, + { + "epoch": 6.71, + "learning_rate": 1.3296433390576403e-06, + "loss": 0.0015, + "step": 51237 + }, + { + "epoch": 6.71, + "learning_rate": 1.3293791188667127e-06, + "loss": 0.0018, + "step": 51238 + }, + { + "epoch": 6.71, + "learning_rate": 1.329114923061442e-06, + "loss": 0.0012, + "step": 51239 + }, + { + "epoch": 6.71, + "learning_rate": 1.3288507516425764e-06, + "loss": 0.0038, + "step": 51240 + }, + { + "epoch": 6.71, + "learning_rate": 1.328586604610853e-06, + "loss": 0.0024, + "step": 51241 + }, + { + "epoch": 6.71, + "learning_rate": 1.3283224819670215e-06, + "loss": 0.0018, + "step": 51242 + }, + { + "epoch": 6.71, + "learning_rate": 1.328058383711819e-06, + "loss": 0.0011, + "step": 51243 + }, + { + "epoch": 6.71, + "learning_rate": 1.3277943098459911e-06, + "loss": 0.0012, + "step": 51244 + }, + { + "epoch": 6.71, + "learning_rate": 1.3275302603702812e-06, + "loss": 0.002, + "step": 51245 + }, + { + "epoch": 6.71, + "learning_rate": 1.3272662352854283e-06, + "loss": 0.0019, + "step": 51246 + }, + { + "epoch": 6.71, + "learning_rate": 1.3270022345921796e-06, + "loss": 0.0014, + "step": 51247 + }, + { + "epoch": 6.71, + "learning_rate": 1.3267382582912747e-06, + "loss": 0.0021, + "step": 51248 + }, + { + "epoch": 6.71, + "learning_rate": 1.3264743063834572e-06, + "loss": 0.003, + "step": 51249 + }, + { + "epoch": 6.71, + "learning_rate": 1.3262103788694658e-06, + "loss": 0.0019, + "step": 51250 + }, + { + "epoch": 6.71, + "learning_rate": 1.3259464757500483e-06, + "loss": 0.0021, + "step": 51251 + }, + { + "epoch": 6.71, + "learning_rate": 1.3256825970259424e-06, + "loss": 0.0023, + "step": 51252 + }, + { + "epoch": 6.71, + "learning_rate": 1.3254187426978949e-06, + "loss": 0.0013, + "step": 51253 + }, + { + "epoch": 6.71, + "learning_rate": 1.325154912766643e-06, + "loss": 0.0023, + "step": 51254 + }, + { + "epoch": 6.71, + "learning_rate": 1.3248911072329318e-06, + "loss": 0.0016, + "step": 51255 + }, + { + "epoch": 6.71, + "learning_rate": 1.324627326097504e-06, + "loss": 0.0012, + "step": 51256 + }, + { + "epoch": 6.71, + "learning_rate": 1.324363569361099e-06, + "loss": 0.0019, + "step": 51257 + }, + { + "epoch": 6.71, + "learning_rate": 1.3240998370244573e-06, + "loss": 0.0013, + "step": 51258 + }, + { + "epoch": 6.71, + "learning_rate": 1.323836129088325e-06, + "loss": 0.0017, + "step": 51259 + }, + { + "epoch": 6.71, + "learning_rate": 1.3235724455534416e-06, + "loss": 0.0017, + "step": 51260 + }, + { + "epoch": 6.71, + "learning_rate": 1.3233087864205464e-06, + "loss": 0.0018, + "step": 51261 + }, + { + "epoch": 6.71, + "learning_rate": 1.3230451516903853e-06, + "loss": 0.0022, + "step": 51262 + }, + { + "epoch": 6.72, + "learning_rate": 1.3227815413636957e-06, + "loss": 0.0021, + "step": 51263 + }, + { + "epoch": 6.72, + "learning_rate": 1.3225179554412238e-06, + "loss": 0.0015, + "step": 51264 + }, + { + "epoch": 6.72, + "learning_rate": 1.3222543939237064e-06, + "loss": 0.0015, + "step": 51265 + }, + { + "epoch": 6.72, + "learning_rate": 1.3219908568118856e-06, + "loss": 0.001, + "step": 51266 + }, + { + "epoch": 6.72, + "learning_rate": 1.3217273441065047e-06, + "loss": 0.0029, + "step": 51267 + }, + { + "epoch": 6.72, + "learning_rate": 1.3214638558083047e-06, + "loss": 0.0016, + "step": 51268 + }, + { + "epoch": 6.72, + "learning_rate": 1.3212003919180217e-06, + "loss": 0.0018, + "step": 51269 + }, + { + "epoch": 6.72, + "learning_rate": 1.3209369524364036e-06, + "loss": 0.0013, + "step": 51270 + }, + { + "epoch": 6.72, + "learning_rate": 1.3206735373641877e-06, + "loss": 0.0018, + "step": 51271 + }, + { + "epoch": 6.72, + "learning_rate": 1.3204101467021158e-06, + "loss": 0.0014, + "step": 51272 + }, + { + "epoch": 6.72, + "learning_rate": 1.320146780450926e-06, + "loss": 0.002, + "step": 51273 + }, + { + "epoch": 6.72, + "learning_rate": 1.3198834386113601e-06, + "loss": 0.0012, + "step": 51274 + }, + { + "epoch": 6.72, + "learning_rate": 1.319620121184163e-06, + "loss": 0.0017, + "step": 51275 + }, + { + "epoch": 6.72, + "learning_rate": 1.3193568281700718e-06, + "loss": 0.0007, + "step": 51276 + }, + { + "epoch": 6.72, + "learning_rate": 1.3190935595698251e-06, + "loss": 0.0019, + "step": 51277 + }, + { + "epoch": 6.72, + "learning_rate": 1.3188303153841675e-06, + "loss": 0.0009, + "step": 51278 + }, + { + "epoch": 6.72, + "learning_rate": 1.3185670956138375e-06, + "loss": 0.0026, + "step": 51279 + }, + { + "epoch": 6.72, + "learning_rate": 1.3183039002595744e-06, + "loss": 0.0014, + "step": 51280 + }, + { + "epoch": 6.72, + "learning_rate": 1.3180407293221175e-06, + "loss": 0.0012, + "step": 51281 + }, + { + "epoch": 6.72, + "learning_rate": 1.317777582802212e-06, + "loss": 0.002, + "step": 51282 + }, + { + "epoch": 6.72, + "learning_rate": 1.317514460700593e-06, + "loss": 0.0009, + "step": 51283 + }, + { + "epoch": 6.72, + "learning_rate": 1.3172513630180017e-06, + "loss": 0.002, + "step": 51284 + }, + { + "epoch": 6.72, + "learning_rate": 1.3169882897551778e-06, + "loss": 0.0019, + "step": 51285 + }, + { + "epoch": 6.72, + "learning_rate": 1.3167252409128638e-06, + "loss": 0.0016, + "step": 51286 + }, + { + "epoch": 6.72, + "learning_rate": 1.3164622164917972e-06, + "loss": 0.0029, + "step": 51287 + }, + { + "epoch": 6.72, + "learning_rate": 1.3161992164927196e-06, + "loss": 0.0012, + "step": 51288 + }, + { + "epoch": 6.72, + "learning_rate": 1.3159362409163668e-06, + "loss": 0.001, + "step": 51289 + }, + { + "epoch": 6.72, + "learning_rate": 1.3156732897634816e-06, + "loss": 0.0013, + "step": 51290 + }, + { + "epoch": 6.72, + "learning_rate": 1.3154103630348047e-06, + "loss": 0.0026, + "step": 51291 + }, + { + "epoch": 6.72, + "learning_rate": 1.3151474607310698e-06, + "loss": 0.0024, + "step": 51292 + }, + { + "epoch": 6.72, + "learning_rate": 1.3148845828530232e-06, + "loss": 0.0012, + "step": 51293 + }, + { + "epoch": 6.72, + "learning_rate": 1.314621729401402e-06, + "loss": 0.0013, + "step": 51294 + }, + { + "epoch": 6.72, + "learning_rate": 1.314358900376943e-06, + "loss": 0.0011, + "step": 51295 + }, + { + "epoch": 6.72, + "learning_rate": 1.314096095780384e-06, + "loss": 0.002, + "step": 51296 + }, + { + "epoch": 6.72, + "learning_rate": 1.3138333156124705e-06, + "loss": 0.0015, + "step": 51297 + }, + { + "epoch": 6.72, + "learning_rate": 1.313570559873939e-06, + "loss": 0.0014, + "step": 51298 + }, + { + "epoch": 6.72, + "learning_rate": 1.3133078285655277e-06, + "loss": 0.002, + "step": 51299 + }, + { + "epoch": 6.72, + "learning_rate": 1.3130451216879736e-06, + "loss": 0.0024, + "step": 51300 + }, + { + "epoch": 6.72, + "learning_rate": 1.3127824392420186e-06, + "loss": 0.0009, + "step": 51301 + }, + { + "epoch": 6.72, + "learning_rate": 1.3125197812284007e-06, + "loss": 0.001, + "step": 51302 + }, + { + "epoch": 6.72, + "learning_rate": 1.3122571476478597e-06, + "loss": 0.0017, + "step": 51303 + }, + { + "epoch": 6.72, + "learning_rate": 1.311994538501129e-06, + "loss": 0.0021, + "step": 51304 + }, + { + "epoch": 6.72, + "learning_rate": 1.3117319537889539e-06, + "loss": 0.0018, + "step": 51305 + }, + { + "epoch": 6.72, + "learning_rate": 1.3114693935120692e-06, + "loss": 0.0018, + "step": 51306 + }, + { + "epoch": 6.72, + "learning_rate": 1.3112068576712133e-06, + "loss": 0.0023, + "step": 51307 + }, + { + "epoch": 6.72, + "learning_rate": 1.3109443462671245e-06, + "loss": 0.0023, + "step": 51308 + }, + { + "epoch": 6.72, + "learning_rate": 1.3106818593005444e-06, + "loss": 0.0017, + "step": 51309 + }, + { + "epoch": 6.72, + "learning_rate": 1.310419396772209e-06, + "loss": 0.0015, + "step": 51310 + }, + { + "epoch": 6.72, + "learning_rate": 1.3101569586828533e-06, + "loss": 0.0027, + "step": 51311 + }, + { + "epoch": 6.72, + "learning_rate": 1.3098945450332212e-06, + "loss": 0.0025, + "step": 51312 + }, + { + "epoch": 6.72, + "learning_rate": 1.3096321558240476e-06, + "loss": 0.0012, + "step": 51313 + }, + { + "epoch": 6.72, + "learning_rate": 1.309369791056072e-06, + "loss": 0.0018, + "step": 51314 + }, + { + "epoch": 6.72, + "learning_rate": 1.3091074507300272e-06, + "loss": 0.0015, + "step": 51315 + }, + { + "epoch": 6.72, + "learning_rate": 1.308845134846658e-06, + "loss": 0.0015, + "step": 51316 + }, + { + "epoch": 6.72, + "learning_rate": 1.3085828434066995e-06, + "loss": 0.0018, + "step": 51317 + }, + { + "epoch": 6.72, + "learning_rate": 1.3083205764108852e-06, + "loss": 0.0005, + "step": 51318 + }, + { + "epoch": 6.72, + "learning_rate": 1.3080583338599574e-06, + "loss": 0.0024, + "step": 51319 + }, + { + "epoch": 6.72, + "learning_rate": 1.3077961157546549e-06, + "loss": 0.0005, + "step": 51320 + }, + { + "epoch": 6.72, + "learning_rate": 1.3075339220957128e-06, + "loss": 0.0014, + "step": 51321 + }, + { + "epoch": 6.72, + "learning_rate": 1.3072717528838685e-06, + "loss": 0.0015, + "step": 51322 + }, + { + "epoch": 6.72, + "learning_rate": 1.307009608119858e-06, + "loss": 0.0012, + "step": 51323 + }, + { + "epoch": 6.72, + "learning_rate": 1.3067474878044218e-06, + "loss": 0.0015, + "step": 51324 + }, + { + "epoch": 6.72, + "learning_rate": 1.3064853919382948e-06, + "loss": 0.0022, + "step": 51325 + }, + { + "epoch": 6.72, + "learning_rate": 1.306223320522213e-06, + "loss": 0.0011, + "step": 51326 + }, + { + "epoch": 6.72, + "learning_rate": 1.305961273556917e-06, + "loss": 0.0013, + "step": 51327 + }, + { + "epoch": 6.72, + "learning_rate": 1.3056992510431421e-06, + "loss": 0.003, + "step": 51328 + }, + { + "epoch": 6.72, + "learning_rate": 1.3054372529816227e-06, + "loss": 0.003, + "step": 51329 + }, + { + "epoch": 6.72, + "learning_rate": 1.3051752793731009e-06, + "loss": 0.0029, + "step": 51330 + }, + { + "epoch": 6.72, + "learning_rate": 1.3049133302183082e-06, + "loss": 0.0013, + "step": 51331 + }, + { + "epoch": 6.72, + "learning_rate": 1.3046514055179848e-06, + "loss": 0.0024, + "step": 51332 + }, + { + "epoch": 6.72, + "learning_rate": 1.3043895052728673e-06, + "loss": 0.0024, + "step": 51333 + }, + { + "epoch": 6.72, + "learning_rate": 1.3041276294836892e-06, + "loss": 0.0022, + "step": 51334 + }, + { + "epoch": 6.72, + "learning_rate": 1.303865778151191e-06, + "loss": 0.0012, + "step": 51335 + }, + { + "epoch": 6.72, + "learning_rate": 1.3036039512761066e-06, + "loss": 0.0015, + "step": 51336 + }, + { + "epoch": 6.72, + "learning_rate": 1.3033421488591736e-06, + "loss": 0.0018, + "step": 51337 + }, + { + "epoch": 6.72, + "learning_rate": 1.3030803709011253e-06, + "loss": 0.0011, + "step": 51338 + }, + { + "epoch": 6.73, + "learning_rate": 1.3028186174027014e-06, + "loss": 0.0014, + "step": 51339 + }, + { + "epoch": 6.73, + "learning_rate": 1.3025568883646379e-06, + "loss": 0.0024, + "step": 51340 + }, + { + "epoch": 6.73, + "learning_rate": 1.3022951837876697e-06, + "loss": 0.0016, + "step": 51341 + }, + { + "epoch": 6.73, + "learning_rate": 1.3020335036725317e-06, + "loss": 0.0009, + "step": 51342 + }, + { + "epoch": 6.73, + "learning_rate": 1.3017718480199636e-06, + "loss": 0.0016, + "step": 51343 + }, + { + "epoch": 6.73, + "learning_rate": 1.301510216830698e-06, + "loss": 0.0014, + "step": 51344 + }, + { + "epoch": 6.73, + "learning_rate": 1.3012486101054722e-06, + "loss": 0.0025, + "step": 51345 + }, + { + "epoch": 6.73, + "learning_rate": 1.30098702784502e-06, + "loss": 0.0021, + "step": 51346 + }, + { + "epoch": 6.73, + "learning_rate": 1.3007254700500804e-06, + "loss": 0.001, + "step": 51347 + }, + { + "epoch": 6.73, + "learning_rate": 1.3004639367213868e-06, + "loss": 0.0018, + "step": 51348 + }, + { + "epoch": 6.73, + "learning_rate": 1.3002024278596725e-06, + "loss": 0.0018, + "step": 51349 + }, + { + "epoch": 6.73, + "learning_rate": 1.2999409434656763e-06, + "loss": 0.0015, + "step": 51350 + }, + { + "epoch": 6.73, + "learning_rate": 1.299679483540135e-06, + "loss": 0.0015, + "step": 51351 + }, + { + "epoch": 6.73, + "learning_rate": 1.2994180480837815e-06, + "loss": 0.0021, + "step": 51352 + }, + { + "epoch": 6.73, + "learning_rate": 1.2991566370973517e-06, + "loss": 0.0018, + "step": 51353 + }, + { + "epoch": 6.73, + "learning_rate": 1.2988952505815778e-06, + "loss": 0.0014, + "step": 51354 + }, + { + "epoch": 6.73, + "learning_rate": 1.2986338885372008e-06, + "loss": 0.0012, + "step": 51355 + }, + { + "epoch": 6.73, + "learning_rate": 1.2983725509649515e-06, + "loss": 0.0021, + "step": 51356 + }, + { + "epoch": 6.73, + "learning_rate": 1.2981112378655647e-06, + "loss": 0.0027, + "step": 51357 + }, + { + "epoch": 6.73, + "learning_rate": 1.2978499492397778e-06, + "loss": 0.0013, + "step": 51358 + }, + { + "epoch": 6.73, + "learning_rate": 1.2975886850883256e-06, + "loss": 0.0018, + "step": 51359 + }, + { + "epoch": 6.73, + "learning_rate": 1.2973274454119411e-06, + "loss": 0.0019, + "step": 51360 + }, + { + "epoch": 6.73, + "learning_rate": 1.2970662302113556e-06, + "loss": 0.0022, + "step": 51361 + }, + { + "epoch": 6.73, + "learning_rate": 1.2968050394873122e-06, + "loss": 0.001, + "step": 51362 + }, + { + "epoch": 6.73, + "learning_rate": 1.2965438732405421e-06, + "loss": 0.0021, + "step": 51363 + }, + { + "epoch": 6.73, + "learning_rate": 1.2962827314717774e-06, + "loss": 0.0025, + "step": 51364 + }, + { + "epoch": 6.73, + "learning_rate": 1.2960216141817527e-06, + "loss": 0.0016, + "step": 51365 + }, + { + "epoch": 6.73, + "learning_rate": 1.2957605213712054e-06, + "loss": 0.002, + "step": 51366 + }, + { + "epoch": 6.73, + "learning_rate": 1.2954994530408682e-06, + "loss": 0.0018, + "step": 51367 + }, + { + "epoch": 6.73, + "learning_rate": 1.295238409191476e-06, + "loss": 0.002, + "step": 51368 + }, + { + "epoch": 6.73, + "learning_rate": 1.2949773898237594e-06, + "loss": 0.0013, + "step": 51369 + }, + { + "epoch": 6.73, + "learning_rate": 1.2947163949384566e-06, + "loss": 0.0032, + "step": 51370 + }, + { + "epoch": 6.73, + "learning_rate": 1.2944554245363016e-06, + "loss": 0.0016, + "step": 51371 + }, + { + "epoch": 6.73, + "learning_rate": 1.294194478618025e-06, + "loss": 0.0013, + "step": 51372 + }, + { + "epoch": 6.73, + "learning_rate": 1.2939335571843625e-06, + "loss": 0.0036, + "step": 51373 + }, + { + "epoch": 6.73, + "learning_rate": 1.2936726602360506e-06, + "loss": 0.001, + "step": 51374 + }, + { + "epoch": 6.73, + "learning_rate": 1.2934117877738195e-06, + "loss": 0.0022, + "step": 51375 + }, + { + "epoch": 6.73, + "learning_rate": 1.2931509397984033e-06, + "loss": 0.0008, + "step": 51376 + }, + { + "epoch": 6.73, + "learning_rate": 1.2928901163105378e-06, + "loss": 0.0014, + "step": 51377 + }, + { + "epoch": 6.73, + "learning_rate": 1.292629317310956e-06, + "loss": 0.0023, + "step": 51378 + }, + { + "epoch": 6.73, + "learning_rate": 1.2923685428003895e-06, + "loss": 0.0012, + "step": 51379 + }, + { + "epoch": 6.73, + "learning_rate": 1.2921077927795712e-06, + "loss": 0.0013, + "step": 51380 + }, + { + "epoch": 6.73, + "learning_rate": 1.2918470672492388e-06, + "loss": 0.0012, + "step": 51381 + }, + { + "epoch": 6.73, + "learning_rate": 1.2915863662101213e-06, + "loss": 0.0026, + "step": 51382 + }, + { + "epoch": 6.73, + "learning_rate": 1.2913256896629523e-06, + "loss": 0.0019, + "step": 51383 + }, + { + "epoch": 6.73, + "learning_rate": 1.2910650376084654e-06, + "loss": 0.0013, + "step": 51384 + }, + { + "epoch": 6.73, + "learning_rate": 1.290804410047397e-06, + "loss": 0.0018, + "step": 51385 + }, + { + "epoch": 6.73, + "learning_rate": 1.2905438069804778e-06, + "loss": 0.0013, + "step": 51386 + }, + { + "epoch": 6.73, + "learning_rate": 1.2902832284084398e-06, + "loss": 0.0025, + "step": 51387 + }, + { + "epoch": 6.73, + "learning_rate": 1.2900226743320143e-06, + "loss": 0.0039, + "step": 51388 + }, + { + "epoch": 6.73, + "learning_rate": 1.2897621447519383e-06, + "loss": 0.0024, + "step": 51389 + }, + { + "epoch": 6.73, + "learning_rate": 1.289501639668942e-06, + "loss": 0.0011, + "step": 51390 + }, + { + "epoch": 6.73, + "learning_rate": 1.2892411590837572e-06, + "loss": 0.0023, + "step": 51391 + }, + { + "epoch": 6.73, + "learning_rate": 1.288980702997119e-06, + "loss": 0.0018, + "step": 51392 + }, + { + "epoch": 6.73, + "learning_rate": 1.2887202714097592e-06, + "loss": 0.0016, + "step": 51393 + }, + { + "epoch": 6.73, + "learning_rate": 1.2884598643224078e-06, + "loss": 0.0022, + "step": 51394 + }, + { + "epoch": 6.73, + "learning_rate": 1.2881994817358011e-06, + "loss": 0.0015, + "step": 51395 + }, + { + "epoch": 6.73, + "learning_rate": 1.2879391236506667e-06, + "loss": 0.0014, + "step": 51396 + }, + { + "epoch": 6.73, + "learning_rate": 1.2876787900677424e-06, + "loss": 0.0016, + "step": 51397 + }, + { + "epoch": 6.73, + "learning_rate": 1.2874184809877576e-06, + "loss": 0.0014, + "step": 51398 + }, + { + "epoch": 6.73, + "learning_rate": 1.2871581964114421e-06, + "loss": 0.0015, + "step": 51399 + }, + { + "epoch": 6.73, + "learning_rate": 1.2868979363395317e-06, + "loss": 0.0018, + "step": 51400 + }, + { + "epoch": 6.73, + "learning_rate": 1.2866377007727571e-06, + "loss": 0.0021, + "step": 51401 + }, + { + "epoch": 6.73, + "learning_rate": 1.2863774897118508e-06, + "loss": 0.0026, + "step": 51402 + }, + { + "epoch": 6.73, + "learning_rate": 1.2861173031575403e-06, + "loss": 0.0018, + "step": 51403 + }, + { + "epoch": 6.73, + "learning_rate": 1.2858571411105635e-06, + "loss": 0.0035, + "step": 51404 + }, + { + "epoch": 6.73, + "learning_rate": 1.285597003571648e-06, + "loss": 0.0013, + "step": 51405 + }, + { + "epoch": 6.73, + "learning_rate": 1.2853368905415287e-06, + "loss": 0.0022, + "step": 51406 + }, + { + "epoch": 6.73, + "learning_rate": 1.2850768020209326e-06, + "loss": 0.0016, + "step": 51407 + }, + { + "epoch": 6.73, + "learning_rate": 1.284816738010597e-06, + "loss": 0.0012, + "step": 51408 + }, + { + "epoch": 6.73, + "learning_rate": 1.284556698511249e-06, + "loss": 0.002, + "step": 51409 + }, + { + "epoch": 6.73, + "learning_rate": 1.2842966835236214e-06, + "loss": 0.0015, + "step": 51410 + }, + { + "epoch": 6.73, + "learning_rate": 1.2840366930484438e-06, + "loss": 0.0023, + "step": 51411 + }, + { + "epoch": 6.73, + "learning_rate": 1.283776727086451e-06, + "loss": 0.0029, + "step": 51412 + }, + { + "epoch": 6.73, + "learning_rate": 1.2835167856383713e-06, + "loss": 0.0009, + "step": 51413 + }, + { + "epoch": 6.73, + "learning_rate": 1.2832568687049352e-06, + "loss": 0.0023, + "step": 51414 + }, + { + "epoch": 6.74, + "learning_rate": 1.2829969762868766e-06, + "loss": 0.0017, + "step": 51415 + }, + { + "epoch": 6.74, + "learning_rate": 1.2827371083849226e-06, + "loss": 0.0018, + "step": 51416 + }, + { + "epoch": 6.74, + "learning_rate": 1.2824772649998086e-06, + "loss": 0.0029, + "step": 51417 + }, + { + "epoch": 6.74, + "learning_rate": 1.2822174461322622e-06, + "loss": 0.0021, + "step": 51418 + }, + { + "epoch": 6.74, + "learning_rate": 1.2819576517830135e-06, + "loss": 0.0018, + "step": 51419 + }, + { + "epoch": 6.74, + "learning_rate": 1.281697881952797e-06, + "loss": 0.0017, + "step": 51420 + }, + { + "epoch": 6.74, + "learning_rate": 1.2814381366423413e-06, + "loss": 0.0026, + "step": 51421 + }, + { + "epoch": 6.74, + "learning_rate": 1.2811784158523733e-06, + "loss": 0.0007, + "step": 51422 + }, + { + "epoch": 6.74, + "learning_rate": 1.2809187195836293e-06, + "loss": 0.0013, + "step": 51423 + }, + { + "epoch": 6.74, + "learning_rate": 1.2806590478368375e-06, + "loss": 0.0014, + "step": 51424 + }, + { + "epoch": 6.74, + "learning_rate": 1.2803994006127273e-06, + "loss": 0.0023, + "step": 51425 + }, + { + "epoch": 6.74, + "learning_rate": 1.2801397779120273e-06, + "loss": 0.002, + "step": 51426 + }, + { + "epoch": 6.74, + "learning_rate": 1.2798801797354698e-06, + "loss": 0.0039, + "step": 51427 + }, + { + "epoch": 6.74, + "learning_rate": 1.2796206060837879e-06, + "loss": 0.0027, + "step": 51428 + }, + { + "epoch": 6.74, + "learning_rate": 1.2793610569577075e-06, + "loss": 0.0011, + "step": 51429 + }, + { + "epoch": 6.74, + "learning_rate": 1.279101532357958e-06, + "loss": 0.001, + "step": 51430 + }, + { + "epoch": 6.74, + "learning_rate": 1.2788420322852734e-06, + "loss": 0.0018, + "step": 51431 + }, + { + "epoch": 6.74, + "learning_rate": 1.2785825567403809e-06, + "loss": 0.0015, + "step": 51432 + }, + { + "epoch": 6.74, + "learning_rate": 1.278323105724011e-06, + "loss": 0.0019, + "step": 51433 + }, + { + "epoch": 6.74, + "learning_rate": 1.2780636792368905e-06, + "loss": 0.002, + "step": 51434 + }, + { + "epoch": 6.74, + "learning_rate": 1.2778042772797528e-06, + "loss": 0.0029, + "step": 51435 + }, + { + "epoch": 6.74, + "learning_rate": 1.2775448998533268e-06, + "loss": 0.0013, + "step": 51436 + }, + { + "epoch": 6.74, + "learning_rate": 1.277285546958339e-06, + "loss": 0.0022, + "step": 51437 + }, + { + "epoch": 6.74, + "learning_rate": 1.2770262185955218e-06, + "loss": 0.0018, + "step": 51438 + }, + { + "epoch": 6.74, + "learning_rate": 1.2767669147656048e-06, + "loss": 0.0017, + "step": 51439 + }, + { + "epoch": 6.74, + "learning_rate": 1.2765076354693173e-06, + "loss": 0.0019, + "step": 51440 + }, + { + "epoch": 6.74, + "learning_rate": 1.2762483807073844e-06, + "loss": 0.0013, + "step": 51441 + }, + { + "epoch": 6.74, + "learning_rate": 1.275989150480541e-06, + "loss": 0.0022, + "step": 51442 + }, + { + "epoch": 6.74, + "learning_rate": 1.2757299447895144e-06, + "loss": 0.0018, + "step": 51443 + }, + { + "epoch": 6.74, + "learning_rate": 1.2754707636350317e-06, + "loss": 0.0026, + "step": 51444 + }, + { + "epoch": 6.74, + "learning_rate": 1.275211607017821e-06, + "loss": 0.0015, + "step": 51445 + }, + { + "epoch": 6.74, + "learning_rate": 1.2749524749386144e-06, + "loss": 0.0018, + "step": 51446 + }, + { + "epoch": 6.74, + "learning_rate": 1.2746933673981409e-06, + "loss": 0.0012, + "step": 51447 + }, + { + "epoch": 6.74, + "learning_rate": 1.2744342843971247e-06, + "loss": 0.0015, + "step": 51448 + }, + { + "epoch": 6.74, + "learning_rate": 1.2741752259362971e-06, + "loss": 0.0018, + "step": 51449 + }, + { + "epoch": 6.74, + "learning_rate": 1.2739161920163889e-06, + "loss": 0.002, + "step": 51450 + }, + { + "epoch": 6.74, + "learning_rate": 1.2736571826381272e-06, + "loss": 0.0024, + "step": 51451 + }, + { + "epoch": 6.74, + "learning_rate": 1.2733981978022402e-06, + "loss": 0.0017, + "step": 51452 + }, + { + "epoch": 6.74, + "learning_rate": 1.2731392375094542e-06, + "loss": 0.0018, + "step": 51453 + }, + { + "epoch": 6.74, + "learning_rate": 1.2728803017605006e-06, + "loss": 0.001, + "step": 51454 + }, + { + "epoch": 6.74, + "learning_rate": 1.272621390556107e-06, + "loss": 0.0016, + "step": 51455 + }, + { + "epoch": 6.74, + "learning_rate": 1.272362503896999e-06, + "loss": 0.0031, + "step": 51456 + }, + { + "epoch": 6.74, + "learning_rate": 1.2721036417839083e-06, + "loss": 0.002, + "step": 51457 + }, + { + "epoch": 6.74, + "learning_rate": 1.2718448042175612e-06, + "loss": 0.0011, + "step": 51458 + }, + { + "epoch": 6.74, + "learning_rate": 1.2715859911986861e-06, + "loss": 0.0011, + "step": 51459 + }, + { + "epoch": 6.74, + "learning_rate": 1.2713272027280087e-06, + "loss": 0.0024, + "step": 51460 + }, + { + "epoch": 6.74, + "learning_rate": 1.2710684388062577e-06, + "loss": 0.0013, + "step": 51461 + }, + { + "epoch": 6.74, + "learning_rate": 1.2708096994341646e-06, + "loss": 0.003, + "step": 51462 + }, + { + "epoch": 6.74, + "learning_rate": 1.2705509846124552e-06, + "loss": 0.0008, + "step": 51463 + }, + { + "epoch": 6.74, + "learning_rate": 1.2702922943418527e-06, + "loss": 0.0016, + "step": 51464 + }, + { + "epoch": 6.74, + "learning_rate": 1.270033628623092e-06, + "loss": 0.0018, + "step": 51465 + }, + { + "epoch": 6.74, + "learning_rate": 1.2697749874568954e-06, + "loss": 0.0017, + "step": 51466 + }, + { + "epoch": 6.74, + "learning_rate": 1.269516370843993e-06, + "loss": 0.0022, + "step": 51467 + }, + { + "epoch": 6.74, + "learning_rate": 1.2692577787851079e-06, + "loss": 0.0016, + "step": 51468 + }, + { + "epoch": 6.74, + "learning_rate": 1.2689992112809723e-06, + "loss": 0.0023, + "step": 51469 + }, + { + "epoch": 6.74, + "learning_rate": 1.26874066833231e-06, + "loss": 0.0017, + "step": 51470 + }, + { + "epoch": 6.74, + "learning_rate": 1.2684821499398514e-06, + "loss": 0.0018, + "step": 51471 + }, + { + "epoch": 6.74, + "learning_rate": 1.2682236561043193e-06, + "loss": 0.0021, + "step": 51472 + }, + { + "epoch": 6.74, + "learning_rate": 1.2679651868264453e-06, + "loss": 0.0012, + "step": 51473 + }, + { + "epoch": 6.74, + "learning_rate": 1.2677067421069545e-06, + "loss": 0.003, + "step": 51474 + }, + { + "epoch": 6.74, + "learning_rate": 1.267448321946574e-06, + "loss": 0.0009, + "step": 51475 + }, + { + "epoch": 6.74, + "learning_rate": 1.2671899263460274e-06, + "loss": 0.0018, + "step": 51476 + }, + { + "epoch": 6.74, + "learning_rate": 1.2669315553060458e-06, + "loss": 0.0022, + "step": 51477 + }, + { + "epoch": 6.74, + "learning_rate": 1.2666732088273548e-06, + "loss": 0.0011, + "step": 51478 + }, + { + "epoch": 6.74, + "learning_rate": 1.2664148869106784e-06, + "loss": 0.0022, + "step": 51479 + }, + { + "epoch": 6.74, + "learning_rate": 1.2661565895567463e-06, + "loss": 0.0019, + "step": 51480 + }, + { + "epoch": 6.74, + "learning_rate": 1.265898316766282e-06, + "loss": 0.0013, + "step": 51481 + }, + { + "epoch": 6.74, + "learning_rate": 1.265640068540016e-06, + "loss": 0.0015, + "step": 51482 + }, + { + "epoch": 6.74, + "learning_rate": 1.2653818448786715e-06, + "loss": 0.0018, + "step": 51483 + }, + { + "epoch": 6.74, + "learning_rate": 1.265123645782974e-06, + "loss": 0.0017, + "step": 51484 + }, + { + "epoch": 6.74, + "learning_rate": 1.2648654712536522e-06, + "loss": 0.0008, + "step": 51485 + }, + { + "epoch": 6.74, + "learning_rate": 1.2646073212914322e-06, + "loss": 0.0034, + "step": 51486 + }, + { + "epoch": 6.74, + "learning_rate": 1.2643491958970366e-06, + "loss": 0.0015, + "step": 51487 + }, + { + "epoch": 6.74, + "learning_rate": 1.264091095071196e-06, + "loss": 0.0017, + "step": 51488 + }, + { + "epoch": 6.74, + "learning_rate": 1.263833018814633e-06, + "loss": 0.0019, + "step": 51489 + }, + { + "epoch": 6.74, + "learning_rate": 1.263574967128075e-06, + "loss": 0.0024, + "step": 51490 + }, + { + "epoch": 6.74, + "learning_rate": 1.2633169400122448e-06, + "loss": 0.003, + "step": 51491 + }, + { + "epoch": 6.75, + "learning_rate": 1.2630589374678703e-06, + "loss": 0.0022, + "step": 51492 + }, + { + "epoch": 6.75, + "learning_rate": 1.2628009594956803e-06, + "loss": 0.0022, + "step": 51493 + }, + { + "epoch": 6.75, + "learning_rate": 1.2625430060963971e-06, + "loss": 0.0016, + "step": 51494 + }, + { + "epoch": 6.75, + "learning_rate": 1.262285077270744e-06, + "loss": 0.0034, + "step": 51495 + }, + { + "epoch": 6.75, + "learning_rate": 1.26202717301945e-06, + "loss": 0.0036, + "step": 51496 + }, + { + "epoch": 6.75, + "learning_rate": 1.2617692933432402e-06, + "loss": 0.0038, + "step": 51497 + }, + { + "epoch": 6.75, + "learning_rate": 1.2615114382428373e-06, + "loss": 0.0014, + "step": 51498 + }, + { + "epoch": 6.75, + "learning_rate": 1.2612536077189697e-06, + "loss": 0.0024, + "step": 51499 + }, + { + "epoch": 6.75, + "learning_rate": 1.260995801772361e-06, + "loss": 0.0021, + "step": 51500 + }, + { + "epoch": 6.75, + "learning_rate": 1.2607380204037356e-06, + "loss": 0.0011, + "step": 51501 + }, + { + "epoch": 6.75, + "learning_rate": 1.260480263613818e-06, + "loss": 0.0013, + "step": 51502 + }, + { + "epoch": 6.75, + "learning_rate": 1.2602225314033335e-06, + "loss": 0.0015, + "step": 51503 + }, + { + "epoch": 6.75, + "learning_rate": 1.25996482377301e-06, + "loss": 0.003, + "step": 51504 + }, + { + "epoch": 6.75, + "learning_rate": 1.2597071407235705e-06, + "loss": 0.0019, + "step": 51505 + }, + { + "epoch": 6.75, + "learning_rate": 1.259449482255737e-06, + "loss": 0.0021, + "step": 51506 + }, + { + "epoch": 6.75, + "learning_rate": 1.2591918483702381e-06, + "loss": 0.0026, + "step": 51507 + }, + { + "epoch": 6.75, + "learning_rate": 1.2589342390677973e-06, + "loss": 0.002, + "step": 51508 + }, + { + "epoch": 6.75, + "learning_rate": 1.258676654349137e-06, + "loss": 0.0016, + "step": 51509 + }, + { + "epoch": 6.75, + "learning_rate": 1.2584190942149822e-06, + "loss": 0.0017, + "step": 51510 + }, + { + "epoch": 6.75, + "learning_rate": 1.25816155866606e-06, + "loss": 0.0014, + "step": 51511 + }, + { + "epoch": 6.75, + "learning_rate": 1.2579040477030935e-06, + "loss": 0.0017, + "step": 51512 + }, + { + "epoch": 6.75, + "learning_rate": 1.257646561326803e-06, + "loss": 0.0015, + "step": 51513 + }, + { + "epoch": 6.75, + "learning_rate": 1.2573890995379167e-06, + "loss": 0.0022, + "step": 51514 + }, + { + "epoch": 6.75, + "learning_rate": 1.2571316623371599e-06, + "loss": 0.0015, + "step": 51515 + }, + { + "epoch": 6.75, + "learning_rate": 1.256874249725255e-06, + "loss": 0.0017, + "step": 51516 + }, + { + "epoch": 6.75, + "learning_rate": 1.2566168617029251e-06, + "loss": 0.001, + "step": 51517 + }, + { + "epoch": 6.75, + "learning_rate": 1.2563594982708938e-06, + "loss": 0.0017, + "step": 51518 + }, + { + "epoch": 6.75, + "learning_rate": 1.2561021594298872e-06, + "loss": 0.0017, + "step": 51519 + }, + { + "epoch": 6.75, + "learning_rate": 1.255844845180627e-06, + "loss": 0.0016, + "step": 51520 + }, + { + "epoch": 6.75, + "learning_rate": 1.2555875555238363e-06, + "loss": 0.0014, + "step": 51521 + }, + { + "epoch": 6.75, + "learning_rate": 1.2553302904602415e-06, + "loss": 0.0015, + "step": 51522 + }, + { + "epoch": 6.75, + "learning_rate": 1.255073049990565e-06, + "loss": 0.0013, + "step": 51523 + }, + { + "epoch": 6.75, + "learning_rate": 1.2548158341155292e-06, + "loss": 0.0011, + "step": 51524 + }, + { + "epoch": 6.75, + "learning_rate": 1.2545586428358558e-06, + "loss": 0.0012, + "step": 51525 + }, + { + "epoch": 6.75, + "learning_rate": 1.254301476152272e-06, + "loss": 0.0017, + "step": 51526 + }, + { + "epoch": 6.75, + "learning_rate": 1.2540443340655007e-06, + "loss": 0.002, + "step": 51527 + }, + { + "epoch": 6.75, + "learning_rate": 1.2537872165762632e-06, + "loss": 0.0026, + "step": 51528 + }, + { + "epoch": 6.75, + "learning_rate": 1.2535301236852825e-06, + "loss": 0.0018, + "step": 51529 + }, + { + "epoch": 6.75, + "learning_rate": 1.2532730553932847e-06, + "loss": 0.0023, + "step": 51530 + }, + { + "epoch": 6.75, + "learning_rate": 1.2530160117009903e-06, + "loss": 0.0016, + "step": 51531 + }, + { + "epoch": 6.75, + "learning_rate": 1.2527589926091221e-06, + "loss": 0.0013, + "step": 51532 + }, + { + "epoch": 6.75, + "learning_rate": 1.2525019981184018e-06, + "loss": 0.0022, + "step": 51533 + }, + { + "epoch": 6.75, + "learning_rate": 1.2522450282295562e-06, + "loss": 0.0012, + "step": 51534 + }, + { + "epoch": 6.75, + "learning_rate": 1.251988082943305e-06, + "loss": 0.0009, + "step": 51535 + }, + { + "epoch": 6.75, + "learning_rate": 1.251731162260369e-06, + "loss": 0.0015, + "step": 51536 + }, + { + "epoch": 6.75, + "learning_rate": 1.2514742661814739e-06, + "loss": 0.0017, + "step": 51537 + }, + { + "epoch": 6.75, + "learning_rate": 1.2512173947073436e-06, + "loss": 0.0024, + "step": 51538 + }, + { + "epoch": 6.75, + "learning_rate": 1.2509605478386989e-06, + "loss": 0.0018, + "step": 51539 + }, + { + "epoch": 6.75, + "learning_rate": 1.250703725576261e-06, + "loss": 0.0009, + "step": 51540 + }, + { + "epoch": 6.75, + "learning_rate": 1.2504469279207509e-06, + "loss": 0.0018, + "step": 51541 + }, + { + "epoch": 6.75, + "learning_rate": 1.2501901548728956e-06, + "loss": 0.0013, + "step": 51542 + }, + { + "epoch": 6.75, + "learning_rate": 1.2499334064334134e-06, + "loss": 0.002, + "step": 51543 + }, + { + "epoch": 6.75, + "learning_rate": 1.249676682603027e-06, + "loss": 0.0014, + "step": 51544 + }, + { + "epoch": 6.75, + "learning_rate": 1.2494199833824593e-06, + "loss": 0.0013, + "step": 51545 + }, + { + "epoch": 6.75, + "learning_rate": 1.249163308772433e-06, + "loss": 0.0007, + "step": 51546 + }, + { + "epoch": 6.75, + "learning_rate": 1.2489066587736664e-06, + "loss": 0.0015, + "step": 51547 + }, + { + "epoch": 6.75, + "learning_rate": 1.2486500333868845e-06, + "loss": 0.0012, + "step": 51548 + }, + { + "epoch": 6.75, + "learning_rate": 1.24839343261281e-06, + "loss": 0.0008, + "step": 51549 + }, + { + "epoch": 6.75, + "learning_rate": 1.2481368564521623e-06, + "loss": 0.0023, + "step": 51550 + }, + { + "epoch": 6.75, + "learning_rate": 1.2478803049056642e-06, + "loss": 0.0018, + "step": 51551 + }, + { + "epoch": 6.75, + "learning_rate": 1.2476237779740353e-06, + "loss": 0.0019, + "step": 51552 + }, + { + "epoch": 6.75, + "learning_rate": 1.247367275657999e-06, + "loss": 0.0016, + "step": 51553 + }, + { + "epoch": 6.75, + "learning_rate": 1.2471107979582776e-06, + "loss": 0.0032, + "step": 51554 + }, + { + "epoch": 6.75, + "learning_rate": 1.2468543448755899e-06, + "loss": 0.0007, + "step": 51555 + }, + { + "epoch": 6.75, + "learning_rate": 1.2465979164106567e-06, + "loss": 0.0017, + "step": 51556 + }, + { + "epoch": 6.75, + "learning_rate": 1.2463415125642032e-06, + "loss": 0.0028, + "step": 51557 + }, + { + "epoch": 6.75, + "learning_rate": 1.246085133336945e-06, + "loss": 0.0016, + "step": 51558 + }, + { + "epoch": 6.75, + "learning_rate": 1.2458287787296086e-06, + "loss": 0.001, + "step": 51559 + }, + { + "epoch": 6.75, + "learning_rate": 1.24557244874291e-06, + "loss": 0.0024, + "step": 51560 + }, + { + "epoch": 6.75, + "learning_rate": 1.245316143377575e-06, + "loss": 0.0013, + "step": 51561 + }, + { + "epoch": 6.75, + "learning_rate": 1.2450598626343235e-06, + "loss": 0.0013, + "step": 51562 + }, + { + "epoch": 6.75, + "learning_rate": 1.2448036065138713e-06, + "loss": 0.0029, + "step": 51563 + }, + { + "epoch": 6.75, + "learning_rate": 1.2445473750169457e-06, + "loss": 0.0012, + "step": 51564 + }, + { + "epoch": 6.75, + "learning_rate": 1.2442911681442637e-06, + "loss": 0.0018, + "step": 51565 + }, + { + "epoch": 6.75, + "learning_rate": 1.244034985896546e-06, + "loss": 0.0011, + "step": 51566 + }, + { + "epoch": 6.75, + "learning_rate": 1.2437788282745123e-06, + "loss": 0.0012, + "step": 51567 + }, + { + "epoch": 6.76, + "learning_rate": 1.243522695278887e-06, + "loss": 0.0037, + "step": 51568 + }, + { + "epoch": 6.76, + "learning_rate": 1.2432665869103844e-06, + "loss": 0.0023, + "step": 51569 + }, + { + "epoch": 6.76, + "learning_rate": 1.2430105031697315e-06, + "loss": 0.001, + "step": 51570 + }, + { + "epoch": 6.76, + "learning_rate": 1.2427544440576422e-06, + "loss": 0.0015, + "step": 51571 + }, + { + "epoch": 6.76, + "learning_rate": 1.2424984095748427e-06, + "loss": 0.0011, + "step": 51572 + }, + { + "epoch": 6.76, + "learning_rate": 1.242242399722049e-06, + "loss": 0.0024, + "step": 51573 + }, + { + "epoch": 6.76, + "learning_rate": 1.2419864144999828e-06, + "loss": 0.0039, + "step": 51574 + }, + { + "epoch": 6.76, + "learning_rate": 1.2417304539093612e-06, + "loss": 0.0016, + "step": 51575 + }, + { + "epoch": 6.76, + "learning_rate": 1.2414745179509091e-06, + "loss": 0.0024, + "step": 51576 + }, + { + "epoch": 6.76, + "learning_rate": 1.2412186066253428e-06, + "loss": 0.0011, + "step": 51577 + }, + { + "epoch": 6.76, + "learning_rate": 1.2409627199333808e-06, + "loss": 0.0013, + "step": 51578 + }, + { + "epoch": 6.76, + "learning_rate": 1.2407068578757463e-06, + "loss": 0.0016, + "step": 51579 + }, + { + "epoch": 6.76, + "learning_rate": 1.240451020453155e-06, + "loss": 0.0012, + "step": 51580 + }, + { + "epoch": 6.76, + "learning_rate": 1.2401952076663315e-06, + "loss": 0.001, + "step": 51581 + }, + { + "epoch": 6.76, + "learning_rate": 1.2399394195159931e-06, + "loss": 0.0009, + "step": 51582 + }, + { + "epoch": 6.76, + "learning_rate": 1.2396836560028557e-06, + "loss": 0.0026, + "step": 51583 + }, + { + "epoch": 6.76, + "learning_rate": 1.2394279171276435e-06, + "loss": 0.0022, + "step": 51584 + }, + { + "epoch": 6.76, + "learning_rate": 1.2391722028910735e-06, + "loss": 0.0024, + "step": 51585 + }, + { + "epoch": 6.76, + "learning_rate": 1.238916513293864e-06, + "loss": 0.0018, + "step": 51586 + }, + { + "epoch": 6.76, + "learning_rate": 1.2386608483367368e-06, + "loss": 0.0015, + "step": 51587 + }, + { + "epoch": 6.76, + "learning_rate": 1.238405208020409e-06, + "loss": 0.0035, + "step": 51588 + }, + { + "epoch": 6.76, + "learning_rate": 1.238149592345601e-06, + "loss": 0.0011, + "step": 51589 + }, + { + "epoch": 6.76, + "learning_rate": 1.237894001313028e-06, + "loss": 0.0013, + "step": 51590 + }, + { + "epoch": 6.76, + "learning_rate": 1.2376384349234117e-06, + "loss": 0.0015, + "step": 51591 + }, + { + "epoch": 6.76, + "learning_rate": 1.2373828931774723e-06, + "loss": 0.0009, + "step": 51592 + }, + { + "epoch": 6.76, + "learning_rate": 1.2371273760759272e-06, + "loss": 0.0013, + "step": 51593 + }, + { + "epoch": 6.76, + "learning_rate": 1.2368718836194927e-06, + "loss": 0.0014, + "step": 51594 + }, + { + "epoch": 6.76, + "learning_rate": 1.2366164158088911e-06, + "loss": 0.0016, + "step": 51595 + }, + { + "epoch": 6.76, + "learning_rate": 1.236360972644839e-06, + "loss": 0.0018, + "step": 51596 + }, + { + "epoch": 6.76, + "learning_rate": 1.2361055541280554e-06, + "loss": 0.0015, + "step": 51597 + }, + { + "epoch": 6.76, + "learning_rate": 1.2358501602592566e-06, + "loss": 0.0009, + "step": 51598 + }, + { + "epoch": 6.76, + "learning_rate": 1.2355947910391642e-06, + "loss": 0.0015, + "step": 51599 + }, + { + "epoch": 6.76, + "learning_rate": 1.2353394464684943e-06, + "loss": 0.0062, + "step": 51600 + }, + { + "epoch": 6.76, + "learning_rate": 1.235084126547963e-06, + "loss": 0.0012, + "step": 51601 + }, + { + "epoch": 6.76, + "learning_rate": 1.234828831278292e-06, + "loss": 0.0039, + "step": 51602 + }, + { + "epoch": 6.76, + "learning_rate": 1.2345735606601993e-06, + "loss": 0.0018, + "step": 51603 + }, + { + "epoch": 6.76, + "learning_rate": 1.2343183146944026e-06, + "loss": 0.0014, + "step": 51604 + }, + { + "epoch": 6.76, + "learning_rate": 1.2340630933816178e-06, + "loss": 0.0024, + "step": 51605 + }, + { + "epoch": 6.76, + "learning_rate": 1.2338078967225619e-06, + "loss": 0.0012, + "step": 51606 + }, + { + "epoch": 6.76, + "learning_rate": 1.233552724717958e-06, + "loss": 0.002, + "step": 51607 + }, + { + "epoch": 6.76, + "learning_rate": 1.2332975773685185e-06, + "loss": 0.0017, + "step": 51608 + }, + { + "epoch": 6.76, + "learning_rate": 1.2330424546749609e-06, + "loss": 0.0022, + "step": 51609 + }, + { + "epoch": 6.76, + "learning_rate": 1.2327873566380066e-06, + "loss": 0.0021, + "step": 51610 + }, + { + "epoch": 6.76, + "learning_rate": 1.232532283258372e-06, + "loss": 0.0014, + "step": 51611 + }, + { + "epoch": 6.76, + "learning_rate": 1.2322772345367705e-06, + "loss": 0.0021, + "step": 51612 + }, + { + "epoch": 6.76, + "learning_rate": 1.232022210473922e-06, + "loss": 0.0019, + "step": 51613 + }, + { + "epoch": 6.76, + "learning_rate": 1.231767211070547e-06, + "loss": 0.0013, + "step": 51614 + }, + { + "epoch": 6.76, + "learning_rate": 1.2315122363273601e-06, + "loss": 0.0021, + "step": 51615 + }, + { + "epoch": 6.76, + "learning_rate": 1.2312572862450767e-06, + "loss": 0.0015, + "step": 51616 + }, + { + "epoch": 6.76, + "learning_rate": 1.2310023608244138e-06, + "loss": 0.0023, + "step": 51617 + }, + { + "epoch": 6.76, + "learning_rate": 1.2307474600660928e-06, + "loss": 0.0022, + "step": 51618 + }, + { + "epoch": 6.76, + "learning_rate": 1.230492583970826e-06, + "loss": 0.0012, + "step": 51619 + }, + { + "epoch": 6.76, + "learning_rate": 1.2302377325393334e-06, + "loss": 0.001, + "step": 51620 + }, + { + "epoch": 6.76, + "learning_rate": 1.2299829057723268e-06, + "loss": 0.0015, + "step": 51621 + }, + { + "epoch": 6.76, + "learning_rate": 1.229728103670529e-06, + "loss": 0.0028, + "step": 51622 + }, + { + "epoch": 6.76, + "learning_rate": 1.2294733262346515e-06, + "loss": 0.0009, + "step": 51623 + }, + { + "epoch": 6.76, + "learning_rate": 1.229218573465415e-06, + "loss": 0.0016, + "step": 51624 + }, + { + "epoch": 6.76, + "learning_rate": 1.2289638453635322e-06, + "loss": 0.0019, + "step": 51625 + }, + { + "epoch": 6.76, + "learning_rate": 1.2287091419297237e-06, + "loss": 0.0016, + "step": 51626 + }, + { + "epoch": 6.76, + "learning_rate": 1.2284544631647043e-06, + "loss": 0.0022, + "step": 51627 + }, + { + "epoch": 6.76, + "learning_rate": 1.2281998090691871e-06, + "loss": 0.0028, + "step": 51628 + }, + { + "epoch": 6.76, + "learning_rate": 1.2279451796438924e-06, + "loss": 0.0019, + "step": 51629 + }, + { + "epoch": 6.76, + "learning_rate": 1.2276905748895352e-06, + "loss": 0.0015, + "step": 51630 + }, + { + "epoch": 6.76, + "learning_rate": 1.2274359948068315e-06, + "loss": 0.0015, + "step": 51631 + }, + { + "epoch": 6.76, + "learning_rate": 1.2271814393964942e-06, + "loss": 0.002, + "step": 51632 + }, + { + "epoch": 6.76, + "learning_rate": 1.2269269086592439e-06, + "loss": 0.0037, + "step": 51633 + }, + { + "epoch": 6.76, + "learning_rate": 1.2266724025957933e-06, + "loss": 0.0019, + "step": 51634 + }, + { + "epoch": 6.76, + "learning_rate": 1.2264179212068606e-06, + "loss": 0.0008, + "step": 51635 + }, + { + "epoch": 6.76, + "learning_rate": 1.2261634644931574e-06, + "loss": 0.0012, + "step": 51636 + }, + { + "epoch": 6.76, + "learning_rate": 1.2259090324554057e-06, + "loss": 0.0014, + "step": 51637 + }, + { + "epoch": 6.76, + "learning_rate": 1.225654625094317e-06, + "loss": 0.0013, + "step": 51638 + }, + { + "epoch": 6.76, + "learning_rate": 1.225400242410608e-06, + "loss": 0.0024, + "step": 51639 + }, + { + "epoch": 6.76, + "learning_rate": 1.2251458844049914e-06, + "loss": 0.0018, + "step": 51640 + }, + { + "epoch": 6.76, + "learning_rate": 1.2248915510781857e-06, + "loss": 0.0014, + "step": 51641 + }, + { + "epoch": 6.76, + "learning_rate": 1.224637242430906e-06, + "loss": 0.0012, + "step": 51642 + }, + { + "epoch": 6.76, + "learning_rate": 1.2243829584638646e-06, + "loss": 0.0009, + "step": 51643 + }, + { + "epoch": 6.76, + "learning_rate": 1.2241286991777813e-06, + "loss": 0.0017, + "step": 51644 + }, + { + "epoch": 6.77, + "learning_rate": 1.2238744645733659e-06, + "loss": 0.0028, + "step": 51645 + }, + { + "epoch": 6.77, + "learning_rate": 1.2236202546513387e-06, + "loss": 0.0017, + "step": 51646 + }, + { + "epoch": 6.77, + "learning_rate": 1.2233660694124128e-06, + "loss": 0.001, + "step": 51647 + }, + { + "epoch": 6.77, + "learning_rate": 1.2231119088572996e-06, + "loss": 0.0016, + "step": 51648 + }, + { + "epoch": 6.77, + "learning_rate": 1.2228577729867185e-06, + "loss": 0.0018, + "step": 51649 + }, + { + "epoch": 6.77, + "learning_rate": 1.2226036618013825e-06, + "loss": 0.0008, + "step": 51650 + }, + { + "epoch": 6.77, + "learning_rate": 1.2223495753020042e-06, + "loss": 0.0032, + "step": 51651 + }, + { + "epoch": 6.77, + "learning_rate": 1.222095513489303e-06, + "loss": 0.002, + "step": 51652 + }, + { + "epoch": 6.77, + "learning_rate": 1.2218414763639908e-06, + "loss": 0.0027, + "step": 51653 + }, + { + "epoch": 6.77, + "learning_rate": 1.2215874639267822e-06, + "loss": 0.0017, + "step": 51654 + }, + { + "epoch": 6.77, + "learning_rate": 1.2213334761783878e-06, + "loss": 0.0009, + "step": 51655 + }, + { + "epoch": 6.77, + "learning_rate": 1.2210795131195263e-06, + "loss": 0.0018, + "step": 51656 + }, + { + "epoch": 6.77, + "learning_rate": 1.2208255747509134e-06, + "loss": 0.0016, + "step": 51657 + }, + { + "epoch": 6.77, + "learning_rate": 1.220571661073261e-06, + "loss": 0.0017, + "step": 51658 + }, + { + "epoch": 6.77, + "learning_rate": 1.2203177720872815e-06, + "loss": 0.0021, + "step": 51659 + }, + { + "epoch": 6.77, + "learning_rate": 1.2200639077936928e-06, + "loss": 0.0013, + "step": 51660 + }, + { + "epoch": 6.77, + "learning_rate": 1.2198100681932068e-06, + "loss": 0.0023, + "step": 51661 + }, + { + "epoch": 6.77, + "learning_rate": 1.2195562532865368e-06, + "loss": 0.0025, + "step": 51662 + }, + { + "epoch": 6.77, + "learning_rate": 1.2193024630743954e-06, + "loss": 0.0017, + "step": 51663 + }, + { + "epoch": 6.77, + "learning_rate": 1.219048697557501e-06, + "loss": 0.0018, + "step": 51664 + }, + { + "epoch": 6.77, + "learning_rate": 1.2187949567365631e-06, + "loss": 0.0015, + "step": 51665 + }, + { + "epoch": 6.77, + "learning_rate": 1.2185412406122953e-06, + "loss": 0.0006, + "step": 51666 + }, + { + "epoch": 6.77, + "learning_rate": 1.2182875491854129e-06, + "loss": 0.0023, + "step": 51667 + }, + { + "epoch": 6.77, + "learning_rate": 1.2180338824566318e-06, + "loss": 0.001, + "step": 51668 + }, + { + "epoch": 6.77, + "learning_rate": 1.2177802404266614e-06, + "loss": 0.0019, + "step": 51669 + }, + { + "epoch": 6.77, + "learning_rate": 1.217526623096218e-06, + "loss": 0.0014, + "step": 51670 + }, + { + "epoch": 6.77, + "learning_rate": 1.2172730304660095e-06, + "loss": 0.0014, + "step": 51671 + }, + { + "epoch": 6.77, + "learning_rate": 1.217019462536756e-06, + "loss": 0.0022, + "step": 51672 + }, + { + "epoch": 6.77, + "learning_rate": 1.2167659193091675e-06, + "loss": 0.0021, + "step": 51673 + }, + { + "epoch": 6.77, + "learning_rate": 1.2165124007839546e-06, + "loss": 0.0012, + "step": 51674 + }, + { + "epoch": 6.77, + "learning_rate": 1.2162589069618359e-06, + "loss": 0.002, + "step": 51675 + }, + { + "epoch": 6.77, + "learning_rate": 1.2160054378435193e-06, + "loss": 0.0015, + "step": 51676 + }, + { + "epoch": 6.77, + "learning_rate": 1.2157519934297213e-06, + "loss": 0.0013, + "step": 51677 + }, + { + "epoch": 6.77, + "learning_rate": 1.2154985737211489e-06, + "loss": 0.002, + "step": 51678 + }, + { + "epoch": 6.77, + "learning_rate": 1.2152451787185226e-06, + "loss": 0.0038, + "step": 51679 + }, + { + "epoch": 6.77, + "learning_rate": 1.214991808422551e-06, + "loss": 0.0021, + "step": 51680 + }, + { + "epoch": 6.77, + "learning_rate": 1.2147384628339475e-06, + "loss": 0.0023, + "step": 51681 + }, + { + "epoch": 6.77, + "learning_rate": 1.214485141953422e-06, + "loss": 0.0013, + "step": 51682 + }, + { + "epoch": 6.77, + "learning_rate": 1.2142318457816915e-06, + "loss": 0.0017, + "step": 51683 + }, + { + "epoch": 6.77, + "learning_rate": 1.2139785743194665e-06, + "loss": 0.0026, + "step": 51684 + }, + { + "epoch": 6.77, + "learning_rate": 1.2137253275674576e-06, + "loss": 0.0019, + "step": 51685 + }, + { + "epoch": 6.77, + "learning_rate": 1.2134721055263777e-06, + "loss": 0.0015, + "step": 51686 + }, + { + "epoch": 6.77, + "learning_rate": 1.2132189081969403e-06, + "loss": 0.0007, + "step": 51687 + }, + { + "epoch": 6.77, + "learning_rate": 1.2129657355798573e-06, + "loss": 0.0018, + "step": 51688 + }, + { + "epoch": 6.77, + "learning_rate": 1.2127125876758394e-06, + "loss": 0.0024, + "step": 51689 + }, + { + "epoch": 6.77, + "learning_rate": 1.212459464485598e-06, + "loss": 0.0021, + "step": 51690 + }, + { + "epoch": 6.77, + "learning_rate": 1.2122063660098483e-06, + "loss": 0.0014, + "step": 51691 + }, + { + "epoch": 6.77, + "learning_rate": 1.2119532922493017e-06, + "loss": 0.0011, + "step": 51692 + }, + { + "epoch": 6.77, + "learning_rate": 1.2117002432046653e-06, + "loss": 0.0013, + "step": 51693 + }, + { + "epoch": 6.77, + "learning_rate": 1.2114472188766557e-06, + "loss": 0.0031, + "step": 51694 + }, + { + "epoch": 6.77, + "learning_rate": 1.2111942192659832e-06, + "loss": 0.0013, + "step": 51695 + }, + { + "epoch": 6.77, + "learning_rate": 1.2109412443733592e-06, + "loss": 0.0016, + "step": 51696 + }, + { + "epoch": 6.77, + "learning_rate": 1.2106882941994924e-06, + "loss": 0.0018, + "step": 51697 + }, + { + "epoch": 6.77, + "learning_rate": 1.2104353687450987e-06, + "loss": 0.0014, + "step": 51698 + }, + { + "epoch": 6.77, + "learning_rate": 1.2101824680108876e-06, + "loss": 0.0024, + "step": 51699 + }, + { + "epoch": 6.77, + "learning_rate": 1.2099295919975684e-06, + "loss": 0.0012, + "step": 51700 + }, + { + "epoch": 6.77, + "learning_rate": 1.209676740705854e-06, + "loss": 0.0021, + "step": 51701 + }, + { + "epoch": 6.77, + "learning_rate": 1.2094239141364582e-06, + "loss": 0.0032, + "step": 51702 + }, + { + "epoch": 6.77, + "learning_rate": 1.2091711122900895e-06, + "loss": 0.0017, + "step": 51703 + }, + { + "epoch": 6.77, + "learning_rate": 1.2089183351674583e-06, + "loss": 0.0023, + "step": 51704 + }, + { + "epoch": 6.77, + "learning_rate": 1.2086655827692739e-06, + "loss": 0.001, + "step": 51705 + }, + { + "epoch": 6.77, + "learning_rate": 1.2084128550962514e-06, + "loss": 0.0017, + "step": 51706 + }, + { + "epoch": 6.77, + "learning_rate": 1.2081601521491005e-06, + "loss": 0.0019, + "step": 51707 + }, + { + "epoch": 6.77, + "learning_rate": 1.207907473928528e-06, + "loss": 0.0015, + "step": 51708 + }, + { + "epoch": 6.77, + "learning_rate": 1.2076548204352501e-06, + "loss": 0.0013, + "step": 51709 + }, + { + "epoch": 6.77, + "learning_rate": 1.2074021916699752e-06, + "loss": 0.0022, + "step": 51710 + }, + { + "epoch": 6.77, + "learning_rate": 1.2071495876334104e-06, + "loss": 0.0019, + "step": 51711 + }, + { + "epoch": 6.77, + "learning_rate": 1.206897008326271e-06, + "loss": 0.0023, + "step": 51712 + }, + { + "epoch": 6.77, + "learning_rate": 1.206644453749265e-06, + "loss": 0.0017, + "step": 51713 + }, + { + "epoch": 6.77, + "learning_rate": 1.206391923903104e-06, + "loss": 0.0019, + "step": 51714 + }, + { + "epoch": 6.77, + "learning_rate": 1.206139418788498e-06, + "loss": 0.0016, + "step": 51715 + }, + { + "epoch": 6.77, + "learning_rate": 1.2058869384061534e-06, + "loss": 0.0011, + "step": 51716 + }, + { + "epoch": 6.77, + "learning_rate": 1.2056344827567868e-06, + "loss": 0.0011, + "step": 51717 + }, + { + "epoch": 6.77, + "learning_rate": 1.205382051841104e-06, + "loss": 0.001, + "step": 51718 + }, + { + "epoch": 6.77, + "learning_rate": 1.2051296456598173e-06, + "loss": 0.0029, + "step": 51719 + }, + { + "epoch": 6.77, + "learning_rate": 1.2048772642136319e-06, + "loss": 0.0015, + "step": 51720 + }, + { + "epoch": 6.78, + "learning_rate": 1.2046249075032624e-06, + "loss": 0.0024, + "step": 51721 + }, + { + "epoch": 6.78, + "learning_rate": 1.2043725755294155e-06, + "loss": 0.0014, + "step": 51722 + }, + { + "epoch": 6.78, + "learning_rate": 1.2041202682928043e-06, + "loss": 0.0009, + "step": 51723 + }, + { + "epoch": 6.78, + "learning_rate": 1.203867985794135e-06, + "loss": 0.0016, + "step": 51724 + }, + { + "epoch": 6.78, + "learning_rate": 1.2036157280341199e-06, + "loss": 0.0015, + "step": 51725 + }, + { + "epoch": 6.78, + "learning_rate": 1.2033634950134676e-06, + "loss": 0.0019, + "step": 51726 + }, + { + "epoch": 6.78, + "learning_rate": 1.2031112867328864e-06, + "loss": 0.0017, + "step": 51727 + }, + { + "epoch": 6.78, + "learning_rate": 1.2028591031930847e-06, + "loss": 0.0024, + "step": 51728 + }, + { + "epoch": 6.78, + "learning_rate": 1.2026069443947753e-06, + "loss": 0.0023, + "step": 51729 + }, + { + "epoch": 6.78, + "learning_rate": 1.202354810338665e-06, + "loss": 0.0027, + "step": 51730 + }, + { + "epoch": 6.78, + "learning_rate": 1.2021027010254617e-06, + "loss": 0.0022, + "step": 51731 + }, + { + "epoch": 6.78, + "learning_rate": 1.2018506164558764e-06, + "loss": 0.0008, + "step": 51732 + }, + { + "epoch": 6.78, + "learning_rate": 1.2015985566306187e-06, + "loss": 0.0025, + "step": 51733 + }, + { + "epoch": 6.78, + "learning_rate": 1.2013465215503973e-06, + "loss": 0.0022, + "step": 51734 + }, + { + "epoch": 6.78, + "learning_rate": 1.20109451121592e-06, + "loss": 0.0007, + "step": 51735 + }, + { + "epoch": 6.78, + "learning_rate": 1.2008425256278943e-06, + "loss": 0.0023, + "step": 51736 + }, + { + "epoch": 6.78, + "learning_rate": 1.200590564787032e-06, + "loss": 0.003, + "step": 51737 + }, + { + "epoch": 6.78, + "learning_rate": 1.2003386286940399e-06, + "loss": 0.0009, + "step": 51738 + }, + { + "epoch": 6.78, + "learning_rate": 1.2000867173496244e-06, + "loss": 0.0021, + "step": 51739 + }, + { + "epoch": 6.78, + "learning_rate": 1.1998348307544993e-06, + "loss": 0.0016, + "step": 51740 + }, + { + "epoch": 6.78, + "learning_rate": 1.1995829689093697e-06, + "loss": 0.002, + "step": 51741 + }, + { + "epoch": 6.78, + "learning_rate": 1.1993311318149449e-06, + "loss": 0.0015, + "step": 51742 + }, + { + "epoch": 6.78, + "learning_rate": 1.1990793194719275e-06, + "loss": 0.0024, + "step": 51743 + }, + { + "epoch": 6.78, + "learning_rate": 1.1988275318810361e-06, + "loss": 0.002, + "step": 51744 + }, + { + "epoch": 6.78, + "learning_rate": 1.1985757690429722e-06, + "loss": 0.0016, + "step": 51745 + }, + { + "epoch": 6.78, + "learning_rate": 1.1983240309584465e-06, + "loss": 0.0029, + "step": 51746 + }, + { + "epoch": 6.78, + "learning_rate": 1.1980723176281629e-06, + "loss": 0.0021, + "step": 51747 + }, + { + "epoch": 6.78, + "learning_rate": 1.197820629052835e-06, + "loss": 0.0016, + "step": 51748 + }, + { + "epoch": 6.78, + "learning_rate": 1.1975689652331667e-06, + "loss": 0.0007, + "step": 51749 + }, + { + "epoch": 6.78, + "learning_rate": 1.197317326169868e-06, + "loss": 0.0011, + "step": 51750 + }, + { + "epoch": 6.78, + "learning_rate": 1.197065711863643e-06, + "loss": 0.0015, + "step": 51751 + }, + { + "epoch": 6.78, + "learning_rate": 1.1968141223152042e-06, + "loss": 0.0027, + "step": 51752 + }, + { + "epoch": 6.78, + "learning_rate": 1.1965625575252571e-06, + "loss": 0.0021, + "step": 51753 + }, + { + "epoch": 6.78, + "learning_rate": 1.1963110174945058e-06, + "loss": 0.0017, + "step": 51754 + }, + { + "epoch": 6.78, + "learning_rate": 1.1960595022236621e-06, + "loss": 0.0014, + "step": 51755 + }, + { + "epoch": 6.78, + "learning_rate": 1.195808011713434e-06, + "loss": 0.002, + "step": 51756 + }, + { + "epoch": 6.78, + "learning_rate": 1.1955565459645269e-06, + "loss": 0.002, + "step": 51757 + }, + { + "epoch": 6.78, + "learning_rate": 1.1953051049776466e-06, + "loss": 0.0013, + "step": 51758 + }, + { + "epoch": 6.78, + "learning_rate": 1.1950536887535024e-06, + "loss": 0.0024, + "step": 51759 + }, + { + "epoch": 6.78, + "learning_rate": 1.1948022972928019e-06, + "loss": 0.0019, + "step": 51760 + }, + { + "epoch": 6.78, + "learning_rate": 1.1945509305962511e-06, + "loss": 0.0025, + "step": 51761 + }, + { + "epoch": 6.78, + "learning_rate": 1.1942995886645547e-06, + "loss": 0.0024, + "step": 51762 + }, + { + "epoch": 6.78, + "learning_rate": 1.1940482714984236e-06, + "loss": 0.0019, + "step": 51763 + }, + { + "epoch": 6.78, + "learning_rate": 1.1937969790985638e-06, + "loss": 0.0022, + "step": 51764 + }, + { + "epoch": 6.78, + "learning_rate": 1.193545711465678e-06, + "loss": 0.0024, + "step": 51765 + }, + { + "epoch": 6.78, + "learning_rate": 1.1932944686004755e-06, + "loss": 0.0022, + "step": 51766 + }, + { + "epoch": 6.78, + "learning_rate": 1.193043250503666e-06, + "loss": 0.002, + "step": 51767 + }, + { + "epoch": 6.78, + "learning_rate": 1.1927920571759532e-06, + "loss": 0.0012, + "step": 51768 + }, + { + "epoch": 6.78, + "learning_rate": 1.1925408886180445e-06, + "loss": 0.0024, + "step": 51769 + }, + { + "epoch": 6.78, + "learning_rate": 1.1922897448306425e-06, + "loss": 0.0015, + "step": 51770 + }, + { + "epoch": 6.78, + "learning_rate": 1.192038625814459e-06, + "loss": 0.0021, + "step": 51771 + }, + { + "epoch": 6.78, + "learning_rate": 1.1917875315701977e-06, + "loss": 0.0018, + "step": 51772 + }, + { + "epoch": 6.78, + "learning_rate": 1.1915364620985625e-06, + "loss": 0.0033, + "step": 51773 + }, + { + "epoch": 6.78, + "learning_rate": 1.1912854174002653e-06, + "loss": 0.0019, + "step": 51774 + }, + { + "epoch": 6.78, + "learning_rate": 1.1910343974760074e-06, + "loss": 0.0022, + "step": 51775 + }, + { + "epoch": 6.78, + "learning_rate": 1.1907834023264942e-06, + "loss": 0.002, + "step": 51776 + }, + { + "epoch": 6.78, + "learning_rate": 1.190532431952436e-06, + "loss": 0.0022, + "step": 51777 + }, + { + "epoch": 6.78, + "learning_rate": 1.1902814863545341e-06, + "loss": 0.0005, + "step": 51778 + }, + { + "epoch": 6.78, + "learning_rate": 1.1900305655334987e-06, + "loss": 0.0008, + "step": 51779 + }, + { + "epoch": 6.78, + "learning_rate": 1.1897796694900332e-06, + "loss": 0.0013, + "step": 51780 + }, + { + "epoch": 6.78, + "learning_rate": 1.1895287982248405e-06, + "loss": 0.0017, + "step": 51781 + }, + { + "epoch": 6.78, + "learning_rate": 1.189277951738631e-06, + "loss": 0.002, + "step": 51782 + }, + { + "epoch": 6.78, + "learning_rate": 1.1890271300321088e-06, + "loss": 0.0016, + "step": 51783 + }, + { + "epoch": 6.78, + "learning_rate": 1.1887763331059776e-06, + "loss": 0.0036, + "step": 51784 + }, + { + "epoch": 6.78, + "learning_rate": 1.1885255609609425e-06, + "loss": 0.0018, + "step": 51785 + }, + { + "epoch": 6.78, + "learning_rate": 1.188274813597712e-06, + "loss": 0.0024, + "step": 51786 + }, + { + "epoch": 6.78, + "learning_rate": 1.1880240910169872e-06, + "loss": 0.0011, + "step": 51787 + }, + { + "epoch": 6.78, + "learning_rate": 1.1877733932194769e-06, + "loss": 0.0025, + "step": 51788 + }, + { + "epoch": 6.78, + "learning_rate": 1.1875227202058837e-06, + "loss": 0.0014, + "step": 51789 + }, + { + "epoch": 6.78, + "learning_rate": 1.187272071976915e-06, + "loss": 0.0018, + "step": 51790 + }, + { + "epoch": 6.78, + "learning_rate": 1.1870214485332732e-06, + "loss": 0.0018, + "step": 51791 + }, + { + "epoch": 6.78, + "learning_rate": 1.1867708498756659e-06, + "loss": 0.0018, + "step": 51792 + }, + { + "epoch": 6.78, + "learning_rate": 1.1865202760047934e-06, + "loss": 0.0015, + "step": 51793 + }, + { + "epoch": 6.78, + "learning_rate": 1.1862697269213652e-06, + "loss": 0.0045, + "step": 51794 + }, + { + "epoch": 6.78, + "learning_rate": 1.1860192026260841e-06, + "loss": 0.003, + "step": 51795 + }, + { + "epoch": 6.78, + "learning_rate": 1.185768703119653e-06, + "loss": 0.0026, + "step": 51796 + }, + { + "epoch": 6.79, + "learning_rate": 1.18551822840278e-06, + "loss": 0.0021, + "step": 51797 + }, + { + "epoch": 6.79, + "learning_rate": 1.1852677784761656e-06, + "loss": 0.0031, + "step": 51798 + }, + { + "epoch": 6.79, + "learning_rate": 1.1850173533405185e-06, + "loss": 0.003, + "step": 51799 + }, + { + "epoch": 6.79, + "learning_rate": 1.18476695299654e-06, + "loss": 0.0018, + "step": 51800 + }, + { + "epoch": 6.79, + "learning_rate": 1.184516577444933e-06, + "loss": 0.0016, + "step": 51801 + }, + { + "epoch": 6.79, + "learning_rate": 1.1842662266864058e-06, + "loss": 0.0011, + "step": 51802 + }, + { + "epoch": 6.79, + "learning_rate": 1.18401590072166e-06, + "loss": 0.0012, + "step": 51803 + }, + { + "epoch": 6.79, + "learning_rate": 1.1837655995513986e-06, + "loss": 0.0017, + "step": 51804 + }, + { + "epoch": 6.79, + "learning_rate": 1.1835153231763285e-06, + "loss": 0.0013, + "step": 51805 + }, + { + "epoch": 6.79, + "learning_rate": 1.1832650715971527e-06, + "loss": 0.0046, + "step": 51806 + }, + { + "epoch": 6.79, + "learning_rate": 1.183014844814574e-06, + "loss": 0.0017, + "step": 51807 + }, + { + "epoch": 6.79, + "learning_rate": 1.182764642829295e-06, + "loss": 0.0017, + "step": 51808 + }, + { + "epoch": 6.79, + "learning_rate": 1.1825144656420196e-06, + "loss": 0.0018, + "step": 51809 + }, + { + "epoch": 6.79, + "learning_rate": 1.1822643132534561e-06, + "loss": 0.0028, + "step": 51810 + }, + { + "epoch": 6.79, + "learning_rate": 1.1820141856643042e-06, + "loss": 0.002, + "step": 51811 + }, + { + "epoch": 6.79, + "learning_rate": 1.1817640828752652e-06, + "loss": 0.0023, + "step": 51812 + }, + { + "epoch": 6.79, + "learning_rate": 1.1815140048870476e-06, + "loss": 0.0007, + "step": 51813 + }, + { + "epoch": 6.79, + "learning_rate": 1.181263951700352e-06, + "loss": 0.001, + "step": 51814 + }, + { + "epoch": 6.79, + "learning_rate": 1.1810139233158812e-06, + "loss": 0.0012, + "step": 51815 + }, + { + "epoch": 6.79, + "learning_rate": 1.1807639197343402e-06, + "loss": 0.0012, + "step": 51816 + }, + { + "epoch": 6.79, + "learning_rate": 1.1805139409564303e-06, + "loss": 0.0016, + "step": 51817 + }, + { + "epoch": 6.79, + "learning_rate": 1.180263986982856e-06, + "loss": 0.0014, + "step": 51818 + }, + { + "epoch": 6.79, + "learning_rate": 1.1800140578143171e-06, + "loss": 0.0015, + "step": 51819 + }, + { + "epoch": 6.79, + "learning_rate": 1.1797641534515192e-06, + "loss": 0.0012, + "step": 51820 + }, + { + "epoch": 6.79, + "learning_rate": 1.179514273895167e-06, + "loss": 0.0013, + "step": 51821 + }, + { + "epoch": 6.79, + "learning_rate": 1.1792644191459612e-06, + "loss": 0.0011, + "step": 51822 + }, + { + "epoch": 6.79, + "learning_rate": 1.1790145892046022e-06, + "loss": 0.0021, + "step": 51823 + }, + { + "epoch": 6.79, + "learning_rate": 1.1787647840717975e-06, + "loss": 0.0014, + "step": 51824 + }, + { + "epoch": 6.79, + "learning_rate": 1.1785150037482474e-06, + "loss": 0.001, + "step": 51825 + }, + { + "epoch": 6.79, + "learning_rate": 1.1782652482346535e-06, + "loss": 0.0023, + "step": 51826 + }, + { + "epoch": 6.79, + "learning_rate": 1.1780155175317165e-06, + "loss": 0.0014, + "step": 51827 + }, + { + "epoch": 6.79, + "learning_rate": 1.1777658116401436e-06, + "loss": 0.0019, + "step": 51828 + }, + { + "epoch": 6.79, + "learning_rate": 1.1775161305606342e-06, + "loss": 0.0031, + "step": 51829 + }, + { + "epoch": 6.79, + "learning_rate": 1.1772664742938887e-06, + "loss": 0.0018, + "step": 51830 + }, + { + "epoch": 6.79, + "learning_rate": 1.1770168428406114e-06, + "loss": 0.0015, + "step": 51831 + }, + { + "epoch": 6.79, + "learning_rate": 1.176767236201507e-06, + "loss": 0.0022, + "step": 51832 + }, + { + "epoch": 6.79, + "learning_rate": 1.176517654377275e-06, + "loss": 0.0009, + "step": 51833 + }, + { + "epoch": 6.79, + "learning_rate": 1.1762680973686159e-06, + "loss": 0.002, + "step": 51834 + }, + { + "epoch": 6.79, + "learning_rate": 1.1760185651762324e-06, + "loss": 0.003, + "step": 51835 + }, + { + "epoch": 6.79, + "learning_rate": 1.1757690578008274e-06, + "loss": 0.0014, + "step": 51836 + }, + { + "epoch": 6.79, + "learning_rate": 1.1755195752431027e-06, + "loss": 0.0019, + "step": 51837 + }, + { + "epoch": 6.79, + "learning_rate": 1.1752701175037573e-06, + "loss": 0.0018, + "step": 51838 + }, + { + "epoch": 6.79, + "learning_rate": 1.1750206845834967e-06, + "loss": 0.0016, + "step": 51839 + }, + { + "epoch": 6.79, + "learning_rate": 1.1747712764830211e-06, + "loss": 0.0017, + "step": 51840 + }, + { + "epoch": 6.79, + "learning_rate": 1.17452189320303e-06, + "loss": 0.0016, + "step": 51841 + }, + { + "epoch": 6.79, + "learning_rate": 1.1742725347442252e-06, + "loss": 0.0018, + "step": 51842 + }, + { + "epoch": 6.79, + "learning_rate": 1.1740232011073084e-06, + "loss": 0.0018, + "step": 51843 + }, + { + "epoch": 6.79, + "learning_rate": 1.173773892292983e-06, + "loss": 0.0019, + "step": 51844 + }, + { + "epoch": 6.79, + "learning_rate": 1.1735246083019492e-06, + "loss": 0.0014, + "step": 51845 + }, + { + "epoch": 6.79, + "learning_rate": 1.1732753491349047e-06, + "loss": 0.002, + "step": 51846 + }, + { + "epoch": 6.79, + "learning_rate": 1.173026114792556e-06, + "loss": 0.0015, + "step": 51847 + }, + { + "epoch": 6.79, + "learning_rate": 1.172776905275601e-06, + "loss": 0.0019, + "step": 51848 + }, + { + "epoch": 6.79, + "learning_rate": 1.1725277205847407e-06, + "loss": 0.0013, + "step": 51849 + }, + { + "epoch": 6.79, + "learning_rate": 1.1722785607206744e-06, + "loss": 0.0011, + "step": 51850 + }, + { + "epoch": 6.79, + "learning_rate": 1.1720294256841069e-06, + "loss": 0.0025, + "step": 51851 + }, + { + "epoch": 6.79, + "learning_rate": 1.1717803154757356e-06, + "loss": 0.0016, + "step": 51852 + }, + { + "epoch": 6.79, + "learning_rate": 1.17153123009626e-06, + "loss": 0.002, + "step": 51853 + }, + { + "epoch": 6.79, + "learning_rate": 1.1712821695463838e-06, + "loss": 0.0014, + "step": 51854 + }, + { + "epoch": 6.79, + "learning_rate": 1.1710331338268078e-06, + "loss": 0.0013, + "step": 51855 + }, + { + "epoch": 6.79, + "learning_rate": 1.1707841229382299e-06, + "loss": 0.0027, + "step": 51856 + }, + { + "epoch": 6.79, + "learning_rate": 1.1705351368813522e-06, + "loss": 0.0009, + "step": 51857 + }, + { + "epoch": 6.79, + "learning_rate": 1.1702861756568728e-06, + "loss": 0.0014, + "step": 51858 + }, + { + "epoch": 6.79, + "learning_rate": 1.1700372392654945e-06, + "loss": 0.0019, + "step": 51859 + }, + { + "epoch": 6.79, + "learning_rate": 1.169788327707917e-06, + "loss": 0.0013, + "step": 51860 + }, + { + "epoch": 6.79, + "learning_rate": 1.169539440984837e-06, + "loss": 0.0018, + "step": 51861 + }, + { + "epoch": 6.79, + "learning_rate": 1.16929057909696e-06, + "loss": 0.0024, + "step": 51862 + }, + { + "epoch": 6.79, + "learning_rate": 1.1690417420449795e-06, + "loss": 0.002, + "step": 51863 + }, + { + "epoch": 6.79, + "learning_rate": 1.1687929298296019e-06, + "loss": 0.0009, + "step": 51864 + }, + { + "epoch": 6.79, + "learning_rate": 1.168544142451522e-06, + "loss": 0.0011, + "step": 51865 + }, + { + "epoch": 6.79, + "learning_rate": 1.1682953799114428e-06, + "loss": 0.0013, + "step": 51866 + }, + { + "epoch": 6.79, + "learning_rate": 1.1680466422100623e-06, + "loss": 0.0013, + "step": 51867 + }, + { + "epoch": 6.79, + "learning_rate": 1.1677979293480813e-06, + "loss": 0.002, + "step": 51868 + }, + { + "epoch": 6.79, + "learning_rate": 1.1675492413261958e-06, + "loss": 0.0017, + "step": 51869 + }, + { + "epoch": 6.79, + "learning_rate": 1.1673005781451097e-06, + "loss": 0.0011, + "step": 51870 + }, + { + "epoch": 6.79, + "learning_rate": 1.1670519398055201e-06, + "loss": 0.0018, + "step": 51871 + }, + { + "epoch": 6.79, + "learning_rate": 1.1668033263081269e-06, + "loss": 0.0011, + "step": 51872 + }, + { + "epoch": 6.79, + "learning_rate": 1.1665547376536257e-06, + "loss": 0.003, + "step": 51873 + }, + { + "epoch": 6.8, + "learning_rate": 1.1663061738427205e-06, + "loss": 0.004, + "step": 51874 + }, + { + "epoch": 6.8, + "learning_rate": 1.1660576348761099e-06, + "loss": 0.0012, + "step": 51875 + }, + { + "epoch": 6.8, + "learning_rate": 1.165809120754492e-06, + "loss": 0.0016, + "step": 51876 + }, + { + "epoch": 6.8, + "learning_rate": 1.165560631478564e-06, + "loss": 0.0012, + "step": 51877 + }, + { + "epoch": 6.8, + "learning_rate": 1.1653121670490274e-06, + "loss": 0.0017, + "step": 51878 + }, + { + "epoch": 6.8, + "learning_rate": 1.165063727466581e-06, + "loss": 0.0015, + "step": 51879 + }, + { + "epoch": 6.8, + "learning_rate": 1.1648153127319196e-06, + "loss": 0.0016, + "step": 51880 + }, + { + "epoch": 6.8, + "learning_rate": 1.1645669228457458e-06, + "loss": 0.0013, + "step": 51881 + }, + { + "epoch": 6.8, + "learning_rate": 1.164318557808758e-06, + "loss": 0.0021, + "step": 51882 + }, + { + "epoch": 6.8, + "learning_rate": 1.1640702176216545e-06, + "loss": 0.0014, + "step": 51883 + }, + { + "epoch": 6.8, + "learning_rate": 1.1638219022851293e-06, + "loss": 0.0017, + "step": 51884 + }, + { + "epoch": 6.8, + "learning_rate": 1.1635736117998852e-06, + "loss": 0.0021, + "step": 51885 + }, + { + "epoch": 6.8, + "learning_rate": 1.1633253461666215e-06, + "loss": 0.0012, + "step": 51886 + }, + { + "epoch": 6.8, + "learning_rate": 1.1630771053860357e-06, + "loss": 0.0015, + "step": 51887 + }, + { + "epoch": 6.8, + "learning_rate": 1.1628288894588225e-06, + "loss": 0.0018, + "step": 51888 + }, + { + "epoch": 6.8, + "learning_rate": 1.1625806983856835e-06, + "loss": 0.0021, + "step": 51889 + }, + { + "epoch": 6.8, + "learning_rate": 1.1623325321673163e-06, + "loss": 0.001, + "step": 51890 + }, + { + "epoch": 6.8, + "learning_rate": 1.162084390804419e-06, + "loss": 0.0019, + "step": 51891 + }, + { + "epoch": 6.8, + "learning_rate": 1.1618362742976852e-06, + "loss": 0.0013, + "step": 51892 + }, + { + "epoch": 6.8, + "learning_rate": 1.1615881826478192e-06, + "loss": 0.0022, + "step": 51893 + }, + { + "epoch": 6.8, + "learning_rate": 1.161340115855516e-06, + "loss": 0.0016, + "step": 51894 + }, + { + "epoch": 6.8, + "learning_rate": 1.1610920739214703e-06, + "loss": 0.0015, + "step": 51895 + }, + { + "epoch": 6.8, + "learning_rate": 1.160844056846383e-06, + "loss": 0.0017, + "step": 51896 + }, + { + "epoch": 6.8, + "learning_rate": 1.1605960646309522e-06, + "loss": 0.001, + "step": 51897 + }, + { + "epoch": 6.8, + "learning_rate": 1.1603480972758753e-06, + "loss": 0.0025, + "step": 51898 + }, + { + "epoch": 6.8, + "learning_rate": 1.160100154781848e-06, + "loss": 0.0012, + "step": 51899 + }, + { + "epoch": 6.8, + "learning_rate": 1.159852237149567e-06, + "loss": 0.0013, + "step": 51900 + }, + { + "epoch": 6.8, + "learning_rate": 1.1596043443797322e-06, + "loss": 0.0021, + "step": 51901 + }, + { + "epoch": 6.8, + "learning_rate": 1.1593564764730393e-06, + "loss": 0.0017, + "step": 51902 + }, + { + "epoch": 6.8, + "learning_rate": 1.1591086334301838e-06, + "loss": 0.0014, + "step": 51903 + }, + { + "epoch": 6.8, + "learning_rate": 1.1588608152518655e-06, + "loss": 0.0007, + "step": 51904 + }, + { + "epoch": 6.8, + "learning_rate": 1.1586130219387814e-06, + "loss": 0.0018, + "step": 51905 + }, + { + "epoch": 6.8, + "learning_rate": 1.158365253491628e-06, + "loss": 0.0022, + "step": 51906 + }, + { + "epoch": 6.8, + "learning_rate": 1.1581175099110975e-06, + "loss": 0.0013, + "step": 51907 + }, + { + "epoch": 6.8, + "learning_rate": 1.157869791197892e-06, + "loss": 0.0015, + "step": 51908 + }, + { + "epoch": 6.8, + "learning_rate": 1.1576220973527085e-06, + "loss": 0.0021, + "step": 51909 + }, + { + "epoch": 6.8, + "learning_rate": 1.1573744283762422e-06, + "loss": 0.0011, + "step": 51910 + }, + { + "epoch": 6.8, + "learning_rate": 1.157126784269187e-06, + "loss": 0.0022, + "step": 51911 + }, + { + "epoch": 6.8, + "learning_rate": 1.1568791650322442e-06, + "loss": 0.002, + "step": 51912 + }, + { + "epoch": 6.8, + "learning_rate": 1.156631570666108e-06, + "loss": 0.0015, + "step": 51913 + }, + { + "epoch": 6.8, + "learning_rate": 1.1563840011714743e-06, + "loss": 0.0007, + "step": 51914 + }, + { + "epoch": 6.8, + "learning_rate": 1.1561364565490374e-06, + "loss": 0.0014, + "step": 51915 + }, + { + "epoch": 6.8, + "learning_rate": 1.1558889367994986e-06, + "loss": 0.0017, + "step": 51916 + }, + { + "epoch": 6.8, + "learning_rate": 1.1556414419235506e-06, + "loss": 0.0013, + "step": 51917 + }, + { + "epoch": 6.8, + "learning_rate": 1.1553939719218876e-06, + "loss": 0.003, + "step": 51918 + }, + { + "epoch": 6.8, + "learning_rate": 1.1551465267952089e-06, + "loss": 0.0016, + "step": 51919 + }, + { + "epoch": 6.8, + "learning_rate": 1.1548991065442105e-06, + "loss": 0.0022, + "step": 51920 + }, + { + "epoch": 6.8, + "learning_rate": 1.1546517111695887e-06, + "loss": 0.0028, + "step": 51921 + }, + { + "epoch": 6.8, + "learning_rate": 1.1544043406720374e-06, + "loss": 0.0022, + "step": 51922 + }, + { + "epoch": 6.8, + "learning_rate": 1.1541569950522501e-06, + "loss": 0.0016, + "step": 51923 + }, + { + "epoch": 6.8, + "learning_rate": 1.153909674310928e-06, + "loss": 0.0013, + "step": 51924 + }, + { + "epoch": 6.8, + "learning_rate": 1.1536623784487633e-06, + "loss": 0.0014, + "step": 51925 + }, + { + "epoch": 6.8, + "learning_rate": 1.1534151074664501e-06, + "loss": 0.002, + "step": 51926 + }, + { + "epoch": 6.8, + "learning_rate": 1.153167861364688e-06, + "loss": 0.0015, + "step": 51927 + }, + { + "epoch": 6.8, + "learning_rate": 1.1529206401441706e-06, + "loss": 0.0031, + "step": 51928 + }, + { + "epoch": 6.8, + "learning_rate": 1.1526734438055899e-06, + "loss": 0.0023, + "step": 51929 + }, + { + "epoch": 6.8, + "learning_rate": 1.1524262723496448e-06, + "loss": 0.0027, + "step": 51930 + }, + { + "epoch": 6.8, + "learning_rate": 1.152179125777031e-06, + "loss": 0.0021, + "step": 51931 + }, + { + "epoch": 6.8, + "learning_rate": 1.151932004088443e-06, + "loss": 0.0019, + "step": 51932 + }, + { + "epoch": 6.8, + "learning_rate": 1.1516849072845748e-06, + "loss": 0.0017, + "step": 51933 + }, + { + "epoch": 6.8, + "learning_rate": 1.1514378353661192e-06, + "loss": 0.0011, + "step": 51934 + }, + { + "epoch": 6.8, + "learning_rate": 1.1511907883337758e-06, + "loss": 0.0016, + "step": 51935 + }, + { + "epoch": 6.8, + "learning_rate": 1.1509437661882372e-06, + "loss": 0.0014, + "step": 51936 + }, + { + "epoch": 6.8, + "learning_rate": 1.1506967689301974e-06, + "loss": 0.0015, + "step": 51937 + }, + { + "epoch": 6.8, + "learning_rate": 1.15044979656035e-06, + "loss": 0.0016, + "step": 51938 + }, + { + "epoch": 6.8, + "learning_rate": 1.1502028490793937e-06, + "loss": 0.0022, + "step": 51939 + }, + { + "epoch": 6.8, + "learning_rate": 1.1499559264880178e-06, + "loss": 0.0011, + "step": 51940 + }, + { + "epoch": 6.8, + "learning_rate": 1.1497090287869217e-06, + "loss": 0.0006, + "step": 51941 + }, + { + "epoch": 6.8, + "learning_rate": 1.149462155976796e-06, + "loss": 0.0015, + "step": 51942 + }, + { + "epoch": 6.8, + "learning_rate": 1.149215308058339e-06, + "loss": 0.001, + "step": 51943 + }, + { + "epoch": 6.8, + "learning_rate": 1.1489684850322424e-06, + "loss": 0.0024, + "step": 51944 + }, + { + "epoch": 6.8, + "learning_rate": 1.148721686899198e-06, + "loss": 0.0025, + "step": 51945 + }, + { + "epoch": 6.8, + "learning_rate": 1.148474913659905e-06, + "loss": 0.0017, + "step": 51946 + }, + { + "epoch": 6.8, + "learning_rate": 1.148228165315055e-06, + "loss": 0.001, + "step": 51947 + }, + { + "epoch": 6.8, + "learning_rate": 1.1479814418653424e-06, + "loss": 0.0015, + "step": 51948 + }, + { + "epoch": 6.8, + "learning_rate": 1.1477347433114572e-06, + "loss": 0.0015, + "step": 51949 + }, + { + "epoch": 6.81, + "learning_rate": 1.1474880696541003e-06, + "loss": 0.0014, + "step": 51950 + }, + { + "epoch": 6.81, + "learning_rate": 1.1472414208939587e-06, + "loss": 0.0015, + "step": 51951 + }, + { + "epoch": 6.81, + "learning_rate": 1.1469947970317308e-06, + "loss": 0.003, + "step": 51952 + }, + { + "epoch": 6.81, + "learning_rate": 1.1467481980681073e-06, + "loss": 0.002, + "step": 51953 + }, + { + "epoch": 6.81, + "learning_rate": 1.146501624003785e-06, + "loss": 0.0011, + "step": 51954 + }, + { + "epoch": 6.81, + "learning_rate": 1.1462550748394552e-06, + "loss": 0.0019, + "step": 51955 + }, + { + "epoch": 6.81, + "learning_rate": 1.146008550575811e-06, + "loss": 0.0015, + "step": 51956 + }, + { + "epoch": 6.81, + "learning_rate": 1.1457620512135448e-06, + "loss": 0.0014, + "step": 51957 + }, + { + "epoch": 6.81, + "learning_rate": 1.1455155767533543e-06, + "loss": 0.0011, + "step": 51958 + }, + { + "epoch": 6.81, + "learning_rate": 1.1452691271959282e-06, + "loss": 0.0016, + "step": 51959 + }, + { + "epoch": 6.81, + "learning_rate": 1.1450227025419603e-06, + "loss": 0.0024, + "step": 51960 + }, + { + "epoch": 6.81, + "learning_rate": 1.1447763027921455e-06, + "loss": 0.0013, + "step": 51961 + }, + { + "epoch": 6.81, + "learning_rate": 1.1445299279471745e-06, + "loss": 0.0018, + "step": 51962 + }, + { + "epoch": 6.81, + "learning_rate": 1.1442835780077433e-06, + "loss": 0.004, + "step": 51963 + }, + { + "epoch": 6.81, + "learning_rate": 1.1440372529745425e-06, + "loss": 0.0011, + "step": 51964 + }, + { + "epoch": 6.81, + "learning_rate": 1.1437909528482639e-06, + "loss": 0.001, + "step": 51965 + }, + { + "epoch": 6.81, + "learning_rate": 1.1435446776296033e-06, + "loss": 0.002, + "step": 51966 + }, + { + "epoch": 6.81, + "learning_rate": 1.1432984273192516e-06, + "loss": 0.001, + "step": 51967 + }, + { + "epoch": 6.81, + "learning_rate": 1.1430522019178992e-06, + "loss": 0.0014, + "step": 51968 + }, + { + "epoch": 6.81, + "learning_rate": 1.1428060014262432e-06, + "loss": 0.0013, + "step": 51969 + }, + { + "epoch": 6.81, + "learning_rate": 1.1425598258449733e-06, + "loss": 0.0018, + "step": 51970 + }, + { + "epoch": 6.81, + "learning_rate": 1.1423136751747831e-06, + "loss": 0.0023, + "step": 51971 + }, + { + "epoch": 6.81, + "learning_rate": 1.1420675494163613e-06, + "loss": 0.0016, + "step": 51972 + }, + { + "epoch": 6.81, + "learning_rate": 1.1418214485704027e-06, + "loss": 0.0017, + "step": 51973 + }, + { + "epoch": 6.81, + "learning_rate": 1.1415753726376011e-06, + "loss": 0.002, + "step": 51974 + }, + { + "epoch": 6.81, + "learning_rate": 1.1413293216186482e-06, + "loss": 0.0027, + "step": 51975 + }, + { + "epoch": 6.81, + "learning_rate": 1.1410832955142327e-06, + "loss": 0.0011, + "step": 51976 + }, + { + "epoch": 6.81, + "learning_rate": 1.14083729432505e-06, + "loss": 0.0013, + "step": 51977 + }, + { + "epoch": 6.81, + "learning_rate": 1.1405913180517903e-06, + "loss": 0.0013, + "step": 51978 + }, + { + "epoch": 6.81, + "learning_rate": 1.140345366695147e-06, + "loss": 0.0008, + "step": 51979 + }, + { + "epoch": 6.81, + "learning_rate": 1.1400994402558075e-06, + "loss": 0.001, + "step": 51980 + }, + { + "epoch": 6.81, + "learning_rate": 1.139853538734469e-06, + "loss": 0.0015, + "step": 51981 + }, + { + "epoch": 6.81, + "learning_rate": 1.1396076621318209e-06, + "loss": 0.0018, + "step": 51982 + }, + { + "epoch": 6.81, + "learning_rate": 1.1393618104485516e-06, + "loss": 0.0008, + "step": 51983 + }, + { + "epoch": 6.81, + "learning_rate": 1.139115983685356e-06, + "loss": 0.0019, + "step": 51984 + }, + { + "epoch": 6.81, + "learning_rate": 1.138870181842926e-06, + "loss": 0.0024, + "step": 51985 + }, + { + "epoch": 6.81, + "learning_rate": 1.138624404921953e-06, + "loss": 0.0013, + "step": 51986 + }, + { + "epoch": 6.81, + "learning_rate": 1.1383786529231256e-06, + "loss": 0.0016, + "step": 51987 + }, + { + "epoch": 6.81, + "learning_rate": 1.1381329258471353e-06, + "loss": 0.0022, + "step": 51988 + }, + { + "epoch": 6.81, + "learning_rate": 1.137887223694676e-06, + "loss": 0.0017, + "step": 51989 + }, + { + "epoch": 6.81, + "learning_rate": 1.137641546466437e-06, + "loss": 0.0014, + "step": 51990 + }, + { + "epoch": 6.81, + "learning_rate": 1.1373958941631068e-06, + "loss": 0.0022, + "step": 51991 + }, + { + "epoch": 6.81, + "learning_rate": 1.1371502667853807e-06, + "loss": 0.0019, + "step": 51992 + }, + { + "epoch": 6.81, + "learning_rate": 1.1369046643339477e-06, + "loss": 0.0007, + "step": 51993 + }, + { + "epoch": 6.81, + "learning_rate": 1.1366590868094961e-06, + "loss": 0.0016, + "step": 51994 + }, + { + "epoch": 6.81, + "learning_rate": 1.1364135342127192e-06, + "loss": 0.001, + "step": 51995 + }, + { + "epoch": 6.81, + "learning_rate": 1.1361680065443093e-06, + "loss": 0.0012, + "step": 51996 + }, + { + "epoch": 6.81, + "learning_rate": 1.1359225038049549e-06, + "loss": 0.0012, + "step": 51997 + }, + { + "epoch": 6.81, + "learning_rate": 1.1356770259953465e-06, + "loss": 0.0015, + "step": 51998 + }, + { + "epoch": 6.81, + "learning_rate": 1.1354315731161715e-06, + "loss": 0.0017, + "step": 51999 + }, + { + "epoch": 6.81, + "learning_rate": 1.1351861451681256e-06, + "loss": 0.0012, + "step": 52000 + }, + { + "epoch": 6.81, + "learning_rate": 1.1349407421518977e-06, + "loss": 0.0013, + "step": 52001 + }, + { + "epoch": 6.81, + "learning_rate": 1.1346953640681758e-06, + "loss": 0.0012, + "step": 52002 + }, + { + "epoch": 6.81, + "learning_rate": 1.1344500109176493e-06, + "loss": 0.0018, + "step": 52003 + }, + { + "epoch": 6.81, + "learning_rate": 1.1342046827010112e-06, + "loss": 0.0013, + "step": 52004 + }, + { + "epoch": 6.81, + "learning_rate": 1.1339593794189497e-06, + "loss": 0.0024, + "step": 52005 + }, + { + "epoch": 6.81, + "learning_rate": 1.1337141010721565e-06, + "loss": 0.002, + "step": 52006 + }, + { + "epoch": 6.81, + "learning_rate": 1.1334688476613176e-06, + "loss": 0.0017, + "step": 52007 + }, + { + "epoch": 6.81, + "learning_rate": 1.1332236191871282e-06, + "loss": 0.0013, + "step": 52008 + }, + { + "epoch": 6.81, + "learning_rate": 1.1329784156502754e-06, + "loss": 0.0011, + "step": 52009 + }, + { + "epoch": 6.81, + "learning_rate": 1.1327332370514466e-06, + "loss": 0.0009, + "step": 52010 + }, + { + "epoch": 6.81, + "learning_rate": 1.1324880833913354e-06, + "loss": 0.0021, + "step": 52011 + }, + { + "epoch": 6.81, + "learning_rate": 1.132242954670628e-06, + "loss": 0.0024, + "step": 52012 + }, + { + "epoch": 6.81, + "learning_rate": 1.1319978508900164e-06, + "loss": 0.0021, + "step": 52013 + }, + { + "epoch": 6.81, + "learning_rate": 1.1317527720501864e-06, + "loss": 0.0015, + "step": 52014 + }, + { + "epoch": 6.81, + "learning_rate": 1.1315077181518308e-06, + "loss": 0.0014, + "step": 52015 + }, + { + "epoch": 6.81, + "learning_rate": 1.1312626891956358e-06, + "loss": 0.0034, + "step": 52016 + }, + { + "epoch": 6.81, + "learning_rate": 1.1310176851822952e-06, + "loss": 0.0021, + "step": 52017 + }, + { + "epoch": 6.81, + "learning_rate": 1.1307727061124918e-06, + "loss": 0.0018, + "step": 52018 + }, + { + "epoch": 6.81, + "learning_rate": 1.1305277519869206e-06, + "loss": 0.0015, + "step": 52019 + }, + { + "epoch": 6.81, + "learning_rate": 1.1302828228062667e-06, + "loss": 0.0014, + "step": 52020 + }, + { + "epoch": 6.81, + "learning_rate": 1.1300379185712207e-06, + "loss": 0.0017, + "step": 52021 + }, + { + "epoch": 6.81, + "learning_rate": 1.1297930392824696e-06, + "loss": 0.0011, + "step": 52022 + }, + { + "epoch": 6.81, + "learning_rate": 1.1295481849407042e-06, + "loss": 0.0014, + "step": 52023 + }, + { + "epoch": 6.81, + "learning_rate": 1.1293033555466115e-06, + "loss": 0.0016, + "step": 52024 + }, + { + "epoch": 6.81, + "learning_rate": 1.12905855110088e-06, + "loss": 0.0024, + "step": 52025 + }, + { + "epoch": 6.82, + "learning_rate": 1.1288137716042002e-06, + "loss": 0.0019, + "step": 52026 + }, + { + "epoch": 6.82, + "learning_rate": 1.1285690170572571e-06, + "loss": 0.0017, + "step": 52027 + }, + { + "epoch": 6.82, + "learning_rate": 1.1283242874607435e-06, + "loss": 0.0014, + "step": 52028 + }, + { + "epoch": 6.82, + "learning_rate": 1.1280795828153458e-06, + "loss": 0.0024, + "step": 52029 + }, + { + "epoch": 6.82, + "learning_rate": 1.1278349031217495e-06, + "loss": 0.0025, + "step": 52030 + }, + { + "epoch": 6.82, + "learning_rate": 1.1275902483806477e-06, + "loss": 0.0009, + "step": 52031 + }, + { + "epoch": 6.82, + "learning_rate": 1.1273456185927245e-06, + "loss": 0.0011, + "step": 52032 + }, + { + "epoch": 6.82, + "learning_rate": 1.127101013758668e-06, + "loss": 0.0013, + "step": 52033 + }, + { + "epoch": 6.82, + "learning_rate": 1.1268564338791698e-06, + "loss": 0.0015, + "step": 52034 + }, + { + "epoch": 6.82, + "learning_rate": 1.1266118789549151e-06, + "loss": 0.0018, + "step": 52035 + }, + { + "epoch": 6.82, + "learning_rate": 1.1263673489865912e-06, + "loss": 0.0013, + "step": 52036 + }, + { + "epoch": 6.82, + "learning_rate": 1.126122843974885e-06, + "loss": 0.0016, + "step": 52037 + }, + { + "epoch": 6.82, + "learning_rate": 1.1258783639204862e-06, + "loss": 0.0031, + "step": 52038 + }, + { + "epoch": 6.82, + "learning_rate": 1.1256339088240841e-06, + "loss": 0.002, + "step": 52039 + }, + { + "epoch": 6.82, + "learning_rate": 1.1253894786863629e-06, + "loss": 0.0019, + "step": 52040 + }, + { + "epoch": 6.82, + "learning_rate": 1.1251450735080105e-06, + "loss": 0.0009, + "step": 52041 + }, + { + "epoch": 6.82, + "learning_rate": 1.1249006932897154e-06, + "loss": 0.0014, + "step": 52042 + }, + { + "epoch": 6.82, + "learning_rate": 1.1246563380321663e-06, + "loss": 0.0018, + "step": 52043 + }, + { + "epoch": 6.82, + "learning_rate": 1.1244120077360477e-06, + "loss": 0.0016, + "step": 52044 + }, + { + "epoch": 6.82, + "learning_rate": 1.1241677024020447e-06, + "loss": 0.0021, + "step": 52045 + }, + { + "epoch": 6.82, + "learning_rate": 1.1239234220308515e-06, + "loss": 0.0014, + "step": 52046 + }, + { + "epoch": 6.82, + "learning_rate": 1.1236791666231494e-06, + "loss": 0.0032, + "step": 52047 + }, + { + "epoch": 6.82, + "learning_rate": 1.1234349361796248e-06, + "loss": 0.0018, + "step": 52048 + }, + { + "epoch": 6.82, + "learning_rate": 1.1231907307009682e-06, + "loss": 0.0024, + "step": 52049 + }, + { + "epoch": 6.82, + "learning_rate": 1.1229465501878655e-06, + "loss": 0.0015, + "step": 52050 + }, + { + "epoch": 6.82, + "learning_rate": 1.1227023946410033e-06, + "loss": 0.0021, + "step": 52051 + }, + { + "epoch": 6.82, + "learning_rate": 1.1224582640610682e-06, + "loss": 0.0018, + "step": 52052 + }, + { + "epoch": 6.82, + "learning_rate": 1.1222141584487444e-06, + "loss": 0.002, + "step": 52053 + }, + { + "epoch": 6.82, + "learning_rate": 1.1219700778047226e-06, + "loss": 0.002, + "step": 52054 + }, + { + "epoch": 6.82, + "learning_rate": 1.1217260221296866e-06, + "loss": 0.0013, + "step": 52055 + }, + { + "epoch": 6.82, + "learning_rate": 1.1214819914243224e-06, + "loss": 0.0017, + "step": 52056 + }, + { + "epoch": 6.82, + "learning_rate": 1.1212379856893186e-06, + "loss": 0.0022, + "step": 52057 + }, + { + "epoch": 6.82, + "learning_rate": 1.1209940049253598e-06, + "loss": 0.0016, + "step": 52058 + }, + { + "epoch": 6.82, + "learning_rate": 1.1207500491331336e-06, + "loss": 0.0021, + "step": 52059 + }, + { + "epoch": 6.82, + "learning_rate": 1.1205061183133203e-06, + "loss": 0.001, + "step": 52060 + }, + { + "epoch": 6.82, + "learning_rate": 1.120262212466614e-06, + "loss": 0.0022, + "step": 52061 + }, + { + "epoch": 6.82, + "learning_rate": 1.1200183315936985e-06, + "loss": 0.0021, + "step": 52062 + }, + { + "epoch": 6.82, + "learning_rate": 1.1197744756952577e-06, + "loss": 0.0027, + "step": 52063 + }, + { + "epoch": 6.82, + "learning_rate": 1.1195306447719767e-06, + "loss": 0.0021, + "step": 52064 + }, + { + "epoch": 6.82, + "learning_rate": 1.1192868388245448e-06, + "loss": 0.0018, + "step": 52065 + }, + { + "epoch": 6.82, + "learning_rate": 1.119043057853646e-06, + "loss": 0.0019, + "step": 52066 + }, + { + "epoch": 6.82, + "learning_rate": 1.1187993018599664e-06, + "loss": 0.0017, + "step": 52067 + }, + { + "epoch": 6.82, + "learning_rate": 1.1185555708441875e-06, + "loss": 0.0011, + "step": 52068 + }, + { + "epoch": 6.82, + "learning_rate": 1.1183118648070003e-06, + "loss": 0.0015, + "step": 52069 + }, + { + "epoch": 6.82, + "learning_rate": 1.118068183749088e-06, + "loss": 0.0014, + "step": 52070 + }, + { + "epoch": 6.82, + "learning_rate": 1.1178245276711341e-06, + "loss": 0.0013, + "step": 52071 + }, + { + "epoch": 6.82, + "learning_rate": 1.1175808965738255e-06, + "loss": 0.0018, + "step": 52072 + }, + { + "epoch": 6.82, + "learning_rate": 1.1173372904578506e-06, + "loss": 0.0017, + "step": 52073 + }, + { + "epoch": 6.82, + "learning_rate": 1.117093709323891e-06, + "loss": 0.0021, + "step": 52074 + }, + { + "epoch": 6.82, + "learning_rate": 1.1168501531726295e-06, + "loss": 0.0025, + "step": 52075 + }, + { + "epoch": 6.82, + "learning_rate": 1.1166066220047566e-06, + "loss": 0.0025, + "step": 52076 + }, + { + "epoch": 6.82, + "learning_rate": 1.1163631158209542e-06, + "loss": 0.0017, + "step": 52077 + }, + { + "epoch": 6.82, + "learning_rate": 1.1161196346219084e-06, + "loss": 0.0015, + "step": 52078 + }, + { + "epoch": 6.82, + "learning_rate": 1.1158761784082995e-06, + "loss": 0.0014, + "step": 52079 + }, + { + "epoch": 6.82, + "learning_rate": 1.1156327471808194e-06, + "loss": 0.0014, + "step": 52080 + }, + { + "epoch": 6.82, + "learning_rate": 1.1153893409401473e-06, + "loss": 0.0011, + "step": 52081 + }, + { + "epoch": 6.82, + "learning_rate": 1.1151459596869686e-06, + "loss": 0.0015, + "step": 52082 + }, + { + "epoch": 6.82, + "learning_rate": 1.114902603421969e-06, + "loss": 0.002, + "step": 52083 + }, + { + "epoch": 6.82, + "learning_rate": 1.1146592721458338e-06, + "loss": 0.0014, + "step": 52084 + }, + { + "epoch": 6.82, + "learning_rate": 1.1144159658592468e-06, + "loss": 0.0013, + "step": 52085 + }, + { + "epoch": 6.82, + "learning_rate": 1.114172684562892e-06, + "loss": 0.002, + "step": 52086 + }, + { + "epoch": 6.82, + "learning_rate": 1.1139294282574509e-06, + "loss": 0.0014, + "step": 52087 + }, + { + "epoch": 6.82, + "learning_rate": 1.113686196943612e-06, + "loss": 0.0016, + "step": 52088 + }, + { + "epoch": 6.82, + "learning_rate": 1.113442990622058e-06, + "loss": 0.0018, + "step": 52089 + }, + { + "epoch": 6.82, + "learning_rate": 1.1131998092934692e-06, + "loss": 0.0012, + "step": 52090 + }, + { + "epoch": 6.82, + "learning_rate": 1.1129566529585356e-06, + "loss": 0.0015, + "step": 52091 + }, + { + "epoch": 6.82, + "learning_rate": 1.1127135216179374e-06, + "loss": 0.0016, + "step": 52092 + }, + { + "epoch": 6.82, + "learning_rate": 1.1124704152723587e-06, + "loss": 0.0014, + "step": 52093 + }, + { + "epoch": 6.82, + "learning_rate": 1.1122273339224844e-06, + "loss": 0.0021, + "step": 52094 + }, + { + "epoch": 6.82, + "learning_rate": 1.111984277568996e-06, + "loss": 0.0036, + "step": 52095 + }, + { + "epoch": 6.82, + "learning_rate": 1.1117412462125799e-06, + "loss": 0.0014, + "step": 52096 + }, + { + "epoch": 6.82, + "learning_rate": 1.1114982398539186e-06, + "loss": 0.0012, + "step": 52097 + }, + { + "epoch": 6.82, + "learning_rate": 1.1112552584936931e-06, + "loss": 0.0013, + "step": 52098 + }, + { + "epoch": 6.82, + "learning_rate": 1.1110123021325903e-06, + "loss": 0.0017, + "step": 52099 + }, + { + "epoch": 6.82, + "learning_rate": 1.110769370771293e-06, + "loss": 0.0024, + "step": 52100 + }, + { + "epoch": 6.82, + "learning_rate": 1.110526464410483e-06, + "loss": 0.0023, + "step": 52101 + }, + { + "epoch": 6.82, + "learning_rate": 1.110283583050842e-06, + "loss": 0.0025, + "step": 52102 + }, + { + "epoch": 6.83, + "learning_rate": 1.1100407266930569e-06, + "loss": 0.0014, + "step": 52103 + }, + { + "epoch": 6.83, + "learning_rate": 1.1097978953378063e-06, + "loss": 0.0014, + "step": 52104 + }, + { + "epoch": 6.83, + "learning_rate": 1.1095550889857787e-06, + "loss": 0.0005, + "step": 52105 + }, + { + "epoch": 6.83, + "learning_rate": 1.109312307637651e-06, + "loss": 0.0015, + "step": 52106 + }, + { + "epoch": 6.83, + "learning_rate": 1.109069551294112e-06, + "loss": 0.0017, + "step": 52107 + }, + { + "epoch": 6.83, + "learning_rate": 1.1088268199558405e-06, + "loss": 0.0011, + "step": 52108 + }, + { + "epoch": 6.83, + "learning_rate": 1.108584113623521e-06, + "loss": 0.0015, + "step": 52109 + }, + { + "epoch": 6.83, + "learning_rate": 1.1083414322978324e-06, + "loss": 0.0008, + "step": 52110 + }, + { + "epoch": 6.83, + "learning_rate": 1.1080987759794627e-06, + "loss": 0.0012, + "step": 52111 + }, + { + "epoch": 6.83, + "learning_rate": 1.1078561446690906e-06, + "loss": 0.0024, + "step": 52112 + }, + { + "epoch": 6.83, + "learning_rate": 1.107613538367398e-06, + "loss": 0.002, + "step": 52113 + }, + { + "epoch": 6.83, + "learning_rate": 1.1073709570750712e-06, + "loss": 0.0014, + "step": 52114 + }, + { + "epoch": 6.83, + "learning_rate": 1.1071284007927886e-06, + "loss": 0.0016, + "step": 52115 + }, + { + "epoch": 6.83, + "learning_rate": 1.1068858695212349e-06, + "loss": 0.002, + "step": 52116 + }, + { + "epoch": 6.83, + "learning_rate": 1.1066433632610907e-06, + "loss": 0.0017, + "step": 52117 + }, + { + "epoch": 6.83, + "learning_rate": 1.1064008820130368e-06, + "loss": 0.003, + "step": 52118 + }, + { + "epoch": 6.83, + "learning_rate": 1.1061584257777592e-06, + "loss": 0.0026, + "step": 52119 + }, + { + "epoch": 6.83, + "learning_rate": 1.1059159945559372e-06, + "loss": 0.0015, + "step": 52120 + }, + { + "epoch": 6.83, + "learning_rate": 1.1056735883482506e-06, + "loss": 0.0014, + "step": 52121 + }, + { + "epoch": 6.83, + "learning_rate": 1.1054312071553863e-06, + "loss": 0.0022, + "step": 52122 + }, + { + "epoch": 6.83, + "learning_rate": 1.1051888509780217e-06, + "loss": 0.0022, + "step": 52123 + }, + { + "epoch": 6.83, + "learning_rate": 1.1049465198168408e-06, + "loss": 0.0019, + "step": 52124 + }, + { + "epoch": 6.83, + "learning_rate": 1.1047042136725216e-06, + "loss": 0.0018, + "step": 52125 + }, + { + "epoch": 6.83, + "learning_rate": 1.1044619325457484e-06, + "loss": 0.003, + "step": 52126 + }, + { + "epoch": 6.83, + "learning_rate": 1.1042196764372036e-06, + "loss": 0.0014, + "step": 52127 + }, + { + "epoch": 6.83, + "learning_rate": 1.1039774453475683e-06, + "loss": 0.0007, + "step": 52128 + }, + { + "epoch": 6.83, + "learning_rate": 1.1037352392775203e-06, + "loss": 0.0016, + "step": 52129 + }, + { + "epoch": 6.83, + "learning_rate": 1.1034930582277448e-06, + "loss": 0.0011, + "step": 52130 + }, + { + "epoch": 6.83, + "learning_rate": 1.1032509021989213e-06, + "loss": 0.0014, + "step": 52131 + }, + { + "epoch": 6.83, + "learning_rate": 1.1030087711917293e-06, + "loss": 0.0011, + "step": 52132 + }, + { + "epoch": 6.83, + "learning_rate": 1.1027666652068536e-06, + "loss": 0.0011, + "step": 52133 + }, + { + "epoch": 6.83, + "learning_rate": 1.1025245842449728e-06, + "loss": 0.0022, + "step": 52134 + }, + { + "epoch": 6.83, + "learning_rate": 1.1022825283067673e-06, + "loss": 0.0019, + "step": 52135 + }, + { + "epoch": 6.83, + "learning_rate": 1.1020404973929166e-06, + "loss": 0.0022, + "step": 52136 + }, + { + "epoch": 6.83, + "learning_rate": 1.1017984915041025e-06, + "loss": 0.0012, + "step": 52137 + }, + { + "epoch": 6.83, + "learning_rate": 1.1015565106410087e-06, + "loss": 0.0009, + "step": 52138 + }, + { + "epoch": 6.83, + "learning_rate": 1.1013145548043137e-06, + "loss": 0.002, + "step": 52139 + }, + { + "epoch": 6.83, + "learning_rate": 1.1010726239946956e-06, + "loss": 0.0023, + "step": 52140 + }, + { + "epoch": 6.83, + "learning_rate": 1.1008307182128386e-06, + "loss": 0.0012, + "step": 52141 + }, + { + "epoch": 6.83, + "learning_rate": 1.1005888374594209e-06, + "loss": 0.0015, + "step": 52142 + }, + { + "epoch": 6.83, + "learning_rate": 1.1003469817351242e-06, + "loss": 0.0016, + "step": 52143 + }, + { + "epoch": 6.83, + "learning_rate": 1.1001051510406246e-06, + "loss": 0.0023, + "step": 52144 + }, + { + "epoch": 6.83, + "learning_rate": 1.0998633453766084e-06, + "loss": 0.0025, + "step": 52145 + }, + { + "epoch": 6.83, + "learning_rate": 1.0996215647437514e-06, + "loss": 0.0014, + "step": 52146 + }, + { + "epoch": 6.83, + "learning_rate": 1.0993798091427332e-06, + "loss": 0.002, + "step": 52147 + }, + { + "epoch": 6.83, + "learning_rate": 1.0991380785742357e-06, + "loss": 0.0019, + "step": 52148 + }, + { + "epoch": 6.83, + "learning_rate": 1.09889637303894e-06, + "loss": 0.002, + "step": 52149 + }, + { + "epoch": 6.83, + "learning_rate": 1.098654692537524e-06, + "loss": 0.0016, + "step": 52150 + }, + { + "epoch": 6.83, + "learning_rate": 1.098413037070668e-06, + "loss": 0.0025, + "step": 52151 + }, + { + "epoch": 6.83, + "learning_rate": 1.0981714066390503e-06, + "loss": 0.0011, + "step": 52152 + }, + { + "epoch": 6.83, + "learning_rate": 1.0979298012433514e-06, + "loss": 0.0016, + "step": 52153 + }, + { + "epoch": 6.83, + "learning_rate": 1.0976882208842521e-06, + "loss": 0.0017, + "step": 52154 + }, + { + "epoch": 6.83, + "learning_rate": 1.0974466655624282e-06, + "loss": 0.0016, + "step": 52155 + }, + { + "epoch": 6.83, + "learning_rate": 1.0972051352785639e-06, + "loss": 0.0029, + "step": 52156 + }, + { + "epoch": 6.83, + "learning_rate": 1.0969636300333363e-06, + "loss": 0.0017, + "step": 52157 + }, + { + "epoch": 6.83, + "learning_rate": 1.0967221498274216e-06, + "loss": 0.0018, + "step": 52158 + }, + { + "epoch": 6.83, + "learning_rate": 1.0964806946615048e-06, + "loss": 0.0011, + "step": 52159 + }, + { + "epoch": 6.83, + "learning_rate": 1.0962392645362586e-06, + "loss": 0.0016, + "step": 52160 + }, + { + "epoch": 6.83, + "learning_rate": 1.0959978594523678e-06, + "loss": 0.0011, + "step": 52161 + }, + { + "epoch": 6.83, + "learning_rate": 1.0957564794105092e-06, + "loss": 0.002, + "step": 52162 + }, + { + "epoch": 6.83, + "learning_rate": 1.0955151244113594e-06, + "loss": 0.0019, + "step": 52163 + }, + { + "epoch": 6.83, + "learning_rate": 1.0952737944556003e-06, + "loss": 0.0022, + "step": 52164 + }, + { + "epoch": 6.83, + "learning_rate": 1.0950324895439103e-06, + "loss": 0.0019, + "step": 52165 + }, + { + "epoch": 6.83, + "learning_rate": 1.0947912096769664e-06, + "loss": 0.0011, + "step": 52166 + }, + { + "epoch": 6.83, + "learning_rate": 1.0945499548554472e-06, + "loss": 0.0016, + "step": 52167 + }, + { + "epoch": 6.83, + "learning_rate": 1.0943087250800332e-06, + "loss": 0.0013, + "step": 52168 + }, + { + "epoch": 6.83, + "learning_rate": 1.0940675203513996e-06, + "loss": 0.0021, + "step": 52169 + }, + { + "epoch": 6.83, + "learning_rate": 1.0938263406702286e-06, + "loss": 0.001, + "step": 52170 + }, + { + "epoch": 6.83, + "learning_rate": 1.093585186037195e-06, + "loss": 0.0018, + "step": 52171 + }, + { + "epoch": 6.83, + "learning_rate": 1.0933440564529806e-06, + "loss": 0.0019, + "step": 52172 + }, + { + "epoch": 6.83, + "learning_rate": 1.0931029519182622e-06, + "loss": 0.0019, + "step": 52173 + }, + { + "epoch": 6.83, + "learning_rate": 1.092861872433717e-06, + "loss": 0.001, + "step": 52174 + }, + { + "epoch": 6.83, + "learning_rate": 1.0926208180000219e-06, + "loss": 0.0021, + "step": 52175 + }, + { + "epoch": 6.83, + "learning_rate": 1.0923797886178577e-06, + "loss": 0.0006, + "step": 52176 + }, + { + "epoch": 6.83, + "learning_rate": 1.0921387842879016e-06, + "loss": 0.0019, + "step": 52177 + }, + { + "epoch": 6.83, + "learning_rate": 1.0918978050108287e-06, + "loss": 0.0012, + "step": 52178 + }, + { + "epoch": 6.84, + "learning_rate": 1.0916568507873216e-06, + "loss": 0.001, + "step": 52179 + }, + { + "epoch": 6.84, + "learning_rate": 1.091415921618052e-06, + "loss": 0.0014, + "step": 52180 + }, + { + "epoch": 6.84, + "learning_rate": 1.091175017503703e-06, + "loss": 0.0016, + "step": 52181 + }, + { + "epoch": 6.84, + "learning_rate": 1.0909341384449479e-06, + "loss": 0.0012, + "step": 52182 + }, + { + "epoch": 6.84, + "learning_rate": 1.0906932844424678e-06, + "loss": 0.0019, + "step": 52183 + }, + { + "epoch": 6.84, + "learning_rate": 1.0904524554969388e-06, + "loss": 0.0014, + "step": 52184 + }, + { + "epoch": 6.84, + "learning_rate": 1.0902116516090378e-06, + "loss": 0.0009, + "step": 52185 + }, + { + "epoch": 6.84, + "learning_rate": 1.08997087277944e-06, + "loss": 0.0012, + "step": 52186 + }, + { + "epoch": 6.84, + "learning_rate": 1.0897301190088272e-06, + "loss": 0.0015, + "step": 52187 + }, + { + "epoch": 6.84, + "learning_rate": 1.089489390297873e-06, + "loss": 0.003, + "step": 52188 + }, + { + "epoch": 6.84, + "learning_rate": 1.0892486866472552e-06, + "loss": 0.0027, + "step": 52189 + }, + { + "epoch": 6.84, + "learning_rate": 1.0890080080576504e-06, + "loss": 0.0013, + "step": 52190 + }, + { + "epoch": 6.84, + "learning_rate": 1.088767354529735e-06, + "loss": 0.0013, + "step": 52191 + }, + { + "epoch": 6.84, + "learning_rate": 1.0885267260641886e-06, + "loss": 0.0011, + "step": 52192 + }, + { + "epoch": 6.84, + "learning_rate": 1.088286122661687e-06, + "loss": 0.001, + "step": 52193 + }, + { + "epoch": 6.84, + "learning_rate": 1.0880455443229032e-06, + "loss": 0.0016, + "step": 52194 + }, + { + "epoch": 6.84, + "learning_rate": 1.0878049910485189e-06, + "loss": 0.0011, + "step": 52195 + }, + { + "epoch": 6.84, + "learning_rate": 1.087564462839209e-06, + "loss": 0.0009, + "step": 52196 + }, + { + "epoch": 6.84, + "learning_rate": 1.0873239596956476e-06, + "loss": 0.0025, + "step": 52197 + }, + { + "epoch": 6.84, + "learning_rate": 1.087083481618515e-06, + "loss": 0.0025, + "step": 52198 + }, + { + "epoch": 6.84, + "learning_rate": 1.086843028608484e-06, + "loss": 0.0012, + "step": 52199 + }, + { + "epoch": 6.84, + "learning_rate": 1.0866026006662334e-06, + "loss": 0.0021, + "step": 52200 + }, + { + "epoch": 6.84, + "learning_rate": 1.0863621977924367e-06, + "loss": 0.0013, + "step": 52201 + }, + { + "epoch": 6.84, + "learning_rate": 1.086121819987771e-06, + "loss": 0.0018, + "step": 52202 + }, + { + "epoch": 6.84, + "learning_rate": 1.085881467252915e-06, + "loss": 0.001, + "step": 52203 + }, + { + "epoch": 6.84, + "learning_rate": 1.0856411395885436e-06, + "loss": 0.0018, + "step": 52204 + }, + { + "epoch": 6.84, + "learning_rate": 1.0854008369953284e-06, + "loss": 0.0015, + "step": 52205 + }, + { + "epoch": 6.84, + "learning_rate": 1.0851605594739511e-06, + "loss": 0.002, + "step": 52206 + }, + { + "epoch": 6.84, + "learning_rate": 1.0849203070250858e-06, + "loss": 0.0017, + "step": 52207 + }, + { + "epoch": 6.84, + "learning_rate": 1.0846800796494062e-06, + "loss": 0.0015, + "step": 52208 + }, + { + "epoch": 6.84, + "learning_rate": 1.0844398773475883e-06, + "loss": 0.0015, + "step": 52209 + }, + { + "epoch": 6.84, + "learning_rate": 1.0841997001203097e-06, + "loss": 0.0016, + "step": 52210 + }, + { + "epoch": 6.84, + "learning_rate": 1.083959547968244e-06, + "loss": 0.0015, + "step": 52211 + }, + { + "epoch": 6.84, + "learning_rate": 1.0837194208920665e-06, + "loss": 0.0015, + "step": 52212 + }, + { + "epoch": 6.84, + "learning_rate": 1.083479318892453e-06, + "loss": 0.0018, + "step": 52213 + }, + { + "epoch": 6.84, + "learning_rate": 1.083239241970081e-06, + "loss": 0.0014, + "step": 52214 + }, + { + "epoch": 6.84, + "learning_rate": 1.0829991901256231e-06, + "loss": 0.0025, + "step": 52215 + }, + { + "epoch": 6.84, + "learning_rate": 1.0827591633597557e-06, + "loss": 0.003, + "step": 52216 + }, + { + "epoch": 6.84, + "learning_rate": 1.0825191616731512e-06, + "loss": 0.0018, + "step": 52217 + }, + { + "epoch": 6.84, + "learning_rate": 1.0822791850664893e-06, + "loss": 0.0024, + "step": 52218 + }, + { + "epoch": 6.84, + "learning_rate": 1.0820392335404416e-06, + "loss": 0.002, + "step": 52219 + }, + { + "epoch": 6.84, + "learning_rate": 1.0817993070956822e-06, + "loss": 0.0024, + "step": 52220 + }, + { + "epoch": 6.84, + "learning_rate": 1.0815594057328882e-06, + "loss": 0.0016, + "step": 52221 + }, + { + "epoch": 6.84, + "learning_rate": 1.0813195294527346e-06, + "loss": 0.0017, + "step": 52222 + }, + { + "epoch": 6.84, + "learning_rate": 1.081079678255894e-06, + "loss": 0.0023, + "step": 52223 + }, + { + "epoch": 6.84, + "learning_rate": 1.0808398521430408e-06, + "loss": 0.0013, + "step": 52224 + }, + { + "epoch": 6.84, + "learning_rate": 1.0806000511148496e-06, + "loss": 0.0015, + "step": 52225 + }, + { + "epoch": 6.84, + "learning_rate": 1.0803602751719988e-06, + "loss": 0.0009, + "step": 52226 + }, + { + "epoch": 6.84, + "learning_rate": 1.080120524315159e-06, + "loss": 0.0032, + "step": 52227 + }, + { + "epoch": 6.84, + "learning_rate": 1.0798807985450033e-06, + "loss": 0.0024, + "step": 52228 + }, + { + "epoch": 6.84, + "learning_rate": 1.0796410978622096e-06, + "loss": 0.002, + "step": 52229 + }, + { + "epoch": 6.84, + "learning_rate": 1.079401422267451e-06, + "loss": 0.0013, + "step": 52230 + }, + { + "epoch": 6.84, + "learning_rate": 1.0791617717614e-06, + "loss": 0.0018, + "step": 52231 + }, + { + "epoch": 6.84, + "learning_rate": 1.0789221463447297e-06, + "loss": 0.0009, + "step": 52232 + }, + { + "epoch": 6.84, + "learning_rate": 1.0786825460181182e-06, + "loss": 0.0026, + "step": 52233 + }, + { + "epoch": 6.84, + "learning_rate": 1.0784429707822363e-06, + "loss": 0.0026, + "step": 52234 + }, + { + "epoch": 6.84, + "learning_rate": 1.0782034206377568e-06, + "loss": 0.0014, + "step": 52235 + }, + { + "epoch": 6.84, + "learning_rate": 1.0779638955853556e-06, + "loss": 0.0015, + "step": 52236 + }, + { + "epoch": 6.84, + "learning_rate": 1.0777243956257066e-06, + "loss": 0.0008, + "step": 52237 + }, + { + "epoch": 6.84, + "learning_rate": 1.077484920759484e-06, + "loss": 0.0027, + "step": 52238 + }, + { + "epoch": 6.84, + "learning_rate": 1.0772454709873592e-06, + "loss": 0.0015, + "step": 52239 + }, + { + "epoch": 6.84, + "learning_rate": 1.077006046310004e-06, + "loss": 0.0024, + "step": 52240 + }, + { + "epoch": 6.84, + "learning_rate": 1.0767666467280967e-06, + "loss": 0.0009, + "step": 52241 + }, + { + "epoch": 6.84, + "learning_rate": 1.0765272722423082e-06, + "loss": 0.0008, + "step": 52242 + }, + { + "epoch": 6.84, + "learning_rate": 1.0762879228533086e-06, + "loss": 0.0022, + "step": 52243 + }, + { + "epoch": 6.84, + "learning_rate": 1.0760485985617764e-06, + "loss": 0.0018, + "step": 52244 + }, + { + "epoch": 6.84, + "learning_rate": 1.0758092993683832e-06, + "loss": 0.0008, + "step": 52245 + }, + { + "epoch": 6.84, + "learning_rate": 1.0755700252737978e-06, + "loss": 0.0017, + "step": 52246 + }, + { + "epoch": 6.84, + "learning_rate": 1.0753307762786969e-06, + "loss": 0.0014, + "step": 52247 + }, + { + "epoch": 6.84, + "learning_rate": 1.0750915523837556e-06, + "loss": 0.0033, + "step": 52248 + }, + { + "epoch": 6.84, + "learning_rate": 1.0748523535896426e-06, + "loss": 0.0011, + "step": 52249 + }, + { + "epoch": 6.84, + "learning_rate": 1.0746131798970338e-06, + "loss": 0.0009, + "step": 52250 + }, + { + "epoch": 6.84, + "learning_rate": 1.0743740313065965e-06, + "loss": 0.0011, + "step": 52251 + }, + { + "epoch": 6.84, + "learning_rate": 1.07413490781901e-06, + "loss": 0.001, + "step": 52252 + }, + { + "epoch": 6.84, + "learning_rate": 1.073895809434944e-06, + "loss": 0.0027, + "step": 52253 + }, + { + "epoch": 6.84, + "learning_rate": 1.0736567361550699e-06, + "loss": 0.0019, + "step": 52254 + }, + { + "epoch": 6.85, + "learning_rate": 1.0734176879800585e-06, + "loss": 0.0016, + "step": 52255 + }, + { + "epoch": 6.85, + "learning_rate": 1.073178664910587e-06, + "loss": 0.0012, + "step": 52256 + }, + { + "epoch": 6.85, + "learning_rate": 1.0729396669473236e-06, + "loss": 0.0012, + "step": 52257 + }, + { + "epoch": 6.85, + "learning_rate": 1.0727006940909435e-06, + "loss": 0.0021, + "step": 52258 + }, + { + "epoch": 6.85, + "learning_rate": 1.072461746342115e-06, + "loss": 0.0013, + "step": 52259 + }, + { + "epoch": 6.85, + "learning_rate": 1.0722228237015154e-06, + "loss": 0.0015, + "step": 52260 + }, + { + "epoch": 6.85, + "learning_rate": 1.0719839261698127e-06, + "loss": 0.001, + "step": 52261 + }, + { + "epoch": 6.85, + "learning_rate": 1.0717450537476782e-06, + "loss": 0.0014, + "step": 52262 + }, + { + "epoch": 6.85, + "learning_rate": 1.0715062064357873e-06, + "loss": 0.002, + "step": 52263 + }, + { + "epoch": 6.85, + "learning_rate": 1.0712673842348099e-06, + "loss": 0.002, + "step": 52264 + }, + { + "epoch": 6.85, + "learning_rate": 1.0710285871454163e-06, + "loss": 0.0014, + "step": 52265 + }, + { + "epoch": 6.85, + "learning_rate": 1.0707898151682783e-06, + "loss": 0.0017, + "step": 52266 + }, + { + "epoch": 6.85, + "learning_rate": 1.0705510683040687e-06, + "loss": 0.0016, + "step": 52267 + }, + { + "epoch": 6.85, + "learning_rate": 1.0703123465534605e-06, + "loss": 0.001, + "step": 52268 + }, + { + "epoch": 6.85, + "learning_rate": 1.0700736499171228e-06, + "loss": 0.0056, + "step": 52269 + }, + { + "epoch": 6.85, + "learning_rate": 1.0698349783957263e-06, + "loss": 0.0009, + "step": 52270 + }, + { + "epoch": 6.85, + "learning_rate": 1.069596331989945e-06, + "loss": 0.002, + "step": 52271 + }, + { + "epoch": 6.85, + "learning_rate": 1.0693577107004482e-06, + "loss": 0.0017, + "step": 52272 + }, + { + "epoch": 6.85, + "learning_rate": 1.0691191145279067e-06, + "loss": 0.0024, + "step": 52273 + }, + { + "epoch": 6.85, + "learning_rate": 1.068880543472991e-06, + "loss": 0.0017, + "step": 52274 + }, + { + "epoch": 6.85, + "learning_rate": 1.068641997536375e-06, + "loss": 0.0018, + "step": 52275 + }, + { + "epoch": 6.85, + "learning_rate": 1.068403476718728e-06, + "loss": 0.0023, + "step": 52276 + }, + { + "epoch": 6.85, + "learning_rate": 1.0681649810207173e-06, + "loss": 0.002, + "step": 52277 + }, + { + "epoch": 6.85, + "learning_rate": 1.0679265104430181e-06, + "loss": 0.0007, + "step": 52278 + }, + { + "epoch": 6.85, + "learning_rate": 1.067688064986302e-06, + "loss": 0.0028, + "step": 52279 + }, + { + "epoch": 6.85, + "learning_rate": 1.0674496446512373e-06, + "loss": 0.0009, + "step": 52280 + }, + { + "epoch": 6.85, + "learning_rate": 1.0672112494384945e-06, + "loss": 0.0013, + "step": 52281 + }, + { + "epoch": 6.85, + "learning_rate": 1.0669728793487422e-06, + "loss": 0.0021, + "step": 52282 + }, + { + "epoch": 6.85, + "learning_rate": 1.0667345343826552e-06, + "loss": 0.0011, + "step": 52283 + }, + { + "epoch": 6.85, + "learning_rate": 1.066496214540902e-06, + "loss": 0.0015, + "step": 52284 + }, + { + "epoch": 6.85, + "learning_rate": 1.0662579198241508e-06, + "loss": 0.002, + "step": 52285 + }, + { + "epoch": 6.85, + "learning_rate": 1.0660196502330744e-06, + "loss": 0.0015, + "step": 52286 + }, + { + "epoch": 6.85, + "learning_rate": 1.0657814057683423e-06, + "loss": 0.0015, + "step": 52287 + }, + { + "epoch": 6.85, + "learning_rate": 1.0655431864306243e-06, + "loss": 0.0026, + "step": 52288 + }, + { + "epoch": 6.85, + "learning_rate": 1.0653049922205882e-06, + "loss": 0.0018, + "step": 52289 + }, + { + "epoch": 6.85, + "learning_rate": 1.065066823138906e-06, + "loss": 0.0027, + "step": 52290 + }, + { + "epoch": 6.85, + "learning_rate": 1.0648286791862506e-06, + "loss": 0.0018, + "step": 52291 + }, + { + "epoch": 6.85, + "learning_rate": 1.0645905603632878e-06, + "loss": 0.0016, + "step": 52292 + }, + { + "epoch": 6.85, + "learning_rate": 1.064352466670686e-06, + "loss": 0.0012, + "step": 52293 + }, + { + "epoch": 6.85, + "learning_rate": 1.0641143981091195e-06, + "loss": 0.0014, + "step": 52294 + }, + { + "epoch": 6.85, + "learning_rate": 1.0638763546792553e-06, + "loss": 0.0012, + "step": 52295 + }, + { + "epoch": 6.85, + "learning_rate": 1.0636383363817626e-06, + "loss": 0.0021, + "step": 52296 + }, + { + "epoch": 6.85, + "learning_rate": 1.063400343217309e-06, + "loss": 0.0011, + "step": 52297 + }, + { + "epoch": 6.85, + "learning_rate": 1.0631623751865684e-06, + "loss": 0.0014, + "step": 52298 + }, + { + "epoch": 6.85, + "learning_rate": 1.0629244322902076e-06, + "loss": 0.0013, + "step": 52299 + }, + { + "epoch": 6.85, + "learning_rate": 1.0626865145288946e-06, + "loss": 0.0008, + "step": 52300 + }, + { + "epoch": 6.85, + "learning_rate": 1.0624486219032993e-06, + "loss": 0.001, + "step": 52301 + }, + { + "epoch": 6.85, + "learning_rate": 1.0622107544140937e-06, + "loss": 0.0022, + "step": 52302 + }, + { + "epoch": 6.85, + "learning_rate": 1.0619729120619437e-06, + "loss": 0.0015, + "step": 52303 + }, + { + "epoch": 6.85, + "learning_rate": 1.0617350948475191e-06, + "loss": 0.0014, + "step": 52304 + }, + { + "epoch": 6.85, + "learning_rate": 1.061497302771487e-06, + "loss": 0.0006, + "step": 52305 + }, + { + "epoch": 6.85, + "learning_rate": 1.061259535834519e-06, + "loss": 0.0014, + "step": 52306 + }, + { + "epoch": 6.85, + "learning_rate": 1.0610217940372825e-06, + "loss": 0.0018, + "step": 52307 + }, + { + "epoch": 6.85, + "learning_rate": 1.0607840773804444e-06, + "loss": 0.0031, + "step": 52308 + }, + { + "epoch": 6.85, + "learning_rate": 1.060546385864678e-06, + "loss": 0.0013, + "step": 52309 + }, + { + "epoch": 6.85, + "learning_rate": 1.060308719490648e-06, + "loss": 0.0016, + "step": 52310 + }, + { + "epoch": 6.85, + "learning_rate": 1.0600710782590218e-06, + "loss": 0.0018, + "step": 52311 + }, + { + "epoch": 6.85, + "learning_rate": 1.0598334621704699e-06, + "loss": 0.0016, + "step": 52312 + }, + { + "epoch": 6.85, + "learning_rate": 1.0595958712256616e-06, + "loss": 0.0008, + "step": 52313 + }, + { + "epoch": 6.85, + "learning_rate": 1.0593583054252643e-06, + "loss": 0.0024, + "step": 52314 + }, + { + "epoch": 6.85, + "learning_rate": 1.0591207647699452e-06, + "loss": 0.0023, + "step": 52315 + }, + { + "epoch": 6.85, + "learning_rate": 1.0588832492603706e-06, + "loss": 0.001, + "step": 52316 + }, + { + "epoch": 6.85, + "learning_rate": 1.0586457588972142e-06, + "loss": 0.0024, + "step": 52317 + }, + { + "epoch": 6.85, + "learning_rate": 1.058408293681139e-06, + "loss": 0.0007, + "step": 52318 + }, + { + "epoch": 6.85, + "learning_rate": 1.0581708536128156e-06, + "loss": 0.0032, + "step": 52319 + }, + { + "epoch": 6.85, + "learning_rate": 1.0579334386929075e-06, + "loss": 0.0017, + "step": 52320 + }, + { + "epoch": 6.85, + "learning_rate": 1.0576960489220867e-06, + "loss": 0.0016, + "step": 52321 + }, + { + "epoch": 6.85, + "learning_rate": 1.0574586843010193e-06, + "loss": 0.0016, + "step": 52322 + }, + { + "epoch": 6.85, + "learning_rate": 1.0572213448303737e-06, + "loss": 0.0018, + "step": 52323 + }, + { + "epoch": 6.85, + "learning_rate": 1.0569840305108158e-06, + "loss": 0.0017, + "step": 52324 + }, + { + "epoch": 6.85, + "learning_rate": 1.0567467413430155e-06, + "loss": 0.0012, + "step": 52325 + }, + { + "epoch": 6.85, + "learning_rate": 1.0565094773276386e-06, + "loss": 0.0009, + "step": 52326 + }, + { + "epoch": 6.85, + "learning_rate": 1.0562722384653513e-06, + "loss": 0.0015, + "step": 52327 + }, + { + "epoch": 6.85, + "learning_rate": 1.056035024756824e-06, + "loss": 0.0014, + "step": 52328 + }, + { + "epoch": 6.85, + "learning_rate": 1.055797836202721e-06, + "loss": 0.0018, + "step": 52329 + }, + { + "epoch": 6.85, + "learning_rate": 1.0555606728037103e-06, + "loss": 0.0012, + "step": 52330 + }, + { + "epoch": 6.85, + "learning_rate": 1.0553235345604585e-06, + "loss": 0.002, + "step": 52331 + }, + { + "epoch": 6.86, + "learning_rate": 1.0550864214736333e-06, + "loss": 0.0019, + "step": 52332 + }, + { + "epoch": 6.86, + "learning_rate": 1.054849333543899e-06, + "loss": 0.0011, + "step": 52333 + }, + { + "epoch": 6.86, + "learning_rate": 1.0546122707719275e-06, + "loss": 0.0013, + "step": 52334 + }, + { + "epoch": 6.86, + "learning_rate": 1.0543752331583813e-06, + "loss": 0.0012, + "step": 52335 + }, + { + "epoch": 6.86, + "learning_rate": 1.0541382207039286e-06, + "loss": 0.0011, + "step": 52336 + }, + { + "epoch": 6.86, + "learning_rate": 1.0539012334092369e-06, + "loss": 0.0017, + "step": 52337 + }, + { + "epoch": 6.86, + "learning_rate": 1.053664271274971e-06, + "loss": 0.0013, + "step": 52338 + }, + { + "epoch": 6.86, + "learning_rate": 1.0534273343017964e-06, + "loss": 0.0022, + "step": 52339 + }, + { + "epoch": 6.86, + "learning_rate": 1.0531904224903833e-06, + "loss": 0.0011, + "step": 52340 + }, + { + "epoch": 6.86, + "learning_rate": 1.0529535358413945e-06, + "loss": 0.0012, + "step": 52341 + }, + { + "epoch": 6.86, + "learning_rate": 1.0527166743554962e-06, + "loss": 0.0024, + "step": 52342 + }, + { + "epoch": 6.86, + "learning_rate": 1.0524798380333579e-06, + "loss": 0.0014, + "step": 52343 + }, + { + "epoch": 6.86, + "learning_rate": 1.0522430268756411e-06, + "loss": 0.0006, + "step": 52344 + }, + { + "epoch": 6.86, + "learning_rate": 1.0520062408830166e-06, + "loss": 0.0015, + "step": 52345 + }, + { + "epoch": 6.86, + "learning_rate": 1.0517694800561484e-06, + "loss": 0.0015, + "step": 52346 + }, + { + "epoch": 6.86, + "learning_rate": 1.0515327443957001e-06, + "loss": 0.0009, + "step": 52347 + }, + { + "epoch": 6.86, + "learning_rate": 1.0512960339023403e-06, + "loss": 0.0017, + "step": 52348 + }, + { + "epoch": 6.86, + "learning_rate": 1.051059348576735e-06, + "loss": 0.0014, + "step": 52349 + }, + { + "epoch": 6.86, + "learning_rate": 1.050822688419546e-06, + "loss": 0.0022, + "step": 52350 + }, + { + "epoch": 6.86, + "learning_rate": 1.050586053431444e-06, + "loss": 0.0011, + "step": 52351 + }, + { + "epoch": 6.86, + "learning_rate": 1.0503494436130924e-06, + "loss": 0.0019, + "step": 52352 + }, + { + "epoch": 6.86, + "learning_rate": 1.0501128589651555e-06, + "loss": 0.0015, + "step": 52353 + }, + { + "epoch": 6.86, + "learning_rate": 1.0498762994882982e-06, + "loss": 0.0015, + "step": 52354 + }, + { + "epoch": 6.86, + "learning_rate": 1.049639765183187e-06, + "loss": 0.0013, + "step": 52355 + }, + { + "epoch": 6.86, + "learning_rate": 1.0494032560504897e-06, + "loss": 0.0007, + "step": 52356 + }, + { + "epoch": 6.86, + "learning_rate": 1.0491667720908682e-06, + "loss": 0.0019, + "step": 52357 + }, + { + "epoch": 6.86, + "learning_rate": 1.0489303133049877e-06, + "loss": 0.0014, + "step": 52358 + }, + { + "epoch": 6.86, + "learning_rate": 1.0486938796935154e-06, + "loss": 0.0013, + "step": 52359 + }, + { + "epoch": 6.86, + "learning_rate": 1.048457471257115e-06, + "loss": 0.0009, + "step": 52360 + }, + { + "epoch": 6.86, + "learning_rate": 1.0482210879964505e-06, + "loss": 0.0019, + "step": 52361 + }, + { + "epoch": 6.86, + "learning_rate": 1.0479847299121858e-06, + "loss": 0.0018, + "step": 52362 + }, + { + "epoch": 6.86, + "learning_rate": 1.0477483970049895e-06, + "loss": 0.0007, + "step": 52363 + }, + { + "epoch": 6.86, + "learning_rate": 1.0475120892755242e-06, + "loss": 0.0023, + "step": 52364 + }, + { + "epoch": 6.86, + "learning_rate": 1.0472758067244527e-06, + "loss": 0.0023, + "step": 52365 + }, + { + "epoch": 6.86, + "learning_rate": 1.047039549352441e-06, + "loss": 0.0027, + "step": 52366 + }, + { + "epoch": 6.86, + "learning_rate": 1.0468033171601554e-06, + "loss": 0.0019, + "step": 52367 + }, + { + "epoch": 6.86, + "learning_rate": 1.04656711014826e-06, + "loss": 0.0021, + "step": 52368 + }, + { + "epoch": 6.86, + "learning_rate": 1.046330928317416e-06, + "loss": 0.0023, + "step": 52369 + }, + { + "epoch": 6.86, + "learning_rate": 1.0460947716682889e-06, + "loss": 0.0019, + "step": 52370 + }, + { + "epoch": 6.86, + "learning_rate": 1.0458586402015446e-06, + "loss": 0.0016, + "step": 52371 + }, + { + "epoch": 6.86, + "learning_rate": 1.045622533917846e-06, + "loss": 0.002, + "step": 52372 + }, + { + "epoch": 6.86, + "learning_rate": 1.0453864528178558e-06, + "loss": 0.0007, + "step": 52373 + }, + { + "epoch": 6.86, + "learning_rate": 1.0451503969022413e-06, + "loss": 0.0014, + "step": 52374 + }, + { + "epoch": 6.86, + "learning_rate": 1.0449143661716643e-06, + "loss": 0.0025, + "step": 52375 + }, + { + "epoch": 6.86, + "learning_rate": 1.0446783606267885e-06, + "loss": 0.0028, + "step": 52376 + }, + { + "epoch": 6.86, + "learning_rate": 1.0444423802682734e-06, + "loss": 0.0031, + "step": 52377 + }, + { + "epoch": 6.86, + "learning_rate": 1.044206425096792e-06, + "loss": 0.0015, + "step": 52378 + }, + { + "epoch": 6.86, + "learning_rate": 1.0439704951130026e-06, + "loss": 0.0016, + "step": 52379 + }, + { + "epoch": 6.86, + "learning_rate": 1.04373459031757e-06, + "loss": 0.0013, + "step": 52380 + }, + { + "epoch": 6.86, + "learning_rate": 1.0434987107111539e-06, + "loss": 0.0012, + "step": 52381 + }, + { + "epoch": 6.86, + "learning_rate": 1.0432628562944235e-06, + "loss": 0.0013, + "step": 52382 + }, + { + "epoch": 6.86, + "learning_rate": 1.0430270270680387e-06, + "loss": 0.0008, + "step": 52383 + }, + { + "epoch": 6.86, + "learning_rate": 1.0427912230326642e-06, + "loss": 0.0011, + "step": 52384 + }, + { + "epoch": 6.86, + "learning_rate": 1.0425554441889596e-06, + "loss": 0.0024, + "step": 52385 + }, + { + "epoch": 6.86, + "learning_rate": 1.0423196905375932e-06, + "loss": 0.0009, + "step": 52386 + }, + { + "epoch": 6.86, + "learning_rate": 1.0420839620792255e-06, + "loss": 0.0016, + "step": 52387 + }, + { + "epoch": 6.86, + "learning_rate": 1.0418482588145184e-06, + "loss": 0.0008, + "step": 52388 + }, + { + "epoch": 6.86, + "learning_rate": 1.0416125807441358e-06, + "loss": 0.0023, + "step": 52389 + }, + { + "epoch": 6.86, + "learning_rate": 1.0413769278687413e-06, + "loss": 0.0016, + "step": 52390 + }, + { + "epoch": 6.86, + "learning_rate": 1.041141300188999e-06, + "loss": 0.0018, + "step": 52391 + }, + { + "epoch": 6.86, + "learning_rate": 1.0409056977055665e-06, + "loss": 0.001, + "step": 52392 + }, + { + "epoch": 6.86, + "learning_rate": 1.0406701204191116e-06, + "loss": 0.0012, + "step": 52393 + }, + { + "epoch": 6.86, + "learning_rate": 1.0404345683302953e-06, + "loss": 0.0025, + "step": 52394 + }, + { + "epoch": 6.86, + "learning_rate": 1.04019904143978e-06, + "loss": 0.0017, + "step": 52395 + }, + { + "epoch": 6.86, + "learning_rate": 1.0399635397482255e-06, + "loss": 0.0027, + "step": 52396 + }, + { + "epoch": 6.86, + "learning_rate": 1.0397280632562978e-06, + "loss": 0.0015, + "step": 52397 + }, + { + "epoch": 6.86, + "learning_rate": 1.0394926119646565e-06, + "loss": 0.0022, + "step": 52398 + }, + { + "epoch": 6.86, + "learning_rate": 1.0392571858739664e-06, + "loss": 0.0011, + "step": 52399 + }, + { + "epoch": 6.86, + "learning_rate": 1.039021784984887e-06, + "loss": 0.0007, + "step": 52400 + }, + { + "epoch": 6.86, + "learning_rate": 1.0387864092980836e-06, + "loss": 0.0016, + "step": 52401 + }, + { + "epoch": 6.86, + "learning_rate": 1.0385510588142155e-06, + "loss": 0.0026, + "step": 52402 + }, + { + "epoch": 6.86, + "learning_rate": 1.0383157335339455e-06, + "loss": 0.0014, + "step": 52403 + }, + { + "epoch": 6.86, + "learning_rate": 1.038080433457934e-06, + "loss": 0.0033, + "step": 52404 + }, + { + "epoch": 6.86, + "learning_rate": 1.0378451585868453e-06, + "loss": 0.0022, + "step": 52405 + }, + { + "epoch": 6.86, + "learning_rate": 1.03760990892134e-06, + "loss": 0.0022, + "step": 52406 + }, + { + "epoch": 6.86, + "learning_rate": 1.037374684462078e-06, + "loss": 0.0017, + "step": 52407 + }, + { + "epoch": 6.87, + "learning_rate": 1.0371394852097238e-06, + "loss": 0.0016, + "step": 52408 + }, + { + "epoch": 6.87, + "learning_rate": 1.0369043111649358e-06, + "loss": 0.0017, + "step": 52409 + }, + { + "epoch": 6.87, + "learning_rate": 1.0366691623283787e-06, + "loss": 0.0014, + "step": 52410 + }, + { + "epoch": 6.87, + "learning_rate": 1.0364340387007122e-06, + "loss": 0.0018, + "step": 52411 + }, + { + "epoch": 6.87, + "learning_rate": 1.0361989402825968e-06, + "loss": 0.0027, + "step": 52412 + }, + { + "epoch": 6.87, + "learning_rate": 1.0359638670746952e-06, + "loss": 0.0006, + "step": 52413 + }, + { + "epoch": 6.87, + "learning_rate": 1.0357288190776682e-06, + "loss": 0.0015, + "step": 52414 + }, + { + "epoch": 6.87, + "learning_rate": 1.0354937962921751e-06, + "loss": 0.001, + "step": 52415 + }, + { + "epoch": 6.87, + "learning_rate": 1.03525879871888e-06, + "loss": 0.0021, + "step": 52416 + }, + { + "epoch": 6.87, + "learning_rate": 1.0350238263584422e-06, + "loss": 0.0022, + "step": 52417 + }, + { + "epoch": 6.87, + "learning_rate": 1.0347888792115224e-06, + "loss": 0.0025, + "step": 52418 + }, + { + "epoch": 6.87, + "learning_rate": 1.0345539572787788e-06, + "loss": 0.0025, + "step": 52419 + }, + { + "epoch": 6.87, + "learning_rate": 1.0343190605608755e-06, + "loss": 0.0018, + "step": 52420 + }, + { + "epoch": 6.87, + "learning_rate": 1.034084189058474e-06, + "loss": 0.0013, + "step": 52421 + }, + { + "epoch": 6.87, + "learning_rate": 1.033849342772234e-06, + "loss": 0.0016, + "step": 52422 + }, + { + "epoch": 6.87, + "learning_rate": 1.0336145217028127e-06, + "loss": 0.0016, + "step": 52423 + }, + { + "epoch": 6.87, + "learning_rate": 1.0333797258508748e-06, + "loss": 0.0013, + "step": 52424 + }, + { + "epoch": 6.87, + "learning_rate": 1.0331449552170791e-06, + "loss": 0.0027, + "step": 52425 + }, + { + "epoch": 6.87, + "learning_rate": 1.032910209802086e-06, + "loss": 0.0017, + "step": 52426 + }, + { + "epoch": 6.87, + "learning_rate": 1.0326754896065527e-06, + "loss": 0.0017, + "step": 52427 + }, + { + "epoch": 6.87, + "learning_rate": 1.0324407946311454e-06, + "loss": 0.001, + "step": 52428 + }, + { + "epoch": 6.87, + "learning_rate": 1.0322061248765192e-06, + "loss": 0.001, + "step": 52429 + }, + { + "epoch": 6.87, + "learning_rate": 1.0319714803433346e-06, + "loss": 0.0012, + "step": 52430 + }, + { + "epoch": 6.87, + "learning_rate": 1.031736861032253e-06, + "loss": 0.0016, + "step": 52431 + }, + { + "epoch": 6.87, + "learning_rate": 1.0315022669439356e-06, + "loss": 0.0021, + "step": 52432 + }, + { + "epoch": 6.87, + "learning_rate": 1.0312676980790404e-06, + "loss": 0.0012, + "step": 52433 + }, + { + "epoch": 6.87, + "learning_rate": 1.0310331544382268e-06, + "loss": 0.0022, + "step": 52434 + }, + { + "epoch": 6.87, + "learning_rate": 1.0307986360221534e-06, + "loss": 0.0024, + "step": 52435 + }, + { + "epoch": 6.87, + "learning_rate": 1.0305641428314828e-06, + "loss": 0.0015, + "step": 52436 + }, + { + "epoch": 6.87, + "learning_rate": 1.0303296748668723e-06, + "loss": 0.0011, + "step": 52437 + }, + { + "epoch": 6.87, + "learning_rate": 1.0300952321289815e-06, + "loss": 0.0013, + "step": 52438 + }, + { + "epoch": 6.87, + "learning_rate": 1.0298608146184707e-06, + "loss": 0.0031, + "step": 52439 + }, + { + "epoch": 6.87, + "learning_rate": 1.0296264223359997e-06, + "loss": 0.0018, + "step": 52440 + }, + { + "epoch": 6.87, + "learning_rate": 1.0293920552822256e-06, + "loss": 0.0024, + "step": 52441 + }, + { + "epoch": 6.87, + "learning_rate": 1.0291577134578057e-06, + "loss": 0.001, + "step": 52442 + }, + { + "epoch": 6.87, + "learning_rate": 1.028923396863406e-06, + "loss": 0.0012, + "step": 52443 + }, + { + "epoch": 6.87, + "learning_rate": 1.0286891054996806e-06, + "loss": 0.0017, + "step": 52444 + }, + { + "epoch": 6.87, + "learning_rate": 1.02845483936729e-06, + "loss": 0.0016, + "step": 52445 + }, + { + "epoch": 6.87, + "learning_rate": 1.0282205984668902e-06, + "loss": 0.001, + "step": 52446 + }, + { + "epoch": 6.87, + "learning_rate": 1.0279863827991443e-06, + "loss": 0.0016, + "step": 52447 + }, + { + "epoch": 6.87, + "learning_rate": 1.0277521923647094e-06, + "loss": 0.0016, + "step": 52448 + }, + { + "epoch": 6.87, + "learning_rate": 1.0275180271642405e-06, + "loss": 0.0011, + "step": 52449 + }, + { + "epoch": 6.87, + "learning_rate": 1.0272838871984026e-06, + "loss": 0.0013, + "step": 52450 + }, + { + "epoch": 6.87, + "learning_rate": 1.0270497724678497e-06, + "loss": 0.0036, + "step": 52451 + }, + { + "epoch": 6.87, + "learning_rate": 1.0268156829732423e-06, + "loss": 0.0019, + "step": 52452 + }, + { + "epoch": 6.87, + "learning_rate": 1.0265816187152355e-06, + "loss": 0.0024, + "step": 52453 + }, + { + "epoch": 6.87, + "learning_rate": 1.0263475796944899e-06, + "loss": 0.0013, + "step": 52454 + }, + { + "epoch": 6.87, + "learning_rate": 1.0261135659116673e-06, + "loss": 0.0012, + "step": 52455 + }, + { + "epoch": 6.87, + "learning_rate": 1.0258795773674202e-06, + "loss": 0.0013, + "step": 52456 + }, + { + "epoch": 6.87, + "learning_rate": 1.0256456140624082e-06, + "loss": 0.0005, + "step": 52457 + }, + { + "epoch": 6.87, + "learning_rate": 1.025411675997292e-06, + "loss": 0.0025, + "step": 52458 + }, + { + "epoch": 6.87, + "learning_rate": 1.0251777631727279e-06, + "loss": 0.0021, + "step": 52459 + }, + { + "epoch": 6.87, + "learning_rate": 1.0249438755893727e-06, + "loss": 0.0019, + "step": 52460 + }, + { + "epoch": 6.87, + "learning_rate": 1.0247100132478826e-06, + "loss": 0.0016, + "step": 52461 + }, + { + "epoch": 6.87, + "learning_rate": 1.0244761761489197e-06, + "loss": 0.0019, + "step": 52462 + }, + { + "epoch": 6.87, + "learning_rate": 1.024242364293141e-06, + "loss": 0.0018, + "step": 52463 + }, + { + "epoch": 6.87, + "learning_rate": 1.0240085776811993e-06, + "loss": 0.0009, + "step": 52464 + }, + { + "epoch": 6.87, + "learning_rate": 1.0237748163137552e-06, + "loss": 0.0019, + "step": 52465 + }, + { + "epoch": 6.87, + "learning_rate": 1.0235410801914692e-06, + "loss": 0.0018, + "step": 52466 + }, + { + "epoch": 6.87, + "learning_rate": 1.0233073693149965e-06, + "loss": 0.0012, + "step": 52467 + }, + { + "epoch": 6.87, + "learning_rate": 1.0230736836849919e-06, + "loss": 0.0019, + "step": 52468 + }, + { + "epoch": 6.87, + "learning_rate": 1.022840023302114e-06, + "loss": 0.0017, + "step": 52469 + }, + { + "epoch": 6.87, + "learning_rate": 1.0226063881670222e-06, + "loss": 0.0015, + "step": 52470 + }, + { + "epoch": 6.87, + "learning_rate": 1.0223727782803716e-06, + "loss": 0.0009, + "step": 52471 + }, + { + "epoch": 6.87, + "learning_rate": 1.0221391936428183e-06, + "loss": 0.0012, + "step": 52472 + }, + { + "epoch": 6.87, + "learning_rate": 1.0219056342550215e-06, + "loss": 0.0016, + "step": 52473 + }, + { + "epoch": 6.87, + "learning_rate": 1.0216721001176367e-06, + "loss": 0.0009, + "step": 52474 + }, + { + "epoch": 6.87, + "learning_rate": 1.0214385912313196e-06, + "loss": 0.0011, + "step": 52475 + }, + { + "epoch": 6.87, + "learning_rate": 1.02120510759673e-06, + "loss": 0.0015, + "step": 52476 + }, + { + "epoch": 6.87, + "learning_rate": 1.0209716492145217e-06, + "loss": 0.0013, + "step": 52477 + }, + { + "epoch": 6.87, + "learning_rate": 1.020738216085354e-06, + "loss": 0.0013, + "step": 52478 + }, + { + "epoch": 6.87, + "learning_rate": 1.020504808209881e-06, + "loss": 0.0019, + "step": 52479 + }, + { + "epoch": 6.87, + "learning_rate": 1.020271425588759e-06, + "loss": 0.0016, + "step": 52480 + }, + { + "epoch": 6.87, + "learning_rate": 1.0200380682226475e-06, + "loss": 0.0011, + "step": 52481 + }, + { + "epoch": 6.87, + "learning_rate": 1.0198047361121999e-06, + "loss": 0.0016, + "step": 52482 + }, + { + "epoch": 6.87, + "learning_rate": 1.0195714292580737e-06, + "loss": 0.0014, + "step": 52483 + }, + { + "epoch": 6.88, + "learning_rate": 1.0193381476609232e-06, + "loss": 0.0019, + "step": 52484 + }, + { + "epoch": 6.88, + "learning_rate": 1.0191048913214075e-06, + "loss": 0.0023, + "step": 52485 + }, + { + "epoch": 6.88, + "learning_rate": 1.0188716602401794e-06, + "loss": 0.0015, + "step": 52486 + }, + { + "epoch": 6.88, + "learning_rate": 1.0186384544178973e-06, + "loss": 0.0019, + "step": 52487 + }, + { + "epoch": 6.88, + "learning_rate": 1.0184052738552153e-06, + "loss": 0.002, + "step": 52488 + }, + { + "epoch": 6.88, + "learning_rate": 1.0181721185527927e-06, + "loss": 0.0028, + "step": 52489 + }, + { + "epoch": 6.88, + "learning_rate": 1.0179389885112812e-06, + "loss": 0.0016, + "step": 52490 + }, + { + "epoch": 6.88, + "learning_rate": 1.0177058837313391e-06, + "loss": 0.0012, + "step": 52491 + }, + { + "epoch": 6.88, + "learning_rate": 1.0174728042136183e-06, + "loss": 0.002, + "step": 52492 + }, + { + "epoch": 6.88, + "learning_rate": 1.0172397499587793e-06, + "loss": 0.0012, + "step": 52493 + }, + { + "epoch": 6.88, + "learning_rate": 1.0170067209674738e-06, + "loss": 0.0012, + "step": 52494 + }, + { + "epoch": 6.88, + "learning_rate": 1.016773717240358e-06, + "loss": 0.0015, + "step": 52495 + }, + { + "epoch": 6.88, + "learning_rate": 1.016540738778089e-06, + "loss": 0.0009, + "step": 52496 + }, + { + "epoch": 6.88, + "learning_rate": 1.0163077855813196e-06, + "loss": 0.0022, + "step": 52497 + }, + { + "epoch": 6.88, + "learning_rate": 1.0160748576507073e-06, + "loss": 0.0013, + "step": 52498 + }, + { + "epoch": 6.88, + "learning_rate": 1.0158419549869048e-06, + "loss": 0.0014, + "step": 52499 + }, + { + "epoch": 6.88, + "learning_rate": 1.0156090775905703e-06, + "loss": 0.0027, + "step": 52500 + }, + { + "epoch": 6.88, + "learning_rate": 1.0153762254623566e-06, + "loss": 0.0012, + "step": 52501 + }, + { + "epoch": 6.88, + "learning_rate": 1.015143398602918e-06, + "loss": 0.0013, + "step": 52502 + }, + { + "epoch": 6.88, + "learning_rate": 1.0149105970129092e-06, + "loss": 0.0021, + "step": 52503 + }, + { + "epoch": 6.88, + "learning_rate": 1.0146778206929875e-06, + "loss": 0.0009, + "step": 52504 + }, + { + "epoch": 6.88, + "learning_rate": 1.0144450696438057e-06, + "loss": 0.0008, + "step": 52505 + }, + { + "epoch": 6.88, + "learning_rate": 1.014212343866019e-06, + "loss": 0.0017, + "step": 52506 + }, + { + "epoch": 6.88, + "learning_rate": 1.0139796433602788e-06, + "loss": 0.0009, + "step": 52507 + }, + { + "epoch": 6.88, + "learning_rate": 1.0137469681272439e-06, + "loss": 0.0013, + "step": 52508 + }, + { + "epoch": 6.88, + "learning_rate": 1.0135143181675676e-06, + "loss": 0.0027, + "step": 52509 + }, + { + "epoch": 6.88, + "learning_rate": 1.0132816934819044e-06, + "loss": 0.0014, + "step": 52510 + }, + { + "epoch": 6.88, + "learning_rate": 1.0130490940709058e-06, + "loss": 0.002, + "step": 52511 + }, + { + "epoch": 6.88, + "learning_rate": 1.0128165199352302e-06, + "loss": 0.0026, + "step": 52512 + }, + { + "epoch": 6.88, + "learning_rate": 1.0125839710755291e-06, + "loss": 0.0022, + "step": 52513 + }, + { + "epoch": 6.88, + "learning_rate": 1.0123514474924546e-06, + "loss": 0.0018, + "step": 52514 + }, + { + "epoch": 6.88, + "learning_rate": 1.0121189491866656e-06, + "loss": 0.0018, + "step": 52515 + }, + { + "epoch": 6.88, + "learning_rate": 1.0118864761588132e-06, + "loss": 0.0013, + "step": 52516 + }, + { + "epoch": 6.88, + "learning_rate": 1.0116540284095522e-06, + "loss": 0.0014, + "step": 52517 + }, + { + "epoch": 6.88, + "learning_rate": 1.0114216059395333e-06, + "loss": 0.0022, + "step": 52518 + }, + { + "epoch": 6.88, + "learning_rate": 1.0111892087494124e-06, + "loss": 0.001, + "step": 52519 + }, + { + "epoch": 6.88, + "learning_rate": 1.0109568368398449e-06, + "loss": 0.0019, + "step": 52520 + }, + { + "epoch": 6.88, + "learning_rate": 1.0107244902114821e-06, + "loss": 0.002, + "step": 52521 + }, + { + "epoch": 6.88, + "learning_rate": 1.0104921688649772e-06, + "loss": 0.0011, + "step": 52522 + }, + { + "epoch": 6.88, + "learning_rate": 1.0102598728009849e-06, + "loss": 0.0018, + "step": 52523 + }, + { + "epoch": 6.88, + "learning_rate": 1.0100276020201594e-06, + "loss": 0.0015, + "step": 52524 + }, + { + "epoch": 6.88, + "learning_rate": 1.0097953565231523e-06, + "loss": 0.002, + "step": 52525 + }, + { + "epoch": 6.88, + "learning_rate": 1.009563136310614e-06, + "loss": 0.0016, + "step": 52526 + }, + { + "epoch": 6.88, + "learning_rate": 1.0093309413832031e-06, + "loss": 0.0021, + "step": 52527 + }, + { + "epoch": 6.88, + "learning_rate": 1.00909877174157e-06, + "loss": 0.0013, + "step": 52528 + }, + { + "epoch": 6.88, + "learning_rate": 1.0088666273863668e-06, + "loss": 0.0022, + "step": 52529 + }, + { + "epoch": 6.88, + "learning_rate": 1.008634508318247e-06, + "loss": 0.0018, + "step": 52530 + }, + { + "epoch": 6.88, + "learning_rate": 1.0084024145378647e-06, + "loss": 0.0023, + "step": 52531 + }, + { + "epoch": 6.88, + "learning_rate": 1.0081703460458725e-06, + "loss": 0.001, + "step": 52532 + }, + { + "epoch": 6.88, + "learning_rate": 1.0079383028429224e-06, + "loss": 0.0009, + "step": 52533 + }, + { + "epoch": 6.88, + "learning_rate": 1.0077062849296649e-06, + "loss": 0.0009, + "step": 52534 + }, + { + "epoch": 6.88, + "learning_rate": 1.007474292306757e-06, + "loss": 0.0016, + "step": 52535 + }, + { + "epoch": 6.88, + "learning_rate": 1.0072423249748486e-06, + "loss": 0.0017, + "step": 52536 + }, + { + "epoch": 6.88, + "learning_rate": 1.0070103829345901e-06, + "loss": 0.0009, + "step": 52537 + }, + { + "epoch": 6.88, + "learning_rate": 1.0067784661866387e-06, + "loss": 0.0017, + "step": 52538 + }, + { + "epoch": 6.88, + "learning_rate": 1.0065465747316427e-06, + "loss": 0.0014, + "step": 52539 + }, + { + "epoch": 6.88, + "learning_rate": 1.0063147085702541e-06, + "loss": 0.0022, + "step": 52540 + }, + { + "epoch": 6.88, + "learning_rate": 1.0060828677031286e-06, + "loss": 0.0018, + "step": 52541 + }, + { + "epoch": 6.88, + "learning_rate": 1.0058510521309139e-06, + "loss": 0.0015, + "step": 52542 + }, + { + "epoch": 6.88, + "learning_rate": 1.005619261854266e-06, + "loss": 0.001, + "step": 52543 + }, + { + "epoch": 6.88, + "learning_rate": 1.0053874968738353e-06, + "loss": 0.0011, + "step": 52544 + }, + { + "epoch": 6.88, + "learning_rate": 1.0051557571902714e-06, + "loss": 0.0015, + "step": 52545 + }, + { + "epoch": 6.88, + "learning_rate": 1.0049240428042295e-06, + "loss": 0.0008, + "step": 52546 + }, + { + "epoch": 6.88, + "learning_rate": 1.00469235371636e-06, + "loss": 0.0024, + "step": 52547 + }, + { + "epoch": 6.88, + "learning_rate": 1.0044606899273136e-06, + "loss": 0.0021, + "step": 52548 + }, + { + "epoch": 6.88, + "learning_rate": 1.0042290514377418e-06, + "loss": 0.0012, + "step": 52549 + }, + { + "epoch": 6.88, + "learning_rate": 1.0039974382482976e-06, + "loss": 0.0013, + "step": 52550 + }, + { + "epoch": 6.88, + "learning_rate": 1.0037658503596293e-06, + "loss": 0.0011, + "step": 52551 + }, + { + "epoch": 6.88, + "learning_rate": 1.003534287772393e-06, + "loss": 0.0015, + "step": 52552 + }, + { + "epoch": 6.88, + "learning_rate": 1.0033027504872362e-06, + "loss": 0.001, + "step": 52553 + }, + { + "epoch": 6.88, + "learning_rate": 1.0030712385048124e-06, + "loss": 0.0016, + "step": 52554 + }, + { + "epoch": 6.88, + "learning_rate": 1.0028397518257716e-06, + "loss": 0.0023, + "step": 52555 + }, + { + "epoch": 6.88, + "learning_rate": 1.002608290450765e-06, + "loss": 0.0016, + "step": 52556 + }, + { + "epoch": 6.88, + "learning_rate": 1.0023768543804413e-06, + "loss": 0.0011, + "step": 52557 + }, + { + "epoch": 6.88, + "learning_rate": 1.0021454436154553e-06, + "loss": 0.0014, + "step": 52558 + }, + { + "epoch": 6.88, + "learning_rate": 1.0019140581564568e-06, + "loss": 0.0011, + "step": 52559 + }, + { + "epoch": 6.88, + "learning_rate": 1.001682698004094e-06, + "loss": 0.0011, + "step": 52560 + }, + { + "epoch": 6.89, + "learning_rate": 1.0014513631590206e-06, + "loss": 0.0013, + "step": 52561 + }, + { + "epoch": 6.89, + "learning_rate": 1.0012200536218842e-06, + "loss": 0.002, + "step": 52562 + }, + { + "epoch": 6.89, + "learning_rate": 1.0009887693933395e-06, + "loss": 0.0014, + "step": 52563 + }, + { + "epoch": 6.89, + "learning_rate": 1.000757510474033e-06, + "loss": 0.0018, + "step": 52564 + }, + { + "epoch": 6.89, + "learning_rate": 1.0005262768646184e-06, + "loss": 0.0018, + "step": 52565 + }, + { + "epoch": 6.89, + "learning_rate": 1.0002950685657442e-06, + "loss": 0.0007, + "step": 52566 + }, + { + "epoch": 6.89, + "learning_rate": 1.000063885578061e-06, + "loss": 0.0028, + "step": 52567 + }, + { + "epoch": 6.89, + "learning_rate": 9.99832727902217e-07, + "loss": 0.0017, + "step": 52568 + }, + { + "epoch": 6.89, + "learning_rate": 9.99601595538866e-07, + "loss": 0.0015, + "step": 52569 + }, + { + "epoch": 6.89, + "learning_rate": 9.993704884886556e-07, + "loss": 0.0015, + "step": 52570 + }, + { + "epoch": 6.89, + "learning_rate": 9.991394067522376e-07, + "loss": 0.0024, + "step": 52571 + }, + { + "epoch": 6.89, + "learning_rate": 9.989083503302576e-07, + "loss": 0.0014, + "step": 52572 + }, + { + "epoch": 6.89, + "learning_rate": 9.98677319223369e-07, + "loss": 0.0014, + "step": 52573 + }, + { + "epoch": 6.89, + "learning_rate": 9.984463134322231e-07, + "loss": 0.0016, + "step": 52574 + }, + { + "epoch": 6.89, + "learning_rate": 9.982153329574674e-07, + "loss": 0.0019, + "step": 52575 + }, + { + "epoch": 6.89, + "learning_rate": 9.9798437779975e-07, + "loss": 0.0015, + "step": 52576 + }, + { + "epoch": 6.89, + "learning_rate": 9.97753447959724e-07, + "loss": 0.001, + "step": 52577 + }, + { + "epoch": 6.89, + "learning_rate": 9.975225434380375e-07, + "loss": 0.0011, + "step": 52578 + }, + { + "epoch": 6.89, + "learning_rate": 9.972916642353369e-07, + "loss": 0.0013, + "step": 52579 + }, + { + "epoch": 6.89, + "learning_rate": 9.97060810352276e-07, + "loss": 0.0014, + "step": 52580 + }, + { + "epoch": 6.89, + "learning_rate": 9.968299817895022e-07, + "loss": 0.0014, + "step": 52581 + }, + { + "epoch": 6.89, + "learning_rate": 9.965991785476647e-07, + "loss": 0.0015, + "step": 52582 + }, + { + "epoch": 6.89, + "learning_rate": 9.963684006274111e-07, + "loss": 0.0018, + "step": 52583 + }, + { + "epoch": 6.89, + "learning_rate": 9.961376480293904e-07, + "loss": 0.0016, + "step": 52584 + }, + { + "epoch": 6.89, + "learning_rate": 9.959069207542572e-07, + "loss": 0.001, + "step": 52585 + }, + { + "epoch": 6.89, + "learning_rate": 9.956762188026547e-07, + "loss": 0.001, + "step": 52586 + }, + { + "epoch": 6.89, + "learning_rate": 9.95445542175232e-07, + "loss": 0.0019, + "step": 52587 + }, + { + "epoch": 6.89, + "learning_rate": 9.952148908726411e-07, + "loss": 0.0013, + "step": 52588 + }, + { + "epoch": 6.89, + "learning_rate": 9.949842648955289e-07, + "loss": 0.0011, + "step": 52589 + }, + { + "epoch": 6.89, + "learning_rate": 9.947536642445433e-07, + "loss": 0.0037, + "step": 52590 + }, + { + "epoch": 6.89, + "learning_rate": 9.945230889203317e-07, + "loss": 0.0019, + "step": 52591 + }, + { + "epoch": 6.89, + "learning_rate": 9.942925389235458e-07, + "loss": 0.0015, + "step": 52592 + }, + { + "epoch": 6.89, + "learning_rate": 9.94062014254834e-07, + "loss": 0.0022, + "step": 52593 + }, + { + "epoch": 6.89, + "learning_rate": 9.938315149148404e-07, + "loss": 0.0019, + "step": 52594 + }, + { + "epoch": 6.89, + "learning_rate": 9.936010409042162e-07, + "loss": 0.0014, + "step": 52595 + }, + { + "epoch": 6.89, + "learning_rate": 9.933705922236113e-07, + "loss": 0.0027, + "step": 52596 + }, + { + "epoch": 6.89, + "learning_rate": 9.93140168873672e-07, + "loss": 0.0013, + "step": 52597 + }, + { + "epoch": 6.89, + "learning_rate": 9.929097708550462e-07, + "loss": 0.0016, + "step": 52598 + }, + { + "epoch": 6.89, + "learning_rate": 9.926793981683803e-07, + "loss": 0.0008, + "step": 52599 + }, + { + "epoch": 6.89, + "learning_rate": 9.92449050814326e-07, + "loss": 0.0015, + "step": 52600 + }, + { + "epoch": 6.89, + "learning_rate": 9.922187287935282e-07, + "loss": 0.0018, + "step": 52601 + }, + { + "epoch": 6.89, + "learning_rate": 9.919884321066342e-07, + "loss": 0.0012, + "step": 52602 + }, + { + "epoch": 6.89, + "learning_rate": 9.917581607542947e-07, + "loss": 0.0025, + "step": 52603 + }, + { + "epoch": 6.89, + "learning_rate": 9.915279147371558e-07, + "loss": 0.0022, + "step": 52604 + }, + { + "epoch": 6.89, + "learning_rate": 9.91297694055866e-07, + "loss": 0.0007, + "step": 52605 + }, + { + "epoch": 6.89, + "learning_rate": 9.91067498711068e-07, + "loss": 0.0017, + "step": 52606 + }, + { + "epoch": 6.89, + "learning_rate": 9.908373287034145e-07, + "loss": 0.0006, + "step": 52607 + }, + { + "epoch": 6.89, + "learning_rate": 9.90607184033552e-07, + "loss": 0.0013, + "step": 52608 + }, + { + "epoch": 6.89, + "learning_rate": 9.90377064702127e-07, + "loss": 0.0029, + "step": 52609 + }, + { + "epoch": 6.89, + "learning_rate": 9.901469707097854e-07, + "loss": 0.0014, + "step": 52610 + }, + { + "epoch": 6.89, + "learning_rate": 9.899169020571774e-07, + "loss": 0.0009, + "step": 52611 + }, + { + "epoch": 6.89, + "learning_rate": 9.896868587449481e-07, + "loss": 0.0015, + "step": 52612 + }, + { + "epoch": 6.89, + "learning_rate": 9.894568407737448e-07, + "loss": 0.0012, + "step": 52613 + }, + { + "epoch": 6.89, + "learning_rate": 9.892268481442124e-07, + "loss": 0.0016, + "step": 52614 + }, + { + "epoch": 6.89, + "learning_rate": 9.889968808570016e-07, + "loss": 0.0009, + "step": 52615 + }, + { + "epoch": 6.89, + "learning_rate": 9.887669389127575e-07, + "loss": 0.001, + "step": 52616 + }, + { + "epoch": 6.89, + "learning_rate": 9.885370223121248e-07, + "loss": 0.0009, + "step": 52617 + }, + { + "epoch": 6.89, + "learning_rate": 9.883071310557513e-07, + "loss": 0.0022, + "step": 52618 + }, + { + "epoch": 6.89, + "learning_rate": 9.88077265144286e-07, + "loss": 0.0016, + "step": 52619 + }, + { + "epoch": 6.89, + "learning_rate": 9.878474245783731e-07, + "loss": 0.0017, + "step": 52620 + }, + { + "epoch": 6.89, + "learning_rate": 9.876176093586597e-07, + "loss": 0.0015, + "step": 52621 + }, + { + "epoch": 6.89, + "learning_rate": 9.8738781948579e-07, + "loss": 0.0017, + "step": 52622 + }, + { + "epoch": 6.89, + "learning_rate": 9.871580549604144e-07, + "loss": 0.0015, + "step": 52623 + }, + { + "epoch": 6.89, + "learning_rate": 9.869283157831765e-07, + "loss": 0.0018, + "step": 52624 + }, + { + "epoch": 6.89, + "learning_rate": 9.866986019547208e-07, + "loss": 0.0018, + "step": 52625 + }, + { + "epoch": 6.89, + "learning_rate": 9.864689134756967e-07, + "loss": 0.0005, + "step": 52626 + }, + { + "epoch": 6.89, + "learning_rate": 9.8623925034675e-07, + "loss": 0.001, + "step": 52627 + }, + { + "epoch": 6.89, + "learning_rate": 9.860096125685225e-07, + "loss": 0.0013, + "step": 52628 + }, + { + "epoch": 6.89, + "learning_rate": 9.85780000141663e-07, + "loss": 0.0019, + "step": 52629 + }, + { + "epoch": 6.89, + "learning_rate": 9.855504130668192e-07, + "loss": 0.0016, + "step": 52630 + }, + { + "epoch": 6.89, + "learning_rate": 9.853208513446355e-07, + "loss": 0.0016, + "step": 52631 + }, + { + "epoch": 6.89, + "learning_rate": 9.850913149757556e-07, + "loss": 0.0019, + "step": 52632 + }, + { + "epoch": 6.89, + "learning_rate": 9.84861803960825e-07, + "loss": 0.0012, + "step": 52633 + }, + { + "epoch": 6.89, + "learning_rate": 9.846323183004923e-07, + "loss": 0.0016, + "step": 52634 + }, + { + "epoch": 6.89, + "learning_rate": 9.844028579954013e-07, + "loss": 0.0013, + "step": 52635 + }, + { + "epoch": 6.89, + "learning_rate": 9.841734230461975e-07, + "loss": 0.0013, + "step": 52636 + }, + { + "epoch": 6.9, + "learning_rate": 9.839440134535238e-07, + "loss": 0.0017, + "step": 52637 + }, + { + "epoch": 6.9, + "learning_rate": 9.837146292180288e-07, + "loss": 0.0026, + "step": 52638 + }, + { + "epoch": 6.9, + "learning_rate": 9.834852703403553e-07, + "loss": 0.0014, + "step": 52639 + }, + { + "epoch": 6.9, + "learning_rate": 9.832559368211504e-07, + "loss": 0.0032, + "step": 52640 + }, + { + "epoch": 6.9, + "learning_rate": 9.830266286610556e-07, + "loss": 0.0014, + "step": 52641 + }, + { + "epoch": 6.9, + "learning_rate": 9.82797345860721e-07, + "loss": 0.0016, + "step": 52642 + }, + { + "epoch": 6.9, + "learning_rate": 9.825680884207888e-07, + "loss": 0.0018, + "step": 52643 + }, + { + "epoch": 6.9, + "learning_rate": 9.823388563419013e-07, + "loss": 0.0026, + "step": 52644 + }, + { + "epoch": 6.9, + "learning_rate": 9.821096496247074e-07, + "loss": 0.0028, + "step": 52645 + }, + { + "epoch": 6.9, + "learning_rate": 9.818804682698502e-07, + "loss": 0.0018, + "step": 52646 + }, + { + "epoch": 6.9, + "learning_rate": 9.816513122779736e-07, + "loss": 0.0007, + "step": 52647 + }, + { + "epoch": 6.9, + "learning_rate": 9.814221816497206e-07, + "loss": 0.0014, + "step": 52648 + }, + { + "epoch": 6.9, + "learning_rate": 9.811930763857402e-07, + "loss": 0.0022, + "step": 52649 + }, + { + "epoch": 6.9, + "learning_rate": 9.809639964866713e-07, + "loss": 0.001, + "step": 52650 + }, + { + "epoch": 6.9, + "learning_rate": 9.80734941953162e-07, + "loss": 0.0009, + "step": 52651 + }, + { + "epoch": 6.9, + "learning_rate": 9.805059127858552e-07, + "loss": 0.0015, + "step": 52652 + }, + { + "epoch": 6.9, + "learning_rate": 9.802769089853958e-07, + "loss": 0.0013, + "step": 52653 + }, + { + "epoch": 6.9, + "learning_rate": 9.80047930552428e-07, + "loss": 0.0008, + "step": 52654 + }, + { + "epoch": 6.9, + "learning_rate": 9.798189774875943e-07, + "loss": 0.0013, + "step": 52655 + }, + { + "epoch": 6.9, + "learning_rate": 9.795900497915378e-07, + "loss": 0.0021, + "step": 52656 + }, + { + "epoch": 6.9, + "learning_rate": 9.793611474649056e-07, + "loss": 0.003, + "step": 52657 + }, + { + "epoch": 6.9, + "learning_rate": 9.791322705083405e-07, + "loss": 0.0005, + "step": 52658 + }, + { + "epoch": 6.9, + "learning_rate": 9.789034189224832e-07, + "loss": 0.0012, + "step": 52659 + }, + { + "epoch": 6.9, + "learning_rate": 9.7867459270798e-07, + "loss": 0.0013, + "step": 52660 + }, + { + "epoch": 6.9, + "learning_rate": 9.784457918654754e-07, + "loss": 0.0018, + "step": 52661 + }, + { + "epoch": 6.9, + "learning_rate": 9.782170163956117e-07, + "loss": 0.0023, + "step": 52662 + }, + { + "epoch": 6.9, + "learning_rate": 9.779882662990325e-07, + "loss": 0.0014, + "step": 52663 + }, + { + "epoch": 6.9, + "learning_rate": 9.777595415763785e-07, + "loss": 0.002, + "step": 52664 + }, + { + "epoch": 6.9, + "learning_rate": 9.77530842228297e-07, + "loss": 0.0021, + "step": 52665 + }, + { + "epoch": 6.9, + "learning_rate": 9.773021682554306e-07, + "loss": 0.0032, + "step": 52666 + }, + { + "epoch": 6.9, + "learning_rate": 9.770735196584192e-07, + "loss": 0.0038, + "step": 52667 + }, + { + "epoch": 6.9, + "learning_rate": 9.768448964379096e-07, + "loss": 0.002, + "step": 52668 + }, + { + "epoch": 6.9, + "learning_rate": 9.766162985945427e-07, + "loss": 0.0009, + "step": 52669 + }, + { + "epoch": 6.9, + "learning_rate": 9.763877261289623e-07, + "loss": 0.0009, + "step": 52670 + }, + { + "epoch": 6.9, + "learning_rate": 9.76159179041809e-07, + "loss": 0.0018, + "step": 52671 + }, + { + "epoch": 6.9, + "learning_rate": 9.75930657333728e-07, + "loss": 0.0009, + "step": 52672 + }, + { + "epoch": 6.9, + "learning_rate": 9.75702161005363e-07, + "loss": 0.0019, + "step": 52673 + }, + { + "epoch": 6.9, + "learning_rate": 9.754736900573548e-07, + "loss": 0.0015, + "step": 52674 + }, + { + "epoch": 6.9, + "learning_rate": 9.752452444903448e-07, + "loss": 0.0014, + "step": 52675 + }, + { + "epoch": 6.9, + "learning_rate": 9.750168243049785e-07, + "loss": 0.0014, + "step": 52676 + }, + { + "epoch": 6.9, + "learning_rate": 9.747884295018973e-07, + "loss": 0.0017, + "step": 52677 + }, + { + "epoch": 6.9, + "learning_rate": 9.745600600817428e-07, + "loss": 0.0016, + "step": 52678 + }, + { + "epoch": 6.9, + "learning_rate": 9.74331716045156e-07, + "loss": 0.0021, + "step": 52679 + }, + { + "epoch": 6.9, + "learning_rate": 9.741033973927815e-07, + "loss": 0.0013, + "step": 52680 + }, + { + "epoch": 6.9, + "learning_rate": 9.738751041252614e-07, + "loss": 0.0015, + "step": 52681 + }, + { + "epoch": 6.9, + "learning_rate": 9.73646836243234e-07, + "loss": 0.0024, + "step": 52682 + }, + { + "epoch": 6.9, + "learning_rate": 9.73418593747345e-07, + "loss": 0.0016, + "step": 52683 + }, + { + "epoch": 6.9, + "learning_rate": 9.731903766382377e-07, + "loss": 0.002, + "step": 52684 + }, + { + "epoch": 6.9, + "learning_rate": 9.729621849165505e-07, + "loss": 0.0011, + "step": 52685 + }, + { + "epoch": 6.9, + "learning_rate": 9.727340185829283e-07, + "loss": 0.0023, + "step": 52686 + }, + { + "epoch": 6.9, + "learning_rate": 9.72505877638007e-07, + "loss": 0.0011, + "step": 52687 + }, + { + "epoch": 6.9, + "learning_rate": 9.722777620824354e-07, + "loss": 0.0018, + "step": 52688 + }, + { + "epoch": 6.9, + "learning_rate": 9.720496719168516e-07, + "loss": 0.0013, + "step": 52689 + }, + { + "epoch": 6.9, + "learning_rate": 9.71821607141895e-07, + "loss": 0.0021, + "step": 52690 + }, + { + "epoch": 6.9, + "learning_rate": 9.715935677582122e-07, + "loss": 0.003, + "step": 52691 + }, + { + "epoch": 6.9, + "learning_rate": 9.7136555376644e-07, + "loss": 0.0026, + "step": 52692 + }, + { + "epoch": 6.9, + "learning_rate": 9.711375651672218e-07, + "loss": 0.0014, + "step": 52693 + }, + { + "epoch": 6.9, + "learning_rate": 9.709096019611964e-07, + "loss": 0.0012, + "step": 52694 + }, + { + "epoch": 6.9, + "learning_rate": 9.706816641490103e-07, + "loss": 0.0017, + "step": 52695 + }, + { + "epoch": 6.9, + "learning_rate": 9.704537517312996e-07, + "loss": 0.0013, + "step": 52696 + }, + { + "epoch": 6.9, + "learning_rate": 9.70225864708708e-07, + "loss": 0.0019, + "step": 52697 + }, + { + "epoch": 6.9, + "learning_rate": 9.699980030818734e-07, + "loss": 0.0029, + "step": 52698 + }, + { + "epoch": 6.9, + "learning_rate": 9.697701668514392e-07, + "loss": 0.0015, + "step": 52699 + }, + { + "epoch": 6.9, + "learning_rate": 9.695423560180472e-07, + "loss": 0.0022, + "step": 52700 + }, + { + "epoch": 6.9, + "learning_rate": 9.693145705823359e-07, + "loss": 0.0014, + "step": 52701 + }, + { + "epoch": 6.9, + "learning_rate": 9.690868105449435e-07, + "loss": 0.0022, + "step": 52702 + }, + { + "epoch": 6.9, + "learning_rate": 9.688590759065163e-07, + "loss": 0.001, + "step": 52703 + }, + { + "epoch": 6.9, + "learning_rate": 9.686313666676904e-07, + "loss": 0.0013, + "step": 52704 + }, + { + "epoch": 6.9, + "learning_rate": 9.684036828291099e-07, + "loss": 0.0016, + "step": 52705 + }, + { + "epoch": 6.9, + "learning_rate": 9.681760243914107e-07, + "loss": 0.0023, + "step": 52706 + }, + { + "epoch": 6.9, + "learning_rate": 9.679483913552367e-07, + "loss": 0.0016, + "step": 52707 + }, + { + "epoch": 6.9, + "learning_rate": 9.677207837212278e-07, + "loss": 0.0017, + "step": 52708 + }, + { + "epoch": 6.9, + "learning_rate": 9.674932014900207e-07, + "loss": 0.0038, + "step": 52709 + }, + { + "epoch": 6.9, + "learning_rate": 9.67265644662261e-07, + "loss": 0.0008, + "step": 52710 + }, + { + "epoch": 6.9, + "learning_rate": 9.670381132385843e-07, + "loss": 0.0015, + "step": 52711 + }, + { + "epoch": 6.9, + "learning_rate": 9.668106072196325e-07, + "loss": 0.0018, + "step": 52712 + }, + { + "epoch": 6.91, + "learning_rate": 9.66583126606042e-07, + "loss": 0.0019, + "step": 52713 + }, + { + "epoch": 6.91, + "learning_rate": 9.66355671398459e-07, + "loss": 0.0012, + "step": 52714 + }, + { + "epoch": 6.91, + "learning_rate": 9.66128241597516e-07, + "loss": 0.0018, + "step": 52715 + }, + { + "epoch": 6.91, + "learning_rate": 9.659008372038591e-07, + "loss": 0.0019, + "step": 52716 + }, + { + "epoch": 6.91, + "learning_rate": 9.656734582181237e-07, + "loss": 0.0015, + "step": 52717 + }, + { + "epoch": 6.91, + "learning_rate": 9.654461046409514e-07, + "loss": 0.0009, + "step": 52718 + }, + { + "epoch": 6.91, + "learning_rate": 9.652187764729815e-07, + "loss": 0.001, + "step": 52719 + }, + { + "epoch": 6.91, + "learning_rate": 9.649914737148513e-07, + "loss": 0.0013, + "step": 52720 + }, + { + "epoch": 6.91, + "learning_rate": 9.647641963672017e-07, + "loss": 0.0015, + "step": 52721 + }, + { + "epoch": 6.91, + "learning_rate": 9.645369444306718e-07, + "loss": 0.0013, + "step": 52722 + }, + { + "epoch": 6.91, + "learning_rate": 9.643097179059013e-07, + "loss": 0.0013, + "step": 52723 + }, + { + "epoch": 6.91, + "learning_rate": 9.640825167935264e-07, + "loss": 0.001, + "step": 52724 + }, + { + "epoch": 6.91, + "learning_rate": 9.638553410941897e-07, + "loss": 0.0018, + "step": 52725 + }, + { + "epoch": 6.91, + "learning_rate": 9.636281908085276e-07, + "loss": 0.0016, + "step": 52726 + }, + { + "epoch": 6.91, + "learning_rate": 9.634010659371817e-07, + "loss": 0.0031, + "step": 52727 + }, + { + "epoch": 6.91, + "learning_rate": 9.631739664807894e-07, + "loss": 0.001, + "step": 52728 + }, + { + "epoch": 6.91, + "learning_rate": 9.629468924399866e-07, + "loss": 0.0028, + "step": 52729 + }, + { + "epoch": 6.91, + "learning_rate": 9.627198438154162e-07, + "loss": 0.001, + "step": 52730 + }, + { + "epoch": 6.91, + "learning_rate": 9.624928206077155e-07, + "loss": 0.0016, + "step": 52731 + }, + { + "epoch": 6.91, + "learning_rate": 9.622658228175208e-07, + "loss": 0.0021, + "step": 52732 + }, + { + "epoch": 6.91, + "learning_rate": 9.620388504454726e-07, + "loss": 0.0012, + "step": 52733 + }, + { + "epoch": 6.91, + "learning_rate": 9.618119034922101e-07, + "loss": 0.0024, + "step": 52734 + }, + { + "epoch": 6.91, + "learning_rate": 9.6158498195837e-07, + "loss": 0.0017, + "step": 52735 + }, + { + "epoch": 6.91, + "learning_rate": 9.613580858445892e-07, + "loss": 0.0021, + "step": 52736 + }, + { + "epoch": 6.91, + "learning_rate": 9.611312151515073e-07, + "loss": 0.0015, + "step": 52737 + }, + { + "epoch": 6.91, + "learning_rate": 9.609043698797638e-07, + "loss": 0.0011, + "step": 52738 + }, + { + "epoch": 6.91, + "learning_rate": 9.606775500299958e-07, + "loss": 0.0008, + "step": 52739 + }, + { + "epoch": 6.91, + "learning_rate": 9.604507556028396e-07, + "loss": 0.0017, + "step": 52740 + }, + { + "epoch": 6.91, + "learning_rate": 9.60223986598936e-07, + "loss": 0.0019, + "step": 52741 + }, + { + "epoch": 6.91, + "learning_rate": 9.599972430189197e-07, + "loss": 0.0013, + "step": 52742 + }, + { + "epoch": 6.91, + "learning_rate": 9.597705248634314e-07, + "loss": 0.0022, + "step": 52743 + }, + { + "epoch": 6.91, + "learning_rate": 9.59543832133104e-07, + "loss": 0.0012, + "step": 52744 + }, + { + "epoch": 6.91, + "learning_rate": 9.593171648285804e-07, + "loss": 0.0008, + "step": 52745 + }, + { + "epoch": 6.91, + "learning_rate": 9.590905229504965e-07, + "loss": 0.0026, + "step": 52746 + }, + { + "epoch": 6.91, + "learning_rate": 9.588639064994864e-07, + "loss": 0.0028, + "step": 52747 + }, + { + "epoch": 6.91, + "learning_rate": 9.586373154761896e-07, + "loss": 0.002, + "step": 52748 + }, + { + "epoch": 6.91, + "learning_rate": 9.584107498812467e-07, + "loss": 0.0021, + "step": 52749 + }, + { + "epoch": 6.91, + "learning_rate": 9.581842097152916e-07, + "loss": 0.0022, + "step": 52750 + }, + { + "epoch": 6.91, + "learning_rate": 9.579576949789626e-07, + "loss": 0.0021, + "step": 52751 + }, + { + "epoch": 6.91, + "learning_rate": 9.577312056728937e-07, + "loss": 0.0022, + "step": 52752 + }, + { + "epoch": 6.91, + "learning_rate": 9.575047417977258e-07, + "loss": 0.0018, + "step": 52753 + }, + { + "epoch": 6.91, + "learning_rate": 9.572783033540944e-07, + "loss": 0.0021, + "step": 52754 + }, + { + "epoch": 6.91, + "learning_rate": 9.57051890342635e-07, + "loss": 0.002, + "step": 52755 + }, + { + "epoch": 6.91, + "learning_rate": 9.568255027639872e-07, + "loss": 0.001, + "step": 52756 + }, + { + "epoch": 6.91, + "learning_rate": 9.565991406187859e-07, + "loss": 0.0012, + "step": 52757 + }, + { + "epoch": 6.91, + "learning_rate": 9.56372803907668e-07, + "loss": 0.0023, + "step": 52758 + }, + { + "epoch": 6.91, + "learning_rate": 9.561464926312658e-07, + "loss": 0.0018, + "step": 52759 + }, + { + "epoch": 6.91, + "learning_rate": 9.55920206790224e-07, + "loss": 0.0023, + "step": 52760 + }, + { + "epoch": 6.91, + "learning_rate": 9.556939463851756e-07, + "loss": 0.0016, + "step": 52761 + }, + { + "epoch": 6.91, + "learning_rate": 9.554677114167553e-07, + "loss": 0.0014, + "step": 52762 + }, + { + "epoch": 6.91, + "learning_rate": 9.552415018856e-07, + "loss": 0.0022, + "step": 52763 + }, + { + "epoch": 6.91, + "learning_rate": 9.550153177923472e-07, + "loss": 0.0007, + "step": 52764 + }, + { + "epoch": 6.91, + "learning_rate": 9.547891591376313e-07, + "loss": 0.0015, + "step": 52765 + }, + { + "epoch": 6.91, + "learning_rate": 9.545630259220883e-07, + "loss": 0.0008, + "step": 52766 + }, + { + "epoch": 6.91, + "learning_rate": 9.54336918146357e-07, + "loss": 0.0014, + "step": 52767 + }, + { + "epoch": 6.91, + "learning_rate": 9.541108358110719e-07, + "loss": 0.0008, + "step": 52768 + }, + { + "epoch": 6.91, + "learning_rate": 9.53884778916867e-07, + "loss": 0.0016, + "step": 52769 + }, + { + "epoch": 6.91, + "learning_rate": 9.536587474643788e-07, + "loss": 0.0019, + "step": 52770 + }, + { + "epoch": 6.91, + "learning_rate": 9.534327414542433e-07, + "loss": 0.0013, + "step": 52771 + }, + { + "epoch": 6.91, + "learning_rate": 9.532067608870976e-07, + "loss": 0.0058, + "step": 52772 + }, + { + "epoch": 6.91, + "learning_rate": 9.52980805763577e-07, + "loss": 0.0012, + "step": 52773 + }, + { + "epoch": 6.91, + "learning_rate": 9.527548760843142e-07, + "loss": 0.0014, + "step": 52774 + }, + { + "epoch": 6.91, + "learning_rate": 9.525289718499475e-07, + "loss": 0.0028, + "step": 52775 + }, + { + "epoch": 6.91, + "learning_rate": 9.523030930611121e-07, + "loss": 0.0016, + "step": 52776 + }, + { + "epoch": 6.91, + "learning_rate": 9.520772397184419e-07, + "loss": 0.0008, + "step": 52777 + }, + { + "epoch": 6.91, + "learning_rate": 9.518514118225708e-07, + "loss": 0.0008, + "step": 52778 + }, + { + "epoch": 6.91, + "learning_rate": 9.516256093741383e-07, + "loss": 0.0012, + "step": 52779 + }, + { + "epoch": 6.91, + "learning_rate": 9.513998323737761e-07, + "loss": 0.0011, + "step": 52780 + }, + { + "epoch": 6.91, + "learning_rate": 9.511740808221193e-07, + "loss": 0.002, + "step": 52781 + }, + { + "epoch": 6.91, + "learning_rate": 9.509483547198028e-07, + "loss": 0.0009, + "step": 52782 + }, + { + "epoch": 6.91, + "learning_rate": 9.507226540674641e-07, + "loss": 0.0016, + "step": 52783 + }, + { + "epoch": 6.91, + "learning_rate": 9.504969788657358e-07, + "loss": 0.0013, + "step": 52784 + }, + { + "epoch": 6.91, + "learning_rate": 9.502713291152521e-07, + "loss": 0.0024, + "step": 52785 + }, + { + "epoch": 6.91, + "learning_rate": 9.500457048166479e-07, + "loss": 0.0019, + "step": 52786 + }, + { + "epoch": 6.91, + "learning_rate": 9.498201059705592e-07, + "loss": 0.001, + "step": 52787 + }, + { + "epoch": 6.91, + "learning_rate": 9.495945325776201e-07, + "loss": 0.0013, + "step": 52788 + }, + { + "epoch": 6.91, + "learning_rate": 9.493689846384624e-07, + "loss": 0.001, + "step": 52789 + }, + { + "epoch": 6.92, + "learning_rate": 9.491434621537244e-07, + "loss": 0.0022, + "step": 52790 + }, + { + "epoch": 6.92, + "learning_rate": 9.489179651240365e-07, + "loss": 0.0007, + "step": 52791 + }, + { + "epoch": 6.92, + "learning_rate": 9.486924935500363e-07, + "loss": 0.0015, + "step": 52792 + }, + { + "epoch": 6.92, + "learning_rate": 9.484670474323565e-07, + "loss": 0.0017, + "step": 52793 + }, + { + "epoch": 6.92, + "learning_rate": 9.482416267716299e-07, + "loss": 0.0011, + "step": 52794 + }, + { + "epoch": 6.92, + "learning_rate": 9.480162315684927e-07, + "loss": 0.0024, + "step": 52795 + }, + { + "epoch": 6.92, + "learning_rate": 9.477908618235788e-07, + "loss": 0.0016, + "step": 52796 + }, + { + "epoch": 6.92, + "learning_rate": 9.475655175375176e-07, + "loss": 0.0023, + "step": 52797 + }, + { + "epoch": 6.92, + "learning_rate": 9.473401987109498e-07, + "loss": 0.0011, + "step": 52798 + }, + { + "epoch": 6.92, + "learning_rate": 9.47114905344505e-07, + "loss": 0.002, + "step": 52799 + }, + { + "epoch": 6.92, + "learning_rate": 9.46889637438817e-07, + "loss": 0.0015, + "step": 52800 + }, + { + "epoch": 6.92, + "learning_rate": 9.466643949945186e-07, + "loss": 0.0009, + "step": 52801 + }, + { + "epoch": 6.92, + "learning_rate": 9.464391780122439e-07, + "loss": 0.0009, + "step": 52802 + }, + { + "epoch": 6.92, + "learning_rate": 9.462139864926289e-07, + "loss": 0.0013, + "step": 52803 + }, + { + "epoch": 6.92, + "learning_rate": 9.459888204363054e-07, + "loss": 0.0029, + "step": 52804 + }, + { + "epoch": 6.92, + "learning_rate": 9.457636798439041e-07, + "loss": 0.0016, + "step": 52805 + }, + { + "epoch": 6.92, + "learning_rate": 9.45538564716062e-07, + "loss": 0.002, + "step": 52806 + }, + { + "epoch": 6.92, + "learning_rate": 9.453134750534099e-07, + "loss": 0.0013, + "step": 52807 + }, + { + "epoch": 6.92, + "learning_rate": 9.450884108565828e-07, + "loss": 0.0024, + "step": 52808 + }, + { + "epoch": 6.92, + "learning_rate": 9.44863372126209e-07, + "loss": 0.0015, + "step": 52809 + }, + { + "epoch": 6.92, + "learning_rate": 9.446383588629282e-07, + "loss": 0.0013, + "step": 52810 + }, + { + "epoch": 6.92, + "learning_rate": 9.444133710673687e-07, + "loss": 0.0013, + "step": 52811 + }, + { + "epoch": 6.92, + "learning_rate": 9.441884087401632e-07, + "loss": 0.0013, + "step": 52812 + }, + { + "epoch": 6.92, + "learning_rate": 9.439634718819446e-07, + "loss": 0.0019, + "step": 52813 + }, + { + "epoch": 6.92, + "learning_rate": 9.437385604933491e-07, + "loss": 0.0028, + "step": 52814 + }, + { + "epoch": 6.92, + "learning_rate": 9.435136745750062e-07, + "loss": 0.0012, + "step": 52815 + }, + { + "epoch": 6.92, + "learning_rate": 9.432888141275476e-07, + "loss": 0.0018, + "step": 52816 + }, + { + "epoch": 6.92, + "learning_rate": 9.430639791516083e-07, + "loss": 0.0014, + "step": 52817 + }, + { + "epoch": 6.92, + "learning_rate": 9.428391696478189e-07, + "loss": 0.0032, + "step": 52818 + }, + { + "epoch": 6.92, + "learning_rate": 9.426143856168124e-07, + "loss": 0.0016, + "step": 52819 + }, + { + "epoch": 6.92, + "learning_rate": 9.423896270592192e-07, + "loss": 0.0015, + "step": 52820 + }, + { + "epoch": 6.92, + "learning_rate": 9.421648939756733e-07, + "loss": 0.0021, + "step": 52821 + }, + { + "epoch": 6.92, + "learning_rate": 9.419401863668076e-07, + "loss": 0.0017, + "step": 52822 + }, + { + "epoch": 6.92, + "learning_rate": 9.417155042332527e-07, + "loss": 0.0015, + "step": 52823 + }, + { + "epoch": 6.92, + "learning_rate": 9.414908475756357e-07, + "loss": 0.0018, + "step": 52824 + }, + { + "epoch": 6.92, + "learning_rate": 9.412662163945985e-07, + "loss": 0.0015, + "step": 52825 + }, + { + "epoch": 6.92, + "learning_rate": 9.410416106907672e-07, + "loss": 0.0017, + "step": 52826 + }, + { + "epoch": 6.92, + "learning_rate": 9.408170304647734e-07, + "loss": 0.0012, + "step": 52827 + }, + { + "epoch": 6.92, + "learning_rate": 9.40592475717248e-07, + "loss": 0.0023, + "step": 52828 + }, + { + "epoch": 6.92, + "learning_rate": 9.403679464488247e-07, + "loss": 0.0017, + "step": 52829 + }, + { + "epoch": 6.92, + "learning_rate": 9.401434426601353e-07, + "loss": 0.0022, + "step": 52830 + }, + { + "epoch": 6.92, + "learning_rate": 9.399189643518069e-07, + "loss": 0.0012, + "step": 52831 + }, + { + "epoch": 6.92, + "learning_rate": 9.396945115244771e-07, + "loss": 0.0009, + "step": 52832 + }, + { + "epoch": 6.92, + "learning_rate": 9.394700841787729e-07, + "loss": 0.0015, + "step": 52833 + }, + { + "epoch": 6.92, + "learning_rate": 9.392456823153273e-07, + "loss": 0.002, + "step": 52834 + }, + { + "epoch": 6.92, + "learning_rate": 9.390213059347686e-07, + "loss": 0.002, + "step": 52835 + }, + { + "epoch": 6.92, + "learning_rate": 9.387969550377307e-07, + "loss": 0.002, + "step": 52836 + }, + { + "epoch": 6.92, + "learning_rate": 9.385726296248443e-07, + "loss": 0.001, + "step": 52837 + }, + { + "epoch": 6.92, + "learning_rate": 9.383483296967411e-07, + "loss": 0.0015, + "step": 52838 + }, + { + "epoch": 6.92, + "learning_rate": 9.381240552540482e-07, + "loss": 0.0015, + "step": 52839 + }, + { + "epoch": 6.92, + "learning_rate": 9.37899806297401e-07, + "loss": 0.0019, + "step": 52840 + }, + { + "epoch": 6.92, + "learning_rate": 9.376755828274275e-07, + "loss": 0.0015, + "step": 52841 + }, + { + "epoch": 6.92, + "learning_rate": 9.374513848447586e-07, + "loss": 0.0018, + "step": 52842 + }, + { + "epoch": 6.92, + "learning_rate": 9.372272123500237e-07, + "loss": 0.0026, + "step": 52843 + }, + { + "epoch": 6.92, + "learning_rate": 9.370030653438566e-07, + "loss": 0.0012, + "step": 52844 + }, + { + "epoch": 6.92, + "learning_rate": 9.36778943826886e-07, + "loss": 0.0027, + "step": 52845 + }, + { + "epoch": 6.92, + "learning_rate": 9.365548477997388e-07, + "loss": 0.0013, + "step": 52846 + }, + { + "epoch": 6.92, + "learning_rate": 9.363307772630503e-07, + "loss": 0.0021, + "step": 52847 + }, + { + "epoch": 6.92, + "learning_rate": 9.361067322174488e-07, + "loss": 0.0012, + "step": 52848 + }, + { + "epoch": 6.92, + "learning_rate": 9.358827126635661e-07, + "loss": 0.0019, + "step": 52849 + }, + { + "epoch": 6.92, + "learning_rate": 9.356587186020294e-07, + "loss": 0.0023, + "step": 52850 + }, + { + "epoch": 6.92, + "learning_rate": 9.354347500334682e-07, + "loss": 0.0009, + "step": 52851 + }, + { + "epoch": 6.92, + "learning_rate": 9.352108069585164e-07, + "loss": 0.0008, + "step": 52852 + }, + { + "epoch": 6.92, + "learning_rate": 9.349868893778014e-07, + "loss": 0.0017, + "step": 52853 + }, + { + "epoch": 6.92, + "learning_rate": 9.347629972919515e-07, + "loss": 0.0019, + "step": 52854 + }, + { + "epoch": 6.92, + "learning_rate": 9.345391307015983e-07, + "loss": 0.0012, + "step": 52855 + }, + { + "epoch": 6.92, + "learning_rate": 9.343152896073726e-07, + "loss": 0.001, + "step": 52856 + }, + { + "epoch": 6.92, + "learning_rate": 9.340914740098995e-07, + "loss": 0.0013, + "step": 52857 + }, + { + "epoch": 6.92, + "learning_rate": 9.338676839098137e-07, + "loss": 0.0033, + "step": 52858 + }, + { + "epoch": 6.92, + "learning_rate": 9.336439193077406e-07, + "loss": 0.001, + "step": 52859 + }, + { + "epoch": 6.92, + "learning_rate": 9.33420180204313e-07, + "loss": 0.0022, + "step": 52860 + }, + { + "epoch": 6.92, + "learning_rate": 9.33196466600158e-07, + "loss": 0.0015, + "step": 52861 + }, + { + "epoch": 6.92, + "learning_rate": 9.329727784959031e-07, + "loss": 0.0014, + "step": 52862 + }, + { + "epoch": 6.92, + "learning_rate": 9.327491158921809e-07, + "loss": 0.002, + "step": 52863 + }, + { + "epoch": 6.92, + "learning_rate": 9.325254787896198e-07, + "loss": 0.0012, + "step": 52864 + }, + { + "epoch": 6.92, + "learning_rate": 9.323018671888473e-07, + "loss": 0.0021, + "step": 52865 + }, + { + "epoch": 6.93, + "learning_rate": 9.320782810904916e-07, + "loss": 0.0019, + "step": 52866 + }, + { + "epoch": 6.93, + "learning_rate": 9.318547204951845e-07, + "loss": 0.0026, + "step": 52867 + }, + { + "epoch": 6.93, + "learning_rate": 9.316311854035509e-07, + "loss": 0.0017, + "step": 52868 + }, + { + "epoch": 6.93, + "learning_rate": 9.31407675816225e-07, + "loss": 0.0016, + "step": 52869 + }, + { + "epoch": 6.93, + "learning_rate": 9.311841917338283e-07, + "loss": 0.0012, + "step": 52870 + }, + { + "epoch": 6.93, + "learning_rate": 9.309607331569959e-07, + "loss": 0.0019, + "step": 52871 + }, + { + "epoch": 6.93, + "learning_rate": 9.307373000863529e-07, + "loss": 0.001, + "step": 52872 + }, + { + "epoch": 6.93, + "learning_rate": 9.305138925225288e-07, + "loss": 0.0019, + "step": 52873 + }, + { + "epoch": 6.93, + "learning_rate": 9.302905104661498e-07, + "loss": 0.0017, + "step": 52874 + }, + { + "epoch": 6.93, + "learning_rate": 9.300671539178474e-07, + "loss": 0.0014, + "step": 52875 + }, + { + "epoch": 6.93, + "learning_rate": 9.29843822878248e-07, + "loss": 0.0012, + "step": 52876 + }, + { + "epoch": 6.93, + "learning_rate": 9.296205173479777e-07, + "loss": 0.0011, + "step": 52877 + }, + { + "epoch": 6.93, + "learning_rate": 9.293972373276694e-07, + "loss": 0.0018, + "step": 52878 + }, + { + "epoch": 6.93, + "learning_rate": 9.291739828179458e-07, + "loss": 0.0026, + "step": 52879 + }, + { + "epoch": 6.93, + "learning_rate": 9.289507538194386e-07, + "loss": 0.0018, + "step": 52880 + }, + { + "epoch": 6.93, + "learning_rate": 9.287275503327731e-07, + "loss": 0.0019, + "step": 52881 + }, + { + "epoch": 6.93, + "learning_rate": 9.285043723585807e-07, + "loss": 0.0014, + "step": 52882 + }, + { + "epoch": 6.93, + "learning_rate": 9.282812198974855e-07, + "loss": 0.0028, + "step": 52883 + }, + { + "epoch": 6.93, + "learning_rate": 9.28058092950117e-07, + "loss": 0.0014, + "step": 52884 + }, + { + "epoch": 6.93, + "learning_rate": 9.278349915171003e-07, + "loss": 0.0018, + "step": 52885 + }, + { + "epoch": 6.93, + "learning_rate": 9.276119155990659e-07, + "loss": 0.0015, + "step": 52886 + }, + { + "epoch": 6.93, + "learning_rate": 9.2738886519664e-07, + "loss": 0.0026, + "step": 52887 + }, + { + "epoch": 6.93, + "learning_rate": 9.2716584031045e-07, + "loss": 0.0013, + "step": 52888 + }, + { + "epoch": 6.93, + "learning_rate": 9.269428409411208e-07, + "loss": 0.0019, + "step": 52889 + }, + { + "epoch": 6.93, + "learning_rate": 9.267198670892818e-07, + "loss": 0.0006, + "step": 52890 + }, + { + "epoch": 6.93, + "learning_rate": 9.264969187555628e-07, + "loss": 0.0014, + "step": 52891 + }, + { + "epoch": 6.93, + "learning_rate": 9.262739959405864e-07, + "loss": 0.0008, + "step": 52892 + }, + { + "epoch": 6.93, + "learning_rate": 9.2605109864498e-07, + "loss": 0.0014, + "step": 52893 + }, + { + "epoch": 6.93, + "learning_rate": 9.258282268693741e-07, + "loss": 0.0032, + "step": 52894 + }, + { + "epoch": 6.93, + "learning_rate": 9.256053806143927e-07, + "loss": 0.0029, + "step": 52895 + }, + { + "epoch": 6.93, + "learning_rate": 9.253825598806609e-07, + "loss": 0.0015, + "step": 52896 + }, + { + "epoch": 6.93, + "learning_rate": 9.251597646688093e-07, + "loss": 0.0031, + "step": 52897 + }, + { + "epoch": 6.93, + "learning_rate": 9.249369949794629e-07, + "loss": 0.0018, + "step": 52898 + }, + { + "epoch": 6.93, + "learning_rate": 9.247142508132467e-07, + "loss": 0.002, + "step": 52899 + }, + { + "epoch": 6.93, + "learning_rate": 9.244915321707881e-07, + "loss": 0.0029, + "step": 52900 + }, + { + "epoch": 6.93, + "learning_rate": 9.242688390527132e-07, + "loss": 0.0017, + "step": 52901 + }, + { + "epoch": 6.93, + "learning_rate": 9.240461714596505e-07, + "loss": 0.0019, + "step": 52902 + }, + { + "epoch": 6.93, + "learning_rate": 9.238235293922248e-07, + "loss": 0.001, + "step": 52903 + }, + { + "epoch": 6.93, + "learning_rate": 9.236009128510603e-07, + "loss": 0.0007, + "step": 52904 + }, + { + "epoch": 6.93, + "learning_rate": 9.233783218367864e-07, + "loss": 0.0016, + "step": 52905 + }, + { + "epoch": 6.93, + "learning_rate": 9.231557563500282e-07, + "loss": 0.0018, + "step": 52906 + }, + { + "epoch": 6.93, + "learning_rate": 9.229332163914106e-07, + "loss": 0.0015, + "step": 52907 + }, + { + "epoch": 6.93, + "learning_rate": 9.227107019615578e-07, + "loss": 0.0017, + "step": 52908 + }, + { + "epoch": 6.93, + "learning_rate": 9.224882130611002e-07, + "loss": 0.0015, + "step": 52909 + }, + { + "epoch": 6.93, + "learning_rate": 9.222657496906606e-07, + "loss": 0.0009, + "step": 52910 + }, + { + "epoch": 6.93, + "learning_rate": 9.220433118508643e-07, + "loss": 0.0021, + "step": 52911 + }, + { + "epoch": 6.93, + "learning_rate": 9.218208995423361e-07, + "loss": 0.0012, + "step": 52912 + }, + { + "epoch": 6.93, + "learning_rate": 9.215985127657068e-07, + "loss": 0.0014, + "step": 52913 + }, + { + "epoch": 6.93, + "learning_rate": 9.213761515215968e-07, + "loss": 0.0014, + "step": 52914 + }, + { + "epoch": 6.93, + "learning_rate": 9.211538158106336e-07, + "loss": 0.0016, + "step": 52915 + }, + { + "epoch": 6.93, + "learning_rate": 9.2093150563344e-07, + "loss": 0.0016, + "step": 52916 + }, + { + "epoch": 6.93, + "learning_rate": 9.207092209906454e-07, + "loss": 0.0017, + "step": 52917 + }, + { + "epoch": 6.93, + "learning_rate": 9.204869618828716e-07, + "loss": 0.0016, + "step": 52918 + }, + { + "epoch": 6.93, + "learning_rate": 9.202647283107435e-07, + "loss": 0.0016, + "step": 52919 + }, + { + "epoch": 6.93, + "learning_rate": 9.200425202748886e-07, + "loss": 0.0017, + "step": 52920 + }, + { + "epoch": 6.93, + "learning_rate": 9.198203377759307e-07, + "loss": 0.0024, + "step": 52921 + }, + { + "epoch": 6.93, + "learning_rate": 9.195981808144938e-07, + "loss": 0.0019, + "step": 52922 + }, + { + "epoch": 6.93, + "learning_rate": 9.19376049391204e-07, + "loss": 0.002, + "step": 52923 + }, + { + "epoch": 6.93, + "learning_rate": 9.191539435066843e-07, + "loss": 0.0012, + "step": 52924 + }, + { + "epoch": 6.93, + "learning_rate": 9.189318631615618e-07, + "loss": 0.0016, + "step": 52925 + }, + { + "epoch": 6.93, + "learning_rate": 9.187098083564605e-07, + "loss": 0.0011, + "step": 52926 + }, + { + "epoch": 6.93, + "learning_rate": 9.184877790920022e-07, + "loss": 0.0016, + "step": 52927 + }, + { + "epoch": 6.93, + "learning_rate": 9.182657753688151e-07, + "loss": 0.0015, + "step": 52928 + }, + { + "epoch": 6.93, + "learning_rate": 9.180437971875223e-07, + "loss": 0.0022, + "step": 52929 + }, + { + "epoch": 6.93, + "learning_rate": 9.178218445487475e-07, + "loss": 0.0011, + "step": 52930 + }, + { + "epoch": 6.93, + "learning_rate": 9.175999174531125e-07, + "loss": 0.0013, + "step": 52931 + }, + { + "epoch": 6.93, + "learning_rate": 9.173780159012468e-07, + "loss": 0.002, + "step": 52932 + }, + { + "epoch": 6.93, + "learning_rate": 9.17156139893769e-07, + "loss": 0.0011, + "step": 52933 + }, + { + "epoch": 6.93, + "learning_rate": 9.169342894313083e-07, + "loss": 0.0021, + "step": 52934 + }, + { + "epoch": 6.93, + "learning_rate": 9.167124645144853e-07, + "loss": 0.0017, + "step": 52935 + }, + { + "epoch": 6.93, + "learning_rate": 9.164906651439254e-07, + "loss": 0.0018, + "step": 52936 + }, + { + "epoch": 6.93, + "learning_rate": 9.162688913202522e-07, + "loss": 0.0038, + "step": 52937 + }, + { + "epoch": 6.93, + "learning_rate": 9.160471430440887e-07, + "loss": 0.0022, + "step": 52938 + }, + { + "epoch": 6.93, + "learning_rate": 9.158254203160577e-07, + "loss": 0.0032, + "step": 52939 + }, + { + "epoch": 6.93, + "learning_rate": 9.156037231367854e-07, + "loss": 0.0011, + "step": 52940 + }, + { + "epoch": 6.93, + "learning_rate": 9.153820515068934e-07, + "loss": 0.002, + "step": 52941 + }, + { + "epoch": 6.94, + "learning_rate": 9.151604054270036e-07, + "loss": 0.0012, + "step": 52942 + }, + { + "epoch": 6.94, + "learning_rate": 9.149387848977443e-07, + "loss": 0.0019, + "step": 52943 + }, + { + "epoch": 6.94, + "learning_rate": 9.147171899197327e-07, + "loss": 0.0016, + "step": 52944 + }, + { + "epoch": 6.94, + "learning_rate": 9.144956204935984e-07, + "loss": 0.0009, + "step": 52945 + }, + { + "epoch": 6.94, + "learning_rate": 9.142740766199587e-07, + "loss": 0.0018, + "step": 52946 + }, + { + "epoch": 6.94, + "learning_rate": 9.140525582994409e-07, + "loss": 0.001, + "step": 52947 + }, + { + "epoch": 6.94, + "learning_rate": 9.138310655326677e-07, + "loss": 0.0015, + "step": 52948 + }, + { + "epoch": 6.94, + "learning_rate": 9.136095983202598e-07, + "loss": 0.0029, + "step": 52949 + }, + { + "epoch": 6.94, + "learning_rate": 9.133881566628389e-07, + "loss": 0.0014, + "step": 52950 + }, + { + "epoch": 6.94, + "learning_rate": 9.131667405610323e-07, + "loss": 0.0014, + "step": 52951 + }, + { + "epoch": 6.94, + "learning_rate": 9.129453500154606e-07, + "loss": 0.0016, + "step": 52952 + }, + { + "epoch": 6.94, + "learning_rate": 9.127239850267455e-07, + "loss": 0.0011, + "step": 52953 + }, + { + "epoch": 6.94, + "learning_rate": 9.125026455955089e-07, + "loss": 0.001, + "step": 52954 + }, + { + "epoch": 6.94, + "learning_rate": 9.122813317223744e-07, + "loss": 0.0028, + "step": 52955 + }, + { + "epoch": 6.94, + "learning_rate": 9.120600434079662e-07, + "loss": 0.0016, + "step": 52956 + }, + { + "epoch": 6.94, + "learning_rate": 9.118387806529061e-07, + "loss": 0.0013, + "step": 52957 + }, + { + "epoch": 6.94, + "learning_rate": 9.116175434578123e-07, + "loss": 0.0026, + "step": 52958 + }, + { + "epoch": 6.94, + "learning_rate": 9.113963318233132e-07, + "loss": 0.0011, + "step": 52959 + }, + { + "epoch": 6.94, + "learning_rate": 9.111751457500273e-07, + "loss": 0.0011, + "step": 52960 + }, + { + "epoch": 6.94, + "learning_rate": 9.109539852385762e-07, + "loss": 0.0013, + "step": 52961 + }, + { + "epoch": 6.94, + "learning_rate": 9.10732850289584e-07, + "loss": 0.0017, + "step": 52962 + }, + { + "epoch": 6.94, + "learning_rate": 9.105117409036724e-07, + "loss": 0.001, + "step": 52963 + }, + { + "epoch": 6.94, + "learning_rate": 9.102906570814618e-07, + "loss": 0.0009, + "step": 52964 + }, + { + "epoch": 6.94, + "learning_rate": 9.10069598823573e-07, + "loss": 0.0033, + "step": 52965 + }, + { + "epoch": 6.94, + "learning_rate": 9.098485661306288e-07, + "loss": 0.0015, + "step": 52966 + }, + { + "epoch": 6.94, + "learning_rate": 9.096275590032544e-07, + "loss": 0.0023, + "step": 52967 + }, + { + "epoch": 6.94, + "learning_rate": 9.094065774420679e-07, + "loss": 0.0022, + "step": 52968 + }, + { + "epoch": 6.94, + "learning_rate": 9.09185621447689e-07, + "loss": 0.0017, + "step": 52969 + }, + { + "epoch": 6.94, + "learning_rate": 9.089646910207439e-07, + "loss": 0.002, + "step": 52970 + }, + { + "epoch": 6.94, + "learning_rate": 9.08743786161852e-07, + "loss": 0.0022, + "step": 52971 + }, + { + "epoch": 6.94, + "learning_rate": 9.085229068716328e-07, + "loss": 0.0017, + "step": 52972 + }, + { + "epoch": 6.94, + "learning_rate": 9.083020531507081e-07, + "loss": 0.0009, + "step": 52973 + }, + { + "epoch": 6.94, + "learning_rate": 9.080812249997006e-07, + "loss": 0.0012, + "step": 52974 + }, + { + "epoch": 6.94, + "learning_rate": 9.078604224192311e-07, + "loss": 0.0011, + "step": 52975 + }, + { + "epoch": 6.94, + "learning_rate": 9.076396454099179e-07, + "loss": 0.0011, + "step": 52976 + }, + { + "epoch": 6.94, + "learning_rate": 9.074188939723838e-07, + "loss": 0.0027, + "step": 52977 + }, + { + "epoch": 6.94, + "learning_rate": 9.071981681072517e-07, + "loss": 0.0015, + "step": 52978 + }, + { + "epoch": 6.94, + "learning_rate": 9.069774678151411e-07, + "loss": 0.0012, + "step": 52979 + }, + { + "epoch": 6.94, + "learning_rate": 9.067567930966715e-07, + "loss": 0.002, + "step": 52980 + }, + { + "epoch": 6.94, + "learning_rate": 9.065361439524634e-07, + "loss": 0.0017, + "step": 52981 + }, + { + "epoch": 6.94, + "learning_rate": 9.063155203831386e-07, + "loss": 0.0023, + "step": 52982 + }, + { + "epoch": 6.94, + "learning_rate": 9.060949223893178e-07, + "loss": 0.0022, + "step": 52983 + }, + { + "epoch": 6.94, + "learning_rate": 9.058743499716183e-07, + "loss": 0.0009, + "step": 52984 + }, + { + "epoch": 6.94, + "learning_rate": 9.05653803130666e-07, + "loss": 0.0009, + "step": 52985 + }, + { + "epoch": 6.94, + "learning_rate": 9.054332818670774e-07, + "loss": 0.002, + "step": 52986 + }, + { + "epoch": 6.94, + "learning_rate": 9.05212786181473e-07, + "loss": 0.0022, + "step": 52987 + }, + { + "epoch": 6.94, + "learning_rate": 9.049923160744723e-07, + "loss": 0.0009, + "step": 52988 + }, + { + "epoch": 6.94, + "learning_rate": 9.047718715466957e-07, + "loss": 0.0007, + "step": 52989 + }, + { + "epoch": 6.94, + "learning_rate": 9.045514525987653e-07, + "loss": 0.001, + "step": 52990 + }, + { + "epoch": 6.94, + "learning_rate": 9.043310592313004e-07, + "loss": 0.0018, + "step": 52991 + }, + { + "epoch": 6.94, + "learning_rate": 9.041106914449182e-07, + "loss": 0.001, + "step": 52992 + }, + { + "epoch": 6.94, + "learning_rate": 9.038903492402418e-07, + "loss": 0.0019, + "step": 52993 + }, + { + "epoch": 6.94, + "learning_rate": 9.036700326178893e-07, + "loss": 0.001, + "step": 52994 + }, + { + "epoch": 6.94, + "learning_rate": 9.034497415784804e-07, + "loss": 0.0027, + "step": 52995 + }, + { + "epoch": 6.94, + "learning_rate": 9.032294761226334e-07, + "loss": 0.0013, + "step": 52996 + }, + { + "epoch": 6.94, + "learning_rate": 9.030092362509702e-07, + "loss": 0.0012, + "step": 52997 + }, + { + "epoch": 6.94, + "learning_rate": 9.0278902196411e-07, + "loss": 0.0012, + "step": 52998 + }, + { + "epoch": 6.94, + "learning_rate": 9.025688332626681e-07, + "loss": 0.0019, + "step": 52999 + }, + { + "epoch": 6.94, + "learning_rate": 9.023486701472683e-07, + "loss": 0.0018, + "step": 53000 + }, + { + "epoch": 6.94, + "learning_rate": 9.021285326185291e-07, + "loss": 0.0011, + "step": 53001 + }, + { + "epoch": 6.94, + "learning_rate": 9.01908420677069e-07, + "loss": 0.0007, + "step": 53002 + }, + { + "epoch": 6.94, + "learning_rate": 9.016883343235072e-07, + "loss": 0.0022, + "step": 53003 + }, + { + "epoch": 6.94, + "learning_rate": 9.014682735584612e-07, + "loss": 0.0013, + "step": 53004 + }, + { + "epoch": 6.94, + "learning_rate": 9.012482383825527e-07, + "loss": 0.0008, + "step": 53005 + }, + { + "epoch": 6.94, + "learning_rate": 9.01028228796399e-07, + "loss": 0.002, + "step": 53006 + }, + { + "epoch": 6.94, + "learning_rate": 9.008082448006161e-07, + "loss": 0.0015, + "step": 53007 + }, + { + "epoch": 6.94, + "learning_rate": 9.005882863958282e-07, + "loss": 0.0021, + "step": 53008 + }, + { + "epoch": 6.94, + "learning_rate": 9.003683535826513e-07, + "loss": 0.0016, + "step": 53009 + }, + { + "epoch": 6.94, + "learning_rate": 9.001484463617016e-07, + "loss": 0.0018, + "step": 53010 + }, + { + "epoch": 6.94, + "learning_rate": 8.999285647335998e-07, + "loss": 0.0015, + "step": 53011 + }, + { + "epoch": 6.94, + "learning_rate": 8.997087086989665e-07, + "loss": 0.0013, + "step": 53012 + }, + { + "epoch": 6.94, + "learning_rate": 8.994888782584165e-07, + "loss": 0.0011, + "step": 53013 + }, + { + "epoch": 6.94, + "learning_rate": 8.992690734125709e-07, + "loss": 0.0009, + "step": 53014 + }, + { + "epoch": 6.94, + "learning_rate": 8.990492941620444e-07, + "loss": 0.001, + "step": 53015 + }, + { + "epoch": 6.94, + "learning_rate": 8.988295405074587e-07, + "loss": 0.0016, + "step": 53016 + }, + { + "epoch": 6.94, + "learning_rate": 8.986098124494303e-07, + "loss": 0.0011, + "step": 53017 + }, + { + "epoch": 6.94, + "learning_rate": 8.983901099885761e-07, + "loss": 0.0017, + "step": 53018 + }, + { + "epoch": 6.95, + "learning_rate": 8.981704331255148e-07, + "loss": 0.0017, + "step": 53019 + }, + { + "epoch": 6.95, + "learning_rate": 8.979507818608657e-07, + "loss": 0.0012, + "step": 53020 + }, + { + "epoch": 6.95, + "learning_rate": 8.977311561952428e-07, + "loss": 0.0028, + "step": 53021 + }, + { + "epoch": 6.95, + "learning_rate": 8.975115561292691e-07, + "loss": 0.0018, + "step": 53022 + }, + { + "epoch": 6.95, + "learning_rate": 8.972919816635572e-07, + "loss": 0.0011, + "step": 53023 + }, + { + "epoch": 6.95, + "learning_rate": 8.970724327987279e-07, + "loss": 0.0028, + "step": 53024 + }, + { + "epoch": 6.95, + "learning_rate": 8.968529095353984e-07, + "loss": 0.0016, + "step": 53025 + }, + { + "epoch": 6.95, + "learning_rate": 8.966334118741837e-07, + "loss": 0.0006, + "step": 53026 + }, + { + "epoch": 6.95, + "learning_rate": 8.964139398157045e-07, + "loss": 0.0019, + "step": 53027 + }, + { + "epoch": 6.95, + "learning_rate": 8.961944933605759e-07, + "loss": 0.0019, + "step": 53028 + }, + { + "epoch": 6.95, + "learning_rate": 8.959750725094163e-07, + "loss": 0.0022, + "step": 53029 + }, + { + "epoch": 6.95, + "learning_rate": 8.957556772628395e-07, + "loss": 0.0014, + "step": 53030 + }, + { + "epoch": 6.95, + "learning_rate": 8.955363076214663e-07, + "loss": 0.0031, + "step": 53031 + }, + { + "epoch": 6.95, + "learning_rate": 8.953169635859116e-07, + "loss": 0.0013, + "step": 53032 + }, + { + "epoch": 6.95, + "learning_rate": 8.95097645156795e-07, + "loss": 0.0022, + "step": 53033 + }, + { + "epoch": 6.95, + "learning_rate": 8.948783523347282e-07, + "loss": 0.0022, + "step": 53034 + }, + { + "epoch": 6.95, + "learning_rate": 8.946590851203341e-07, + "loss": 0.0024, + "step": 53035 + }, + { + "epoch": 6.95, + "learning_rate": 8.944398435142266e-07, + "loss": 0.0016, + "step": 53036 + }, + { + "epoch": 6.95, + "learning_rate": 8.942206275170217e-07, + "loss": 0.001, + "step": 53037 + }, + { + "epoch": 6.95, + "learning_rate": 8.940014371293349e-07, + "loss": 0.0016, + "step": 53038 + }, + { + "epoch": 6.95, + "learning_rate": 8.937822723517853e-07, + "loss": 0.0012, + "step": 53039 + }, + { + "epoch": 6.95, + "learning_rate": 8.935631331849881e-07, + "loss": 0.0023, + "step": 53040 + }, + { + "epoch": 6.95, + "learning_rate": 8.933440196295584e-07, + "loss": 0.0012, + "step": 53041 + }, + { + "epoch": 6.95, + "learning_rate": 8.931249316861157e-07, + "loss": 0.002, + "step": 53042 + }, + { + "epoch": 6.95, + "learning_rate": 8.929058693552717e-07, + "loss": 0.0023, + "step": 53043 + }, + { + "epoch": 6.95, + "learning_rate": 8.92686832637647e-07, + "loss": 0.002, + "step": 53044 + }, + { + "epoch": 6.95, + "learning_rate": 8.924678215338556e-07, + "loss": 0.0009, + "step": 53045 + }, + { + "epoch": 6.95, + "learning_rate": 8.922488360445126e-07, + "loss": 0.0019, + "step": 53046 + }, + { + "epoch": 6.95, + "learning_rate": 8.92029876170235e-07, + "loss": 0.0026, + "step": 53047 + }, + { + "epoch": 6.95, + "learning_rate": 8.918109419116394e-07, + "loss": 0.0015, + "step": 53048 + }, + { + "epoch": 6.95, + "learning_rate": 8.915920332693373e-07, + "loss": 0.0015, + "step": 53049 + }, + { + "epoch": 6.95, + "learning_rate": 8.913731502439504e-07, + "loss": 0.0015, + "step": 53050 + }, + { + "epoch": 6.95, + "learning_rate": 8.911542928360917e-07, + "loss": 0.0011, + "step": 53051 + }, + { + "epoch": 6.95, + "learning_rate": 8.909354610463761e-07, + "loss": 0.0016, + "step": 53052 + }, + { + "epoch": 6.95, + "learning_rate": 8.907166548754176e-07, + "loss": 0.0009, + "step": 53053 + }, + { + "epoch": 6.95, + "learning_rate": 8.904978743238335e-07, + "loss": 0.0014, + "step": 53054 + }, + { + "epoch": 6.95, + "learning_rate": 8.902791193922411e-07, + "loss": 0.0019, + "step": 53055 + }, + { + "epoch": 6.95, + "learning_rate": 8.900603900812532e-07, + "loss": 0.0021, + "step": 53056 + }, + { + "epoch": 6.95, + "learning_rate": 8.898416863914839e-07, + "loss": 0.0009, + "step": 53057 + }, + { + "epoch": 6.95, + "learning_rate": 8.896230083235513e-07, + "loss": 0.0019, + "step": 53058 + }, + { + "epoch": 6.95, + "learning_rate": 8.894043558780685e-07, + "loss": 0.0015, + "step": 53059 + }, + { + "epoch": 6.95, + "learning_rate": 8.891857290556505e-07, + "loss": 0.0016, + "step": 53060 + }, + { + "epoch": 6.95, + "learning_rate": 8.889671278569112e-07, + "loss": 0.0015, + "step": 53061 + }, + { + "epoch": 6.95, + "learning_rate": 8.887485522824691e-07, + "loss": 0.0017, + "step": 53062 + }, + { + "epoch": 6.95, + "learning_rate": 8.885300023329358e-07, + "loss": 0.001, + "step": 53063 + }, + { + "epoch": 6.95, + "learning_rate": 8.883114780089242e-07, + "loss": 0.001, + "step": 53064 + }, + { + "epoch": 6.95, + "learning_rate": 8.880929793110515e-07, + "loss": 0.0014, + "step": 53065 + }, + { + "epoch": 6.95, + "learning_rate": 8.878745062399352e-07, + "loss": 0.0013, + "step": 53066 + }, + { + "epoch": 6.95, + "learning_rate": 8.876560587961858e-07, + "loss": 0.0008, + "step": 53067 + }, + { + "epoch": 6.95, + "learning_rate": 8.874376369804183e-07, + "loss": 0.0016, + "step": 53068 + }, + { + "epoch": 6.95, + "learning_rate": 8.872192407932456e-07, + "loss": 0.0019, + "step": 53069 + }, + { + "epoch": 6.95, + "learning_rate": 8.87000870235285e-07, + "loss": 0.0015, + "step": 53070 + }, + { + "epoch": 6.95, + "learning_rate": 8.867825253071493e-07, + "loss": 0.001, + "step": 53071 + }, + { + "epoch": 6.95, + "learning_rate": 8.865642060094503e-07, + "loss": 0.0028, + "step": 53072 + }, + { + "epoch": 6.95, + "learning_rate": 8.863459123428064e-07, + "loss": 0.0013, + "step": 53073 + }, + { + "epoch": 6.95, + "learning_rate": 8.861276443078292e-07, + "loss": 0.001, + "step": 53074 + }, + { + "epoch": 6.95, + "learning_rate": 8.859094019051318e-07, + "loss": 0.0011, + "step": 53075 + }, + { + "epoch": 6.95, + "learning_rate": 8.856911851353279e-07, + "loss": 0.0016, + "step": 53076 + }, + { + "epoch": 6.95, + "learning_rate": 8.854729939990336e-07, + "loss": 0.0042, + "step": 53077 + }, + { + "epoch": 6.95, + "learning_rate": 8.852548284968621e-07, + "loss": 0.0015, + "step": 53078 + }, + { + "epoch": 6.95, + "learning_rate": 8.850366886294248e-07, + "loss": 0.0019, + "step": 53079 + }, + { + "epoch": 6.95, + "learning_rate": 8.848185743973359e-07, + "loss": 0.0017, + "step": 53080 + }, + { + "epoch": 6.95, + "learning_rate": 8.846004858012103e-07, + "loss": 0.0018, + "step": 53081 + }, + { + "epoch": 6.95, + "learning_rate": 8.843824228416609e-07, + "loss": 0.0024, + "step": 53082 + }, + { + "epoch": 6.95, + "learning_rate": 8.841643855192982e-07, + "loss": 0.0012, + "step": 53083 + }, + { + "epoch": 6.95, + "learning_rate": 8.839463738347398e-07, + "loss": 0.0027, + "step": 53084 + }, + { + "epoch": 6.95, + "learning_rate": 8.83728387788596e-07, + "loss": 0.0033, + "step": 53085 + }, + { + "epoch": 6.95, + "learning_rate": 8.835104273814798e-07, + "loss": 0.0015, + "step": 53086 + }, + { + "epoch": 6.95, + "learning_rate": 8.832924926140074e-07, + "loss": 0.0012, + "step": 53087 + }, + { + "epoch": 6.95, + "learning_rate": 8.83074583486786e-07, + "loss": 0.0013, + "step": 53088 + }, + { + "epoch": 6.95, + "learning_rate": 8.828567000004351e-07, + "loss": 0.0025, + "step": 53089 + }, + { + "epoch": 6.95, + "learning_rate": 8.826388421555632e-07, + "loss": 0.001, + "step": 53090 + }, + { + "epoch": 6.95, + "learning_rate": 8.824210099527819e-07, + "loss": 0.0019, + "step": 53091 + }, + { + "epoch": 6.95, + "learning_rate": 8.822032033927086e-07, + "loss": 0.0024, + "step": 53092 + }, + { + "epoch": 6.95, + "learning_rate": 8.819854224759539e-07, + "loss": 0.0018, + "step": 53093 + }, + { + "epoch": 6.95, + "learning_rate": 8.817676672031283e-07, + "loss": 0.001, + "step": 53094 + }, + { + "epoch": 6.96, + "learning_rate": 8.815499375748437e-07, + "loss": 0.0015, + "step": 53095 + }, + { + "epoch": 6.96, + "learning_rate": 8.813322335917174e-07, + "loss": 0.0012, + "step": 53096 + }, + { + "epoch": 6.96, + "learning_rate": 8.811145552543554e-07, + "loss": 0.0015, + "step": 53097 + }, + { + "epoch": 6.96, + "learning_rate": 8.808969025633762e-07, + "loss": 0.0024, + "step": 53098 + }, + { + "epoch": 6.96, + "learning_rate": 8.806792755193861e-07, + "loss": 0.0013, + "step": 53099 + }, + { + "epoch": 6.96, + "learning_rate": 8.804616741230021e-07, + "loss": 0.0012, + "step": 53100 + }, + { + "epoch": 6.96, + "learning_rate": 8.802440983748339e-07, + "loss": 0.0017, + "step": 53101 + }, + { + "epoch": 6.96, + "learning_rate": 8.800265482754932e-07, + "loss": 0.0017, + "step": 53102 + }, + { + "epoch": 6.96, + "learning_rate": 8.798090238255908e-07, + "loss": 0.0013, + "step": 53103 + }, + { + "epoch": 6.96, + "learning_rate": 8.795915250257414e-07, + "loss": 0.0013, + "step": 53104 + }, + { + "epoch": 6.96, + "learning_rate": 8.793740518765559e-07, + "loss": 0.0013, + "step": 53105 + }, + { + "epoch": 6.96, + "learning_rate": 8.791566043786426e-07, + "loss": 0.001, + "step": 53106 + }, + { + "epoch": 6.96, + "learning_rate": 8.789391825326177e-07, + "loss": 0.0012, + "step": 53107 + }, + { + "epoch": 6.96, + "learning_rate": 8.787217863390885e-07, + "loss": 0.0016, + "step": 53108 + }, + { + "epoch": 6.96, + "learning_rate": 8.78504415798671e-07, + "loss": 0.0015, + "step": 53109 + }, + { + "epoch": 6.96, + "learning_rate": 8.782870709119729e-07, + "loss": 0.0021, + "step": 53110 + }, + { + "epoch": 6.96, + "learning_rate": 8.780697516796055e-07, + "loss": 0.0014, + "step": 53111 + }, + { + "epoch": 6.96, + "learning_rate": 8.778524581021842e-07, + "loss": 0.0019, + "step": 53112 + }, + { + "epoch": 6.96, + "learning_rate": 8.776351901803148e-07, + "loss": 0.0026, + "step": 53113 + }, + { + "epoch": 6.96, + "learning_rate": 8.774179479146106e-07, + "loss": 0.0017, + "step": 53114 + }, + { + "epoch": 6.96, + "learning_rate": 8.772007313056841e-07, + "loss": 0.0025, + "step": 53115 + }, + { + "epoch": 6.96, + "learning_rate": 8.769835403541438e-07, + "loss": 0.0011, + "step": 53116 + }, + { + "epoch": 6.96, + "learning_rate": 8.767663750606026e-07, + "loss": 0.0021, + "step": 53117 + }, + { + "epoch": 6.96, + "learning_rate": 8.765492354256677e-07, + "loss": 0.001, + "step": 53118 + }, + { + "epoch": 6.96, + "learning_rate": 8.763321214499521e-07, + "loss": 0.0011, + "step": 53119 + }, + { + "epoch": 6.96, + "learning_rate": 8.761150331340684e-07, + "loss": 0.0017, + "step": 53120 + }, + { + "epoch": 6.96, + "learning_rate": 8.758979704786253e-07, + "loss": 0.0024, + "step": 53121 + }, + { + "epoch": 6.96, + "learning_rate": 8.75680933484232e-07, + "loss": 0.0015, + "step": 53122 + }, + { + "epoch": 6.96, + "learning_rate": 8.754639221515016e-07, + "loss": 0.0012, + "step": 53123 + }, + { + "epoch": 6.96, + "learning_rate": 8.752469364810435e-07, + "loss": 0.0011, + "step": 53124 + }, + { + "epoch": 6.96, + "learning_rate": 8.75029976473466e-07, + "loss": 0.0019, + "step": 53125 + }, + { + "epoch": 6.96, + "learning_rate": 8.748130421293799e-07, + "loss": 0.001, + "step": 53126 + }, + { + "epoch": 6.96, + "learning_rate": 8.74596133449398e-07, + "loss": 0.0015, + "step": 53127 + }, + { + "epoch": 6.96, + "learning_rate": 8.743792504341286e-07, + "loss": 0.0022, + "step": 53128 + }, + { + "epoch": 6.96, + "learning_rate": 8.741623930841802e-07, + "loss": 0.002, + "step": 53129 + }, + { + "epoch": 6.96, + "learning_rate": 8.739455614001635e-07, + "loss": 0.0025, + "step": 53130 + }, + { + "epoch": 6.96, + "learning_rate": 8.737287553826912e-07, + "loss": 0.0011, + "step": 53131 + }, + { + "epoch": 6.96, + "learning_rate": 8.735119750323706e-07, + "loss": 0.0022, + "step": 53132 + }, + { + "epoch": 6.96, + "learning_rate": 8.73295220349809e-07, + "loss": 0.0007, + "step": 53133 + }, + { + "epoch": 6.96, + "learning_rate": 8.730784913356205e-07, + "loss": 0.0013, + "step": 53134 + }, + { + "epoch": 6.96, + "learning_rate": 8.728617879904133e-07, + "loss": 0.0023, + "step": 53135 + }, + { + "epoch": 6.96, + "learning_rate": 8.72645110314796e-07, + "loss": 0.0019, + "step": 53136 + }, + { + "epoch": 6.96, + "learning_rate": 8.724284583093768e-07, + "loss": 0.0018, + "step": 53137 + }, + { + "epoch": 6.96, + "learning_rate": 8.722118319747685e-07, + "loss": 0.0015, + "step": 53138 + }, + { + "epoch": 6.96, + "learning_rate": 8.719952313115776e-07, + "loss": 0.0013, + "step": 53139 + }, + { + "epoch": 6.96, + "learning_rate": 8.717786563204145e-07, + "loss": 0.0019, + "step": 53140 + }, + { + "epoch": 6.96, + "learning_rate": 8.715621070018842e-07, + "loss": 0.0011, + "step": 53141 + }, + { + "epoch": 6.96, + "learning_rate": 8.71345583356602e-07, + "loss": 0.0028, + "step": 53142 + }, + { + "epoch": 6.96, + "learning_rate": 8.711290853851751e-07, + "loss": 0.0015, + "step": 53143 + }, + { + "epoch": 6.96, + "learning_rate": 8.709126130882117e-07, + "loss": 0.0015, + "step": 53144 + }, + { + "epoch": 6.96, + "learning_rate": 8.706961664663183e-07, + "loss": 0.0012, + "step": 53145 + }, + { + "epoch": 6.96, + "learning_rate": 8.704797455201075e-07, + "loss": 0.0012, + "step": 53146 + }, + { + "epoch": 6.96, + "learning_rate": 8.702633502501867e-07, + "loss": 0.0013, + "step": 53147 + }, + { + "epoch": 6.96, + "learning_rate": 8.700469806571621e-07, + "loss": 0.001, + "step": 53148 + }, + { + "epoch": 6.96, + "learning_rate": 8.698306367416454e-07, + "loss": 0.0008, + "step": 53149 + }, + { + "epoch": 6.96, + "learning_rate": 8.696143185042439e-07, + "loss": 0.0034, + "step": 53150 + }, + { + "epoch": 6.96, + "learning_rate": 8.69398025945567e-07, + "loss": 0.0016, + "step": 53151 + }, + { + "epoch": 6.96, + "learning_rate": 8.691817590662188e-07, + "loss": 0.001, + "step": 53152 + }, + { + "epoch": 6.96, + "learning_rate": 8.689655178668111e-07, + "loss": 0.0016, + "step": 53153 + }, + { + "epoch": 6.96, + "learning_rate": 8.687493023479531e-07, + "loss": 0.0017, + "step": 53154 + }, + { + "epoch": 6.96, + "learning_rate": 8.685331125102525e-07, + "loss": 0.0016, + "step": 53155 + }, + { + "epoch": 6.96, + "learning_rate": 8.683169483543142e-07, + "loss": 0.001, + "step": 53156 + }, + { + "epoch": 6.96, + "learning_rate": 8.681008098807486e-07, + "loss": 0.0009, + "step": 53157 + }, + { + "epoch": 6.96, + "learning_rate": 8.678846970901644e-07, + "loss": 0.0014, + "step": 53158 + }, + { + "epoch": 6.96, + "learning_rate": 8.676686099831677e-07, + "loss": 0.0016, + "step": 53159 + }, + { + "epoch": 6.96, + "learning_rate": 8.674525485603657e-07, + "loss": 0.0015, + "step": 53160 + }, + { + "epoch": 6.96, + "learning_rate": 8.672365128223681e-07, + "loss": 0.002, + "step": 53161 + }, + { + "epoch": 6.96, + "learning_rate": 8.67020502769782e-07, + "loss": 0.0026, + "step": 53162 + }, + { + "epoch": 6.96, + "learning_rate": 8.668045184032115e-07, + "loss": 0.0012, + "step": 53163 + }, + { + "epoch": 6.96, + "learning_rate": 8.665885597232682e-07, + "loss": 0.0027, + "step": 53164 + }, + { + "epoch": 6.96, + "learning_rate": 8.663726267305595e-07, + "loss": 0.0011, + "step": 53165 + }, + { + "epoch": 6.96, + "learning_rate": 8.661567194256914e-07, + "loss": 0.0009, + "step": 53166 + }, + { + "epoch": 6.96, + "learning_rate": 8.659408378092704e-07, + "loss": 0.0014, + "step": 53167 + }, + { + "epoch": 6.96, + "learning_rate": 8.657249818819036e-07, + "loss": 0.0016, + "step": 53168 + }, + { + "epoch": 6.96, + "learning_rate": 8.655091516442005e-07, + "loss": 0.0012, + "step": 53169 + }, + { + "epoch": 6.96, + "learning_rate": 8.652933470967661e-07, + "loss": 0.002, + "step": 53170 + }, + { + "epoch": 6.97, + "learning_rate": 8.650775682402058e-07, + "loss": 0.0015, + "step": 53171 + }, + { + "epoch": 6.97, + "learning_rate": 8.648618150751309e-07, + "loss": 0.0007, + "step": 53172 + }, + { + "epoch": 6.97, + "learning_rate": 8.646460876021445e-07, + "loss": 0.0041, + "step": 53173 + }, + { + "epoch": 6.97, + "learning_rate": 8.644303858218539e-07, + "loss": 0.0021, + "step": 53174 + }, + { + "epoch": 6.97, + "learning_rate": 8.642147097348674e-07, + "loss": 0.001, + "step": 53175 + }, + { + "epoch": 6.97, + "learning_rate": 8.639990593417891e-07, + "loss": 0.0021, + "step": 53176 + }, + { + "epoch": 6.97, + "learning_rate": 8.637834346432284e-07, + "loss": 0.0011, + "step": 53177 + }, + { + "epoch": 6.97, + "learning_rate": 8.635678356397903e-07, + "loss": 0.0015, + "step": 53178 + }, + { + "epoch": 6.97, + "learning_rate": 8.6335226233208e-07, + "loss": 0.0021, + "step": 53179 + }, + { + "epoch": 6.97, + "learning_rate": 8.63136714720706e-07, + "loss": 0.0011, + "step": 53180 + }, + { + "epoch": 6.97, + "learning_rate": 8.62921192806273e-07, + "loss": 0.0013, + "step": 53181 + }, + { + "epoch": 6.97, + "learning_rate": 8.627056965893887e-07, + "loss": 0.0012, + "step": 53182 + }, + { + "epoch": 6.97, + "learning_rate": 8.624902260706558e-07, + "loss": 0.0022, + "step": 53183 + }, + { + "epoch": 6.97, + "learning_rate": 8.622747812506837e-07, + "loss": 0.0015, + "step": 53184 + }, + { + "epoch": 6.97, + "learning_rate": 8.620593621300766e-07, + "loss": 0.0016, + "step": 53185 + }, + { + "epoch": 6.97, + "learning_rate": 8.618439687094415e-07, + "loss": 0.0025, + "step": 53186 + }, + { + "epoch": 6.97, + "learning_rate": 8.616286009893826e-07, + "loss": 0.0007, + "step": 53187 + }, + { + "epoch": 6.97, + "learning_rate": 8.614132589705082e-07, + "loss": 0.0015, + "step": 53188 + }, + { + "epoch": 6.97, + "learning_rate": 8.611979426534222e-07, + "loss": 0.0012, + "step": 53189 + }, + { + "epoch": 6.97, + "learning_rate": 8.609826520387299e-07, + "loss": 0.0006, + "step": 53190 + }, + { + "epoch": 6.97, + "learning_rate": 8.607673871270361e-07, + "loss": 0.0016, + "step": 53191 + }, + { + "epoch": 6.97, + "learning_rate": 8.605521479189493e-07, + "loss": 0.001, + "step": 53192 + }, + { + "epoch": 6.97, + "learning_rate": 8.603369344150725e-07, + "loss": 0.0015, + "step": 53193 + }, + { + "epoch": 6.97, + "learning_rate": 8.601217466160094e-07, + "loss": 0.0018, + "step": 53194 + }, + { + "epoch": 6.97, + "learning_rate": 8.599065845223675e-07, + "loss": 0.0031, + "step": 53195 + }, + { + "epoch": 6.97, + "learning_rate": 8.596914481347529e-07, + "loss": 0.0015, + "step": 53196 + }, + { + "epoch": 6.97, + "learning_rate": 8.594763374537695e-07, + "loss": 0.0011, + "step": 53197 + }, + { + "epoch": 6.97, + "learning_rate": 8.592612524800204e-07, + "loss": 0.0013, + "step": 53198 + }, + { + "epoch": 6.97, + "learning_rate": 8.590461932141148e-07, + "loss": 0.0009, + "step": 53199 + }, + { + "epoch": 6.97, + "learning_rate": 8.588311596566546e-07, + "loss": 0.0026, + "step": 53200 + }, + { + "epoch": 6.97, + "learning_rate": 8.586161518082437e-07, + "loss": 0.0011, + "step": 53201 + }, + { + "epoch": 6.97, + "learning_rate": 8.584011696694883e-07, + "loss": 0.0016, + "step": 53202 + }, + { + "epoch": 6.97, + "learning_rate": 8.581862132409935e-07, + "loss": 0.0008, + "step": 53203 + }, + { + "epoch": 6.97, + "learning_rate": 8.579712825233632e-07, + "loss": 0.0007, + "step": 53204 + }, + { + "epoch": 6.97, + "learning_rate": 8.577563775172015e-07, + "loss": 0.0019, + "step": 53205 + }, + { + "epoch": 6.97, + "learning_rate": 8.575414982231101e-07, + "loss": 0.001, + "step": 53206 + }, + { + "epoch": 6.97, + "learning_rate": 8.573266446416995e-07, + "loss": 0.0015, + "step": 53207 + }, + { + "epoch": 6.97, + "learning_rate": 8.571118167735715e-07, + "loss": 0.001, + "step": 53208 + }, + { + "epoch": 6.97, + "learning_rate": 8.56897014619329e-07, + "loss": 0.0017, + "step": 53209 + }, + { + "epoch": 6.97, + "learning_rate": 8.566822381795747e-07, + "loss": 0.0014, + "step": 53210 + }, + { + "epoch": 6.97, + "learning_rate": 8.564674874549172e-07, + "loss": 0.0018, + "step": 53211 + }, + { + "epoch": 6.97, + "learning_rate": 8.562527624459582e-07, + "loss": 0.0017, + "step": 53212 + }, + { + "epoch": 6.97, + "learning_rate": 8.560380631532994e-07, + "loss": 0.0012, + "step": 53213 + }, + { + "epoch": 6.97, + "learning_rate": 8.558233895775481e-07, + "loss": 0.0022, + "step": 53214 + }, + { + "epoch": 6.97, + "learning_rate": 8.556087417193059e-07, + "loss": 0.0021, + "step": 53215 + }, + { + "epoch": 6.97, + "learning_rate": 8.553941195791782e-07, + "loss": 0.0017, + "step": 53216 + }, + { + "epoch": 6.97, + "learning_rate": 8.551795231577653e-07, + "loss": 0.0014, + "step": 53217 + }, + { + "epoch": 6.97, + "learning_rate": 8.549649524556725e-07, + "loss": 0.0013, + "step": 53218 + }, + { + "epoch": 6.97, + "learning_rate": 8.547504074735069e-07, + "loss": 0.0013, + "step": 53219 + }, + { + "epoch": 6.97, + "learning_rate": 8.545358882118681e-07, + "loss": 0.0021, + "step": 53220 + }, + { + "epoch": 6.97, + "learning_rate": 8.54321394671358e-07, + "loss": 0.0021, + "step": 53221 + }, + { + "epoch": 6.97, + "learning_rate": 8.541069268525837e-07, + "loss": 0.0017, + "step": 53222 + }, + { + "epoch": 6.97, + "learning_rate": 8.53892484756147e-07, + "loss": 0.0025, + "step": 53223 + }, + { + "epoch": 6.97, + "learning_rate": 8.536780683826507e-07, + "loss": 0.0012, + "step": 53224 + }, + { + "epoch": 6.97, + "learning_rate": 8.534636777326955e-07, + "loss": 0.0019, + "step": 53225 + }, + { + "epoch": 6.97, + "learning_rate": 8.532493128068887e-07, + "loss": 0.0008, + "step": 53226 + }, + { + "epoch": 6.97, + "learning_rate": 8.530349736058307e-07, + "loss": 0.0006, + "step": 53227 + }, + { + "epoch": 6.97, + "learning_rate": 8.528206601301236e-07, + "loss": 0.0017, + "step": 53228 + }, + { + "epoch": 6.97, + "learning_rate": 8.526063723803713e-07, + "loss": 0.002, + "step": 53229 + }, + { + "epoch": 6.97, + "learning_rate": 8.523921103571775e-07, + "loss": 0.0016, + "step": 53230 + }, + { + "epoch": 6.97, + "learning_rate": 8.521778740611442e-07, + "loss": 0.0005, + "step": 53231 + }, + { + "epoch": 6.97, + "learning_rate": 8.51963663492874e-07, + "loss": 0.0019, + "step": 53232 + }, + { + "epoch": 6.97, + "learning_rate": 8.517494786529667e-07, + "loss": 0.0007, + "step": 53233 + }, + { + "epoch": 6.97, + "learning_rate": 8.515353195420284e-07, + "loss": 0.0012, + "step": 53234 + }, + { + "epoch": 6.97, + "learning_rate": 8.513211861606596e-07, + "loss": 0.0014, + "step": 53235 + }, + { + "epoch": 6.97, + "learning_rate": 8.51107078509461e-07, + "loss": 0.0023, + "step": 53236 + }, + { + "epoch": 6.97, + "learning_rate": 8.508929965890389e-07, + "loss": 0.0029, + "step": 53237 + }, + { + "epoch": 6.97, + "learning_rate": 8.506789403999927e-07, + "loss": 0.0011, + "step": 53238 + }, + { + "epoch": 6.97, + "learning_rate": 8.504649099429229e-07, + "loss": 0.0026, + "step": 53239 + }, + { + "epoch": 6.97, + "learning_rate": 8.502509052184349e-07, + "loss": 0.001, + "step": 53240 + }, + { + "epoch": 6.97, + "learning_rate": 8.50036926227128e-07, + "loss": 0.0011, + "step": 53241 + }, + { + "epoch": 6.97, + "learning_rate": 8.498229729696061e-07, + "loss": 0.0009, + "step": 53242 + }, + { + "epoch": 6.97, + "learning_rate": 8.496090454464701e-07, + "loss": 0.0009, + "step": 53243 + }, + { + "epoch": 6.97, + "learning_rate": 8.493951436583192e-07, + "loss": 0.0029, + "step": 53244 + }, + { + "epoch": 6.97, + "learning_rate": 8.491812676057587e-07, + "loss": 0.0014, + "step": 53245 + }, + { + "epoch": 6.97, + "learning_rate": 8.489674172893891e-07, + "loss": 0.0012, + "step": 53246 + }, + { + "epoch": 6.97, + "learning_rate": 8.487535927098112e-07, + "loss": 0.0025, + "step": 53247 + }, + { + "epoch": 6.98, + "learning_rate": 8.485397938676254e-07, + "loss": 0.0008, + "step": 53248 + }, + { + "epoch": 6.98, + "learning_rate": 8.483260207634348e-07, + "loss": 0.0014, + "step": 53249 + }, + { + "epoch": 6.98, + "learning_rate": 8.481122733978386e-07, + "loss": 0.0018, + "step": 53250 + }, + { + "epoch": 6.98, + "learning_rate": 8.478985517714411e-07, + "loss": 0.0017, + "step": 53251 + }, + { + "epoch": 6.98, + "learning_rate": 8.476848558848394e-07, + "loss": 0.0018, + "step": 53252 + }, + { + "epoch": 6.98, + "learning_rate": 8.474711857386386e-07, + "loss": 0.0024, + "step": 53253 + }, + { + "epoch": 6.98, + "learning_rate": 8.472575413334383e-07, + "loss": 0.002, + "step": 53254 + }, + { + "epoch": 6.98, + "learning_rate": 8.47043922669838e-07, + "loss": 0.0013, + "step": 53255 + }, + { + "epoch": 6.98, + "learning_rate": 8.468303297484382e-07, + "loss": 0.0011, + "step": 53256 + }, + { + "epoch": 6.98, + "learning_rate": 8.466167625698418e-07, + "loss": 0.0011, + "step": 53257 + }, + { + "epoch": 6.98, + "learning_rate": 8.464032211346485e-07, + "loss": 0.0016, + "step": 53258 + }, + { + "epoch": 6.98, + "learning_rate": 8.461897054434564e-07, + "loss": 0.002, + "step": 53259 + }, + { + "epoch": 6.98, + "learning_rate": 8.459762154968709e-07, + "loss": 0.0015, + "step": 53260 + }, + { + "epoch": 6.98, + "learning_rate": 8.45762751295488e-07, + "loss": 0.0032, + "step": 53261 + }, + { + "epoch": 6.98, + "learning_rate": 8.455493128399106e-07, + "loss": 0.0014, + "step": 53262 + }, + { + "epoch": 6.98, + "learning_rate": 8.453359001307371e-07, + "loss": 0.0016, + "step": 53263 + }, + { + "epoch": 6.98, + "learning_rate": 8.451225131685714e-07, + "loss": 0.0038, + "step": 53264 + }, + { + "epoch": 6.98, + "learning_rate": 8.449091519540098e-07, + "loss": 0.0013, + "step": 53265 + }, + { + "epoch": 6.98, + "learning_rate": 8.446958164876551e-07, + "loss": 0.0025, + "step": 53266 + }, + { + "epoch": 6.98, + "learning_rate": 8.444825067701023e-07, + "loss": 0.0013, + "step": 53267 + }, + { + "epoch": 6.98, + "learning_rate": 8.442692228019577e-07, + "loss": 0.0013, + "step": 53268 + }, + { + "epoch": 6.98, + "learning_rate": 8.440559645838186e-07, + "loss": 0.0012, + "step": 53269 + }, + { + "epoch": 6.98, + "learning_rate": 8.438427321162846e-07, + "loss": 0.0013, + "step": 53270 + }, + { + "epoch": 6.98, + "learning_rate": 8.436295253999527e-07, + "loss": 0.0017, + "step": 53271 + }, + { + "epoch": 6.98, + "learning_rate": 8.43416344435426e-07, + "loss": 0.0011, + "step": 53272 + }, + { + "epoch": 6.98, + "learning_rate": 8.43203189223305e-07, + "loss": 0.0012, + "step": 53273 + }, + { + "epoch": 6.98, + "learning_rate": 8.42990059764186e-07, + "loss": 0.0017, + "step": 53274 + }, + { + "epoch": 6.98, + "learning_rate": 8.427769560586696e-07, + "loss": 0.0015, + "step": 53275 + }, + { + "epoch": 6.98, + "learning_rate": 8.425638781073564e-07, + "loss": 0.0016, + "step": 53276 + }, + { + "epoch": 6.98, + "learning_rate": 8.423508259108448e-07, + "loss": 0.0021, + "step": 53277 + }, + { + "epoch": 6.98, + "learning_rate": 8.421377994697321e-07, + "loss": 0.0008, + "step": 53278 + }, + { + "epoch": 6.98, + "learning_rate": 8.419247987846213e-07, + "loss": 0.0016, + "step": 53279 + }, + { + "epoch": 6.98, + "learning_rate": 8.417118238561084e-07, + "loss": 0.0017, + "step": 53280 + }, + { + "epoch": 6.98, + "learning_rate": 8.414988746847941e-07, + "loss": 0.0014, + "step": 53281 + }, + { + "epoch": 6.98, + "learning_rate": 8.412859512712745e-07, + "loss": 0.0016, + "step": 53282 + }, + { + "epoch": 6.98, + "learning_rate": 8.410730536161505e-07, + "loss": 0.0012, + "step": 53283 + }, + { + "epoch": 6.98, + "learning_rate": 8.408601817200224e-07, + "loss": 0.0029, + "step": 53284 + }, + { + "epoch": 6.98, + "learning_rate": 8.406473355834877e-07, + "loss": 0.0014, + "step": 53285 + }, + { + "epoch": 6.98, + "learning_rate": 8.404345152071414e-07, + "loss": 0.0023, + "step": 53286 + }, + { + "epoch": 6.98, + "learning_rate": 8.402217205915886e-07, + "loss": 0.002, + "step": 53287 + }, + { + "epoch": 6.98, + "learning_rate": 8.400089517374233e-07, + "loss": 0.0007, + "step": 53288 + }, + { + "epoch": 6.98, + "learning_rate": 8.39796208645246e-07, + "loss": 0.0021, + "step": 53289 + }, + { + "epoch": 6.98, + "learning_rate": 8.395834913156519e-07, + "loss": 0.0015, + "step": 53290 + }, + { + "epoch": 6.98, + "learning_rate": 8.393707997492428e-07, + "loss": 0.0013, + "step": 53291 + }, + { + "epoch": 6.98, + "learning_rate": 8.39158133946616e-07, + "loss": 0.0011, + "step": 53292 + }, + { + "epoch": 6.98, + "learning_rate": 8.389454939083674e-07, + "loss": 0.0015, + "step": 53293 + }, + { + "epoch": 6.98, + "learning_rate": 8.387328796350969e-07, + "loss": 0.0022, + "step": 53294 + }, + { + "epoch": 6.98, + "learning_rate": 8.385202911274037e-07, + "loss": 0.002, + "step": 53295 + }, + { + "epoch": 6.98, + "learning_rate": 8.383077283858842e-07, + "loss": 0.0018, + "step": 53296 + }, + { + "epoch": 6.98, + "learning_rate": 8.380951914111369e-07, + "loss": 0.0015, + "step": 53297 + }, + { + "epoch": 6.98, + "learning_rate": 8.378826802037565e-07, + "loss": 0.002, + "step": 53298 + }, + { + "epoch": 6.98, + "learning_rate": 8.376701947643451e-07, + "loss": 0.0017, + "step": 53299 + }, + { + "epoch": 6.98, + "learning_rate": 8.374577350934987e-07, + "loss": 0.0011, + "step": 53300 + }, + { + "epoch": 6.98, + "learning_rate": 8.372453011918124e-07, + "loss": 0.0037, + "step": 53301 + }, + { + "epoch": 6.98, + "learning_rate": 8.37032893059887e-07, + "loss": 0.002, + "step": 53302 + }, + { + "epoch": 6.98, + "learning_rate": 8.368205106983196e-07, + "loss": 0.002, + "step": 53303 + }, + { + "epoch": 6.98, + "learning_rate": 8.366081541077065e-07, + "loss": 0.0015, + "step": 53304 + }, + { + "epoch": 6.98, + "learning_rate": 8.363958232886426e-07, + "loss": 0.0019, + "step": 53305 + }, + { + "epoch": 6.98, + "learning_rate": 8.361835182417277e-07, + "loss": 0.0013, + "step": 53306 + }, + { + "epoch": 6.98, + "learning_rate": 8.359712389675601e-07, + "loss": 0.0014, + "step": 53307 + }, + { + "epoch": 6.98, + "learning_rate": 8.357589854667359e-07, + "loss": 0.0014, + "step": 53308 + }, + { + "epoch": 6.98, + "learning_rate": 8.355467577398491e-07, + "loss": 0.0012, + "step": 53309 + }, + { + "epoch": 6.98, + "learning_rate": 8.353345557875014e-07, + "loss": 0.001, + "step": 53310 + }, + { + "epoch": 6.98, + "learning_rate": 8.35122379610287e-07, + "loss": 0.0014, + "step": 53311 + }, + { + "epoch": 6.98, + "learning_rate": 8.349102292088029e-07, + "loss": 0.0021, + "step": 53312 + }, + { + "epoch": 6.98, + "learning_rate": 8.346981045836444e-07, + "loss": 0.0008, + "step": 53313 + }, + { + "epoch": 6.98, + "learning_rate": 8.344860057354109e-07, + "loss": 0.0024, + "step": 53314 + }, + { + "epoch": 6.98, + "learning_rate": 8.342739326646965e-07, + "loss": 0.0023, + "step": 53315 + }, + { + "epoch": 6.98, + "learning_rate": 8.340618853720983e-07, + "loss": 0.0019, + "step": 53316 + }, + { + "epoch": 6.98, + "learning_rate": 8.338498638582116e-07, + "loss": 0.0006, + "step": 53317 + }, + { + "epoch": 6.98, + "learning_rate": 8.336378681236368e-07, + "loss": 0.0024, + "step": 53318 + }, + { + "epoch": 6.98, + "learning_rate": 8.334258981689669e-07, + "loss": 0.001, + "step": 53319 + }, + { + "epoch": 6.98, + "learning_rate": 8.332139539947981e-07, + "loss": 0.0029, + "step": 53320 + }, + { + "epoch": 6.98, + "learning_rate": 8.330020356017254e-07, + "loss": 0.0017, + "step": 53321 + }, + { + "epoch": 6.98, + "learning_rate": 8.327901429903484e-07, + "loss": 0.0019, + "step": 53322 + }, + { + "epoch": 6.98, + "learning_rate": 8.32578276161261e-07, + "loss": 0.0015, + "step": 53323 + }, + { + "epoch": 6.99, + "learning_rate": 8.323664351150573e-07, + "loss": 0.0014, + "step": 53324 + }, + { + "epoch": 6.99, + "learning_rate": 8.321546198523356e-07, + "loss": 0.0011, + "step": 53325 + }, + { + "epoch": 6.99, + "learning_rate": 8.319428303736898e-07, + "loss": 0.001, + "step": 53326 + }, + { + "epoch": 6.99, + "learning_rate": 8.317310666797185e-07, + "loss": 0.0024, + "step": 53327 + }, + { + "epoch": 6.99, + "learning_rate": 8.315193287710132e-07, + "loss": 0.0022, + "step": 53328 + }, + { + "epoch": 6.99, + "learning_rate": 8.313076166481737e-07, + "loss": 0.0012, + "step": 53329 + }, + { + "epoch": 6.99, + "learning_rate": 8.310959303117926e-07, + "loss": 0.0007, + "step": 53330 + }, + { + "epoch": 6.99, + "learning_rate": 8.308842697624664e-07, + "loss": 0.0017, + "step": 53331 + }, + { + "epoch": 6.99, + "learning_rate": 8.306726350007888e-07, + "loss": 0.0017, + "step": 53332 + }, + { + "epoch": 6.99, + "learning_rate": 8.304610260273572e-07, + "loss": 0.0016, + "step": 53333 + }, + { + "epoch": 6.99, + "learning_rate": 8.302494428427654e-07, + "loss": 0.0027, + "step": 53334 + }, + { + "epoch": 6.99, + "learning_rate": 8.300378854476087e-07, + "loss": 0.0012, + "step": 53335 + }, + { + "epoch": 6.99, + "learning_rate": 8.2982635384248e-07, + "loss": 0.003, + "step": 53336 + }, + { + "epoch": 6.99, + "learning_rate": 8.296148480279775e-07, + "loss": 0.0012, + "step": 53337 + }, + { + "epoch": 6.99, + "learning_rate": 8.294033680046965e-07, + "loss": 0.0019, + "step": 53338 + }, + { + "epoch": 6.99, + "learning_rate": 8.291919137732296e-07, + "loss": 0.0009, + "step": 53339 + }, + { + "epoch": 6.99, + "learning_rate": 8.289804853341698e-07, + "loss": 0.0009, + "step": 53340 + }, + { + "epoch": 6.99, + "learning_rate": 8.287690826881167e-07, + "loss": 0.0027, + "step": 53341 + }, + { + "epoch": 6.99, + "learning_rate": 8.285577058356619e-07, + "loss": 0.0011, + "step": 53342 + }, + { + "epoch": 6.99, + "learning_rate": 8.283463547773985e-07, + "loss": 0.0016, + "step": 53343 + }, + { + "epoch": 6.99, + "learning_rate": 8.281350295139234e-07, + "loss": 0.0012, + "step": 53344 + }, + { + "epoch": 6.99, + "learning_rate": 8.279237300458309e-07, + "loss": 0.0013, + "step": 53345 + }, + { + "epoch": 6.99, + "learning_rate": 8.277124563737149e-07, + "loss": 0.0017, + "step": 53346 + }, + { + "epoch": 6.99, + "learning_rate": 8.27501208498167e-07, + "loss": 0.0024, + "step": 53347 + }, + { + "epoch": 6.99, + "learning_rate": 8.272899864197836e-07, + "loss": 0.0015, + "step": 53348 + }, + { + "epoch": 6.99, + "learning_rate": 8.270787901391597e-07, + "loss": 0.0013, + "step": 53349 + }, + { + "epoch": 6.99, + "learning_rate": 8.268676196568892e-07, + "loss": 0.0013, + "step": 53350 + }, + { + "epoch": 6.99, + "learning_rate": 8.266564749735629e-07, + "loss": 0.001, + "step": 53351 + }, + { + "epoch": 6.99, + "learning_rate": 8.264453560897778e-07, + "loss": 0.001, + "step": 53352 + }, + { + "epoch": 6.99, + "learning_rate": 8.262342630061271e-07, + "loss": 0.0016, + "step": 53353 + }, + { + "epoch": 6.99, + "learning_rate": 8.260231957232045e-07, + "loss": 0.002, + "step": 53354 + }, + { + "epoch": 6.99, + "learning_rate": 8.258121542416008e-07, + "loss": 0.0013, + "step": 53355 + }, + { + "epoch": 6.99, + "learning_rate": 8.256011385619144e-07, + "loss": 0.0013, + "step": 53356 + }, + { + "epoch": 6.99, + "learning_rate": 8.253901486847349e-07, + "loss": 0.0015, + "step": 53357 + }, + { + "epoch": 6.99, + "learning_rate": 8.251791846106561e-07, + "loss": 0.0008, + "step": 53358 + }, + { + "epoch": 6.99, + "learning_rate": 8.249682463402719e-07, + "loss": 0.0017, + "step": 53359 + }, + { + "epoch": 6.99, + "learning_rate": 8.247573338741788e-07, + "loss": 0.0007, + "step": 53360 + }, + { + "epoch": 6.99, + "learning_rate": 8.24546447212966e-07, + "loss": 0.0018, + "step": 53361 + }, + { + "epoch": 6.99, + "learning_rate": 8.243355863572278e-07, + "loss": 0.0021, + "step": 53362 + }, + { + "epoch": 6.99, + "learning_rate": 8.241247513075556e-07, + "loss": 0.0015, + "step": 53363 + }, + { + "epoch": 6.99, + "learning_rate": 8.239139420645458e-07, + "loss": 0.0009, + "step": 53364 + }, + { + "epoch": 6.99, + "learning_rate": 8.237031586287892e-07, + "loss": 0.0015, + "step": 53365 + }, + { + "epoch": 6.99, + "learning_rate": 8.23492401000876e-07, + "loss": 0.0027, + "step": 53366 + }, + { + "epoch": 6.99, + "learning_rate": 8.232816691814049e-07, + "loss": 0.0014, + "step": 53367 + }, + { + "epoch": 6.99, + "learning_rate": 8.230709631709655e-07, + "loss": 0.001, + "step": 53368 + }, + { + "epoch": 6.99, + "learning_rate": 8.228602829701493e-07, + "loss": 0.002, + "step": 53369 + }, + { + "epoch": 6.99, + "learning_rate": 8.226496285795482e-07, + "loss": 0.0012, + "step": 53370 + }, + { + "epoch": 6.99, + "learning_rate": 8.224389999997573e-07, + "loss": 0.0019, + "step": 53371 + }, + { + "epoch": 6.99, + "learning_rate": 8.222283972313694e-07, + "loss": 0.0023, + "step": 53372 + }, + { + "epoch": 6.99, + "learning_rate": 8.220178202749751e-07, + "loss": 0.0008, + "step": 53373 + }, + { + "epoch": 6.99, + "learning_rate": 8.218072691311652e-07, + "loss": 0.0016, + "step": 53374 + }, + { + "epoch": 6.99, + "learning_rate": 8.215967438005346e-07, + "loss": 0.002, + "step": 53375 + }, + { + "epoch": 6.99, + "learning_rate": 8.213862442836751e-07, + "loss": 0.0015, + "step": 53376 + }, + { + "epoch": 6.99, + "learning_rate": 8.211757705811785e-07, + "loss": 0.001, + "step": 53377 + }, + { + "epoch": 6.99, + "learning_rate": 8.209653226936332e-07, + "loss": 0.0022, + "step": 53378 + }, + { + "epoch": 6.99, + "learning_rate": 8.207549006216375e-07, + "loss": 0.0017, + "step": 53379 + }, + { + "epoch": 6.99, + "learning_rate": 8.205445043657789e-07, + "loss": 0.001, + "step": 53380 + }, + { + "epoch": 6.99, + "learning_rate": 8.20334133926648e-07, + "loss": 0.002, + "step": 53381 + }, + { + "epoch": 6.99, + "learning_rate": 8.201237893048386e-07, + "loss": 0.0028, + "step": 53382 + }, + { + "epoch": 6.99, + "learning_rate": 8.199134705009448e-07, + "loss": 0.003, + "step": 53383 + }, + { + "epoch": 6.99, + "learning_rate": 8.19703177515555e-07, + "loss": 0.0013, + "step": 53384 + }, + { + "epoch": 6.99, + "learning_rate": 8.194929103492611e-07, + "loss": 0.0018, + "step": 53385 + }, + { + "epoch": 6.99, + "learning_rate": 8.192826690026523e-07, + "loss": 0.0008, + "step": 53386 + }, + { + "epoch": 6.99, + "learning_rate": 8.190724534763239e-07, + "loss": 0.0012, + "step": 53387 + }, + { + "epoch": 6.99, + "learning_rate": 8.188622637708654e-07, + "loss": 0.002, + "step": 53388 + }, + { + "epoch": 6.99, + "learning_rate": 8.186520998868663e-07, + "loss": 0.0013, + "step": 53389 + }, + { + "epoch": 6.99, + "learning_rate": 8.184419618249217e-07, + "loss": 0.0009, + "step": 53390 + }, + { + "epoch": 6.99, + "learning_rate": 8.18231849585619e-07, + "loss": 0.0018, + "step": 53391 + }, + { + "epoch": 6.99, + "learning_rate": 8.180217631695497e-07, + "loss": 0.0018, + "step": 53392 + }, + { + "epoch": 6.99, + "learning_rate": 8.178117025773036e-07, + "loss": 0.0019, + "step": 53393 + }, + { + "epoch": 6.99, + "learning_rate": 8.176016678094767e-07, + "loss": 0.0013, + "step": 53394 + }, + { + "epoch": 6.99, + "learning_rate": 8.173916588666552e-07, + "loss": 0.0008, + "step": 53395 + }, + { + "epoch": 6.99, + "learning_rate": 8.1718167574943e-07, + "loss": 0.0019, + "step": 53396 + }, + { + "epoch": 6.99, + "learning_rate": 8.169717184583914e-07, + "loss": 0.0015, + "step": 53397 + }, + { + "epoch": 6.99, + "learning_rate": 8.167617869941324e-07, + "loss": 0.0009, + "step": 53398 + }, + { + "epoch": 6.99, + "learning_rate": 8.165518813572415e-07, + "loss": 0.0021, + "step": 53399 + }, + { + "epoch": 7.0, + "learning_rate": 8.163420015483082e-07, + "loss": 0.0012, + "step": 53400 + }, + { + "epoch": 7.0, + "learning_rate": 8.161321475679252e-07, + "loss": 0.0011, + "step": 53401 + }, + { + "epoch": 7.0, + "learning_rate": 8.159223194166821e-07, + "loss": 0.0013, + "step": 53402 + }, + { + "epoch": 7.0, + "learning_rate": 8.157125170951663e-07, + "loss": 0.001, + "step": 53403 + }, + { + "epoch": 7.0, + "learning_rate": 8.155027406039717e-07, + "loss": 0.0015, + "step": 53404 + }, + { + "epoch": 7.0, + "learning_rate": 8.152929899436845e-07, + "loss": 0.0012, + "step": 53405 + }, + { + "epoch": 7.0, + "learning_rate": 8.150832651148988e-07, + "loss": 0.0012, + "step": 53406 + }, + { + "epoch": 7.0, + "learning_rate": 8.148735661182017e-07, + "loss": 0.0017, + "step": 53407 + }, + { + "epoch": 7.0, + "learning_rate": 8.146638929541828e-07, + "loss": 0.0013, + "step": 53408 + }, + { + "epoch": 7.0, + "learning_rate": 8.144542456234339e-07, + "loss": 0.001, + "step": 53409 + }, + { + "epoch": 7.0, + "learning_rate": 8.142446241265434e-07, + "loss": 0.0016, + "step": 53410 + }, + { + "epoch": 7.0, + "learning_rate": 8.140350284640997e-07, + "loss": 0.0016, + "step": 53411 + }, + { + "epoch": 7.0, + "learning_rate": 8.138254586366923e-07, + "loss": 0.0014, + "step": 53412 + }, + { + "epoch": 7.0, + "learning_rate": 8.136159146449141e-07, + "loss": 0.0016, + "step": 53413 + }, + { + "epoch": 7.0, + "learning_rate": 8.134063964893491e-07, + "loss": 0.0015, + "step": 53414 + }, + { + "epoch": 7.0, + "learning_rate": 8.131969041705912e-07, + "loss": 0.003, + "step": 53415 + }, + { + "epoch": 7.0, + "learning_rate": 8.129874376892266e-07, + "loss": 0.0012, + "step": 53416 + }, + { + "epoch": 7.0, + "learning_rate": 8.127779970458471e-07, + "loss": 0.0017, + "step": 53417 + }, + { + "epoch": 7.0, + "learning_rate": 8.125685822410401e-07, + "loss": 0.0014, + "step": 53418 + }, + { + "epoch": 7.0, + "learning_rate": 8.123591932753949e-07, + "loss": 0.0009, + "step": 53419 + }, + { + "epoch": 7.0, + "learning_rate": 8.121498301494979e-07, + "loss": 0.0024, + "step": 53420 + }, + { + "epoch": 7.0, + "learning_rate": 8.119404928639429e-07, + "loss": 0.0009, + "step": 53421 + }, + { + "epoch": 7.0, + "learning_rate": 8.117311814193152e-07, + "loss": 0.0015, + "step": 53422 + }, + { + "epoch": 7.0, + "learning_rate": 8.11521895816203e-07, + "loss": 0.0014, + "step": 53423 + }, + { + "epoch": 7.0, + "learning_rate": 8.113126360551981e-07, + "loss": 0.0011, + "step": 53424 + }, + { + "epoch": 7.0, + "learning_rate": 8.111034021368847e-07, + "loss": 0.0018, + "step": 53425 + }, + { + "epoch": 7.0, + "learning_rate": 8.108941940618554e-07, + "loss": 0.0024, + "step": 53426 + }, + { + "epoch": 7.0, + "learning_rate": 8.106850118306975e-07, + "loss": 0.0023, + "step": 53427 + }, + { + "epoch": 7.0, + "learning_rate": 8.104758554439962e-07, + "loss": 0.002, + "step": 53428 + }, + { + "epoch": 7.0, + "learning_rate": 8.102667249023444e-07, + "loss": 0.001, + "step": 53429 + }, + { + "epoch": 7.0, + "learning_rate": 8.100576202063282e-07, + "loss": 0.0007, + "step": 53430 + }, + { + "epoch": 7.0, + "learning_rate": 8.098485413565326e-07, + "loss": 0.0015, + "step": 53431 + }, + { + "epoch": 7.0, + "learning_rate": 8.096394883535519e-07, + "loss": 0.0013, + "step": 53432 + }, + { + "epoch": 7.0, + "learning_rate": 8.094304611979698e-07, + "loss": 0.0012, + "step": 53433 + }, + { + "epoch": 7.0, + "learning_rate": 8.092214598903747e-07, + "loss": 0.0011, + "step": 53434 + }, + { + "epoch": 7.0, + "learning_rate": 8.09012484431354e-07, + "loss": 0.0026, + "step": 53435 + }, + { + "epoch": 7.0, + "learning_rate": 8.088035348214962e-07, + "loss": 0.0012, + "step": 53436 + }, + { + "epoch": 7.0, + "learning_rate": 8.085946110613907e-07, + "loss": 0.0011, + "step": 53437 + }, + { + "epoch": 7.0, + "learning_rate": 8.083857131516226e-07, + "loss": 0.001, + "step": 53438 + }, + { + "epoch": 7.0, + "learning_rate": 8.081768410927792e-07, + "loss": 0.0005, + "step": 53439 + }, + { + "epoch": 7.0, + "learning_rate": 8.079679948854502e-07, + "loss": 0.0008, + "step": 53440 + }, + { + "epoch": 7.0, + "learning_rate": 8.077591745302216e-07, + "loss": 0.0004, + "step": 53441 + }, + { + "epoch": 7.0, + "learning_rate": 8.075503800276818e-07, + "loss": 0.0006, + "step": 53442 + }, + { + "epoch": 7.0, + "learning_rate": 8.07341611378415e-07, + "loss": 0.0005, + "step": 53443 + }, + { + "epoch": 7.0, + "learning_rate": 8.071328685830116e-07, + "loss": 0.0005, + "step": 53444 + }, + { + "epoch": 7.0, + "learning_rate": 8.069241516420579e-07, + "loss": 0.0012, + "step": 53445 + }, + { + "epoch": 7.0, + "learning_rate": 8.067154605561378e-07, + "loss": 0.0009, + "step": 53446 + }, + { + "epoch": 7.0, + "learning_rate": 8.06506795325842e-07, + "loss": 0.0008, + "step": 53447 + }, + { + "epoch": 7.0, + "learning_rate": 8.062981559517569e-07, + "loss": 0.0008, + "step": 53448 + }, + { + "epoch": 7.0, + "learning_rate": 8.060895424344694e-07, + "loss": 0.0006, + "step": 53449 + }, + { + "epoch": 7.0, + "learning_rate": 8.058809547745638e-07, + "loss": 0.0002, + "step": 53450 + }, + { + "epoch": 7.0, + "learning_rate": 8.056723929726296e-07, + "loss": 0.0014, + "step": 53451 + }, + { + "epoch": 7.0, + "learning_rate": 8.054638570292527e-07, + "loss": 0.0007, + "step": 53452 + }, + { + "epoch": 7.0, + "learning_rate": 8.052553469450197e-07, + "loss": 0.0004, + "step": 53453 + }, + { + "epoch": 7.0, + "learning_rate": 8.050468627205132e-07, + "loss": 0.0004, + "step": 53454 + }, + { + "epoch": 7.0, + "learning_rate": 8.048384043563251e-07, + "loss": 0.0004, + "step": 53455 + }, + { + "epoch": 7.0, + "learning_rate": 8.046299718530392e-07, + "loss": 0.0012, + "step": 53456 + }, + { + "epoch": 7.0, + "learning_rate": 8.044215652112409e-07, + "loss": 0.0006, + "step": 53457 + }, + { + "epoch": 7.0, + "learning_rate": 8.04213184431516e-07, + "loss": 0.0006, + "step": 53458 + }, + { + "epoch": 7.0, + "learning_rate": 8.040048295144542e-07, + "loss": 0.0004, + "step": 53459 + }, + { + "epoch": 7.0, + "learning_rate": 8.037965004606396e-07, + "loss": 0.0004, + "step": 53460 + }, + { + "epoch": 7.0, + "learning_rate": 8.03588197270656e-07, + "loss": 0.0005, + "step": 53461 + }, + { + "epoch": 7.0, + "learning_rate": 8.033799199450909e-07, + "loss": 0.0004, + "step": 53462 + }, + { + "epoch": 7.0, + "learning_rate": 8.031716684845303e-07, + "loss": 0.0008, + "step": 53463 + }, + { + "epoch": 7.0, + "learning_rate": 8.029634428895605e-07, + "loss": 0.0007, + "step": 53464 + }, + { + "epoch": 7.0, + "learning_rate": 8.027552431607643e-07, + "loss": 0.0003, + "step": 53465 + }, + { + "epoch": 7.0, + "learning_rate": 8.025470692987314e-07, + "loss": 0.0005, + "step": 53466 + }, + { + "epoch": 7.0, + "learning_rate": 8.023389213040445e-07, + "loss": 0.0005, + "step": 53467 + }, + { + "epoch": 7.0, + "learning_rate": 8.021307991772887e-07, + "loss": 0.0006, + "step": 53468 + }, + { + "epoch": 7.0, + "learning_rate": 8.019227029190513e-07, + "loss": 0.0004, + "step": 53469 + }, + { + "epoch": 7.0, + "learning_rate": 8.017146325299141e-07, + "loss": 0.0004, + "step": 53470 + }, + { + "epoch": 7.0, + "learning_rate": 8.015065880104666e-07, + "loss": 0.0015, + "step": 53471 + }, + { + "epoch": 7.0, + "learning_rate": 8.01298569361293e-07, + "loss": 0.0005, + "step": 53472 + }, + { + "epoch": 7.0, + "learning_rate": 8.010905765829746e-07, + "loss": 0.0005, + "step": 53473 + }, + { + "epoch": 7.0, + "learning_rate": 8.008826096761013e-07, + "loss": 0.0007, + "step": 53474 + }, + { + "epoch": 7.0, + "learning_rate": 8.006746686412558e-07, + "loss": 0.0004, + "step": 53475 + }, + { + "epoch": 7.0, + "learning_rate": 8.004667534790222e-07, + "loss": 0.0003, + "step": 53476 + }, + { + "epoch": 7.01, + "learning_rate": 8.002588641899855e-07, + "loss": 0.0006, + "step": 53477 + }, + { + "epoch": 7.01, + "learning_rate": 8.00051000774732e-07, + "loss": 0.0005, + "step": 53478 + }, + { + "epoch": 7.01, + "learning_rate": 7.998431632338433e-07, + "loss": 0.0008, + "step": 53479 + }, + { + "epoch": 7.01, + "learning_rate": 7.996353515679067e-07, + "loss": 0.0007, + "step": 53480 + }, + { + "epoch": 7.01, + "learning_rate": 7.994275657775053e-07, + "loss": 0.0004, + "step": 53481 + }, + { + "epoch": 7.01, + "learning_rate": 7.99219805863225e-07, + "loss": 0.0004, + "step": 53482 + }, + { + "epoch": 7.01, + "learning_rate": 7.990120718256489e-07, + "loss": 0.0005, + "step": 53483 + }, + { + "epoch": 7.01, + "learning_rate": 7.988043636653608e-07, + "loss": 0.0012, + "step": 53484 + }, + { + "epoch": 7.01, + "learning_rate": 7.985966813829449e-07, + "loss": 0.0008, + "step": 53485 + }, + { + "epoch": 7.01, + "learning_rate": 7.983890249789872e-07, + "loss": 0.0004, + "step": 53486 + }, + { + "epoch": 7.01, + "learning_rate": 7.981813944540706e-07, + "loss": 0.0006, + "step": 53487 + }, + { + "epoch": 7.01, + "learning_rate": 7.979737898087769e-07, + "loss": 0.0003, + "step": 53488 + }, + { + "epoch": 7.01, + "learning_rate": 7.977662110436935e-07, + "loss": 0.0006, + "step": 53489 + }, + { + "epoch": 7.01, + "learning_rate": 7.975586581594009e-07, + "loss": 0.0004, + "step": 53490 + }, + { + "epoch": 7.01, + "learning_rate": 7.973511311564853e-07, + "loss": 0.0008, + "step": 53491 + }, + { + "epoch": 7.01, + "learning_rate": 7.971436300355306e-07, + "loss": 0.0009, + "step": 53492 + }, + { + "epoch": 7.01, + "learning_rate": 7.969361547971177e-07, + "loss": 0.0004, + "step": 53493 + }, + { + "epoch": 7.01, + "learning_rate": 7.967287054418327e-07, + "loss": 0.0003, + "step": 53494 + }, + { + "epoch": 7.01, + "learning_rate": 7.965212819702594e-07, + "loss": 0.0005, + "step": 53495 + }, + { + "epoch": 7.01, + "learning_rate": 7.963138843829765e-07, + "loss": 0.0005, + "step": 53496 + }, + { + "epoch": 7.01, + "learning_rate": 7.961065126805734e-07, + "loss": 0.0007, + "step": 53497 + }, + { + "epoch": 7.01, + "learning_rate": 7.958991668636307e-07, + "loss": 0.0007, + "step": 53498 + }, + { + "epoch": 7.01, + "learning_rate": 7.956918469327313e-07, + "loss": 0.0003, + "step": 53499 + }, + { + "epoch": 7.01, + "learning_rate": 7.954845528884569e-07, + "loss": 0.0008, + "step": 53500 + }, + { + "epoch": 7.01, + "learning_rate": 7.952772847313917e-07, + "loss": 0.0008, + "step": 53501 + }, + { + "epoch": 7.01, + "learning_rate": 7.950700424621205e-07, + "loss": 0.0004, + "step": 53502 + }, + { + "epoch": 7.01, + "learning_rate": 7.948628260812252e-07, + "loss": 0.0006, + "step": 53503 + }, + { + "epoch": 7.01, + "learning_rate": 7.946556355892865e-07, + "loss": 0.0005, + "step": 53504 + }, + { + "epoch": 7.01, + "learning_rate": 7.944484709868893e-07, + "loss": 0.0004, + "step": 53505 + }, + { + "epoch": 7.01, + "learning_rate": 7.942413322746157e-07, + "loss": 0.0006, + "step": 53506 + }, + { + "epoch": 7.01, + "learning_rate": 7.940342194530493e-07, + "loss": 0.0004, + "step": 53507 + }, + { + "epoch": 7.01, + "learning_rate": 7.938271325227687e-07, + "loss": 0.0004, + "step": 53508 + }, + { + "epoch": 7.01, + "learning_rate": 7.936200714843601e-07, + "loss": 0.0005, + "step": 53509 + }, + { + "epoch": 7.01, + "learning_rate": 7.934130363384062e-07, + "loss": 0.0005, + "step": 53510 + }, + { + "epoch": 7.01, + "learning_rate": 7.932060270854858e-07, + "loss": 0.0005, + "step": 53511 + }, + { + "epoch": 7.01, + "learning_rate": 7.929990437261825e-07, + "loss": 0.0003, + "step": 53512 + }, + { + "epoch": 7.01, + "learning_rate": 7.927920862610816e-07, + "loss": 0.0009, + "step": 53513 + }, + { + "epoch": 7.01, + "learning_rate": 7.925851546907615e-07, + "loss": 0.0004, + "step": 53514 + }, + { + "epoch": 7.01, + "learning_rate": 7.92378249015805e-07, + "loss": 0.0014, + "step": 53515 + }, + { + "epoch": 7.01, + "learning_rate": 7.92171369236795e-07, + "loss": 0.0006, + "step": 53516 + }, + { + "epoch": 7.01, + "learning_rate": 7.919645153543132e-07, + "loss": 0.0007, + "step": 53517 + }, + { + "epoch": 7.01, + "learning_rate": 7.917576873689404e-07, + "loss": 0.0005, + "step": 53518 + }, + { + "epoch": 7.01, + "learning_rate": 7.91550885281257e-07, + "loss": 0.0006, + "step": 53519 + }, + { + "epoch": 7.01, + "learning_rate": 7.913441090918495e-07, + "loss": 0.0006, + "step": 53520 + }, + { + "epoch": 7.01, + "learning_rate": 7.911373588012949e-07, + "loss": 0.0003, + "step": 53521 + }, + { + "epoch": 7.01, + "learning_rate": 7.909306344101764e-07, + "loss": 0.0003, + "step": 53522 + }, + { + "epoch": 7.01, + "learning_rate": 7.90723935919071e-07, + "loss": 0.0007, + "step": 53523 + }, + { + "epoch": 7.01, + "learning_rate": 7.905172633285685e-07, + "loss": 0.0002, + "step": 53524 + }, + { + "epoch": 7.01, + "learning_rate": 7.90310616639246e-07, + "loss": 0.0006, + "step": 53525 + }, + { + "epoch": 7.01, + "learning_rate": 7.901039958516832e-07, + "loss": 0.0007, + "step": 53526 + }, + { + "epoch": 7.01, + "learning_rate": 7.898974009664617e-07, + "loss": 0.0003, + "step": 53527 + }, + { + "epoch": 7.01, + "learning_rate": 7.896908319841645e-07, + "loss": 0.0005, + "step": 53528 + }, + { + "epoch": 7.01, + "learning_rate": 7.894842889053722e-07, + "loss": 0.0005, + "step": 53529 + }, + { + "epoch": 7.01, + "learning_rate": 7.892777717306621e-07, + "loss": 0.0013, + "step": 53530 + }, + { + "epoch": 7.01, + "learning_rate": 7.890712804606204e-07, + "loss": 0.0005, + "step": 53531 + }, + { + "epoch": 7.01, + "learning_rate": 7.888648150958244e-07, + "loss": 0.0007, + "step": 53532 + }, + { + "epoch": 7.01, + "learning_rate": 7.88658375636856e-07, + "loss": 0.0007, + "step": 53533 + }, + { + "epoch": 7.01, + "learning_rate": 7.884519620842934e-07, + "loss": 0.0004, + "step": 53534 + }, + { + "epoch": 7.01, + "learning_rate": 7.882455744387196e-07, + "loss": 0.0008, + "step": 53535 + }, + { + "epoch": 7.01, + "learning_rate": 7.880392127007153e-07, + "loss": 0.0006, + "step": 53536 + }, + { + "epoch": 7.01, + "learning_rate": 7.87832876870861e-07, + "loss": 0.0014, + "step": 53537 + }, + { + "epoch": 7.01, + "learning_rate": 7.87626566949734e-07, + "loss": 0.0005, + "step": 53538 + }, + { + "epoch": 7.01, + "learning_rate": 7.874202829379185e-07, + "loss": 0.0007, + "step": 53539 + }, + { + "epoch": 7.01, + "learning_rate": 7.872140248359939e-07, + "loss": 0.0005, + "step": 53540 + }, + { + "epoch": 7.01, + "learning_rate": 7.870077926445385e-07, + "loss": 0.0009, + "step": 53541 + }, + { + "epoch": 7.01, + "learning_rate": 7.86801586364131e-07, + "loss": 0.0006, + "step": 53542 + }, + { + "epoch": 7.01, + "learning_rate": 7.865954059953552e-07, + "loss": 0.0012, + "step": 53543 + }, + { + "epoch": 7.01, + "learning_rate": 7.863892515387894e-07, + "loss": 0.0007, + "step": 53544 + }, + { + "epoch": 7.01, + "learning_rate": 7.861831229950113e-07, + "loss": 0.0005, + "step": 53545 + }, + { + "epoch": 7.01, + "learning_rate": 7.859770203646022e-07, + "loss": 0.0003, + "step": 53546 + }, + { + "epoch": 7.01, + "learning_rate": 7.857709436481442e-07, + "loss": 0.0003, + "step": 53547 + }, + { + "epoch": 7.01, + "learning_rate": 7.855648928462145e-07, + "loss": 0.0004, + "step": 53548 + }, + { + "epoch": 7.01, + "learning_rate": 7.853588679593938e-07, + "loss": 0.001, + "step": 53549 + }, + { + "epoch": 7.01, + "learning_rate": 7.851528689882582e-07, + "loss": 0.0005, + "step": 53550 + }, + { + "epoch": 7.01, + "learning_rate": 7.849468959333906e-07, + "loss": 0.0005, + "step": 53551 + }, + { + "epoch": 7.01, + "learning_rate": 7.847409487953694e-07, + "loss": 0.0005, + "step": 53552 + }, + { + "epoch": 7.02, + "learning_rate": 7.845350275747721e-07, + "loss": 0.0005, + "step": 53553 + }, + { + "epoch": 7.02, + "learning_rate": 7.843291322721813e-07, + "loss": 0.0004, + "step": 53554 + }, + { + "epoch": 7.02, + "learning_rate": 7.841232628881734e-07, + "loss": 0.001, + "step": 53555 + }, + { + "epoch": 7.02, + "learning_rate": 7.839174194233267e-07, + "loss": 0.0003, + "step": 53556 + }, + { + "epoch": 7.02, + "learning_rate": 7.837116018782232e-07, + "loss": 0.0004, + "step": 53557 + }, + { + "epoch": 7.02, + "learning_rate": 7.835058102534377e-07, + "loss": 0.0006, + "step": 53558 + }, + { + "epoch": 7.02, + "learning_rate": 7.833000445495531e-07, + "loss": 0.0006, + "step": 53559 + }, + { + "epoch": 7.02, + "learning_rate": 7.830943047671469e-07, + "loss": 0.0003, + "step": 53560 + }, + { + "epoch": 7.02, + "learning_rate": 7.828885909067951e-07, + "loss": 0.0005, + "step": 53561 + }, + { + "epoch": 7.02, + "learning_rate": 7.826829029690808e-07, + "loss": 0.0005, + "step": 53562 + }, + { + "epoch": 7.02, + "learning_rate": 7.824772409545789e-07, + "loss": 0.0007, + "step": 53563 + }, + { + "epoch": 7.02, + "learning_rate": 7.82271604863869e-07, + "loss": 0.0007, + "step": 53564 + }, + { + "epoch": 7.02, + "learning_rate": 7.820659946975284e-07, + "loss": 0.0005, + "step": 53565 + }, + { + "epoch": 7.02, + "learning_rate": 7.818604104561378e-07, + "loss": 0.0008, + "step": 53566 + }, + { + "epoch": 7.02, + "learning_rate": 7.816548521402712e-07, + "loss": 0.0003, + "step": 53567 + }, + { + "epoch": 7.02, + "learning_rate": 7.814493197505124e-07, + "loss": 0.0004, + "step": 53568 + }, + { + "epoch": 7.02, + "learning_rate": 7.812438132874345e-07, + "loss": 0.0003, + "step": 53569 + }, + { + "epoch": 7.02, + "learning_rate": 7.810383327516191e-07, + "loss": 0.0004, + "step": 53570 + }, + { + "epoch": 7.02, + "learning_rate": 7.808328781436414e-07, + "loss": 0.0005, + "step": 53571 + }, + { + "epoch": 7.02, + "learning_rate": 7.80627449464082e-07, + "loss": 0.0006, + "step": 53572 + }, + { + "epoch": 7.02, + "learning_rate": 7.804220467135138e-07, + "loss": 0.0007, + "step": 53573 + }, + { + "epoch": 7.02, + "learning_rate": 7.802166698925195e-07, + "loss": 0.0002, + "step": 53574 + }, + { + "epoch": 7.02, + "learning_rate": 7.800113190016756e-07, + "loss": 0.0002, + "step": 53575 + }, + { + "epoch": 7.02, + "learning_rate": 7.79805994041557e-07, + "loss": 0.0004, + "step": 53576 + }, + { + "epoch": 7.02, + "learning_rate": 7.796006950127444e-07, + "loss": 0.0006, + "step": 53577 + }, + { + "epoch": 7.02, + "learning_rate": 7.793954219158117e-07, + "loss": 0.0003, + "step": 53578 + }, + { + "epoch": 7.02, + "learning_rate": 7.791901747513408e-07, + "loss": 0.0007, + "step": 53579 + }, + { + "epoch": 7.02, + "learning_rate": 7.789849535199045e-07, + "loss": 0.0004, + "step": 53580 + }, + { + "epoch": 7.02, + "learning_rate": 7.787797582220846e-07, + "loss": 0.0005, + "step": 53581 + }, + { + "epoch": 7.02, + "learning_rate": 7.785745888584539e-07, + "loss": 0.0005, + "step": 53582 + }, + { + "epoch": 7.02, + "learning_rate": 7.783694454295921e-07, + "loss": 0.0002, + "step": 53583 + }, + { + "epoch": 7.02, + "learning_rate": 7.78164327936074e-07, + "loss": 0.0003, + "step": 53584 + }, + { + "epoch": 7.02, + "learning_rate": 7.779592363784783e-07, + "loss": 0.0006, + "step": 53585 + }, + { + "epoch": 7.02, + "learning_rate": 7.77754170757381e-07, + "loss": 0.0006, + "step": 53586 + }, + { + "epoch": 7.02, + "learning_rate": 7.775491310733596e-07, + "loss": 0.0005, + "step": 53587 + }, + { + "epoch": 7.02, + "learning_rate": 7.773441173269869e-07, + "loss": 0.0008, + "step": 53588 + }, + { + "epoch": 7.02, + "learning_rate": 7.771391295188457e-07, + "loss": 0.0004, + "step": 53589 + }, + { + "epoch": 7.02, + "learning_rate": 7.769341676495101e-07, + "loss": 0.0006, + "step": 53590 + }, + { + "epoch": 7.02, + "learning_rate": 7.767292317195552e-07, + "loss": 0.0005, + "step": 53591 + }, + { + "epoch": 7.02, + "learning_rate": 7.76524321729557e-07, + "loss": 0.0005, + "step": 53592 + }, + { + "epoch": 7.02, + "learning_rate": 7.763194376800942e-07, + "loss": 0.0005, + "step": 53593 + }, + { + "epoch": 7.02, + "learning_rate": 7.761145795717428e-07, + "loss": 0.0015, + "step": 53594 + }, + { + "epoch": 7.02, + "learning_rate": 7.759097474050747e-07, + "loss": 0.0004, + "step": 53595 + }, + { + "epoch": 7.02, + "learning_rate": 7.757049411806727e-07, + "loss": 0.0003, + "step": 53596 + }, + { + "epoch": 7.02, + "learning_rate": 7.755001608991076e-07, + "loss": 0.001, + "step": 53597 + }, + { + "epoch": 7.02, + "learning_rate": 7.752954065609586e-07, + "loss": 0.0009, + "step": 53598 + }, + { + "epoch": 7.02, + "learning_rate": 7.750906781667977e-07, + "loss": 0.0005, + "step": 53599 + }, + { + "epoch": 7.02, + "learning_rate": 7.748859757172022e-07, + "loss": 0.0005, + "step": 53600 + }, + { + "epoch": 7.02, + "learning_rate": 7.746812992127517e-07, + "loss": 0.0004, + "step": 53601 + }, + { + "epoch": 7.02, + "learning_rate": 7.744766486540178e-07, + "loss": 0.0007, + "step": 53602 + }, + { + "epoch": 7.02, + "learning_rate": 7.742720240415758e-07, + "loss": 0.0006, + "step": 53603 + }, + { + "epoch": 7.02, + "learning_rate": 7.740674253760039e-07, + "loss": 0.0004, + "step": 53604 + }, + { + "epoch": 7.02, + "learning_rate": 7.738628526578762e-07, + "loss": 0.0007, + "step": 53605 + }, + { + "epoch": 7.02, + "learning_rate": 7.736583058877678e-07, + "loss": 0.0004, + "step": 53606 + }, + { + "epoch": 7.02, + "learning_rate": 7.734537850662526e-07, + "loss": 0.0009, + "step": 53607 + }, + { + "epoch": 7.02, + "learning_rate": 7.732492901939081e-07, + "loss": 0.0004, + "step": 53608 + }, + { + "epoch": 7.02, + "learning_rate": 7.730448212713093e-07, + "loss": 0.0004, + "step": 53609 + }, + { + "epoch": 7.02, + "learning_rate": 7.728403782990291e-07, + "loss": 0.0007, + "step": 53610 + }, + { + "epoch": 7.02, + "learning_rate": 7.726359612776436e-07, + "loss": 0.0003, + "step": 53611 + }, + { + "epoch": 7.02, + "learning_rate": 7.7243157020773e-07, + "loss": 0.0004, + "step": 53612 + }, + { + "epoch": 7.02, + "learning_rate": 7.722272050898616e-07, + "loss": 0.0005, + "step": 53613 + }, + { + "epoch": 7.02, + "learning_rate": 7.720228659246121e-07, + "loss": 0.0004, + "step": 53614 + }, + { + "epoch": 7.02, + "learning_rate": 7.718185527125555e-07, + "loss": 0.0008, + "step": 53615 + }, + { + "epoch": 7.02, + "learning_rate": 7.716142654542691e-07, + "loss": 0.0006, + "step": 53616 + }, + { + "epoch": 7.02, + "learning_rate": 7.71410004150327e-07, + "loss": 0.0004, + "step": 53617 + }, + { + "epoch": 7.02, + "learning_rate": 7.712057688013008e-07, + "loss": 0.0006, + "step": 53618 + }, + { + "epoch": 7.02, + "learning_rate": 7.71001559407768e-07, + "loss": 0.0005, + "step": 53619 + }, + { + "epoch": 7.02, + "learning_rate": 7.707973759703025e-07, + "loss": 0.0007, + "step": 53620 + }, + { + "epoch": 7.02, + "learning_rate": 7.70593218489476e-07, + "loss": 0.0002, + "step": 53621 + }, + { + "epoch": 7.02, + "learning_rate": 7.703890869658659e-07, + "loss": 0.0004, + "step": 53622 + }, + { + "epoch": 7.02, + "learning_rate": 7.701849814000429e-07, + "loss": 0.0007, + "step": 53623 + }, + { + "epoch": 7.02, + "learning_rate": 7.699809017925863e-07, + "loss": 0.0006, + "step": 53624 + }, + { + "epoch": 7.02, + "learning_rate": 7.697768481440659e-07, + "loss": 0.0006, + "step": 53625 + }, + { + "epoch": 7.02, + "learning_rate": 7.695728204550546e-07, + "loss": 0.0006, + "step": 53626 + }, + { + "epoch": 7.02, + "learning_rate": 7.693688187261306e-07, + "loss": 0.0007, + "step": 53627 + }, + { + "epoch": 7.02, + "learning_rate": 7.691648429578647e-07, + "loss": 0.0004, + "step": 53628 + }, + { + "epoch": 7.03, + "learning_rate": 7.689608931508319e-07, + "loss": 0.0003, + "step": 53629 + }, + { + "epoch": 7.03, + "learning_rate": 7.68756969305603e-07, + "loss": 0.001, + "step": 53630 + }, + { + "epoch": 7.03, + "learning_rate": 7.68553071422754e-07, + "loss": 0.0003, + "step": 53631 + }, + { + "epoch": 7.03, + "learning_rate": 7.68349199502858e-07, + "loss": 0.0003, + "step": 53632 + }, + { + "epoch": 7.03, + "learning_rate": 7.681453535464889e-07, + "loss": 0.0005, + "step": 53633 + }, + { + "epoch": 7.03, + "learning_rate": 7.679415335542184e-07, + "loss": 0.0005, + "step": 53634 + }, + { + "epoch": 7.03, + "learning_rate": 7.677377395266216e-07, + "loss": 0.0004, + "step": 53635 + }, + { + "epoch": 7.03, + "learning_rate": 7.675339714642705e-07, + "loss": 0.0007, + "step": 53636 + }, + { + "epoch": 7.03, + "learning_rate": 7.673302293677398e-07, + "loss": 0.0005, + "step": 53637 + }, + { + "epoch": 7.03, + "learning_rate": 7.671265132375982e-07, + "loss": 0.0005, + "step": 53638 + }, + { + "epoch": 7.03, + "learning_rate": 7.669228230744241e-07, + "loss": 0.0002, + "step": 53639 + }, + { + "epoch": 7.03, + "learning_rate": 7.66719158878787e-07, + "loss": 0.0009, + "step": 53640 + }, + { + "epoch": 7.03, + "learning_rate": 7.665155206512598e-07, + "loss": 0.0004, + "step": 53641 + }, + { + "epoch": 7.03, + "learning_rate": 7.663119083924175e-07, + "loss": 0.0009, + "step": 53642 + }, + { + "epoch": 7.03, + "learning_rate": 7.661083221028298e-07, + "loss": 0.0004, + "step": 53643 + }, + { + "epoch": 7.03, + "learning_rate": 7.659047617830718e-07, + "loss": 0.0002, + "step": 53644 + }, + { + "epoch": 7.03, + "learning_rate": 7.657012274337128e-07, + "loss": 0.0002, + "step": 53645 + }, + { + "epoch": 7.03, + "learning_rate": 7.654977190553303e-07, + "loss": 0.0003, + "step": 53646 + }, + { + "epoch": 7.03, + "learning_rate": 7.652942366484928e-07, + "loss": 0.0005, + "step": 53647 + }, + { + "epoch": 7.03, + "learning_rate": 7.65090780213773e-07, + "loss": 0.0003, + "step": 53648 + }, + { + "epoch": 7.03, + "learning_rate": 7.648873497517428e-07, + "loss": 0.0003, + "step": 53649 + }, + { + "epoch": 7.03, + "learning_rate": 7.646839452629773e-07, + "loss": 0.0006, + "step": 53650 + }, + { + "epoch": 7.03, + "learning_rate": 7.644805667480459e-07, + "loss": 0.0007, + "step": 53651 + }, + { + "epoch": 7.03, + "learning_rate": 7.642772142075205e-07, + "loss": 0.0005, + "step": 53652 + }, + { + "epoch": 7.03, + "learning_rate": 7.640738876419717e-07, + "loss": 0.001, + "step": 53653 + }, + { + "epoch": 7.03, + "learning_rate": 7.638705870519747e-07, + "loss": 0.0003, + "step": 53654 + }, + { + "epoch": 7.03, + "learning_rate": 7.636673124380999e-07, + "loss": 0.0005, + "step": 53655 + }, + { + "epoch": 7.03, + "learning_rate": 7.634640638009205e-07, + "loss": 0.001, + "step": 53656 + }, + { + "epoch": 7.03, + "learning_rate": 7.632608411410037e-07, + "loss": 0.0011, + "step": 53657 + }, + { + "epoch": 7.03, + "learning_rate": 7.630576444589255e-07, + "loss": 0.0002, + "step": 53658 + }, + { + "epoch": 7.03, + "learning_rate": 7.628544737552568e-07, + "loss": 0.0006, + "step": 53659 + }, + { + "epoch": 7.03, + "learning_rate": 7.626513290305659e-07, + "loss": 0.0009, + "step": 53660 + }, + { + "epoch": 7.03, + "learning_rate": 7.624482102854269e-07, + "loss": 0.0005, + "step": 53661 + }, + { + "epoch": 7.03, + "learning_rate": 7.622451175204115e-07, + "loss": 0.0004, + "step": 53662 + }, + { + "epoch": 7.03, + "learning_rate": 7.620420507360892e-07, + "loss": 0.0007, + "step": 53663 + }, + { + "epoch": 7.03, + "learning_rate": 7.618390099330308e-07, + "loss": 0.0006, + "step": 53664 + }, + { + "epoch": 7.03, + "learning_rate": 7.61635995111808e-07, + "loss": 0.0006, + "step": 53665 + }, + { + "epoch": 7.03, + "learning_rate": 7.614330062729935e-07, + "loss": 0.0005, + "step": 53666 + }, + { + "epoch": 7.03, + "learning_rate": 7.61230043417156e-07, + "loss": 0.0004, + "step": 53667 + }, + { + "epoch": 7.03, + "learning_rate": 7.61027106544866e-07, + "loss": 0.0003, + "step": 53668 + }, + { + "epoch": 7.03, + "learning_rate": 7.608241956566964e-07, + "loss": 0.0007, + "step": 53669 + }, + { + "epoch": 7.03, + "learning_rate": 7.606213107532179e-07, + "loss": 0.001, + "step": 53670 + }, + { + "epoch": 7.03, + "learning_rate": 7.604184518349988e-07, + "loss": 0.0006, + "step": 53671 + }, + { + "epoch": 7.03, + "learning_rate": 7.6021561890261e-07, + "loss": 0.0007, + "step": 53672 + }, + { + "epoch": 7.03, + "learning_rate": 7.60012811956623e-07, + "loss": 0.0007, + "step": 53673 + }, + { + "epoch": 7.03, + "learning_rate": 7.598100309976098e-07, + "loss": 0.0002, + "step": 53674 + }, + { + "epoch": 7.03, + "learning_rate": 7.596072760261352e-07, + "loss": 0.0005, + "step": 53675 + }, + { + "epoch": 7.03, + "learning_rate": 7.594045470427747e-07, + "loss": 0.0008, + "step": 53676 + }, + { + "epoch": 7.03, + "learning_rate": 7.592018440480975e-07, + "loss": 0.0004, + "step": 53677 + }, + { + "epoch": 7.03, + "learning_rate": 7.589991670426733e-07, + "loss": 0.0006, + "step": 53678 + }, + { + "epoch": 7.03, + "learning_rate": 7.587965160270716e-07, + "loss": 0.0005, + "step": 53679 + }, + { + "epoch": 7.03, + "learning_rate": 7.58593891001862e-07, + "loss": 0.0005, + "step": 53680 + }, + { + "epoch": 7.03, + "learning_rate": 7.58391291967615e-07, + "loss": 0.0005, + "step": 53681 + }, + { + "epoch": 7.03, + "learning_rate": 7.581887189249016e-07, + "loss": 0.0008, + "step": 53682 + }, + { + "epoch": 7.03, + "learning_rate": 7.579861718742876e-07, + "loss": 0.0005, + "step": 53683 + }, + { + "epoch": 7.03, + "learning_rate": 7.577836508163483e-07, + "loss": 0.0007, + "step": 53684 + }, + { + "epoch": 7.03, + "learning_rate": 7.575811557516488e-07, + "loss": 0.0003, + "step": 53685 + }, + { + "epoch": 7.03, + "learning_rate": 7.573786866807608e-07, + "loss": 0.0007, + "step": 53686 + }, + { + "epoch": 7.03, + "learning_rate": 7.571762436042517e-07, + "loss": 0.0005, + "step": 53687 + }, + { + "epoch": 7.03, + "learning_rate": 7.569738265226922e-07, + "loss": 0.0007, + "step": 53688 + }, + { + "epoch": 7.03, + "learning_rate": 7.567714354366528e-07, + "loss": 0.0006, + "step": 53689 + }, + { + "epoch": 7.03, + "learning_rate": 7.56569070346701e-07, + "loss": 0.0006, + "step": 53690 + }, + { + "epoch": 7.03, + "learning_rate": 7.563667312534062e-07, + "loss": 0.0007, + "step": 53691 + }, + { + "epoch": 7.03, + "learning_rate": 7.56164418157338e-07, + "loss": 0.0006, + "step": 53692 + }, + { + "epoch": 7.03, + "learning_rate": 7.559621310590659e-07, + "loss": 0.0013, + "step": 53693 + }, + { + "epoch": 7.03, + "learning_rate": 7.557598699591584e-07, + "loss": 0.0003, + "step": 53694 + }, + { + "epoch": 7.03, + "learning_rate": 7.555576348581817e-07, + "loss": 0.0006, + "step": 53695 + }, + { + "epoch": 7.03, + "learning_rate": 7.553554257567086e-07, + "loss": 0.0003, + "step": 53696 + }, + { + "epoch": 7.03, + "learning_rate": 7.551532426553054e-07, + "loss": 0.0003, + "step": 53697 + }, + { + "epoch": 7.03, + "learning_rate": 7.549510855545406e-07, + "loss": 0.0003, + "step": 53698 + }, + { + "epoch": 7.03, + "learning_rate": 7.547489544549835e-07, + "loss": 0.0003, + "step": 53699 + }, + { + "epoch": 7.03, + "learning_rate": 7.545468493572039e-07, + "loss": 0.0007, + "step": 53700 + }, + { + "epoch": 7.03, + "learning_rate": 7.54344770261769e-07, + "loss": 0.0007, + "step": 53701 + }, + { + "epoch": 7.03, + "learning_rate": 7.541427171692472e-07, + "loss": 0.0003, + "step": 53702 + }, + { + "epoch": 7.03, + "learning_rate": 7.53940690080206e-07, + "loss": 0.0005, + "step": 53703 + }, + { + "epoch": 7.03, + "learning_rate": 7.537386889952147e-07, + "loss": 0.0001, + "step": 53704 + }, + { + "epoch": 7.03, + "learning_rate": 7.535367139148408e-07, + "loss": 0.0004, + "step": 53705 + }, + { + "epoch": 7.04, + "learning_rate": 7.533347648396516e-07, + "loss": 0.0008, + "step": 53706 + }, + { + "epoch": 7.04, + "learning_rate": 7.531328417702166e-07, + "loss": 0.0005, + "step": 53707 + }, + { + "epoch": 7.04, + "learning_rate": 7.529309447071043e-07, + "loss": 0.0004, + "step": 53708 + }, + { + "epoch": 7.04, + "learning_rate": 7.527290736508786e-07, + "loss": 0.001, + "step": 53709 + }, + { + "epoch": 7.04, + "learning_rate": 7.525272286021102e-07, + "loss": 0.0009, + "step": 53710 + }, + { + "epoch": 7.04, + "learning_rate": 7.523254095613686e-07, + "loss": 0.0006, + "step": 53711 + }, + { + "epoch": 7.04, + "learning_rate": 7.52123616529219e-07, + "loss": 0.0004, + "step": 53712 + }, + { + "epoch": 7.04, + "learning_rate": 7.519218495062286e-07, + "loss": 0.0005, + "step": 53713 + }, + { + "epoch": 7.04, + "learning_rate": 7.517201084929648e-07, + "loss": 0.0007, + "step": 53714 + }, + { + "epoch": 7.04, + "learning_rate": 7.515183934899973e-07, + "loss": 0.0006, + "step": 53715 + }, + { + "epoch": 7.04, + "learning_rate": 7.513167044978919e-07, + "loss": 0.0003, + "step": 53716 + }, + { + "epoch": 7.04, + "learning_rate": 7.511150415172131e-07, + "loss": 0.0004, + "step": 53717 + }, + { + "epoch": 7.04, + "learning_rate": 7.509134045485334e-07, + "loss": 0.0002, + "step": 53718 + }, + { + "epoch": 7.04, + "learning_rate": 7.507117935924146e-07, + "loss": 0.0005, + "step": 53719 + }, + { + "epoch": 7.04, + "learning_rate": 7.505102086494287e-07, + "loss": 0.0003, + "step": 53720 + }, + { + "epoch": 7.04, + "learning_rate": 7.503086497201406e-07, + "loss": 0.0004, + "step": 53721 + }, + { + "epoch": 7.04, + "learning_rate": 7.501071168051144e-07, + "loss": 0.0008, + "step": 53722 + }, + { + "epoch": 7.04, + "learning_rate": 7.499056099049217e-07, + "loss": 0.0005, + "step": 53723 + }, + { + "epoch": 7.04, + "learning_rate": 7.497041290201268e-07, + "loss": 0.0003, + "step": 53724 + }, + { + "epoch": 7.04, + "learning_rate": 7.495026741512945e-07, + "loss": 0.0011, + "step": 53725 + }, + { + "epoch": 7.04, + "learning_rate": 7.493012452989945e-07, + "loss": 0.0004, + "step": 53726 + }, + { + "epoch": 7.04, + "learning_rate": 7.49099842463793e-07, + "loss": 0.0005, + "step": 53727 + }, + { + "epoch": 7.04, + "learning_rate": 7.48898465646255e-07, + "loss": 0.0004, + "step": 53728 + }, + { + "epoch": 7.04, + "learning_rate": 7.486971148469457e-07, + "loss": 0.0005, + "step": 53729 + }, + { + "epoch": 7.04, + "learning_rate": 7.484957900664336e-07, + "loss": 0.0004, + "step": 53730 + }, + { + "epoch": 7.04, + "learning_rate": 7.482944913052858e-07, + "loss": 0.0004, + "step": 53731 + }, + { + "epoch": 7.04, + "learning_rate": 7.480932185640666e-07, + "loss": 0.0014, + "step": 53732 + }, + { + "epoch": 7.04, + "learning_rate": 7.478919718433408e-07, + "loss": 0.0009, + "step": 53733 + }, + { + "epoch": 7.04, + "learning_rate": 7.476907511436781e-07, + "loss": 0.0005, + "step": 53734 + }, + { + "epoch": 7.04, + "learning_rate": 7.474895564656437e-07, + "loss": 0.0005, + "step": 53735 + }, + { + "epoch": 7.04, + "learning_rate": 7.472883878098003e-07, + "loss": 0.0004, + "step": 53736 + }, + { + "epoch": 7.04, + "learning_rate": 7.470872451767152e-07, + "loss": 0.0004, + "step": 53737 + }, + { + "epoch": 7.04, + "learning_rate": 7.46886128566956e-07, + "loss": 0.0005, + "step": 53738 + }, + { + "epoch": 7.04, + "learning_rate": 7.466850379810864e-07, + "loss": 0.0003, + "step": 53739 + }, + { + "epoch": 7.04, + "learning_rate": 7.464839734196705e-07, + "loss": 0.0004, + "step": 53740 + }, + { + "epoch": 7.04, + "learning_rate": 7.462829348832757e-07, + "loss": 0.0008, + "step": 53741 + }, + { + "epoch": 7.04, + "learning_rate": 7.460819223724691e-07, + "loss": 0.0004, + "step": 53742 + }, + { + "epoch": 7.04, + "learning_rate": 7.458809358878149e-07, + "loss": 0.0003, + "step": 53743 + }, + { + "epoch": 7.04, + "learning_rate": 7.456799754298772e-07, + "loss": 0.0005, + "step": 53744 + }, + { + "epoch": 7.04, + "learning_rate": 7.454790409992197e-07, + "loss": 0.0005, + "step": 53745 + }, + { + "epoch": 7.04, + "learning_rate": 7.452781325964121e-07, + "loss": 0.0009, + "step": 53746 + }, + { + "epoch": 7.04, + "learning_rate": 7.45077250222016e-07, + "loss": 0.0004, + "step": 53747 + }, + { + "epoch": 7.04, + "learning_rate": 7.448763938765968e-07, + "loss": 0.0007, + "step": 53748 + }, + { + "epoch": 7.04, + "learning_rate": 7.446755635607206e-07, + "loss": 0.0006, + "step": 53749 + }, + { + "epoch": 7.04, + "learning_rate": 7.444747592749523e-07, + "loss": 0.0009, + "step": 53750 + }, + { + "epoch": 7.04, + "learning_rate": 7.442739810198552e-07, + "loss": 0.0007, + "step": 53751 + }, + { + "epoch": 7.04, + "learning_rate": 7.440732287959928e-07, + "loss": 0.0007, + "step": 53752 + }, + { + "epoch": 7.04, + "learning_rate": 7.438725026039317e-07, + "loss": 0.0005, + "step": 53753 + }, + { + "epoch": 7.04, + "learning_rate": 7.436718024442379e-07, + "loss": 0.0008, + "step": 53754 + }, + { + "epoch": 7.04, + "learning_rate": 7.434711283174744e-07, + "loss": 0.0003, + "step": 53755 + }, + { + "epoch": 7.04, + "learning_rate": 7.432704802242041e-07, + "loss": 0.0009, + "step": 53756 + }, + { + "epoch": 7.04, + "learning_rate": 7.430698581649942e-07, + "loss": 0.0007, + "step": 53757 + }, + { + "epoch": 7.04, + "learning_rate": 7.428692621404065e-07, + "loss": 0.0005, + "step": 53758 + }, + { + "epoch": 7.04, + "learning_rate": 7.426686921510074e-07, + "loss": 0.0003, + "step": 53759 + }, + { + "epoch": 7.04, + "learning_rate": 7.424681481973573e-07, + "loss": 0.0004, + "step": 53760 + }, + { + "epoch": 7.04, + "learning_rate": 7.422676302800236e-07, + "loss": 0.0004, + "step": 53761 + }, + { + "epoch": 7.04, + "learning_rate": 7.420671383995692e-07, + "loss": 0.0004, + "step": 53762 + }, + { + "epoch": 7.04, + "learning_rate": 7.41866672556557e-07, + "loss": 0.0009, + "step": 53763 + }, + { + "epoch": 7.04, + "learning_rate": 7.416662327515511e-07, + "loss": 0.0005, + "step": 53764 + }, + { + "epoch": 7.04, + "learning_rate": 7.414658189851187e-07, + "loss": 0.0003, + "step": 53765 + }, + { + "epoch": 7.04, + "learning_rate": 7.412654312578193e-07, + "loss": 0.0002, + "step": 53766 + }, + { + "epoch": 7.04, + "learning_rate": 7.410650695702171e-07, + "loss": 0.001, + "step": 53767 + }, + { + "epoch": 7.04, + "learning_rate": 7.40864733922878e-07, + "loss": 0.0007, + "step": 53768 + }, + { + "epoch": 7.04, + "learning_rate": 7.40664424316363e-07, + "loss": 0.0014, + "step": 53769 + }, + { + "epoch": 7.04, + "learning_rate": 7.40464140751237e-07, + "loss": 0.0003, + "step": 53770 + }, + { + "epoch": 7.04, + "learning_rate": 7.402638832280606e-07, + "loss": 0.0004, + "step": 53771 + }, + { + "epoch": 7.04, + "learning_rate": 7.400636517474003e-07, + "loss": 0.0004, + "step": 53772 + }, + { + "epoch": 7.04, + "learning_rate": 7.398634463098175e-07, + "loss": 0.0005, + "step": 53773 + }, + { + "epoch": 7.04, + "learning_rate": 7.396632669158743e-07, + "loss": 0.0004, + "step": 53774 + }, + { + "epoch": 7.04, + "learning_rate": 7.394631135661356e-07, + "loss": 0.0004, + "step": 53775 + }, + { + "epoch": 7.04, + "learning_rate": 7.392629862611655e-07, + "loss": 0.0007, + "step": 53776 + }, + { + "epoch": 7.04, + "learning_rate": 7.390628850015246e-07, + "loss": 0.0006, + "step": 53777 + }, + { + "epoch": 7.04, + "learning_rate": 7.388628097877759e-07, + "loss": 0.0003, + "step": 53778 + }, + { + "epoch": 7.04, + "learning_rate": 7.38662760620481e-07, + "loss": 0.0005, + "step": 53779 + }, + { + "epoch": 7.04, + "learning_rate": 7.384627375002052e-07, + "loss": 0.0009, + "step": 53780 + }, + { + "epoch": 7.04, + "learning_rate": 7.382627404275101e-07, + "loss": 0.0003, + "step": 53781 + }, + { + "epoch": 7.05, + "learning_rate": 7.380627694029562e-07, + "loss": 0.0004, + "step": 53782 + }, + { + "epoch": 7.05, + "learning_rate": 7.378628244271092e-07, + "loss": 0.0004, + "step": 53783 + }, + { + "epoch": 7.05, + "learning_rate": 7.376629055005291e-07, + "loss": 0.0003, + "step": 53784 + }, + { + "epoch": 7.05, + "learning_rate": 7.37463012623778e-07, + "loss": 0.0006, + "step": 53785 + }, + { + "epoch": 7.05, + "learning_rate": 7.372631457974211e-07, + "loss": 0.0005, + "step": 53786 + }, + { + "epoch": 7.05, + "learning_rate": 7.370633050220155e-07, + "loss": 0.0007, + "step": 53787 + }, + { + "epoch": 7.05, + "learning_rate": 7.368634902981275e-07, + "loss": 0.0005, + "step": 53788 + }, + { + "epoch": 7.05, + "learning_rate": 7.366637016263189e-07, + "loss": 0.0004, + "step": 53789 + }, + { + "epoch": 7.05, + "learning_rate": 7.364639390071482e-07, + "loss": 0.0006, + "step": 53790 + }, + { + "epoch": 7.05, + "learning_rate": 7.362642024411804e-07, + "loss": 0.0005, + "step": 53791 + }, + { + "epoch": 7.05, + "learning_rate": 7.360644919289772e-07, + "loss": 0.0008, + "step": 53792 + }, + { + "epoch": 7.05, + "learning_rate": 7.358648074710984e-07, + "loss": 0.0008, + "step": 53793 + }, + { + "epoch": 7.05, + "learning_rate": 7.356651490681055e-07, + "loss": 0.0005, + "step": 53794 + }, + { + "epoch": 7.05, + "learning_rate": 7.354655167205616e-07, + "loss": 0.0002, + "step": 53795 + }, + { + "epoch": 7.05, + "learning_rate": 7.352659104290271e-07, + "loss": 0.0003, + "step": 53796 + }, + { + "epoch": 7.05, + "learning_rate": 7.350663301940652e-07, + "loss": 0.0006, + "step": 53797 + }, + { + "epoch": 7.05, + "learning_rate": 7.348667760162343e-07, + "loss": 0.0007, + "step": 53798 + }, + { + "epoch": 7.05, + "learning_rate": 7.346672478960981e-07, + "loss": 0.0002, + "step": 53799 + }, + { + "epoch": 7.05, + "learning_rate": 7.344677458342176e-07, + "loss": 0.0015, + "step": 53800 + }, + { + "epoch": 7.05, + "learning_rate": 7.342682698311532e-07, + "loss": 0.0011, + "step": 53801 + }, + { + "epoch": 7.05, + "learning_rate": 7.340688198874635e-07, + "loss": 0.0009, + "step": 53802 + }, + { + "epoch": 7.05, + "learning_rate": 7.338693960037147e-07, + "loss": 0.0005, + "step": 53803 + }, + { + "epoch": 7.05, + "learning_rate": 7.336699981804629e-07, + "loss": 0.001, + "step": 53804 + }, + { + "epoch": 7.05, + "learning_rate": 7.334706264182711e-07, + "loss": 0.0002, + "step": 53805 + }, + { + "epoch": 7.05, + "learning_rate": 7.332712807177001e-07, + "loss": 0.0004, + "step": 53806 + }, + { + "epoch": 7.05, + "learning_rate": 7.330719610793091e-07, + "loss": 0.0004, + "step": 53807 + }, + { + "epoch": 7.05, + "learning_rate": 7.328726675036613e-07, + "loss": 0.0024, + "step": 53808 + }, + { + "epoch": 7.05, + "learning_rate": 7.326733999913149e-07, + "loss": 0.0006, + "step": 53809 + }, + { + "epoch": 7.05, + "learning_rate": 7.324741585428297e-07, + "loss": 0.0003, + "step": 53810 + }, + { + "epoch": 7.05, + "learning_rate": 7.322749431587695e-07, + "loss": 0.0004, + "step": 53811 + }, + { + "epoch": 7.05, + "learning_rate": 7.320757538396917e-07, + "loss": 0.0003, + "step": 53812 + }, + { + "epoch": 7.05, + "learning_rate": 7.318765905861569e-07, + "loss": 0.0007, + "step": 53813 + }, + { + "epoch": 7.05, + "learning_rate": 7.316774533987259e-07, + "loss": 0.0008, + "step": 53814 + }, + { + "epoch": 7.05, + "learning_rate": 7.314783422779593e-07, + "loss": 0.0004, + "step": 53815 + }, + { + "epoch": 7.05, + "learning_rate": 7.312792572244165e-07, + "loss": 0.0009, + "step": 53816 + }, + { + "epoch": 7.05, + "learning_rate": 7.310801982386551e-07, + "loss": 0.0006, + "step": 53817 + }, + { + "epoch": 7.05, + "learning_rate": 7.308811653212366e-07, + "loss": 0.0005, + "step": 53818 + }, + { + "epoch": 7.05, + "learning_rate": 7.306821584727241e-07, + "loss": 0.0003, + "step": 53819 + }, + { + "epoch": 7.05, + "learning_rate": 7.304831776936738e-07, + "loss": 0.0004, + "step": 53820 + }, + { + "epoch": 7.05, + "learning_rate": 7.302842229846441e-07, + "loss": 0.0009, + "step": 53821 + }, + { + "epoch": 7.05, + "learning_rate": 7.300852943461977e-07, + "loss": 0.0006, + "step": 53822 + }, + { + "epoch": 7.05, + "learning_rate": 7.298863917788935e-07, + "loss": 0.0004, + "step": 53823 + }, + { + "epoch": 7.05, + "learning_rate": 7.296875152832894e-07, + "loss": 0.0003, + "step": 53824 + }, + { + "epoch": 7.05, + "learning_rate": 7.294886648599453e-07, + "loss": 0.0016, + "step": 53825 + }, + { + "epoch": 7.05, + "learning_rate": 7.292898405094207e-07, + "loss": 0.0001, + "step": 53826 + }, + { + "epoch": 7.05, + "learning_rate": 7.29091042232275e-07, + "loss": 0.0009, + "step": 53827 + }, + { + "epoch": 7.05, + "learning_rate": 7.288922700290657e-07, + "loss": 0.0014, + "step": 53828 + }, + { + "epoch": 7.05, + "learning_rate": 7.286935239003534e-07, + "loss": 0.0005, + "step": 53829 + }, + { + "epoch": 7.05, + "learning_rate": 7.284948038466988e-07, + "loss": 0.0007, + "step": 53830 + }, + { + "epoch": 7.05, + "learning_rate": 7.282961098686581e-07, + "loss": 0.0004, + "step": 53831 + }, + { + "epoch": 7.05, + "learning_rate": 7.280974419667896e-07, + "loss": 0.0017, + "step": 53832 + }, + { + "epoch": 7.05, + "learning_rate": 7.278988001416542e-07, + "loss": 0.0004, + "step": 53833 + }, + { + "epoch": 7.05, + "learning_rate": 7.277001843938103e-07, + "loss": 0.0006, + "step": 53834 + }, + { + "epoch": 7.05, + "learning_rate": 7.275015947238162e-07, + "loss": 0.0002, + "step": 53835 + }, + { + "epoch": 7.05, + "learning_rate": 7.273030311322271e-07, + "loss": 0.001, + "step": 53836 + }, + { + "epoch": 7.05, + "learning_rate": 7.27104493619607e-07, + "loss": 0.0003, + "step": 53837 + }, + { + "epoch": 7.05, + "learning_rate": 7.269059821865121e-07, + "loss": 0.0006, + "step": 53838 + }, + { + "epoch": 7.05, + "learning_rate": 7.267074968334997e-07, + "loss": 0.0008, + "step": 53839 + }, + { + "epoch": 7.05, + "learning_rate": 7.265090375611261e-07, + "loss": 0.0005, + "step": 53840 + }, + { + "epoch": 7.05, + "learning_rate": 7.263106043699542e-07, + "loss": 0.0008, + "step": 53841 + }, + { + "epoch": 7.05, + "learning_rate": 7.2611219726054e-07, + "loss": 0.0003, + "step": 53842 + }, + { + "epoch": 7.05, + "learning_rate": 7.259138162334422e-07, + "loss": 0.0005, + "step": 53843 + }, + { + "epoch": 7.05, + "learning_rate": 7.257154612892159e-07, + "loss": 0.0002, + "step": 53844 + }, + { + "epoch": 7.05, + "learning_rate": 7.255171324284227e-07, + "loss": 0.0006, + "step": 53845 + }, + { + "epoch": 7.05, + "learning_rate": 7.253188296516179e-07, + "loss": 0.0004, + "step": 53846 + }, + { + "epoch": 7.05, + "learning_rate": 7.251205529593586e-07, + "loss": 0.0003, + "step": 53847 + }, + { + "epoch": 7.05, + "learning_rate": 7.249223023522067e-07, + "loss": 0.0004, + "step": 53848 + }, + { + "epoch": 7.05, + "learning_rate": 7.247240778307152e-07, + "loss": 0.0004, + "step": 53849 + }, + { + "epoch": 7.05, + "learning_rate": 7.245258793954424e-07, + "loss": 0.0005, + "step": 53850 + }, + { + "epoch": 7.05, + "learning_rate": 7.243277070469489e-07, + "loss": 0.0004, + "step": 53851 + }, + { + "epoch": 7.05, + "learning_rate": 7.241295607857879e-07, + "loss": 0.0003, + "step": 53852 + }, + { + "epoch": 7.05, + "learning_rate": 7.239314406125198e-07, + "loss": 0.0007, + "step": 53853 + }, + { + "epoch": 7.05, + "learning_rate": 7.237333465276996e-07, + "loss": 0.0004, + "step": 53854 + }, + { + "epoch": 7.05, + "learning_rate": 7.23535278531885e-07, + "loss": 0.0005, + "step": 53855 + }, + { + "epoch": 7.05, + "learning_rate": 7.233372366256353e-07, + "loss": 0.0008, + "step": 53856 + }, + { + "epoch": 7.05, + "learning_rate": 7.231392208095045e-07, + "loss": 0.0003, + "step": 53857 + }, + { + "epoch": 7.06, + "learning_rate": 7.229412310840511e-07, + "loss": 0.0004, + "step": 53858 + }, + { + "epoch": 7.06, + "learning_rate": 7.227432674498302e-07, + "loss": 0.0002, + "step": 53859 + }, + { + "epoch": 7.06, + "learning_rate": 7.225453299074015e-07, + "loss": 0.0006, + "step": 53860 + }, + { + "epoch": 7.06, + "learning_rate": 7.223474184573175e-07, + "loss": 0.0004, + "step": 53861 + }, + { + "epoch": 7.06, + "learning_rate": 7.221495331001393e-07, + "loss": 0.0008, + "step": 53862 + }, + { + "epoch": 7.06, + "learning_rate": 7.219516738364207e-07, + "loss": 0.0005, + "step": 53863 + }, + { + "epoch": 7.06, + "learning_rate": 7.217538406667191e-07, + "loss": 0.0017, + "step": 53864 + }, + { + "epoch": 7.06, + "learning_rate": 7.215560335915917e-07, + "loss": 0.0004, + "step": 53865 + }, + { + "epoch": 7.06, + "learning_rate": 7.213582526115936e-07, + "loss": 0.0004, + "step": 53866 + }, + { + "epoch": 7.06, + "learning_rate": 7.211604977272801e-07, + "loss": 0.0004, + "step": 53867 + }, + { + "epoch": 7.06, + "learning_rate": 7.209627689392096e-07, + "loss": 0.0009, + "step": 53868 + }, + { + "epoch": 7.06, + "learning_rate": 7.207650662479371e-07, + "loss": 0.0006, + "step": 53869 + }, + { + "epoch": 7.06, + "learning_rate": 7.205673896540178e-07, + "loss": 0.0003, + "step": 53870 + }, + { + "epoch": 7.06, + "learning_rate": 7.2036973915801e-07, + "loss": 0.0008, + "step": 53871 + }, + { + "epoch": 7.06, + "learning_rate": 7.201721147604657e-07, + "loss": 0.0004, + "step": 53872 + }, + { + "epoch": 7.06, + "learning_rate": 7.199745164619454e-07, + "loss": 0.0007, + "step": 53873 + }, + { + "epoch": 7.06, + "learning_rate": 7.197769442630032e-07, + "loss": 0.0006, + "step": 53874 + }, + { + "epoch": 7.06, + "learning_rate": 7.195793981641919e-07, + "loss": 0.0003, + "step": 53875 + }, + { + "epoch": 7.06, + "learning_rate": 7.1938187816607e-07, + "loss": 0.0003, + "step": 53876 + }, + { + "epoch": 7.06, + "learning_rate": 7.191843842691937e-07, + "loss": 0.0005, + "step": 53877 + }, + { + "epoch": 7.06, + "learning_rate": 7.189869164741148e-07, + "loss": 0.0004, + "step": 53878 + }, + { + "epoch": 7.06, + "learning_rate": 7.187894747813928e-07, + "loss": 0.0007, + "step": 53879 + }, + { + "epoch": 7.06, + "learning_rate": 7.185920591915818e-07, + "loss": 0.0002, + "step": 53880 + }, + { + "epoch": 7.06, + "learning_rate": 7.183946697052357e-07, + "loss": 0.0005, + "step": 53881 + }, + { + "epoch": 7.06, + "learning_rate": 7.181973063229086e-07, + "loss": 0.0008, + "step": 53882 + }, + { + "epoch": 7.06, + "learning_rate": 7.179999690451566e-07, + "loss": 0.0002, + "step": 53883 + }, + { + "epoch": 7.06, + "learning_rate": 7.178026578725383e-07, + "loss": 0.0003, + "step": 53884 + }, + { + "epoch": 7.06, + "learning_rate": 7.176053728056053e-07, + "loss": 0.0004, + "step": 53885 + }, + { + "epoch": 7.06, + "learning_rate": 7.174081138449107e-07, + "loss": 0.0008, + "step": 53886 + }, + { + "epoch": 7.06, + "learning_rate": 7.172108809910139e-07, + "loss": 0.0007, + "step": 53887 + }, + { + "epoch": 7.06, + "learning_rate": 7.170136742444667e-07, + "loss": 0.0004, + "step": 53888 + }, + { + "epoch": 7.06, + "learning_rate": 7.168164936058242e-07, + "loss": 0.0004, + "step": 53889 + }, + { + "epoch": 7.06, + "learning_rate": 7.166193390756393e-07, + "loss": 0.0005, + "step": 53890 + }, + { + "epoch": 7.06, + "learning_rate": 7.164222106544705e-07, + "loss": 0.0003, + "step": 53891 + }, + { + "epoch": 7.06, + "learning_rate": 7.162251083428696e-07, + "loss": 0.0006, + "step": 53892 + }, + { + "epoch": 7.06, + "learning_rate": 7.160280321413882e-07, + "loss": 0.0003, + "step": 53893 + }, + { + "epoch": 7.06, + "learning_rate": 7.158309820505848e-07, + "loss": 0.0002, + "step": 53894 + }, + { + "epoch": 7.06, + "learning_rate": 7.156339580710137e-07, + "loss": 0.0004, + "step": 53895 + }, + { + "epoch": 7.06, + "learning_rate": 7.154369602032285e-07, + "loss": 0.0006, + "step": 53896 + }, + { + "epoch": 7.06, + "learning_rate": 7.152399884477801e-07, + "loss": 0.0005, + "step": 53897 + }, + { + "epoch": 7.06, + "learning_rate": 7.150430428052257e-07, + "loss": 0.0005, + "step": 53898 + }, + { + "epoch": 7.06, + "learning_rate": 7.148461232761194e-07, + "loss": 0.0008, + "step": 53899 + }, + { + "epoch": 7.06, + "learning_rate": 7.146492298610131e-07, + "loss": 0.0003, + "step": 53900 + }, + { + "epoch": 7.06, + "learning_rate": 7.144523625604605e-07, + "loss": 0.0005, + "step": 53901 + }, + { + "epoch": 7.06, + "learning_rate": 7.142555213750168e-07, + "loss": 0.0002, + "step": 53902 + }, + { + "epoch": 7.06, + "learning_rate": 7.140587063052351e-07, + "loss": 0.0006, + "step": 53903 + }, + { + "epoch": 7.06, + "learning_rate": 7.138619173516692e-07, + "loss": 0.0008, + "step": 53904 + }, + { + "epoch": 7.06, + "learning_rate": 7.136651545148699e-07, + "loss": 0.0006, + "step": 53905 + }, + { + "epoch": 7.06, + "learning_rate": 7.134684177953943e-07, + "loss": 0.0002, + "step": 53906 + }, + { + "epoch": 7.06, + "learning_rate": 7.132717071937956e-07, + "loss": 0.0003, + "step": 53907 + }, + { + "epoch": 7.06, + "learning_rate": 7.130750227106242e-07, + "loss": 0.0004, + "step": 53908 + }, + { + "epoch": 7.06, + "learning_rate": 7.128783643464343e-07, + "loss": 0.0008, + "step": 53909 + }, + { + "epoch": 7.06, + "learning_rate": 7.126817321017809e-07, + "loss": 0.0006, + "step": 53910 + }, + { + "epoch": 7.06, + "learning_rate": 7.124851259772159e-07, + "loss": 0.0005, + "step": 53911 + }, + { + "epoch": 7.06, + "learning_rate": 7.122885459732898e-07, + "loss": 0.0003, + "step": 53912 + }, + { + "epoch": 7.06, + "learning_rate": 7.120919920905589e-07, + "loss": 0.0008, + "step": 53913 + }, + { + "epoch": 7.06, + "learning_rate": 7.11895464329575e-07, + "loss": 0.0003, + "step": 53914 + }, + { + "epoch": 7.06, + "learning_rate": 7.116989626908899e-07, + "loss": 0.0004, + "step": 53915 + }, + { + "epoch": 7.06, + "learning_rate": 7.115024871750554e-07, + "loss": 0.0006, + "step": 53916 + }, + { + "epoch": 7.06, + "learning_rate": 7.113060377826264e-07, + "loss": 0.0012, + "step": 53917 + }, + { + "epoch": 7.06, + "learning_rate": 7.111096145141549e-07, + "loss": 0.001, + "step": 53918 + }, + { + "epoch": 7.06, + "learning_rate": 7.109132173701938e-07, + "loss": 0.0003, + "step": 53919 + }, + { + "epoch": 7.06, + "learning_rate": 7.107168463512926e-07, + "loss": 0.0003, + "step": 53920 + }, + { + "epoch": 7.06, + "learning_rate": 7.105205014580074e-07, + "loss": 0.0012, + "step": 53921 + }, + { + "epoch": 7.06, + "learning_rate": 7.103241826908891e-07, + "loss": 0.0004, + "step": 53922 + }, + { + "epoch": 7.06, + "learning_rate": 7.101278900504893e-07, + "loss": 0.0006, + "step": 53923 + }, + { + "epoch": 7.06, + "learning_rate": 7.099316235373577e-07, + "loss": 0.0005, + "step": 53924 + }, + { + "epoch": 7.06, + "learning_rate": 7.097353831520503e-07, + "loss": 0.0003, + "step": 53925 + }, + { + "epoch": 7.06, + "learning_rate": 7.095391688951181e-07, + "loss": 0.0006, + "step": 53926 + }, + { + "epoch": 7.06, + "learning_rate": 7.093429807671104e-07, + "loss": 0.0008, + "step": 53927 + }, + { + "epoch": 7.06, + "learning_rate": 7.091468187685801e-07, + "loss": 0.0006, + "step": 53928 + }, + { + "epoch": 7.06, + "learning_rate": 7.089506829000814e-07, + "loss": 0.0005, + "step": 53929 + }, + { + "epoch": 7.06, + "learning_rate": 7.087545731621647e-07, + "loss": 0.0005, + "step": 53930 + }, + { + "epoch": 7.06, + "learning_rate": 7.085584895553809e-07, + "loss": 0.0008, + "step": 53931 + }, + { + "epoch": 7.06, + "learning_rate": 7.083624320802795e-07, + "loss": 0.0005, + "step": 53932 + }, + { + "epoch": 7.06, + "learning_rate": 7.081664007374168e-07, + "loss": 0.0005, + "step": 53933 + }, + { + "epoch": 7.06, + "learning_rate": 7.079703955273399e-07, + "loss": 0.0005, + "step": 53934 + }, + { + "epoch": 7.07, + "learning_rate": 7.077744164506006e-07, + "loss": 0.0005, + "step": 53935 + }, + { + "epoch": 7.07, + "learning_rate": 7.07578463507752e-07, + "loss": 0.0011, + "step": 53936 + }, + { + "epoch": 7.07, + "learning_rate": 7.073825366993436e-07, + "loss": 0.0006, + "step": 53937 + }, + { + "epoch": 7.07, + "learning_rate": 7.071866360259261e-07, + "loss": 0.0004, + "step": 53938 + }, + { + "epoch": 7.07, + "learning_rate": 7.069907614880533e-07, + "loss": 0.0006, + "step": 53939 + }, + { + "epoch": 7.07, + "learning_rate": 7.067949130862717e-07, + "loss": 0.0004, + "step": 53940 + }, + { + "epoch": 7.07, + "learning_rate": 7.065990908211362e-07, + "loss": 0.0008, + "step": 53941 + }, + { + "epoch": 7.07, + "learning_rate": 7.064032946931964e-07, + "loss": 0.0007, + "step": 53942 + }, + { + "epoch": 7.07, + "learning_rate": 7.062075247030009e-07, + "loss": 0.0001, + "step": 53943 + }, + { + "epoch": 7.07, + "learning_rate": 7.060117808511035e-07, + "loss": 0.0007, + "step": 53944 + }, + { + "epoch": 7.07, + "learning_rate": 7.058160631380518e-07, + "loss": 0.0004, + "step": 53945 + }, + { + "epoch": 7.07, + "learning_rate": 7.056203715643983e-07, + "loss": 0.0006, + "step": 53946 + }, + { + "epoch": 7.07, + "learning_rate": 7.054247061306918e-07, + "loss": 0.0007, + "step": 53947 + }, + { + "epoch": 7.07, + "learning_rate": 7.052290668374839e-07, + "loss": 0.0005, + "step": 53948 + }, + { + "epoch": 7.07, + "learning_rate": 7.05033453685322e-07, + "loss": 0.0005, + "step": 53949 + }, + { + "epoch": 7.07, + "learning_rate": 7.048378666747613e-07, + "loss": 0.0004, + "step": 53950 + }, + { + "epoch": 7.07, + "learning_rate": 7.046423058063479e-07, + "loss": 0.0003, + "step": 53951 + }, + { + "epoch": 7.07, + "learning_rate": 7.044467710806335e-07, + "loss": 0.0005, + "step": 53952 + }, + { + "epoch": 7.07, + "learning_rate": 7.042512624981679e-07, + "loss": 0.0006, + "step": 53953 + }, + { + "epoch": 7.07, + "learning_rate": 7.040557800595005e-07, + "loss": 0.0006, + "step": 53954 + }, + { + "epoch": 7.07, + "learning_rate": 7.038603237651797e-07, + "loss": 0.0003, + "step": 53955 + }, + { + "epoch": 7.07, + "learning_rate": 7.036648936157586e-07, + "loss": 0.0005, + "step": 53956 + }, + { + "epoch": 7.07, + "learning_rate": 7.034694896117844e-07, + "loss": 0.0006, + "step": 53957 + }, + { + "epoch": 7.07, + "learning_rate": 7.032741117538055e-07, + "loss": 0.0006, + "step": 53958 + }, + { + "epoch": 7.07, + "learning_rate": 7.03078760042375e-07, + "loss": 0.0009, + "step": 53959 + }, + { + "epoch": 7.07, + "learning_rate": 7.028834344780388e-07, + "loss": 0.0004, + "step": 53960 + }, + { + "epoch": 7.07, + "learning_rate": 7.026881350613502e-07, + "loss": 0.0006, + "step": 53961 + }, + { + "epoch": 7.07, + "learning_rate": 7.024928617928528e-07, + "loss": 0.0004, + "step": 53962 + }, + { + "epoch": 7.07, + "learning_rate": 7.02297614673102e-07, + "loss": 0.0003, + "step": 53963 + }, + { + "epoch": 7.07, + "learning_rate": 7.021023937026439e-07, + "loss": 0.0011, + "step": 53964 + }, + { + "epoch": 7.07, + "learning_rate": 7.019071988820269e-07, + "loss": 0.0004, + "step": 53965 + }, + { + "epoch": 7.07, + "learning_rate": 7.017120302117996e-07, + "loss": 0.0011, + "step": 53966 + }, + { + "epoch": 7.07, + "learning_rate": 7.015168876925138e-07, + "loss": 0.0005, + "step": 53967 + }, + { + "epoch": 7.07, + "learning_rate": 7.013217713247167e-07, + "loss": 0.0013, + "step": 53968 + }, + { + "epoch": 7.07, + "learning_rate": 7.011266811089568e-07, + "loss": 0.0005, + "step": 53969 + }, + { + "epoch": 7.07, + "learning_rate": 7.009316170457813e-07, + "loss": 0.0009, + "step": 53970 + }, + { + "epoch": 7.07, + "learning_rate": 7.007365791357412e-07, + "loss": 0.0004, + "step": 53971 + }, + { + "epoch": 7.07, + "learning_rate": 7.005415673793847e-07, + "loss": 0.0007, + "step": 53972 + }, + { + "epoch": 7.07, + "learning_rate": 7.003465817772604e-07, + "loss": 0.0004, + "step": 53973 + }, + { + "epoch": 7.07, + "learning_rate": 7.001516223299143e-07, + "loss": 0.0006, + "step": 53974 + }, + { + "epoch": 7.07, + "learning_rate": 6.999566890378973e-07, + "loss": 0.0005, + "step": 53975 + }, + { + "epoch": 7.07, + "learning_rate": 6.997617819017578e-07, + "loss": 0.0004, + "step": 53976 + }, + { + "epoch": 7.07, + "learning_rate": 6.995669009220418e-07, + "loss": 0.0009, + "step": 53977 + }, + { + "epoch": 7.07, + "learning_rate": 6.993720460992992e-07, + "loss": 0.0003, + "step": 53978 + }, + { + "epoch": 7.07, + "learning_rate": 6.991772174340783e-07, + "loss": 0.0005, + "step": 53979 + }, + { + "epoch": 7.07, + "learning_rate": 6.989824149269254e-07, + "loss": 0.0005, + "step": 53980 + }, + { + "epoch": 7.07, + "learning_rate": 6.987876385783888e-07, + "loss": 0.0007, + "step": 53981 + }, + { + "epoch": 7.07, + "learning_rate": 6.985928883890147e-07, + "loss": 0.0006, + "step": 53982 + }, + { + "epoch": 7.07, + "learning_rate": 6.983981643593562e-07, + "loss": 0.0011, + "step": 53983 + }, + { + "epoch": 7.07, + "learning_rate": 6.982034664899562e-07, + "loss": 0.0004, + "step": 53984 + }, + { + "epoch": 7.07, + "learning_rate": 6.98008794781363e-07, + "loss": 0.0004, + "step": 53985 + }, + { + "epoch": 7.07, + "learning_rate": 6.978141492341261e-07, + "loss": 0.0009, + "step": 53986 + }, + { + "epoch": 7.07, + "learning_rate": 6.976195298487909e-07, + "loss": 0.0002, + "step": 53987 + }, + { + "epoch": 7.07, + "learning_rate": 6.974249366259056e-07, + "loss": 0.0001, + "step": 53988 + }, + { + "epoch": 7.07, + "learning_rate": 6.972303695660166e-07, + "loss": 0.0004, + "step": 53989 + }, + { + "epoch": 7.07, + "learning_rate": 6.970358286696721e-07, + "loss": 0.0006, + "step": 53990 + }, + { + "epoch": 7.07, + "learning_rate": 6.968413139374197e-07, + "loss": 0.0001, + "step": 53991 + }, + { + "epoch": 7.07, + "learning_rate": 6.966468253698045e-07, + "loss": 0.001, + "step": 53992 + }, + { + "epoch": 7.07, + "learning_rate": 6.964523629673736e-07, + "loss": 0.0009, + "step": 53993 + }, + { + "epoch": 7.07, + "learning_rate": 6.962579267306768e-07, + "loss": 0.0003, + "step": 53994 + }, + { + "epoch": 7.07, + "learning_rate": 6.960635166602592e-07, + "loss": 0.0004, + "step": 53995 + }, + { + "epoch": 7.07, + "learning_rate": 6.95869132756668e-07, + "loss": 0.0005, + "step": 53996 + }, + { + "epoch": 7.07, + "learning_rate": 6.956747750204473e-07, + "loss": 0.0003, + "step": 53997 + }, + { + "epoch": 7.07, + "learning_rate": 6.954804434521478e-07, + "loss": 0.0005, + "step": 53998 + }, + { + "epoch": 7.07, + "learning_rate": 6.952861380523134e-07, + "loss": 0.0004, + "step": 53999 + }, + { + "epoch": 7.07, + "learning_rate": 6.950918588214906e-07, + "loss": 0.0004, + "step": 54000 + }, + { + "epoch": 7.07, + "learning_rate": 6.948976057602274e-07, + "loss": 0.0003, + "step": 54001 + }, + { + "epoch": 7.07, + "learning_rate": 6.947033788690693e-07, + "loss": 0.001, + "step": 54002 + }, + { + "epoch": 7.07, + "learning_rate": 6.945091781485602e-07, + "loss": 0.0005, + "step": 54003 + }, + { + "epoch": 7.07, + "learning_rate": 6.943150035992508e-07, + "loss": 0.0002, + "step": 54004 + }, + { + "epoch": 7.07, + "learning_rate": 6.941208552216838e-07, + "loss": 0.0005, + "step": 54005 + }, + { + "epoch": 7.07, + "learning_rate": 6.939267330164067e-07, + "loss": 0.0007, + "step": 54006 + }, + { + "epoch": 7.07, + "learning_rate": 6.937326369839659e-07, + "loss": 0.0005, + "step": 54007 + }, + { + "epoch": 7.07, + "learning_rate": 6.935385671249051e-07, + "loss": 0.0007, + "step": 54008 + }, + { + "epoch": 7.07, + "learning_rate": 6.933445234397718e-07, + "loss": 0.0008, + "step": 54009 + }, + { + "epoch": 7.07, + "learning_rate": 6.931505059291121e-07, + "loss": 0.0006, + "step": 54010 + }, + { + "epoch": 7.08, + "learning_rate": 6.929565145934714e-07, + "loss": 0.0004, + "step": 54011 + }, + { + "epoch": 7.08, + "learning_rate": 6.927625494333934e-07, + "loss": 0.0003, + "step": 54012 + }, + { + "epoch": 7.08, + "learning_rate": 6.925686104494256e-07, + "loss": 0.0006, + "step": 54013 + }, + { + "epoch": 7.08, + "learning_rate": 6.92374697642112e-07, + "loss": 0.0003, + "step": 54014 + }, + { + "epoch": 7.08, + "learning_rate": 6.921808110120009e-07, + "loss": 0.0006, + "step": 54015 + }, + { + "epoch": 7.08, + "learning_rate": 6.919869505596333e-07, + "loss": 0.0004, + "step": 54016 + }, + { + "epoch": 7.08, + "learning_rate": 6.917931162855596e-07, + "loss": 0.0005, + "step": 54017 + }, + { + "epoch": 7.08, + "learning_rate": 6.915993081903205e-07, + "loss": 0.0004, + "step": 54018 + }, + { + "epoch": 7.08, + "learning_rate": 6.914055262744634e-07, + "loss": 0.0005, + "step": 54019 + }, + { + "epoch": 7.08, + "learning_rate": 6.912117705385302e-07, + "loss": 0.0004, + "step": 54020 + }, + { + "epoch": 7.08, + "learning_rate": 6.910180409830713e-07, + "loss": 0.0004, + "step": 54021 + }, + { + "epoch": 7.08, + "learning_rate": 6.908243376086265e-07, + "loss": 0.0007, + "step": 54022 + }, + { + "epoch": 7.08, + "learning_rate": 6.90630660415742e-07, + "loss": 0.0005, + "step": 54023 + }, + { + "epoch": 7.08, + "learning_rate": 6.904370094049651e-07, + "loss": 0.0002, + "step": 54024 + }, + { + "epoch": 7.08, + "learning_rate": 6.902433845768363e-07, + "loss": 0.0006, + "step": 54025 + }, + { + "epoch": 7.08, + "learning_rate": 6.900497859319033e-07, + "loss": 0.0002, + "step": 54026 + }, + { + "epoch": 7.08, + "learning_rate": 6.898562134707087e-07, + "loss": 0.0008, + "step": 54027 + }, + { + "epoch": 7.08, + "learning_rate": 6.89662667193799e-07, + "loss": 0.0005, + "step": 54028 + }, + { + "epoch": 7.08, + "learning_rate": 6.894691471017167e-07, + "loss": 0.0007, + "step": 54029 + }, + { + "epoch": 7.08, + "learning_rate": 6.892756531950073e-07, + "loss": 0.0032, + "step": 54030 + }, + { + "epoch": 7.08, + "learning_rate": 6.890821854742124e-07, + "loss": 0.0002, + "step": 54031 + }, + { + "epoch": 7.08, + "learning_rate": 6.888887439398795e-07, + "loss": 0.001, + "step": 54032 + }, + { + "epoch": 7.08, + "learning_rate": 6.886953285925513e-07, + "loss": 0.0007, + "step": 54033 + }, + { + "epoch": 7.08, + "learning_rate": 6.885019394327697e-07, + "loss": 0.0003, + "step": 54034 + }, + { + "epoch": 7.08, + "learning_rate": 6.88308576461083e-07, + "loss": 0.0007, + "step": 54035 + }, + { + "epoch": 7.08, + "learning_rate": 6.8811523967803e-07, + "loss": 0.0008, + "step": 54036 + }, + { + "epoch": 7.08, + "learning_rate": 6.879219290841588e-07, + "loss": 0.0008, + "step": 54037 + }, + { + "epoch": 7.08, + "learning_rate": 6.877286446800113e-07, + "loss": 0.0002, + "step": 54038 + }, + { + "epoch": 7.08, + "learning_rate": 6.875353864661305e-07, + "loss": 0.0004, + "step": 54039 + }, + { + "epoch": 7.08, + "learning_rate": 6.873421544430603e-07, + "loss": 0.0007, + "step": 54040 + }, + { + "epoch": 7.08, + "learning_rate": 6.871489486113459e-07, + "loss": 0.0002, + "step": 54041 + }, + { + "epoch": 7.08, + "learning_rate": 6.869557689715267e-07, + "loss": 0.0011, + "step": 54042 + }, + { + "epoch": 7.08, + "learning_rate": 6.867626155241503e-07, + "loss": 0.0003, + "step": 54043 + }, + { + "epoch": 7.08, + "learning_rate": 6.865694882697583e-07, + "loss": 0.0005, + "step": 54044 + }, + { + "epoch": 7.08, + "learning_rate": 6.863763872088936e-07, + "loss": 0.0003, + "step": 54045 + }, + { + "epoch": 7.08, + "learning_rate": 6.86183312342098e-07, + "loss": 0.0006, + "step": 54046 + }, + { + "epoch": 7.08, + "learning_rate": 6.859902636699145e-07, + "loss": 0.0005, + "step": 54047 + }, + { + "epoch": 7.08, + "learning_rate": 6.857972411928904e-07, + "loss": 0.0011, + "step": 54048 + }, + { + "epoch": 7.08, + "learning_rate": 6.856042449115652e-07, + "loss": 0.0004, + "step": 54049 + }, + { + "epoch": 7.08, + "learning_rate": 6.854112748264807e-07, + "loss": 0.0003, + "step": 54050 + }, + { + "epoch": 7.08, + "learning_rate": 6.852183309381821e-07, + "loss": 0.0005, + "step": 54051 + }, + { + "epoch": 7.08, + "learning_rate": 6.8502541324721e-07, + "loss": 0.0006, + "step": 54052 + }, + { + "epoch": 7.08, + "learning_rate": 6.848325217541096e-07, + "loss": 0.0004, + "step": 54053 + }, + { + "epoch": 7.08, + "learning_rate": 6.846396564594193e-07, + "loss": 0.001, + "step": 54054 + }, + { + "epoch": 7.08, + "learning_rate": 6.844468173636853e-07, + "loss": 0.0005, + "step": 54055 + }, + { + "epoch": 7.08, + "learning_rate": 6.842540044674483e-07, + "loss": 0.0003, + "step": 54056 + }, + { + "epoch": 7.08, + "learning_rate": 6.840612177712503e-07, + "loss": 0.0003, + "step": 54057 + }, + { + "epoch": 7.08, + "learning_rate": 6.838684572756338e-07, + "loss": 0.0006, + "step": 54058 + }, + { + "epoch": 7.08, + "learning_rate": 6.836757229811419e-07, + "loss": 0.0007, + "step": 54059 + }, + { + "epoch": 7.08, + "learning_rate": 6.834830148883165e-07, + "loss": 0.0011, + "step": 54060 + }, + { + "epoch": 7.08, + "learning_rate": 6.83290332997698e-07, + "loss": 0.0005, + "step": 54061 + }, + { + "epoch": 7.08, + "learning_rate": 6.830976773098296e-07, + "loss": 0.0009, + "step": 54062 + }, + { + "epoch": 7.08, + "learning_rate": 6.82905047825253e-07, + "loss": 0.0007, + "step": 54063 + }, + { + "epoch": 7.08, + "learning_rate": 6.827124445445099e-07, + "loss": 0.0006, + "step": 54064 + }, + { + "epoch": 7.08, + "learning_rate": 6.82519867468141e-07, + "loss": 0.0004, + "step": 54065 + }, + { + "epoch": 7.08, + "learning_rate": 6.823273165966893e-07, + "loss": 0.0004, + "step": 54066 + }, + { + "epoch": 7.08, + "learning_rate": 6.821347919306964e-07, + "loss": 0.0006, + "step": 54067 + }, + { + "epoch": 7.08, + "learning_rate": 6.819422934707032e-07, + "loss": 0.0006, + "step": 54068 + }, + { + "epoch": 7.08, + "learning_rate": 6.817498212172501e-07, + "loss": 0.0004, + "step": 54069 + }, + { + "epoch": 7.08, + "learning_rate": 6.815573751708781e-07, + "loss": 0.0006, + "step": 54070 + }, + { + "epoch": 7.08, + "learning_rate": 6.813649553321322e-07, + "loss": 0.0009, + "step": 54071 + }, + { + "epoch": 7.08, + "learning_rate": 6.811725617015519e-07, + "loss": 0.0004, + "step": 54072 + }, + { + "epoch": 7.08, + "learning_rate": 6.809801942796746e-07, + "loss": 0.0005, + "step": 54073 + }, + { + "epoch": 7.08, + "learning_rate": 6.807878530670465e-07, + "loss": 0.001, + "step": 54074 + }, + { + "epoch": 7.08, + "learning_rate": 6.80595538064206e-07, + "loss": 0.0005, + "step": 54075 + }, + { + "epoch": 7.08, + "learning_rate": 6.804032492716939e-07, + "loss": 0.0006, + "step": 54076 + }, + { + "epoch": 7.08, + "learning_rate": 6.802109866900497e-07, + "loss": 0.0009, + "step": 54077 + }, + { + "epoch": 7.08, + "learning_rate": 6.800187503198186e-07, + "loss": 0.0008, + "step": 54078 + }, + { + "epoch": 7.08, + "learning_rate": 6.798265401615378e-07, + "loss": 0.0009, + "step": 54079 + }, + { + "epoch": 7.08, + "learning_rate": 6.796343562157459e-07, + "loss": 0.0005, + "step": 54080 + }, + { + "epoch": 7.08, + "learning_rate": 6.79442198482988e-07, + "loss": 0.0007, + "step": 54081 + }, + { + "epoch": 7.08, + "learning_rate": 6.792500669638025e-07, + "loss": 0.0007, + "step": 54082 + }, + { + "epoch": 7.08, + "learning_rate": 6.790579616587301e-07, + "loss": 0.0004, + "step": 54083 + }, + { + "epoch": 7.08, + "learning_rate": 6.788658825683103e-07, + "loss": 0.0005, + "step": 54084 + }, + { + "epoch": 7.08, + "learning_rate": 6.78673829693084e-07, + "loss": 0.0009, + "step": 54085 + }, + { + "epoch": 7.08, + "learning_rate": 6.784818030335927e-07, + "loss": 0.0006, + "step": 54086 + }, + { + "epoch": 7.09, + "learning_rate": 6.782898025903739e-07, + "loss": 0.0004, + "step": 54087 + }, + { + "epoch": 7.09, + "learning_rate": 6.780978283639672e-07, + "loss": 0.0006, + "step": 54088 + }, + { + "epoch": 7.09, + "learning_rate": 6.779058803549154e-07, + "loss": 0.0008, + "step": 54089 + }, + { + "epoch": 7.09, + "learning_rate": 6.777139585637571e-07, + "loss": 0.0003, + "step": 54090 + }, + { + "epoch": 7.09, + "learning_rate": 6.775220629910296e-07, + "loss": 0.0003, + "step": 54091 + }, + { + "epoch": 7.09, + "learning_rate": 6.773301936372756e-07, + "loss": 0.0006, + "step": 54092 + }, + { + "epoch": 7.09, + "learning_rate": 6.771383505030349e-07, + "loss": 0.0003, + "step": 54093 + }, + { + "epoch": 7.09, + "learning_rate": 6.76946533588847e-07, + "loss": 0.0004, + "step": 54094 + }, + { + "epoch": 7.09, + "learning_rate": 6.767547428952504e-07, + "loss": 0.0004, + "step": 54095 + }, + { + "epoch": 7.09, + "learning_rate": 6.765629784227823e-07, + "loss": 0.0006, + "step": 54096 + }, + { + "epoch": 7.09, + "learning_rate": 6.763712401719868e-07, + "loss": 0.0007, + "step": 54097 + }, + { + "epoch": 7.09, + "learning_rate": 6.761795281434003e-07, + "loss": 0.0004, + "step": 54098 + }, + { + "epoch": 7.09, + "learning_rate": 6.759878423375599e-07, + "loss": 0.0004, + "step": 54099 + }, + { + "epoch": 7.09, + "learning_rate": 6.757961827550097e-07, + "loss": 0.0005, + "step": 54100 + }, + { + "epoch": 7.09, + "learning_rate": 6.75604549396286e-07, + "loss": 0.0003, + "step": 54101 + }, + { + "epoch": 7.09, + "learning_rate": 6.75412942261926e-07, + "loss": 0.0006, + "step": 54102 + }, + { + "epoch": 7.09, + "learning_rate": 6.752213613524727e-07, + "loss": 0.0006, + "step": 54103 + }, + { + "epoch": 7.09, + "learning_rate": 6.750298066684613e-07, + "loss": 0.0005, + "step": 54104 + }, + { + "epoch": 7.09, + "learning_rate": 6.748382782104335e-07, + "loss": 0.0012, + "step": 54105 + }, + { + "epoch": 7.09, + "learning_rate": 6.746467759789254e-07, + "loss": 0.0004, + "step": 54106 + }, + { + "epoch": 7.09, + "learning_rate": 6.744552999744757e-07, + "loss": 0.0006, + "step": 54107 + }, + { + "epoch": 7.09, + "learning_rate": 6.742638501976262e-07, + "loss": 0.0002, + "step": 54108 + }, + { + "epoch": 7.09, + "learning_rate": 6.740724266489118e-07, + "loss": 0.0011, + "step": 54109 + }, + { + "epoch": 7.09, + "learning_rate": 6.738810293288734e-07, + "loss": 0.0008, + "step": 54110 + }, + { + "epoch": 7.09, + "learning_rate": 6.736896582380448e-07, + "loss": 0.0005, + "step": 54111 + }, + { + "epoch": 7.09, + "learning_rate": 6.734983133769701e-07, + "loss": 0.0018, + "step": 54112 + }, + { + "epoch": 7.09, + "learning_rate": 6.733069947461823e-07, + "loss": 0.0004, + "step": 54113 + }, + { + "epoch": 7.09, + "learning_rate": 6.731157023462242e-07, + "loss": 0.0006, + "step": 54114 + }, + { + "epoch": 7.09, + "learning_rate": 6.7292443617763e-07, + "loss": 0.0002, + "step": 54115 + }, + { + "epoch": 7.09, + "learning_rate": 6.727331962409401e-07, + "loss": 0.0011, + "step": 54116 + }, + { + "epoch": 7.09, + "learning_rate": 6.725419825366919e-07, + "loss": 0.0007, + "step": 54117 + }, + { + "epoch": 7.09, + "learning_rate": 6.723507950654218e-07, + "loss": 0.001, + "step": 54118 + }, + { + "epoch": 7.09, + "learning_rate": 6.721596338276682e-07, + "loss": 0.001, + "step": 54119 + }, + { + "epoch": 7.09, + "learning_rate": 6.719684988239695e-07, + "loss": 0.0005, + "step": 54120 + }, + { + "epoch": 7.09, + "learning_rate": 6.71777390054863e-07, + "loss": 0.0004, + "step": 54121 + }, + { + "epoch": 7.09, + "learning_rate": 6.715863075208839e-07, + "loss": 0.0004, + "step": 54122 + }, + { + "epoch": 7.09, + "learning_rate": 6.713952512225719e-07, + "loss": 0.0004, + "step": 54123 + }, + { + "epoch": 7.09, + "learning_rate": 6.712042211604652e-07, + "loss": 0.0012, + "step": 54124 + }, + { + "epoch": 7.09, + "learning_rate": 6.710132173351003e-07, + "loss": 0.0003, + "step": 54125 + }, + { + "epoch": 7.09, + "learning_rate": 6.708222397470132e-07, + "loss": 0.0002, + "step": 54126 + }, + { + "epoch": 7.09, + "learning_rate": 6.706312883967403e-07, + "loss": 0.0008, + "step": 54127 + }, + { + "epoch": 7.09, + "learning_rate": 6.704403632848222e-07, + "loss": 0.0005, + "step": 54128 + }, + { + "epoch": 7.09, + "learning_rate": 6.702494644117941e-07, + "loss": 0.0004, + "step": 54129 + }, + { + "epoch": 7.09, + "learning_rate": 6.700585917781899e-07, + "loss": 0.0007, + "step": 54130 + }, + { + "epoch": 7.09, + "learning_rate": 6.698677453845515e-07, + "loss": 0.0002, + "step": 54131 + }, + { + "epoch": 7.09, + "learning_rate": 6.696769252314129e-07, + "loss": 0.0003, + "step": 54132 + }, + { + "epoch": 7.09, + "learning_rate": 6.694861313193113e-07, + "loss": 0.0007, + "step": 54133 + }, + { + "epoch": 7.09, + "learning_rate": 6.692953636487798e-07, + "loss": 0.0002, + "step": 54134 + }, + { + "epoch": 7.09, + "learning_rate": 6.691046222203612e-07, + "loss": 0.0007, + "step": 54135 + }, + { + "epoch": 7.09, + "learning_rate": 6.689139070345885e-07, + "loss": 0.0002, + "step": 54136 + }, + { + "epoch": 7.09, + "learning_rate": 6.687232180919989e-07, + "loss": 0.0005, + "step": 54137 + }, + { + "epoch": 7.09, + "learning_rate": 6.685325553931266e-07, + "loss": 0.0005, + "step": 54138 + }, + { + "epoch": 7.09, + "learning_rate": 6.683419189385121e-07, + "loss": 0.0006, + "step": 54139 + }, + { + "epoch": 7.09, + "learning_rate": 6.681513087286873e-07, + "loss": 0.0004, + "step": 54140 + }, + { + "epoch": 7.09, + "learning_rate": 6.679607247641917e-07, + "loss": 0.0007, + "step": 54141 + }, + { + "epoch": 7.09, + "learning_rate": 6.677701670455572e-07, + "loss": 0.0005, + "step": 54142 + }, + { + "epoch": 7.09, + "learning_rate": 6.675796355733232e-07, + "loss": 0.0006, + "step": 54143 + }, + { + "epoch": 7.09, + "learning_rate": 6.673891303480251e-07, + "loss": 0.0006, + "step": 54144 + }, + { + "epoch": 7.09, + "learning_rate": 6.671986513701967e-07, + "loss": 0.0003, + "step": 54145 + }, + { + "epoch": 7.09, + "learning_rate": 6.670081986403753e-07, + "loss": 0.0006, + "step": 54146 + }, + { + "epoch": 7.09, + "learning_rate": 6.668177721590974e-07, + "loss": 0.0003, + "step": 54147 + }, + { + "epoch": 7.09, + "learning_rate": 6.666273719268979e-07, + "loss": 0.0003, + "step": 54148 + }, + { + "epoch": 7.09, + "learning_rate": 6.664369979443097e-07, + "loss": 0.0002, + "step": 54149 + }, + { + "epoch": 7.09, + "learning_rate": 6.662466502118725e-07, + "loss": 0.0003, + "step": 54150 + }, + { + "epoch": 7.09, + "learning_rate": 6.660563287301192e-07, + "loss": 0.0008, + "step": 54151 + }, + { + "epoch": 7.09, + "learning_rate": 6.658660334995859e-07, + "loss": 0.0009, + "step": 54152 + }, + { + "epoch": 7.09, + "learning_rate": 6.656757645208056e-07, + "loss": 0.0006, + "step": 54153 + }, + { + "epoch": 7.09, + "learning_rate": 6.654855217943156e-07, + "loss": 0.0006, + "step": 54154 + }, + { + "epoch": 7.09, + "learning_rate": 6.652953053206512e-07, + "loss": 0.0007, + "step": 54155 + }, + { + "epoch": 7.09, + "learning_rate": 6.651051151003462e-07, + "loss": 0.0004, + "step": 54156 + }, + { + "epoch": 7.09, + "learning_rate": 6.649149511339348e-07, + "loss": 0.0006, + "step": 54157 + }, + { + "epoch": 7.09, + "learning_rate": 6.647248134219542e-07, + "loss": 0.0005, + "step": 54158 + }, + { + "epoch": 7.09, + "learning_rate": 6.645347019649384e-07, + "loss": 0.0003, + "step": 54159 + }, + { + "epoch": 7.09, + "learning_rate": 6.643446167634226e-07, + "loss": 0.0003, + "step": 54160 + }, + { + "epoch": 7.09, + "learning_rate": 6.641545578179376e-07, + "loss": 0.0007, + "step": 54161 + }, + { + "epoch": 7.09, + "learning_rate": 6.639645251290228e-07, + "loss": 0.0009, + "step": 54162 + }, + { + "epoch": 7.09, + "learning_rate": 6.6377451869721e-07, + "loss": 0.0006, + "step": 54163 + }, + { + "epoch": 7.1, + "learning_rate": 6.635845385230333e-07, + "loss": 0.0012, + "step": 54164 + }, + { + "epoch": 7.1, + "learning_rate": 6.633945846070288e-07, + "loss": 0.0003, + "step": 54165 + }, + { + "epoch": 7.1, + "learning_rate": 6.632046569497308e-07, + "loss": 0.0003, + "step": 54166 + }, + { + "epoch": 7.1, + "learning_rate": 6.630147555516697e-07, + "loss": 0.0006, + "step": 54167 + }, + { + "epoch": 7.1, + "learning_rate": 6.628248804133852e-07, + "loss": 0.0007, + "step": 54168 + }, + { + "epoch": 7.1, + "learning_rate": 6.626350315354058e-07, + "loss": 0.0002, + "step": 54169 + }, + { + "epoch": 7.1, + "learning_rate": 6.624452089182709e-07, + "loss": 0.0005, + "step": 54170 + }, + { + "epoch": 7.1, + "learning_rate": 6.622554125625103e-07, + "loss": 0.0004, + "step": 54171 + }, + { + "epoch": 7.1, + "learning_rate": 6.620656424686588e-07, + "loss": 0.0005, + "step": 54172 + }, + { + "epoch": 7.1, + "learning_rate": 6.618758986372508e-07, + "loss": 0.0003, + "step": 54173 + }, + { + "epoch": 7.1, + "learning_rate": 6.616861810688202e-07, + "loss": 0.0006, + "step": 54174 + }, + { + "epoch": 7.1, + "learning_rate": 6.614964897638997e-07, + "loss": 0.0003, + "step": 54175 + }, + { + "epoch": 7.1, + "learning_rate": 6.613068247230214e-07, + "loss": 0.0003, + "step": 54176 + }, + { + "epoch": 7.1, + "learning_rate": 6.611171859467225e-07, + "loss": 0.0003, + "step": 54177 + }, + { + "epoch": 7.1, + "learning_rate": 6.609275734355324e-07, + "loss": 0.0007, + "step": 54178 + }, + { + "epoch": 7.1, + "learning_rate": 6.607379871899866e-07, + "loss": 0.0013, + "step": 54179 + }, + { + "epoch": 7.1, + "learning_rate": 6.605484272106177e-07, + "loss": 0.0008, + "step": 54180 + }, + { + "epoch": 7.1, + "learning_rate": 6.603588934979599e-07, + "loss": 0.0004, + "step": 54181 + }, + { + "epoch": 7.1, + "learning_rate": 6.60169386052546e-07, + "loss": 0.0013, + "step": 54182 + }, + { + "epoch": 7.1, + "learning_rate": 6.599799048749078e-07, + "loss": 0.0002, + "step": 54183 + }, + { + "epoch": 7.1, + "learning_rate": 6.597904499655772e-07, + "loss": 0.0006, + "step": 54184 + }, + { + "epoch": 7.1, + "learning_rate": 6.596010213250903e-07, + "loss": 0.0008, + "step": 54185 + }, + { + "epoch": 7.1, + "learning_rate": 6.594116189539778e-07, + "loss": 0.0004, + "step": 54186 + }, + { + "epoch": 7.1, + "learning_rate": 6.592222428527706e-07, + "loss": 0.0005, + "step": 54187 + }, + { + "epoch": 7.1, + "learning_rate": 6.590328930220058e-07, + "loss": 0.0004, + "step": 54188 + }, + { + "epoch": 7.1, + "learning_rate": 6.588435694622119e-07, + "loss": 0.0007, + "step": 54189 + }, + { + "epoch": 7.1, + "learning_rate": 6.586542721739253e-07, + "loss": 0.0004, + "step": 54190 + }, + { + "epoch": 7.1, + "learning_rate": 6.584650011576744e-07, + "loss": 0.0004, + "step": 54191 + }, + { + "epoch": 7.1, + "learning_rate": 6.58275756413993e-07, + "loss": 0.0009, + "step": 54192 + }, + { + "epoch": 7.1, + "learning_rate": 6.580865379434143e-07, + "loss": 0.0006, + "step": 54193 + }, + { + "epoch": 7.1, + "learning_rate": 6.578973457464711e-07, + "loss": 0.0004, + "step": 54194 + }, + { + "epoch": 7.1, + "learning_rate": 6.577081798236906e-07, + "loss": 0.0004, + "step": 54195 + }, + { + "epoch": 7.1, + "learning_rate": 6.575190401756115e-07, + "loss": 0.0001, + "step": 54196 + }, + { + "epoch": 7.1, + "learning_rate": 6.573299268027622e-07, + "loss": 0.0003, + "step": 54197 + }, + { + "epoch": 7.1, + "learning_rate": 6.571408397056744e-07, + "loss": 0.0003, + "step": 54198 + }, + { + "epoch": 7.1, + "learning_rate": 6.569517788848789e-07, + "loss": 0.0013, + "step": 54199 + }, + { + "epoch": 7.1, + "learning_rate": 6.567627443409097e-07, + "loss": 0.0005, + "step": 54200 + }, + { + "epoch": 7.1, + "learning_rate": 6.565737360742986e-07, + "loss": 0.0003, + "step": 54201 + }, + { + "epoch": 7.1, + "learning_rate": 6.563847540855773e-07, + "loss": 0.0003, + "step": 54202 + }, + { + "epoch": 7.1, + "learning_rate": 6.561957983752731e-07, + "loss": 0.0006, + "step": 54203 + }, + { + "epoch": 7.1, + "learning_rate": 6.560068689439236e-07, + "loss": 0.0006, + "step": 54204 + }, + { + "epoch": 7.1, + "learning_rate": 6.558179657920571e-07, + "loss": 0.0006, + "step": 54205 + }, + { + "epoch": 7.1, + "learning_rate": 6.556290889202044e-07, + "loss": 0.0006, + "step": 54206 + }, + { + "epoch": 7.1, + "learning_rate": 6.554402383288961e-07, + "loss": 0.0009, + "step": 54207 + }, + { + "epoch": 7.1, + "learning_rate": 6.552514140186649e-07, + "loss": 0.0007, + "step": 54208 + }, + { + "epoch": 7.1, + "learning_rate": 6.55062615990043e-07, + "loss": 0.0003, + "step": 54209 + }, + { + "epoch": 7.1, + "learning_rate": 6.548738442435576e-07, + "loss": 0.0006, + "step": 54210 + }, + { + "epoch": 7.1, + "learning_rate": 6.546850987797426e-07, + "loss": 0.0003, + "step": 54211 + }, + { + "epoch": 7.1, + "learning_rate": 6.544963795991288e-07, + "loss": 0.0003, + "step": 54212 + }, + { + "epoch": 7.1, + "learning_rate": 6.543076867022458e-07, + "loss": 0.0009, + "step": 54213 + }, + { + "epoch": 7.1, + "learning_rate": 6.541190200896241e-07, + "loss": 0.0004, + "step": 54214 + }, + { + "epoch": 7.1, + "learning_rate": 6.539303797617968e-07, + "loss": 0.0002, + "step": 54215 + }, + { + "epoch": 7.1, + "learning_rate": 6.537417657192912e-07, + "loss": 0.0006, + "step": 54216 + }, + { + "epoch": 7.1, + "learning_rate": 6.535531779626403e-07, + "loss": 0.0003, + "step": 54217 + }, + { + "epoch": 7.1, + "learning_rate": 6.533646164923713e-07, + "loss": 0.0004, + "step": 54218 + }, + { + "epoch": 7.1, + "learning_rate": 6.531760813090183e-07, + "loss": 0.0003, + "step": 54219 + }, + { + "epoch": 7.1, + "learning_rate": 6.529875724131107e-07, + "loss": 0.0002, + "step": 54220 + }, + { + "epoch": 7.1, + "learning_rate": 6.527990898051762e-07, + "loss": 0.0004, + "step": 54221 + }, + { + "epoch": 7.1, + "learning_rate": 6.52610633485744e-07, + "loss": 0.0004, + "step": 54222 + }, + { + "epoch": 7.1, + "learning_rate": 6.524222034553507e-07, + "loss": 0.0003, + "step": 54223 + }, + { + "epoch": 7.1, + "learning_rate": 6.522337997145212e-07, + "loss": 0.0003, + "step": 54224 + }, + { + "epoch": 7.1, + "learning_rate": 6.520454222637862e-07, + "loss": 0.0003, + "step": 54225 + }, + { + "epoch": 7.1, + "learning_rate": 6.518570711036743e-07, + "loss": 0.0002, + "step": 54226 + }, + { + "epoch": 7.1, + "learning_rate": 6.516687462347182e-07, + "loss": 0.0007, + "step": 54227 + }, + { + "epoch": 7.1, + "learning_rate": 6.514804476574466e-07, + "loss": 0.0007, + "step": 54228 + }, + { + "epoch": 7.1, + "learning_rate": 6.512921753723855e-07, + "loss": 0.0008, + "step": 54229 + }, + { + "epoch": 7.1, + "learning_rate": 6.511039293800703e-07, + "loss": 0.0005, + "step": 54230 + }, + { + "epoch": 7.1, + "learning_rate": 6.509157096810259e-07, + "loss": 0.0004, + "step": 54231 + }, + { + "epoch": 7.1, + "learning_rate": 6.507275162757842e-07, + "loss": 0.0005, + "step": 54232 + }, + { + "epoch": 7.1, + "learning_rate": 6.505393491648715e-07, + "loss": 0.0003, + "step": 54233 + }, + { + "epoch": 7.1, + "learning_rate": 6.503512083488195e-07, + "loss": 0.0003, + "step": 54234 + }, + { + "epoch": 7.1, + "learning_rate": 6.501630938281578e-07, + "loss": 0.0004, + "step": 54235 + }, + { + "epoch": 7.1, + "learning_rate": 6.499750056034148e-07, + "loss": 0.0003, + "step": 54236 + }, + { + "epoch": 7.1, + "learning_rate": 6.497869436751181e-07, + "loss": 0.0002, + "step": 54237 + }, + { + "epoch": 7.1, + "learning_rate": 6.495989080437981e-07, + "loss": 0.0004, + "step": 54238 + }, + { + "epoch": 7.1, + "learning_rate": 6.494108987099845e-07, + "loss": 0.0004, + "step": 54239 + }, + { + "epoch": 7.11, + "learning_rate": 6.492229156742036e-07, + "loss": 0.0007, + "step": 54240 + }, + { + "epoch": 7.11, + "learning_rate": 6.490349589369849e-07, + "loss": 0.0011, + "step": 54241 + }, + { + "epoch": 7.11, + "learning_rate": 6.488470284988579e-07, + "loss": 0.0009, + "step": 54242 + }, + { + "epoch": 7.11, + "learning_rate": 6.486591243603513e-07, + "loss": 0.0004, + "step": 54243 + }, + { + "epoch": 7.11, + "learning_rate": 6.484712465219912e-07, + "loss": 0.0005, + "step": 54244 + }, + { + "epoch": 7.11, + "learning_rate": 6.482833949843082e-07, + "loss": 0.0002, + "step": 54245 + }, + { + "epoch": 7.11, + "learning_rate": 6.480955697478308e-07, + "loss": 0.0009, + "step": 54246 + }, + { + "epoch": 7.11, + "learning_rate": 6.479077708130877e-07, + "loss": 0.0005, + "step": 54247 + }, + { + "epoch": 7.11, + "learning_rate": 6.477199981806049e-07, + "loss": 0.0007, + "step": 54248 + }, + { + "epoch": 7.11, + "learning_rate": 6.475322518509098e-07, + "loss": 0.0005, + "step": 54249 + }, + { + "epoch": 7.11, + "learning_rate": 6.473445318245352e-07, + "loss": 0.0005, + "step": 54250 + }, + { + "epoch": 7.11, + "learning_rate": 6.471568381020043e-07, + "loss": 0.0004, + "step": 54251 + }, + { + "epoch": 7.11, + "learning_rate": 6.469691706838465e-07, + "loss": 0.0003, + "step": 54252 + }, + { + "epoch": 7.11, + "learning_rate": 6.467815295705915e-07, + "loss": 0.0005, + "step": 54253 + }, + { + "epoch": 7.11, + "learning_rate": 6.465939147627632e-07, + "loss": 0.0005, + "step": 54254 + }, + { + "epoch": 7.11, + "learning_rate": 6.464063262608921e-07, + "loss": 0.0005, + "step": 54255 + }, + { + "epoch": 7.11, + "learning_rate": 6.46218764065506e-07, + "loss": 0.0004, + "step": 54256 + }, + { + "epoch": 7.11, + "learning_rate": 6.460312281771308e-07, + "loss": 0.0011, + "step": 54257 + }, + { + "epoch": 7.11, + "learning_rate": 6.458437185962951e-07, + "loss": 0.0004, + "step": 54258 + }, + { + "epoch": 7.11, + "learning_rate": 6.456562353235263e-07, + "loss": 0.0012, + "step": 54259 + }, + { + "epoch": 7.11, + "learning_rate": 6.454687783593494e-07, + "loss": 0.0007, + "step": 54260 + }, + { + "epoch": 7.11, + "learning_rate": 6.452813477042952e-07, + "loss": 0.0006, + "step": 54261 + }, + { + "epoch": 7.11, + "learning_rate": 6.450939433588887e-07, + "loss": 0.0008, + "step": 54262 + }, + { + "epoch": 7.11, + "learning_rate": 6.449065653236575e-07, + "loss": 0.0003, + "step": 54263 + }, + { + "epoch": 7.11, + "learning_rate": 6.447192135991265e-07, + "loss": 0.0009, + "step": 54264 + }, + { + "epoch": 7.11, + "learning_rate": 6.445318881858242e-07, + "loss": 0.0002, + "step": 54265 + }, + { + "epoch": 7.11, + "learning_rate": 6.443445890842803e-07, + "loss": 0.0005, + "step": 54266 + }, + { + "epoch": 7.11, + "learning_rate": 6.441573162950188e-07, + "loss": 0.0005, + "step": 54267 + }, + { + "epoch": 7.11, + "learning_rate": 6.439700698185646e-07, + "loss": 0.0003, + "step": 54268 + }, + { + "epoch": 7.11, + "learning_rate": 6.437828496554477e-07, + "loss": 0.0007, + "step": 54269 + }, + { + "epoch": 7.11, + "learning_rate": 6.43595655806194e-07, + "loss": 0.0003, + "step": 54270 + }, + { + "epoch": 7.11, + "learning_rate": 6.434084882713287e-07, + "loss": 0.0005, + "step": 54271 + }, + { + "epoch": 7.11, + "learning_rate": 6.432213470513781e-07, + "loss": 0.0009, + "step": 54272 + }, + { + "epoch": 7.11, + "learning_rate": 6.430342321468697e-07, + "loss": 0.0004, + "step": 54273 + }, + { + "epoch": 7.11, + "learning_rate": 6.428471435583306e-07, + "loss": 0.0008, + "step": 54274 + }, + { + "epoch": 7.11, + "learning_rate": 6.426600812862827e-07, + "loss": 0.0016, + "step": 54275 + }, + { + "epoch": 7.11, + "learning_rate": 6.424730453312556e-07, + "loss": 0.0004, + "step": 54276 + }, + { + "epoch": 7.11, + "learning_rate": 6.422860356937755e-07, + "loss": 0.0004, + "step": 54277 + }, + { + "epoch": 7.11, + "learning_rate": 6.420990523743687e-07, + "loss": 0.0004, + "step": 54278 + }, + { + "epoch": 7.11, + "learning_rate": 6.41912095373558e-07, + "loss": 0.0002, + "step": 54279 + }, + { + "epoch": 7.11, + "learning_rate": 6.417251646918732e-07, + "loss": 0.0004, + "step": 54280 + }, + { + "epoch": 7.11, + "learning_rate": 6.415382603298371e-07, + "loss": 0.0004, + "step": 54281 + }, + { + "epoch": 7.11, + "learning_rate": 6.413513822879768e-07, + "loss": 0.0007, + "step": 54282 + }, + { + "epoch": 7.11, + "learning_rate": 6.411645305668157e-07, + "loss": 0.0003, + "step": 54283 + }, + { + "epoch": 7.11, + "learning_rate": 6.40977705166882e-07, + "loss": 0.0005, + "step": 54284 + }, + { + "epoch": 7.11, + "learning_rate": 6.407909060887008e-07, + "loss": 0.0007, + "step": 54285 + }, + { + "epoch": 7.11, + "learning_rate": 6.406041333327972e-07, + "loss": 0.0007, + "step": 54286 + }, + { + "epoch": 7.11, + "learning_rate": 6.404173868996921e-07, + "loss": 0.0004, + "step": 54287 + }, + { + "epoch": 7.11, + "learning_rate": 6.402306667899182e-07, + "loss": 0.0002, + "step": 54288 + }, + { + "epoch": 7.11, + "learning_rate": 6.400439730039965e-07, + "loss": 0.001, + "step": 54289 + }, + { + "epoch": 7.11, + "learning_rate": 6.398573055424528e-07, + "loss": 0.0004, + "step": 54290 + }, + { + "epoch": 7.11, + "learning_rate": 6.396706644058115e-07, + "loss": 0.0009, + "step": 54291 + }, + { + "epoch": 7.11, + "learning_rate": 6.394840495945986e-07, + "loss": 0.0006, + "step": 54292 + }, + { + "epoch": 7.11, + "learning_rate": 6.392974611093384e-07, + "loss": 0.0005, + "step": 54293 + }, + { + "epoch": 7.11, + "learning_rate": 6.391108989505545e-07, + "loss": 0.0008, + "step": 54294 + }, + { + "epoch": 7.11, + "learning_rate": 6.389243631187736e-07, + "loss": 0.0002, + "step": 54295 + }, + { + "epoch": 7.11, + "learning_rate": 6.387378536145194e-07, + "loss": 0.0006, + "step": 54296 + }, + { + "epoch": 7.11, + "learning_rate": 6.385513704383173e-07, + "loss": 0.0002, + "step": 54297 + }, + { + "epoch": 7.11, + "learning_rate": 6.383649135906889e-07, + "loss": 0.0009, + "step": 54298 + }, + { + "epoch": 7.11, + "learning_rate": 6.381784830721605e-07, + "loss": 0.0005, + "step": 54299 + }, + { + "epoch": 7.11, + "learning_rate": 6.379920788832584e-07, + "loss": 0.0004, + "step": 54300 + }, + { + "epoch": 7.11, + "learning_rate": 6.378057010245054e-07, + "loss": 0.0019, + "step": 54301 + }, + { + "epoch": 7.11, + "learning_rate": 6.376193494964234e-07, + "loss": 0.0007, + "step": 54302 + }, + { + "epoch": 7.11, + "learning_rate": 6.374330242995397e-07, + "loss": 0.0007, + "step": 54303 + }, + { + "epoch": 7.11, + "learning_rate": 6.372467254343772e-07, + "loss": 0.0006, + "step": 54304 + }, + { + "epoch": 7.11, + "learning_rate": 6.370604529014591e-07, + "loss": 0.0006, + "step": 54305 + }, + { + "epoch": 7.11, + "learning_rate": 6.36874206701309e-07, + "loss": 0.0005, + "step": 54306 + }, + { + "epoch": 7.11, + "learning_rate": 6.366879868344534e-07, + "loss": 0.0005, + "step": 54307 + }, + { + "epoch": 7.11, + "learning_rate": 6.365017933014129e-07, + "loss": 0.0006, + "step": 54308 + }, + { + "epoch": 7.11, + "learning_rate": 6.363156261027126e-07, + "loss": 0.0006, + "step": 54309 + }, + { + "epoch": 7.11, + "learning_rate": 6.361294852388744e-07, + "loss": 0.0009, + "step": 54310 + }, + { + "epoch": 7.11, + "learning_rate": 6.359433707104257e-07, + "loss": 0.0003, + "step": 54311 + }, + { + "epoch": 7.11, + "learning_rate": 6.357572825178882e-07, + "loss": 0.0005, + "step": 54312 + }, + { + "epoch": 7.11, + "learning_rate": 6.355712206617837e-07, + "loss": 0.0002, + "step": 54313 + }, + { + "epoch": 7.11, + "learning_rate": 6.353851851426352e-07, + "loss": 0.0006, + "step": 54314 + }, + { + "epoch": 7.11, + "learning_rate": 6.351991759609688e-07, + "loss": 0.0005, + "step": 54315 + }, + { + "epoch": 7.12, + "learning_rate": 6.350131931173065e-07, + "loss": 0.0006, + "step": 54316 + }, + { + "epoch": 7.12, + "learning_rate": 6.348272366121689e-07, + "loss": 0.0003, + "step": 54317 + }, + { + "epoch": 7.12, + "learning_rate": 6.346413064460821e-07, + "loss": 0.0003, + "step": 54318 + }, + { + "epoch": 7.12, + "learning_rate": 6.344554026195682e-07, + "loss": 0.0004, + "step": 54319 + }, + { + "epoch": 7.12, + "learning_rate": 6.342695251331488e-07, + "loss": 0.0006, + "step": 54320 + }, + { + "epoch": 7.12, + "learning_rate": 6.340836739873479e-07, + "loss": 0.0005, + "step": 54321 + }, + { + "epoch": 7.12, + "learning_rate": 6.338978491826874e-07, + "loss": 0.0005, + "step": 54322 + }, + { + "epoch": 7.12, + "learning_rate": 6.337120507196914e-07, + "loss": 0.0007, + "step": 54323 + }, + { + "epoch": 7.12, + "learning_rate": 6.335262785988816e-07, + "loss": 0.0005, + "step": 54324 + }, + { + "epoch": 7.12, + "learning_rate": 6.333405328207786e-07, + "loss": 0.0004, + "step": 54325 + }, + { + "epoch": 7.12, + "learning_rate": 6.331548133859089e-07, + "loss": 0.0004, + "step": 54326 + }, + { + "epoch": 7.12, + "learning_rate": 6.329691202947918e-07, + "loss": 0.0005, + "step": 54327 + }, + { + "epoch": 7.12, + "learning_rate": 6.327834535479505e-07, + "loss": 0.0003, + "step": 54328 + }, + { + "epoch": 7.12, + "learning_rate": 6.325978131459054e-07, + "loss": 0.0005, + "step": 54329 + }, + { + "epoch": 7.12, + "learning_rate": 6.324121990891807e-07, + "loss": 0.0004, + "step": 54330 + }, + { + "epoch": 7.12, + "learning_rate": 6.322266113782971e-07, + "loss": 0.0006, + "step": 54331 + }, + { + "epoch": 7.12, + "learning_rate": 6.320410500137797e-07, + "loss": 0.0004, + "step": 54332 + }, + { + "epoch": 7.12, + "learning_rate": 6.318555149961448e-07, + "loss": 0.0006, + "step": 54333 + }, + { + "epoch": 7.12, + "learning_rate": 6.316700063259196e-07, + "loss": 0.0007, + "step": 54334 + }, + { + "epoch": 7.12, + "learning_rate": 6.314845240036238e-07, + "loss": 0.0006, + "step": 54335 + }, + { + "epoch": 7.12, + "learning_rate": 6.312990680297792e-07, + "loss": 0.0004, + "step": 54336 + }, + { + "epoch": 7.12, + "learning_rate": 6.311136384049044e-07, + "loss": 0.0004, + "step": 54337 + }, + { + "epoch": 7.12, + "learning_rate": 6.309282351295265e-07, + "loss": 0.0005, + "step": 54338 + }, + { + "epoch": 7.12, + "learning_rate": 6.30742858204163e-07, + "loss": 0.0008, + "step": 54339 + }, + { + "epoch": 7.12, + "learning_rate": 6.305575076293346e-07, + "loss": 0.0004, + "step": 54340 + }, + { + "epoch": 7.12, + "learning_rate": 6.303721834055665e-07, + "loss": 0.0003, + "step": 54341 + }, + { + "epoch": 7.12, + "learning_rate": 6.301868855333759e-07, + "loss": 0.0007, + "step": 54342 + }, + { + "epoch": 7.12, + "learning_rate": 6.30001614013287e-07, + "loss": 0.0014, + "step": 54343 + }, + { + "epoch": 7.12, + "learning_rate": 6.298163688458181e-07, + "loss": 0.0007, + "step": 54344 + }, + { + "epoch": 7.12, + "learning_rate": 6.296311500314933e-07, + "loss": 0.0013, + "step": 54345 + }, + { + "epoch": 7.12, + "learning_rate": 6.294459575708322e-07, + "loss": 0.0003, + "step": 54346 + }, + { + "epoch": 7.12, + "learning_rate": 6.292607914643556e-07, + "loss": 0.001, + "step": 54347 + }, + { + "epoch": 7.12, + "learning_rate": 6.290756517125817e-07, + "loss": 0.0004, + "step": 54348 + }, + { + "epoch": 7.12, + "learning_rate": 6.288905383160349e-07, + "loss": 0.0008, + "step": 54349 + }, + { + "epoch": 7.12, + "learning_rate": 6.287054512752355e-07, + "loss": 0.0009, + "step": 54350 + }, + { + "epoch": 7.12, + "learning_rate": 6.285203905907022e-07, + "loss": 0.0004, + "step": 54351 + }, + { + "epoch": 7.12, + "learning_rate": 6.283353562629546e-07, + "loss": 0.0004, + "step": 54352 + }, + { + "epoch": 7.12, + "learning_rate": 6.281503482925144e-07, + "loss": 0.0005, + "step": 54353 + }, + { + "epoch": 7.12, + "learning_rate": 6.279653666799046e-07, + "loss": 0.0008, + "step": 54354 + }, + { + "epoch": 7.12, + "learning_rate": 6.277804114256425e-07, + "loss": 0.0007, + "step": 54355 + }, + { + "epoch": 7.12, + "learning_rate": 6.275954825302478e-07, + "loss": 0.0003, + "step": 54356 + }, + { + "epoch": 7.12, + "learning_rate": 6.274105799942431e-07, + "loss": 0.001, + "step": 54357 + }, + { + "epoch": 7.12, + "learning_rate": 6.272257038181473e-07, + "loss": 0.0006, + "step": 54358 + }, + { + "epoch": 7.12, + "learning_rate": 6.270408540024786e-07, + "loss": 0.0005, + "step": 54359 + }, + { + "epoch": 7.12, + "learning_rate": 6.268560305477589e-07, + "loss": 0.0006, + "step": 54360 + }, + { + "epoch": 7.12, + "learning_rate": 6.266712334545089e-07, + "loss": 0.0006, + "step": 54361 + }, + { + "epoch": 7.12, + "learning_rate": 6.264864627232458e-07, + "loss": 0.0007, + "step": 54362 + }, + { + "epoch": 7.12, + "learning_rate": 6.263017183544895e-07, + "loss": 0.0002, + "step": 54363 + }, + { + "epoch": 7.12, + "learning_rate": 6.261170003487593e-07, + "loss": 0.0003, + "step": 54364 + }, + { + "epoch": 7.12, + "learning_rate": 6.259323087065783e-07, + "loss": 0.0005, + "step": 54365 + }, + { + "epoch": 7.12, + "learning_rate": 6.257476434284637e-07, + "loss": 0.0004, + "step": 54366 + }, + { + "epoch": 7.12, + "learning_rate": 6.25563004514933e-07, + "loss": 0.0003, + "step": 54367 + }, + { + "epoch": 7.12, + "learning_rate": 6.253783919665079e-07, + "loss": 0.0007, + "step": 54368 + }, + { + "epoch": 7.12, + "learning_rate": 6.251938057837082e-07, + "loss": 0.0003, + "step": 54369 + }, + { + "epoch": 7.12, + "learning_rate": 6.250092459670498e-07, + "loss": 0.0006, + "step": 54370 + }, + { + "epoch": 7.12, + "learning_rate": 6.248247125170537e-07, + "loss": 0.0006, + "step": 54371 + }, + { + "epoch": 7.12, + "learning_rate": 6.246402054342404e-07, + "loss": 0.0002, + "step": 54372 + }, + { + "epoch": 7.12, + "learning_rate": 6.244557247191263e-07, + "loss": 0.0002, + "step": 54373 + }, + { + "epoch": 7.12, + "learning_rate": 6.242712703722298e-07, + "loss": 0.0006, + "step": 54374 + }, + { + "epoch": 7.12, + "learning_rate": 6.240868423940716e-07, + "loss": 0.0006, + "step": 54375 + }, + { + "epoch": 7.12, + "learning_rate": 6.239024407851713e-07, + "loss": 0.0003, + "step": 54376 + }, + { + "epoch": 7.12, + "learning_rate": 6.237180655460451e-07, + "loss": 0.0004, + "step": 54377 + }, + { + "epoch": 7.12, + "learning_rate": 6.235337166772137e-07, + "loss": 0.0003, + "step": 54378 + }, + { + "epoch": 7.12, + "learning_rate": 6.233493941791924e-07, + "loss": 0.0011, + "step": 54379 + }, + { + "epoch": 7.12, + "learning_rate": 6.231650980525039e-07, + "loss": 0.0005, + "step": 54380 + }, + { + "epoch": 7.12, + "learning_rate": 6.229808282976635e-07, + "loss": 0.0004, + "step": 54381 + }, + { + "epoch": 7.12, + "learning_rate": 6.227965849151885e-07, + "loss": 0.0005, + "step": 54382 + }, + { + "epoch": 7.12, + "learning_rate": 6.226123679056007e-07, + "loss": 0.0006, + "step": 54383 + }, + { + "epoch": 7.12, + "learning_rate": 6.224281772694151e-07, + "loss": 0.0003, + "step": 54384 + }, + { + "epoch": 7.12, + "learning_rate": 6.222440130071506e-07, + "loss": 0.0002, + "step": 54385 + }, + { + "epoch": 7.12, + "learning_rate": 6.220598751193263e-07, + "loss": 0.0007, + "step": 54386 + }, + { + "epoch": 7.12, + "learning_rate": 6.218757636064576e-07, + "loss": 0.0004, + "step": 54387 + }, + { + "epoch": 7.12, + "learning_rate": 6.216916784690663e-07, + "loss": 0.0003, + "step": 54388 + }, + { + "epoch": 7.12, + "learning_rate": 6.215076197076674e-07, + "loss": 0.0007, + "step": 54389 + }, + { + "epoch": 7.12, + "learning_rate": 6.213235873227775e-07, + "loss": 0.0002, + "step": 54390 + }, + { + "epoch": 7.12, + "learning_rate": 6.211395813149157e-07, + "loss": 0.0007, + "step": 54391 + }, + { + "epoch": 7.12, + "learning_rate": 6.209556016846008e-07, + "loss": 0.0004, + "step": 54392 + }, + { + "epoch": 7.13, + "learning_rate": 6.207716484323489e-07, + "loss": 0.0009, + "step": 54393 + }, + { + "epoch": 7.13, + "learning_rate": 6.205877215586753e-07, + "loss": 0.0008, + "step": 54394 + }, + { + "epoch": 7.13, + "learning_rate": 6.204038210641006e-07, + "loss": 0.0005, + "step": 54395 + }, + { + "epoch": 7.13, + "learning_rate": 6.202199469491388e-07, + "loss": 0.0005, + "step": 54396 + }, + { + "epoch": 7.13, + "learning_rate": 6.200360992143106e-07, + "loss": 0.0015, + "step": 54397 + }, + { + "epoch": 7.13, + "learning_rate": 6.198522778601302e-07, + "loss": 0.0003, + "step": 54398 + }, + { + "epoch": 7.13, + "learning_rate": 6.19668482887118e-07, + "loss": 0.0005, + "step": 54399 + }, + { + "epoch": 7.13, + "learning_rate": 6.194847142957872e-07, + "loss": 0.0003, + "step": 54400 + }, + { + "epoch": 7.13, + "learning_rate": 6.193009720866572e-07, + "loss": 0.0006, + "step": 54401 + }, + { + "epoch": 7.13, + "learning_rate": 6.19117256260241e-07, + "loss": 0.0004, + "step": 54402 + }, + { + "epoch": 7.13, + "learning_rate": 6.189335668170604e-07, + "loss": 0.0005, + "step": 54403 + }, + { + "epoch": 7.13, + "learning_rate": 6.187499037576294e-07, + "loss": 0.0002, + "step": 54404 + }, + { + "epoch": 7.13, + "learning_rate": 6.185662670824632e-07, + "loss": 0.0004, + "step": 54405 + }, + { + "epoch": 7.13, + "learning_rate": 6.183826567920814e-07, + "loss": 0.0006, + "step": 54406 + }, + { + "epoch": 7.13, + "learning_rate": 6.181990728869969e-07, + "loss": 0.0005, + "step": 54407 + }, + { + "epoch": 7.13, + "learning_rate": 6.180155153677303e-07, + "loss": 0.0002, + "step": 54408 + }, + { + "epoch": 7.13, + "learning_rate": 6.178319842347924e-07, + "loss": 0.0004, + "step": 54409 + }, + { + "epoch": 7.13, + "learning_rate": 6.17648479488705e-07, + "loss": 0.0004, + "step": 54410 + }, + { + "epoch": 7.13, + "learning_rate": 6.174650011299821e-07, + "loss": 0.0006, + "step": 54411 + }, + { + "epoch": 7.13, + "learning_rate": 6.172815491591378e-07, + "loss": 0.0004, + "step": 54412 + }, + { + "epoch": 7.13, + "learning_rate": 6.170981235766893e-07, + "loss": 0.0007, + "step": 54413 + }, + { + "epoch": 7.13, + "learning_rate": 6.16914724383153e-07, + "loss": 0.0006, + "step": 54414 + }, + { + "epoch": 7.13, + "learning_rate": 6.167313515790451e-07, + "loss": 0.0006, + "step": 54415 + }, + { + "epoch": 7.13, + "learning_rate": 6.165480051648787e-07, + "loss": 0.0003, + "step": 54416 + }, + { + "epoch": 7.13, + "learning_rate": 6.163646851411731e-07, + "loss": 0.0004, + "step": 54417 + }, + { + "epoch": 7.13, + "learning_rate": 6.161813915084403e-07, + "loss": 0.0005, + "step": 54418 + }, + { + "epoch": 7.13, + "learning_rate": 6.159981242671997e-07, + "loss": 0.0005, + "step": 54419 + }, + { + "epoch": 7.13, + "learning_rate": 6.158148834179645e-07, + "loss": 0.0003, + "step": 54420 + }, + { + "epoch": 7.13, + "learning_rate": 6.156316689612484e-07, + "loss": 0.0002, + "step": 54421 + }, + { + "epoch": 7.13, + "learning_rate": 6.154484808975703e-07, + "loss": 0.0002, + "step": 54422 + }, + { + "epoch": 7.13, + "learning_rate": 6.152653192274427e-07, + "loss": 0.0006, + "step": 54423 + }, + { + "epoch": 7.13, + "learning_rate": 6.15082183951381e-07, + "loss": 0.0008, + "step": 54424 + }, + { + "epoch": 7.13, + "learning_rate": 6.148990750699024e-07, + "loss": 0.0002, + "step": 54425 + }, + { + "epoch": 7.13, + "learning_rate": 6.147159925835199e-07, + "loss": 0.0005, + "step": 54426 + }, + { + "epoch": 7.13, + "learning_rate": 6.145329364927499e-07, + "loss": 0.001, + "step": 54427 + }, + { + "epoch": 7.13, + "learning_rate": 6.143499067981041e-07, + "loss": 0.0002, + "step": 54428 + }, + { + "epoch": 7.13, + "learning_rate": 6.141669035000985e-07, + "loss": 0.0008, + "step": 54429 + }, + { + "epoch": 7.13, + "learning_rate": 6.139839265992509e-07, + "loss": 0.0004, + "step": 54430 + }, + { + "epoch": 7.13, + "learning_rate": 6.13800976096074e-07, + "loss": 0.0008, + "step": 54431 + }, + { + "epoch": 7.13, + "learning_rate": 6.136180519910795e-07, + "loss": 0.0004, + "step": 54432 + }, + { + "epoch": 7.13, + "learning_rate": 6.134351542847872e-07, + "loss": 0.0003, + "step": 54433 + }, + { + "epoch": 7.13, + "learning_rate": 6.132522829777077e-07, + "loss": 0.0006, + "step": 54434 + }, + { + "epoch": 7.13, + "learning_rate": 6.130694380703572e-07, + "loss": 0.0005, + "step": 54435 + }, + { + "epoch": 7.13, + "learning_rate": 6.128866195632466e-07, + "loss": 0.0003, + "step": 54436 + }, + { + "epoch": 7.13, + "learning_rate": 6.127038274568941e-07, + "loss": 0.0004, + "step": 54437 + }, + { + "epoch": 7.13, + "learning_rate": 6.125210617518129e-07, + "loss": 0.0009, + "step": 54438 + }, + { + "epoch": 7.13, + "learning_rate": 6.123383224485135e-07, + "loss": 0.0015, + "step": 54439 + }, + { + "epoch": 7.13, + "learning_rate": 6.121556095475134e-07, + "loss": 0.0003, + "step": 54440 + }, + { + "epoch": 7.13, + "learning_rate": 6.119729230493277e-07, + "loss": 0.0003, + "step": 54441 + }, + { + "epoch": 7.13, + "learning_rate": 6.117902629544681e-07, + "loss": 0.0006, + "step": 54442 + }, + { + "epoch": 7.13, + "learning_rate": 6.116076292634476e-07, + "loss": 0.0006, + "step": 54443 + }, + { + "epoch": 7.13, + "learning_rate": 6.114250219767792e-07, + "loss": 0.0006, + "step": 54444 + }, + { + "epoch": 7.13, + "learning_rate": 6.112424410949802e-07, + "loss": 0.0002, + "step": 54445 + }, + { + "epoch": 7.13, + "learning_rate": 6.110598866185613e-07, + "loss": 0.001, + "step": 54446 + }, + { + "epoch": 7.13, + "learning_rate": 6.108773585480355e-07, + "loss": 0.0004, + "step": 54447 + }, + { + "epoch": 7.13, + "learning_rate": 6.106948568839188e-07, + "loss": 0.0007, + "step": 54448 + }, + { + "epoch": 7.13, + "learning_rate": 6.105123816267223e-07, + "loss": 0.0003, + "step": 54449 + }, + { + "epoch": 7.13, + "learning_rate": 6.103299327769597e-07, + "loss": 0.001, + "step": 54450 + }, + { + "epoch": 7.13, + "learning_rate": 6.101475103351428e-07, + "loss": 0.0009, + "step": 54451 + }, + { + "epoch": 7.13, + "learning_rate": 6.099651143017859e-07, + "loss": 0.0008, + "step": 54452 + }, + { + "epoch": 7.13, + "learning_rate": 6.097827446774041e-07, + "loss": 0.0012, + "step": 54453 + }, + { + "epoch": 7.13, + "learning_rate": 6.096004014625068e-07, + "loss": 0.0004, + "step": 54454 + }, + { + "epoch": 7.13, + "learning_rate": 6.094180846576081e-07, + "loss": 0.0007, + "step": 54455 + }, + { + "epoch": 7.13, + "learning_rate": 6.092357942632221e-07, + "loss": 0.0004, + "step": 54456 + }, + { + "epoch": 7.13, + "learning_rate": 6.090535302798605e-07, + "loss": 0.0004, + "step": 54457 + }, + { + "epoch": 7.13, + "learning_rate": 6.088712927080353e-07, + "loss": 0.0005, + "step": 54458 + }, + { + "epoch": 7.13, + "learning_rate": 6.086890815482583e-07, + "loss": 0.0003, + "step": 54459 + }, + { + "epoch": 7.13, + "learning_rate": 6.085068968010454e-07, + "loss": 0.0003, + "step": 54460 + }, + { + "epoch": 7.13, + "learning_rate": 6.083247384669055e-07, + "loss": 0.0007, + "step": 54461 + }, + { + "epoch": 7.13, + "learning_rate": 6.081426065463514e-07, + "loss": 0.0003, + "step": 54462 + }, + { + "epoch": 7.13, + "learning_rate": 6.079605010398959e-07, + "loss": 0.0005, + "step": 54463 + }, + { + "epoch": 7.13, + "learning_rate": 6.07778421948052e-07, + "loss": 0.0008, + "step": 54464 + }, + { + "epoch": 7.13, + "learning_rate": 6.075963692713316e-07, + "loss": 0.0002, + "step": 54465 + }, + { + "epoch": 7.13, + "learning_rate": 6.074143430102452e-07, + "loss": 0.0003, + "step": 54466 + }, + { + "epoch": 7.13, + "learning_rate": 6.072323431653072e-07, + "loss": 0.0006, + "step": 54467 + }, + { + "epoch": 7.13, + "learning_rate": 6.070503697370278e-07, + "loss": 0.0003, + "step": 54468 + }, + { + "epoch": 7.14, + "learning_rate": 6.068684227259192e-07, + "loss": 0.0006, + "step": 54469 + }, + { + "epoch": 7.14, + "learning_rate": 6.066865021324919e-07, + "loss": 0.0009, + "step": 54470 + }, + { + "epoch": 7.14, + "learning_rate": 6.06504607957259e-07, + "loss": 0.0003, + "step": 54471 + }, + { + "epoch": 7.14, + "learning_rate": 6.063227402007321e-07, + "loss": 0.0006, + "step": 54472 + }, + { + "epoch": 7.14, + "learning_rate": 6.061408988634209e-07, + "loss": 0.0003, + "step": 54473 + }, + { + "epoch": 7.14, + "learning_rate": 6.059590839458385e-07, + "loss": 0.0012, + "step": 54474 + }, + { + "epoch": 7.14, + "learning_rate": 6.057772954484964e-07, + "loss": 0.0011, + "step": 54475 + }, + { + "epoch": 7.14, + "learning_rate": 6.055955333719066e-07, + "loss": 0.0008, + "step": 54476 + }, + { + "epoch": 7.14, + "learning_rate": 6.054137977165786e-07, + "loss": 0.0006, + "step": 54477 + }, + { + "epoch": 7.14, + "learning_rate": 6.052320884830221e-07, + "loss": 0.0003, + "step": 54478 + }, + { + "epoch": 7.14, + "learning_rate": 6.050504056717521e-07, + "loss": 0.0004, + "step": 54479 + }, + { + "epoch": 7.14, + "learning_rate": 6.048687492832772e-07, + "loss": 0.0003, + "step": 54480 + }, + { + "epoch": 7.14, + "learning_rate": 6.04687119318107e-07, + "loss": 0.0006, + "step": 54481 + }, + { + "epoch": 7.14, + "learning_rate": 6.045055157767555e-07, + "loss": 0.0003, + "step": 54482 + }, + { + "epoch": 7.14, + "learning_rate": 6.043239386597321e-07, + "loss": 0.0005, + "step": 54483 + }, + { + "epoch": 7.14, + "learning_rate": 6.041423879675456e-07, + "loss": 0.0004, + "step": 54484 + }, + { + "epoch": 7.14, + "learning_rate": 6.0396086370071e-07, + "loss": 0.0004, + "step": 54485 + }, + { + "epoch": 7.14, + "learning_rate": 6.037793658597324e-07, + "loss": 0.0003, + "step": 54486 + }, + { + "epoch": 7.14, + "learning_rate": 6.03597894445127e-07, + "loss": 0.0002, + "step": 54487 + }, + { + "epoch": 7.14, + "learning_rate": 6.034164494574013e-07, + "loss": 0.0007, + "step": 54488 + }, + { + "epoch": 7.14, + "learning_rate": 6.032350308970658e-07, + "loss": 0.0007, + "step": 54489 + }, + { + "epoch": 7.14, + "learning_rate": 6.030536387646324e-07, + "loss": 0.0004, + "step": 54490 + }, + { + "epoch": 7.14, + "learning_rate": 6.028722730606107e-07, + "loss": 0.0004, + "step": 54491 + }, + { + "epoch": 7.14, + "learning_rate": 6.026909337855091e-07, + "loss": 0.0005, + "step": 54492 + }, + { + "epoch": 7.14, + "learning_rate": 6.025096209398384e-07, + "loss": 0.0006, + "step": 54493 + }, + { + "epoch": 7.14, + "learning_rate": 6.023283345241104e-07, + "loss": 0.0002, + "step": 54494 + }, + { + "epoch": 7.14, + "learning_rate": 6.021470745388313e-07, + "loss": 0.0005, + "step": 54495 + }, + { + "epoch": 7.14, + "learning_rate": 6.019658409845153e-07, + "loss": 0.0003, + "step": 54496 + }, + { + "epoch": 7.14, + "learning_rate": 6.017846338616673e-07, + "loss": 0.0007, + "step": 54497 + }, + { + "epoch": 7.14, + "learning_rate": 6.016034531708026e-07, + "loss": 0.0003, + "step": 54498 + }, + { + "epoch": 7.14, + "learning_rate": 6.014222989124263e-07, + "loss": 0.0005, + "step": 54499 + }, + { + "epoch": 7.14, + "learning_rate": 6.012411710870502e-07, + "loss": 0.0005, + "step": 54500 + }, + { + "epoch": 7.14, + "learning_rate": 6.010600696951807e-07, + "loss": 0.0002, + "step": 54501 + }, + { + "epoch": 7.14, + "learning_rate": 6.008789947373317e-07, + "loss": 0.0004, + "step": 54502 + }, + { + "epoch": 7.14, + "learning_rate": 6.006979462140084e-07, + "loss": 0.0005, + "step": 54503 + }, + { + "epoch": 7.14, + "learning_rate": 6.005169241257214e-07, + "loss": 0.0006, + "step": 54504 + }, + { + "epoch": 7.14, + "learning_rate": 6.003359284729815e-07, + "loss": 0.0003, + "step": 54505 + }, + { + "epoch": 7.14, + "learning_rate": 6.001549592562938e-07, + "loss": 0.0002, + "step": 54506 + }, + { + "epoch": 7.14, + "learning_rate": 5.999740164761714e-07, + "loss": 0.0005, + "step": 54507 + }, + { + "epoch": 7.14, + "learning_rate": 5.997931001331226e-07, + "loss": 0.0004, + "step": 54508 + }, + { + "epoch": 7.14, + "learning_rate": 5.996122102276525e-07, + "loss": 0.0008, + "step": 54509 + }, + { + "epoch": 7.14, + "learning_rate": 5.994313467602741e-07, + "loss": 0.0006, + "step": 54510 + }, + { + "epoch": 7.14, + "learning_rate": 5.992505097314949e-07, + "loss": 0.0009, + "step": 54511 + }, + { + "epoch": 7.14, + "learning_rate": 5.99069699141821e-07, + "loss": 0.0001, + "step": 54512 + }, + { + "epoch": 7.14, + "learning_rate": 5.988889149917643e-07, + "loss": 0.0004, + "step": 54513 + }, + { + "epoch": 7.14, + "learning_rate": 5.987081572818332e-07, + "loss": 0.0002, + "step": 54514 + }, + { + "epoch": 7.14, + "learning_rate": 5.98527426012534e-07, + "loss": 0.0003, + "step": 54515 + }, + { + "epoch": 7.14, + "learning_rate": 5.983467211843719e-07, + "loss": 0.0004, + "step": 54516 + }, + { + "epoch": 7.14, + "learning_rate": 5.98166042797863e-07, + "loss": 0.0004, + "step": 54517 + }, + { + "epoch": 7.14, + "learning_rate": 5.979853908535105e-07, + "loss": 0.0004, + "step": 54518 + }, + { + "epoch": 7.14, + "learning_rate": 5.978047653518248e-07, + "loss": 0.0004, + "step": 54519 + }, + { + "epoch": 7.14, + "learning_rate": 5.976241662933102e-07, + "loss": 0.0006, + "step": 54520 + }, + { + "epoch": 7.14, + "learning_rate": 5.974435936784784e-07, + "loss": 0.0009, + "step": 54521 + }, + { + "epoch": 7.14, + "learning_rate": 5.972630475078356e-07, + "loss": 0.0002, + "step": 54522 + }, + { + "epoch": 7.14, + "learning_rate": 5.970825277818904e-07, + "loss": 0.0005, + "step": 54523 + }, + { + "epoch": 7.14, + "learning_rate": 5.969020345011478e-07, + "loss": 0.0005, + "step": 54524 + }, + { + "epoch": 7.14, + "learning_rate": 5.967215676661187e-07, + "loss": 0.0003, + "step": 54525 + }, + { + "epoch": 7.14, + "learning_rate": 5.965411272773103e-07, + "loss": 0.0006, + "step": 54526 + }, + { + "epoch": 7.14, + "learning_rate": 5.963607133352268e-07, + "loss": 0.0003, + "step": 54527 + }, + { + "epoch": 7.14, + "learning_rate": 5.961803258403787e-07, + "loss": 0.0006, + "step": 54528 + }, + { + "epoch": 7.14, + "learning_rate": 5.959999647932735e-07, + "loss": 0.0006, + "step": 54529 + }, + { + "epoch": 7.14, + "learning_rate": 5.958196301944186e-07, + "loss": 0.0002, + "step": 54530 + }, + { + "epoch": 7.14, + "learning_rate": 5.956393220443179e-07, + "loss": 0.0005, + "step": 54531 + }, + { + "epoch": 7.14, + "learning_rate": 5.954590403434824e-07, + "loss": 0.0003, + "step": 54532 + }, + { + "epoch": 7.14, + "learning_rate": 5.95278785092418e-07, + "loss": 0.0006, + "step": 54533 + }, + { + "epoch": 7.14, + "learning_rate": 5.9509855629163e-07, + "loss": 0.0004, + "step": 54534 + }, + { + "epoch": 7.14, + "learning_rate": 5.949183539416259e-07, + "loss": 0.0008, + "step": 54535 + }, + { + "epoch": 7.14, + "learning_rate": 5.947381780429151e-07, + "loss": 0.0003, + "step": 54536 + }, + { + "epoch": 7.14, + "learning_rate": 5.945580285960018e-07, + "loss": 0.0003, + "step": 54537 + }, + { + "epoch": 7.14, + "learning_rate": 5.943779056013909e-07, + "loss": 0.0002, + "step": 54538 + }, + { + "epoch": 7.14, + "learning_rate": 5.941978090595923e-07, + "loss": 0.0003, + "step": 54539 + }, + { + "epoch": 7.14, + "learning_rate": 5.940177389711122e-07, + "loss": 0.0006, + "step": 54540 + }, + { + "epoch": 7.14, + "learning_rate": 5.938376953364566e-07, + "loss": 0.0002, + "step": 54541 + }, + { + "epoch": 7.14, + "learning_rate": 5.93657678156132e-07, + "loss": 0.0003, + "step": 54542 + }, + { + "epoch": 7.14, + "learning_rate": 5.934776874306425e-07, + "loss": 0.0004, + "step": 54543 + }, + { + "epoch": 7.14, + "learning_rate": 5.932977231604975e-07, + "loss": 0.0006, + "step": 54544 + }, + { + "epoch": 7.15, + "learning_rate": 5.931177853462022e-07, + "loss": 0.0004, + "step": 54545 + }, + { + "epoch": 7.15, + "learning_rate": 5.929378739882596e-07, + "loss": 0.0003, + "step": 54546 + }, + { + "epoch": 7.15, + "learning_rate": 5.927579890871804e-07, + "loss": 0.0003, + "step": 54547 + }, + { + "epoch": 7.15, + "learning_rate": 5.925781306434686e-07, + "loss": 0.0003, + "step": 54548 + }, + { + "epoch": 7.15, + "learning_rate": 5.923982986576282e-07, + "loss": 0.0009, + "step": 54549 + }, + { + "epoch": 7.15, + "learning_rate": 5.922184931301678e-07, + "loss": 0.0006, + "step": 54550 + }, + { + "epoch": 7.15, + "learning_rate": 5.920387140615913e-07, + "loss": 0.001, + "step": 54551 + }, + { + "epoch": 7.15, + "learning_rate": 5.918589614524051e-07, + "loss": 0.0002, + "step": 54552 + }, + { + "epoch": 7.15, + "learning_rate": 5.916792353031153e-07, + "loss": 0.0002, + "step": 54553 + }, + { + "epoch": 7.15, + "learning_rate": 5.91499535614225e-07, + "loss": 0.0008, + "step": 54554 + }, + { + "epoch": 7.15, + "learning_rate": 5.913198623862426e-07, + "loss": 0.0003, + "step": 54555 + }, + { + "epoch": 7.15, + "learning_rate": 5.911402156196722e-07, + "loss": 0.0005, + "step": 54556 + }, + { + "epoch": 7.15, + "learning_rate": 5.909605953150188e-07, + "loss": 0.0002, + "step": 54557 + }, + { + "epoch": 7.15, + "learning_rate": 5.907810014727866e-07, + "loss": 0.0002, + "step": 54558 + }, + { + "epoch": 7.15, + "learning_rate": 5.906014340934829e-07, + "loss": 0.0007, + "step": 54559 + }, + { + "epoch": 7.15, + "learning_rate": 5.904218931776106e-07, + "loss": 0.0005, + "step": 54560 + }, + { + "epoch": 7.15, + "learning_rate": 5.902423787256762e-07, + "loss": 0.0003, + "step": 54561 + }, + { + "epoch": 7.15, + "learning_rate": 5.900628907381833e-07, + "loss": 0.0005, + "step": 54562 + }, + { + "epoch": 7.15, + "learning_rate": 5.898834292156386e-07, + "loss": 0.0007, + "step": 54563 + }, + { + "epoch": 7.15, + "learning_rate": 5.897039941585447e-07, + "loss": 0.0002, + "step": 54564 + }, + { + "epoch": 7.15, + "learning_rate": 5.895245855674081e-07, + "loss": 0.0004, + "step": 54565 + }, + { + "epoch": 7.15, + "learning_rate": 5.893452034427316e-07, + "loss": 0.0004, + "step": 54566 + }, + { + "epoch": 7.15, + "learning_rate": 5.891658477850216e-07, + "loss": 0.0006, + "step": 54567 + }, + { + "epoch": 7.15, + "learning_rate": 5.889865185947807e-07, + "loss": 0.0003, + "step": 54568 + }, + { + "epoch": 7.15, + "learning_rate": 5.888072158725133e-07, + "loss": 0.0006, + "step": 54569 + }, + { + "epoch": 7.15, + "learning_rate": 5.886279396187256e-07, + "loss": 0.0009, + "step": 54570 + }, + { + "epoch": 7.15, + "learning_rate": 5.884486898339193e-07, + "loss": 0.0003, + "step": 54571 + }, + { + "epoch": 7.15, + "learning_rate": 5.882694665186018e-07, + "loss": 0.0005, + "step": 54572 + }, + { + "epoch": 7.15, + "learning_rate": 5.880902696732748e-07, + "loss": 0.0004, + "step": 54573 + }, + { + "epoch": 7.15, + "learning_rate": 5.879110992984416e-07, + "loss": 0.0007, + "step": 54574 + }, + { + "epoch": 7.15, + "learning_rate": 5.877319553946082e-07, + "loss": 0.0006, + "step": 54575 + }, + { + "epoch": 7.15, + "learning_rate": 5.875528379622775e-07, + "loss": 0.0006, + "step": 54576 + }, + { + "epoch": 7.15, + "learning_rate": 5.873737470019514e-07, + "loss": 0.0004, + "step": 54577 + }, + { + "epoch": 7.15, + "learning_rate": 5.871946825141373e-07, + "loss": 0.0003, + "step": 54578 + }, + { + "epoch": 7.15, + "learning_rate": 5.870156444993369e-07, + "loss": 0.0002, + "step": 54579 + }, + { + "epoch": 7.15, + "learning_rate": 5.868366329580533e-07, + "loss": 0.0008, + "step": 54580 + }, + { + "epoch": 7.15, + "learning_rate": 5.866576478907893e-07, + "loss": 0.0007, + "step": 54581 + }, + { + "epoch": 7.15, + "learning_rate": 5.8647868929805e-07, + "loss": 0.0002, + "step": 54582 + }, + { + "epoch": 7.15, + "learning_rate": 5.862997571803386e-07, + "loss": 0.0006, + "step": 54583 + }, + { + "epoch": 7.15, + "learning_rate": 5.861208515381578e-07, + "loss": 0.0004, + "step": 54584 + }, + { + "epoch": 7.15, + "learning_rate": 5.859419723720106e-07, + "loss": 0.0005, + "step": 54585 + }, + { + "epoch": 7.15, + "learning_rate": 5.857631196823998e-07, + "loss": 0.0012, + "step": 54586 + }, + { + "epoch": 7.15, + "learning_rate": 5.855842934698308e-07, + "loss": 0.0007, + "step": 54587 + }, + { + "epoch": 7.15, + "learning_rate": 5.85405493734803e-07, + "loss": 0.0007, + "step": 54588 + }, + { + "epoch": 7.15, + "learning_rate": 5.852267204778206e-07, + "loss": 0.0002, + "step": 54589 + }, + { + "epoch": 7.15, + "learning_rate": 5.850479736993874e-07, + "loss": 0.0007, + "step": 54590 + }, + { + "epoch": 7.15, + "learning_rate": 5.848692534000056e-07, + "loss": 0.0003, + "step": 54591 + }, + { + "epoch": 7.15, + "learning_rate": 5.846905595801766e-07, + "loss": 0.0006, + "step": 54592 + }, + { + "epoch": 7.15, + "learning_rate": 5.845118922404036e-07, + "loss": 0.0006, + "step": 54593 + }, + { + "epoch": 7.15, + "learning_rate": 5.843332513811905e-07, + "loss": 0.0018, + "step": 54594 + }, + { + "epoch": 7.15, + "learning_rate": 5.841546370030393e-07, + "loss": 0.0003, + "step": 54595 + }, + { + "epoch": 7.15, + "learning_rate": 5.839760491064505e-07, + "loss": 0.0004, + "step": 54596 + }, + { + "epoch": 7.15, + "learning_rate": 5.837974876919295e-07, + "loss": 0.0006, + "step": 54597 + }, + { + "epoch": 7.15, + "learning_rate": 5.836189527599756e-07, + "loss": 0.0003, + "step": 54598 + }, + { + "epoch": 7.15, + "learning_rate": 5.83440444311092e-07, + "loss": 0.0003, + "step": 54599 + }, + { + "epoch": 7.15, + "learning_rate": 5.832619623457802e-07, + "loss": 0.0004, + "step": 54600 + }, + { + "epoch": 7.15, + "learning_rate": 5.830835068645435e-07, + "loss": 0.0005, + "step": 54601 + }, + { + "epoch": 7.15, + "learning_rate": 5.829050778678824e-07, + "loss": 0.0003, + "step": 54602 + }, + { + "epoch": 7.15, + "learning_rate": 5.827266753562999e-07, + "loss": 0.0006, + "step": 54603 + }, + { + "epoch": 7.15, + "learning_rate": 5.825482993302945e-07, + "loss": 0.0005, + "step": 54604 + }, + { + "epoch": 7.15, + "learning_rate": 5.823699497903723e-07, + "loss": 0.0003, + "step": 54605 + }, + { + "epoch": 7.15, + "learning_rate": 5.821916267370342e-07, + "loss": 0.0002, + "step": 54606 + }, + { + "epoch": 7.15, + "learning_rate": 5.820133301707798e-07, + "loss": 0.0005, + "step": 54607 + }, + { + "epoch": 7.15, + "learning_rate": 5.818350600921108e-07, + "loss": 0.0006, + "step": 54608 + }, + { + "epoch": 7.15, + "learning_rate": 5.816568165015301e-07, + "loss": 0.0004, + "step": 54609 + }, + { + "epoch": 7.15, + "learning_rate": 5.814785993995387e-07, + "loss": 0.0008, + "step": 54610 + }, + { + "epoch": 7.15, + "learning_rate": 5.813004087866348e-07, + "loss": 0.0007, + "step": 54611 + }, + { + "epoch": 7.15, + "learning_rate": 5.811222446633236e-07, + "loss": 0.0006, + "step": 54612 + }, + { + "epoch": 7.15, + "learning_rate": 5.809441070301048e-07, + "loss": 0.0003, + "step": 54613 + }, + { + "epoch": 7.15, + "learning_rate": 5.80765995887479e-07, + "loss": 0.0008, + "step": 54614 + }, + { + "epoch": 7.15, + "learning_rate": 5.805879112359447e-07, + "loss": 0.0009, + "step": 54615 + }, + { + "epoch": 7.15, + "learning_rate": 5.804098530760072e-07, + "loss": 0.0006, + "step": 54616 + }, + { + "epoch": 7.15, + "learning_rate": 5.802318214081659e-07, + "loss": 0.0003, + "step": 54617 + }, + { + "epoch": 7.15, + "learning_rate": 5.800538162329205e-07, + "loss": 0.0003, + "step": 54618 + }, + { + "epoch": 7.15, + "learning_rate": 5.798758375507707e-07, + "loss": 0.0003, + "step": 54619 + }, + { + "epoch": 7.15, + "learning_rate": 5.796978853622204e-07, + "loss": 0.0005, + "step": 54620 + }, + { + "epoch": 7.15, + "learning_rate": 5.795199596677681e-07, + "loss": 0.0002, + "step": 54621 + }, + { + "epoch": 7.16, + "learning_rate": 5.793420604679146e-07, + "loss": 0.0002, + "step": 54622 + }, + { + "epoch": 7.16, + "learning_rate": 5.791641877631582e-07, + "loss": 0.001, + "step": 54623 + }, + { + "epoch": 7.16, + "learning_rate": 5.78986341554002e-07, + "loss": 0.0003, + "step": 54624 + }, + { + "epoch": 7.16, + "learning_rate": 5.788085218409456e-07, + "loss": 0.0003, + "step": 54625 + }, + { + "epoch": 7.16, + "learning_rate": 5.786307286244863e-07, + "loss": 0.0008, + "step": 54626 + }, + { + "epoch": 7.16, + "learning_rate": 5.784529619051261e-07, + "loss": 0.0004, + "step": 54627 + }, + { + "epoch": 7.16, + "learning_rate": 5.782752216833676e-07, + "loss": 0.0004, + "step": 54628 + }, + { + "epoch": 7.16, + "learning_rate": 5.780975079597074e-07, + "loss": 0.0009, + "step": 54629 + }, + { + "epoch": 7.16, + "learning_rate": 5.77919820734647e-07, + "loss": 0.0015, + "step": 54630 + }, + { + "epoch": 7.16, + "learning_rate": 5.777421600086841e-07, + "loss": 0.0005, + "step": 54631 + }, + { + "epoch": 7.16, + "learning_rate": 5.775645257823203e-07, + "loss": 0.0002, + "step": 54632 + }, + { + "epoch": 7.16, + "learning_rate": 5.773869180560543e-07, + "loss": 0.0006, + "step": 54633 + }, + { + "epoch": 7.16, + "learning_rate": 5.772093368303855e-07, + "loss": 0.0003, + "step": 54634 + }, + { + "epoch": 7.16, + "learning_rate": 5.770317821058147e-07, + "loss": 0.0004, + "step": 54635 + }, + { + "epoch": 7.16, + "learning_rate": 5.768542538828393e-07, + "loss": 0.0007, + "step": 54636 + }, + { + "epoch": 7.16, + "learning_rate": 5.766767521619587e-07, + "loss": 0.0003, + "step": 54637 + }, + { + "epoch": 7.16, + "learning_rate": 5.76499276943675e-07, + "loss": 0.0005, + "step": 54638 + }, + { + "epoch": 7.16, + "learning_rate": 5.763218282284833e-07, + "loss": 0.0008, + "step": 54639 + }, + { + "epoch": 7.16, + "learning_rate": 5.761444060168853e-07, + "loss": 0.0006, + "step": 54640 + }, + { + "epoch": 7.16, + "learning_rate": 5.759670103093806e-07, + "loss": 0.0009, + "step": 54641 + }, + { + "epoch": 7.16, + "learning_rate": 5.757896411064645e-07, + "loss": 0.0009, + "step": 54642 + }, + { + "epoch": 7.16, + "learning_rate": 5.756122984086399e-07, + "loss": 0.0004, + "step": 54643 + }, + { + "epoch": 7.16, + "learning_rate": 5.754349822164041e-07, + "loss": 0.001, + "step": 54644 + }, + { + "epoch": 7.16, + "learning_rate": 5.752576925302545e-07, + "loss": 0.0002, + "step": 54645 + }, + { + "epoch": 7.16, + "learning_rate": 5.750804293506907e-07, + "loss": 0.0006, + "step": 54646 + }, + { + "epoch": 7.16, + "learning_rate": 5.7490319267821e-07, + "loss": 0.0002, + "step": 54647 + }, + { + "epoch": 7.16, + "learning_rate": 5.747259825133145e-07, + "loss": 0.0009, + "step": 54648 + }, + { + "epoch": 7.16, + "learning_rate": 5.74548798856499e-07, + "loss": 0.0006, + "step": 54649 + }, + { + "epoch": 7.16, + "learning_rate": 5.743716417082634e-07, + "loss": 0.0005, + "step": 54650 + }, + { + "epoch": 7.16, + "learning_rate": 5.741945110691061e-07, + "loss": 0.0006, + "step": 54651 + }, + { + "epoch": 7.16, + "learning_rate": 5.740174069395244e-07, + "loss": 0.0004, + "step": 54652 + }, + { + "epoch": 7.16, + "learning_rate": 5.738403293200179e-07, + "loss": 0.0003, + "step": 54653 + }, + { + "epoch": 7.16, + "learning_rate": 5.736632782110807e-07, + "loss": 0.0008, + "step": 54654 + }, + { + "epoch": 7.16, + "learning_rate": 5.734862536132157e-07, + "loss": 0.0007, + "step": 54655 + }, + { + "epoch": 7.16, + "learning_rate": 5.733092555269193e-07, + "loss": 0.0005, + "step": 54656 + }, + { + "epoch": 7.16, + "learning_rate": 5.731322839526865e-07, + "loss": 0.0002, + "step": 54657 + }, + { + "epoch": 7.16, + "learning_rate": 5.72955338891017e-07, + "loss": 0.0007, + "step": 54658 + }, + { + "epoch": 7.16, + "learning_rate": 5.727784203424102e-07, + "loss": 0.0004, + "step": 54659 + }, + { + "epoch": 7.16, + "learning_rate": 5.726015283073616e-07, + "loss": 0.0006, + "step": 54660 + }, + { + "epoch": 7.16, + "learning_rate": 5.724246627863684e-07, + "loss": 0.0003, + "step": 54661 + }, + { + "epoch": 7.16, + "learning_rate": 5.722478237799301e-07, + "loss": 0.0006, + "step": 54662 + }, + { + "epoch": 7.16, + "learning_rate": 5.72071011288543e-07, + "loss": 0.0004, + "step": 54663 + }, + { + "epoch": 7.16, + "learning_rate": 5.718942253127036e-07, + "loss": 0.0003, + "step": 54664 + }, + { + "epoch": 7.16, + "learning_rate": 5.717174658529079e-07, + "loss": 0.0003, + "step": 54665 + }, + { + "epoch": 7.16, + "learning_rate": 5.715407329096567e-07, + "loss": 0.0008, + "step": 54666 + }, + { + "epoch": 7.16, + "learning_rate": 5.713640264834452e-07, + "loss": 0.0006, + "step": 54667 + }, + { + "epoch": 7.16, + "learning_rate": 5.711873465747708e-07, + "loss": 0.0004, + "step": 54668 + }, + { + "epoch": 7.16, + "learning_rate": 5.710106931841253e-07, + "loss": 0.0004, + "step": 54669 + }, + { + "epoch": 7.16, + "learning_rate": 5.708340663120149e-07, + "loss": 0.0013, + "step": 54670 + }, + { + "epoch": 7.16, + "learning_rate": 5.706574659589303e-07, + "loss": 0.0004, + "step": 54671 + }, + { + "epoch": 7.16, + "learning_rate": 5.70480892125369e-07, + "loss": 0.0003, + "step": 54672 + }, + { + "epoch": 7.16, + "learning_rate": 5.703043448118284e-07, + "loss": 0.0006, + "step": 54673 + }, + { + "epoch": 7.16, + "learning_rate": 5.701278240188046e-07, + "loss": 0.0005, + "step": 54674 + }, + { + "epoch": 7.16, + "learning_rate": 5.69951329746794e-07, + "loss": 0.0002, + "step": 54675 + }, + { + "epoch": 7.16, + "learning_rate": 5.697748619962929e-07, + "loss": 0.0004, + "step": 54676 + }, + { + "epoch": 7.16, + "learning_rate": 5.695984207677985e-07, + "loss": 0.0005, + "step": 54677 + }, + { + "epoch": 7.16, + "learning_rate": 5.694220060618061e-07, + "loss": 0.0011, + "step": 54678 + }, + { + "epoch": 7.16, + "learning_rate": 5.692456178788131e-07, + "loss": 0.0004, + "step": 54679 + }, + { + "epoch": 7.16, + "learning_rate": 5.690692562193123e-07, + "loss": 0.0008, + "step": 54680 + }, + { + "epoch": 7.16, + "learning_rate": 5.688929210838024e-07, + "loss": 0.0003, + "step": 54681 + }, + { + "epoch": 7.16, + "learning_rate": 5.687166124727795e-07, + "loss": 0.0008, + "step": 54682 + }, + { + "epoch": 7.16, + "learning_rate": 5.685403303867398e-07, + "loss": 0.0004, + "step": 54683 + }, + { + "epoch": 7.16, + "learning_rate": 5.683640748261765e-07, + "loss": 0.0002, + "step": 54684 + }, + { + "epoch": 7.16, + "learning_rate": 5.681878457915879e-07, + "loss": 0.0009, + "step": 54685 + }, + { + "epoch": 7.16, + "learning_rate": 5.680116432834692e-07, + "loss": 0.001, + "step": 54686 + }, + { + "epoch": 7.16, + "learning_rate": 5.678354673023157e-07, + "loss": 0.001, + "step": 54687 + }, + { + "epoch": 7.16, + "learning_rate": 5.676593178486211e-07, + "loss": 0.0002, + "step": 54688 + }, + { + "epoch": 7.16, + "learning_rate": 5.674831949228831e-07, + "loss": 0.0003, + "step": 54689 + }, + { + "epoch": 7.16, + "learning_rate": 5.67307098525598e-07, + "loss": 0.0006, + "step": 54690 + }, + { + "epoch": 7.16, + "learning_rate": 5.671310286572562e-07, + "loss": 0.0006, + "step": 54691 + }, + { + "epoch": 7.16, + "learning_rate": 5.669549853183565e-07, + "loss": 0.0009, + "step": 54692 + }, + { + "epoch": 7.16, + "learning_rate": 5.66778968509395e-07, + "loss": 0.0005, + "step": 54693 + }, + { + "epoch": 7.16, + "learning_rate": 5.666029782308657e-07, + "loss": 0.0004, + "step": 54694 + }, + { + "epoch": 7.16, + "learning_rate": 5.664270144832628e-07, + "loss": 0.0005, + "step": 54695 + }, + { + "epoch": 7.16, + "learning_rate": 5.662510772670804e-07, + "loss": 0.0004, + "step": 54696 + }, + { + "epoch": 7.16, + "learning_rate": 5.660751665828157e-07, + "loss": 0.0002, + "step": 54697 + }, + { + "epoch": 7.17, + "learning_rate": 5.658992824309628e-07, + "loss": 0.0003, + "step": 54698 + }, + { + "epoch": 7.17, + "learning_rate": 5.657234248120136e-07, + "loss": 0.0004, + "step": 54699 + }, + { + "epoch": 7.17, + "learning_rate": 5.655475937264665e-07, + "loss": 0.0004, + "step": 54700 + }, + { + "epoch": 7.17, + "learning_rate": 5.653717891748145e-07, + "loss": 0.0004, + "step": 54701 + }, + { + "epoch": 7.17, + "learning_rate": 5.651960111575494e-07, + "loss": 0.0003, + "step": 54702 + }, + { + "epoch": 7.17, + "learning_rate": 5.650202596751708e-07, + "loss": 0.0003, + "step": 54703 + }, + { + "epoch": 7.17, + "learning_rate": 5.648445347281683e-07, + "loss": 0.0002, + "step": 54704 + }, + { + "epoch": 7.17, + "learning_rate": 5.646688363170405e-07, + "loss": 0.0003, + "step": 54705 + }, + { + "epoch": 7.17, + "learning_rate": 5.64493164442278e-07, + "loss": 0.0004, + "step": 54706 + }, + { + "epoch": 7.17, + "learning_rate": 5.643175191043749e-07, + "loss": 0.0006, + "step": 54707 + }, + { + "epoch": 7.17, + "learning_rate": 5.641419003038273e-07, + "loss": 0.0005, + "step": 54708 + }, + { + "epoch": 7.17, + "learning_rate": 5.639663080411295e-07, + "loss": 0.0004, + "step": 54709 + }, + { + "epoch": 7.17, + "learning_rate": 5.637907423167733e-07, + "loss": 0.0002, + "step": 54710 + }, + { + "epoch": 7.17, + "learning_rate": 5.636152031312503e-07, + "loss": 0.0013, + "step": 54711 + }, + { + "epoch": 7.17, + "learning_rate": 5.634396904850604e-07, + "loss": 0.0007, + "step": 54712 + }, + { + "epoch": 7.17, + "learning_rate": 5.632642043786917e-07, + "loss": 0.0009, + "step": 54713 + }, + { + "epoch": 7.17, + "learning_rate": 5.63088744812641e-07, + "loss": 0.0005, + "step": 54714 + }, + { + "epoch": 7.17, + "learning_rate": 5.629133117873998e-07, + "loss": 0.0002, + "step": 54715 + }, + { + "epoch": 7.17, + "learning_rate": 5.627379053034632e-07, + "loss": 0.0004, + "step": 54716 + }, + { + "epoch": 7.17, + "learning_rate": 5.625625253613243e-07, + "loss": 0.0004, + "step": 54717 + }, + { + "epoch": 7.17, + "learning_rate": 5.623871719614749e-07, + "loss": 0.001, + "step": 54718 + }, + { + "epoch": 7.17, + "learning_rate": 5.622118451044079e-07, + "loss": 0.0003, + "step": 54719 + }, + { + "epoch": 7.17, + "learning_rate": 5.620365447906196e-07, + "loss": 0.0003, + "step": 54720 + }, + { + "epoch": 7.17, + "learning_rate": 5.618612710205995e-07, + "loss": 0.0008, + "step": 54721 + }, + { + "epoch": 7.17, + "learning_rate": 5.616860237948418e-07, + "loss": 0.0004, + "step": 54722 + }, + { + "epoch": 7.17, + "learning_rate": 5.615108031138405e-07, + "loss": 0.0003, + "step": 54723 + }, + { + "epoch": 7.17, + "learning_rate": 5.613356089780863e-07, + "loss": 0.0012, + "step": 54724 + }, + { + "epoch": 7.17, + "learning_rate": 5.611604413880745e-07, + "loss": 0.0014, + "step": 54725 + }, + { + "epoch": 7.17, + "learning_rate": 5.609853003442945e-07, + "loss": 0.0004, + "step": 54726 + }, + { + "epoch": 7.17, + "learning_rate": 5.608101858472426e-07, + "loss": 0.0006, + "step": 54727 + }, + { + "epoch": 7.17, + "learning_rate": 5.606350978974085e-07, + "loss": 0.0005, + "step": 54728 + }, + { + "epoch": 7.17, + "learning_rate": 5.604600364952862e-07, + "loss": 0.0008, + "step": 54729 + }, + { + "epoch": 7.17, + "learning_rate": 5.602850016413653e-07, + "loss": 0.0004, + "step": 54730 + }, + { + "epoch": 7.17, + "learning_rate": 5.60109993336142e-07, + "loss": 0.0007, + "step": 54731 + }, + { + "epoch": 7.17, + "learning_rate": 5.599350115801072e-07, + "loss": 0.0008, + "step": 54732 + }, + { + "epoch": 7.17, + "learning_rate": 5.597600563737504e-07, + "loss": 0.0001, + "step": 54733 + }, + { + "epoch": 7.17, + "learning_rate": 5.595851277175679e-07, + "loss": 0.0007, + "step": 54734 + }, + { + "epoch": 7.17, + "learning_rate": 5.594102256120471e-07, + "loss": 0.0003, + "step": 54735 + }, + { + "epoch": 7.17, + "learning_rate": 5.592353500576841e-07, + "loss": 0.0003, + "step": 54736 + }, + { + "epoch": 7.17, + "learning_rate": 5.590605010549688e-07, + "loss": 0.0012, + "step": 54737 + }, + { + "epoch": 7.17, + "learning_rate": 5.588856786043917e-07, + "loss": 0.0007, + "step": 54738 + }, + { + "epoch": 7.17, + "learning_rate": 5.587108827064469e-07, + "loss": 0.0002, + "step": 54739 + }, + { + "epoch": 7.17, + "learning_rate": 5.585361133616263e-07, + "loss": 0.001, + "step": 54740 + }, + { + "epoch": 7.17, + "learning_rate": 5.583613705704172e-07, + "loss": 0.0002, + "step": 54741 + }, + { + "epoch": 7.17, + "learning_rate": 5.581866543333159e-07, + "loss": 0.0005, + "step": 54742 + }, + { + "epoch": 7.17, + "learning_rate": 5.580119646508119e-07, + "loss": 0.0005, + "step": 54743 + }, + { + "epoch": 7.17, + "learning_rate": 5.578373015233962e-07, + "loss": 0.001, + "step": 54744 + }, + { + "epoch": 7.17, + "learning_rate": 5.576626649515593e-07, + "loss": 0.0002, + "step": 54745 + }, + { + "epoch": 7.17, + "learning_rate": 5.57488054935793e-07, + "loss": 0.0002, + "step": 54746 + }, + { + "epoch": 7.17, + "learning_rate": 5.573134714765904e-07, + "loss": 0.0007, + "step": 54747 + }, + { + "epoch": 7.17, + "learning_rate": 5.571389145744411e-07, + "loss": 0.0004, + "step": 54748 + }, + { + "epoch": 7.17, + "learning_rate": 5.569643842298334e-07, + "loss": 0.0007, + "step": 54749 + }, + { + "epoch": 7.17, + "learning_rate": 5.567898804432625e-07, + "loss": 0.0003, + "step": 54750 + }, + { + "epoch": 7.17, + "learning_rate": 5.566154032152172e-07, + "loss": 0.0003, + "step": 54751 + }, + { + "epoch": 7.17, + "learning_rate": 5.564409525461878e-07, + "loss": 0.0004, + "step": 54752 + }, + { + "epoch": 7.17, + "learning_rate": 5.562665284366642e-07, + "loss": 0.0006, + "step": 54753 + }, + { + "epoch": 7.17, + "learning_rate": 5.560921308871392e-07, + "loss": 0.001, + "step": 54754 + }, + { + "epoch": 7.17, + "learning_rate": 5.559177598981025e-07, + "loss": 0.0003, + "step": 54755 + }, + { + "epoch": 7.17, + "learning_rate": 5.557434154700425e-07, + "loss": 0.0011, + "step": 54756 + }, + { + "epoch": 7.17, + "learning_rate": 5.555690976034511e-07, + "loss": 0.0006, + "step": 54757 + }, + { + "epoch": 7.17, + "learning_rate": 5.553948062988213e-07, + "loss": 0.0009, + "step": 54758 + }, + { + "epoch": 7.17, + "learning_rate": 5.552205415566392e-07, + "loss": 0.0008, + "step": 54759 + }, + { + "epoch": 7.17, + "learning_rate": 5.550463033773967e-07, + "loss": 0.0003, + "step": 54760 + }, + { + "epoch": 7.17, + "learning_rate": 5.548720917615824e-07, + "loss": 0.0006, + "step": 54761 + }, + { + "epoch": 7.17, + "learning_rate": 5.54697906709688e-07, + "loss": 0.0006, + "step": 54762 + }, + { + "epoch": 7.17, + "learning_rate": 5.545237482222044e-07, + "loss": 0.0008, + "step": 54763 + }, + { + "epoch": 7.17, + "learning_rate": 5.543496162996164e-07, + "loss": 0.0013, + "step": 54764 + }, + { + "epoch": 7.17, + "learning_rate": 5.541755109424196e-07, + "loss": 0.0002, + "step": 54765 + }, + { + "epoch": 7.17, + "learning_rate": 5.540014321511012e-07, + "loss": 0.0006, + "step": 54766 + }, + { + "epoch": 7.17, + "learning_rate": 5.538273799261507e-07, + "loss": 0.0006, + "step": 54767 + }, + { + "epoch": 7.17, + "learning_rate": 5.536533542680555e-07, + "loss": 0.0002, + "step": 54768 + }, + { + "epoch": 7.17, + "learning_rate": 5.534793551773077e-07, + "loss": 0.0005, + "step": 54769 + }, + { + "epoch": 7.17, + "learning_rate": 5.533053826543977e-07, + "loss": 0.0002, + "step": 54770 + }, + { + "epoch": 7.17, + "learning_rate": 5.531314366998131e-07, + "loss": 0.0019, + "step": 54771 + }, + { + "epoch": 7.17, + "learning_rate": 5.529575173140422e-07, + "loss": 0.0002, + "step": 54772 + }, + { + "epoch": 7.17, + "learning_rate": 5.52783624497576e-07, + "loss": 0.0002, + "step": 54773 + }, + { + "epoch": 7.18, + "learning_rate": 5.526097582509038e-07, + "loss": 0.0003, + "step": 54774 + }, + { + "epoch": 7.18, + "learning_rate": 5.52435918574512e-07, + "loss": 0.0004, + "step": 54775 + }, + { + "epoch": 7.18, + "learning_rate": 5.522621054688904e-07, + "loss": 0.0004, + "step": 54776 + }, + { + "epoch": 7.18, + "learning_rate": 5.520883189345306e-07, + "loss": 0.0008, + "step": 54777 + }, + { + "epoch": 7.18, + "learning_rate": 5.519145589719177e-07, + "loss": 0.0007, + "step": 54778 + }, + { + "epoch": 7.18, + "learning_rate": 5.517408255815427e-07, + "loss": 0.0005, + "step": 54779 + }, + { + "epoch": 7.18, + "learning_rate": 5.515671187638916e-07, + "loss": 0.0002, + "step": 54780 + }, + { + "epoch": 7.18, + "learning_rate": 5.513934385194575e-07, + "loss": 0.0008, + "step": 54781 + }, + { + "epoch": 7.18, + "learning_rate": 5.512197848487255e-07, + "loss": 0.0002, + "step": 54782 + }, + { + "epoch": 7.18, + "learning_rate": 5.510461577521831e-07, + "loss": 0.0002, + "step": 54783 + }, + { + "epoch": 7.18, + "learning_rate": 5.50872557230322e-07, + "loss": 0.0002, + "step": 54784 + }, + { + "epoch": 7.18, + "learning_rate": 5.506989832836284e-07, + "loss": 0.0005, + "step": 54785 + }, + { + "epoch": 7.18, + "learning_rate": 5.505254359125911e-07, + "loss": 0.0002, + "step": 54786 + }, + { + "epoch": 7.18, + "learning_rate": 5.503519151176962e-07, + "loss": 0.0007, + "step": 54787 + }, + { + "epoch": 7.18, + "learning_rate": 5.501784208994343e-07, + "loss": 0.0003, + "step": 54788 + }, + { + "epoch": 7.18, + "learning_rate": 5.500049532582919e-07, + "loss": 0.0004, + "step": 54789 + }, + { + "epoch": 7.18, + "learning_rate": 5.498315121947573e-07, + "loss": 0.0007, + "step": 54790 + }, + { + "epoch": 7.18, + "learning_rate": 5.49658097709318e-07, + "loss": 0.0003, + "step": 54791 + }, + { + "epoch": 7.18, + "learning_rate": 5.494847098024636e-07, + "loss": 0.0006, + "step": 54792 + }, + { + "epoch": 7.18, + "learning_rate": 5.493113484746792e-07, + "loss": 0.0003, + "step": 54793 + }, + { + "epoch": 7.18, + "learning_rate": 5.491380137264546e-07, + "loss": 0.001, + "step": 54794 + }, + { + "epoch": 7.18, + "learning_rate": 5.489647055582736e-07, + "loss": 0.0002, + "step": 54795 + }, + { + "epoch": 7.18, + "learning_rate": 5.487914239706282e-07, + "loss": 0.0004, + "step": 54796 + }, + { + "epoch": 7.18, + "learning_rate": 5.486181689640035e-07, + "loss": 0.0004, + "step": 54797 + }, + { + "epoch": 7.18, + "learning_rate": 5.484449405388848e-07, + "loss": 0.0004, + "step": 54798 + }, + { + "epoch": 7.18, + "learning_rate": 5.482717386957636e-07, + "loss": 0.0007, + "step": 54799 + }, + { + "epoch": 7.18, + "learning_rate": 5.480985634351222e-07, + "loss": 0.0006, + "step": 54800 + }, + { + "epoch": 7.18, + "learning_rate": 5.479254147574531e-07, + "loss": 0.0003, + "step": 54801 + }, + { + "epoch": 7.18, + "learning_rate": 5.477522926632395e-07, + "loss": 0.0003, + "step": 54802 + }, + { + "epoch": 7.18, + "learning_rate": 5.475791971529687e-07, + "loss": 0.0009, + "step": 54803 + }, + { + "epoch": 7.18, + "learning_rate": 5.474061282271281e-07, + "loss": 0.0005, + "step": 54804 + }, + { + "epoch": 7.18, + "learning_rate": 5.472330858862051e-07, + "loss": 0.0012, + "step": 54805 + }, + { + "epoch": 7.18, + "learning_rate": 5.470600701306849e-07, + "loss": 0.0008, + "step": 54806 + }, + { + "epoch": 7.18, + "learning_rate": 5.46887080961055e-07, + "loss": 0.0006, + "step": 54807 + }, + { + "epoch": 7.18, + "learning_rate": 5.467141183778024e-07, + "loss": 0.0006, + "step": 54808 + }, + { + "epoch": 7.18, + "learning_rate": 5.465411823814137e-07, + "loss": 0.0003, + "step": 54809 + }, + { + "epoch": 7.18, + "learning_rate": 5.463682729723719e-07, + "loss": 0.0003, + "step": 54810 + }, + { + "epoch": 7.18, + "learning_rate": 5.461953901511663e-07, + "loss": 0.0004, + "step": 54811 + }, + { + "epoch": 7.18, + "learning_rate": 5.460225339182846e-07, + "loss": 0.0004, + "step": 54812 + }, + { + "epoch": 7.18, + "learning_rate": 5.458497042742106e-07, + "loss": 0.0012, + "step": 54813 + }, + { + "epoch": 7.18, + "learning_rate": 5.456769012194285e-07, + "loss": 0.0003, + "step": 54814 + }, + { + "epoch": 7.18, + "learning_rate": 5.45504124754429e-07, + "loss": 0.0004, + "step": 54815 + }, + { + "epoch": 7.18, + "learning_rate": 5.453313748796962e-07, + "loss": 0.0003, + "step": 54816 + }, + { + "epoch": 7.18, + "learning_rate": 5.451586515957152e-07, + "loss": 0.0004, + "step": 54817 + }, + { + "epoch": 7.18, + "learning_rate": 5.449859549029701e-07, + "loss": 0.0008, + "step": 54818 + }, + { + "epoch": 7.18, + "learning_rate": 5.448132848019505e-07, + "loss": 0.0003, + "step": 54819 + }, + { + "epoch": 7.18, + "learning_rate": 5.446406412931393e-07, + "loss": 0.0013, + "step": 54820 + }, + { + "epoch": 7.18, + "learning_rate": 5.444680243770218e-07, + "loss": 0.0012, + "step": 54821 + }, + { + "epoch": 7.18, + "learning_rate": 5.44295434054084e-07, + "loss": 0.0003, + "step": 54822 + }, + { + "epoch": 7.18, + "learning_rate": 5.441228703248136e-07, + "loss": 0.0007, + "step": 54823 + }, + { + "epoch": 7.18, + "learning_rate": 5.439503331896934e-07, + "loss": 0.0008, + "step": 54824 + }, + { + "epoch": 7.18, + "learning_rate": 5.437778226492107e-07, + "loss": 0.0005, + "step": 54825 + }, + { + "epoch": 7.18, + "learning_rate": 5.436053387038464e-07, + "loss": 0.0002, + "step": 54826 + }, + { + "epoch": 7.18, + "learning_rate": 5.434328813540912e-07, + "loss": 0.0003, + "step": 54827 + }, + { + "epoch": 7.18, + "learning_rate": 5.432604506004269e-07, + "loss": 0.0002, + "step": 54828 + }, + { + "epoch": 7.18, + "learning_rate": 5.430880464433364e-07, + "loss": 0.0003, + "step": 54829 + }, + { + "epoch": 7.18, + "learning_rate": 5.429156688833104e-07, + "loss": 0.0003, + "step": 54830 + }, + { + "epoch": 7.18, + "learning_rate": 5.427433179208297e-07, + "loss": 0.0002, + "step": 54831 + }, + { + "epoch": 7.18, + "learning_rate": 5.425709935563794e-07, + "loss": 0.0002, + "step": 54832 + }, + { + "epoch": 7.18, + "learning_rate": 5.423986957904426e-07, + "loss": 0.0006, + "step": 54833 + }, + { + "epoch": 7.18, + "learning_rate": 5.422264246235076e-07, + "loss": 0.0007, + "step": 54834 + }, + { + "epoch": 7.18, + "learning_rate": 5.420541800560575e-07, + "loss": 0.0004, + "step": 54835 + }, + { + "epoch": 7.18, + "learning_rate": 5.418819620885774e-07, + "loss": 0.0004, + "step": 54836 + }, + { + "epoch": 7.18, + "learning_rate": 5.41709770721548e-07, + "loss": 0.0003, + "step": 54837 + }, + { + "epoch": 7.18, + "learning_rate": 5.415376059554578e-07, + "loss": 0.001, + "step": 54838 + }, + { + "epoch": 7.18, + "learning_rate": 5.413654677907898e-07, + "loss": 0.0007, + "step": 54839 + }, + { + "epoch": 7.18, + "learning_rate": 5.411933562280269e-07, + "loss": 0.0006, + "step": 54840 + }, + { + "epoch": 7.18, + "learning_rate": 5.410212712676533e-07, + "loss": 0.0005, + "step": 54841 + }, + { + "epoch": 7.18, + "learning_rate": 5.40849212910155e-07, + "loss": 0.0008, + "step": 54842 + }, + { + "epoch": 7.18, + "learning_rate": 5.406771811560141e-07, + "loss": 0.0007, + "step": 54843 + }, + { + "epoch": 7.18, + "learning_rate": 5.405051760057134e-07, + "loss": 0.0004, + "step": 54844 + }, + { + "epoch": 7.18, + "learning_rate": 5.403331974597392e-07, + "loss": 0.0007, + "step": 54845 + }, + { + "epoch": 7.18, + "learning_rate": 5.401612455185745e-07, + "loss": 0.0005, + "step": 54846 + }, + { + "epoch": 7.18, + "learning_rate": 5.399893201827033e-07, + "loss": 0.0003, + "step": 54847 + }, + { + "epoch": 7.18, + "learning_rate": 5.398174214526064e-07, + "loss": 0.0009, + "step": 54848 + }, + { + "epoch": 7.18, + "learning_rate": 5.396455493287711e-07, + "loss": 0.0004, + "step": 54849 + }, + { + "epoch": 7.18, + "learning_rate": 5.394737038116782e-07, + "loss": 0.0005, + "step": 54850 + }, + { + "epoch": 7.19, + "learning_rate": 5.393018849018117e-07, + "loss": 0.0013, + "step": 54851 + }, + { + "epoch": 7.19, + "learning_rate": 5.391300925996546e-07, + "loss": 0.0012, + "step": 54852 + }, + { + "epoch": 7.19, + "learning_rate": 5.389583269056908e-07, + "loss": 0.0003, + "step": 54853 + }, + { + "epoch": 7.19, + "learning_rate": 5.387865878204024e-07, + "loss": 0.0004, + "step": 54854 + }, + { + "epoch": 7.19, + "learning_rate": 5.386148753442722e-07, + "loss": 0.0003, + "step": 54855 + }, + { + "epoch": 7.19, + "learning_rate": 5.384431894777842e-07, + "loss": 0.0007, + "step": 54856 + }, + { + "epoch": 7.19, + "learning_rate": 5.382715302214215e-07, + "loss": 0.0003, + "step": 54857 + }, + { + "epoch": 7.19, + "learning_rate": 5.380998975756668e-07, + "loss": 0.001, + "step": 54858 + }, + { + "epoch": 7.19, + "learning_rate": 5.379282915410011e-07, + "loss": 0.0003, + "step": 54859 + }, + { + "epoch": 7.19, + "learning_rate": 5.377567121179083e-07, + "loss": 0.0007, + "step": 54860 + }, + { + "epoch": 7.19, + "learning_rate": 5.375851593068715e-07, + "loss": 0.0007, + "step": 54861 + }, + { + "epoch": 7.19, + "learning_rate": 5.374136331083724e-07, + "loss": 0.0012, + "step": 54862 + }, + { + "epoch": 7.19, + "learning_rate": 5.372421335228928e-07, + "loss": 0.0005, + "step": 54863 + }, + { + "epoch": 7.19, + "learning_rate": 5.370706605509168e-07, + "loss": 0.0006, + "step": 54864 + }, + { + "epoch": 7.19, + "learning_rate": 5.368992141929252e-07, + "loss": 0.0003, + "step": 54865 + }, + { + "epoch": 7.19, + "learning_rate": 5.367277944493998e-07, + "loss": 0.0004, + "step": 54866 + }, + { + "epoch": 7.19, + "learning_rate": 5.365564013208246e-07, + "loss": 0.0001, + "step": 54867 + }, + { + "epoch": 7.19, + "learning_rate": 5.363850348076794e-07, + "loss": 0.0006, + "step": 54868 + }, + { + "epoch": 7.19, + "learning_rate": 5.362136949104491e-07, + "loss": 0.0004, + "step": 54869 + }, + { + "epoch": 7.19, + "learning_rate": 5.360423816296134e-07, + "loss": 0.0013, + "step": 54870 + }, + { + "epoch": 7.19, + "learning_rate": 5.358710949656531e-07, + "loss": 0.0012, + "step": 54871 + }, + { + "epoch": 7.19, + "learning_rate": 5.356998349190523e-07, + "loss": 0.0006, + "step": 54872 + }, + { + "epoch": 7.19, + "learning_rate": 5.355286014902927e-07, + "loss": 0.0004, + "step": 54873 + }, + { + "epoch": 7.19, + "learning_rate": 5.353573946798541e-07, + "loss": 0.0003, + "step": 54874 + }, + { + "epoch": 7.19, + "learning_rate": 5.351862144882181e-07, + "loss": 0.0005, + "step": 54875 + }, + { + "epoch": 7.19, + "learning_rate": 5.350150609158678e-07, + "loss": 0.0004, + "step": 54876 + }, + { + "epoch": 7.19, + "learning_rate": 5.348439339632816e-07, + "loss": 0.0004, + "step": 54877 + }, + { + "epoch": 7.19, + "learning_rate": 5.346728336309447e-07, + "loss": 0.0004, + "step": 54878 + }, + { + "epoch": 7.19, + "learning_rate": 5.345017599193347e-07, + "loss": 0.0004, + "step": 54879 + }, + { + "epoch": 7.19, + "learning_rate": 5.343307128289365e-07, + "loss": 0.0004, + "step": 54880 + }, + { + "epoch": 7.19, + "learning_rate": 5.341596923602288e-07, + "loss": 0.0006, + "step": 54881 + }, + { + "epoch": 7.19, + "learning_rate": 5.339886985136933e-07, + "loss": 0.0004, + "step": 54882 + }, + { + "epoch": 7.19, + "learning_rate": 5.338177312898085e-07, + "loss": 0.0002, + "step": 54883 + }, + { + "epoch": 7.19, + "learning_rate": 5.336467906890585e-07, + "loss": 0.0004, + "step": 54884 + }, + { + "epoch": 7.19, + "learning_rate": 5.33475876711923e-07, + "loss": 0.0009, + "step": 54885 + }, + { + "epoch": 7.19, + "learning_rate": 5.333049893588815e-07, + "loss": 0.0003, + "step": 54886 + }, + { + "epoch": 7.19, + "learning_rate": 5.33134128630417e-07, + "loss": 0.0002, + "step": 54887 + }, + { + "epoch": 7.19, + "learning_rate": 5.329632945270069e-07, + "loss": 0.0004, + "step": 54888 + }, + { + "epoch": 7.19, + "learning_rate": 5.327924870491341e-07, + "loss": 0.0005, + "step": 54889 + }, + { + "epoch": 7.19, + "learning_rate": 5.326217061972793e-07, + "loss": 0.0003, + "step": 54890 + }, + { + "epoch": 7.19, + "learning_rate": 5.324509519719201e-07, + "loss": 0.0003, + "step": 54891 + }, + { + "epoch": 7.19, + "learning_rate": 5.322802243735392e-07, + "loss": 0.0007, + "step": 54892 + }, + { + "epoch": 7.19, + "learning_rate": 5.321095234026163e-07, + "loss": 0.0005, + "step": 54893 + }, + { + "epoch": 7.19, + "learning_rate": 5.3193884905963e-07, + "loss": 0.0006, + "step": 54894 + }, + { + "epoch": 7.19, + "learning_rate": 5.317682013450631e-07, + "loss": 0.0003, + "step": 54895 + }, + { + "epoch": 7.19, + "learning_rate": 5.315975802593942e-07, + "loss": 0.0004, + "step": 54896 + }, + { + "epoch": 7.19, + "learning_rate": 5.314269858031019e-07, + "loss": 0.0005, + "step": 54897 + }, + { + "epoch": 7.19, + "learning_rate": 5.312564179766666e-07, + "loss": 0.0007, + "step": 54898 + }, + { + "epoch": 7.19, + "learning_rate": 5.310858767805671e-07, + "loss": 0.0005, + "step": 54899 + }, + { + "epoch": 7.19, + "learning_rate": 5.309153622152862e-07, + "loss": 0.0007, + "step": 54900 + }, + { + "epoch": 7.19, + "learning_rate": 5.307448742813026e-07, + "loss": 0.0003, + "step": 54901 + }, + { + "epoch": 7.19, + "learning_rate": 5.305744129790924e-07, + "loss": 0.0004, + "step": 54902 + }, + { + "epoch": 7.19, + "learning_rate": 5.304039783091386e-07, + "loss": 0.0003, + "step": 54903 + }, + { + "epoch": 7.19, + "learning_rate": 5.302335702719196e-07, + "loss": 0.0014, + "step": 54904 + }, + { + "epoch": 7.19, + "learning_rate": 5.300631888679153e-07, + "loss": 0.0003, + "step": 54905 + }, + { + "epoch": 7.19, + "learning_rate": 5.298928340976018e-07, + "loss": 0.0005, + "step": 54906 + }, + { + "epoch": 7.19, + "learning_rate": 5.297225059614619e-07, + "loss": 0.0005, + "step": 54907 + }, + { + "epoch": 7.19, + "learning_rate": 5.295522044599722e-07, + "loss": 0.0004, + "step": 54908 + }, + { + "epoch": 7.19, + "learning_rate": 5.29381929593612e-07, + "loss": 0.0007, + "step": 54909 + }, + { + "epoch": 7.19, + "learning_rate": 5.292116813628612e-07, + "loss": 0.0011, + "step": 54910 + }, + { + "epoch": 7.19, + "learning_rate": 5.290414597681992e-07, + "loss": 0.0011, + "step": 54911 + }, + { + "epoch": 7.19, + "learning_rate": 5.288712648101036e-07, + "loss": 0.0005, + "step": 54912 + }, + { + "epoch": 7.19, + "learning_rate": 5.287010964890516e-07, + "loss": 0.0007, + "step": 54913 + }, + { + "epoch": 7.19, + "learning_rate": 5.285309548055251e-07, + "loss": 0.0003, + "step": 54914 + }, + { + "epoch": 7.19, + "learning_rate": 5.283608397600015e-07, + "loss": 0.0006, + "step": 54915 + }, + { + "epoch": 7.19, + "learning_rate": 5.281907513529583e-07, + "loss": 0.0005, + "step": 54916 + }, + { + "epoch": 7.19, + "learning_rate": 5.280206895848716e-07, + "loss": 0.0002, + "step": 54917 + }, + { + "epoch": 7.19, + "learning_rate": 5.278506544562256e-07, + "loss": 0.0002, + "step": 54918 + }, + { + "epoch": 7.19, + "learning_rate": 5.276806459674932e-07, + "loss": 0.0003, + "step": 54919 + }, + { + "epoch": 7.19, + "learning_rate": 5.275106641191552e-07, + "loss": 0.0003, + "step": 54920 + }, + { + "epoch": 7.19, + "learning_rate": 5.273407089116878e-07, + "loss": 0.0004, + "step": 54921 + }, + { + "epoch": 7.19, + "learning_rate": 5.271707803455717e-07, + "loss": 0.0002, + "step": 54922 + }, + { + "epoch": 7.19, + "learning_rate": 5.270008784212833e-07, + "loss": 0.0009, + "step": 54923 + }, + { + "epoch": 7.19, + "learning_rate": 5.268310031393009e-07, + "loss": 0.0003, + "step": 54924 + }, + { + "epoch": 7.19, + "learning_rate": 5.266611545000999e-07, + "loss": 0.0002, + "step": 54925 + }, + { + "epoch": 7.19, + "learning_rate": 5.26491332504162e-07, + "loss": 0.0006, + "step": 54926 + }, + { + "epoch": 7.2, + "learning_rate": 5.263215371519625e-07, + "loss": 0.0004, + "step": 54927 + }, + { + "epoch": 7.2, + "learning_rate": 5.261517684439776e-07, + "loss": 0.0003, + "step": 54928 + }, + { + "epoch": 7.2, + "learning_rate": 5.259820263806869e-07, + "loss": 0.0003, + "step": 54929 + }, + { + "epoch": 7.2, + "learning_rate": 5.258123109625679e-07, + "loss": 0.0004, + "step": 54930 + }, + { + "epoch": 7.2, + "learning_rate": 5.256426221900968e-07, + "loss": 0.0004, + "step": 54931 + }, + { + "epoch": 7.2, + "learning_rate": 5.25472960063752e-07, + "loss": 0.0005, + "step": 54932 + }, + { + "epoch": 7.2, + "learning_rate": 5.253033245840078e-07, + "loss": 0.0002, + "step": 54933 + }, + { + "epoch": 7.2, + "learning_rate": 5.251337157513458e-07, + "loss": 0.0006, + "step": 54934 + }, + { + "epoch": 7.2, + "learning_rate": 5.249641335662403e-07, + "loss": 0.0003, + "step": 54935 + }, + { + "epoch": 7.2, + "learning_rate": 5.247945780291674e-07, + "loss": 0.0006, + "step": 54936 + }, + { + "epoch": 7.2, + "learning_rate": 5.246250491406057e-07, + "loss": 0.0006, + "step": 54937 + }, + { + "epoch": 7.2, + "learning_rate": 5.244555469010326e-07, + "loss": 0.0006, + "step": 54938 + }, + { + "epoch": 7.2, + "learning_rate": 5.242860713109232e-07, + "loss": 0.0008, + "step": 54939 + }, + { + "epoch": 7.2, + "learning_rate": 5.241166223707539e-07, + "loss": 0.0003, + "step": 54940 + }, + { + "epoch": 7.2, + "learning_rate": 5.23947200081002e-07, + "loss": 0.0008, + "step": 54941 + }, + { + "epoch": 7.2, + "learning_rate": 5.237778044421438e-07, + "loss": 0.0003, + "step": 54942 + }, + { + "epoch": 7.2, + "learning_rate": 5.236084354546566e-07, + "loss": 0.0004, + "step": 54943 + }, + { + "epoch": 7.2, + "learning_rate": 5.234390931190148e-07, + "loss": 0.0003, + "step": 54944 + }, + { + "epoch": 7.2, + "learning_rate": 5.232697774356976e-07, + "loss": 0.0009, + "step": 54945 + }, + { + "epoch": 7.2, + "learning_rate": 5.231004884051782e-07, + "loss": 0.0006, + "step": 54946 + }, + { + "epoch": 7.2, + "learning_rate": 5.229312260279351e-07, + "loss": 0.0008, + "step": 54947 + }, + { + "epoch": 7.2, + "learning_rate": 5.227619903044424e-07, + "loss": 0.0003, + "step": 54948 + }, + { + "epoch": 7.2, + "learning_rate": 5.225927812351772e-07, + "loss": 0.0004, + "step": 54949 + }, + { + "epoch": 7.2, + "learning_rate": 5.224235988206161e-07, + "loss": 0.0006, + "step": 54950 + }, + { + "epoch": 7.2, + "learning_rate": 5.22254443061232e-07, + "loss": 0.0005, + "step": 54951 + }, + { + "epoch": 7.2, + "learning_rate": 5.220853139575033e-07, + "loss": 0.0003, + "step": 54952 + }, + { + "epoch": 7.2, + "learning_rate": 5.219162115099042e-07, + "loss": 0.0015, + "step": 54953 + }, + { + "epoch": 7.2, + "learning_rate": 5.21747135718913e-07, + "loss": 0.0005, + "step": 54954 + }, + { + "epoch": 7.2, + "learning_rate": 5.215780865850018e-07, + "loss": 0.0005, + "step": 54955 + }, + { + "epoch": 7.2, + "learning_rate": 5.214090641086466e-07, + "loss": 0.0005, + "step": 54956 + }, + { + "epoch": 7.2, + "learning_rate": 5.212400682903262e-07, + "loss": 0.0003, + "step": 54957 + }, + { + "epoch": 7.2, + "learning_rate": 5.210710991305124e-07, + "loss": 0.0004, + "step": 54958 + }, + { + "epoch": 7.2, + "learning_rate": 5.209021566296801e-07, + "loss": 0.0008, + "step": 54959 + }, + { + "epoch": 7.2, + "learning_rate": 5.20733240788307e-07, + "loss": 0.0008, + "step": 54960 + }, + { + "epoch": 7.2, + "learning_rate": 5.20564351606867e-07, + "loss": 0.0004, + "step": 54961 + }, + { + "epoch": 7.2, + "learning_rate": 5.203954890858343e-07, + "loss": 0.0003, + "step": 54962 + }, + { + "epoch": 7.2, + "learning_rate": 5.202266532256839e-07, + "loss": 0.0004, + "step": 54963 + }, + { + "epoch": 7.2, + "learning_rate": 5.20057844026891e-07, + "loss": 0.0008, + "step": 54964 + }, + { + "epoch": 7.2, + "learning_rate": 5.19889061489931e-07, + "loss": 0.0003, + "step": 54965 + }, + { + "epoch": 7.2, + "learning_rate": 5.1972030561528e-07, + "loss": 0.0003, + "step": 54966 + }, + { + "epoch": 7.2, + "learning_rate": 5.195515764034076e-07, + "loss": 0.0004, + "step": 54967 + }, + { + "epoch": 7.2, + "learning_rate": 5.193828738547934e-07, + "loss": 0.0002, + "step": 54968 + }, + { + "epoch": 7.2, + "learning_rate": 5.192141979699106e-07, + "loss": 0.0007, + "step": 54969 + }, + { + "epoch": 7.2, + "learning_rate": 5.190455487492329e-07, + "loss": 0.0005, + "step": 54970 + }, + { + "epoch": 7.2, + "learning_rate": 5.188769261932336e-07, + "loss": 0.0005, + "step": 54971 + }, + { + "epoch": 7.2, + "learning_rate": 5.187083303023888e-07, + "loss": 0.0006, + "step": 54972 + }, + { + "epoch": 7.2, + "learning_rate": 5.185397610771726e-07, + "loss": 0.0011, + "step": 54973 + }, + { + "epoch": 7.2, + "learning_rate": 5.183712185180556e-07, + "loss": 0.0003, + "step": 54974 + }, + { + "epoch": 7.2, + "learning_rate": 5.182027026255154e-07, + "loss": 0.0002, + "step": 54975 + }, + { + "epoch": 7.2, + "learning_rate": 5.180342134000272e-07, + "loss": 0.0006, + "step": 54976 + }, + { + "epoch": 7.2, + "learning_rate": 5.178657508420626e-07, + "loss": 0.0002, + "step": 54977 + }, + { + "epoch": 7.2, + "learning_rate": 5.176973149520937e-07, + "loss": 0.0003, + "step": 54978 + }, + { + "epoch": 7.2, + "learning_rate": 5.175289057305976e-07, + "loss": 0.0005, + "step": 54979 + }, + { + "epoch": 7.2, + "learning_rate": 5.173605231780465e-07, + "loss": 0.0006, + "step": 54980 + }, + { + "epoch": 7.2, + "learning_rate": 5.171921672949143e-07, + "loss": 0.0009, + "step": 54981 + }, + { + "epoch": 7.2, + "learning_rate": 5.170238380816728e-07, + "loss": 0.0006, + "step": 54982 + }, + { + "epoch": 7.2, + "learning_rate": 5.168555355387972e-07, + "loss": 0.0007, + "step": 54983 + }, + { + "epoch": 7.2, + "learning_rate": 5.166872596667616e-07, + "loss": 0.0009, + "step": 54984 + }, + { + "epoch": 7.2, + "learning_rate": 5.165190104660378e-07, + "loss": 0.0005, + "step": 54985 + }, + { + "epoch": 7.2, + "learning_rate": 5.163507879370955e-07, + "loss": 0.0006, + "step": 54986 + }, + { + "epoch": 7.2, + "learning_rate": 5.161825920804153e-07, + "loss": 0.0006, + "step": 54987 + }, + { + "epoch": 7.2, + "learning_rate": 5.160144228964658e-07, + "loss": 0.0005, + "step": 54988 + }, + { + "epoch": 7.2, + "learning_rate": 5.158462803857222e-07, + "loss": 0.0004, + "step": 54989 + }, + { + "epoch": 7.2, + "learning_rate": 5.156781645486531e-07, + "loss": 0.0004, + "step": 54990 + }, + { + "epoch": 7.2, + "learning_rate": 5.155100753857356e-07, + "loss": 0.0005, + "step": 54991 + }, + { + "epoch": 7.2, + "learning_rate": 5.153420128974418e-07, + "loss": 0.0013, + "step": 54992 + }, + { + "epoch": 7.2, + "learning_rate": 5.151739770842412e-07, + "loss": 0.0004, + "step": 54993 + }, + { + "epoch": 7.2, + "learning_rate": 5.150059679466101e-07, + "loss": 0.0008, + "step": 54994 + }, + { + "epoch": 7.2, + "learning_rate": 5.148379854850206e-07, + "loss": 0.0003, + "step": 54995 + }, + { + "epoch": 7.2, + "learning_rate": 5.14670029699943e-07, + "loss": 0.0003, + "step": 54996 + }, + { + "epoch": 7.2, + "learning_rate": 5.145021005918493e-07, + "loss": 0.0004, + "step": 54997 + }, + { + "epoch": 7.2, + "learning_rate": 5.143341981612149e-07, + "loss": 0.0005, + "step": 54998 + }, + { + "epoch": 7.2, + "learning_rate": 5.141663224085102e-07, + "loss": 0.0004, + "step": 54999 + }, + { + "epoch": 7.2, + "learning_rate": 5.139984733342085e-07, + "loss": 0.0006, + "step": 55000 + }, + { + "epoch": 7.2, + "learning_rate": 5.138306509387792e-07, + "loss": 0.0004, + "step": 55001 + }, + { + "epoch": 7.2, + "learning_rate": 5.136628552226974e-07, + "loss": 0.0007, + "step": 55002 + }, + { + "epoch": 7.21, + "learning_rate": 5.134950861864329e-07, + "loss": 0.0003, + "step": 55003 + }, + { + "epoch": 7.21, + "learning_rate": 5.133273438304598e-07, + "loss": 0.0003, + "step": 55004 + }, + { + "epoch": 7.21, + "learning_rate": 5.131596281552464e-07, + "loss": 0.0005, + "step": 55005 + }, + { + "epoch": 7.21, + "learning_rate": 5.12991939161267e-07, + "loss": 0.0004, + "step": 55006 + }, + { + "epoch": 7.21, + "learning_rate": 5.128242768489933e-07, + "loss": 0.0005, + "step": 55007 + }, + { + "epoch": 7.21, + "learning_rate": 5.126566412188949e-07, + "loss": 0.0004, + "step": 55008 + }, + { + "epoch": 7.21, + "learning_rate": 5.124890322714438e-07, + "loss": 0.0005, + "step": 55009 + }, + { + "epoch": 7.21, + "learning_rate": 5.123214500071138e-07, + "loss": 0.0003, + "step": 55010 + }, + { + "epoch": 7.21, + "learning_rate": 5.121538944263737e-07, + "loss": 0.0004, + "step": 55011 + }, + { + "epoch": 7.21, + "learning_rate": 5.119863655296964e-07, + "loss": 0.0003, + "step": 55012 + }, + { + "epoch": 7.21, + "learning_rate": 5.118188633175514e-07, + "loss": 0.0006, + "step": 55013 + }, + { + "epoch": 7.21, + "learning_rate": 5.116513877904105e-07, + "loss": 0.0012, + "step": 55014 + }, + { + "epoch": 7.21, + "learning_rate": 5.114839389487458e-07, + "loss": 0.0011, + "step": 55015 + }, + { + "epoch": 7.21, + "learning_rate": 5.113165167930257e-07, + "loss": 0.0003, + "step": 55016 + }, + { + "epoch": 7.21, + "learning_rate": 5.111491213237229e-07, + "loss": 0.0006, + "step": 55017 + }, + { + "epoch": 7.21, + "learning_rate": 5.109817525413097e-07, + "loss": 0.0006, + "step": 55018 + }, + { + "epoch": 7.21, + "learning_rate": 5.10814410446252e-07, + "loss": 0.0007, + "step": 55019 + }, + { + "epoch": 7.21, + "learning_rate": 5.106470950390263e-07, + "loss": 0.0007, + "step": 55020 + }, + { + "epoch": 7.21, + "learning_rate": 5.104798063200977e-07, + "loss": 0.0005, + "step": 55021 + }, + { + "epoch": 7.21, + "learning_rate": 5.103125442899415e-07, + "loss": 0.0004, + "step": 55022 + }, + { + "epoch": 7.21, + "learning_rate": 5.10145308949025e-07, + "loss": 0.0004, + "step": 55023 + }, + { + "epoch": 7.21, + "learning_rate": 5.099781002978188e-07, + "loss": 0.0003, + "step": 55024 + }, + { + "epoch": 7.21, + "learning_rate": 5.09810918336795e-07, + "loss": 0.0009, + "step": 55025 + }, + { + "epoch": 7.21, + "learning_rate": 5.09643763066422e-07, + "loss": 0.0013, + "step": 55026 + }, + { + "epoch": 7.21, + "learning_rate": 5.094766344871715e-07, + "loss": 0.0006, + "step": 55027 + }, + { + "epoch": 7.21, + "learning_rate": 5.093095325995112e-07, + "loss": 0.0002, + "step": 55028 + }, + { + "epoch": 7.21, + "learning_rate": 5.091424574039127e-07, + "loss": 0.0003, + "step": 55029 + }, + { + "epoch": 7.21, + "learning_rate": 5.089754089008459e-07, + "loss": 0.0002, + "step": 55030 + }, + { + "epoch": 7.21, + "learning_rate": 5.088083870907801e-07, + "loss": 0.0002, + "step": 55031 + }, + { + "epoch": 7.21, + "learning_rate": 5.086413919741851e-07, + "loss": 0.0004, + "step": 55032 + }, + { + "epoch": 7.21, + "learning_rate": 5.084744235515316e-07, + "loss": 0.0005, + "step": 55033 + }, + { + "epoch": 7.21, + "learning_rate": 5.083074818232881e-07, + "loss": 0.0002, + "step": 55034 + }, + { + "epoch": 7.21, + "learning_rate": 5.081405667899242e-07, + "loss": 0.0003, + "step": 55035 + }, + { + "epoch": 7.21, + "learning_rate": 5.079736784519085e-07, + "loss": 0.0009, + "step": 55036 + }, + { + "epoch": 7.21, + "learning_rate": 5.078068168097128e-07, + "loss": 0.0004, + "step": 55037 + }, + { + "epoch": 7.21, + "learning_rate": 5.076399818638045e-07, + "loss": 0.0004, + "step": 55038 + }, + { + "epoch": 7.21, + "learning_rate": 5.074731736146532e-07, + "loss": 0.0006, + "step": 55039 + }, + { + "epoch": 7.21, + "learning_rate": 5.073063920627274e-07, + "loss": 0.0002, + "step": 55040 + }, + { + "epoch": 7.21, + "learning_rate": 5.071396372084969e-07, + "loss": 0.0003, + "step": 55041 + }, + { + "epoch": 7.21, + "learning_rate": 5.069729090524322e-07, + "loss": 0.0007, + "step": 55042 + }, + { + "epoch": 7.21, + "learning_rate": 5.068062075949986e-07, + "loss": 0.0002, + "step": 55043 + }, + { + "epoch": 7.21, + "learning_rate": 5.06639532836668e-07, + "loss": 0.0011, + "step": 55044 + }, + { + "epoch": 7.21, + "learning_rate": 5.064728847779087e-07, + "loss": 0.0006, + "step": 55045 + }, + { + "epoch": 7.21, + "learning_rate": 5.063062634191895e-07, + "loss": 0.0006, + "step": 55046 + }, + { + "epoch": 7.21, + "learning_rate": 5.061396687609765e-07, + "loss": 0.0006, + "step": 55047 + }, + { + "epoch": 7.21, + "learning_rate": 5.059731008037417e-07, + "loss": 0.0004, + "step": 55048 + }, + { + "epoch": 7.21, + "learning_rate": 5.058065595479511e-07, + "loss": 0.0004, + "step": 55049 + }, + { + "epoch": 7.21, + "learning_rate": 5.056400449940735e-07, + "loss": 0.0005, + "step": 55050 + }, + { + "epoch": 7.21, + "learning_rate": 5.054735571425762e-07, + "loss": 0.0004, + "step": 55051 + }, + { + "epoch": 7.21, + "learning_rate": 5.053070959939321e-07, + "loss": 0.0003, + "step": 55052 + }, + { + "epoch": 7.21, + "learning_rate": 5.051406615486054e-07, + "loss": 0.0001, + "step": 55053 + }, + { + "epoch": 7.21, + "learning_rate": 5.049742538070646e-07, + "loss": 0.0006, + "step": 55054 + }, + { + "epoch": 7.21, + "learning_rate": 5.04807872769777e-07, + "loss": 0.0003, + "step": 55055 + }, + { + "epoch": 7.21, + "learning_rate": 5.046415184372134e-07, + "loss": 0.0012, + "step": 55056 + }, + { + "epoch": 7.21, + "learning_rate": 5.044751908098389e-07, + "loss": 0.0006, + "step": 55057 + }, + { + "epoch": 7.21, + "learning_rate": 5.043088898881221e-07, + "loss": 0.002, + "step": 55058 + }, + { + "epoch": 7.21, + "learning_rate": 5.041426156725315e-07, + "loss": 0.0003, + "step": 55059 + }, + { + "epoch": 7.21, + "learning_rate": 5.039763681635345e-07, + "loss": 0.0006, + "step": 55060 + }, + { + "epoch": 7.21, + "learning_rate": 5.038101473615975e-07, + "loss": 0.0003, + "step": 55061 + }, + { + "epoch": 7.21, + "learning_rate": 5.036439532671878e-07, + "loss": 0.0003, + "step": 55062 + }, + { + "epoch": 7.21, + "learning_rate": 5.034777858807738e-07, + "loss": 0.0006, + "step": 55063 + }, + { + "epoch": 7.21, + "learning_rate": 5.033116452028252e-07, + "loss": 0.0007, + "step": 55064 + }, + { + "epoch": 7.21, + "learning_rate": 5.031455312338051e-07, + "loss": 0.0005, + "step": 55065 + }, + { + "epoch": 7.21, + "learning_rate": 5.02979443974182e-07, + "loss": 0.0003, + "step": 55066 + }, + { + "epoch": 7.21, + "learning_rate": 5.028133834244253e-07, + "loss": 0.0002, + "step": 55067 + }, + { + "epoch": 7.21, + "learning_rate": 5.026473495849993e-07, + "loss": 0.0002, + "step": 55068 + }, + { + "epoch": 7.21, + "learning_rate": 5.024813424563724e-07, + "loss": 0.0011, + "step": 55069 + }, + { + "epoch": 7.21, + "learning_rate": 5.023153620390098e-07, + "loss": 0.0006, + "step": 55070 + }, + { + "epoch": 7.21, + "learning_rate": 5.0214940833338e-07, + "loss": 0.0006, + "step": 55071 + }, + { + "epoch": 7.21, + "learning_rate": 5.019834813399493e-07, + "loss": 0.0005, + "step": 55072 + }, + { + "epoch": 7.21, + "learning_rate": 5.018175810591841e-07, + "loss": 0.0003, + "step": 55073 + }, + { + "epoch": 7.21, + "learning_rate": 5.016517074915495e-07, + "loss": 0.0003, + "step": 55074 + }, + { + "epoch": 7.21, + "learning_rate": 5.014858606375161e-07, + "loss": 0.0021, + "step": 55075 + }, + { + "epoch": 7.21, + "learning_rate": 5.013200404975471e-07, + "loss": 0.0005, + "step": 55076 + }, + { + "epoch": 7.21, + "learning_rate": 5.011542470721109e-07, + "loss": 0.0005, + "step": 55077 + }, + { + "epoch": 7.21, + "learning_rate": 5.009884803616704e-07, + "loss": 0.0003, + "step": 55078 + }, + { + "epoch": 7.21, + "learning_rate": 5.008227403666954e-07, + "loss": 0.0004, + "step": 55079 + }, + { + "epoch": 7.22, + "learning_rate": 5.006570270876498e-07, + "loss": 0.0018, + "step": 55080 + }, + { + "epoch": 7.22, + "learning_rate": 5.004913405250001e-07, + "loss": 0.0002, + "step": 55081 + }, + { + "epoch": 7.22, + "learning_rate": 5.003256806792134e-07, + "loss": 0.0009, + "step": 55082 + }, + { + "epoch": 7.22, + "learning_rate": 5.001600475507551e-07, + "loss": 0.0008, + "step": 55083 + }, + { + "epoch": 7.22, + "learning_rate": 4.999944411400892e-07, + "loss": 0.0005, + "step": 55084 + }, + { + "epoch": 7.22, + "learning_rate": 4.998288614476853e-07, + "loss": 0.0005, + "step": 55085 + }, + { + "epoch": 7.22, + "learning_rate": 4.996633084740043e-07, + "loss": 0.0001, + "step": 55086 + }, + { + "epoch": 7.22, + "learning_rate": 4.994977822195168e-07, + "loss": 0.0006, + "step": 55087 + }, + { + "epoch": 7.22, + "learning_rate": 4.993322826846847e-07, + "loss": 0.0006, + "step": 55088 + }, + { + "epoch": 7.22, + "learning_rate": 4.991668098699742e-07, + "loss": 0.0005, + "step": 55089 + }, + { + "epoch": 7.22, + "learning_rate": 4.990013637758518e-07, + "loss": 0.0003, + "step": 55090 + }, + { + "epoch": 7.22, + "learning_rate": 4.988359444027824e-07, + "loss": 0.0004, + "step": 55091 + }, + { + "epoch": 7.22, + "learning_rate": 4.986705517512302e-07, + "loss": 0.0005, + "step": 55092 + }, + { + "epoch": 7.22, + "learning_rate": 4.985051858216605e-07, + "loss": 0.0004, + "step": 55093 + }, + { + "epoch": 7.22, + "learning_rate": 4.983398466145406e-07, + "loss": 0.0005, + "step": 55094 + }, + { + "epoch": 7.22, + "learning_rate": 4.981745341303323e-07, + "loss": 0.0004, + "step": 55095 + }, + { + "epoch": 7.22, + "learning_rate": 4.980092483695031e-07, + "loss": 0.0006, + "step": 55096 + }, + { + "epoch": 7.22, + "learning_rate": 4.978439893325159e-07, + "loss": 0.0003, + "step": 55097 + }, + { + "epoch": 7.22, + "learning_rate": 4.976787570198371e-07, + "loss": 0.0006, + "step": 55098 + }, + { + "epoch": 7.22, + "learning_rate": 4.975135514319318e-07, + "loss": 0.0002, + "step": 55099 + }, + { + "epoch": 7.22, + "learning_rate": 4.973483725692618e-07, + "loss": 0.0007, + "step": 55100 + }, + { + "epoch": 7.22, + "learning_rate": 4.971832204322946e-07, + "loss": 0.0019, + "step": 55101 + }, + { + "epoch": 7.22, + "learning_rate": 4.970180950214943e-07, + "loss": 0.0005, + "step": 55102 + }, + { + "epoch": 7.22, + "learning_rate": 4.968529963373237e-07, + "loss": 0.0005, + "step": 55103 + }, + { + "epoch": 7.22, + "learning_rate": 4.966879243802469e-07, + "loss": 0.0008, + "step": 55104 + }, + { + "epoch": 7.22, + "learning_rate": 4.965228791507304e-07, + "loss": 0.0006, + "step": 55105 + }, + { + "epoch": 7.22, + "learning_rate": 4.963578606492359e-07, + "loss": 0.0004, + "step": 55106 + }, + { + "epoch": 7.22, + "learning_rate": 4.961928688762308e-07, + "loss": 0.0005, + "step": 55107 + }, + { + "epoch": 7.22, + "learning_rate": 4.960279038321747e-07, + "loss": 0.0003, + "step": 55108 + }, + { + "epoch": 7.22, + "learning_rate": 4.958629655175362e-07, + "loss": 0.0002, + "step": 55109 + }, + { + "epoch": 7.22, + "learning_rate": 4.956980539327771e-07, + "loss": 0.0004, + "step": 55110 + }, + { + "epoch": 7.22, + "learning_rate": 4.955331690783605e-07, + "loss": 0.0014, + "step": 55111 + }, + { + "epoch": 7.22, + "learning_rate": 4.953683109547502e-07, + "loss": 0.002, + "step": 55112 + }, + { + "epoch": 7.22, + "learning_rate": 4.952034795624106e-07, + "loss": 0.0001, + "step": 55113 + }, + { + "epoch": 7.22, + "learning_rate": 4.950386749018055e-07, + "loss": 0.0004, + "step": 55114 + }, + { + "epoch": 7.22, + "learning_rate": 4.948738969733968e-07, + "loss": 0.0007, + "step": 55115 + }, + { + "epoch": 7.22, + "learning_rate": 4.947091457776509e-07, + "loss": 0.0003, + "step": 55116 + }, + { + "epoch": 7.22, + "learning_rate": 4.945444213150274e-07, + "loss": 0.0003, + "step": 55117 + }, + { + "epoch": 7.22, + "learning_rate": 4.943797235859927e-07, + "loss": 0.0002, + "step": 55118 + }, + { + "epoch": 7.22, + "learning_rate": 4.942150525910094e-07, + "loss": 0.0021, + "step": 55119 + }, + { + "epoch": 7.22, + "learning_rate": 4.940504083305375e-07, + "loss": 0.0009, + "step": 55120 + }, + { + "epoch": 7.22, + "learning_rate": 4.938857908050454e-07, + "loss": 0.0009, + "step": 55121 + }, + { + "epoch": 7.22, + "learning_rate": 4.937212000149927e-07, + "loss": 0.0003, + "step": 55122 + }, + { + "epoch": 7.22, + "learning_rate": 4.935566359608412e-07, + "loss": 0.0008, + "step": 55123 + }, + { + "epoch": 7.22, + "learning_rate": 4.933920986430574e-07, + "loss": 0.0004, + "step": 55124 + }, + { + "epoch": 7.22, + "learning_rate": 4.932275880621018e-07, + "loss": 0.0006, + "step": 55125 + }, + { + "epoch": 7.22, + "learning_rate": 4.930631042184364e-07, + "loss": 0.0004, + "step": 55126 + }, + { + "epoch": 7.22, + "learning_rate": 4.928986471125252e-07, + "loss": 0.0002, + "step": 55127 + }, + { + "epoch": 7.22, + "learning_rate": 4.92734216744829e-07, + "loss": 0.0004, + "step": 55128 + }, + { + "epoch": 7.22, + "learning_rate": 4.925698131158129e-07, + "loss": 0.0002, + "step": 55129 + }, + { + "epoch": 7.22, + "learning_rate": 4.924054362259389e-07, + "loss": 0.0001, + "step": 55130 + }, + { + "epoch": 7.22, + "learning_rate": 4.922410860756654e-07, + "loss": 0.0005, + "step": 55131 + }, + { + "epoch": 7.22, + "learning_rate": 4.920767626654599e-07, + "loss": 0.0004, + "step": 55132 + }, + { + "epoch": 7.22, + "learning_rate": 4.919124659957819e-07, + "loss": 0.0007, + "step": 55133 + }, + { + "epoch": 7.22, + "learning_rate": 4.917481960670944e-07, + "loss": 0.0006, + "step": 55134 + }, + { + "epoch": 7.22, + "learning_rate": 4.915839528798572e-07, + "loss": 0.0001, + "step": 55135 + }, + { + "epoch": 7.22, + "learning_rate": 4.914197364345341e-07, + "loss": 0.0004, + "step": 55136 + }, + { + "epoch": 7.22, + "learning_rate": 4.912555467315872e-07, + "loss": 0.0004, + "step": 55137 + }, + { + "epoch": 7.22, + "learning_rate": 4.91091383771477e-07, + "loss": 0.0005, + "step": 55138 + }, + { + "epoch": 7.22, + "learning_rate": 4.909272475546656e-07, + "loss": 0.0011, + "step": 55139 + }, + { + "epoch": 7.22, + "learning_rate": 4.907631380816158e-07, + "loss": 0.0002, + "step": 55140 + }, + { + "epoch": 7.22, + "learning_rate": 4.905990553527883e-07, + "loss": 0.0003, + "step": 55141 + }, + { + "epoch": 7.22, + "learning_rate": 4.904349993686453e-07, + "loss": 0.0003, + "step": 55142 + }, + { + "epoch": 7.22, + "learning_rate": 4.902709701296449e-07, + "loss": 0.0005, + "step": 55143 + }, + { + "epoch": 7.22, + "learning_rate": 4.901069676362535e-07, + "loss": 0.001, + "step": 55144 + }, + { + "epoch": 7.22, + "learning_rate": 4.899429918889298e-07, + "loss": 0.0004, + "step": 55145 + }, + { + "epoch": 7.22, + "learning_rate": 4.897790428881322e-07, + "loss": 0.0008, + "step": 55146 + }, + { + "epoch": 7.22, + "learning_rate": 4.896151206343269e-07, + "loss": 0.0005, + "step": 55147 + }, + { + "epoch": 7.22, + "learning_rate": 4.894512251279737e-07, + "loss": 0.0004, + "step": 55148 + }, + { + "epoch": 7.22, + "learning_rate": 4.89287356369531e-07, + "loss": 0.0004, + "step": 55149 + }, + { + "epoch": 7.22, + "learning_rate": 4.891235143594586e-07, + "loss": 0.0023, + "step": 55150 + }, + { + "epoch": 7.22, + "learning_rate": 4.889596990982226e-07, + "loss": 0.0006, + "step": 55151 + }, + { + "epoch": 7.22, + "learning_rate": 4.887959105862817e-07, + "loss": 0.0011, + "step": 55152 + }, + { + "epoch": 7.22, + "learning_rate": 4.886321488240953e-07, + "loss": 0.0003, + "step": 55153 + }, + { + "epoch": 7.22, + "learning_rate": 4.884684138121232e-07, + "loss": 0.0006, + "step": 55154 + }, + { + "epoch": 7.22, + "learning_rate": 4.883047055508284e-07, + "loss": 0.0009, + "step": 55155 + }, + { + "epoch": 7.23, + "learning_rate": 4.881410240406692e-07, + "loss": 0.0006, + "step": 55156 + }, + { + "epoch": 7.23, + "learning_rate": 4.879773692821077e-07, + "loss": 0.0008, + "step": 55157 + }, + { + "epoch": 7.23, + "learning_rate": 4.878137412756023e-07, + "loss": 0.0003, + "step": 55158 + }, + { + "epoch": 7.23, + "learning_rate": 4.876501400216149e-07, + "loss": 0.0002, + "step": 55159 + }, + { + "epoch": 7.23, + "learning_rate": 4.874865655206052e-07, + "loss": 0.0004, + "step": 55160 + }, + { + "epoch": 7.23, + "learning_rate": 4.873230177730326e-07, + "loss": 0.0004, + "step": 55161 + }, + { + "epoch": 7.23, + "learning_rate": 4.871594967793559e-07, + "loss": 0.0005, + "step": 55162 + }, + { + "epoch": 7.23, + "learning_rate": 4.869960025400388e-07, + "loss": 0.0008, + "step": 55163 + }, + { + "epoch": 7.23, + "learning_rate": 4.868325350555381e-07, + "loss": 0.0005, + "step": 55164 + }, + { + "epoch": 7.23, + "learning_rate": 4.86669094326313e-07, + "loss": 0.0003, + "step": 55165 + }, + { + "epoch": 7.23, + "learning_rate": 4.865056803528268e-07, + "loss": 0.0003, + "step": 55166 + }, + { + "epoch": 7.23, + "learning_rate": 4.863422931355354e-07, + "loss": 0.0007, + "step": 55167 + }, + { + "epoch": 7.23, + "learning_rate": 4.861789326749012e-07, + "loss": 0.0002, + "step": 55168 + }, + { + "epoch": 7.23, + "learning_rate": 4.8601559897138e-07, + "loss": 0.0008, + "step": 55169 + }, + { + "epoch": 7.23, + "learning_rate": 4.858522920254338e-07, + "loss": 0.0002, + "step": 55170 + }, + { + "epoch": 7.23, + "learning_rate": 4.856890118375224e-07, + "loss": 0.0002, + "step": 55171 + }, + { + "epoch": 7.23, + "learning_rate": 4.855257584081018e-07, + "loss": 0.0002, + "step": 55172 + }, + { + "epoch": 7.23, + "learning_rate": 4.853625317376343e-07, + "loss": 0.0004, + "step": 55173 + }, + { + "epoch": 7.23, + "learning_rate": 4.85199331826579e-07, + "loss": 0.0005, + "step": 55174 + }, + { + "epoch": 7.23, + "learning_rate": 4.850361586753938e-07, + "loss": 0.0006, + "step": 55175 + }, + { + "epoch": 7.23, + "learning_rate": 4.84873012284538e-07, + "loss": 0.0007, + "step": 55176 + }, + { + "epoch": 7.23, + "learning_rate": 4.84709892654468e-07, + "loss": 0.0004, + "step": 55177 + }, + { + "epoch": 7.23, + "learning_rate": 4.845467997856468e-07, + "loss": 0.0009, + "step": 55178 + }, + { + "epoch": 7.23, + "learning_rate": 4.843837336785306e-07, + "loss": 0.0007, + "step": 55179 + }, + { + "epoch": 7.23, + "learning_rate": 4.84220694333577e-07, + "loss": 0.0003, + "step": 55180 + }, + { + "epoch": 7.23, + "learning_rate": 4.840576817512477e-07, + "loss": 0.0004, + "step": 55181 + }, + { + "epoch": 7.23, + "learning_rate": 4.838946959319979e-07, + "loss": 0.0006, + "step": 55182 + }, + { + "epoch": 7.23, + "learning_rate": 4.837317368762895e-07, + "loss": 0.0006, + "step": 55183 + }, + { + "epoch": 7.23, + "learning_rate": 4.835688045845777e-07, + "loss": 0.0005, + "step": 55184 + }, + { + "epoch": 7.23, + "learning_rate": 4.83405899057321e-07, + "loss": 0.001, + "step": 55185 + }, + { + "epoch": 7.23, + "learning_rate": 4.8324302029498e-07, + "loss": 0.0005, + "step": 55186 + }, + { + "epoch": 7.23, + "learning_rate": 4.830801682980102e-07, + "loss": 0.0003, + "step": 55187 + }, + { + "epoch": 7.23, + "learning_rate": 4.82917343066871e-07, + "loss": 0.0005, + "step": 55188 + }, + { + "epoch": 7.23, + "learning_rate": 4.827545446020199e-07, + "loss": 0.001, + "step": 55189 + }, + { + "epoch": 7.23, + "learning_rate": 4.825917729039154e-07, + "loss": 0.0005, + "step": 55190 + }, + { + "epoch": 7.23, + "learning_rate": 4.824290279730148e-07, + "loss": 0.0006, + "step": 55191 + }, + { + "epoch": 7.23, + "learning_rate": 4.822663098097736e-07, + "loss": 0.0007, + "step": 55192 + }, + { + "epoch": 7.23, + "learning_rate": 4.821036184146522e-07, + "loss": 0.0003, + "step": 55193 + }, + { + "epoch": 7.23, + "learning_rate": 4.819409537881092e-07, + "loss": 0.0004, + "step": 55194 + }, + { + "epoch": 7.23, + "learning_rate": 4.817783159305999e-07, + "loss": 0.0003, + "step": 55195 + }, + { + "epoch": 7.23, + "learning_rate": 4.816157048425807e-07, + "loss": 0.0005, + "step": 55196 + }, + { + "epoch": 7.23, + "learning_rate": 4.814531205245121e-07, + "loss": 0.0003, + "step": 55197 + }, + { + "epoch": 7.23, + "learning_rate": 4.812905629768494e-07, + "loss": 0.0015, + "step": 55198 + }, + { + "epoch": 7.23, + "learning_rate": 4.8112803220005e-07, + "loss": 0.0006, + "step": 55199 + }, + { + "epoch": 7.23, + "learning_rate": 4.809655281945702e-07, + "loss": 0.0002, + "step": 55200 + }, + { + "epoch": 7.23, + "learning_rate": 4.808030509608697e-07, + "loss": 0.0011, + "step": 55201 + }, + { + "epoch": 7.23, + "learning_rate": 4.806406004994025e-07, + "loss": 0.0008, + "step": 55202 + }, + { + "epoch": 7.23, + "learning_rate": 4.804781768106259e-07, + "loss": 0.0006, + "step": 55203 + }, + { + "epoch": 7.23, + "learning_rate": 4.803157798949976e-07, + "loss": 0.0008, + "step": 55204 + }, + { + "epoch": 7.23, + "learning_rate": 4.801534097529758e-07, + "loss": 0.0006, + "step": 55205 + }, + { + "epoch": 7.23, + "learning_rate": 4.79991066385015e-07, + "loss": 0.0004, + "step": 55206 + }, + { + "epoch": 7.23, + "learning_rate": 4.798287497915732e-07, + "loss": 0.0008, + "step": 55207 + }, + { + "epoch": 7.23, + "learning_rate": 4.796664599731038e-07, + "loss": 0.0003, + "step": 55208 + }, + { + "epoch": 7.23, + "learning_rate": 4.795041969300674e-07, + "loss": 0.0002, + "step": 55209 + }, + { + "epoch": 7.23, + "learning_rate": 4.793419606629179e-07, + "loss": 0.0007, + "step": 55210 + }, + { + "epoch": 7.23, + "learning_rate": 4.791797511721108e-07, + "loss": 0.0003, + "step": 55211 + }, + { + "epoch": 7.23, + "learning_rate": 4.790175684581056e-07, + "loss": 0.0007, + "step": 55212 + }, + { + "epoch": 7.23, + "learning_rate": 4.788554125213562e-07, + "loss": 0.0005, + "step": 55213 + }, + { + "epoch": 7.23, + "learning_rate": 4.786932833623192e-07, + "loss": 0.0005, + "step": 55214 + }, + { + "epoch": 7.23, + "learning_rate": 4.785311809814474e-07, + "loss": 0.0004, + "step": 55215 + }, + { + "epoch": 7.23, + "learning_rate": 4.783691053792017e-07, + "loss": 0.0003, + "step": 55216 + }, + { + "epoch": 7.23, + "learning_rate": 4.78207056556037e-07, + "loss": 0.0003, + "step": 55217 + }, + { + "epoch": 7.23, + "learning_rate": 4.780450345124066e-07, + "loss": 0.001, + "step": 55218 + }, + { + "epoch": 7.23, + "learning_rate": 4.778830392487666e-07, + "loss": 0.0005, + "step": 55219 + }, + { + "epoch": 7.23, + "learning_rate": 4.777210707655755e-07, + "loss": 0.0005, + "step": 55220 + }, + { + "epoch": 7.23, + "learning_rate": 4.775591290632863e-07, + "loss": 0.0008, + "step": 55221 + }, + { + "epoch": 7.23, + "learning_rate": 4.773972141423544e-07, + "loss": 0.0003, + "step": 55222 + }, + { + "epoch": 7.23, + "learning_rate": 4.772353260032347e-07, + "loss": 0.0004, + "step": 55223 + }, + { + "epoch": 7.23, + "learning_rate": 4.770734646463859e-07, + "loss": 0.0006, + "step": 55224 + }, + { + "epoch": 7.23, + "learning_rate": 4.769116300722598e-07, + "loss": 0.0003, + "step": 55225 + }, + { + "epoch": 7.23, + "learning_rate": 4.767498222813116e-07, + "loss": 0.0002, + "step": 55226 + }, + { + "epoch": 7.23, + "learning_rate": 4.765880412739965e-07, + "loss": 0.0011, + "step": 55227 + }, + { + "epoch": 7.23, + "learning_rate": 4.7642628705077297e-07, + "loss": 0.0003, + "step": 55228 + }, + { + "epoch": 7.23, + "learning_rate": 4.762645596120929e-07, + "loss": 0.0009, + "step": 55229 + }, + { + "epoch": 7.23, + "learning_rate": 4.7610285895841045e-07, + "loss": 0.0003, + "step": 55230 + }, + { + "epoch": 7.23, + "learning_rate": 4.759411850901818e-07, + "loss": 0.0009, + "step": 55231 + }, + { + "epoch": 7.24, + "learning_rate": 4.7577953800786227e-07, + "loss": 0.0002, + "step": 55232 + }, + { + "epoch": 7.24, + "learning_rate": 4.756179177119058e-07, + "loss": 0.0007, + "step": 55233 + }, + { + "epoch": 7.24, + "learning_rate": 4.7545632420276545e-07, + "loss": 0.0008, + "step": 55234 + }, + { + "epoch": 7.24, + "learning_rate": 4.7529475748089746e-07, + "loss": 0.0003, + "step": 55235 + }, + { + "epoch": 7.24, + "learning_rate": 4.7513321754675714e-07, + "loss": 0.0005, + "step": 55236 + }, + { + "epoch": 7.24, + "learning_rate": 4.749717044007951e-07, + "loss": 0.0001, + "step": 55237 + }, + { + "epoch": 7.24, + "learning_rate": 4.748102180434677e-07, + "loss": 0.0003, + "step": 55238 + }, + { + "epoch": 7.24, + "learning_rate": 4.746487584752313e-07, + "loss": 0.0006, + "step": 55239 + }, + { + "epoch": 7.24, + "learning_rate": 4.744873256965377e-07, + "loss": 0.0004, + "step": 55240 + }, + { + "epoch": 7.24, + "learning_rate": 4.7432591970784093e-07, + "loss": 0.0006, + "step": 55241 + }, + { + "epoch": 7.24, + "learning_rate": 4.7416454050959404e-07, + "loss": 0.0006, + "step": 55242 + }, + { + "epoch": 7.24, + "learning_rate": 4.740031881022533e-07, + "loss": 0.0013, + "step": 55243 + }, + { + "epoch": 7.24, + "learning_rate": 4.738418624862717e-07, + "loss": 0.0004, + "step": 55244 + }, + { + "epoch": 7.24, + "learning_rate": 4.736805636621011e-07, + "loss": 0.0005, + "step": 55245 + }, + { + "epoch": 7.24, + "learning_rate": 4.7351929163019785e-07, + "loss": 0.001, + "step": 55246 + }, + { + "epoch": 7.24, + "learning_rate": 4.733580463910148e-07, + "loss": 0.001, + "step": 55247 + }, + { + "epoch": 7.24, + "learning_rate": 4.7319682794500274e-07, + "loss": 0.0003, + "step": 55248 + }, + { + "epoch": 7.24, + "learning_rate": 4.730356362926192e-07, + "loss": 0.0009, + "step": 55249 + }, + { + "epoch": 7.24, + "learning_rate": 4.7287447143431476e-07, + "loss": 0.0007, + "step": 55250 + }, + { + "epoch": 7.24, + "learning_rate": 4.727133333705436e-07, + "loss": 0.0005, + "step": 55251 + }, + { + "epoch": 7.24, + "learning_rate": 4.725522221017598e-07, + "loss": 0.0006, + "step": 55252 + }, + { + "epoch": 7.24, + "learning_rate": 4.7239113762841407e-07, + "loss": 0.0006, + "step": 55253 + }, + { + "epoch": 7.24, + "learning_rate": 4.7223007995096157e-07, + "loss": 0.0005, + "step": 55254 + }, + { + "epoch": 7.24, + "learning_rate": 4.720690490698554e-07, + "loss": 0.0008, + "step": 55255 + }, + { + "epoch": 7.24, + "learning_rate": 4.7190804498554734e-07, + "loss": 0.0005, + "step": 55256 + }, + { + "epoch": 7.24, + "learning_rate": 4.717470676984881e-07, + "loss": 0.0006, + "step": 55257 + }, + { + "epoch": 7.24, + "learning_rate": 4.715861172091352e-07, + "loss": 0.0005, + "step": 55258 + }, + { + "epoch": 7.24, + "learning_rate": 4.7142519351793816e-07, + "loss": 0.0007, + "step": 55259 + }, + { + "epoch": 7.24, + "learning_rate": 4.712642966253511e-07, + "loss": 0.0005, + "step": 55260 + }, + { + "epoch": 7.24, + "learning_rate": 4.7110342653182373e-07, + "loss": 0.0008, + "step": 55261 + }, + { + "epoch": 7.24, + "learning_rate": 4.709425832378123e-07, + "loss": 0.0002, + "step": 55262 + }, + { + "epoch": 7.24, + "learning_rate": 4.7078176674376755e-07, + "loss": 0.0004, + "step": 55263 + }, + { + "epoch": 7.24, + "learning_rate": 4.706209770501424e-07, + "loss": 0.0005, + "step": 55264 + }, + { + "epoch": 7.24, + "learning_rate": 4.704602141573855e-07, + "loss": 0.0006, + "step": 55265 + }, + { + "epoch": 7.24, + "learning_rate": 4.702994780659542e-07, + "loss": 0.0004, + "step": 55266 + }, + { + "epoch": 7.24, + "learning_rate": 4.7013876877629704e-07, + "loss": 0.0004, + "step": 55267 + }, + { + "epoch": 7.24, + "learning_rate": 4.69978086288867e-07, + "loss": 0.0006, + "step": 55268 + }, + { + "epoch": 7.24, + "learning_rate": 4.69817430604117e-07, + "loss": 0.0004, + "step": 55269 + }, + { + "epoch": 7.24, + "learning_rate": 4.6965680172249674e-07, + "loss": 0.0004, + "step": 55270 + }, + { + "epoch": 7.24, + "learning_rate": 4.6949619964446024e-07, + "loss": 0.0006, + "step": 55271 + }, + { + "epoch": 7.24, + "learning_rate": 4.6933562437045833e-07, + "loss": 0.0003, + "step": 55272 + }, + { + "epoch": 7.24, + "learning_rate": 4.6917507590094056e-07, + "loss": 0.0002, + "step": 55273 + }, + { + "epoch": 7.24, + "learning_rate": 4.690145542363622e-07, + "loss": 0.0001, + "step": 55274 + }, + { + "epoch": 7.24, + "learning_rate": 4.688540593771729e-07, + "loss": 0.0005, + "step": 55275 + }, + { + "epoch": 7.24, + "learning_rate": 4.686935913238222e-07, + "loss": 0.0007, + "step": 55276 + }, + { + "epoch": 7.24, + "learning_rate": 4.6853315007676534e-07, + "loss": 0.0004, + "step": 55277 + }, + { + "epoch": 7.24, + "learning_rate": 4.683727356364509e-07, + "loss": 0.001, + "step": 55278 + }, + { + "epoch": 7.24, + "learning_rate": 4.682123480033307e-07, + "loss": 0.001, + "step": 55279 + }, + { + "epoch": 7.24, + "learning_rate": 4.680519871778533e-07, + "loss": 0.0005, + "step": 55280 + }, + { + "epoch": 7.24, + "learning_rate": 4.678916531604738e-07, + "loss": 0.0003, + "step": 55281 + }, + { + "epoch": 7.24, + "learning_rate": 4.677313459516419e-07, + "loss": 0.0002, + "step": 55282 + }, + { + "epoch": 7.24, + "learning_rate": 4.6757106555180734e-07, + "loss": 0.0002, + "step": 55283 + }, + { + "epoch": 7.24, + "learning_rate": 4.674108119614207e-07, + "loss": 0.0003, + "step": 55284 + }, + { + "epoch": 7.24, + "learning_rate": 4.6725058518093504e-07, + "loss": 0.0008, + "step": 55285 + }, + { + "epoch": 7.24, + "learning_rate": 4.670903852108e-07, + "loss": 0.0008, + "step": 55286 + }, + { + "epoch": 7.24, + "learning_rate": 4.669302120514652e-07, + "loss": 0.0004, + "step": 55287 + }, + { + "epoch": 7.24, + "learning_rate": 4.6677006570338025e-07, + "loss": 0.0006, + "step": 55288 + }, + { + "epoch": 7.24, + "learning_rate": 4.666099461669971e-07, + "loss": 0.0005, + "step": 55289 + }, + { + "epoch": 7.24, + "learning_rate": 4.664498534427664e-07, + "loss": 0.0008, + "step": 55290 + }, + { + "epoch": 7.24, + "learning_rate": 4.662897875311367e-07, + "loss": 0.0007, + "step": 55291 + }, + { + "epoch": 7.24, + "learning_rate": 4.6612974843255997e-07, + "loss": 0.0006, + "step": 55292 + }, + { + "epoch": 7.24, + "learning_rate": 4.659697361474857e-07, + "loss": 0.0006, + "step": 55293 + }, + { + "epoch": 7.24, + "learning_rate": 4.658097506763648e-07, + "loss": 0.0004, + "step": 55294 + }, + { + "epoch": 7.24, + "learning_rate": 4.656497920196446e-07, + "loss": 0.0004, + "step": 55295 + }, + { + "epoch": 7.24, + "learning_rate": 4.6548986017777685e-07, + "loss": 0.0006, + "step": 55296 + }, + { + "epoch": 7.24, + "learning_rate": 4.653299551512125e-07, + "loss": 0.0003, + "step": 55297 + }, + { + "epoch": 7.24, + "learning_rate": 4.6517007694039995e-07, + "loss": 0.0003, + "step": 55298 + }, + { + "epoch": 7.24, + "learning_rate": 4.6501022554578666e-07, + "loss": 0.0006, + "step": 55299 + }, + { + "epoch": 7.24, + "learning_rate": 4.648504009678256e-07, + "loss": 0.0011, + "step": 55300 + }, + { + "epoch": 7.24, + "learning_rate": 4.6469060320696533e-07, + "loss": 0.0002, + "step": 55301 + }, + { + "epoch": 7.24, + "learning_rate": 4.6453083226365546e-07, + "loss": 0.0004, + "step": 55302 + }, + { + "epoch": 7.24, + "learning_rate": 4.643710881383423e-07, + "loss": 0.0008, + "step": 55303 + }, + { + "epoch": 7.24, + "learning_rate": 4.642113708314788e-07, + "loss": 0.0004, + "step": 55304 + }, + { + "epoch": 7.24, + "learning_rate": 4.6405168034351467e-07, + "loss": 0.0003, + "step": 55305 + }, + { + "epoch": 7.24, + "learning_rate": 4.638920166748961e-07, + "loss": 0.0005, + "step": 55306 + }, + { + "epoch": 7.24, + "learning_rate": 4.637323798260729e-07, + "loss": 0.0008, + "step": 55307 + }, + { + "epoch": 7.24, + "learning_rate": 4.6357276979749566e-07, + "loss": 0.0005, + "step": 55308 + }, + { + "epoch": 7.25, + "learning_rate": 4.6341318658961186e-07, + "loss": 0.0005, + "step": 55309 + }, + { + "epoch": 7.25, + "learning_rate": 4.6325363020287006e-07, + "loss": 0.0004, + "step": 55310 + }, + { + "epoch": 7.25, + "learning_rate": 4.630941006377199e-07, + "loss": 0.0007, + "step": 55311 + }, + { + "epoch": 7.25, + "learning_rate": 4.629345978946109e-07, + "loss": 0.0006, + "step": 55312 + }, + { + "epoch": 7.25, + "learning_rate": 4.627751219739884e-07, + "loss": 0.0004, + "step": 55313 + }, + { + "epoch": 7.25, + "learning_rate": 4.6261567287630425e-07, + "loss": 0.0006, + "step": 55314 + }, + { + "epoch": 7.25, + "learning_rate": 4.624562506020047e-07, + "loss": 0.0006, + "step": 55315 + }, + { + "epoch": 7.25, + "learning_rate": 4.622968551515394e-07, + "loss": 0.0002, + "step": 55316 + }, + { + "epoch": 7.25, + "learning_rate": 4.6213748652535696e-07, + "loss": 0.0005, + "step": 55317 + }, + { + "epoch": 7.25, + "learning_rate": 4.619781447239036e-07, + "loss": 0.0002, + "step": 55318 + }, + { + "epoch": 7.25, + "learning_rate": 4.6181882974763006e-07, + "loss": 0.0008, + "step": 55319 + }, + { + "epoch": 7.25, + "learning_rate": 4.616595415969838e-07, + "loss": 0.0002, + "step": 55320 + }, + { + "epoch": 7.25, + "learning_rate": 4.615002802724111e-07, + "loss": 0.0005, + "step": 55321 + }, + { + "epoch": 7.25, + "learning_rate": 4.6134104577436054e-07, + "loss": 0.0005, + "step": 55322 + }, + { + "epoch": 7.25, + "learning_rate": 4.6118183810328063e-07, + "loss": 0.0004, + "step": 55323 + }, + { + "epoch": 7.25, + "learning_rate": 4.610226572596177e-07, + "loss": 0.0008, + "step": 55324 + }, + { + "epoch": 7.25, + "learning_rate": 4.6086350324382247e-07, + "loss": 0.0006, + "step": 55325 + }, + { + "epoch": 7.25, + "learning_rate": 4.60704376056339e-07, + "loss": 0.0004, + "step": 55326 + }, + { + "epoch": 7.25, + "learning_rate": 4.6054527569761807e-07, + "loss": 0.0013, + "step": 55327 + }, + { + "epoch": 7.25, + "learning_rate": 4.603862021681049e-07, + "loss": 0.0003, + "step": 55328 + }, + { + "epoch": 7.25, + "learning_rate": 4.60227155468248e-07, + "loss": 0.0003, + "step": 55329 + }, + { + "epoch": 7.25, + "learning_rate": 4.600681355984926e-07, + "loss": 0.0004, + "step": 55330 + }, + { + "epoch": 7.25, + "learning_rate": 4.599091425592894e-07, + "loss": 0.0011, + "step": 55331 + }, + { + "epoch": 7.25, + "learning_rate": 4.5975017635108253e-07, + "loss": 0.0012, + "step": 55332 + }, + { + "epoch": 7.25, + "learning_rate": 4.595912369743205e-07, + "loss": 0.0013, + "step": 55333 + }, + { + "epoch": 7.25, + "learning_rate": 4.594323244294507e-07, + "loss": 0.0005, + "step": 55334 + }, + { + "epoch": 7.25, + "learning_rate": 4.5927343871691734e-07, + "loss": 0.0004, + "step": 55335 + }, + { + "epoch": 7.25, + "learning_rate": 4.5911457983717213e-07, + "loss": 0.0003, + "step": 55336 + }, + { + "epoch": 7.25, + "learning_rate": 4.589557477906581e-07, + "loss": 0.0004, + "step": 55337 + }, + { + "epoch": 7.25, + "learning_rate": 4.587969425778216e-07, + "loss": 0.0007, + "step": 55338 + }, + { + "epoch": 7.25, + "learning_rate": 4.5863816419911224e-07, + "loss": 0.0003, + "step": 55339 + }, + { + "epoch": 7.25, + "learning_rate": 4.584794126549752e-07, + "loss": 0.0004, + "step": 55340 + }, + { + "epoch": 7.25, + "learning_rate": 4.583206879458546e-07, + "loss": 0.0001, + "step": 55341 + }, + { + "epoch": 7.25, + "learning_rate": 4.5816199007220007e-07, + "loss": 0.0016, + "step": 55342 + }, + { + "epoch": 7.25, + "learning_rate": 4.5800331903445796e-07, + "loss": 0.0003, + "step": 55343 + }, + { + "epoch": 7.25, + "learning_rate": 4.5784467483307225e-07, + "loss": 0.0003, + "step": 55344 + }, + { + "epoch": 7.25, + "learning_rate": 4.576860574684905e-07, + "loss": 0.0012, + "step": 55345 + }, + { + "epoch": 7.25, + "learning_rate": 4.575274669411567e-07, + "loss": 0.0004, + "step": 55346 + }, + { + "epoch": 7.25, + "learning_rate": 4.5736890325152163e-07, + "loss": 0.0004, + "step": 55347 + }, + { + "epoch": 7.25, + "learning_rate": 4.5721036640002714e-07, + "loss": 0.0002, + "step": 55348 + }, + { + "epoch": 7.25, + "learning_rate": 4.5705185638711956e-07, + "loss": 0.0005, + "step": 55349 + }, + { + "epoch": 7.25, + "learning_rate": 4.5689337321324633e-07, + "loss": 0.0004, + "step": 55350 + }, + { + "epoch": 7.25, + "learning_rate": 4.567349168788526e-07, + "loss": 0.0005, + "step": 55351 + }, + { + "epoch": 7.25, + "learning_rate": 4.5657648738438365e-07, + "loss": 0.0002, + "step": 55352 + }, + { + "epoch": 7.25, + "learning_rate": 4.564180847302846e-07, + "loss": 0.0003, + "step": 55353 + }, + { + "epoch": 7.25, + "learning_rate": 4.5625970891700177e-07, + "loss": 0.0004, + "step": 55354 + }, + { + "epoch": 7.25, + "learning_rate": 4.561013599449804e-07, + "loss": 0.0002, + "step": 55355 + }, + { + "epoch": 7.25, + "learning_rate": 4.5594303781466453e-07, + "loss": 0.0003, + "step": 55356 + }, + { + "epoch": 7.25, + "learning_rate": 4.557847425265005e-07, + "loss": 0.0003, + "step": 55357 + }, + { + "epoch": 7.25, + "learning_rate": 4.556264740809357e-07, + "loss": 0.0003, + "step": 55358 + }, + { + "epoch": 7.25, + "learning_rate": 4.554682324784121e-07, + "loss": 0.0004, + "step": 55359 + }, + { + "epoch": 7.25, + "learning_rate": 4.553100177193748e-07, + "loss": 0.0004, + "step": 55360 + }, + { + "epoch": 7.25, + "learning_rate": 4.551518298042712e-07, + "loss": 0.0004, + "step": 55361 + }, + { + "epoch": 7.25, + "learning_rate": 4.5499366873354545e-07, + "loss": 0.0004, + "step": 55362 + }, + { + "epoch": 7.25, + "learning_rate": 4.548355345076405e-07, + "loss": 0.0003, + "step": 55363 + }, + { + "epoch": 7.25, + "learning_rate": 4.5467742712700156e-07, + "loss": 0.0002, + "step": 55364 + }, + { + "epoch": 7.25, + "learning_rate": 4.5451934659207606e-07, + "loss": 0.0003, + "step": 55365 + }, + { + "epoch": 7.25, + "learning_rate": 4.543612929033059e-07, + "loss": 0.0007, + "step": 55366 + }, + { + "epoch": 7.25, + "learning_rate": 4.542032660611351e-07, + "loss": 0.0005, + "step": 55367 + }, + { + "epoch": 7.25, + "learning_rate": 4.5404526606600885e-07, + "loss": 0.0004, + "step": 55368 + }, + { + "epoch": 7.25, + "learning_rate": 4.538872929183735e-07, + "loss": 0.0004, + "step": 55369 + }, + { + "epoch": 7.25, + "learning_rate": 4.5372934661867205e-07, + "loss": 0.0008, + "step": 55370 + }, + { + "epoch": 7.25, + "learning_rate": 4.535714271673475e-07, + "loss": 0.0012, + "step": 55371 + }, + { + "epoch": 7.25, + "learning_rate": 4.534135345648449e-07, + "loss": 0.0004, + "step": 55372 + }, + { + "epoch": 7.25, + "learning_rate": 4.532556688116085e-07, + "loss": 0.0005, + "step": 55373 + }, + { + "epoch": 7.25, + "learning_rate": 4.530978299080835e-07, + "loss": 0.0003, + "step": 55374 + }, + { + "epoch": 7.25, + "learning_rate": 4.5294001785470943e-07, + "loss": 0.0005, + "step": 55375 + }, + { + "epoch": 7.25, + "learning_rate": 4.52782232651936e-07, + "loss": 0.0009, + "step": 55376 + }, + { + "epoch": 7.25, + "learning_rate": 4.5262447430020284e-07, + "loss": 0.0005, + "step": 55377 + }, + { + "epoch": 7.25, + "learning_rate": 4.524667427999563e-07, + "loss": 0.0008, + "step": 55378 + }, + { + "epoch": 7.25, + "learning_rate": 4.5230903815163595e-07, + "loss": 0.0005, + "step": 55379 + }, + { + "epoch": 7.25, + "learning_rate": 4.521513603556893e-07, + "loss": 0.0006, + "step": 55380 + }, + { + "epoch": 7.25, + "learning_rate": 4.519937094125593e-07, + "loss": 0.0006, + "step": 55381 + }, + { + "epoch": 7.25, + "learning_rate": 4.518360853226877e-07, + "loss": 0.0002, + "step": 55382 + }, + { + "epoch": 7.25, + "learning_rate": 4.516784880865177e-07, + "loss": 0.0002, + "step": 55383 + }, + { + "epoch": 7.25, + "learning_rate": 4.5152091770449546e-07, + "loss": 0.0005, + "step": 55384 + }, + { + "epoch": 7.26, + "learning_rate": 4.5136337417706175e-07, + "loss": 0.0003, + "step": 55385 + }, + { + "epoch": 7.26, + "learning_rate": 4.5120585750466073e-07, + "loss": 0.0004, + "step": 55386 + }, + { + "epoch": 7.26, + "learning_rate": 4.5104836768773195e-07, + "loss": 0.0006, + "step": 55387 + }, + { + "epoch": 7.26, + "learning_rate": 4.50890904726724e-07, + "loss": 0.0004, + "step": 55388 + }, + { + "epoch": 7.26, + "learning_rate": 4.5073346862207544e-07, + "loss": 0.0005, + "step": 55389 + }, + { + "epoch": 7.26, + "learning_rate": 4.505760593742303e-07, + "loss": 0.0005, + "step": 55390 + }, + { + "epoch": 7.26, + "learning_rate": 4.504186769836305e-07, + "loss": 0.0003, + "step": 55391 + }, + { + "epoch": 7.26, + "learning_rate": 4.502613214507212e-07, + "loss": 0.0003, + "step": 55392 + }, + { + "epoch": 7.26, + "learning_rate": 4.501039927759443e-07, + "loss": 0.0004, + "step": 55393 + }, + { + "epoch": 7.26, + "learning_rate": 4.499466909597405e-07, + "loss": 0.0007, + "step": 55394 + }, + { + "epoch": 7.26, + "learning_rate": 4.4978941600255175e-07, + "loss": 0.0003, + "step": 55395 + }, + { + "epoch": 7.26, + "learning_rate": 4.496321679048232e-07, + "loss": 0.0008, + "step": 55396 + }, + { + "epoch": 7.26, + "learning_rate": 4.4947494666699563e-07, + "loss": 0.0005, + "step": 55397 + }, + { + "epoch": 7.26, + "learning_rate": 4.4931775228950983e-07, + "loss": 0.0005, + "step": 55398 + }, + { + "epoch": 7.26, + "learning_rate": 4.4916058477281087e-07, + "loss": 0.0002, + "step": 55399 + }, + { + "epoch": 7.26, + "learning_rate": 4.4900344411733853e-07, + "loss": 0.0004, + "step": 55400 + }, + { + "epoch": 7.26, + "learning_rate": 4.4884633032353464e-07, + "loss": 0.001, + "step": 55401 + }, + { + "epoch": 7.26, + "learning_rate": 4.486892433918422e-07, + "loss": 0.0005, + "step": 55402 + }, + { + "epoch": 7.26, + "learning_rate": 4.4853218332270186e-07, + "loss": 0.0009, + "step": 55403 + }, + { + "epoch": 7.26, + "learning_rate": 4.4837515011655785e-07, + "loss": 0.0005, + "step": 55404 + }, + { + "epoch": 7.26, + "learning_rate": 4.4821814377384974e-07, + "loss": 0.0003, + "step": 55405 + }, + { + "epoch": 7.26, + "learning_rate": 4.4806116429501834e-07, + "loss": 0.0006, + "step": 55406 + }, + { + "epoch": 7.26, + "learning_rate": 4.479042116805066e-07, + "loss": 0.0002, + "step": 55407 + }, + { + "epoch": 7.26, + "learning_rate": 4.4774728593075635e-07, + "loss": 0.0004, + "step": 55408 + }, + { + "epoch": 7.26, + "learning_rate": 4.475903870462084e-07, + "loss": 0.0005, + "step": 55409 + }, + { + "epoch": 7.26, + "learning_rate": 4.474335150273024e-07, + "loss": 0.0008, + "step": 55410 + }, + { + "epoch": 7.26, + "learning_rate": 4.472766698744824e-07, + "loss": 0.0005, + "step": 55411 + }, + { + "epoch": 7.26, + "learning_rate": 4.471198515881858e-07, + "loss": 0.0003, + "step": 55412 + }, + { + "epoch": 7.26, + "learning_rate": 4.4696306016885793e-07, + "loss": 0.0003, + "step": 55413 + }, + { + "epoch": 7.26, + "learning_rate": 4.468062956169361e-07, + "loss": 0.0005, + "step": 55414 + }, + { + "epoch": 7.26, + "learning_rate": 4.4664955793286446e-07, + "loss": 0.0007, + "step": 55415 + }, + { + "epoch": 7.26, + "learning_rate": 4.4649284711708264e-07, + "loss": 0.0009, + "step": 55416 + }, + { + "epoch": 7.26, + "learning_rate": 4.46336163170028e-07, + "loss": 0.0003, + "step": 55417 + }, + { + "epoch": 7.26, + "learning_rate": 4.46179506092147e-07, + "loss": 0.0006, + "step": 55418 + }, + { + "epoch": 7.26, + "learning_rate": 4.4602287588387694e-07, + "loss": 0.0005, + "step": 55419 + }, + { + "epoch": 7.26, + "learning_rate": 4.458662725456586e-07, + "loss": 0.0007, + "step": 55420 + }, + { + "epoch": 7.26, + "learning_rate": 4.457096960779317e-07, + "loss": 0.0006, + "step": 55421 + }, + { + "epoch": 7.26, + "learning_rate": 4.45553146481138e-07, + "loss": 0.0009, + "step": 55422 + }, + { + "epoch": 7.26, + "learning_rate": 4.453966237557161e-07, + "loss": 0.0004, + "step": 55423 + }, + { + "epoch": 7.26, + "learning_rate": 4.45240127902109e-07, + "loss": 0.0007, + "step": 55424 + }, + { + "epoch": 7.26, + "learning_rate": 4.4508365892075414e-07, + "loss": 0.0001, + "step": 55425 + }, + { + "epoch": 7.26, + "learning_rate": 4.449272168120944e-07, + "loss": 0.0002, + "step": 55426 + }, + { + "epoch": 7.26, + "learning_rate": 4.4477080157656725e-07, + "loss": 0.0002, + "step": 55427 + }, + { + "epoch": 7.26, + "learning_rate": 4.446144132146135e-07, + "loss": 0.0008, + "step": 55428 + }, + { + "epoch": 7.26, + "learning_rate": 4.4445805172667165e-07, + "loss": 0.0004, + "step": 55429 + }, + { + "epoch": 7.26, + "learning_rate": 4.4430171711318357e-07, + "loss": 0.0003, + "step": 55430 + }, + { + "epoch": 7.26, + "learning_rate": 4.441454093745878e-07, + "loss": 0.0003, + "step": 55431 + }, + { + "epoch": 7.26, + "learning_rate": 4.4398912851132405e-07, + "loss": 0.0007, + "step": 55432 + }, + { + "epoch": 7.26, + "learning_rate": 4.4383287452383294e-07, + "loss": 0.0006, + "step": 55433 + }, + { + "epoch": 7.26, + "learning_rate": 4.436766474125509e-07, + "loss": 0.0009, + "step": 55434 + }, + { + "epoch": 7.26, + "learning_rate": 4.435204471779209e-07, + "loss": 0.0009, + "step": 55435 + }, + { + "epoch": 7.26, + "learning_rate": 4.433642738203814e-07, + "loss": 0.0015, + "step": 55436 + }, + { + "epoch": 7.26, + "learning_rate": 4.432081273403688e-07, + "loss": 0.0004, + "step": 55437 + }, + { + "epoch": 7.26, + "learning_rate": 4.4305200773832714e-07, + "loss": 0.0006, + "step": 55438 + }, + { + "epoch": 7.26, + "learning_rate": 4.4289591501469166e-07, + "loss": 0.0013, + "step": 55439 + }, + { + "epoch": 7.26, + "learning_rate": 4.427398491699009e-07, + "loss": 0.0005, + "step": 55440 + }, + { + "epoch": 7.26, + "learning_rate": 4.4258381020439666e-07, + "loss": 0.0007, + "step": 55441 + }, + { + "epoch": 7.26, + "learning_rate": 4.4242779811861757e-07, + "loss": 0.0005, + "step": 55442 + }, + { + "epoch": 7.26, + "learning_rate": 4.4227181291300105e-07, + "loss": 0.0004, + "step": 55443 + }, + { + "epoch": 7.26, + "learning_rate": 4.4211585458798337e-07, + "loss": 0.0005, + "step": 55444 + }, + { + "epoch": 7.26, + "learning_rate": 4.4195992314400747e-07, + "loss": 0.0004, + "step": 55445 + }, + { + "epoch": 7.26, + "learning_rate": 4.418040185815109e-07, + "loss": 0.0006, + "step": 55446 + }, + { + "epoch": 7.26, + "learning_rate": 4.416481409009321e-07, + "loss": 0.0002, + "step": 55447 + }, + { + "epoch": 7.26, + "learning_rate": 4.4149229010270636e-07, + "loss": 0.0008, + "step": 55448 + }, + { + "epoch": 7.26, + "learning_rate": 4.4133646618727655e-07, + "loss": 0.0003, + "step": 55449 + }, + { + "epoch": 7.26, + "learning_rate": 4.4118066915507906e-07, + "loss": 0.0013, + "step": 55450 + }, + { + "epoch": 7.26, + "learning_rate": 4.4102489900655243e-07, + "loss": 0.0013, + "step": 55451 + }, + { + "epoch": 7.26, + "learning_rate": 4.4086915574213184e-07, + "loss": 0.0003, + "step": 55452 + }, + { + "epoch": 7.26, + "learning_rate": 4.4071343936225917e-07, + "loss": 0.0015, + "step": 55453 + }, + { + "epoch": 7.26, + "learning_rate": 4.4055774986737186e-07, + "loss": 0.0005, + "step": 55454 + }, + { + "epoch": 7.26, + "learning_rate": 4.40402087257904e-07, + "loss": 0.0002, + "step": 55455 + }, + { + "epoch": 7.26, + "learning_rate": 4.4024645153429746e-07, + "loss": 0.0004, + "step": 55456 + }, + { + "epoch": 7.26, + "learning_rate": 4.4009084269698964e-07, + "loss": 0.0003, + "step": 55457 + }, + { + "epoch": 7.26, + "learning_rate": 4.399352607464169e-07, + "loss": 0.0007, + "step": 55458 + }, + { + "epoch": 7.26, + "learning_rate": 4.3977970568301775e-07, + "loss": 0.0004, + "step": 55459 + }, + { + "epoch": 7.26, + "learning_rate": 4.396241775072274e-07, + "loss": 0.0003, + "step": 55460 + }, + { + "epoch": 7.26, + "learning_rate": 4.394686762194855e-07, + "loss": 0.0002, + "step": 55461 + }, + { + "epoch": 7.27, + "learning_rate": 4.3931320182022954e-07, + "loss": 0.0007, + "step": 55462 + }, + { + "epoch": 7.27, + "learning_rate": 4.391577543098957e-07, + "loss": 0.0004, + "step": 55463 + }, + { + "epoch": 7.27, + "learning_rate": 4.3900233368892154e-07, + "loss": 0.0006, + "step": 55464 + }, + { + "epoch": 7.27, + "learning_rate": 4.388469399577444e-07, + "loss": 0.0013, + "step": 55465 + }, + { + "epoch": 7.27, + "learning_rate": 4.3869157311679955e-07, + "loss": 0.0001, + "step": 55466 + }, + { + "epoch": 7.27, + "learning_rate": 4.385362331665266e-07, + "loss": 0.0005, + "step": 55467 + }, + { + "epoch": 7.27, + "learning_rate": 4.383809201073619e-07, + "loss": 0.0007, + "step": 55468 + }, + { + "epoch": 7.27, + "learning_rate": 4.382256339397406e-07, + "loss": 0.0004, + "step": 55469 + }, + { + "epoch": 7.27, + "learning_rate": 4.380703746641024e-07, + "loss": 0.0003, + "step": 55470 + }, + { + "epoch": 7.27, + "learning_rate": 4.3791514228087914e-07, + "loss": 0.0004, + "step": 55471 + }, + { + "epoch": 7.27, + "learning_rate": 4.3775993679051277e-07, + "loss": 0.0008, + "step": 55472 + }, + { + "epoch": 7.27, + "learning_rate": 4.376047581934373e-07, + "loss": 0.0005, + "step": 55473 + }, + { + "epoch": 7.27, + "learning_rate": 4.3744960649008907e-07, + "loss": 0.0004, + "step": 55474 + }, + { + "epoch": 7.27, + "learning_rate": 4.372944816809022e-07, + "loss": 0.0002, + "step": 55475 + }, + { + "epoch": 7.27, + "learning_rate": 4.3713938376631737e-07, + "loss": 0.0009, + "step": 55476 + }, + { + "epoch": 7.27, + "learning_rate": 4.3698431274676766e-07, + "loss": 0.0004, + "step": 55477 + }, + { + "epoch": 7.27, + "learning_rate": 4.368292686226927e-07, + "loss": 0.0007, + "step": 55478 + }, + { + "epoch": 7.27, + "learning_rate": 4.3667425139452324e-07, + "loss": 0.0004, + "step": 55479 + }, + { + "epoch": 7.27, + "learning_rate": 4.3651926106270004e-07, + "loss": 0.0004, + "step": 55480 + }, + { + "epoch": 7.27, + "learning_rate": 4.363642976276572e-07, + "loss": 0.0005, + "step": 55481 + }, + { + "epoch": 7.27, + "learning_rate": 4.362093610898288e-07, + "loss": 0.0007, + "step": 55482 + }, + { + "epoch": 7.27, + "learning_rate": 4.360544514496534e-07, + "loss": 0.0003, + "step": 55483 + }, + { + "epoch": 7.27, + "learning_rate": 4.358995687075662e-07, + "loss": 0.0004, + "step": 55484 + }, + { + "epoch": 7.27, + "learning_rate": 4.3574471286400246e-07, + "loss": 0.0002, + "step": 55485 + }, + { + "epoch": 7.27, + "learning_rate": 4.355898839193951e-07, + "loss": 0.0006, + "step": 55486 + }, + { + "epoch": 7.27, + "learning_rate": 4.354350818741837e-07, + "loss": 0.0003, + "step": 55487 + }, + { + "epoch": 7.27, + "learning_rate": 4.352803067288003e-07, + "loss": 0.0005, + "step": 55488 + }, + { + "epoch": 7.27, + "learning_rate": 4.3512555848368333e-07, + "loss": 0.0008, + "step": 55489 + }, + { + "epoch": 7.27, + "learning_rate": 4.349708371392647e-07, + "loss": 0.0003, + "step": 55490 + }, + { + "epoch": 7.27, + "learning_rate": 4.3481614269598294e-07, + "loss": 0.0004, + "step": 55491 + }, + { + "epoch": 7.27, + "learning_rate": 4.34661475154271e-07, + "loss": 0.0007, + "step": 55492 + }, + { + "epoch": 7.27, + "learning_rate": 4.345068345145642e-07, + "loss": 0.0006, + "step": 55493 + }, + { + "epoch": 7.27, + "learning_rate": 4.3435222077729654e-07, + "loss": 0.0002, + "step": 55494 + }, + { + "epoch": 7.27, + "learning_rate": 4.341976339429055e-07, + "loss": 0.0009, + "step": 55495 + }, + { + "epoch": 7.27, + "learning_rate": 4.340430740118229e-07, + "loss": 0.0003, + "step": 55496 + }, + { + "epoch": 7.27, + "learning_rate": 4.338885409844851e-07, + "loss": 0.0005, + "step": 55497 + }, + { + "epoch": 7.27, + "learning_rate": 4.3373403486132613e-07, + "loss": 0.0004, + "step": 55498 + }, + { + "epoch": 7.27, + "learning_rate": 4.335795556427802e-07, + "loss": 0.0004, + "step": 55499 + }, + { + "epoch": 7.27, + "learning_rate": 4.3342510332928354e-07, + "loss": 0.0005, + "step": 55500 + }, + { + "epoch": 7.27, + "learning_rate": 4.332706779212692e-07, + "loss": 0.0004, + "step": 55501 + }, + { + "epoch": 7.27, + "learning_rate": 4.3311627941917125e-07, + "loss": 0.0006, + "step": 55502 + }, + { + "epoch": 7.27, + "learning_rate": 4.3296190782342483e-07, + "loss": 0.0004, + "step": 55503 + }, + { + "epoch": 7.27, + "learning_rate": 4.3280756313446306e-07, + "loss": 0.0004, + "step": 55504 + }, + { + "epoch": 7.27, + "learning_rate": 4.326532453527199e-07, + "loss": 0.0003, + "step": 55505 + }, + { + "epoch": 7.27, + "learning_rate": 4.3249895447863175e-07, + "loss": 0.0003, + "step": 55506 + }, + { + "epoch": 7.27, + "learning_rate": 4.323446905126294e-07, + "loss": 0.0007, + "step": 55507 + }, + { + "epoch": 7.27, + "learning_rate": 4.321904534551491e-07, + "loss": 0.0002, + "step": 55508 + }, + { + "epoch": 7.27, + "learning_rate": 4.320362433066227e-07, + "loss": 0.0002, + "step": 55509 + }, + { + "epoch": 7.27, + "learning_rate": 4.3188206006748336e-07, + "loss": 0.0002, + "step": 55510 + }, + { + "epoch": 7.27, + "learning_rate": 4.317279037381694e-07, + "loss": 0.0006, + "step": 55511 + }, + { + "epoch": 7.27, + "learning_rate": 4.315737743191095e-07, + "loss": 0.0005, + "step": 55512 + }, + { + "epoch": 7.27, + "learning_rate": 4.3141967181073886e-07, + "loss": 0.0002, + "step": 55513 + }, + { + "epoch": 7.27, + "learning_rate": 4.312655962134915e-07, + "loss": 0.0011, + "step": 55514 + }, + { + "epoch": 7.27, + "learning_rate": 4.3111154752779936e-07, + "loss": 0.0003, + "step": 55515 + }, + { + "epoch": 7.27, + "learning_rate": 4.3095752575409767e-07, + "loss": 0.0004, + "step": 55516 + }, + { + "epoch": 7.27, + "learning_rate": 4.30803530892816e-07, + "loss": 0.0006, + "step": 55517 + }, + { + "epoch": 7.27, + "learning_rate": 4.306495629443919e-07, + "loss": 0.0003, + "step": 55518 + }, + { + "epoch": 7.27, + "learning_rate": 4.30495621909256e-07, + "loss": 0.0005, + "step": 55519 + }, + { + "epoch": 7.27, + "learning_rate": 4.3034170778784025e-07, + "loss": 0.0004, + "step": 55520 + }, + { + "epoch": 7.27, + "learning_rate": 4.301878205805787e-07, + "loss": 0.0003, + "step": 55521 + }, + { + "epoch": 7.27, + "learning_rate": 4.3003396028790556e-07, + "loss": 0.0005, + "step": 55522 + }, + { + "epoch": 7.27, + "learning_rate": 4.298801269102526e-07, + "loss": 0.0003, + "step": 55523 + }, + { + "epoch": 7.27, + "learning_rate": 4.2972632044805175e-07, + "loss": 0.0003, + "step": 55524 + }, + { + "epoch": 7.27, + "learning_rate": 4.295725409017348e-07, + "loss": 0.0003, + "step": 55525 + }, + { + "epoch": 7.27, + "learning_rate": 4.294187882717371e-07, + "loss": 0.0004, + "step": 55526 + }, + { + "epoch": 7.27, + "learning_rate": 4.2926506255848933e-07, + "loss": 0.0008, + "step": 55527 + }, + { + "epoch": 7.27, + "learning_rate": 4.291113637624222e-07, + "loss": 0.0005, + "step": 55528 + }, + { + "epoch": 7.27, + "learning_rate": 4.289576918839711e-07, + "loss": 0.0012, + "step": 55529 + }, + { + "epoch": 7.27, + "learning_rate": 4.288040469235666e-07, + "loss": 0.0005, + "step": 55530 + }, + { + "epoch": 7.27, + "learning_rate": 4.2865042888164174e-07, + "loss": 0.0002, + "step": 55531 + }, + { + "epoch": 7.27, + "learning_rate": 4.284968377586252e-07, + "loss": 0.0004, + "step": 55532 + }, + { + "epoch": 7.27, + "learning_rate": 4.2834327355495306e-07, + "loss": 0.0005, + "step": 55533 + }, + { + "epoch": 7.27, + "learning_rate": 4.2818973627105745e-07, + "loss": 0.0005, + "step": 55534 + }, + { + "epoch": 7.27, + "learning_rate": 4.280362259073667e-07, + "loss": 0.0006, + "step": 55535 + }, + { + "epoch": 7.27, + "learning_rate": 4.2788274246431396e-07, + "loss": 0.0006, + "step": 55536 + }, + { + "epoch": 7.27, + "learning_rate": 4.277292859423332e-07, + "loss": 0.0007, + "step": 55537 + }, + { + "epoch": 7.28, + "learning_rate": 4.2757585634185304e-07, + "loss": 0.0006, + "step": 55538 + }, + { + "epoch": 7.28, + "learning_rate": 4.274224536633065e-07, + "loss": 0.0004, + "step": 55539 + }, + { + "epoch": 7.28, + "learning_rate": 4.2726907790712313e-07, + "loss": 0.0003, + "step": 55540 + }, + { + "epoch": 7.28, + "learning_rate": 4.27115729073736e-07, + "loss": 0.0002, + "step": 55541 + }, + { + "epoch": 7.28, + "learning_rate": 4.2696240716357694e-07, + "loss": 0.0007, + "step": 55542 + }, + { + "epoch": 7.28, + "learning_rate": 4.268091121770757e-07, + "loss": 0.0009, + "step": 55543 + }, + { + "epoch": 7.28, + "learning_rate": 4.266558441146629e-07, + "loss": 0.0002, + "step": 55544 + }, + { + "epoch": 7.28, + "learning_rate": 4.265026029767727e-07, + "loss": 0.0004, + "step": 55545 + }, + { + "epoch": 7.28, + "learning_rate": 4.2634938876383367e-07, + "loss": 0.0003, + "step": 55546 + }, + { + "epoch": 7.28, + "learning_rate": 4.261962014762766e-07, + "loss": 0.0002, + "step": 55547 + }, + { + "epoch": 7.28, + "learning_rate": 4.260430411145333e-07, + "loss": 0.0006, + "step": 55548 + }, + { + "epoch": 7.28, + "learning_rate": 4.258899076790346e-07, + "loss": 0.0002, + "step": 55549 + }, + { + "epoch": 7.28, + "learning_rate": 4.257368011702112e-07, + "loss": 0.0002, + "step": 55550 + }, + { + "epoch": 7.28, + "learning_rate": 4.255837215884906e-07, + "loss": 0.0005, + "step": 55551 + }, + { + "epoch": 7.28, + "learning_rate": 4.2543066893430795e-07, + "loss": 0.0007, + "step": 55552 + }, + { + "epoch": 7.28, + "learning_rate": 4.2527764320809184e-07, + "loss": 0.0003, + "step": 55553 + }, + { + "epoch": 7.28, + "learning_rate": 4.251246444102708e-07, + "loss": 0.0007, + "step": 55554 + }, + { + "epoch": 7.28, + "learning_rate": 4.249716725412778e-07, + "loss": 0.0004, + "step": 55555 + }, + { + "epoch": 7.28, + "learning_rate": 4.248187276015425e-07, + "loss": 0.0011, + "step": 55556 + }, + { + "epoch": 7.28, + "learning_rate": 4.246658095914946e-07, + "loss": 0.0002, + "step": 55557 + }, + { + "epoch": 7.28, + "learning_rate": 4.2451291851156373e-07, + "loss": 0.0003, + "step": 55558 + }, + { + "epoch": 7.28, + "learning_rate": 4.2436005436217954e-07, + "loss": 0.0008, + "step": 55559 + }, + { + "epoch": 7.28, + "learning_rate": 4.2420721714377386e-07, + "loss": 0.0002, + "step": 55560 + }, + { + "epoch": 7.28, + "learning_rate": 4.2405440685677646e-07, + "loss": 0.0006, + "step": 55561 + }, + { + "epoch": 7.28, + "learning_rate": 4.239016235016136e-07, + "loss": 0.0004, + "step": 55562 + }, + { + "epoch": 7.28, + "learning_rate": 4.237488670787193e-07, + "loss": 0.0002, + "step": 55563 + }, + { + "epoch": 7.28, + "learning_rate": 4.2359613758852116e-07, + "loss": 0.0005, + "step": 55564 + }, + { + "epoch": 7.28, + "learning_rate": 4.234434350314476e-07, + "loss": 0.0004, + "step": 55565 + }, + { + "epoch": 7.28, + "learning_rate": 4.232907594079305e-07, + "loss": 0.0005, + "step": 55566 + }, + { + "epoch": 7.28, + "learning_rate": 4.231381107183974e-07, + "loss": 0.0006, + "step": 55567 + }, + { + "epoch": 7.28, + "learning_rate": 4.2298548896327895e-07, + "loss": 0.0004, + "step": 55568 + }, + { + "epoch": 7.28, + "learning_rate": 4.228328941430049e-07, + "loss": 0.0002, + "step": 55569 + }, + { + "epoch": 7.28, + "learning_rate": 4.2268032625800146e-07, + "loss": 0.0003, + "step": 55570 + }, + { + "epoch": 7.28, + "learning_rate": 4.2252778530870065e-07, + "loss": 0.0003, + "step": 55571 + }, + { + "epoch": 7.28, + "learning_rate": 4.223752712955298e-07, + "loss": 0.0003, + "step": 55572 + }, + { + "epoch": 7.28, + "learning_rate": 4.222227842189197e-07, + "loss": 0.0007, + "step": 55573 + }, + { + "epoch": 7.28, + "learning_rate": 4.220703240792967e-07, + "loss": 0.0003, + "step": 55574 + }, + { + "epoch": 7.28, + "learning_rate": 4.2191789087708936e-07, + "loss": 0.0007, + "step": 55575 + }, + { + "epoch": 7.28, + "learning_rate": 4.217654846127306e-07, + "loss": 0.0004, + "step": 55576 + }, + { + "epoch": 7.28, + "learning_rate": 4.2161310528664565e-07, + "loss": 0.0004, + "step": 55577 + }, + { + "epoch": 7.28, + "learning_rate": 4.2146075289926316e-07, + "loss": 0.0007, + "step": 55578 + }, + { + "epoch": 7.28, + "learning_rate": 4.2130842745101266e-07, + "loss": 0.0003, + "step": 55579 + }, + { + "epoch": 7.28, + "learning_rate": 4.211561289423227e-07, + "loss": 0.001, + "step": 55580 + }, + { + "epoch": 7.28, + "learning_rate": 4.210038573736208e-07, + "loss": 0.0008, + "step": 55581 + }, + { + "epoch": 7.28, + "learning_rate": 4.2085161274533434e-07, + "loss": 0.0004, + "step": 55582 + }, + { + "epoch": 7.28, + "learning_rate": 4.2069939505789415e-07, + "loss": 0.0003, + "step": 55583 + }, + { + "epoch": 7.28, + "learning_rate": 4.2054720431172645e-07, + "loss": 0.001, + "step": 55584 + }, + { + "epoch": 7.28, + "learning_rate": 4.2039504050725763e-07, + "loss": 0.0003, + "step": 55585 + }, + { + "epoch": 7.28, + "learning_rate": 4.202429036449185e-07, + "loss": 0.0006, + "step": 55586 + }, + { + "epoch": 7.28, + "learning_rate": 4.200907937251375e-07, + "loss": 0.0012, + "step": 55587 + }, + { + "epoch": 7.28, + "learning_rate": 4.199387107483399e-07, + "loss": 0.0003, + "step": 55588 + }, + { + "epoch": 7.28, + "learning_rate": 4.197866547149554e-07, + "loss": 0.0005, + "step": 55589 + }, + { + "epoch": 7.28, + "learning_rate": 4.196346256254091e-07, + "loss": 0.0006, + "step": 55590 + }, + { + "epoch": 7.28, + "learning_rate": 4.194826234801319e-07, + "loss": 0.0002, + "step": 55591 + }, + { + "epoch": 7.28, + "learning_rate": 4.1933064827954894e-07, + "loss": 0.0004, + "step": 55592 + }, + { + "epoch": 7.28, + "learning_rate": 4.191787000240877e-07, + "loss": 0.0008, + "step": 55593 + }, + { + "epoch": 7.28, + "learning_rate": 4.1902677871417776e-07, + "loss": 0.0013, + "step": 55594 + }, + { + "epoch": 7.28, + "learning_rate": 4.188748843502433e-07, + "loss": 0.0002, + "step": 55595 + }, + { + "epoch": 7.28, + "learning_rate": 4.187230169327139e-07, + "loss": 0.0003, + "step": 55596 + }, + { + "epoch": 7.28, + "learning_rate": 4.1857117646201375e-07, + "loss": 0.0004, + "step": 55597 + }, + { + "epoch": 7.28, + "learning_rate": 4.1841936293857357e-07, + "loss": 0.0007, + "step": 55598 + }, + { + "epoch": 7.28, + "learning_rate": 4.182675763628197e-07, + "loss": 0.0003, + "step": 55599 + }, + { + "epoch": 7.28, + "learning_rate": 4.1811581673517733e-07, + "loss": 0.0004, + "step": 55600 + }, + { + "epoch": 7.28, + "learning_rate": 4.179640840560728e-07, + "loss": 0.0004, + "step": 55601 + }, + { + "epoch": 7.28, + "learning_rate": 4.178123783259347e-07, + "loss": 0.0003, + "step": 55602 + }, + { + "epoch": 7.28, + "learning_rate": 4.176606995451904e-07, + "loss": 0.0006, + "step": 55603 + }, + { + "epoch": 7.28, + "learning_rate": 4.1750904771426405e-07, + "loss": 0.0009, + "step": 55604 + }, + { + "epoch": 7.28, + "learning_rate": 4.1735742283358195e-07, + "loss": 0.0007, + "step": 55605 + }, + { + "epoch": 7.28, + "learning_rate": 4.1720582490357373e-07, + "loss": 0.0005, + "step": 55606 + }, + { + "epoch": 7.28, + "learning_rate": 4.1705425392466247e-07, + "loss": 0.0005, + "step": 55607 + }, + { + "epoch": 7.28, + "learning_rate": 4.1690270989727557e-07, + "loss": 0.0003, + "step": 55608 + }, + { + "epoch": 7.28, + "learning_rate": 4.167511928218393e-07, + "loss": 0.0004, + "step": 55609 + }, + { + "epoch": 7.28, + "learning_rate": 4.1659970269878115e-07, + "loss": 0.0002, + "step": 55610 + }, + { + "epoch": 7.28, + "learning_rate": 4.164482395285263e-07, + "loss": 0.0007, + "step": 55611 + }, + { + "epoch": 7.28, + "learning_rate": 4.162968033114989e-07, + "loss": 0.0005, + "step": 55612 + }, + { + "epoch": 7.28, + "learning_rate": 4.1614539404812746e-07, + "loss": 0.0004, + "step": 55613 + }, + { + "epoch": 7.29, + "learning_rate": 4.1599401173883725e-07, + "loss": 0.0004, + "step": 55614 + }, + { + "epoch": 7.29, + "learning_rate": 4.158426563840534e-07, + "loss": 0.0002, + "step": 55615 + }, + { + "epoch": 7.29, + "learning_rate": 4.1569132798420007e-07, + "loss": 0.0007, + "step": 55616 + }, + { + "epoch": 7.29, + "learning_rate": 4.155400265397058e-07, + "loss": 0.0003, + "step": 55617 + }, + { + "epoch": 7.29, + "learning_rate": 4.153887520509947e-07, + "loss": 0.0003, + "step": 55618 + }, + { + "epoch": 7.29, + "learning_rate": 4.15237504518492e-07, + "loss": 0.0002, + "step": 55619 + }, + { + "epoch": 7.29, + "learning_rate": 4.1508628394262285e-07, + "loss": 0.0008, + "step": 55620 + }, + { + "epoch": 7.29, + "learning_rate": 4.149350903238136e-07, + "loss": 0.0009, + "step": 55621 + }, + { + "epoch": 7.29, + "learning_rate": 4.1478392366248956e-07, + "loss": 0.0003, + "step": 55622 + }, + { + "epoch": 7.29, + "learning_rate": 4.146327839590758e-07, + "loss": 0.0003, + "step": 55623 + }, + { + "epoch": 7.29, + "learning_rate": 4.144816712139943e-07, + "loss": 0.0005, + "step": 55624 + }, + { + "epoch": 7.29, + "learning_rate": 4.143305854276747e-07, + "loss": 0.0003, + "step": 55625 + }, + { + "epoch": 7.29, + "learning_rate": 4.1417952660053995e-07, + "loss": 0.0003, + "step": 55626 + }, + { + "epoch": 7.29, + "learning_rate": 4.140284947330142e-07, + "loss": 0.0004, + "step": 55627 + }, + { + "epoch": 7.29, + "learning_rate": 4.1387748982552265e-07, + "loss": 0.0002, + "step": 55628 + }, + { + "epoch": 7.29, + "learning_rate": 4.137265118784917e-07, + "loss": 0.0004, + "step": 55629 + }, + { + "epoch": 7.29, + "learning_rate": 4.1357556089234197e-07, + "loss": 0.0005, + "step": 55630 + }, + { + "epoch": 7.29, + "learning_rate": 4.134246368675021e-07, + "loss": 0.0004, + "step": 55631 + }, + { + "epoch": 7.29, + "learning_rate": 4.13273739804394e-07, + "loss": 0.0003, + "step": 55632 + }, + { + "epoch": 7.29, + "learning_rate": 4.1312286970344394e-07, + "loss": 0.0004, + "step": 55633 + }, + { + "epoch": 7.29, + "learning_rate": 4.1297202656507496e-07, + "loss": 0.0009, + "step": 55634 + }, + { + "epoch": 7.29, + "learning_rate": 4.1282121038971113e-07, + "loss": 0.0013, + "step": 55635 + }, + { + "epoch": 7.29, + "learning_rate": 4.1267042117777765e-07, + "loss": 0.0008, + "step": 55636 + }, + { + "epoch": 7.29, + "learning_rate": 4.1251965892969756e-07, + "loss": 0.0011, + "step": 55637 + }, + { + "epoch": 7.29, + "learning_rate": 4.1236892364589607e-07, + "loss": 0.0006, + "step": 55638 + }, + { + "epoch": 7.29, + "learning_rate": 4.1221821532679617e-07, + "loss": 0.0005, + "step": 55639 + }, + { + "epoch": 7.29, + "learning_rate": 4.1206753397282195e-07, + "loss": 0.0003, + "step": 55640 + }, + { + "epoch": 7.29, + "learning_rate": 4.119168795843964e-07, + "loss": 0.0002, + "step": 55641 + }, + { + "epoch": 7.29, + "learning_rate": 4.1176625216194476e-07, + "loss": 0.0003, + "step": 55642 + }, + { + "epoch": 7.29, + "learning_rate": 4.116156517058889e-07, + "loss": 0.0008, + "step": 55643 + }, + { + "epoch": 7.29, + "learning_rate": 4.1146507821665515e-07, + "loss": 0.0004, + "step": 55644 + }, + { + "epoch": 7.29, + "learning_rate": 4.113145316946643e-07, + "loss": 0.0006, + "step": 55645 + }, + { + "epoch": 7.29, + "learning_rate": 4.1116401214034153e-07, + "loss": 0.0002, + "step": 55646 + }, + { + "epoch": 7.29, + "learning_rate": 4.110135195541076e-07, + "loss": 0.0004, + "step": 55647 + }, + { + "epoch": 7.29, + "learning_rate": 4.108630539363889e-07, + "loss": 0.0009, + "step": 55648 + }, + { + "epoch": 7.29, + "learning_rate": 4.107126152876073e-07, + "loss": 0.0005, + "step": 55649 + }, + { + "epoch": 7.29, + "learning_rate": 4.105622036081847e-07, + "loss": 0.0005, + "step": 55650 + }, + { + "epoch": 7.29, + "learning_rate": 4.104118188985462e-07, + "loss": 0.0001, + "step": 55651 + }, + { + "epoch": 7.29, + "learning_rate": 4.102614611591127e-07, + "loss": 0.0012, + "step": 55652 + }, + { + "epoch": 7.29, + "learning_rate": 4.101111303903094e-07, + "loss": 0.001, + "step": 55653 + }, + { + "epoch": 7.29, + "learning_rate": 4.099608265925581e-07, + "loss": 0.0008, + "step": 55654 + }, + { + "epoch": 7.29, + "learning_rate": 4.0981054976627965e-07, + "loss": 0.0005, + "step": 55655 + }, + { + "epoch": 7.29, + "learning_rate": 4.096602999119004e-07, + "loss": 0.0002, + "step": 55656 + }, + { + "epoch": 7.29, + "learning_rate": 4.095100770298399e-07, + "loss": 0.0004, + "step": 55657 + }, + { + "epoch": 7.29, + "learning_rate": 4.093598811205202e-07, + "loss": 0.0004, + "step": 55658 + }, + { + "epoch": 7.29, + "learning_rate": 4.092097121843674e-07, + "loss": 0.0003, + "step": 55659 + }, + { + "epoch": 7.29, + "learning_rate": 4.090595702218003e-07, + "loss": 0.0002, + "step": 55660 + }, + { + "epoch": 7.29, + "learning_rate": 4.0890945523324287e-07, + "loss": 0.0004, + "step": 55661 + }, + { + "epoch": 7.29, + "learning_rate": 4.0875936721911593e-07, + "loss": 0.0004, + "step": 55662 + }, + { + "epoch": 7.29, + "learning_rate": 4.086093061798424e-07, + "loss": 0.0005, + "step": 55663 + }, + { + "epoch": 7.29, + "learning_rate": 4.0845927211584536e-07, + "loss": 0.0004, + "step": 55664 + }, + { + "epoch": 7.29, + "learning_rate": 4.0830926502754553e-07, + "loss": 0.0008, + "step": 55665 + }, + { + "epoch": 7.29, + "learning_rate": 4.081592849153648e-07, + "loss": 0.0002, + "step": 55666 + }, + { + "epoch": 7.29, + "learning_rate": 4.0800933177972625e-07, + "loss": 0.001, + "step": 55667 + }, + { + "epoch": 7.29, + "learning_rate": 4.0785940562105055e-07, + "loss": 0.0006, + "step": 55668 + }, + { + "epoch": 7.29, + "learning_rate": 4.077095064397596e-07, + "loss": 0.0004, + "step": 55669 + }, + { + "epoch": 7.29, + "learning_rate": 4.075596342362731e-07, + "loss": 0.0002, + "step": 55670 + }, + { + "epoch": 7.29, + "learning_rate": 4.0740978901101624e-07, + "loss": 0.0002, + "step": 55671 + }, + { + "epoch": 7.29, + "learning_rate": 4.072599707644076e-07, + "loss": 0.0006, + "step": 55672 + }, + { + "epoch": 7.29, + "learning_rate": 4.0711017949686906e-07, + "loss": 0.0002, + "step": 55673 + }, + { + "epoch": 7.29, + "learning_rate": 4.0696041520882137e-07, + "loss": 0.0005, + "step": 55674 + }, + { + "epoch": 7.29, + "learning_rate": 4.068106779006886e-07, + "loss": 0.0004, + "step": 55675 + }, + { + "epoch": 7.29, + "learning_rate": 4.0666096757289055e-07, + "loss": 0.0003, + "step": 55676 + }, + { + "epoch": 7.29, + "learning_rate": 4.065112842258456e-07, + "loss": 0.0003, + "step": 55677 + }, + { + "epoch": 7.29, + "learning_rate": 4.06361627859978e-07, + "loss": 0.0005, + "step": 55678 + }, + { + "epoch": 7.29, + "learning_rate": 4.0621199847570734e-07, + "loss": 0.0007, + "step": 55679 + }, + { + "epoch": 7.29, + "learning_rate": 4.060623960734544e-07, + "loss": 0.0003, + "step": 55680 + }, + { + "epoch": 7.29, + "learning_rate": 4.059128206536389e-07, + "loss": 0.0005, + "step": 55681 + }, + { + "epoch": 7.29, + "learning_rate": 4.0576327221668486e-07, + "loss": 0.0003, + "step": 55682 + }, + { + "epoch": 7.29, + "learning_rate": 4.056137507630098e-07, + "loss": 0.0003, + "step": 55683 + }, + { + "epoch": 7.29, + "learning_rate": 4.0546425629303445e-07, + "loss": 0.0004, + "step": 55684 + }, + { + "epoch": 7.29, + "learning_rate": 4.0531478880717954e-07, + "loss": 0.0006, + "step": 55685 + }, + { + "epoch": 7.29, + "learning_rate": 4.0516534830586707e-07, + "loss": 0.0002, + "step": 55686 + }, + { + "epoch": 7.29, + "learning_rate": 4.050159347895155e-07, + "loss": 0.0014, + "step": 55687 + }, + { + "epoch": 7.29, + "learning_rate": 4.0486654825854677e-07, + "loss": 0.0002, + "step": 55688 + }, + { + "epoch": 7.29, + "learning_rate": 4.047171887133783e-07, + "loss": 0.0002, + "step": 55689 + }, + { + "epoch": 7.29, + "learning_rate": 4.0456785615443304e-07, + "loss": 0.0002, + "step": 55690 + }, + { + "epoch": 7.3, + "learning_rate": 4.0441855058212853e-07, + "loss": 0.0002, + "step": 55691 + }, + { + "epoch": 7.3, + "learning_rate": 4.042692719968866e-07, + "loss": 0.0003, + "step": 55692 + }, + { + "epoch": 7.3, + "learning_rate": 4.041200203991258e-07, + "loss": 0.0006, + "step": 55693 + }, + { + "epoch": 7.3, + "learning_rate": 4.03970795789268e-07, + "loss": 0.0004, + "step": 55694 + }, + { + "epoch": 7.3, + "learning_rate": 4.0382159816772957e-07, + "loss": 0.0004, + "step": 55695 + }, + { + "epoch": 7.3, + "learning_rate": 4.036724275349324e-07, + "loss": 0.0003, + "step": 55696 + }, + { + "epoch": 7.3, + "learning_rate": 4.03523283891295e-07, + "loss": 0.0005, + "step": 55697 + }, + { + "epoch": 7.3, + "learning_rate": 4.033741672372382e-07, + "loss": 0.0006, + "step": 55698 + }, + { + "epoch": 7.3, + "learning_rate": 4.032250775731805e-07, + "loss": 0.0005, + "step": 55699 + }, + { + "epoch": 7.3, + "learning_rate": 4.030760148995394e-07, + "loss": 0.0008, + "step": 55700 + }, + { + "epoch": 7.3, + "learning_rate": 4.0292697921673786e-07, + "loss": 0.0004, + "step": 55701 + }, + { + "epoch": 7.3, + "learning_rate": 4.027779705251933e-07, + "loss": 0.0009, + "step": 55702 + }, + { + "epoch": 7.3, + "learning_rate": 4.026289888253243e-07, + "loss": 0.0006, + "step": 55703 + }, + { + "epoch": 7.3, + "learning_rate": 4.0248003411754944e-07, + "loss": 0.0002, + "step": 55704 + }, + { + "epoch": 7.3, + "learning_rate": 4.023311064022895e-07, + "loss": 0.0013, + "step": 55705 + }, + { + "epoch": 7.3, + "learning_rate": 4.021822056799607e-07, + "loss": 0.0003, + "step": 55706 + }, + { + "epoch": 7.3, + "learning_rate": 4.020333319509839e-07, + "loss": 0.0003, + "step": 55707 + }, + { + "epoch": 7.3, + "learning_rate": 4.0188448521577663e-07, + "loss": 0.0005, + "step": 55708 + }, + { + "epoch": 7.3, + "learning_rate": 4.017356654747595e-07, + "loss": 0.0006, + "step": 55709 + }, + { + "epoch": 7.3, + "learning_rate": 4.0158687272835005e-07, + "loss": 0.0005, + "step": 55710 + }, + { + "epoch": 7.3, + "learning_rate": 4.014381069769657e-07, + "loss": 0.0005, + "step": 55711 + }, + { + "epoch": 7.3, + "learning_rate": 4.0128936822102393e-07, + "loss": 0.0004, + "step": 55712 + }, + { + "epoch": 7.3, + "learning_rate": 4.011406564609466e-07, + "loss": 0.0004, + "step": 55713 + }, + { + "epoch": 7.3, + "learning_rate": 4.0099197169715e-07, + "loss": 0.0002, + "step": 55714 + }, + { + "epoch": 7.3, + "learning_rate": 4.0084331393005047e-07, + "loss": 0.0008, + "step": 55715 + }, + { + "epoch": 7.3, + "learning_rate": 4.0069468316007e-07, + "loss": 0.001, + "step": 55716 + }, + { + "epoch": 7.3, + "learning_rate": 4.005460793876226e-07, + "loss": 0.0011, + "step": 55717 + }, + { + "epoch": 7.3, + "learning_rate": 4.0039750261313015e-07, + "loss": 0.0003, + "step": 55718 + }, + { + "epoch": 7.3, + "learning_rate": 4.0024895283700795e-07, + "loss": 0.0004, + "step": 55719 + }, + { + "epoch": 7.3, + "learning_rate": 4.001004300596734e-07, + "loss": 0.0004, + "step": 55720 + }, + { + "epoch": 7.3, + "learning_rate": 3.999519342815461e-07, + "loss": 0.0004, + "step": 55721 + }, + { + "epoch": 7.3, + "learning_rate": 3.998034655030425e-07, + "loss": 0.0004, + "step": 55722 + }, + { + "epoch": 7.3, + "learning_rate": 3.9965502372458e-07, + "loss": 0.0006, + "step": 55723 + }, + { + "epoch": 7.3, + "learning_rate": 3.995066089465771e-07, + "loss": 0.0006, + "step": 55724 + }, + { + "epoch": 7.3, + "learning_rate": 3.993582211694513e-07, + "loss": 0.0005, + "step": 55725 + }, + { + "epoch": 7.3, + "learning_rate": 3.9920986039361786e-07, + "loss": 0.0004, + "step": 55726 + }, + { + "epoch": 7.3, + "learning_rate": 3.990615266194953e-07, + "loss": 0.001, + "step": 55727 + }, + { + "epoch": 7.3, + "learning_rate": 3.98913219847501e-07, + "loss": 0.0009, + "step": 55728 + }, + { + "epoch": 7.3, + "learning_rate": 3.987649400780524e-07, + "loss": 0.0004, + "step": 55729 + }, + { + "epoch": 7.3, + "learning_rate": 3.986166873115671e-07, + "loss": 0.0004, + "step": 55730 + }, + { + "epoch": 7.3, + "learning_rate": 3.984684615484591e-07, + "loss": 0.0004, + "step": 55731 + }, + { + "epoch": 7.3, + "learning_rate": 3.9832026278914913e-07, + "loss": 0.0004, + "step": 55732 + }, + { + "epoch": 7.3, + "learning_rate": 3.981720910340514e-07, + "loss": 0.0005, + "step": 55733 + }, + { + "epoch": 7.3, + "learning_rate": 3.980239462835822e-07, + "loss": 0.0005, + "step": 55734 + }, + { + "epoch": 7.3, + "learning_rate": 3.978758285381612e-07, + "loss": 0.0007, + "step": 55735 + }, + { + "epoch": 7.3, + "learning_rate": 3.977277377982025e-07, + "loss": 0.0002, + "step": 55736 + }, + { + "epoch": 7.3, + "learning_rate": 3.9757967406412243e-07, + "loss": 0.0012, + "step": 55737 + }, + { + "epoch": 7.3, + "learning_rate": 3.974316373363385e-07, + "loss": 0.0003, + "step": 55738 + }, + { + "epoch": 7.3, + "learning_rate": 3.9728362761526583e-07, + "loss": 0.0003, + "step": 55739 + }, + { + "epoch": 7.3, + "learning_rate": 3.9713564490132304e-07, + "loss": 0.0012, + "step": 55740 + }, + { + "epoch": 7.3, + "learning_rate": 3.9698768919492535e-07, + "loss": 0.0002, + "step": 55741 + }, + { + "epoch": 7.3, + "learning_rate": 3.9683976049648574e-07, + "loss": 0.0003, + "step": 55742 + }, + { + "epoch": 7.3, + "learning_rate": 3.9669185880642503e-07, + "loss": 0.0003, + "step": 55743 + }, + { + "epoch": 7.3, + "learning_rate": 3.9654398412515725e-07, + "loss": 0.0003, + "step": 55744 + }, + { + "epoch": 7.3, + "learning_rate": 3.963961364530977e-07, + "loss": 0.0006, + "step": 55745 + }, + { + "epoch": 7.3, + "learning_rate": 3.962483157906616e-07, + "loss": 0.0009, + "step": 55746 + }, + { + "epoch": 7.3, + "learning_rate": 3.9610052213826743e-07, + "loss": 0.0005, + "step": 55747 + }, + { + "epoch": 7.3, + "learning_rate": 3.9595275549632826e-07, + "loss": 0.0017, + "step": 55748 + }, + { + "epoch": 7.3, + "learning_rate": 3.958050158652604e-07, + "loss": 0.0006, + "step": 55749 + }, + { + "epoch": 7.3, + "learning_rate": 3.9565730324547803e-07, + "loss": 0.0003, + "step": 55750 + }, + { + "epoch": 7.3, + "learning_rate": 3.955096176373996e-07, + "loss": 0.0002, + "step": 55751 + }, + { + "epoch": 7.3, + "learning_rate": 3.9536195904143926e-07, + "loss": 0.0003, + "step": 55752 + }, + { + "epoch": 7.3, + "learning_rate": 3.9521432745801224e-07, + "loss": 0.0006, + "step": 55753 + }, + { + "epoch": 7.3, + "learning_rate": 3.950667228875316e-07, + "loss": 0.0002, + "step": 55754 + }, + { + "epoch": 7.3, + "learning_rate": 3.9491914533041576e-07, + "loss": 0.0008, + "step": 55755 + }, + { + "epoch": 7.3, + "learning_rate": 3.9477159478707894e-07, + "loss": 0.0004, + "step": 55756 + }, + { + "epoch": 7.3, + "learning_rate": 3.94624071257933e-07, + "loss": 0.0005, + "step": 55757 + }, + { + "epoch": 7.3, + "learning_rate": 3.944765747433976e-07, + "loss": 0.0004, + "step": 55758 + }, + { + "epoch": 7.3, + "learning_rate": 3.9432910524388466e-07, + "loss": 0.0002, + "step": 55759 + }, + { + "epoch": 7.3, + "learning_rate": 3.941816627598105e-07, + "loss": 0.0009, + "step": 55760 + }, + { + "epoch": 7.3, + "learning_rate": 3.94034247291587e-07, + "loss": 0.0006, + "step": 55761 + }, + { + "epoch": 7.3, + "learning_rate": 3.9388685883963054e-07, + "loss": 0.0004, + "step": 55762 + }, + { + "epoch": 7.3, + "learning_rate": 3.9373949740435734e-07, + "loss": 0.0003, + "step": 55763 + }, + { + "epoch": 7.3, + "learning_rate": 3.935921629861794e-07, + "loss": 0.0004, + "step": 55764 + }, + { + "epoch": 7.3, + "learning_rate": 3.9344485558551194e-07, + "loss": 0.0002, + "step": 55765 + }, + { + "epoch": 7.3, + "learning_rate": 3.932975752027701e-07, + "loss": 0.0004, + "step": 55766 + }, + { + "epoch": 7.31, + "learning_rate": 3.93150321838367e-07, + "loss": 0.0002, + "step": 55767 + }, + { + "epoch": 7.31, + "learning_rate": 3.9300309549271667e-07, + "loss": 0.0003, + "step": 55768 + }, + { + "epoch": 7.31, + "learning_rate": 3.928558961662321e-07, + "loss": 0.0008, + "step": 55769 + }, + { + "epoch": 7.31, + "learning_rate": 3.927087238593308e-07, + "loss": 0.0003, + "step": 55770 + }, + { + "epoch": 7.31, + "learning_rate": 3.925615785724246e-07, + "loss": 0.0004, + "step": 55771 + }, + { + "epoch": 7.31, + "learning_rate": 3.9241446030592547e-07, + "loss": 0.0009, + "step": 55772 + }, + { + "epoch": 7.31, + "learning_rate": 3.9226736906024966e-07, + "loss": 0.0004, + "step": 55773 + }, + { + "epoch": 7.31, + "learning_rate": 3.9212030483581133e-07, + "loss": 0.0004, + "step": 55774 + }, + { + "epoch": 7.31, + "learning_rate": 3.9197326763302237e-07, + "loss": 0.0003, + "step": 55775 + }, + { + "epoch": 7.31, + "learning_rate": 3.9182625745229686e-07, + "loss": 0.0009, + "step": 55776 + }, + { + "epoch": 7.31, + "learning_rate": 3.9167927429404784e-07, + "loss": 0.0004, + "step": 55777 + }, + { + "epoch": 7.31, + "learning_rate": 3.915323181586905e-07, + "loss": 0.0005, + "step": 55778 + }, + { + "epoch": 7.31, + "learning_rate": 3.913853890466368e-07, + "loss": 0.0007, + "step": 55779 + }, + { + "epoch": 7.31, + "learning_rate": 3.912384869582986e-07, + "loss": 0.0008, + "step": 55780 + }, + { + "epoch": 7.31, + "learning_rate": 3.9109161189409104e-07, + "loss": 0.0013, + "step": 55781 + }, + { + "epoch": 7.31, + "learning_rate": 3.909447638544273e-07, + "loss": 0.0006, + "step": 55782 + }, + { + "epoch": 7.31, + "learning_rate": 3.9079794283971795e-07, + "loss": 0.0006, + "step": 55783 + }, + { + "epoch": 7.31, + "learning_rate": 3.9065114885037836e-07, + "loss": 0.0005, + "step": 55784 + }, + { + "epoch": 7.31, + "learning_rate": 3.905043818868215e-07, + "loss": 0.0007, + "step": 55785 + }, + { + "epoch": 7.31, + "learning_rate": 3.903576419494592e-07, + "loss": 0.0003, + "step": 55786 + }, + { + "epoch": 7.31, + "learning_rate": 3.902109290387046e-07, + "loss": 0.0003, + "step": 55787 + }, + { + "epoch": 7.31, + "learning_rate": 3.9006424315496947e-07, + "loss": 0.0005, + "step": 55788 + }, + { + "epoch": 7.31, + "learning_rate": 3.899175842986669e-07, + "loss": 0.0004, + "step": 55789 + }, + { + "epoch": 7.31, + "learning_rate": 3.897709524702098e-07, + "loss": 0.0005, + "step": 55790 + }, + { + "epoch": 7.31, + "learning_rate": 3.896243476700101e-07, + "loss": 0.0004, + "step": 55791 + }, + { + "epoch": 7.31, + "learning_rate": 3.8947776989847863e-07, + "loss": 0.0004, + "step": 55792 + }, + { + "epoch": 7.31, + "learning_rate": 3.8933121915603166e-07, + "loss": 0.0003, + "step": 55793 + }, + { + "epoch": 7.31, + "learning_rate": 3.8918469544307667e-07, + "loss": 0.0007, + "step": 55794 + }, + { + "epoch": 7.31, + "learning_rate": 3.890381987600289e-07, + "loss": 0.001, + "step": 55795 + }, + { + "epoch": 7.31, + "learning_rate": 3.88891729107298e-07, + "loss": 0.0004, + "step": 55796 + }, + { + "epoch": 7.31, + "learning_rate": 3.887452864852992e-07, + "loss": 0.001, + "step": 55797 + }, + { + "epoch": 7.31, + "learning_rate": 3.885988708944421e-07, + "loss": 0.0002, + "step": 55798 + }, + { + "epoch": 7.31, + "learning_rate": 3.8845248233513766e-07, + "loss": 0.0005, + "step": 55799 + }, + { + "epoch": 7.31, + "learning_rate": 3.8830612080779983e-07, + "loss": 0.0005, + "step": 55800 + }, + { + "epoch": 7.31, + "learning_rate": 3.8815978631283946e-07, + "loss": 0.0006, + "step": 55801 + }, + { + "epoch": 7.31, + "learning_rate": 3.880134788506673e-07, + "loss": 0.0006, + "step": 55802 + }, + { + "epoch": 7.31, + "learning_rate": 3.878671984216953e-07, + "loss": 0.0004, + "step": 55803 + }, + { + "epoch": 7.31, + "learning_rate": 3.877209450263353e-07, + "loss": 0.0005, + "step": 55804 + }, + { + "epoch": 7.31, + "learning_rate": 3.87574718664997e-07, + "loss": 0.0003, + "step": 55805 + }, + { + "epoch": 7.31, + "learning_rate": 3.8742851933809447e-07, + "loss": 0.0011, + "step": 55806 + }, + { + "epoch": 7.31, + "learning_rate": 3.872823470460352e-07, + "loss": 0.0003, + "step": 55807 + }, + { + "epoch": 7.31, + "learning_rate": 3.871362017892344e-07, + "loss": 0.0001, + "step": 55808 + }, + { + "epoch": 7.31, + "learning_rate": 3.8699008356810175e-07, + "loss": 0.0006, + "step": 55809 + }, + { + "epoch": 7.31, + "learning_rate": 3.8684399238304695e-07, + "loss": 0.0002, + "step": 55810 + }, + { + "epoch": 7.31, + "learning_rate": 3.866979282344796e-07, + "loss": 0.0009, + "step": 55811 + }, + { + "epoch": 7.31, + "learning_rate": 3.865518911228139e-07, + "loss": 0.0004, + "step": 55812 + }, + { + "epoch": 7.31, + "learning_rate": 3.8640588104845946e-07, + "loss": 0.0004, + "step": 55813 + }, + { + "epoch": 7.31, + "learning_rate": 3.86259898011826e-07, + "loss": 0.0009, + "step": 55814 + }, + { + "epoch": 7.31, + "learning_rate": 3.861139420133242e-07, + "loss": 0.001, + "step": 55815 + }, + { + "epoch": 7.31, + "learning_rate": 3.8596801305336493e-07, + "loss": 0.0004, + "step": 55816 + }, + { + "epoch": 7.31, + "learning_rate": 3.858221111323601e-07, + "loss": 0.0009, + "step": 55817 + }, + { + "epoch": 7.31, + "learning_rate": 3.8567623625071715e-07, + "loss": 0.0002, + "step": 55818 + }, + { + "epoch": 7.31, + "learning_rate": 3.8553038840884794e-07, + "loss": 0.0018, + "step": 55819 + }, + { + "epoch": 7.31, + "learning_rate": 3.8538456760716325e-07, + "loss": 0.0004, + "step": 55820 + }, + { + "epoch": 7.31, + "learning_rate": 3.8523877384607275e-07, + "loss": 0.0008, + "step": 55821 + }, + { + "epoch": 7.31, + "learning_rate": 3.8509300712598394e-07, + "loss": 0.0002, + "step": 55822 + }, + { + "epoch": 7.31, + "learning_rate": 3.84947267447312e-07, + "loss": 0.0005, + "step": 55823 + }, + { + "epoch": 7.31, + "learning_rate": 3.848015548104622e-07, + "loss": 0.0005, + "step": 55824 + }, + { + "epoch": 7.31, + "learning_rate": 3.846558692158464e-07, + "loss": 0.0003, + "step": 55825 + }, + { + "epoch": 7.31, + "learning_rate": 3.8451021066387315e-07, + "loss": 0.0004, + "step": 55826 + }, + { + "epoch": 7.31, + "learning_rate": 3.8436457915495215e-07, + "loss": 0.0005, + "step": 55827 + }, + { + "epoch": 7.31, + "learning_rate": 3.8421897468949643e-07, + "loss": 0.0003, + "step": 55828 + }, + { + "epoch": 7.31, + "learning_rate": 3.8407339726791113e-07, + "loss": 0.0002, + "step": 55829 + }, + { + "epoch": 7.31, + "learning_rate": 3.839278468906071e-07, + "loss": 0.0009, + "step": 55830 + }, + { + "epoch": 7.31, + "learning_rate": 3.8378232355799404e-07, + "loss": 0.0006, + "step": 55831 + }, + { + "epoch": 7.31, + "learning_rate": 3.836368272704827e-07, + "loss": 0.0003, + "step": 55832 + }, + { + "epoch": 7.31, + "learning_rate": 3.8349135802847937e-07, + "loss": 0.0008, + "step": 55833 + }, + { + "epoch": 7.31, + "learning_rate": 3.833459158323938e-07, + "loss": 0.0017, + "step": 55834 + }, + { + "epoch": 7.31, + "learning_rate": 3.832005006826367e-07, + "loss": 0.0007, + "step": 55835 + }, + { + "epoch": 7.31, + "learning_rate": 3.830551125796167e-07, + "loss": 0.0003, + "step": 55836 + }, + { + "epoch": 7.31, + "learning_rate": 3.829097515237401e-07, + "loss": 0.0003, + "step": 55837 + }, + { + "epoch": 7.31, + "learning_rate": 3.827644175154177e-07, + "loss": 0.0004, + "step": 55838 + }, + { + "epoch": 7.31, + "learning_rate": 3.826191105550603e-07, + "loss": 0.0002, + "step": 55839 + }, + { + "epoch": 7.31, + "learning_rate": 3.824738306430731e-07, + "loss": 0.0003, + "step": 55840 + }, + { + "epoch": 7.31, + "learning_rate": 3.8232857777986687e-07, + "loss": 0.0003, + "step": 55841 + }, + { + "epoch": 7.31, + "learning_rate": 3.82183351965848e-07, + "loss": 0.0004, + "step": 55842 + }, + { + "epoch": 7.32, + "learning_rate": 3.820381532014283e-07, + "loss": 0.0008, + "step": 55843 + }, + { + "epoch": 7.32, + "learning_rate": 3.8189298148701313e-07, + "loss": 0.0002, + "step": 55844 + }, + { + "epoch": 7.32, + "learning_rate": 3.8174783682301096e-07, + "loss": 0.0004, + "step": 55845 + }, + { + "epoch": 7.32, + "learning_rate": 3.816027192098315e-07, + "loss": 0.0007, + "step": 55846 + }, + { + "epoch": 7.32, + "learning_rate": 3.8145762864788327e-07, + "loss": 0.0004, + "step": 55847 + }, + { + "epoch": 7.32, + "learning_rate": 3.8131256513757044e-07, + "loss": 0.0003, + "step": 55848 + }, + { + "epoch": 7.32, + "learning_rate": 3.8116752867930486e-07, + "loss": 0.0004, + "step": 55849 + }, + { + "epoch": 7.32, + "learning_rate": 3.810225192734951e-07, + "loss": 0.0008, + "step": 55850 + }, + { + "epoch": 7.32, + "learning_rate": 3.808775369205464e-07, + "loss": 0.0005, + "step": 55851 + }, + { + "epoch": 7.32, + "learning_rate": 3.8073258162086736e-07, + "loss": 0.0004, + "step": 55852 + }, + { + "epoch": 7.32, + "learning_rate": 3.8058765337486423e-07, + "loss": 0.0008, + "step": 55853 + }, + { + "epoch": 7.32, + "learning_rate": 3.804427521829479e-07, + "loss": 0.0002, + "step": 55854 + }, + { + "epoch": 7.32, + "learning_rate": 3.8029787804552465e-07, + "loss": 0.0004, + "step": 55855 + }, + { + "epoch": 7.32, + "learning_rate": 3.801530309629997e-07, + "loss": 0.0007, + "step": 55856 + }, + { + "epoch": 7.32, + "learning_rate": 3.8000821093578166e-07, + "loss": 0.0006, + "step": 55857 + }, + { + "epoch": 7.32, + "learning_rate": 3.7986341796428017e-07, + "loss": 0.0004, + "step": 55858 + }, + { + "epoch": 7.32, + "learning_rate": 3.7971865204889825e-07, + "loss": 0.0004, + "step": 55859 + }, + { + "epoch": 7.32, + "learning_rate": 3.795739131900467e-07, + "loss": 0.0002, + "step": 55860 + }, + { + "epoch": 7.32, + "learning_rate": 3.794292013881307e-07, + "loss": 0.0009, + "step": 55861 + }, + { + "epoch": 7.32, + "learning_rate": 3.792845166435577e-07, + "loss": 0.0005, + "step": 55862 + }, + { + "epoch": 7.32, + "learning_rate": 3.791398589567363e-07, + "loss": 0.0009, + "step": 55863 + }, + { + "epoch": 7.32, + "learning_rate": 3.7899522832806957e-07, + "loss": 0.0006, + "step": 55864 + }, + { + "epoch": 7.32, + "learning_rate": 3.788506247579671e-07, + "loss": 0.0003, + "step": 55865 + }, + { + "epoch": 7.32, + "learning_rate": 3.787060482468352e-07, + "loss": 0.0006, + "step": 55866 + }, + { + "epoch": 7.32, + "learning_rate": 3.785614987950803e-07, + "loss": 0.001, + "step": 55867 + }, + { + "epoch": 7.32, + "learning_rate": 3.784169764031087e-07, + "loss": 0.0004, + "step": 55868 + }, + { + "epoch": 7.32, + "learning_rate": 3.7827248107132675e-07, + "loss": 0.0002, + "step": 55869 + }, + { + "epoch": 7.32, + "learning_rate": 3.781280128001408e-07, + "loss": 0.0004, + "step": 55870 + }, + { + "epoch": 7.32, + "learning_rate": 3.779835715899571e-07, + "loss": 0.0002, + "step": 55871 + }, + { + "epoch": 7.32, + "learning_rate": 3.7783915744118215e-07, + "loss": 0.0009, + "step": 55872 + }, + { + "epoch": 7.32, + "learning_rate": 3.776947703542233e-07, + "loss": 0.0003, + "step": 55873 + }, + { + "epoch": 7.32, + "learning_rate": 3.775504103294858e-07, + "loss": 0.0003, + "step": 55874 + }, + { + "epoch": 7.32, + "learning_rate": 3.7740607736737486e-07, + "loss": 0.0004, + "step": 55875 + }, + { + "epoch": 7.32, + "learning_rate": 3.7726177146829577e-07, + "loss": 0.0004, + "step": 55876 + }, + { + "epoch": 7.32, + "learning_rate": 3.7711749263265705e-07, + "loss": 0.0006, + "step": 55877 + }, + { + "epoch": 7.32, + "learning_rate": 3.7697324086086285e-07, + "loss": 0.0005, + "step": 55878 + }, + { + "epoch": 7.32, + "learning_rate": 3.7682901615331725e-07, + "loss": 0.0005, + "step": 55879 + }, + { + "epoch": 7.32, + "learning_rate": 3.766848185104288e-07, + "loss": 0.0007, + "step": 55880 + }, + { + "epoch": 7.32, + "learning_rate": 3.765406479326006e-07, + "loss": 0.0011, + "step": 55881 + }, + { + "epoch": 7.32, + "learning_rate": 3.7639650442024114e-07, + "loss": 0.0006, + "step": 55882 + }, + { + "epoch": 7.32, + "learning_rate": 3.762523879737534e-07, + "loss": 0.0004, + "step": 55883 + }, + { + "epoch": 7.32, + "learning_rate": 3.761082985935427e-07, + "loss": 0.0008, + "step": 55884 + }, + { + "epoch": 7.32, + "learning_rate": 3.7596423628001644e-07, + "loss": 0.0004, + "step": 55885 + }, + { + "epoch": 7.32, + "learning_rate": 3.7582020103357763e-07, + "loss": 0.0002, + "step": 55886 + }, + { + "epoch": 7.32, + "learning_rate": 3.756761928546304e-07, + "loss": 0.0002, + "step": 55887 + }, + { + "epoch": 7.32, + "learning_rate": 3.7553221174358333e-07, + "loss": 0.0005, + "step": 55888 + }, + { + "epoch": 7.32, + "learning_rate": 3.753882577008383e-07, + "loss": 0.0008, + "step": 55889 + }, + { + "epoch": 7.32, + "learning_rate": 3.7524433072680276e-07, + "loss": 0.0004, + "step": 55890 + }, + { + "epoch": 7.32, + "learning_rate": 3.7510043082187754e-07, + "loss": 0.0012, + "step": 55891 + }, + { + "epoch": 7.32, + "learning_rate": 3.7495655798647e-07, + "loss": 0.0009, + "step": 55892 + }, + { + "epoch": 7.32, + "learning_rate": 3.748127122209866e-07, + "loss": 0.0006, + "step": 55893 + }, + { + "epoch": 7.32, + "learning_rate": 3.746688935258291e-07, + "loss": 0.0006, + "step": 55894 + }, + { + "epoch": 7.32, + "learning_rate": 3.745251019014018e-07, + "loss": 0.0007, + "step": 55895 + }, + { + "epoch": 7.32, + "learning_rate": 3.743813373481109e-07, + "loss": 0.0002, + "step": 55896 + }, + { + "epoch": 7.32, + "learning_rate": 3.7423759986636057e-07, + "loss": 0.0004, + "step": 55897 + }, + { + "epoch": 7.32, + "learning_rate": 3.740938894565538e-07, + "loss": 0.0003, + "step": 55898 + }, + { + "epoch": 7.32, + "learning_rate": 3.7395020611909473e-07, + "loss": 0.0006, + "step": 55899 + }, + { + "epoch": 7.32, + "learning_rate": 3.738065498543886e-07, + "loss": 0.0006, + "step": 55900 + }, + { + "epoch": 7.32, + "learning_rate": 3.7366292066283947e-07, + "loss": 0.0007, + "step": 55901 + }, + { + "epoch": 7.32, + "learning_rate": 3.735193185448494e-07, + "loss": 0.0005, + "step": 55902 + }, + { + "epoch": 7.32, + "learning_rate": 3.733757435008234e-07, + "loss": 0.0006, + "step": 55903 + }, + { + "epoch": 7.32, + "learning_rate": 3.732321955311668e-07, + "loss": 0.0004, + "step": 55904 + }, + { + "epoch": 7.32, + "learning_rate": 3.7308867463628275e-07, + "loss": 0.0005, + "step": 55905 + }, + { + "epoch": 7.32, + "learning_rate": 3.7294518081657295e-07, + "loss": 0.0007, + "step": 55906 + }, + { + "epoch": 7.32, + "learning_rate": 3.7280171407244157e-07, + "loss": 0.0004, + "step": 55907 + }, + { + "epoch": 7.32, + "learning_rate": 3.726582744042939e-07, + "loss": 0.0002, + "step": 55908 + }, + { + "epoch": 7.32, + "learning_rate": 3.725148618125318e-07, + "loss": 0.0006, + "step": 55909 + }, + { + "epoch": 7.32, + "learning_rate": 3.7237147629755833e-07, + "loss": 0.0008, + "step": 55910 + }, + { + "epoch": 7.32, + "learning_rate": 3.722281178597775e-07, + "loss": 0.0005, + "step": 55911 + }, + { + "epoch": 7.32, + "learning_rate": 3.7208478649959356e-07, + "loss": 0.001, + "step": 55912 + }, + { + "epoch": 7.32, + "learning_rate": 3.7194148221740833e-07, + "loss": 0.0016, + "step": 55913 + }, + { + "epoch": 7.32, + "learning_rate": 3.7179820501362154e-07, + "loss": 0.0008, + "step": 55914 + }, + { + "epoch": 7.32, + "learning_rate": 3.716549548886428e-07, + "loss": 0.0003, + "step": 55915 + }, + { + "epoch": 7.32, + "learning_rate": 3.715117318428718e-07, + "loss": 0.001, + "step": 55916 + }, + { + "epoch": 7.32, + "learning_rate": 3.713685358767116e-07, + "loss": 0.0003, + "step": 55917 + }, + { + "epoch": 7.32, + "learning_rate": 3.712253669905619e-07, + "loss": 0.0005, + "step": 55918 + }, + { + "epoch": 7.32, + "learning_rate": 3.7108222518483115e-07, + "loss": 0.0004, + "step": 55919 + }, + { + "epoch": 7.33, + "learning_rate": 3.7093911045991804e-07, + "loss": 0.0003, + "step": 55920 + }, + { + "epoch": 7.33, + "learning_rate": 3.707960228162255e-07, + "loss": 0.0003, + "step": 55921 + }, + { + "epoch": 7.33, + "learning_rate": 3.7065296225415546e-07, + "loss": 0.0002, + "step": 55922 + }, + { + "epoch": 7.33, + "learning_rate": 3.7050992877411316e-07, + "loss": 0.0007, + "step": 55923 + }, + { + "epoch": 7.33, + "learning_rate": 3.703669223764983e-07, + "loss": 0.0003, + "step": 55924 + }, + { + "epoch": 7.33, + "learning_rate": 3.702239430617116e-07, + "loss": 0.0004, + "step": 55925 + }, + { + "epoch": 7.33, + "learning_rate": 3.7008099083015834e-07, + "loss": 0.0005, + "step": 55926 + }, + { + "epoch": 7.33, + "learning_rate": 3.699380656822404e-07, + "loss": 0.0003, + "step": 55927 + }, + { + "epoch": 7.33, + "learning_rate": 3.6979516761835865e-07, + "loss": 0.0003, + "step": 55928 + }, + { + "epoch": 7.33, + "learning_rate": 3.6965229663891377e-07, + "loss": 0.0007, + "step": 55929 + }, + { + "epoch": 7.33, + "learning_rate": 3.6950945274431104e-07, + "loss": 0.0004, + "step": 55930 + }, + { + "epoch": 7.33, + "learning_rate": 3.6936663593494904e-07, + "loss": 0.0015, + "step": 55931 + }, + { + "epoch": 7.33, + "learning_rate": 3.6922384621123185e-07, + "loss": 0.0007, + "step": 55932 + }, + { + "epoch": 7.33, + "learning_rate": 3.69081083573557e-07, + "loss": 0.0002, + "step": 55933 + }, + { + "epoch": 7.33, + "learning_rate": 3.6893834802233076e-07, + "loss": 0.0004, + "step": 55934 + }, + { + "epoch": 7.33, + "learning_rate": 3.687956395579528e-07, + "loss": 0.0004, + "step": 55935 + }, + { + "epoch": 7.33, + "learning_rate": 3.6865295818082183e-07, + "loss": 0.0007, + "step": 55936 + }, + { + "epoch": 7.33, + "learning_rate": 3.685103038913429e-07, + "loss": 0.0007, + "step": 55937 + }, + { + "epoch": 7.33, + "learning_rate": 3.683676766899158e-07, + "loss": 0.0002, + "step": 55938 + }, + { + "epoch": 7.33, + "learning_rate": 3.682250765769424e-07, + "loss": 0.0004, + "step": 55939 + }, + { + "epoch": 7.33, + "learning_rate": 3.680825035528235e-07, + "loss": 0.0004, + "step": 55940 + }, + { + "epoch": 7.33, + "learning_rate": 3.6793995761795765e-07, + "loss": 0.0002, + "step": 55941 + }, + { + "epoch": 7.33, + "learning_rate": 3.67797438772749e-07, + "loss": 0.0006, + "step": 55942 + }, + { + "epoch": 7.33, + "learning_rate": 3.676549470175972e-07, + "loss": 0.0003, + "step": 55943 + }, + { + "epoch": 7.33, + "learning_rate": 3.675124823529019e-07, + "loss": 0.0007, + "step": 55944 + }, + { + "epoch": 7.33, + "learning_rate": 3.673700447790662e-07, + "loss": 0.0003, + "step": 55945 + }, + { + "epoch": 7.33, + "learning_rate": 3.6722763429648867e-07, + "loss": 0.0003, + "step": 55946 + }, + { + "epoch": 7.33, + "learning_rate": 3.6708525090557003e-07, + "loss": 0.0005, + "step": 55947 + }, + { + "epoch": 7.33, + "learning_rate": 3.6694289460671106e-07, + "loss": 0.0008, + "step": 55948 + }, + { + "epoch": 7.33, + "learning_rate": 3.668005654003126e-07, + "loss": 0.0006, + "step": 55949 + }, + { + "epoch": 7.33, + "learning_rate": 3.6665826328677436e-07, + "loss": 0.0006, + "step": 55950 + }, + { + "epoch": 7.33, + "learning_rate": 3.6651598826649703e-07, + "loss": 0.0004, + "step": 55951 + }, + { + "epoch": 7.33, + "learning_rate": 3.6637374033988037e-07, + "loss": 0.0002, + "step": 55952 + }, + { + "epoch": 7.33, + "learning_rate": 3.6623151950732516e-07, + "loss": 0.0002, + "step": 55953 + }, + { + "epoch": 7.33, + "learning_rate": 3.6608932576922996e-07, + "loss": 0.0002, + "step": 55954 + }, + { + "epoch": 7.33, + "learning_rate": 3.6594715912599666e-07, + "loss": 0.0015, + "step": 55955 + }, + { + "epoch": 7.33, + "learning_rate": 3.6580501957802164e-07, + "loss": 0.0007, + "step": 55956 + }, + { + "epoch": 7.33, + "learning_rate": 3.6566290712570897e-07, + "loss": 0.0013, + "step": 55957 + }, + { + "epoch": 7.33, + "learning_rate": 3.65520821769455e-07, + "loss": 0.0004, + "step": 55958 + }, + { + "epoch": 7.33, + "learning_rate": 3.653787635096617e-07, + "loss": 0.0011, + "step": 55959 + }, + { + "epoch": 7.33, + "learning_rate": 3.6523673234672543e-07, + "loss": 0.0005, + "step": 55960 + }, + { + "epoch": 7.33, + "learning_rate": 3.650947282810502e-07, + "loss": 0.0003, + "step": 55961 + }, + { + "epoch": 7.33, + "learning_rate": 3.6495275131303243e-07, + "loss": 0.0003, + "step": 55962 + }, + { + "epoch": 7.33, + "learning_rate": 3.648108014430729e-07, + "loss": 0.0004, + "step": 55963 + }, + { + "epoch": 7.33, + "learning_rate": 3.6466887867156796e-07, + "loss": 0.0008, + "step": 55964 + }, + { + "epoch": 7.33, + "learning_rate": 3.645269829989195e-07, + "loss": 0.0009, + "step": 55965 + }, + { + "epoch": 7.33, + "learning_rate": 3.6438511442552617e-07, + "loss": 0.0004, + "step": 55966 + }, + { + "epoch": 7.33, + "learning_rate": 3.642432729517853e-07, + "loss": 0.0011, + "step": 55967 + }, + { + "epoch": 7.33, + "learning_rate": 3.6410145857809774e-07, + "loss": 0.0003, + "step": 55968 + }, + { + "epoch": 7.33, + "learning_rate": 3.639596713048621e-07, + "loss": 0.0009, + "step": 55969 + }, + { + "epoch": 7.33, + "learning_rate": 3.638179111324769e-07, + "loss": 0.0002, + "step": 55970 + }, + { + "epoch": 7.33, + "learning_rate": 3.636761780613407e-07, + "loss": 0.0003, + "step": 55971 + }, + { + "epoch": 7.33, + "learning_rate": 3.63534472091851e-07, + "loss": 0.0003, + "step": 55972 + }, + { + "epoch": 7.33, + "learning_rate": 3.6339279322440745e-07, + "loss": 0.0007, + "step": 55973 + }, + { + "epoch": 7.33, + "learning_rate": 3.6325114145940974e-07, + "loss": 0.0005, + "step": 55974 + }, + { + "epoch": 7.33, + "learning_rate": 3.6310951679725427e-07, + "loss": 0.0006, + "step": 55975 + }, + { + "epoch": 7.33, + "learning_rate": 3.629679192383406e-07, + "loss": 0.0006, + "step": 55976 + }, + { + "epoch": 7.33, + "learning_rate": 3.628263487830663e-07, + "loss": 0.0004, + "step": 55977 + }, + { + "epoch": 7.33, + "learning_rate": 3.626848054318299e-07, + "loss": 0.0005, + "step": 55978 + }, + { + "epoch": 7.33, + "learning_rate": 3.625432891850267e-07, + "loss": 0.0004, + "step": 55979 + }, + { + "epoch": 7.33, + "learning_rate": 3.624018000430596e-07, + "loss": 0.0003, + "step": 55980 + }, + { + "epoch": 7.33, + "learning_rate": 3.62260338006325e-07, + "loss": 0.0004, + "step": 55981 + }, + { + "epoch": 7.33, + "learning_rate": 3.6211890307521923e-07, + "loss": 0.0003, + "step": 55982 + }, + { + "epoch": 7.33, + "learning_rate": 3.619774952501398e-07, + "loss": 0.0005, + "step": 55983 + }, + { + "epoch": 7.33, + "learning_rate": 3.618361145314864e-07, + "loss": 0.001, + "step": 55984 + }, + { + "epoch": 7.33, + "learning_rate": 3.616947609196553e-07, + "loss": 0.001, + "step": 55985 + }, + { + "epoch": 7.33, + "learning_rate": 3.615534344150451e-07, + "loss": 0.0009, + "step": 55986 + }, + { + "epoch": 7.33, + "learning_rate": 3.6141213501805103e-07, + "loss": 0.0003, + "step": 55987 + }, + { + "epoch": 7.33, + "learning_rate": 3.61270862729074e-07, + "loss": 0.001, + "step": 55988 + }, + { + "epoch": 7.33, + "learning_rate": 3.6112961754850904e-07, + "loss": 0.0006, + "step": 55989 + }, + { + "epoch": 7.33, + "learning_rate": 3.6098839947675156e-07, + "loss": 0.0002, + "step": 55990 + }, + { + "epoch": 7.33, + "learning_rate": 3.608472085142023e-07, + "loss": 0.0006, + "step": 55991 + }, + { + "epoch": 7.33, + "learning_rate": 3.607060446612576e-07, + "loss": 0.0008, + "step": 55992 + }, + { + "epoch": 7.33, + "learning_rate": 3.605649079183138e-07, + "loss": 0.001, + "step": 55993 + }, + { + "epoch": 7.33, + "learning_rate": 3.604237982857661e-07, + "loss": 0.0003, + "step": 55994 + }, + { + "epoch": 7.33, + "learning_rate": 3.602827157640143e-07, + "loss": 0.0003, + "step": 55995 + }, + { + "epoch": 7.34, + "learning_rate": 3.6014166035345465e-07, + "loss": 0.0006, + "step": 55996 + }, + { + "epoch": 7.34, + "learning_rate": 3.6000063205448355e-07, + "loss": 0.0004, + "step": 55997 + }, + { + "epoch": 7.34, + "learning_rate": 3.5985963086749507e-07, + "loss": 0.0005, + "step": 55998 + }, + { + "epoch": 7.34, + "learning_rate": 3.5971865679289006e-07, + "loss": 0.0002, + "step": 55999 + }, + { + "epoch": 7.34, + "learning_rate": 3.5957770983106267e-07, + "loss": 0.0004, + "step": 56000 + }, + { + "epoch": 7.34, + "learning_rate": 3.594367899824092e-07, + "loss": 0.0004, + "step": 56001 + }, + { + "epoch": 7.34, + "learning_rate": 3.592958972473248e-07, + "loss": 0.0002, + "step": 56002 + }, + { + "epoch": 7.34, + "learning_rate": 3.5915503162620936e-07, + "loss": 0.0007, + "step": 56003 + }, + { + "epoch": 7.34, + "learning_rate": 3.5901419311945684e-07, + "loss": 0.0005, + "step": 56004 + }, + { + "epoch": 7.34, + "learning_rate": 3.588733817274637e-07, + "loss": 0.0002, + "step": 56005 + }, + { + "epoch": 7.34, + "learning_rate": 3.5873259745062506e-07, + "loss": 0.0005, + "step": 56006 + }, + { + "epoch": 7.34, + "learning_rate": 3.5859184028933734e-07, + "loss": 0.0005, + "step": 56007 + }, + { + "epoch": 7.34, + "learning_rate": 3.5845111024399804e-07, + "loss": 0.0007, + "step": 56008 + }, + { + "epoch": 7.34, + "learning_rate": 3.58310407314999e-07, + "loss": 0.0004, + "step": 56009 + }, + { + "epoch": 7.34, + "learning_rate": 3.58169731502741e-07, + "loss": 0.0002, + "step": 56010 + }, + { + "epoch": 7.34, + "learning_rate": 3.5802908280761605e-07, + "loss": 0.0005, + "step": 56011 + }, + { + "epoch": 7.34, + "learning_rate": 3.578884612300204e-07, + "loss": 0.0005, + "step": 56012 + }, + { + "epoch": 7.34, + "learning_rate": 3.5774786677035044e-07, + "loss": 0.0003, + "step": 56013 + }, + { + "epoch": 7.34, + "learning_rate": 3.576072994290014e-07, + "loss": 0.0006, + "step": 56014 + }, + { + "epoch": 7.34, + "learning_rate": 3.5746675920636855e-07, + "loss": 0.0005, + "step": 56015 + }, + { + "epoch": 7.34, + "learning_rate": 3.5732624610284703e-07, + "loss": 0.0012, + "step": 56016 + }, + { + "epoch": 7.34, + "learning_rate": 3.5718576011883e-07, + "loss": 0.0002, + "step": 56017 + }, + { + "epoch": 7.34, + "learning_rate": 3.5704530125471703e-07, + "loss": 0.0012, + "step": 56018 + }, + { + "epoch": 7.34, + "learning_rate": 3.569048695109001e-07, + "loss": 0.0002, + "step": 56019 + }, + { + "epoch": 7.34, + "learning_rate": 3.567644648877744e-07, + "loss": 0.0003, + "step": 56020 + }, + { + "epoch": 7.34, + "learning_rate": 3.5662408738573407e-07, + "loss": 0.0003, + "step": 56021 + }, + { + "epoch": 7.34, + "learning_rate": 3.5648373700517545e-07, + "loss": 0.0004, + "step": 56022 + }, + { + "epoch": 7.34, + "learning_rate": 3.5634341374649275e-07, + "loss": 0.0003, + "step": 56023 + }, + { + "epoch": 7.34, + "learning_rate": 3.562031176100811e-07, + "loss": 0.0003, + "step": 56024 + }, + { + "epoch": 7.34, + "learning_rate": 3.5606284859633357e-07, + "loss": 0.0004, + "step": 56025 + }, + { + "epoch": 7.34, + "learning_rate": 3.559226067056476e-07, + "loss": 0.0005, + "step": 56026 + }, + { + "epoch": 7.34, + "learning_rate": 3.557823919384151e-07, + "loss": 0.0002, + "step": 56027 + }, + { + "epoch": 7.34, + "learning_rate": 3.556422042950314e-07, + "loss": 0.0007, + "step": 56028 + }, + { + "epoch": 7.34, + "learning_rate": 3.5550204377588824e-07, + "loss": 0.001, + "step": 56029 + }, + { + "epoch": 7.34, + "learning_rate": 3.5536191038138433e-07, + "loss": 0.0006, + "step": 56030 + }, + { + "epoch": 7.34, + "learning_rate": 3.552218041119104e-07, + "loss": 0.0003, + "step": 56031 + }, + { + "epoch": 7.34, + "learning_rate": 3.5508172496786175e-07, + "loss": 0.0006, + "step": 56032 + }, + { + "epoch": 7.34, + "learning_rate": 3.5494167294963246e-07, + "loss": 0.0002, + "step": 56033 + }, + { + "epoch": 7.34, + "learning_rate": 3.548016480576144e-07, + "loss": 0.0004, + "step": 56034 + }, + { + "epoch": 7.34, + "learning_rate": 3.546616502922051e-07, + "loss": 0.0006, + "step": 56035 + }, + { + "epoch": 7.34, + "learning_rate": 3.545216796537954e-07, + "loss": 0.0003, + "step": 56036 + }, + { + "epoch": 7.34, + "learning_rate": 3.543817361427793e-07, + "loss": 0.0008, + "step": 56037 + }, + { + "epoch": 7.34, + "learning_rate": 3.542418197595521e-07, + "loss": 0.0005, + "step": 56038 + }, + { + "epoch": 7.34, + "learning_rate": 3.541019305045057e-07, + "loss": 0.0003, + "step": 56039 + }, + { + "epoch": 7.34, + "learning_rate": 3.5396206837803315e-07, + "loss": 0.0002, + "step": 56040 + }, + { + "epoch": 7.34, + "learning_rate": 3.538222333805297e-07, + "loss": 0.0005, + "step": 56041 + }, + { + "epoch": 7.34, + "learning_rate": 3.536824255123883e-07, + "loss": 0.0005, + "step": 56042 + }, + { + "epoch": 7.34, + "learning_rate": 3.535426447739998e-07, + "loss": 0.0006, + "step": 56043 + }, + { + "epoch": 7.34, + "learning_rate": 3.5340289116575946e-07, + "loss": 0.0004, + "step": 56044 + }, + { + "epoch": 7.34, + "learning_rate": 3.53263164688058e-07, + "loss": 0.001, + "step": 56045 + }, + { + "epoch": 7.34, + "learning_rate": 3.531234653412929e-07, + "loss": 0.0004, + "step": 56046 + }, + { + "epoch": 7.34, + "learning_rate": 3.5298379312585287e-07, + "loss": 0.0004, + "step": 56047 + }, + { + "epoch": 7.34, + "learning_rate": 3.5284414804213183e-07, + "loss": 0.0009, + "step": 56048 + }, + { + "epoch": 7.34, + "learning_rate": 3.527045300905241e-07, + "loss": 0.0006, + "step": 56049 + }, + { + "epoch": 7.34, + "learning_rate": 3.5256493927142144e-07, + "loss": 0.0003, + "step": 56050 + }, + { + "epoch": 7.34, + "learning_rate": 3.5242537558521474e-07, + "loss": 0.0003, + "step": 56051 + }, + { + "epoch": 7.34, + "learning_rate": 3.522858390322981e-07, + "loss": 0.0001, + "step": 56052 + }, + { + "epoch": 7.34, + "learning_rate": 3.521463296130645e-07, + "loss": 0.0003, + "step": 56053 + }, + { + "epoch": 7.34, + "learning_rate": 3.520068473279059e-07, + "loss": 0.0005, + "step": 56054 + }, + { + "epoch": 7.34, + "learning_rate": 3.518673921772131e-07, + "loss": 0.0003, + "step": 56055 + }, + { + "epoch": 7.34, + "learning_rate": 3.517279641613791e-07, + "loss": 0.0008, + "step": 56056 + }, + { + "epoch": 7.34, + "learning_rate": 3.5158856328079694e-07, + "loss": 0.0005, + "step": 56057 + }, + { + "epoch": 7.34, + "learning_rate": 3.5144918953585963e-07, + "loss": 0.0003, + "step": 56058 + }, + { + "epoch": 7.34, + "learning_rate": 3.513098429269557e-07, + "loss": 0.0003, + "step": 56059 + }, + { + "epoch": 7.34, + "learning_rate": 3.5117052345448043e-07, + "loss": 0.0004, + "step": 56060 + }, + { + "epoch": 7.34, + "learning_rate": 3.5103123111882353e-07, + "loss": 0.0004, + "step": 56061 + }, + { + "epoch": 7.34, + "learning_rate": 3.50891965920378e-07, + "loss": 0.001, + "step": 56062 + }, + { + "epoch": 7.34, + "learning_rate": 3.507527278595335e-07, + "loss": 0.0011, + "step": 56063 + }, + { + "epoch": 7.34, + "learning_rate": 3.506135169366842e-07, + "loss": 0.0013, + "step": 56064 + }, + { + "epoch": 7.34, + "learning_rate": 3.504743331522209e-07, + "loss": 0.0004, + "step": 56065 + }, + { + "epoch": 7.34, + "learning_rate": 3.5033517650653324e-07, + "loss": 0.0006, + "step": 56066 + }, + { + "epoch": 7.34, + "learning_rate": 3.501960470000143e-07, + "loss": 0.0003, + "step": 56067 + }, + { + "epoch": 7.34, + "learning_rate": 3.5005694463305594e-07, + "loss": 0.0004, + "step": 56068 + }, + { + "epoch": 7.34, + "learning_rate": 3.49917869406049e-07, + "loss": 0.0002, + "step": 56069 + }, + { + "epoch": 7.34, + "learning_rate": 3.497788213193842e-07, + "loss": 0.0007, + "step": 56070 + }, + { + "epoch": 7.34, + "learning_rate": 3.4963980037345134e-07, + "loss": 0.0006, + "step": 56071 + }, + { + "epoch": 7.35, + "learning_rate": 3.495008065686434e-07, + "loss": 0.0003, + "step": 56072 + }, + { + "epoch": 7.35, + "learning_rate": 3.4936183990535e-07, + "loss": 0.0003, + "step": 56073 + }, + { + "epoch": 7.35, + "learning_rate": 3.49222900383962e-07, + "loss": 0.0017, + "step": 56074 + }, + { + "epoch": 7.35, + "learning_rate": 3.490839880048724e-07, + "loss": 0.0003, + "step": 56075 + }, + { + "epoch": 7.35, + "learning_rate": 3.489451027684698e-07, + "loss": 0.0004, + "step": 56076 + }, + { + "epoch": 7.35, + "learning_rate": 3.48806244675145e-07, + "loss": 0.0004, + "step": 56077 + }, + { + "epoch": 7.35, + "learning_rate": 3.4866741372528657e-07, + "loss": 0.0006, + "step": 56078 + }, + { + "epoch": 7.35, + "learning_rate": 3.485286099192886e-07, + "loss": 0.001, + "step": 56079 + }, + { + "epoch": 7.35, + "learning_rate": 3.483898332575397e-07, + "loss": 0.0005, + "step": 56080 + }, + { + "epoch": 7.35, + "learning_rate": 3.482510837404318e-07, + "loss": 0.0004, + "step": 56081 + }, + { + "epoch": 7.35, + "learning_rate": 3.4811236136835123e-07, + "loss": 0.0002, + "step": 56082 + }, + { + "epoch": 7.35, + "learning_rate": 3.479736661416933e-07, + "loss": 0.0004, + "step": 56083 + }, + { + "epoch": 7.35, + "learning_rate": 3.478349980608442e-07, + "loss": 0.0002, + "step": 56084 + }, + { + "epoch": 7.35, + "learning_rate": 3.47696357126196e-07, + "loss": 0.0005, + "step": 56085 + }, + { + "epoch": 7.35, + "learning_rate": 3.4755774333813607e-07, + "loss": 0.0001, + "step": 56086 + }, + { + "epoch": 7.35, + "learning_rate": 3.474191566970575e-07, + "loss": 0.0003, + "step": 56087 + }, + { + "epoch": 7.35, + "learning_rate": 3.4728059720334775e-07, + "loss": 0.0003, + "step": 56088 + }, + { + "epoch": 7.35, + "learning_rate": 3.4714206485739644e-07, + "loss": 0.0006, + "step": 56089 + }, + { + "epoch": 7.35, + "learning_rate": 3.470035596595944e-07, + "loss": 0.0006, + "step": 56090 + }, + { + "epoch": 7.35, + "learning_rate": 3.468650816103325e-07, + "loss": 0.0002, + "step": 56091 + }, + { + "epoch": 7.35, + "learning_rate": 3.4672663070999703e-07, + "loss": 0.0005, + "step": 56092 + }, + { + "epoch": 7.35, + "learning_rate": 3.465882069589799e-07, + "loss": 0.0004, + "step": 56093 + }, + { + "epoch": 7.35, + "learning_rate": 3.4644981035766854e-07, + "loss": 0.0003, + "step": 56094 + }, + { + "epoch": 7.35, + "learning_rate": 3.4631144090645275e-07, + "loss": 0.0003, + "step": 56095 + }, + { + "epoch": 7.35, + "learning_rate": 3.461730986057232e-07, + "loss": 0.0002, + "step": 56096 + }, + { + "epoch": 7.35, + "learning_rate": 3.460347834558664e-07, + "loss": 0.0005, + "step": 56097 + }, + { + "epoch": 7.35, + "learning_rate": 3.45896495457273e-07, + "loss": 0.0002, + "step": 56098 + }, + { + "epoch": 7.35, + "learning_rate": 3.4575823461033276e-07, + "loss": 0.0006, + "step": 56099 + }, + { + "epoch": 7.35, + "learning_rate": 3.456200009154309e-07, + "loss": 0.0005, + "step": 56100 + }, + { + "epoch": 7.35, + "learning_rate": 3.4548179437295934e-07, + "loss": 0.002, + "step": 56101 + }, + { + "epoch": 7.35, + "learning_rate": 3.453436149833078e-07, + "loss": 0.0007, + "step": 56102 + }, + { + "epoch": 7.35, + "learning_rate": 3.4520546274686263e-07, + "loss": 0.0011, + "step": 56103 + }, + { + "epoch": 7.35, + "learning_rate": 3.4506733766401235e-07, + "loss": 0.0005, + "step": 56104 + }, + { + "epoch": 7.35, + "learning_rate": 3.4492923973514556e-07, + "loss": 0.0002, + "step": 56105 + }, + { + "epoch": 7.35, + "learning_rate": 3.44791168960652e-07, + "loss": 0.0001, + "step": 56106 + }, + { + "epoch": 7.35, + "learning_rate": 3.4465312534092024e-07, + "loss": 0.0006, + "step": 56107 + }, + { + "epoch": 7.35, + "learning_rate": 3.4451510887633653e-07, + "loss": 0.0006, + "step": 56108 + }, + { + "epoch": 7.35, + "learning_rate": 3.4437711956728845e-07, + "loss": 0.0002, + "step": 56109 + }, + { + "epoch": 7.35, + "learning_rate": 3.442391574141657e-07, + "loss": 0.0004, + "step": 56110 + }, + { + "epoch": 7.35, + "learning_rate": 3.4410122241735675e-07, + "loss": 0.0007, + "step": 56111 + }, + { + "epoch": 7.35, + "learning_rate": 3.439633145772503e-07, + "loss": 0.0015, + "step": 56112 + }, + { + "epoch": 7.35, + "learning_rate": 3.438254338942304e-07, + "loss": 0.0004, + "step": 56113 + }, + { + "epoch": 7.35, + "learning_rate": 3.4368758036868897e-07, + "loss": 0.0004, + "step": 56114 + }, + { + "epoch": 7.35, + "learning_rate": 3.4354975400101134e-07, + "loss": 0.0004, + "step": 56115 + }, + { + "epoch": 7.35, + "learning_rate": 3.4341195479158593e-07, + "loss": 0.0005, + "step": 56116 + }, + { + "epoch": 7.35, + "learning_rate": 3.4327418274080036e-07, + "loss": 0.0005, + "step": 56117 + }, + { + "epoch": 7.35, + "learning_rate": 3.4313643784904204e-07, + "loss": 0.0003, + "step": 56118 + }, + { + "epoch": 7.35, + "learning_rate": 3.4299872011669843e-07, + "loss": 0.0003, + "step": 56119 + }, + { + "epoch": 7.35, + "learning_rate": 3.428610295441548e-07, + "loss": 0.0004, + "step": 56120 + }, + { + "epoch": 7.35, + "learning_rate": 3.427233661318008e-07, + "loss": 0.0002, + "step": 56121 + }, + { + "epoch": 7.35, + "learning_rate": 3.4258572988002506e-07, + "loss": 0.0005, + "step": 56122 + }, + { + "epoch": 7.35, + "learning_rate": 3.4244812078921163e-07, + "loss": 0.0004, + "step": 56123 + }, + { + "epoch": 7.35, + "learning_rate": 3.4231053885974696e-07, + "loss": 0.0004, + "step": 56124 + }, + { + "epoch": 7.35, + "learning_rate": 3.4217298409202184e-07, + "loss": 0.0005, + "step": 56125 + }, + { + "epoch": 7.35, + "learning_rate": 3.4203545648642036e-07, + "loss": 0.0004, + "step": 56126 + }, + { + "epoch": 7.35, + "learning_rate": 3.4189795604333e-07, + "loss": 0.0002, + "step": 56127 + }, + { + "epoch": 7.35, + "learning_rate": 3.41760482763136e-07, + "loss": 0.0003, + "step": 56128 + }, + { + "epoch": 7.35, + "learning_rate": 3.4162303664622697e-07, + "loss": 0.0005, + "step": 56129 + }, + { + "epoch": 7.35, + "learning_rate": 3.4148561769298926e-07, + "loss": 0.0017, + "step": 56130 + }, + { + "epoch": 7.35, + "learning_rate": 3.413482259038081e-07, + "loss": 0.0003, + "step": 56131 + }, + { + "epoch": 7.35, + "learning_rate": 3.41210861279071e-07, + "loss": 0.0002, + "step": 56132 + }, + { + "epoch": 7.35, + "learning_rate": 3.4107352381916423e-07, + "loss": 0.0004, + "step": 56133 + }, + { + "epoch": 7.35, + "learning_rate": 3.4093621352447423e-07, + "loss": 0.0003, + "step": 56134 + }, + { + "epoch": 7.35, + "learning_rate": 3.407989303953874e-07, + "loss": 0.0003, + "step": 56135 + }, + { + "epoch": 7.35, + "learning_rate": 3.406616744322866e-07, + "loss": 0.0012, + "step": 56136 + }, + { + "epoch": 7.35, + "learning_rate": 3.4052444563556277e-07, + "loss": 0.0002, + "step": 56137 + }, + { + "epoch": 7.35, + "learning_rate": 3.4038724400559884e-07, + "loss": 0.0007, + "step": 56138 + }, + { + "epoch": 7.35, + "learning_rate": 3.4025006954278017e-07, + "loss": 0.0004, + "step": 56139 + }, + { + "epoch": 7.35, + "learning_rate": 3.4011292224749525e-07, + "loss": 0.0008, + "step": 56140 + }, + { + "epoch": 7.35, + "learning_rate": 3.3997580212012825e-07, + "loss": 0.0002, + "step": 56141 + }, + { + "epoch": 7.35, + "learning_rate": 3.398387091610655e-07, + "loss": 0.0004, + "step": 56142 + }, + { + "epoch": 7.35, + "learning_rate": 3.3970164337069013e-07, + "loss": 0.0004, + "step": 56143 + }, + { + "epoch": 7.35, + "learning_rate": 3.3956460474938945e-07, + "loss": 0.0005, + "step": 56144 + }, + { + "epoch": 7.35, + "learning_rate": 3.3942759329754994e-07, + "loss": 0.0005, + "step": 56145 + }, + { + "epoch": 7.35, + "learning_rate": 3.3929060901555677e-07, + "loss": 0.0009, + "step": 56146 + }, + { + "epoch": 7.35, + "learning_rate": 3.391536519037919e-07, + "loss": 0.0004, + "step": 56147 + }, + { + "epoch": 7.35, + "learning_rate": 3.39016721962645e-07, + "loss": 0.0006, + "step": 56148 + }, + { + "epoch": 7.36, + "learning_rate": 3.3887981919249915e-07, + "loss": 0.001, + "step": 56149 + }, + { + "epoch": 7.36, + "learning_rate": 3.387429435937384e-07, + "loss": 0.0003, + "step": 56150 + }, + { + "epoch": 7.36, + "learning_rate": 3.38606095166748e-07, + "loss": 0.0005, + "step": 56151 + }, + { + "epoch": 7.36, + "learning_rate": 3.384692739119144e-07, + "loss": 0.0004, + "step": 56152 + }, + { + "epoch": 7.36, + "learning_rate": 3.383324798296217e-07, + "loss": 0.0005, + "step": 56153 + }, + { + "epoch": 7.36, + "learning_rate": 3.3819571292025286e-07, + "loss": 0.0003, + "step": 56154 + }, + { + "epoch": 7.36, + "learning_rate": 3.3805897318419433e-07, + "loss": 0.0004, + "step": 56155 + }, + { + "epoch": 7.36, + "learning_rate": 3.3792226062183245e-07, + "loss": 0.0004, + "step": 56156 + }, + { + "epoch": 7.36, + "learning_rate": 3.3778557523354795e-07, + "loss": 0.0005, + "step": 56157 + }, + { + "epoch": 7.36, + "learning_rate": 3.3764891701972835e-07, + "loss": 0.0003, + "step": 56158 + }, + { + "epoch": 7.36, + "learning_rate": 3.3751228598075445e-07, + "loss": 0.0004, + "step": 56159 + }, + { + "epoch": 7.36, + "learning_rate": 3.373756821170149e-07, + "loss": 0.0008, + "step": 56160 + }, + { + "epoch": 7.36, + "learning_rate": 3.372391054288915e-07, + "loss": 0.0007, + "step": 56161 + }, + { + "epoch": 7.36, + "learning_rate": 3.3710255591676623e-07, + "loss": 0.0002, + "step": 56162 + }, + { + "epoch": 7.36, + "learning_rate": 3.3696603358102764e-07, + "loss": 0.0008, + "step": 56163 + }, + { + "epoch": 7.36, + "learning_rate": 3.3682953842205767e-07, + "loss": 0.0005, + "step": 56164 + }, + { + "epoch": 7.36, + "learning_rate": 3.366930704402382e-07, + "loss": 0.0002, + "step": 56165 + }, + { + "epoch": 7.36, + "learning_rate": 3.365566296359557e-07, + "loss": 0.0011, + "step": 56166 + }, + { + "epoch": 7.36, + "learning_rate": 3.3642021600959305e-07, + "loss": 0.0003, + "step": 56167 + }, + { + "epoch": 7.36, + "learning_rate": 3.3628382956153447e-07, + "loss": 0.0001, + "step": 56168 + }, + { + "epoch": 7.36, + "learning_rate": 3.36147470292163e-07, + "loss": 0.0005, + "step": 56169 + }, + { + "epoch": 7.36, + "learning_rate": 3.360111382018616e-07, + "loss": 0.0003, + "step": 56170 + }, + { + "epoch": 7.36, + "learning_rate": 3.3587483329101443e-07, + "loss": 0.0007, + "step": 56171 + }, + { + "epoch": 7.36, + "learning_rate": 3.3573855556000457e-07, + "loss": 0.0001, + "step": 56172 + }, + { + "epoch": 7.36, + "learning_rate": 3.356023050092161e-07, + "loss": 0.0013, + "step": 56173 + }, + { + "epoch": 7.36, + "learning_rate": 3.354660816390298e-07, + "loss": 0.0006, + "step": 56174 + }, + { + "epoch": 7.36, + "learning_rate": 3.353298854498321e-07, + "loss": 0.0005, + "step": 56175 + }, + { + "epoch": 7.36, + "learning_rate": 3.351937164420038e-07, + "loss": 0.0002, + "step": 56176 + }, + { + "epoch": 7.36, + "learning_rate": 3.3505757461592903e-07, + "loss": 0.0005, + "step": 56177 + }, + { + "epoch": 7.36, + "learning_rate": 3.3492145997198855e-07, + "loss": 0.0009, + "step": 56178 + }, + { + "epoch": 7.36, + "learning_rate": 3.3478537251056763e-07, + "loss": 0.0005, + "step": 56179 + }, + { + "epoch": 7.36, + "learning_rate": 3.346493122320493e-07, + "loss": 0.0007, + "step": 56180 + }, + { + "epoch": 7.36, + "learning_rate": 3.345132791368133e-07, + "loss": 0.0005, + "step": 56181 + }, + { + "epoch": 7.36, + "learning_rate": 3.343772732252448e-07, + "loss": 0.0013, + "step": 56182 + }, + { + "epoch": 7.36, + "learning_rate": 3.342412944977258e-07, + "loss": 0.0004, + "step": 56183 + }, + { + "epoch": 7.36, + "learning_rate": 3.341053429546393e-07, + "loss": 0.0002, + "step": 56184 + }, + { + "epoch": 7.36, + "learning_rate": 3.339694185963649e-07, + "loss": 0.0006, + "step": 56185 + }, + { + "epoch": 7.36, + "learning_rate": 3.3383352142328797e-07, + "loss": 0.0005, + "step": 56186 + }, + { + "epoch": 7.36, + "learning_rate": 3.3369765143578816e-07, + "loss": 0.001, + "step": 56187 + }, + { + "epoch": 7.36, + "learning_rate": 3.3356180863424956e-07, + "loss": 0.0007, + "step": 56188 + }, + { + "epoch": 7.36, + "learning_rate": 3.334259930190531e-07, + "loss": 0.0006, + "step": 56189 + }, + { + "epoch": 7.36, + "learning_rate": 3.332902045905828e-07, + "loss": 0.0002, + "step": 56190 + }, + { + "epoch": 7.36, + "learning_rate": 3.331544433492184e-07, + "loss": 0.001, + "step": 56191 + }, + { + "epoch": 7.36, + "learning_rate": 3.3301870929534295e-07, + "loss": 0.0002, + "step": 56192 + }, + { + "epoch": 7.36, + "learning_rate": 3.3288300242933616e-07, + "loss": 0.0006, + "step": 56193 + }, + { + "epoch": 7.36, + "learning_rate": 3.3274732275158206e-07, + "loss": 0.0006, + "step": 56194 + }, + { + "epoch": 7.36, + "learning_rate": 3.3261167026246156e-07, + "loss": 0.0003, + "step": 56195 + }, + { + "epoch": 7.36, + "learning_rate": 3.324760449623543e-07, + "loss": 0.0004, + "step": 56196 + }, + { + "epoch": 7.36, + "learning_rate": 3.3234044685164446e-07, + "loss": 0.0006, + "step": 56197 + }, + { + "epoch": 7.36, + "learning_rate": 3.3220487593071173e-07, + "loss": 0.001, + "step": 56198 + }, + { + "epoch": 7.36, + "learning_rate": 3.320693321999391e-07, + "loss": 0.0003, + "step": 56199 + }, + { + "epoch": 7.36, + "learning_rate": 3.3193381565970627e-07, + "loss": 0.0001, + "step": 56200 + }, + { + "epoch": 7.36, + "learning_rate": 3.317983263103941e-07, + "loss": 0.0008, + "step": 56201 + }, + { + "epoch": 7.36, + "learning_rate": 3.3166286415238446e-07, + "loss": 0.0002, + "step": 56202 + }, + { + "epoch": 7.36, + "learning_rate": 3.315274291860593e-07, + "loss": 0.0003, + "step": 56203 + }, + { + "epoch": 7.36, + "learning_rate": 3.3139202141179606e-07, + "loss": 0.0004, + "step": 56204 + }, + { + "epoch": 7.36, + "learning_rate": 3.3125664082998e-07, + "loss": 0.0007, + "step": 56205 + }, + { + "epoch": 7.36, + "learning_rate": 3.3112128744098857e-07, + "loss": 0.0006, + "step": 56206 + }, + { + "epoch": 7.36, + "learning_rate": 3.3098596124520486e-07, + "loss": 0.0007, + "step": 56207 + }, + { + "epoch": 7.36, + "learning_rate": 3.3085066224300633e-07, + "loss": 0.0006, + "step": 56208 + }, + { + "epoch": 7.36, + "learning_rate": 3.3071539043477597e-07, + "loss": 0.0006, + "step": 56209 + }, + { + "epoch": 7.36, + "learning_rate": 3.305801458208946e-07, + "loss": 0.0003, + "step": 56210 + }, + { + "epoch": 7.36, + "learning_rate": 3.304449284017408e-07, + "loss": 0.0005, + "step": 56211 + }, + { + "epoch": 7.36, + "learning_rate": 3.303097381776954e-07, + "loss": 0.0009, + "step": 56212 + }, + { + "epoch": 7.36, + "learning_rate": 3.301745751491403e-07, + "loss": 0.0006, + "step": 56213 + }, + { + "epoch": 7.36, + "learning_rate": 3.300394393164541e-07, + "loss": 0.0006, + "step": 56214 + }, + { + "epoch": 7.36, + "learning_rate": 3.2990433068001647e-07, + "loss": 0.0004, + "step": 56215 + }, + { + "epoch": 7.36, + "learning_rate": 3.2976924924020716e-07, + "loss": 0.0004, + "step": 56216 + }, + { + "epoch": 7.36, + "learning_rate": 3.29634194997408e-07, + "loss": 0.0007, + "step": 56217 + }, + { + "epoch": 7.36, + "learning_rate": 3.2949916795199767e-07, + "loss": 0.0003, + "step": 56218 + }, + { + "epoch": 7.36, + "learning_rate": 3.293641681043558e-07, + "loss": 0.0003, + "step": 56219 + }, + { + "epoch": 7.36, + "learning_rate": 3.2922919545486096e-07, + "loss": 0.0002, + "step": 56220 + }, + { + "epoch": 7.36, + "learning_rate": 3.290942500038952e-07, + "loss": 0.0007, + "step": 56221 + }, + { + "epoch": 7.36, + "learning_rate": 3.2895933175183694e-07, + "loss": 0.0005, + "step": 56222 + }, + { + "epoch": 7.36, + "learning_rate": 3.2882444069906594e-07, + "loss": 0.0004, + "step": 56223 + }, + { + "epoch": 7.36, + "learning_rate": 3.286895768459597e-07, + "loss": 0.0008, + "step": 56224 + }, + { + "epoch": 7.37, + "learning_rate": 3.285547401929001e-07, + "loss": 0.0002, + "step": 56225 + }, + { + "epoch": 7.37, + "learning_rate": 3.284199307402658e-07, + "loss": 0.0004, + "step": 56226 + }, + { + "epoch": 7.37, + "learning_rate": 3.282851484884342e-07, + "loss": 0.0005, + "step": 56227 + }, + { + "epoch": 7.37, + "learning_rate": 3.281503934377861e-07, + "loss": 0.0004, + "step": 56228 + }, + { + "epoch": 7.37, + "learning_rate": 3.280156655887001e-07, + "loss": 0.0011, + "step": 56229 + }, + { + "epoch": 7.37, + "learning_rate": 3.2788096494155483e-07, + "loss": 0.0002, + "step": 56230 + }, + { + "epoch": 7.37, + "learning_rate": 3.2774629149672775e-07, + "loss": 0.0011, + "step": 56231 + }, + { + "epoch": 7.37, + "learning_rate": 3.276116452546008e-07, + "loss": 0.0003, + "step": 56232 + }, + { + "epoch": 7.37, + "learning_rate": 3.2747702621555133e-07, + "loss": 0.0005, + "step": 56233 + }, + { + "epoch": 7.37, + "learning_rate": 3.2734243437995696e-07, + "loss": 0.0004, + "step": 56234 + }, + { + "epoch": 7.37, + "learning_rate": 3.272078697481962e-07, + "loss": 0.0005, + "step": 56235 + }, + { + "epoch": 7.37, + "learning_rate": 3.270733323206499e-07, + "loss": 0.001, + "step": 56236 + }, + { + "epoch": 7.37, + "learning_rate": 3.2693882209769435e-07, + "loss": 0.0003, + "step": 56237 + }, + { + "epoch": 7.37, + "learning_rate": 3.268043390797082e-07, + "loss": 0.0004, + "step": 56238 + }, + { + "epoch": 7.37, + "learning_rate": 3.2666988326706785e-07, + "loss": 0.0006, + "step": 56239 + }, + { + "epoch": 7.37, + "learning_rate": 3.265354546601551e-07, + "loss": 0.0003, + "step": 56240 + }, + { + "epoch": 7.37, + "learning_rate": 3.2640105325934536e-07, + "loss": 0.0003, + "step": 56241 + }, + { + "epoch": 7.37, + "learning_rate": 3.262666790650182e-07, + "loss": 0.0004, + "step": 56242 + }, + { + "epoch": 7.37, + "learning_rate": 3.2613233207755e-07, + "loss": 0.0009, + "step": 56243 + }, + { + "epoch": 7.37, + "learning_rate": 3.2599801229732055e-07, + "loss": 0.0006, + "step": 56244 + }, + { + "epoch": 7.37, + "learning_rate": 3.2586371972470723e-07, + "loss": 0.0004, + "step": 56245 + }, + { + "epoch": 7.37, + "learning_rate": 3.2572945436008527e-07, + "loss": 0.0008, + "step": 56246 + }, + { + "epoch": 7.37, + "learning_rate": 3.255952162038345e-07, + "loss": 0.0004, + "step": 56247 + }, + { + "epoch": 7.37, + "learning_rate": 3.254610052563323e-07, + "loss": 0.0004, + "step": 56248 + }, + { + "epoch": 7.37, + "learning_rate": 3.253268215179561e-07, + "loss": 0.0005, + "step": 56249 + }, + { + "epoch": 7.37, + "learning_rate": 3.2519266498908243e-07, + "loss": 0.0006, + "step": 56250 + }, + { + "epoch": 7.37, + "learning_rate": 3.2505853567008973e-07, + "loss": 0.0004, + "step": 56251 + }, + { + "epoch": 7.37, + "learning_rate": 3.2492443356135326e-07, + "loss": 0.0002, + "step": 56252 + }, + { + "epoch": 7.37, + "learning_rate": 3.247903586632528e-07, + "loss": 0.0008, + "step": 56253 + }, + { + "epoch": 7.37, + "learning_rate": 3.2465631097616355e-07, + "loss": 0.001, + "step": 56254 + }, + { + "epoch": 7.37, + "learning_rate": 3.245222905004641e-07, + "loss": 0.0003, + "step": 56255 + }, + { + "epoch": 7.37, + "learning_rate": 3.2438829723652974e-07, + "loss": 0.0003, + "step": 56256 + }, + { + "epoch": 7.37, + "learning_rate": 3.2425433118473905e-07, + "loss": 0.0003, + "step": 56257 + }, + { + "epoch": 7.37, + "learning_rate": 3.241203923454661e-07, + "loss": 0.0004, + "step": 56258 + }, + { + "epoch": 7.37, + "learning_rate": 3.2398648071909065e-07, + "loss": 0.0003, + "step": 56259 + }, + { + "epoch": 7.37, + "learning_rate": 3.23852596305988e-07, + "loss": 0.0009, + "step": 56260 + }, + { + "epoch": 7.37, + "learning_rate": 3.2371873910653326e-07, + "loss": 0.0008, + "step": 56261 + }, + { + "epoch": 7.37, + "learning_rate": 3.235849091211063e-07, + "loss": 0.0006, + "step": 56262 + }, + { + "epoch": 7.37, + "learning_rate": 3.234511063500789e-07, + "loss": 0.0002, + "step": 56263 + }, + { + "epoch": 7.37, + "learning_rate": 3.23317330793832e-07, + "loss": 0.0007, + "step": 56264 + }, + { + "epoch": 7.37, + "learning_rate": 3.231835824527396e-07, + "loss": 0.0005, + "step": 56265 + }, + { + "epoch": 7.37, + "learning_rate": 3.2304986132717706e-07, + "loss": 0.0004, + "step": 56266 + }, + { + "epoch": 7.37, + "learning_rate": 3.22916167417523e-07, + "loss": 0.0002, + "step": 56267 + }, + { + "epoch": 7.37, + "learning_rate": 3.227825007241503e-07, + "loss": 0.0003, + "step": 56268 + }, + { + "epoch": 7.37, + "learning_rate": 3.2264886124743653e-07, + "loss": 0.0007, + "step": 56269 + }, + { + "epoch": 7.37, + "learning_rate": 3.225152489877592e-07, + "loss": 0.0004, + "step": 56270 + }, + { + "epoch": 7.37, + "learning_rate": 3.223816639454913e-07, + "loss": 0.0003, + "step": 56271 + }, + { + "epoch": 7.37, + "learning_rate": 3.2224810612101033e-07, + "loss": 0.0007, + "step": 56272 + }, + { + "epoch": 7.37, + "learning_rate": 3.221145755146893e-07, + "loss": 0.0003, + "step": 56273 + }, + { + "epoch": 7.37, + "learning_rate": 3.2198107212690565e-07, + "loss": 0.0002, + "step": 56274 + }, + { + "epoch": 7.37, + "learning_rate": 3.2184759595803693e-07, + "loss": 0.0003, + "step": 56275 + }, + { + "epoch": 7.37, + "learning_rate": 3.2171414700845614e-07, + "loss": 0.0004, + "step": 56276 + }, + { + "epoch": 7.37, + "learning_rate": 3.2158072527853745e-07, + "loss": 0.0005, + "step": 56277 + }, + { + "epoch": 7.37, + "learning_rate": 3.214473307686583e-07, + "loss": 0.0003, + "step": 56278 + }, + { + "epoch": 7.37, + "learning_rate": 3.2131396347919395e-07, + "loss": 0.0002, + "step": 56279 + }, + { + "epoch": 7.37, + "learning_rate": 3.211806234105175e-07, + "loss": 0.0003, + "step": 56280 + }, + { + "epoch": 7.37, + "learning_rate": 3.210473105630052e-07, + "loss": 0.0003, + "step": 56281 + }, + { + "epoch": 7.37, + "learning_rate": 3.209140249370324e-07, + "loss": 0.0012, + "step": 56282 + }, + { + "epoch": 7.37, + "learning_rate": 3.207807665329743e-07, + "loss": 0.0002, + "step": 56283 + }, + { + "epoch": 7.37, + "learning_rate": 3.2064753535120284e-07, + "loss": 0.0004, + "step": 56284 + }, + { + "epoch": 7.37, + "learning_rate": 3.205143313920944e-07, + "loss": 0.0006, + "step": 56285 + }, + { + "epoch": 7.37, + "learning_rate": 3.203811546560265e-07, + "loss": 0.001, + "step": 56286 + }, + { + "epoch": 7.37, + "learning_rate": 3.2024800514336984e-07, + "loss": 0.0002, + "step": 56287 + }, + { + "epoch": 7.37, + "learning_rate": 3.2011488285450087e-07, + "loss": 0.0003, + "step": 56288 + }, + { + "epoch": 7.37, + "learning_rate": 3.199817877897915e-07, + "loss": 0.0003, + "step": 56289 + }, + { + "epoch": 7.37, + "learning_rate": 3.198487199496203e-07, + "loss": 0.001, + "step": 56290 + }, + { + "epoch": 7.37, + "learning_rate": 3.197156793343581e-07, + "loss": 0.0008, + "step": 56291 + }, + { + "epoch": 7.37, + "learning_rate": 3.1958266594437904e-07, + "loss": 0.0002, + "step": 56292 + }, + { + "epoch": 7.37, + "learning_rate": 3.194496797800595e-07, + "loss": 0.0005, + "step": 56293 + }, + { + "epoch": 7.37, + "learning_rate": 3.1931672084177133e-07, + "loss": 0.0002, + "step": 56294 + }, + { + "epoch": 7.37, + "learning_rate": 3.191837891298899e-07, + "loss": 0.0008, + "step": 56295 + }, + { + "epoch": 7.37, + "learning_rate": 3.1905088464478707e-07, + "loss": 0.0003, + "step": 56296 + }, + { + "epoch": 7.37, + "learning_rate": 3.189180073868392e-07, + "loss": 0.0004, + "step": 56297 + }, + { + "epoch": 7.37, + "learning_rate": 3.1878515735641936e-07, + "loss": 0.0002, + "step": 56298 + }, + { + "epoch": 7.37, + "learning_rate": 3.1865233455390164e-07, + "loss": 0.0002, + "step": 56299 + }, + { + "epoch": 7.37, + "learning_rate": 3.185195389796558e-07, + "loss": 0.0007, + "step": 56300 + }, + { + "epoch": 7.38, + "learning_rate": 3.1838677063406045e-07, + "loss": 0.0003, + "step": 56301 + }, + { + "epoch": 7.38, + "learning_rate": 3.182540295174863e-07, + "loss": 0.0007, + "step": 56302 + }, + { + "epoch": 7.38, + "learning_rate": 3.1812131563030755e-07, + "loss": 0.0008, + "step": 56303 + }, + { + "epoch": 7.38, + "learning_rate": 3.1798862897289506e-07, + "loss": 0.0006, + "step": 56304 + }, + { + "epoch": 7.38, + "learning_rate": 3.1785596954562623e-07, + "loss": 0.0003, + "step": 56305 + }, + { + "epoch": 7.38, + "learning_rate": 3.1772333734887085e-07, + "loss": 0.0003, + "step": 56306 + }, + { + "epoch": 7.38, + "learning_rate": 3.17590732383003e-07, + "loss": 0.0005, + "step": 56307 + }, + { + "epoch": 7.38, + "learning_rate": 3.174581546483946e-07, + "loss": 0.0003, + "step": 56308 + }, + { + "epoch": 7.38, + "learning_rate": 3.1732560414542093e-07, + "loss": 0.0005, + "step": 56309 + }, + { + "epoch": 7.38, + "learning_rate": 3.171930808744528e-07, + "loss": 0.0003, + "step": 56310 + }, + { + "epoch": 7.38, + "learning_rate": 3.1706058483586324e-07, + "loss": 0.0004, + "step": 56311 + }, + { + "epoch": 7.38, + "learning_rate": 3.169281160300264e-07, + "loss": 0.0004, + "step": 56312 + }, + { + "epoch": 7.38, + "learning_rate": 3.167956744573131e-07, + "loss": 0.0003, + "step": 56313 + }, + { + "epoch": 7.38, + "learning_rate": 3.1666326011809635e-07, + "loss": 0.0006, + "step": 56314 + }, + { + "epoch": 7.38, + "learning_rate": 3.16530873012747e-07, + "loss": 0.0003, + "step": 56315 + }, + { + "epoch": 7.38, + "learning_rate": 3.163985131416403e-07, + "loss": 0.0007, + "step": 56316 + }, + { + "epoch": 7.38, + "learning_rate": 3.16266180505147e-07, + "loss": 0.0005, + "step": 56317 + }, + { + "epoch": 7.38, + "learning_rate": 3.1613387510363913e-07, + "loss": 0.0005, + "step": 56318 + }, + { + "epoch": 7.38, + "learning_rate": 3.1600159693748743e-07, + "loss": 0.0003, + "step": 56319 + }, + { + "epoch": 7.38, + "learning_rate": 3.158693460070683e-07, + "loss": 0.0009, + "step": 56320 + }, + { + "epoch": 7.38, + "learning_rate": 3.1573712231275033e-07, + "loss": 0.0002, + "step": 56321 + }, + { + "epoch": 7.38, + "learning_rate": 3.1560492585490545e-07, + "loss": 0.0005, + "step": 56322 + }, + { + "epoch": 7.38, + "learning_rate": 3.1547275663390553e-07, + "loss": 0.0008, + "step": 56323 + }, + { + "epoch": 7.38, + "learning_rate": 3.1534061465012364e-07, + "loss": 0.0003, + "step": 56324 + }, + { + "epoch": 7.38, + "learning_rate": 3.1520849990392956e-07, + "loss": 0.0002, + "step": 56325 + }, + { + "epoch": 7.38, + "learning_rate": 3.150764123956962e-07, + "loss": 0.0013, + "step": 56326 + }, + { + "epoch": 7.38, + "learning_rate": 3.1494435212579556e-07, + "loss": 0.0006, + "step": 56327 + }, + { + "epoch": 7.38, + "learning_rate": 3.148123190945973e-07, + "loss": 0.0006, + "step": 56328 + }, + { + "epoch": 7.38, + "learning_rate": 3.146803133024734e-07, + "loss": 0.0004, + "step": 56329 + }, + { + "epoch": 7.38, + "learning_rate": 3.1454833474979575e-07, + "loss": 0.0008, + "step": 56330 + }, + { + "epoch": 7.38, + "learning_rate": 3.1441638343693516e-07, + "loss": 0.0003, + "step": 56331 + }, + { + "epoch": 7.38, + "learning_rate": 3.1428445936426245e-07, + "loss": 0.0004, + "step": 56332 + }, + { + "epoch": 7.38, + "learning_rate": 3.1415256253214953e-07, + "loss": 0.0005, + "step": 56333 + }, + { + "epoch": 7.38, + "learning_rate": 3.1402069294096617e-07, + "loss": 0.0007, + "step": 56334 + }, + { + "epoch": 7.38, + "learning_rate": 3.138888505910842e-07, + "loss": 0.0006, + "step": 56335 + }, + { + "epoch": 7.38, + "learning_rate": 3.137570354828734e-07, + "loss": 0.0005, + "step": 56336 + }, + { + "epoch": 7.38, + "learning_rate": 3.1362524761670563e-07, + "loss": 0.0012, + "step": 56337 + }, + { + "epoch": 7.38, + "learning_rate": 3.1349348699295065e-07, + "loss": 0.0004, + "step": 56338 + }, + { + "epoch": 7.38, + "learning_rate": 3.133617536119804e-07, + "loss": 0.0003, + "step": 56339 + }, + { + "epoch": 7.38, + "learning_rate": 3.132300474741623e-07, + "loss": 0.0009, + "step": 56340 + }, + { + "epoch": 7.38, + "learning_rate": 3.130983685798705e-07, + "loss": 0.0004, + "step": 56341 + }, + { + "epoch": 7.38, + "learning_rate": 3.129667169294726e-07, + "loss": 0.0003, + "step": 56342 + }, + { + "epoch": 7.38, + "learning_rate": 3.1283509252334145e-07, + "loss": 0.0007, + "step": 56343 + }, + { + "epoch": 7.38, + "learning_rate": 3.1270349536184465e-07, + "loss": 0.0006, + "step": 56344 + }, + { + "epoch": 7.38, + "learning_rate": 3.125719254453541e-07, + "loss": 0.0004, + "step": 56345 + }, + { + "epoch": 7.38, + "learning_rate": 3.1244038277423837e-07, + "loss": 0.0006, + "step": 56346 + }, + { + "epoch": 7.38, + "learning_rate": 3.1230886734886946e-07, + "loss": 0.0002, + "step": 56347 + }, + { + "epoch": 7.38, + "learning_rate": 3.1217737916961475e-07, + "loss": 0.0002, + "step": 56348 + }, + { + "epoch": 7.38, + "learning_rate": 3.1204591823684514e-07, + "loss": 0.0004, + "step": 56349 + }, + { + "epoch": 7.38, + "learning_rate": 3.119144845509314e-07, + "loss": 0.001, + "step": 56350 + }, + { + "epoch": 7.38, + "learning_rate": 3.117830781122411e-07, + "loss": 0.0005, + "step": 56351 + }, + { + "epoch": 7.38, + "learning_rate": 3.11651698921146e-07, + "loss": 0.0004, + "step": 56352 + }, + { + "epoch": 7.38, + "learning_rate": 3.115203469780137e-07, + "loss": 0.0006, + "step": 56353 + }, + { + "epoch": 7.38, + "learning_rate": 3.11389022283215e-07, + "loss": 0.0006, + "step": 56354 + }, + { + "epoch": 7.38, + "learning_rate": 3.112577248371185e-07, + "loss": 0.0004, + "step": 56355 + }, + { + "epoch": 7.38, + "learning_rate": 3.111264546400938e-07, + "loss": 0.0003, + "step": 56356 + }, + { + "epoch": 7.38, + "learning_rate": 3.109952116925097e-07, + "loss": 0.0005, + "step": 56357 + }, + { + "epoch": 7.38, + "learning_rate": 3.108639959947357e-07, + "loss": 0.0004, + "step": 56358 + }, + { + "epoch": 7.38, + "learning_rate": 3.107328075471416e-07, + "loss": 0.0003, + "step": 56359 + }, + { + "epoch": 7.38, + "learning_rate": 3.1060164635009495e-07, + "loss": 0.0003, + "step": 56360 + }, + { + "epoch": 7.38, + "learning_rate": 3.104705124039653e-07, + "loss": 0.0003, + "step": 56361 + }, + { + "epoch": 7.38, + "learning_rate": 3.1033940570912023e-07, + "loss": 0.0007, + "step": 56362 + }, + { + "epoch": 7.38, + "learning_rate": 3.102083262659306e-07, + "loss": 0.0004, + "step": 56363 + }, + { + "epoch": 7.38, + "learning_rate": 3.100772740747637e-07, + "loss": 0.0004, + "step": 56364 + }, + { + "epoch": 7.38, + "learning_rate": 3.099462491359884e-07, + "loss": 0.0003, + "step": 56365 + }, + { + "epoch": 7.38, + "learning_rate": 3.0981525144997417e-07, + "loss": 0.0006, + "step": 56366 + }, + { + "epoch": 7.38, + "learning_rate": 3.096842810170875e-07, + "loss": 0.001, + "step": 56367 + }, + { + "epoch": 7.38, + "learning_rate": 3.09553337837698e-07, + "loss": 0.0005, + "step": 56368 + }, + { + "epoch": 7.38, + "learning_rate": 3.094224219121744e-07, + "loss": 0.0004, + "step": 56369 + }, + { + "epoch": 7.38, + "learning_rate": 3.0929153324088303e-07, + "loss": 0.0005, + "step": 56370 + }, + { + "epoch": 7.38, + "learning_rate": 3.091606718241946e-07, + "loss": 0.0003, + "step": 56371 + }, + { + "epoch": 7.38, + "learning_rate": 3.090298376624734e-07, + "loss": 0.0004, + "step": 56372 + }, + { + "epoch": 7.38, + "learning_rate": 3.088990307560913e-07, + "loss": 0.0012, + "step": 56373 + }, + { + "epoch": 7.38, + "learning_rate": 3.0876825110541467e-07, + "loss": 0.0004, + "step": 56374 + }, + { + "epoch": 7.38, + "learning_rate": 3.0863749871081096e-07, + "loss": 0.0008, + "step": 56375 + }, + { + "epoch": 7.38, + "learning_rate": 3.085067735726477e-07, + "loss": 0.0003, + "step": 56376 + }, + { + "epoch": 7.38, + "learning_rate": 3.0837607569129346e-07, + "loss": 0.0008, + "step": 56377 + }, + { + "epoch": 7.39, + "learning_rate": 3.0824540506711575e-07, + "loss": 0.0015, + "step": 56378 + }, + { + "epoch": 7.39, + "learning_rate": 3.0811476170048204e-07, + "loss": 0.0024, + "step": 56379 + }, + { + "epoch": 7.39, + "learning_rate": 3.079841455917587e-07, + "loss": 0.0009, + "step": 56380 + }, + { + "epoch": 7.39, + "learning_rate": 3.078535567413143e-07, + "loss": 0.0008, + "step": 56381 + }, + { + "epoch": 7.39, + "learning_rate": 3.077229951495153e-07, + "loss": 0.0005, + "step": 56382 + }, + { + "epoch": 7.39, + "learning_rate": 3.075924608167291e-07, + "loss": 0.0003, + "step": 56383 + }, + { + "epoch": 7.39, + "learning_rate": 3.074619537433221e-07, + "loss": 0.0004, + "step": 56384 + }, + { + "epoch": 7.39, + "learning_rate": 3.073314739296629e-07, + "loss": 0.0005, + "step": 56385 + }, + { + "epoch": 7.39, + "learning_rate": 3.0720102137611894e-07, + "loss": 0.0006, + "step": 56386 + }, + { + "epoch": 7.39, + "learning_rate": 3.0707059608305555e-07, + "loss": 0.0009, + "step": 56387 + }, + { + "epoch": 7.39, + "learning_rate": 3.0694019805083796e-07, + "loss": 0.0003, + "step": 56388 + }, + { + "epoch": 7.39, + "learning_rate": 3.06809827279837e-07, + "loss": 0.0002, + "step": 56389 + }, + { + "epoch": 7.39, + "learning_rate": 3.066794837704168e-07, + "loss": 0.0004, + "step": 56390 + }, + { + "epoch": 7.39, + "learning_rate": 3.0654916752294374e-07, + "loss": 0.0006, + "step": 56391 + }, + { + "epoch": 7.39, + "learning_rate": 3.064188785377853e-07, + "loss": 0.0003, + "step": 56392 + }, + { + "epoch": 7.39, + "learning_rate": 3.0628861681530784e-07, + "loss": 0.0002, + "step": 56393 + }, + { + "epoch": 7.39, + "learning_rate": 3.061583823558767e-07, + "loss": 0.0007, + "step": 56394 + }, + { + "epoch": 7.39, + "learning_rate": 3.0602817515986036e-07, + "loss": 0.0009, + "step": 56395 + }, + { + "epoch": 7.39, + "learning_rate": 3.0589799522762087e-07, + "loss": 0.0001, + "step": 56396 + }, + { + "epoch": 7.39, + "learning_rate": 3.057678425595301e-07, + "loss": 0.0005, + "step": 56397 + }, + { + "epoch": 7.39, + "learning_rate": 3.0563771715594994e-07, + "loss": 0.0003, + "step": 56398 + }, + { + "epoch": 7.39, + "learning_rate": 3.055076190172457e-07, + "loss": 0.0002, + "step": 56399 + }, + { + "epoch": 7.39, + "learning_rate": 3.0537754814378707e-07, + "loss": 0.0001, + "step": 56400 + }, + { + "epoch": 7.39, + "learning_rate": 3.052475045359371e-07, + "loss": 0.0005, + "step": 56401 + }, + { + "epoch": 7.39, + "learning_rate": 3.0511748819406326e-07, + "loss": 0.0008, + "step": 56402 + }, + { + "epoch": 7.39, + "learning_rate": 3.0498749911852867e-07, + "loss": 0.0009, + "step": 56403 + }, + { + "epoch": 7.39, + "learning_rate": 3.048575373097007e-07, + "loss": 0.0007, + "step": 56404 + }, + { + "epoch": 7.39, + "learning_rate": 3.0472760276794467e-07, + "loss": 0.0002, + "step": 56405 + }, + { + "epoch": 7.39, + "learning_rate": 3.0459769549362583e-07, + "loss": 0.0003, + "step": 56406 + }, + { + "epoch": 7.39, + "learning_rate": 3.0446781548710947e-07, + "loss": 0.0009, + "step": 56407 + }, + { + "epoch": 7.39, + "learning_rate": 3.0433796274876193e-07, + "loss": 0.0005, + "step": 56408 + }, + { + "epoch": 7.39, + "learning_rate": 3.042081372789485e-07, + "loss": 0.0004, + "step": 56409 + }, + { + "epoch": 7.39, + "learning_rate": 3.040783390780322e-07, + "loss": 0.0007, + "step": 56410 + }, + { + "epoch": 7.39, + "learning_rate": 3.039485681463783e-07, + "loss": 0.0006, + "step": 56411 + }, + { + "epoch": 7.39, + "learning_rate": 3.038188244843543e-07, + "loss": 0.0011, + "step": 56412 + }, + { + "epoch": 7.39, + "learning_rate": 3.0368910809232324e-07, + "loss": 0.0008, + "step": 56413 + }, + { + "epoch": 7.39, + "learning_rate": 3.035594189706481e-07, + "loss": 0.0007, + "step": 56414 + }, + { + "epoch": 7.39, + "learning_rate": 3.034297571196987e-07, + "loss": 0.0002, + "step": 56415 + }, + { + "epoch": 7.39, + "learning_rate": 3.0330012253983463e-07, + "loss": 0.0004, + "step": 56416 + }, + { + "epoch": 7.39, + "learning_rate": 3.031705152314235e-07, + "loss": 0.0006, + "step": 56417 + }, + { + "epoch": 7.39, + "learning_rate": 3.0304093519482827e-07, + "loss": 0.0003, + "step": 56418 + }, + { + "epoch": 7.39, + "learning_rate": 3.0291138243041417e-07, + "loss": 0.0002, + "step": 56419 + }, + { + "epoch": 7.39, + "learning_rate": 3.0278185693854657e-07, + "loss": 0.0005, + "step": 56420 + }, + { + "epoch": 7.39, + "learning_rate": 3.026523587195873e-07, + "loss": 0.0003, + "step": 56421 + }, + { + "epoch": 7.39, + "learning_rate": 3.025228877739017e-07, + "loss": 0.0004, + "step": 56422 + }, + { + "epoch": 7.39, + "learning_rate": 3.02393444101855e-07, + "loss": 0.0006, + "step": 56423 + }, + { + "epoch": 7.39, + "learning_rate": 3.022640277038114e-07, + "loss": 0.0008, + "step": 56424 + }, + { + "epoch": 7.39, + "learning_rate": 3.021346385801327e-07, + "loss": 0.0003, + "step": 56425 + }, + { + "epoch": 7.39, + "learning_rate": 3.020052767311821e-07, + "loss": 0.0006, + "step": 56426 + }, + { + "epoch": 7.39, + "learning_rate": 3.0187594215732583e-07, + "loss": 0.0009, + "step": 56427 + }, + { + "epoch": 7.39, + "learning_rate": 3.017466348589282e-07, + "loss": 0.0005, + "step": 56428 + }, + { + "epoch": 7.39, + "learning_rate": 3.0161735483635213e-07, + "loss": 0.0003, + "step": 56429 + }, + { + "epoch": 7.39, + "learning_rate": 3.014881020899596e-07, + "loss": 0.0004, + "step": 56430 + }, + { + "epoch": 7.39, + "learning_rate": 3.013588766201159e-07, + "loss": 0.0008, + "step": 56431 + }, + { + "epoch": 7.39, + "learning_rate": 3.0122967842718397e-07, + "loss": 0.0007, + "step": 56432 + }, + { + "epoch": 7.39, + "learning_rate": 3.0110050751152585e-07, + "loss": 0.0008, + "step": 56433 + }, + { + "epoch": 7.39, + "learning_rate": 3.009713638735079e-07, + "loss": 0.0006, + "step": 56434 + }, + { + "epoch": 7.39, + "learning_rate": 3.008422475134909e-07, + "loss": 0.0005, + "step": 56435 + }, + { + "epoch": 7.39, + "learning_rate": 3.007131584318379e-07, + "loss": 0.0002, + "step": 56436 + }, + { + "epoch": 7.39, + "learning_rate": 3.0058409662891306e-07, + "loss": 0.0004, + "step": 56437 + }, + { + "epoch": 7.39, + "learning_rate": 3.0045506210507726e-07, + "loss": 0.0003, + "step": 56438 + }, + { + "epoch": 7.39, + "learning_rate": 3.003260548606979e-07, + "loss": 0.0002, + "step": 56439 + }, + { + "epoch": 7.39, + "learning_rate": 3.0019707489613357e-07, + "loss": 0.0006, + "step": 56440 + }, + { + "epoch": 7.39, + "learning_rate": 3.000681222117474e-07, + "loss": 0.0004, + "step": 56441 + }, + { + "epoch": 7.39, + "learning_rate": 2.999391968079046e-07, + "loss": 0.0007, + "step": 56442 + }, + { + "epoch": 7.39, + "learning_rate": 2.998102986849671e-07, + "loss": 0.0009, + "step": 56443 + }, + { + "epoch": 7.39, + "learning_rate": 2.9968142784329577e-07, + "loss": 0.0007, + "step": 56444 + }, + { + "epoch": 7.39, + "learning_rate": 2.9955258428325254e-07, + "loss": 0.0003, + "step": 56445 + }, + { + "epoch": 7.39, + "learning_rate": 2.9942376800520256e-07, + "loss": 0.0004, + "step": 56446 + }, + { + "epoch": 7.39, + "learning_rate": 2.9929497900950565e-07, + "loss": 0.0005, + "step": 56447 + }, + { + "epoch": 7.39, + "learning_rate": 2.9916621729652483e-07, + "loss": 0.0003, + "step": 56448 + }, + { + "epoch": 7.39, + "learning_rate": 2.9903748286662316e-07, + "loss": 0.0006, + "step": 56449 + }, + { + "epoch": 7.39, + "learning_rate": 2.989087757201614e-07, + "loss": 0.0003, + "step": 56450 + }, + { + "epoch": 7.39, + "learning_rate": 2.9878009585750376e-07, + "loss": 0.0003, + "step": 56451 + }, + { + "epoch": 7.39, + "learning_rate": 2.986514432790089e-07, + "loss": 0.0006, + "step": 56452 + }, + { + "epoch": 7.39, + "learning_rate": 2.9852281798503966e-07, + "loss": 0.0007, + "step": 56453 + }, + { + "epoch": 7.4, + "learning_rate": 2.9839421997596043e-07, + "loss": 0.0004, + "step": 56454 + }, + { + "epoch": 7.4, + "learning_rate": 2.9826564925212966e-07, + "loss": 0.0002, + "step": 56455 + }, + { + "epoch": 7.4, + "learning_rate": 2.9813710581390933e-07, + "loss": 0.0007, + "step": 56456 + }, + { + "epoch": 7.4, + "learning_rate": 2.9800858966166357e-07, + "loss": 0.0004, + "step": 56457 + }, + { + "epoch": 7.4, + "learning_rate": 2.9788010079575104e-07, + "loss": 0.0002, + "step": 56458 + }, + { + "epoch": 7.4, + "learning_rate": 2.977516392165336e-07, + "loss": 0.0002, + "step": 56459 + }, + { + "epoch": 7.4, + "learning_rate": 2.9762320492437327e-07, + "loss": 0.0004, + "step": 56460 + }, + { + "epoch": 7.4, + "learning_rate": 2.9749479791963075e-07, + "loss": 0.0003, + "step": 56461 + }, + { + "epoch": 7.4, + "learning_rate": 2.973664182026681e-07, + "loss": 0.0006, + "step": 56462 + }, + { + "epoch": 7.4, + "learning_rate": 2.9723806577384496e-07, + "loss": 0.0005, + "step": 56463 + }, + { + "epoch": 7.4, + "learning_rate": 2.9710974063352326e-07, + "loss": 0.0003, + "step": 56464 + }, + { + "epoch": 7.4, + "learning_rate": 2.9698144278206387e-07, + "loss": 0.0014, + "step": 56465 + }, + { + "epoch": 7.4, + "learning_rate": 2.968531722198287e-07, + "loss": 0.0005, + "step": 56466 + }, + { + "epoch": 7.4, + "learning_rate": 2.967249289471763e-07, + "loss": 0.0005, + "step": 56467 + }, + { + "epoch": 7.4, + "learning_rate": 2.965967129644665e-07, + "loss": 0.0003, + "step": 56468 + }, + { + "epoch": 7.4, + "learning_rate": 2.964685242720644e-07, + "loss": 0.0004, + "step": 56469 + }, + { + "epoch": 7.4, + "learning_rate": 2.963403628703265e-07, + "loss": 0.0009, + "step": 56470 + }, + { + "epoch": 7.4, + "learning_rate": 2.9621222875961477e-07, + "loss": 0.0007, + "step": 56471 + }, + { + "epoch": 7.4, + "learning_rate": 2.960841219402888e-07, + "loss": 0.0009, + "step": 56472 + }, + { + "epoch": 7.4, + "learning_rate": 2.959560424127095e-07, + "loss": 0.0006, + "step": 56473 + }, + { + "epoch": 7.4, + "learning_rate": 2.958279901772387e-07, + "loss": 0.0005, + "step": 56474 + }, + { + "epoch": 7.4, + "learning_rate": 2.9569996523423406e-07, + "loss": 0.0004, + "step": 56475 + }, + { + "epoch": 7.4, + "learning_rate": 2.9557196758405514e-07, + "loss": 0.0002, + "step": 56476 + }, + { + "epoch": 7.4, + "learning_rate": 2.9544399722706507e-07, + "loss": 0.0006, + "step": 56477 + }, + { + "epoch": 7.4, + "learning_rate": 2.953160541636213e-07, + "loss": 0.0008, + "step": 56478 + }, + { + "epoch": 7.4, + "learning_rate": 2.9518813839408357e-07, + "loss": 0.0003, + "step": 56479 + }, + { + "epoch": 7.4, + "learning_rate": 2.9506024991881375e-07, + "loss": 0.0005, + "step": 56480 + }, + { + "epoch": 7.4, + "learning_rate": 2.949323887381694e-07, + "loss": 0.0005, + "step": 56481 + }, + { + "epoch": 7.4, + "learning_rate": 2.9480455485251137e-07, + "loss": 0.0004, + "step": 56482 + }, + { + "epoch": 7.4, + "learning_rate": 2.9467674826219704e-07, + "loss": 0.0002, + "step": 56483 + }, + { + "epoch": 7.4, + "learning_rate": 2.945489689675896e-07, + "loss": 0.0002, + "step": 56484 + }, + { + "epoch": 7.4, + "learning_rate": 2.9442121696904527e-07, + "loss": 0.0003, + "step": 56485 + }, + { + "epoch": 7.4, + "learning_rate": 2.9429349226692495e-07, + "loss": 0.0003, + "step": 56486 + }, + { + "epoch": 7.4, + "learning_rate": 2.9416579486158723e-07, + "loss": 0.0002, + "step": 56487 + }, + { + "epoch": 7.4, + "learning_rate": 2.940381247533908e-07, + "loss": 0.0003, + "step": 56488 + }, + { + "epoch": 7.4, + "learning_rate": 2.9391048194269633e-07, + "loss": 0.0004, + "step": 56489 + }, + { + "epoch": 7.4, + "learning_rate": 2.9378286642986143e-07, + "loss": 0.0003, + "step": 56490 + }, + { + "epoch": 7.4, + "learning_rate": 2.9365527821524464e-07, + "loss": 0.0005, + "step": 56491 + }, + { + "epoch": 7.4, + "learning_rate": 2.935277172992057e-07, + "loss": 0.0018, + "step": 56492 + }, + { + "epoch": 7.4, + "learning_rate": 2.934001836821032e-07, + "loss": 0.0004, + "step": 56493 + }, + { + "epoch": 7.4, + "learning_rate": 2.9327267736429574e-07, + "loss": 0.0009, + "step": 56494 + }, + { + "epoch": 7.4, + "learning_rate": 2.9314519834614197e-07, + "loss": 0.0003, + "step": 56495 + }, + { + "epoch": 7.4, + "learning_rate": 2.9301774662800043e-07, + "loss": 0.0002, + "step": 56496 + }, + { + "epoch": 7.4, + "learning_rate": 2.9289032221022974e-07, + "loss": 0.0003, + "step": 56497 + }, + { + "epoch": 7.4, + "learning_rate": 2.927629250931874e-07, + "loss": 0.0006, + "step": 56498 + }, + { + "epoch": 7.4, + "learning_rate": 2.9263555527723306e-07, + "loss": 0.0007, + "step": 56499 + }, + { + "epoch": 7.4, + "learning_rate": 2.9250821276272547e-07, + "loss": 0.0004, + "step": 56500 + }, + { + "epoch": 7.4, + "learning_rate": 2.9238089755001973e-07, + "loss": 0.0007, + "step": 56501 + }, + { + "epoch": 7.4, + "learning_rate": 2.922536096394757e-07, + "loss": 0.0008, + "step": 56502 + }, + { + "epoch": 7.4, + "learning_rate": 2.9212634903145187e-07, + "loss": 0.0002, + "step": 56503 + }, + { + "epoch": 7.4, + "learning_rate": 2.919991157263058e-07, + "loss": 0.0005, + "step": 56504 + }, + { + "epoch": 7.4, + "learning_rate": 2.9187190972439605e-07, + "loss": 0.0002, + "step": 56505 + }, + { + "epoch": 7.4, + "learning_rate": 2.917447310260779e-07, + "loss": 0.0007, + "step": 56506 + }, + { + "epoch": 7.4, + "learning_rate": 2.916175796317122e-07, + "loss": 0.0003, + "step": 56507 + }, + { + "epoch": 7.4, + "learning_rate": 2.914904555416553e-07, + "loss": 0.0004, + "step": 56508 + }, + { + "epoch": 7.4, + "learning_rate": 2.9136335875626367e-07, + "loss": 0.0004, + "step": 56509 + }, + { + "epoch": 7.4, + "learning_rate": 2.9123628927589466e-07, + "loss": 0.0003, + "step": 56510 + }, + { + "epoch": 7.4, + "learning_rate": 2.9110924710090803e-07, + "loss": 0.0004, + "step": 56511 + }, + { + "epoch": 7.4, + "learning_rate": 2.909822322316591e-07, + "loss": 0.0008, + "step": 56512 + }, + { + "epoch": 7.4, + "learning_rate": 2.9085524466850535e-07, + "loss": 0.0003, + "step": 56513 + }, + { + "epoch": 7.4, + "learning_rate": 2.907282844118031e-07, + "loss": 0.0021, + "step": 56514 + }, + { + "epoch": 7.4, + "learning_rate": 2.9060135146191216e-07, + "loss": 0.0006, + "step": 56515 + }, + { + "epoch": 7.4, + "learning_rate": 2.904744458191888e-07, + "loss": 0.0005, + "step": 56516 + }, + { + "epoch": 7.4, + "learning_rate": 2.903475674839873e-07, + "loss": 0.0009, + "step": 56517 + }, + { + "epoch": 7.4, + "learning_rate": 2.902207164566662e-07, + "loss": 0.0009, + "step": 56518 + }, + { + "epoch": 7.4, + "learning_rate": 2.900938927375829e-07, + "loss": 0.0002, + "step": 56519 + }, + { + "epoch": 7.4, + "learning_rate": 2.8996709632709394e-07, + "loss": 0.0004, + "step": 56520 + }, + { + "epoch": 7.4, + "learning_rate": 2.8984032722555347e-07, + "loss": 0.0003, + "step": 56521 + }, + { + "epoch": 7.4, + "learning_rate": 2.8971358543332215e-07, + "loss": 0.0003, + "step": 56522 + }, + { + "epoch": 7.4, + "learning_rate": 2.895868709507532e-07, + "loss": 0.0004, + "step": 56523 + }, + { + "epoch": 7.4, + "learning_rate": 2.894601837782052e-07, + "loss": 0.0004, + "step": 56524 + }, + { + "epoch": 7.4, + "learning_rate": 2.8933352391603107e-07, + "loss": 0.0004, + "step": 56525 + }, + { + "epoch": 7.4, + "learning_rate": 2.892068913645896e-07, + "loss": 0.0003, + "step": 56526 + }, + { + "epoch": 7.4, + "learning_rate": 2.890802861242381e-07, + "loss": 0.0002, + "step": 56527 + }, + { + "epoch": 7.4, + "learning_rate": 2.889537081953309e-07, + "loss": 0.0008, + "step": 56528 + }, + { + "epoch": 7.4, + "learning_rate": 2.888271575782231e-07, + "loss": 0.0005, + "step": 56529 + }, + { + "epoch": 7.41, + "learning_rate": 2.8870063427327343e-07, + "loss": 0.0002, + "step": 56530 + }, + { + "epoch": 7.41, + "learning_rate": 2.88574138280836e-07, + "loss": 0.0002, + "step": 56531 + }, + { + "epoch": 7.41, + "learning_rate": 2.8844766960126616e-07, + "loss": 0.0003, + "step": 56532 + }, + { + "epoch": 7.41, + "learning_rate": 2.88321228234919e-07, + "loss": 0.0003, + "step": 56533 + }, + { + "epoch": 7.41, + "learning_rate": 2.8819481418215225e-07, + "loss": 0.0003, + "step": 56534 + }, + { + "epoch": 7.41, + "learning_rate": 2.88068427443321e-07, + "loss": 0.0002, + "step": 56535 + }, + { + "epoch": 7.41, + "learning_rate": 2.8794206801877836e-07, + "loss": 0.0005, + "step": 56536 + }, + { + "epoch": 7.41, + "learning_rate": 2.878157359088818e-07, + "loss": 0.0004, + "step": 56537 + }, + { + "epoch": 7.41, + "learning_rate": 2.876894311139866e-07, + "loss": 0.0008, + "step": 56538 + }, + { + "epoch": 7.41, + "learning_rate": 2.875631536344492e-07, + "loss": 0.0008, + "step": 56539 + }, + { + "epoch": 7.41, + "learning_rate": 2.874369034706215e-07, + "loss": 0.0007, + "step": 56540 + }, + { + "epoch": 7.41, + "learning_rate": 2.8731068062285983e-07, + "loss": 0.0004, + "step": 56541 + }, + { + "epoch": 7.41, + "learning_rate": 2.871844850915217e-07, + "loss": 0.0006, + "step": 56542 + }, + { + "epoch": 7.41, + "learning_rate": 2.8705831687695916e-07, + "loss": 0.0004, + "step": 56543 + }, + { + "epoch": 7.41, + "learning_rate": 2.8693217597952626e-07, + "loss": 0.0006, + "step": 56544 + }, + { + "epoch": 7.41, + "learning_rate": 2.868060623995816e-07, + "loss": 0.0002, + "step": 56545 + }, + { + "epoch": 7.41, + "learning_rate": 2.8667997613747614e-07, + "loss": 0.0008, + "step": 56546 + }, + { + "epoch": 7.41, + "learning_rate": 2.865539171935661e-07, + "loss": 0.0003, + "step": 56547 + }, + { + "epoch": 7.41, + "learning_rate": 2.8642788556820564e-07, + "loss": 0.0003, + "step": 56548 + }, + { + "epoch": 7.41, + "learning_rate": 2.863018812617502e-07, + "loss": 0.0005, + "step": 56549 + }, + { + "epoch": 7.41, + "learning_rate": 2.8617590427455377e-07, + "loss": 0.0003, + "step": 56550 + }, + { + "epoch": 7.41, + "learning_rate": 2.860499546069706e-07, + "loss": 0.0001, + "step": 56551 + }, + { + "epoch": 7.41, + "learning_rate": 2.8592403225935373e-07, + "loss": 0.0002, + "step": 56552 + }, + { + "epoch": 7.41, + "learning_rate": 2.857981372320584e-07, + "loss": 0.0006, + "step": 56553 + }, + { + "epoch": 7.41, + "learning_rate": 2.856722695254388e-07, + "loss": 0.0003, + "step": 56554 + }, + { + "epoch": 7.41, + "learning_rate": 2.855464291398491e-07, + "loss": 0.0004, + "step": 56555 + }, + { + "epoch": 7.41, + "learning_rate": 2.854206160756412e-07, + "loss": 0.0009, + "step": 56556 + }, + { + "epoch": 7.41, + "learning_rate": 2.852948303331715e-07, + "loss": 0.0004, + "step": 56557 + }, + { + "epoch": 7.41, + "learning_rate": 2.8516907191279195e-07, + "loss": 0.0003, + "step": 56558 + }, + { + "epoch": 7.41, + "learning_rate": 2.8504334081485783e-07, + "loss": 0.0006, + "step": 56559 + }, + { + "epoch": 7.41, + "learning_rate": 2.8491763703972107e-07, + "loss": 0.0003, + "step": 56560 + }, + { + "epoch": 7.41, + "learning_rate": 2.8479196058773696e-07, + "loss": 0.0003, + "step": 56561 + }, + { + "epoch": 7.41, + "learning_rate": 2.846663114592585e-07, + "loss": 0.0005, + "step": 56562 + }, + { + "epoch": 7.41, + "learning_rate": 2.8454068965463655e-07, + "loss": 0.0008, + "step": 56563 + }, + { + "epoch": 7.41, + "learning_rate": 2.844150951742286e-07, + "loss": 0.0003, + "step": 56564 + }, + { + "epoch": 7.41, + "learning_rate": 2.842895280183855e-07, + "loss": 0.0004, + "step": 56565 + }, + { + "epoch": 7.41, + "learning_rate": 2.841639881874614e-07, + "loss": 0.0002, + "step": 56566 + }, + { + "epoch": 7.41, + "learning_rate": 2.8403847568180707e-07, + "loss": 0.0007, + "step": 56567 + }, + { + "epoch": 7.41, + "learning_rate": 2.8391299050177787e-07, + "loss": 0.0004, + "step": 56568 + }, + { + "epoch": 7.41, + "learning_rate": 2.8378753264772575e-07, + "loss": 0.0006, + "step": 56569 + }, + { + "epoch": 7.41, + "learning_rate": 2.8366210212000365e-07, + "loss": 0.0005, + "step": 56570 + }, + { + "epoch": 7.41, + "learning_rate": 2.8353669891896473e-07, + "loss": 0.0005, + "step": 56571 + }, + { + "epoch": 7.41, + "learning_rate": 2.834113230449609e-07, + "loss": 0.0003, + "step": 56572 + }, + { + "epoch": 7.41, + "learning_rate": 2.832859744983463e-07, + "loss": 0.0003, + "step": 56573 + }, + { + "epoch": 7.41, + "learning_rate": 2.8316065327947285e-07, + "loss": 0.0017, + "step": 56574 + }, + { + "epoch": 7.41, + "learning_rate": 2.8303535938869143e-07, + "loss": 0.0008, + "step": 56575 + }, + { + "epoch": 7.41, + "learning_rate": 2.829100928263562e-07, + "loss": 0.0005, + "step": 56576 + }, + { + "epoch": 7.41, + "learning_rate": 2.8278485359281795e-07, + "loss": 0.0004, + "step": 56577 + }, + { + "epoch": 7.41, + "learning_rate": 2.826596416884297e-07, + "loss": 0.0006, + "step": 56578 + }, + { + "epoch": 7.41, + "learning_rate": 2.8253445711354577e-07, + "loss": 0.0005, + "step": 56579 + }, + { + "epoch": 7.41, + "learning_rate": 2.824092998685135e-07, + "loss": 0.0006, + "step": 56580 + }, + { + "epoch": 7.41, + "learning_rate": 2.8228416995368937e-07, + "loss": 0.0004, + "step": 56581 + }, + { + "epoch": 7.41, + "learning_rate": 2.821590673694241e-07, + "loss": 0.0004, + "step": 56582 + }, + { + "epoch": 7.41, + "learning_rate": 2.820339921160664e-07, + "loss": 0.001, + "step": 56583 + }, + { + "epoch": 7.41, + "learning_rate": 2.8190894419397266e-07, + "loss": 0.0007, + "step": 56584 + }, + { + "epoch": 7.41, + "learning_rate": 2.817839236034925e-07, + "loss": 0.0003, + "step": 56585 + }, + { + "epoch": 7.41, + "learning_rate": 2.816589303449768e-07, + "loss": 0.0003, + "step": 56586 + }, + { + "epoch": 7.41, + "learning_rate": 2.815339644187787e-07, + "loss": 0.0005, + "step": 56587 + }, + { + "epoch": 7.41, + "learning_rate": 2.814090258252478e-07, + "loss": 0.0003, + "step": 56588 + }, + { + "epoch": 7.41, + "learning_rate": 2.812841145647371e-07, + "loss": 0.0004, + "step": 56589 + }, + { + "epoch": 7.41, + "learning_rate": 2.8115923063759656e-07, + "loss": 0.0004, + "step": 56590 + }, + { + "epoch": 7.41, + "learning_rate": 2.8103437404417786e-07, + "loss": 0.0003, + "step": 56591 + }, + { + "epoch": 7.41, + "learning_rate": 2.8090954478483313e-07, + "loss": 0.0004, + "step": 56592 + }, + { + "epoch": 7.41, + "learning_rate": 2.807847428599131e-07, + "loss": 0.0006, + "step": 56593 + }, + { + "epoch": 7.41, + "learning_rate": 2.806599682697675e-07, + "loss": 0.0004, + "step": 56594 + }, + { + "epoch": 7.41, + "learning_rate": 2.805352210147494e-07, + "loss": 0.0011, + "step": 56595 + }, + { + "epoch": 7.41, + "learning_rate": 2.804105010952074e-07, + "loss": 0.0003, + "step": 56596 + }, + { + "epoch": 7.41, + "learning_rate": 2.802858085114946e-07, + "loss": 0.0006, + "step": 56597 + }, + { + "epoch": 7.41, + "learning_rate": 2.801611432639584e-07, + "loss": 0.0003, + "step": 56598 + }, + { + "epoch": 7.41, + "learning_rate": 2.800365053529519e-07, + "loss": 0.0002, + "step": 56599 + }, + { + "epoch": 7.41, + "learning_rate": 2.7991189477882595e-07, + "loss": 0.0008, + "step": 56600 + }, + { + "epoch": 7.41, + "learning_rate": 2.7978731154192917e-07, + "loss": 0.0006, + "step": 56601 + }, + { + "epoch": 7.41, + "learning_rate": 2.7966275564261345e-07, + "loss": 0.0003, + "step": 56602 + }, + { + "epoch": 7.41, + "learning_rate": 2.7953822708122855e-07, + "loss": 0.0005, + "step": 56603 + }, + { + "epoch": 7.41, + "learning_rate": 2.794137258581253e-07, + "loss": 0.0004, + "step": 56604 + }, + { + "epoch": 7.41, + "learning_rate": 2.792892519736534e-07, + "loss": 0.0004, + "step": 56605 + }, + { + "epoch": 7.41, + "learning_rate": 2.791648054281615e-07, + "loss": 0.0003, + "step": 56606 + }, + { + "epoch": 7.42, + "learning_rate": 2.7904038622200147e-07, + "loss": 0.0004, + "step": 56607 + }, + { + "epoch": 7.42, + "learning_rate": 2.7891599435552306e-07, + "loss": 0.0002, + "step": 56608 + }, + { + "epoch": 7.42, + "learning_rate": 2.7879162982907495e-07, + "loss": 0.0005, + "step": 56609 + }, + { + "epoch": 7.42, + "learning_rate": 2.7866729264300893e-07, + "loss": 0.0004, + "step": 56610 + }, + { + "epoch": 7.42, + "learning_rate": 2.7854298279767265e-07, + "loss": 0.0008, + "step": 56611 + }, + { + "epoch": 7.42, + "learning_rate": 2.784187002934169e-07, + "loss": 0.0001, + "step": 56612 + }, + { + "epoch": 7.42, + "learning_rate": 2.782944451305891e-07, + "loss": 0.0002, + "step": 56613 + }, + { + "epoch": 7.42, + "learning_rate": 2.781702173095424e-07, + "loss": 0.0006, + "step": 56614 + }, + { + "epoch": 7.42, + "learning_rate": 2.7804601683062426e-07, + "loss": 0.0005, + "step": 56615 + }, + { + "epoch": 7.42, + "learning_rate": 2.7792184369418443e-07, + "loss": 0.0013, + "step": 56616 + }, + { + "epoch": 7.42, + "learning_rate": 2.777976979005692e-07, + "loss": 0.0008, + "step": 56617 + }, + { + "epoch": 7.42, + "learning_rate": 2.7767357945013283e-07, + "loss": 0.0001, + "step": 56618 + }, + { + "epoch": 7.42, + "learning_rate": 2.775494883432206e-07, + "loss": 0.0004, + "step": 56619 + }, + { + "epoch": 7.42, + "learning_rate": 2.774254245801833e-07, + "loss": 0.0005, + "step": 56620 + }, + { + "epoch": 7.42, + "learning_rate": 2.7730138816136844e-07, + "loss": 0.0003, + "step": 56621 + }, + { + "epoch": 7.42, + "learning_rate": 2.7717737908712683e-07, + "loss": 0.0004, + "step": 56622 + }, + { + "epoch": 7.42, + "learning_rate": 2.7705339735780487e-07, + "loss": 0.0002, + "step": 56623 + }, + { + "epoch": 7.42, + "learning_rate": 2.769294429737523e-07, + "loss": 0.0003, + "step": 56624 + }, + { + "epoch": 7.42, + "learning_rate": 2.768055159353178e-07, + "loss": 0.0003, + "step": 56625 + }, + { + "epoch": 7.42, + "learning_rate": 2.7668161624285096e-07, + "loss": 0.0004, + "step": 56626 + }, + { + "epoch": 7.42, + "learning_rate": 2.765577438966993e-07, + "loss": 0.0004, + "step": 56627 + }, + { + "epoch": 7.42, + "learning_rate": 2.764338988972104e-07, + "loss": 0.0004, + "step": 56628 + }, + { + "epoch": 7.42, + "learning_rate": 2.763100812447339e-07, + "loss": 0.0004, + "step": 56629 + }, + { + "epoch": 7.42, + "learning_rate": 2.7618629093961733e-07, + "loss": 0.0006, + "step": 56630 + }, + { + "epoch": 7.42, + "learning_rate": 2.7606252798220935e-07, + "loss": 0.0005, + "step": 56631 + }, + { + "epoch": 7.42, + "learning_rate": 2.7593879237285737e-07, + "loss": 0.0004, + "step": 56632 + }, + { + "epoch": 7.42, + "learning_rate": 2.7581508411190893e-07, + "loss": 0.0004, + "step": 56633 + }, + { + "epoch": 7.42, + "learning_rate": 2.7569140319971376e-07, + "loss": 0.0003, + "step": 56634 + }, + { + "epoch": 7.42, + "learning_rate": 2.755677496366183e-07, + "loss": 0.001, + "step": 56635 + }, + { + "epoch": 7.42, + "learning_rate": 2.754441234229699e-07, + "loss": 0.0003, + "step": 56636 + }, + { + "epoch": 7.42, + "learning_rate": 2.753205245591184e-07, + "loss": 0.0008, + "step": 56637 + }, + { + "epoch": 7.42, + "learning_rate": 2.7519695304541016e-07, + "loss": 0.0007, + "step": 56638 + }, + { + "epoch": 7.42, + "learning_rate": 2.750734088821916e-07, + "loss": 0.0004, + "step": 56639 + }, + { + "epoch": 7.42, + "learning_rate": 2.749498920698113e-07, + "loss": 0.0006, + "step": 56640 + }, + { + "epoch": 7.42, + "learning_rate": 2.748264026086167e-07, + "loss": 0.0006, + "step": 56641 + }, + { + "epoch": 7.42, + "learning_rate": 2.747029404989554e-07, + "loss": 0.0003, + "step": 56642 + }, + { + "epoch": 7.42, + "learning_rate": 2.7457950574117375e-07, + "loss": 0.0007, + "step": 56643 + }, + { + "epoch": 7.42, + "learning_rate": 2.7445609833561925e-07, + "loss": 0.0002, + "step": 56644 + }, + { + "epoch": 7.42, + "learning_rate": 2.743327182826394e-07, + "loss": 0.0005, + "step": 56645 + }, + { + "epoch": 7.42, + "learning_rate": 2.742093655825806e-07, + "loss": 0.0004, + "step": 56646 + }, + { + "epoch": 7.42, + "learning_rate": 2.7408604023579033e-07, + "loss": 0.0006, + "step": 56647 + }, + { + "epoch": 7.42, + "learning_rate": 2.73962742242615e-07, + "loss": 0.0005, + "step": 56648 + }, + { + "epoch": 7.42, + "learning_rate": 2.73839471603401e-07, + "loss": 0.0003, + "step": 56649 + }, + { + "epoch": 7.42, + "learning_rate": 2.7371622831849685e-07, + "loss": 0.0006, + "step": 56650 + }, + { + "epoch": 7.42, + "learning_rate": 2.735930123882458e-07, + "loss": 0.0005, + "step": 56651 + }, + { + "epoch": 7.42, + "learning_rate": 2.734698238129985e-07, + "loss": 0.0001, + "step": 56652 + }, + { + "epoch": 7.42, + "learning_rate": 2.7334666259309804e-07, + "loss": 0.0005, + "step": 56653 + }, + { + "epoch": 7.42, + "learning_rate": 2.7322352872889313e-07, + "loss": 0.001, + "step": 56654 + }, + { + "epoch": 7.42, + "learning_rate": 2.7310042222072786e-07, + "loss": 0.0007, + "step": 56655 + }, + { + "epoch": 7.42, + "learning_rate": 2.7297734306894975e-07, + "loss": 0.0005, + "step": 56656 + }, + { + "epoch": 7.42, + "learning_rate": 2.728542912739063e-07, + "loss": 0.0001, + "step": 56657 + }, + { + "epoch": 7.42, + "learning_rate": 2.7273126683594165e-07, + "loss": 0.0003, + "step": 56658 + }, + { + "epoch": 7.42, + "learning_rate": 2.726082697554011e-07, + "loss": 0.0005, + "step": 56659 + }, + { + "epoch": 7.42, + "learning_rate": 2.724853000326322e-07, + "loss": 0.0005, + "step": 56660 + }, + { + "epoch": 7.42, + "learning_rate": 2.7236235766798126e-07, + "loss": 0.0009, + "step": 56661 + }, + { + "epoch": 7.42, + "learning_rate": 2.7223944266179356e-07, + "loss": 0.0005, + "step": 56662 + }, + { + "epoch": 7.42, + "learning_rate": 2.7211655501441224e-07, + "loss": 0.0006, + "step": 56663 + }, + { + "epoch": 7.42, + "learning_rate": 2.7199369472618696e-07, + "loss": 0.0002, + "step": 56664 + }, + { + "epoch": 7.42, + "learning_rate": 2.718708617974608e-07, + "loss": 0.0006, + "step": 56665 + }, + { + "epoch": 7.42, + "learning_rate": 2.71748056228579e-07, + "loss": 0.0007, + "step": 56666 + }, + { + "epoch": 7.42, + "learning_rate": 2.7162527801988804e-07, + "loss": 0.0004, + "step": 56667 + }, + { + "epoch": 7.42, + "learning_rate": 2.715025271717342e-07, + "loss": 0.0005, + "step": 56668 + }, + { + "epoch": 7.42, + "learning_rate": 2.713798036844606e-07, + "loss": 0.0003, + "step": 56669 + }, + { + "epoch": 7.42, + "learning_rate": 2.712571075584136e-07, + "loss": 0.0003, + "step": 56670 + }, + { + "epoch": 7.42, + "learning_rate": 2.711344387939363e-07, + "loss": 0.0003, + "step": 56671 + }, + { + "epoch": 7.42, + "learning_rate": 2.710117973913773e-07, + "loss": 0.0004, + "step": 56672 + }, + { + "epoch": 7.42, + "learning_rate": 2.708891833510796e-07, + "loss": 0.0006, + "step": 56673 + }, + { + "epoch": 7.42, + "learning_rate": 2.7076659667338634e-07, + "loss": 0.0007, + "step": 56674 + }, + { + "epoch": 7.42, + "learning_rate": 2.70644037358645e-07, + "loss": 0.0004, + "step": 56675 + }, + { + "epoch": 7.42, + "learning_rate": 2.7052150540719857e-07, + "loss": 0.0004, + "step": 56676 + }, + { + "epoch": 7.42, + "learning_rate": 2.7039900081939353e-07, + "loss": 0.0006, + "step": 56677 + }, + { + "epoch": 7.42, + "learning_rate": 2.7027652359556956e-07, + "loss": 0.0016, + "step": 56678 + }, + { + "epoch": 7.42, + "learning_rate": 2.701540737360786e-07, + "loss": 0.0006, + "step": 56679 + }, + { + "epoch": 7.42, + "learning_rate": 2.700316512412593e-07, + "loss": 0.001, + "step": 56680 + }, + { + "epoch": 7.42, + "learning_rate": 2.699092561114591e-07, + "loss": 0.0002, + "step": 56681 + }, + { + "epoch": 7.42, + "learning_rate": 2.6978688834702005e-07, + "loss": 0.0005, + "step": 56682 + }, + { + "epoch": 7.43, + "learning_rate": 2.6966454794828734e-07, + "loss": 0.0003, + "step": 56683 + }, + { + "epoch": 7.43, + "learning_rate": 2.6954223491560627e-07, + "loss": 0.0009, + "step": 56684 + }, + { + "epoch": 7.43, + "learning_rate": 2.694199492493166e-07, + "loss": 0.0004, + "step": 56685 + }, + { + "epoch": 7.43, + "learning_rate": 2.69297690949768e-07, + "loss": 0.001, + "step": 56686 + }, + { + "epoch": 7.43, + "learning_rate": 2.691754600173002e-07, + "loss": 0.0004, + "step": 56687 + }, + { + "epoch": 7.43, + "learning_rate": 2.690532564522586e-07, + "loss": 0.0002, + "step": 56688 + }, + { + "epoch": 7.43, + "learning_rate": 2.6893108025498605e-07, + "loss": 0.0004, + "step": 56689 + }, + { + "epoch": 7.43, + "learning_rate": 2.6880893142582577e-07, + "loss": 0.0014, + "step": 56690 + }, + { + "epoch": 7.43, + "learning_rate": 2.6868680996512297e-07, + "loss": 0.0007, + "step": 56691 + }, + { + "epoch": 7.43, + "learning_rate": 2.6856471587322074e-07, + "loss": 0.0004, + "step": 56692 + }, + { + "epoch": 7.43, + "learning_rate": 2.68442649150461e-07, + "loss": 0.0003, + "step": 56693 + }, + { + "epoch": 7.43, + "learning_rate": 2.683206097971902e-07, + "loss": 0.0006, + "step": 56694 + }, + { + "epoch": 7.43, + "learning_rate": 2.68198597813748e-07, + "loss": 0.0003, + "step": 56695 + }, + { + "epoch": 7.43, + "learning_rate": 2.680766132004797e-07, + "loss": 0.0007, + "step": 56696 + }, + { + "epoch": 7.43, + "learning_rate": 2.679546559577262e-07, + "loss": 0.0007, + "step": 56697 + }, + { + "epoch": 7.43, + "learning_rate": 2.678327260858327e-07, + "loss": 0.0005, + "step": 56698 + }, + { + "epoch": 7.43, + "learning_rate": 2.6771082358514225e-07, + "loss": 0.0005, + "step": 56699 + }, + { + "epoch": 7.43, + "learning_rate": 2.675889484559957e-07, + "loss": 0.0005, + "step": 56700 + }, + { + "epoch": 7.43, + "learning_rate": 2.674671006987362e-07, + "loss": 0.0007, + "step": 56701 + }, + { + "epoch": 7.43, + "learning_rate": 2.6734528031370887e-07, + "loss": 0.0007, + "step": 56702 + }, + { + "epoch": 7.43, + "learning_rate": 2.672234873012547e-07, + "loss": 0.0002, + "step": 56703 + }, + { + "epoch": 7.43, + "learning_rate": 2.6710172166171553e-07, + "loss": 0.0001, + "step": 56704 + }, + { + "epoch": 7.43, + "learning_rate": 2.6697998339543343e-07, + "loss": 0.0004, + "step": 56705 + }, + { + "epoch": 7.43, + "learning_rate": 2.668582725027535e-07, + "loss": 0.0007, + "step": 56706 + }, + { + "epoch": 7.43, + "learning_rate": 2.667365889840157e-07, + "loss": 0.001, + "step": 56707 + }, + { + "epoch": 7.43, + "learning_rate": 2.6661493283956283e-07, + "loss": 0.0003, + "step": 56708 + }, + { + "epoch": 7.43, + "learning_rate": 2.6649330406973704e-07, + "loss": 0.0005, + "step": 56709 + }, + { + "epoch": 7.43, + "learning_rate": 2.6637170267488134e-07, + "loss": 0.0002, + "step": 56710 + }, + { + "epoch": 7.43, + "learning_rate": 2.662501286553354e-07, + "loss": 0.0005, + "step": 56711 + }, + { + "epoch": 7.43, + "learning_rate": 2.6612858201144453e-07, + "loss": 0.0004, + "step": 56712 + }, + { + "epoch": 7.43, + "learning_rate": 2.6600706274354624e-07, + "loss": 0.0005, + "step": 56713 + }, + { + "epoch": 7.43, + "learning_rate": 2.6588557085198696e-07, + "loss": 0.0004, + "step": 56714 + }, + { + "epoch": 7.43, + "learning_rate": 2.657641063371053e-07, + "loss": 0.0012, + "step": 56715 + }, + { + "epoch": 7.43, + "learning_rate": 2.656426691992431e-07, + "loss": 0.0004, + "step": 56716 + }, + { + "epoch": 7.43, + "learning_rate": 2.655212594387435e-07, + "loss": 0.0003, + "step": 56717 + }, + { + "epoch": 7.43, + "learning_rate": 2.653998770559474e-07, + "loss": 0.0004, + "step": 56718 + }, + { + "epoch": 7.43, + "learning_rate": 2.652785220511944e-07, + "loss": 0.0006, + "step": 56719 + }, + { + "epoch": 7.43, + "learning_rate": 2.651571944248277e-07, + "loss": 0.0007, + "step": 56720 + }, + { + "epoch": 7.43, + "learning_rate": 2.65035894177188e-07, + "loss": 0.0003, + "step": 56721 + }, + { + "epoch": 7.43, + "learning_rate": 2.649146213086151e-07, + "loss": 0.0005, + "step": 56722 + }, + { + "epoch": 7.43, + "learning_rate": 2.647933758194532e-07, + "loss": 0.0002, + "step": 56723 + }, + { + "epoch": 7.43, + "learning_rate": 2.6467215771004087e-07, + "loss": 0.001, + "step": 56724 + }, + { + "epoch": 7.43, + "learning_rate": 2.6455096698072003e-07, + "loss": 0.0003, + "step": 56725 + }, + { + "epoch": 7.43, + "learning_rate": 2.644298036318305e-07, + "loss": 0.0003, + "step": 56726 + }, + { + "epoch": 7.43, + "learning_rate": 2.643086676637152e-07, + "loss": 0.0003, + "step": 56727 + }, + { + "epoch": 7.43, + "learning_rate": 2.641875590767118e-07, + "loss": 0.0009, + "step": 56728 + }, + { + "epoch": 7.43, + "learning_rate": 2.6406647787116215e-07, + "loss": 0.0009, + "step": 56729 + }, + { + "epoch": 7.43, + "learning_rate": 2.6394542404740817e-07, + "loss": 0.0002, + "step": 56730 + }, + { + "epoch": 7.43, + "learning_rate": 2.638243976057886e-07, + "loss": 0.0006, + "step": 56731 + }, + { + "epoch": 7.43, + "learning_rate": 2.6370339854664526e-07, + "loss": 0.0003, + "step": 56732 + }, + { + "epoch": 7.43, + "learning_rate": 2.6358242687031576e-07, + "loss": 0.0004, + "step": 56733 + }, + { + "epoch": 7.43, + "learning_rate": 2.634614825771442e-07, + "loss": 0.0003, + "step": 56734 + }, + { + "epoch": 7.43, + "learning_rate": 2.63340565667467e-07, + "loss": 0.001, + "step": 56735 + }, + { + "epoch": 7.43, + "learning_rate": 2.6321967614162723e-07, + "loss": 0.0013, + "step": 56736 + }, + { + "epoch": 7.43, + "learning_rate": 2.6309881399996353e-07, + "loss": 0.0003, + "step": 56737 + }, + { + "epoch": 7.43, + "learning_rate": 2.6297797924281556e-07, + "loss": 0.0002, + "step": 56738 + }, + { + "epoch": 7.43, + "learning_rate": 2.6285717187052304e-07, + "loss": 0.001, + "step": 56739 + }, + { + "epoch": 7.43, + "learning_rate": 2.627363918834269e-07, + "loss": 0.0007, + "step": 56740 + }, + { + "epoch": 7.43, + "learning_rate": 2.6261563928186685e-07, + "loss": 0.0003, + "step": 56741 + }, + { + "epoch": 7.43, + "learning_rate": 2.6249491406618033e-07, + "loss": 0.0006, + "step": 56742 + }, + { + "epoch": 7.43, + "learning_rate": 2.623742162367082e-07, + "loss": 0.0005, + "step": 56743 + }, + { + "epoch": 7.43, + "learning_rate": 2.6225354579379026e-07, + "loss": 0.0005, + "step": 56744 + }, + { + "epoch": 7.43, + "learning_rate": 2.6213290273776614e-07, + "loss": 0.0003, + "step": 56745 + }, + { + "epoch": 7.43, + "learning_rate": 2.620122870689756e-07, + "loss": 0.0002, + "step": 56746 + }, + { + "epoch": 7.43, + "learning_rate": 2.6189169878775514e-07, + "loss": 0.0003, + "step": 56747 + }, + { + "epoch": 7.43, + "learning_rate": 2.617711378944465e-07, + "loss": 0.0004, + "step": 56748 + }, + { + "epoch": 7.43, + "learning_rate": 2.616506043893885e-07, + "loss": 0.0008, + "step": 56749 + }, + { + "epoch": 7.43, + "learning_rate": 2.6153009827291854e-07, + "loss": 0.0007, + "step": 56750 + }, + { + "epoch": 7.43, + "learning_rate": 2.6140961954537635e-07, + "loss": 0.0002, + "step": 56751 + }, + { + "epoch": 7.43, + "learning_rate": 2.612891682071028e-07, + "loss": 0.0007, + "step": 56752 + }, + { + "epoch": 7.43, + "learning_rate": 2.611687442584332e-07, + "loss": 0.0002, + "step": 56753 + }, + { + "epoch": 7.43, + "learning_rate": 2.610483476997083e-07, + "loss": 0.0002, + "step": 56754 + }, + { + "epoch": 7.43, + "learning_rate": 2.609279785312657e-07, + "loss": 0.0003, + "step": 56755 + }, + { + "epoch": 7.43, + "learning_rate": 2.608076367534451e-07, + "loss": 0.0004, + "step": 56756 + }, + { + "epoch": 7.43, + "learning_rate": 2.606873223665851e-07, + "loss": 0.0003, + "step": 56757 + }, + { + "epoch": 7.43, + "learning_rate": 2.6056703537102214e-07, + "loss": 0.0006, + "step": 56758 + }, + { + "epoch": 7.44, + "learning_rate": 2.60446775767097e-07, + "loss": 0.0004, + "step": 56759 + }, + { + "epoch": 7.44, + "learning_rate": 2.603265435551461e-07, + "loss": 0.0005, + "step": 56760 + }, + { + "epoch": 7.44, + "learning_rate": 2.6020633873550807e-07, + "loss": 0.0002, + "step": 56761 + }, + { + "epoch": 7.44, + "learning_rate": 2.6008616130852036e-07, + "loss": 0.0004, + "step": 56762 + }, + { + "epoch": 7.44, + "learning_rate": 2.599660112745228e-07, + "loss": 0.0007, + "step": 56763 + }, + { + "epoch": 7.44, + "learning_rate": 2.5984588863385175e-07, + "loss": 0.0009, + "step": 56764 + }, + { + "epoch": 7.44, + "learning_rate": 2.5972579338684465e-07, + "loss": 0.0003, + "step": 56765 + }, + { + "epoch": 7.44, + "learning_rate": 2.596057255338402e-07, + "loss": 0.0008, + "step": 56766 + }, + { + "epoch": 7.44, + "learning_rate": 2.594856850751759e-07, + "loss": 0.0002, + "step": 56767 + }, + { + "epoch": 7.44, + "learning_rate": 2.5936567201119036e-07, + "loss": 0.0009, + "step": 56768 + }, + { + "epoch": 7.44, + "learning_rate": 2.592456863422199e-07, + "loss": 0.0009, + "step": 56769 + }, + { + "epoch": 7.44, + "learning_rate": 2.5912572806860213e-07, + "loss": 0.0005, + "step": 56770 + }, + { + "epoch": 7.44, + "learning_rate": 2.590057971906745e-07, + "loss": 0.0003, + "step": 56771 + }, + { + "epoch": 7.44, + "learning_rate": 2.588858937087746e-07, + "loss": 0.0005, + "step": 56772 + }, + { + "epoch": 7.44, + "learning_rate": 2.5876601762323873e-07, + "loss": 0.0008, + "step": 56773 + }, + { + "epoch": 7.44, + "learning_rate": 2.586461689344044e-07, + "loss": 0.0003, + "step": 56774 + }, + { + "epoch": 7.44, + "learning_rate": 2.585263476426103e-07, + "loss": 0.0003, + "step": 56775 + }, + { + "epoch": 7.44, + "learning_rate": 2.584065537481906e-07, + "loss": 0.0005, + "step": 56776 + }, + { + "epoch": 7.44, + "learning_rate": 2.5828678725148494e-07, + "loss": 0.0007, + "step": 56777 + }, + { + "epoch": 7.44, + "learning_rate": 2.5816704815282753e-07, + "loss": 0.001, + "step": 56778 + }, + { + "epoch": 7.44, + "learning_rate": 2.580473364525582e-07, + "loss": 0.0003, + "step": 56779 + }, + { + "epoch": 7.44, + "learning_rate": 2.5792765215101103e-07, + "loss": 0.0005, + "step": 56780 + }, + { + "epoch": 7.44, + "learning_rate": 2.578079952485235e-07, + "loss": 0.0003, + "step": 56781 + }, + { + "epoch": 7.44, + "learning_rate": 2.576883657454321e-07, + "loss": 0.0002, + "step": 56782 + }, + { + "epoch": 7.44, + "learning_rate": 2.575687636420743e-07, + "loss": 0.0004, + "step": 56783 + }, + { + "epoch": 7.44, + "learning_rate": 2.5744918893878534e-07, + "loss": 0.0009, + "step": 56784 + }, + { + "epoch": 7.44, + "learning_rate": 2.5732964163590057e-07, + "loss": 0.0007, + "step": 56785 + }, + { + "epoch": 7.44, + "learning_rate": 2.5721012173375746e-07, + "loss": 0.0006, + "step": 56786 + }, + { + "epoch": 7.44, + "learning_rate": 2.5709062923269243e-07, + "loss": 0.0001, + "step": 56787 + }, + { + "epoch": 7.44, + "learning_rate": 2.569711641330419e-07, + "loss": 0.0006, + "step": 56788 + }, + { + "epoch": 7.44, + "learning_rate": 2.5685172643514e-07, + "loss": 0.0008, + "step": 56789 + }, + { + "epoch": 7.44, + "learning_rate": 2.5673231613932425e-07, + "loss": 0.0007, + "step": 56790 + }, + { + "epoch": 7.44, + "learning_rate": 2.5661293324592995e-07, + "loss": 0.0011, + "step": 56791 + }, + { + "epoch": 7.44, + "learning_rate": 2.564935777552935e-07, + "loss": 0.0004, + "step": 56792 + }, + { + "epoch": 7.44, + "learning_rate": 2.56374249667749e-07, + "loss": 0.0003, + "step": 56793 + }, + { + "epoch": 7.44, + "learning_rate": 2.5625494898363415e-07, + "loss": 0.001, + "step": 56794 + }, + { + "epoch": 7.44, + "learning_rate": 2.5613567570328293e-07, + "loss": 0.0021, + "step": 56795 + }, + { + "epoch": 7.44, + "learning_rate": 2.560164298270307e-07, + "loss": 0.0005, + "step": 56796 + }, + { + "epoch": 7.44, + "learning_rate": 2.5589721135521386e-07, + "loss": 0.0004, + "step": 56797 + }, + { + "epoch": 7.44, + "learning_rate": 2.557780202881666e-07, + "loss": 0.0003, + "step": 56798 + }, + { + "epoch": 7.44, + "learning_rate": 2.5565885662622526e-07, + "loss": 0.0007, + "step": 56799 + }, + { + "epoch": 7.44, + "learning_rate": 2.5553972036972407e-07, + "loss": 0.0007, + "step": 56800 + }, + { + "epoch": 7.44, + "learning_rate": 2.5542061151899835e-07, + "loss": 0.0003, + "step": 56801 + }, + { + "epoch": 7.44, + "learning_rate": 2.5530153007438443e-07, + "loss": 0.0005, + "step": 56802 + }, + { + "epoch": 7.44, + "learning_rate": 2.551824760362154e-07, + "loss": 0.0001, + "step": 56803 + }, + { + "epoch": 7.44, + "learning_rate": 2.550634494048265e-07, + "loss": 0.0004, + "step": 56804 + }, + { + "epoch": 7.44, + "learning_rate": 2.5494445018055315e-07, + "loss": 0.0004, + "step": 56805 + }, + { + "epoch": 7.44, + "learning_rate": 2.548254783637294e-07, + "loss": 0.0003, + "step": 56806 + }, + { + "epoch": 7.44, + "learning_rate": 2.547065339546906e-07, + "loss": 0.0007, + "step": 56807 + }, + { + "epoch": 7.44, + "learning_rate": 2.5458761695376867e-07, + "loss": 0.0007, + "step": 56808 + }, + { + "epoch": 7.44, + "learning_rate": 2.544687273613011e-07, + "loss": 0.0005, + "step": 56809 + }, + { + "epoch": 7.44, + "learning_rate": 2.5434986517762216e-07, + "loss": 0.0002, + "step": 56810 + }, + { + "epoch": 7.44, + "learning_rate": 2.5423103040306486e-07, + "loss": 0.0005, + "step": 56811 + }, + { + "epoch": 7.44, + "learning_rate": 2.541122230379622e-07, + "loss": 0.0005, + "step": 56812 + }, + { + "epoch": 7.44, + "learning_rate": 2.5399344308265184e-07, + "loss": 0.0004, + "step": 56813 + }, + { + "epoch": 7.44, + "learning_rate": 2.538746905374656e-07, + "loss": 0.0004, + "step": 56814 + }, + { + "epoch": 7.44, + "learning_rate": 2.5375596540273774e-07, + "loss": 0.0002, + "step": 56815 + }, + { + "epoch": 7.44, + "learning_rate": 2.536372676788024e-07, + "loss": 0.0004, + "step": 56816 + }, + { + "epoch": 7.44, + "learning_rate": 2.535185973659926e-07, + "loss": 0.0003, + "step": 56817 + }, + { + "epoch": 7.44, + "learning_rate": 2.533999544646437e-07, + "loss": 0.0008, + "step": 56818 + }, + { + "epoch": 7.44, + "learning_rate": 2.532813389750866e-07, + "loss": 0.0008, + "step": 56819 + }, + { + "epoch": 7.44, + "learning_rate": 2.531627508976575e-07, + "loss": 0.0004, + "step": 56820 + }, + { + "epoch": 7.44, + "learning_rate": 2.530441902326897e-07, + "loss": 0.0003, + "step": 56821 + }, + { + "epoch": 7.44, + "learning_rate": 2.5292565698051607e-07, + "loss": 0.001, + "step": 56822 + }, + { + "epoch": 7.44, + "learning_rate": 2.528071511414687e-07, + "loss": 0.0004, + "step": 56823 + }, + { + "epoch": 7.44, + "learning_rate": 2.5268867271588395e-07, + "loss": 0.0002, + "step": 56824 + }, + { + "epoch": 7.44, + "learning_rate": 2.5257022170409264e-07, + "loss": 0.0004, + "step": 56825 + }, + { + "epoch": 7.44, + "learning_rate": 2.5245179810642786e-07, + "loss": 0.0003, + "step": 56826 + }, + { + "epoch": 7.44, + "learning_rate": 2.5233340192322265e-07, + "loss": 0.0003, + "step": 56827 + }, + { + "epoch": 7.44, + "learning_rate": 2.522150331548123e-07, + "loss": 0.0004, + "step": 56828 + }, + { + "epoch": 7.44, + "learning_rate": 2.5209669180152774e-07, + "loss": 0.0003, + "step": 56829 + }, + { + "epoch": 7.44, + "learning_rate": 2.519783778637019e-07, + "loss": 0.0005, + "step": 56830 + }, + { + "epoch": 7.44, + "learning_rate": 2.518600913416669e-07, + "loss": 0.0004, + "step": 56831 + }, + { + "epoch": 7.44, + "learning_rate": 2.517418322357568e-07, + "loss": 0.0008, + "step": 56832 + }, + { + "epoch": 7.44, + "learning_rate": 2.5162360054630465e-07, + "loss": 0.0005, + "step": 56833 + }, + { + "epoch": 7.44, + "learning_rate": 2.5150539627364133e-07, + "loss": 0.0003, + "step": 56834 + }, + { + "epoch": 7.44, + "learning_rate": 2.5138721941809885e-07, + "loss": 0.0006, + "step": 56835 + }, + { + "epoch": 7.45, + "learning_rate": 2.512690699800124e-07, + "loss": 0.0002, + "step": 56836 + }, + { + "epoch": 7.45, + "learning_rate": 2.5115094795971185e-07, + "loss": 0.0002, + "step": 56837 + }, + { + "epoch": 7.45, + "learning_rate": 2.5103285335753014e-07, + "loss": 0.0004, + "step": 56838 + }, + { + "epoch": 7.45, + "learning_rate": 2.5091478617379926e-07, + "loss": 0.0006, + "step": 56839 + }, + { + "epoch": 7.45, + "learning_rate": 2.507967464088523e-07, + "loss": 0.0006, + "step": 56840 + }, + { + "epoch": 7.45, + "learning_rate": 2.50678734063019e-07, + "loss": 0.0006, + "step": 56841 + }, + { + "epoch": 7.45, + "learning_rate": 2.505607491366324e-07, + "loss": 0.0002, + "step": 56842 + }, + { + "epoch": 7.45, + "learning_rate": 2.504427916300245e-07, + "loss": 0.0008, + "step": 56843 + }, + { + "epoch": 7.45, + "learning_rate": 2.503248615435283e-07, + "loss": 0.0003, + "step": 56844 + }, + { + "epoch": 7.45, + "learning_rate": 2.502069588774736e-07, + "loss": 0.0004, + "step": 56845 + }, + { + "epoch": 7.45, + "learning_rate": 2.500890836321923e-07, + "loss": 0.0003, + "step": 56846 + }, + { + "epoch": 7.45, + "learning_rate": 2.4997123580801754e-07, + "loss": 0.0004, + "step": 56847 + }, + { + "epoch": 7.45, + "learning_rate": 2.4985341540527787e-07, + "loss": 0.0003, + "step": 56848 + }, + { + "epoch": 7.45, + "learning_rate": 2.497356224243075e-07, + "loss": 0.0006, + "step": 56849 + }, + { + "epoch": 7.45, + "learning_rate": 2.4961785686543504e-07, + "loss": 0.0003, + "step": 56850 + }, + { + "epoch": 7.45, + "learning_rate": 2.4950011872899473e-07, + "loss": 0.0002, + "step": 56851 + }, + { + "epoch": 7.45, + "learning_rate": 2.493824080153151e-07, + "loss": 0.0006, + "step": 56852 + }, + { + "epoch": 7.45, + "learning_rate": 2.492647247247282e-07, + "loss": 0.0004, + "step": 56853 + }, + { + "epoch": 7.45, + "learning_rate": 2.491470688575648e-07, + "loss": 0.0005, + "step": 56854 + }, + { + "epoch": 7.45, + "learning_rate": 2.490294404141558e-07, + "loss": 0.0002, + "step": 56855 + }, + { + "epoch": 7.45, + "learning_rate": 2.489118393948331e-07, + "loss": 0.0019, + "step": 56856 + }, + { + "epoch": 7.45, + "learning_rate": 2.487942657999276e-07, + "loss": 0.0009, + "step": 56857 + }, + { + "epoch": 7.45, + "learning_rate": 2.486767196297668e-07, + "loss": 0.0004, + "step": 56858 + }, + { + "epoch": 7.45, + "learning_rate": 2.485592008846838e-07, + "loss": 0.0004, + "step": 56859 + }, + { + "epoch": 7.45, + "learning_rate": 2.4844170956501044e-07, + "loss": 0.0005, + "step": 56860 + }, + { + "epoch": 7.45, + "learning_rate": 2.4832424567107327e-07, + "loss": 0.0004, + "step": 56861 + }, + { + "epoch": 7.45, + "learning_rate": 2.4820680920320637e-07, + "loss": 0.0006, + "step": 56862 + }, + { + "epoch": 7.45, + "learning_rate": 2.4808940016173844e-07, + "loss": 0.0008, + "step": 56863 + }, + { + "epoch": 7.45, + "learning_rate": 2.47972018546998e-07, + "loss": 0.001, + "step": 56864 + }, + { + "epoch": 7.45, + "learning_rate": 2.478546643593183e-07, + "loss": 0.0005, + "step": 56865 + }, + { + "epoch": 7.45, + "learning_rate": 2.4773733759902773e-07, + "loss": 0.002, + "step": 56866 + }, + { + "epoch": 7.45, + "learning_rate": 2.476200382664573e-07, + "loss": 0.0007, + "step": 56867 + }, + { + "epoch": 7.45, + "learning_rate": 2.475027663619356e-07, + "loss": 0.0002, + "step": 56868 + }, + { + "epoch": 7.45, + "learning_rate": 2.4738552188579233e-07, + "loss": 0.0004, + "step": 56869 + }, + { + "epoch": 7.45, + "learning_rate": 2.472683048383595e-07, + "loss": 0.0005, + "step": 56870 + }, + { + "epoch": 7.45, + "learning_rate": 2.471511152199646e-07, + "loss": 0.0007, + "step": 56871 + }, + { + "epoch": 7.45, + "learning_rate": 2.4703395303093735e-07, + "loss": 0.0005, + "step": 56872 + }, + { + "epoch": 7.45, + "learning_rate": 2.469168182716075e-07, + "loss": 0.0003, + "step": 56873 + }, + { + "epoch": 7.45, + "learning_rate": 2.467997109423059e-07, + "loss": 0.0007, + "step": 56874 + }, + { + "epoch": 7.45, + "learning_rate": 2.46682631043359e-07, + "loss": 0.0002, + "step": 56875 + }, + { + "epoch": 7.45, + "learning_rate": 2.4656557857509865e-07, + "loss": 0.0005, + "step": 56876 + }, + { + "epoch": 7.45, + "learning_rate": 2.4644855353785357e-07, + "loss": 0.0004, + "step": 56877 + }, + { + "epoch": 7.45, + "learning_rate": 2.4633155593195233e-07, + "loss": 0.001, + "step": 56878 + }, + { + "epoch": 7.45, + "learning_rate": 2.4621458575772475e-07, + "loss": 0.0004, + "step": 56879 + }, + { + "epoch": 7.45, + "learning_rate": 2.460976430154982e-07, + "loss": 0.0008, + "step": 56880 + }, + { + "epoch": 7.45, + "learning_rate": 2.4598072770560256e-07, + "loss": 0.0003, + "step": 56881 + }, + { + "epoch": 7.45, + "learning_rate": 2.4586383982836747e-07, + "loss": 0.0007, + "step": 56882 + }, + { + "epoch": 7.45, + "learning_rate": 2.4574697938412165e-07, + "loss": 0.0005, + "step": 56883 + }, + { + "epoch": 7.45, + "learning_rate": 2.456301463731914e-07, + "loss": 0.0005, + "step": 56884 + }, + { + "epoch": 7.45, + "learning_rate": 2.455133407959076e-07, + "loss": 0.0014, + "step": 56885 + }, + { + "epoch": 7.45, + "learning_rate": 2.4539656265259673e-07, + "loss": 0.0002, + "step": 56886 + }, + { + "epoch": 7.45, + "learning_rate": 2.452798119435906e-07, + "loss": 0.0004, + "step": 56887 + }, + { + "epoch": 7.45, + "learning_rate": 2.4516308866921356e-07, + "loss": 0.0011, + "step": 56888 + }, + { + "epoch": 7.45, + "learning_rate": 2.450463928297975e-07, + "loss": 0.0005, + "step": 56889 + }, + { + "epoch": 7.45, + "learning_rate": 2.4492972442566875e-07, + "loss": 0.0002, + "step": 56890 + }, + { + "epoch": 7.45, + "learning_rate": 2.4481308345715606e-07, + "loss": 0.0006, + "step": 56891 + }, + { + "epoch": 7.45, + "learning_rate": 2.446964699245857e-07, + "loss": 0.0004, + "step": 56892 + }, + { + "epoch": 7.45, + "learning_rate": 2.4457988382828756e-07, + "loss": 0.0011, + "step": 56893 + }, + { + "epoch": 7.45, + "learning_rate": 2.4446332516859015e-07, + "loss": 0.0004, + "step": 56894 + }, + { + "epoch": 7.45, + "learning_rate": 2.4434679394581774e-07, + "loss": 0.0007, + "step": 56895 + }, + { + "epoch": 7.45, + "learning_rate": 2.442302901603022e-07, + "loss": 0.0011, + "step": 56896 + }, + { + "epoch": 7.45, + "learning_rate": 2.4411381381236886e-07, + "loss": 0.0006, + "step": 56897 + }, + { + "epoch": 7.45, + "learning_rate": 2.439973649023464e-07, + "loss": 0.0008, + "step": 56898 + }, + { + "epoch": 7.45, + "learning_rate": 2.438809434305611e-07, + "loss": 0.0004, + "step": 56899 + }, + { + "epoch": 7.45, + "learning_rate": 2.437645493973417e-07, + "loss": 0.0003, + "step": 56900 + }, + { + "epoch": 7.45, + "learning_rate": 2.436481828030146e-07, + "loss": 0.0008, + "step": 56901 + }, + { + "epoch": 7.45, + "learning_rate": 2.435318436479084e-07, + "loss": 0.0012, + "step": 56902 + }, + { + "epoch": 7.45, + "learning_rate": 2.4341553193234725e-07, + "loss": 0.0002, + "step": 56903 + }, + { + "epoch": 7.45, + "learning_rate": 2.43299247656662e-07, + "loss": 0.0004, + "step": 56904 + }, + { + "epoch": 7.45, + "learning_rate": 2.4318299082117804e-07, + "loss": 0.0002, + "step": 56905 + }, + { + "epoch": 7.45, + "learning_rate": 2.4306676142622277e-07, + "loss": 0.0008, + "step": 56906 + }, + { + "epoch": 7.45, + "learning_rate": 2.4295055947212156e-07, + "loss": 0.0002, + "step": 56907 + }, + { + "epoch": 7.45, + "learning_rate": 2.4283438495920185e-07, + "loss": 0.0006, + "step": 56908 + }, + { + "epoch": 7.45, + "learning_rate": 2.427182378877913e-07, + "loss": 0.0002, + "step": 56909 + }, + { + "epoch": 7.45, + "learning_rate": 2.426021182582172e-07, + "loss": 0.0003, + "step": 56910 + }, + { + "epoch": 7.45, + "learning_rate": 2.4248602607080283e-07, + "loss": 0.0002, + "step": 56911 + }, + { + "epoch": 7.46, + "learning_rate": 2.4236996132587897e-07, + "loss": 0.0003, + "step": 56912 + }, + { + "epoch": 7.46, + "learning_rate": 2.422539240237698e-07, + "loss": 0.0005, + "step": 56913 + }, + { + "epoch": 7.46, + "learning_rate": 2.421379141648017e-07, + "loss": 0.0004, + "step": 56914 + }, + { + "epoch": 7.46, + "learning_rate": 2.420219317493e-07, + "loss": 0.0005, + "step": 56915 + }, + { + "epoch": 7.46, + "learning_rate": 2.419059767775922e-07, + "loss": 0.0002, + "step": 56916 + }, + { + "epoch": 7.46, + "learning_rate": 2.417900492500058e-07, + "loss": 0.0002, + "step": 56917 + }, + { + "epoch": 7.46, + "learning_rate": 2.416741491668628e-07, + "loss": 0.0004, + "step": 56918 + }, + { + "epoch": 7.46, + "learning_rate": 2.4155827652849297e-07, + "loss": 0.0004, + "step": 56919 + }, + { + "epoch": 7.46, + "learning_rate": 2.414424313352204e-07, + "loss": 0.0001, + "step": 56920 + }, + { + "epoch": 7.46, + "learning_rate": 2.413266135873726e-07, + "loss": 0.0004, + "step": 56921 + }, + { + "epoch": 7.46, + "learning_rate": 2.412108232852728e-07, + "loss": 0.0001, + "step": 56922 + }, + { + "epoch": 7.46, + "learning_rate": 2.410950604292472e-07, + "loss": 0.0012, + "step": 56923 + }, + { + "epoch": 7.46, + "learning_rate": 2.409793250196235e-07, + "loss": 0.0006, + "step": 56924 + }, + { + "epoch": 7.46, + "learning_rate": 2.408636170567258e-07, + "loss": 0.0004, + "step": 56925 + }, + { + "epoch": 7.46, + "learning_rate": 2.4074793654087825e-07, + "loss": 0.0004, + "step": 56926 + }, + { + "epoch": 7.46, + "learning_rate": 2.406322834724084e-07, + "loss": 0.0005, + "step": 56927 + }, + { + "epoch": 7.46, + "learning_rate": 2.4051665785164047e-07, + "loss": 0.0008, + "step": 56928 + }, + { + "epoch": 7.46, + "learning_rate": 2.404010596788986e-07, + "loss": 0.0003, + "step": 56929 + }, + { + "epoch": 7.46, + "learning_rate": 2.402854889545092e-07, + "loss": 0.0024, + "step": 56930 + }, + { + "epoch": 7.46, + "learning_rate": 2.401699456787987e-07, + "loss": 0.0003, + "step": 56931 + }, + { + "epoch": 7.46, + "learning_rate": 2.4005442985209017e-07, + "loss": 0.001, + "step": 56932 + }, + { + "epoch": 7.46, + "learning_rate": 2.3993894147470774e-07, + "loss": 0.0005, + "step": 56933 + }, + { + "epoch": 7.46, + "learning_rate": 2.398234805469779e-07, + "loss": 0.0004, + "step": 56934 + }, + { + "epoch": 7.46, + "learning_rate": 2.397080470692248e-07, + "loss": 0.0005, + "step": 56935 + }, + { + "epoch": 7.46, + "learning_rate": 2.395926410417726e-07, + "loss": 0.0007, + "step": 56936 + }, + { + "epoch": 7.46, + "learning_rate": 2.3947726246494776e-07, + "loss": 0.0003, + "step": 56937 + }, + { + "epoch": 7.46, + "learning_rate": 2.3936191133907104e-07, + "loss": 0.0004, + "step": 56938 + }, + { + "epoch": 7.46, + "learning_rate": 2.392465876644712e-07, + "loss": 0.0004, + "step": 56939 + }, + { + "epoch": 7.46, + "learning_rate": 2.3913129144146894e-07, + "loss": 0.0007, + "step": 56940 + }, + { + "epoch": 7.46, + "learning_rate": 2.390160226703908e-07, + "loss": 0.0008, + "step": 56941 + }, + { + "epoch": 7.46, + "learning_rate": 2.389007813515598e-07, + "loss": 0.0006, + "step": 56942 + }, + { + "epoch": 7.46, + "learning_rate": 2.387855674853012e-07, + "loss": 0.0008, + "step": 56943 + }, + { + "epoch": 7.46, + "learning_rate": 2.386703810719382e-07, + "loss": 0.0002, + "step": 56944 + }, + { + "epoch": 7.46, + "learning_rate": 2.385552221117948e-07, + "loss": 0.0004, + "step": 56945 + }, + { + "epoch": 7.46, + "learning_rate": 2.3844009060519537e-07, + "loss": 0.0003, + "step": 56946 + }, + { + "epoch": 7.46, + "learning_rate": 2.3832498655246394e-07, + "loss": 0.0004, + "step": 56947 + }, + { + "epoch": 7.46, + "learning_rate": 2.3820990995392258e-07, + "loss": 0.0002, + "step": 56948 + }, + { + "epoch": 7.46, + "learning_rate": 2.3809486080989652e-07, + "loss": 0.0003, + "step": 56949 + }, + { + "epoch": 7.46, + "learning_rate": 2.3797983912070888e-07, + "loss": 0.0007, + "step": 56950 + }, + { + "epoch": 7.46, + "learning_rate": 2.378648448866816e-07, + "loss": 0.0005, + "step": 56951 + }, + { + "epoch": 7.46, + "learning_rate": 2.377498781081411e-07, + "loss": 0.0002, + "step": 56952 + }, + { + "epoch": 7.46, + "learning_rate": 2.3763493878540821e-07, + "loss": 0.0006, + "step": 56953 + }, + { + "epoch": 7.46, + "learning_rate": 2.3752002691880826e-07, + "loss": 0.0005, + "step": 56954 + }, + { + "epoch": 7.46, + "learning_rate": 2.374051425086632e-07, + "loss": 0.0005, + "step": 56955 + }, + { + "epoch": 7.46, + "learning_rate": 2.372902855552961e-07, + "loss": 0.0003, + "step": 56956 + }, + { + "epoch": 7.46, + "learning_rate": 2.3717545605903003e-07, + "loss": 0.0003, + "step": 56957 + }, + { + "epoch": 7.46, + "learning_rate": 2.3706065402018807e-07, + "loss": 0.0003, + "step": 56958 + }, + { + "epoch": 7.46, + "learning_rate": 2.369458794390933e-07, + "loss": 0.0001, + "step": 56959 + }, + { + "epoch": 7.46, + "learning_rate": 2.3683113231606768e-07, + "loss": 0.0003, + "step": 56960 + }, + { + "epoch": 7.46, + "learning_rate": 2.367164126514354e-07, + "loss": 0.0007, + "step": 56961 + }, + { + "epoch": 7.46, + "learning_rate": 2.3660172044551731e-07, + "loss": 0.0006, + "step": 56962 + }, + { + "epoch": 7.46, + "learning_rate": 2.364870556986387e-07, + "loss": 0.0003, + "step": 56963 + }, + { + "epoch": 7.46, + "learning_rate": 2.3637241841111935e-07, + "loss": 0.001, + "step": 56964 + }, + { + "epoch": 7.46, + "learning_rate": 2.3625780858328228e-07, + "loss": 0.0004, + "step": 56965 + }, + { + "epoch": 7.46, + "learning_rate": 2.361432262154506e-07, + "loss": 0.0005, + "step": 56966 + }, + { + "epoch": 7.46, + "learning_rate": 2.3602867130794627e-07, + "loss": 0.0008, + "step": 56967 + }, + { + "epoch": 7.46, + "learning_rate": 2.3591414386109014e-07, + "loss": 0.0004, + "step": 56968 + }, + { + "epoch": 7.46, + "learning_rate": 2.3579964387520638e-07, + "loss": 0.0004, + "step": 56969 + }, + { + "epoch": 7.46, + "learning_rate": 2.3568517135061697e-07, + "loss": 0.0002, + "step": 56970 + }, + { + "epoch": 7.46, + "learning_rate": 2.3557072628764278e-07, + "loss": 0.0001, + "step": 56971 + }, + { + "epoch": 7.46, + "learning_rate": 2.3545630868660464e-07, + "loss": 0.0008, + "step": 56972 + }, + { + "epoch": 7.46, + "learning_rate": 2.353419185478256e-07, + "loss": 0.0003, + "step": 56973 + }, + { + "epoch": 7.46, + "learning_rate": 2.3522755587162882e-07, + "loss": 0.0005, + "step": 56974 + }, + { + "epoch": 7.46, + "learning_rate": 2.3511322065833396e-07, + "loss": 0.0002, + "step": 56975 + }, + { + "epoch": 7.46, + "learning_rate": 2.3499891290826305e-07, + "loss": 0.0003, + "step": 56976 + }, + { + "epoch": 7.46, + "learning_rate": 2.34884632621738e-07, + "loss": 0.0004, + "step": 56977 + }, + { + "epoch": 7.46, + "learning_rate": 2.347703797990797e-07, + "loss": 0.0006, + "step": 56978 + }, + { + "epoch": 7.46, + "learning_rate": 2.3465615444061008e-07, + "loss": 0.0002, + "step": 56979 + }, + { + "epoch": 7.46, + "learning_rate": 2.345419565466489e-07, + "loss": 0.0002, + "step": 56980 + }, + { + "epoch": 7.46, + "learning_rate": 2.3442778611751926e-07, + "loss": 0.0002, + "step": 56981 + }, + { + "epoch": 7.46, + "learning_rate": 2.34313643153542e-07, + "loss": 0.0009, + "step": 56982 + }, + { + "epoch": 7.46, + "learning_rate": 2.3419952765503572e-07, + "loss": 0.0008, + "step": 56983 + }, + { + "epoch": 7.46, + "learning_rate": 2.3408543962232355e-07, + "loss": 0.0006, + "step": 56984 + }, + { + "epoch": 7.46, + "learning_rate": 2.3397137905572743e-07, + "loss": 0.0007, + "step": 56985 + }, + { + "epoch": 7.46, + "learning_rate": 2.3385734595556596e-07, + "loss": 0.0006, + "step": 56986 + }, + { + "epoch": 7.46, + "learning_rate": 2.3374334032216117e-07, + "loss": 0.0003, + "step": 56987 + }, + { + "epoch": 7.47, + "learning_rate": 2.3362936215583164e-07, + "loss": 0.0001, + "step": 56988 + }, + { + "epoch": 7.47, + "learning_rate": 2.335154114569016e-07, + "loss": 0.0002, + "step": 56989 + }, + { + "epoch": 7.47, + "learning_rate": 2.3340148822568852e-07, + "loss": 0.0004, + "step": 56990 + }, + { + "epoch": 7.47, + "learning_rate": 2.332875924625122e-07, + "loss": 0.0003, + "step": 56991 + }, + { + "epoch": 7.47, + "learning_rate": 2.331737241676968e-07, + "loss": 0.0005, + "step": 56992 + }, + { + "epoch": 7.47, + "learning_rate": 2.330598833415587e-07, + "loss": 0.001, + "step": 56993 + }, + { + "epoch": 7.47, + "learning_rate": 2.3294606998441993e-07, + "loss": 0.001, + "step": 56994 + }, + { + "epoch": 7.47, + "learning_rate": 2.3283228409659908e-07, + "loss": 0.0002, + "step": 56995 + }, + { + "epoch": 7.47, + "learning_rate": 2.3271852567841924e-07, + "loss": 0.0013, + "step": 56996 + }, + { + "epoch": 7.47, + "learning_rate": 2.3260479473019683e-07, + "loss": 0.0001, + "step": 56997 + }, + { + "epoch": 7.47, + "learning_rate": 2.3249109125225488e-07, + "loss": 0.0004, + "step": 56998 + }, + { + "epoch": 7.47, + "learning_rate": 2.3237741524490988e-07, + "loss": 0.0003, + "step": 56999 + }, + { + "epoch": 7.47, + "learning_rate": 2.3226376670848482e-07, + "loss": 0.0006, + "step": 57000 + }, + { + "epoch": 7.47, + "learning_rate": 2.321501456432973e-07, + "loss": 0.0005, + "step": 57001 + }, + { + "epoch": 7.47, + "learning_rate": 2.320365520496659e-07, + "loss": 0.0005, + "step": 57002 + }, + { + "epoch": 7.47, + "learning_rate": 2.3192298592791374e-07, + "loss": 0.0002, + "step": 57003 + }, + { + "epoch": 7.47, + "learning_rate": 2.318094472783572e-07, + "loss": 0.0002, + "step": 57004 + }, + { + "epoch": 7.47, + "learning_rate": 2.31695936101316e-07, + "loss": 0.0002, + "step": 57005 + }, + { + "epoch": 7.47, + "learning_rate": 2.3158245239710886e-07, + "loss": 0.0004, + "step": 57006 + }, + { + "epoch": 7.47, + "learning_rate": 2.3146899616605655e-07, + "loss": 0.0003, + "step": 57007 + }, + { + "epoch": 7.47, + "learning_rate": 2.3135556740847887e-07, + "loss": 0.0006, + "step": 57008 + }, + { + "epoch": 7.47, + "learning_rate": 2.312421661246922e-07, + "loss": 0.0005, + "step": 57009 + }, + { + "epoch": 7.47, + "learning_rate": 2.3112879231501628e-07, + "loss": 0.0007, + "step": 57010 + }, + { + "epoch": 7.47, + "learning_rate": 2.31015445979772e-07, + "loss": 0.0003, + "step": 57011 + }, + { + "epoch": 7.47, + "learning_rate": 2.3090212711927573e-07, + "loss": 0.0005, + "step": 57012 + }, + { + "epoch": 7.47, + "learning_rate": 2.3078883573384724e-07, + "loss": 0.0006, + "step": 57013 + }, + { + "epoch": 7.47, + "learning_rate": 2.3067557182380407e-07, + "loss": 0.0002, + "step": 57014 + }, + { + "epoch": 7.47, + "learning_rate": 2.3056233538946593e-07, + "loss": 0.0006, + "step": 57015 + }, + { + "epoch": 7.47, + "learning_rate": 2.3044912643115148e-07, + "loss": 0.0003, + "step": 57016 + }, + { + "epoch": 7.47, + "learning_rate": 2.3033594494917823e-07, + "loss": 0.0005, + "step": 57017 + }, + { + "epoch": 7.47, + "learning_rate": 2.3022279094386368e-07, + "loss": 0.0005, + "step": 57018 + }, + { + "epoch": 7.47, + "learning_rate": 2.3010966441552873e-07, + "loss": 0.0003, + "step": 57019 + }, + { + "epoch": 7.47, + "learning_rate": 2.2999656536448977e-07, + "loss": 0.0002, + "step": 57020 + }, + { + "epoch": 7.47, + "learning_rate": 2.2988349379106655e-07, + "loss": 0.0012, + "step": 57021 + }, + { + "epoch": 7.47, + "learning_rate": 2.2977044969557328e-07, + "loss": 0.0006, + "step": 57022 + }, + { + "epoch": 7.47, + "learning_rate": 2.2965743307833187e-07, + "loss": 0.0005, + "step": 57023 + }, + { + "epoch": 7.47, + "learning_rate": 2.295444439396588e-07, + "loss": 0.0007, + "step": 57024 + }, + { + "epoch": 7.47, + "learning_rate": 2.2943148227987045e-07, + "loss": 0.0006, + "step": 57025 + }, + { + "epoch": 7.47, + "learning_rate": 2.2931854809928655e-07, + "loss": 0.0008, + "step": 57026 + }, + { + "epoch": 7.47, + "learning_rate": 2.2920564139822356e-07, + "loss": 0.0004, + "step": 57027 + }, + { + "epoch": 7.47, + "learning_rate": 2.2909276217699895e-07, + "loss": 0.0006, + "step": 57028 + }, + { + "epoch": 7.47, + "learning_rate": 2.2897991043593138e-07, + "loss": 0.0004, + "step": 57029 + }, + { + "epoch": 7.47, + "learning_rate": 2.2886708617533725e-07, + "loss": 0.0004, + "step": 57030 + }, + { + "epoch": 7.47, + "learning_rate": 2.2875428939553413e-07, + "loss": 0.0005, + "step": 57031 + }, + { + "epoch": 7.47, + "learning_rate": 2.286415200968395e-07, + "loss": 0.0003, + "step": 57032 + }, + { + "epoch": 7.47, + "learning_rate": 2.2852877827956977e-07, + "loss": 0.0003, + "step": 57033 + }, + { + "epoch": 7.47, + "learning_rate": 2.2841606394404248e-07, + "loss": 0.0008, + "step": 57034 + }, + { + "epoch": 7.47, + "learning_rate": 2.2830337709057516e-07, + "loss": 0.0005, + "step": 57035 + }, + { + "epoch": 7.47, + "learning_rate": 2.281907177194842e-07, + "loss": 0.0003, + "step": 57036 + }, + { + "epoch": 7.47, + "learning_rate": 2.2807808583108605e-07, + "loss": 0.0003, + "step": 57037 + }, + { + "epoch": 7.47, + "learning_rate": 2.2796548142569707e-07, + "loss": 0.0003, + "step": 57038 + }, + { + "epoch": 7.47, + "learning_rate": 2.2785290450363596e-07, + "loss": 0.0003, + "step": 57039 + }, + { + "epoch": 7.47, + "learning_rate": 2.2774035506521907e-07, + "loss": 0.0002, + "step": 57040 + }, + { + "epoch": 7.47, + "learning_rate": 2.2762783311075955e-07, + "loss": 0.0004, + "step": 57041 + }, + { + "epoch": 7.47, + "learning_rate": 2.275153386405782e-07, + "loss": 0.0002, + "step": 57042 + }, + { + "epoch": 7.47, + "learning_rate": 2.274028716549892e-07, + "loss": 0.0009, + "step": 57043 + }, + { + "epoch": 7.47, + "learning_rate": 2.2729043215431013e-07, + "loss": 0.0001, + "step": 57044 + }, + { + "epoch": 7.47, + "learning_rate": 2.2717802013885405e-07, + "loss": 0.0004, + "step": 57045 + }, + { + "epoch": 7.47, + "learning_rate": 2.2706563560894067e-07, + "loss": 0.0002, + "step": 57046 + }, + { + "epoch": 7.47, + "learning_rate": 2.2695327856488536e-07, + "loss": 0.0003, + "step": 57047 + }, + { + "epoch": 7.47, + "learning_rate": 2.2684094900700228e-07, + "loss": 0.0003, + "step": 57048 + }, + { + "epoch": 7.47, + "learning_rate": 2.2672864693560782e-07, + "loss": 0.0005, + "step": 57049 + }, + { + "epoch": 7.47, + "learning_rate": 2.2661637235102064e-07, + "loss": 0.0002, + "step": 57050 + }, + { + "epoch": 7.47, + "learning_rate": 2.265041252535538e-07, + "loss": 0.0003, + "step": 57051 + }, + { + "epoch": 7.47, + "learning_rate": 2.2639190564352265e-07, + "loss": 0.0005, + "step": 57052 + }, + { + "epoch": 7.47, + "learning_rate": 2.2627971352124468e-07, + "loss": 0.0002, + "step": 57053 + }, + { + "epoch": 7.47, + "learning_rate": 2.2616754888703518e-07, + "loss": 0.0005, + "step": 57054 + }, + { + "epoch": 7.47, + "learning_rate": 2.2605541174120839e-07, + "loss": 0.001, + "step": 57055 + }, + { + "epoch": 7.47, + "learning_rate": 2.2594330208407954e-07, + "loss": 0.0005, + "step": 57056 + }, + { + "epoch": 7.47, + "learning_rate": 2.2583121991596624e-07, + "loss": 0.0003, + "step": 57057 + }, + { + "epoch": 7.47, + "learning_rate": 2.257191652371804e-07, + "loss": 0.0005, + "step": 57058 + }, + { + "epoch": 7.47, + "learning_rate": 2.2560713804804068e-07, + "loss": 0.0003, + "step": 57059 + }, + { + "epoch": 7.47, + "learning_rate": 2.2549513834885795e-07, + "loss": 0.0002, + "step": 57060 + }, + { + "epoch": 7.47, + "learning_rate": 2.2538316613995083e-07, + "loss": 0.0001, + "step": 57061 + }, + { + "epoch": 7.47, + "learning_rate": 2.2527122142163348e-07, + "loss": 0.0003, + "step": 57062 + }, + { + "epoch": 7.47, + "learning_rate": 2.2515930419422016e-07, + "loss": 0.0002, + "step": 57063 + }, + { + "epoch": 7.47, + "learning_rate": 2.2504741445802503e-07, + "loss": 0.0008, + "step": 57064 + }, + { + "epoch": 7.48, + "learning_rate": 2.249355522133645e-07, + "loss": 0.0005, + "step": 57065 + }, + { + "epoch": 7.48, + "learning_rate": 2.2482371746055165e-07, + "loss": 0.0009, + "step": 57066 + }, + { + "epoch": 7.48, + "learning_rate": 2.2471191019990067e-07, + "loss": 0.0006, + "step": 57067 + }, + { + "epoch": 7.48, + "learning_rate": 2.2460013043172913e-07, + "loss": 0.0006, + "step": 57068 + }, + { + "epoch": 7.48, + "learning_rate": 2.2448837815634784e-07, + "loss": 0.0003, + "step": 57069 + }, + { + "epoch": 7.48, + "learning_rate": 2.243766533740721e-07, + "loss": 0.0001, + "step": 57070 + }, + { + "epoch": 7.48, + "learning_rate": 2.2426495608521615e-07, + "loss": 0.0002, + "step": 57071 + }, + { + "epoch": 7.48, + "learning_rate": 2.2415328629009525e-07, + "loss": 0.0009, + "step": 57072 + }, + { + "epoch": 7.48, + "learning_rate": 2.240416439890225e-07, + "loss": 0.0008, + "step": 57073 + }, + { + "epoch": 7.48, + "learning_rate": 2.239300291823121e-07, + "loss": 0.0003, + "step": 57074 + }, + { + "epoch": 7.48, + "learning_rate": 2.2381844187027713e-07, + "loss": 0.0003, + "step": 57075 + }, + { + "epoch": 7.48, + "learning_rate": 2.2370688205323287e-07, + "loss": 0.0005, + "step": 57076 + }, + { + "epoch": 7.48, + "learning_rate": 2.2359534973149245e-07, + "loss": 0.0002, + "step": 57077 + }, + { + "epoch": 7.48, + "learning_rate": 2.2348384490537e-07, + "loss": 0.0006, + "step": 57078 + }, + { + "epoch": 7.48, + "learning_rate": 2.2337236757517754e-07, + "loss": 0.0008, + "step": 57079 + }, + { + "epoch": 7.48, + "learning_rate": 2.2326091774123037e-07, + "loss": 0.0002, + "step": 57080 + }, + { + "epoch": 7.48, + "learning_rate": 2.2314949540384046e-07, + "loss": 0.0006, + "step": 57081 + }, + { + "epoch": 7.48, + "learning_rate": 2.2303810056332197e-07, + "loss": 0.0001, + "step": 57082 + }, + { + "epoch": 7.48, + "learning_rate": 2.2292673321998804e-07, + "loss": 0.0006, + "step": 57083 + }, + { + "epoch": 7.48, + "learning_rate": 2.2281539337415282e-07, + "loss": 0.0004, + "step": 57084 + }, + { + "epoch": 7.48, + "learning_rate": 2.2270408102612828e-07, + "loss": 0.0009, + "step": 57085 + }, + { + "epoch": 7.48, + "learning_rate": 2.2259279617622863e-07, + "loss": 0.0002, + "step": 57086 + }, + { + "epoch": 7.48, + "learning_rate": 2.2248153882476475e-07, + "loss": 0.0007, + "step": 57087 + }, + { + "epoch": 7.48, + "learning_rate": 2.2237030897205192e-07, + "loss": 0.0002, + "step": 57088 + }, + { + "epoch": 7.48, + "learning_rate": 2.222591066184021e-07, + "loss": 0.0003, + "step": 57089 + }, + { + "epoch": 7.48, + "learning_rate": 2.2214793176412618e-07, + "loss": 0.0008, + "step": 57090 + }, + { + "epoch": 7.48, + "learning_rate": 2.2203678440953945e-07, + "loss": 0.0003, + "step": 57091 + }, + { + "epoch": 7.48, + "learning_rate": 2.219256645549539e-07, + "loss": 0.0009, + "step": 57092 + }, + { + "epoch": 7.48, + "learning_rate": 2.2181457220068147e-07, + "loss": 0.0005, + "step": 57093 + }, + { + "epoch": 7.48, + "learning_rate": 2.2170350734703527e-07, + "loss": 0.0001, + "step": 57094 + }, + { + "epoch": 7.48, + "learning_rate": 2.2159246999432615e-07, + "loss": 0.0002, + "step": 57095 + }, + { + "epoch": 7.48, + "learning_rate": 2.214814601428683e-07, + "loss": 0.0007, + "step": 57096 + }, + { + "epoch": 7.48, + "learning_rate": 2.2137047779297262e-07, + "loss": 0.0004, + "step": 57097 + }, + { + "epoch": 7.48, + "learning_rate": 2.2125952294495212e-07, + "loss": 0.0006, + "step": 57098 + }, + { + "epoch": 7.48, + "learning_rate": 2.2114859559911882e-07, + "loss": 0.0006, + "step": 57099 + }, + { + "epoch": 7.48, + "learning_rate": 2.210376957557836e-07, + "loss": 0.0004, + "step": 57100 + }, + { + "epoch": 7.48, + "learning_rate": 2.2092682341525951e-07, + "loss": 0.0002, + "step": 57101 + }, + { + "epoch": 7.48, + "learning_rate": 2.2081597857785743e-07, + "loss": 0.0006, + "step": 57102 + }, + { + "epoch": 7.48, + "learning_rate": 2.207051612438904e-07, + "loss": 0.0006, + "step": 57103 + }, + { + "epoch": 7.48, + "learning_rate": 2.2059437141366825e-07, + "loss": 0.0007, + "step": 57104 + }, + { + "epoch": 7.48, + "learning_rate": 2.204836090875051e-07, + "loss": 0.0002, + "step": 57105 + }, + { + "epoch": 7.48, + "learning_rate": 2.2037287426570963e-07, + "loss": 0.0006, + "step": 57106 + }, + { + "epoch": 7.48, + "learning_rate": 2.2026216694859493e-07, + "loss": 0.0007, + "step": 57107 + }, + { + "epoch": 7.48, + "learning_rate": 2.2015148713647293e-07, + "loss": 0.0003, + "step": 57108 + }, + { + "epoch": 7.48, + "learning_rate": 2.2004083482965454e-07, + "loss": 0.0003, + "step": 57109 + }, + { + "epoch": 7.48, + "learning_rate": 2.1993021002844837e-07, + "loss": 0.0004, + "step": 57110 + }, + { + "epoch": 7.48, + "learning_rate": 2.1981961273316976e-07, + "loss": 0.0007, + "step": 57111 + }, + { + "epoch": 7.48, + "learning_rate": 2.197090429441262e-07, + "loss": 0.0026, + "step": 57112 + }, + { + "epoch": 7.48, + "learning_rate": 2.195985006616308e-07, + "loss": 0.0006, + "step": 57113 + }, + { + "epoch": 7.48, + "learning_rate": 2.194879858859933e-07, + "loss": 0.0007, + "step": 57114 + }, + { + "epoch": 7.48, + "learning_rate": 2.1937749861752455e-07, + "loss": 0.0001, + "step": 57115 + }, + { + "epoch": 7.48, + "learning_rate": 2.1926703885653767e-07, + "loss": 0.0007, + "step": 57116 + }, + { + "epoch": 7.48, + "learning_rate": 2.1915660660333904e-07, + "loss": 0.0003, + "step": 57117 + }, + { + "epoch": 7.48, + "learning_rate": 2.1904620185824288e-07, + "loss": 0.0005, + "step": 57118 + }, + { + "epoch": 7.48, + "learning_rate": 2.1893582462155894e-07, + "loss": 0.0002, + "step": 57119 + }, + { + "epoch": 7.48, + "learning_rate": 2.1882547489359694e-07, + "loss": 0.0003, + "step": 57120 + }, + { + "epoch": 7.48, + "learning_rate": 2.187151526746667e-07, + "loss": 0.0002, + "step": 57121 + }, + { + "epoch": 7.48, + "learning_rate": 2.18604857965079e-07, + "loss": 0.0004, + "step": 57122 + }, + { + "epoch": 7.48, + "learning_rate": 2.184945907651459e-07, + "loss": 0.0005, + "step": 57123 + }, + { + "epoch": 7.48, + "learning_rate": 2.1838435107517487e-07, + "loss": 0.0006, + "step": 57124 + }, + { + "epoch": 7.48, + "learning_rate": 2.1827413889547567e-07, + "loss": 0.0007, + "step": 57125 + }, + { + "epoch": 7.48, + "learning_rate": 2.1816395422636028e-07, + "loss": 0.0008, + "step": 57126 + }, + { + "epoch": 7.48, + "learning_rate": 2.1805379706813846e-07, + "loss": 0.0006, + "step": 57127 + }, + { + "epoch": 7.48, + "learning_rate": 2.1794366742111995e-07, + "loss": 0.0003, + "step": 57128 + }, + { + "epoch": 7.48, + "learning_rate": 2.1783356528561228e-07, + "loss": 0.0005, + "step": 57129 + }, + { + "epoch": 7.48, + "learning_rate": 2.1772349066192745e-07, + "loss": 0.0004, + "step": 57130 + }, + { + "epoch": 7.48, + "learning_rate": 2.1761344355037517e-07, + "loss": 0.0004, + "step": 57131 + }, + { + "epoch": 7.48, + "learning_rate": 2.17503423951263e-07, + "loss": 0.0003, + "step": 57132 + }, + { + "epoch": 7.48, + "learning_rate": 2.173934318649029e-07, + "loss": 0.0017, + "step": 57133 + }, + { + "epoch": 7.48, + "learning_rate": 2.1728346729160134e-07, + "loss": 0.0008, + "step": 57134 + }, + { + "epoch": 7.48, + "learning_rate": 2.171735302316702e-07, + "loss": 0.0006, + "step": 57135 + }, + { + "epoch": 7.48, + "learning_rate": 2.170636206854171e-07, + "loss": 0.0004, + "step": 57136 + }, + { + "epoch": 7.48, + "learning_rate": 2.1695373865315063e-07, + "loss": 0.0004, + "step": 57137 + }, + { + "epoch": 7.48, + "learning_rate": 2.1684388413518165e-07, + "loss": 0.0004, + "step": 57138 + }, + { + "epoch": 7.48, + "learning_rate": 2.1673405713181882e-07, + "loss": 0.0009, + "step": 57139 + }, + { + "epoch": 7.48, + "learning_rate": 2.166242576433697e-07, + "loss": 0.0003, + "step": 57140 + }, + { + "epoch": 7.49, + "learning_rate": 2.16514485670144e-07, + "loss": 0.0009, + "step": 57141 + }, + { + "epoch": 7.49, + "learning_rate": 2.1640474121245148e-07, + "loss": 0.0006, + "step": 57142 + }, + { + "epoch": 7.49, + "learning_rate": 2.1629502427059857e-07, + "loss": 0.0005, + "step": 57143 + }, + { + "epoch": 7.49, + "learning_rate": 2.161853348448939e-07, + "loss": 0.0005, + "step": 57144 + }, + { + "epoch": 7.49, + "learning_rate": 2.1607567293564835e-07, + "loss": 0.0001, + "step": 57145 + }, + { + "epoch": 7.49, + "learning_rate": 2.1596603854316945e-07, + "loss": 0.0002, + "step": 57146 + }, + { + "epoch": 7.49, + "learning_rate": 2.158564316677636e-07, + "loss": 0.0002, + "step": 57147 + }, + { + "epoch": 7.49, + "learning_rate": 2.1574685230973946e-07, + "loss": 0.0002, + "step": 57148 + }, + { + "epoch": 7.49, + "learning_rate": 2.15637300469409e-07, + "loss": 0.0005, + "step": 57149 + }, + { + "epoch": 7.49, + "learning_rate": 2.1552777614707642e-07, + "loss": 0.0003, + "step": 57150 + }, + { + "epoch": 7.49, + "learning_rate": 2.1541827934305148e-07, + "loss": 0.0002, + "step": 57151 + }, + { + "epoch": 7.49, + "learning_rate": 2.153088100576395e-07, + "loss": 0.001, + "step": 57152 + }, + { + "epoch": 7.49, + "learning_rate": 2.151993682911524e-07, + "loss": 0.0002, + "step": 57153 + }, + { + "epoch": 7.49, + "learning_rate": 2.1508995404389666e-07, + "loss": 0.0006, + "step": 57154 + }, + { + "epoch": 7.49, + "learning_rate": 2.1498056731617755e-07, + "loss": 0.0002, + "step": 57155 + }, + { + "epoch": 7.49, + "learning_rate": 2.1487120810830487e-07, + "loss": 0.0003, + "step": 57156 + }, + { + "epoch": 7.49, + "learning_rate": 2.1476187642058722e-07, + "loss": 0.0004, + "step": 57157 + }, + { + "epoch": 7.49, + "learning_rate": 2.1465257225332992e-07, + "loss": 0.0002, + "step": 57158 + }, + { + "epoch": 7.49, + "learning_rate": 2.145432956068394e-07, + "loss": 0.0003, + "step": 57159 + }, + { + "epoch": 7.49, + "learning_rate": 2.144340464814254e-07, + "loss": 0.0006, + "step": 57160 + }, + { + "epoch": 7.49, + "learning_rate": 2.143248248773966e-07, + "loss": 0.0002, + "step": 57161 + }, + { + "epoch": 7.49, + "learning_rate": 2.1421563079505715e-07, + "loss": 0.0003, + "step": 57162 + }, + { + "epoch": 7.49, + "learning_rate": 2.1410646423471348e-07, + "loss": 0.0003, + "step": 57163 + }, + { + "epoch": 7.49, + "learning_rate": 2.139973251966765e-07, + "loss": 0.0007, + "step": 57164 + }, + { + "epoch": 7.49, + "learning_rate": 2.1388821368125036e-07, + "loss": 0.001, + "step": 57165 + }, + { + "epoch": 7.49, + "learning_rate": 2.137791296887426e-07, + "loss": 0.0003, + "step": 57166 + }, + { + "epoch": 7.49, + "learning_rate": 2.1367007321945854e-07, + "loss": 0.0006, + "step": 57167 + }, + { + "epoch": 7.49, + "learning_rate": 2.1356104427370683e-07, + "loss": 0.0005, + "step": 57168 + }, + { + "epoch": 7.49, + "learning_rate": 2.1345204285179387e-07, + "loss": 0.0004, + "step": 57169 + }, + { + "epoch": 7.49, + "learning_rate": 2.1334306895402611e-07, + "loss": 0.0006, + "step": 57170 + }, + { + "epoch": 7.49, + "learning_rate": 2.1323412258070886e-07, + "loss": 0.0002, + "step": 57171 + }, + { + "epoch": 7.49, + "learning_rate": 2.1312520373214962e-07, + "loss": 0.0007, + "step": 57172 + }, + { + "epoch": 7.49, + "learning_rate": 2.1301631240865594e-07, + "loss": 0.0002, + "step": 57173 + }, + { + "epoch": 7.49, + "learning_rate": 2.1290744861053203e-07, + "loss": 0.0005, + "step": 57174 + }, + { + "epoch": 7.49, + "learning_rate": 2.127986123380843e-07, + "loss": 0.0007, + "step": 57175 + }, + { + "epoch": 7.49, + "learning_rate": 2.1268980359161917e-07, + "loss": 0.0003, + "step": 57176 + }, + { + "epoch": 7.49, + "learning_rate": 2.1258102237144308e-07, + "loss": 0.0003, + "step": 57177 + }, + { + "epoch": 7.49, + "learning_rate": 2.124722686778613e-07, + "loss": 0.0009, + "step": 57178 + }, + { + "epoch": 7.49, + "learning_rate": 2.123635425111814e-07, + "loss": 0.0005, + "step": 57179 + }, + { + "epoch": 7.49, + "learning_rate": 2.1225484387170535e-07, + "loss": 0.0004, + "step": 57180 + }, + { + "epoch": 7.49, + "learning_rate": 2.121461727597429e-07, + "loss": 0.0002, + "step": 57181 + }, + { + "epoch": 7.49, + "learning_rate": 2.120375291755983e-07, + "loss": 0.0002, + "step": 57182 + }, + { + "epoch": 7.49, + "learning_rate": 2.1192891311957675e-07, + "loss": 0.0004, + "step": 57183 + }, + { + "epoch": 7.49, + "learning_rate": 2.1182032459198366e-07, + "loss": 0.0006, + "step": 57184 + }, + { + "epoch": 7.49, + "learning_rate": 2.117117635931254e-07, + "loss": 0.0003, + "step": 57185 + }, + { + "epoch": 7.49, + "learning_rate": 2.1160323012330508e-07, + "loss": 0.0002, + "step": 57186 + }, + { + "epoch": 7.49, + "learning_rate": 2.1149472418283024e-07, + "loss": 0.0005, + "step": 57187 + }, + { + "epoch": 7.49, + "learning_rate": 2.1138624577200618e-07, + "loss": 0.0003, + "step": 57188 + }, + { + "epoch": 7.49, + "learning_rate": 2.1127779489113598e-07, + "loss": 0.001, + "step": 57189 + }, + { + "epoch": 7.49, + "learning_rate": 2.1116937154052497e-07, + "loss": 0.0005, + "step": 57190 + }, + { + "epoch": 7.49, + "learning_rate": 2.1106097572047956e-07, + "loss": 0.0018, + "step": 57191 + }, + { + "epoch": 7.49, + "learning_rate": 2.1095260743130396e-07, + "loss": 0.0002, + "step": 57192 + }, + { + "epoch": 7.49, + "learning_rate": 2.1084426667330238e-07, + "loss": 0.0005, + "step": 57193 + }, + { + "epoch": 7.49, + "learning_rate": 2.107359534467801e-07, + "loss": 0.0008, + "step": 57194 + }, + { + "epoch": 7.49, + "learning_rate": 2.1062766775204247e-07, + "loss": 0.0004, + "step": 57195 + }, + { + "epoch": 7.49, + "learning_rate": 2.1051940958939253e-07, + "loss": 0.0003, + "step": 57196 + }, + { + "epoch": 7.49, + "learning_rate": 2.1041117895913564e-07, + "loss": 0.0003, + "step": 57197 + }, + { + "epoch": 7.49, + "learning_rate": 2.10302975861576e-07, + "loss": 0.0004, + "step": 57198 + }, + { + "epoch": 7.49, + "learning_rate": 2.1019480029701777e-07, + "loss": 0.0004, + "step": 57199 + }, + { + "epoch": 7.49, + "learning_rate": 2.100866522657663e-07, + "loss": 0.0001, + "step": 57200 + }, + { + "epoch": 7.49, + "learning_rate": 2.0997853176812354e-07, + "loss": 0.0004, + "step": 57201 + }, + { + "epoch": 7.49, + "learning_rate": 2.0987043880439372e-07, + "loss": 0.0006, + "step": 57202 + }, + { + "epoch": 7.49, + "learning_rate": 2.0976237337488437e-07, + "loss": 0.0013, + "step": 57203 + }, + { + "epoch": 7.49, + "learning_rate": 2.0965433547989634e-07, + "loss": 0.0003, + "step": 57204 + }, + { + "epoch": 7.49, + "learning_rate": 2.0954632511973272e-07, + "loss": 0.0004, + "step": 57205 + }, + { + "epoch": 7.49, + "learning_rate": 2.0943834229469994e-07, + "loss": 0.0002, + "step": 57206 + }, + { + "epoch": 7.49, + "learning_rate": 2.093303870051011e-07, + "loss": 0.0005, + "step": 57207 + }, + { + "epoch": 7.49, + "learning_rate": 2.092224592512382e-07, + "loss": 0.0004, + "step": 57208 + }, + { + "epoch": 7.49, + "learning_rate": 2.091145590334154e-07, + "loss": 0.0008, + "step": 57209 + }, + { + "epoch": 7.49, + "learning_rate": 2.0900668635193688e-07, + "loss": 0.0006, + "step": 57210 + }, + { + "epoch": 7.49, + "learning_rate": 2.0889884120710692e-07, + "loss": 0.0008, + "step": 57211 + }, + { + "epoch": 7.49, + "learning_rate": 2.087910235992252e-07, + "loss": 0.0004, + "step": 57212 + }, + { + "epoch": 7.49, + "learning_rate": 2.086832335285982e-07, + "loss": 0.0005, + "step": 57213 + }, + { + "epoch": 7.49, + "learning_rate": 2.08575470995529e-07, + "loss": 0.0002, + "step": 57214 + }, + { + "epoch": 7.49, + "learning_rate": 2.0846773600031955e-07, + "loss": 0.0002, + "step": 57215 + }, + { + "epoch": 7.49, + "learning_rate": 2.0836002854327297e-07, + "loss": 0.0006, + "step": 57216 + }, + { + "epoch": 7.5, + "learning_rate": 2.0825234862469124e-07, + "loss": 0.0005, + "step": 57217 + }, + { + "epoch": 7.5, + "learning_rate": 2.0814469624487966e-07, + "loss": 0.0006, + "step": 57218 + }, + { + "epoch": 7.5, + "learning_rate": 2.080370714041402e-07, + "loss": 0.0005, + "step": 57219 + }, + { + "epoch": 7.5, + "learning_rate": 2.0792947410277374e-07, + "loss": 0.0003, + "step": 57220 + }, + { + "epoch": 7.5, + "learning_rate": 2.078219043410845e-07, + "loss": 0.0005, + "step": 57221 + }, + { + "epoch": 7.5, + "learning_rate": 2.0771436211937558e-07, + "loss": 0.0002, + "step": 57222 + }, + { + "epoch": 7.5, + "learning_rate": 2.076068474379489e-07, + "loss": 0.0004, + "step": 57223 + }, + { + "epoch": 7.5, + "learning_rate": 2.0749936029710428e-07, + "loss": 0.0003, + "step": 57224 + }, + { + "epoch": 7.5, + "learning_rate": 2.07391900697147e-07, + "loss": 0.0005, + "step": 57225 + }, + { + "epoch": 7.5, + "learning_rate": 2.0728446863837903e-07, + "loss": 0.0002, + "step": 57226 + }, + { + "epoch": 7.5, + "learning_rate": 2.071770641211024e-07, + "loss": 0.0004, + "step": 57227 + }, + { + "epoch": 7.5, + "learning_rate": 2.0706968714561792e-07, + "loss": 0.0005, + "step": 57228 + }, + { + "epoch": 7.5, + "learning_rate": 2.0696233771222874e-07, + "loss": 0.0003, + "step": 57229 + }, + { + "epoch": 7.5, + "learning_rate": 2.0685501582123678e-07, + "loss": 0.0003, + "step": 57230 + }, + { + "epoch": 7.5, + "learning_rate": 2.067477214729441e-07, + "loss": 0.0002, + "step": 57231 + }, + { + "epoch": 7.5, + "learning_rate": 2.0664045466765038e-07, + "loss": 0.0012, + "step": 57232 + }, + { + "epoch": 7.5, + "learning_rate": 2.0653321540565985e-07, + "loss": 0.0003, + "step": 57233 + }, + { + "epoch": 7.5, + "learning_rate": 2.0642600368727338e-07, + "loss": 0.0002, + "step": 57234 + }, + { + "epoch": 7.5, + "learning_rate": 2.0631881951279077e-07, + "loss": 0.0003, + "step": 57235 + }, + { + "epoch": 7.5, + "learning_rate": 2.0621166288251504e-07, + "loss": 0.0006, + "step": 57236 + }, + { + "epoch": 7.5, + "learning_rate": 2.0610453379674932e-07, + "loss": 0.0006, + "step": 57237 + }, + { + "epoch": 7.5, + "learning_rate": 2.0599743225579227e-07, + "loss": 0.0003, + "step": 57238 + }, + { + "epoch": 7.5, + "learning_rate": 2.0589035825994586e-07, + "loss": 0.0006, + "step": 57239 + }, + { + "epoch": 7.5, + "learning_rate": 2.0578331180950984e-07, + "loss": 0.0002, + "step": 57240 + }, + { + "epoch": 7.5, + "learning_rate": 2.0567629290478841e-07, + "loss": 0.0007, + "step": 57241 + }, + { + "epoch": 7.5, + "learning_rate": 2.0556930154608133e-07, + "loss": 0.0008, + "step": 57242 + }, + { + "epoch": 7.5, + "learning_rate": 2.0546233773368728e-07, + "loss": 0.0002, + "step": 57243 + }, + { + "epoch": 7.5, + "learning_rate": 2.0535540146790934e-07, + "loss": 0.0005, + "step": 57244 + }, + { + "epoch": 7.5, + "learning_rate": 2.0524849274904833e-07, + "loss": 0.0006, + "step": 57245 + }, + { + "epoch": 7.5, + "learning_rate": 2.0514161157740297e-07, + "loss": 0.0003, + "step": 57246 + }, + { + "epoch": 7.5, + "learning_rate": 2.0503475795327632e-07, + "loss": 0.0005, + "step": 57247 + }, + { + "epoch": 7.5, + "learning_rate": 2.0492793187696813e-07, + "loss": 0.0005, + "step": 57248 + }, + { + "epoch": 7.5, + "learning_rate": 2.048211333487782e-07, + "loss": 0.0002, + "step": 57249 + }, + { + "epoch": 7.5, + "learning_rate": 2.0471436236900731e-07, + "loss": 0.001, + "step": 57250 + }, + { + "epoch": 7.5, + "learning_rate": 2.0460761893795422e-07, + "loss": 0.0005, + "step": 57251 + }, + { + "epoch": 7.5, + "learning_rate": 2.0450090305592197e-07, + "loss": 0.0004, + "step": 57252 + }, + { + "epoch": 7.5, + "learning_rate": 2.0439421472321032e-07, + "loss": 0.0002, + "step": 57253 + }, + { + "epoch": 7.5, + "learning_rate": 2.042875539401168e-07, + "loss": 0.0005, + "step": 57254 + }, + { + "epoch": 7.5, + "learning_rate": 2.041809207069423e-07, + "loss": 0.0002, + "step": 57255 + }, + { + "epoch": 7.5, + "learning_rate": 2.0407431502398878e-07, + "loss": 0.0004, + "step": 57256 + }, + { + "epoch": 7.5, + "learning_rate": 2.039677368915527e-07, + "loss": 0.0005, + "step": 57257 + }, + { + "epoch": 7.5, + "learning_rate": 2.0386118630993824e-07, + "loss": 0.0005, + "step": 57258 + }, + { + "epoch": 7.5, + "learning_rate": 2.037546632794396e-07, + "loss": 0.0005, + "step": 57259 + }, + { + "epoch": 7.5, + "learning_rate": 2.0364816780036212e-07, + "loss": 0.0005, + "step": 57260 + }, + { + "epoch": 7.5, + "learning_rate": 2.035416998730011e-07, + "loss": 0.0003, + "step": 57261 + }, + { + "epoch": 7.5, + "learning_rate": 2.034352594976574e-07, + "loss": 0.0002, + "step": 57262 + }, + { + "epoch": 7.5, + "learning_rate": 2.0332884667462972e-07, + "loss": 0.0003, + "step": 57263 + }, + { + "epoch": 7.5, + "learning_rate": 2.0322246140421997e-07, + "loss": 0.0003, + "step": 57264 + }, + { + "epoch": 7.5, + "learning_rate": 2.0311610368672352e-07, + "loss": 0.0008, + "step": 57265 + }, + { + "epoch": 7.5, + "learning_rate": 2.0300977352244123e-07, + "loss": 0.0002, + "step": 57266 + }, + { + "epoch": 7.5, + "learning_rate": 2.0290347091167396e-07, + "loss": 0.0007, + "step": 57267 + }, + { + "epoch": 7.5, + "learning_rate": 2.0279719585471702e-07, + "loss": 0.0005, + "step": 57268 + }, + { + "epoch": 7.5, + "learning_rate": 2.026909483518724e-07, + "loss": 0.0002, + "step": 57269 + }, + { + "epoch": 7.5, + "learning_rate": 2.0258472840343658e-07, + "loss": 0.0002, + "step": 57270 + }, + { + "epoch": 7.5, + "learning_rate": 2.0247853600971034e-07, + "loss": 0.0003, + "step": 57271 + }, + { + "epoch": 7.5, + "learning_rate": 2.0237237117099129e-07, + "loss": 0.0004, + "step": 57272 + }, + { + "epoch": 7.5, + "learning_rate": 2.0226623388757916e-07, + "loss": 0.0004, + "step": 57273 + }, + { + "epoch": 7.5, + "learning_rate": 2.0216012415977038e-07, + "loss": 0.0004, + "step": 57274 + }, + { + "epoch": 7.5, + "learning_rate": 2.020540419878647e-07, + "loss": 0.0002, + "step": 57275 + }, + { + "epoch": 7.5, + "learning_rate": 2.0194798737216082e-07, + "loss": 0.0002, + "step": 57276 + }, + { + "epoch": 7.5, + "learning_rate": 2.0184196031295512e-07, + "loss": 0.0003, + "step": 57277 + }, + { + "epoch": 7.5, + "learning_rate": 2.0173596081054847e-07, + "loss": 0.0003, + "step": 57278 + }, + { + "epoch": 7.5, + "learning_rate": 2.016299888652362e-07, + "loss": 0.0004, + "step": 57279 + }, + { + "epoch": 7.5, + "learning_rate": 2.015240444773192e-07, + "loss": 0.0006, + "step": 57280 + }, + { + "epoch": 7.5, + "learning_rate": 2.0141812764709389e-07, + "loss": 0.0002, + "step": 57281 + }, + { + "epoch": 7.5, + "learning_rate": 2.013122383748567e-07, + "loss": 0.0004, + "step": 57282 + }, + { + "epoch": 7.5, + "learning_rate": 2.0120637666090846e-07, + "loss": 0.0005, + "step": 57283 + }, + { + "epoch": 7.5, + "learning_rate": 2.0110054250554566e-07, + "loss": 0.0003, + "step": 57284 + }, + { + "epoch": 7.5, + "learning_rate": 2.0099473590906472e-07, + "loss": 0.0001, + "step": 57285 + }, + { + "epoch": 7.5, + "learning_rate": 2.0088895687176535e-07, + "loss": 0.0003, + "step": 57286 + }, + { + "epoch": 7.5, + "learning_rate": 2.0078320539394402e-07, + "loss": 0.001, + "step": 57287 + }, + { + "epoch": 7.5, + "learning_rate": 2.0067748147589828e-07, + "loss": 0.0004, + "step": 57288 + }, + { + "epoch": 7.5, + "learning_rate": 2.0057178511792342e-07, + "loss": 0.0004, + "step": 57289 + }, + { + "epoch": 7.5, + "learning_rate": 2.004661163203192e-07, + "loss": 0.0003, + "step": 57290 + }, + { + "epoch": 7.5, + "learning_rate": 2.003604750833832e-07, + "loss": 0.0003, + "step": 57291 + }, + { + "epoch": 7.5, + "learning_rate": 2.0025486140741069e-07, + "loss": 0.0003, + "step": 57292 + }, + { + "epoch": 7.5, + "learning_rate": 2.0014927529269924e-07, + "loss": 0.0007, + "step": 57293 + }, + { + "epoch": 7.51, + "learning_rate": 2.0004371673954747e-07, + "loss": 0.0007, + "step": 57294 + }, + { + "epoch": 7.51, + "learning_rate": 1.9993818574825076e-07, + "loss": 0.0011, + "step": 57295 + }, + { + "epoch": 7.51, + "learning_rate": 1.9983268231910547e-07, + "loss": 0.0013, + "step": 57296 + }, + { + "epoch": 7.51, + "learning_rate": 1.9972720645240806e-07, + "loss": 0.0003, + "step": 57297 + }, + { + "epoch": 7.51, + "learning_rate": 1.996217581484572e-07, + "loss": 0.0002, + "step": 57298 + }, + { + "epoch": 7.51, + "learning_rate": 1.9951633740754705e-07, + "loss": 0.0006, + "step": 57299 + }, + { + "epoch": 7.51, + "learning_rate": 1.9941094422997632e-07, + "loss": 0.0004, + "step": 57300 + }, + { + "epoch": 7.51, + "learning_rate": 1.9930557861603915e-07, + "loss": 0.0004, + "step": 57301 + }, + { + "epoch": 7.51, + "learning_rate": 1.9920024056603427e-07, + "loss": 0.0004, + "step": 57302 + }, + { + "epoch": 7.51, + "learning_rate": 1.9909493008025694e-07, + "loss": 0.0004, + "step": 57303 + }, + { + "epoch": 7.51, + "learning_rate": 1.9898964715900248e-07, + "loss": 0.0002, + "step": 57304 + }, + { + "epoch": 7.51, + "learning_rate": 1.9888439180256736e-07, + "loss": 0.0005, + "step": 57305 + }, + { + "epoch": 7.51, + "learning_rate": 1.9877916401124907e-07, + "loss": 0.0001, + "step": 57306 + }, + { + "epoch": 7.51, + "learning_rate": 1.9867396378534186e-07, + "loss": 0.0003, + "step": 57307 + }, + { + "epoch": 7.51, + "learning_rate": 1.9856879112514215e-07, + "loss": 0.0004, + "step": 57308 + }, + { + "epoch": 7.51, + "learning_rate": 1.9846364603094636e-07, + "loss": 0.0003, + "step": 57309 + }, + { + "epoch": 7.51, + "learning_rate": 1.983585285030487e-07, + "loss": 0.0007, + "step": 57310 + }, + { + "epoch": 7.51, + "learning_rate": 1.9825343854174562e-07, + "loss": 0.0006, + "step": 57311 + }, + { + "epoch": 7.51, + "learning_rate": 1.981483761473324e-07, + "loss": 0.0009, + "step": 57312 + }, + { + "epoch": 7.51, + "learning_rate": 1.980433413201066e-07, + "loss": 0.0006, + "step": 57313 + }, + { + "epoch": 7.51, + "learning_rate": 1.9793833406036134e-07, + "loss": 0.0006, + "step": 57314 + }, + { + "epoch": 7.51, + "learning_rate": 1.97833354368393e-07, + "loss": 0.0002, + "step": 57315 + }, + { + "epoch": 7.51, + "learning_rate": 1.9772840224449475e-07, + "loss": 0.0004, + "step": 57316 + }, + { + "epoch": 7.51, + "learning_rate": 1.9762347768896406e-07, + "loss": 0.0004, + "step": 57317 + }, + { + "epoch": 7.51, + "learning_rate": 1.975185807020963e-07, + "loss": 0.0005, + "step": 57318 + }, + { + "epoch": 7.51, + "learning_rate": 1.974137112841834e-07, + "loss": 0.0004, + "step": 57319 + }, + { + "epoch": 7.51, + "learning_rate": 1.9730886943552408e-07, + "loss": 0.0007, + "step": 57320 + }, + { + "epoch": 7.51, + "learning_rate": 1.972040551564114e-07, + "loss": 0.0002, + "step": 57321 + }, + { + "epoch": 7.51, + "learning_rate": 1.9709926844713956e-07, + "loss": 0.0004, + "step": 57322 + }, + { + "epoch": 7.51, + "learning_rate": 1.9699450930800502e-07, + "loss": 0.0004, + "step": 57323 + }, + { + "epoch": 7.51, + "learning_rate": 1.9688977773930083e-07, + "loss": 0.0003, + "step": 57324 + }, + { + "epoch": 7.51, + "learning_rate": 1.9678507374132239e-07, + "loss": 0.0007, + "step": 57325 + }, + { + "epoch": 7.51, + "learning_rate": 1.9668039731436383e-07, + "loss": 0.0004, + "step": 57326 + }, + { + "epoch": 7.51, + "learning_rate": 1.965757484587183e-07, + "loss": 0.0004, + "step": 57327 + }, + { + "epoch": 7.51, + "learning_rate": 1.964711271746833e-07, + "loss": 0.0003, + "step": 57328 + }, + { + "epoch": 7.51, + "learning_rate": 1.9636653346255085e-07, + "loss": 0.0006, + "step": 57329 + }, + { + "epoch": 7.51, + "learning_rate": 1.9626196732261516e-07, + "loss": 0.0005, + "step": 57330 + }, + { + "epoch": 7.51, + "learning_rate": 1.961574287551693e-07, + "loss": 0.0007, + "step": 57331 + }, + { + "epoch": 7.51, + "learning_rate": 1.9605291776050973e-07, + "loss": 0.0004, + "step": 57332 + }, + { + "epoch": 7.51, + "learning_rate": 1.9594843433892953e-07, + "loss": 0.0002, + "step": 57333 + }, + { + "epoch": 7.51, + "learning_rate": 1.9584397849072178e-07, + "loss": 0.0005, + "step": 57334 + }, + { + "epoch": 7.51, + "learning_rate": 1.9573955021618074e-07, + "loss": 0.0006, + "step": 57335 + }, + { + "epoch": 7.51, + "learning_rate": 1.9563514951560058e-07, + "loss": 0.0001, + "step": 57336 + }, + { + "epoch": 7.51, + "learning_rate": 1.955307763892744e-07, + "loss": 0.0006, + "step": 57337 + }, + { + "epoch": 7.51, + "learning_rate": 1.9542643083749535e-07, + "loss": 0.0004, + "step": 57338 + }, + { + "epoch": 7.51, + "learning_rate": 1.9532211286055756e-07, + "loss": 0.001, + "step": 57339 + }, + { + "epoch": 7.51, + "learning_rate": 1.952178224587542e-07, + "loss": 0.0007, + "step": 57340 + }, + { + "epoch": 7.51, + "learning_rate": 1.9511355963237833e-07, + "loss": 0.0002, + "step": 57341 + }, + { + "epoch": 7.51, + "learning_rate": 1.9500932438172305e-07, + "loss": 0.0004, + "step": 57342 + }, + { + "epoch": 7.51, + "learning_rate": 1.9490511670708368e-07, + "loss": 0.0009, + "step": 57343 + }, + { + "epoch": 7.51, + "learning_rate": 1.9480093660874887e-07, + "loss": 0.0006, + "step": 57344 + }, + { + "epoch": 7.51, + "learning_rate": 1.946967840870162e-07, + "loss": 0.0006, + "step": 57345 + }, + { + "epoch": 7.51, + "learning_rate": 1.945926591421765e-07, + "loss": 0.0004, + "step": 57346 + }, + { + "epoch": 7.51, + "learning_rate": 1.944885617745218e-07, + "loss": 0.0004, + "step": 57347 + }, + { + "epoch": 7.51, + "learning_rate": 1.943844919843474e-07, + "loss": 0.0006, + "step": 57348 + }, + { + "epoch": 7.51, + "learning_rate": 1.9428044977194417e-07, + "loss": 0.0003, + "step": 57349 + }, + { + "epoch": 7.51, + "learning_rate": 1.9417643513760298e-07, + "loss": 0.0002, + "step": 57350 + }, + { + "epoch": 7.51, + "learning_rate": 1.9407244808162028e-07, + "loss": 0.0004, + "step": 57351 + }, + { + "epoch": 7.51, + "learning_rate": 1.9396848860428697e-07, + "loss": 0.0012, + "step": 57352 + }, + { + "epoch": 7.51, + "learning_rate": 1.9386455670589389e-07, + "loss": 0.0007, + "step": 57353 + }, + { + "epoch": 7.51, + "learning_rate": 1.9376065238673414e-07, + "loss": 0.0006, + "step": 57354 + }, + { + "epoch": 7.51, + "learning_rate": 1.9365677564710083e-07, + "loss": 0.0003, + "step": 57355 + }, + { + "epoch": 7.51, + "learning_rate": 1.9355292648728597e-07, + "loss": 0.0004, + "step": 57356 + }, + { + "epoch": 7.51, + "learning_rate": 1.9344910490758152e-07, + "loss": 0.0003, + "step": 57357 + }, + { + "epoch": 7.51, + "learning_rate": 1.9334531090827835e-07, + "loss": 0.0005, + "step": 57358 + }, + { + "epoch": 7.51, + "learning_rate": 1.932415444896707e-07, + "loss": 0.0004, + "step": 57359 + }, + { + "epoch": 7.51, + "learning_rate": 1.9313780565204832e-07, + "loss": 0.0006, + "step": 57360 + }, + { + "epoch": 7.51, + "learning_rate": 1.930340943957032e-07, + "loss": 0.0003, + "step": 57361 + }, + { + "epoch": 7.51, + "learning_rate": 1.9293041072092734e-07, + "loss": 0.0004, + "step": 57362 + }, + { + "epoch": 7.51, + "learning_rate": 1.928267546280127e-07, + "loss": 0.0004, + "step": 57363 + }, + { + "epoch": 7.51, + "learning_rate": 1.927231261172513e-07, + "loss": 0.0005, + "step": 57364 + }, + { + "epoch": 7.51, + "learning_rate": 1.926195251889318e-07, + "loss": 0.0005, + "step": 57365 + }, + { + "epoch": 7.51, + "learning_rate": 1.9251595184334836e-07, + "loss": 0.0007, + "step": 57366 + }, + { + "epoch": 7.51, + "learning_rate": 1.9241240608079303e-07, + "loss": 0.0003, + "step": 57367 + }, + { + "epoch": 7.51, + "learning_rate": 1.9230888790155444e-07, + "loss": 0.0005, + "step": 57368 + }, + { + "epoch": 7.51, + "learning_rate": 1.922053973059246e-07, + "loss": 0.0002, + "step": 57369 + }, + { + "epoch": 7.52, + "learning_rate": 1.9210193429419543e-07, + "loss": 0.0006, + "step": 57370 + }, + { + "epoch": 7.52, + "learning_rate": 1.9199849886665677e-07, + "loss": 0.0005, + "step": 57371 + }, + { + "epoch": 7.52, + "learning_rate": 1.918950910236006e-07, + "loss": 0.0004, + "step": 57372 + }, + { + "epoch": 7.52, + "learning_rate": 1.9179171076531667e-07, + "loss": 0.0006, + "step": 57373 + }, + { + "epoch": 7.52, + "learning_rate": 1.9168835809209584e-07, + "loss": 0.0003, + "step": 57374 + }, + { + "epoch": 7.52, + "learning_rate": 1.9158503300423015e-07, + "loss": 0.0009, + "step": 57375 + }, + { + "epoch": 7.52, + "learning_rate": 1.9148173550200933e-07, + "loss": 0.0009, + "step": 57376 + }, + { + "epoch": 7.52, + "learning_rate": 1.9137846558572204e-07, + "loss": 0.0002, + "step": 57377 + }, + { + "epoch": 7.52, + "learning_rate": 1.912752232556614e-07, + "loss": 0.0003, + "step": 57378 + }, + { + "epoch": 7.52, + "learning_rate": 1.9117200851211825e-07, + "loss": 0.0004, + "step": 57379 + }, + { + "epoch": 7.52, + "learning_rate": 1.9106882135538018e-07, + "loss": 0.0005, + "step": 57380 + }, + { + "epoch": 7.52, + "learning_rate": 1.9096566178573805e-07, + "loss": 0.0002, + "step": 57381 + }, + { + "epoch": 7.52, + "learning_rate": 1.9086252980348385e-07, + "loss": 0.0017, + "step": 57382 + }, + { + "epoch": 7.52, + "learning_rate": 1.9075942540890624e-07, + "loss": 0.0003, + "step": 57383 + }, + { + "epoch": 7.52, + "learning_rate": 1.9065634860229498e-07, + "loss": 0.0007, + "step": 57384 + }, + { + "epoch": 7.52, + "learning_rate": 1.9055329938394095e-07, + "loss": 0.0004, + "step": 57385 + }, + { + "epoch": 7.52, + "learning_rate": 1.9045027775413282e-07, + "loss": 0.0004, + "step": 57386 + }, + { + "epoch": 7.52, + "learning_rate": 1.9034728371316148e-07, + "loss": 0.0004, + "step": 57387 + }, + { + "epoch": 7.52, + "learning_rate": 1.9024431726131443e-07, + "loss": 0.001, + "step": 57388 + }, + { + "epoch": 7.52, + "learning_rate": 1.901413783988837e-07, + "loss": 0.0007, + "step": 57389 + }, + { + "epoch": 7.52, + "learning_rate": 1.9003846712615904e-07, + "loss": 0.0002, + "step": 57390 + }, + { + "epoch": 7.52, + "learning_rate": 1.89935583443428e-07, + "loss": 0.0005, + "step": 57391 + }, + { + "epoch": 7.52, + "learning_rate": 1.8983272735098035e-07, + "loss": 0.0005, + "step": 57392 + }, + { + "epoch": 7.52, + "learning_rate": 1.8972989884910587e-07, + "loss": 0.0003, + "step": 57393 + }, + { + "epoch": 7.52, + "learning_rate": 1.896270979380943e-07, + "loss": 0.001, + "step": 57394 + }, + { + "epoch": 7.52, + "learning_rate": 1.8952432461823324e-07, + "loss": 0.0007, + "step": 57395 + }, + { + "epoch": 7.52, + "learning_rate": 1.894215788898124e-07, + "loss": 0.0006, + "step": 57396 + }, + { + "epoch": 7.52, + "learning_rate": 1.8931886075312267e-07, + "loss": 0.001, + "step": 57397 + }, + { + "epoch": 7.52, + "learning_rate": 1.892161702084494e-07, + "loss": 0.0002, + "step": 57398 + }, + { + "epoch": 7.52, + "learning_rate": 1.8911350725608346e-07, + "loss": 0.0005, + "step": 57399 + }, + { + "epoch": 7.52, + "learning_rate": 1.8901087189631352e-07, + "loss": 0.0004, + "step": 57400 + }, + { + "epoch": 7.52, + "learning_rate": 1.889082641294282e-07, + "loss": 0.0004, + "step": 57401 + }, + { + "epoch": 7.52, + "learning_rate": 1.8880568395571619e-07, + "loss": 0.0009, + "step": 57402 + }, + { + "epoch": 7.52, + "learning_rate": 1.8870313137546503e-07, + "loss": 0.0003, + "step": 57403 + }, + { + "epoch": 7.52, + "learning_rate": 1.886006063889645e-07, + "loss": 0.0004, + "step": 57404 + }, + { + "epoch": 7.52, + "learning_rate": 1.8849810899650212e-07, + "loss": 0.0006, + "step": 57405 + }, + { + "epoch": 7.52, + "learning_rate": 1.8839563919836657e-07, + "loss": 0.0003, + "step": 57406 + }, + { + "epoch": 7.52, + "learning_rate": 1.8829319699484427e-07, + "loss": 0.0005, + "step": 57407 + }, + { + "epoch": 7.52, + "learning_rate": 1.881907823862261e-07, + "loss": 0.0002, + "step": 57408 + }, + { + "epoch": 7.52, + "learning_rate": 1.8808839537279965e-07, + "loss": 0.0005, + "step": 57409 + }, + { + "epoch": 7.52, + "learning_rate": 1.879860359548502e-07, + "loss": 0.0002, + "step": 57410 + }, + { + "epoch": 7.52, + "learning_rate": 1.8788370413266865e-07, + "loss": 0.0003, + "step": 57411 + }, + { + "epoch": 7.52, + "learning_rate": 1.8778139990654143e-07, + "loss": 0.0002, + "step": 57412 + }, + { + "epoch": 7.52, + "learning_rate": 1.8767912327675607e-07, + "loss": 0.0002, + "step": 57413 + }, + { + "epoch": 7.52, + "learning_rate": 1.8757687424360128e-07, + "loss": 0.0009, + "step": 57414 + }, + { + "epoch": 7.52, + "learning_rate": 1.8747465280736344e-07, + "loss": 0.0001, + "step": 57415 + }, + { + "epoch": 7.52, + "learning_rate": 1.8737245896833123e-07, + "loss": 0.0002, + "step": 57416 + }, + { + "epoch": 7.52, + "learning_rate": 1.872702927267911e-07, + "loss": 0.0005, + "step": 57417 + }, + { + "epoch": 7.52, + "learning_rate": 1.8716815408303168e-07, + "loss": 0.0007, + "step": 57418 + }, + { + "epoch": 7.52, + "learning_rate": 1.8706604303733723e-07, + "loss": 0.0006, + "step": 57419 + }, + { + "epoch": 7.52, + "learning_rate": 1.8696395958999857e-07, + "loss": 0.0003, + "step": 57420 + }, + { + "epoch": 7.52, + "learning_rate": 1.8686190374129998e-07, + "loss": 0.0002, + "step": 57421 + }, + { + "epoch": 7.52, + "learning_rate": 1.867598754915312e-07, + "loss": 0.0002, + "step": 57422 + }, + { + "epoch": 7.52, + "learning_rate": 1.8665787484097642e-07, + "loss": 0.0002, + "step": 57423 + }, + { + "epoch": 7.52, + "learning_rate": 1.8655590178992432e-07, + "loss": 0.0003, + "step": 57424 + }, + { + "epoch": 7.52, + "learning_rate": 1.8645395633866136e-07, + "loss": 0.0003, + "step": 57425 + }, + { + "epoch": 7.52, + "learning_rate": 1.8635203848747395e-07, + "loss": 0.0004, + "step": 57426 + }, + { + "epoch": 7.52, + "learning_rate": 1.8625014823664745e-07, + "loss": 0.0003, + "step": 57427 + }, + { + "epoch": 7.52, + "learning_rate": 1.8614828558647157e-07, + "loss": 0.0004, + "step": 57428 + }, + { + "epoch": 7.52, + "learning_rate": 1.8604645053723058e-07, + "loss": 0.0002, + "step": 57429 + }, + { + "epoch": 7.52, + "learning_rate": 1.859446430892109e-07, + "loss": 0.0001, + "step": 57430 + }, + { + "epoch": 7.52, + "learning_rate": 1.8584286324269896e-07, + "loss": 0.0003, + "step": 57431 + }, + { + "epoch": 7.52, + "learning_rate": 1.8574111099798232e-07, + "loss": 0.0007, + "step": 57432 + }, + { + "epoch": 7.52, + "learning_rate": 1.856393863553463e-07, + "loss": 0.0006, + "step": 57433 + }, + { + "epoch": 7.52, + "learning_rate": 1.855376893150751e-07, + "loss": 0.0006, + "step": 57434 + }, + { + "epoch": 7.52, + "learning_rate": 1.854360198774585e-07, + "loss": 0.0005, + "step": 57435 + }, + { + "epoch": 7.52, + "learning_rate": 1.8533437804277964e-07, + "loss": 0.0003, + "step": 57436 + }, + { + "epoch": 7.52, + "learning_rate": 1.8523276381132605e-07, + "loss": 0.0006, + "step": 57437 + }, + { + "epoch": 7.52, + "learning_rate": 1.851311771833808e-07, + "loss": 0.0003, + "step": 57438 + }, + { + "epoch": 7.52, + "learning_rate": 1.8502961815923259e-07, + "loss": 0.0004, + "step": 57439 + }, + { + "epoch": 7.52, + "learning_rate": 1.8492808673916563e-07, + "loss": 0.0007, + "step": 57440 + }, + { + "epoch": 7.52, + "learning_rate": 1.8482658292346633e-07, + "loss": 0.0005, + "step": 57441 + }, + { + "epoch": 7.52, + "learning_rate": 1.8472510671241783e-07, + "loss": 0.0005, + "step": 57442 + }, + { + "epoch": 7.52, + "learning_rate": 1.8462365810630878e-07, + "loss": 0.0007, + "step": 57443 + }, + { + "epoch": 7.52, + "learning_rate": 1.845222371054234e-07, + "loss": 0.0004, + "step": 57444 + }, + { + "epoch": 7.52, + "learning_rate": 1.8442084371004588e-07, + "loss": 0.0006, + "step": 57445 + }, + { + "epoch": 7.53, + "learning_rate": 1.8431947792046155e-07, + "loss": 0.0002, + "step": 57446 + }, + { + "epoch": 7.53, + "learning_rate": 1.842181397369569e-07, + "loss": 0.0009, + "step": 57447 + }, + { + "epoch": 7.53, + "learning_rate": 1.841168291598161e-07, + "loss": 0.0004, + "step": 57448 + }, + { + "epoch": 7.53, + "learning_rate": 1.8401554618932226e-07, + "loss": 0.0016, + "step": 57449 + }, + { + "epoch": 7.53, + "learning_rate": 1.8391429082576406e-07, + "loss": 0.0011, + "step": 57450 + }, + { + "epoch": 7.53, + "learning_rate": 1.8381306306942348e-07, + "loss": 0.0005, + "step": 57451 + }, + { + "epoch": 7.53, + "learning_rate": 1.8371186292058585e-07, + "loss": 0.0002, + "step": 57452 + }, + { + "epoch": 7.53, + "learning_rate": 1.8361069037953538e-07, + "loss": 0.0003, + "step": 57453 + }, + { + "epoch": 7.53, + "learning_rate": 1.835095454465563e-07, + "loss": 0.0007, + "step": 57454 + }, + { + "epoch": 7.53, + "learning_rate": 1.8340842812193615e-07, + "loss": 0.0006, + "step": 57455 + }, + { + "epoch": 7.53, + "learning_rate": 1.8330733840595584e-07, + "loss": 0.0003, + "step": 57456 + }, + { + "epoch": 7.53, + "learning_rate": 1.8320627629890064e-07, + "loss": 0.0004, + "step": 57457 + }, + { + "epoch": 7.53, + "learning_rate": 1.8310524180105483e-07, + "loss": 0.0002, + "step": 57458 + }, + { + "epoch": 7.53, + "learning_rate": 1.830042349127037e-07, + "loss": 0.0003, + "step": 57459 + }, + { + "epoch": 7.53, + "learning_rate": 1.8290325563413037e-07, + "loss": 0.0005, + "step": 57460 + }, + { + "epoch": 7.53, + "learning_rate": 1.8280230396561794e-07, + "loss": 0.001, + "step": 57461 + }, + { + "epoch": 7.53, + "learning_rate": 1.8270137990745174e-07, + "loss": 0.0005, + "step": 57462 + }, + { + "epoch": 7.53, + "learning_rate": 1.8260048345991487e-07, + "loss": 0.0003, + "step": 57463 + }, + { + "epoch": 7.53, + "learning_rate": 1.8249961462329158e-07, + "loss": 0.0004, + "step": 57464 + }, + { + "epoch": 7.53, + "learning_rate": 1.8239877339786494e-07, + "loss": 0.0003, + "step": 57465 + }, + { + "epoch": 7.53, + "learning_rate": 1.8229795978391916e-07, + "loss": 0.0003, + "step": 57466 + }, + { + "epoch": 7.53, + "learning_rate": 1.821971737817374e-07, + "loss": 0.0012, + "step": 57467 + }, + { + "epoch": 7.53, + "learning_rate": 1.8209641539160383e-07, + "loss": 0.0004, + "step": 57468 + }, + { + "epoch": 7.53, + "learning_rate": 1.8199568461380047e-07, + "loss": 0.0006, + "step": 57469 + }, + { + "epoch": 7.53, + "learning_rate": 1.8189498144861262e-07, + "loss": 0.0004, + "step": 57470 + }, + { + "epoch": 7.53, + "learning_rate": 1.817943058963212e-07, + "loss": 0.0006, + "step": 57471 + }, + { + "epoch": 7.53, + "learning_rate": 1.816936579572104e-07, + "loss": 0.0002, + "step": 57472 + }, + { + "epoch": 7.53, + "learning_rate": 1.8159303763156334e-07, + "loss": 0.0004, + "step": 57473 + }, + { + "epoch": 7.53, + "learning_rate": 1.8149244491966421e-07, + "loss": 0.0006, + "step": 57474 + }, + { + "epoch": 7.53, + "learning_rate": 1.8139187982179285e-07, + "loss": 0.0004, + "step": 57475 + }, + { + "epoch": 7.53, + "learning_rate": 1.8129134233823564e-07, + "loss": 0.0003, + "step": 57476 + }, + { + "epoch": 7.53, + "learning_rate": 1.8119083246927237e-07, + "loss": 0.0006, + "step": 57477 + }, + { + "epoch": 7.53, + "learning_rate": 1.8109035021518728e-07, + "loss": 0.0007, + "step": 57478 + }, + { + "epoch": 7.53, + "learning_rate": 1.8098989557626346e-07, + "loss": 0.0002, + "step": 57479 + }, + { + "epoch": 7.53, + "learning_rate": 1.8088946855278178e-07, + "loss": 0.0005, + "step": 57480 + }, + { + "epoch": 7.53, + "learning_rate": 1.8078906914502647e-07, + "loss": 0.0005, + "step": 57481 + }, + { + "epoch": 7.53, + "learning_rate": 1.8068869735327843e-07, + "loss": 0.0003, + "step": 57482 + }, + { + "epoch": 7.53, + "learning_rate": 1.8058835317782076e-07, + "loss": 0.0008, + "step": 57483 + }, + { + "epoch": 7.53, + "learning_rate": 1.8048803661893544e-07, + "loss": 0.0005, + "step": 57484 + }, + { + "epoch": 7.53, + "learning_rate": 1.8038774767690447e-07, + "loss": 0.0013, + "step": 57485 + }, + { + "epoch": 7.53, + "learning_rate": 1.8028748635200876e-07, + "loss": 0.0005, + "step": 57486 + }, + { + "epoch": 7.53, + "learning_rate": 1.8018725264453364e-07, + "loss": 0.0005, + "step": 57487 + }, + { + "epoch": 7.53, + "learning_rate": 1.8008704655475772e-07, + "loss": 0.0004, + "step": 57488 + }, + { + "epoch": 7.53, + "learning_rate": 1.7998686808296418e-07, + "loss": 0.0002, + "step": 57489 + }, + { + "epoch": 7.53, + "learning_rate": 1.7988671722943495e-07, + "loss": 0.0003, + "step": 57490 + }, + { + "epoch": 7.53, + "learning_rate": 1.797865939944521e-07, + "loss": 0.0002, + "step": 57491 + }, + { + "epoch": 7.53, + "learning_rate": 1.7968649837829534e-07, + "loss": 0.0003, + "step": 57492 + }, + { + "epoch": 7.53, + "learning_rate": 1.7958643038124778e-07, + "loss": 0.0001, + "step": 57493 + }, + { + "epoch": 7.53, + "learning_rate": 1.7948639000359037e-07, + "loss": 0.0017, + "step": 57494 + }, + { + "epoch": 7.53, + "learning_rate": 1.7938637724560393e-07, + "loss": 0.0006, + "step": 57495 + }, + { + "epoch": 7.53, + "learning_rate": 1.792863921075716e-07, + "loss": 0.0007, + "step": 57496 + }, + { + "epoch": 7.53, + "learning_rate": 1.7918643458977203e-07, + "loss": 0.0003, + "step": 57497 + }, + { + "epoch": 7.53, + "learning_rate": 1.7908650469248834e-07, + "loss": 0.0003, + "step": 57498 + }, + { + "epoch": 7.53, + "learning_rate": 1.789866024160003e-07, + "loss": 0.0003, + "step": 57499 + }, + { + "epoch": 7.53, + "learning_rate": 1.788867277605899e-07, + "loss": 0.0003, + "step": 57500 + }, + { + "epoch": 7.53, + "learning_rate": 1.7878688072653805e-07, + "loss": 0.0009, + "step": 57501 + }, + { + "epoch": 7.53, + "learning_rate": 1.786870613141245e-07, + "loss": 0.0008, + "step": 57502 + }, + { + "epoch": 7.53, + "learning_rate": 1.7858726952363014e-07, + "loss": 0.0004, + "step": 57503 + }, + { + "epoch": 7.53, + "learning_rate": 1.78487505355337e-07, + "loss": 0.0005, + "step": 57504 + }, + { + "epoch": 7.53, + "learning_rate": 1.7838776880952368e-07, + "loss": 0.0003, + "step": 57505 + }, + { + "epoch": 7.53, + "learning_rate": 1.7828805988647336e-07, + "loss": 0.0005, + "step": 57506 + }, + { + "epoch": 7.53, + "learning_rate": 1.7818837858646243e-07, + "loss": 0.0005, + "step": 57507 + }, + { + "epoch": 7.53, + "learning_rate": 1.7808872490977514e-07, + "loss": 0.0005, + "step": 57508 + }, + { + "epoch": 7.53, + "learning_rate": 1.7798909885669013e-07, + "loss": 0.0002, + "step": 57509 + }, + { + "epoch": 7.53, + "learning_rate": 1.7788950042748832e-07, + "loss": 0.001, + "step": 57510 + }, + { + "epoch": 7.53, + "learning_rate": 1.7778992962244834e-07, + "loss": 0.0004, + "step": 57511 + }, + { + "epoch": 7.53, + "learning_rate": 1.776903864418511e-07, + "loss": 0.0003, + "step": 57512 + }, + { + "epoch": 7.53, + "learning_rate": 1.7759087088597748e-07, + "loss": 0.0002, + "step": 57513 + }, + { + "epoch": 7.53, + "learning_rate": 1.7749138295510503e-07, + "loss": 0.0005, + "step": 57514 + }, + { + "epoch": 7.53, + "learning_rate": 1.7739192264951576e-07, + "loss": 0.0003, + "step": 57515 + }, + { + "epoch": 7.53, + "learning_rate": 1.7729248996948945e-07, + "loss": 0.0005, + "step": 57516 + }, + { + "epoch": 7.53, + "learning_rate": 1.7719308491530473e-07, + "loss": 0.0002, + "step": 57517 + }, + { + "epoch": 7.53, + "learning_rate": 1.770937074872403e-07, + "loss": 0.0003, + "step": 57518 + }, + { + "epoch": 7.53, + "learning_rate": 1.7699435768557592e-07, + "loss": 0.0004, + "step": 57519 + }, + { + "epoch": 7.53, + "learning_rate": 1.7689503551059362e-07, + "loss": 0.0003, + "step": 57520 + }, + { + "epoch": 7.53, + "learning_rate": 1.7679574096257091e-07, + "loss": 0.0012, + "step": 57521 + }, + { + "epoch": 7.53, + "learning_rate": 1.7669647404178536e-07, + "loss": 0.0003, + "step": 57522 + }, + { + "epoch": 7.54, + "learning_rate": 1.7659723474852008e-07, + "loss": 0.0003, + "step": 57523 + }, + { + "epoch": 7.54, + "learning_rate": 1.7649802308305042e-07, + "loss": 0.0008, + "step": 57524 + }, + { + "epoch": 7.54, + "learning_rate": 1.7639883904565835e-07, + "loss": 0.0005, + "step": 57525 + }, + { + "epoch": 7.54, + "learning_rate": 1.7629968263661924e-07, + "loss": 0.0003, + "step": 57526 + }, + { + "epoch": 7.54, + "learning_rate": 1.7620055385621616e-07, + "loss": 0.0005, + "step": 57527 + }, + { + "epoch": 7.54, + "learning_rate": 1.7610145270472556e-07, + "loss": 0.0004, + "step": 57528 + }, + { + "epoch": 7.54, + "learning_rate": 1.7600237918242502e-07, + "loss": 0.0005, + "step": 57529 + }, + { + "epoch": 7.54, + "learning_rate": 1.759033332895954e-07, + "loss": 0.0008, + "step": 57530 + }, + { + "epoch": 7.54, + "learning_rate": 1.758043150265154e-07, + "loss": 0.0007, + "step": 57531 + }, + { + "epoch": 7.54, + "learning_rate": 1.7570532439346145e-07, + "loss": 0.0005, + "step": 57532 + }, + { + "epoch": 7.54, + "learning_rate": 1.756063613907144e-07, + "loss": 0.0011, + "step": 57533 + }, + { + "epoch": 7.54, + "learning_rate": 1.7550742601854965e-07, + "loss": 0.0008, + "step": 57534 + }, + { + "epoch": 7.54, + "learning_rate": 1.7540851827724802e-07, + "loss": 0.0004, + "step": 57535 + }, + { + "epoch": 7.54, + "learning_rate": 1.7530963816708713e-07, + "loss": 0.0005, + "step": 57536 + }, + { + "epoch": 7.54, + "learning_rate": 1.7521078568834337e-07, + "loss": 0.0002, + "step": 57537 + }, + { + "epoch": 7.54, + "learning_rate": 1.7511196084129655e-07, + "loss": 0.0003, + "step": 57538 + }, + { + "epoch": 7.54, + "learning_rate": 1.7501316362622534e-07, + "loss": 0.0002, + "step": 57539 + }, + { + "epoch": 7.54, + "learning_rate": 1.7491439404340615e-07, + "loss": 0.0001, + "step": 57540 + }, + { + "epoch": 7.54, + "learning_rate": 1.7481565209311547e-07, + "loss": 0.0007, + "step": 57541 + }, + { + "epoch": 7.54, + "learning_rate": 1.7471693777563304e-07, + "loss": 0.0003, + "step": 57542 + }, + { + "epoch": 7.54, + "learning_rate": 1.7461825109123643e-07, + "loss": 0.0003, + "step": 57543 + }, + { + "epoch": 7.54, + "learning_rate": 1.7451959204020318e-07, + "loss": 0.0005, + "step": 57544 + }, + { + "epoch": 7.54, + "learning_rate": 1.7442096062280867e-07, + "loss": 0.0008, + "step": 57545 + }, + { + "epoch": 7.54, + "learning_rate": 1.7432235683933375e-07, + "loss": 0.0003, + "step": 57546 + }, + { + "epoch": 7.54, + "learning_rate": 1.7422378069005374e-07, + "loss": 0.0005, + "step": 57547 + }, + { + "epoch": 7.54, + "learning_rate": 1.7412523217524625e-07, + "loss": 0.0007, + "step": 57548 + }, + { + "epoch": 7.54, + "learning_rate": 1.7402671129518655e-07, + "loss": 0.0009, + "step": 57549 + }, + { + "epoch": 7.54, + "learning_rate": 1.739282180501556e-07, + "loss": 0.0004, + "step": 57550 + }, + { + "epoch": 7.54, + "learning_rate": 1.7382975244042756e-07, + "loss": 0.0002, + "step": 57551 + }, + { + "epoch": 7.54, + "learning_rate": 1.737313144662789e-07, + "loss": 0.0002, + "step": 57552 + }, + { + "epoch": 7.54, + "learning_rate": 1.7363290412798827e-07, + "loss": 0.0002, + "step": 57553 + }, + { + "epoch": 7.54, + "learning_rate": 1.7353452142583216e-07, + "loss": 0.0004, + "step": 57554 + }, + { + "epoch": 7.54, + "learning_rate": 1.7343616636008698e-07, + "loss": 0.0004, + "step": 57555 + }, + { + "epoch": 7.54, + "learning_rate": 1.733378389310292e-07, + "loss": 0.0007, + "step": 57556 + }, + { + "epoch": 7.54, + "learning_rate": 1.7323953913893522e-07, + "loss": 0.0004, + "step": 57557 + }, + { + "epoch": 7.54, + "learning_rate": 1.7314126698408263e-07, + "loss": 0.0004, + "step": 57558 + }, + { + "epoch": 7.54, + "learning_rate": 1.7304302246674676e-07, + "loss": 0.0003, + "step": 57559 + }, + { + "epoch": 7.54, + "learning_rate": 1.7294480558720294e-07, + "loss": 0.0002, + "step": 57560 + }, + { + "epoch": 7.54, + "learning_rate": 1.7284661634572875e-07, + "loss": 0.0009, + "step": 57561 + }, + { + "epoch": 7.54, + "learning_rate": 1.727484547426006e-07, + "loss": 0.0005, + "step": 57562 + }, + { + "epoch": 7.54, + "learning_rate": 1.7265032077809386e-07, + "loss": 0.0007, + "step": 57563 + }, + { + "epoch": 7.54, + "learning_rate": 1.7255221445248381e-07, + "loss": 0.0002, + "step": 57564 + }, + { + "epoch": 7.54, + "learning_rate": 1.724541357660492e-07, + "loss": 0.0004, + "step": 57565 + }, + { + "epoch": 7.54, + "learning_rate": 1.7235608471906308e-07, + "loss": 0.0006, + "step": 57566 + }, + { + "epoch": 7.54, + "learning_rate": 1.72258061311803e-07, + "loss": 0.0013, + "step": 57567 + }, + { + "epoch": 7.54, + "learning_rate": 1.7216006554454213e-07, + "loss": 0.0005, + "step": 57568 + }, + { + "epoch": 7.54, + "learning_rate": 1.7206209741755908e-07, + "loss": 0.001, + "step": 57569 + }, + { + "epoch": 7.54, + "learning_rate": 1.719641569311281e-07, + "loss": 0.0002, + "step": 57570 + }, + { + "epoch": 7.54, + "learning_rate": 1.7186624408552343e-07, + "loss": 0.0003, + "step": 57571 + }, + { + "epoch": 7.54, + "learning_rate": 1.717683588810215e-07, + "loss": 0.0003, + "step": 57572 + }, + { + "epoch": 7.54, + "learning_rate": 1.7167050131789765e-07, + "loss": 0.0003, + "step": 57573 + }, + { + "epoch": 7.54, + "learning_rate": 1.715726713964283e-07, + "loss": 0.0003, + "step": 57574 + }, + { + "epoch": 7.54, + "learning_rate": 1.714748691168877e-07, + "loss": 0.0003, + "step": 57575 + }, + { + "epoch": 7.54, + "learning_rate": 1.7137709447954898e-07, + "loss": 0.0003, + "step": 57576 + }, + { + "epoch": 7.54, + "learning_rate": 1.7127934748468967e-07, + "loss": 0.001, + "step": 57577 + }, + { + "epoch": 7.54, + "learning_rate": 1.7118162813258398e-07, + "loss": 0.0005, + "step": 57578 + }, + { + "epoch": 7.54, + "learning_rate": 1.7108393642350618e-07, + "loss": 0.0009, + "step": 57579 + }, + { + "epoch": 7.54, + "learning_rate": 1.7098627235773158e-07, + "loss": 0.0003, + "step": 57580 + }, + { + "epoch": 7.54, + "learning_rate": 1.708886359355344e-07, + "loss": 0.0003, + "step": 57581 + }, + { + "epoch": 7.54, + "learning_rate": 1.7079102715719e-07, + "loss": 0.0006, + "step": 57582 + }, + { + "epoch": 7.54, + "learning_rate": 1.7069344602297144e-07, + "loss": 0.0004, + "step": 57583 + }, + { + "epoch": 7.54, + "learning_rate": 1.7059589253315412e-07, + "loss": 0.0003, + "step": 57584 + }, + { + "epoch": 7.54, + "learning_rate": 1.7049836668801334e-07, + "loss": 0.0002, + "step": 57585 + }, + { + "epoch": 7.54, + "learning_rate": 1.7040086848782222e-07, + "loss": 0.0006, + "step": 57586 + }, + { + "epoch": 7.54, + "learning_rate": 1.7030339793285498e-07, + "loss": 0.0002, + "step": 57587 + }, + { + "epoch": 7.54, + "learning_rate": 1.7020595502338698e-07, + "loss": 0.0006, + "step": 57588 + }, + { + "epoch": 7.54, + "learning_rate": 1.7010853975969022e-07, + "loss": 0.0004, + "step": 57589 + }, + { + "epoch": 7.54, + "learning_rate": 1.7001115214204e-07, + "loss": 0.0006, + "step": 57590 + }, + { + "epoch": 7.54, + "learning_rate": 1.6991379217070947e-07, + "loss": 0.0002, + "step": 57591 + }, + { + "epoch": 7.54, + "learning_rate": 1.6981645984597396e-07, + "loss": 0.0003, + "step": 57592 + }, + { + "epoch": 7.54, + "learning_rate": 1.6971915516810543e-07, + "loss": 0.0007, + "step": 57593 + }, + { + "epoch": 7.54, + "learning_rate": 1.6962187813737817e-07, + "loss": 0.0005, + "step": 57594 + }, + { + "epoch": 7.54, + "learning_rate": 1.695246287540653e-07, + "loss": 0.0007, + "step": 57595 + }, + { + "epoch": 7.54, + "learning_rate": 1.6942740701844207e-07, + "loss": 0.0003, + "step": 57596 + }, + { + "epoch": 7.54, + "learning_rate": 1.6933021293078055e-07, + "loss": 0.0002, + "step": 57597 + }, + { + "epoch": 7.54, + "learning_rate": 1.6923304649135386e-07, + "loss": 0.0006, + "step": 57598 + }, + { + "epoch": 7.55, + "learning_rate": 1.691359077004351e-07, + "loss": 0.0002, + "step": 57599 + }, + { + "epoch": 7.55, + "learning_rate": 1.6903879655829957e-07, + "loss": 0.0006, + "step": 57600 + }, + { + "epoch": 7.55, + "learning_rate": 1.6894171306521822e-07, + "loss": 0.0003, + "step": 57601 + }, + { + "epoch": 7.55, + "learning_rate": 1.6884465722146415e-07, + "loss": 0.0005, + "step": 57602 + }, + { + "epoch": 7.55, + "learning_rate": 1.6874762902731268e-07, + "loss": 0.0006, + "step": 57603 + }, + { + "epoch": 7.55, + "learning_rate": 1.686506284830336e-07, + "loss": 0.0003, + "step": 57604 + }, + { + "epoch": 7.55, + "learning_rate": 1.6855365558890223e-07, + "loss": 0.0002, + "step": 57605 + }, + { + "epoch": 7.55, + "learning_rate": 1.684567103451884e-07, + "loss": 0.0003, + "step": 57606 + }, + { + "epoch": 7.55, + "learning_rate": 1.6835979275216628e-07, + "loss": 0.0004, + "step": 57607 + }, + { + "epoch": 7.55, + "learning_rate": 1.6826290281011016e-07, + "loss": 0.0003, + "step": 57608 + }, + { + "epoch": 7.55, + "learning_rate": 1.6816604051929087e-07, + "loss": 0.0006, + "step": 57609 + }, + { + "epoch": 7.55, + "learning_rate": 1.6806920587998043e-07, + "loss": 0.0004, + "step": 57610 + }, + { + "epoch": 7.55, + "learning_rate": 1.6797239889245197e-07, + "loss": 0.0005, + "step": 57611 + }, + { + "epoch": 7.55, + "learning_rate": 1.6787561955697862e-07, + "loss": 0.0002, + "step": 57612 + }, + { + "epoch": 7.55, + "learning_rate": 1.6777886787383125e-07, + "loss": 0.0005, + "step": 57613 + }, + { + "epoch": 7.55, + "learning_rate": 1.6768214384328075e-07, + "loss": 0.0003, + "step": 57614 + }, + { + "epoch": 7.55, + "learning_rate": 1.6758544746560134e-07, + "loss": 0.0005, + "step": 57615 + }, + { + "epoch": 7.55, + "learning_rate": 1.6748877874106506e-07, + "loss": 0.0002, + "step": 57616 + }, + { + "epoch": 7.55, + "learning_rate": 1.6739213766994167e-07, + "loss": 0.0005, + "step": 57617 + }, + { + "epoch": 7.55, + "learning_rate": 1.672955242525043e-07, + "loss": 0.0004, + "step": 57618 + }, + { + "epoch": 7.55, + "learning_rate": 1.6719893848902492e-07, + "loss": 0.0003, + "step": 57619 + }, + { + "epoch": 7.55, + "learning_rate": 1.6710238037977555e-07, + "loss": 0.0002, + "step": 57620 + }, + { + "epoch": 7.55, + "learning_rate": 1.6700584992502712e-07, + "loss": 0.0003, + "step": 57621 + }, + { + "epoch": 7.55, + "learning_rate": 1.6690934712505048e-07, + "loss": 0.0005, + "step": 57622 + }, + { + "epoch": 7.55, + "learning_rate": 1.6681287198011763e-07, + "loss": 0.0001, + "step": 57623 + }, + { + "epoch": 7.55, + "learning_rate": 1.6671642449050063e-07, + "loss": 0.0004, + "step": 57624 + }, + { + "epoch": 7.55, + "learning_rate": 1.666200046564692e-07, + "loss": 0.0002, + "step": 57625 + }, + { + "epoch": 7.55, + "learning_rate": 1.6652361247829651e-07, + "loss": 0.0001, + "step": 57626 + }, + { + "epoch": 7.55, + "learning_rate": 1.664272479562512e-07, + "loss": 0.0004, + "step": 57627 + }, + { + "epoch": 7.55, + "learning_rate": 1.663309110906064e-07, + "loss": 0.0002, + "step": 57628 + }, + { + "epoch": 7.55, + "learning_rate": 1.6623460188163076e-07, + "loss": 0.0004, + "step": 57629 + }, + { + "epoch": 7.55, + "learning_rate": 1.6613832032959854e-07, + "loss": 0.0005, + "step": 57630 + }, + { + "epoch": 7.55, + "learning_rate": 1.6604206643477839e-07, + "loss": 0.0004, + "step": 57631 + }, + { + "epoch": 7.55, + "learning_rate": 1.659458401974401e-07, + "loss": 0.0003, + "step": 57632 + }, + { + "epoch": 7.55, + "learning_rate": 1.6584964161785565e-07, + "loss": 0.0004, + "step": 57633 + }, + { + "epoch": 7.55, + "learning_rate": 1.65753470696296e-07, + "loss": 0.0002, + "step": 57634 + }, + { + "epoch": 7.55, + "learning_rate": 1.6565732743303197e-07, + "loss": 0.0011, + "step": 57635 + }, + { + "epoch": 7.55, + "learning_rate": 1.6556121182833118e-07, + "loss": 0.0003, + "step": 57636 + }, + { + "epoch": 7.55, + "learning_rate": 1.6546512388246672e-07, + "loss": 0.0014, + "step": 57637 + }, + { + "epoch": 7.55, + "learning_rate": 1.6536906359570725e-07, + "loss": 0.0009, + "step": 57638 + }, + { + "epoch": 7.55, + "learning_rate": 1.652730309683237e-07, + "loss": 0.0004, + "step": 57639 + }, + { + "epoch": 7.55, + "learning_rate": 1.6517702600058693e-07, + "loss": 0.0003, + "step": 57640 + }, + { + "epoch": 7.55, + "learning_rate": 1.6508104869276454e-07, + "loss": 0.0011, + "step": 57641 + }, + { + "epoch": 7.55, + "learning_rate": 1.649850990451285e-07, + "loss": 0.0006, + "step": 57642 + }, + { + "epoch": 7.55, + "learning_rate": 1.648891770579486e-07, + "loss": 0.0011, + "step": 57643 + }, + { + "epoch": 7.55, + "learning_rate": 1.6479328273149353e-07, + "loss": 0.0007, + "step": 57644 + }, + { + "epoch": 7.55, + "learning_rate": 1.6469741606603417e-07, + "loss": 0.0002, + "step": 57645 + }, + { + "epoch": 7.55, + "learning_rate": 1.6460157706184032e-07, + "loss": 0.0003, + "step": 57646 + }, + { + "epoch": 7.55, + "learning_rate": 1.6450576571917954e-07, + "loss": 0.0002, + "step": 57647 + }, + { + "epoch": 7.55, + "learning_rate": 1.644099820383227e-07, + "loss": 0.0006, + "step": 57648 + }, + { + "epoch": 7.55, + "learning_rate": 1.6431422601953962e-07, + "loss": 0.0004, + "step": 57649 + }, + { + "epoch": 7.55, + "learning_rate": 1.6421849766309784e-07, + "loss": 0.0003, + "step": 57650 + }, + { + "epoch": 7.55, + "learning_rate": 1.6412279696926824e-07, + "loss": 0.0004, + "step": 57651 + }, + { + "epoch": 7.55, + "learning_rate": 1.6402712393831955e-07, + "loss": 0.0003, + "step": 57652 + }, + { + "epoch": 7.55, + "learning_rate": 1.639314785705204e-07, + "loss": 0.0004, + "step": 57653 + }, + { + "epoch": 7.55, + "learning_rate": 1.6383586086614167e-07, + "loss": 0.0002, + "step": 57654 + }, + { + "epoch": 7.55, + "learning_rate": 1.6374027082544984e-07, + "loss": 0.0004, + "step": 57655 + }, + { + "epoch": 7.55, + "learning_rate": 1.6364470844871362e-07, + "loss": 0.0006, + "step": 57656 + }, + { + "epoch": 7.55, + "learning_rate": 1.6354917373620493e-07, + "loss": 0.0005, + "step": 57657 + }, + { + "epoch": 7.55, + "learning_rate": 1.6345366668818918e-07, + "loss": 0.0005, + "step": 57658 + }, + { + "epoch": 7.55, + "learning_rate": 1.6335818730493503e-07, + "loss": 0.0004, + "step": 57659 + }, + { + "epoch": 7.55, + "learning_rate": 1.6326273558671446e-07, + "loss": 0.0012, + "step": 57660 + }, + { + "epoch": 7.55, + "learning_rate": 1.6316731153379173e-07, + "loss": 0.0007, + "step": 57661 + }, + { + "epoch": 7.55, + "learning_rate": 1.6307191514643773e-07, + "loss": 0.0004, + "step": 57662 + }, + { + "epoch": 7.55, + "learning_rate": 1.629765464249211e-07, + "loss": 0.0004, + "step": 57663 + }, + { + "epoch": 7.55, + "learning_rate": 1.6288120536950723e-07, + "loss": 0.0007, + "step": 57664 + }, + { + "epoch": 7.55, + "learning_rate": 1.6278589198046813e-07, + "loss": 0.0004, + "step": 57665 + }, + { + "epoch": 7.55, + "learning_rate": 1.6269060625806909e-07, + "loss": 0.0004, + "step": 57666 + }, + { + "epoch": 7.55, + "learning_rate": 1.6259534820257772e-07, + "loss": 0.0003, + "step": 57667 + }, + { + "epoch": 7.55, + "learning_rate": 1.625001178142638e-07, + "loss": 0.0003, + "step": 57668 + }, + { + "epoch": 7.55, + "learning_rate": 1.6240491509339485e-07, + "loss": 0.0007, + "step": 57669 + }, + { + "epoch": 7.55, + "learning_rate": 1.6230974004023847e-07, + "loss": 0.0009, + "step": 57670 + }, + { + "epoch": 7.55, + "learning_rate": 1.622145926550611e-07, + "loss": 0.0003, + "step": 57671 + }, + { + "epoch": 7.55, + "learning_rate": 1.6211947293813146e-07, + "loss": 0.0005, + "step": 57672 + }, + { + "epoch": 7.55, + "learning_rate": 1.6202438088971705e-07, + "loss": 0.0001, + "step": 57673 + }, + { + "epoch": 7.55, + "learning_rate": 1.6192931651008547e-07, + "loss": 0.0006, + "step": 57674 + }, + { + "epoch": 7.56, + "learning_rate": 1.6183427979950317e-07, + "loss": 0.0004, + "step": 57675 + }, + { + "epoch": 7.56, + "learning_rate": 1.6173927075823882e-07, + "loss": 0.0005, + "step": 57676 + }, + { + "epoch": 7.56, + "learning_rate": 1.6164428938655884e-07, + "loss": 0.0001, + "step": 57677 + }, + { + "epoch": 7.56, + "learning_rate": 1.6154933568472975e-07, + "loss": 0.0006, + "step": 57678 + }, + { + "epoch": 7.56, + "learning_rate": 1.6145440965301906e-07, + "loss": 0.0007, + "step": 57679 + }, + { + "epoch": 7.56, + "learning_rate": 1.6135951129169436e-07, + "loss": 0.0002, + "step": 57680 + }, + { + "epoch": 7.56, + "learning_rate": 1.612646406010232e-07, + "loss": 0.0003, + "step": 57681 + }, + { + "epoch": 7.56, + "learning_rate": 1.6116979758126872e-07, + "loss": 0.0003, + "step": 57682 + }, + { + "epoch": 7.56, + "learning_rate": 1.6107498223270178e-07, + "loss": 0.0007, + "step": 57683 + }, + { + "epoch": 7.56, + "learning_rate": 1.6098019455558777e-07, + "loss": 0.0004, + "step": 57684 + }, + { + "epoch": 7.56, + "learning_rate": 1.6088543455019312e-07, + "loss": 0.0004, + "step": 57685 + }, + { + "epoch": 7.56, + "learning_rate": 1.6079070221678316e-07, + "loss": 0.0005, + "step": 57686 + }, + { + "epoch": 7.56, + "learning_rate": 1.606959975556255e-07, + "loss": 0.0001, + "step": 57687 + }, + { + "epoch": 7.56, + "learning_rate": 1.6060132056698762e-07, + "loss": 0.0004, + "step": 57688 + }, + { + "epoch": 7.56, + "learning_rate": 1.6050667125113383e-07, + "loss": 0.0003, + "step": 57689 + }, + { + "epoch": 7.56, + "learning_rate": 1.6041204960833058e-07, + "loss": 0.0006, + "step": 57690 + }, + { + "epoch": 7.56, + "learning_rate": 1.6031745563884538e-07, + "loss": 0.0002, + "step": 57691 + }, + { + "epoch": 7.56, + "learning_rate": 1.602228893429425e-07, + "loss": 0.0005, + "step": 57692 + }, + { + "epoch": 7.56, + "learning_rate": 1.601283507208884e-07, + "loss": 0.0003, + "step": 57693 + }, + { + "epoch": 7.56, + "learning_rate": 1.600338397729495e-07, + "loss": 0.0006, + "step": 57694 + }, + { + "epoch": 7.56, + "learning_rate": 1.599393564993923e-07, + "loss": 0.0008, + "step": 57695 + }, + { + "epoch": 7.56, + "learning_rate": 1.5984490090048099e-07, + "loss": 0.0005, + "step": 57696 + }, + { + "epoch": 7.56, + "learning_rate": 1.5975047297648317e-07, + "loss": 0.0007, + "step": 57697 + }, + { + "epoch": 7.56, + "learning_rate": 1.596560727276608e-07, + "loss": 0.0005, + "step": 57698 + }, + { + "epoch": 7.56, + "learning_rate": 1.5956170015428375e-07, + "loss": 0.0004, + "step": 57699 + }, + { + "epoch": 7.56, + "learning_rate": 1.5946735525661506e-07, + "loss": 0.0001, + "step": 57700 + }, + { + "epoch": 7.56, + "learning_rate": 1.59373038034919e-07, + "loss": 0.0004, + "step": 57701 + }, + { + "epoch": 7.56, + "learning_rate": 1.5927874848946312e-07, + "loss": 0.0003, + "step": 57702 + }, + { + "epoch": 7.56, + "learning_rate": 1.5918448662051277e-07, + "loss": 0.0005, + "step": 57703 + }, + { + "epoch": 7.56, + "learning_rate": 1.5909025242833e-07, + "loss": 0.001, + "step": 57704 + }, + { + "epoch": 7.56, + "learning_rate": 1.589960459131834e-07, + "loss": 0.0004, + "step": 57705 + }, + { + "epoch": 7.56, + "learning_rate": 1.5890186707533505e-07, + "loss": 0.0002, + "step": 57706 + }, + { + "epoch": 7.56, + "learning_rate": 1.588077159150514e-07, + "loss": 0.0009, + "step": 57707 + }, + { + "epoch": 7.56, + "learning_rate": 1.5871359243259776e-07, + "loss": 0.0002, + "step": 57708 + }, + { + "epoch": 7.56, + "learning_rate": 1.5861949662823727e-07, + "loss": 0.0003, + "step": 57709 + }, + { + "epoch": 7.56, + "learning_rate": 1.585254285022364e-07, + "loss": 0.0006, + "step": 57710 + }, + { + "epoch": 7.56, + "learning_rate": 1.5843138805485825e-07, + "loss": 0.0003, + "step": 57711 + }, + { + "epoch": 7.56, + "learning_rate": 1.5833737528636816e-07, + "loss": 0.0012, + "step": 57712 + }, + { + "epoch": 7.56, + "learning_rate": 1.5824339019702927e-07, + "loss": 0.0004, + "step": 57713 + }, + { + "epoch": 7.56, + "learning_rate": 1.5814943278710693e-07, + "loss": 0.0004, + "step": 57714 + }, + { + "epoch": 7.56, + "learning_rate": 1.5805550305686423e-07, + "loss": 0.0006, + "step": 57715 + }, + { + "epoch": 7.56, + "learning_rate": 1.5796160100656766e-07, + "loss": 0.0004, + "step": 57716 + }, + { + "epoch": 7.56, + "learning_rate": 1.578677266364792e-07, + "loss": 0.0005, + "step": 57717 + }, + { + "epoch": 7.56, + "learning_rate": 1.5777387994686423e-07, + "loss": 0.0004, + "step": 57718 + }, + { + "epoch": 7.56, + "learning_rate": 1.5768006093798695e-07, + "loss": 0.0006, + "step": 57719 + }, + { + "epoch": 7.56, + "learning_rate": 1.5758626961010938e-07, + "loss": 0.0002, + "step": 57720 + }, + { + "epoch": 7.56, + "learning_rate": 1.5749250596349574e-07, + "loss": 0.0003, + "step": 57721 + }, + { + "epoch": 7.56, + "learning_rate": 1.573987699984103e-07, + "loss": 0.0003, + "step": 57722 + }, + { + "epoch": 7.56, + "learning_rate": 1.5730506171511728e-07, + "loss": 0.0002, + "step": 57723 + }, + { + "epoch": 7.56, + "learning_rate": 1.572113811138798e-07, + "loss": 0.0003, + "step": 57724 + }, + { + "epoch": 7.56, + "learning_rate": 1.5711772819496095e-07, + "loss": 0.0004, + "step": 57725 + }, + { + "epoch": 7.56, + "learning_rate": 1.5702410295862392e-07, + "loss": 0.0005, + "step": 57726 + }, + { + "epoch": 7.56, + "learning_rate": 1.569305054051329e-07, + "loss": 0.0003, + "step": 57727 + }, + { + "epoch": 7.56, + "learning_rate": 1.5683693553475098e-07, + "loss": 0.0002, + "step": 57728 + }, + { + "epoch": 7.56, + "learning_rate": 1.5674339334773914e-07, + "loss": 0.0005, + "step": 57729 + }, + { + "epoch": 7.56, + "learning_rate": 1.566498788443649e-07, + "loss": 0.0004, + "step": 57730 + }, + { + "epoch": 7.56, + "learning_rate": 1.5655639202488692e-07, + "loss": 0.0003, + "step": 57731 + }, + { + "epoch": 7.56, + "learning_rate": 1.564629328895706e-07, + "loss": 0.0006, + "step": 57732 + }, + { + "epoch": 7.56, + "learning_rate": 1.563695014386779e-07, + "loss": 0.0006, + "step": 57733 + }, + { + "epoch": 7.56, + "learning_rate": 1.5627609767247197e-07, + "loss": 0.0003, + "step": 57734 + }, + { + "epoch": 7.56, + "learning_rate": 1.5618272159121596e-07, + "loss": 0.0004, + "step": 57735 + }, + { + "epoch": 7.56, + "learning_rate": 1.560893731951696e-07, + "loss": 0.0007, + "step": 57736 + }, + { + "epoch": 7.56, + "learning_rate": 1.5599605248459937e-07, + "loss": 0.0003, + "step": 57737 + }, + { + "epoch": 7.56, + "learning_rate": 1.5590275945976508e-07, + "loss": 0.0002, + "step": 57738 + }, + { + "epoch": 7.56, + "learning_rate": 1.5580949412093204e-07, + "loss": 0.0003, + "step": 57739 + }, + { + "epoch": 7.56, + "learning_rate": 1.5571625646835786e-07, + "loss": 0.0018, + "step": 57740 + }, + { + "epoch": 7.56, + "learning_rate": 1.5562304650230899e-07, + "loss": 0.0003, + "step": 57741 + }, + { + "epoch": 7.56, + "learning_rate": 1.5552986422304627e-07, + "loss": 0.0007, + "step": 57742 + }, + { + "epoch": 7.56, + "learning_rate": 1.554367096308318e-07, + "loss": 0.0003, + "step": 57743 + }, + { + "epoch": 7.56, + "learning_rate": 1.553435827259264e-07, + "loss": 0.0003, + "step": 57744 + }, + { + "epoch": 7.56, + "learning_rate": 1.5525048350859328e-07, + "loss": 0.0002, + "step": 57745 + }, + { + "epoch": 7.56, + "learning_rate": 1.551574119790944e-07, + "loss": 0.0005, + "step": 57746 + }, + { + "epoch": 7.56, + "learning_rate": 1.5506436813768955e-07, + "loss": 0.0003, + "step": 57747 + }, + { + "epoch": 7.56, + "learning_rate": 1.5497135198464297e-07, + "loss": 0.0001, + "step": 57748 + }, + { + "epoch": 7.56, + "learning_rate": 1.5487836352021446e-07, + "loss": 0.0003, + "step": 57749 + }, + { + "epoch": 7.56, + "learning_rate": 1.5478540274466713e-07, + "loss": 0.0002, + "step": 57750 + }, + { + "epoch": 7.56, + "learning_rate": 1.546924696582608e-07, + "loss": 0.0004, + "step": 57751 + }, + { + "epoch": 7.57, + "learning_rate": 1.5459956426125855e-07, + "loss": 0.0005, + "step": 57752 + }, + { + "epoch": 7.57, + "learning_rate": 1.545066865539202e-07, + "loss": 0.0002, + "step": 57753 + }, + { + "epoch": 7.57, + "learning_rate": 1.5441383653650777e-07, + "loss": 0.0005, + "step": 57754 + }, + { + "epoch": 7.57, + "learning_rate": 1.5432101420928214e-07, + "loss": 0.0001, + "step": 57755 + }, + { + "epoch": 7.57, + "learning_rate": 1.5422821957250422e-07, + "loss": 0.0002, + "step": 57756 + }, + { + "epoch": 7.57, + "learning_rate": 1.541354526264349e-07, + "loss": 0.0007, + "step": 57757 + }, + { + "epoch": 7.57, + "learning_rate": 1.5404271337133624e-07, + "loss": 0.0006, + "step": 57758 + }, + { + "epoch": 7.57, + "learning_rate": 1.5395000180746576e-07, + "loss": 0.0009, + "step": 57759 + }, + { + "epoch": 7.57, + "learning_rate": 1.5385731793508884e-07, + "loss": 0.0008, + "step": 57760 + }, + { + "epoch": 7.57, + "learning_rate": 1.5376466175446303e-07, + "loss": 0.0002, + "step": 57761 + }, + { + "epoch": 7.57, + "learning_rate": 1.5367203326585033e-07, + "loss": 0.0021, + "step": 57762 + }, + { + "epoch": 7.57, + "learning_rate": 1.5357943246950947e-07, + "loss": 0.0014, + "step": 57763 + }, + { + "epoch": 7.57, + "learning_rate": 1.534868593657035e-07, + "loss": 0.0002, + "step": 57764 + }, + { + "epoch": 7.57, + "learning_rate": 1.5339431395469007e-07, + "loss": 0.0007, + "step": 57765 + }, + { + "epoch": 7.57, + "learning_rate": 1.5330179623673113e-07, + "loss": 0.0005, + "step": 57766 + }, + { + "epoch": 7.57, + "learning_rate": 1.5320930621208763e-07, + "loss": 0.0005, + "step": 57767 + }, + { + "epoch": 7.57, + "learning_rate": 1.531168438810182e-07, + "loss": 0.0005, + "step": 57768 + }, + { + "epoch": 7.57, + "learning_rate": 1.530244092437827e-07, + "loss": 0.0001, + "step": 57769 + }, + { + "epoch": 7.57, + "learning_rate": 1.5293200230064086e-07, + "loss": 0.0003, + "step": 57770 + }, + { + "epoch": 7.57, + "learning_rate": 1.5283962305185474e-07, + "loss": 0.0002, + "step": 57771 + }, + { + "epoch": 7.57, + "learning_rate": 1.5274727149768187e-07, + "loss": 0.0004, + "step": 57772 + }, + { + "epoch": 7.57, + "learning_rate": 1.5265494763838428e-07, + "loss": 0.0003, + "step": 57773 + }, + { + "epoch": 7.57, + "learning_rate": 1.5256265147421846e-07, + "loss": 0.0003, + "step": 57774 + }, + { + "epoch": 7.57, + "learning_rate": 1.5247038300544748e-07, + "loss": 0.0003, + "step": 57775 + }, + { + "epoch": 7.57, + "learning_rate": 1.5237814223232783e-07, + "loss": 0.0006, + "step": 57776 + }, + { + "epoch": 7.57, + "learning_rate": 1.5228592915512154e-07, + "loss": 0.0007, + "step": 57777 + }, + { + "epoch": 7.57, + "learning_rate": 1.5219374377408503e-07, + "loss": 0.0011, + "step": 57778 + }, + { + "epoch": 7.57, + "learning_rate": 1.5210158608948032e-07, + "loss": 0.0002, + "step": 57779 + }, + { + "epoch": 7.57, + "learning_rate": 1.5200945610156614e-07, + "loss": 0.0004, + "step": 57780 + }, + { + "epoch": 7.57, + "learning_rate": 1.5191735381059892e-07, + "loss": 0.0003, + "step": 57781 + }, + { + "epoch": 7.57, + "learning_rate": 1.5182527921684065e-07, + "loss": 0.0004, + "step": 57782 + }, + { + "epoch": 7.57, + "learning_rate": 1.5173323232054892e-07, + "loss": 0.0008, + "step": 57783 + }, + { + "epoch": 7.57, + "learning_rate": 1.5164121312198466e-07, + "loss": 0.0004, + "step": 57784 + }, + { + "epoch": 7.57, + "learning_rate": 1.5154922162140318e-07, + "loss": 0.0002, + "step": 57785 + }, + { + "epoch": 7.57, + "learning_rate": 1.5145725781906539e-07, + "loss": 0.0006, + "step": 57786 + }, + { + "epoch": 7.57, + "learning_rate": 1.513653217152311e-07, + "loss": 0.0006, + "step": 57787 + }, + { + "epoch": 7.57, + "learning_rate": 1.5127341331015567e-07, + "loss": 0.0003, + "step": 57788 + }, + { + "epoch": 7.57, + "learning_rate": 1.5118153260409996e-07, + "loss": 0.0005, + "step": 57789 + }, + { + "epoch": 7.57, + "learning_rate": 1.5108967959732156e-07, + "loss": 0.0005, + "step": 57790 + }, + { + "epoch": 7.57, + "learning_rate": 1.5099785429007918e-07, + "loss": 0.0002, + "step": 57791 + }, + { + "epoch": 7.57, + "learning_rate": 1.5090605668263037e-07, + "loss": 0.0006, + "step": 57792 + }, + { + "epoch": 7.57, + "learning_rate": 1.508142867752349e-07, + "loss": 0.0002, + "step": 57793 + }, + { + "epoch": 7.57, + "learning_rate": 1.5072254456814818e-07, + "loss": 0.0004, + "step": 57794 + }, + { + "epoch": 7.57, + "learning_rate": 1.5063083006163103e-07, + "loss": 0.0006, + "step": 57795 + }, + { + "epoch": 7.57, + "learning_rate": 1.5053914325594e-07, + "loss": 0.0004, + "step": 57796 + }, + { + "epoch": 7.57, + "learning_rate": 1.5044748415133147e-07, + "loss": 0.0005, + "step": 57797 + }, + { + "epoch": 7.57, + "learning_rate": 1.503558527480664e-07, + "loss": 0.0003, + "step": 57798 + }, + { + "epoch": 7.57, + "learning_rate": 1.5026424904640124e-07, + "loss": 0.0003, + "step": 57799 + }, + { + "epoch": 7.57, + "learning_rate": 1.5017267304659354e-07, + "loss": 0.0006, + "step": 57800 + }, + { + "epoch": 7.57, + "learning_rate": 1.500811247488998e-07, + "loss": 0.0009, + "step": 57801 + }, + { + "epoch": 7.57, + "learning_rate": 1.4998960415357867e-07, + "loss": 0.0004, + "step": 57802 + }, + { + "epoch": 7.57, + "learning_rate": 1.4989811126088661e-07, + "loss": 0.0008, + "step": 57803 + }, + { + "epoch": 7.57, + "learning_rate": 1.4980664607108232e-07, + "loss": 0.0005, + "step": 57804 + }, + { + "epoch": 7.57, + "learning_rate": 1.4971520858442113e-07, + "loss": 0.0002, + "step": 57805 + }, + { + "epoch": 7.57, + "learning_rate": 1.4962379880116286e-07, + "loss": 0.0004, + "step": 57806 + }, + { + "epoch": 7.57, + "learning_rate": 1.4953241672156282e-07, + "loss": 0.0003, + "step": 57807 + }, + { + "epoch": 7.57, + "learning_rate": 1.4944106234587863e-07, + "loss": 0.0002, + "step": 57808 + }, + { + "epoch": 7.57, + "learning_rate": 1.4934973567436561e-07, + "loss": 0.0002, + "step": 57809 + }, + { + "epoch": 7.57, + "learning_rate": 1.4925843670728246e-07, + "loss": 0.0005, + "step": 57810 + }, + { + "epoch": 7.57, + "learning_rate": 1.4916716544488562e-07, + "loss": 0.0004, + "step": 57811 + }, + { + "epoch": 7.57, + "learning_rate": 1.4907592188743047e-07, + "loss": 0.0005, + "step": 57812 + }, + { + "epoch": 7.57, + "learning_rate": 1.4898470603517566e-07, + "loss": 0.0003, + "step": 57813 + }, + { + "epoch": 7.57, + "learning_rate": 1.4889351788837658e-07, + "loss": 0.0003, + "step": 57814 + }, + { + "epoch": 7.57, + "learning_rate": 1.4880235744729077e-07, + "loss": 0.0004, + "step": 57815 + }, + { + "epoch": 7.57, + "learning_rate": 1.4871122471217247e-07, + "loss": 0.0008, + "step": 57816 + }, + { + "epoch": 7.57, + "learning_rate": 1.4862011968327927e-07, + "loss": 0.0011, + "step": 57817 + }, + { + "epoch": 7.57, + "learning_rate": 1.4852904236086874e-07, + "loss": 0.0002, + "step": 57818 + }, + { + "epoch": 7.57, + "learning_rate": 1.4843799274519398e-07, + "loss": 0.0011, + "step": 57819 + }, + { + "epoch": 7.57, + "learning_rate": 1.4834697083651372e-07, + "loss": 0.0004, + "step": 57820 + }, + { + "epoch": 7.57, + "learning_rate": 1.482559766350822e-07, + "loss": 0.0004, + "step": 57821 + }, + { + "epoch": 7.57, + "learning_rate": 1.4816501014115692e-07, + "loss": 0.0003, + "step": 57822 + }, + { + "epoch": 7.57, + "learning_rate": 1.480740713549933e-07, + "loss": 0.0004, + "step": 57823 + }, + { + "epoch": 7.57, + "learning_rate": 1.4798316027684555e-07, + "loss": 0.0007, + "step": 57824 + }, + { + "epoch": 7.57, + "learning_rate": 1.4789227690697017e-07, + "loss": 0.0004, + "step": 57825 + }, + { + "epoch": 7.57, + "learning_rate": 1.4780142124562468e-07, + "loss": 0.0005, + "step": 57826 + }, + { + "epoch": 7.57, + "learning_rate": 1.4771059329306225e-07, + "loss": 0.0004, + "step": 57827 + }, + { + "epoch": 7.58, + "learning_rate": 1.4761979304953823e-07, + "loss": 0.0003, + "step": 57828 + }, + { + "epoch": 7.58, + "learning_rate": 1.4752902051531014e-07, + "loss": 0.0002, + "step": 57829 + }, + { + "epoch": 7.58, + "learning_rate": 1.474382756906323e-07, + "loss": 0.0006, + "step": 57830 + }, + { + "epoch": 7.58, + "learning_rate": 1.4734755857575888e-07, + "loss": 0.0009, + "step": 57831 + }, + { + "epoch": 7.58, + "learning_rate": 1.4725686917094529e-07, + "loss": 0.0005, + "step": 57832 + }, + { + "epoch": 7.58, + "learning_rate": 1.4716620747644795e-07, + "loss": 0.0002, + "step": 57833 + }, + { + "epoch": 7.58, + "learning_rate": 1.470755734925211e-07, + "loss": 0.0007, + "step": 57834 + }, + { + "epoch": 7.58, + "learning_rate": 1.4698496721941902e-07, + "loss": 0.0003, + "step": 57835 + }, + { + "epoch": 7.58, + "learning_rate": 1.4689438865739593e-07, + "loss": 0.0002, + "step": 57836 + }, + { + "epoch": 7.58, + "learning_rate": 1.4680383780670937e-07, + "loss": 0.0004, + "step": 57837 + }, + { + "epoch": 7.58, + "learning_rate": 1.467133146676114e-07, + "loss": 0.0004, + "step": 57838 + }, + { + "epoch": 7.58, + "learning_rate": 1.4662281924035739e-07, + "loss": 0.0003, + "step": 57839 + }, + { + "epoch": 7.58, + "learning_rate": 1.4653235152520263e-07, + "loss": 0.0004, + "step": 57840 + }, + { + "epoch": 7.58, + "learning_rate": 1.4644191152240138e-07, + "loss": 0.0011, + "step": 57841 + }, + { + "epoch": 7.58, + "learning_rate": 1.4635149923220682e-07, + "loss": 0.0009, + "step": 57842 + }, + { + "epoch": 7.58, + "learning_rate": 1.4626111465487313e-07, + "loss": 0.0006, + "step": 57843 + }, + { + "epoch": 7.58, + "learning_rate": 1.461707577906568e-07, + "loss": 0.0003, + "step": 57844 + }, + { + "epoch": 7.58, + "learning_rate": 1.4608042863980987e-07, + "loss": 0.0004, + "step": 57845 + }, + { + "epoch": 7.58, + "learning_rate": 1.4599012720258655e-07, + "loss": 0.0002, + "step": 57846 + }, + { + "epoch": 7.58, + "learning_rate": 1.4589985347924108e-07, + "loss": 0.0002, + "step": 57847 + }, + { + "epoch": 7.58, + "learning_rate": 1.458096074700288e-07, + "loss": 0.0005, + "step": 57848 + }, + { + "epoch": 7.58, + "learning_rate": 1.457193891752018e-07, + "loss": 0.001, + "step": 57849 + }, + { + "epoch": 7.58, + "learning_rate": 1.4562919859501313e-07, + "loss": 0.0003, + "step": 57850 + }, + { + "epoch": 7.58, + "learning_rate": 1.4553903572971818e-07, + "loss": 0.0013, + "step": 57851 + }, + { + "epoch": 7.58, + "learning_rate": 1.4544890057957007e-07, + "loss": 0.0004, + "step": 57852 + }, + { + "epoch": 7.58, + "learning_rate": 1.4535879314482303e-07, + "loss": 0.0005, + "step": 57853 + }, + { + "epoch": 7.58, + "learning_rate": 1.4526871342572802e-07, + "loss": 0.0005, + "step": 57854 + }, + { + "epoch": 7.58, + "learning_rate": 1.4517866142254034e-07, + "loss": 0.0003, + "step": 57855 + }, + { + "epoch": 7.58, + "learning_rate": 1.4508863713551425e-07, + "loss": 0.0004, + "step": 57856 + }, + { + "epoch": 7.58, + "learning_rate": 1.4499864056489954e-07, + "loss": 0.0002, + "step": 57857 + }, + { + "epoch": 7.58, + "learning_rate": 1.449086717109527e-07, + "loss": 0.0003, + "step": 57858 + }, + { + "epoch": 7.58, + "learning_rate": 1.4481873057392459e-07, + "loss": 0.0004, + "step": 57859 + }, + { + "epoch": 7.58, + "learning_rate": 1.447288171540695e-07, + "loss": 0.0006, + "step": 57860 + }, + { + "epoch": 7.58, + "learning_rate": 1.4463893145163944e-07, + "loss": 0.0007, + "step": 57861 + }, + { + "epoch": 7.58, + "learning_rate": 1.445490734668875e-07, + "loss": 0.0006, + "step": 57862 + }, + { + "epoch": 7.58, + "learning_rate": 1.4445924320006687e-07, + "loss": 0.0005, + "step": 57863 + }, + { + "epoch": 7.58, + "learning_rate": 1.4436944065143067e-07, + "loss": 0.0003, + "step": 57864 + }, + { + "epoch": 7.58, + "learning_rate": 1.4427966582122978e-07, + "loss": 0.0002, + "step": 57865 + }, + { + "epoch": 7.58, + "learning_rate": 1.4418991870971622e-07, + "loss": 0.0005, + "step": 57866 + }, + { + "epoch": 7.58, + "learning_rate": 1.4410019931714537e-07, + "loss": 0.0002, + "step": 57867 + }, + { + "epoch": 7.58, + "learning_rate": 1.440105076437659e-07, + "loss": 0.0003, + "step": 57868 + }, + { + "epoch": 7.58, + "learning_rate": 1.4392084368983427e-07, + "loss": 0.0002, + "step": 57869 + }, + { + "epoch": 7.58, + "learning_rate": 1.438312074555992e-07, + "loss": 0.0004, + "step": 57870 + }, + { + "epoch": 7.58, + "learning_rate": 1.437415989413138e-07, + "loss": 0.0003, + "step": 57871 + }, + { + "epoch": 7.58, + "learning_rate": 1.4365201814723117e-07, + "loss": 0.0005, + "step": 57872 + }, + { + "epoch": 7.58, + "learning_rate": 1.4356246507360228e-07, + "loss": 0.0011, + "step": 57873 + }, + { + "epoch": 7.58, + "learning_rate": 1.43472939720678e-07, + "loss": 0.0003, + "step": 57874 + }, + { + "epoch": 7.58, + "learning_rate": 1.433834420887126e-07, + "loss": 0.0004, + "step": 57875 + }, + { + "epoch": 7.58, + "learning_rate": 1.4329397217795582e-07, + "loss": 0.0007, + "step": 57876 + }, + { + "epoch": 7.58, + "learning_rate": 1.4320452998865863e-07, + "loss": 0.0002, + "step": 57877 + }, + { + "epoch": 7.58, + "learning_rate": 1.4311511552107526e-07, + "loss": 0.0007, + "step": 57878 + }, + { + "epoch": 7.58, + "learning_rate": 1.430257287754544e-07, + "loss": 0.0009, + "step": 57879 + }, + { + "epoch": 7.58, + "learning_rate": 1.4293636975205026e-07, + "loss": 0.0003, + "step": 57880 + }, + { + "epoch": 7.58, + "learning_rate": 1.4284703845111158e-07, + "loss": 0.0005, + "step": 57881 + }, + { + "epoch": 7.58, + "learning_rate": 1.4275773487289146e-07, + "loss": 0.0004, + "step": 57882 + }, + { + "epoch": 7.58, + "learning_rate": 1.4266845901763971e-07, + "loss": 0.0003, + "step": 57883 + }, + { + "epoch": 7.58, + "learning_rate": 1.4257921088560833e-07, + "loss": 0.0002, + "step": 57884 + }, + { + "epoch": 7.58, + "learning_rate": 1.4248999047704826e-07, + "loss": 0.0005, + "step": 57885 + }, + { + "epoch": 7.58, + "learning_rate": 1.4240079779220928e-07, + "loss": 0.0006, + "step": 57886 + }, + { + "epoch": 7.58, + "learning_rate": 1.4231163283134452e-07, + "loss": 0.0006, + "step": 57887 + }, + { + "epoch": 7.58, + "learning_rate": 1.422224955947027e-07, + "loss": 0.0007, + "step": 57888 + }, + { + "epoch": 7.58, + "learning_rate": 1.4213338608253358e-07, + "loss": 0.0002, + "step": 57889 + }, + { + "epoch": 7.58, + "learning_rate": 1.420443042950903e-07, + "loss": 0.0008, + "step": 57890 + }, + { + "epoch": 7.58, + "learning_rate": 1.419552502326238e-07, + "loss": 0.0007, + "step": 57891 + }, + { + "epoch": 7.58, + "learning_rate": 1.4186622389538164e-07, + "loss": 0.0004, + "step": 57892 + }, + { + "epoch": 7.58, + "learning_rate": 1.417772252836158e-07, + "loss": 0.0004, + "step": 57893 + }, + { + "epoch": 7.58, + "learning_rate": 1.4168825439757727e-07, + "loss": 0.0005, + "step": 57894 + }, + { + "epoch": 7.58, + "learning_rate": 1.415993112375158e-07, + "loss": 0.0007, + "step": 57895 + }, + { + "epoch": 7.58, + "learning_rate": 1.4151039580368009e-07, + "loss": 0.0004, + "step": 57896 + }, + { + "epoch": 7.58, + "learning_rate": 1.4142150809632215e-07, + "loss": 0.001, + "step": 57897 + }, + { + "epoch": 7.58, + "learning_rate": 1.4133264811569182e-07, + "loss": 0.0003, + "step": 57898 + }, + { + "epoch": 7.58, + "learning_rate": 1.4124381586203772e-07, + "loss": 0.0003, + "step": 57899 + }, + { + "epoch": 7.58, + "learning_rate": 1.411550113356097e-07, + "loss": 0.0006, + "step": 57900 + }, + { + "epoch": 7.58, + "learning_rate": 1.410662345366587e-07, + "loss": 0.0002, + "step": 57901 + }, + { + "epoch": 7.58, + "learning_rate": 1.4097748546543443e-07, + "loss": 0.0001, + "step": 57902 + }, + { + "epoch": 7.58, + "learning_rate": 1.4088876412218567e-07, + "loss": 0.0007, + "step": 57903 + }, + { + "epoch": 7.59, + "learning_rate": 1.4080007050716215e-07, + "loss": 0.0002, + "step": 57904 + }, + { + "epoch": 7.59, + "learning_rate": 1.407114046206137e-07, + "loss": 0.0003, + "step": 57905 + }, + { + "epoch": 7.59, + "learning_rate": 1.4062276646278905e-07, + "loss": 0.0006, + "step": 57906 + }, + { + "epoch": 7.59, + "learning_rate": 1.4053415603393795e-07, + "loss": 0.0004, + "step": 57907 + }, + { + "epoch": 7.59, + "learning_rate": 1.4044557333430908e-07, + "loss": 0.0007, + "step": 57908 + }, + { + "epoch": 7.59, + "learning_rate": 1.4035701836415228e-07, + "loss": 0.0002, + "step": 57909 + }, + { + "epoch": 7.59, + "learning_rate": 1.4026849112371732e-07, + "loss": 0.001, + "step": 57910 + }, + { + "epoch": 7.59, + "learning_rate": 1.4017999161325068e-07, + "loss": 0.0003, + "step": 57911 + }, + { + "epoch": 7.59, + "learning_rate": 1.4009151983300217e-07, + "loss": 0.0002, + "step": 57912 + }, + { + "epoch": 7.59, + "learning_rate": 1.400030757832227e-07, + "loss": 0.0008, + "step": 57913 + }, + { + "epoch": 7.59, + "learning_rate": 1.3991465946415984e-07, + "loss": 0.0007, + "step": 57914 + }, + { + "epoch": 7.59, + "learning_rate": 1.3982627087606116e-07, + "loss": 0.0004, + "step": 57915 + }, + { + "epoch": 7.59, + "learning_rate": 1.3973791001917536e-07, + "loss": 0.0012, + "step": 57916 + }, + { + "epoch": 7.59, + "learning_rate": 1.3964957689375225e-07, + "loss": 0.0005, + "step": 57917 + }, + { + "epoch": 7.59, + "learning_rate": 1.3956127150004052e-07, + "loss": 0.0005, + "step": 57918 + }, + { + "epoch": 7.59, + "learning_rate": 1.394729938382855e-07, + "loss": 0.0003, + "step": 57919 + }, + { + "epoch": 7.59, + "learning_rate": 1.3938474390873924e-07, + "loss": 0.0004, + "step": 57920 + }, + { + "epoch": 7.59, + "learning_rate": 1.392965217116482e-07, + "loss": 0.0005, + "step": 57921 + }, + { + "epoch": 7.59, + "learning_rate": 1.3920832724725996e-07, + "loss": 0.0003, + "step": 57922 + }, + { + "epoch": 7.59, + "learning_rate": 1.3912016051582323e-07, + "loss": 0.0006, + "step": 57923 + }, + { + "epoch": 7.59, + "learning_rate": 1.3903202151758554e-07, + "loss": 0.0003, + "step": 57924 + }, + { + "epoch": 7.59, + "learning_rate": 1.3894391025279674e-07, + "loss": 0.0004, + "step": 57925 + }, + { + "epoch": 7.59, + "learning_rate": 1.3885582672170218e-07, + "loss": 0.0006, + "step": 57926 + }, + { + "epoch": 7.59, + "learning_rate": 1.387677709245494e-07, + "loss": 0.0009, + "step": 57927 + }, + { + "epoch": 7.59, + "learning_rate": 1.3867974286158824e-07, + "loss": 0.0005, + "step": 57928 + }, + { + "epoch": 7.59, + "learning_rate": 1.3859174253306517e-07, + "loss": 0.0003, + "step": 57929 + }, + { + "epoch": 7.59, + "learning_rate": 1.3850376993922776e-07, + "loss": 0.0002, + "step": 57930 + }, + { + "epoch": 7.59, + "learning_rate": 1.3841582508032248e-07, + "loss": 0.0007, + "step": 57931 + }, + { + "epoch": 7.59, + "learning_rate": 1.3832790795659912e-07, + "loss": 0.0004, + "step": 57932 + }, + { + "epoch": 7.59, + "learning_rate": 1.3824001856830194e-07, + "loss": 0.0005, + "step": 57933 + }, + { + "epoch": 7.59, + "learning_rate": 1.3815215691567964e-07, + "loss": 0.0009, + "step": 57934 + }, + { + "epoch": 7.59, + "learning_rate": 1.3806432299897977e-07, + "loss": 0.0003, + "step": 57935 + }, + { + "epoch": 7.59, + "learning_rate": 1.3797651681844882e-07, + "loss": 0.0004, + "step": 57936 + }, + { + "epoch": 7.59, + "learning_rate": 1.3788873837433324e-07, + "loss": 0.0005, + "step": 57937 + }, + { + "epoch": 7.59, + "learning_rate": 1.3780098766688067e-07, + "loss": 0.0001, + "step": 57938 + }, + { + "epoch": 7.59, + "learning_rate": 1.377132646963375e-07, + "loss": 0.0005, + "step": 57939 + }, + { + "epoch": 7.59, + "learning_rate": 1.3762556946295135e-07, + "loss": 0.0007, + "step": 57940 + }, + { + "epoch": 7.59, + "learning_rate": 1.3753790196696758e-07, + "loss": 0.0005, + "step": 57941 + }, + { + "epoch": 7.59, + "learning_rate": 1.3745026220863266e-07, + "loss": 0.0004, + "step": 57942 + }, + { + "epoch": 7.59, + "learning_rate": 1.3736265018819416e-07, + "loss": 0.0008, + "step": 57943 + }, + { + "epoch": 7.59, + "learning_rate": 1.3727506590589746e-07, + "loss": 0.0004, + "step": 57944 + }, + { + "epoch": 7.59, + "learning_rate": 1.371875093619901e-07, + "loss": 0.0004, + "step": 57945 + }, + { + "epoch": 7.59, + "learning_rate": 1.3709998055671636e-07, + "loss": 0.0002, + "step": 57946 + }, + { + "epoch": 7.59, + "learning_rate": 1.3701247949032493e-07, + "loss": 0.0009, + "step": 57947 + }, + { + "epoch": 7.59, + "learning_rate": 1.3692500616306116e-07, + "loss": 0.0004, + "step": 57948 + }, + { + "epoch": 7.59, + "learning_rate": 1.3683756057517038e-07, + "loss": 0.0007, + "step": 57949 + }, + { + "epoch": 7.59, + "learning_rate": 1.3675014272689802e-07, + "loss": 0.0005, + "step": 57950 + }, + { + "epoch": 7.59, + "learning_rate": 1.366627526184905e-07, + "loss": 0.0005, + "step": 57951 + }, + { + "epoch": 7.59, + "learning_rate": 1.365753902501943e-07, + "loss": 0.0004, + "step": 57952 + }, + { + "epoch": 7.59, + "learning_rate": 1.3648805562225477e-07, + "loss": 0.0002, + "step": 57953 + }, + { + "epoch": 7.59, + "learning_rate": 1.3640074873491616e-07, + "loss": 0.0005, + "step": 57954 + }, + { + "epoch": 7.59, + "learning_rate": 1.3631346958842716e-07, + "loss": 0.0004, + "step": 57955 + }, + { + "epoch": 7.59, + "learning_rate": 1.3622621818302872e-07, + "loss": 0.0003, + "step": 57956 + }, + { + "epoch": 7.59, + "learning_rate": 1.361389945189706e-07, + "loss": 0.0009, + "step": 57957 + }, + { + "epoch": 7.59, + "learning_rate": 1.3605179859649486e-07, + "loss": 0.0005, + "step": 57958 + }, + { + "epoch": 7.59, + "learning_rate": 1.3596463041585018e-07, + "loss": 0.0003, + "step": 57959 + }, + { + "epoch": 7.59, + "learning_rate": 1.3587748997727858e-07, + "loss": 0.0004, + "step": 57960 + }, + { + "epoch": 7.59, + "learning_rate": 1.357903772810254e-07, + "loss": 0.0003, + "step": 57961 + }, + { + "epoch": 7.59, + "learning_rate": 1.3570329232733713e-07, + "loss": 0.0007, + "step": 57962 + }, + { + "epoch": 7.59, + "learning_rate": 1.3561623511645916e-07, + "loss": 0.0009, + "step": 57963 + }, + { + "epoch": 7.59, + "learning_rate": 1.355292056486346e-07, + "loss": 0.0006, + "step": 57964 + }, + { + "epoch": 7.59, + "learning_rate": 1.3544220392410768e-07, + "loss": 0.0002, + "step": 57965 + }, + { + "epoch": 7.59, + "learning_rate": 1.3535522994312488e-07, + "loss": 0.0002, + "step": 57966 + }, + { + "epoch": 7.59, + "learning_rate": 1.3526828370593048e-07, + "loss": 0.0006, + "step": 57967 + }, + { + "epoch": 7.59, + "learning_rate": 1.351813652127687e-07, + "loss": 0.0002, + "step": 57968 + }, + { + "epoch": 7.59, + "learning_rate": 1.350944744638838e-07, + "loss": 0.0004, + "step": 57969 + }, + { + "epoch": 7.59, + "learning_rate": 1.3500761145952e-07, + "loss": 0.0004, + "step": 57970 + }, + { + "epoch": 7.59, + "learning_rate": 1.3492077619992273e-07, + "loss": 0.0003, + "step": 57971 + }, + { + "epoch": 7.59, + "learning_rate": 1.3483396868533504e-07, + "loss": 0.0002, + "step": 57972 + }, + { + "epoch": 7.59, + "learning_rate": 1.3474718891600126e-07, + "loss": 0.0006, + "step": 57973 + }, + { + "epoch": 7.59, + "learning_rate": 1.3466043689216668e-07, + "loss": 0.0004, + "step": 57974 + }, + { + "epoch": 7.59, + "learning_rate": 1.3457371261407338e-07, + "loss": 0.0004, + "step": 57975 + }, + { + "epoch": 7.59, + "learning_rate": 1.3448701608196556e-07, + "loss": 0.0006, + "step": 57976 + }, + { + "epoch": 7.59, + "learning_rate": 1.3440034729608865e-07, + "loss": 0.0003, + "step": 57977 + }, + { + "epoch": 7.59, + "learning_rate": 1.343137062566846e-07, + "loss": 0.0008, + "step": 57978 + }, + { + "epoch": 7.59, + "learning_rate": 1.3422709296399882e-07, + "loss": 0.0009, + "step": 57979 + }, + { + "epoch": 7.59, + "learning_rate": 1.3414050741827333e-07, + "loss": 0.0002, + "step": 57980 + }, + { + "epoch": 7.6, + "learning_rate": 1.3405394961975237e-07, + "loss": 0.0008, + "step": 57981 + }, + { + "epoch": 7.6, + "learning_rate": 1.3396741956867908e-07, + "loss": 0.0004, + "step": 57982 + }, + { + "epoch": 7.6, + "learning_rate": 1.3388091726529773e-07, + "loss": 0.0011, + "step": 57983 + }, + { + "epoch": 7.6, + "learning_rate": 1.3379444270985033e-07, + "loss": 0.0003, + "step": 57984 + }, + { + "epoch": 7.6, + "learning_rate": 1.337079959025811e-07, + "loss": 0.0003, + "step": 57985 + }, + { + "epoch": 7.6, + "learning_rate": 1.3362157684373211e-07, + "loss": 0.0005, + "step": 57986 + }, + { + "epoch": 7.6, + "learning_rate": 1.335351855335476e-07, + "loss": 0.0008, + "step": 57987 + }, + { + "epoch": 7.6, + "learning_rate": 1.334488219722696e-07, + "loss": 0.0006, + "step": 57988 + }, + { + "epoch": 7.6, + "learning_rate": 1.3336248616014126e-07, + "loss": 0.0002, + "step": 57989 + }, + { + "epoch": 7.6, + "learning_rate": 1.3327617809740568e-07, + "loss": 0.0003, + "step": 57990 + }, + { + "epoch": 7.6, + "learning_rate": 1.331898977843049e-07, + "loss": 0.0006, + "step": 57991 + }, + { + "epoch": 7.6, + "learning_rate": 1.331036452210821e-07, + "loss": 0.0002, + "step": 57992 + }, + { + "epoch": 7.6, + "learning_rate": 1.3301742040798149e-07, + "loss": 0.0004, + "step": 57993 + }, + { + "epoch": 7.6, + "learning_rate": 1.3293122334524288e-07, + "loss": 0.0006, + "step": 57994 + }, + { + "epoch": 7.6, + "learning_rate": 1.3284505403310944e-07, + "loss": 0.0004, + "step": 57995 + }, + { + "epoch": 7.6, + "learning_rate": 1.3275891247182427e-07, + "loss": 0.0009, + "step": 57996 + }, + { + "epoch": 7.6, + "learning_rate": 1.326727986616283e-07, + "loss": 0.0003, + "step": 57997 + }, + { + "epoch": 7.6, + "learning_rate": 1.3258671260276578e-07, + "loss": 0.001, + "step": 57998 + }, + { + "epoch": 7.6, + "learning_rate": 1.3250065429547653e-07, + "loss": 0.0006, + "step": 57999 + }, + { + "epoch": 7.6, + "learning_rate": 1.324146237400048e-07, + "loss": 0.0005, + "step": 58000 + }, + { + "epoch": 7.6, + "learning_rate": 1.3232862093659037e-07, + "loss": 0.0004, + "step": 58001 + }, + { + "epoch": 7.6, + "learning_rate": 1.3224264588547753e-07, + "loss": 0.0005, + "step": 58002 + }, + { + "epoch": 7.6, + "learning_rate": 1.3215669858690606e-07, + "loss": 0.0004, + "step": 58003 + }, + { + "epoch": 7.6, + "learning_rate": 1.3207077904111798e-07, + "loss": 0.001, + "step": 58004 + }, + { + "epoch": 7.6, + "learning_rate": 1.3198488724835645e-07, + "loss": 0.0007, + "step": 58005 + }, + { + "epoch": 7.6, + "learning_rate": 1.3189902320886126e-07, + "loss": 0.0002, + "step": 58006 + }, + { + "epoch": 7.6, + "learning_rate": 1.3181318692287448e-07, + "loss": 0.0004, + "step": 58007 + }, + { + "epoch": 7.6, + "learning_rate": 1.317273783906381e-07, + "loss": 0.0003, + "step": 58008 + }, + { + "epoch": 7.6, + "learning_rate": 1.3164159761239193e-07, + "loss": 0.0006, + "step": 58009 + }, + { + "epoch": 7.6, + "learning_rate": 1.315558445883791e-07, + "loss": 0.0002, + "step": 58010 + }, + { + "epoch": 7.6, + "learning_rate": 1.3147011931883835e-07, + "loss": 0.0002, + "step": 58011 + }, + { + "epoch": 7.6, + "learning_rate": 1.313844218040139e-07, + "loss": 0.0004, + "step": 58012 + }, + { + "epoch": 7.6, + "learning_rate": 1.3129875204414555e-07, + "loss": 0.0001, + "step": 58013 + }, + { + "epoch": 7.6, + "learning_rate": 1.3121311003947422e-07, + "loss": 0.0003, + "step": 58014 + }, + { + "epoch": 7.6, + "learning_rate": 1.3112749579023865e-07, + "loss": 0.0004, + "step": 58015 + }, + { + "epoch": 7.6, + "learning_rate": 1.3104190929668303e-07, + "loss": 0.0004, + "step": 58016 + }, + { + "epoch": 7.6, + "learning_rate": 1.309563505590461e-07, + "loss": 0.0014, + "step": 58017 + }, + { + "epoch": 7.6, + "learning_rate": 1.3087081957756764e-07, + "loss": 0.0016, + "step": 58018 + }, + { + "epoch": 7.6, + "learning_rate": 1.307853163524908e-07, + "loss": 0.0003, + "step": 58019 + }, + { + "epoch": 7.6, + "learning_rate": 1.306998408840543e-07, + "loss": 0.0015, + "step": 58020 + }, + { + "epoch": 7.6, + "learning_rate": 1.3061439317249903e-07, + "loss": 0.0008, + "step": 58021 + }, + { + "epoch": 7.6, + "learning_rate": 1.305289732180659e-07, + "loss": 0.0002, + "step": 58022 + }, + { + "epoch": 7.6, + "learning_rate": 1.3044358102099254e-07, + "loss": 0.0004, + "step": 58023 + }, + { + "epoch": 7.6, + "learning_rate": 1.3035821658152314e-07, + "loss": 0.0002, + "step": 58024 + }, + { + "epoch": 7.6, + "learning_rate": 1.3027287989989424e-07, + "loss": 0.0008, + "step": 58025 + }, + { + "epoch": 7.6, + "learning_rate": 1.3018757097634783e-07, + "loss": 0.0003, + "step": 58026 + }, + { + "epoch": 7.6, + "learning_rate": 1.3010228981112261e-07, + "loss": 0.0005, + "step": 58027 + }, + { + "epoch": 7.6, + "learning_rate": 1.300170364044606e-07, + "loss": 0.0005, + "step": 58028 + }, + { + "epoch": 7.6, + "learning_rate": 1.2993181075659945e-07, + "loss": 0.0005, + "step": 58029 + }, + { + "epoch": 7.6, + "learning_rate": 1.2984661286777888e-07, + "loss": 0.0002, + "step": 58030 + }, + { + "epoch": 7.6, + "learning_rate": 1.2976144273823877e-07, + "loss": 0.0002, + "step": 58031 + }, + { + "epoch": 7.6, + "learning_rate": 1.2967630036822e-07, + "loss": 0.0004, + "step": 58032 + }, + { + "epoch": 7.6, + "learning_rate": 1.2959118575796015e-07, + "loss": 0.0006, + "step": 58033 + }, + { + "epoch": 7.6, + "learning_rate": 1.2950609890770015e-07, + "loss": 0.0002, + "step": 58034 + }, + { + "epoch": 7.6, + "learning_rate": 1.2942103981767763e-07, + "loss": 0.0009, + "step": 58035 + }, + { + "epoch": 7.6, + "learning_rate": 1.2933600848813454e-07, + "loss": 0.0006, + "step": 58036 + }, + { + "epoch": 7.6, + "learning_rate": 1.2925100491930742e-07, + "loss": 0.0004, + "step": 58037 + }, + { + "epoch": 7.6, + "learning_rate": 1.2916602911143495e-07, + "loss": 0.0006, + "step": 58038 + }, + { + "epoch": 7.6, + "learning_rate": 1.2908108106475913e-07, + "loss": 0.0005, + "step": 58039 + }, + { + "epoch": 7.6, + "learning_rate": 1.2899616077951648e-07, + "loss": 0.0003, + "step": 58040 + }, + { + "epoch": 7.6, + "learning_rate": 1.289112682559468e-07, + "loss": 0.0008, + "step": 58041 + }, + { + "epoch": 7.6, + "learning_rate": 1.2882640349428766e-07, + "loss": 0.0003, + "step": 58042 + }, + { + "epoch": 7.6, + "learning_rate": 1.2874156649477888e-07, + "loss": 0.0003, + "step": 58043 + }, + { + "epoch": 7.6, + "learning_rate": 1.2865675725765914e-07, + "loss": 0.0002, + "step": 58044 + }, + { + "epoch": 7.6, + "learning_rate": 1.2857197578316715e-07, + "loss": 0.0007, + "step": 58045 + }, + { + "epoch": 7.6, + "learning_rate": 1.284872220715394e-07, + "loss": 0.0003, + "step": 58046 + }, + { + "epoch": 7.6, + "learning_rate": 1.284024961230168e-07, + "loss": 0.0002, + "step": 58047 + }, + { + "epoch": 7.6, + "learning_rate": 1.2831779793783583e-07, + "loss": 0.0005, + "step": 58048 + }, + { + "epoch": 7.6, + "learning_rate": 1.2823312751623518e-07, + "loss": 0.0003, + "step": 58049 + }, + { + "epoch": 7.6, + "learning_rate": 1.2814848485845354e-07, + "loss": 0.0003, + "step": 58050 + }, + { + "epoch": 7.6, + "learning_rate": 1.2806386996472852e-07, + "loss": 0.0003, + "step": 58051 + }, + { + "epoch": 7.6, + "learning_rate": 1.2797928283529882e-07, + "loss": 0.0002, + "step": 58052 + }, + { + "epoch": 7.6, + "learning_rate": 1.2789472347040088e-07, + "loss": 0.0015, + "step": 58053 + }, + { + "epoch": 7.6, + "learning_rate": 1.2781019187027234e-07, + "loss": 0.0002, + "step": 58054 + }, + { + "epoch": 7.6, + "learning_rate": 1.2772568803515406e-07, + "loss": 0.0004, + "step": 58055 + }, + { + "epoch": 7.6, + "learning_rate": 1.2764121196528033e-07, + "loss": 0.0008, + "step": 58056 + }, + { + "epoch": 7.61, + "learning_rate": 1.2755676366088988e-07, + "loss": 0.0001, + "step": 58057 + }, + { + "epoch": 7.61, + "learning_rate": 1.2747234312222023e-07, + "loss": 0.0001, + "step": 58058 + }, + { + "epoch": 7.61, + "learning_rate": 1.2738795034951013e-07, + "loss": 0.0002, + "step": 58059 + }, + { + "epoch": 7.61, + "learning_rate": 1.2730358534299492e-07, + "loss": 0.0004, + "step": 58060 + }, + { + "epoch": 7.61, + "learning_rate": 1.272192481029122e-07, + "loss": 0.0005, + "step": 58061 + }, + { + "epoch": 7.61, + "learning_rate": 1.2713493862949955e-07, + "loss": 0.0004, + "step": 58062 + }, + { + "epoch": 7.61, + "learning_rate": 1.2705065692299456e-07, + "loss": 0.0006, + "step": 58063 + }, + { + "epoch": 7.61, + "learning_rate": 1.2696640298363372e-07, + "loss": 0.0005, + "step": 58064 + }, + { + "epoch": 7.61, + "learning_rate": 1.2688217681165348e-07, + "loss": 0.0007, + "step": 58065 + }, + { + "epoch": 7.61, + "learning_rate": 1.2679797840729258e-07, + "loss": 0.0003, + "step": 58066 + }, + { + "epoch": 7.61, + "learning_rate": 1.2671380777078523e-07, + "loss": 0.0003, + "step": 58067 + }, + { + "epoch": 7.61, + "learning_rate": 1.2662966490237015e-07, + "loss": 0.0006, + "step": 58068 + }, + { + "epoch": 7.61, + "learning_rate": 1.2654554980228383e-07, + "loss": 0.0002, + "step": 58069 + }, + { + "epoch": 7.61, + "learning_rate": 1.2646146247076163e-07, + "loss": 0.0009, + "step": 58070 + }, + { + "epoch": 7.61, + "learning_rate": 1.2637740290804224e-07, + "loss": 0.0003, + "step": 58071 + }, + { + "epoch": 7.61, + "learning_rate": 1.262933711143588e-07, + "loss": 0.0003, + "step": 58072 + }, + { + "epoch": 7.61, + "learning_rate": 1.2620936708995003e-07, + "loss": 0.0004, + "step": 58073 + }, + { + "epoch": 7.61, + "learning_rate": 1.2612539083505237e-07, + "loss": 0.0009, + "step": 58074 + }, + { + "epoch": 7.61, + "learning_rate": 1.2604144234990012e-07, + "loss": 0.0006, + "step": 58075 + }, + { + "epoch": 7.61, + "learning_rate": 1.2595752163472973e-07, + "loss": 0.0008, + "step": 58076 + }, + { + "epoch": 7.61, + "learning_rate": 1.2587362868977994e-07, + "loss": 0.0004, + "step": 58077 + }, + { + "epoch": 7.61, + "learning_rate": 1.2578976351528382e-07, + "loss": 0.0006, + "step": 58078 + }, + { + "epoch": 7.61, + "learning_rate": 1.2570592611147904e-07, + "loss": 0.0003, + "step": 58079 + }, + { + "epoch": 7.61, + "learning_rate": 1.2562211647859866e-07, + "loss": 0.0005, + "step": 58080 + }, + { + "epoch": 7.61, + "learning_rate": 1.255383346168826e-07, + "loss": 0.0006, + "step": 58081 + }, + { + "epoch": 7.61, + "learning_rate": 1.254545805265628e-07, + "loss": 0.0004, + "step": 58082 + }, + { + "epoch": 7.61, + "learning_rate": 1.2537085420787575e-07, + "loss": 0.001, + "step": 58083 + }, + { + "epoch": 7.61, + "learning_rate": 1.2528715566105798e-07, + "loss": 0.0005, + "step": 58084 + }, + { + "epoch": 7.61, + "learning_rate": 1.2520348488634371e-07, + "loss": 0.0002, + "step": 58085 + }, + { + "epoch": 7.61, + "learning_rate": 1.251198418839694e-07, + "loss": 0.0004, + "step": 58086 + }, + { + "epoch": 7.61, + "learning_rate": 1.2503622665416938e-07, + "loss": 0.0006, + "step": 58087 + }, + { + "epoch": 7.61, + "learning_rate": 1.2495263919717893e-07, + "loss": 0.0004, + "step": 58088 + }, + { + "epoch": 7.61, + "learning_rate": 1.2486907951323346e-07, + "loss": 0.0005, + "step": 58089 + }, + { + "epoch": 7.61, + "learning_rate": 1.2478554760256835e-07, + "loss": 0.0004, + "step": 58090 + }, + { + "epoch": 7.61, + "learning_rate": 1.247020434654167e-07, + "loss": 0.0011, + "step": 58091 + }, + { + "epoch": 7.61, + "learning_rate": 1.2461856710201504e-07, + "loss": 0.0008, + "step": 58092 + }, + { + "epoch": 7.61, + "learning_rate": 1.245351185125987e-07, + "loss": 0.0002, + "step": 58093 + }, + { + "epoch": 7.61, + "learning_rate": 1.2445169769740084e-07, + "loss": 0.0002, + "step": 58094 + }, + { + "epoch": 7.61, + "learning_rate": 1.2436830465665682e-07, + "loss": 0.0003, + "step": 58095 + }, + { + "epoch": 7.61, + "learning_rate": 1.2428493939060093e-07, + "loss": 0.0006, + "step": 58096 + }, + { + "epoch": 7.61, + "learning_rate": 1.2420160189946738e-07, + "loss": 0.0004, + "step": 58097 + }, + { + "epoch": 7.61, + "learning_rate": 1.2411829218349158e-07, + "loss": 0.0006, + "step": 58098 + }, + { + "epoch": 7.61, + "learning_rate": 1.2403501024290665e-07, + "loss": 0.0004, + "step": 58099 + }, + { + "epoch": 7.61, + "learning_rate": 1.2395175607794795e-07, + "loss": 0.0005, + "step": 58100 + }, + { + "epoch": 7.61, + "learning_rate": 1.2386852968884867e-07, + "loss": 0.0009, + "step": 58101 + }, + { + "epoch": 7.61, + "learning_rate": 1.2378533107584302e-07, + "loss": 0.0005, + "step": 58102 + }, + { + "epoch": 7.61, + "learning_rate": 1.2370216023916526e-07, + "loss": 0.0005, + "step": 58103 + }, + { + "epoch": 7.61, + "learning_rate": 1.2361901717904966e-07, + "loss": 0.0004, + "step": 58104 + }, + { + "epoch": 7.61, + "learning_rate": 1.235359018957294e-07, + "loss": 0.0002, + "step": 58105 + }, + { + "epoch": 7.61, + "learning_rate": 1.2345281438943756e-07, + "loss": 0.0002, + "step": 58106 + }, + { + "epoch": 7.61, + "learning_rate": 1.2336975466040957e-07, + "loss": 0.0004, + "step": 58107 + }, + { + "epoch": 7.61, + "learning_rate": 1.2328672270887743e-07, + "loss": 0.0003, + "step": 58108 + }, + { + "epoch": 7.61, + "learning_rate": 1.2320371853507652e-07, + "loss": 0.0001, + "step": 58109 + }, + { + "epoch": 7.61, + "learning_rate": 1.2312074213923885e-07, + "loss": 0.0003, + "step": 58110 + }, + { + "epoch": 7.61, + "learning_rate": 1.230377935215976e-07, + "loss": 0.0004, + "step": 58111 + }, + { + "epoch": 7.61, + "learning_rate": 1.2295487268238704e-07, + "loss": 0.0005, + "step": 58112 + }, + { + "epoch": 7.61, + "learning_rate": 1.2287197962184028e-07, + "loss": 0.0004, + "step": 58113 + }, + { + "epoch": 7.61, + "learning_rate": 1.2278911434018936e-07, + "loss": 0.0003, + "step": 58114 + }, + { + "epoch": 7.61, + "learning_rate": 1.2270627683766856e-07, + "loss": 0.0002, + "step": 58115 + }, + { + "epoch": 7.61, + "learning_rate": 1.2262346711451102e-07, + "loss": 0.0005, + "step": 58116 + }, + { + "epoch": 7.61, + "learning_rate": 1.2254068517094764e-07, + "loss": 0.0003, + "step": 58117 + }, + { + "epoch": 7.61, + "learning_rate": 1.224579310072127e-07, + "loss": 0.0003, + "step": 58118 + }, + { + "epoch": 7.61, + "learning_rate": 1.2237520462353937e-07, + "loss": 0.0003, + "step": 58119 + }, + { + "epoch": 7.61, + "learning_rate": 1.2229250602015963e-07, + "loss": 0.0003, + "step": 58120 + }, + { + "epoch": 7.61, + "learning_rate": 1.2220983519730668e-07, + "loss": 0.0001, + "step": 58121 + }, + { + "epoch": 7.61, + "learning_rate": 1.221271921552114e-07, + "loss": 0.0004, + "step": 58122 + }, + { + "epoch": 7.61, + "learning_rate": 1.220445768941081e-07, + "loss": 0.0002, + "step": 58123 + }, + { + "epoch": 7.61, + "learning_rate": 1.2196198941422876e-07, + "loss": 0.0003, + "step": 58124 + }, + { + "epoch": 7.61, + "learning_rate": 1.2187942971580435e-07, + "loss": 0.0002, + "step": 58125 + }, + { + "epoch": 7.61, + "learning_rate": 1.21796897799068e-07, + "loss": 0.0003, + "step": 58126 + }, + { + "epoch": 7.61, + "learning_rate": 1.2171439366425287e-07, + "loss": 0.0005, + "step": 58127 + }, + { + "epoch": 7.61, + "learning_rate": 1.2163191731158986e-07, + "loss": 0.0005, + "step": 58128 + }, + { + "epoch": 7.61, + "learning_rate": 1.215494687413099e-07, + "loss": 0.0004, + "step": 58129 + }, + { + "epoch": 7.61, + "learning_rate": 1.2146704795364618e-07, + "loss": 0.0002, + "step": 58130 + }, + { + "epoch": 7.61, + "learning_rate": 1.2138465494883068e-07, + "loss": 0.0003, + "step": 58131 + }, + { + "epoch": 7.61, + "learning_rate": 1.2130228972709546e-07, + "loss": 0.0001, + "step": 58132 + }, + { + "epoch": 7.62, + "learning_rate": 1.2121995228867034e-07, + "loss": 0.001, + "step": 58133 + }, + { + "epoch": 7.62, + "learning_rate": 1.2113764263378847e-07, + "loss": 0.0002, + "step": 58134 + }, + { + "epoch": 7.62, + "learning_rate": 1.2105536076268186e-07, + "loss": 0.0003, + "step": 58135 + }, + { + "epoch": 7.62, + "learning_rate": 1.2097310667558037e-07, + "loss": 0.0006, + "step": 58136 + }, + { + "epoch": 7.62, + "learning_rate": 1.2089088037271491e-07, + "loss": 0.0005, + "step": 58137 + }, + { + "epoch": 7.62, + "learning_rate": 1.208086818543186e-07, + "loss": 0.0009, + "step": 58138 + }, + { + "epoch": 7.62, + "learning_rate": 1.2072651112062238e-07, + "loss": 0.0002, + "step": 58139 + }, + { + "epoch": 7.62, + "learning_rate": 1.206443681718561e-07, + "loss": 0.0004, + "step": 58140 + }, + { + "epoch": 7.62, + "learning_rate": 1.2056225300825064e-07, + "loss": 0.0007, + "step": 58141 + }, + { + "epoch": 7.62, + "learning_rate": 1.2048016563003806e-07, + "loss": 0.0004, + "step": 58142 + }, + { + "epoch": 7.62, + "learning_rate": 1.203981060374493e-07, + "loss": 0.0002, + "step": 58143 + }, + { + "epoch": 7.62, + "learning_rate": 1.2031607423071524e-07, + "loss": 0.0005, + "step": 58144 + }, + { + "epoch": 7.62, + "learning_rate": 1.202340702100657e-07, + "loss": 0.0004, + "step": 58145 + }, + { + "epoch": 7.62, + "learning_rate": 1.2015209397573169e-07, + "loss": 0.0012, + "step": 58146 + }, + { + "epoch": 7.62, + "learning_rate": 1.2007014552794405e-07, + "loss": 0.0003, + "step": 58147 + }, + { + "epoch": 7.62, + "learning_rate": 1.1998822486693152e-07, + "loss": 0.0007, + "step": 58148 + }, + { + "epoch": 7.62, + "learning_rate": 1.1990633199292834e-07, + "loss": 0.0004, + "step": 58149 + }, + { + "epoch": 7.62, + "learning_rate": 1.19824466906161e-07, + "loss": 0.0002, + "step": 58150 + }, + { + "epoch": 7.62, + "learning_rate": 1.1974262960686157e-07, + "loss": 0.0003, + "step": 58151 + }, + { + "epoch": 7.62, + "learning_rate": 1.1966082009525982e-07, + "loss": 0.0004, + "step": 58152 + }, + { + "epoch": 7.62, + "learning_rate": 1.1957903837158445e-07, + "loss": 0.0004, + "step": 58153 + }, + { + "epoch": 7.62, + "learning_rate": 1.1949728443606868e-07, + "loss": 0.0002, + "step": 58154 + }, + { + "epoch": 7.62, + "learning_rate": 1.1941555828894004e-07, + "loss": 0.0007, + "step": 58155 + }, + { + "epoch": 7.62, + "learning_rate": 1.1933385993042834e-07, + "loss": 0.0026, + "step": 58156 + }, + { + "epoch": 7.62, + "learning_rate": 1.1925218936076566e-07, + "loss": 0.0005, + "step": 58157 + }, + { + "epoch": 7.62, + "learning_rate": 1.1917054658017957e-07, + "loss": 0.0007, + "step": 58158 + }, + { + "epoch": 7.62, + "learning_rate": 1.1908893158889879e-07, + "loss": 0.0003, + "step": 58159 + }, + { + "epoch": 7.62, + "learning_rate": 1.1900734438715533e-07, + "loss": 0.0006, + "step": 58160 + }, + { + "epoch": 7.62, + "learning_rate": 1.1892578497517682e-07, + "loss": 0.001, + "step": 58161 + }, + { + "epoch": 7.62, + "learning_rate": 1.1884425335319416e-07, + "loss": 0.0002, + "step": 58162 + }, + { + "epoch": 7.62, + "learning_rate": 1.1876274952143496e-07, + "loss": 0.002, + "step": 58163 + }, + { + "epoch": 7.62, + "learning_rate": 1.1868127348012903e-07, + "loss": 0.0001, + "step": 58164 + }, + { + "epoch": 7.62, + "learning_rate": 1.185998252295073e-07, + "loss": 0.0005, + "step": 58165 + }, + { + "epoch": 7.62, + "learning_rate": 1.1851840476979626e-07, + "loss": 0.0008, + "step": 58166 + }, + { + "epoch": 7.62, + "learning_rate": 1.1843701210122682e-07, + "loss": 0.0004, + "step": 58167 + }, + { + "epoch": 7.62, + "learning_rate": 1.1835564722402659e-07, + "loss": 0.0004, + "step": 58168 + }, + { + "epoch": 7.62, + "learning_rate": 1.1827431013842427e-07, + "loss": 0.0002, + "step": 58169 + }, + { + "epoch": 7.62, + "learning_rate": 1.1819300084465079e-07, + "loss": 0.0005, + "step": 58170 + }, + { + "epoch": 7.62, + "learning_rate": 1.1811171934293152e-07, + "loss": 0.0004, + "step": 58171 + }, + { + "epoch": 7.62, + "learning_rate": 1.1803046563349741e-07, + "loss": 0.0002, + "step": 58172 + }, + { + "epoch": 7.62, + "learning_rate": 1.1794923971657713e-07, + "loss": 0.0005, + "step": 58173 + }, + { + "epoch": 7.62, + "learning_rate": 1.1786804159239717e-07, + "loss": 0.0005, + "step": 58174 + }, + { + "epoch": 7.62, + "learning_rate": 1.1778687126118738e-07, + "loss": 0.0006, + "step": 58175 + }, + { + "epoch": 7.62, + "learning_rate": 1.1770572872317643e-07, + "loss": 0.0006, + "step": 58176 + }, + { + "epoch": 7.62, + "learning_rate": 1.1762461397859082e-07, + "loss": 0.0002, + "step": 58177 + }, + { + "epoch": 7.62, + "learning_rate": 1.1754352702766037e-07, + "loss": 0.0008, + "step": 58178 + }, + { + "epoch": 7.62, + "learning_rate": 1.1746246787061266e-07, + "loss": 0.0007, + "step": 58179 + }, + { + "epoch": 7.62, + "learning_rate": 1.173814365076753e-07, + "loss": 0.0004, + "step": 58180 + }, + { + "epoch": 7.62, + "learning_rate": 1.17300432939077e-07, + "loss": 0.0003, + "step": 58181 + }, + { + "epoch": 7.62, + "learning_rate": 1.1721945716504424e-07, + "loss": 0.0004, + "step": 58182 + }, + { + "epoch": 7.62, + "learning_rate": 1.1713850918580461e-07, + "loss": 0.0003, + "step": 58183 + }, + { + "epoch": 7.62, + "learning_rate": 1.1705758900158793e-07, + "loss": 0.0005, + "step": 58184 + }, + { + "epoch": 7.62, + "learning_rate": 1.1697669661261957e-07, + "loss": 0.0005, + "step": 58185 + }, + { + "epoch": 7.62, + "learning_rate": 1.1689583201912936e-07, + "loss": 0.0007, + "step": 58186 + }, + { + "epoch": 7.62, + "learning_rate": 1.1681499522134154e-07, + "loss": 0.0002, + "step": 58187 + }, + { + "epoch": 7.62, + "learning_rate": 1.1673418621948596e-07, + "loss": 0.0009, + "step": 58188 + }, + { + "epoch": 7.62, + "learning_rate": 1.1665340501379019e-07, + "loss": 0.0002, + "step": 58189 + }, + { + "epoch": 7.62, + "learning_rate": 1.1657265160447961e-07, + "loss": 0.0024, + "step": 58190 + }, + { + "epoch": 7.62, + "learning_rate": 1.1649192599178183e-07, + "loss": 0.0005, + "step": 58191 + }, + { + "epoch": 7.62, + "learning_rate": 1.1641122817592443e-07, + "loss": 0.0004, + "step": 58192 + }, + { + "epoch": 7.62, + "learning_rate": 1.163305581571339e-07, + "loss": 0.0002, + "step": 58193 + }, + { + "epoch": 7.62, + "learning_rate": 1.1624991593563673e-07, + "loss": 0.0003, + "step": 58194 + }, + { + "epoch": 7.62, + "learning_rate": 1.1616930151166161e-07, + "loss": 0.0003, + "step": 58195 + }, + { + "epoch": 7.62, + "learning_rate": 1.1608871488543282e-07, + "loss": 0.001, + "step": 58196 + }, + { + "epoch": 7.62, + "learning_rate": 1.1600815605717908e-07, + "loss": 0.0004, + "step": 58197 + }, + { + "epoch": 7.62, + "learning_rate": 1.1592762502712573e-07, + "loss": 0.0005, + "step": 58198 + }, + { + "epoch": 7.62, + "learning_rate": 1.158471217955004e-07, + "loss": 0.0003, + "step": 58199 + }, + { + "epoch": 7.62, + "learning_rate": 1.1576664636252733e-07, + "loss": 0.0006, + "step": 58200 + }, + { + "epoch": 7.62, + "learning_rate": 1.1568619872843523e-07, + "loss": 0.0005, + "step": 58201 + }, + { + "epoch": 7.62, + "learning_rate": 1.1560577889344838e-07, + "loss": 0.0006, + "step": 58202 + }, + { + "epoch": 7.62, + "learning_rate": 1.1552538685779435e-07, + "loss": 0.0003, + "step": 58203 + }, + { + "epoch": 7.62, + "learning_rate": 1.1544502262169966e-07, + "loss": 0.0002, + "step": 58204 + }, + { + "epoch": 7.62, + "learning_rate": 1.1536468618538854e-07, + "loss": 0.0003, + "step": 58205 + }, + { + "epoch": 7.62, + "learning_rate": 1.152843775490875e-07, + "loss": 0.0004, + "step": 58206 + }, + { + "epoch": 7.62, + "learning_rate": 1.1520409671302301e-07, + "loss": 0.0003, + "step": 58207 + }, + { + "epoch": 7.62, + "learning_rate": 1.1512384367742158e-07, + "loss": 0.0007, + "step": 58208 + }, + { + "epoch": 7.62, + "learning_rate": 1.1504361844250744e-07, + "loss": 0.0002, + "step": 58209 + }, + { + "epoch": 7.63, + "learning_rate": 1.14963421008506e-07, + "loss": 0.0006, + "step": 58210 + }, + { + "epoch": 7.63, + "learning_rate": 1.1488325137564371e-07, + "loss": 0.0002, + "step": 58211 + }, + { + "epoch": 7.63, + "learning_rate": 1.1480310954414598e-07, + "loss": 0.0006, + "step": 58212 + }, + { + "epoch": 7.63, + "learning_rate": 1.1472299551423816e-07, + "loss": 0.0003, + "step": 58213 + }, + { + "epoch": 7.63, + "learning_rate": 1.1464290928614563e-07, + "loss": 0.0005, + "step": 58214 + }, + { + "epoch": 7.63, + "learning_rate": 1.1456285086009378e-07, + "loss": 0.0002, + "step": 58215 + }, + { + "epoch": 7.63, + "learning_rate": 1.1448282023630797e-07, + "loss": 0.0003, + "step": 58216 + }, + { + "epoch": 7.63, + "learning_rate": 1.1440281741501136e-07, + "loss": 0.0002, + "step": 58217 + }, + { + "epoch": 7.63, + "learning_rate": 1.1432284239643043e-07, + "loss": 0.0006, + "step": 58218 + }, + { + "epoch": 7.63, + "learning_rate": 1.1424289518079057e-07, + "loss": 0.0003, + "step": 58219 + }, + { + "epoch": 7.63, + "learning_rate": 1.1416297576831603e-07, + "loss": 0.0004, + "step": 58220 + }, + { + "epoch": 7.63, + "learning_rate": 1.1408308415923219e-07, + "loss": 0.0001, + "step": 58221 + }, + { + "epoch": 7.63, + "learning_rate": 1.1400322035376221e-07, + "loss": 0.0003, + "step": 58222 + }, + { + "epoch": 7.63, + "learning_rate": 1.1392338435213257e-07, + "loss": 0.0002, + "step": 58223 + }, + { + "epoch": 7.63, + "learning_rate": 1.1384357615456753e-07, + "loss": 0.0009, + "step": 58224 + }, + { + "epoch": 7.63, + "learning_rate": 1.1376379576128915e-07, + "loss": 0.0003, + "step": 58225 + }, + { + "epoch": 7.63, + "learning_rate": 1.1368404317252502e-07, + "loss": 0.0005, + "step": 58226 + }, + { + "epoch": 7.63, + "learning_rate": 1.1360431838849828e-07, + "loss": 0.0004, + "step": 58227 + }, + { + "epoch": 7.63, + "learning_rate": 1.135246214094321e-07, + "loss": 0.0002, + "step": 58228 + }, + { + "epoch": 7.63, + "learning_rate": 1.1344495223555074e-07, + "loss": 0.0003, + "step": 58229 + }, + { + "epoch": 7.63, + "learning_rate": 1.1336531086707958e-07, + "loss": 0.0007, + "step": 58230 + }, + { + "epoch": 7.63, + "learning_rate": 1.1328569730424288e-07, + "loss": 0.0002, + "step": 58231 + }, + { + "epoch": 7.63, + "learning_rate": 1.132061115472638e-07, + "loss": 0.0006, + "step": 58232 + }, + { + "epoch": 7.63, + "learning_rate": 1.1312655359636437e-07, + "loss": 0.0003, + "step": 58233 + }, + { + "epoch": 7.63, + "learning_rate": 1.130470234517711e-07, + "loss": 0.0005, + "step": 58234 + }, + { + "epoch": 7.63, + "learning_rate": 1.1296752111370712e-07, + "loss": 0.0001, + "step": 58235 + }, + { + "epoch": 7.63, + "learning_rate": 1.1288804658239339e-07, + "loss": 0.0005, + "step": 58236 + }, + { + "epoch": 7.63, + "learning_rate": 1.1280859985805748e-07, + "loss": 0.0006, + "step": 58237 + }, + { + "epoch": 7.63, + "learning_rate": 1.1272918094092034e-07, + "loss": 0.0003, + "step": 58238 + }, + { + "epoch": 7.63, + "learning_rate": 1.1264978983120511e-07, + "loss": 0.0007, + "step": 58239 + }, + { + "epoch": 7.63, + "learning_rate": 1.1257042652913719e-07, + "loss": 0.0004, + "step": 58240 + }, + { + "epoch": 7.63, + "learning_rate": 1.1249109103493639e-07, + "loss": 0.0006, + "step": 58241 + }, + { + "epoch": 7.63, + "learning_rate": 1.1241178334883029e-07, + "loss": 0.0006, + "step": 58242 + }, + { + "epoch": 7.63, + "learning_rate": 1.1233250347103763e-07, + "loss": 0.0004, + "step": 58243 + }, + { + "epoch": 7.63, + "learning_rate": 1.1225325140178378e-07, + "loss": 0.0003, + "step": 58244 + }, + { + "epoch": 7.63, + "learning_rate": 1.1217402714129189e-07, + "loss": 0.0004, + "step": 58245 + }, + { + "epoch": 7.63, + "learning_rate": 1.1209483068978289e-07, + "loss": 0.0012, + "step": 58246 + }, + { + "epoch": 7.63, + "learning_rate": 1.1201566204748216e-07, + "loss": 0.0002, + "step": 58247 + }, + { + "epoch": 7.63, + "learning_rate": 1.1193652121460841e-07, + "loss": 0.0002, + "step": 58248 + }, + { + "epoch": 7.63, + "learning_rate": 1.1185740819138813e-07, + "loss": 0.0003, + "step": 58249 + }, + { + "epoch": 7.63, + "learning_rate": 1.1177832297804225e-07, + "loss": 0.0004, + "step": 58250 + }, + { + "epoch": 7.63, + "learning_rate": 1.1169926557479282e-07, + "loss": 0.0004, + "step": 58251 + }, + { + "epoch": 7.63, + "learning_rate": 1.11620235981863e-07, + "loss": 0.0012, + "step": 58252 + }, + { + "epoch": 7.63, + "learning_rate": 1.1154123419947482e-07, + "loss": 0.0002, + "step": 58253 + }, + { + "epoch": 7.63, + "learning_rate": 1.1146226022785034e-07, + "loss": 0.0002, + "step": 58254 + }, + { + "epoch": 7.63, + "learning_rate": 1.1138331406721159e-07, + "loss": 0.0005, + "step": 58255 + }, + { + "epoch": 7.63, + "learning_rate": 1.1130439571778062e-07, + "loss": 0.0006, + "step": 58256 + }, + { + "epoch": 7.63, + "learning_rate": 1.1122550517977948e-07, + "loss": 0.0003, + "step": 58257 + }, + { + "epoch": 7.63, + "learning_rate": 1.111466424534302e-07, + "loss": 0.0005, + "step": 58258 + }, + { + "epoch": 7.63, + "learning_rate": 1.1106780753895374e-07, + "loss": 0.0003, + "step": 58259 + }, + { + "epoch": 7.63, + "learning_rate": 1.1098900043657213e-07, + "loss": 0.0003, + "step": 58260 + }, + { + "epoch": 7.63, + "learning_rate": 1.1091022114650851e-07, + "loss": 0.0003, + "step": 58261 + }, + { + "epoch": 7.63, + "learning_rate": 1.1083146966898273e-07, + "loss": 0.0013, + "step": 58262 + }, + { + "epoch": 7.63, + "learning_rate": 1.107527460042157e-07, + "loss": 0.0007, + "step": 58263 + }, + { + "epoch": 7.63, + "learning_rate": 1.1067405015243172e-07, + "loss": 0.0007, + "step": 58264 + }, + { + "epoch": 7.63, + "learning_rate": 1.1059538211384946e-07, + "loss": 0.0003, + "step": 58265 + }, + { + "epoch": 7.63, + "learning_rate": 1.105167418886921e-07, + "loss": 0.0004, + "step": 58266 + }, + { + "epoch": 7.63, + "learning_rate": 1.1043812947717836e-07, + "loss": 0.0004, + "step": 58267 + }, + { + "epoch": 7.63, + "learning_rate": 1.1035954487953138e-07, + "loss": 0.0006, + "step": 58268 + }, + { + "epoch": 7.63, + "learning_rate": 1.1028098809597209e-07, + "loss": 0.0003, + "step": 58269 + }, + { + "epoch": 7.63, + "learning_rate": 1.1020245912671923e-07, + "loss": 0.0003, + "step": 58270 + }, + { + "epoch": 7.63, + "learning_rate": 1.1012395797199704e-07, + "loss": 0.0002, + "step": 58271 + }, + { + "epoch": 7.63, + "learning_rate": 1.1004548463202314e-07, + "loss": 0.0004, + "step": 58272 + }, + { + "epoch": 7.63, + "learning_rate": 1.0996703910702067e-07, + "loss": 0.0003, + "step": 58273 + }, + { + "epoch": 7.63, + "learning_rate": 1.0988862139720835e-07, + "loss": 0.0005, + "step": 58274 + }, + { + "epoch": 7.63, + "learning_rate": 1.0981023150280823e-07, + "loss": 0.0001, + "step": 58275 + }, + { + "epoch": 7.63, + "learning_rate": 1.0973186942404013e-07, + "loss": 0.0004, + "step": 58276 + }, + { + "epoch": 7.63, + "learning_rate": 1.0965353516112387e-07, + "loss": 0.0009, + "step": 58277 + }, + { + "epoch": 7.63, + "learning_rate": 1.0957522871428039e-07, + "loss": 0.0004, + "step": 58278 + }, + { + "epoch": 7.63, + "learning_rate": 1.0949695008373062e-07, + "loss": 0.0004, + "step": 58279 + }, + { + "epoch": 7.63, + "learning_rate": 1.0941869926969439e-07, + "loss": 0.0006, + "step": 58280 + }, + { + "epoch": 7.63, + "learning_rate": 1.0934047627239042e-07, + "loss": 0.001, + "step": 58281 + }, + { + "epoch": 7.63, + "learning_rate": 1.0926228109203962e-07, + "loss": 0.0003, + "step": 58282 + }, + { + "epoch": 7.63, + "learning_rate": 1.0918411372886184e-07, + "loss": 0.0009, + "step": 58283 + }, + { + "epoch": 7.63, + "learning_rate": 1.09105974183078e-07, + "loss": 0.0004, + "step": 58284 + }, + { + "epoch": 7.63, + "learning_rate": 1.090278624549057e-07, + "loss": 0.0008, + "step": 58285 + }, + { + "epoch": 7.64, + "learning_rate": 1.089497785445659e-07, + "loss": 0.0004, + "step": 58286 + }, + { + "epoch": 7.64, + "learning_rate": 1.0887172245227951e-07, + "loss": 0.0002, + "step": 58287 + }, + { + "epoch": 7.64, + "learning_rate": 1.0879369417826414e-07, + "loss": 0.0003, + "step": 58288 + }, + { + "epoch": 7.64, + "learning_rate": 1.087156937227396e-07, + "loss": 0.0002, + "step": 58289 + }, + { + "epoch": 7.64, + "learning_rate": 1.0863772108592463e-07, + "loss": 0.0003, + "step": 58290 + }, + { + "epoch": 7.64, + "learning_rate": 1.0855977626804016e-07, + "loss": 0.0003, + "step": 58291 + }, + { + "epoch": 7.64, + "learning_rate": 1.0848185926930377e-07, + "loss": 0.0006, + "step": 58292 + }, + { + "epoch": 7.64, + "learning_rate": 1.084039700899353e-07, + "loss": 0.0007, + "step": 58293 + }, + { + "epoch": 7.64, + "learning_rate": 1.0832610873015458e-07, + "loss": 0.0007, + "step": 58294 + }, + { + "epoch": 7.64, + "learning_rate": 1.0824827519018033e-07, + "loss": 0.0005, + "step": 58295 + }, + { + "epoch": 7.64, + "learning_rate": 1.0817046947023013e-07, + "loss": 0.0005, + "step": 58296 + }, + { + "epoch": 7.64, + "learning_rate": 1.0809269157052494e-07, + "loss": 0.0005, + "step": 58297 + }, + { + "epoch": 7.64, + "learning_rate": 1.0801494149128122e-07, + "loss": 0.0004, + "step": 58298 + }, + { + "epoch": 7.64, + "learning_rate": 1.0793721923271883e-07, + "loss": 0.0003, + "step": 58299 + }, + { + "epoch": 7.64, + "learning_rate": 1.0785952479505646e-07, + "loss": 0.0006, + "step": 58300 + }, + { + "epoch": 7.64, + "learning_rate": 1.0778185817851172e-07, + "loss": 0.0003, + "step": 58301 + }, + { + "epoch": 7.64, + "learning_rate": 1.0770421938330444e-07, + "loss": 0.0005, + "step": 58302 + }, + { + "epoch": 7.64, + "learning_rate": 1.0762660840965333e-07, + "loss": 0.0009, + "step": 58303 + }, + { + "epoch": 7.64, + "learning_rate": 1.0754902525777378e-07, + "loss": 0.0003, + "step": 58304 + }, + { + "epoch": 7.64, + "learning_rate": 1.074714699278867e-07, + "loss": 0.0002, + "step": 58305 + }, + { + "epoch": 7.64, + "learning_rate": 1.0739394242020861e-07, + "loss": 0.0002, + "step": 58306 + }, + { + "epoch": 7.64, + "learning_rate": 1.0731644273495823e-07, + "loss": 0.0006, + "step": 58307 + }, + { + "epoch": 7.64, + "learning_rate": 1.0723897087235424e-07, + "loss": 0.0002, + "step": 58308 + }, + { + "epoch": 7.64, + "learning_rate": 1.0716152683261316e-07, + "loss": 0.0002, + "step": 58309 + }, + { + "epoch": 7.64, + "learning_rate": 1.070841106159537e-07, + "loss": 0.0004, + "step": 58310 + }, + { + "epoch": 7.64, + "learning_rate": 1.0700672222259345e-07, + "loss": 0.0003, + "step": 58311 + }, + { + "epoch": 7.64, + "learning_rate": 1.0692936165275113e-07, + "loss": 0.001, + "step": 58312 + }, + { + "epoch": 7.64, + "learning_rate": 1.0685202890664104e-07, + "loss": 0.0003, + "step": 58313 + }, + { + "epoch": 7.64, + "learning_rate": 1.0677472398448407e-07, + "loss": 0.0001, + "step": 58314 + }, + { + "epoch": 7.64, + "learning_rate": 1.0669744688649675e-07, + "loss": 0.0005, + "step": 58315 + }, + { + "epoch": 7.64, + "learning_rate": 1.0662019761289444e-07, + "loss": 0.001, + "step": 58316 + }, + { + "epoch": 7.64, + "learning_rate": 1.0654297616389587e-07, + "loss": 0.0008, + "step": 58317 + }, + { + "epoch": 7.64, + "learning_rate": 1.0646578253971973e-07, + "loss": 0.0007, + "step": 58318 + }, + { + "epoch": 7.64, + "learning_rate": 1.0638861674058143e-07, + "loss": 0.0003, + "step": 58319 + }, + { + "epoch": 7.64, + "learning_rate": 1.0631147876669745e-07, + "loss": 0.0002, + "step": 58320 + }, + { + "epoch": 7.64, + "learning_rate": 1.0623436861828651e-07, + "loss": 0.0007, + "step": 58321 + }, + { + "epoch": 7.64, + "learning_rate": 1.061572862955651e-07, + "loss": 0.0003, + "step": 58322 + }, + { + "epoch": 7.64, + "learning_rate": 1.060802317987486e-07, + "loss": 0.0002, + "step": 58323 + }, + { + "epoch": 7.64, + "learning_rate": 1.0600320512805352e-07, + "loss": 0.001, + "step": 58324 + }, + { + "epoch": 7.64, + "learning_rate": 1.0592620628369854e-07, + "loss": 0.0004, + "step": 58325 + }, + { + "epoch": 7.64, + "learning_rate": 1.0584923526590019e-07, + "loss": 0.0009, + "step": 58326 + }, + { + "epoch": 7.64, + "learning_rate": 1.0577229207487271e-07, + "loss": 0.0002, + "step": 58327 + }, + { + "epoch": 7.64, + "learning_rate": 1.0569537671083374e-07, + "loss": 0.0008, + "step": 58328 + }, + { + "epoch": 7.64, + "learning_rate": 1.0561848917399974e-07, + "loss": 0.0007, + "step": 58329 + }, + { + "epoch": 7.64, + "learning_rate": 1.0554162946458723e-07, + "loss": 0.0006, + "step": 58330 + }, + { + "epoch": 7.64, + "learning_rate": 1.0546479758281158e-07, + "loss": 0.0007, + "step": 58331 + }, + { + "epoch": 7.64, + "learning_rate": 1.0538799352888928e-07, + "loss": 0.0003, + "step": 58332 + }, + { + "epoch": 7.64, + "learning_rate": 1.0531121730303684e-07, + "loss": 0.0003, + "step": 58333 + }, + { + "epoch": 7.64, + "learning_rate": 1.0523446890546962e-07, + "loss": 0.0003, + "step": 58334 + }, + { + "epoch": 7.64, + "learning_rate": 1.0515774833640302e-07, + "loss": 0.0005, + "step": 58335 + }, + { + "epoch": 7.64, + "learning_rate": 1.0508105559605353e-07, + "loss": 0.0003, + "step": 58336 + }, + { + "epoch": 7.64, + "learning_rate": 1.0500439068463653e-07, + "loss": 0.0005, + "step": 58337 + }, + { + "epoch": 7.64, + "learning_rate": 1.049277536023674e-07, + "loss": 0.0005, + "step": 58338 + }, + { + "epoch": 7.64, + "learning_rate": 1.0485114434946264e-07, + "loss": 0.0004, + "step": 58339 + }, + { + "epoch": 7.64, + "learning_rate": 1.0477456292613653e-07, + "loss": 0.0013, + "step": 58340 + }, + { + "epoch": 7.64, + "learning_rate": 1.0469800933260555e-07, + "loss": 0.0002, + "step": 58341 + }, + { + "epoch": 7.64, + "learning_rate": 1.0462148356908508e-07, + "loss": 0.0008, + "step": 58342 + }, + { + "epoch": 7.64, + "learning_rate": 1.045449856357883e-07, + "loss": 0.0002, + "step": 58343 + }, + { + "epoch": 7.64, + "learning_rate": 1.044685155329328e-07, + "loss": 0.0002, + "step": 58344 + }, + { + "epoch": 7.64, + "learning_rate": 1.0439207326073286e-07, + "loss": 0.0004, + "step": 58345 + }, + { + "epoch": 7.64, + "learning_rate": 1.0431565881940276e-07, + "loss": 0.0004, + "step": 58346 + }, + { + "epoch": 7.64, + "learning_rate": 1.0423927220915675e-07, + "loss": 0.0006, + "step": 58347 + }, + { + "epoch": 7.64, + "learning_rate": 1.0416291343021245e-07, + "loss": 0.0003, + "step": 58348 + }, + { + "epoch": 7.64, + "learning_rate": 1.0408658248278192e-07, + "loss": 0.0006, + "step": 58349 + }, + { + "epoch": 7.64, + "learning_rate": 1.0401027936708163e-07, + "loss": 0.0005, + "step": 58350 + }, + { + "epoch": 7.64, + "learning_rate": 1.0393400408332477e-07, + "loss": 0.0006, + "step": 58351 + }, + { + "epoch": 7.64, + "learning_rate": 1.0385775663172782e-07, + "loss": 0.0007, + "step": 58352 + }, + { + "epoch": 7.64, + "learning_rate": 1.0378153701250283e-07, + "loss": 0.0008, + "step": 58353 + }, + { + "epoch": 7.64, + "learning_rate": 1.0370534522586517e-07, + "loss": 0.0003, + "step": 58354 + }, + { + "epoch": 7.64, + "learning_rate": 1.0362918127202915e-07, + "loss": 0.0003, + "step": 58355 + }, + { + "epoch": 7.64, + "learning_rate": 1.0355304515121012e-07, + "loss": 0.0004, + "step": 58356 + }, + { + "epoch": 7.64, + "learning_rate": 1.0347693686362014e-07, + "loss": 0.0009, + "step": 58357 + }, + { + "epoch": 7.64, + "learning_rate": 1.0340085640947462e-07, + "loss": 0.0006, + "step": 58358 + }, + { + "epoch": 7.64, + "learning_rate": 1.0332480378898669e-07, + "loss": 0.0003, + "step": 58359 + }, + { + "epoch": 7.64, + "learning_rate": 1.0324877900237063e-07, + "loss": 0.0004, + "step": 58360 + }, + { + "epoch": 7.64, + "learning_rate": 1.0317278204984182e-07, + "loss": 0.0008, + "step": 58361 + }, + { + "epoch": 7.65, + "learning_rate": 1.0309681293161122e-07, + "loss": 0.001, + "step": 58362 + }, + { + "epoch": 7.65, + "learning_rate": 1.030208716478942e-07, + "loss": 0.001, + "step": 58363 + }, + { + "epoch": 7.65, + "learning_rate": 1.0294495819890394e-07, + "loss": 0.0006, + "step": 58364 + }, + { + "epoch": 7.65, + "learning_rate": 1.0286907258485357e-07, + "loss": 0.0003, + "step": 58365 + }, + { + "epoch": 7.65, + "learning_rate": 1.0279321480595739e-07, + "loss": 0.0002, + "step": 58366 + }, + { + "epoch": 7.65, + "learning_rate": 1.0271738486242744e-07, + "loss": 0.0007, + "step": 58367 + }, + { + "epoch": 7.65, + "learning_rate": 1.026415827544791e-07, + "loss": 0.0002, + "step": 58368 + }, + { + "epoch": 7.65, + "learning_rate": 1.0256580848232334e-07, + "loss": 0.0003, + "step": 58369 + }, + { + "epoch": 7.65, + "learning_rate": 1.0249006204617329e-07, + "loss": 0.0007, + "step": 58370 + }, + { + "epoch": 7.65, + "learning_rate": 1.0241434344624435e-07, + "loss": 0.0005, + "step": 58371 + }, + { + "epoch": 7.65, + "learning_rate": 1.0233865268274745e-07, + "loss": 0.0007, + "step": 58372 + }, + { + "epoch": 7.65, + "learning_rate": 1.0226298975589688e-07, + "loss": 0.0005, + "step": 58373 + }, + { + "epoch": 7.65, + "learning_rate": 1.0218735466590357e-07, + "loss": 0.0003, + "step": 58374 + }, + { + "epoch": 7.65, + "learning_rate": 1.0211174741298068e-07, + "loss": 0.0013, + "step": 58375 + }, + { + "epoch": 7.65, + "learning_rate": 1.020361679973425e-07, + "loss": 0.0004, + "step": 58376 + }, + { + "epoch": 7.65, + "learning_rate": 1.0196061641920107e-07, + "loss": 0.0008, + "step": 58377 + }, + { + "epoch": 7.65, + "learning_rate": 1.0188509267876623e-07, + "loss": 0.0007, + "step": 58378 + }, + { + "epoch": 7.65, + "learning_rate": 1.0180959677625446e-07, + "loss": 0.0006, + "step": 58379 + }, + { + "epoch": 7.65, + "learning_rate": 1.017341287118756e-07, + "loss": 0.0002, + "step": 58380 + }, + { + "epoch": 7.65, + "learning_rate": 1.016586884858417e-07, + "loss": 0.0003, + "step": 58381 + }, + { + "epoch": 7.65, + "learning_rate": 1.0158327609836482e-07, + "loss": 0.0003, + "step": 58382 + }, + { + "epoch": 7.65, + "learning_rate": 1.0150789154965923e-07, + "loss": 0.001, + "step": 58383 + }, + { + "epoch": 7.65, + "learning_rate": 1.0143253483993587e-07, + "loss": 0.0003, + "step": 58384 + }, + { + "epoch": 7.65, + "learning_rate": 1.0135720596940568e-07, + "loss": 0.0003, + "step": 58385 + }, + { + "epoch": 7.65, + "learning_rate": 1.0128190493828182e-07, + "loss": 0.0003, + "step": 58386 + }, + { + "epoch": 7.65, + "learning_rate": 1.0120663174677525e-07, + "loss": 0.0003, + "step": 58387 + }, + { + "epoch": 7.65, + "learning_rate": 1.011313863950969e-07, + "loss": 0.0008, + "step": 58388 + }, + { + "epoch": 7.65, + "learning_rate": 1.0105616888345993e-07, + "loss": 0.0007, + "step": 58389 + }, + { + "epoch": 7.65, + "learning_rate": 1.0098097921207639e-07, + "loss": 0.0006, + "step": 58390 + }, + { + "epoch": 7.65, + "learning_rate": 1.0090581738115612e-07, + "loss": 0.0002, + "step": 58391 + }, + { + "epoch": 7.65, + "learning_rate": 1.0083068339091007e-07, + "loss": 0.0006, + "step": 58392 + }, + { + "epoch": 7.65, + "learning_rate": 1.0075557724155138e-07, + "loss": 0.0002, + "step": 58393 + }, + { + "epoch": 7.65, + "learning_rate": 1.006804989332899e-07, + "loss": 0.0008, + "step": 58394 + }, + { + "epoch": 7.65, + "learning_rate": 1.0060544846633769e-07, + "loss": 0.0004, + "step": 58395 + }, + { + "epoch": 7.65, + "learning_rate": 1.0053042584090678e-07, + "loss": 0.0005, + "step": 58396 + }, + { + "epoch": 7.65, + "learning_rate": 1.0045543105720479e-07, + "loss": 0.0002, + "step": 58397 + }, + { + "epoch": 7.65, + "learning_rate": 1.00380464115446e-07, + "loss": 0.0003, + "step": 58398 + }, + { + "epoch": 7.65, + "learning_rate": 1.0030552501584023e-07, + "loss": 0.0005, + "step": 58399 + }, + { + "epoch": 7.65, + "learning_rate": 1.0023061375859621e-07, + "loss": 0.0003, + "step": 58400 + }, + { + "epoch": 7.65, + "learning_rate": 1.0015573034392823e-07, + "loss": 0.0002, + "step": 58401 + }, + { + "epoch": 7.65, + "learning_rate": 1.0008087477204498e-07, + "loss": 0.0003, + "step": 58402 + }, + { + "epoch": 7.65, + "learning_rate": 1.0000604704315519e-07, + "loss": 0.0002, + "step": 58403 + }, + { + "epoch": 7.65, + "learning_rate": 9.993124715747316e-08, + "loss": 0.0004, + "step": 58404 + }, + { + "epoch": 7.65, + "learning_rate": 9.985647511520647e-08, + "loss": 0.0002, + "step": 58405 + }, + { + "epoch": 7.65, + "learning_rate": 9.978173091656607e-08, + "loss": 0.0002, + "step": 58406 + }, + { + "epoch": 7.65, + "learning_rate": 9.970701456176291e-08, + "loss": 0.0003, + "step": 58407 + }, + { + "epoch": 7.65, + "learning_rate": 9.96323260510057e-08, + "loss": 0.0003, + "step": 58408 + }, + { + "epoch": 7.65, + "learning_rate": 9.95576653845065e-08, + "loss": 0.0006, + "step": 58409 + }, + { + "epoch": 7.65, + "learning_rate": 9.948303256247404e-08, + "loss": 0.0006, + "step": 58410 + }, + { + "epoch": 7.65, + "learning_rate": 9.940842758511705e-08, + "loss": 0.0006, + "step": 58411 + }, + { + "epoch": 7.65, + "learning_rate": 9.933385045264754e-08, + "loss": 0.0002, + "step": 58412 + }, + { + "epoch": 7.65, + "learning_rate": 9.925930116527427e-08, + "loss": 0.0004, + "step": 58413 + }, + { + "epoch": 7.65, + "learning_rate": 9.918477972320706e-08, + "loss": 0.0008, + "step": 58414 + }, + { + "epoch": 7.65, + "learning_rate": 9.911028612665464e-08, + "loss": 0.0005, + "step": 58415 + }, + { + "epoch": 7.65, + "learning_rate": 9.903582037582793e-08, + "loss": 0.0007, + "step": 58416 + }, + { + "epoch": 7.65, + "learning_rate": 9.896138247093567e-08, + "loss": 0.0005, + "step": 58417 + }, + { + "epoch": 7.65, + "learning_rate": 9.88869724121877e-08, + "loss": 0.0005, + "step": 58418 + }, + { + "epoch": 7.65, + "learning_rate": 9.881259019979273e-08, + "loss": 0.0002, + "step": 58419 + }, + { + "epoch": 7.65, + "learning_rate": 9.873823583395947e-08, + "loss": 0.0005, + "step": 58420 + }, + { + "epoch": 7.65, + "learning_rate": 9.866390931489778e-08, + "loss": 0.0002, + "step": 58421 + }, + { + "epoch": 7.65, + "learning_rate": 9.858961064281636e-08, + "loss": 0.0001, + "step": 58422 + }, + { + "epoch": 7.65, + "learning_rate": 9.851533981792505e-08, + "loss": 0.0007, + "step": 58423 + }, + { + "epoch": 7.65, + "learning_rate": 9.844109684043146e-08, + "loss": 0.0006, + "step": 58424 + }, + { + "epoch": 7.65, + "learning_rate": 9.836688171054542e-08, + "loss": 0.0005, + "step": 58425 + }, + { + "epoch": 7.65, + "learning_rate": 9.829269442847566e-08, + "loss": 0.0003, + "step": 58426 + }, + { + "epoch": 7.65, + "learning_rate": 9.821853499442979e-08, + "loss": 0.0005, + "step": 58427 + }, + { + "epoch": 7.65, + "learning_rate": 9.814440340861653e-08, + "loss": 0.0005, + "step": 58428 + }, + { + "epoch": 7.65, + "learning_rate": 9.80702996712457e-08, + "loss": 0.0007, + "step": 58429 + }, + { + "epoch": 7.65, + "learning_rate": 9.799622378252493e-08, + "loss": 0.0007, + "step": 58430 + }, + { + "epoch": 7.65, + "learning_rate": 9.792217574266072e-08, + "loss": 0.0002, + "step": 58431 + }, + { + "epoch": 7.65, + "learning_rate": 9.784815555186511e-08, + "loss": 0.0002, + "step": 58432 + }, + { + "epoch": 7.65, + "learning_rate": 9.77741632103435e-08, + "loss": 0.0002, + "step": 58433 + }, + { + "epoch": 7.65, + "learning_rate": 9.770019871830461e-08, + "loss": 0.0002, + "step": 58434 + }, + { + "epoch": 7.65, + "learning_rate": 9.762626207595715e-08, + "loss": 0.0003, + "step": 58435 + }, + { + "epoch": 7.65, + "learning_rate": 9.755235328350765e-08, + "loss": 0.0006, + "step": 58436 + }, + { + "epoch": 7.65, + "learning_rate": 9.74784723411648e-08, + "loss": 0.0006, + "step": 58437 + }, + { + "epoch": 7.65, + "learning_rate": 9.740461924913735e-08, + "loss": 0.0002, + "step": 58438 + }, + { + "epoch": 7.66, + "learning_rate": 9.733079400763068e-08, + "loss": 0.0006, + "step": 58439 + }, + { + "epoch": 7.66, + "learning_rate": 9.725699661685572e-08, + "loss": 0.0002, + "step": 58440 + }, + { + "epoch": 7.66, + "learning_rate": 9.718322707701677e-08, + "loss": 0.0006, + "step": 58441 + }, + { + "epoch": 7.66, + "learning_rate": 9.710948538832366e-08, + "loss": 0.0004, + "step": 58442 + }, + { + "epoch": 7.66, + "learning_rate": 9.703577155098176e-08, + "loss": 0.0001, + "step": 58443 + }, + { + "epoch": 7.66, + "learning_rate": 9.696208556519982e-08, + "loss": 0.0004, + "step": 58444 + }, + { + "epoch": 7.66, + "learning_rate": 9.688842743118432e-08, + "loss": 0.0002, + "step": 58445 + }, + { + "epoch": 7.66, + "learning_rate": 9.681479714914288e-08, + "loss": 0.0002, + "step": 58446 + }, + { + "epoch": 7.66, + "learning_rate": 9.6741194719282e-08, + "loss": 0.0002, + "step": 58447 + }, + { + "epoch": 7.66, + "learning_rate": 9.66676201418093e-08, + "loss": 0.0005, + "step": 58448 + }, + { + "epoch": 7.66, + "learning_rate": 9.659407341693239e-08, + "loss": 0.0004, + "step": 58449 + }, + { + "epoch": 7.66, + "learning_rate": 9.652055454485665e-08, + "loss": 0.0005, + "step": 58450 + }, + { + "epoch": 7.66, + "learning_rate": 9.64470635257897e-08, + "loss": 0.0003, + "step": 58451 + }, + { + "epoch": 7.66, + "learning_rate": 9.637360035993804e-08, + "loss": 0.0002, + "step": 58452 + }, + { + "epoch": 7.66, + "learning_rate": 9.630016504750817e-08, + "loss": 0.0007, + "step": 58453 + }, + { + "epoch": 7.66, + "learning_rate": 9.622675758870547e-08, + "loss": 0.0005, + "step": 58454 + }, + { + "epoch": 7.66, + "learning_rate": 9.61533779837387e-08, + "loss": 0.0002, + "step": 58455 + }, + { + "epoch": 7.66, + "learning_rate": 9.608002623281321e-08, + "loss": 0.0002, + "step": 58456 + }, + { + "epoch": 7.66, + "learning_rate": 9.600670233613552e-08, + "loss": 0.0003, + "step": 58457 + }, + { + "epoch": 7.66, + "learning_rate": 9.59334062939099e-08, + "loss": 0.0004, + "step": 58458 + }, + { + "epoch": 7.66, + "learning_rate": 9.58601381063462e-08, + "loss": 0.0003, + "step": 58459 + }, + { + "epoch": 7.66, + "learning_rate": 9.578689777364758e-08, + "loss": 0.0002, + "step": 58460 + }, + { + "epoch": 7.66, + "learning_rate": 9.571368529602166e-08, + "loss": 0.0002, + "step": 58461 + }, + { + "epoch": 7.66, + "learning_rate": 9.56405006736727e-08, + "loss": 0.0005, + "step": 58462 + }, + { + "epoch": 7.66, + "learning_rate": 9.556734390680721e-08, + "loss": 0.0004, + "step": 58463 + }, + { + "epoch": 7.66, + "learning_rate": 9.549421499563171e-08, + "loss": 0.0006, + "step": 58464 + }, + { + "epoch": 7.66, + "learning_rate": 9.542111394035048e-08, + "loss": 0.0003, + "step": 58465 + }, + { + "epoch": 7.66, + "learning_rate": 9.53480407411711e-08, + "loss": 0.0006, + "step": 58466 + }, + { + "epoch": 7.66, + "learning_rate": 9.527499539829676e-08, + "loss": 0.0003, + "step": 58467 + }, + { + "epoch": 7.66, + "learning_rate": 9.520197791193509e-08, + "loss": 0.0002, + "step": 58468 + }, + { + "epoch": 7.66, + "learning_rate": 9.512898828228923e-08, + "loss": 0.001, + "step": 58469 + }, + { + "epoch": 7.66, + "learning_rate": 9.505602650956569e-08, + "loss": 0.0003, + "step": 58470 + }, + { + "epoch": 7.66, + "learning_rate": 9.498309259397099e-08, + "loss": 0.0008, + "step": 58471 + }, + { + "epoch": 7.66, + "learning_rate": 9.491018653570827e-08, + "loss": 0.0005, + "step": 58472 + }, + { + "epoch": 7.66, + "learning_rate": 9.483730833498295e-08, + "loss": 0.0004, + "step": 58473 + }, + { + "epoch": 7.66, + "learning_rate": 9.47644579920004e-08, + "loss": 0.0003, + "step": 58474 + }, + { + "epoch": 7.66, + "learning_rate": 9.469163550696491e-08, + "loss": 0.0003, + "step": 58475 + }, + { + "epoch": 7.66, + "learning_rate": 9.461884088008188e-08, + "loss": 0.0003, + "step": 58476 + }, + { + "epoch": 7.66, + "learning_rate": 9.454607411155558e-08, + "loss": 0.0005, + "step": 58477 + }, + { + "epoch": 7.66, + "learning_rate": 9.44733352015914e-08, + "loss": 0.0003, + "step": 58478 + }, + { + "epoch": 7.66, + "learning_rate": 9.440062415039364e-08, + "loss": 0.0003, + "step": 58479 + }, + { + "epoch": 7.66, + "learning_rate": 9.432794095816656e-08, + "loss": 0.0004, + "step": 58480 + }, + { + "epoch": 7.66, + "learning_rate": 9.425528562511333e-08, + "loss": 0.0006, + "step": 58481 + }, + { + "epoch": 7.66, + "learning_rate": 9.418265815144157e-08, + "loss": 0.0003, + "step": 58482 + }, + { + "epoch": 7.66, + "learning_rate": 9.411005853735222e-08, + "loss": 0.0005, + "step": 58483 + }, + { + "epoch": 7.66, + "learning_rate": 9.40374867830518e-08, + "loss": 0.0004, + "step": 58484 + }, + { + "epoch": 7.66, + "learning_rate": 9.396494288874235e-08, + "loss": 0.0003, + "step": 58485 + }, + { + "epoch": 7.66, + "learning_rate": 9.38924268546304e-08, + "loss": 0.0003, + "step": 58486 + }, + { + "epoch": 7.66, + "learning_rate": 9.381993868091687e-08, + "loss": 0.0003, + "step": 58487 + }, + { + "epoch": 7.66, + "learning_rate": 9.374747836780828e-08, + "loss": 0.0003, + "step": 58488 + }, + { + "epoch": 7.66, + "learning_rate": 9.36750459155067e-08, + "loss": 0.0011, + "step": 58489 + }, + { + "epoch": 7.66, + "learning_rate": 9.360264132421637e-08, + "loss": 0.0004, + "step": 58490 + }, + { + "epoch": 7.66, + "learning_rate": 9.353026459414161e-08, + "loss": 0.0003, + "step": 58491 + }, + { + "epoch": 7.66, + "learning_rate": 9.345791572548557e-08, + "loss": 0.001, + "step": 58492 + }, + { + "epoch": 7.66, + "learning_rate": 9.338559471845144e-08, + "loss": 0.0003, + "step": 58493 + }, + { + "epoch": 7.66, + "learning_rate": 9.331330157324347e-08, + "loss": 0.0003, + "step": 58494 + }, + { + "epoch": 7.66, + "learning_rate": 9.324103629006375e-08, + "loss": 0.0005, + "step": 58495 + }, + { + "epoch": 7.66, + "learning_rate": 9.316879886911545e-08, + "loss": 0.0005, + "step": 58496 + }, + { + "epoch": 7.66, + "learning_rate": 9.309658931060283e-08, + "loss": 0.001, + "step": 58497 + }, + { + "epoch": 7.66, + "learning_rate": 9.302440761472797e-08, + "loss": 0.0003, + "step": 58498 + }, + { + "epoch": 7.66, + "learning_rate": 9.295225378169514e-08, + "loss": 0.0002, + "step": 58499 + }, + { + "epoch": 7.66, + "learning_rate": 9.28801278117053e-08, + "loss": 0.0013, + "step": 58500 + }, + { + "epoch": 7.66, + "learning_rate": 9.280802970496384e-08, + "loss": 0.0002, + "step": 58501 + }, + { + "epoch": 7.66, + "learning_rate": 9.273595946167168e-08, + "loss": 0.001, + "step": 58502 + }, + { + "epoch": 7.66, + "learning_rate": 9.266391708203204e-08, + "loss": 0.0004, + "step": 58503 + }, + { + "epoch": 7.66, + "learning_rate": 9.259190256624695e-08, + "loss": 0.0005, + "step": 58504 + }, + { + "epoch": 7.66, + "learning_rate": 9.25199159145207e-08, + "loss": 0.0004, + "step": 58505 + }, + { + "epoch": 7.66, + "learning_rate": 9.244795712705424e-08, + "loss": 0.0005, + "step": 58506 + }, + { + "epoch": 7.66, + "learning_rate": 9.237602620405073e-08, + "loss": 0.0007, + "step": 58507 + }, + { + "epoch": 7.66, + "learning_rate": 9.230412314571002e-08, + "loss": 0.0013, + "step": 58508 + }, + { + "epoch": 7.66, + "learning_rate": 9.223224795223862e-08, + "loss": 0.0004, + "step": 58509 + }, + { + "epoch": 7.66, + "learning_rate": 9.216040062383525e-08, + "loss": 0.0003, + "step": 58510 + }, + { + "epoch": 7.66, + "learning_rate": 9.208858116070307e-08, + "loss": 0.0004, + "step": 58511 + }, + { + "epoch": 7.66, + "learning_rate": 9.201678956304416e-08, + "loss": 0.0004, + "step": 58512 + }, + { + "epoch": 7.66, + "learning_rate": 9.194502583106057e-08, + "loss": 0.0005, + "step": 58513 + }, + { + "epoch": 7.66, + "learning_rate": 9.187328996495437e-08, + "loss": 0.0003, + "step": 58514 + }, + { + "epoch": 7.67, + "learning_rate": 9.18015819649265e-08, + "loss": 0.0003, + "step": 58515 + }, + { + "epoch": 7.67, + "learning_rate": 9.172990183117902e-08, + "loss": 0.0004, + "step": 58516 + }, + { + "epoch": 7.67, + "learning_rate": 9.1658249563914e-08, + "loss": 0.0002, + "step": 58517 + }, + { + "epoch": 7.67, + "learning_rate": 9.158662516333238e-08, + "loss": 0.0001, + "step": 58518 + }, + { + "epoch": 7.67, + "learning_rate": 9.151502862963513e-08, + "loss": 0.0003, + "step": 58519 + }, + { + "epoch": 7.67, + "learning_rate": 9.144345996302539e-08, + "loss": 0.0004, + "step": 58520 + }, + { + "epoch": 7.67, + "learning_rate": 9.137191916370192e-08, + "loss": 0.0002, + "step": 58521 + }, + { + "epoch": 7.67, + "learning_rate": 9.130040623186898e-08, + "loss": 0.0008, + "step": 58522 + }, + { + "epoch": 7.67, + "learning_rate": 9.122892116772308e-08, + "loss": 0.0007, + "step": 58523 + }, + { + "epoch": 7.67, + "learning_rate": 9.115746397147074e-08, + "loss": 0.0003, + "step": 58524 + }, + { + "epoch": 7.67, + "learning_rate": 9.108603464330956e-08, + "loss": 0.0003, + "step": 58525 + }, + { + "epoch": 7.67, + "learning_rate": 9.101463318344161e-08, + "loss": 0.0004, + "step": 58526 + }, + { + "epoch": 7.67, + "learning_rate": 9.094325959206562e-08, + "loss": 0.0002, + "step": 58527 + }, + { + "epoch": 7.67, + "learning_rate": 9.087191386938588e-08, + "loss": 0.0012, + "step": 58528 + }, + { + "epoch": 7.67, + "learning_rate": 9.08005960156011e-08, + "loss": 0.001, + "step": 58529 + }, + { + "epoch": 7.67, + "learning_rate": 9.072930603091112e-08, + "loss": 0.0004, + "step": 58530 + }, + { + "epoch": 7.67, + "learning_rate": 9.065804391551691e-08, + "loss": 0.0005, + "step": 58531 + }, + { + "epoch": 7.67, + "learning_rate": 9.058680966962052e-08, + "loss": 0.0005, + "step": 58532 + }, + { + "epoch": 7.67, + "learning_rate": 9.051560329342069e-08, + "loss": 0.0006, + "step": 58533 + }, + { + "epoch": 7.67, + "learning_rate": 9.044442478711724e-08, + "loss": 0.0001, + "step": 58534 + }, + { + "epoch": 7.67, + "learning_rate": 9.037327415091112e-08, + "loss": 0.0004, + "step": 58535 + }, + { + "epoch": 7.67, + "learning_rate": 9.030215138500332e-08, + "loss": 0.0008, + "step": 58536 + }, + { + "epoch": 7.67, + "learning_rate": 9.023105648959252e-08, + "loss": 0.0005, + "step": 58537 + }, + { + "epoch": 7.67, + "learning_rate": 9.01599894648797e-08, + "loss": 0.0004, + "step": 58538 + }, + { + "epoch": 7.67, + "learning_rate": 9.008895031106357e-08, + "loss": 0.0007, + "step": 58539 + }, + { + "epoch": 7.67, + "learning_rate": 9.001793902834511e-08, + "loss": 0.0005, + "step": 58540 + }, + { + "epoch": 7.67, + "learning_rate": 8.994695561692301e-08, + "loss": 0.0002, + "step": 58541 + }, + { + "epoch": 7.67, + "learning_rate": 8.987600007699715e-08, + "loss": 0.0003, + "step": 58542 + }, + { + "epoch": 7.67, + "learning_rate": 8.980507240876735e-08, + "loss": 0.0008, + "step": 58543 + }, + { + "epoch": 7.67, + "learning_rate": 8.973417261243345e-08, + "loss": 0.0003, + "step": 58544 + }, + { + "epoch": 7.67, + "learning_rate": 8.966330068819418e-08, + "loss": 0.0005, + "step": 58545 + }, + { + "epoch": 7.67, + "learning_rate": 8.959245663624938e-08, + "loss": 0.0005, + "step": 58546 + }, + { + "epoch": 7.67, + "learning_rate": 8.952164045679779e-08, + "loss": 0.0012, + "step": 58547 + }, + { + "epoch": 7.67, + "learning_rate": 8.945085215003924e-08, + "loss": 0.0009, + "step": 58548 + }, + { + "epoch": 7.67, + "learning_rate": 8.938009171617134e-08, + "loss": 0.0002, + "step": 58549 + }, + { + "epoch": 7.67, + "learning_rate": 8.930935915539506e-08, + "loss": 0.0002, + "step": 58550 + }, + { + "epoch": 7.67, + "learning_rate": 8.923865446790803e-08, + "loss": 0.0002, + "step": 58551 + }, + { + "epoch": 7.67, + "learning_rate": 8.916797765391005e-08, + "loss": 0.0004, + "step": 58552 + }, + { + "epoch": 7.67, + "learning_rate": 8.909732871359878e-08, + "loss": 0.0009, + "step": 58553 + }, + { + "epoch": 7.67, + "learning_rate": 8.902670764717403e-08, + "loss": 0.0003, + "step": 58554 + }, + { + "epoch": 7.67, + "learning_rate": 8.895611445483343e-08, + "loss": 0.0007, + "step": 58555 + }, + { + "epoch": 7.67, + "learning_rate": 8.888554913677572e-08, + "loss": 0.0013, + "step": 58556 + }, + { + "epoch": 7.67, + "learning_rate": 8.881501169320073e-08, + "loss": 0.0011, + "step": 58557 + }, + { + "epoch": 7.67, + "learning_rate": 8.874450212430386e-08, + "loss": 0.0005, + "step": 58558 + }, + { + "epoch": 7.67, + "learning_rate": 8.867402043028716e-08, + "loss": 0.0006, + "step": 58559 + }, + { + "epoch": 7.67, + "learning_rate": 8.860356661134606e-08, + "loss": 0.0005, + "step": 58560 + }, + { + "epoch": 7.67, + "learning_rate": 8.853314066767926e-08, + "loss": 0.0005, + "step": 58561 + }, + { + "epoch": 7.67, + "learning_rate": 8.846274259948551e-08, + "loss": 0.0002, + "step": 58562 + }, + { + "epoch": 7.67, + "learning_rate": 8.839237240696352e-08, + "loss": 0.0015, + "step": 58563 + }, + { + "epoch": 7.67, + "learning_rate": 8.832203009030871e-08, + "loss": 0.0005, + "step": 58564 + }, + { + "epoch": 7.67, + "learning_rate": 8.82517156497209e-08, + "loss": 0.0005, + "step": 58565 + }, + { + "epoch": 7.67, + "learning_rate": 8.818142908539772e-08, + "loss": 0.0003, + "step": 58566 + }, + { + "epoch": 7.67, + "learning_rate": 8.811117039753569e-08, + "loss": 0.0005, + "step": 58567 + }, + { + "epoch": 7.67, + "learning_rate": 8.804093958633353e-08, + "loss": 0.0007, + "step": 58568 + }, + { + "epoch": 7.67, + "learning_rate": 8.797073665198774e-08, + "loss": 0.0003, + "step": 58569 + }, + { + "epoch": 7.67, + "learning_rate": 8.790056159469706e-08, + "loss": 0.0005, + "step": 58570 + }, + { + "epoch": 7.67, + "learning_rate": 8.783041441465911e-08, + "loss": 0.0003, + "step": 58571 + }, + { + "epoch": 7.67, + "learning_rate": 8.77602951120693e-08, + "loss": 0.0004, + "step": 58572 + }, + { + "epoch": 7.67, + "learning_rate": 8.769020368712522e-08, + "loss": 0.0004, + "step": 58573 + }, + { + "epoch": 7.67, + "learning_rate": 8.762014014002563e-08, + "loss": 0.0005, + "step": 58574 + }, + { + "epoch": 7.67, + "learning_rate": 8.75501044709659e-08, + "loss": 0.0005, + "step": 58575 + }, + { + "epoch": 7.67, + "learning_rate": 8.748009668014368e-08, + "loss": 0.0005, + "step": 58576 + }, + { + "epoch": 7.67, + "learning_rate": 8.741011676775546e-08, + "loss": 0.0005, + "step": 58577 + }, + { + "epoch": 7.67, + "learning_rate": 8.734016473399886e-08, + "loss": 0.0002, + "step": 58578 + }, + { + "epoch": 7.67, + "learning_rate": 8.72702405790704e-08, + "loss": 0.0003, + "step": 58579 + }, + { + "epoch": 7.67, + "learning_rate": 8.720034430316548e-08, + "loss": 0.0005, + "step": 58580 + }, + { + "epoch": 7.67, + "learning_rate": 8.713047590648283e-08, + "loss": 0.0003, + "step": 58581 + }, + { + "epoch": 7.67, + "learning_rate": 8.706063538921783e-08, + "loss": 0.0006, + "step": 58582 + }, + { + "epoch": 7.67, + "learning_rate": 8.6990822751567e-08, + "loss": 0.0005, + "step": 58583 + }, + { + "epoch": 7.67, + "learning_rate": 8.692103799372576e-08, + "loss": 0.0005, + "step": 58584 + }, + { + "epoch": 7.67, + "learning_rate": 8.68512811158917e-08, + "loss": 0.0006, + "step": 58585 + }, + { + "epoch": 7.67, + "learning_rate": 8.678155211826134e-08, + "loss": 0.0006, + "step": 58586 + }, + { + "epoch": 7.67, + "learning_rate": 8.671185100102785e-08, + "loss": 0.0004, + "step": 58587 + }, + { + "epoch": 7.67, + "learning_rate": 8.66421777643911e-08, + "loss": 0.0004, + "step": 58588 + }, + { + "epoch": 7.67, + "learning_rate": 8.657253240854424e-08, + "loss": 0.0007, + "step": 58589 + }, + { + "epoch": 7.67, + "learning_rate": 8.65029149336849e-08, + "loss": 0.0003, + "step": 58590 + }, + { + "epoch": 7.68, + "learning_rate": 8.643332534000848e-08, + "loss": 0.001, + "step": 58591 + }, + { + "epoch": 7.68, + "learning_rate": 8.636376362770927e-08, + "loss": 0.0002, + "step": 58592 + }, + { + "epoch": 7.68, + "learning_rate": 8.629422979698487e-08, + "loss": 0.0005, + "step": 58593 + }, + { + "epoch": 7.68, + "learning_rate": 8.62247238480307e-08, + "loss": 0.0004, + "step": 58594 + }, + { + "epoch": 7.68, + "learning_rate": 8.615524578103996e-08, + "loss": 0.0006, + "step": 58595 + }, + { + "epoch": 7.68, + "learning_rate": 8.608579559621022e-08, + "loss": 0.0007, + "step": 58596 + }, + { + "epoch": 7.68, + "learning_rate": 8.601637329373691e-08, + "loss": 0.0005, + "step": 58597 + }, + { + "epoch": 7.68, + "learning_rate": 8.594697887381431e-08, + "loss": 0.0004, + "step": 58598 + }, + { + "epoch": 7.68, + "learning_rate": 8.587761233663672e-08, + "loss": 0.0003, + "step": 58599 + }, + { + "epoch": 7.68, + "learning_rate": 8.580827368240174e-08, + "loss": 0.0003, + "step": 58600 + }, + { + "epoch": 7.68, + "learning_rate": 8.573896291130257e-08, + "loss": 0.0004, + "step": 58601 + }, + { + "epoch": 7.68, + "learning_rate": 8.56696800235346e-08, + "loss": 0.0003, + "step": 58602 + }, + { + "epoch": 7.68, + "learning_rate": 8.560042501929211e-08, + "loss": 0.0004, + "step": 58603 + }, + { + "epoch": 7.68, + "learning_rate": 8.553119789877163e-08, + "loss": 0.0005, + "step": 58604 + }, + { + "epoch": 7.68, + "learning_rate": 8.54619986621663e-08, + "loss": 0.0005, + "step": 58605 + }, + { + "epoch": 7.68, + "learning_rate": 8.539282730967158e-08, + "loss": 0.0007, + "step": 58606 + }, + { + "epoch": 7.68, + "learning_rate": 8.53236838414806e-08, + "loss": 0.0003, + "step": 58607 + }, + { + "epoch": 7.68, + "learning_rate": 8.525456825778877e-08, + "loss": 0.0005, + "step": 58608 + }, + { + "epoch": 7.68, + "learning_rate": 8.51854805587915e-08, + "loss": 0.0002, + "step": 58609 + }, + { + "epoch": 7.68, + "learning_rate": 8.511642074468086e-08, + "loss": 0.0004, + "step": 58610 + }, + { + "epoch": 7.68, + "learning_rate": 8.504738881565221e-08, + "loss": 0.0005, + "step": 58611 + }, + { + "epoch": 7.68, + "learning_rate": 8.497838477190102e-08, + "loss": 0.0003, + "step": 58612 + }, + { + "epoch": 7.68, + "learning_rate": 8.490940861361929e-08, + "loss": 0.0005, + "step": 58613 + }, + { + "epoch": 7.68, + "learning_rate": 8.484046034100247e-08, + "loss": 0.0004, + "step": 58614 + }, + { + "epoch": 7.68, + "learning_rate": 8.477153995424259e-08, + "loss": 0.0003, + "step": 58615 + }, + { + "epoch": 7.68, + "learning_rate": 8.470264745353619e-08, + "loss": 0.0005, + "step": 58616 + }, + { + "epoch": 7.68, + "learning_rate": 8.463378283907419e-08, + "loss": 0.0002, + "step": 58617 + }, + { + "epoch": 7.68, + "learning_rate": 8.456494611105203e-08, + "loss": 0.0003, + "step": 58618 + }, + { + "epoch": 7.68, + "learning_rate": 8.449613726966399e-08, + "loss": 0.0002, + "step": 58619 + }, + { + "epoch": 7.68, + "learning_rate": 8.442735631510213e-08, + "loss": 0.0012, + "step": 58620 + }, + { + "epoch": 7.68, + "learning_rate": 8.435860324755962e-08, + "loss": 0.0004, + "step": 58621 + }, + { + "epoch": 7.68, + "learning_rate": 8.428987806723077e-08, + "loss": 0.001, + "step": 58622 + }, + { + "epoch": 7.68, + "learning_rate": 8.422118077430874e-08, + "loss": 0.0003, + "step": 58623 + }, + { + "epoch": 7.68, + "learning_rate": 8.415251136898673e-08, + "loss": 0.0004, + "step": 58624 + }, + { + "epoch": 7.68, + "learning_rate": 8.408386985145678e-08, + "loss": 0.0002, + "step": 58625 + }, + { + "epoch": 7.68, + "learning_rate": 8.401525622191431e-08, + "loss": 0.0006, + "step": 58626 + }, + { + "epoch": 7.68, + "learning_rate": 8.394667048055027e-08, + "loss": 0.0006, + "step": 58627 + }, + { + "epoch": 7.68, + "learning_rate": 8.387811262755784e-08, + "loss": 0.0005, + "step": 58628 + }, + { + "epoch": 7.68, + "learning_rate": 8.380958266313022e-08, + "loss": 0.0004, + "step": 58629 + }, + { + "epoch": 7.68, + "learning_rate": 8.374108058746056e-08, + "loss": 0.0006, + "step": 58630 + }, + { + "epoch": 7.68, + "learning_rate": 8.367260640074093e-08, + "loss": 0.0004, + "step": 58631 + }, + { + "epoch": 7.68, + "learning_rate": 8.360416010316341e-08, + "loss": 0.0003, + "step": 58632 + }, + { + "epoch": 7.68, + "learning_rate": 8.353574169492118e-08, + "loss": 0.0006, + "step": 58633 + }, + { + "epoch": 7.68, + "learning_rate": 8.346735117620741e-08, + "loss": 0.0002, + "step": 58634 + }, + { + "epoch": 7.68, + "learning_rate": 8.339898854721306e-08, + "loss": 0.0006, + "step": 58635 + }, + { + "epoch": 7.68, + "learning_rate": 8.333065380813132e-08, + "loss": 0.0004, + "step": 58636 + }, + { + "epoch": 7.68, + "learning_rate": 8.326234695915314e-08, + "loss": 0.0001, + "step": 58637 + }, + { + "epoch": 7.68, + "learning_rate": 8.319406800047281e-08, + "loss": 0.0002, + "step": 58638 + }, + { + "epoch": 7.68, + "learning_rate": 8.312581693228016e-08, + "loss": 0.0002, + "step": 58639 + }, + { + "epoch": 7.68, + "learning_rate": 8.305759375476841e-08, + "loss": 0.0003, + "step": 58640 + }, + { + "epoch": 7.68, + "learning_rate": 8.298939846812849e-08, + "loss": 0.0006, + "step": 58641 + }, + { + "epoch": 7.68, + "learning_rate": 8.292123107255357e-08, + "loss": 0.0006, + "step": 58642 + }, + { + "epoch": 7.68, + "learning_rate": 8.285309156823352e-08, + "loss": 0.0003, + "step": 58643 + }, + { + "epoch": 7.68, + "learning_rate": 8.278497995536261e-08, + "loss": 0.0006, + "step": 58644 + }, + { + "epoch": 7.68, + "learning_rate": 8.27168962341296e-08, + "loss": 0.0001, + "step": 58645 + }, + { + "epoch": 7.68, + "learning_rate": 8.264884040472764e-08, + "loss": 0.0002, + "step": 58646 + }, + { + "epoch": 7.68, + "learning_rate": 8.258081246734773e-08, + "loss": 0.0003, + "step": 58647 + }, + { + "epoch": 7.68, + "learning_rate": 8.25128124221808e-08, + "loss": 0.0003, + "step": 58648 + }, + { + "epoch": 7.68, + "learning_rate": 8.24448402694189e-08, + "loss": 0.0007, + "step": 58649 + }, + { + "epoch": 7.68, + "learning_rate": 8.237689600925303e-08, + "loss": 0.0004, + "step": 58650 + }, + { + "epoch": 7.68, + "learning_rate": 8.230897964187301e-08, + "loss": 0.0003, + "step": 58651 + }, + { + "epoch": 7.68, + "learning_rate": 8.224109116747092e-08, + "loss": 0.0005, + "step": 58652 + }, + { + "epoch": 7.68, + "learning_rate": 8.217323058623883e-08, + "loss": 0.0003, + "step": 58653 + }, + { + "epoch": 7.68, + "learning_rate": 8.210539789836435e-08, + "loss": 0.0005, + "step": 58654 + }, + { + "epoch": 7.68, + "learning_rate": 8.20375931040418e-08, + "loss": 0.0003, + "step": 58655 + }, + { + "epoch": 7.68, + "learning_rate": 8.19698162034599e-08, + "loss": 0.0004, + "step": 58656 + }, + { + "epoch": 7.68, + "learning_rate": 8.19020671968096e-08, + "loss": 0.0006, + "step": 58657 + }, + { + "epoch": 7.68, + "learning_rate": 8.183434608428187e-08, + "loss": 0.0006, + "step": 58658 + }, + { + "epoch": 7.68, + "learning_rate": 8.176665286606766e-08, + "loss": 0.0005, + "step": 58659 + }, + { + "epoch": 7.68, + "learning_rate": 8.169898754235573e-08, + "loss": 0.0003, + "step": 58660 + }, + { + "epoch": 7.68, + "learning_rate": 8.1631350113337e-08, + "loss": 0.0004, + "step": 58661 + }, + { + "epoch": 7.68, + "learning_rate": 8.156374057920246e-08, + "loss": 0.0003, + "step": 58662 + }, + { + "epoch": 7.68, + "learning_rate": 8.149615894014084e-08, + "loss": 0.001, + "step": 58663 + }, + { + "epoch": 7.68, + "learning_rate": 8.142860519634421e-08, + "loss": 0.0011, + "step": 58664 + }, + { + "epoch": 7.68, + "learning_rate": 8.13610793480002e-08, + "loss": 0.0003, + "step": 58665 + }, + { + "epoch": 7.68, + "learning_rate": 8.129358139530086e-08, + "loss": 0.0009, + "step": 58666 + }, + { + "epoch": 7.68, + "learning_rate": 8.122611133843495e-08, + "loss": 0.0008, + "step": 58667 + }, + { + "epoch": 7.69, + "learning_rate": 8.115866917759119e-08, + "loss": 0.0005, + "step": 58668 + }, + { + "epoch": 7.69, + "learning_rate": 8.109125491296277e-08, + "loss": 0.0002, + "step": 58669 + }, + { + "epoch": 7.69, + "learning_rate": 8.102386854473509e-08, + "loss": 0.0006, + "step": 58670 + }, + { + "epoch": 7.69, + "learning_rate": 8.095651007310135e-08, + "loss": 0.0005, + "step": 58671 + }, + { + "epoch": 7.69, + "learning_rate": 8.088917949824693e-08, + "loss": 0.0004, + "step": 58672 + }, + { + "epoch": 7.69, + "learning_rate": 8.082187682036502e-08, + "loss": 0.0001, + "step": 58673 + }, + { + "epoch": 7.69, + "learning_rate": 8.075460203964324e-08, + "loss": 0.0005, + "step": 58674 + }, + { + "epoch": 7.69, + "learning_rate": 8.068735515627036e-08, + "loss": 0.0009, + "step": 58675 + }, + { + "epoch": 7.69, + "learning_rate": 8.062013617043618e-08, + "loss": 0.0002, + "step": 58676 + }, + { + "epoch": 7.69, + "learning_rate": 8.055294508232946e-08, + "loss": 0.001, + "step": 58677 + }, + { + "epoch": 7.69, + "learning_rate": 8.048578189214007e-08, + "loss": 0.0003, + "step": 58678 + }, + { + "epoch": 7.69, + "learning_rate": 8.041864660005561e-08, + "loss": 0.0004, + "step": 58679 + }, + { + "epoch": 7.69, + "learning_rate": 8.035153920626593e-08, + "loss": 0.0001, + "step": 58680 + }, + { + "epoch": 7.69, + "learning_rate": 8.028445971095866e-08, + "loss": 0.0003, + "step": 58681 + }, + { + "epoch": 7.69, + "learning_rate": 8.021740811432366e-08, + "loss": 0.0013, + "step": 58682 + }, + { + "epoch": 7.69, + "learning_rate": 8.015038441654855e-08, + "loss": 0.0009, + "step": 58683 + }, + { + "epoch": 7.69, + "learning_rate": 8.008338861782317e-08, + "loss": 0.0002, + "step": 58684 + }, + { + "epoch": 7.69, + "learning_rate": 8.001642071833404e-08, + "loss": 0.0002, + "step": 58685 + }, + { + "epoch": 7.69, + "learning_rate": 7.994948071827103e-08, + "loss": 0.0006, + "step": 58686 + }, + { + "epoch": 7.69, + "learning_rate": 7.988256861782063e-08, + "loss": 0.0009, + "step": 58687 + }, + { + "epoch": 7.69, + "learning_rate": 7.981568441717268e-08, + "loss": 0.0002, + "step": 58688 + }, + { + "epoch": 7.69, + "learning_rate": 7.974882811651597e-08, + "loss": 0.0003, + "step": 58689 + }, + { + "epoch": 7.69, + "learning_rate": 7.968199971603586e-08, + "loss": 0.0004, + "step": 58690 + }, + { + "epoch": 7.69, + "learning_rate": 7.961519921592221e-08, + "loss": 0.0009, + "step": 58691 + }, + { + "epoch": 7.69, + "learning_rate": 7.954842661636375e-08, + "loss": 0.0004, + "step": 58692 + }, + { + "epoch": 7.69, + "learning_rate": 7.94816819175459e-08, + "loss": 0.0007, + "step": 58693 + }, + { + "epoch": 7.69, + "learning_rate": 7.941496511965851e-08, + "loss": 0.0005, + "step": 58694 + }, + { + "epoch": 7.69, + "learning_rate": 7.934827622288699e-08, + "loss": 0.0002, + "step": 58695 + }, + { + "epoch": 7.69, + "learning_rate": 7.928161522742117e-08, + "loss": 0.0009, + "step": 58696 + }, + { + "epoch": 7.69, + "learning_rate": 7.921498213344647e-08, + "loss": 0.0004, + "step": 58697 + }, + { + "epoch": 7.69, + "learning_rate": 7.914837694115162e-08, + "loss": 0.0003, + "step": 58698 + }, + { + "epoch": 7.69, + "learning_rate": 7.908179965072315e-08, + "loss": 0.0008, + "step": 58699 + }, + { + "epoch": 7.69, + "learning_rate": 7.90152502623498e-08, + "loss": 0.0005, + "step": 58700 + }, + { + "epoch": 7.69, + "learning_rate": 7.894872877621806e-08, + "loss": 0.0002, + "step": 58701 + }, + { + "epoch": 7.69, + "learning_rate": 7.888223519251337e-08, + "loss": 0.0014, + "step": 58702 + }, + { + "epoch": 7.69, + "learning_rate": 7.881576951142555e-08, + "loss": 0.0004, + "step": 58703 + }, + { + "epoch": 7.69, + "learning_rate": 7.874933173313892e-08, + "loss": 0.0004, + "step": 58704 + }, + { + "epoch": 7.69, + "learning_rate": 7.868292185784221e-08, + "loss": 0.0008, + "step": 58705 + }, + { + "epoch": 7.69, + "learning_rate": 7.861653988572082e-08, + "loss": 0.0004, + "step": 58706 + }, + { + "epoch": 7.69, + "learning_rate": 7.855018581696239e-08, + "loss": 0.0003, + "step": 58707 + }, + { + "epoch": 7.69, + "learning_rate": 7.848385965175343e-08, + "loss": 0.0004, + "step": 58708 + }, + { + "epoch": 7.69, + "learning_rate": 7.841756139027934e-08, + "loss": 0.0002, + "step": 58709 + }, + { + "epoch": 7.69, + "learning_rate": 7.835129103272776e-08, + "loss": 0.0007, + "step": 58710 + }, + { + "epoch": 7.69, + "learning_rate": 7.82850485792852e-08, + "loss": 0.0004, + "step": 58711 + }, + { + "epoch": 7.69, + "learning_rate": 7.821883403013819e-08, + "loss": 0.0004, + "step": 58712 + }, + { + "epoch": 7.69, + "learning_rate": 7.815264738547212e-08, + "loss": 0.0004, + "step": 58713 + }, + { + "epoch": 7.69, + "learning_rate": 7.808648864547242e-08, + "loss": 0.0003, + "step": 58714 + }, + { + "epoch": 7.69, + "learning_rate": 7.802035781032669e-08, + "loss": 0.0003, + "step": 58715 + }, + { + "epoch": 7.69, + "learning_rate": 7.795425488022035e-08, + "loss": 0.0003, + "step": 58716 + }, + { + "epoch": 7.69, + "learning_rate": 7.788817985533881e-08, + "loss": 0.0003, + "step": 58717 + }, + { + "epoch": 7.69, + "learning_rate": 7.782213273586858e-08, + "loss": 0.0002, + "step": 58718 + }, + { + "epoch": 7.69, + "learning_rate": 7.775611352199619e-08, + "loss": 0.0006, + "step": 58719 + }, + { + "epoch": 7.69, + "learning_rate": 7.769012221390481e-08, + "loss": 0.0006, + "step": 58720 + }, + { + "epoch": 7.69, + "learning_rate": 7.76241588117832e-08, + "loss": 0.0008, + "step": 58721 + }, + { + "epoch": 7.69, + "learning_rate": 7.755822331581342e-08, + "loss": 0.0002, + "step": 58722 + }, + { + "epoch": 7.69, + "learning_rate": 7.749231572618421e-08, + "loss": 0.0004, + "step": 58723 + }, + { + "epoch": 7.69, + "learning_rate": 7.742643604307876e-08, + "loss": 0.0002, + "step": 58724 + }, + { + "epoch": 7.69, + "learning_rate": 7.736058426668247e-08, + "loss": 0.0006, + "step": 58725 + }, + { + "epoch": 7.69, + "learning_rate": 7.729476039718298e-08, + "loss": 0.0008, + "step": 58726 + }, + { + "epoch": 7.69, + "learning_rate": 7.722896443476236e-08, + "loss": 0.0005, + "step": 58727 + }, + { + "epoch": 7.69, + "learning_rate": 7.716319637960712e-08, + "loss": 0.0006, + "step": 58728 + }, + { + "epoch": 7.69, + "learning_rate": 7.709745623190157e-08, + "loss": 0.0006, + "step": 58729 + }, + { + "epoch": 7.69, + "learning_rate": 7.70317439918311e-08, + "loss": 0.0002, + "step": 58730 + }, + { + "epoch": 7.69, + "learning_rate": 7.696605965958003e-08, + "loss": 0.0004, + "step": 58731 + }, + { + "epoch": 7.69, + "learning_rate": 7.690040323533487e-08, + "loss": 0.0007, + "step": 58732 + }, + { + "epoch": 7.69, + "learning_rate": 7.68347747192777e-08, + "loss": 0.0004, + "step": 58733 + }, + { + "epoch": 7.69, + "learning_rate": 7.67691741115939e-08, + "loss": 0.0005, + "step": 58734 + }, + { + "epoch": 7.69, + "learning_rate": 7.670360141247002e-08, + "loss": 0.0005, + "step": 58735 + }, + { + "epoch": 7.69, + "learning_rate": 7.663805662208812e-08, + "loss": 0.0009, + "step": 58736 + }, + { + "epoch": 7.69, + "learning_rate": 7.65725397406325e-08, + "loss": 0.0011, + "step": 58737 + }, + { + "epoch": 7.69, + "learning_rate": 7.650705076828858e-08, + "loss": 0.0004, + "step": 58738 + }, + { + "epoch": 7.69, + "learning_rate": 7.644158970524063e-08, + "loss": 0.0002, + "step": 58739 + }, + { + "epoch": 7.69, + "learning_rate": 7.637615655167074e-08, + "loss": 0.0002, + "step": 58740 + }, + { + "epoch": 7.69, + "learning_rate": 7.631075130776544e-08, + "loss": 0.0004, + "step": 58741 + }, + { + "epoch": 7.69, + "learning_rate": 7.62453739737079e-08, + "loss": 0.0002, + "step": 58742 + }, + { + "epoch": 7.69, + "learning_rate": 7.618002454968132e-08, + "loss": 0.0004, + "step": 58743 + }, + { + "epoch": 7.7, + "learning_rate": 7.61147030358711e-08, + "loss": 0.0003, + "step": 58744 + }, + { + "epoch": 7.7, + "learning_rate": 7.60494094324582e-08, + "loss": 0.0004, + "step": 58745 + }, + { + "epoch": 7.7, + "learning_rate": 7.598414373962915e-08, + "loss": 0.0005, + "step": 58746 + }, + { + "epoch": 7.7, + "learning_rate": 7.591890595756601e-08, + "loss": 0.0003, + "step": 58747 + }, + { + "epoch": 7.7, + "learning_rate": 7.585369608645088e-08, + "loss": 0.0003, + "step": 58748 + }, + { + "epoch": 7.7, + "learning_rate": 7.578851412647025e-08, + "loss": 0.0006, + "step": 58749 + }, + { + "epoch": 7.7, + "learning_rate": 7.572336007780623e-08, + "loss": 0.0004, + "step": 58750 + }, + { + "epoch": 7.7, + "learning_rate": 7.565823394064087e-08, + "loss": 0.0005, + "step": 58751 + }, + { + "epoch": 7.7, + "learning_rate": 7.559313571515847e-08, + "loss": 0.0004, + "step": 58752 + }, + { + "epoch": 7.7, + "learning_rate": 7.552806540154111e-08, + "loss": 0.0005, + "step": 58753 + }, + { + "epoch": 7.7, + "learning_rate": 7.546302299997421e-08, + "loss": 0.0007, + "step": 58754 + }, + { + "epoch": 7.7, + "learning_rate": 7.539800851063872e-08, + "loss": 0.0004, + "step": 58755 + }, + { + "epoch": 7.7, + "learning_rate": 7.533302193371672e-08, + "loss": 0.0041, + "step": 58756 + }, + { + "epoch": 7.7, + "learning_rate": 7.526806326939252e-08, + "loss": 0.0003, + "step": 58757 + }, + { + "epoch": 7.7, + "learning_rate": 7.520313251784928e-08, + "loss": 0.0005, + "step": 58758 + }, + { + "epoch": 7.7, + "learning_rate": 7.513822967926798e-08, + "loss": 0.0007, + "step": 58759 + }, + { + "epoch": 7.7, + "learning_rate": 7.507335475383071e-08, + "loss": 0.0007, + "step": 58760 + }, + { + "epoch": 7.7, + "learning_rate": 7.500850774172285e-08, + "loss": 0.0004, + "step": 58761 + }, + { + "epoch": 7.7, + "learning_rate": 7.494368864312429e-08, + "loss": 0.0002, + "step": 58762 + }, + { + "epoch": 7.7, + "learning_rate": 7.48788974582182e-08, + "loss": 0.0002, + "step": 58763 + }, + { + "epoch": 7.7, + "learning_rate": 7.481413418718664e-08, + "loss": 0.0003, + "step": 58764 + }, + { + "epoch": 7.7, + "learning_rate": 7.474939883021171e-08, + "loss": 0.0004, + "step": 58765 + }, + { + "epoch": 7.7, + "learning_rate": 7.468469138747547e-08, + "loss": 0.0007, + "step": 58766 + }, + { + "epoch": 7.7, + "learning_rate": 7.462001185916001e-08, + "loss": 0.0004, + "step": 58767 + }, + { + "epoch": 7.7, + "learning_rate": 7.45553602454474e-08, + "loss": 0.0004, + "step": 58768 + }, + { + "epoch": 7.7, + "learning_rate": 7.449073654651972e-08, + "loss": 0.0004, + "step": 58769 + }, + { + "epoch": 7.7, + "learning_rate": 7.442614076255794e-08, + "loss": 0.0003, + "step": 58770 + }, + { + "epoch": 7.7, + "learning_rate": 7.436157289374301e-08, + "loss": 0.0006, + "step": 58771 + }, + { + "epoch": 7.7, + "learning_rate": 7.429703294025925e-08, + "loss": 0.0006, + "step": 58772 + }, + { + "epoch": 7.7, + "learning_rate": 7.423252090228539e-08, + "loss": 0.0003, + "step": 58773 + }, + { + "epoch": 7.7, + "learning_rate": 7.41680367800035e-08, + "loss": 0.0002, + "step": 58774 + }, + { + "epoch": 7.7, + "learning_rate": 7.410358057359679e-08, + "loss": 0.0002, + "step": 58775 + }, + { + "epoch": 7.7, + "learning_rate": 7.403915228324398e-08, + "loss": 0.0003, + "step": 58776 + }, + { + "epoch": 7.7, + "learning_rate": 7.397475190912828e-08, + "loss": 0.0005, + "step": 58777 + }, + { + "epoch": 7.7, + "learning_rate": 7.391037945143065e-08, + "loss": 0.0006, + "step": 58778 + }, + { + "epoch": 7.7, + "learning_rate": 7.384603491032982e-08, + "loss": 0.0005, + "step": 58779 + }, + { + "epoch": 7.7, + "learning_rate": 7.378171828601011e-08, + "loss": 0.001, + "step": 58780 + }, + { + "epoch": 7.7, + "learning_rate": 7.371742957864913e-08, + "loss": 0.0005, + "step": 58781 + }, + { + "epoch": 7.7, + "learning_rate": 7.365316878843009e-08, + "loss": 0.0008, + "step": 58782 + }, + { + "epoch": 7.7, + "learning_rate": 7.358893591553285e-08, + "loss": 0.0004, + "step": 58783 + }, + { + "epoch": 7.7, + "learning_rate": 7.352473096013835e-08, + "loss": 0.0003, + "step": 58784 + }, + { + "epoch": 7.7, + "learning_rate": 7.346055392242645e-08, + "loss": 0.0005, + "step": 58785 + }, + { + "epoch": 7.7, + "learning_rate": 7.339640480257814e-08, + "loss": 0.0002, + "step": 58786 + }, + { + "epoch": 7.7, + "learning_rate": 7.333228360077438e-08, + "loss": 0.0005, + "step": 58787 + }, + { + "epoch": 7.7, + "learning_rate": 7.32681903171939e-08, + "loss": 0.0003, + "step": 58788 + }, + { + "epoch": 7.7, + "learning_rate": 7.32041249520199e-08, + "loss": 0.0004, + "step": 58789 + }, + { + "epoch": 7.7, + "learning_rate": 7.314008750542889e-08, + "loss": 0.0004, + "step": 58790 + }, + { + "epoch": 7.7, + "learning_rate": 7.307607797760297e-08, + "loss": 0.0002, + "step": 58791 + }, + { + "epoch": 7.7, + "learning_rate": 7.301209636872309e-08, + "loss": 0.0004, + "step": 58792 + }, + { + "epoch": 7.7, + "learning_rate": 7.294814267896688e-08, + "loss": 0.0004, + "step": 58793 + }, + { + "epoch": 7.7, + "learning_rate": 7.288421690851533e-08, + "loss": 0.0004, + "step": 58794 + }, + { + "epoch": 7.7, + "learning_rate": 7.282031905754938e-08, + "loss": 0.0003, + "step": 58795 + }, + { + "epoch": 7.7, + "learning_rate": 7.275644912624668e-08, + "loss": 0.0013, + "step": 58796 + }, + { + "epoch": 7.7, + "learning_rate": 7.26926071147882e-08, + "loss": 0.0004, + "step": 58797 + }, + { + "epoch": 7.7, + "learning_rate": 7.262879302335268e-08, + "loss": 0.0004, + "step": 58798 + }, + { + "epoch": 7.7, + "learning_rate": 7.256500685211997e-08, + "loss": 0.0004, + "step": 58799 + }, + { + "epoch": 7.7, + "learning_rate": 7.250124860126995e-08, + "loss": 0.0007, + "step": 58800 + }, + { + "epoch": 7.7, + "learning_rate": 7.243751827098133e-08, + "loss": 0.0002, + "step": 58801 + }, + { + "epoch": 7.7, + "learning_rate": 7.237381586143399e-08, + "loss": 0.0002, + "step": 58802 + }, + { + "epoch": 7.7, + "learning_rate": 7.231014137280556e-08, + "loss": 0.0005, + "step": 58803 + }, + { + "epoch": 7.7, + "learning_rate": 7.224649480527702e-08, + "loss": 0.0004, + "step": 58804 + }, + { + "epoch": 7.7, + "learning_rate": 7.218287615902597e-08, + "loss": 0.0005, + "step": 58805 + }, + { + "epoch": 7.7, + "learning_rate": 7.21192854342323e-08, + "loss": 0.0003, + "step": 58806 + }, + { + "epoch": 7.7, + "learning_rate": 7.205572263107474e-08, + "loss": 0.0004, + "step": 58807 + }, + { + "epoch": 7.7, + "learning_rate": 7.199218774973094e-08, + "loss": 0.0005, + "step": 58808 + }, + { + "epoch": 7.7, + "learning_rate": 7.192868079038185e-08, + "loss": 0.0002, + "step": 58809 + }, + { + "epoch": 7.7, + "learning_rate": 7.186520175320289e-08, + "loss": 0.0001, + "step": 58810 + }, + { + "epoch": 7.7, + "learning_rate": 7.180175063837503e-08, + "loss": 0.0003, + "step": 58811 + }, + { + "epoch": 7.7, + "learning_rate": 7.173832744607701e-08, + "loss": 0.0002, + "step": 58812 + }, + { + "epoch": 7.7, + "learning_rate": 7.167493217648424e-08, + "loss": 0.0004, + "step": 58813 + }, + { + "epoch": 7.7, + "learning_rate": 7.161156482977882e-08, + "loss": 0.0002, + "step": 58814 + }, + { + "epoch": 7.7, + "learning_rate": 7.154822540613615e-08, + "loss": 0.0001, + "step": 58815 + }, + { + "epoch": 7.7, + "learning_rate": 7.148491390573609e-08, + "loss": 0.0015, + "step": 58816 + }, + { + "epoch": 7.7, + "learning_rate": 7.142163032875515e-08, + "loss": 0.0015, + "step": 58817 + }, + { + "epoch": 7.7, + "learning_rate": 7.13583746753721e-08, + "loss": 0.0004, + "step": 58818 + }, + { + "epoch": 7.7, + "learning_rate": 7.129514694576567e-08, + "loss": 0.0003, + "step": 58819 + }, + { + "epoch": 7.71, + "learning_rate": 7.12319471401135e-08, + "loss": 0.0004, + "step": 58820 + }, + { + "epoch": 7.71, + "learning_rate": 7.1168775258591e-08, + "loss": 0.0005, + "step": 58821 + }, + { + "epoch": 7.71, + "learning_rate": 7.110563130137915e-08, + "loss": 0.0003, + "step": 58822 + }, + { + "epoch": 7.71, + "learning_rate": 7.104251526865336e-08, + "loss": 0.0005, + "step": 58823 + }, + { + "epoch": 7.71, + "learning_rate": 7.097942716059126e-08, + "loss": 0.0009, + "step": 58824 + }, + { + "epoch": 7.71, + "learning_rate": 7.091636697737159e-08, + "loss": 0.0003, + "step": 58825 + }, + { + "epoch": 7.71, + "learning_rate": 7.085333471916978e-08, + "loss": 0.0004, + "step": 58826 + }, + { + "epoch": 7.71, + "learning_rate": 7.079033038616568e-08, + "loss": 0.0004, + "step": 58827 + }, + { + "epoch": 7.71, + "learning_rate": 7.072735397853359e-08, + "loss": 0.0003, + "step": 58828 + }, + { + "epoch": 7.71, + "learning_rate": 7.066440549645226e-08, + "loss": 0.0005, + "step": 58829 + }, + { + "epoch": 7.71, + "learning_rate": 7.060148494009934e-08, + "loss": 0.0007, + "step": 58830 + }, + { + "epoch": 7.71, + "learning_rate": 7.053859230965021e-08, + "loss": 0.0005, + "step": 58831 + }, + { + "epoch": 7.71, + "learning_rate": 7.047572760528254e-08, + "loss": 0.0002, + "step": 58832 + }, + { + "epoch": 7.71, + "learning_rate": 7.041289082717395e-08, + "loss": 0.0005, + "step": 58833 + }, + { + "epoch": 7.71, + "learning_rate": 7.035008197549986e-08, + "loss": 0.0002, + "step": 58834 + }, + { + "epoch": 7.71, + "learning_rate": 7.02873010504379e-08, + "loss": 0.0009, + "step": 58835 + }, + { + "epoch": 7.71, + "learning_rate": 7.022454805216239e-08, + "loss": 0.0002, + "step": 58836 + }, + { + "epoch": 7.71, + "learning_rate": 7.016182298085316e-08, + "loss": 0.0001, + "step": 58837 + }, + { + "epoch": 7.71, + "learning_rate": 7.009912583668455e-08, + "loss": 0.0003, + "step": 58838 + }, + { + "epoch": 7.71, + "learning_rate": 7.003645661983415e-08, + "loss": 0.0004, + "step": 58839 + }, + { + "epoch": 7.71, + "learning_rate": 6.99738153304752e-08, + "loss": 0.0004, + "step": 58840 + }, + { + "epoch": 7.71, + "learning_rate": 6.991120196878754e-08, + "loss": 0.0004, + "step": 58841 + }, + { + "epoch": 7.71, + "learning_rate": 6.984861653494657e-08, + "loss": 0.0014, + "step": 58842 + }, + { + "epoch": 7.71, + "learning_rate": 6.978605902912661e-08, + "loss": 0.0003, + "step": 58843 + }, + { + "epoch": 7.71, + "learning_rate": 6.972352945150418e-08, + "loss": 0.0002, + "step": 58844 + }, + { + "epoch": 7.71, + "learning_rate": 6.966102780225692e-08, + "loss": 0.0004, + "step": 58845 + }, + { + "epoch": 7.71, + "learning_rate": 6.959855408155802e-08, + "loss": 0.0003, + "step": 58846 + }, + { + "epoch": 7.71, + "learning_rate": 6.953610828958401e-08, + "loss": 0.0003, + "step": 58847 + }, + { + "epoch": 7.71, + "learning_rate": 6.947369042651253e-08, + "loss": 0.0003, + "step": 58848 + }, + { + "epoch": 7.71, + "learning_rate": 6.941130049251566e-08, + "loss": 0.0004, + "step": 58849 + }, + { + "epoch": 7.71, + "learning_rate": 6.934893848777214e-08, + "loss": 0.0004, + "step": 58850 + }, + { + "epoch": 7.71, + "learning_rate": 6.928660441245516e-08, + "loss": 0.0004, + "step": 58851 + }, + { + "epoch": 7.71, + "learning_rate": 6.922429826674016e-08, + "loss": 0.0009, + "step": 58852 + }, + { + "epoch": 7.71, + "learning_rate": 6.916202005080363e-08, + "loss": 0.0006, + "step": 58853 + }, + { + "epoch": 7.71, + "learning_rate": 6.909976976481992e-08, + "loss": 0.0002, + "step": 58854 + }, + { + "epoch": 7.71, + "learning_rate": 6.90375474089644e-08, + "loss": 0.0002, + "step": 58855 + }, + { + "epoch": 7.71, + "learning_rate": 6.897535298341252e-08, + "loss": 0.0003, + "step": 58856 + }, + { + "epoch": 7.71, + "learning_rate": 6.891318648833745e-08, + "loss": 0.0003, + "step": 58857 + }, + { + "epoch": 7.71, + "learning_rate": 6.885104792391684e-08, + "loss": 0.0006, + "step": 58858 + }, + { + "epoch": 7.71, + "learning_rate": 6.878893729032166e-08, + "loss": 0.0007, + "step": 58859 + }, + { + "epoch": 7.71, + "learning_rate": 6.872685458772954e-08, + "loss": 0.0005, + "step": 58860 + }, + { + "epoch": 7.71, + "learning_rate": 6.866479981631479e-08, + "loss": 0.0002, + "step": 58861 + }, + { + "epoch": 7.71, + "learning_rate": 6.860277297625173e-08, + "loss": 0.0004, + "step": 58862 + }, + { + "epoch": 7.71, + "learning_rate": 6.854077406771243e-08, + "loss": 0.0003, + "step": 58863 + }, + { + "epoch": 7.71, + "learning_rate": 6.847880309087562e-08, + "loss": 0.0007, + "step": 58864 + }, + { + "epoch": 7.71, + "learning_rate": 6.84168600459123e-08, + "loss": 0.0004, + "step": 58865 + }, + { + "epoch": 7.71, + "learning_rate": 6.835494493299789e-08, + "loss": 0.0003, + "step": 58866 + }, + { + "epoch": 7.71, + "learning_rate": 6.829305775230666e-08, + "loss": 0.0002, + "step": 58867 + }, + { + "epoch": 7.71, + "learning_rate": 6.823119850401183e-08, + "loss": 0.0003, + "step": 58868 + }, + { + "epoch": 7.71, + "learning_rate": 6.816936718828771e-08, + "loss": 0.0005, + "step": 58869 + }, + { + "epoch": 7.71, + "learning_rate": 6.810756380530859e-08, + "loss": 0.0002, + "step": 58870 + }, + { + "epoch": 7.71, + "learning_rate": 6.804578835524767e-08, + "loss": 0.0002, + "step": 58871 + }, + { + "epoch": 7.71, + "learning_rate": 6.798404083827925e-08, + "loss": 0.0006, + "step": 58872 + }, + { + "epoch": 7.71, + "learning_rate": 6.792232125457654e-08, + "loss": 0.0003, + "step": 58873 + }, + { + "epoch": 7.71, + "learning_rate": 6.786062960431495e-08, + "loss": 0.0007, + "step": 58874 + }, + { + "epoch": 7.71, + "learning_rate": 6.779896588766433e-08, + "loss": 0.0003, + "step": 58875 + }, + { + "epoch": 7.71, + "learning_rate": 6.773733010480122e-08, + "loss": 0.0002, + "step": 58876 + }, + { + "epoch": 7.71, + "learning_rate": 6.76757222558977e-08, + "loss": 0.0003, + "step": 58877 + }, + { + "epoch": 7.71, + "learning_rate": 6.761414234112807e-08, + "loss": 0.0003, + "step": 58878 + }, + { + "epoch": 7.71, + "learning_rate": 6.755259036066331e-08, + "loss": 0.0003, + "step": 58879 + }, + { + "epoch": 7.71, + "learning_rate": 6.749106631467883e-08, + "loss": 0.0004, + "step": 58880 + }, + { + "epoch": 7.71, + "learning_rate": 6.742957020334784e-08, + "loss": 0.0007, + "step": 58881 + }, + { + "epoch": 7.71, + "learning_rate": 6.736810202684018e-08, + "loss": 0.0005, + "step": 58882 + }, + { + "epoch": 7.71, + "learning_rate": 6.73066617853324e-08, + "loss": 0.0002, + "step": 58883 + }, + { + "epoch": 7.71, + "learning_rate": 6.724524947899436e-08, + "loss": 0.0002, + "step": 58884 + }, + { + "epoch": 7.71, + "learning_rate": 6.718386510800034e-08, + "loss": 0.0002, + "step": 58885 + }, + { + "epoch": 7.71, + "learning_rate": 6.712250867252357e-08, + "loss": 0.0002, + "step": 58886 + }, + { + "epoch": 7.71, + "learning_rate": 6.706118017273499e-08, + "loss": 0.0004, + "step": 58887 + }, + { + "epoch": 7.71, + "learning_rate": 6.699987960880894e-08, + "loss": 0.0005, + "step": 58888 + }, + { + "epoch": 7.71, + "learning_rate": 6.693860698091637e-08, + "loss": 0.0007, + "step": 58889 + }, + { + "epoch": 7.71, + "learning_rate": 6.687736228922937e-08, + "loss": 0.0006, + "step": 58890 + }, + { + "epoch": 7.71, + "learning_rate": 6.681614553392113e-08, + "loss": 0.0004, + "step": 58891 + }, + { + "epoch": 7.71, + "learning_rate": 6.675495671516374e-08, + "loss": 0.0004, + "step": 58892 + }, + { + "epoch": 7.71, + "learning_rate": 6.669379583312819e-08, + "loss": 0.0001, + "step": 58893 + }, + { + "epoch": 7.71, + "learning_rate": 6.663266288798764e-08, + "loss": 0.0003, + "step": 58894 + }, + { + "epoch": 7.71, + "learning_rate": 6.65715578799142e-08, + "loss": 0.0005, + "step": 58895 + }, + { + "epoch": 7.71, + "learning_rate": 6.651048080907995e-08, + "loss": 0.0005, + "step": 58896 + }, + { + "epoch": 7.72, + "learning_rate": 6.644943167565476e-08, + "loss": 0.0005, + "step": 58897 + }, + { + "epoch": 7.72, + "learning_rate": 6.638841047981182e-08, + "loss": 0.0002, + "step": 58898 + }, + { + "epoch": 7.72, + "learning_rate": 6.63274172217232e-08, + "loss": 0.0004, + "step": 58899 + }, + { + "epoch": 7.72, + "learning_rate": 6.626645190155877e-08, + "loss": 0.0005, + "step": 58900 + }, + { + "epoch": 7.72, + "learning_rate": 6.620551451949175e-08, + "loss": 0.0003, + "step": 58901 + }, + { + "epoch": 7.72, + "learning_rate": 6.614460507569199e-08, + "loss": 0.0002, + "step": 58902 + }, + { + "epoch": 7.72, + "learning_rate": 6.608372357033265e-08, + "loss": 0.0002, + "step": 58903 + }, + { + "epoch": 7.72, + "learning_rate": 6.602287000358254e-08, + "loss": 0.0007, + "step": 58904 + }, + { + "epoch": 7.72, + "learning_rate": 6.596204437561371e-08, + "loss": 0.0007, + "step": 58905 + }, + { + "epoch": 7.72, + "learning_rate": 6.590124668659936e-08, + "loss": 0.0004, + "step": 58906 + }, + { + "epoch": 7.72, + "learning_rate": 6.584047693670714e-08, + "loss": 0.0004, + "step": 58907 + }, + { + "epoch": 7.72, + "learning_rate": 6.577973512611135e-08, + "loss": 0.0004, + "step": 58908 + }, + { + "epoch": 7.72, + "learning_rate": 6.571902125497965e-08, + "loss": 0.0006, + "step": 58909 + }, + { + "epoch": 7.72, + "learning_rate": 6.56583353234841e-08, + "loss": 0.0003, + "step": 58910 + }, + { + "epoch": 7.72, + "learning_rate": 6.559767733179568e-08, + "loss": 0.0008, + "step": 58911 + }, + { + "epoch": 7.72, + "learning_rate": 6.553704728008537e-08, + "loss": 0.0005, + "step": 58912 + }, + { + "epoch": 7.72, + "learning_rate": 6.547644516852191e-08, + "loss": 0.0004, + "step": 58913 + }, + { + "epoch": 7.72, + "learning_rate": 6.541587099727853e-08, + "loss": 0.0005, + "step": 58914 + }, + { + "epoch": 7.72, + "learning_rate": 6.535532476652284e-08, + "loss": 0.0003, + "step": 58915 + }, + { + "epoch": 7.72, + "learning_rate": 6.529480647642583e-08, + "loss": 0.0004, + "step": 58916 + }, + { + "epoch": 7.72, + "learning_rate": 6.523431612715847e-08, + "loss": 0.0005, + "step": 58917 + }, + { + "epoch": 7.72, + "learning_rate": 6.517385371889173e-08, + "loss": 0.0004, + "step": 58918 + }, + { + "epoch": 7.72, + "learning_rate": 6.511341925179327e-08, + "loss": 0.0006, + "step": 58919 + }, + { + "epoch": 7.72, + "learning_rate": 6.505301272603404e-08, + "loss": 0.0003, + "step": 58920 + }, + { + "epoch": 7.72, + "learning_rate": 6.499263414178503e-08, + "loss": 0.0002, + "step": 58921 + }, + { + "epoch": 7.72, + "learning_rate": 6.493228349921498e-08, + "loss": 0.0003, + "step": 58922 + }, + { + "epoch": 7.72, + "learning_rate": 6.487196079849378e-08, + "loss": 0.0004, + "step": 58923 + }, + { + "epoch": 7.72, + "learning_rate": 6.481166603979017e-08, + "loss": 0.001, + "step": 58924 + }, + { + "epoch": 7.72, + "learning_rate": 6.475139922327622e-08, + "loss": 0.0005, + "step": 58925 + }, + { + "epoch": 7.72, + "learning_rate": 6.469116034911849e-08, + "loss": 0.0004, + "step": 58926 + }, + { + "epoch": 7.72, + "learning_rate": 6.463094941748793e-08, + "loss": 0.0006, + "step": 58927 + }, + { + "epoch": 7.72, + "learning_rate": 6.457076642855441e-08, + "loss": 0.0002, + "step": 58928 + }, + { + "epoch": 7.72, + "learning_rate": 6.451061138248671e-08, + "loss": 0.0009, + "step": 58929 + }, + { + "epoch": 7.72, + "learning_rate": 6.445048427945355e-08, + "loss": 0.0005, + "step": 58930 + }, + { + "epoch": 7.72, + "learning_rate": 6.43903851196237e-08, + "loss": 0.0006, + "step": 58931 + }, + { + "epoch": 7.72, + "learning_rate": 6.433031390316702e-08, + "loss": 0.001, + "step": 58932 + }, + { + "epoch": 7.72, + "learning_rate": 6.427027063025226e-08, + "loss": 0.0002, + "step": 58933 + }, + { + "epoch": 7.72, + "learning_rate": 6.421025530104929e-08, + "loss": 0.0006, + "step": 58934 + }, + { + "epoch": 7.72, + "learning_rate": 6.415026791572465e-08, + "loss": 0.0005, + "step": 58935 + }, + { + "epoch": 7.72, + "learning_rate": 6.409030847444819e-08, + "loss": 0.0003, + "step": 58936 + }, + { + "epoch": 7.72, + "learning_rate": 6.403037697738978e-08, + "loss": 0.0002, + "step": 58937 + }, + { + "epoch": 7.72, + "learning_rate": 6.397047342471596e-08, + "loss": 0.0002, + "step": 58938 + }, + { + "epoch": 7.72, + "learning_rate": 6.391059781659658e-08, + "loss": 0.0003, + "step": 58939 + }, + { + "epoch": 7.72, + "learning_rate": 6.385075015319819e-08, + "loss": 0.0005, + "step": 58940 + }, + { + "epoch": 7.72, + "learning_rate": 6.379093043469175e-08, + "loss": 0.0007, + "step": 58941 + }, + { + "epoch": 7.72, + "learning_rate": 6.37311386612438e-08, + "loss": 0.0002, + "step": 58942 + }, + { + "epoch": 7.72, + "learning_rate": 6.36713748330231e-08, + "loss": 0.0006, + "step": 58943 + }, + { + "epoch": 7.72, + "learning_rate": 6.361163895019728e-08, + "loss": 0.0005, + "step": 58944 + }, + { + "epoch": 7.72, + "learning_rate": 6.35519310129351e-08, + "loss": 0.0001, + "step": 58945 + }, + { + "epoch": 7.72, + "learning_rate": 6.34922510214031e-08, + "loss": 0.0003, + "step": 58946 + }, + { + "epoch": 7.72, + "learning_rate": 6.343259897577004e-08, + "loss": 0.0007, + "step": 58947 + }, + { + "epoch": 7.72, + "learning_rate": 6.337297487620464e-08, + "loss": 0.0002, + "step": 58948 + }, + { + "epoch": 7.72, + "learning_rate": 6.331337872287236e-08, + "loss": 0.0004, + "step": 58949 + }, + { + "epoch": 7.72, + "learning_rate": 6.325381051594303e-08, + "loss": 0.0001, + "step": 58950 + }, + { + "epoch": 7.72, + "learning_rate": 6.319427025558211e-08, + "loss": 0.0004, + "step": 58951 + }, + { + "epoch": 7.72, + "learning_rate": 6.313475794195834e-08, + "loss": 0.0002, + "step": 58952 + }, + { + "epoch": 7.72, + "learning_rate": 6.307527357524045e-08, + "loss": 0.001, + "step": 58953 + }, + { + "epoch": 7.72, + "learning_rate": 6.301581715559279e-08, + "loss": 0.0003, + "step": 58954 + }, + { + "epoch": 7.72, + "learning_rate": 6.29563886831841e-08, + "loss": 0.0002, + "step": 58955 + }, + { + "epoch": 7.72, + "learning_rate": 6.289698815818202e-08, + "loss": 0.0002, + "step": 58956 + }, + { + "epoch": 7.72, + "learning_rate": 6.283761558075197e-08, + "loss": 0.0005, + "step": 58957 + }, + { + "epoch": 7.72, + "learning_rate": 6.277827095106271e-08, + "loss": 0.0004, + "step": 58958 + }, + { + "epoch": 7.72, + "learning_rate": 6.271895426927965e-08, + "loss": 0.0002, + "step": 58959 + }, + { + "epoch": 7.72, + "learning_rate": 6.265966553557046e-08, + "loss": 0.0005, + "step": 58960 + }, + { + "epoch": 7.72, + "learning_rate": 6.260040475010276e-08, + "loss": 0.0005, + "step": 58961 + }, + { + "epoch": 7.72, + "learning_rate": 6.254117191304088e-08, + "loss": 0.0004, + "step": 58962 + }, + { + "epoch": 7.72, + "learning_rate": 6.248196702455356e-08, + "loss": 0.0004, + "step": 58963 + }, + { + "epoch": 7.72, + "learning_rate": 6.242279008480734e-08, + "loss": 0.0006, + "step": 58964 + }, + { + "epoch": 7.72, + "learning_rate": 6.236364109396654e-08, + "loss": 0.0005, + "step": 58965 + }, + { + "epoch": 7.72, + "learning_rate": 6.230452005219878e-08, + "loss": 0.0004, + "step": 58966 + }, + { + "epoch": 7.72, + "learning_rate": 6.224542695967173e-08, + "loss": 0.0003, + "step": 58967 + }, + { + "epoch": 7.72, + "learning_rate": 6.218636181654857e-08, + "loss": 0.0004, + "step": 58968 + }, + { + "epoch": 7.72, + "learning_rate": 6.212732462299809e-08, + "loss": 0.0007, + "step": 58969 + }, + { + "epoch": 7.72, + "learning_rate": 6.206831537918456e-08, + "loss": 0.0007, + "step": 58970 + }, + { + "epoch": 7.72, + "learning_rate": 6.200933408527454e-08, + "loss": 0.0009, + "step": 58971 + }, + { + "epoch": 7.72, + "learning_rate": 6.195038074143455e-08, + "loss": 0.0003, + "step": 58972 + }, + { + "epoch": 7.73, + "learning_rate": 6.189145534783003e-08, + "loss": 0.0002, + "step": 58973 + }, + { + "epoch": 7.73, + "learning_rate": 6.183255790462639e-08, + "loss": 0.0002, + "step": 58974 + }, + { + "epoch": 7.73, + "learning_rate": 6.177368841198906e-08, + "loss": 0.0002, + "step": 58975 + }, + { + "epoch": 7.73, + "learning_rate": 6.171484687008456e-08, + "loss": 0.0005, + "step": 58976 + }, + { + "epoch": 7.73, + "learning_rate": 6.16560332790772e-08, + "loss": 0.0003, + "step": 58977 + }, + { + "epoch": 7.73, + "learning_rate": 6.159724763913356e-08, + "loss": 0.0009, + "step": 58978 + }, + { + "epoch": 7.73, + "learning_rate": 6.153848995041789e-08, + "loss": 0.0005, + "step": 58979 + }, + { + "epoch": 7.73, + "learning_rate": 6.147976021309676e-08, + "loss": 0.0002, + "step": 58980 + }, + { + "epoch": 7.73, + "learning_rate": 6.142105842733448e-08, + "loss": 0.001, + "step": 58981 + }, + { + "epoch": 7.73, + "learning_rate": 6.136238459329535e-08, + "loss": 0.0004, + "step": 58982 + }, + { + "epoch": 7.73, + "learning_rate": 6.130373871114592e-08, + "loss": 0.0008, + "step": 58983 + }, + { + "epoch": 7.73, + "learning_rate": 6.124512078104939e-08, + "loss": 0.0005, + "step": 58984 + }, + { + "epoch": 7.73, + "learning_rate": 6.118653080317228e-08, + "loss": 0.0002, + "step": 58985 + }, + { + "epoch": 7.73, + "learning_rate": 6.112796877768001e-08, + "loss": 0.0004, + "step": 58986 + }, + { + "epoch": 7.73, + "learning_rate": 6.10694347047347e-08, + "loss": 0.0004, + "step": 58987 + }, + { + "epoch": 7.73, + "learning_rate": 6.101092858450286e-08, + "loss": 0.0002, + "step": 58988 + }, + { + "epoch": 7.73, + "learning_rate": 6.09524504171477e-08, + "loss": 0.0005, + "step": 58989 + }, + { + "epoch": 7.73, + "learning_rate": 6.089400020283575e-08, + "loss": 0.0005, + "step": 58990 + }, + { + "epoch": 7.73, + "learning_rate": 6.083557794173023e-08, + "loss": 0.0004, + "step": 58991 + }, + { + "epoch": 7.73, + "learning_rate": 6.077718363399432e-08, + "loss": 0.0005, + "step": 58992 + }, + { + "epoch": 7.73, + "learning_rate": 6.071881727979235e-08, + "loss": 0.0005, + "step": 58993 + }, + { + "epoch": 7.73, + "learning_rate": 6.066047887929084e-08, + "loss": 0.0002, + "step": 58994 + }, + { + "epoch": 7.73, + "learning_rate": 6.060216843265298e-08, + "loss": 0.0003, + "step": 58995 + }, + { + "epoch": 7.73, + "learning_rate": 6.054388594004202e-08, + "loss": 0.0007, + "step": 58996 + }, + { + "epoch": 7.73, + "learning_rate": 6.048563140162111e-08, + "loss": 0.0002, + "step": 58997 + }, + { + "epoch": 7.73, + "learning_rate": 6.042740481755571e-08, + "loss": 0.0003, + "step": 58998 + }, + { + "epoch": 7.73, + "learning_rate": 6.03692061880079e-08, + "loss": 0.0003, + "step": 58999 + }, + { + "epoch": 7.73, + "learning_rate": 6.03110355131431e-08, + "loss": 0.0002, + "step": 59000 + }, + { + "epoch": 7.73, + "learning_rate": 6.025289279312452e-08, + "loss": 0.0004, + "step": 59001 + }, + { + "epoch": 7.73, + "learning_rate": 6.019477802811536e-08, + "loss": 0.0005, + "step": 59002 + }, + { + "epoch": 7.73, + "learning_rate": 6.013669121827881e-08, + "loss": 0.002, + "step": 59003 + }, + { + "epoch": 7.73, + "learning_rate": 6.007863236377809e-08, + "loss": 0.0002, + "step": 59004 + }, + { + "epoch": 7.73, + "learning_rate": 6.00206014647775e-08, + "loss": 0.0003, + "step": 59005 + }, + { + "epoch": 7.73, + "learning_rate": 5.996259852143915e-08, + "loss": 0.0003, + "step": 59006 + }, + { + "epoch": 7.73, + "learning_rate": 5.990462353392734e-08, + "loss": 0.0004, + "step": 59007 + }, + { + "epoch": 7.73, + "learning_rate": 5.984667650240417e-08, + "loss": 0.0006, + "step": 59008 + }, + { + "epoch": 7.73, + "learning_rate": 5.978875742703282e-08, + "loss": 0.0006, + "step": 59009 + }, + { + "epoch": 7.73, + "learning_rate": 5.973086630797542e-08, + "loss": 0.0007, + "step": 59010 + }, + { + "epoch": 7.73, + "learning_rate": 5.967300314539737e-08, + "loss": 0.0004, + "step": 59011 + }, + { + "epoch": 7.73, + "learning_rate": 5.961516793945743e-08, + "loss": 0.0003, + "step": 59012 + }, + { + "epoch": 7.73, + "learning_rate": 5.955736069032214e-08, + "loss": 0.0005, + "step": 59013 + }, + { + "epoch": 7.73, + "learning_rate": 5.949958139815248e-08, + "loss": 0.0026, + "step": 59014 + }, + { + "epoch": 7.73, + "learning_rate": 5.944183006310944e-08, + "loss": 0.0004, + "step": 59015 + }, + { + "epoch": 7.73, + "learning_rate": 5.9384106685357325e-08, + "loss": 0.0004, + "step": 59016 + }, + { + "epoch": 7.73, + "learning_rate": 5.9326411265059333e-08, + "loss": 0.0002, + "step": 59017 + }, + { + "epoch": 7.73, + "learning_rate": 5.926874380237535e-08, + "loss": 0.0007, + "step": 59018 + }, + { + "epoch": 7.73, + "learning_rate": 5.921110429746857e-08, + "loss": 0.0002, + "step": 59019 + }, + { + "epoch": 7.73, + "learning_rate": 5.915349275050219e-08, + "loss": 0.0002, + "step": 59020 + }, + { + "epoch": 7.73, + "learning_rate": 5.90959091616361e-08, + "loss": 0.0009, + "step": 59021 + }, + { + "epoch": 7.73, + "learning_rate": 5.90383535310346e-08, + "loss": 0.0009, + "step": 59022 + }, + { + "epoch": 7.73, + "learning_rate": 5.898082585885645e-08, + "loss": 0.0002, + "step": 59023 + }, + { + "epoch": 7.73, + "learning_rate": 5.892332614526708e-08, + "loss": 0.0002, + "step": 59024 + }, + { + "epoch": 7.73, + "learning_rate": 5.8865854390425245e-08, + "loss": 0.0005, + "step": 59025 + }, + { + "epoch": 7.73, + "learning_rate": 5.8808410594495267e-08, + "loss": 0.0004, + "step": 59026 + }, + { + "epoch": 7.73, + "learning_rate": 5.8750994757635905e-08, + "loss": 0.0006, + "step": 59027 + }, + { + "epoch": 7.73, + "learning_rate": 5.869360688001036e-08, + "loss": 0.0003, + "step": 59028 + }, + { + "epoch": 7.73, + "learning_rate": 5.863624696177961e-08, + "loss": 0.0003, + "step": 59029 + }, + { + "epoch": 7.73, + "learning_rate": 5.857891500310575e-08, + "loss": 0.0011, + "step": 59030 + }, + { + "epoch": 7.73, + "learning_rate": 5.852161100414755e-08, + "loss": 0.0004, + "step": 59031 + }, + { + "epoch": 7.73, + "learning_rate": 5.846433496506931e-08, + "loss": 0.0002, + "step": 59032 + }, + { + "epoch": 7.73, + "learning_rate": 5.8407086886030916e-08, + "loss": 0.0001, + "step": 59033 + }, + { + "epoch": 7.73, + "learning_rate": 5.834986676719223e-08, + "loss": 0.0003, + "step": 59034 + }, + { + "epoch": 7.73, + "learning_rate": 5.829267460871535e-08, + "loss": 0.0003, + "step": 59035 + }, + { + "epoch": 7.73, + "learning_rate": 5.823551041076125e-08, + "loss": 0.0004, + "step": 59036 + }, + { + "epoch": 7.73, + "learning_rate": 5.817837417348982e-08, + "loss": 0.0009, + "step": 59037 + }, + { + "epoch": 7.73, + "learning_rate": 5.812126589706313e-08, + "loss": 0.0003, + "step": 59038 + }, + { + "epoch": 7.73, + "learning_rate": 5.8064185581639955e-08, + "loss": 0.0004, + "step": 59039 + }, + { + "epoch": 7.73, + "learning_rate": 5.8007133227382385e-08, + "loss": 0.0004, + "step": 59040 + }, + { + "epoch": 7.73, + "learning_rate": 5.7950108834450293e-08, + "loss": 0.0003, + "step": 59041 + }, + { + "epoch": 7.73, + "learning_rate": 5.789311240300355e-08, + "loss": 0.0008, + "step": 59042 + }, + { + "epoch": 7.73, + "learning_rate": 5.783614393320425e-08, + "loss": 0.0003, + "step": 59043 + }, + { + "epoch": 7.73, + "learning_rate": 5.777920342521004e-08, + "loss": 0.0004, + "step": 59044 + }, + { + "epoch": 7.73, + "learning_rate": 5.772229087918302e-08, + "loss": 0.0005, + "step": 59045 + }, + { + "epoch": 7.73, + "learning_rate": 5.7665406295283055e-08, + "loss": 0.0005, + "step": 59046 + }, + { + "epoch": 7.73, + "learning_rate": 5.7608549673667804e-08, + "loss": 0.0005, + "step": 59047 + }, + { + "epoch": 7.73, + "learning_rate": 5.755172101450046e-08, + "loss": 0.0008, + "step": 59048 + }, + { + "epoch": 7.74, + "learning_rate": 5.749492031793979e-08, + "loss": 0.0004, + "step": 59049 + }, + { + "epoch": 7.74, + "learning_rate": 5.7438147584144564e-08, + "loss": 0.0007, + "step": 59050 + }, + { + "epoch": 7.74, + "learning_rate": 5.738140281327576e-08, + "loss": 0.0004, + "step": 59051 + }, + { + "epoch": 7.74, + "learning_rate": 5.7324686005491015e-08, + "loss": 0.0004, + "step": 59052 + }, + { + "epoch": 7.74, + "learning_rate": 5.7267997160952436e-08, + "loss": 0.0002, + "step": 59053 + }, + { + "epoch": 7.74, + "learning_rate": 5.721133627981656e-08, + "loss": 0.0005, + "step": 59054 + }, + { + "epoch": 7.74, + "learning_rate": 5.715470336224549e-08, + "loss": 0.0002, + "step": 59055 + }, + { + "epoch": 7.74, + "learning_rate": 5.709809840839797e-08, + "loss": 0.0007, + "step": 59056 + }, + { + "epoch": 7.74, + "learning_rate": 5.704152141843167e-08, + "loss": 0.0002, + "step": 59057 + }, + { + "epoch": 7.74, + "learning_rate": 5.6984972392505335e-08, + "loss": 0.0007, + "step": 59058 + }, + { + "epoch": 7.74, + "learning_rate": 5.6928451330781063e-08, + "loss": 0.0004, + "step": 59059 + }, + { + "epoch": 7.74, + "learning_rate": 5.68719582334154e-08, + "loss": 0.0007, + "step": 59060 + }, + { + "epoch": 7.74, + "learning_rate": 5.681549310056822e-08, + "loss": 0.0003, + "step": 59061 + }, + { + "epoch": 7.74, + "learning_rate": 5.6759055932397166e-08, + "loss": 0.0003, + "step": 59062 + }, + { + "epoch": 7.74, + "learning_rate": 5.670264672906323e-08, + "loss": 0.0005, + "step": 59063 + }, + { + "epoch": 7.74, + "learning_rate": 5.664626549072294e-08, + "loss": 0.0005, + "step": 59064 + }, + { + "epoch": 7.74, + "learning_rate": 5.658991221753396e-08, + "loss": 0.0006, + "step": 59065 + }, + { + "epoch": 7.74, + "learning_rate": 5.653358690965838e-08, + "loss": 0.0008, + "step": 59066 + }, + { + "epoch": 7.74, + "learning_rate": 5.6477289567251625e-08, + "loss": 0.0014, + "step": 59067 + }, + { + "epoch": 7.74, + "learning_rate": 5.6421020190473575e-08, + "loss": 0.0004, + "step": 59068 + }, + { + "epoch": 7.74, + "learning_rate": 5.6364778779480764e-08, + "loss": 0.0003, + "step": 59069 + }, + { + "epoch": 7.74, + "learning_rate": 5.6308565334433074e-08, + "loss": 0.0004, + "step": 59070 + }, + { + "epoch": 7.74, + "learning_rate": 5.625237985548815e-08, + "loss": 0.0005, + "step": 59071 + }, + { + "epoch": 7.74, + "learning_rate": 5.6196222342803644e-08, + "loss": 0.0006, + "step": 59072 + }, + { + "epoch": 7.74, + "learning_rate": 5.614009279653832e-08, + "loss": 0.0002, + "step": 59073 + }, + { + "epoch": 7.74, + "learning_rate": 5.608399121684871e-08, + "loss": 0.0004, + "step": 59074 + }, + { + "epoch": 7.74, + "learning_rate": 5.602791760389359e-08, + "loss": 0.001, + "step": 59075 + }, + { + "epoch": 7.74, + "learning_rate": 5.59718719578306e-08, + "loss": 0.0005, + "step": 59076 + }, + { + "epoch": 7.74, + "learning_rate": 5.591585427881741e-08, + "loss": 0.0002, + "step": 59077 + }, + { + "epoch": 7.74, + "learning_rate": 5.5859864567010533e-08, + "loss": 0.0007, + "step": 59078 + }, + { + "epoch": 7.74, + "learning_rate": 5.580390282256876e-08, + "loss": 0.0002, + "step": 59079 + }, + { + "epoch": 7.74, + "learning_rate": 5.574796904564861e-08, + "loss": 0.0002, + "step": 59080 + }, + { + "epoch": 7.74, + "learning_rate": 5.5692063236407746e-08, + "loss": 0.0006, + "step": 59081 + }, + { + "epoch": 7.74, + "learning_rate": 5.5636185395003815e-08, + "loss": 0.0013, + "step": 59082 + }, + { + "epoch": 7.74, + "learning_rate": 5.558033552159336e-08, + "loss": 0.0003, + "step": 59083 + }, + { + "epoch": 7.74, + "learning_rate": 5.552451361633293e-08, + "loss": 0.0005, + "step": 59084 + }, + { + "epoch": 7.74, + "learning_rate": 5.5468719679381276e-08, + "loss": 0.0004, + "step": 59085 + }, + { + "epoch": 7.74, + "learning_rate": 5.541295371089384e-08, + "loss": 0.0011, + "step": 59086 + }, + { + "epoch": 7.74, + "learning_rate": 5.535721571102826e-08, + "loss": 0.0004, + "step": 59087 + }, + { + "epoch": 7.74, + "learning_rate": 5.530150567994108e-08, + "loss": 0.0005, + "step": 59088 + }, + { + "epoch": 7.74, + "learning_rate": 5.524582361778774e-08, + "loss": 0.0005, + "step": 59089 + }, + { + "epoch": 7.74, + "learning_rate": 5.5190169524727e-08, + "loss": 0.0007, + "step": 59090 + }, + { + "epoch": 7.74, + "learning_rate": 5.5134543400913175e-08, + "loss": 0.0005, + "step": 59091 + }, + { + "epoch": 7.74, + "learning_rate": 5.5078945246505036e-08, + "loss": 0.0003, + "step": 59092 + }, + { + "epoch": 7.74, + "learning_rate": 5.502337506165689e-08, + "loss": 0.0004, + "step": 59093 + }, + { + "epoch": 7.74, + "learning_rate": 5.49678328465264e-08, + "loss": 0.0006, + "step": 59094 + }, + { + "epoch": 7.74, + "learning_rate": 5.4912318601269e-08, + "loss": 0.0003, + "step": 59095 + }, + { + "epoch": 7.74, + "learning_rate": 5.485683232604122e-08, + "loss": 0.0006, + "step": 59096 + }, + { + "epoch": 7.74, + "learning_rate": 5.480137402099961e-08, + "loss": 0.0002, + "step": 59097 + }, + { + "epoch": 7.74, + "learning_rate": 5.4745943686298484e-08, + "loss": 0.0009, + "step": 59098 + }, + { + "epoch": 7.74, + "learning_rate": 5.46905413220955e-08, + "loss": 0.0002, + "step": 59099 + }, + { + "epoch": 7.74, + "learning_rate": 5.463516692854498e-08, + "loss": 0.0002, + "step": 59100 + }, + { + "epoch": 7.74, + "learning_rate": 5.457982050580457e-08, + "loss": 0.0003, + "step": 59101 + }, + { + "epoch": 7.74, + "learning_rate": 5.452450205402748e-08, + "loss": 0.0005, + "step": 59102 + }, + { + "epoch": 7.74, + "learning_rate": 5.446921157337137e-08, + "loss": 0.0004, + "step": 59103 + }, + { + "epoch": 7.74, + "learning_rate": 5.4413949063990554e-08, + "loss": 0.0001, + "step": 59104 + }, + { + "epoch": 7.74, + "learning_rate": 5.435871452604158e-08, + "loss": 0.0002, + "step": 59105 + }, + { + "epoch": 7.74, + "learning_rate": 5.430350795967876e-08, + "loss": 0.0008, + "step": 59106 + }, + { + "epoch": 7.74, + "learning_rate": 5.424832936505753e-08, + "loss": 0.0003, + "step": 59107 + }, + { + "epoch": 7.74, + "learning_rate": 5.419317874233332e-08, + "loss": 0.0004, + "step": 59108 + }, + { + "epoch": 7.74, + "learning_rate": 5.413805609166156e-08, + "loss": 0.0003, + "step": 59109 + }, + { + "epoch": 7.74, + "learning_rate": 5.4082961413196574e-08, + "loss": 0.0002, + "step": 59110 + }, + { + "epoch": 7.74, + "learning_rate": 5.402789470709269e-08, + "loss": 0.0003, + "step": 59111 + }, + { + "epoch": 7.74, + "learning_rate": 5.3972855973507546e-08, + "loss": 0.0008, + "step": 59112 + }, + { + "epoch": 7.74, + "learning_rate": 5.3917845212592134e-08, + "loss": 0.0004, + "step": 59113 + }, + { + "epoch": 7.74, + "learning_rate": 5.3862862424504114e-08, + "loss": 0.0004, + "step": 59114 + }, + { + "epoch": 7.74, + "learning_rate": 5.380790760939669e-08, + "loss": 0.0006, + "step": 59115 + }, + { + "epoch": 7.74, + "learning_rate": 5.37529807674253e-08, + "loss": 0.002, + "step": 59116 + }, + { + "epoch": 7.74, + "learning_rate": 5.369808189874426e-08, + "loss": 0.0002, + "step": 59117 + }, + { + "epoch": 7.74, + "learning_rate": 5.364321100350789e-08, + "loss": 0.0008, + "step": 59118 + }, + { + "epoch": 7.74, + "learning_rate": 5.35883680818694e-08, + "loss": 0.0005, + "step": 59119 + }, + { + "epoch": 7.74, + "learning_rate": 5.353355313398423e-08, + "loss": 0.0003, + "step": 59120 + }, + { + "epoch": 7.74, + "learning_rate": 5.347876616000669e-08, + "loss": 0.0004, + "step": 59121 + }, + { + "epoch": 7.74, + "learning_rate": 5.3424007160089996e-08, + "loss": 0.0004, + "step": 59122 + }, + { + "epoch": 7.74, + "learning_rate": 5.336927613438847e-08, + "loss": 0.0004, + "step": 59123 + }, + { + "epoch": 7.74, + "learning_rate": 5.331457308305643e-08, + "loss": 0.0003, + "step": 59124 + }, + { + "epoch": 7.74, + "learning_rate": 5.325989800624709e-08, + "loss": 0.0014, + "step": 59125 + }, + { + "epoch": 7.75, + "learning_rate": 5.320525090411477e-08, + "loss": 0.0003, + "step": 59126 + }, + { + "epoch": 7.75, + "learning_rate": 5.3150631776813786e-08, + "loss": 0.0002, + "step": 59127 + }, + { + "epoch": 7.75, + "learning_rate": 5.309604062449625e-08, + "loss": 0.0004, + "step": 59128 + }, + { + "epoch": 7.75, + "learning_rate": 5.3041477447316466e-08, + "loss": 0.0007, + "step": 59129 + }, + { + "epoch": 7.75, + "learning_rate": 5.298694224542766e-08, + "loss": 0.0001, + "step": 59130 + }, + { + "epoch": 7.75, + "learning_rate": 5.2932435018983044e-08, + "loss": 0.0002, + "step": 59131 + }, + { + "epoch": 7.75, + "learning_rate": 5.2877955768136925e-08, + "loss": 0.0002, + "step": 59132 + }, + { + "epoch": 7.75, + "learning_rate": 5.282350449304141e-08, + "loss": 0.0003, + "step": 59133 + }, + { + "epoch": 7.75, + "learning_rate": 5.2769081193850826e-08, + "loss": 0.0004, + "step": 59134 + }, + { + "epoch": 7.75, + "learning_rate": 5.2714685870716155e-08, + "loss": 0.0003, + "step": 59135 + }, + { + "epoch": 7.75, + "learning_rate": 5.266031852379283e-08, + "loss": 0.0004, + "step": 59136 + }, + { + "epoch": 7.75, + "learning_rate": 5.2605979153231845e-08, + "loss": 0.0006, + "step": 59137 + }, + { + "epoch": 7.75, + "learning_rate": 5.255166775918752e-08, + "loss": 0.001, + "step": 59138 + }, + { + "epoch": 7.75, + "learning_rate": 5.2497384341811954e-08, + "loss": 0.0002, + "step": 59139 + }, + { + "epoch": 7.75, + "learning_rate": 5.2443128901257247e-08, + "loss": 0.0005, + "step": 59140 + }, + { + "epoch": 7.75, + "learning_rate": 5.2388901437676606e-08, + "loss": 0.0003, + "step": 59141 + }, + { + "epoch": 7.75, + "learning_rate": 5.233470195122214e-08, + "loss": 0.0007, + "step": 59142 + }, + { + "epoch": 7.75, + "learning_rate": 5.228053044204706e-08, + "loss": 0.0002, + "step": 59143 + }, + { + "epoch": 7.75, + "learning_rate": 5.222638691030346e-08, + "loss": 0.0002, + "step": 59144 + }, + { + "epoch": 7.75, + "learning_rate": 5.217227135614344e-08, + "loss": 0.0005, + "step": 59145 + }, + { + "epoch": 7.75, + "learning_rate": 5.2118183779718e-08, + "loss": 0.0005, + "step": 59146 + }, + { + "epoch": 7.75, + "learning_rate": 5.2064124181181455e-08, + "loss": 0.0004, + "step": 59147 + }, + { + "epoch": 7.75, + "learning_rate": 5.201009256068479e-08, + "loss": 0.0008, + "step": 59148 + }, + { + "epoch": 7.75, + "learning_rate": 5.1956088918379e-08, + "loss": 0.0004, + "step": 59149 + }, + { + "epoch": 7.75, + "learning_rate": 5.1902113254418406e-08, + "loss": 0.0004, + "step": 59150 + }, + { + "epoch": 7.75, + "learning_rate": 5.184816556895289e-08, + "loss": 0.0006, + "step": 59151 + }, + { + "epoch": 7.75, + "learning_rate": 5.179424586213455e-08, + "loss": 0.0002, + "step": 59152 + }, + { + "epoch": 7.75, + "learning_rate": 5.174035413411438e-08, + "loss": 0.0006, + "step": 59153 + }, + { + "epoch": 7.75, + "learning_rate": 5.168649038504558e-08, + "loss": 0.0004, + "step": 59154 + }, + { + "epoch": 7.75, + "learning_rate": 5.163265461507805e-08, + "loss": 0.0005, + "step": 59155 + }, + { + "epoch": 7.75, + "learning_rate": 5.157884682436387e-08, + "loss": 0.0003, + "step": 59156 + }, + { + "epoch": 7.75, + "learning_rate": 5.152506701305515e-08, + "loss": 0.0002, + "step": 59157 + }, + { + "epoch": 7.75, + "learning_rate": 5.1471315181301774e-08, + "loss": 0.0002, + "step": 59158 + }, + { + "epoch": 7.75, + "learning_rate": 5.141759132925583e-08, + "loss": 0.0004, + "step": 59159 + }, + { + "epoch": 7.75, + "learning_rate": 5.136389545706833e-08, + "loss": 0.0003, + "step": 59160 + }, + { + "epoch": 7.75, + "learning_rate": 5.131022756488913e-08, + "loss": 0.0005, + "step": 59161 + }, + { + "epoch": 7.75, + "learning_rate": 5.125658765287145e-08, + "loss": 0.0004, + "step": 59162 + }, + { + "epoch": 7.75, + "learning_rate": 5.1202975721164064e-08, + "loss": 0.0003, + "step": 59163 + }, + { + "epoch": 7.75, + "learning_rate": 5.114939176991907e-08, + "loss": 0.001, + "step": 59164 + }, + { + "epoch": 7.75, + "learning_rate": 5.109583579928634e-08, + "loss": 0.0004, + "step": 59165 + }, + { + "epoch": 7.75, + "learning_rate": 5.104230780941577e-08, + "loss": 0.0003, + "step": 59166 + }, + { + "epoch": 7.75, + "learning_rate": 5.098880780045945e-08, + "loss": 0.0004, + "step": 59167 + }, + { + "epoch": 7.75, + "learning_rate": 5.093533577256837e-08, + "loss": 0.0003, + "step": 59168 + }, + { + "epoch": 7.75, + "learning_rate": 5.0881891725890196e-08, + "loss": 0.0013, + "step": 59169 + }, + { + "epoch": 7.75, + "learning_rate": 5.082847566057814e-08, + "loss": 0.0004, + "step": 59170 + }, + { + "epoch": 7.75, + "learning_rate": 5.077508757678096e-08, + "loss": 0.0004, + "step": 59171 + }, + { + "epoch": 7.75, + "learning_rate": 5.072172747464854e-08, + "loss": 0.0006, + "step": 59172 + }, + { + "epoch": 7.75, + "learning_rate": 5.066839535433188e-08, + "loss": 0.0005, + "step": 59173 + }, + { + "epoch": 7.75, + "learning_rate": 5.061509121598085e-08, + "loss": 0.0003, + "step": 59174 + }, + { + "epoch": 7.75, + "learning_rate": 5.056181505974422e-08, + "loss": 0.0003, + "step": 59175 + }, + { + "epoch": 7.75, + "learning_rate": 5.050856688577299e-08, + "loss": 0.0003, + "step": 59176 + }, + { + "epoch": 7.75, + "learning_rate": 5.0455346694215924e-08, + "loss": 0.0005, + "step": 59177 + }, + { + "epoch": 7.75, + "learning_rate": 5.040215448522401e-08, + "loss": 0.0004, + "step": 59178 + }, + { + "epoch": 7.75, + "learning_rate": 5.034899025894602e-08, + "loss": 0.0002, + "step": 59179 + }, + { + "epoch": 7.75, + "learning_rate": 5.0295854015531834e-08, + "loss": 0.0002, + "step": 59180 + }, + { + "epoch": 7.75, + "learning_rate": 5.024274575513022e-08, + "loss": 0.0003, + "step": 59181 + }, + { + "epoch": 7.75, + "learning_rate": 5.018966547789217e-08, + "loss": 0.0004, + "step": 59182 + }, + { + "epoch": 7.75, + "learning_rate": 5.0136613183965345e-08, + "loss": 0.0002, + "step": 59183 + }, + { + "epoch": 7.75, + "learning_rate": 5.008358887349851e-08, + "loss": 0.0007, + "step": 59184 + }, + { + "epoch": 7.75, + "learning_rate": 5.003059254664267e-08, + "loss": 0.0004, + "step": 59185 + }, + { + "epoch": 7.75, + "learning_rate": 4.9977624203545463e-08, + "loss": 0.0002, + "step": 59186 + }, + { + "epoch": 7.75, + "learning_rate": 4.992468384435678e-08, + "loss": 0.0004, + "step": 59187 + }, + { + "epoch": 7.75, + "learning_rate": 4.9871771469224285e-08, + "loss": 0.0004, + "step": 59188 + }, + { + "epoch": 7.75, + "learning_rate": 4.981888707829785e-08, + "loss": 0.0002, + "step": 59189 + }, + { + "epoch": 7.75, + "learning_rate": 4.976603067172625e-08, + "loss": 0.0016, + "step": 59190 + }, + { + "epoch": 7.75, + "learning_rate": 4.971320224965826e-08, + "loss": 0.0006, + "step": 59191 + }, + { + "epoch": 7.75, + "learning_rate": 4.966040181224041e-08, + "loss": 0.0004, + "step": 59192 + }, + { + "epoch": 7.75, + "learning_rate": 4.960762935962482e-08, + "loss": 0.0003, + "step": 59193 + }, + { + "epoch": 7.75, + "learning_rate": 4.955488489195692e-08, + "loss": 0.0003, + "step": 59194 + }, + { + "epoch": 7.75, + "learning_rate": 4.9502168409385485e-08, + "loss": 0.0005, + "step": 59195 + }, + { + "epoch": 7.75, + "learning_rate": 4.944947991205928e-08, + "loss": 0.0002, + "step": 59196 + }, + { + "epoch": 7.75, + "learning_rate": 4.9396819400128194e-08, + "loss": 0.0002, + "step": 59197 + }, + { + "epoch": 7.75, + "learning_rate": 4.9344186873737654e-08, + "loss": 0.0004, + "step": 59198 + }, + { + "epoch": 7.75, + "learning_rate": 4.9291582333035324e-08, + "loss": 0.0003, + "step": 59199 + }, + { + "epoch": 7.75, + "learning_rate": 4.92390057781722e-08, + "loss": 0.0005, + "step": 59200 + }, + { + "epoch": 7.75, + "learning_rate": 4.918645720929371e-08, + "loss": 0.0003, + "step": 59201 + }, + { + "epoch": 7.76, + "learning_rate": 4.913393662654864e-08, + "loss": 0.0004, + "step": 59202 + }, + { + "epoch": 7.76, + "learning_rate": 4.9081444030084635e-08, + "loss": 0.0004, + "step": 59203 + }, + { + "epoch": 7.76, + "learning_rate": 4.902897942004825e-08, + "loss": 0.0004, + "step": 59204 + }, + { + "epoch": 7.76, + "learning_rate": 4.8976542796589365e-08, + "loss": 0.0003, + "step": 59205 + }, + { + "epoch": 7.76, + "learning_rate": 4.892413415985342e-08, + "loss": 0.0004, + "step": 59206 + }, + { + "epoch": 7.76, + "learning_rate": 4.8871753509988075e-08, + "loss": 0.0005, + "step": 59207 + }, + { + "epoch": 7.76, + "learning_rate": 4.881940084714099e-08, + "loss": 0.0002, + "step": 59208 + }, + { + "epoch": 7.76, + "learning_rate": 4.876707617145982e-08, + "loss": 0.0004, + "step": 59209 + }, + { + "epoch": 7.76, + "learning_rate": 4.8714779483091115e-08, + "loss": 0.0005, + "step": 59210 + }, + { + "epoch": 7.76, + "learning_rate": 4.8662510782181424e-08, + "loss": 0.0008, + "step": 59211 + }, + { + "epoch": 7.76, + "learning_rate": 4.861027006887953e-08, + "loss": 0.0006, + "step": 59212 + }, + { + "epoch": 7.76, + "learning_rate": 4.8558057343331966e-08, + "loss": 0.0014, + "step": 59213 + }, + { + "epoch": 7.76, + "learning_rate": 4.850587260568307e-08, + "loss": 0.0003, + "step": 59214 + }, + { + "epoch": 7.76, + "learning_rate": 4.845371585608272e-08, + "loss": 0.0009, + "step": 59215 + }, + { + "epoch": 7.76, + "learning_rate": 4.840158709467635e-08, + "loss": 0.0002, + "step": 59216 + }, + { + "epoch": 7.76, + "learning_rate": 4.8349486321610517e-08, + "loss": 0.0009, + "step": 59217 + }, + { + "epoch": 7.76, + "learning_rate": 4.8297413537030656e-08, + "loss": 0.001, + "step": 59218 + }, + { + "epoch": 7.76, + "learning_rate": 4.824536874108554e-08, + "loss": 0.0002, + "step": 59219 + }, + { + "epoch": 7.76, + "learning_rate": 4.8193351933920604e-08, + "loss": 0.0002, + "step": 59220 + }, + { + "epoch": 7.76, + "learning_rate": 4.8141363115681295e-08, + "loss": 0.0006, + "step": 59221 + }, + { + "epoch": 7.76, + "learning_rate": 4.808940228651304e-08, + "loss": 0.0004, + "step": 59222 + }, + { + "epoch": 7.76, + "learning_rate": 4.803746944656573e-08, + "loss": 0.0008, + "step": 59223 + }, + { + "epoch": 7.76, + "learning_rate": 4.798556459598147e-08, + "loss": 0.0003, + "step": 59224 + }, + { + "epoch": 7.76, + "learning_rate": 4.7933687734909026e-08, + "loss": 0.0002, + "step": 59225 + }, + { + "epoch": 7.76, + "learning_rate": 4.7881838863492736e-08, + "loss": 0.0005, + "step": 59226 + }, + { + "epoch": 7.76, + "learning_rate": 4.783001798187803e-08, + "loss": 0.0007, + "step": 59227 + }, + { + "epoch": 7.76, + "learning_rate": 4.777822509021257e-08, + "loss": 0.0007, + "step": 59228 + }, + { + "epoch": 7.76, + "learning_rate": 4.772646018863958e-08, + "loss": 0.0003, + "step": 59229 + }, + { + "epoch": 7.76, + "learning_rate": 4.767472327730671e-08, + "loss": 0.0002, + "step": 59230 + }, + { + "epoch": 7.76, + "learning_rate": 4.762301435635941e-08, + "loss": 0.0002, + "step": 59231 + }, + { + "epoch": 7.76, + "learning_rate": 4.7571333425942e-08, + "loss": 0.0002, + "step": 59232 + }, + { + "epoch": 7.76, + "learning_rate": 4.751968048619881e-08, + "loss": 0.0002, + "step": 59233 + }, + { + "epoch": 7.76, + "learning_rate": 4.746805553727751e-08, + "loss": 0.0007, + "step": 59234 + }, + { + "epoch": 7.76, + "learning_rate": 4.741645857932242e-08, + "loss": 0.0003, + "step": 59235 + }, + { + "epoch": 7.76, + "learning_rate": 4.7364889612478985e-08, + "loss": 0.0002, + "step": 59236 + }, + { + "epoch": 7.76, + "learning_rate": 4.731334863689041e-08, + "loss": 0.0003, + "step": 59237 + }, + { + "epoch": 7.76, + "learning_rate": 4.7261835652704366e-08, + "loss": 0.0004, + "step": 59238 + }, + { + "epoch": 7.76, + "learning_rate": 4.721035066006407e-08, + "loss": 0.0002, + "step": 59239 + }, + { + "epoch": 7.76, + "learning_rate": 4.715889365911497e-08, + "loss": 0.0004, + "step": 59240 + }, + { + "epoch": 7.76, + "learning_rate": 4.710746465000027e-08, + "loss": 0.0001, + "step": 59241 + }, + { + "epoch": 7.76, + "learning_rate": 4.705606363286652e-08, + "loss": 0.0004, + "step": 59242 + }, + { + "epoch": 7.76, + "learning_rate": 4.7004690607858064e-08, + "loss": 0.0009, + "step": 59243 + }, + { + "epoch": 7.76, + "learning_rate": 4.6953345575118104e-08, + "loss": 0.0004, + "step": 59244 + }, + { + "epoch": 7.76, + "learning_rate": 4.6902028534790975e-08, + "loss": 0.0003, + "step": 59245 + }, + { + "epoch": 7.76, + "learning_rate": 4.685073948702323e-08, + "loss": 0.0004, + "step": 59246 + }, + { + "epoch": 7.76, + "learning_rate": 4.679947843195809e-08, + "loss": 0.0002, + "step": 59247 + }, + { + "epoch": 7.76, + "learning_rate": 4.674824536973877e-08, + "loss": 0.0005, + "step": 59248 + }, + { + "epoch": 7.76, + "learning_rate": 4.6697040300509585e-08, + "loss": 0.0006, + "step": 59249 + }, + { + "epoch": 7.76, + "learning_rate": 4.6645863224416e-08, + "loss": 0.0009, + "step": 59250 + }, + { + "epoch": 7.76, + "learning_rate": 4.659471414160011e-08, + "loss": 0.0003, + "step": 59251 + }, + { + "epoch": 7.76, + "learning_rate": 4.6543593052206236e-08, + "loss": 0.0003, + "step": 59252 + }, + { + "epoch": 7.76, + "learning_rate": 4.649249995637872e-08, + "loss": 0.0008, + "step": 59253 + }, + { + "epoch": 7.76, + "learning_rate": 4.6441434854260783e-08, + "loss": 0.002, + "step": 59254 + }, + { + "epoch": 7.76, + "learning_rate": 4.639039774599674e-08, + "loss": 0.0005, + "step": 59255 + }, + { + "epoch": 7.76, + "learning_rate": 4.633938863172982e-08, + "loss": 0.0002, + "step": 59256 + }, + { + "epoch": 7.76, + "learning_rate": 4.628840751160213e-08, + "loss": 0.0005, + "step": 59257 + }, + { + "epoch": 7.76, + "learning_rate": 4.6237454385759105e-08, + "loss": 0.0002, + "step": 59258 + }, + { + "epoch": 7.76, + "learning_rate": 4.618652925434397e-08, + "loss": 0.0004, + "step": 59259 + }, + { + "epoch": 7.76, + "learning_rate": 4.613563211749772e-08, + "loss": 0.0002, + "step": 59260 + }, + { + "epoch": 7.76, + "learning_rate": 4.608476297536579e-08, + "loss": 0.0006, + "step": 59261 + }, + { + "epoch": 7.76, + "learning_rate": 4.60339218280903e-08, + "loss": 0.0008, + "step": 59262 + }, + { + "epoch": 7.76, + "learning_rate": 4.598310867581446e-08, + "loss": 0.0002, + "step": 59263 + }, + { + "epoch": 7.76, + "learning_rate": 4.593232351868038e-08, + "loss": 0.0005, + "step": 59264 + }, + { + "epoch": 7.76, + "learning_rate": 4.588156635683238e-08, + "loss": 0.0002, + "step": 59265 + }, + { + "epoch": 7.76, + "learning_rate": 4.5830837190411484e-08, + "loss": 0.0005, + "step": 59266 + }, + { + "epoch": 7.76, + "learning_rate": 4.5780136019561995e-08, + "loss": 0.0004, + "step": 59267 + }, + { + "epoch": 7.76, + "learning_rate": 4.572946284442492e-08, + "loss": 0.0002, + "step": 59268 + }, + { + "epoch": 7.76, + "learning_rate": 4.567881766514459e-08, + "loss": 0.0003, + "step": 59269 + }, + { + "epoch": 7.76, + "learning_rate": 4.5628200481862e-08, + "loss": 0.0006, + "step": 59270 + }, + { + "epoch": 7.76, + "learning_rate": 4.557761129472038e-08, + "loss": 0.0003, + "step": 59271 + }, + { + "epoch": 7.76, + "learning_rate": 4.5527050103860716e-08, + "loss": 0.0002, + "step": 59272 + }, + { + "epoch": 7.76, + "learning_rate": 4.547651690942623e-08, + "loss": 0.0003, + "step": 59273 + }, + { + "epoch": 7.76, + "learning_rate": 4.542601171156014e-08, + "loss": 0.0002, + "step": 59274 + }, + { + "epoch": 7.76, + "learning_rate": 4.537553451040233e-08, + "loss": 0.0003, + "step": 59275 + }, + { + "epoch": 7.76, + "learning_rate": 4.532508530609603e-08, + "loss": 0.0006, + "step": 59276 + }, + { + "epoch": 7.76, + "learning_rate": 4.5274664098782225e-08, + "loss": 0.0004, + "step": 59277 + }, + { + "epoch": 7.76, + "learning_rate": 4.522427088860415e-08, + "loss": 0.0003, + "step": 59278 + }, + { + "epoch": 7.77, + "learning_rate": 4.517390567570168e-08, + "loss": 0.0004, + "step": 59279 + }, + { + "epoch": 7.77, + "learning_rate": 4.512356846021915e-08, + "loss": 0.0002, + "step": 59280 + }, + { + "epoch": 7.77, + "learning_rate": 4.507325924229533e-08, + "loss": 0.0006, + "step": 59281 + }, + { + "epoch": 7.77, + "learning_rate": 4.502297802207234e-08, + "loss": 0.0003, + "step": 59282 + }, + { + "epoch": 7.77, + "learning_rate": 4.497272479969339e-08, + "loss": 0.0019, + "step": 59283 + }, + { + "epoch": 7.77, + "learning_rate": 4.492249957529726e-08, + "loss": 0.0005, + "step": 59284 + }, + { + "epoch": 7.77, + "learning_rate": 4.4872302349027175e-08, + "loss": 0.0001, + "step": 59285 + }, + { + "epoch": 7.77, + "learning_rate": 4.482213312102302e-08, + "loss": 0.0003, + "step": 59286 + }, + { + "epoch": 7.77, + "learning_rate": 4.477199189142689e-08, + "loss": 0.0006, + "step": 59287 + }, + { + "epoch": 7.77, + "learning_rate": 4.4721878660379803e-08, + "loss": 0.0002, + "step": 59288 + }, + { + "epoch": 7.77, + "learning_rate": 4.467179342802164e-08, + "loss": 0.0004, + "step": 59289 + }, + { + "epoch": 7.77, + "learning_rate": 4.4621736194494504e-08, + "loss": 0.0009, + "step": 59290 + }, + { + "epoch": 7.77, + "learning_rate": 4.4571706959938286e-08, + "loss": 0.0002, + "step": 59291 + }, + { + "epoch": 7.77, + "learning_rate": 4.452170572449399e-08, + "loss": 0.0009, + "step": 59292 + }, + { + "epoch": 7.77, + "learning_rate": 4.447173248830261e-08, + "loss": 0.0003, + "step": 59293 + }, + { + "epoch": 7.77, + "learning_rate": 4.442178725150292e-08, + "loss": 0.0032, + "step": 59294 + }, + { + "epoch": 7.77, + "learning_rate": 4.437187001423815e-08, + "loss": 0.0003, + "step": 59295 + }, + { + "epoch": 7.77, + "learning_rate": 4.432198077664707e-08, + "loss": 0.0016, + "step": 59296 + }, + { + "epoch": 7.77, + "learning_rate": 4.427211953887067e-08, + "loss": 0.0003, + "step": 59297 + }, + { + "epoch": 7.77, + "learning_rate": 4.422228630104775e-08, + "loss": 0.0005, + "step": 59298 + }, + { + "epoch": 7.77, + "learning_rate": 4.4172481063319284e-08, + "loss": 0.0002, + "step": 59299 + }, + { + "epoch": 7.77, + "learning_rate": 4.4122703825826285e-08, + "loss": 0.0006, + "step": 59300 + }, + { + "epoch": 7.77, + "learning_rate": 4.407295458870753e-08, + "loss": 0.0004, + "step": 59301 + }, + { + "epoch": 7.77, + "learning_rate": 4.4023233352104014e-08, + "loss": 0.0004, + "step": 59302 + }, + { + "epoch": 7.77, + "learning_rate": 4.3973540116154514e-08, + "loss": 0.0005, + "step": 59303 + }, + { + "epoch": 7.77, + "learning_rate": 4.392387488100003e-08, + "loss": 0.0002, + "step": 59304 + }, + { + "epoch": 7.77, + "learning_rate": 4.3874237646779336e-08, + "loss": 0.0004, + "step": 59305 + }, + { + "epoch": 7.77, + "learning_rate": 4.382462841363122e-08, + "loss": 0.0005, + "step": 59306 + }, + { + "epoch": 7.77, + "learning_rate": 4.377504718169667e-08, + "loss": 0.0005, + "step": 59307 + }, + { + "epoch": 7.77, + "learning_rate": 4.3725493951114474e-08, + "loss": 0.0007, + "step": 59308 + }, + { + "epoch": 7.77, + "learning_rate": 4.36759687220234e-08, + "loss": 0.0001, + "step": 59309 + }, + { + "epoch": 7.77, + "learning_rate": 4.362647149456445e-08, + "loss": 0.0002, + "step": 59310 + }, + { + "epoch": 7.77, + "learning_rate": 4.357700226887529e-08, + "loss": 0.0006, + "step": 59311 + }, + { + "epoch": 7.77, + "learning_rate": 4.352756104509692e-08, + "loss": 0.0002, + "step": 59312 + }, + { + "epoch": 7.77, + "learning_rate": 4.34781478233659e-08, + "loss": 0.0002, + "step": 59313 + }, + { + "epoch": 7.77, + "learning_rate": 4.3428762603822115e-08, + "loss": 0.0005, + "step": 59314 + }, + { + "epoch": 7.77, + "learning_rate": 4.337940538660545e-08, + "loss": 0.0002, + "step": 59315 + }, + { + "epoch": 7.77, + "learning_rate": 4.3330076171853584e-08, + "loss": 0.0005, + "step": 59316 + }, + { + "epoch": 7.77, + "learning_rate": 4.3280774959705286e-08, + "loss": 0.0003, + "step": 59317 + }, + { + "epoch": 7.77, + "learning_rate": 4.3231501750300444e-08, + "loss": 0.0004, + "step": 59318 + }, + { + "epoch": 7.77, + "learning_rate": 4.318225654377672e-08, + "loss": 0.0005, + "step": 59319 + }, + { + "epoch": 7.77, + "learning_rate": 4.31330393402718e-08, + "loss": 0.0006, + "step": 59320 + }, + { + "epoch": 7.77, + "learning_rate": 4.308385013992555e-08, + "loss": 0.0007, + "step": 59321 + }, + { + "epoch": 7.77, + "learning_rate": 4.303468894287566e-08, + "loss": 0.0004, + "step": 59322 + }, + { + "epoch": 7.77, + "learning_rate": 4.2985555749260886e-08, + "loss": 0.0008, + "step": 59323 + }, + { + "epoch": 7.77, + "learning_rate": 4.2936450559218913e-08, + "loss": 0.0004, + "step": 59324 + }, + { + "epoch": 7.77, + "learning_rate": 4.28873733728874e-08, + "loss": 0.0005, + "step": 59325 + }, + { + "epoch": 7.77, + "learning_rate": 4.2838324190405125e-08, + "loss": 0.0004, + "step": 59326 + }, + { + "epoch": 7.77, + "learning_rate": 4.2789303011909754e-08, + "loss": 0.0004, + "step": 59327 + }, + { + "epoch": 7.77, + "learning_rate": 4.274030983754007e-08, + "loss": 0.0007, + "step": 59328 + }, + { + "epoch": 7.77, + "learning_rate": 4.269134466743152e-08, + "loss": 0.0007, + "step": 59329 + }, + { + "epoch": 7.77, + "learning_rate": 4.264240750172399e-08, + "loss": 0.0011, + "step": 59330 + }, + { + "epoch": 7.77, + "learning_rate": 4.259349834055404e-08, + "loss": 0.0004, + "step": 59331 + }, + { + "epoch": 7.77, + "learning_rate": 4.2544617184060447e-08, + "loss": 0.0004, + "step": 59332 + }, + { + "epoch": 7.77, + "learning_rate": 4.249576403237865e-08, + "loss": 0.0003, + "step": 59333 + }, + { + "epoch": 7.77, + "learning_rate": 4.244693888564855e-08, + "loss": 0.0004, + "step": 59334 + }, + { + "epoch": 7.77, + "learning_rate": 4.2398141744005586e-08, + "loss": 0.0003, + "step": 59335 + }, + { + "epoch": 7.77, + "learning_rate": 4.2349372607587424e-08, + "loss": 0.0003, + "step": 59336 + }, + { + "epoch": 7.77, + "learning_rate": 4.230063147653174e-08, + "loss": 0.0004, + "step": 59337 + }, + { + "epoch": 7.77, + "learning_rate": 4.2251918350975084e-08, + "loss": 0.0007, + "step": 59338 + }, + { + "epoch": 7.77, + "learning_rate": 4.220323323105513e-08, + "loss": 0.0003, + "step": 59339 + }, + { + "epoch": 7.77, + "learning_rate": 4.215457611690732e-08, + "loss": 0.0007, + "step": 59340 + }, + { + "epoch": 7.77, + "learning_rate": 4.2105947008670434e-08, + "loss": 0.0004, + "step": 59341 + }, + { + "epoch": 7.77, + "learning_rate": 4.205734590647881e-08, + "loss": 0.0007, + "step": 59342 + }, + { + "epoch": 7.77, + "learning_rate": 4.2008772810472334e-08, + "loss": 0.0011, + "step": 59343 + }, + { + "epoch": 7.77, + "learning_rate": 4.196022772078534e-08, + "loss": 0.0008, + "step": 59344 + }, + { + "epoch": 7.77, + "learning_rate": 4.191171063755439e-08, + "loss": 0.0002, + "step": 59345 + }, + { + "epoch": 7.77, + "learning_rate": 4.186322156091715e-08, + "loss": 0.0002, + "step": 59346 + }, + { + "epoch": 7.77, + "learning_rate": 4.1814760491009075e-08, + "loss": 0.0004, + "step": 59347 + }, + { + "epoch": 7.77, + "learning_rate": 4.17663274279656e-08, + "loss": 0.0005, + "step": 59348 + }, + { + "epoch": 7.77, + "learning_rate": 4.1717922371925515e-08, + "loss": 0.0002, + "step": 59349 + }, + { + "epoch": 7.77, + "learning_rate": 4.166954532302314e-08, + "loss": 0.0004, + "step": 59350 + }, + { + "epoch": 7.77, + "learning_rate": 4.1621196281393936e-08, + "loss": 0.0009, + "step": 59351 + }, + { + "epoch": 7.77, + "learning_rate": 4.157287524717557e-08, + "loss": 0.0003, + "step": 59352 + }, + { + "epoch": 7.77, + "learning_rate": 4.1524582220503483e-08, + "loss": 0.0003, + "step": 59353 + }, + { + "epoch": 7.77, + "learning_rate": 4.147631720151313e-08, + "loss": 0.0002, + "step": 59354 + }, + { + "epoch": 7.78, + "learning_rate": 4.142808019033995e-08, + "loss": 0.0003, + "step": 59355 + }, + { + "epoch": 7.78, + "learning_rate": 4.137987118711939e-08, + "loss": 0.0004, + "step": 59356 + }, + { + "epoch": 7.78, + "learning_rate": 4.1331690191989126e-08, + "loss": 0.0002, + "step": 59357 + }, + { + "epoch": 7.78, + "learning_rate": 4.128353720508238e-08, + "loss": 0.0003, + "step": 59358 + }, + { + "epoch": 7.78, + "learning_rate": 4.12354122265346e-08, + "loss": 0.0003, + "step": 59359 + }, + { + "epoch": 7.78, + "learning_rate": 4.118731525648234e-08, + "loss": 0.0002, + "step": 59360 + }, + { + "epoch": 7.78, + "learning_rate": 4.113924629506105e-08, + "loss": 0.0005, + "step": 59361 + }, + { + "epoch": 7.78, + "learning_rate": 4.109120534240507e-08, + "loss": 0.0003, + "step": 59362 + }, + { + "epoch": 7.78, + "learning_rate": 4.104319239864873e-08, + "loss": 0.0006, + "step": 59363 + }, + { + "epoch": 7.78, + "learning_rate": 4.09952074639286e-08, + "loss": 0.0013, + "step": 59364 + }, + { + "epoch": 7.78, + "learning_rate": 4.094725053838011e-08, + "loss": 0.0009, + "step": 59365 + }, + { + "epoch": 7.78, + "learning_rate": 4.08993216221365e-08, + "loss": 0.0005, + "step": 59366 + }, + { + "epoch": 7.78, + "learning_rate": 4.0851420715332103e-08, + "loss": 0.0003, + "step": 59367 + }, + { + "epoch": 7.78, + "learning_rate": 4.0803547818104586e-08, + "loss": 0.0009, + "step": 59368 + }, + { + "epoch": 7.78, + "learning_rate": 4.075570293058606e-08, + "loss": 0.0004, + "step": 59369 + }, + { + "epoch": 7.78, + "learning_rate": 4.070788605291087e-08, + "loss": 0.0004, + "step": 59370 + }, + { + "epoch": 7.78, + "learning_rate": 4.0660097185215575e-08, + "loss": 0.0005, + "step": 59371 + }, + { + "epoch": 7.78, + "learning_rate": 4.0612336327632286e-08, + "loss": 0.0003, + "step": 59372 + }, + { + "epoch": 7.78, + "learning_rate": 4.056460348029756e-08, + "loss": 0.001, + "step": 59373 + }, + { + "epoch": 7.78, + "learning_rate": 4.051689864334352e-08, + "loss": 0.0009, + "step": 59374 + }, + { + "epoch": 7.78, + "learning_rate": 4.0469221816905604e-08, + "loss": 0.0004, + "step": 59375 + }, + { + "epoch": 7.78, + "learning_rate": 4.042157300111815e-08, + "loss": 0.0005, + "step": 59376 + }, + { + "epoch": 7.78, + "learning_rate": 4.03739521961144e-08, + "loss": 0.0004, + "step": 59377 + }, + { + "epoch": 7.78, + "learning_rate": 4.032635940202867e-08, + "loss": 0.0007, + "step": 59378 + }, + { + "epoch": 7.78, + "learning_rate": 4.02787946189942e-08, + "loss": 0.0008, + "step": 59379 + }, + { + "epoch": 7.78, + "learning_rate": 4.023125784714532e-08, + "loss": 0.0003, + "step": 59380 + }, + { + "epoch": 7.78, + "learning_rate": 4.018374908661527e-08, + "loss": 0.0003, + "step": 59381 + }, + { + "epoch": 7.78, + "learning_rate": 4.0136268337538363e-08, + "loss": 0.0006, + "step": 59382 + }, + { + "epoch": 7.78, + "learning_rate": 4.008881560004785e-08, + "loss": 0.0004, + "step": 59383 + }, + { + "epoch": 7.78, + "learning_rate": 4.004139087427805e-08, + "loss": 0.0008, + "step": 59384 + }, + { + "epoch": 7.78, + "learning_rate": 3.99939941603611e-08, + "loss": 0.0004, + "step": 59385 + }, + { + "epoch": 7.78, + "learning_rate": 3.994662545842909e-08, + "loss": 0.0002, + "step": 59386 + }, + { + "epoch": 7.78, + "learning_rate": 3.98992847686186e-08, + "loss": 0.0007, + "step": 59387 + }, + { + "epoch": 7.78, + "learning_rate": 3.985197209105951e-08, + "loss": 0.0002, + "step": 59388 + }, + { + "epoch": 7.78, + "learning_rate": 3.980468742588728e-08, + "loss": 0.0002, + "step": 59389 + }, + { + "epoch": 7.78, + "learning_rate": 3.975743077323402e-08, + "loss": 0.0002, + "step": 59390 + }, + { + "epoch": 7.78, + "learning_rate": 3.9710202133231845e-08, + "loss": 0.0002, + "step": 59391 + }, + { + "epoch": 7.78, + "learning_rate": 3.96630015060151e-08, + "loss": 0.0004, + "step": 59392 + }, + { + "epoch": 7.78, + "learning_rate": 3.961582889171589e-08, + "loss": 0.0008, + "step": 59393 + }, + { + "epoch": 7.78, + "learning_rate": 3.956868429046634e-08, + "loss": 0.0002, + "step": 59394 + }, + { + "epoch": 7.78, + "learning_rate": 3.9521567702399675e-08, + "loss": 0.0003, + "step": 59395 + }, + { + "epoch": 7.78, + "learning_rate": 3.9474479127648014e-08, + "loss": 0.0007, + "step": 59396 + }, + { + "epoch": 7.78, + "learning_rate": 3.942741856634347e-08, + "loss": 0.0003, + "step": 59397 + }, + { + "epoch": 7.78, + "learning_rate": 3.938038601861816e-08, + "loss": 0.0003, + "step": 59398 + }, + { + "epoch": 7.78, + "learning_rate": 3.9333381484606413e-08, + "loss": 0.0002, + "step": 59399 + }, + { + "epoch": 7.78, + "learning_rate": 3.928640496443925e-08, + "loss": 0.0006, + "step": 59400 + }, + { + "epoch": 7.78, + "learning_rate": 3.923945645824767e-08, + "loss": 0.0002, + "step": 59401 + }, + { + "epoch": 7.78, + "learning_rate": 3.9192535966166013e-08, + "loss": 0.0006, + "step": 59402 + }, + { + "epoch": 7.78, + "learning_rate": 3.914564348832417e-08, + "loss": 0.0006, + "step": 59403 + }, + { + "epoch": 7.78, + "learning_rate": 3.909877902485426e-08, + "loss": 0.0002, + "step": 59404 + }, + { + "epoch": 7.78, + "learning_rate": 3.9051942575889514e-08, + "loss": 0.0003, + "step": 59405 + }, + { + "epoch": 7.78, + "learning_rate": 3.900513414155982e-08, + "loss": 0.0003, + "step": 59406 + }, + { + "epoch": 7.78, + "learning_rate": 3.895835372199841e-08, + "loss": 0.0006, + "step": 59407 + }, + { + "epoch": 7.78, + "learning_rate": 3.8911601317335176e-08, + "loss": 0.0003, + "step": 59408 + }, + { + "epoch": 7.78, + "learning_rate": 3.886487692770336e-08, + "loss": 0.0003, + "step": 59409 + }, + { + "epoch": 7.78, + "learning_rate": 3.881818055323394e-08, + "loss": 0.0008, + "step": 59410 + }, + { + "epoch": 7.78, + "learning_rate": 3.877151219405795e-08, + "loss": 0.0002, + "step": 59411 + }, + { + "epoch": 7.78, + "learning_rate": 3.872487185030638e-08, + "loss": 0.0002, + "step": 59412 + }, + { + "epoch": 7.78, + "learning_rate": 3.867825952211024e-08, + "loss": 0.0003, + "step": 59413 + }, + { + "epoch": 7.78, + "learning_rate": 3.863167520960165e-08, + "loss": 0.0003, + "step": 59414 + }, + { + "epoch": 7.78, + "learning_rate": 3.85851189129105e-08, + "loss": 0.0003, + "step": 59415 + }, + { + "epoch": 7.78, + "learning_rate": 3.853859063216781e-08, + "loss": 0.0002, + "step": 59416 + }, + { + "epoch": 7.78, + "learning_rate": 3.849209036750567e-08, + "loss": 0.0004, + "step": 59417 + }, + { + "epoch": 7.78, + "learning_rate": 3.84456181190529e-08, + "loss": 0.0002, + "step": 59418 + }, + { + "epoch": 7.78, + "learning_rate": 3.8399173886942695e-08, + "loss": 0.0004, + "step": 59419 + }, + { + "epoch": 7.78, + "learning_rate": 3.8352757671303864e-08, + "loss": 0.0005, + "step": 59420 + }, + { + "epoch": 7.78, + "learning_rate": 3.83063694722674e-08, + "loss": 0.0007, + "step": 59421 + }, + { + "epoch": 7.78, + "learning_rate": 3.826000928996432e-08, + "loss": 0.0009, + "step": 59422 + }, + { + "epoch": 7.78, + "learning_rate": 3.82136771245234e-08, + "loss": 0.0003, + "step": 59423 + }, + { + "epoch": 7.78, + "learning_rate": 3.8167372976076755e-08, + "loss": 0.0003, + "step": 59424 + }, + { + "epoch": 7.78, + "learning_rate": 3.812109684475318e-08, + "loss": 0.0014, + "step": 59425 + }, + { + "epoch": 7.78, + "learning_rate": 3.8074848730684786e-08, + "loss": 0.0003, + "step": 59426 + }, + { + "epoch": 7.78, + "learning_rate": 3.8028628633999254e-08, + "loss": 0.0002, + "step": 59427 + }, + { + "epoch": 7.78, + "learning_rate": 3.798243655482869e-08, + "loss": 0.0005, + "step": 59428 + }, + { + "epoch": 7.78, + "learning_rate": 3.793627249330078e-08, + "loss": 0.0004, + "step": 59429 + }, + { + "epoch": 7.78, + "learning_rate": 3.789013644954764e-08, + "loss": 0.001, + "step": 59430 + }, + { + "epoch": 7.79, + "learning_rate": 3.7844028423696945e-08, + "loss": 0.0006, + "step": 59431 + }, + { + "epoch": 7.79, + "learning_rate": 3.779794841587969e-08, + "loss": 0.0006, + "step": 59432 + }, + { + "epoch": 7.79, + "learning_rate": 3.775189642622579e-08, + "loss": 0.0003, + "step": 59433 + }, + { + "epoch": 7.79, + "learning_rate": 3.770587245486401e-08, + "loss": 0.0004, + "step": 59434 + }, + { + "epoch": 7.79, + "learning_rate": 3.7659876501924263e-08, + "loss": 0.0002, + "step": 59435 + }, + { + "epoch": 7.79, + "learning_rate": 3.7613908567534216e-08, + "loss": 0.0005, + "step": 59436 + }, + { + "epoch": 7.79, + "learning_rate": 3.756796865182599e-08, + "loss": 0.0003, + "step": 59437 + }, + { + "epoch": 7.79, + "learning_rate": 3.752205675492726e-08, + "loss": 0.0002, + "step": 59438 + }, + { + "epoch": 7.79, + "learning_rate": 3.74761728769657e-08, + "loss": 0.0003, + "step": 59439 + }, + { + "epoch": 7.79, + "learning_rate": 3.7430317018073423e-08, + "loss": 0.0004, + "step": 59440 + }, + { + "epoch": 7.79, + "learning_rate": 3.7384489178377005e-08, + "loss": 0.0005, + "step": 59441 + }, + { + "epoch": 7.79, + "learning_rate": 3.733868935800633e-08, + "loss": 0.0008, + "step": 59442 + }, + { + "epoch": 7.79, + "learning_rate": 3.729291755709019e-08, + "loss": 0.0007, + "step": 59443 + }, + { + "epoch": 7.79, + "learning_rate": 3.724717377575737e-08, + "loss": 0.0004, + "step": 59444 + }, + { + "epoch": 7.79, + "learning_rate": 3.720145801413555e-08, + "loss": 0.0011, + "step": 59445 + }, + { + "epoch": 7.79, + "learning_rate": 3.7155770272355726e-08, + "loss": 0.0012, + "step": 59446 + }, + { + "epoch": 7.79, + "learning_rate": 3.7110110550543366e-08, + "loss": 0.0002, + "step": 59447 + }, + { + "epoch": 7.79, + "learning_rate": 3.706447884882835e-08, + "loss": 0.0007, + "step": 59448 + }, + { + "epoch": 7.79, + "learning_rate": 3.7018875167339484e-08, + "loss": 0.0003, + "step": 59449 + }, + { + "epoch": 7.79, + "learning_rate": 3.697329950620443e-08, + "loss": 0.0002, + "step": 59450 + }, + { + "epoch": 7.79, + "learning_rate": 3.692775186554976e-08, + "loss": 0.0007, + "step": 59451 + }, + { + "epoch": 7.79, + "learning_rate": 3.6882232245506465e-08, + "loss": 0.0004, + "step": 59452 + }, + { + "epoch": 7.79, + "learning_rate": 3.683674064620224e-08, + "loss": 0.0002, + "step": 59453 + }, + { + "epoch": 7.79, + "learning_rate": 3.679127706776364e-08, + "loss": 0.0002, + "step": 59454 + }, + { + "epoch": 7.79, + "learning_rate": 3.674584151031835e-08, + "loss": 0.0002, + "step": 59455 + }, + { + "epoch": 7.79, + "learning_rate": 3.670043397399514e-08, + "loss": 0.0003, + "step": 59456 + }, + { + "epoch": 7.79, + "learning_rate": 3.665505445892059e-08, + "loss": 0.0014, + "step": 59457 + }, + { + "epoch": 7.79, + "learning_rate": 3.660970296522459e-08, + "loss": 0.0008, + "step": 59458 + }, + { + "epoch": 7.79, + "learning_rate": 3.65643794930326e-08, + "loss": 0.0004, + "step": 59459 + }, + { + "epoch": 7.79, + "learning_rate": 3.651908404247229e-08, + "loss": 0.0003, + "step": 59460 + }, + { + "epoch": 7.79, + "learning_rate": 3.647381661367244e-08, + "loss": 0.0003, + "step": 59461 + }, + { + "epoch": 7.79, + "learning_rate": 3.642857720675852e-08, + "loss": 0.0002, + "step": 59462 + }, + { + "epoch": 7.79, + "learning_rate": 3.63833658218582e-08, + "loss": 0.0007, + "step": 59463 + }, + { + "epoch": 7.79, + "learning_rate": 3.633818245910026e-08, + "loss": 0.0005, + "step": 59464 + }, + { + "epoch": 7.79, + "learning_rate": 3.629302711861016e-08, + "loss": 0.0008, + "step": 59465 + }, + { + "epoch": 7.79, + "learning_rate": 3.6247899800514466e-08, + "loss": 0.0003, + "step": 59466 + }, + { + "epoch": 7.79, + "learning_rate": 3.620280050494196e-08, + "loss": 0.0005, + "step": 59467 + }, + { + "epoch": 7.79, + "learning_rate": 3.61577292320181e-08, + "loss": 0.0004, + "step": 59468 + }, + { + "epoch": 7.79, + "learning_rate": 3.611268598187056e-08, + "loss": 0.0003, + "step": 59469 + }, + { + "epoch": 7.79, + "learning_rate": 3.6067670754624806e-08, + "loss": 0.0003, + "step": 59470 + }, + { + "epoch": 7.79, + "learning_rate": 3.60226835504085e-08, + "loss": 0.0005, + "step": 59471 + }, + { + "epoch": 7.79, + "learning_rate": 3.597772436934821e-08, + "loss": 0.0005, + "step": 59472 + }, + { + "epoch": 7.79, + "learning_rate": 3.593279321156828e-08, + "loss": 0.0009, + "step": 59473 + }, + { + "epoch": 7.79, + "learning_rate": 3.588789007719862e-08, + "loss": 0.0003, + "step": 59474 + }, + { + "epoch": 7.79, + "learning_rate": 3.584301496636244e-08, + "loss": 0.0004, + "step": 59475 + }, + { + "epoch": 7.79, + "learning_rate": 3.579816787918744e-08, + "loss": 0.0005, + "step": 59476 + }, + { + "epoch": 7.79, + "learning_rate": 3.575334881580017e-08, + "loss": 0.0002, + "step": 59477 + }, + { + "epoch": 7.79, + "learning_rate": 3.570855777632498e-08, + "loss": 0.0004, + "step": 59478 + }, + { + "epoch": 7.79, + "learning_rate": 3.5663794760889546e-08, + "loss": 0.0005, + "step": 59479 + }, + { + "epoch": 7.79, + "learning_rate": 3.561905976961932e-08, + "loss": 0.0004, + "step": 59480 + }, + { + "epoch": 7.79, + "learning_rate": 3.557435280263866e-08, + "loss": 0.0003, + "step": 59481 + }, + { + "epoch": 7.79, + "learning_rate": 3.552967386007522e-08, + "loss": 0.0004, + "step": 59482 + }, + { + "epoch": 7.79, + "learning_rate": 3.548502294205447e-08, + "loss": 0.0002, + "step": 59483 + }, + { + "epoch": 7.79, + "learning_rate": 3.544040004870075e-08, + "loss": 0.0002, + "step": 59484 + }, + { + "epoch": 7.79, + "learning_rate": 3.539580518014063e-08, + "loss": 0.0006, + "step": 59485 + }, + { + "epoch": 7.79, + "learning_rate": 3.535123833649845e-08, + "loss": 0.0005, + "step": 59486 + }, + { + "epoch": 7.79, + "learning_rate": 3.530669951790078e-08, + "loss": 0.0003, + "step": 59487 + }, + { + "epoch": 7.79, + "learning_rate": 3.5262188724473065e-08, + "loss": 0.0009, + "step": 59488 + }, + { + "epoch": 7.79, + "learning_rate": 3.521770595633855e-08, + "loss": 0.0008, + "step": 59489 + }, + { + "epoch": 7.79, + "learning_rate": 3.5173251213623804e-08, + "loss": 0.0003, + "step": 59490 + }, + { + "epoch": 7.79, + "learning_rate": 3.5128824496454275e-08, + "loss": 0.0005, + "step": 59491 + }, + { + "epoch": 7.79, + "learning_rate": 3.50844258049543e-08, + "loss": 0.0006, + "step": 59492 + }, + { + "epoch": 7.79, + "learning_rate": 3.504005513924824e-08, + "loss": 0.0008, + "step": 59493 + }, + { + "epoch": 7.79, + "learning_rate": 3.4995712499461546e-08, + "loss": 0.0007, + "step": 59494 + }, + { + "epoch": 7.79, + "learning_rate": 3.4951397885718554e-08, + "loss": 0.0008, + "step": 59495 + }, + { + "epoch": 7.79, + "learning_rate": 3.4907111298143615e-08, + "loss": 0.0002, + "step": 59496 + }, + { + "epoch": 7.79, + "learning_rate": 3.486285273686218e-08, + "loss": 0.0002, + "step": 59497 + }, + { + "epoch": 7.79, + "learning_rate": 3.481862220199861e-08, + "loss": 0.0002, + "step": 59498 + }, + { + "epoch": 7.79, + "learning_rate": 3.477441969367723e-08, + "loss": 0.0003, + "step": 59499 + }, + { + "epoch": 7.79, + "learning_rate": 3.473024521202128e-08, + "loss": 0.0004, + "step": 59500 + }, + { + "epoch": 7.79, + "learning_rate": 3.468609875715623e-08, + "loss": 0.0006, + "step": 59501 + }, + { + "epoch": 7.79, + "learning_rate": 3.46419803292064e-08, + "loss": 0.0003, + "step": 59502 + }, + { + "epoch": 7.79, + "learning_rate": 3.459788992829505e-08, + "loss": 0.0006, + "step": 59503 + }, + { + "epoch": 7.79, + "learning_rate": 3.455382755454539e-08, + "loss": 0.0003, + "step": 59504 + }, + { + "epoch": 7.79, + "learning_rate": 3.4509793208084006e-08, + "loss": 0.0003, + "step": 59505 + }, + { + "epoch": 7.79, + "learning_rate": 3.44657868890319e-08, + "loss": 0.0003, + "step": 59506 + }, + { + "epoch": 7.79, + "learning_rate": 3.442180859751454e-08, + "loss": 0.0006, + "step": 59507 + }, + { + "epoch": 7.8, + "learning_rate": 3.4377858333655145e-08, + "loss": 0.0008, + "step": 59508 + }, + { + "epoch": 7.8, + "learning_rate": 3.433393609757696e-08, + "loss": 0.0008, + "step": 59509 + }, + { + "epoch": 7.8, + "learning_rate": 3.429004188940432e-08, + "loss": 0.0002, + "step": 59510 + }, + { + "epoch": 7.8, + "learning_rate": 3.424617570926048e-08, + "loss": 0.0003, + "step": 59511 + }, + { + "epoch": 7.8, + "learning_rate": 3.420233755726754e-08, + "loss": 0.0004, + "step": 59512 + }, + { + "epoch": 7.8, + "learning_rate": 3.415852743355097e-08, + "loss": 0.0005, + "step": 59513 + }, + { + "epoch": 7.8, + "learning_rate": 3.411474533823178e-08, + "loss": 0.0003, + "step": 59514 + }, + { + "epoch": 7.8, + "learning_rate": 3.407099127143543e-08, + "loss": 0.0003, + "step": 59515 + }, + { + "epoch": 7.8, + "learning_rate": 3.4027265233282925e-08, + "loss": 0.0003, + "step": 59516 + }, + { + "epoch": 7.8, + "learning_rate": 3.398356722389751e-08, + "loss": 0.0005, + "step": 59517 + }, + { + "epoch": 7.8, + "learning_rate": 3.393989724340352e-08, + "loss": 0.0002, + "step": 59518 + }, + { + "epoch": 7.8, + "learning_rate": 3.389625529192198e-08, + "loss": 0.0002, + "step": 59519 + }, + { + "epoch": 7.8, + "learning_rate": 3.3852641369577224e-08, + "loss": 0.0005, + "step": 59520 + }, + { + "epoch": 7.8, + "learning_rate": 3.3809055476490274e-08, + "loss": 0.0005, + "step": 59521 + }, + { + "epoch": 7.8, + "learning_rate": 3.376549761278547e-08, + "loss": 0.0005, + "step": 59522 + }, + { + "epoch": 7.8, + "learning_rate": 3.372196777858494e-08, + "loss": 0.0002, + "step": 59523 + }, + { + "epoch": 7.8, + "learning_rate": 3.36784659740097e-08, + "loss": 0.0003, + "step": 59524 + }, + { + "epoch": 7.8, + "learning_rate": 3.363499219918298e-08, + "loss": 0.0004, + "step": 59525 + }, + { + "epoch": 7.8, + "learning_rate": 3.359154645422802e-08, + "loss": 0.0004, + "step": 59526 + }, + { + "epoch": 7.8, + "learning_rate": 3.354812873926472e-08, + "loss": 0.0003, + "step": 59527 + }, + { + "epoch": 7.8, + "learning_rate": 3.350473905441742e-08, + "loss": 0.0002, + "step": 59528 + }, + { + "epoch": 7.8, + "learning_rate": 3.346137739980715e-08, + "loss": 0.0002, + "step": 59529 + }, + { + "epoch": 7.8, + "learning_rate": 3.341804377555713e-08, + "loss": 0.0002, + "step": 59530 + }, + { + "epoch": 7.8, + "learning_rate": 3.3374738181786157e-08, + "loss": 0.0004, + "step": 59531 + }, + { + "epoch": 7.8, + "learning_rate": 3.333146061861969e-08, + "loss": 0.0003, + "step": 59532 + }, + { + "epoch": 7.8, + "learning_rate": 3.328821108617653e-08, + "loss": 0.0006, + "step": 59533 + }, + { + "epoch": 7.8, + "learning_rate": 3.324498958457989e-08, + "loss": 0.0003, + "step": 59534 + }, + { + "epoch": 7.8, + "learning_rate": 3.32017961139508e-08, + "loss": 0.0011, + "step": 59535 + }, + { + "epoch": 7.8, + "learning_rate": 3.3158630674411385e-08, + "loss": 0.0011, + "step": 59536 + }, + { + "epoch": 7.8, + "learning_rate": 3.311549326608265e-08, + "loss": 0.0003, + "step": 59537 + }, + { + "epoch": 7.8, + "learning_rate": 3.3072383889085625e-08, + "loss": 0.0002, + "step": 59538 + }, + { + "epoch": 7.8, + "learning_rate": 3.3029302543540196e-08, + "loss": 0.0002, + "step": 59539 + }, + { + "epoch": 7.8, + "learning_rate": 3.298624922957072e-08, + "loss": 0.0003, + "step": 59540 + }, + { + "epoch": 7.8, + "learning_rate": 3.29432239472971e-08, + "loss": 0.0005, + "step": 59541 + }, + { + "epoch": 7.8, + "learning_rate": 3.290022669683923e-08, + "loss": 0.0003, + "step": 59542 + }, + { + "epoch": 7.8, + "learning_rate": 3.2857257478318136e-08, + "loss": 0.0002, + "step": 59543 + }, + { + "epoch": 7.8, + "learning_rate": 3.281431629185594e-08, + "loss": 0.0004, + "step": 59544 + }, + { + "epoch": 7.8, + "learning_rate": 3.2771403137572546e-08, + "loss": 0.0004, + "step": 59545 + }, + { + "epoch": 7.8, + "learning_rate": 3.272851801558896e-08, + "loss": 0.0003, + "step": 59546 + }, + { + "epoch": 7.8, + "learning_rate": 3.26856609260251e-08, + "loss": 0.0004, + "step": 59547 + }, + { + "epoch": 7.8, + "learning_rate": 3.264283186900197e-08, + "loss": 0.0002, + "step": 59548 + }, + { + "epoch": 7.8, + "learning_rate": 3.2600030844640586e-08, + "loss": 0.0005, + "step": 59549 + }, + { + "epoch": 7.8, + "learning_rate": 3.2557257853060855e-08, + "loss": 0.0006, + "step": 59550 + }, + { + "epoch": 7.8, + "learning_rate": 3.251451289438268e-08, + "loss": 0.0006, + "step": 59551 + }, + { + "epoch": 7.8, + "learning_rate": 3.247179596872707e-08, + "loss": 0.0002, + "step": 59552 + }, + { + "epoch": 7.8, + "learning_rate": 3.242910707621283e-08, + "loss": 0.0009, + "step": 59553 + }, + { + "epoch": 7.8, + "learning_rate": 3.2386446216962076e-08, + "loss": 0.0008, + "step": 59554 + }, + { + "epoch": 7.8, + "learning_rate": 3.2343813391092494e-08, + "loss": 0.0003, + "step": 59555 + }, + { + "epoch": 7.8, + "learning_rate": 3.230120859872621e-08, + "loss": 0.0003, + "step": 59556 + }, + { + "epoch": 7.8, + "learning_rate": 3.2258631839982016e-08, + "loss": 0.0003, + "step": 59557 + }, + { + "epoch": 7.8, + "learning_rate": 3.22160831149787e-08, + "loss": 0.0006, + "step": 59558 + }, + { + "epoch": 7.8, + "learning_rate": 3.217356242383729e-08, + "loss": 0.0003, + "step": 59559 + }, + { + "epoch": 7.8, + "learning_rate": 3.213106976667657e-08, + "loss": 0.0004, + "step": 59560 + }, + { + "epoch": 7.8, + "learning_rate": 3.208860514361645e-08, + "loss": 0.0004, + "step": 59561 + }, + { + "epoch": 7.8, + "learning_rate": 3.2046168554776824e-08, + "loss": 0.0003, + "step": 59562 + }, + { + "epoch": 7.8, + "learning_rate": 3.200376000027649e-08, + "loss": 0.0003, + "step": 59563 + }, + { + "epoch": 7.8, + "learning_rate": 3.196137948023426e-08, + "loss": 0.0004, + "step": 59564 + }, + { + "epoch": 7.8, + "learning_rate": 3.191902699477001e-08, + "loss": 0.0004, + "step": 59565 + }, + { + "epoch": 7.8, + "learning_rate": 3.187670254400255e-08, + "loss": 0.0003, + "step": 59566 + }, + { + "epoch": 7.8, + "learning_rate": 3.183440612805178e-08, + "loss": 0.0003, + "step": 59567 + }, + { + "epoch": 7.8, + "learning_rate": 3.179213774703538e-08, + "loss": 0.0003, + "step": 59568 + }, + { + "epoch": 7.8, + "learning_rate": 3.174989740107215e-08, + "loss": 0.0006, + "step": 59569 + }, + { + "epoch": 7.8, + "learning_rate": 3.170768509028199e-08, + "loss": 0.0002, + "step": 59570 + }, + { + "epoch": 7.8, + "learning_rate": 3.166550081478259e-08, + "loss": 0.0004, + "step": 59571 + }, + { + "epoch": 7.8, + "learning_rate": 3.162334457469385e-08, + "loss": 0.0002, + "step": 59572 + }, + { + "epoch": 7.8, + "learning_rate": 3.158121637013345e-08, + "loss": 0.0004, + "step": 59573 + }, + { + "epoch": 7.8, + "learning_rate": 3.153911620122019e-08, + "loss": 0.0002, + "step": 59574 + }, + { + "epoch": 7.8, + "learning_rate": 3.1497044068072857e-08, + "loss": 0.0007, + "step": 59575 + }, + { + "epoch": 7.8, + "learning_rate": 3.145499997080803e-08, + "loss": 0.0004, + "step": 59576 + }, + { + "epoch": 7.8, + "learning_rate": 3.1412983909545616e-08, + "loss": 0.0022, + "step": 59577 + }, + { + "epoch": 7.8, + "learning_rate": 3.1370995884403285e-08, + "loss": 0.0009, + "step": 59578 + }, + { + "epoch": 7.8, + "learning_rate": 3.132903589549985e-08, + "loss": 0.0003, + "step": 59579 + }, + { + "epoch": 7.8, + "learning_rate": 3.128710394295187e-08, + "loss": 0.0005, + "step": 59580 + }, + { + "epoch": 7.8, + "learning_rate": 3.124520002687814e-08, + "loss": 0.0003, + "step": 59581 + }, + { + "epoch": 7.8, + "learning_rate": 3.1203324147396355e-08, + "loss": 0.0007, + "step": 59582 + }, + { + "epoch": 7.8, + "learning_rate": 3.1161476304624186e-08, + "loss": 0.001, + "step": 59583 + }, + { + "epoch": 7.81, + "learning_rate": 3.111965649868043e-08, + "loss": 0.0026, + "step": 59584 + }, + { + "epoch": 7.81, + "learning_rate": 3.1077864729680554e-08, + "loss": 0.0003, + "step": 59585 + }, + { + "epoch": 7.81, + "learning_rate": 3.1036100997744454e-08, + "loss": 0.0002, + "step": 59586 + }, + { + "epoch": 7.81, + "learning_rate": 3.09943653029876e-08, + "loss": 0.0005, + "step": 59587 + }, + { + "epoch": 7.81, + "learning_rate": 3.095265764552879e-08, + "loss": 0.0007, + "step": 59588 + }, + { + "epoch": 7.81, + "learning_rate": 3.091097802548459e-08, + "loss": 0.0003, + "step": 59589 + }, + { + "epoch": 7.81, + "learning_rate": 3.0869326442971584e-08, + "loss": 0.0008, + "step": 59590 + }, + { + "epoch": 7.81, + "learning_rate": 3.0827702898108545e-08, + "loss": 0.0008, + "step": 59591 + }, + { + "epoch": 7.81, + "learning_rate": 3.0786107391010955e-08, + "loss": 0.0005, + "step": 59592 + }, + { + "epoch": 7.81, + "learning_rate": 3.07445399217976e-08, + "loss": 0.0002, + "step": 59593 + }, + { + "epoch": 7.81, + "learning_rate": 3.070300049058395e-08, + "loss": 0.0005, + "step": 59594 + }, + { + "epoch": 7.81, + "learning_rate": 3.066148909748656e-08, + "loss": 0.0007, + "step": 59595 + }, + { + "epoch": 7.81, + "learning_rate": 3.062000574262314e-08, + "loss": 0.0008, + "step": 59596 + }, + { + "epoch": 7.81, + "learning_rate": 3.0578550426111353e-08, + "loss": 0.0007, + "step": 59597 + }, + { + "epoch": 7.81, + "learning_rate": 3.053712314806556e-08, + "loss": 0.0003, + "step": 59598 + }, + { + "epoch": 7.81, + "learning_rate": 3.0495723908602335e-08, + "loss": 0.0005, + "step": 59599 + }, + { + "epoch": 7.81, + "learning_rate": 3.0454352707840475e-08, + "loss": 0.0003, + "step": 59600 + }, + { + "epoch": 7.81, + "learning_rate": 3.0413009545893215e-08, + "loss": 0.0006, + "step": 59601 + }, + { + "epoch": 7.81, + "learning_rate": 3.037169442287935e-08, + "loss": 0.0002, + "step": 59602 + }, + { + "epoch": 7.81, + "learning_rate": 3.033040733891435e-08, + "loss": 0.001, + "step": 59603 + }, + { + "epoch": 7.81, + "learning_rate": 3.028914829411367e-08, + "loss": 0.0003, + "step": 59604 + }, + { + "epoch": 7.81, + "learning_rate": 3.0247917288594995e-08, + "loss": 0.0002, + "step": 59605 + }, + { + "epoch": 7.81, + "learning_rate": 3.020671432247157e-08, + "loss": 0.0002, + "step": 59606 + }, + { + "epoch": 7.81, + "learning_rate": 3.01655393958622e-08, + "loss": 0.0003, + "step": 59607 + }, + { + "epoch": 7.81, + "learning_rate": 3.012439250888011e-08, + "loss": 0.0004, + "step": 59608 + }, + { + "epoch": 7.81, + "learning_rate": 3.008327366164299e-08, + "loss": 0.001, + "step": 59609 + }, + { + "epoch": 7.81, + "learning_rate": 3.004218285426519e-08, + "loss": 0.0006, + "step": 59610 + }, + { + "epoch": 7.81, + "learning_rate": 3.000112008686329e-08, + "loss": 0.0005, + "step": 59611 + }, + { + "epoch": 7.81, + "learning_rate": 2.996008535955275e-08, + "loss": 0.0005, + "step": 59612 + }, + { + "epoch": 7.81, + "learning_rate": 2.991907867244792e-08, + "loss": 0.0009, + "step": 59613 + }, + { + "epoch": 7.81, + "learning_rate": 2.987810002566427e-08, + "loss": 0.0002, + "step": 59614 + }, + { + "epoch": 7.81, + "learning_rate": 2.983714941931837e-08, + "loss": 0.0005, + "step": 59615 + }, + { + "epoch": 7.81, + "learning_rate": 2.9796226853523456e-08, + "loss": 0.0003, + "step": 59616 + }, + { + "epoch": 7.81, + "learning_rate": 2.9755332328397225e-08, + "loss": 0.0008, + "step": 59617 + }, + { + "epoch": 7.81, + "learning_rate": 2.971446584405291e-08, + "loss": 0.0004, + "step": 59618 + }, + { + "epoch": 7.81, + "learning_rate": 2.9673627400604866e-08, + "loss": 0.0005, + "step": 59619 + }, + { + "epoch": 7.81, + "learning_rate": 2.9632816998168555e-08, + "loss": 0.0003, + "step": 59620 + }, + { + "epoch": 7.81, + "learning_rate": 2.9592034636860557e-08, + "loss": 0.0026, + "step": 59621 + }, + { + "epoch": 7.81, + "learning_rate": 2.9551280316792997e-08, + "loss": 0.0003, + "step": 59622 + }, + { + "epoch": 7.81, + "learning_rate": 2.951055403808134e-08, + "loss": 0.0002, + "step": 59623 + }, + { + "epoch": 7.81, + "learning_rate": 2.946985580083994e-08, + "loss": 0.0005, + "step": 59624 + }, + { + "epoch": 7.81, + "learning_rate": 2.9429185605184263e-08, + "loss": 0.0005, + "step": 59625 + }, + { + "epoch": 7.81, + "learning_rate": 2.9388543451228658e-08, + "loss": 0.0003, + "step": 59626 + }, + { + "epoch": 7.81, + "learning_rate": 2.9347929339085258e-08, + "loss": 0.0002, + "step": 59627 + }, + { + "epoch": 7.81, + "learning_rate": 2.9307343268870637e-08, + "loss": 0.0002, + "step": 59628 + }, + { + "epoch": 7.81, + "learning_rate": 2.9266785240698038e-08, + "loss": 0.0006, + "step": 59629 + }, + { + "epoch": 7.81, + "learning_rate": 2.9226255254681812e-08, + "loss": 0.001, + "step": 59630 + }, + { + "epoch": 7.81, + "learning_rate": 2.9185753310936317e-08, + "loss": 0.0006, + "step": 59631 + }, + { + "epoch": 7.81, + "learning_rate": 2.914527940957368e-08, + "loss": 0.0002, + "step": 59632 + }, + { + "epoch": 7.81, + "learning_rate": 2.9104833550710476e-08, + "loss": 0.0006, + "step": 59633 + }, + { + "epoch": 7.81, + "learning_rate": 2.906441573445773e-08, + "loss": 0.0004, + "step": 59634 + }, + { + "epoch": 7.81, + "learning_rate": 2.90240259609309e-08, + "loss": 0.0007, + "step": 59635 + }, + { + "epoch": 7.81, + "learning_rate": 2.8983664230243237e-08, + "loss": 0.0002, + "step": 59636 + }, + { + "epoch": 7.81, + "learning_rate": 2.8943330542506864e-08, + "loss": 0.0003, + "step": 59637 + }, + { + "epoch": 7.81, + "learning_rate": 2.890302489783725e-08, + "loss": 0.0006, + "step": 59638 + }, + { + "epoch": 7.81, + "learning_rate": 2.8862747296346527e-08, + "loss": 0.0005, + "step": 59639 + }, + { + "epoch": 7.81, + "learning_rate": 2.8822497738149047e-08, + "loss": 0.0005, + "step": 59640 + }, + { + "epoch": 7.81, + "learning_rate": 2.878227622335694e-08, + "loss": 0.0009, + "step": 59641 + }, + { + "epoch": 7.81, + "learning_rate": 2.8742082752082344e-08, + "loss": 0.0014, + "step": 59642 + }, + { + "epoch": 7.81, + "learning_rate": 2.870191732444183e-08, + "loss": 0.0002, + "step": 59643 + }, + { + "epoch": 7.81, + "learning_rate": 2.866177994054531e-08, + "loss": 0.0006, + "step": 59644 + }, + { + "epoch": 7.81, + "learning_rate": 2.8621670600506023e-08, + "loss": 0.0002, + "step": 59645 + }, + { + "epoch": 7.81, + "learning_rate": 2.858158930443722e-08, + "loss": 0.0002, + "step": 59646 + }, + { + "epoch": 7.81, + "learning_rate": 2.8541536052452134e-08, + "loss": 0.0007, + "step": 59647 + }, + { + "epoch": 7.81, + "learning_rate": 2.8501510844662905e-08, + "loss": 0.0002, + "step": 59648 + }, + { + "epoch": 7.81, + "learning_rate": 2.8461513681182772e-08, + "loss": 0.0003, + "step": 59649 + }, + { + "epoch": 7.81, + "learning_rate": 2.842154456212276e-08, + "loss": 0.001, + "step": 59650 + }, + { + "epoch": 7.81, + "learning_rate": 2.838160348759611e-08, + "loss": 0.001, + "step": 59651 + }, + { + "epoch": 7.81, + "learning_rate": 2.8341690457716065e-08, + "loss": 0.0003, + "step": 59652 + }, + { + "epoch": 7.81, + "learning_rate": 2.8301805472593647e-08, + "loss": 0.0002, + "step": 59653 + }, + { + "epoch": 7.81, + "learning_rate": 2.826194853234099e-08, + "loss": 0.0007, + "step": 59654 + }, + { + "epoch": 7.81, + "learning_rate": 2.8222119637070223e-08, + "loss": 0.0004, + "step": 59655 + }, + { + "epoch": 7.81, + "learning_rate": 2.818231878689459e-08, + "loss": 0.0006, + "step": 59656 + }, + { + "epoch": 7.81, + "learning_rate": 2.8142545981925118e-08, + "loss": 0.0007, + "step": 59657 + }, + { + "epoch": 7.81, + "learning_rate": 2.8102801222272824e-08, + "loss": 0.001, + "step": 59658 + }, + { + "epoch": 7.81, + "learning_rate": 2.8063084508050952e-08, + "loss": 0.0007, + "step": 59659 + }, + { + "epoch": 7.82, + "learning_rate": 2.8023395839370527e-08, + "loss": 0.0001, + "step": 59660 + }, + { + "epoch": 7.82, + "learning_rate": 2.798373521634368e-08, + "loss": 0.0006, + "step": 59661 + }, + { + "epoch": 7.82, + "learning_rate": 2.7944102639081427e-08, + "loss": 0.0006, + "step": 59662 + }, + { + "epoch": 7.82, + "learning_rate": 2.79044981076948e-08, + "loss": 0.0003, + "step": 59663 + }, + { + "epoch": 7.82, + "learning_rate": 2.7864921622295925e-08, + "loss": 0.0006, + "step": 59664 + }, + { + "epoch": 7.82, + "learning_rate": 2.782537318299583e-08, + "loss": 0.001, + "step": 59665 + }, + { + "epoch": 7.82, + "learning_rate": 2.7785852789906644e-08, + "loss": 0.0003, + "step": 59666 + }, + { + "epoch": 7.82, + "learning_rate": 2.7746360443138274e-08, + "loss": 0.0004, + "step": 59667 + }, + { + "epoch": 7.82, + "learning_rate": 2.7706896142801752e-08, + "loss": 0.0006, + "step": 59668 + }, + { + "epoch": 7.82, + "learning_rate": 2.7667459889008098e-08, + "loss": 0.0004, + "step": 59669 + }, + { + "epoch": 7.82, + "learning_rate": 2.7628051681869438e-08, + "loss": 0.0002, + "step": 59670 + }, + { + "epoch": 7.82, + "learning_rate": 2.758867152149569e-08, + "loss": 0.0006, + "step": 59671 + }, + { + "epoch": 7.82, + "learning_rate": 2.7549319407997877e-08, + "loss": 0.0002, + "step": 59672 + }, + { + "epoch": 7.82, + "learning_rate": 2.7509995341485907e-08, + "loss": 0.0003, + "step": 59673 + }, + { + "epoch": 7.82, + "learning_rate": 2.7470699322071915e-08, + "loss": 0.0007, + "step": 59674 + }, + { + "epoch": 7.82, + "learning_rate": 2.7431431349865812e-08, + "loss": 0.0004, + "step": 59675 + }, + { + "epoch": 7.82, + "learning_rate": 2.73921914249764e-08, + "loss": 0.0003, + "step": 59676 + }, + { + "epoch": 7.82, + "learning_rate": 2.735297954751581e-08, + "loss": 0.0008, + "step": 59677 + }, + { + "epoch": 7.82, + "learning_rate": 2.731379571759507e-08, + "loss": 0.0002, + "step": 59678 + }, + { + "epoch": 7.82, + "learning_rate": 2.7274639935321867e-08, + "loss": 0.0002, + "step": 59679 + }, + { + "epoch": 7.82, + "learning_rate": 2.7235512200808334e-08, + "loss": 0.0006, + "step": 59680 + }, + { + "epoch": 7.82, + "learning_rate": 2.7196412514163273e-08, + "loss": 0.0013, + "step": 59681 + }, + { + "epoch": 7.82, + "learning_rate": 2.7157340875497707e-08, + "loss": 0.0008, + "step": 59682 + }, + { + "epoch": 7.82, + "learning_rate": 2.7118297284920437e-08, + "loss": 0.0006, + "step": 59683 + }, + { + "epoch": 7.82, + "learning_rate": 2.7079281742542484e-08, + "loss": 0.0003, + "step": 59684 + }, + { + "epoch": 7.82, + "learning_rate": 2.7040294248472655e-08, + "loss": 0.0004, + "step": 59685 + }, + { + "epoch": 7.82, + "learning_rate": 2.7001334802820857e-08, + "loss": 0.0004, + "step": 59686 + }, + { + "epoch": 7.82, + "learning_rate": 2.6962403405697e-08, + "loss": 0.0004, + "step": 59687 + }, + { + "epoch": 7.82, + "learning_rate": 2.6923500057209895e-08, + "loss": 0.0004, + "step": 59688 + }, + { + "epoch": 7.82, + "learning_rate": 2.6884624757469445e-08, + "loss": 0.0002, + "step": 59689 + }, + { + "epoch": 7.82, + "learning_rate": 2.684577750658557e-08, + "loss": 0.0005, + "step": 59690 + }, + { + "epoch": 7.82, + "learning_rate": 2.6806958304665952e-08, + "loss": 0.0005, + "step": 59691 + }, + { + "epoch": 7.82, + "learning_rate": 2.6768167151820513e-08, + "loss": 0.0008, + "step": 59692 + }, + { + "epoch": 7.82, + "learning_rate": 2.6729404048159157e-08, + "loss": 0.0003, + "step": 59693 + }, + { + "epoch": 7.82, + "learning_rate": 2.669066899379069e-08, + "loss": 0.0003, + "step": 59694 + }, + { + "epoch": 7.82, + "learning_rate": 2.6651961988822805e-08, + "loss": 0.0003, + "step": 59695 + }, + { + "epoch": 7.82, + "learning_rate": 2.661328303336541e-08, + "loss": 0.0005, + "step": 59696 + }, + { + "epoch": 7.82, + "learning_rate": 2.6574632127527312e-08, + "loss": 0.0007, + "step": 59697 + }, + { + "epoch": 7.82, + "learning_rate": 2.6536009271416196e-08, + "loss": 0.0005, + "step": 59698 + }, + { + "epoch": 7.82, + "learning_rate": 2.649741446514198e-08, + "loss": 0.0004, + "step": 59699 + }, + { + "epoch": 7.82, + "learning_rate": 2.645884770881235e-08, + "loss": 0.0003, + "step": 59700 + }, + { + "epoch": 7.82, + "learning_rate": 2.6420309002536116e-08, + "loss": 0.0006, + "step": 59701 + }, + { + "epoch": 7.82, + "learning_rate": 2.6381798346422072e-08, + "loss": 0.0004, + "step": 59702 + }, + { + "epoch": 7.82, + "learning_rate": 2.6343315740577913e-08, + "loss": 0.0002, + "step": 59703 + }, + { + "epoch": 7.82, + "learning_rate": 2.630486118511133e-08, + "loss": 0.0009, + "step": 59704 + }, + { + "epoch": 7.82, + "learning_rate": 2.6266434680132236e-08, + "loss": 0.0005, + "step": 59705 + }, + { + "epoch": 7.82, + "learning_rate": 2.622803622574721e-08, + "loss": 0.0007, + "step": 59706 + }, + { + "epoch": 7.82, + "learning_rate": 2.6189665822065057e-08, + "loss": 0.0012, + "step": 59707 + }, + { + "epoch": 7.82, + "learning_rate": 2.615132346919236e-08, + "loss": 0.0006, + "step": 59708 + }, + { + "epoch": 7.82, + "learning_rate": 2.6113009167239022e-08, + "loss": 0.0003, + "step": 59709 + }, + { + "epoch": 7.82, + "learning_rate": 2.6074722916311634e-08, + "loss": 0.0003, + "step": 59710 + }, + { + "epoch": 7.82, + "learning_rate": 2.6036464716516775e-08, + "loss": 0.0009, + "step": 59711 + }, + { + "epoch": 7.82, + "learning_rate": 2.5998234567964355e-08, + "loss": 0.0004, + "step": 59712 + }, + { + "epoch": 7.82, + "learning_rate": 2.5960032470760953e-08, + "loss": 0.0008, + "step": 59713 + }, + { + "epoch": 7.82, + "learning_rate": 2.5921858425013158e-08, + "loss": 0.0002, + "step": 59714 + }, + { + "epoch": 7.82, + "learning_rate": 2.5883712430828657e-08, + "loss": 0.0012, + "step": 59715 + }, + { + "epoch": 7.82, + "learning_rate": 2.584559448831514e-08, + "loss": 0.0003, + "step": 59716 + }, + { + "epoch": 7.82, + "learning_rate": 2.58075045975803e-08, + "loss": 0.0006, + "step": 59717 + }, + { + "epoch": 7.82, + "learning_rate": 2.576944275873072e-08, + "loss": 0.0005, + "step": 59718 + }, + { + "epoch": 7.82, + "learning_rate": 2.5731408971872986e-08, + "loss": 0.0002, + "step": 59719 + }, + { + "epoch": 7.82, + "learning_rate": 2.5693403237114777e-08, + "loss": 0.0011, + "step": 59720 + }, + { + "epoch": 7.82, + "learning_rate": 2.5655425554562685e-08, + "loss": 0.0003, + "step": 59721 + }, + { + "epoch": 7.82, + "learning_rate": 2.5617475924323287e-08, + "loss": 0.0002, + "step": 59722 + }, + { + "epoch": 7.82, + "learning_rate": 2.5579554346504276e-08, + "loss": 0.0005, + "step": 59723 + }, + { + "epoch": 7.82, + "learning_rate": 2.5541660821211124e-08, + "loss": 0.0002, + "step": 59724 + }, + { + "epoch": 7.82, + "learning_rate": 2.5503795348550407e-08, + "loss": 0.0005, + "step": 59725 + }, + { + "epoch": 7.82, + "learning_rate": 2.5465957928629825e-08, + "loss": 0.0005, + "step": 59726 + }, + { + "epoch": 7.82, + "learning_rate": 2.5428148561554844e-08, + "loss": 0.0007, + "step": 59727 + }, + { + "epoch": 7.82, + "learning_rate": 2.5390367247432047e-08, + "loss": 0.0004, + "step": 59728 + }, + { + "epoch": 7.82, + "learning_rate": 2.5352613986368012e-08, + "loss": 0.0003, + "step": 59729 + }, + { + "epoch": 7.82, + "learning_rate": 2.5314888778467105e-08, + "loss": 0.0004, + "step": 59730 + }, + { + "epoch": 7.82, + "learning_rate": 2.5277191623838128e-08, + "loss": 0.0005, + "step": 59731 + }, + { + "epoch": 7.82, + "learning_rate": 2.5239522522586545e-08, + "loss": 0.0004, + "step": 59732 + }, + { + "epoch": 7.82, + "learning_rate": 2.5201881474815616e-08, + "loss": 0.0005, + "step": 59733 + }, + { + "epoch": 7.82, + "learning_rate": 2.5164268480635246e-08, + "loss": 0.0002, + "step": 59734 + }, + { + "epoch": 7.82, + "learning_rate": 2.512668354014758e-08, + "loss": 0.0002, + "step": 59735 + }, + { + "epoch": 7.82, + "learning_rate": 2.5089126653460304e-08, + "loss": 0.0006, + "step": 59736 + }, + { + "epoch": 7.83, + "learning_rate": 2.5051597820680007e-08, + "loss": 0.0007, + "step": 59737 + }, + { + "epoch": 7.83, + "learning_rate": 2.5014097041908824e-08, + "loss": 0.0009, + "step": 59738 + }, + { + "epoch": 7.83, + "learning_rate": 2.497662431725556e-08, + "loss": 0.0006, + "step": 59739 + }, + { + "epoch": 7.83, + "learning_rate": 2.4939179646824574e-08, + "loss": 0.0003, + "step": 59740 + }, + { + "epoch": 7.83, + "learning_rate": 2.4901763030720228e-08, + "loss": 0.0003, + "step": 59741 + }, + { + "epoch": 7.83, + "learning_rate": 2.48643744690491e-08, + "loss": 0.0002, + "step": 59742 + }, + { + "epoch": 7.83, + "learning_rate": 2.4827013961915558e-08, + "loss": 0.0011, + "step": 59743 + }, + { + "epoch": 7.83, + "learning_rate": 2.4789681509423958e-08, + "loss": 0.0004, + "step": 59744 + }, + { + "epoch": 7.83, + "learning_rate": 2.475237711168088e-08, + "loss": 0.0004, + "step": 59745 + }, + { + "epoch": 7.83, + "learning_rate": 2.471510076879069e-08, + "loss": 0.0007, + "step": 59746 + }, + { + "epoch": 7.83, + "learning_rate": 2.4677852480857746e-08, + "loss": 0.0003, + "step": 59747 + }, + { + "epoch": 7.83, + "learning_rate": 2.4640632247987517e-08, + "loss": 0.0006, + "step": 59748 + }, + { + "epoch": 7.83, + "learning_rate": 2.4603440070283257e-08, + "loss": 0.0007, + "step": 59749 + }, + { + "epoch": 7.83, + "learning_rate": 2.4566275947851548e-08, + "loss": 0.0004, + "step": 59750 + }, + { + "epoch": 7.83, + "learning_rate": 2.452913988079564e-08, + "loss": 0.0003, + "step": 59751 + }, + { + "epoch": 7.83, + "learning_rate": 2.4492031869221e-08, + "loss": 0.0003, + "step": 59752 + }, + { + "epoch": 7.83, + "learning_rate": 2.4454951913229774e-08, + "loss": 0.0003, + "step": 59753 + }, + { + "epoch": 7.83, + "learning_rate": 2.441790001292854e-08, + "loss": 0.0012, + "step": 59754 + }, + { + "epoch": 7.83, + "learning_rate": 2.4380876168421664e-08, + "loss": 0.0006, + "step": 59755 + }, + { + "epoch": 7.83, + "learning_rate": 2.434388037981128e-08, + "loss": 0.0003, + "step": 59756 + }, + { + "epoch": 7.83, + "learning_rate": 2.4306912647202862e-08, + "loss": 0.0007, + "step": 59757 + }, + { + "epoch": 7.83, + "learning_rate": 2.426997297069966e-08, + "loss": 0.0002, + "step": 59758 + }, + { + "epoch": 7.83, + "learning_rate": 2.423306135040604e-08, + "loss": 0.001, + "step": 59759 + }, + { + "epoch": 7.83, + "learning_rate": 2.4196177786426357e-08, + "loss": 0.0007, + "step": 59760 + }, + { + "epoch": 7.83, + "learning_rate": 2.4159322278862753e-08, + "loss": 0.0002, + "step": 59761 + }, + { + "epoch": 7.83, + "learning_rate": 2.4122494827820696e-08, + "loss": 0.0006, + "step": 59762 + }, + { + "epoch": 7.83, + "learning_rate": 2.4085695433402334e-08, + "loss": 0.0003, + "step": 59763 + }, + { + "epoch": 7.83, + "learning_rate": 2.4048924095710913e-08, + "loss": 0.0002, + "step": 59764 + }, + { + "epoch": 7.83, + "learning_rate": 2.4012180814851904e-08, + "loss": 0.0003, + "step": 59765 + }, + { + "epoch": 7.83, + "learning_rate": 2.397546559092745e-08, + "loss": 0.0002, + "step": 59766 + }, + { + "epoch": 7.83, + "learning_rate": 2.3938778424039687e-08, + "loss": 0.0008, + "step": 59767 + }, + { + "epoch": 7.83, + "learning_rate": 2.390211931429298e-08, + "loss": 0.0004, + "step": 59768 + }, + { + "epoch": 7.83, + "learning_rate": 2.386548826179169e-08, + "loss": 0.0005, + "step": 59769 + }, + { + "epoch": 7.83, + "learning_rate": 2.382888526663574e-08, + "loss": 0.0001, + "step": 59770 + }, + { + "epoch": 7.83, + "learning_rate": 2.37923103289317e-08, + "loss": 0.0002, + "step": 59771 + }, + { + "epoch": 7.83, + "learning_rate": 2.375576344877839e-08, + "loss": 0.0006, + "step": 59772 + }, + { + "epoch": 7.83, + "learning_rate": 2.371924462628239e-08, + "loss": 0.0002, + "step": 59773 + }, + { + "epoch": 7.83, + "learning_rate": 2.3682753861543616e-08, + "loss": 0.0006, + "step": 59774 + }, + { + "epoch": 7.83, + "learning_rate": 2.3646291154666435e-08, + "loss": 0.0004, + "step": 59775 + }, + { + "epoch": 7.83, + "learning_rate": 2.360985650575298e-08, + "loss": 0.0002, + "step": 59776 + }, + { + "epoch": 7.83, + "learning_rate": 2.35734499149054e-08, + "loss": 0.0002, + "step": 59777 + }, + { + "epoch": 7.83, + "learning_rate": 2.3537071382225828e-08, + "loss": 0.0002, + "step": 59778 + }, + { + "epoch": 7.83, + "learning_rate": 2.350072090781641e-08, + "loss": 0.0003, + "step": 59779 + }, + { + "epoch": 7.83, + "learning_rate": 2.3464398491780393e-08, + "loss": 0.0004, + "step": 59780 + }, + { + "epoch": 7.83, + "learning_rate": 2.342810413421992e-08, + "loss": 0.0007, + "step": 59781 + }, + { + "epoch": 7.83, + "learning_rate": 2.339183783523713e-08, + "loss": 0.0002, + "step": 59782 + }, + { + "epoch": 7.83, + "learning_rate": 2.3355599594933054e-08, + "loss": 0.0006, + "step": 59783 + }, + { + "epoch": 7.83, + "learning_rate": 2.3319389413409832e-08, + "loss": 0.0008, + "step": 59784 + }, + { + "epoch": 7.83, + "learning_rate": 2.3283207290769606e-08, + "loss": 0.0004, + "step": 59785 + }, + { + "epoch": 7.83, + "learning_rate": 2.3247053227113402e-08, + "loss": 0.0007, + "step": 59786 + }, + { + "epoch": 7.83, + "learning_rate": 2.321092722254448e-08, + "loss": 0.0004, + "step": 59787 + }, + { + "epoch": 7.83, + "learning_rate": 2.317482927716275e-08, + "loss": 0.0006, + "step": 59788 + }, + { + "epoch": 7.83, + "learning_rate": 2.3138759391071464e-08, + "loss": 0.0005, + "step": 59789 + }, + { + "epoch": 7.83, + "learning_rate": 2.3102717564370546e-08, + "loss": 0.0003, + "step": 59790 + }, + { + "epoch": 7.83, + "learning_rate": 2.3066703797162137e-08, + "loss": 0.0005, + "step": 59791 + }, + { + "epoch": 7.83, + "learning_rate": 2.3030718089547267e-08, + "loss": 0.0003, + "step": 59792 + }, + { + "epoch": 7.83, + "learning_rate": 2.2994760441626963e-08, + "loss": 0.0004, + "step": 59793 + }, + { + "epoch": 7.83, + "learning_rate": 2.2958830853503365e-08, + "loss": 0.0006, + "step": 59794 + }, + { + "epoch": 7.83, + "learning_rate": 2.292292932527529e-08, + "loss": 0.0002, + "step": 59795 + }, + { + "epoch": 7.83, + "learning_rate": 2.288705585704598e-08, + "loss": 0.0003, + "step": 59796 + }, + { + "epoch": 7.83, + "learning_rate": 2.2851210448916472e-08, + "loss": 0.0002, + "step": 59797 + }, + { + "epoch": 7.83, + "learning_rate": 2.2815393100985573e-08, + "loss": 0.0006, + "step": 59798 + }, + { + "epoch": 7.83, + "learning_rate": 2.2779603813355422e-08, + "loss": 0.0004, + "step": 59799 + }, + { + "epoch": 7.83, + "learning_rate": 2.2743842586125942e-08, + "loss": 0.0006, + "step": 59800 + }, + { + "epoch": 7.83, + "learning_rate": 2.2708109419398162e-08, + "loss": 0.0003, + "step": 59801 + }, + { + "epoch": 7.83, + "learning_rate": 2.2672404313272e-08, + "loss": 0.0003, + "step": 59802 + }, + { + "epoch": 7.83, + "learning_rate": 2.26367272678496e-08, + "loss": 0.0006, + "step": 59803 + }, + { + "epoch": 7.83, + "learning_rate": 2.260107828322866e-08, + "loss": 0.0007, + "step": 59804 + }, + { + "epoch": 7.83, + "learning_rate": 2.256545735951243e-08, + "loss": 0.0002, + "step": 59805 + }, + { + "epoch": 7.83, + "learning_rate": 2.2529864496798615e-08, + "loss": 0.001, + "step": 59806 + }, + { + "epoch": 7.83, + "learning_rate": 2.2494299695188238e-08, + "loss": 0.0008, + "step": 59807 + }, + { + "epoch": 7.83, + "learning_rate": 2.2458762954781222e-08, + "loss": 0.0006, + "step": 59808 + }, + { + "epoch": 7.83, + "learning_rate": 2.2423254275678595e-08, + "loss": 0.0002, + "step": 59809 + }, + { + "epoch": 7.83, + "learning_rate": 2.238777365797806e-08, + "loss": 0.0003, + "step": 59810 + }, + { + "epoch": 7.83, + "learning_rate": 2.235232110178065e-08, + "loss": 0.0003, + "step": 59811 + }, + { + "epoch": 7.83, + "learning_rate": 2.2316896607186278e-08, + "loss": 0.0005, + "step": 59812 + }, + { + "epoch": 7.84, + "learning_rate": 2.2281500174294866e-08, + "loss": 0.0006, + "step": 59813 + }, + { + "epoch": 7.84, + "learning_rate": 2.224613180320412e-08, + "loss": 0.0008, + "step": 59814 + }, + { + "epoch": 7.84, + "learning_rate": 2.221079149401506e-08, + "loss": 0.0003, + "step": 59815 + }, + { + "epoch": 7.84, + "learning_rate": 2.2175479246827613e-08, + "loss": 0.0003, + "step": 59816 + }, + { + "epoch": 7.84, + "learning_rate": 2.2140195061740587e-08, + "loss": 0.0002, + "step": 59817 + }, + { + "epoch": 7.84, + "learning_rate": 2.2104938938851685e-08, + "loss": 0.0002, + "step": 59818 + }, + { + "epoch": 7.84, + "learning_rate": 2.206971087826193e-08, + "loss": 0.0013, + "step": 59819 + }, + { + "epoch": 7.84, + "learning_rate": 2.2034510880070138e-08, + "loss": 0.0003, + "step": 59820 + }, + { + "epoch": 7.84, + "learning_rate": 2.1999338944374006e-08, + "loss": 0.0002, + "step": 59821 + }, + { + "epoch": 7.84, + "learning_rate": 2.196419507127345e-08, + "loss": 0.0004, + "step": 59822 + }, + { + "epoch": 7.84, + "learning_rate": 2.192907926086729e-08, + "loss": 0.0011, + "step": 59823 + }, + { + "epoch": 7.84, + "learning_rate": 2.189399151325544e-08, + "loss": 0.0002, + "step": 59824 + }, + { + "epoch": 7.84, + "learning_rate": 2.185893182853449e-08, + "loss": 0.0005, + "step": 59825 + }, + { + "epoch": 7.84, + "learning_rate": 2.182390020680325e-08, + "loss": 0.0006, + "step": 59826 + }, + { + "epoch": 7.84, + "learning_rate": 2.178889664816164e-08, + "loss": 0.0005, + "step": 59827 + }, + { + "epoch": 7.84, + "learning_rate": 2.1753921152707357e-08, + "loss": 0.0005, + "step": 59828 + }, + { + "epoch": 7.84, + "learning_rate": 2.1718973720539216e-08, + "loss": 0.0004, + "step": 59829 + }, + { + "epoch": 7.84, + "learning_rate": 2.1684054351753802e-08, + "loss": 0.0004, + "step": 59830 + }, + { + "epoch": 7.84, + "learning_rate": 2.1649163046452147e-08, + "loss": 0.0002, + "step": 59831 + }, + { + "epoch": 7.84, + "learning_rate": 2.1614299804729733e-08, + "loss": 0.0005, + "step": 59832 + }, + { + "epoch": 7.84, + "learning_rate": 2.1579464626686475e-08, + "loss": 0.0002, + "step": 59833 + }, + { + "epoch": 7.84, + "learning_rate": 2.1544657512418966e-08, + "loss": 0.0005, + "step": 59834 + }, + { + "epoch": 7.84, + "learning_rate": 2.1509878462026012e-08, + "loss": 0.0003, + "step": 59835 + }, + { + "epoch": 7.84, + "learning_rate": 2.1475127475605317e-08, + "loss": 0.0012, + "step": 59836 + }, + { + "epoch": 7.84, + "learning_rate": 2.1440404553253468e-08, + "loss": 0.0004, + "step": 59837 + }, + { + "epoch": 7.84, + "learning_rate": 2.1405709695070386e-08, + "loss": 0.0001, + "step": 59838 + }, + { + "epoch": 7.84, + "learning_rate": 2.137104290115155e-08, + "loss": 0.0004, + "step": 59839 + }, + { + "epoch": 7.84, + "learning_rate": 2.133640417159577e-08, + "loss": 0.0003, + "step": 59840 + }, + { + "epoch": 7.84, + "learning_rate": 2.1301793506499634e-08, + "loss": 0.0009, + "step": 59841 + }, + { + "epoch": 7.84, + "learning_rate": 2.1267210905960846e-08, + "loss": 0.0003, + "step": 59842 + }, + { + "epoch": 7.84, + "learning_rate": 2.1232656370077097e-08, + "loss": 0.0007, + "step": 59843 + }, + { + "epoch": 7.84, + "learning_rate": 2.1198129898944986e-08, + "loss": 0.0006, + "step": 59844 + }, + { + "epoch": 7.84, + "learning_rate": 2.1163631492661095e-08, + "loss": 0.0006, + "step": 59845 + }, + { + "epoch": 7.84, + "learning_rate": 2.112916115132313e-08, + "loss": 0.0004, + "step": 59846 + }, + { + "epoch": 7.84, + "learning_rate": 2.1094718875028786e-08, + "loss": 0.0006, + "step": 59847 + }, + { + "epoch": 7.84, + "learning_rate": 2.1060304663873543e-08, + "loss": 0.0004, + "step": 59848 + }, + { + "epoch": 7.84, + "learning_rate": 2.10259185179551e-08, + "loss": 0.0003, + "step": 59849 + }, + { + "epoch": 7.84, + "learning_rate": 2.0991560437368942e-08, + "loss": 0.0005, + "step": 59850 + }, + { + "epoch": 7.84, + "learning_rate": 2.095723042221387e-08, + "loss": 0.0003, + "step": 59851 + }, + { + "epoch": 7.84, + "learning_rate": 2.092292847258426e-08, + "loss": 0.0002, + "step": 59852 + }, + { + "epoch": 7.84, + "learning_rate": 2.0888654588577805e-08, + "loss": 0.0002, + "step": 59853 + }, + { + "epoch": 7.84, + "learning_rate": 2.08544087702911e-08, + "loss": 0.0007, + "step": 59854 + }, + { + "epoch": 7.84, + "learning_rate": 2.0820191017819623e-08, + "loss": 0.0003, + "step": 59855 + }, + { + "epoch": 7.84, + "learning_rate": 2.0786001331258855e-08, + "loss": 0.0005, + "step": 59856 + }, + { + "epoch": 7.84, + "learning_rate": 2.07518397107076e-08, + "loss": 0.0002, + "step": 59857 + }, + { + "epoch": 7.84, + "learning_rate": 2.0717706156259122e-08, + "loss": 0.0006, + "step": 59858 + }, + { + "epoch": 7.84, + "learning_rate": 2.0683600668012226e-08, + "loss": 0.0003, + "step": 59859 + }, + { + "epoch": 7.84, + "learning_rate": 2.0649523246060176e-08, + "loss": 0.0001, + "step": 59860 + }, + { + "epoch": 7.84, + "learning_rate": 2.061547389049956e-08, + "loss": 0.0015, + "step": 59861 + }, + { + "epoch": 7.84, + "learning_rate": 2.0581452601426965e-08, + "loss": 0.0008, + "step": 59862 + }, + { + "epoch": 7.84, + "learning_rate": 2.0547459378937873e-08, + "loss": 0.0005, + "step": 59863 + }, + { + "epoch": 7.84, + "learning_rate": 2.051349422312776e-08, + "loss": 0.0007, + "step": 59864 + }, + { + "epoch": 7.84, + "learning_rate": 2.0479557134091e-08, + "loss": 0.0005, + "step": 59865 + }, + { + "epoch": 7.84, + "learning_rate": 2.0445648111924178e-08, + "loss": 0.0003, + "step": 59866 + }, + { + "epoch": 7.84, + "learning_rate": 2.0411767156723882e-08, + "loss": 0.0004, + "step": 59867 + }, + { + "epoch": 7.84, + "learning_rate": 2.0377914268582267e-08, + "loss": 0.0005, + "step": 59868 + }, + { + "epoch": 7.84, + "learning_rate": 2.0344089447598135e-08, + "loss": 0.0004, + "step": 59869 + }, + { + "epoch": 7.84, + "learning_rate": 2.0310292693863644e-08, + "loss": 0.0006, + "step": 59870 + }, + { + "epoch": 7.84, + "learning_rate": 2.0276524007475374e-08, + "loss": 0.0004, + "step": 59871 + }, + { + "epoch": 7.84, + "learning_rate": 2.0242783388527698e-08, + "loss": 0.0005, + "step": 59872 + }, + { + "epoch": 7.84, + "learning_rate": 2.0209070837116096e-08, + "loss": 0.0003, + "step": 59873 + }, + { + "epoch": 7.84, + "learning_rate": 2.0175386353334934e-08, + "loss": 0.0003, + "step": 59874 + }, + { + "epoch": 7.84, + "learning_rate": 2.0141729937278586e-08, + "loss": 0.0004, + "step": 59875 + }, + { + "epoch": 7.84, + "learning_rate": 2.0108101589042527e-08, + "loss": 0.0003, + "step": 59876 + }, + { + "epoch": 7.84, + "learning_rate": 2.0074501308721128e-08, + "loss": 0.0005, + "step": 59877 + }, + { + "epoch": 7.84, + "learning_rate": 2.0040929096407647e-08, + "loss": 0.0001, + "step": 59878 + }, + { + "epoch": 7.84, + "learning_rate": 2.000738495219867e-08, + "loss": 0.0008, + "step": 59879 + }, + { + "epoch": 7.84, + "learning_rate": 1.9973868876187462e-08, + "loss": 0.0008, + "step": 59880 + }, + { + "epoch": 7.84, + "learning_rate": 1.994038086846839e-08, + "loss": 0.0003, + "step": 59881 + }, + { + "epoch": 7.84, + "learning_rate": 1.990692092913471e-08, + "loss": 0.0004, + "step": 59882 + }, + { + "epoch": 7.84, + "learning_rate": 1.9873489058281904e-08, + "loss": 0.0004, + "step": 59883 + }, + { + "epoch": 7.84, + "learning_rate": 1.9840085256003226e-08, + "loss": 0.0008, + "step": 59884 + }, + { + "epoch": 7.84, + "learning_rate": 1.980670952239305e-08, + "loss": 0.0005, + "step": 59885 + }, + { + "epoch": 7.84, + "learning_rate": 1.9773361857544636e-08, + "loss": 0.0006, + "step": 59886 + }, + { + "epoch": 7.84, + "learning_rate": 1.974004226155346e-08, + "loss": 0.0006, + "step": 59887 + }, + { + "epoch": 7.84, + "learning_rate": 1.970675073451056e-08, + "loss": 0.0003, + "step": 59888 + }, + { + "epoch": 7.85, + "learning_rate": 1.9673487276511417e-08, + "loss": 0.0003, + "step": 59889 + }, + { + "epoch": 7.85, + "learning_rate": 1.9640251887649285e-08, + "loss": 0.0005, + "step": 59890 + }, + { + "epoch": 7.85, + "learning_rate": 1.960704456801632e-08, + "loss": 0.0003, + "step": 59891 + }, + { + "epoch": 7.85, + "learning_rate": 1.9573865317707997e-08, + "loss": 0.0005, + "step": 59892 + }, + { + "epoch": 7.85, + "learning_rate": 1.9540714136817573e-08, + "loss": 0.0007, + "step": 59893 + }, + { + "epoch": 7.85, + "learning_rate": 1.950759102543609e-08, + "loss": 0.0003, + "step": 59894 + }, + { + "epoch": 7.85, + "learning_rate": 1.9474495983657914e-08, + "loss": 0.0002, + "step": 59895 + }, + { + "epoch": 7.85, + "learning_rate": 1.9441429011577418e-08, + "loss": 0.0002, + "step": 59896 + }, + { + "epoch": 7.85, + "learning_rate": 1.9408390109285636e-08, + "loss": 0.0005, + "step": 59897 + }, + { + "epoch": 7.85, + "learning_rate": 1.937537927687583e-08, + "loss": 0.0004, + "step": 59898 + }, + { + "epoch": 7.85, + "learning_rate": 1.9342396514441254e-08, + "loss": 0.0007, + "step": 59899 + }, + { + "epoch": 7.85, + "learning_rate": 1.930944182207517e-08, + "loss": 0.0003, + "step": 59900 + }, + { + "epoch": 7.85, + "learning_rate": 1.9276515199870836e-08, + "loss": 0.0004, + "step": 59901 + }, + { + "epoch": 7.85, + "learning_rate": 1.9243616647918183e-08, + "loss": 0.0002, + "step": 59902 + }, + { + "epoch": 7.85, + "learning_rate": 1.9210746166311578e-08, + "loss": 0.0005, + "step": 59903 + }, + { + "epoch": 7.85, + "learning_rate": 1.917790375514428e-08, + "loss": 0.0005, + "step": 59904 + }, + { + "epoch": 7.85, + "learning_rate": 1.9145089414506213e-08, + "loss": 0.0004, + "step": 59905 + }, + { + "epoch": 7.85, + "learning_rate": 1.911230314449175e-08, + "loss": 0.0004, + "step": 59906 + }, + { + "epoch": 7.85, + "learning_rate": 1.907954494519304e-08, + "loss": 0.0005, + "step": 59907 + }, + { + "epoch": 7.85, + "learning_rate": 1.904681481670112e-08, + "loss": 0.0007, + "step": 59908 + }, + { + "epoch": 7.85, + "learning_rate": 1.9014112759108138e-08, + "loss": 0.0006, + "step": 59909 + }, + { + "epoch": 7.85, + "learning_rate": 1.8981438772506244e-08, + "loss": 0.0004, + "step": 59910 + }, + { + "epoch": 7.85, + "learning_rate": 1.8948792856988694e-08, + "loss": 0.0002, + "step": 59911 + }, + { + "epoch": 7.85, + "learning_rate": 1.8916175012645423e-08, + "loss": 0.0002, + "step": 59912 + }, + { + "epoch": 7.85, + "learning_rate": 1.888358523956857e-08, + "loss": 0.0004, + "step": 59913 + }, + { + "epoch": 7.85, + "learning_rate": 1.885102353785029e-08, + "loss": 0.0002, + "step": 59914 + }, + { + "epoch": 7.85, + "learning_rate": 1.8818489907582727e-08, + "loss": 0.0002, + "step": 59915 + }, + { + "epoch": 7.85, + "learning_rate": 1.8785984348855812e-08, + "loss": 0.0003, + "step": 59916 + }, + { + "epoch": 7.85, + "learning_rate": 1.8753506861761694e-08, + "loss": 0.0004, + "step": 59917 + }, + { + "epoch": 7.85, + "learning_rate": 1.8721057446392522e-08, + "loss": 0.0012, + "step": 59918 + }, + { + "epoch": 7.85, + "learning_rate": 1.868863610283822e-08, + "loss": 0.0006, + "step": 59919 + }, + { + "epoch": 7.85, + "learning_rate": 1.865624283119094e-08, + "loss": 0.0003, + "step": 59920 + }, + { + "epoch": 7.85, + "learning_rate": 1.862387763154061e-08, + "loss": 0.0002, + "step": 59921 + }, + { + "epoch": 7.85, + "learning_rate": 1.859154050398049e-08, + "loss": 0.0005, + "step": 59922 + }, + { + "epoch": 7.85, + "learning_rate": 1.855923144859939e-08, + "loss": 0.0005, + "step": 59923 + }, + { + "epoch": 7.85, + "learning_rate": 1.8526950465488357e-08, + "loss": 0.0014, + "step": 59924 + }, + { + "epoch": 7.85, + "learning_rate": 1.8494697554739536e-08, + "loss": 0.0005, + "step": 59925 + }, + { + "epoch": 7.85, + "learning_rate": 1.846247271644286e-08, + "loss": 0.0008, + "step": 59926 + }, + { + "epoch": 7.85, + "learning_rate": 1.843027595068825e-08, + "loss": 0.0005, + "step": 59927 + }, + { + "epoch": 7.85, + "learning_rate": 1.8398107257567855e-08, + "loss": 0.0003, + "step": 59928 + }, + { + "epoch": 7.85, + "learning_rate": 1.8365966637170495e-08, + "loss": 0.0003, + "step": 59929 + }, + { + "epoch": 7.85, + "learning_rate": 1.833385408958721e-08, + "loss": 0.0004, + "step": 59930 + }, + { + "epoch": 7.85, + "learning_rate": 1.8301769614909037e-08, + "loss": 0.0004, + "step": 59931 + }, + { + "epoch": 7.85, + "learning_rate": 1.8269713213225902e-08, + "loss": 0.0004, + "step": 59932 + }, + { + "epoch": 7.85, + "learning_rate": 1.8237684884626628e-08, + "loss": 0.0004, + "step": 59933 + }, + { + "epoch": 7.85, + "learning_rate": 1.8205684629202248e-08, + "loss": 0.0005, + "step": 59934 + }, + { + "epoch": 7.85, + "learning_rate": 1.8173712447043802e-08, + "loss": 0.0002, + "step": 59935 + }, + { + "epoch": 7.85, + "learning_rate": 1.8141768338238997e-08, + "loss": 0.0003, + "step": 59936 + }, + { + "epoch": 7.85, + "learning_rate": 1.8109852302879984e-08, + "loss": 0.0003, + "step": 59937 + }, + { + "epoch": 7.85, + "learning_rate": 1.8077964341055576e-08, + "loss": 0.0004, + "step": 59938 + }, + { + "epoch": 7.85, + "learning_rate": 1.8046104452854596e-08, + "loss": 0.0007, + "step": 59939 + }, + { + "epoch": 7.85, + "learning_rate": 1.801427263836697e-08, + "loss": 0.0002, + "step": 59940 + }, + { + "epoch": 7.85, + "learning_rate": 1.7982468897683736e-08, + "loss": 0.0003, + "step": 59941 + }, + { + "epoch": 7.85, + "learning_rate": 1.7950693230892603e-08, + "loss": 0.0002, + "step": 59942 + }, + { + "epoch": 7.85, + "learning_rate": 1.7918945638083497e-08, + "loss": 0.0004, + "step": 59943 + }, + { + "epoch": 7.85, + "learning_rate": 1.7887226119346346e-08, + "loss": 0.0003, + "step": 59944 + }, + { + "epoch": 7.85, + "learning_rate": 1.785553467476886e-08, + "loss": 0.0003, + "step": 59945 + }, + { + "epoch": 7.85, + "learning_rate": 1.7823871304442074e-08, + "loss": 0.0003, + "step": 59946 + }, + { + "epoch": 7.85, + "learning_rate": 1.77922360084537e-08, + "loss": 0.0006, + "step": 59947 + }, + { + "epoch": 7.85, + "learning_rate": 1.7760628786892552e-08, + "loss": 0.0004, + "step": 59948 + }, + { + "epoch": 7.85, + "learning_rate": 1.7729049639849673e-08, + "loss": 0.0005, + "step": 59949 + }, + { + "epoch": 7.85, + "learning_rate": 1.7697498567410543e-08, + "loss": 0.0007, + "step": 59950 + }, + { + "epoch": 7.85, + "learning_rate": 1.7665975569666205e-08, + "loss": 0.0004, + "step": 59951 + }, + { + "epoch": 7.85, + "learning_rate": 1.7634480646704366e-08, + "loss": 0.0006, + "step": 59952 + }, + { + "epoch": 7.85, + "learning_rate": 1.7603013798614956e-08, + "loss": 0.001, + "step": 59953 + }, + { + "epoch": 7.85, + "learning_rate": 1.7571575025484565e-08, + "loss": 0.0005, + "step": 59954 + }, + { + "epoch": 7.85, + "learning_rate": 1.7540164327403132e-08, + "loss": 0.0002, + "step": 59955 + }, + { + "epoch": 7.85, + "learning_rate": 1.7508781704457246e-08, + "loss": 0.0004, + "step": 59956 + }, + { + "epoch": 7.85, + "learning_rate": 1.7477427156737947e-08, + "loss": 0.0011, + "step": 59957 + }, + { + "epoch": 7.85, + "learning_rate": 1.7446100684330724e-08, + "loss": 0.0006, + "step": 59958 + }, + { + "epoch": 7.85, + "learning_rate": 1.7414802287324396e-08, + "loss": 0.0004, + "step": 59959 + }, + { + "epoch": 7.85, + "learning_rate": 1.7383531965808885e-08, + "loss": 0.0002, + "step": 59960 + }, + { + "epoch": 7.85, + "learning_rate": 1.7352289719869686e-08, + "loss": 0.0007, + "step": 59961 + }, + { + "epoch": 7.85, + "learning_rate": 1.732107554959561e-08, + "loss": 0.0003, + "step": 59962 + }, + { + "epoch": 7.85, + "learning_rate": 1.728988945507326e-08, + "loss": 0.0007, + "step": 59963 + }, + { + "epoch": 7.85, + "learning_rate": 1.7258731436392562e-08, + "loss": 0.0005, + "step": 59964 + }, + { + "epoch": 7.85, + "learning_rate": 1.722760149364011e-08, + "loss": 0.0003, + "step": 59965 + }, + { + "epoch": 7.86, + "learning_rate": 1.7196499626903617e-08, + "loss": 0.0002, + "step": 59966 + }, + { + "epoch": 7.86, + "learning_rate": 1.7165425836269678e-08, + "loss": 0.0007, + "step": 59967 + }, + { + "epoch": 7.86, + "learning_rate": 1.713438012182711e-08, + "loss": 0.0004, + "step": 59968 + }, + { + "epoch": 7.86, + "learning_rate": 1.7103362483661402e-08, + "loss": 0.0002, + "step": 59969 + }, + { + "epoch": 7.86, + "learning_rate": 1.707237292186248e-08, + "loss": 0.0006, + "step": 59970 + }, + { + "epoch": 7.86, + "learning_rate": 1.7041411436514722e-08, + "loss": 0.0003, + "step": 59971 + }, + { + "epoch": 7.86, + "learning_rate": 1.7010478027706946e-08, + "loss": 0.0002, + "step": 59972 + }, + { + "epoch": 7.86, + "learning_rate": 1.6979572695524637e-08, + "loss": 0.0005, + "step": 59973 + }, + { + "epoch": 7.86, + "learning_rate": 1.694869544005662e-08, + "loss": 0.0006, + "step": 59974 + }, + { + "epoch": 7.86, + "learning_rate": 1.6917846261388372e-08, + "loss": 0.0001, + "step": 59975 + }, + { + "epoch": 7.86, + "learning_rate": 1.6887025159606497e-08, + "loss": 0.0003, + "step": 59976 + }, + { + "epoch": 7.86, + "learning_rate": 1.68562321347987e-08, + "loss": 0.0002, + "step": 59977 + }, + { + "epoch": 7.86, + "learning_rate": 1.682546718705158e-08, + "loss": 0.0005, + "step": 59978 + }, + { + "epoch": 7.86, + "learning_rate": 1.6794730316450625e-08, + "loss": 0.0002, + "step": 59979 + }, + { + "epoch": 7.86, + "learning_rate": 1.676402152308354e-08, + "loss": 0.0005, + "step": 59980 + }, + { + "epoch": 7.86, + "learning_rate": 1.6733340807035814e-08, + "loss": 0.0003, + "step": 59981 + }, + { + "epoch": 7.86, + "learning_rate": 1.6702688168392933e-08, + "loss": 0.0003, + "step": 59982 + }, + { + "epoch": 7.86, + "learning_rate": 1.6672063607243716e-08, + "loss": 0.0002, + "step": 59983 + }, + { + "epoch": 7.86, + "learning_rate": 1.664146712367143e-08, + "loss": 0.0016, + "step": 59984 + }, + { + "epoch": 7.86, + "learning_rate": 1.6610898717763778e-08, + "loss": 0.0002, + "step": 59985 + }, + { + "epoch": 7.86, + "learning_rate": 1.6580358389606256e-08, + "loss": 0.0004, + "step": 59986 + }, + { + "epoch": 7.86, + "learning_rate": 1.654984613928434e-08, + "loss": 0.0004, + "step": 59987 + }, + { + "epoch": 7.86, + "learning_rate": 1.651936196688575e-08, + "loss": 0.0004, + "step": 59988 + }, + { + "epoch": 7.86, + "learning_rate": 1.6488905872493744e-08, + "loss": 0.0004, + "step": 59989 + }, + { + "epoch": 7.86, + "learning_rate": 1.645847785619492e-08, + "loss": 0.0004, + "step": 59990 + }, + { + "epoch": 7.86, + "learning_rate": 1.642807791807477e-08, + "loss": 0.0005, + "step": 59991 + }, + { + "epoch": 7.86, + "learning_rate": 1.6397706058219888e-08, + "loss": 0.0007, + "step": 59992 + }, + { + "epoch": 7.86, + "learning_rate": 1.636736227671465e-08, + "loss": 0.0003, + "step": 59993 + }, + { + "epoch": 7.86, + "learning_rate": 1.6337046573644544e-08, + "loss": 0.0009, + "step": 59994 + }, + { + "epoch": 7.86, + "learning_rate": 1.630675894909395e-08, + "loss": 0.0007, + "step": 59995 + }, + { + "epoch": 7.86, + "learning_rate": 1.627649940314946e-08, + "loss": 0.0004, + "step": 59996 + }, + { + "epoch": 7.86, + "learning_rate": 1.6246267935895456e-08, + "loss": 0.001, + "step": 59997 + }, + { + "epoch": 7.86, + "learning_rate": 1.621606454741742e-08, + "loss": 0.0002, + "step": 59998 + }, + { + "epoch": 7.86, + "learning_rate": 1.6185889237800845e-08, + "loss": 0.0007, + "step": 59999 + }, + { + "epoch": 7.86, + "learning_rate": 1.6155742007128993e-08, + "loss": 0.0004, + "step": 60000 + }, + { + "epoch": 7.86, + "learning_rate": 1.6125622855487355e-08, + "loss": 0.0006, + "step": 60001 + }, + { + "epoch": 7.86, + "learning_rate": 1.6095531782960305e-08, + "loss": 0.0002, + "step": 60002 + }, + { + "epoch": 7.86, + "learning_rate": 1.606546878963333e-08, + "loss": 0.0002, + "step": 60003 + }, + { + "epoch": 7.86, + "learning_rate": 1.6035433875590812e-08, + "loss": 0.0004, + "step": 60004 + }, + { + "epoch": 7.86, + "learning_rate": 1.600542704091712e-08, + "loss": 0.0002, + "step": 60005 + }, + { + "epoch": 7.86, + "learning_rate": 1.597544828569664e-08, + "loss": 0.0003, + "step": 60006 + }, + { + "epoch": 7.86, + "learning_rate": 1.594549761001263e-08, + "loss": 0.0003, + "step": 60007 + }, + { + "epoch": 7.86, + "learning_rate": 1.5915575013950578e-08, + "loss": 0.0003, + "step": 60008 + }, + { + "epoch": 7.86, + "learning_rate": 1.5885680497593758e-08, + "loss": 0.0009, + "step": 60009 + }, + { + "epoch": 7.86, + "learning_rate": 1.5855814061027652e-08, + "loss": 0.001, + "step": 60010 + }, + { + "epoch": 7.86, + "learning_rate": 1.5825975704335528e-08, + "loss": 0.0002, + "step": 60011 + }, + { + "epoch": 7.86, + "learning_rate": 1.5796165427600652e-08, + "loss": 0.0004, + "step": 60012 + }, + { + "epoch": 7.86, + "learning_rate": 1.5766383230907402e-08, + "loss": 0.0003, + "step": 60013 + }, + { + "epoch": 7.86, + "learning_rate": 1.573662911433904e-08, + "loss": 0.0003, + "step": 60014 + }, + { + "epoch": 7.86, + "learning_rate": 1.570690307798106e-08, + "loss": 0.0011, + "step": 60015 + }, + { + "epoch": 7.86, + "learning_rate": 1.567720512191451e-08, + "loss": 0.0003, + "step": 60016 + }, + { + "epoch": 7.86, + "learning_rate": 1.5647535246224864e-08, + "loss": 0.0009, + "step": 60017 + }, + { + "epoch": 7.86, + "learning_rate": 1.5617893450994292e-08, + "loss": 0.0005, + "step": 60018 + }, + { + "epoch": 7.86, + "learning_rate": 1.5588279736307166e-08, + "loss": 0.0004, + "step": 60019 + }, + { + "epoch": 7.86, + "learning_rate": 1.555869410224564e-08, + "loss": 0.0006, + "step": 60020 + }, + { + "epoch": 7.86, + "learning_rate": 1.55291365488941e-08, + "loss": 0.0002, + "step": 60021 + }, + { + "epoch": 7.86, + "learning_rate": 1.549960707633469e-08, + "loss": 0.0006, + "step": 60022 + }, + { + "epoch": 7.86, + "learning_rate": 1.5470105684651794e-08, + "loss": 0.0007, + "step": 60023 + }, + { + "epoch": 7.86, + "learning_rate": 1.5440632373926457e-08, + "loss": 0.0003, + "step": 60024 + }, + { + "epoch": 7.86, + "learning_rate": 1.5411187144243055e-08, + "loss": 0.0004, + "step": 60025 + }, + { + "epoch": 7.86, + "learning_rate": 1.538176999568375e-08, + "loss": 0.0005, + "step": 60026 + }, + { + "epoch": 7.86, + "learning_rate": 1.535238092833291e-08, + "loss": 0.0003, + "step": 60027 + }, + { + "epoch": 7.86, + "learning_rate": 1.5323019942270477e-08, + "loss": 0.0003, + "step": 60028 + }, + { + "epoch": 7.86, + "learning_rate": 1.529368703757972e-08, + "loss": 0.0002, + "step": 60029 + }, + { + "epoch": 7.86, + "learning_rate": 1.526438221434501e-08, + "loss": 0.0006, + "step": 60030 + }, + { + "epoch": 7.86, + "learning_rate": 1.5235105472647393e-08, + "loss": 0.0002, + "step": 60031 + }, + { + "epoch": 7.86, + "learning_rate": 1.520585681256903e-08, + "loss": 0.0001, + "step": 60032 + }, + { + "epoch": 7.86, + "learning_rate": 1.5176636234192077e-08, + "loss": 0.0004, + "step": 60033 + }, + { + "epoch": 7.86, + "learning_rate": 1.51474437375998e-08, + "loss": 0.0001, + "step": 60034 + }, + { + "epoch": 7.86, + "learning_rate": 1.5118279322873243e-08, + "loss": 0.0005, + "step": 60035 + }, + { + "epoch": 7.86, + "learning_rate": 1.508914299009456e-08, + "loss": 0.0009, + "step": 60036 + }, + { + "epoch": 7.86, + "learning_rate": 1.5060034739345918e-08, + "loss": 0.0004, + "step": 60037 + }, + { + "epoch": 7.86, + "learning_rate": 1.5030954570710576e-08, + "loss": 0.0003, + "step": 60038 + }, + { + "epoch": 7.86, + "learning_rate": 1.500190248426736e-08, + "loss": 0.0004, + "step": 60039 + }, + { + "epoch": 7.86, + "learning_rate": 1.497287848009954e-08, + "loss": 0.0007, + "step": 60040 + }, + { + "epoch": 7.86, + "learning_rate": 1.494388255828927e-08, + "loss": 0.0004, + "step": 60041 + }, + { + "epoch": 7.87, + "learning_rate": 1.4914914718917593e-08, + "loss": 0.0005, + "step": 60042 + }, + { + "epoch": 7.87, + "learning_rate": 1.4885974962066674e-08, + "loss": 0.0006, + "step": 60043 + }, + { + "epoch": 7.87, + "learning_rate": 1.4857063287816442e-08, + "loss": 0.0002, + "step": 60044 + }, + { + "epoch": 7.87, + "learning_rate": 1.4828179696249057e-08, + "loss": 0.0004, + "step": 60045 + }, + { + "epoch": 7.87, + "learning_rate": 1.4799324187445564e-08, + "loss": 0.0006, + "step": 60046 + }, + { + "epoch": 7.87, + "learning_rate": 1.4770496761487007e-08, + "loss": 0.0002, + "step": 60047 + }, + { + "epoch": 7.87, + "learning_rate": 1.4741697418455548e-08, + "loss": 0.0007, + "step": 60048 + }, + { + "epoch": 7.87, + "learning_rate": 1.4712926158431117e-08, + "loss": 0.0006, + "step": 60049 + }, + { + "epoch": 7.87, + "learning_rate": 1.4684182981494765e-08, + "loss": 0.0004, + "step": 60050 + }, + { + "epoch": 7.87, + "learning_rate": 1.4655467887726426e-08, + "loss": 0.0003, + "step": 60051 + }, + { + "epoch": 7.87, + "learning_rate": 1.4626780877208258e-08, + "loss": 0.0006, + "step": 60052 + }, + { + "epoch": 7.87, + "learning_rate": 1.4598121950021304e-08, + "loss": 0.0005, + "step": 60053 + }, + { + "epoch": 7.87, + "learning_rate": 1.4569491106245504e-08, + "loss": 0.0004, + "step": 60054 + }, + { + "epoch": 7.87, + "learning_rate": 1.454088834595968e-08, + "loss": 0.0003, + "step": 60055 + }, + { + "epoch": 7.87, + "learning_rate": 1.4512313669247103e-08, + "loss": 0.0005, + "step": 60056 + }, + { + "epoch": 7.87, + "learning_rate": 1.4483767076186595e-08, + "loss": 0.0003, + "step": 60057 + }, + { + "epoch": 7.87, + "learning_rate": 1.4455248566859203e-08, + "loss": 0.0006, + "step": 60058 + }, + { + "epoch": 7.87, + "learning_rate": 1.4426758141343756e-08, + "loss": 0.0004, + "step": 60059 + }, + { + "epoch": 7.87, + "learning_rate": 1.4398295799722406e-08, + "loss": 0.0007, + "step": 60060 + }, + { + "epoch": 7.87, + "learning_rate": 1.4369861542072872e-08, + "loss": 0.0005, + "step": 60061 + }, + { + "epoch": 7.87, + "learning_rate": 1.4341455368477309e-08, + "loss": 0.0006, + "step": 60062 + }, + { + "epoch": 7.87, + "learning_rate": 1.4313077279013432e-08, + "loss": 0.001, + "step": 60063 + }, + { + "epoch": 7.87, + "learning_rate": 1.4284727273763399e-08, + "loss": 0.0002, + "step": 60064 + }, + { + "epoch": 7.87, + "learning_rate": 1.4256405352806035e-08, + "loss": 0.0004, + "step": 60065 + }, + { + "epoch": 7.87, + "learning_rate": 1.4228111516220167e-08, + "loss": 0.0002, + "step": 60066 + }, + { + "epoch": 7.87, + "learning_rate": 1.4199845764085729e-08, + "loss": 0.0007, + "step": 60067 + }, + { + "epoch": 7.87, + "learning_rate": 1.4171608096482658e-08, + "loss": 0.0002, + "step": 60068 + }, + { + "epoch": 7.87, + "learning_rate": 1.4143398513489781e-08, + "loss": 0.0005, + "step": 60069 + }, + { + "epoch": 7.87, + "learning_rate": 1.4115217015187033e-08, + "loss": 0.0013, + "step": 60070 + }, + { + "epoch": 7.87, + "learning_rate": 1.4087063601654349e-08, + "loss": 0.0001, + "step": 60071 + }, + { + "epoch": 7.87, + "learning_rate": 1.4058938272969447e-08, + "loss": 0.0008, + "step": 60072 + }, + { + "epoch": 7.87, + "learning_rate": 1.4030841029211151e-08, + "loss": 0.0005, + "step": 60073 + }, + { + "epoch": 7.87, + "learning_rate": 1.4002771870460508e-08, + "loss": 0.0003, + "step": 60074 + }, + { + "epoch": 7.87, + "learning_rate": 1.3974730796795232e-08, + "loss": 0.0003, + "step": 60075 + }, + { + "epoch": 7.87, + "learning_rate": 1.3946717808293042e-08, + "loss": 0.0003, + "step": 60076 + }, + { + "epoch": 7.87, + "learning_rate": 1.3918732905034982e-08, + "loss": 0.0009, + "step": 60077 + }, + { + "epoch": 7.87, + "learning_rate": 1.3890776087098767e-08, + "loss": 0.0006, + "step": 60078 + }, + { + "epoch": 7.87, + "learning_rate": 1.3862847354562114e-08, + "loss": 0.0005, + "step": 60079 + }, + { + "epoch": 7.87, + "learning_rate": 1.3834946707504959e-08, + "loss": 0.0003, + "step": 60080 + }, + { + "epoch": 7.87, + "learning_rate": 1.3807074146005018e-08, + "loss": 0.0008, + "step": 60081 + }, + { + "epoch": 7.87, + "learning_rate": 1.3779229670141115e-08, + "loss": 0.0004, + "step": 60082 + }, + { + "epoch": 7.87, + "learning_rate": 1.3751413279992076e-08, + "loss": 0.0004, + "step": 60083 + }, + { + "epoch": 7.87, + "learning_rate": 1.3723624975634509e-08, + "loss": 0.0004, + "step": 60084 + }, + { + "epoch": 7.87, + "learning_rate": 1.3695864757147237e-08, + "loss": 0.0004, + "step": 60085 + }, + { + "epoch": 7.87, + "learning_rate": 1.3668132624607978e-08, + "loss": 0.0005, + "step": 60086 + }, + { + "epoch": 7.87, + "learning_rate": 1.3640428578096665e-08, + "loss": 0.0002, + "step": 60087 + }, + { + "epoch": 7.87, + "learning_rate": 1.3612752617689906e-08, + "loss": 0.001, + "step": 60088 + }, + { + "epoch": 7.87, + "learning_rate": 1.3585104743464305e-08, + "loss": 0.0003, + "step": 60089 + }, + { + "epoch": 7.87, + "learning_rate": 1.35574849554998e-08, + "loss": 0.0003, + "step": 60090 + }, + { + "epoch": 7.87, + "learning_rate": 1.3529893253871884e-08, + "loss": 0.0005, + "step": 60091 + }, + { + "epoch": 7.87, + "learning_rate": 1.3502329638660494e-08, + "loss": 0.0007, + "step": 60092 + }, + { + "epoch": 7.87, + "learning_rate": 1.3474794109941125e-08, + "loss": 0.0004, + "step": 60093 + }, + { + "epoch": 7.87, + "learning_rate": 1.3447286667792603e-08, + "loss": 0.0002, + "step": 60094 + }, + { + "epoch": 7.87, + "learning_rate": 1.3419807312291534e-08, + "loss": 0.0004, + "step": 60095 + }, + { + "epoch": 7.87, + "learning_rate": 1.339235604351563e-08, + "loss": 0.0004, + "step": 60096 + }, + { + "epoch": 7.87, + "learning_rate": 1.3364932861541502e-08, + "loss": 0.0003, + "step": 60097 + }, + { + "epoch": 7.87, + "learning_rate": 1.3337537766447973e-08, + "loss": 0.0001, + "step": 60098 + }, + { + "epoch": 7.87, + "learning_rate": 1.3310170758309427e-08, + "loss": 0.0004, + "step": 60099 + }, + { + "epoch": 7.87, + "learning_rate": 1.3282831837204691e-08, + "loss": 0.0005, + "step": 60100 + }, + { + "epoch": 7.87, + "learning_rate": 1.325552100321037e-08, + "loss": 0.0004, + "step": 60101 + }, + { + "epoch": 7.87, + "learning_rate": 1.322823825640307e-08, + "loss": 0.0003, + "step": 60102 + }, + { + "epoch": 7.87, + "learning_rate": 1.3200983596859395e-08, + "loss": 0.0006, + "step": 60103 + }, + { + "epoch": 7.87, + "learning_rate": 1.3173757024655953e-08, + "loss": 0.0003, + "step": 60104 + }, + { + "epoch": 7.87, + "learning_rate": 1.3146558539869348e-08, + "loss": 0.0004, + "step": 60105 + }, + { + "epoch": 7.87, + "learning_rate": 1.3119388142576183e-08, + "loss": 0.0005, + "step": 60106 + }, + { + "epoch": 7.87, + "learning_rate": 1.3092245832853068e-08, + "loss": 0.0004, + "step": 60107 + }, + { + "epoch": 7.87, + "learning_rate": 1.3065131610776605e-08, + "loss": 0.0005, + "step": 60108 + }, + { + "epoch": 7.87, + "learning_rate": 1.3038045476422289e-08, + "loss": 0.0006, + "step": 60109 + }, + { + "epoch": 7.87, + "learning_rate": 1.3010987429866728e-08, + "loss": 0.0002, + "step": 60110 + }, + { + "epoch": 7.87, + "learning_rate": 1.2983957471185415e-08, + "loss": 0.0005, + "step": 60111 + }, + { + "epoch": 7.87, + "learning_rate": 1.2956955600454957e-08, + "loss": 0.0002, + "step": 60112 + }, + { + "epoch": 7.87, + "learning_rate": 1.2929981817751957e-08, + "loss": 0.0006, + "step": 60113 + }, + { + "epoch": 7.87, + "learning_rate": 1.2903036123150803e-08, + "loss": 0.0003, + "step": 60114 + }, + { + "epoch": 7.87, + "learning_rate": 1.2876118516728097e-08, + "loss": 0.0004, + "step": 60115 + }, + { + "epoch": 7.87, + "learning_rate": 1.2849228998560448e-08, + "loss": 0.0003, + "step": 60116 + }, + { + "epoch": 7.87, + "learning_rate": 1.2822367568721128e-08, + "loss": 0.0004, + "step": 60117 + }, + { + "epoch": 7.88, + "learning_rate": 1.2795534227287854e-08, + "loss": 0.0003, + "step": 60118 + }, + { + "epoch": 7.88, + "learning_rate": 1.2768728974335009e-08, + "loss": 0.0003, + "step": 60119 + }, + { + "epoch": 7.88, + "learning_rate": 1.274195180993809e-08, + "loss": 0.0003, + "step": 60120 + }, + { + "epoch": 7.88, + "learning_rate": 1.2715202734172593e-08, + "loss": 0.0009, + "step": 60121 + }, + { + "epoch": 7.88, + "learning_rate": 1.2688481747115123e-08, + "loss": 0.0004, + "step": 60122 + }, + { + "epoch": 7.88, + "learning_rate": 1.2661788848837842e-08, + "loss": 0.0006, + "step": 60123 + }, + { + "epoch": 7.88, + "learning_rate": 1.2635124039417357e-08, + "loss": 0.0005, + "step": 60124 + }, + { + "epoch": 7.88, + "learning_rate": 1.2608487318929164e-08, + "loss": 0.0005, + "step": 60125 + }, + { + "epoch": 7.88, + "learning_rate": 1.2581878687448756e-08, + "loss": 0.001, + "step": 60126 + }, + { + "epoch": 7.88, + "learning_rate": 1.25552981450483e-08, + "loss": 0.0004, + "step": 60127 + }, + { + "epoch": 7.88, + "learning_rate": 1.2528745691805511e-08, + "loss": 0.0005, + "step": 60128 + }, + { + "epoch": 7.88, + "learning_rate": 1.2502221327792552e-08, + "loss": 0.0004, + "step": 60129 + }, + { + "epoch": 7.88, + "learning_rate": 1.247572505308603e-08, + "loss": 0.0003, + "step": 60130 + }, + { + "epoch": 7.88, + "learning_rate": 1.2449256867759218e-08, + "loss": 0.0002, + "step": 60131 + }, + { + "epoch": 7.88, + "learning_rate": 1.2422816771887614e-08, + "loss": 0.0018, + "step": 60132 + }, + { + "epoch": 7.88, + "learning_rate": 1.239640476554449e-08, + "loss": 0.0004, + "step": 60133 + }, + { + "epoch": 7.88, + "learning_rate": 1.2370020848805341e-08, + "loss": 0.0007, + "step": 60134 + }, + { + "epoch": 7.88, + "learning_rate": 1.2343665021742335e-08, + "loss": 0.0002, + "step": 60135 + }, + { + "epoch": 7.88, + "learning_rate": 1.2317337284432074e-08, + "loss": 0.0005, + "step": 60136 + }, + { + "epoch": 7.88, + "learning_rate": 1.2291037636946723e-08, + "loss": 0.0006, + "step": 60137 + }, + { + "epoch": 7.88, + "learning_rate": 1.2264766079361778e-08, + "loss": 0.0002, + "step": 60138 + }, + { + "epoch": 7.88, + "learning_rate": 1.2238522611749404e-08, + "loss": 0.0006, + "step": 60139 + }, + { + "epoch": 7.88, + "learning_rate": 1.2212307234183984e-08, + "loss": 0.0004, + "step": 60140 + }, + { + "epoch": 7.88, + "learning_rate": 1.2186119946741014e-08, + "loss": 0.0002, + "step": 60141 + }, + { + "epoch": 7.88, + "learning_rate": 1.2159960749491551e-08, + "loss": 0.0004, + "step": 60142 + }, + { + "epoch": 7.88, + "learning_rate": 1.2133829642511086e-08, + "loss": 0.0005, + "step": 60143 + }, + { + "epoch": 7.88, + "learning_rate": 1.2107726625871785e-08, + "loss": 0.0005, + "step": 60144 + }, + { + "epoch": 7.88, + "learning_rate": 1.2081651699646923e-08, + "loss": 0.0003, + "step": 60145 + }, + { + "epoch": 7.88, + "learning_rate": 1.2055604863911996e-08, + "loss": 0.0002, + "step": 60146 + }, + { + "epoch": 7.88, + "learning_rate": 1.2029586118738058e-08, + "loss": 0.0008, + "step": 60147 + }, + { + "epoch": 7.88, + "learning_rate": 1.2003595464199492e-08, + "loss": 0.0002, + "step": 60148 + }, + { + "epoch": 7.88, + "learning_rate": 1.1977632900368463e-08, + "loss": 0.0005, + "step": 60149 + }, + { + "epoch": 7.88, + "learning_rate": 1.1951698427318248e-08, + "loss": 0.0003, + "step": 60150 + }, + { + "epoch": 7.88, + "learning_rate": 1.192579204512212e-08, + "loss": 0.0003, + "step": 60151 + }, + { + "epoch": 7.88, + "learning_rate": 1.1899913753852243e-08, + "loss": 0.0007, + "step": 60152 + }, + { + "epoch": 7.88, + "learning_rate": 1.1874063553583005e-08, + "loss": 0.0007, + "step": 60153 + }, + { + "epoch": 7.88, + "learning_rate": 1.1848241444385455e-08, + "loss": 0.0007, + "step": 60154 + }, + { + "epoch": 7.88, + "learning_rate": 1.1822447426333983e-08, + "loss": 0.0003, + "step": 60155 + }, + { + "epoch": 7.88, + "learning_rate": 1.179668149949853e-08, + "loss": 0.0003, + "step": 60156 + }, + { + "epoch": 7.88, + "learning_rate": 1.1770943663953483e-08, + "loss": 0.0003, + "step": 60157 + }, + { + "epoch": 7.88, + "learning_rate": 1.1745233919771003e-08, + "loss": 0.0002, + "step": 60158 + }, + { + "epoch": 7.88, + "learning_rate": 1.171955226702326e-08, + "loss": 0.0003, + "step": 60159 + }, + { + "epoch": 7.88, + "learning_rate": 1.1693898705782414e-08, + "loss": 0.0004, + "step": 60160 + }, + { + "epoch": 7.88, + "learning_rate": 1.166827323611952e-08, + "loss": 0.0003, + "step": 60161 + }, + { + "epoch": 7.88, + "learning_rate": 1.1642675858107854e-08, + "loss": 0.0003, + "step": 60162 + }, + { + "epoch": 7.88, + "learning_rate": 1.1617106571820691e-08, + "loss": 0.0002, + "step": 60163 + }, + { + "epoch": 7.88, + "learning_rate": 1.1591565377326863e-08, + "loss": 0.0005, + "step": 60164 + }, + { + "epoch": 7.88, + "learning_rate": 1.1566052274700756e-08, + "loss": 0.0002, + "step": 60165 + }, + { + "epoch": 7.88, + "learning_rate": 1.1540567264013425e-08, + "loss": 0.0006, + "step": 60166 + }, + { + "epoch": 7.88, + "learning_rate": 1.1515110345335923e-08, + "loss": 0.0005, + "step": 60167 + }, + { + "epoch": 7.88, + "learning_rate": 1.1489681518740414e-08, + "loss": 0.0002, + "step": 60168 + }, + { + "epoch": 7.88, + "learning_rate": 1.1464280784297954e-08, + "loss": 0.0003, + "step": 60169 + }, + { + "epoch": 7.88, + "learning_rate": 1.1438908142080707e-08, + "loss": 0.0005, + "step": 60170 + }, + { + "epoch": 7.88, + "learning_rate": 1.1413563592159726e-08, + "loss": 0.0003, + "step": 60171 + }, + { + "epoch": 7.88, + "learning_rate": 1.1388247134604957e-08, + "loss": 0.0003, + "step": 60172 + }, + { + "epoch": 7.88, + "learning_rate": 1.1362958769489674e-08, + "loss": 0.0004, + "step": 60173 + }, + { + "epoch": 7.88, + "learning_rate": 1.133769849688382e-08, + "loss": 0.0002, + "step": 60174 + }, + { + "epoch": 7.88, + "learning_rate": 1.1312466316859561e-08, + "loss": 0.0008, + "step": 60175 + }, + { + "epoch": 7.88, + "learning_rate": 1.128726222948684e-08, + "loss": 0.0006, + "step": 60176 + }, + { + "epoch": 7.88, + "learning_rate": 1.1262086234836711e-08, + "loss": 0.0002, + "step": 60177 + }, + { + "epoch": 7.88, + "learning_rate": 1.1236938332979119e-08, + "loss": 0.0005, + "step": 60178 + }, + { + "epoch": 7.88, + "learning_rate": 1.1211818523986229e-08, + "loss": 0.0009, + "step": 60179 + }, + { + "epoch": 7.88, + "learning_rate": 1.1186726807929094e-08, + "loss": 0.0002, + "step": 60180 + }, + { + "epoch": 7.88, + "learning_rate": 1.1161663184876547e-08, + "loss": 0.0006, + "step": 60181 + }, + { + "epoch": 7.88, + "learning_rate": 1.1136627654899646e-08, + "loss": 0.0003, + "step": 60182 + }, + { + "epoch": 7.88, + "learning_rate": 1.1111620218069441e-08, + "loss": 0.0002, + "step": 60183 + }, + { + "epoch": 7.88, + "learning_rate": 1.1086640874455878e-08, + "loss": 0.0002, + "step": 60184 + }, + { + "epoch": 7.88, + "learning_rate": 1.1061689624128902e-08, + "loss": 0.0006, + "step": 60185 + }, + { + "epoch": 7.88, + "learning_rate": 1.1036766467159565e-08, + "loss": 0.0002, + "step": 60186 + }, + { + "epoch": 7.88, + "learning_rate": 1.1011871403616703e-08, + "loss": 0.0006, + "step": 60187 + }, + { + "epoch": 7.88, + "learning_rate": 1.0987004433571369e-08, + "loss": 0.0004, + "step": 60188 + }, + { + "epoch": 7.88, + "learning_rate": 1.0962165557092397e-08, + "loss": 0.0008, + "step": 60189 + }, + { + "epoch": 7.88, + "learning_rate": 1.0937354774251952e-08, + "loss": 0.001, + "step": 60190 + }, + { + "epoch": 7.88, + "learning_rate": 1.0912572085116646e-08, + "loss": 0.0003, + "step": 60191 + }, + { + "epoch": 7.88, + "learning_rate": 1.0887817489758645e-08, + "loss": 0.0002, + "step": 60192 + }, + { + "epoch": 7.88, + "learning_rate": 1.0863090988246783e-08, + "loss": 0.0012, + "step": 60193 + }, + { + "epoch": 7.88, + "learning_rate": 1.0838392580649893e-08, + "loss": 0.0008, + "step": 60194 + }, + { + "epoch": 7.89, + "learning_rate": 1.0813722267039029e-08, + "loss": 0.0007, + "step": 60195 + }, + { + "epoch": 7.89, + "learning_rate": 1.0789080047481915e-08, + "loss": 0.0005, + "step": 60196 + }, + { + "epoch": 7.89, + "learning_rate": 1.0764465922048495e-08, + "loss": 0.0006, + "step": 60197 + }, + { + "epoch": 7.89, + "learning_rate": 1.0739879890807603e-08, + "loss": 0.0003, + "step": 60198 + }, + { + "epoch": 7.89, + "learning_rate": 1.0715321953830292e-08, + "loss": 0.0002, + "step": 60199 + }, + { + "epoch": 7.89, + "learning_rate": 1.0690792111183179e-08, + "loss": 0.0002, + "step": 60200 + }, + { + "epoch": 7.89, + "learning_rate": 1.0666290362936205e-08, + "loss": 0.0004, + "step": 60201 + }, + { + "epoch": 7.89, + "learning_rate": 1.0641816709159314e-08, + "loss": 0.0003, + "step": 60202 + }, + { + "epoch": 7.89, + "learning_rate": 1.061737114991912e-08, + "loss": 0.0004, + "step": 60203 + }, + { + "epoch": 7.89, + "learning_rate": 1.0592953685285568e-08, + "loss": 0.0004, + "step": 60204 + }, + { + "epoch": 7.89, + "learning_rate": 1.05685643153286e-08, + "loss": 0.0004, + "step": 60205 + }, + { + "epoch": 7.89, + "learning_rate": 1.054420304011483e-08, + "loss": 0.0005, + "step": 60206 + }, + { + "epoch": 7.89, + "learning_rate": 1.0519869859713094e-08, + "loss": 0.0005, + "step": 60207 + }, + { + "epoch": 7.89, + "learning_rate": 1.0495564774192223e-08, + "loss": 0.0003, + "step": 60208 + }, + { + "epoch": 7.89, + "learning_rate": 1.0471287783622164e-08, + "loss": 0.0004, + "step": 60209 + }, + { + "epoch": 7.89, + "learning_rate": 1.0447038888068418e-08, + "loss": 0.0005, + "step": 60210 + }, + { + "epoch": 7.89, + "learning_rate": 1.0422818087599818e-08, + "loss": 0.0004, + "step": 60211 + }, + { + "epoch": 7.89, + "learning_rate": 1.039862538228631e-08, + "loss": 0.0008, + "step": 60212 + }, + { + "epoch": 7.89, + "learning_rate": 1.0374460772194506e-08, + "loss": 0.0001, + "step": 60213 + }, + { + "epoch": 7.89, + "learning_rate": 1.0350324257392131e-08, + "loss": 0.0003, + "step": 60214 + }, + { + "epoch": 7.89, + "learning_rate": 1.0326215837946908e-08, + "loss": 0.0003, + "step": 60215 + }, + { + "epoch": 7.89, + "learning_rate": 1.0302135513927669e-08, + "loss": 0.0008, + "step": 60216 + }, + { + "epoch": 7.89, + "learning_rate": 1.0278083285402141e-08, + "loss": 0.0011, + "step": 60217 + }, + { + "epoch": 7.89, + "learning_rate": 1.0254059152436935e-08, + "loss": 0.0007, + "step": 60218 + }, + { + "epoch": 7.89, + "learning_rate": 1.0230063115100885e-08, + "loss": 0.0005, + "step": 60219 + }, + { + "epoch": 7.89, + "learning_rate": 1.0206095173460606e-08, + "loss": 0.0003, + "step": 60220 + }, + { + "epoch": 7.89, + "learning_rate": 1.018215532758382e-08, + "loss": 0.0003, + "step": 60221 + }, + { + "epoch": 7.89, + "learning_rate": 1.0158243577537142e-08, + "loss": 0.0005, + "step": 60222 + }, + { + "epoch": 7.89, + "learning_rate": 1.0134359923388292e-08, + "loss": 0.0004, + "step": 60223 + }, + { + "epoch": 7.89, + "learning_rate": 1.0110504365204998e-08, + "loss": 0.0005, + "step": 60224 + }, + { + "epoch": 7.89, + "learning_rate": 1.0086676903054982e-08, + "loss": 0.0003, + "step": 60225 + }, + { + "epoch": 7.89, + "learning_rate": 1.0062877537002635e-08, + "loss": 0.001, + "step": 60226 + }, + { + "epoch": 7.89, + "learning_rate": 1.0039106267116794e-08, + "loss": 0.0002, + "step": 60227 + }, + { + "epoch": 7.89, + "learning_rate": 1.001536309346407e-08, + "loss": 0.0002, + "step": 60228 + }, + { + "epoch": 7.89, + "learning_rate": 9.991648016111078e-09, + "loss": 0.0002, + "step": 60229 + }, + { + "epoch": 7.89, + "learning_rate": 9.96796103512443e-09, + "loss": 0.0003, + "step": 60230 + }, + { + "epoch": 7.89, + "learning_rate": 9.944302150571849e-09, + "loss": 0.0002, + "step": 60231 + }, + { + "epoch": 7.89, + "learning_rate": 9.920671362517731e-09, + "loss": 0.0003, + "step": 60232 + }, + { + "epoch": 7.89, + "learning_rate": 9.897068671030907e-09, + "loss": 0.0009, + "step": 60233 + }, + { + "epoch": 7.89, + "learning_rate": 9.873494076175771e-09, + "loss": 0.0007, + "step": 60234 + }, + { + "epoch": 7.89, + "learning_rate": 9.849947578018936e-09, + "loss": 0.0007, + "step": 60235 + }, + { + "epoch": 7.89, + "learning_rate": 9.826429176628127e-09, + "loss": 0.0007, + "step": 60236 + }, + { + "epoch": 7.89, + "learning_rate": 9.802938872067736e-09, + "loss": 0.0006, + "step": 60237 + }, + { + "epoch": 7.89, + "learning_rate": 9.779476664405484e-09, + "loss": 0.0015, + "step": 60238 + }, + { + "epoch": 7.89, + "learning_rate": 9.756042553705768e-09, + "loss": 0.0008, + "step": 60239 + }, + { + "epoch": 7.89, + "learning_rate": 9.732636540035201e-09, + "loss": 0.0005, + "step": 60240 + }, + { + "epoch": 7.89, + "learning_rate": 9.709258623459283e-09, + "loss": 0.0005, + "step": 60241 + }, + { + "epoch": 7.89, + "learning_rate": 9.685908804044631e-09, + "loss": 0.001, + "step": 60242 + }, + { + "epoch": 7.89, + "learning_rate": 9.662587081855634e-09, + "loss": 0.0006, + "step": 60243 + }, + { + "epoch": 7.89, + "learning_rate": 9.63929345696002e-09, + "loss": 0.0004, + "step": 60244 + }, + { + "epoch": 7.89, + "learning_rate": 9.616027929419958e-09, + "loss": 0.0004, + "step": 60245 + }, + { + "epoch": 7.89, + "learning_rate": 9.592790499304283e-09, + "loss": 0.0003, + "step": 60246 + }, + { + "epoch": 7.89, + "learning_rate": 9.569581166676279e-09, + "loss": 0.0003, + "step": 60247 + }, + { + "epoch": 7.89, + "learning_rate": 9.546399931601446e-09, + "loss": 0.0006, + "step": 60248 + }, + { + "epoch": 7.89, + "learning_rate": 9.523246794146401e-09, + "loss": 0.0003, + "step": 60249 + }, + { + "epoch": 7.89, + "learning_rate": 9.500121754374425e-09, + "loss": 0.0004, + "step": 60250 + }, + { + "epoch": 7.89, + "learning_rate": 9.477024812351022e-09, + "loss": 0.0004, + "step": 60251 + }, + { + "epoch": 7.89, + "learning_rate": 9.453955968142802e-09, + "loss": 0.0007, + "step": 60252 + }, + { + "epoch": 7.89, + "learning_rate": 9.430915221811943e-09, + "loss": 0.0001, + "step": 60253 + }, + { + "epoch": 7.89, + "learning_rate": 9.407902573425054e-09, + "loss": 0.0004, + "step": 60254 + }, + { + "epoch": 7.89, + "learning_rate": 9.38491802304653e-09, + "loss": 0.0004, + "step": 60255 + }, + { + "epoch": 7.89, + "learning_rate": 9.361961570740763e-09, + "loss": 0.0003, + "step": 60256 + }, + { + "epoch": 7.89, + "learning_rate": 9.339033216572146e-09, + "loss": 0.0004, + "step": 60257 + }, + { + "epoch": 7.89, + "learning_rate": 9.316132960606183e-09, + "loss": 0.0006, + "step": 60258 + }, + { + "epoch": 7.89, + "learning_rate": 9.293260802906156e-09, + "loss": 0.0002, + "step": 60259 + }, + { + "epoch": 7.89, + "learning_rate": 9.27041674353757e-09, + "loss": 0.0005, + "step": 60260 + }, + { + "epoch": 7.89, + "learning_rate": 9.247600782562594e-09, + "loss": 0.0002, + "step": 60261 + }, + { + "epoch": 7.89, + "learning_rate": 9.224812920046734e-09, + "loss": 0.0003, + "step": 60262 + }, + { + "epoch": 7.89, + "learning_rate": 9.20205315605549e-09, + "loss": 0.0005, + "step": 60263 + }, + { + "epoch": 7.89, + "learning_rate": 9.179321490649928e-09, + "loss": 0.0009, + "step": 60264 + }, + { + "epoch": 7.89, + "learning_rate": 9.15661792389666e-09, + "loss": 0.0006, + "step": 60265 + }, + { + "epoch": 7.89, + "learning_rate": 9.133942455857858e-09, + "loss": 0.0004, + "step": 60266 + }, + { + "epoch": 7.89, + "learning_rate": 9.111295086597915e-09, + "loss": 0.0002, + "step": 60267 + }, + { + "epoch": 7.89, + "learning_rate": 9.088675816180115e-09, + "loss": 0.0002, + "step": 60268 + }, + { + "epoch": 7.89, + "learning_rate": 9.066084644667739e-09, + "loss": 0.0001, + "step": 60269 + }, + { + "epoch": 7.89, + "learning_rate": 9.043521572126291e-09, + "loss": 0.0006, + "step": 60270 + }, + { + "epoch": 7.9, + "learning_rate": 9.020986598616832e-09, + "loss": 0.0004, + "step": 60271 + }, + { + "epoch": 7.9, + "learning_rate": 8.998479724204868e-09, + "loss": 0.0004, + "step": 60272 + }, + { + "epoch": 7.9, + "learning_rate": 8.976000948951457e-09, + "loss": 0.0005, + "step": 60273 + }, + { + "epoch": 7.9, + "learning_rate": 8.953550272920996e-09, + "loss": 0.0004, + "step": 60274 + }, + { + "epoch": 7.9, + "learning_rate": 8.931127696177876e-09, + "loss": 0.0003, + "step": 60275 + }, + { + "epoch": 7.9, + "learning_rate": 8.90873321878316e-09, + "loss": 0.0016, + "step": 60276 + }, + { + "epoch": 7.9, + "learning_rate": 8.886366840800132e-09, + "loss": 0.0007, + "step": 60277 + }, + { + "epoch": 7.9, + "learning_rate": 8.86402856229318e-09, + "loss": 0.0005, + "step": 60278 + }, + { + "epoch": 7.9, + "learning_rate": 8.841718383323372e-09, + "loss": 0.0003, + "step": 60279 + }, + { + "epoch": 7.9, + "learning_rate": 8.819436303953987e-09, + "loss": 0.0005, + "step": 60280 + }, + { + "epoch": 7.9, + "learning_rate": 8.79718232424831e-09, + "loss": 0.0005, + "step": 60281 + }, + { + "epoch": 7.9, + "learning_rate": 8.774956444268511e-09, + "loss": 0.0005, + "step": 60282 + }, + { + "epoch": 7.9, + "learning_rate": 8.752758664076765e-09, + "loss": 0.0004, + "step": 60283 + }, + { + "epoch": 7.9, + "learning_rate": 8.730588983736354e-09, + "loss": 0.0005, + "step": 60284 + }, + { + "epoch": 7.9, + "learning_rate": 8.70844740330834e-09, + "loss": 0.0005, + "step": 60285 + }, + { + "epoch": 7.9, + "learning_rate": 8.686333922856006e-09, + "loss": 0.0004, + "step": 60286 + }, + { + "epoch": 7.9, + "learning_rate": 8.664248542441523e-09, + "loss": 0.0006, + "step": 60287 + }, + { + "epoch": 7.9, + "learning_rate": 8.642191262127064e-09, + "loss": 0.0007, + "step": 60288 + }, + { + "epoch": 7.9, + "learning_rate": 8.620162081973694e-09, + "loss": 0.0008, + "step": 60289 + }, + { + "epoch": 7.9, + "learning_rate": 8.598161002044692e-09, + "loss": 0.0001, + "step": 60290 + }, + { + "epoch": 7.9, + "learning_rate": 8.576188022401122e-09, + "loss": 0.0004, + "step": 60291 + }, + { + "epoch": 7.9, + "learning_rate": 8.554243143104046e-09, + "loss": 0.0007, + "step": 60292 + }, + { + "epoch": 7.9, + "learning_rate": 8.532326364216747e-09, + "loss": 0.0006, + "step": 60293 + }, + { + "epoch": 7.9, + "learning_rate": 8.510437685800289e-09, + "loss": 0.0004, + "step": 60294 + }, + { + "epoch": 7.9, + "learning_rate": 8.48857710791684e-09, + "loss": 0.0006, + "step": 60295 + }, + { + "epoch": 7.9, + "learning_rate": 8.466744630626356e-09, + "loss": 0.0002, + "step": 60296 + }, + { + "epoch": 7.9, + "learning_rate": 8.444940253991008e-09, + "loss": 0.0003, + "step": 60297 + }, + { + "epoch": 7.9, + "learning_rate": 8.423163978072969e-09, + "loss": 0.0004, + "step": 60298 + }, + { + "epoch": 7.9, + "learning_rate": 8.401415802931079e-09, + "loss": 0.0005, + "step": 60299 + }, + { + "epoch": 7.9, + "learning_rate": 8.379695728629734e-09, + "loss": 0.0004, + "step": 60300 + }, + { + "epoch": 7.9, + "learning_rate": 8.358003755227773e-09, + "loss": 0.0005, + "step": 60301 + }, + { + "epoch": 7.9, + "learning_rate": 8.33633988278737e-09, + "loss": 0.0007, + "step": 60302 + }, + { + "epoch": 7.9, + "learning_rate": 8.314704111367366e-09, + "loss": 0.0003, + "step": 60303 + }, + { + "epoch": 7.9, + "learning_rate": 8.293096441032156e-09, + "loss": 0.0004, + "step": 60304 + }, + { + "epoch": 7.9, + "learning_rate": 8.27151687183947e-09, + "loss": 0.0001, + "step": 60305 + }, + { + "epoch": 7.9, + "learning_rate": 8.249965403851479e-09, + "loss": 0.0004, + "step": 60306 + }, + { + "epoch": 7.9, + "learning_rate": 8.228442037128137e-09, + "loss": 0.001, + "step": 60307 + }, + { + "epoch": 7.9, + "learning_rate": 8.206946771729397e-09, + "loss": 0.0004, + "step": 60308 + }, + { + "epoch": 7.9, + "learning_rate": 8.185479607717429e-09, + "loss": 0.0005, + "step": 60309 + }, + { + "epoch": 7.9, + "learning_rate": 8.164040545152186e-09, + "loss": 0.0003, + "step": 60310 + }, + { + "epoch": 7.9, + "learning_rate": 8.14262958409251e-09, + "loss": 0.0008, + "step": 60311 + }, + { + "epoch": 7.9, + "learning_rate": 8.121246724599463e-09, + "loss": 0.0003, + "step": 60312 + }, + { + "epoch": 7.9, + "learning_rate": 8.099891966734108e-09, + "loss": 0.0008, + "step": 60313 + }, + { + "epoch": 7.9, + "learning_rate": 8.078565310555286e-09, + "loss": 0.0003, + "step": 60314 + }, + { + "epoch": 7.9, + "learning_rate": 8.05726675612406e-09, + "loss": 0.0006, + "step": 60315 + }, + { + "epoch": 7.9, + "learning_rate": 8.035996303498162e-09, + "loss": 0.0006, + "step": 60316 + }, + { + "epoch": 7.9, + "learning_rate": 8.014753952740872e-09, + "loss": 0.0004, + "step": 60317 + }, + { + "epoch": 7.9, + "learning_rate": 7.993539703908815e-09, + "loss": 0.0006, + "step": 60318 + }, + { + "epoch": 7.9, + "learning_rate": 7.97235355706305e-09, + "loss": 0.0003, + "step": 60319 + }, + { + "epoch": 7.9, + "learning_rate": 7.951195512262422e-09, + "loss": 0.0003, + "step": 60320 + }, + { + "epoch": 7.9, + "learning_rate": 7.93006556956799e-09, + "loss": 0.0002, + "step": 60321 + }, + { + "epoch": 7.9, + "learning_rate": 7.908963729037489e-09, + "loss": 0.0008, + "step": 60322 + }, + { + "epoch": 7.9, + "learning_rate": 7.887889990730868e-09, + "loss": 0.0003, + "step": 60323 + }, + { + "epoch": 7.9, + "learning_rate": 7.866844354706971e-09, + "loss": 0.0003, + "step": 60324 + }, + { + "epoch": 7.9, + "learning_rate": 7.845826821025749e-09, + "loss": 0.0006, + "step": 60325 + }, + { + "epoch": 7.9, + "learning_rate": 7.824837389747153e-09, + "loss": 0.0003, + "step": 60326 + }, + { + "epoch": 7.9, + "learning_rate": 7.803876060927806e-09, + "loss": 0.0007, + "step": 60327 + }, + { + "epoch": 7.9, + "learning_rate": 7.78294283462877e-09, + "loss": 0.0003, + "step": 60328 + }, + { + "epoch": 7.9, + "learning_rate": 7.762037710907777e-09, + "loss": 0.0004, + "step": 60329 + }, + { + "epoch": 7.9, + "learning_rate": 7.741160689823668e-09, + "loss": 0.0003, + "step": 60330 + }, + { + "epoch": 7.9, + "learning_rate": 7.720311771435285e-09, + "loss": 0.0008, + "step": 60331 + }, + { + "epoch": 7.9, + "learning_rate": 7.69949095580147e-09, + "loss": 0.0002, + "step": 60332 + }, + { + "epoch": 7.9, + "learning_rate": 7.678698242981064e-09, + "loss": 0.0004, + "step": 60333 + }, + { + "epoch": 7.9, + "learning_rate": 7.65793363303291e-09, + "loss": 0.0006, + "step": 60334 + }, + { + "epoch": 7.9, + "learning_rate": 7.637197126013628e-09, + "loss": 0.0002, + "step": 60335 + }, + { + "epoch": 7.9, + "learning_rate": 7.616488721982062e-09, + "loss": 0.0002, + "step": 60336 + }, + { + "epoch": 7.9, + "learning_rate": 7.595808420998163e-09, + "loss": 0.0003, + "step": 60337 + }, + { + "epoch": 7.9, + "learning_rate": 7.575156223118552e-09, + "loss": 0.0005, + "step": 60338 + }, + { + "epoch": 7.9, + "learning_rate": 7.55453212840207e-09, + "loss": 0.0002, + "step": 60339 + }, + { + "epoch": 7.9, + "learning_rate": 7.53393613690645e-09, + "loss": 0.0003, + "step": 60340 + }, + { + "epoch": 7.9, + "learning_rate": 7.513368248688313e-09, + "loss": 0.0012, + "step": 60341 + }, + { + "epoch": 7.9, + "learning_rate": 7.49282846380761e-09, + "loss": 0.0005, + "step": 60342 + }, + { + "epoch": 7.9, + "learning_rate": 7.472316782320966e-09, + "loss": 0.0004, + "step": 60343 + }, + { + "epoch": 7.9, + "learning_rate": 7.451833204286107e-09, + "loss": 0.0003, + "step": 60344 + }, + { + "epoch": 7.9, + "learning_rate": 7.431377729760769e-09, + "loss": 0.0002, + "step": 60345 + }, + { + "epoch": 7.9, + "learning_rate": 7.4109503588026824e-09, + "loss": 0.0012, + "step": 60346 + }, + { + "epoch": 7.91, + "learning_rate": 7.390551091469577e-09, + "loss": 0.0007, + "step": 60347 + }, + { + "epoch": 7.91, + "learning_rate": 7.370179927818078e-09, + "loss": 0.0003, + "step": 60348 + }, + { + "epoch": 7.91, + "learning_rate": 7.349836867905913e-09, + "loss": 0.0003, + "step": 60349 + }, + { + "epoch": 7.91, + "learning_rate": 7.329521911790816e-09, + "loss": 0.0007, + "step": 60350 + }, + { + "epoch": 7.91, + "learning_rate": 7.309235059528297e-09, + "loss": 0.0001, + "step": 60351 + }, + { + "epoch": 7.91, + "learning_rate": 7.2889763111771985e-09, + "loss": 0.0006, + "step": 60352 + }, + { + "epoch": 7.91, + "learning_rate": 7.268745666793031e-09, + "loss": 0.0003, + "step": 60353 + }, + { + "epoch": 7.91, + "learning_rate": 7.248543126433527e-09, + "loss": 0.0006, + "step": 60354 + }, + { + "epoch": 7.91, + "learning_rate": 7.228368690155308e-09, + "loss": 0.0006, + "step": 60355 + }, + { + "epoch": 7.91, + "learning_rate": 7.2082223580149935e-09, + "loss": 0.001, + "step": 60356 + }, + { + "epoch": 7.91, + "learning_rate": 7.188104130070317e-09, + "loss": 0.0012, + "step": 60357 + }, + { + "epoch": 7.91, + "learning_rate": 7.168014006375679e-09, + "loss": 0.0001, + "step": 60358 + }, + { + "epoch": 7.91, + "learning_rate": 7.147951986989921e-09, + "loss": 0.0007, + "step": 60359 + }, + { + "epoch": 7.91, + "learning_rate": 7.1279180719674435e-09, + "loss": 0.0003, + "step": 60360 + }, + { + "epoch": 7.91, + "learning_rate": 7.107912261365979e-09, + "loss": 0.0003, + "step": 60361 + }, + { + "epoch": 7.91, + "learning_rate": 7.087934555242149e-09, + "loss": 0.0002, + "step": 60362 + }, + { + "epoch": 7.91, + "learning_rate": 7.067984953650353e-09, + "loss": 0.0002, + "step": 60363 + }, + { + "epoch": 7.91, + "learning_rate": 7.048063456648324e-09, + "loss": 0.0002, + "step": 60364 + }, + { + "epoch": 7.91, + "learning_rate": 7.028170064290463e-09, + "loss": 0.0003, + "step": 60365 + }, + { + "epoch": 7.91, + "learning_rate": 7.0083047766345e-09, + "loss": 0.0004, + "step": 60366 + }, + { + "epoch": 7.91, + "learning_rate": 6.988467593735948e-09, + "loss": 0.0003, + "step": 60367 + }, + { + "epoch": 7.91, + "learning_rate": 6.968658515649207e-09, + "loss": 0.0004, + "step": 60368 + }, + { + "epoch": 7.91, + "learning_rate": 6.9488775424320085e-09, + "loss": 0.0011, + "step": 60369 + }, + { + "epoch": 7.91, + "learning_rate": 6.9291246741376435e-09, + "loss": 0.0002, + "step": 60370 + }, + { + "epoch": 7.91, + "learning_rate": 6.909399910823844e-09, + "loss": 0.0003, + "step": 60371 + }, + { + "epoch": 7.91, + "learning_rate": 6.88970325254501e-09, + "loss": 0.0002, + "step": 60372 + }, + { + "epoch": 7.91, + "learning_rate": 6.870034699356654e-09, + "loss": 0.0002, + "step": 60373 + }, + { + "epoch": 7.91, + "learning_rate": 6.850394251314285e-09, + "loss": 0.0004, + "step": 60374 + }, + { + "epoch": 7.91, + "learning_rate": 6.830781908472306e-09, + "loss": 0.0005, + "step": 60375 + }, + { + "epoch": 7.91, + "learning_rate": 6.811197670887338e-09, + "loss": 0.0005, + "step": 60376 + }, + { + "epoch": 7.91, + "learning_rate": 6.79164153861378e-09, + "loss": 0.0004, + "step": 60377 + }, + { + "epoch": 7.91, + "learning_rate": 6.772113511706036e-09, + "loss": 0.0004, + "step": 60378 + }, + { + "epoch": 7.91, + "learning_rate": 6.7526135902207244e-09, + "loss": 0.0003, + "step": 60379 + }, + { + "epoch": 7.91, + "learning_rate": 6.733141774210028e-09, + "loss": 0.0004, + "step": 60380 + }, + { + "epoch": 7.91, + "learning_rate": 6.713698063730567e-09, + "loss": 0.0008, + "step": 60381 + }, + { + "epoch": 7.91, + "learning_rate": 6.6942824588367425e-09, + "loss": 0.0003, + "step": 60382 + }, + { + "epoch": 7.91, + "learning_rate": 6.674894959584066e-09, + "loss": 0.0004, + "step": 60383 + }, + { + "epoch": 7.91, + "learning_rate": 6.655535566024718e-09, + "loss": 0.0003, + "step": 60384 + }, + { + "epoch": 7.91, + "learning_rate": 6.636204278215319e-09, + "loss": 0.0002, + "step": 60385 + }, + { + "epoch": 7.91, + "learning_rate": 6.616901096209161e-09, + "loss": 0.0004, + "step": 60386 + }, + { + "epoch": 7.91, + "learning_rate": 6.597626020060643e-09, + "loss": 0.0011, + "step": 60387 + }, + { + "epoch": 7.91, + "learning_rate": 6.578379049824168e-09, + "loss": 0.0011, + "step": 60388 + }, + { + "epoch": 7.91, + "learning_rate": 6.559160185554137e-09, + "loss": 0.0003, + "step": 60389 + }, + { + "epoch": 7.91, + "learning_rate": 6.5399694273038386e-09, + "loss": 0.0002, + "step": 60390 + }, + { + "epoch": 7.91, + "learning_rate": 6.520806775127675e-09, + "loss": 0.0008, + "step": 60391 + }, + { + "epoch": 7.91, + "learning_rate": 6.501672229080047e-09, + "loss": 0.0003, + "step": 60392 + }, + { + "epoch": 7.91, + "learning_rate": 6.482565789214245e-09, + "loss": 0.0007, + "step": 60393 + }, + { + "epoch": 7.91, + "learning_rate": 6.463487455583561e-09, + "loss": 0.0004, + "step": 60394 + }, + { + "epoch": 7.91, + "learning_rate": 6.444437228242395e-09, + "loss": 0.0005, + "step": 60395 + }, + { + "epoch": 7.91, + "learning_rate": 6.425415107244037e-09, + "loss": 0.0005, + "step": 60396 + }, + { + "epoch": 7.91, + "learning_rate": 6.406421092641779e-09, + "loss": 0.0006, + "step": 60397 + }, + { + "epoch": 7.91, + "learning_rate": 6.387455184488911e-09, + "loss": 0.0003, + "step": 60398 + }, + { + "epoch": 7.91, + "learning_rate": 6.368517382839834e-09, + "loss": 0.0005, + "step": 60399 + }, + { + "epoch": 7.91, + "learning_rate": 6.3496076877478385e-09, + "loss": 0.0007, + "step": 60400 + }, + { + "epoch": 7.91, + "learning_rate": 6.3307260992639955e-09, + "loss": 0.0003, + "step": 60401 + }, + { + "epoch": 7.91, + "learning_rate": 6.311872617443815e-09, + "loss": 0.0003, + "step": 60402 + }, + { + "epoch": 7.91, + "learning_rate": 6.2930472423383685e-09, + "loss": 0.0003, + "step": 60403 + }, + { + "epoch": 7.91, + "learning_rate": 6.274249974002056e-09, + "loss": 0.0013, + "step": 60404 + }, + { + "epoch": 7.91, + "learning_rate": 6.255480812487058e-09, + "loss": 0.0005, + "step": 60405 + }, + { + "epoch": 7.91, + "learning_rate": 6.236739757846666e-09, + "loss": 0.0005, + "step": 60406 + }, + { + "epoch": 7.91, + "learning_rate": 6.21802681013306e-09, + "loss": 0.0004, + "step": 60407 + }, + { + "epoch": 7.91, + "learning_rate": 6.199341969398421e-09, + "loss": 0.0002, + "step": 60408 + }, + { + "epoch": 7.91, + "learning_rate": 6.180685235697148e-09, + "loss": 0.0005, + "step": 60409 + }, + { + "epoch": 7.91, + "learning_rate": 6.162056609079203e-09, + "loss": 0.0006, + "step": 60410 + }, + { + "epoch": 7.91, + "learning_rate": 6.143456089598987e-09, + "loss": 0.0008, + "step": 60411 + }, + { + "epoch": 7.91, + "learning_rate": 6.124883677307569e-09, + "loss": 0.0006, + "step": 60412 + }, + { + "epoch": 7.91, + "learning_rate": 6.106339372258241e-09, + "loss": 0.0005, + "step": 60413 + }, + { + "epoch": 7.91, + "learning_rate": 6.087823174503182e-09, + "loss": 0.0004, + "step": 60414 + }, + { + "epoch": 7.91, + "learning_rate": 6.069335084093464e-09, + "loss": 0.001, + "step": 60415 + }, + { + "epoch": 7.91, + "learning_rate": 6.050875101081266e-09, + "loss": 0.0005, + "step": 60416 + }, + { + "epoch": 7.91, + "learning_rate": 6.032443225518769e-09, + "loss": 0.0006, + "step": 60417 + }, + { + "epoch": 7.91, + "learning_rate": 6.014039457458154e-09, + "loss": 0.0006, + "step": 60418 + }, + { + "epoch": 7.91, + "learning_rate": 5.99566379695049e-09, + "loss": 0.0006, + "step": 60419 + }, + { + "epoch": 7.91, + "learning_rate": 5.977316244047959e-09, + "loss": 0.0005, + "step": 60420 + }, + { + "epoch": 7.91, + "learning_rate": 5.958996798801631e-09, + "loss": 0.0002, + "step": 60421 + }, + { + "epoch": 7.91, + "learning_rate": 5.940705461263685e-09, + "loss": 0.0001, + "step": 60422 + }, + { + "epoch": 7.91, + "learning_rate": 5.922442231485193e-09, + "loss": 0.0012, + "step": 60423 + }, + { + "epoch": 7.92, + "learning_rate": 5.904207109518334e-09, + "loss": 0.0006, + "step": 60424 + }, + { + "epoch": 7.92, + "learning_rate": 5.886000095411959e-09, + "loss": 0.0002, + "step": 60425 + }, + { + "epoch": 7.92, + "learning_rate": 5.8678211892204685e-09, + "loss": 0.0003, + "step": 60426 + }, + { + "epoch": 7.92, + "learning_rate": 5.849670390993822e-09, + "loss": 0.0004, + "step": 60427 + }, + { + "epoch": 7.92, + "learning_rate": 5.831547700781981e-09, + "loss": 0.0006, + "step": 60428 + }, + { + "epoch": 7.92, + "learning_rate": 5.813453118636015e-09, + "loss": 0.0007, + "step": 60429 + }, + { + "epoch": 7.92, + "learning_rate": 5.7953866446092134e-09, + "loss": 0.0005, + "step": 60430 + }, + { + "epoch": 7.92, + "learning_rate": 5.777348278749318e-09, + "loss": 0.0003, + "step": 60431 + }, + { + "epoch": 7.92, + "learning_rate": 5.759338021108507e-09, + "loss": 0.0004, + "step": 60432 + }, + { + "epoch": 7.92, + "learning_rate": 5.741355871737852e-09, + "loss": 0.0005, + "step": 60433 + }, + { + "epoch": 7.92, + "learning_rate": 5.723401830687314e-09, + "loss": 0.0004, + "step": 60434 + }, + { + "epoch": 7.92, + "learning_rate": 5.705475898007961e-09, + "loss": 0.0013, + "step": 60435 + }, + { + "epoch": 7.92, + "learning_rate": 5.6875780737486454e-09, + "loss": 0.0008, + "step": 60436 + }, + { + "epoch": 7.92, + "learning_rate": 5.669708357961545e-09, + "loss": 0.0017, + "step": 60437 + }, + { + "epoch": 7.92, + "learning_rate": 5.651866750695511e-09, + "loss": 0.0003, + "step": 60438 + }, + { + "epoch": 7.92, + "learning_rate": 5.634053252001615e-09, + "loss": 0.0003, + "step": 60439 + }, + { + "epoch": 7.92, + "learning_rate": 5.616267861929814e-09, + "loss": 0.0011, + "step": 60440 + }, + { + "epoch": 7.92, + "learning_rate": 5.59851058053007e-09, + "loss": 0.0006, + "step": 60441 + }, + { + "epoch": 7.92, + "learning_rate": 5.580781407852342e-09, + "loss": 0.0005, + "step": 60442 + }, + { + "epoch": 7.92, + "learning_rate": 5.563080343946592e-09, + "loss": 0.0004, + "step": 60443 + }, + { + "epoch": 7.92, + "learning_rate": 5.5454073888627776e-09, + "loss": 0.0003, + "step": 60444 + }, + { + "epoch": 7.92, + "learning_rate": 5.52776254264975e-09, + "loss": 0.0008, + "step": 60445 + }, + { + "epoch": 7.92, + "learning_rate": 5.51014580535747e-09, + "loss": 0.0005, + "step": 60446 + }, + { + "epoch": 7.92, + "learning_rate": 5.492557177035896e-09, + "loss": 0.0003, + "step": 60447 + }, + { + "epoch": 7.92, + "learning_rate": 5.474996657733877e-09, + "loss": 0.0002, + "step": 60448 + }, + { + "epoch": 7.92, + "learning_rate": 5.4574642475013765e-09, + "loss": 0.0003, + "step": 60449 + }, + { + "epoch": 7.92, + "learning_rate": 5.4399599463883515e-09, + "loss": 0.0004, + "step": 60450 + }, + { + "epoch": 7.92, + "learning_rate": 5.422483754442542e-09, + "loss": 0.0016, + "step": 60451 + }, + { + "epoch": 7.92, + "learning_rate": 5.4050356717128e-09, + "loss": 0.0003, + "step": 60452 + }, + { + "epoch": 7.92, + "learning_rate": 5.387615698250193e-09, + "loss": 0.0002, + "step": 60453 + }, + { + "epoch": 7.92, + "learning_rate": 5.370223834102462e-09, + "loss": 0.0003, + "step": 60454 + }, + { + "epoch": 7.92, + "learning_rate": 5.352860079319566e-09, + "loss": 0.0005, + "step": 60455 + }, + { + "epoch": 7.92, + "learning_rate": 5.335524433948136e-09, + "loss": 0.0008, + "step": 60456 + }, + { + "epoch": 7.92, + "learning_rate": 5.318216898039241e-09, + "loss": 0.0004, + "step": 60457 + }, + { + "epoch": 7.92, + "learning_rate": 5.300937471640621e-09, + "loss": 0.0002, + "step": 60458 + }, + { + "epoch": 7.92, + "learning_rate": 5.283686154800016e-09, + "loss": 0.0003, + "step": 60459 + }, + { + "epoch": 7.92, + "learning_rate": 5.266462947567386e-09, + "loss": 0.0005, + "step": 60460 + }, + { + "epoch": 7.92, + "learning_rate": 5.249267849990469e-09, + "loss": 0.0005, + "step": 60461 + }, + { + "epoch": 7.92, + "learning_rate": 5.2321008621181166e-09, + "loss": 0.001, + "step": 60462 + }, + { + "epoch": 7.92, + "learning_rate": 5.214961983996958e-09, + "loss": 0.0003, + "step": 60463 + }, + { + "epoch": 7.92, + "learning_rate": 5.197851215678062e-09, + "loss": 0.001, + "step": 60464 + }, + { + "epoch": 7.92, + "learning_rate": 5.18076855720695e-09, + "loss": 0.0018, + "step": 60465 + }, + { + "epoch": 7.92, + "learning_rate": 5.16371400863358e-09, + "loss": 0.0006, + "step": 60466 + }, + { + "epoch": 7.92, + "learning_rate": 5.146687570003472e-09, + "loss": 0.0002, + "step": 60467 + }, + { + "epoch": 7.92, + "learning_rate": 5.129689241367697e-09, + "loss": 0.0002, + "step": 60468 + }, + { + "epoch": 7.92, + "learning_rate": 5.1127190227717725e-09, + "loss": 0.0005, + "step": 60469 + }, + { + "epoch": 7.92, + "learning_rate": 5.095776914263439e-09, + "loss": 0.0005, + "step": 60470 + }, + { + "epoch": 7.92, + "learning_rate": 5.078862915891547e-09, + "loss": 0.0003, + "step": 60471 + }, + { + "epoch": 7.92, + "learning_rate": 5.061977027703835e-09, + "loss": 0.0006, + "step": 60472 + }, + { + "epoch": 7.92, + "learning_rate": 5.045119249745822e-09, + "loss": 0.0003, + "step": 60473 + }, + { + "epoch": 7.92, + "learning_rate": 5.028289582066359e-09, + "loss": 0.0007, + "step": 60474 + }, + { + "epoch": 7.92, + "learning_rate": 5.011488024713185e-09, + "loss": 0.0004, + "step": 60475 + }, + { + "epoch": 7.92, + "learning_rate": 4.9947145777329285e-09, + "loss": 0.0002, + "step": 60476 + }, + { + "epoch": 7.92, + "learning_rate": 4.97796924117222e-09, + "loss": 0.0003, + "step": 60477 + }, + { + "epoch": 7.92, + "learning_rate": 4.961252015078799e-09, + "loss": 0.0005, + "step": 60478 + }, + { + "epoch": 7.92, + "learning_rate": 4.9445628994992946e-09, + "loss": 0.0007, + "step": 60479 + }, + { + "epoch": 7.92, + "learning_rate": 4.927901894481446e-09, + "loss": 0.0004, + "step": 60480 + }, + { + "epoch": 7.92, + "learning_rate": 4.9112690000718835e-09, + "loss": 0.0007, + "step": 60481 + }, + { + "epoch": 7.92, + "learning_rate": 4.894664216316125e-09, + "loss": 0.0003, + "step": 60482 + }, + { + "epoch": 7.92, + "learning_rate": 4.878087543263021e-09, + "loss": 0.0004, + "step": 60483 + }, + { + "epoch": 7.92, + "learning_rate": 4.8615389809569815e-09, + "loss": 0.0005, + "step": 60484 + }, + { + "epoch": 7.92, + "learning_rate": 4.8450185294468546e-09, + "loss": 0.0006, + "step": 60485 + }, + { + "epoch": 7.92, + "learning_rate": 4.828526188777049e-09, + "loss": 0.0005, + "step": 60486 + }, + { + "epoch": 7.92, + "learning_rate": 4.8120619589941964e-09, + "loss": 0.0004, + "step": 60487 + }, + { + "epoch": 7.92, + "learning_rate": 4.795625840146034e-09, + "loss": 0.0005, + "step": 60488 + }, + { + "epoch": 7.92, + "learning_rate": 4.779217832276972e-09, + "loss": 0.0005, + "step": 60489 + }, + { + "epoch": 7.92, + "learning_rate": 4.76283793543475e-09, + "loss": 0.0004, + "step": 60490 + }, + { + "epoch": 7.92, + "learning_rate": 4.746486149664886e-09, + "loss": 0.0002, + "step": 60491 + }, + { + "epoch": 7.92, + "learning_rate": 4.730162475012901e-09, + "loss": 0.0004, + "step": 60492 + }, + { + "epoch": 7.92, + "learning_rate": 4.7138669115254224e-09, + "loss": 0.0005, + "step": 60493 + }, + { + "epoch": 7.92, + "learning_rate": 4.69759945924797e-09, + "loss": 0.0001, + "step": 60494 + }, + { + "epoch": 7.92, + "learning_rate": 4.681360118226063e-09, + "loss": 0.0005, + "step": 60495 + }, + { + "epoch": 7.92, + "learning_rate": 4.665148888505222e-09, + "loss": 0.0005, + "step": 60496 + }, + { + "epoch": 7.92, + "learning_rate": 4.6489657701320746e-09, + "loss": 0.0002, + "step": 60497 + }, + { + "epoch": 7.92, + "learning_rate": 4.63281076315103e-09, + "loss": 0.0003, + "step": 60498 + }, + { + "epoch": 7.92, + "learning_rate": 4.6166838676076074e-09, + "loss": 0.0002, + "step": 60499 + }, + { + "epoch": 7.93, + "learning_rate": 4.6005850835473266e-09, + "loss": 0.0003, + "step": 60500 + }, + { + "epoch": 7.93, + "learning_rate": 4.584514411015706e-09, + "loss": 0.0003, + "step": 60501 + }, + { + "epoch": 7.93, + "learning_rate": 4.568471850058265e-09, + "loss": 0.0003, + "step": 60502 + }, + { + "epoch": 7.93, + "learning_rate": 4.5524574007194124e-09, + "loss": 0.0002, + "step": 60503 + }, + { + "epoch": 7.93, + "learning_rate": 4.536471063044668e-09, + "loss": 0.0003, + "step": 60504 + }, + { + "epoch": 7.93, + "learning_rate": 4.5205128370795495e-09, + "loss": 0.0008, + "step": 60505 + }, + { + "epoch": 7.93, + "learning_rate": 4.5045827228673565e-09, + "loss": 0.0004, + "step": 60506 + }, + { + "epoch": 7.93, + "learning_rate": 4.488680720453609e-09, + "loss": 0.0003, + "step": 60507 + }, + { + "epoch": 7.93, + "learning_rate": 4.472806829883825e-09, + "loss": 0.0009, + "step": 60508 + }, + { + "epoch": 7.93, + "learning_rate": 4.456961051201303e-09, + "loss": 0.0004, + "step": 60509 + }, + { + "epoch": 7.93, + "learning_rate": 4.441143384451563e-09, + "loss": 0.0006, + "step": 60510 + }, + { + "epoch": 7.93, + "learning_rate": 4.425353829679013e-09, + "loss": 0.0002, + "step": 60511 + }, + { + "epoch": 7.93, + "learning_rate": 4.409592386926953e-09, + "loss": 0.0003, + "step": 60512 + }, + { + "epoch": 7.93, + "learning_rate": 4.393859056240901e-09, + "loss": 0.0003, + "step": 60513 + }, + { + "epoch": 7.93, + "learning_rate": 4.378153837665267e-09, + "loss": 0.0003, + "step": 60514 + }, + { + "epoch": 7.93, + "learning_rate": 4.362476731244458e-09, + "loss": 0.0006, + "step": 60515 + }, + { + "epoch": 7.93, + "learning_rate": 4.346827737020665e-09, + "loss": 0.0003, + "step": 60516 + }, + { + "epoch": 7.93, + "learning_rate": 4.331206855039405e-09, + "loss": 0.0004, + "step": 60517 + }, + { + "epoch": 7.93, + "learning_rate": 4.315614085345088e-09, + "loss": 0.0006, + "step": 60518 + }, + { + "epoch": 7.93, + "learning_rate": 4.300049427979902e-09, + "loss": 0.0004, + "step": 60519 + }, + { + "epoch": 7.93, + "learning_rate": 4.284512882989367e-09, + "loss": 0.0012, + "step": 60520 + }, + { + "epoch": 7.93, + "learning_rate": 4.2690044504156705e-09, + "loss": 0.0005, + "step": 60521 + }, + { + "epoch": 7.93, + "learning_rate": 4.253524130303222e-09, + "loss": 0.0005, + "step": 60522 + }, + { + "epoch": 7.93, + "learning_rate": 4.238071922695319e-09, + "loss": 0.0005, + "step": 60523 + }, + { + "epoch": 7.93, + "learning_rate": 4.2226478276352624e-09, + "loss": 0.0005, + "step": 60524 + }, + { + "epoch": 7.93, + "learning_rate": 4.2072518451674596e-09, + "loss": 0.0006, + "step": 60525 + }, + { + "epoch": 7.93, + "learning_rate": 4.191883975334099e-09, + "loss": 0.0012, + "step": 60526 + }, + { + "epoch": 7.93, + "learning_rate": 4.1765442181784795e-09, + "loss": 0.0003, + "step": 60527 + }, + { + "epoch": 7.93, + "learning_rate": 4.161232573745011e-09, + "loss": 0.0005, + "step": 60528 + }, + { + "epoch": 7.93, + "learning_rate": 4.145949042074771e-09, + "loss": 0.0004, + "step": 60529 + }, + { + "epoch": 7.93, + "learning_rate": 4.130693623212168e-09, + "loss": 0.0005, + "step": 60530 + }, + { + "epoch": 7.93, + "learning_rate": 4.11546631719939e-09, + "loss": 0.0003, + "step": 60531 + }, + { + "epoch": 7.93, + "learning_rate": 4.1002671240797374e-09, + "loss": 0.0009, + "step": 60532 + }, + { + "epoch": 7.93, + "learning_rate": 4.085096043895398e-09, + "loss": 0.0003, + "step": 60533 + }, + { + "epoch": 7.93, + "learning_rate": 4.06995307668967e-09, + "loss": 0.0002, + "step": 60534 + }, + { + "epoch": 7.93, + "learning_rate": 4.0548382225047426e-09, + "loss": 0.0008, + "step": 60535 + }, + { + "epoch": 7.93, + "learning_rate": 4.0397514813839135e-09, + "loss": 0.0008, + "step": 60536 + }, + { + "epoch": 7.93, + "learning_rate": 4.0246928533682614e-09, + "loss": 0.001, + "step": 60537 + }, + { + "epoch": 7.93, + "learning_rate": 4.009662338501086e-09, + "loss": 0.0004, + "step": 60538 + }, + { + "epoch": 7.93, + "learning_rate": 3.994659936824574e-09, + "loss": 0.0008, + "step": 60539 + }, + { + "epoch": 7.93, + "learning_rate": 3.979685648380915e-09, + "loss": 0.0005, + "step": 60540 + }, + { + "epoch": 7.93, + "learning_rate": 3.9647394732111875e-09, + "loss": 0.0005, + "step": 60541 + }, + { + "epoch": 7.93, + "learning_rate": 3.9498214113597995e-09, + "loss": 0.0006, + "step": 60542 + }, + { + "epoch": 7.93, + "learning_rate": 3.93493146286561e-09, + "loss": 0.0008, + "step": 60543 + }, + { + "epoch": 7.93, + "learning_rate": 3.920069627773026e-09, + "loss": 0.0002, + "step": 60544 + }, + { + "epoch": 7.93, + "learning_rate": 3.905235906123128e-09, + "loss": 0.0002, + "step": 60545 + }, + { + "epoch": 7.93, + "learning_rate": 3.890430297956993e-09, + "loss": 0.0004, + "step": 60546 + }, + { + "epoch": 7.93, + "learning_rate": 3.875652803315699e-09, + "loss": 0.0007, + "step": 60547 + }, + { + "epoch": 7.93, + "learning_rate": 3.8609034222436556e-09, + "loss": 0.0004, + "step": 60548 + }, + { + "epoch": 7.93, + "learning_rate": 3.84618215477972e-09, + "loss": 0.0002, + "step": 60549 + }, + { + "epoch": 7.93, + "learning_rate": 3.8314890009660825e-09, + "loss": 0.0009, + "step": 60550 + }, + { + "epoch": 7.93, + "learning_rate": 3.816823960843819e-09, + "loss": 0.0002, + "step": 60551 + }, + { + "epoch": 7.93, + "learning_rate": 3.802187034454008e-09, + "loss": 0.0005, + "step": 60552 + }, + { + "epoch": 7.93, + "learning_rate": 3.787578221838839e-09, + "loss": 0.0003, + "step": 60553 + }, + { + "epoch": 7.93, + "learning_rate": 3.77299752303939e-09, + "loss": 0.0005, + "step": 60554 + }, + { + "epoch": 7.93, + "learning_rate": 3.7584449380945185e-09, + "loss": 0.0008, + "step": 60555 + }, + { + "epoch": 7.93, + "learning_rate": 3.743920467048634e-09, + "loss": 0.0003, + "step": 60556 + }, + { + "epoch": 7.93, + "learning_rate": 3.7294241099394835e-09, + "loss": 0.0002, + "step": 60557 + }, + { + "epoch": 7.93, + "learning_rate": 3.7149558668092556e-09, + "loss": 0.0004, + "step": 60558 + }, + { + "epoch": 7.93, + "learning_rate": 3.7005157376979183e-09, + "loss": 0.0005, + "step": 60559 + }, + { + "epoch": 7.93, + "learning_rate": 3.6861037226465503e-09, + "loss": 0.0003, + "step": 60560 + }, + { + "epoch": 7.93, + "learning_rate": 3.6717198216962292e-09, + "loss": 0.0002, + "step": 60561 + }, + { + "epoch": 7.93, + "learning_rate": 3.6573640348869233e-09, + "loss": 0.0003, + "step": 60562 + }, + { + "epoch": 7.93, + "learning_rate": 3.6430363622586007e-09, + "loss": 0.001, + "step": 60563 + }, + { + "epoch": 7.93, + "learning_rate": 3.6287368038512295e-09, + "loss": 0.0004, + "step": 60564 + }, + { + "epoch": 7.93, + "learning_rate": 3.614465359706998e-09, + "loss": 0.0009, + "step": 60565 + }, + { + "epoch": 7.93, + "learning_rate": 3.6002220298636537e-09, + "loss": 0.0002, + "step": 60566 + }, + { + "epoch": 7.93, + "learning_rate": 3.5860068143633853e-09, + "loss": 0.0004, + "step": 60567 + }, + { + "epoch": 7.93, + "learning_rate": 3.5718197132439405e-09, + "loss": 0.0006, + "step": 60568 + }, + { + "epoch": 7.93, + "learning_rate": 3.557660726547507e-09, + "loss": 0.0003, + "step": 60569 + }, + { + "epoch": 7.93, + "learning_rate": 3.5435298543118336e-09, + "loss": 0.0003, + "step": 60570 + }, + { + "epoch": 7.93, + "learning_rate": 3.5294270965779974e-09, + "loss": 0.0004, + "step": 60571 + }, + { + "epoch": 7.93, + "learning_rate": 3.5153524533859673e-09, + "loss": 0.0002, + "step": 60572 + }, + { + "epoch": 7.93, + "learning_rate": 3.5013059247746004e-09, + "loss": 0.0004, + "step": 60573 + }, + { + "epoch": 7.93, + "learning_rate": 3.4872875107827554e-09, + "loss": 0.0006, + "step": 60574 + }, + { + "epoch": 7.93, + "learning_rate": 3.4732972114503993e-09, + "loss": 0.0003, + "step": 60575 + }, + { + "epoch": 7.94, + "learning_rate": 3.459335026817501e-09, + "loss": 0.0002, + "step": 60576 + }, + { + "epoch": 7.94, + "learning_rate": 3.4454009569229173e-09, + "loss": 0.0003, + "step": 60577 + }, + { + "epoch": 7.94, + "learning_rate": 3.431495001805507e-09, + "loss": 0.0002, + "step": 60578 + }, + { + "epoch": 7.94, + "learning_rate": 3.4176171615052377e-09, + "loss": 0.0004, + "step": 60579 + }, + { + "epoch": 7.94, + "learning_rate": 3.403767436059857e-09, + "loss": 0.0005, + "step": 60580 + }, + { + "epoch": 7.94, + "learning_rate": 3.3899458255104434e-09, + "loss": 0.0005, + "step": 60581 + }, + { + "epoch": 7.94, + "learning_rate": 3.3761523298925236e-09, + "loss": 0.0002, + "step": 60582 + }, + { + "epoch": 7.94, + "learning_rate": 3.3623869492482862e-09, + "loss": 0.0004, + "step": 60583 + }, + { + "epoch": 7.94, + "learning_rate": 3.348649683614369e-09, + "loss": 0.0002, + "step": 60584 + }, + { + "epoch": 7.94, + "learning_rate": 3.334940533029629e-09, + "loss": 0.0003, + "step": 60585 + }, + { + "epoch": 7.94, + "learning_rate": 3.3212594975340352e-09, + "loss": 0.0002, + "step": 60586 + }, + { + "epoch": 7.94, + "learning_rate": 3.3076065771642243e-09, + "loss": 0.0004, + "step": 60587 + }, + { + "epoch": 7.94, + "learning_rate": 3.293981771959054e-09, + "loss": 0.0005, + "step": 60588 + }, + { + "epoch": 7.94, + "learning_rate": 3.2803850819584925e-09, + "loss": 0.0002, + "step": 60589 + }, + { + "epoch": 7.94, + "learning_rate": 3.266816507198067e-09, + "loss": 0.0006, + "step": 60590 + }, + { + "epoch": 7.94, + "learning_rate": 3.2532760477188565e-09, + "loss": 0.0003, + "step": 60591 + }, + { + "epoch": 7.94, + "learning_rate": 3.239763703556387e-09, + "loss": 0.0004, + "step": 60592 + }, + { + "epoch": 7.94, + "learning_rate": 3.2262794747506265e-09, + "loss": 0.0005, + "step": 60593 + }, + { + "epoch": 7.94, + "learning_rate": 3.212823361337103e-09, + "loss": 0.0007, + "step": 60594 + }, + { + "epoch": 7.94, + "learning_rate": 3.199395363356894e-09, + "loss": 0.0006, + "step": 60595 + }, + { + "epoch": 7.94, + "learning_rate": 3.185995480844417e-09, + "loss": 0.0007, + "step": 60596 + }, + { + "epoch": 7.94, + "learning_rate": 3.17262371383964e-09, + "loss": 0.0014, + "step": 60597 + }, + { + "epoch": 7.94, + "learning_rate": 3.15928006238031e-09, + "loss": 0.0003, + "step": 60598 + }, + { + "epoch": 7.94, + "learning_rate": 3.1459645265019544e-09, + "loss": 0.0003, + "step": 60599 + }, + { + "epoch": 7.94, + "learning_rate": 3.1326771062434313e-09, + "loss": 0.0008, + "step": 60600 + }, + { + "epoch": 7.94, + "learning_rate": 3.119417801642488e-09, + "loss": 0.0004, + "step": 60601 + }, + { + "epoch": 7.94, + "learning_rate": 3.106186612735762e-09, + "loss": 0.001, + "step": 60602 + }, + { + "epoch": 7.94, + "learning_rate": 3.0929835395598905e-09, + "loss": 0.0005, + "step": 60603 + }, + { + "epoch": 7.94, + "learning_rate": 3.0798085821526216e-09, + "loss": 0.0011, + "step": 60604 + }, + { + "epoch": 7.94, + "learning_rate": 3.066661740551702e-09, + "loss": 0.0006, + "step": 60605 + }, + { + "epoch": 7.94, + "learning_rate": 3.0535430147937695e-09, + "loss": 0.0005, + "step": 60606 + }, + { + "epoch": 7.94, + "learning_rate": 3.0404524049143513e-09, + "loss": 0.0003, + "step": 60607 + }, + { + "epoch": 7.94, + "learning_rate": 3.027389910951195e-09, + "loss": 0.0002, + "step": 60608 + }, + { + "epoch": 7.94, + "learning_rate": 3.0143555329420482e-09, + "loss": 0.0016, + "step": 60609 + }, + { + "epoch": 7.94, + "learning_rate": 3.001349270922438e-09, + "loss": 0.0005, + "step": 60610 + }, + { + "epoch": 7.94, + "learning_rate": 2.9883711249290015e-09, + "loss": 0.0008, + "step": 60611 + }, + { + "epoch": 7.94, + "learning_rate": 2.9754210949972662e-09, + "loss": 0.0004, + "step": 60612 + }, + { + "epoch": 7.94, + "learning_rate": 2.96249918116609e-09, + "loss": 0.0006, + "step": 60613 + }, + { + "epoch": 7.94, + "learning_rate": 2.9496053834709994e-09, + "loss": 0.0002, + "step": 60614 + }, + { + "epoch": 7.94, + "learning_rate": 2.9367397019464117e-09, + "loss": 0.0002, + "step": 60615 + }, + { + "epoch": 7.94, + "learning_rate": 2.9239021366311848e-09, + "loss": 0.0005, + "step": 60616 + }, + { + "epoch": 7.94, + "learning_rate": 2.9110926875586256e-09, + "loss": 0.0011, + "step": 60617 + }, + { + "epoch": 7.94, + "learning_rate": 2.8983113547675912e-09, + "loss": 0.0003, + "step": 60618 + }, + { + "epoch": 7.94, + "learning_rate": 2.885558138291389e-09, + "loss": 0.0002, + "step": 60619 + }, + { + "epoch": 7.94, + "learning_rate": 2.872833038167766e-09, + "loss": 0.0004, + "step": 60620 + }, + { + "epoch": 7.94, + "learning_rate": 2.86013605443225e-09, + "loss": 0.0003, + "step": 60621 + }, + { + "epoch": 7.94, + "learning_rate": 2.8474671871203676e-09, + "loss": 0.0005, + "step": 60622 + }, + { + "epoch": 7.94, + "learning_rate": 2.8348264362676458e-09, + "loss": 0.0004, + "step": 60623 + }, + { + "epoch": 7.94, + "learning_rate": 2.822213801908502e-09, + "loss": 0.0004, + "step": 60624 + }, + { + "epoch": 7.94, + "learning_rate": 2.809629284080684e-09, + "loss": 0.0004, + "step": 60625 + }, + { + "epoch": 7.94, + "learning_rate": 2.7970728828174977e-09, + "loss": 0.002, + "step": 60626 + }, + { + "epoch": 7.94, + "learning_rate": 2.7845445981555806e-09, + "loss": 0.0002, + "step": 60627 + }, + { + "epoch": 7.94, + "learning_rate": 2.7720444301304604e-09, + "loss": 0.0007, + "step": 60628 + }, + { + "epoch": 7.94, + "learning_rate": 2.7595723787754435e-09, + "loss": 0.0003, + "step": 60629 + }, + { + "epoch": 7.94, + "learning_rate": 2.747128444127167e-09, + "loss": 0.0003, + "step": 60630 + }, + { + "epoch": 7.94, + "learning_rate": 2.7347126262200484e-09, + "loss": 0.0003, + "step": 60631 + }, + { + "epoch": 7.94, + "learning_rate": 2.7223249250896144e-09, + "loss": 0.0002, + "step": 60632 + }, + { + "epoch": 7.94, + "learning_rate": 2.709965340770282e-09, + "loss": 0.0009, + "step": 60633 + }, + { + "epoch": 7.94, + "learning_rate": 2.697633873296468e-09, + "loss": 0.0004, + "step": 60634 + }, + { + "epoch": 7.94, + "learning_rate": 2.6853305227025894e-09, + "loss": 0.0004, + "step": 60635 + }, + { + "epoch": 7.94, + "learning_rate": 2.6730552890241736e-09, + "loss": 0.0002, + "step": 60636 + }, + { + "epoch": 7.94, + "learning_rate": 2.660808172296747e-09, + "loss": 0.0007, + "step": 60637 + }, + { + "epoch": 7.94, + "learning_rate": 2.648589172551397e-09, + "loss": 0.0003, + "step": 60638 + }, + { + "epoch": 7.94, + "learning_rate": 2.63639828982587e-09, + "loss": 0.0003, + "step": 60639 + }, + { + "epoch": 7.94, + "learning_rate": 2.624235524152363e-09, + "loss": 0.0004, + "step": 60640 + }, + { + "epoch": 7.94, + "learning_rate": 2.6121008755664036e-09, + "loss": 0.0005, + "step": 60641 + }, + { + "epoch": 7.94, + "learning_rate": 2.5999943441012977e-09, + "loss": 0.0002, + "step": 60642 + }, + { + "epoch": 7.94, + "learning_rate": 2.587915929790352e-09, + "loss": 0.0004, + "step": 60643 + }, + { + "epoch": 7.94, + "learning_rate": 2.5758656326702047e-09, + "loss": 0.0003, + "step": 60644 + }, + { + "epoch": 7.94, + "learning_rate": 2.5638434527719413e-09, + "loss": 0.0007, + "step": 60645 + }, + { + "epoch": 7.94, + "learning_rate": 2.551849390131089e-09, + "loss": 0.0006, + "step": 60646 + }, + { + "epoch": 7.94, + "learning_rate": 2.5398834447798447e-09, + "loss": 0.0001, + "step": 60647 + }, + { + "epoch": 7.94, + "learning_rate": 2.527945616753735e-09, + "loss": 0.0005, + "step": 60648 + }, + { + "epoch": 7.94, + "learning_rate": 2.516035906084957e-09, + "loss": 0.0006, + "step": 60649 + }, + { + "epoch": 7.94, + "learning_rate": 2.504154312806817e-09, + "loss": 0.0005, + "step": 60650 + }, + { + "epoch": 7.94, + "learning_rate": 2.492300836954842e-09, + "loss": 0.0009, + "step": 60651 + }, + { + "epoch": 7.94, + "learning_rate": 2.4804754785590078e-09, + "loss": 0.0006, + "step": 60652 + }, + { + "epoch": 7.95, + "learning_rate": 2.4686782376559528e-09, + "loss": 0.0004, + "step": 60653 + }, + { + "epoch": 7.95, + "learning_rate": 2.456909114276762e-09, + "loss": 0.0002, + "step": 60654 + }, + { + "epoch": 7.95, + "learning_rate": 2.445168108454743e-09, + "loss": 0.0002, + "step": 60655 + }, + { + "epoch": 7.95, + "learning_rate": 2.433455220224312e-09, + "loss": 0.0006, + "step": 60656 + }, + { + "epoch": 7.95, + "learning_rate": 2.421770449616556e-09, + "loss": 0.0002, + "step": 60657 + }, + { + "epoch": 7.95, + "learning_rate": 2.410113796665892e-09, + "loss": 0.001, + "step": 60658 + }, + { + "epoch": 7.95, + "learning_rate": 2.398485261403405e-09, + "loss": 0.0016, + "step": 60659 + }, + { + "epoch": 7.95, + "learning_rate": 2.386884843863513e-09, + "loss": 0.0001, + "step": 60660 + }, + { + "epoch": 7.95, + "learning_rate": 2.3753125440773017e-09, + "loss": 0.0002, + "step": 60661 + }, + { + "epoch": 7.95, + "learning_rate": 2.3637683620791885e-09, + "loss": 0.0004, + "step": 60662 + }, + { + "epoch": 7.95, + "learning_rate": 2.3522522979002594e-09, + "loss": 0.0007, + "step": 60663 + }, + { + "epoch": 7.95, + "learning_rate": 2.340764351572711e-09, + "loss": 0.0013, + "step": 60664 + }, + { + "epoch": 7.95, + "learning_rate": 2.3293045231298493e-09, + "loss": 0.0003, + "step": 60665 + }, + { + "epoch": 7.95, + "learning_rate": 2.3178728126027617e-09, + "loss": 0.0009, + "step": 60666 + }, + { + "epoch": 7.95, + "learning_rate": 2.3064692200247542e-09, + "loss": 0.001, + "step": 60667 + }, + { + "epoch": 7.95, + "learning_rate": 2.295093745428023e-09, + "loss": 0.0003, + "step": 60668 + }, + { + "epoch": 7.95, + "learning_rate": 2.2837463888425447e-09, + "loss": 0.001, + "step": 60669 + }, + { + "epoch": 7.95, + "learning_rate": 2.2724271503027362e-09, + "loss": 0.0003, + "step": 60670 + }, + { + "epoch": 7.95, + "learning_rate": 2.261136029838573e-09, + "loss": 0.0002, + "step": 60671 + }, + { + "epoch": 7.95, + "learning_rate": 2.2498730274833624e-09, + "loss": 0.0009, + "step": 60672 + }, + { + "epoch": 7.95, + "learning_rate": 2.23863814326708e-09, + "loss": 0.0011, + "step": 60673 + }, + { + "epoch": 7.95, + "learning_rate": 2.2274313772230326e-09, + "loss": 0.0005, + "step": 60674 + }, + { + "epoch": 7.95, + "learning_rate": 2.2162527293811966e-09, + "loss": 0.0012, + "step": 60675 + }, + { + "epoch": 7.95, + "learning_rate": 2.205102199773768e-09, + "loss": 0.0007, + "step": 60676 + }, + { + "epoch": 7.95, + "learning_rate": 2.1939797884329427e-09, + "loss": 0.0007, + "step": 60677 + }, + { + "epoch": 7.95, + "learning_rate": 2.182885495388698e-09, + "loss": 0.0002, + "step": 60678 + }, + { + "epoch": 7.95, + "learning_rate": 2.1718193206721193e-09, + "loss": 0.0003, + "step": 60679 + }, + { + "epoch": 7.95, + "learning_rate": 2.1607812643154036e-09, + "loss": 0.0003, + "step": 60680 + }, + { + "epoch": 7.95, + "learning_rate": 2.1497713263485266e-09, + "loss": 0.0002, + "step": 60681 + }, + { + "epoch": 7.95, + "learning_rate": 2.1387895068036847e-09, + "loss": 0.0003, + "step": 60682 + }, + { + "epoch": 7.95, + "learning_rate": 2.127835805710854e-09, + "loss": 0.0006, + "step": 60683 + }, + { + "epoch": 7.95, + "learning_rate": 2.116910223101121e-09, + "loss": 0.0008, + "step": 60684 + }, + { + "epoch": 7.95, + "learning_rate": 2.106012759005571e-09, + "loss": 0.0007, + "step": 60685 + }, + { + "epoch": 7.95, + "learning_rate": 2.0951434134530712e-09, + "loss": 0.0002, + "step": 60686 + }, + { + "epoch": 7.95, + "learning_rate": 2.0843021864769273e-09, + "loss": 0.0001, + "step": 60687 + }, + { + "epoch": 7.95, + "learning_rate": 2.073489078106006e-09, + "loss": 0.0008, + "step": 60688 + }, + { + "epoch": 7.95, + "learning_rate": 2.062704088370282e-09, + "loss": 0.0006, + "step": 60689 + }, + { + "epoch": 7.95, + "learning_rate": 2.051947217301953e-09, + "loss": 0.0002, + "step": 60690 + }, + { + "epoch": 7.95, + "learning_rate": 2.0412184649287737e-09, + "loss": 0.0007, + "step": 60691 + }, + { + "epoch": 7.95, + "learning_rate": 2.030517831282941e-09, + "loss": 0.0004, + "step": 60692 + }, + { + "epoch": 7.95, + "learning_rate": 2.019845316394431e-09, + "loss": 0.0002, + "step": 60693 + }, + { + "epoch": 7.95, + "learning_rate": 2.0092009202921093e-09, + "loss": 0.0007, + "step": 60694 + }, + { + "epoch": 7.95, + "learning_rate": 1.998584643005952e-09, + "loss": 0.0004, + "step": 60695 + }, + { + "epoch": 7.95, + "learning_rate": 1.987996484567045e-09, + "loss": 0.0005, + "step": 60696 + }, + { + "epoch": 7.95, + "learning_rate": 1.977436445004255e-09, + "loss": 0.0009, + "step": 60697 + }, + { + "epoch": 7.95, + "learning_rate": 1.9669045243475573e-09, + "loss": 0.0005, + "step": 60698 + }, + { + "epoch": 7.95, + "learning_rate": 1.9564007226258173e-09, + "loss": 0.0007, + "step": 60699 + }, + { + "epoch": 7.95, + "learning_rate": 1.9459250398701223e-09, + "loss": 0.0002, + "step": 60700 + }, + { + "epoch": 7.95, + "learning_rate": 1.9354774761093378e-09, + "loss": 0.0001, + "step": 60701 + }, + { + "epoch": 7.95, + "learning_rate": 1.9250580313712186e-09, + "loss": 0.0004, + "step": 60702 + }, + { + "epoch": 7.95, + "learning_rate": 1.914666705687962e-09, + "loss": 0.0004, + "step": 60703 + }, + { + "epoch": 7.95, + "learning_rate": 1.904303499086213e-09, + "loss": 0.0002, + "step": 60704 + }, + { + "epoch": 7.95, + "learning_rate": 1.893968411597058e-09, + "loss": 0.0002, + "step": 60705 + }, + { + "epoch": 7.95, + "learning_rate": 1.8836614432482525e-09, + "loss": 0.0008, + "step": 60706 + }, + { + "epoch": 7.95, + "learning_rate": 1.873382594068662e-09, + "loss": 0.0007, + "step": 60707 + }, + { + "epoch": 7.95, + "learning_rate": 1.8631318640882632e-09, + "loss": 0.0004, + "step": 60708 + }, + { + "epoch": 7.95, + "learning_rate": 1.8529092533359217e-09, + "loss": 0.0006, + "step": 60709 + }, + { + "epoch": 7.95, + "learning_rate": 1.8427147618393925e-09, + "loss": 0.0009, + "step": 60710 + }, + { + "epoch": 7.95, + "learning_rate": 1.832548389627542e-09, + "loss": 0.0003, + "step": 60711 + }, + { + "epoch": 7.95, + "learning_rate": 1.8224101367292358e-09, + "loss": 0.0003, + "step": 60712 + }, + { + "epoch": 7.95, + "learning_rate": 1.8123000031733395e-09, + "loss": 0.0004, + "step": 60713 + }, + { + "epoch": 7.95, + "learning_rate": 1.802217988987609e-09, + "loss": 0.0002, + "step": 60714 + }, + { + "epoch": 7.95, + "learning_rate": 1.79216409420091e-09, + "loss": 0.0007, + "step": 60715 + }, + { + "epoch": 7.95, + "learning_rate": 1.782138318840998e-09, + "loss": 0.0003, + "step": 60716 + }, + { + "epoch": 7.95, + "learning_rate": 1.7721406629367388e-09, + "loss": 0.0004, + "step": 60717 + }, + { + "epoch": 7.95, + "learning_rate": 1.7621711265158882e-09, + "loss": 0.0002, + "step": 60718 + }, + { + "epoch": 7.95, + "learning_rate": 1.7522297096062012e-09, + "loss": 0.0005, + "step": 60719 + }, + { + "epoch": 7.95, + "learning_rate": 1.7423164122365444e-09, + "loss": 0.0003, + "step": 60720 + }, + { + "epoch": 7.95, + "learning_rate": 1.7324312344335624e-09, + "loss": 0.0002, + "step": 60721 + }, + { + "epoch": 7.95, + "learning_rate": 1.7225741762261216e-09, + "loss": 0.0007, + "step": 60722 + }, + { + "epoch": 7.95, + "learning_rate": 1.712745237640867e-09, + "loss": 0.0007, + "step": 60723 + }, + { + "epoch": 7.95, + "learning_rate": 1.7029444187066646e-09, + "loss": 0.0005, + "step": 60724 + }, + { + "epoch": 7.95, + "learning_rate": 1.6931717194501595e-09, + "loss": 0.0003, + "step": 60725 + }, + { + "epoch": 7.95, + "learning_rate": 1.6834271398991075e-09, + "loss": 0.0002, + "step": 60726 + }, + { + "epoch": 7.95, + "learning_rate": 1.6737106800812642e-09, + "loss": 0.0008, + "step": 60727 + }, + { + "epoch": 7.95, + "learning_rate": 1.6640223400232747e-09, + "loss": 0.0007, + "step": 60728 + }, + { + "epoch": 7.96, + "learning_rate": 1.6543621197528947e-09, + "loss": 0.0003, + "step": 60729 + }, + { + "epoch": 7.96, + "learning_rate": 1.6447300192978798e-09, + "loss": 0.0006, + "step": 60730 + }, + { + "epoch": 7.96, + "learning_rate": 1.6351260386837652e-09, + "loss": 0.0004, + "step": 60731 + }, + { + "epoch": 7.96, + "learning_rate": 1.6255501779394168e-09, + "loss": 0.0003, + "step": 60732 + }, + { + "epoch": 7.96, + "learning_rate": 1.616002437089259e-09, + "loss": 0.0006, + "step": 60733 + }, + { + "epoch": 7.96, + "learning_rate": 1.6064828161632685e-09, + "loss": 0.0004, + "step": 60734 + }, + { + "epoch": 7.96, + "learning_rate": 1.5969913151858696e-09, + "loss": 0.0003, + "step": 60735 + }, + { + "epoch": 7.96, + "learning_rate": 1.587527934183708e-09, + "loss": 0.0001, + "step": 60736 + }, + { + "epoch": 7.96, + "learning_rate": 1.5780926731856493e-09, + "loss": 0.0003, + "step": 60737 + }, + { + "epoch": 7.96, + "learning_rate": 1.5686855322150086e-09, + "loss": 0.0006, + "step": 60738 + }, + { + "epoch": 7.96, + "learning_rate": 1.5593065113017615e-09, + "loss": 0.0008, + "step": 60739 + }, + { + "epoch": 7.96, + "learning_rate": 1.5499556104692225e-09, + "loss": 0.0003, + "step": 60740 + }, + { + "epoch": 7.96, + "learning_rate": 1.5406328297451478e-09, + "loss": 0.0002, + "step": 60741 + }, + { + "epoch": 7.96, + "learning_rate": 1.5313381691561824e-09, + "loss": 0.0006, + "step": 60742 + }, + { + "epoch": 7.96, + "learning_rate": 1.5220716287278615e-09, + "loss": 0.0003, + "step": 60743 + }, + { + "epoch": 7.96, + "learning_rate": 1.51283320848572e-09, + "loss": 0.0002, + "step": 60744 + }, + { + "epoch": 7.96, + "learning_rate": 1.5036229084575139e-09, + "loss": 0.0003, + "step": 60745 + }, + { + "epoch": 7.96, + "learning_rate": 1.4944407286665574e-09, + "loss": 0.0004, + "step": 60746 + }, + { + "epoch": 7.96, + "learning_rate": 1.4852866691406065e-09, + "loss": 0.0002, + "step": 60747 + }, + { + "epoch": 7.96, + "learning_rate": 1.4761607299040859e-09, + "loss": 0.0003, + "step": 60748 + }, + { + "epoch": 7.96, + "learning_rate": 1.4670629109847511e-09, + "loss": 0.0009, + "step": 60749 + }, + { + "epoch": 7.96, + "learning_rate": 1.4579932124059171e-09, + "loss": 0.0005, + "step": 60750 + }, + { + "epoch": 7.96, + "learning_rate": 1.4489516341942289e-09, + "loss": 0.0008, + "step": 60751 + }, + { + "epoch": 7.96, + "learning_rate": 1.4399381763752218e-09, + "loss": 0.0004, + "step": 60752 + }, + { + "epoch": 7.96, + "learning_rate": 1.4309528389733207e-09, + "loss": 0.0006, + "step": 60753 + }, + { + "epoch": 7.96, + "learning_rate": 1.4219956220140607e-09, + "loss": 0.0004, + "step": 60754 + }, + { + "epoch": 7.96, + "learning_rate": 1.4130665255240873e-09, + "loss": 0.0004, + "step": 60755 + }, + { + "epoch": 7.96, + "learning_rate": 1.4041655495267147e-09, + "loss": 0.0008, + "step": 60756 + }, + { + "epoch": 7.96, + "learning_rate": 1.395292694048589e-09, + "loss": 0.0004, + "step": 60757 + }, + { + "epoch": 7.96, + "learning_rate": 1.3864479591130243e-09, + "loss": 0.0003, + "step": 60758 + }, + { + "epoch": 7.96, + "learning_rate": 1.3776313447455558e-09, + "loss": 0.0003, + "step": 60759 + }, + { + "epoch": 7.96, + "learning_rate": 1.3688428509706086e-09, + "loss": 0.0005, + "step": 60760 + }, + { + "epoch": 7.96, + "learning_rate": 1.360082477814828e-09, + "loss": 0.0006, + "step": 60761 + }, + { + "epoch": 7.96, + "learning_rate": 1.3513502252993082e-09, + "loss": 0.0009, + "step": 60762 + }, + { + "epoch": 7.96, + "learning_rate": 1.342646093451805e-09, + "loss": 0.0006, + "step": 60763 + }, + { + "epoch": 7.96, + "learning_rate": 1.3339700822956324e-09, + "loss": 0.0003, + "step": 60764 + }, + { + "epoch": 7.96, + "learning_rate": 1.3253221918552162e-09, + "loss": 0.0006, + "step": 60765 + }, + { + "epoch": 7.96, + "learning_rate": 1.3167024221549807e-09, + "loss": 0.0002, + "step": 60766 + }, + { + "epoch": 7.96, + "learning_rate": 1.308110773219351e-09, + "loss": 0.0006, + "step": 60767 + }, + { + "epoch": 7.96, + "learning_rate": 1.2995472450716419e-09, + "loss": 0.0004, + "step": 60768 + }, + { + "epoch": 7.96, + "learning_rate": 1.291011837736278e-09, + "loss": 0.0003, + "step": 60769 + }, + { + "epoch": 7.96, + "learning_rate": 1.2825045512376844e-09, + "loss": 0.0009, + "step": 60770 + }, + { + "epoch": 7.96, + "learning_rate": 1.274025385600286e-09, + "loss": 0.0003, + "step": 60771 + }, + { + "epoch": 7.96, + "learning_rate": 1.2655743408462873e-09, + "loss": 0.0009, + "step": 60772 + }, + { + "epoch": 7.96, + "learning_rate": 1.2571514170012234e-09, + "loss": 0.0002, + "step": 60773 + }, + { + "epoch": 7.96, + "learning_rate": 1.2487566140872987e-09, + "loss": 0.001, + "step": 60774 + }, + { + "epoch": 7.96, + "learning_rate": 1.240389932128938e-09, + "loss": 0.0014, + "step": 60775 + }, + { + "epoch": 7.96, + "learning_rate": 1.2320513711494564e-09, + "loss": 0.0003, + "step": 60776 + }, + { + "epoch": 7.96, + "learning_rate": 1.2237409311721682e-09, + "loss": 0.0004, + "step": 60777 + }, + { + "epoch": 7.96, + "learning_rate": 1.2154586122214985e-09, + "loss": 0.0006, + "step": 60778 + }, + { + "epoch": 7.96, + "learning_rate": 1.2072044143196514e-09, + "loss": 0.0002, + "step": 60779 + }, + { + "epoch": 7.96, + "learning_rate": 1.198978337488832e-09, + "loss": 0.0003, + "step": 60780 + }, + { + "epoch": 7.96, + "learning_rate": 1.190780381754575e-09, + "loss": 0.0005, + "step": 60781 + }, + { + "epoch": 7.96, + "learning_rate": 1.1826105471390847e-09, + "loss": 0.0007, + "step": 60782 + }, + { + "epoch": 7.96, + "learning_rate": 1.174468833664566e-09, + "loss": 0.0002, + "step": 60783 + }, + { + "epoch": 7.96, + "learning_rate": 1.166355241353223e-09, + "loss": 0.0002, + "step": 60784 + }, + { + "epoch": 7.96, + "learning_rate": 1.1582697702305913e-09, + "loss": 0.0004, + "step": 60785 + }, + { + "epoch": 7.96, + "learning_rate": 1.1502124203177645e-09, + "loss": 0.0003, + "step": 60786 + }, + { + "epoch": 7.96, + "learning_rate": 1.1421831916369474e-09, + "loss": 0.001, + "step": 60787 + }, + { + "epoch": 7.96, + "learning_rate": 1.134182084211455e-09, + "loss": 0.0003, + "step": 60788 + }, + { + "epoch": 7.96, + "learning_rate": 1.126209098063491e-09, + "loss": 0.0002, + "step": 60789 + }, + { + "epoch": 7.96, + "learning_rate": 1.1182642332152604e-09, + "loss": 0.0011, + "step": 60790 + }, + { + "epoch": 7.96, + "learning_rate": 1.110347489690078e-09, + "loss": 0.0002, + "step": 60791 + }, + { + "epoch": 7.96, + "learning_rate": 1.1024588675090375e-09, + "loss": 0.0006, + "step": 60792 + }, + { + "epoch": 7.96, + "learning_rate": 1.094598366695454e-09, + "loss": 0.0004, + "step": 60793 + }, + { + "epoch": 7.96, + "learning_rate": 1.0867659872704217e-09, + "loss": 0.0003, + "step": 60794 + }, + { + "epoch": 7.96, + "learning_rate": 1.0789617292572551e-09, + "loss": 0.0004, + "step": 60795 + }, + { + "epoch": 7.96, + "learning_rate": 1.0711855926759385e-09, + "loss": 0.0003, + "step": 60796 + }, + { + "epoch": 7.96, + "learning_rate": 1.0634375775497864e-09, + "loss": 0.0007, + "step": 60797 + }, + { + "epoch": 7.96, + "learning_rate": 1.0557176838998929e-09, + "loss": 0.0003, + "step": 60798 + }, + { + "epoch": 7.96, + "learning_rate": 1.0480259117495727e-09, + "loss": 0.0006, + "step": 60799 + }, + { + "epoch": 7.96, + "learning_rate": 1.0403622611177e-09, + "loss": 0.0002, + "step": 60800 + }, + { + "epoch": 7.96, + "learning_rate": 1.0327267320275892e-09, + "loss": 0.001, + "step": 60801 + }, + { + "epoch": 7.96, + "learning_rate": 1.0251193245014447e-09, + "loss": 0.0004, + "step": 60802 + }, + { + "epoch": 7.96, + "learning_rate": 1.0175400385592505e-09, + "loss": 0.0018, + "step": 60803 + }, + { + "epoch": 7.96, + "learning_rate": 1.009988874222101e-09, + "loss": 0.0005, + "step": 60804 + }, + { + "epoch": 7.97, + "learning_rate": 1.0024658315122005e-09, + "loss": 0.0004, + "step": 60805 + }, + { + "epoch": 7.97, + "learning_rate": 9.949709104506433e-10, + "loss": 0.0005, + "step": 60806 + }, + { + "epoch": 7.97, + "learning_rate": 9.875041110585237e-10, + "loss": 0.0003, + "step": 60807 + }, + { + "epoch": 7.97, + "learning_rate": 9.800654333558256e-10, + "loss": 0.0005, + "step": 60808 + }, + { + "epoch": 7.97, + "learning_rate": 9.726548773647537e-10, + "loss": 0.001, + "step": 60809 + }, + { + "epoch": 7.97, + "learning_rate": 9.652724431052917e-10, + "loss": 0.0005, + "step": 60810 + }, + { + "epoch": 7.97, + "learning_rate": 9.579181305985341e-10, + "loss": 0.0005, + "step": 60811 + }, + { + "epoch": 7.97, + "learning_rate": 9.50591939865575e-10, + "loss": 0.0002, + "step": 60812 + }, + { + "epoch": 7.97, + "learning_rate": 9.432938709263983e-10, + "loss": 0.0002, + "step": 60813 + }, + { + "epoch": 7.97, + "learning_rate": 9.360239238009883e-10, + "loss": 0.0006, + "step": 60814 + }, + { + "epoch": 7.97, + "learning_rate": 9.287820985115492e-10, + "loss": 0.0005, + "step": 60815 + }, + { + "epoch": 7.97, + "learning_rate": 9.215683950769549e-10, + "loss": 0.0002, + "step": 60816 + }, + { + "epoch": 7.97, + "learning_rate": 9.143828135182997e-10, + "loss": 0.0005, + "step": 60817 + }, + { + "epoch": 7.97, + "learning_rate": 9.072253538555676e-10, + "loss": 0.0006, + "step": 60818 + }, + { + "epoch": 7.97, + "learning_rate": 9.000960161076322e-10, + "loss": 0.0002, + "step": 60819 + }, + { + "epoch": 7.97, + "learning_rate": 8.929948002966981e-10, + "loss": 0.0004, + "step": 60820 + }, + { + "epoch": 7.97, + "learning_rate": 8.859217064416392e-10, + "loss": 0.0002, + "step": 60821 + }, + { + "epoch": 7.97, + "learning_rate": 8.788767345624393e-10, + "loss": 0.0009, + "step": 60822 + }, + { + "epoch": 7.97, + "learning_rate": 8.718598846790827e-10, + "loss": 0.0004, + "step": 60823 + }, + { + "epoch": 7.97, + "learning_rate": 8.648711568104429e-10, + "loss": 0.0002, + "step": 60824 + }, + { + "epoch": 7.97, + "learning_rate": 8.579105509776142e-10, + "loss": 0.0002, + "step": 60825 + }, + { + "epoch": 7.97, + "learning_rate": 8.509780671983603e-10, + "loss": 0.0006, + "step": 60826 + }, + { + "epoch": 7.97, + "learning_rate": 8.440737054937754e-10, + "loss": 0.0003, + "step": 60827 + }, + { + "epoch": 7.97, + "learning_rate": 8.371974658838434e-10, + "loss": 0.0003, + "step": 60828 + }, + { + "epoch": 7.97, + "learning_rate": 8.303493483852176e-10, + "loss": 0.0013, + "step": 60829 + }, + { + "epoch": 7.97, + "learning_rate": 8.235293530201027e-10, + "loss": 0.001, + "step": 60830 + }, + { + "epoch": 7.97, + "learning_rate": 8.16737479805152e-10, + "loss": 0.0005, + "step": 60831 + }, + { + "epoch": 7.97, + "learning_rate": 8.099737287614595e-10, + "loss": 0.0002, + "step": 60832 + }, + { + "epoch": 7.97, + "learning_rate": 8.032380999067891e-10, + "loss": 0.0005, + "step": 60833 + }, + { + "epoch": 7.97, + "learning_rate": 7.965305932600142e-10, + "loss": 0.0003, + "step": 60834 + }, + { + "epoch": 7.97, + "learning_rate": 7.898512088411192e-10, + "loss": 0.0006, + "step": 60835 + }, + { + "epoch": 7.97, + "learning_rate": 7.831999466689777e-10, + "loss": 0.0006, + "step": 60836 + }, + { + "epoch": 7.97, + "learning_rate": 7.765768067602431e-10, + "loss": 0.0004, + "step": 60837 + }, + { + "epoch": 7.97, + "learning_rate": 7.699817891360095e-10, + "loss": 0.001, + "step": 60838 + }, + { + "epoch": 7.97, + "learning_rate": 7.634148938129304e-10, + "loss": 0.0009, + "step": 60839 + }, + { + "epoch": 7.97, + "learning_rate": 7.568761208109899e-10, + "loss": 0.0004, + "step": 60840 + }, + { + "epoch": 7.97, + "learning_rate": 7.503654701479512e-10, + "loss": 0.0005, + "step": 60841 + }, + { + "epoch": 7.97, + "learning_rate": 7.438829418404681e-10, + "loss": 0.0004, + "step": 60842 + }, + { + "epoch": 7.97, + "learning_rate": 7.374285359096345e-10, + "loss": 0.0005, + "step": 60843 + }, + { + "epoch": 7.97, + "learning_rate": 7.310022523721038e-10, + "loss": 0.0009, + "step": 60844 + }, + { + "epoch": 7.97, + "learning_rate": 7.2460409124675e-10, + "loss": 0.0004, + "step": 60845 + }, + { + "epoch": 7.97, + "learning_rate": 7.182340525502262e-10, + "loss": 0.0002, + "step": 60846 + }, + { + "epoch": 7.97, + "learning_rate": 7.118921363002962e-10, + "loss": 0.0004, + "step": 60847 + }, + { + "epoch": 7.97, + "learning_rate": 7.055783425169438e-10, + "loss": 0.0005, + "step": 60848 + }, + { + "epoch": 7.97, + "learning_rate": 6.992926712168225e-10, + "loss": 0.0006, + "step": 60849 + }, + { + "epoch": 7.97, + "learning_rate": 6.930351224165854e-10, + "loss": 0.0005, + "step": 60850 + }, + { + "epoch": 7.97, + "learning_rate": 6.868056961351066e-10, + "loss": 0.0004, + "step": 60851 + }, + { + "epoch": 7.97, + "learning_rate": 6.806043923890393e-10, + "loss": 0.0003, + "step": 60852 + }, + { + "epoch": 7.97, + "learning_rate": 6.744312111972573e-10, + "loss": 0.0002, + "step": 60853 + }, + { + "epoch": 7.97, + "learning_rate": 6.682861525753037e-10, + "loss": 0.0004, + "step": 60854 + }, + { + "epoch": 7.97, + "learning_rate": 6.621692165420523e-10, + "loss": 0.0007, + "step": 60855 + }, + { + "epoch": 7.97, + "learning_rate": 6.560804031130464e-10, + "loss": 0.0002, + "step": 60856 + }, + { + "epoch": 7.97, + "learning_rate": 6.500197123060492e-10, + "loss": 0.0002, + "step": 60857 + }, + { + "epoch": 7.97, + "learning_rate": 6.439871441399348e-10, + "loss": 0.0004, + "step": 60858 + }, + { + "epoch": 7.97, + "learning_rate": 6.379826986291359e-10, + "loss": 0.0004, + "step": 60859 + }, + { + "epoch": 7.97, + "learning_rate": 6.320063757914164e-10, + "loss": 0.0006, + "step": 60860 + }, + { + "epoch": 7.97, + "learning_rate": 6.260581756434292e-10, + "loss": 0.0005, + "step": 60861 + }, + { + "epoch": 7.97, + "learning_rate": 6.201380982029381e-10, + "loss": 0.0006, + "step": 60862 + }, + { + "epoch": 7.97, + "learning_rate": 6.142461434854863e-10, + "loss": 0.0004, + "step": 60863 + }, + { + "epoch": 7.97, + "learning_rate": 6.08382311507727e-10, + "loss": 0.0002, + "step": 60864 + }, + { + "epoch": 7.97, + "learning_rate": 6.025466022863136e-10, + "loss": 0.0004, + "step": 60865 + }, + { + "epoch": 7.97, + "learning_rate": 5.967390158390096e-10, + "loss": 0.0002, + "step": 60866 + }, + { + "epoch": 7.97, + "learning_rate": 5.909595521791378e-10, + "loss": 0.0008, + "step": 60867 + }, + { + "epoch": 7.97, + "learning_rate": 5.852082113255719e-10, + "loss": 0.0002, + "step": 60868 + }, + { + "epoch": 7.97, + "learning_rate": 5.794849932938551e-10, + "loss": 0.0005, + "step": 60869 + }, + { + "epoch": 7.97, + "learning_rate": 5.737898980995304e-10, + "loss": 0.0004, + "step": 60870 + }, + { + "epoch": 7.97, + "learning_rate": 5.681229257581411e-10, + "loss": 0.0003, + "step": 60871 + }, + { + "epoch": 7.97, + "learning_rate": 5.624840762874506e-10, + "loss": 0.0003, + "step": 60872 + }, + { + "epoch": 7.97, + "learning_rate": 5.568733497018919e-10, + "loss": 0.0007, + "step": 60873 + }, + { + "epoch": 7.97, + "learning_rate": 5.512907460181183e-10, + "loss": 0.0006, + "step": 60874 + }, + { + "epoch": 7.97, + "learning_rate": 5.457362652505627e-10, + "loss": 0.0004, + "step": 60875 + }, + { + "epoch": 7.97, + "learning_rate": 5.402099074158784e-10, + "loss": 0.0003, + "step": 60876 + }, + { + "epoch": 7.97, + "learning_rate": 5.347116725296087e-10, + "loss": 0.0002, + "step": 60877 + }, + { + "epoch": 7.97, + "learning_rate": 5.292415606072965e-10, + "loss": 0.0004, + "step": 60878 + }, + { + "epoch": 7.97, + "learning_rate": 5.237995716633748e-10, + "loss": 0.0006, + "step": 60879 + }, + { + "epoch": 7.97, + "learning_rate": 5.183857057133867e-10, + "loss": 0.0004, + "step": 60880 + }, + { + "epoch": 7.97, + "learning_rate": 5.129999627739857e-10, + "loss": 0.0003, + "step": 60881 + }, + { + "epoch": 7.98, + "learning_rate": 5.076423428584942e-10, + "loss": 0.0005, + "step": 60882 + }, + { + "epoch": 7.98, + "learning_rate": 5.023128459824556e-10, + "loss": 0.0005, + "step": 60883 + }, + { + "epoch": 7.98, + "learning_rate": 4.970114721614128e-10, + "loss": 0.0003, + "step": 60884 + }, + { + "epoch": 7.98, + "learning_rate": 4.91738221410909e-10, + "loss": 0.0007, + "step": 60885 + }, + { + "epoch": 7.98, + "learning_rate": 4.864930937442669e-10, + "loss": 0.0003, + "step": 60886 + }, + { + "epoch": 7.98, + "learning_rate": 4.812760891770296e-10, + "loss": 0.0004, + "step": 60887 + }, + { + "epoch": 7.98, + "learning_rate": 4.7608720772363e-10, + "loss": 0.0001, + "step": 60888 + }, + { + "epoch": 7.98, + "learning_rate": 4.709264493985011e-10, + "loss": 0.0007, + "step": 60889 + }, + { + "epoch": 7.98, + "learning_rate": 4.657938142160756e-10, + "loss": 0.0005, + "step": 60890 + }, + { + "epoch": 7.98, + "learning_rate": 4.6068930219189676e-10, + "loss": 0.0005, + "step": 60891 + }, + { + "epoch": 7.98, + "learning_rate": 4.5561291333817704e-10, + "loss": 0.0005, + "step": 60892 + }, + { + "epoch": 7.98, + "learning_rate": 4.5056464767156974e-10, + "loss": 0.0005, + "step": 60893 + }, + { + "epoch": 7.98, + "learning_rate": 4.455445052053975e-10, + "loss": 0.0003, + "step": 60894 + }, + { + "epoch": 7.98, + "learning_rate": 4.405524859540933e-10, + "loss": 0.0002, + "step": 60895 + }, + { + "epoch": 7.98, + "learning_rate": 4.3558858992986953e-10, + "loss": 0.0003, + "step": 60896 + }, + { + "epoch": 7.98, + "learning_rate": 4.306528171493796e-10, + "loss": 0.0004, + "step": 60897 + }, + { + "epoch": 7.98, + "learning_rate": 4.2574516762372563e-10, + "loss": 0.0004, + "step": 60898 + }, + { + "epoch": 7.98, + "learning_rate": 4.208656413695611e-10, + "loss": 0.0005, + "step": 60899 + }, + { + "epoch": 7.98, + "learning_rate": 4.1601423839798815e-10, + "loss": 0.0003, + "step": 60900 + }, + { + "epoch": 7.98, + "learning_rate": 4.1119095872454994e-10, + "loss": 0.0006, + "step": 60901 + }, + { + "epoch": 7.98, + "learning_rate": 4.0639580236256914e-10, + "loss": 0.0005, + "step": 60902 + }, + { + "epoch": 7.98, + "learning_rate": 4.016287693242582e-10, + "loss": 0.0002, + "step": 60903 + }, + { + "epoch": 7.98, + "learning_rate": 3.9688985962405004e-10, + "loss": 0.0003, + "step": 60904 + }, + { + "epoch": 7.98, + "learning_rate": 3.9217907327526726e-10, + "loss": 0.0009, + "step": 60905 + }, + { + "epoch": 7.98, + "learning_rate": 3.8749641029123265e-10, + "loss": 0.0006, + "step": 60906 + }, + { + "epoch": 7.98, + "learning_rate": 3.8284187068415855e-10, + "loss": 0.0007, + "step": 60907 + }, + { + "epoch": 7.98, + "learning_rate": 3.7821545446847797e-10, + "loss": 0.0009, + "step": 60908 + }, + { + "epoch": 7.98, + "learning_rate": 3.7361716165529306e-10, + "loss": 0.0002, + "step": 60909 + }, + { + "epoch": 7.98, + "learning_rate": 3.6904699226014696e-10, + "loss": 0.0006, + "step": 60910 + }, + { + "epoch": 7.98, + "learning_rate": 3.645049462941419e-10, + "loss": 0.0012, + "step": 60911 + }, + { + "epoch": 7.98, + "learning_rate": 3.5999102377060057e-10, + "loss": 0.0008, + "step": 60912 + }, + { + "epoch": 7.98, + "learning_rate": 3.5550522470173545e-10, + "loss": 0.0004, + "step": 60913 + }, + { + "epoch": 7.98, + "learning_rate": 3.5104754910086914e-10, + "loss": 0.0005, + "step": 60914 + }, + { + "epoch": 7.98, + "learning_rate": 3.4661799698021414e-10, + "loss": 0.0005, + "step": 60915 + }, + { + "epoch": 7.98, + "learning_rate": 3.422165683519829e-10, + "loss": 0.0002, + "step": 60916 + }, + { + "epoch": 7.98, + "learning_rate": 3.378432632283879e-10, + "loss": 0.0007, + "step": 60917 + }, + { + "epoch": 7.98, + "learning_rate": 3.3349808162275177e-10, + "loss": 0.0003, + "step": 60918 + }, + { + "epoch": 7.98, + "learning_rate": 3.291810235461768e-10, + "loss": 0.0007, + "step": 60919 + }, + { + "epoch": 7.98, + "learning_rate": 3.248920890108753e-10, + "loss": 0.0001, + "step": 60920 + }, + { + "epoch": 7.98, + "learning_rate": 3.2063127803017016e-10, + "loss": 0.0018, + "step": 60921 + }, + { + "epoch": 7.98, + "learning_rate": 3.163985906151634e-10, + "loss": 0.0008, + "step": 60922 + }, + { + "epoch": 7.98, + "learning_rate": 3.1219402677695743e-10, + "loss": 0.0002, + "step": 60923 + }, + { + "epoch": 7.98, + "learning_rate": 3.080175865288748e-10, + "loss": 0.0002, + "step": 60924 + }, + { + "epoch": 7.98, + "learning_rate": 3.038692698820178e-10, + "loss": 0.0005, + "step": 60925 + }, + { + "epoch": 7.98, + "learning_rate": 2.9974907684748865e-10, + "loss": 0.0002, + "step": 60926 + }, + { + "epoch": 7.98, + "learning_rate": 2.9565700743749983e-10, + "loss": 0.0005, + "step": 60927 + }, + { + "epoch": 7.98, + "learning_rate": 2.915930616642637e-10, + "loss": 0.0003, + "step": 60928 + }, + { + "epoch": 7.98, + "learning_rate": 2.8755723953777235e-10, + "loss": 0.0006, + "step": 60929 + }, + { + "epoch": 7.98, + "learning_rate": 2.8354954107023826e-10, + "loss": 0.0007, + "step": 60930 + }, + { + "epoch": 7.98, + "learning_rate": 2.7956996627276357e-10, + "loss": 0.0003, + "step": 60931 + }, + { + "epoch": 7.98, + "learning_rate": 2.756185151553403e-10, + "loss": 0.0007, + "step": 60932 + }, + { + "epoch": 7.98, + "learning_rate": 2.716951877312912e-10, + "loss": 0.0005, + "step": 60933 + }, + { + "epoch": 7.98, + "learning_rate": 2.6779998401060827e-10, + "loss": 0.0006, + "step": 60934 + }, + { + "epoch": 7.98, + "learning_rate": 2.6393290400439363e-10, + "loss": 0.0011, + "step": 60935 + }, + { + "epoch": 7.98, + "learning_rate": 2.600939477226394e-10, + "loss": 0.0003, + "step": 60936 + }, + { + "epoch": 7.98, + "learning_rate": 2.56283115177558e-10, + "loss": 0.0011, + "step": 60937 + }, + { + "epoch": 7.98, + "learning_rate": 2.525004063780312e-10, + "loss": 0.0007, + "step": 60938 + }, + { + "epoch": 7.98, + "learning_rate": 2.4874582133738167e-10, + "loss": 0.0004, + "step": 60939 + }, + { + "epoch": 7.98, + "learning_rate": 2.45019360063381e-10, + "loss": 0.0004, + "step": 60940 + }, + { + "epoch": 7.98, + "learning_rate": 2.4132102256824163e-10, + "loss": 0.0011, + "step": 60941 + }, + { + "epoch": 7.98, + "learning_rate": 2.376508088619556e-10, + "loss": 0.0003, + "step": 60942 + }, + { + "epoch": 7.98, + "learning_rate": 2.3400871895451483e-10, + "loss": 0.0004, + "step": 60943 + }, + { + "epoch": 7.98, + "learning_rate": 2.303947528559114e-10, + "loss": 0.0004, + "step": 60944 + }, + { + "epoch": 7.98, + "learning_rate": 2.2680891057835775e-10, + "loss": 0.0002, + "step": 60945 + }, + { + "epoch": 7.98, + "learning_rate": 2.232511921285152e-10, + "loss": 0.0004, + "step": 60946 + }, + { + "epoch": 7.98, + "learning_rate": 2.1972159751970644e-10, + "loss": 0.0003, + "step": 60947 + }, + { + "epoch": 7.98, + "learning_rate": 2.1622012675970306e-10, + "loss": 0.0006, + "step": 60948 + }, + { + "epoch": 7.98, + "learning_rate": 2.1274677985960723e-10, + "loss": 0.0007, + "step": 60949 + }, + { + "epoch": 7.98, + "learning_rate": 2.0930155682830078e-10, + "loss": 0.0004, + "step": 60950 + }, + { + "epoch": 7.98, + "learning_rate": 2.058844576757757e-10, + "loss": 0.0006, + "step": 60951 + }, + { + "epoch": 7.98, + "learning_rate": 2.0249548241091377e-10, + "loss": 0.0002, + "step": 60952 + }, + { + "epoch": 7.98, + "learning_rate": 1.9913463104481722e-10, + "loss": 0.0006, + "step": 60953 + }, + { + "epoch": 7.98, + "learning_rate": 1.9580190358636787e-10, + "loss": 0.0003, + "step": 60954 + }, + { + "epoch": 7.98, + "learning_rate": 1.9249730004444744e-10, + "loss": 0.0005, + "step": 60955 + }, + { + "epoch": 7.98, + "learning_rate": 1.89220820429048e-10, + "loss": 0.0002, + "step": 60956 + }, + { + "epoch": 7.98, + "learning_rate": 1.8597246474794106e-10, + "loss": 0.0005, + "step": 60957 + }, + { + "epoch": 7.99, + "learning_rate": 1.827522330122289e-10, + "loss": 0.0004, + "step": 60958 + }, + { + "epoch": 7.99, + "learning_rate": 1.7956012522968302e-10, + "loss": 0.0003, + "step": 60959 + }, + { + "epoch": 7.99, + "learning_rate": 1.7639614140918525e-10, + "loss": 0.0004, + "step": 60960 + }, + { + "epoch": 7.99, + "learning_rate": 1.732602815607276e-10, + "loss": 0.0001, + "step": 60961 + }, + { + "epoch": 7.99, + "learning_rate": 1.7015254569208161e-10, + "loss": 0.0004, + "step": 60962 + }, + { + "epoch": 7.99, + "learning_rate": 1.6707293381212908e-10, + "loss": 0.0011, + "step": 60963 + }, + { + "epoch": 7.99, + "learning_rate": 1.6402144592975178e-10, + "loss": 0.0003, + "step": 60964 + }, + { + "epoch": 7.99, + "learning_rate": 1.609980820538315e-10, + "loss": 0.0009, + "step": 60965 + }, + { + "epoch": 7.99, + "learning_rate": 1.5800284219325002e-10, + "loss": 0.0002, + "step": 60966 + }, + { + "epoch": 7.99, + "learning_rate": 1.5503572635466867e-10, + "loss": 0.0004, + "step": 60967 + }, + { + "epoch": 7.99, + "learning_rate": 1.520967345480795e-10, + "loss": 0.0003, + "step": 60968 + }, + { + "epoch": 7.99, + "learning_rate": 1.4918586678125403e-10, + "loss": 0.0003, + "step": 60969 + }, + { + "epoch": 7.99, + "learning_rate": 1.4630312306196382e-10, + "loss": 0.0009, + "step": 60970 + }, + { + "epoch": 7.99, + "learning_rate": 1.4344850339798045e-10, + "loss": 0.0007, + "step": 60971 + }, + { + "epoch": 7.99, + "learning_rate": 1.4062200779929592e-10, + "loss": 0.0002, + "step": 60972 + }, + { + "epoch": 7.99, + "learning_rate": 1.3782363627257155e-10, + "loss": 0.0005, + "step": 60973 + }, + { + "epoch": 7.99, + "learning_rate": 1.3505338882446873e-10, + "loss": 0.0002, + "step": 60974 + }, + { + "epoch": 7.99, + "learning_rate": 1.323112654649794e-10, + "loss": 0.0005, + "step": 60975 + }, + { + "epoch": 7.99, + "learning_rate": 1.2959726620076496e-10, + "loss": 0.0003, + "step": 60976 + }, + { + "epoch": 7.99, + "learning_rate": 1.2691139103959694e-10, + "loss": 0.0004, + "step": 60977 + }, + { + "epoch": 7.99, + "learning_rate": 1.2425363998924688e-10, + "loss": 0.0005, + "step": 60978 + }, + { + "epoch": 7.99, + "learning_rate": 1.2162401305637616e-10, + "loss": 0.0009, + "step": 60979 + }, + { + "epoch": 7.99, + "learning_rate": 1.1902251024875633e-10, + "loss": 0.0003, + "step": 60980 + }, + { + "epoch": 7.99, + "learning_rate": 1.1644913157415893e-10, + "loss": 0.0004, + "step": 60981 + }, + { + "epoch": 7.99, + "learning_rate": 1.1390387703924532e-10, + "loss": 0.0006, + "step": 60982 + }, + { + "epoch": 7.99, + "learning_rate": 1.1138674665178706e-10, + "loss": 0.0008, + "step": 60983 + }, + { + "epoch": 7.99, + "learning_rate": 1.0889774041844548e-10, + "loss": 0.0005, + "step": 60984 + }, + { + "epoch": 7.99, + "learning_rate": 1.0643685834588191e-10, + "loss": 0.0003, + "step": 60985 + }, + { + "epoch": 7.99, + "learning_rate": 1.0400410044186793e-10, + "loss": 0.0009, + "step": 60986 + }, + { + "epoch": 7.99, + "learning_rate": 1.0159946671306486e-10, + "loss": 0.0005, + "step": 60987 + }, + { + "epoch": 7.99, + "learning_rate": 9.922295716613406e-11, + "loss": 0.0011, + "step": 60988 + }, + { + "epoch": 7.99, + "learning_rate": 9.687457180662663e-11, + "loss": 0.0004, + "step": 60989 + }, + { + "epoch": 7.99, + "learning_rate": 9.455431064342436e-11, + "loss": 0.001, + "step": 60990 + }, + { + "epoch": 7.99, + "learning_rate": 9.226217368096812e-11, + "loss": 0.0003, + "step": 60991 + }, + { + "epoch": 7.99, + "learning_rate": 8.999816092591929e-11, + "loss": 0.0002, + "step": 60992 + }, + { + "epoch": 7.99, + "learning_rate": 8.77622723860494e-11, + "loss": 0.0007, + "step": 60993 + }, + { + "epoch": 7.99, + "learning_rate": 8.555450806690958e-11, + "loss": 0.0002, + "step": 60994 + }, + { + "epoch": 7.99, + "learning_rate": 8.337486797405092e-11, + "loss": 0.0006, + "step": 60995 + }, + { + "epoch": 7.99, + "learning_rate": 8.12233521141348e-11, + "loss": 0.0006, + "step": 60996 + }, + { + "epoch": 7.99, + "learning_rate": 7.909996049382252e-11, + "loss": 0.0004, + "step": 60997 + }, + { + "epoch": 7.99, + "learning_rate": 7.70046931186652e-11, + "loss": 0.0003, + "step": 60998 + }, + { + "epoch": 7.99, + "learning_rate": 7.493754999421398e-11, + "loss": 0.0001, + "step": 60999 + }, + { + "epoch": 7.99, + "learning_rate": 7.289853112601997e-11, + "loss": 0.0002, + "step": 61000 + }, + { + "epoch": 7.99, + "learning_rate": 7.088763652074449e-11, + "loss": 0.0005, + "step": 61001 + }, + { + "epoch": 7.99, + "learning_rate": 6.890486618282843e-11, + "loss": 0.0005, + "step": 61002 + }, + { + "epoch": 7.99, + "learning_rate": 6.695022011893315e-11, + "loss": 0.0002, + "step": 61003 + }, + { + "epoch": 7.99, + "learning_rate": 6.502369833460976e-11, + "loss": 0.0005, + "step": 61004 + }, + { + "epoch": 7.99, + "learning_rate": 6.312530083429913e-11, + "loss": 0.0005, + "step": 61005 + }, + { + "epoch": 7.99, + "learning_rate": 6.125502762466263e-11, + "loss": 0.0003, + "step": 61006 + }, + { + "epoch": 7.99, + "learning_rate": 5.941287871014112e-11, + "loss": 0.0001, + "step": 61007 + }, + { + "epoch": 7.99, + "learning_rate": 5.7598854095175515e-11, + "loss": 0.0003, + "step": 61008 + }, + { + "epoch": 7.99, + "learning_rate": 5.581295378642715e-11, + "loss": 0.0008, + "step": 61009 + }, + { + "epoch": 7.99, + "learning_rate": 5.405517778833691e-11, + "loss": 0.0003, + "step": 61010 + }, + { + "epoch": 7.99, + "learning_rate": 5.2325526105345694e-11, + "loss": 0.0003, + "step": 61011 + }, + { + "epoch": 7.99, + "learning_rate": 5.0623998741894385e-11, + "loss": 0.0004, + "step": 61012 + }, + { + "epoch": 7.99, + "learning_rate": 4.895059570464433e-11, + "loss": 0.0002, + "step": 61013 + }, + { + "epoch": 7.99, + "learning_rate": 4.73053169969262e-11, + "loss": 0.0002, + "step": 61014 + }, + { + "epoch": 7.99, + "learning_rate": 4.5688162623180876e-11, + "loss": 0.0002, + "step": 61015 + }, + { + "epoch": 7.99, + "learning_rate": 4.409913258895948e-11, + "loss": 0.0005, + "step": 61016 + }, + { + "epoch": 7.99, + "learning_rate": 4.253822689759268e-11, + "loss": 0.0002, + "step": 61017 + }, + { + "epoch": 7.99, + "learning_rate": 4.1005445554631596e-11, + "loss": 0.0004, + "step": 61018 + }, + { + "epoch": 7.99, + "learning_rate": 3.950078856340689e-11, + "loss": 0.0005, + "step": 61019 + }, + { + "epoch": 7.99, + "learning_rate": 3.802425592946968e-11, + "loss": 0.0004, + "step": 61020 + }, + { + "epoch": 7.99, + "learning_rate": 3.657584765504041e-11, + "loss": 0.0002, + "step": 61021 + }, + { + "epoch": 7.99, + "learning_rate": 3.51555637456702e-11, + "loss": 0.0002, + "step": 61022 + }, + { + "epoch": 7.99, + "learning_rate": 3.376340420468971e-11, + "loss": 0.0005, + "step": 61023 + }, + { + "epoch": 7.99, + "learning_rate": 3.239936903653984e-11, + "loss": 0.0003, + "step": 61024 + }, + { + "epoch": 7.99, + "learning_rate": 3.1063458244551256e-11, + "loss": 0.0001, + "step": 61025 + }, + { + "epoch": 7.99, + "learning_rate": 2.9755671832054637e-11, + "loss": 0.0002, + "step": 61026 + }, + { + "epoch": 7.99, + "learning_rate": 2.8476009803490856e-11, + "loss": 0.0011, + "step": 61027 + }, + { + "epoch": 7.99, + "learning_rate": 2.7224472163300818e-11, + "loss": 0.0005, + "step": 61028 + }, + { + "epoch": 7.99, + "learning_rate": 2.600105891370497e-11, + "loss": 0.0004, + "step": 61029 + }, + { + "epoch": 7.99, + "learning_rate": 2.4805770058033974e-11, + "loss": 0.0005, + "step": 61030 + }, + { + "epoch": 7.99, + "learning_rate": 2.3638605599618503e-11, + "loss": 0.0004, + "step": 61031 + }, + { + "epoch": 7.99, + "learning_rate": 2.249956554289945e-11, + "loss": 0.0005, + "step": 61032 + }, + { + "epoch": 7.99, + "learning_rate": 2.1388649890097258e-11, + "loss": 0.0004, + "step": 61033 + }, + { + "epoch": 8.0, + "learning_rate": 2.0305858644542596e-11, + "loss": 0.0007, + "step": 61034 + }, + { + "epoch": 8.0, + "learning_rate": 1.9251191809566138e-11, + "loss": 0.0003, + "step": 61035 + }, + { + "epoch": 8.0, + "learning_rate": 1.822464938849855e-11, + "loss": 0.0005, + "step": 61036 + }, + { + "epoch": 8.0, + "learning_rate": 1.7226231382450055e-11, + "loss": 0.0004, + "step": 61037 + }, + { + "epoch": 8.0, + "learning_rate": 1.6255937796971764e-11, + "loss": 0.0003, + "step": 61038 + }, + { + "epoch": 8.0, + "learning_rate": 1.5313768632063686e-11, + "loss": 0.0005, + "step": 61039 + }, + { + "epoch": 8.0, + "learning_rate": 1.4399723891056483e-11, + "loss": 0.0007, + "step": 61040 + }, + { + "epoch": 8.0, + "learning_rate": 1.351380357839105e-11, + "loss": 0.0005, + "step": 61041 + }, + { + "epoch": 8.0, + "learning_rate": 1.2656007694067385e-11, + "loss": 0.0009, + "step": 61042 + }, + { + "epoch": 8.0, + "learning_rate": 1.1826336242526382e-11, + "loss": 0.0003, + "step": 61043 + }, + { + "epoch": 8.0, + "learning_rate": 1.1024789224878263e-11, + "loss": 0.0009, + "step": 61044 + }, + { + "epoch": 8.0, + "learning_rate": 1.0251366643343475e-11, + "loss": 0.0002, + "step": 61045 + }, + { + "epoch": 8.0, + "learning_rate": 9.506068500142463e-12, + "loss": 0.0004, + "step": 61046 + }, + { + "epoch": 8.0, + "learning_rate": 8.788894798605897e-12, + "loss": 0.0003, + "step": 61047 + }, + { + "epoch": 8.0, + "learning_rate": 8.099845539843997e-12, + "loss": 0.0005, + "step": 61048 + }, + { + "epoch": 8.0, + "learning_rate": 7.43892072496699e-12, + "loss": 0.0004, + "step": 61049 + }, + { + "epoch": 8.0, + "learning_rate": 6.806120357305546e-12, + "loss": 0.0004, + "step": 61050 + }, + { + "epoch": 8.0, + "learning_rate": 6.2014444368596604e-12, + "loss": 0.0004, + "step": 61051 + }, + { + "epoch": 8.0, + "learning_rate": 5.624892968070228e-12, + "loss": 0.0005, + "step": 61052 + }, + { + "epoch": 8.0, + "learning_rate": 5.076465949827025e-12, + "loss": 0.0002, + "step": 61053 + }, + { + "epoch": 8.0, + "learning_rate": 4.556163385460721e-12, + "loss": 0.0002, + "step": 61054 + }, + { + "epoch": 8.0, + "learning_rate": 4.063985274971316e-12, + "loss": 0.0003, + "step": 61055 + }, + { + "epoch": 8.0, + "learning_rate": 3.5999316205792557e-12, + "loss": 0.0005, + "step": 61056 + }, + { + "epoch": 8.0, + "learning_rate": 3.164002424504986e-12, + "loss": 0.0002, + "step": 61057 + }, + { + "epoch": 8.0, + "learning_rate": 2.7561976856382844e-12, + "loss": 0.0002, + "step": 61058 + }, + { + "epoch": 8.0, + "learning_rate": 2.3765174073098195e-12, + "loss": 0.0008, + "step": 61059 + }, + { + "epoch": 8.0, + "learning_rate": 2.0249615895195918e-12, + "loss": 0.0003, + "step": 61060 + }, + { + "epoch": 8.0, + "learning_rate": 1.7015302344880468e-12, + "loss": 0.0003, + "step": 61061 + }, + { + "epoch": 8.0, + "learning_rate": 1.406223341104962e-12, + "loss": 0.0006, + "step": 61062 + }, + { + "epoch": 8.0, + "learning_rate": 1.139040911590783e-12, + "loss": 0.0003, + "step": 61063 + }, + { + "epoch": 8.0, + "learning_rate": 8.999829459455101e-13, + "loss": 0.0004, + "step": 61064 + }, + { + "epoch": 8.0, + "learning_rate": 6.890494452793661e-13, + "loss": 0.0001, + "step": 61065 + }, + { + "epoch": 8.0, + "learning_rate": 5.062404107025742e-13, + "loss": 0.0002, + "step": 61066 + }, + { + "epoch": 8.0, + "learning_rate": 3.5155584110491136e-13, + "loss": 0.0005, + "step": 61067 + }, + { + "epoch": 8.0, + "learning_rate": 2.2499573870682356e-13, + "loss": 0.0003, + "step": 61068 + }, + { + "epoch": 8.0, + "learning_rate": 1.2656010350831082e-13, + "loss": 0.0007, + "step": 61069 + }, + { + "epoch": 8.0, + "learning_rate": 5.6248934399150134e-14, + "loss": 0.0002, + "step": 61070 + }, + { + "epoch": 8.0, + "learning_rate": 1.4062233599787534e-14, + "loss": 0.0005, + "step": 61071 + }, + { + "epoch": 8.0, + "learning_rate": 0.0, + "loss": 0.0002, + "step": 61072 + }, + { + "epoch": 8.0, + "step": 61072, + "total_flos": 1.7162385804053237e+20, + "train_loss": 0.04017619166928129, + "train_runtime": 233548.1884, + "train_samples_per_second": 33.47, + "train_steps_per_second": 0.261 + } + ], + "logging_steps": 1.0, + "max_steps": 61072, + "num_input_tokens_seen": 0, + "num_train_epochs": 8, + "save_steps": 20000000, + "total_flos": 1.7162385804053237e+20, + "train_batch_size": 2, + "trial_name": null, + "trial_params": null +}